article
stringlengths 0
745k
| abstract
stringlengths 0
92.7k
|
|---|---|
for over a decade extensive research has been undertaken on negative refractive indexed metamaterials @xcite both in the microwave and optical regimes .
the research in optical metamaterials has been stimulated by the possibility of light manipulation methods and potential applications .
most of the experimental work with metamaterials utilizes the plasmonic resonance in metallic structures embedded into a host dielectric .
the disadvantage of this type of metamaterial arises from the high loss value that results from an induced plasmonic resonance in the optical range .
currently , the reduction or compensation of these losses has been the the focus of intensive research .
conceptually , the compensation of losses can be achieved in two ways .
first , by adding gain into a host material containing the metallic nanostructures .
second , by adding a gain material into the nanostructures themselves . in this paper , we consider the optical pulse dynamics in a negative index metamaterial using both types of loss compensation techniques .
to understand the general features of the pulse dynamics in such a system we consider a simplified mathematical model .
the propagation of an electromagnetic wave in matter is described by the system of maxwell s equations : @xmath0 by choosing a system of coordinates in such a way as to assume that a wave propagates in the @xmath1-direction , the magnetic field is oriented in @xmath2-direction , @xmath3 , and the electric field is oriented in the @xmath4-direction , @xmath5 , we transform the system of vector equations ( [ vector : maxwell ] ) into a system of scalar equations : @xmath6 we start our analysis of the system from the initial case of an electromagnetic wave as it propagates through a nanostructured host dielectric without a gain material .
we also assume that all resonances of the host material are far from the resonances of the metallic nanostructures . in this case ,
@xmath7 and @xmath8 can be represented in the standard form through the medium polarization , @xmath9 , and the magnetization , @xmath10 : @xmath11 to close the system of equations we need to add material equations that describe the interaction of @xmath9 and @xmath10 with an electromagnetic wave .
the total medium response can be represented as sum of the responses from the nanostructures and the gain material : @xmath12 and @xmath13 . to describe such an interaction
, we consider the simplified model represented by distributed linear electric and magnetic resonators .
the plasmonic oscillations result in an oscillation of the polarization described by following equation @xcite : @xmath14 here @xmath15 is the effective plasma frequency , @xmath16 is the frequency of the dimensional quantization that results from the geometry of the resonator , and @xmath17 accounts for losses . in the popular fishnet structures the dielectric permittivity is negative in operating frequency range . in this case
we can choose @xmath18 .
to describe the magnetic response we assume that a magnetic resonator in a leading order operates as an @xmath19-circuit @xcite .
this allows us to write a partial differential equation that relates the magnetization and the magnetic field as follows @xcite : @xmath20 where @xmath21 , @xmath22 are effective inductance and capacity , @xmath23 is a tompson frequency , @xmath24 is a dissipative constant , and @xmath25 describes coupling of nanoresonator with external magnetic flux @xcite . the system ( [ scalar : maxwell ] ) , ( [ b : and : d : general ] ) , ( [ polarization_eqn ] ) , ( [ magnetization_eqn ] ) is a generalization of the well known maxwell - lorentz system of equations , which describes the resonant interaction of an electromagnetic field with a system of linear oscillators that responds only to the electric component of incident light .
this system has played a fundamental role in understanding the interaction of light with resonant atoms @xcite .
the host material is characterized by the dielectric permittivity , @xmath26 , and the magnetic permeability , @xmath27 . for most materials
the permeability in the optical range is @xmath28 , that is , @xmath29 .
now , by taking the fourier transform of the equations in ( [ scalar : maxwell ] ) and substituting in the expressions for @xmath30 and @xmath31 we obtain the following equations for fourier components @xmath32 and @xmath33 in terms of @xmath34 and @xmath35 : @xmath36 in order to describe the response from the nanostructures we introduce both an effective dielectric and a magnetic susceptibility .
we consider here the case of a `` diluted '' metamaterial when near neighbor interactions of the nanostructures are much weaker than their interaction via the electromagnetic field . the nanostructure contribution to the effective permittivity and permeability
can be obtained from equations ( [ polarization_eqn ] ) and ( [ magnetization_eqn ] ) by applying the fourier transform , that is , @xmath37 ultrashort optical pulses are well described by a slowly varying envelope approximation .
this approximation allows one to simplify the system ( [ scalar : maxwell ] ) , ( [ b : and : d : general ] ) , ( [ polarization_eqn ] ) , ( [ magnetization_eqn ] ) by representing @xmath42 , @xmath43 , @xmath9 and @xmath10 in the following form : @xmath44 where @xmath45 , @xmath46 , @xmath47 and @xmath48 are slowly - varying functions of @xmath49 and @xmath1 .
it should be noted that the constant @xmath50 stands for the projection of the wave vector on @xmath1 axis , and that it can be both positive and negative for positive and negative index materials respectively .
since the width of wave packet is much less than carrier frequency , the inequalities @xmath53 and @xmath54 still hold . by expanding the expression in the square brackets in equations ( [ fourier : slow : e ] ) we obtain @xmath55 & \approx & k^2_0 + 2k_0 k+k^2 - \frac{(\omega^2_0 + 2 \omega_0 \omega + \omega^2)}{c^2 } [ n^2_0 ( \omega_0 ) \nonumber \\ & + & \left[\frac{\partial n^2}{\partial \omega}\right]_{\omega_0 } \omega + \frac{1}{2 } \left[\frac{\partial^2 n^2}{\partial \omega^2}\right]_{\omega_0 } \omega^2],\end{aligned}\ ] ] where @xmath56 is a square of the effective complex index of refraction .
note that in the above when we divide both sides by @xmath50 the expansion in the fourier domain is of second order , where the second order contribution comes from @xmath65 term . in our further analysis
we will ignore this dispersive term by dropping this second order term .
thus , we have the following slowly - varying governing equations for the system : @xmath66 it follows from equations ( [ svepa : h ] ) that the magnetic field is proportional to the electric field .
therefore , it is sufficient to analyze only the equation for electric field .
also , note that if the carrier frequency is in the frequency range that determines a positive real value for the phase , then @xmath67 and the imaginary coefficients will all have a sign opposite that which that they are for the case when @xmath68 .
as noted in the previous section , the metallic nanoresonators are lossy . to compensate for these losses
we propose that active atoms be embedded in the dielectric host medium or into nanostructures .
it should be noted that here we consider coherent amplification , where the coherent state of the atomic ensemble is preserved during transition from an excited to a ground state .
the atomic inclusions that we consider here will be two - level , that is , they have an excited and a ground state . in the case
when amplifying atoms are included to host dielectric , the polarization , @xmath69 , is represented as polarizability , @xmath70 , of a two - level single multiplied by atomic density , @xmath71 : @xmath72 .
the bloch equations , which describe the interaction of polarizability and population difference , @xmath73 , of two - level atoms with the electric field is given in the slowly - varying envelope approximation as : @xmath74 where @xmath75 is the carrier frequency , @xmath76 is the frequency corresponding to the transition between the excited and ground state , @xmath77 is the dipole moment .
we consider the case when the spectral pulse width , @xmath78 , is within the negative index frequency , @xmath79 , domain and that the pulse width , @xmath80 , is smaller than the characteristic time of the atomic polarization relaxation , @xmath81 . in an experimental situation , @xmath82 @xcite ,
the value of @xmath81 varies in a broad range . for @xmath83 ions in a glass matrix at room temperature ,
@xmath84 and @xmath85 for @xmath86 at @xmath87 @xcite .
we consider the situation when all these requirements are fulfilled and coherent amplification takes place . to fulfill these requirements and select an optimal value for the carrier frequency we analyze the dispersion relation ( [ dispersion : law ] ) .
the negative index property is determined by the magnetic response of the metamaterial .
the parameter @xmath88 determines the strength of this response and therefore controls the negative index property .
this parameter depends on the design of the nanostructure .
the case of a double split ring was considered in @xcite , and from these papers it follows that the value of @xmath88 can be chosen from the interval @xmath89 $ ] . in real experimental situations of doubly resonant metamaterials ,
the values of @xmath16 and @xmath90 are close . in the case of a fishnet structure
there is no resonance in the electric response and @xmath16 can be chosen to be zero @xmath18 @xcite . in our study
we chose @xmath91 and all the frequencies are normalized relative to @xmath90 .
thus we chose @xmath92 , @xmath93 , @xmath94 and the dissipative constants were chosen as @xmath95 and @xmath96 .
this choice of parameters approximately corresponds to the case of silver .
the value of @xmath17 is taken from @xcite and the value of @xmath24 was extracted from experimental results presented in @xcite .
the functions @xmath97 and @xmath98 are presented in figure [ dispersion : relation ] .
the function @xmath97 is negative on the interval @xmath99 $ ] .
the function @xmath98 is positive , which corresponds to the presence of losses , and has a flat local minima around @xmath100 . the neighborhood of this flat local minima of @xmath98 and interval of negative values for @xmath97 overlap .
if the carrier frequency , @xmath75 , is chosen near this local minima then the effect of dispersion of dissipation can be neglected . since @xmath101 ,
the phase velocity is negative , while on the other hand the group velocity , @xmath62 , is positive @xmath102 for this interval of frequencies .
therefore , the phase and group velocities are oppositely directed for @xmath75 selected as described above .
if the resonant frequency of the active atoms is close to such an @xmath75 , the material possesses the negative index property and the dispersion of losses can be neglected .
the balance between coherent amplification and dissipation leads to the formation of a stationary optical pulse .
the shape and velocity of such pulse can be found using technique suggested in @xcite . in the following dimensionless variables : @xmath103 , @xmath104 , @xmath105 , @xmath106 , @xmath107
we can obtain the following governing equations for the system @xmath108 where @xmath109 by examining the behavior of the system near the excited state , @xmath110 , we will be able to study the linear behavior of the front .
the system of equations near the excited state has the following approximation in the comoving system of coordinates @xmath111 , @xmath112 : @xmath113 where @xmath114 .
taking the fourier transform of this system gives us the dispersion relation : @xmath115 thus , near the excited state an amplitude of the electric field can be approximated as @xmath116 where @xmath117 is the phase of the integral .
we are interested in the behavior of the traveling wave solution for large distances so as to determine solitary wave solutions .
the method for determining this type of a solution for integrals of this form is the stationary phase technique .
the phase , @xmath118 , has a fixed point at : @xmath119 .
that is , @xmath120 for these values of @xmath121 we see that the phase has the following forms : @xmath122 thus , we see that when the positive square root is chosen the solution will grow exponentially for large @xmath1 , which can not occur , and thus the contribution from this term must be zero . for the second stationary point
there is an exponential decay , which implies that we will have a strong dependence on @xmath1 . in our traveling wave solution
, the shape of the pulse for large @xmath1 should be determined by @xmath123 , and therefore this strong dependence of @xmath1 must be eliminated .
thus , we need the following to hold : @xmath124im@xmath125\sqrt{\beta } - \gamma=0 , $ ] that is , @xmath126 ^ 2}\\ v_{fr}=\frac{4 \mathrm{im } [ i\sqrt{1+i\delta}]^2}{4\mathrm{im } [ i\sqrt{1+i\delta}]^2+\gamma^2 } \nonumber\end{aligned}\ ] ] using this expression for @xmath127 and substituting in @xmath128 into the phase of the integral in expression ( [ e : integral ] ) we see that near the excited state the electric field can be approximated as follows : @xmath129 z ) }
i(z ) , \label{formoffront}\end{aligned}\ ] ] where @xmath130 is the integral of higher order terms from the expansion of the phase around that stationary point @xmath128 .
note that @xmath131= \lambda \beta_{cr}+ 2 \mathrm{re}[i\sqrt{1+i\delta}\sqrt{\beta_{cr}}]$ ] .
let us use the following notation to simplify this expression : @xmath132 , which allows us to simplify this term : @xmath133= \frac{\lambda \gamma^2 - 4\gamma \delta '' \delta'}{4(\delta')^2}.\end{aligned}\ ] ] also note that if @xmath134 , i.e. if the coefficient in front of the polarization in the first equation in the system of governing equations is not complex , and if we ignore detuning , @xmath135 , then the phase will be zero . also , note that the form of the front depends on @xmath136 , which is determined by the initial condition for the electric field .
in addition , we see that the shape of the pulse is determined by the variable @xmath123 , and the imaginary part of @xmath128 , that is the shape of the pulse is proportional to @xmath137im@xmath138\tau ) = \exp(-(2(\delta')^2/\gamma ) \tau)$ ] .
we also see that because @xmath128 is complex this solution will oscillate in @xmath123 with period determined by re@xmath139 $ ] .
thus , as we have noted before , when @xmath140 the phase is zero , and no oscillations exist . when numerical computations are performed this can be observed .
we have obtained two very important pieces of information from the linear analysis .
firstly , we have determined the velocity , @xmath141 , of the front , and for this reason we are now able to solve the ordinary differential equation determined by moving into a co - moving frame of coordinates .
secondly , we now know that the front oscillates as it propagates , so that before we move into the co - moving frame we must express the electric field and polarization envelopes as : @xmath142 and @xmath143 . plugging these into eqs .
( [ dimensionless : pde ] ) gives us the following system in the co - moving frame @xmath144 @xmath145 near @xmath110 we can observe that : @xmath146 with the velocity determined by our previous analysis and recall that @xmath147 the characteristic equation for the above system is degenerate .
to see this , let us look at the simpler case when @xmath135 , that is , when the carrier frequency is equal to the transition frequency of the atomic inclusions .
for this case we have the following ode @xmath148 which is degenerate with solutions of the form : @xmath149 note that to solve the nonlinear ode we need both @xmath150 and @xmath151 .
if we can determine @xmath152 from the initial condition given for @xmath153 , then we know @xmath151 . unfortunately , the above linear analysis does not allow us to do this because the characteristic equation is degenerate and our solution near the excited state does not allow us to simply express @xmath152 in terms of the @xmath150 .
however , if we look at the linearization around @xmath154 , the ground state , we see that : @xmath155 and using the velocity computed in the previous analysis our characteristic equation tells us that @xmath156 can be expressed as : @xmath157 where @xmath158 since we are analyzing the behavior of the front near the ground state we know that the shape of the pulse must decay exponentially here , and therefore @xmath159 .
thus , @xmath160 thus , given a choice for @xmath150 we see that : @xmath161 and we are able to numerically solve the nonlinear ordinary differential equation .
these solutions can be seen in figs .
( [ electricfield : ode])-([population : ode ] ) to match the asymptotic behavior let us write the form of the solution near the excited state in the form : @xmath162 this asymptotic form must match the condition obtained using the stationary phase technique .
since the position of the solution described by the stationary phase technique is described through z while the position of the front solution given by the ode solution is determined through the shift variable @xmath163 , by comparing the asymptotic of these two equations we will be able to express @xmath163 as a function of z. thus , if we add a shift term , @xmath163 to the ode solution we can compare asymptotic : @xmath164 thus , @xmath165 the position of the front is given by @xmath163 while the width of the pulse is given by @xmath166 .
since the transition from the stable to unstable states takes place on the order @xmath167 this implies that @xmath166 is of order @xmath167 .
on the other hand , @xmath163 the position of the pulse is dependent on the size of the incident pulse , @xmath168 , that is @xmath169 .
since the initialization is small this means that @xmath163 is large .
thus , @xmath170 and @xmath171 from this we see that @xmath172 and @xmath163 , the shift variable , can be determined iteratively .
let us return to the dimensionless system of partial differential equations ( [ dimensionless : pde ] ) .
to solve this system numerically we use a strang splitting method , where we have a step that solves the ordinary differential equations , and we have a separate step that solves the partial differential dimensionless maxwell s equation .
the ordinary differential equation step is solved using midpoint method , and the partial differential equation step is solved using the maccormack method . at the boundary we consider a small gaussian pulse @xmath173 and we also assume that the system is initially in the excited state , @xmath174 .
the result of the numerical computations can be seen in fig .
( [ intensity : numerical ] ) . as was noted in the section on linear analysis and in the previous section
, the pulse should oscillate as in moves through the medium . in the numerical computations
this does occur , and the period of these oscillations if given by the phase of the approximation of the front given in the expression ( [ formoffront ] ) .
a comparison of the velocity computed analytically and the velocity of the pulse generated by the system numerically can be seen in fig .
( [ analytic : numerical : complexzero ] ) .
another approach toward loss compensation in materials with embedded nanostructures is to add the amplification capability directly into the nanostructures .
there are two main motivations in favor of this approach .
first , the small size of the nanostructures will induce highly intense local electric fields that result from the interaction of the nanostructures with an external electromagnetic field .
this strong local field will enable a strong response from the gain material if it is inserted directly into a nanostructure . secondly
, since the nanostructures are the source of loss , the losses in the system occur on the scale of a nanostructure . in the case that was considered in the previous subsection ,
the loss was compensated by a gain material distributed over the entire host medium .
thus , a strong loss occurred on the scale of a nanostructure , while the compensatation acted on a mean - field scale .
the scales of loss and loss compensation are not in agreement when performed in this way . a more appropriate way to compensate for loss would not only balance the gain and loss of energy , but would also match the scales of these two procedures .
we propose that this can be achieved by embedding a gain material directly into the nanostructures , and in this subsection we focus on modeling such a method of loss compensation . in our analysis
we consider the simplest case where nanostructure in a leading order can be viewed as an @xmath19 circuit .
we consider nanostructures that produce spatially homogeneous effective local fields .
this simplification allows us to avoid a consideration of the spatial dynamics for a local field , and to take into account only temporal dynamics .
the external electric field generates a current in the @xmath19 circuit .
the current in turn establishes a local electric field across a `` capacitor '' .
thus , if we place active atoms or quantum dots in the part of the structure that is responsible for the electric capacity , we will be able to utilize the strong local electric field and directly compensate for loss .
we first derive an equation that describes the oscillatory processes in an @xmath19 circuit that contains active atoms in the capacitor . in the ideal case without losses ,
the potential difference at the capacitor can be expressed in the form : @xmath175 here @xmath176 is potential difference on the inductance and @xmath177 is the electromotive force due to magnetic flux variations through the area of the @xmath19 circuit
. we can express @xmath176 via the electric field , @xmath178 , and the polarization of the medium , @xmath179 , inside the capacitor : @xmath180 here @xmath77 is the size of the gap in a @xmath19 circuit , @xmath178 is the electric field induced by charges at the edges of a @xmath19 circuit , and @xmath181 is the polarization of the medium inside the capacitor .
the electric field in the capacitor is @xmath182 where @xmath183 is the area of the metallic plates formed at the end of the @xmath19 circuit .
taking into account the expression for @xmath42 , we see that the expression for @xmath184 has the following form : @xmath185 the potential difference on the inductance and the electromotive force can be expressed as : @xmath186 which means that the equation describing this @xmath19-circuit reads @xmath187 that is , @xmath188 recall that the capacity is given by : @xmath189 , which allows us to make the following substitution : @xmath190 from the expression for thompson s frequency , @xmath191 , we obtain system of equations @xmath192 if we consider a system with loss we add the potential drop due to resistance , @xmath193 , to the system and the governing equations will read : @xmath194 we know that the polarization effects due to the presence of active atomic inclusions has the form : @xmath195 where @xmath196 is the density of the atomic population , @xmath197 is an off - diagonal element of the density matrix and @xmath198 is the dipole moment . since these atoms are contained within the capacitor , we can express the atomic density as : @xmath199 where @xmath200 is the population and @xmath201 is the volume of the capacitor .
thus , the polarization term that comes from atomic inclusions placed inside the capacitor of the @xmath19 circuit can be expressed as : @xmath202 since the size of the `` capacitor '' is small it follows that the polarization contribution is quite large . recalling that the magnetization , @xmath10 , can be expressed as a function of the current in the @xmath19-circuit , @xmath203 the material equation determined by these nanostructures with inclusions placed inside the capacitor is therefore : @xmath204 where @xmath205 , @xmath206 , and @xmath207 .
now , as in the previous case we will be considering structures responding as an @xmath19-circuit to the magnetic field and these structures can have a resonant response to external electric field .
the polarization effects comes from plasmonic oscillations and the geometry of the nanostructures : @xmath208 where @xmath42 is the external electric field , @xmath209 is the polarization contribution from the nanostructures , @xmath210 is the losses due to plasmonic oscillations , @xmath16 is the frequency of the dimensional quantization that comes from the geometry of the resonator , and @xmath15 is the plasma frequency . now that we have the material equations for this case we need to derive the slowly - varying envelope equations . from eq .
( [ magnetization : comp ] ) , we take the fourier transform to see that : @xmath211 thus , @xmath212 and we see that the effective permeability of the system , @xmath213 , is the same as in the case of mean - field amplification . by taking the fourier trasform of eq .
( [ polarization : nano ] ) we obtain @xmath214 to see that the effective permittivity , @xmath215 , is also the same as that for mean - field amplification , which is determined by the nano - structures . recalling the fourier transform of maxwell s equations we have : @xmath216 where @xmath35 comes from the nanostructures .
note that the polarization effects from the active atoms interact mainly with the local electric field inside the capacitor , which is much stronger than the external electric field .
thus , we have @xmath217\hat{e}&= & -\frac{i 4\pi \omega^2 k \omega^2_t \alpha}{c(\omega^2_t -\omega^2 -i\omega \gamma)}\hat{p}_a \\
\left[k^2 -\frac{\omega^2 \tilde{\varepsilon}(\omega)\tilde{\mu}(\omega)}{c^2}\right ] \hat{h } & = & -\frac{i ( 4\pi)^2 \omega^3 \epsilon_0(\omega ) \omega^2_t \alpha}{c^2(\omega^2_t
-\omega^2 -i\omega\gamma ) } \hat{p}_a\end{aligned}\ ] ] thus , the expansion of the bracketed terms , and the eventual resulting slowly - varying envelope approximation is similar to the previous case .
@xmath218 \mathcal{e } = - \frac{i 4\pi \omega^2_0 k_0 \omega^2_t \alpha}{c(\omega^2_t -\omega^2_0 -i\omega_0 \gamma)}\mathcal{p}_a \\ i\left[\frac{\partial}{\partial x } + \frac{\phi}{v_g}\frac{\partial}{\partial t } + \gamma -i\omega \right ] \mathcal{h } = -\frac{i ( 4\pi)^2 \omega^3_0 \epsilon_0(\omega_0 ) \omega^2_t \alpha}{c^2(\omega^2_t -\omega^2_0 -i\omega_0\gamma ) } \mathcal{p}_a\end{aligned}\ ] ] in addition to the above material equations we must recall that the time derivative of the local electric field is directly proportional to the magnetization that results from the nanostructures , that is , @xmath219 thus , we have that the fourier transform of the local electric field can be written as the direct sum of the fourier transforms of the external electric field and polarization contribution from the active atomic inclusions , that is , @xmath220 where as before @xmath43 is the external magnetic field .
we now add to these slowly - varying equations those for the polarization and relative atomic population as determined in the previous case .
the difference we must keep in mind is that these amplifying structures are embedded into the nano - structures , and therefore the electric field in these equations is the local electric field rather than the external .
the slowly - varying governing equations for the amplifying material : @xmath221 where as before @xmath76 corresponds to the transition frequency and @xmath75 is carrier frequency .
thus , @xmath222 is the detuning . to simplify this expression
let us replace @xmath223 .
thus the slowly - varying envelope approximation for the governing equations are : @xmath218 \mathcal{h } = -\frac{i ( 4\pi)^2 \omega^3_0 \epsilon_0(\omega_0 ) \omega^2_t \alpha}{c^2(\omega^2_t -\omega^2_0 -i\omega_0\gamma ) } \mathcal{p}_a \\ \frac{\partial \mathcal{p}_a}{\partial t } = \frac{2}{\hbar}n ( \eta \mathcal{h } + \lambda \mathcal{p}_a ) - i(\omega_{12 } -\omega_0 ) \mathcal{p}_a \\
\frac{\partial n}{\partial t } = -\frac{1}{\hbar } ( ( \eta \mathcal{h } + \lambda \mathcal{p}_a)^ * \mathcal{p}_a + ( \eta \mathcal{h } + \lambda \mathcal{p}_a ) \mathcal{p}^*_a),\end{aligned}\ ] ] that is @xmath224 \mathcal{h } = \tilde{a } \mathcal{p}_a \\
\frac{\partial \mathcal{p}_a}{\partial t } = \frac{2}{\hbar}(\eta n \mathcal{h } + \lambda n \mathcal{p}_a ) - i(\omega_{12 } -\omega_0 ) \mathcal{p}_a \\
\frac{\partial n}{\partial t } = -\frac{1 } { \hbar } ( \eta^ * \mathcal{h}^ * \mathcal{p}_a + \eta \mathcal{h } \mathcal{p}^*_a + ( \lambda^*+\lambda)|\mathcal{p}_a|^2),\end{aligned}\ ] ] where @xmath225 is @xmath226 times the complex coefficient in front of the polarization from the previous equations .
this is done to make notation simpler . to determine the governing equations in dimensionless variables let us first multiply the first equation in this system of equations by @xmath227 and let us relabel @xmath228 for convenience .
thus we will have : @xmath224 \mathcal{h } = a \mathcal{p}_a \\
\frac{\partial \mathcal{p}_a}{\partial t } = \frac{2}{\hbar } ( n \mathcal{h } + \lambda n \mathcal{p}_a ) - i(\omega_{12 } -\omega_0 ) \mathcal{p}_a \\
\frac{\partial n}{\partial t } = -\frac{1 } { \hbar } ( \mathcal{h}^ * \mathcal{p}_a + \mathcal{h } \mathcal{p}^*_a + ( \lambda^*+\lambda)|\mathcal{p}_a|^2),\end{aligned}\ ] ] where @xmath229 .
now , let us write @xmath230 and introduce the following dimensionless variables : @xmath231 , @xmath232 , @xmath233 , @xmath234 , @xmath235 .
thus in dimensionless variables the governing equations read @xmath236 where @xmath237 , @xmath238 , @xmath239 , @xmath240 , @xmath241 , @xmath242 .
this is a new system of equations describing electromagnetic field interaction with nanostructures and the interaction of nanostructures with gain material via the induced local fields .
note that the usual conservation law holds in this case : @xmath243 .
to compute velocity of a steady wave solution we will use technique described in near the excited state @xmath244 .
this reduces the system of equations to @xmath245 in a co - moving frame @xmath246 , @xmath247 @xmath248 where @xmath249 .
the dispersion relation for this equation therefore reads : @xmath250 we know from the previous case that to obtain the velocity of the front we must find the zeros of the first derivative of the dispersion , which in this case are : @xmath251 using these to evaluate the imaginary part of the phase as before we obtain : @xmath252 = -\beta \lambda ' + \gamma \mp 2 \mathrm{re}[i\sqrt{1+i\tilde{\alpha } } ] \sqrt{\beta}\end{aligned}\ ] ] where @xmath253 .
thus , the velocity is given @xmath254+\sqrt{4(\mathrm{re}[i\sqrt{1+i\tilde{\alpha}}])^2 + 4 \lambda ' \gamma}}{2 \lambda ' } \end{aligned}\ ] ] if we expand around @xmath255 we see that up to second order @xmath256 } -\frac{\gamma^2 \lambda'}{8 \mathrm{re}[i\sqrt{1+i\tilde{\alpha}}]^3 } + \frac{\gamma^3 ( \lambda')^2}{16 \mathrm{re}[i\sqrt{1+i\tilde{\alpha}}]^5}\end{aligned}\ ] ] from this we see that when @xmath255 the velocity is the same as the most basic case if @xmath257 as well , that is @xmath258 .
in this paper we theoretically studied the propagation of a steady state ultrashort pulse in a dissipative medium .
we considered the case when the medium consists of lossy metallic nanostructures embedded to a gain material and the case when the gain material is embedded directly to a nanostructures .
we derived a system of governing equations and found analytic expressions for the velocity and spatial period of an optical pulse coupled with the polarization wave .
we also proposed a simple technique that describes the shape of this pulse using a simple system of ordinary differential equations .
we would like to thank andrei sarychev and alexander kildishev for enlightening discussions .
a.i.m appreciates support and hospitality of the university of arizona department of mathematics during his work on this manuscript .
this work was partially supported by nsf ( grant dms-0509589 ) , aro - muri award 50342-ph - mur and state of arizona ( proposition 301 ) , rfbr ( grant no .
09 - 02 - 00701-a ) .
s. v. manakov and v. g. kamensky , in nonlinear evolution equations : integrability and spectral methods , edited by a. degasperis , a. p. fordy , and m. lakshmanan , manchester university press , manchester , 1990 , p. 477 .
kildishev a.v .
yuan h .- k .
xiao s. drachev v. p. chettiar , u.k .
, , and v. m. shalaev .
dual - band negative index metamaterial : double negative at 813 nm and single negative at 772 nm . , 32:16711673 , 2007 . | in this paper we theoretically study propagation of steady state ultrashort pulse in dissipative medium .
we considered two cases ( i ) medium consists of lossy metallic nanostructures embedded into a gain material and ( ii ) the gain material is embedded directly into the nanostructures .
we found the shape and velocity of an optical pulse coupled with the polarization wave . |
the cross sections of hadronic collisions are important and fundamental quantities in high energy particle and nuclear physics and have been studied in the last 40 years in experiments covering many orders of magnitude in center - of - mass energies . in this report
we present two methods for measuring the inelastic proton - proton cross section with the cms detector .
data collected in 2010 were used in both analyses but with different luminosity and pile - up conditions .
low pile - up data were used to count events with activity in either of the hadron forward calorimeters ( hf ) in the first method presented . the other method used high luminosity data to count pile - up events in a given bunch crossing as a function of bunch luminosity and fitted with a poisson distribution to evaluate the cross section .
both measurements are corrected to hadron level definitions of the inelastic @xmath1 cross section . from these definitions
, extrapolations are performed to the total inelastic @xmath1 cross section with various monte carlo models also used in cosmic - rays physics .
the goal of the first method @xcite is to count events with as loose selection as possible to detect the largest possible cross section which translates to the smallest possible extrapolation . in the event
counting , we required one reconstructed energy deposit in either side of the hf calorimeters with at least 5 gev total energy .
the possible signal events were triggered by two proton bunches entering cms and the background was estimated from the number of selected events triggered by a single bunch entering cms . with the hf detectors only , more than one inelastic interaction in the same bunch- crossing can not be separated , which means the need of low pile - up data ( @xmath2 ) for small pile - up correction factor .
using generator level monte carlo , the inelastic interactions can be characterized with the variable @xmath3 .
the final state particles are ordered in rapidity and the largest rapidity gap is used to assign the particles into two systems .
the invariant masses of the two systems are calculated and the higher mass system is called @xmath4 , then the variable @xmath3 is given by @xmath5 . in case of single diffractive events , @xmath3 is the fractional momentum loss of the scattered proton . studying the hf selection efficiency as a function of @xmath3 showed that events with small @xmath3 can escape detection . for @xmath6 ,
the efficiency of detection is more than @xmath7 .
the definition of the inelastic @xmath1 cross section with @xmath6 is @xmath8 where @xmath9 is the number of events selected by the hf calorimeters after subtracting the background , @xmath10 is the fraction of selected events that are low mass ( contamination ) , @xmath11 is the pile - up correction factor , @xmath12 is the integrated luminosity and @xmath13 is the efficiency to detect an inelastic event with @xmath6 , namely the fraction of high mass events that are selected .
the efficiency ( @xmath14 ) and contamination ( @xmath15 ) correction factors were determined from three monte carlo generators using the full detector simulation of cms : pythia 6 @xcite , pythia 8 @xcite and phojet @xcite . to calculate the pile - up correction factor ( @xmath11 ) an iterative method was used where the average number of collisions per bunch - crossing , the pile - up ( @xmath16 ) was measured from the data directly .
several low pile - up datasets were used to obtain the cross section .
the integrated luminosity values for these datasets were obtained on the basis of van der meer scans , which carry a 4% normalization uncertainty dominating the uncertainties in the present analyses .
the result of the analysis for events with @xmath17 is obtained by averaging the 5 gev hf threshold cross section values measured at different pile - up conditions .
the systematic uncertainties of the result take into account noisy tower exclusion , run - by - run luminosity variations , varying the hf energy threshold and the model dependence .
the final result for the inelastic @xmath1 cross section with @xmath17 : @xmath18
the second method @xcite is based on the assumption that the number of inelastic @xmath1 interactions in a given bunch crossing follows a poisson probability distribution : @xmath19 where @xmath20 is the bunch crossing luminosity and @xmath21 is the total inelastic @xmath1 cross section .
two data samples , an inclusive di - electron and a single muon , were collected with the cms triggers .
the specific trigger requirements are not important as long as their efficiencies do not depend on the instantaneous luminosity .
using these data samples the number of reconstructed vertices were counted in each bunch crossing in bins of luminosity .
every vertex had to fulfill the quality requirements to be counted as a visible vertex .
the quality requirements are the following : the transverse position of the vertex between @xmath22 cm , the minimum distance between two vertices 0.1 cm , @xmath23 where the @xmath24 are the quality of the tracks associated with the vertex , at least 2 , 3 or 4 tracks with @xmath25 mev/@xmath26 in @xmath27 associated with the vertex and each track should have at least 2 pixel and 5 strip hits . to obtain the true number of vertices from the visible number of vertices a bin - by - bin correction factor was applied using the full simulation of the cms detector accounting for vertex reconstruction efficiency , vertex merging and fakes .
after the corrections applied , the distributions of the fraction of pile - up events as a function of the single bunch luminosity for @xmath28 pile - up events are well fitted with a poisson distribution .
the 9 values of the cross section are fitted together to get the visible inelastic @xmath1 cross section .
the main sources of systematic uncertainties are the use of different datasets , change in the fit limits , the minimum distance between vertices , the vertex quality requirement and the application of an analytic method for the corrections .
the final results with at least 2 , 3 or 4 charged particles with @xmath27 and @xmath29 mev/@xmath26 of the visible inelastic @xmath1 cross section are the following : @xmath30 @xmath31 @xmath32
additional monte carlo models were used for the extrapolation from the different hadron level definitions of the cross section to the total inelastic @xmath1 cross section .
the considered models were pythia 6 @xcite , pythia 8 @xcite , phojet @xcite , epos 1.99 @xcite , sybill 2.1 @xcite , qgsjet 1 and qgsjet - ii @xcite which use different models / tunings for the hard parton - parton and for the diffractive scattering cross sections .
every model show a similar trend for the measured hadron level cross sections but there are substancial differences in the expextations for the total inelastic @xmath1 cross section . for the extrapolation from the inelastic @xmath1 cross section with @xmath17 to the total inelastic @xmath1 cross section all generators were considered except qgsjet 1 since this model used in cosmic - rays physics is only describing one of the hemispheres of the collision , which is not suitable for efficiency calculations when a single energy deposit in either one of the two hf calorimeters is required .
the final result of the first analysis is @xmath33 the extrapolation from the visible cross section with at least 2,3 or 4 charged particles with @xmath25 mev/@xmath26 and @xmath27 has been computed using only the models in agreement with the measured points ( not phojet or sybill ) .
the final result of the second analysis is @xmath34 rt0.6
cms measured the inelastic @xmath1 cross section at @xmath35 tev with two independent method using two different subdetectors . the results for the inelastic @xmath1 cross section with @xmath17 are in very good agreement with the result of the atlas collaboration in the same kinematic range @xcite .
the extrapolations to the total inelastic @xmath1 cross section rely entirely on the models and their description of the low mass region . within the large extrapolation uncertainties the results of cms presented here are in agreement with the results from atlas , alice and totem collaborations as shown in figure [ fig : summary ] .
the author wishes to thank to the hungarian scientific research fund ( k 81614 ) for support .
cms collaboration , cms physics analysis summary , * cms - pas - qcd-11 - 002 * https://cdsweb.cern.ch/record/1433413[cdsweb.cern.ch/record/1433413 ] t. sjostrand , s. mrenna and p. z. skands , jhep * 0605 * ( 2006 ) 026 http://arxiv.org/abs/hep-ph/0603175[hep-ph/0603175 ] t. sjostrand , s. mrenna and p. z. skands , comput .
* 178 * ( 2008 ) 852 http://arxiv.org/abs/0710.3820[arxiv:0710.3820 ] r. engel , j. ranft and s. roesler , phys .
d * 52 * ( 1995 ) 1459 http://arxiv.org/abs/hep-ph/9502319[hep-ph/9502319 ] cms collaboration , cms physics analysis summary , * cms - pas - fwd-11 - 001 * https://cdsweb.cern.ch/record/1373466[cdsweb.cern.ch/record/1373466 ] k. werner , f. -m .
liu and t. pierog , phys .
c * 74 * ( 2006 ) 044902 http://arxiv.org/abs/hep-ph/0506232[hep-ph/0506232 ] e. -j .
ahn , r. engel , t. k. gaisser , p. lipari and t. stanev , phys .
d * 80 * ( 2009 ) 094003 http://arxiv.org/abs/0906.4113[arxiv:0906.4113 ] s. ostapchenko , phys .
d * 83 * ( 2011 ) 014018 http://arxiv.org/abs/1010.1869[arxiv:1010.1869 ] atlas collaboration , g. aad _ et al .
_ , nature commun .
* 2 * ( 2011 ) 46 http://arxiv.org/abs/1104.0326[arxiv:1104.0326 ] m. g. poghosyan , j. phys .
g g * 38 * ( 2011 ) 124044 http://arxiv.org/abs/1109.4510[arxiv:1109.4510 ] g. antchev , p. aspell , i. atanassov , v. avati , j. baechler , v. berardi , m. berretti and e. bossini _ et al .
_ , europhys .
* 96 * ( 2011 ) 21002 http://arxiv.org/abs/1110.1395[arxiv:1110.1395 ] | we present measurements of the total inelastic pp cross section at obtained with the cms detector .
two different methods are used . in runs with low event pile - up
, we exploit the large pseudorapidity coverage ( @xmath0 ) of the cms calorimeters to obtain the cross section for events with any activity in the acceptance range .
in addition , runs with high event pile - up are used by fitting a poisson distribution with the total visible cross section as parameter to the number of reconstructed primary vertices .
both measurements are corrected to a hadron level definition of the inelastic cross section . |
the understanding of hadron properties from an underlying lagrangian or hamiltonian is a major issue in nuclear and particle physics .
it demands both a relativistic framework to deal with quasi - massless particles ( the pion , up and down quarks , etc . ) or with high momentum and high energy experiments , and a nonperturbative framework .
the latter is mandatory in order to calculate for instance the mass of a bound state from the pole of the scattering amplitude or from an eigenstate equation .
one may already gain some physical insights from nonrelativistic studies like the nonrelativistic constituent quark model in particle physics , or the study of few - nucleon systems , based on the nonrelativistic nucleon - nucleon or three - nucleon potentials , in nuclear physics .
this has given rise to numerous studies in the last 40 years .
the extension to relativistic calculations may either rely on the use of relativistic equations like the four - dimensional bethe - salpeter equation ( including its various three - dimensional quasipotential reductions ) and the dyson - schwinger equation , or on hamiltonian dynamics , using one of its three forms proposed by dirac in 1949 @xcite .
we shall follow in this review the path pointed out by dirac and choose light - front dynamics ( lfd ) as a basis of our approach .
a natural testing ground for the use of lfd is the study of few - body systems .
the properties of this particular form of dynamics are indeed very well - suited to make a tight connection with nonrelativistic considerations . in the standard form of lfd
, the state vector of a physical system is defined not at a fixed moment of time but on the light - front plane given by the equation @xmath2 . the nonrelativistic limit reached by taking @xmath3 leads thus naturally to the ordinary equal - time formulation @xmath4 , giving rise , in particular , to the schrdinger equation for the nonrelativistic wave function .
one can always decompose the state vector in fock components .
since the physical vacuum is trivial in lfd , i.e. it coincides with the vacuum for not interacting particles , this decomposition does not include the vacuum fluctuations but contains the physical states only .
few - body systems are represented just by the first few components of this expansion . moreover
, each fock component has a probabilistic interpretation similar to that of the nonrelativistic wave function . in this formulation
, the elementary kernel of the eigenstate equation defining the state vector of the physical system is calculated in light - front time - ordered perturbation theory .
all four - momenta are on the mass shell , while all intermediate states are off the energy shell .
the eigenstate equation to be solved is then three - dimensional , in direct analogy with the nonrelativistic schrdinger equation . according to the dirac s classification
, the ten generators of the poincar group , given by space - time translations ( four generators ) , space rotations ( three generators ) , and lorentz boosts ( three generators ) , can be separated into kinematical and dynamical operators .
the kinematical operators leave the light - front plane invariant and are independent of dynamics , i.e. of the interaction hamiltonian of the system , while the dynamical ones change the light - front position and depend therefore on the interaction . among the kinematical operators , one finds , in lfd , the boost along the @xmath5 axis .
this property is of particular interest when one calculates electromagnetic observables at high momentum transfer , since once one knows the state vector in one reference frame , it is easy to calculate it in any other frame .
one has of course to pay some price for that .
the spatial rotations in the @xmath6 and @xmath7 planes become dynamical , in contrast to the case of equal - time dynamics .
this is a direct consequence of the violation of rotational invariance caused by the non - invariant definition of the light - front plane orientation .
this violation should be kept under control .
[ [ control - of - rotational - invariance . ] ] control of rotational invariance .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + while rotational invariance should be recovered automatically in any exact calculation , this is not _ a priori _ the case if the fock expansion is truncated .
the control of the violation of rotational symmetry is very difficult in practice , when using the standard form of lfd .
to avoid such an unpleasant feature of the latter , we shall use below the covariant formulation of lfd ( clfd ) @xcite , which provides a simple , practical , and very powerful tool in order to describe physical systems as well as their electromagnetic amplitudes . in this formulation
, the state vector is defined on the plane characterized by the invariant equation @xmath0{$\cdot$}}x=0 $ ] , where @xmath1 is an arbitrary light - like ( @xmath8 ) four - vector .
the standard lfd on the plane @xmath9 is recovered by considering the particular choice @xmath10 .
the covariance of our approach relies on the invariance of the light - front plane equation under any lorentz transformation of both @xmath1 and @xmath11 .
this implies in particular that @xmath1 can not be kept the same in any reference frame , as it is the case in the standard formulation of lfd .
there is of course equivalence , in principle , between the standard and covariant forms of lfd in any exact calculation . calculated physical observables must coincide in both approaches , though their derivation in clfd in most cases is much simpler and more transparent .
indeed , the relation between clfd and standard lfd reminds that between the feynman graph technique and old - fashioned perturbation theory .
in approximate calculations however , clfd has a definite advantage in the sense that it enables a direct handle on the contributions which violate rotational invariance .
these ones depend explicitly on the orientation of the light - front surface ( i.e. on @xmath1 ) and can thus be separated covariantly from true physical contributions .
this is of particular interest when one considers for instance electromagnetic observables in few - body systems @xcite , or field theory on the light front .
the interest to the application of lfd to field - theoretical problems is also very old .
it originates from the study of deep inelastic scattering experiments .
indeed , it is easy to see that in this kinematical domain , all events are close to the light - front plane .
this legitimates the use of the infinite momentum frame in the calculation of high energy observables @xcite .
the calculation is done in perturbation theory .
the extension to qed is natural , since perturbation theory is also applicable in that case @xcite .
the use of lfd in field theory is important in order to extend these calculations to the nonperturbative domain @xcite .
as already mentioned , the structure of the vacuum in lfd enables a well - defined expansion of the state vector in fock components . in field theory
, this expansion is in principle infinite .
from a practical point of view , it should be truncated to a limited number of components . we thus have to worry about the convergence of this expansion , and the various ways to speed it up if one wants to be able to make meaningful predictions .
the truncation of the fock expansion induces however two pernicious features in the study of field theory on the light front .
the first one , which we have already addressed in the discussion of few - body systems , is the violation of rotational invariance due to the particular choice of the orientation of the light - front plane .
the second one is the appearance of uncancelled divergences which calls for an appropriate renormalization scheme .
[ [ appropriate - renormalization - scheme . ] ] appropriate renormalization scheme .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the truncation of the fock expansion complicates the renormalization procedure , in contrast to that in standard perturbation theory .
indeed , the full cancellation of field - theoretical divergences which appear in a given fock sector requires taking into account contributions from other sectors .
if even a part of the latter is beyond our approximation , some divergences may leave uncancelled . for instance , looking at fig . [ self ] for the calculation of the fermion propagator in the second order of perturbation theory
, one immediately realizes that the cancellation of divergences between the self - energy contribution ( of 2nd order in the fock decomposition ) and the fermion mass counterterm ( of 1st order one ) involves two different fock sectors . .
[ self ] ] this means that , as a necessary condition for the cancellation of divergences , any mass counterterm should be associated with the number of particles present ( or `` in flight '' ) in a given fock sector .
in other words , all mass counterterms must depend on the fock sector under consideration , as advocated first in ref .
this is also true for the renormalization of the bare coupling constant .
the presence of uncancelled divergences reflects itself in possible dependence of approximately calculated observables on the regularization parameters ( e. g. , cutoffs ) .
in other words , calculated physical observables are not anymore scale invariant .
this prevents to make any physical predictions if we can not control the renormalization procedure in one way or another .
we have developed an appropriate renormalization procedure the so - called fock sector dependent renormalization ( fsdr ) scheme in order to keep the cancellation of field - theoretical divergences under permanent control .
this scheme relies also directly on clfd in order to define the renormalization conditions imposed on true physical observables @xcite .
we shall detail in the following the main features of clfd , as well as fsdr , and apply our general framework to various physical systems .
the state vector , @xmath12 , of a compound system corresponds to definite values for its mass @xmath13 , its four - momentum @xmath14 , and its total angular momentum @xmath15 with the projection @xmath16 onto the @xmath5 axis in the rest frame , i.e. , the state vector forms a representation of the poincar group .
it satisfies the following eigenstate equations : @xmath17 where @xmath18 is the four - momentum operator , @xmath19 is the pauli - lubanski vector @xmath20 and @xmath21 is the four - dimensional angular momentum operator which is represented as a sum of the free and interaction parts : @xmath22 in terms of the interaction hamiltonian @xmath23 we have @xmath24{$\cdot$}}x)\ d^4x.\ ] ] similarly to @xmath21 , the momentum operator also can be split into the free and interaction parts : @xmath25 with @xmath26{$\cdot$}}x)\,d^4x.\ ] ] from the general transformation properties of both the state vector and the light - front plane , it follows @xcite that @xmath27 where @xmath28 the equation ( [ kt12 ] ) is called the _
angular condition_. we can use it in order to replace the operator @xmath29 entering into eq .
( [ kt18 ] ) by @xmath30 . introducing the notations
@xmath31 we obtain , instead of eqs .
( [ kt16 ] ) and ( [ kt17 ] ) , @xmath32 _ these equations do not contain the interaction hamiltonian , once _
@xmath33 _ satisfies eqs . _
( [ kt14 ] ) _ and _ ( [ kt15 ] ) .
the construction of the state vector of a physical system with definite total angular momentum becomes therefore _ a purely kinematical problem_. indeed , the transformation properties of the state vector under rotations of the coordinate system are fully determined by its total angular momentum , while the dynamical part of the latter is separated out by means of the angular condition .
the dynamical dependence of the state vector on the light - front plane orientation turns now into its explicit dependence on the four - vector @xmath1 @xcite .
such a separation , in a covariant way , of kinematical and dynamical transformations is a definite advantage of clfd , as compared to standard lfd .
according to the general properties of lfd , we decompose the state vector of a physical system in fock sectors .
we have @xmath34 where @xmath35 is the state containing @xmath36 free particles with the four - momenta @xmath37 and @xmath38 s are relativistic @xmath36-body wave functions or the so - called fock components . here and below we will omit , for shortness , all spin indices in the notation of the state vector .
note the particular overall momentum conservation law given by the @xmath39-function .
it follows from the general transformation properties of the light - front plane @xmath0{$\cdot$}}x=0 $ ] under four - dimensional translations .
the quantity @xmath40 is a measure of how far the @xmath36-body system is off the energy shell .
it is completely determined by this conservation law and the on - mass - shell condition for each individual particle momentum .
we get @xmath41{$\cdot$}}p \ \tau_n=(s_n - m^2),\ ] ] where @xmath42 the phase space volume element is represented schematically by @xmath43 .
it involves integrations over the components of the constituent four - momenta @xmath44 and over @xmath45 in infinite limits .
the state @xmath46 can be written as @xmath47 where @xmath48 is a generic notation for free particle creation operators . to completely determine the state vector , we normalize it according to @xmath49 it is convenient to introduce , instead of the wave functions @xmath38
, the vertex functions @xmath50 ( which we will also refer to as fock components ) , defined by @xmath51{$\cdot$}}p \ \tau_n \phi_n.\ ] ] in the particular case of a fermion coupled to bosons , it is convenient to extract from @xmath50 the fermion bispinors , and make the replacement @xmath52 where @xmath53 is the four - momentum of the constituent fermion . when fock space is truncated , it is necessary to keep track of the order of truncation @xmath54 ( i.e. the maximal number of particles admitted in the fock sectors ) in the calculation of the vertex function . for this purpose we will use the notation @xmath55 for the @xmath36-body vertex function . in the lfd graph technique ,
it is represented by a @xmath56-leg vertex with one incoming double line corresponding to the physical state and @xmath36 outgoing single lines corresponding to constituents . by its spin structure and transformation properties
it is completely analogous to a @xmath36-body wave function . as an example , we show such a vertex in fig . [ gamman ] for the case of a physical fermion state composed from one constituent fermion and @xmath57 bosons . for simplicity
, we shall use below the same notation for the vertex functions of both fermion and boson physical states .
-body vertex function for a fock space truncation of order @xmath54 , for a physical fermion ( double straight line ) made of a constituent fermion ( single straight line ) coupled to bosons(wavy lines ) . ] with the decomposition ( [ fock ] ) , the normalization condition ( [ normep ] ) writes @xmath58 where @xmath59 is the contribution of the @xmath36-body fock sector to the full norm of the state vector , equal to 1 .
the explicit formulas for @xmath59 in terms of the vertex functions for some important particular cases can be found in @xcite .
the system of coupled equations for the fock components of the state vector can be obtained from eq .
( [ kt15 ] ) by substituting there the fock decomposition ( [ fock ] ) and calculating the matrix elements of the operator @xmath60 in fock space . with the expressions ( [ pshr ] ) and
( [ pintham ] ) , we get the eigenstate equation @xcite : @xmath61{$\cdot$}}p)\int \tilde{h}^{int}(\omega\tau)\frac{d\tau}{2\pi } \phi(p)= -\left[\left(\hat{p}^{(0)}\right)^2-m^{2}\right]\phi(p),\nonumber \\\end{aligned}\ ] ] where @xmath62 is the interaction hamiltonian in momentum space : @xmath63{$\cdot$}}x)\tau}d^4x.\ ] ] using the general momentum conservation law in eq .
( [ fock ] ) , we conclude that the operator in the square brackets on the right - hand side of eq .
( [ eq1b ] ) simply multiplies each fock component of the state vector by the factor @xmath64{$\cdot$}}p)\tau_n$ ] .
it is therefore convenient to introduce the notation @xmath65{$\cdot$}}p)\hat{\tau}\phi(p),\ ] ] where @xmath66 is the operator which , acting on a given component @xmath67 of @xmath68 , gives @xmath69 .
@xmath70 has a fock decomposition which is obtained from eq .
( [ fock ] ) by the replacement of the wave functions @xmath71 by the vertex functions @xmath72 .
we can thus cast the eigenstate equation in the form @xmath73\frac{d\tau}{\tau } { \cal g}(p).\ ] ] the physical mass @xmath13 of the compound system is found from the condition that the eigenvalue is 1 .
this equation is quite general and equivalent to the eigenstate equation ( [ kt15 ] ) .
it is nonperturbative .
as follows from the angular condition , the spin structure of the wave functions @xmath67 is very simple , since its construction does not require the knowledge of dynamics .
it should incorporate however @xmath1-dependent components .
it is convenient to decompose each wave function @xmath67 into invariant amplitudes constructed from all available particle four - momenta ( including the four - vector @xmath1 ! ) and spin structures ( matrices , bispinors , etc . ) . in the yukawa model for instance
, we have for the one- and two - body components @xcite : @xmath74{$\cdot$}}p}\right ] u(p ) , \label{onetwo}\end{aligned}\ ] ] since no other independent spin structures can be constructed .
here @xmath75 , @xmath76 , and @xmath77 are scalar functions determined by dynamics . for a spin
@xmath78 physical fermion composed from a constituent spin @xmath78 fermion coupled to scalar bosons , the number of invariant amplitudes for the two - body fock component coincides with the number of independent amplitudes of the reaction @xmath79 , which is @xmath80 , due to parity conservation .
similar expansions can be done for the three - body component of the same system or for fock components in qed , as we shall see in sec .
[ appli ] .
in the standard renormalization theory , the bare parameters are determined by fixing some physical quantities like the particle masses and the physical coupling constants .
the bare parameters are thus expressed through the physical ones .
this identification implies in fact the following two important questions which are usually never clarified in lfd calculations , but are at the heart of our scheme . _
( i ) _ in order to express the bare parameters through the physical ones , and vice - versa , one should be able to calculate observables or , in other words , physical amplitudes . in lfd ,
any physical amplitude is represented as a sum of partial contributions , each depending on the light - front plane orientation . since
an observable quantity can not depend on the latter , this spurious dependence must cancel in the whole sum , as already mentioned .
such a situation indeed takes place , for instance , in perturbation theory , provided the regularization of divergencies in lfd amplitudes is done in a rotationally invariant way @xcite . in nonperturbative lfd calculations , which are always approximate
, the dependence on the light - front plane orientation may survive even in calculated physical amplitudes .
for this reason , the identification of such amplitudes with observable quantities becomes ambiguous and expressing the calculated amplitudes through the physical parameters turns into a non - trivial problem .
_ ( ii ) _ the explicit form of the relationship between the bare and physical parameters depends on the approximation which is made .
this is a trivial statement in perturbation theory where the order of approximation is distinctly determined by the power of the coupling constant . in our nonperturbative approach based on the truncated fock decomposition of the state vector , an analogous parameter ( the power of the coupling constant ) is absent . at the same time , to make calculations compatible with the order of truncation , one has to trace somehow the level of approximation .
this implies that , on general grounds , the bare parameters should depend on the fock sector in which they are considered .
moreover , this dependence must be such that all divergent contributions are cancelled , as already mentioned in sec .
[ deltam ] .
how this should be done is the crucial point of the nonperturbative formulation of field theory on the light - front .
the use of our fsdr scheme in clfd is a unique opportunity to answer both questions . for clarity , we shall take , as a background , a model of interacting fermions and bosons like the yukawa model or qed , with the aim to calculate the physical fermion state vector .
the basis of fock space is formed by a set of fock sectors , each containing one constituent fermion and a certain number of bosons ( 0 , 1 , 2 , ... ) .
the truncation is made by retaining only those fock sectors where the maximal total number of constituent particles does not exceed @xmath54 .
the consideration of antiparticle degrees of freedom will be discussed in the simple case of a scalar model in sec .
[ scalar ] .
we will assume that the interaction hamiltonian @xmath23 is constructed through the bare fields satisfying the free dirac or klein - gordon equations with the corresponding _ physical _ masses , while the mass renormalization is performed by introducing , into @xmath23 , appropriate mass counterterms .
we emphasize that in spite of that we consider hereafter several particular forms of interaction , our scheme is applicable to physical systems with arbitrary interaction admitting a fock decomposition of the state vector .
the simple example of the renormalization of the fermion self - energy within the two - body fock space truncation , presented in sec .
[ deltam ] , can serve as a guideline to set up our general rule . in this example
, the mass counterterm should be labelled with a subscript and denoted by @xmath81 , in order to indicate that it is introduced to cancel , at @xmath82 , where @xmath83 is the constituent fermion mass , the self - energy contribution which belongs to the two - body fock sector .
let us denote by @xmath84 the mass counterterm in the most general case . since we truncate our fock space to order @xmath54
, one should make sure that , at any light - front time , the total number of particles is at most @xmath54 .
our first rule is thus : * _ in any amplitude where the mass counterterm @xmath85 appears , the value of @xmath86 is such that the total number of bosons in flight plus @xmath86 equals the maximal number of the fock sectors considered in the calculation , _ i.e. _ @xmath54 .
_ for instance , in the typical contribution indicated in fig
. [ gammadeltan ] , the mass counterterm is @xmath87 . for the mass counterterm of the lowest order
, we simply have @xmath88 since the fermion mass is not renormalized at all if the fermion can not fluctuate in more than one particle ! the general strategy we developed above for the calculation of the mass counterterm should be also applied to the calculation of the bare coupling constant , with however a bit of caution , since this one may enter in two different types of contributions .
the first one appears in the calculation of the state vector itself , when eq .
( [ eq3 ] ) is solved . in that case ,
any boson - fermion coupling constant is associated with the emission or the absorption of a boson which participates in the particle counting , in accordance with the rules detailed above , since it is a part of the state vector .
the second one appears in the calculation of the boson - fermion scattering amplitude or of the boson - fermion three - point green s function ( 3pgf ) like the electromagnetic form factor .
since the external boson is an ( asymptotic ) free field rather than a part of the state vector , the particle counting rule advocated above should therefore not include the external boson line .
following the same reasoning developed above for the calculation of mass counterterms , we can formulate the following general rule for the calculation of the bare coupling constants * _ in any amplitude which couples constituents inside the state vector one should attach to each vertex the internal bare coupling constant @xmath89 .
the value of @xmath86 is such that the total number of bosons in flight before ( after ) the vertex - if the latter corresponds to the boson emission ( absorption ) - plus @xmath86 equals the maximal number of the fock sectors considered in the calculation , it i.e. @xmath54 . _ the calculation of external bare coupling constants proceeds in the same spirit , with the final rule : * _ in any amplitude which couples constituents of the state vector with an external field , one should attach to the vertex involving this external field the external bare coupling constant @xmath90 .
the value of @xmath86 is such that , at the light - front time corresponding to the vertex , the total number of internal bosons in flight ( those emitted and absorbed by particles entering the state vector ) plus @xmath86 equals the maximal number of the fock sectors considered in the calculation , i.e. @xmath54 . _ the lowest order bare coupling constants are @xmath91 the first one is trivial , because no fermion - boson interaction is allowed in the one - body fock space truncation .
the second one reflects the fact that the external bare coupling constant , in the same approximation , is not renormalized at all , since a single fermion can not be `` dressed '' .
some illustrations of the rules concerning the internal and external bare coupling constants are given in fig .
[ gammag0n ] . ]
though we relied on the fermion - boson model when considering the above fsdr procedure , the latter can be easily extended to other systems with additional counterterms and bare parameters .
once proper bare coupling constants and mass counterterms have been identified , one should fix them from a set of renormalization conditions .
in perturbation theory , there are three types of quantities to be determined : the mass counterterms , the bare coupling constants , and the norms of the fermion and boson fields .
usually , the on - mass - shell renormalization is applied , with the following conditions . for each field ,
the mass counterterm is fixed from the requirement that the corresponding two - point green s function has a pole at @xmath92 , where @xmath83 is the physical mass of the particle .
the field normalization is fixed from the condition that the residue of the two - point green s function at the pole is @xmath93 .
the bare coupling constant is determined by requiring that the on - mass - shell 3pgf is given by the product of the physical coupling constant and the elementary vertex .
the renormalization conditions in lfd are of slightly different form , although they rely on the same grounds .
the fermion mass counterterm is fixed from the eigenvalue equation ( [ kt15 ] ) by demanding that the mass , @xmath13 , of the physical state is identical to the mass of the constituent fermion which we called @xmath83 .
solving a similar equation for the physical boson state vector allows us to find the boson mass counterterm from quite analogous requirements .
the determination of the internal bare coupling constant needs some care .
it is found , as in perturbation theory , by relating the on - energy - shell two - body vertex function @xmath94 to the physical coupling constant @xmath95 . as follows from the momentum conservation law ,
taking @xmath94 on the energy shell is equivalent to setting @xmath96 , or @xmath97 . in order to fix the relationship between @xmath94 and @xmath95 one needs to take into account the renormalization factors coming from radiative corrections to all legs of the two - body vertex function @xcite .
these factors do also depend on the order of the fock space truncation , as detailed in @xcite . in the case of a fermion coupled to bosons , without polarization effects generated by antifermion contributions
, this relationship reads @xmath98 where @xmath99 is the one - body contribution to the norm of the state vector , as given in eq .
( [ inor ] ) , calculated for the fock space truncation of order @xmath100 .
( [ gamma2r ] ) admits simple physical interpretation .
each leg of the on - energy - shell two - body vertex function contributes for an individual factor @xmath101 to the physical coupling constant , where @xmath102 is the field strength normalization factor @xcite .
the physical fermion state is normalized to @xmath93 , so that its factor @xmath102 equals @xmath93
. the constituent boson line is not renormalized - since we do not consider antifermions - so that for the bosonic line we also have @xmath103 .
if we did not neglect the antifermionic degrees of freedom , it would contribute by a non - unity factor @xmath104 .
finally , we have shown in @xcite that the field strength normalization factor of the constituent fermion is just the weight of the one - body component in the norm of the physical state , i.e. @xmath105 . according to our fsdr scheme , the normalization factor of the constituent fermion should correspond to the fock space truncation of order @xmath100 , since , by definition , the two - body vertex function contains one extra boson in flight in the final state . the condition ( [ gamma2r ] ) has two important consequences .
the first one is that the two - body vertex function at @xmath106 should be independent of the four - vector @xmath1 which determines the orientation of the light - front plane . with the spin decomposition ( [ onetwo ] ) , this implies that the component @xmath107 at @xmath106 should be identically zero . while this property is automatically verified in the case of the two - body fock space truncation - if using a regularization scheme which does not violate rotational invariance - this is not guaranteed for calculations within higher order truncations .
indeed , nothing prevents @xmath94 to be @xmath1-dependent , since it is an off - shell object , but this dependence must completely disappear on the energy shell , i.e. for @xmath106
. it would be so if no fock space truncation occurs .
the latter , in approximations higher than the two - body one ( i. e. for @xmath108 ) , may cause some @xmath1-dependence of @xmath94 even on the energy shell , which immediately makes the general renormalization condition ( [ gamma2r ] ) ambiguous .
if so , one has to insert new counterterms into the light - front interaction hamiltonian , which explicitly depend on @xmath1 and cancel the @xmath1-dependence of @xmath109 .
note that the explicit covariance of clfd allows to separate the terms which depend on the light - front plane orientation ( i.e. on @xmath1 ) from other contributions and establish the structure of these counterterms .
this is not possible in ordinary lfd .
we should thus enforce the condition ( [ gamma2r ] ) by introducing , into the interaction hamiltonian , appropriate @xmath1 dependent counterterms .
for instance , in the yukawa model without antifermion contributions and within the three - body fock space truncation ( see below , sec .
[ yukawa3 ] ) , we need one additional counterterm of the form @xcite @xmath110{$\cdot$}}\partial } \psi\varphi,\ ] ] where @xmath111 is a constant adjusted to make eq .
( [ gamma2r ] ) true , @xmath112 is the fermion ( boson ) field , and @xmath113{$\cdot$}}\partial)$ ] is the reversal derivative operator , fully analogous to the operator @xmath114 in ordinary light - front dynamics . the second consequence of eq .
( [ gamma2r ] ) is that the two - body vertex function at @xmath106 should be a constant .
this is a non - trivial requirement , since @xmath94 , as well as the components @xmath76 and @xmath107 in eq .
( [ onetwo ] ) , do depend on two invariant kinematical variables which are usually chosen as the longitudinal constituent momentum fraction , @xmath11 , and the square of its transverse momentum , @xmath115 .
the on - mass - shell condition @xmath116 where @xmath117 is the constituent boson mass , fixes one variable ( say , @xmath118 ) only , while the second variable ( @xmath11 ) remains arbitrary .
hence , under the condition ( [ s2 ] ) , if we assume @xmath118 fixed , @xmath94 should be independent of @xmath11 .
again , this property is verified in the two - body fock space truncation , since in this approximation our equations are equivalent to perturbation theory of order @xmath119 .
it is not guaranteed for higher order calculations .
we shall come back to this point in sec .
[ appli ] . in practice ,
we shall fix @xmath109 at some preset value @xmath120 and verify that the physical observables are not sensitive to the choice of @xmath120 . to summarize , we can thus list the normalization conditions in clfd , for a calculation done in a fock space truncation of order @xmath54 , without considering fermion - antifermion polarization corrections : _ _ * the mass counterterm @xmath121 is fixed by solving the eigenstate equation ( [ eq3 ] ) in the limit @xmath122 . *
the state vector is normalized according to the standard condition ( [ inor ] ) . * the internal bare coupling constant @xmath123 is fixed from the condition that the @xmath1-independent part of the two - body vertex function at @xmath124 and at a fixed value of @xmath11 , denoted by @xmath120 , is given by the right - hand side of eq .
( [ gamma2r ] ) . * the external bare coupling constant @xmath125 is fixed from the condition that the @xmath1-independent part of the on - energy - shell 3pgf is proportional to the elementary vertex , with the proportionality coefficient being the physical coupling constant . * the @xmath1-dependent counterterms in the hamiltonian
are fixed by the conditions that the @xmath1-dependent parts of the two - body vertex function and the 3pgf , both taken on the energy shell , turn to zero . *
the values of all bare parameters and counterterms for @xmath126 are determined from successive calculations within the @xmath127 fock space truncations .
the explicit solution of the eigenvalue equation ( [ eq3 ] ) requires to define first a regularization scheme , since loop contributions are _ a priori _ divergent .
we shall use in the following applications the pauli - villars ( pv ) regularization scheme , first applied to lfd in @xcite .
it has the nice feature of being rotationally invariant @xcite , and it can be implemented rather easily in calculations within the two- and three - body fock space truncations @xcite .
besides that , in this regularization scheme , all contact interactions inherent to lfd are absent .
it however necessitates to extend fock space in order to embrace pv fermions and pv bosons on equal grounds with the physical particles . before discussing the calculation of the properties of compound systems , it is instructive to look at the structure of the fermion self - energy in the simple yukawa model in clfd .
the results are very similar to those for qed @xcite .
since our formalism is explicitly covariant , we can write down immediately the general structure of the self - energy of a fermion with the off - shell four - momentum @xmath128 ( @xmath129 ) . it writes @xmath130{$\cdot$}}k}+ { \cal c}_1\sigma,\ ] ] where @xmath131{$\cdot$}}k)$ ] .
the coefficients in this expansion depend on @xmath132 only .
they are given by @xmath133,\\ \label{abc2 } { \cal b}(k^2)&=&\frac{m}{4 \omega{\makebox[0.08cm]{$\cdot$}}k } \mbox{tr}\left[\sigma({{\not\!}k}){{\not\!}\omega}\right],\\ \label{abc3 } { \cal c}(k^2)&=&\frac{1}{4m}\mbox{tr}\left[\sigma({{\not\!}k})\left({{\not\!}k}-{\not\!}{\omega}\frac{k^2}{\omega{\makebox[0.08cm]{$\cdot$}}k}\right)\right],\\ \label{c } { \cal c}_1(k^2 ) & = & \mbox{tr}\left[\sigma({{\not\!}k})\sigma\right].\end{aligned}\ ] ] in the two - body approximation , when @xmath134 is entirely given by the loop diagram shown in fig . [ self ] , the coefficient @xmath135 is identically zero .
the coefficient @xmath136 should also be zero since the two - point green s function should be equivalent to the one calculated in the four - dimensional feynman approach , and is therefore independent of @xmath1 provided one uses a rotationally invariant regularization @xcite .
note that the coefficient @xmath136 is not _ a priori _ chiral invariant in the sense that if the mass of the constituent fermion goes to zero , @xmath136 does not vanish , in contrast to @xmath137 and @xmath138 , as it should . using the pv regularization scheme , @xmath137 and @xmath138 depend logarithmically on the pv boson mass . without regularization ,
the coefficient @xmath139 diverges quadratically at high momenta .
it is however identical to zero when the pv regularization scheme is used with one pv fermion and one pv boson only .
this simple case provides a good starting point to understand how our general framework should be applied in practice .
the eigenstate equation one has to solve is shown graphically in fig .
this equation is written in accordance with the prescriptions of our fsdr scheme .
note the use of the fock sector dependent mass counterterms and the bare coupling constants .
the two - body vertex function writes in that case @xmath140u(p),\ ] ] where @xmath141 is the polarization vector of the photon . to simplify notations
, we omit the superscript `` ( 2 ) '' at @xmath142 .
we shall use in the following the feynman gauge .
the number of independent invariant amplitudes for this vertex function coincides with that for the reaction @xmath143 .
however , one should take into account that in the feynman gauge the vector boson wave function has _
four _ independent components .
so , the total number of invariant amplitudes is @xmath144 .
we choose the following set of invariant amplitudes @xcite : @xmath145{$\cdot$}}p } + b_3\frac{m{\not\!}{\omega}\gamma^{\mu}}{\omega{\makebox[0.08cm]{$\cdot$}}p}+ b_4\frac{m^2{\not\!}{\omega}\omega^{\mu}}{(\omega{\makebox[0.08cm]{$\cdot$}}p)^2}\nonumber \\ & & + b_5\frac{p^{\mu}}{m}+b_6\frac{k_1^{\mu}}{m}+ b_7\frac{m{\not\!}{\omega}p^{\mu}}{\omega{\makebox[0.08cm]{$\cdot$}}p } + b_8\frac{m{\not\!}{\omega}k_1^{\mu}}{\omega{\makebox[0.08cm]{$\cdot$}}p}.\end{aligned}\ ] ] in the two - body approximation , and using the pv regularization scheme , we find @xmath146 where @xmath75 is defined in eq .
( [ oneone ] ) .
these components refer to the physical ones .
the ones associated with pv bosons and/or fermions can be found easily @xcite . in this approximation ,
the vertex function @xmath147 is a constant matrix proportional to @xmath148 . with these results
, one can calculate the norms of the fock sectors entering into the normalization condition ( [ inor ] ) : @xmath149 where the expression for @xmath150 can be found in @xcite .
it depends logarithmically on the mass of the pv boson used to regularize the loop integral .
the normalization condition thus fixes @xmath75 : @xmath151 the renormalization condition ( [ gamma2r ] ) enables us to calculate @xmath152 as a function of the physical coupling constant denoted by @xmath153 .
this condition , for @xmath154 , writes simply @xmath155 since @xmath156 . with eq .
( [ psi1 ] ) we get @xmath157 and @xmath158 with all these quantities , one can easily calculate the electromagnetic form factors given by the diagrams shown in fig .
note that since the state vector is normalized to @xmath93 , we find for the external bare coupling constant : @xmath159 at zero momentum transfer , the pauli form factor gives the anomalous magnetic moment of the electron .
we get @xcite @xmath160 this expression exactly coincides with the well - known schwinger term . one may wonder why our nonperturbative approach exactly recovers the perturbative result @xcite .
the reason is simple .
for the two - body fock space truncation we consider here , the irreducible contributions to the two - point green s function or to the electromagnetic form factors are just identical to the corresponding ones in the second order of perturbation theory . and their re - summation to all orders of the coupling constant just defines the physical mass of the electron , which is also the same , by construction , in both approaches .
our result ( [ schwinger ] ) is also independent of the pv fermion and boson masses , when they tend to infinity .
this is of course necessary in order to preserve the scale invariance of physical observables .
note that at large enough pv masses , the one - body part @xmath99 of the normalization condition is negative , while the two - body part @xmath161 exceeds @xmath93 .
this implies also that @xmath162 is negative .
these features are just an artefact of the regularization scheme which is used .
they show that indeed @xmath99 and @xmath161 , and , more generally , any @xmath59 in eq .
( [ inor ] ) , as well as the bare coupling constants , are not physical observables and can therefore be scale dependent .
finding the state vector for the three - body fock space truncation is the first non - trivial nonperturbative calculation .
we begin our consideration of the three - body approximation with the study of a toy model : a heavy scalar boson @xmath163 with mass @xmath83 , interacting with light scalar bosons @xmath164 with mass @xmath117 .
we will calculate the state vector of the heavy boson and represent the former as a sum of the following fock sectors , in symbolic form : @xmath165 where @xmath166 denotes the antiparticle .
the interaction hamiltonian is @xmath167 where @xmath168 and @xmath169 are the mass counterterms which , in this model , have the dimension of mass squared .
since the physical coupling constant @xmath95 has the dimension of mass , it is convenient to define a dimensionless coupling constant @xmath170 .
the system of eigenstate equations for the vertex functions is represented graphically in fig .
[ syseq3 ] .
the two three - body vertices @xmath171 and @xmath172 are trivially expressed through the two - body vertex @xmath173 . due to this fact
, we can obtain a closed equation involving @xmath173 only . in this model , we encounter two types of irreducible divergent contributions : the heavy and light boson self - energies , as shown in fig .
[ bbse ] . both of them diverge logarithmically at high momenta ; the divergences are expected to be cancelled by the corresponding mass counterterms .
the renormalization condition should now incorporate the light boson field strength normalization factor . instead of eq .
( [ gamma2r ] ) , we have @xmath174 where the factor @xmath175 stands for the combined contribution of the heavy and light boson strength normalization factors , calculated according to the fsdr prescription , extended to include antiparticle degrees of freedom @xcite .
this factor is completely finite in a pure scalar system , and is expressed in terms of derivatives of the heavy and light boson self - energies .
the eigenstate equation for @xmath173 with the renormalization condition ( [ eq4 - 3 ] ) has been solved numerically in the limit @xmath122 for the following set of parameters : @xmath176 , @xmath177 , @xmath178 .
note that this value of @xmath179 is far from the perturbative domain .
for convenience , @xmath173 is parameterized as a function of two variables , @xmath180 and @xmath11 .
the physical domain corresponds to @xmath181 , @xmath182 .
the renormalization point is defined by @xmath124 and @xmath183 , where we chose @xmath184 .
we show in fig .
[ g2sx ] the characteristic dependence of @xmath173 on one of its arguments , while the second argument is fixed .
in contrast to the two - body case , where @xmath185 is a constant , now it exhibits rather nontrivial behavior , especially as a function of @xmath180 at fixed @xmath11 . at fixed @xmath180 ( upper plot ) and as a function of @xmath180 at fixed @xmath11 ( lower plot ) .
] to check the consistency of the renormalization condition ( [ eq4 - 3 ] ) , we show in fig . [ g2x_st ] the dependence of @xmath173 on @xmath11 at fixed @xmath124 .
it is distinctly seen that if the fock space includes all the four fock sectors , eq .
( [ eq4 - 1 ] ) , @xmath186 is almost a constant , as it ought to be .
if one however neglects one of the three - body sectors , either @xmath187 or @xmath188 , @xmath186 strongly depends on @xmath11 , which makes the renormalization condition ambiguous because physical results turn out to be sensitive to the choice of @xmath120 . ) at @xmath124 as a function of @xmath11 .
the solid line is the calculation with all the four fock sectors , eq .
( [ eq4 - 1 ] ) , in the state vector .
the dash - dotted and dotted lines are the same but without the @xmath187 or @xmath188 sector , respectively . ]
the above result is very non - trivial .
let us consider , for instance , a part of the three - body contributions to @xmath173 calculated in perturbation theory , i.e. in terms of the two - body vertex function @xmath185 found in the lower , two - body , approximation , as shown in fig .
[ anti ] . their sum ( but not each of them ! ) is exactly @xmath11-independent at @xmath124 , since @xmath185 is a constant .
this statement follows from the fact that the sum of the two contributions on the energy shell coincides with the corresponding on - mass - shell feynman amplitude , and the latter is a constant . in our calculation of order @xmath189 , the contributions analogous to those in fig .
[ anti ] , but with @xmath173 instead of @xmath185 appear in the eigenstate equation and determine the @xmath11-dependence of @xmath190 . from fig .
[ g2x_st ] it is seen that the latter dependence is surprisingly weak , though @xmath173 is not a constant , as fig .
[ g2sx ] clearly demonstrates . at the same time , if one calculates the amplitudes of the diagrams in fig .
[ anti ] with an arbitrary chosen function @xmath191 instead of @xmath185 , their sum would hardly be a constant at @xmath124 .
such a result may indicate that the set of contributions we considered in the case of the three - body fock space truncation forms an almost consistent set , in the sense that it meets all consistency requirements .
this constitutes a basis of a well - controlled , nonperturbative fock expansion of the the state vector , in the same spirit as an expansion in @xmath119 in perturbation theory . the yukawa model is much closer to reality ( e.g. for instance to qed ) than the scalar model discussed in the previous section .
simultaneously , it is much more involved from the point of view of renormalization , because of complicated spin structure of the vertex functions and stronger divergences to be regularized and renormalized . in this section
, we apply our fsdr renormalization scheme to the yukawa model within the three - body fock space truncation , but , in order to avoid extra complications , without incorporating antifermion degrees of freedom .
our purpose is to demonstrate the capabilities of the fsdr scheme in solving a true nonperturbative problem for particles with spin .
for example , in contrast to the two - body fock space truncation ( in this approximation the yukawa model is almost equivalent to qed , as considered in sec .
[ qed2 ] ) , where we have only one irreducible contribution to the fermion self - energy , we now should consider all the graphs for the self - energy , which contain one fermion and two bosons in intermediate states , including overlapping self - energy type diagrams .
the number of such irreducible graphs is infinite .
some of them are shown in fig .
the solution of the eigenstate equation for the state vector automatically generates these contributions to all orders of the coupling constant .
, width=226 ] we start with the following fock decomposition for the state vector @xmath192 where the symbols @xmath193 and @xmath164 denote , respectively , the constituent fermion and boson with masses @xmath83 and @xmath117 .
the interaction hamiltonian reads @xmath194 where @xmath195 is defined by eq .
( [ zomega ] ) .
the regularization is done in a rotationally invariant way by introducing one pv fermion with mass @xmath196 and one pv boson with mass @xmath197 .
the system of equations for the vertex functions in graphical representation can be obtained from that for the scalar case , shown in fig .
[ syseq3 ] , by changing @xmath198 , @xmath199{$\cdot$}}p$ ] and by setting @xmath200 , @xmath201 , since we do not consider here the dressing of the boson line .
after that , the expressions for the light - front diagrams are written according to the clfd graph technique rules , taking into account that certain lines correspond now to a particle with spin 1/2 .
expressing @xmath171 through @xmath173 , as in the scalar case , we get a closed matrix equation for the two - body vertex .
its spin structure is written similarly to eq .
( [ onetwo ] ) : @xmath202{$\cdot$}}p}\right]u(p),\ ] ] where @xmath203 and @xmath204 are scalar functions . for clarity ,
we indicate here only the physical components . the pv fermion and boson components can be calculated as well @xcite .
the three - body component is completely determined by four scalar functions , like , for instance @xmath205 where these functions are denoted @xmath206 , and @xmath207 are basis spin structures .
it is convenient to construct the the latter ones as follows @xcite : @xmath208{$\cdot$}}p } } , \nonumber \\
s_2&=&\displaystyle{\frac{m{\not\!}\omega}{\omega{\makebox[0.08cm]{$\cdot$}}p } } , \nonumber\\ s_3&=&ic_{ps}\left[2x_1 -(1-x_1)\displaystyle{\frac{m{\not\!}\omega}{\omega{\makebox[0.08cm]{$\cdot$}}p}}\right]\gamma_5 , \nonumber\\ s_4&=&i\,c_{ps}\displaystyle{\frac{m{\not\!}\omega}{\omega{\makebox[0.08cm]{$\cdot$}}p}}\gamma_5\end{aligned}\ ] ] with @xmath209{$\cdot$}}k_1}{\omega { \makebox[0.08cm]{$\cdot$}}p}$ ] , while @xmath210 is the following pseudoscalar : @xmath211{$\cdot$}}p } e^{\mu\nu\rho\gamma}k_{2\mu}k_{3\nu}p_{\rho}\omega_{\gamma}.\ ] ] the function @xmath210 can only be constructed with four independent four - vectors .
this is the case in lfd for @xmath212 . in the nonrelativistic limit
, one would need @xmath213 .
we can then construct two additional spin structures @xmath214 and @xmath215 of the same parity as @xmath216 and @xmath217 by combining @xmath210 with parity negative matrices constructed from @xmath216 , @xmath217 , and @xmath218 matrices . in our computational procedure
we take the limit @xmath219 analytically and then study the limit @xmath220 numerically . as a result ,
the calculated vertex functions depend parametrically on @xmath197 .
the main question we are interested in concerns the behavior of observables as a function of @xmath197 : do they remain finite and physically reasonable at @xmath221 or not ? for this purpose , we calculate the fermion anomalous magnetic moment , using the state vector ( [ eq4 - 5 ] ) . it can be extracted from the fermion - boson 3pgf in the three - body approximation , as shown in fig .
the detail of the calculation can be found in @xcite . we just recall here the main numerical results . the anomalous magnetic moment is calculated for a typical set of physical parameters @xmath222 gev , @xmath223 gev , and two values of the coupling constant @xmath224 and @xmath225 .
this mimics , to some extent , a physical nucleon coupled to scalar `` pions '' .
the typical pion - nucleon coupling constant is given by @xmath226 where @xmath227 is a typical momentum scale , and @xmath228 and @xmath229 are the axial coupling constant and the pion decay constant , respectively . for @xmath230
gev we just get @xmath231 . , for two different values of the coupling constant , @xmath232 ( upper plot ) and @xmath225 ( lower plot ) .
the dashed and dotted lines are , respectively , the two- and three - body contributions , while the solid line is the total result .
the value of the anomalous magnetic moment calculated in the @xmath154 approximation is shown by the thin line on the right axis.,title="fig:",width=321 ] , for two different values of the coupling constant , @xmath232 ( upper plot ) and @xmath225 ( lower plot ) .
the dashed and dotted lines are , respectively , the two- and three - body contributions , while the solid line is the total result . the value of the anomalous magnetic moment calculated in the @xmath154 approximation is shown by the thin line on the right axis.,title="fig:",width=321 ] we plot in fig .
[ amm ] the anomalous magnetic moment as a function of @xmath233 $ ] , for the two different values of @xmath179 mentioned above .
we show also on each of these plots the value of the anomalous magnetic moment calculated for the @xmath154 truncation , which coincides with the anomalous magnetic moment obtained in the second order of perturbation theory . the results for @xmath232 show rather good convergence as @xmath221 .
the contribution of the three - body fock sector to the anomalous magnetic moment is sizeable but small , indicating that the fock decomposition ( [ fock ] ) converges rapidly .
this may show that once higher fock components are small , we can achieve a practically converging calculation of the anomalous magnetic moment .
note that this value of @xmath179 is not particularly small : it is about 30 times the electromagnetic coupling , and is about the size of the typical pion - nucleon coupling in a nucleus .
when @xmath179 increases , we see that the contribution of the three - body sector considerably increases . for @xmath234 the three - body contribution to the anomalous magnetic moment
starts to dominate at large values of @xmath197 .
the dependence of the anomalous magnetic moment on the pv boson mass @xmath197 becomes more appreciable , although it keeps rather small .
, of the state vector , as a function of the pv boson mass @xmath197 , for @xmath235 ( upper plot ) and @xmath234 ( lower plot).,title="fig:",width=321 ] , of the state vector , as a function of the pv boson mass @xmath197 , for @xmath235 ( upper plot ) and @xmath234 ( lower plot).,title="fig:",width=321 ] in order to have a more physical insight into the relative importance of different fock sectors in the decomposition ( [ fock ] ) for the state vector , we plot in fig .
[ n123 ] the contributions of the one- , two- , and three - body fock sectors to the norm of the state vector for the two values of the coupling constant , considered in this work .
we see again that at @xmath236 the three - body contribution to the norm is small , while it is not negligible and increases with @xmath197 , when @xmath234 .
according to renormalization theory , the pv boson mass should be much larger than any intrinsic momentum scale present in the calculation of physical observables . with this limitation
, physical observables should be independent of any variation of the pv boson mass , within an accuracy which can be increased at will .
this is what we found in our numerical calculation for small enough values of @xmath179 .
( upper plot ) and @xmath204 ( lower plot ) of the two - body vertex function calculated at @xmath124 , as a function of @xmath11 , for @xmath232 ( dashed line ) and @xmath237 ( solid line ) , for a typical value of @xmath238 gev.,title="fig:",width=321 ] ( upper plot ) and @xmath204 ( lower plot ) of the two - body vertex function calculated at @xmath124 , as a function of @xmath11 , for @xmath232 ( dashed line ) and @xmath237 ( solid line ) , for a typical value of @xmath238 gev.,title="fig:",width=321 ] in order to understand the possible origin of the residual dependence of the anomalous magnetic moment on @xmath197 , we plot in figs . [ b1on ] the two - body spin components @xmath203 and @xmath204 calculated at @xmath124 , as a function of @xmath11 . as we already mentioned in sec
. [ renorcond ] , the spin components for physical particles @xmath239 and @xmath240 should be independent of @xmath11 in an exact calculation . moreover , @xmath240 should be zero .
it is here fixed to zero at a given value of @xmath241 , by the adjustment of the constant @xmath111 .
we clearly see in these plots that the on - shell @xmath203 is not a constant , although its dependence on @xmath11 is always weak , while @xmath204 is not identically zero , although its value is relatively smaller than that of @xmath203 for @xmath236 , and starts to be not negligible for @xmath237 .
a similar situation is observed in the scalar case , when we remove `` by hands '' , from the state vector , the fock sector with the antiparticle ( see the dash - dotted curve in fig .
[ g2x_st ] ) .
it is instructive to study the properties of @xmath242 in perturbation theory .
this can be done by calculating the amplitudes of the diagrams shown in fig .
[ anti ] ( but for fermions , of course ! ) . note that in our fock space truncation ( [ eq4 - 5 ] ) , the contribution of the @xmath243 intermediate state ( the left diagram in fig .
[ anti ] ) is automatically taken into account by the solution of the eigenstate equation , while the contribution of the @xmath244 state ( the right diagram in the same figure ) is absent .
if one calculates the sum of both contributions within perturbation theory @xcite , one finds @xmath245 this is a first indication that the expected properties of the on - shell functions @xmath246 are indeed recovered , when antifermion degrees of freedom are involved .
this is also a confirmation of similar features found in sec .
[ scalar ] for the scalar system .
the calculation of baryon properties within the framework of chiral perturbation theory is a subject of active theoretical developments .
since the nucleon mass is not zero in the chiral limit , all momentum scales are _ a priori _ involved in the calculation of baryon properties ( like masses or electro - weak observables ) beyond tree level .
this is at variance with the meson sector for which a meaningful power expansion of any physical amplitude can be done . while there is not much freedom , thanks to chiral symmetry , for the construction of the effective lagrangian in chiral perturbation theory in terms of the pion field or more precisely in terms of the u field defined by @xmath247 , where @xmath248 is the pion decay constant and @xmath249 are the pauli matrices , one should settle an appropriate approximation scheme in order to calculate baryon properties . up to now , two main strategies have been adopted .
the first one is to force the bare ( and hence the physical ) nucleon mass to be infinite , in heavy baryon chiral perturbation theory @xcite . in this case , by construction , an expansion in characteristic momenta can be developed .
the second one is to use a specific regularization scheme @xcite in order to separate contributions which exhibit a meaningful power expansion , and hide the other parts in appropriate counterterms . in both cases
however , the explicit calculation of baryon properties relies on an extra approximation in the sense that physical amplitudes are further calculated by expanding the effective lagrangian , denoted by @xmath250 , in a finite number of pion fields . moreover , it has recently been realized that the contribution of pion - nucleon resonances , like the @xmath251 and roper resonances , may play an important role in the understanding of the nucleon properties at low energies @xcite .
these resonances are just added `` by hand '' in the chiral effective lagrangian .
this is also the case for the most important @xmath252 resonances , like the @xmath16 and @xmath253 resonances .
since in the chiral limit the pion mass is zero , any calculation of @xmath254 systems demands a relativistic framework to get , for instance , the right analytical properties of the physical amplitudes .
the calculation of compound systems , like a physical nucleon composed of a bare nucleon coupled to many pions , relies also on a nonperturbative eigenstate equation . while the mass of the system can be determined in leading order from the iteration of the @xmath254 self - energy calculated in the first order of perturbation theory , as indicated in fig .
[ self_gen](a ) , this is in general not possible , in particular , for @xmath254 irreducible contributions , as shown in fig .
[ self_gen](b ) . ] the general framework we have developed above is particularly suited to deal with these requirements .
this leads to the formulation of light - front chiral effective field theory ( lf@xmath255eft ) @xcite with a specific effective lagrangian @xmath256 .
the decomposition of the state vector in a finite number of fock components implies to consider an effective lagrangian which enables all possible elementary couplings between the pion and nucleon fields compatible with the fock space truncation .
this is indeed easy to achieve in chiral perturbation theory , since each derivative of the u field involves one derivative of the pion field . in the chiral limit ,
the chiral effective lagrangian of order @xmath14 , @xmath257 , involves @xmath14 derivatives and therefore at least @xmath14 degrees of the pion field . in order to calculate the state vector in the @xmath54-body approximation , with one fermion and @xmath258 pions
, one has therefore to include contributions up to @xmath259 pion fields in the effective lagrangian , as shown in fig .
[ counting ] .
we thus should calculate the state vector in the @xmath54-body approximation with an effective lagrangian denoted by @xmath260 and given by @xmath261 pion fields in the initial and final states . [ counting ] ]
while the effective lagrangian in lf@xmath255eft can be mapped out to the cpt lagrangian of order @xmath14 , the calculation of the state vector does not rely on any momentum decomposition .
it relies only on an expansion in the number of pions in flight at a given light - front time . in other words , it relies on an expansion in the fluctuation time , @xmath262 , of such a contribution . from general arguments , the more particles we have at a given light - front time , the smaller the fluctuation time is . at low energies ,
when all processes have characteristic interaction times larger than @xmath262 , this expansion should be meaningful .
it is interesting to illustrate the general features of lf@xmath255eft calculations . at order @xmath154
, we already have to deal with irreducible contributions , as shown in fig . [ self_gen](b ) .
it leads to non - trivial renormalization of the coupling constant .
the calculation at order @xmath189 incorporates explicitly contributions coming from @xmath263 interactions , as well as all low energy @xmath254 resonances , like the @xmath251 or roper resonances , as shown schematically in fig .
( [ delta ] ) . indeed , in the @xmath264 fock sector , the @xmath254 state can couple to both @xmath265 as well as @xmath266 states .
we can generate therefore all @xmath254 resonances in the intermediate state without the need to include them explicitly , provided the effective lagrangian has the right dynamics to generate these resonances .
this is the case , by the construction , in cpt .
resonances and @xmath251 as well as roper nucleon resonances . [ delta ] ] preliminary results obtained by using the pv regularization scheme can be found in @xcite .
the understanding of the properties of relativistic compound systems from an elementary hamiltonian in nuclear and particle physics demands to develop a nonperturbative framework .
this framework should include a well - defined strategy for approximate calculations of these properties and a systematic way to improve the accuracy .
we have described in this review a general framework based on light - front dynamics . in this scheme ,
the state vector of any system of interacting particles is decomposed in fock components . since for obvious practical reasons this decomposition should be truncated to take into account a finite number of fock components ,
we have shown how to control in a systematic way the convergence of such expansion .
our formalism relies , first , on the covariant formulation of light - front dynamics , and , second , on a systematic nonperturbative renormalization scheme in order to avoid any uncancelled divergences .
the applications we have presented on qed , on a purely scalar system , on the yukawa model , and on chiral effective field theory on the light - front have shown the flexibility , the real advantages , and the nice features of our formalism .
several developments will soon achieve to settle a complete framework to deal with field theory on the light front .
this includes first the full account of antiparticle degrees of freedom in order to recover , order by order in the fock expansion , the scale invariance of physical observables for arbitrary values of the coupling constant .
this scale invariance should be checked within a given regularization scheme .
we have used up to now the pauli - villars regularization scheme . while this scheme is systematic and
can be applied to a variety of physical systems , it may be cumbersome to implement from a numerical point of view in higher order calculations , since it involves many non - physical components including pauli - villars fields .
it also demands to perform calculations with very large mass scales ( the pauli - villars fermion and boson masses ) .
the use of the taylor - lagrange renormalization scheme @xcite has proven to be a very natural scheme in light - front dynamics .
it should be applied to more involved calculations .
finally , it is now well - recognized that the understanding of spontaneous symmetry breaking phenomena on the light - front can be achieved by taking into account nonperturbative zero mode contributions to field operators @xcite .
it remains to include these zero modes in the general framework we developed in this review .
three of us ( v.a.k . , a.v.s . and
are sincerely grateful for the warm hospitality of the laboratoire de physique corpusculaire , universit blaise pascal , in clermont - ferrand , where a part of the present study was performed . this work has been supported by grants from cnrs / in2p3 and the russian academy of science .
p. a. m. dirac , rev .
* 21 * , 392 ( 1949 ) v.a .
karmanov , zh .
. fiz . * 71 * , 399 ( 1976 ) ; [ transl .
: sov . phys .
jetp 44 , 210 ( 1976 ) ] j. carbonell , b. desplanques , v.a .
karmanov and j .- f .
mathiot , phys .
reports * 300 * , 215 ( 1998 ) s. fubini and g. furlan , physics * 1 * , 229 ( 1965 ) ; + l. susskind and g. frye , phys . rev . * 164 * , 2003 ( 1967 ) ; + l. susskind , phys . rev . * 165 * , 1535 ( 1968 ) s.j .
brodsky , r. roskies and r. suaya , phys .
* d8 * , 4574 ( 1973 ) s.j .
brodsky , h.c .
pauli and s.s .
pinsky , phys . rep . *
301 * , 299 ( 1998 ) r. perry , a. harindranath and k. wilson , phys .
lett . * 65 * , 2959 ( 1990 ) v.a .
karmanov , j .- f .
mathiot and a.v .
smirnov , phys .
* d77 * , 085028 ( 2008 ) v.a .
karmanov , zh .
. fiz . * 83 * , 3 ( 1982 ) [ trans.:jetp * 56 * , 1 ( 1982 ) ] d. bernard , th .
cousin , v.a .
karmanov and j .- f .
mathiot , phys . rev . *
d65 * , 025016 ( 2001 ) v.a .
karmanov , j .- f .
mathiot , a.v .
smirnov , phys .
* d75 * , 045012 ( 2007 ) j. hiller and s.j .
brodsky , phys .
* d59 * , 016006 ( 1999 ) v.a .
karmanov , j .- f .
mathiot , and a.v .
smirnov , in preparation v.a .
karmanov , nucl .
a644 * , 165 ( 1998 ) .
karmanov , j .- f .
mathiot , and a.v . smirnov .
arxiv hep - th/1006.5282 m.e .
peskin , d.v .
schroeder , an introduction to quantum field theory
, perseus books , ( 1995 ) s.j .
brodsky , j.r .
hiller and g. mccartor , phys .
d 64 * , 114023 ( 2001 ) v.a .
karmanov , j .- f .
mathiot , and a.v .
phys . rev . * d69
* , 045009 ( 2004 ) s. chabysheva and j. hiller , phys . rev .
* d 79 * , 114017 ( 2009 ) v.a .
karmanov , j .- f .
mathiot and a.v .
smirnov , nucl .
* b * ( proc . suppl . )
* 199 * , 35 ( 2010 ) e. jenkins and a.v .
manohar , _ phys .
* b255 * , 558 ( 1991 ) r.j .
ellis and h.b .
tang , _ phys .
_ * c57 * ( 1998 ) 3356 ; + t. becher and h. leutwyler , _ eur .
phys . j. _
* c9 * ( 1999 ) 643 e. jenkins and a.v .
manohar , _ phys .
_ * b259 * ( 1991 ) 353 j .- f .
mathiot , _ chiral effective field theory on the light - front _ , contribution to the workshop international workshop on effective field theories : from the pion to the upsilon , valencia ( spain ) , february 2009 , proceedings of science , sissa , trieste , pos(eft09)014 n. tsirova , v.a .
karmanov , j .- f .
mathiot , _ baryon structure in chiral effective field theory on the light - front _ , contribution to the workshop
chiral dynamics 2009 , bern ( switzerland ) , july 2009 , proceedings of science , sissa , trieste , pos(cd09)093 p. grang and e. werner , _ quantum fields as operator valued distributions and causality _ , arxiv : math - ph/0612011 and nucl .
b , proc . supp .
* 161 * ( 2006 ) 75 + p. grang , j .- f .
mathiot , b. mutet , e. werner , phys . rev .
* d80 * ( 2009)105012 ; phys .
d82 * ( 2010 ) 025012 t. heinzl , s. krusche and e. werner , phys .
lett . * 256 * , 55 ( 1991 ) ; + t. heinzl , s. krusche , s. simbrger , e. werner , z. phys . ,
* c56 * , 415 ( 1992 ) | we present a general framework to calculate the properties of relativistic compound systems from the knowledge of an elementary hamiltonian .
our framework provides a well - controlled nonperturbative calculational scheme which can be systematically improved .
the state vector of a physical system is calculated in light - front dynamics . from the general properties of this form of dynamics
, the state vector can be further decomposed in well - defined fock components . in order to control the convergence of this expansion
, we advocate the use of the covariant formulation of light - front dynamics . in this formulation
, the state vector is projected on an arbitrary light - front plane @xmath0{$\cdot$}}x=0 $ ] defined by a light - like four - vector @xmath1 .
this enables us to control any violation of rotational invariance due to the truncation of the fock expansion .
we then present a general nonperturbative renormalization scheme in order to avoid field - theoretical divergences which may remain uncancelled due to this truncation .
this general framework has been applied to a large variety of models . as a starting point , we consider qed for the two - body fock space truncation and calculate the anomalous magnetic moment of the electron . we show that it coincides , in this approximation , with the well - known schwinger term .
then we investigate the properties of a purely scalar system in the three - body approximation , where we highlight the role of antiparticle degrees of freedom . as a non - trivial example of our framework
, we calculate the structure of a physical fermion in the yukawa model , for the three - body fock space truncation ( but still without antifermion contributions ) .
we finally show why our approach is also well - suited to describe effective field theories like chiral perturbation theory in the baryonic sector . |
measurements of the spectra of high energy @xmath3rays , neutrinos and cosmic rays have emerged as a useful constraint on particle physics theories which predict topological defects ( for recent summaries see e.g. refs .
although it does not appear likely that the emission of high energy particles by topological defects can explain the observed spectra , the current data can be used as upper bounds to constrain theories of particle physics beyond the standard model . of particular interest for cosmology
are theories giving rise to cosmic strings .
some time ago , the spectrum of cosmic rays from non - superconducting strings was computed @xcite under the assumption of the _ standard _ scaling picture @xcite emerging from studies of cosmic string dynamics in an expanding universe ( see refs .
@xcite for recent reviews ) . in the standard picture
, the long string network evolves into string loops which then decay predominantly by gravitational radiation .
it was found that for strings with @xmath4 ( required for strings to be relevant for cosmic structure formation @xcite ) , where @xmath5 is the mass per unit length in the string and @xmath6 is the newton gravitational constant , the predicted particle fluxes are substantially lower than the observational detections or limits .
recently , however , vincent , hindmarsh and sakellariadou @xcite have challenged the standard picture of string evolution .
they claim that the small - scale structure on the strings does not scale with the expansion of the universe . in the _ vhs _ scenario ,
the long strings lose their energy directly into particles instead of string loops .
this leads to a greater production rate of particles and hence to the expectation of greater fluxes of @xmath3rays , neutrinos , and cosmic rays at earth .
expanding on these points further , detailed numerical simulations @xcite have demonstrated that in an expanding universe , the network of _ long _ cosmic strings ( _ long _ meaning with curvature radius greater than the hubble radius ) approaches a scaling solution in which the number of long string segments crossing each hubble radius approaches a constant value @xmath7 . in the vhs scenario , the string network maintains this scaling solution by the direct radiation of scalar and gauge particles . in the standard scenario ,
the string network maintains the scaling solution by continuously giving off part of its energy in the form of string loops with radius smaller than the hubble radius .
based on certain theoretical arguments ( see e.g. @xcite ) it is expected that the distribution of loops in the standard scenario also eventually takes on a scaling solution , i.e. when all lengths are scaled to the hubble radius , the statistical properties of the distribution of string loops are independent of time . however , the resolution of string network simulations ( nambu action strings ) is not yet good enough to be able to verify the standard scenario .
moreover , recent field theory simulations @xcite provide some evidence that the string defects give off their energy not in the form of string loops , but directly by scalar and gauge particle radiation , thus corroborating the vhs scenario ( these latter simulations remain controversial @xcite ) . in the standard cosmic string scenario ,
the string loops oscillate due to the relativistic tension , and decay slowly by the emission of gravitational radiation .
only a small fraction of the energy is released in the standard scenario in the form of scalar and gauge particle radiation .
this occurs via the process of cusp annihilation @xcite , during the final loop collapse or by the evaporation of black holes created from string loops @xcite .
it has been shown @xcite that for gut - scale ( @xmath4 ) strings the flux of ultra - high energy ( uhe ) particles in the standard picture from these mechanisms is at or below the relevant observational measurements or limits
. the predicted flux increases , and may be observable , if the symmetry breaking scale is significantly lower . in the vhs scenario , in contrast ,
all of the string energy is released directly from the long strings as scalar and gauge particle radiation .
additionally , the predicted flux increases as the symmetry breaking scale increases .
hence , the flux of high energy @xmath3rays , neutrinos and cosmic rays in the vhs scenario is expected to be much larger than the fluxes from loops in the standard scenario .
indeed , it has been remarked @xcite that the fluxes from gut scale strings in the vhs scenario will be above the observational limits . in this paper , we point out that this limit comes only from the neutrino emission at high energy and the cascades produced by the electromagnetic emission , whilst the nearest expected vhs string is too distant to produce an observable flux of cosmic rays or @xmath8-rays . in the vhs scenario the highest energy fluxes ( above the gzk cutoff ) of cosmic rays and @xmath8-rays are suppressed due to the interaction of these particles with the photons of the cosmic radiation backgrounds .
the result is that no isotropic flux of cosmic rays and @xmath8-rays from a network of vhs strings is expected to be observed .
however , the neutrino emission from low @xmath9 vhs strings may produce an observable signal .
we present a detailed calculation in this paper of the neutrino , photon and cosmic ray fluxes for a range of string and decay scales .
we also show that the predictions depend sensitively on the initial energy @xmath1 characterizing the decay of the particle radiation , which in turn depends on the presently unknown physics in the regime between the scale of electroweak symmetry breaking and the scale of string formation .
we find that decreasing @xmath1 leads to a decrease in the upper cutoff in the predicted spectrum and an increase in the cosmic ray fluxes at intermediate energies .
thus , even the additional freedom of decreasing @xmath1 can not make gut strings obeying vhs dynamics consistent with the observational constraints . nevertheless , in the event that a segment of long string is substantially closer than the expected average distance to the nearest string , a highly anisotropic flux of cosmic rays protons , antiprotons and @xmath8-rays could be observed above the gzk cutoff in the vhs scenario .
this possibility , however , is not supported by the current observations of events above the gzk cutoff which exhibit large scale isotropy and at most only small scale clustering in arrival angle @xcite . in this paper
we perform for the first time the calculation of the tau neutrinos directly produced by the hadronic decays of the x - particles .
the calculation of the direct @xmath10 spectrum has been not previously been included in the analysis of any cosmic string scenario .
this was because it was assumed that the @xmath10 spectrum was highly suppressed at all energies compared with the @xmath11 and @xmath12 spectra .
instead only the much weaker @xmath10 spectrum generated by the cascade of the x - decay @xmath11 and @xmath12 off the cosmic relic neutrino background was described . recently
, however , macgibbon , wichoski , and webber @xcite have shown that , in hadronic jets at accelerator energies , the tau neutrinos are a significant fraction of the total neutrino spectrum at high @xmath13 , where @xmath13 is the ratio of the neutrino energy to the jet energy . extrapolating their fragmentation function parametrization for @xmath10 to uhe energies ,
we show here that , in the vhs and all other cosmic string scenarios , the directly produced tau neutrinos generate a signal at earth comparable at the highest energies with the expected @xmath11 and @xmath12 signal .
this flux is orders of magnitude greater than the contribution from the cascade @xmath10 .
one consequence of our result is that the detection of a significant fraction of @xmath10 in uhe neutrino events may be due to hadronic decay at the source and not @xmath14 oscillation in transit . in the following sections , we present the neutrino , photon and cosmic ray spectra for vanilla emission taking @xmath15 and omitting susy particles in the jet decay .
however , we discuss in section [ sec : ext ] the effects of including a susy sector , non - zero neutrino mass , neutrino oscillation or other extensions to the standard model . in the appendix , we perform the full computation of the flux of secondary muon and electron neutrinos resulting from muon decay in the particle jets .
we use units in which @xmath16 , unless otherwise stated , and take a hubble constant of @xmath17 .
if the network of long strings scales , then the corresponding energy density @xmath18 at time @xmath19 is @xmath20 where @xmath5 is the mass per unit length in the string .
this equation defines the constant @xmath7 which counts the number of strings per volume @xmath21 in the scaling solution .
the value of @xmath7 can be determined from simulations of cosmic string network evolution .
current work @xcite gives @xmath22 . in the vhs scenario ,
the scaling solution is maintained not by the production of string loops , but by the radiation of gauge and scalar particles ( which we collectively call x - particles in the following ) .
making use of the equation of state of a string , the continuity equation becomes @xmath23 where @xmath24 and @xmath25 are the number density and mass , respectively , of the x - particles .
@xmath25 is effectively the symmetry breaking scale at which the string formed , @xmath26 ( e.g. @xmath27 gev for @xmath4 ) . strictly speaking , @xmath28 with the exact factor depending on the form of the potential of the scalar field .
the decay of the high energy x - particles will lead to the production of jets analogous to the qcd jets seen in accelerators . to simplify the discussion we assume that all jets in the x - decay have the same initial energy , @xmath1 . in this case
, the decay of a single x - particle will lead to @xmath29 jets , and the number density of jets generated by the energy release of long strings is @xmath30 the largest uncertainty in the calculation of the flux of high energy particles generated by strings comes from our ignorance of the structure of the jets at ultra - high initial energies . jet production has been studied in detail in qcd . from such studies , the fragmentation function of jets into photons , neutrinos and baryons
is known @xcite , at least to a good approximation , up to a few tev . however , in our case the particles producing the jets are superheavy scalar and gauge particles .
following refs .
@xcite , the fragmentation functions of such jets are usually calculated by extrapolating the qcd fragmentation functions to higher energies ( in which case @xmath1 is taken to be @xmath31 ) .
this is appropriate if there is no new physics which does not match this extrapolation between the electroweak symmetry breaking scale and the unification scale ( the scale at which the defects are produced ) , but it is not justified if there is new physics which does not sufficiently match the extrapolation .
we will parametrize ignorance about new physics and its effects on the jets from the x - particle decay by introducing a new scale @xmath1 ( which may be the scale of the new physics ) as the scale to which we can satisfactorily extrapolate the fragmentation functions from the qcd regime .
we will then assume that the initial scalar or gauge x - particle produces @xmath29 jets with initial energy @xmath1 each .
( in reality , there may be intermediate decay steps between the @xmath25 and @xmath1 decay scales . ) while this may or may not be the precise description of the decay , it will at least help further quantify the dependence of the final fluxes on the uncertainties in the fragmentation process .
the initial jet particle decays into quarks and leptons on a time scale of @xmath32 , where @xmath33 is the coupling constant associated with the physics at the energy scale @xmath1 .
the quarks then hadronize on a strong interaction time scale .
more than 90% of the total energy in a qcd jet goes into pions , with the majority of the remainder going into baryons ( mainly neutrons , antineutrons , protons and antiprotons ) . on astrophysical time scales , the neutral pions decay into two photons , the charged pions decay into electron and muon neutrinos and electrons or positrons , and the neutrons decay into protons and leptons .
in addition , a relatively small but still significant number of tau neutrinos are produced by the decay of heavy quarks and tau leptons .
non - jet leptons @xmath34 produced by the decay @xmath35 may also influence the observable @xmath8-ray flux below @xmath36 ev ( see section [ sec : obs ] ) .
the distribution of energies @xmath37 of the primary qcd jet decay products ( predominantly equal numbers of @xmath38 and @xmath39 ) can be approximated by the fragmentation function @xcite @xmath40 where @xmath41 is the fraction of the jet energy carried by the decay product and @xmath37 continues down to @xmath42 ev .
the two body decays of the primary @xmath38 or @xmath39 decay particles lead , on the average , to two photons and two primary muon neutrinos for every three pions .
the energy distribution of the decay products can be obtained by integrating ( [ primdecay ] ) from @xmath43 to @xmath44 with the invariant measure @xmath45 , @xmath46\ , , \ ] ] where the constant @xmath47 depends on the decay process being considered ( see below ) .
( [ secdecay ] ) applies to the final spectrum of photons produced in the jet .
the photons result from @xmath48 decay . in this case , both of the decay particles have vanishing rest mass , and it is possible for a single photon to carry away the entire pion energy in the lab frame .
hence , @xmath49 for the photons .
electron and muon neutrinos result from the @xmath50 and @xmath39 decay . in the decay process @xmath51 , one of the decay products ( the muon or antimuon ) has nonvanishing rest mass . in this case ,
the integration limits when integrating ( [ primdecay ] ) are not @xmath13 and @xmath44 ( see e.g. @xcite and @xcite ) .
the correction to the upper integration end for @xmath52 and @xmath12 is negligible , but not so the change in the lower integration limit for @xmath12 which becomes @xmath43 with @xmath53 when @xmath54 mev , the charged pion mass , and @xmath55 mev is the muon mass . equation ( [ secdecay ] ) describes the @xmath12 and @xmath56 neutrinos immediately produced by the decay of the charged pions . the final spectrum of @xmath12 and @xmath56 neutrinos , however , is comprised of the @xmath12 and @xmath56 neutrinos produced by the subsequent decay of the @xmath57 and @xmath58 , together with the contribution from those immediately produced by the decay of the @xmath50 and @xmath39 .
the final spectrum of @xmath52 and @xmath59 neutrinos is produced by the decay of the muons and antimuons .
taking into consideration the spin polarization of the muons , we show in the appendix that the final spectra of the @xmath52 and @xmath12 neutrinos generated by jet decay are approximately @xmath60 \ ; , \qquad \quad e_{\nu_{\mu } } \le \frac{m_j}{\alpha } \ ; , \end{aligned}\ ] ] for the muon neutrinos produced in the first stage of the pion decay @xmath61 \ ; , \qquad \quad e_{\nu_{\mu } } \le m_j \ ; , \end{aligned}\ ] ] for the muon neutrinos produced by the decay of muons in the pion decay , and @xmath62 \ ; , \qquad \quad e_{\nu_{e } } \le m_j \ ; , \end{aligned}\ ] ] for the final electron neutrinos . equation ( [ primdecay ] ) also applies to the primary baryonic decay products such as protons and neutrons . assuming that 3% of the total energy of the jet goes into baryons and antibaryons and noting that most of the energy of decaying neutrons and antineutrons is transferred to daughter protons and antiprotons , the final distribution of protons and antiprotons is approximately @xmath63 equation ( [ primdecay ] ) is one approximation to the numerically computed qcd fragmentation functions , taken from ref .
it is derived simply by requiring that the number of decay products scales as @xmath64 and applying energy conservation and scale invariance .
further the ratio of nucleon to pions is assumed at all @xmath13 to equal the ratio of total nucleons to total pions per jet . even with these reservations
, we find that eqs .
( [ appres1 ] ) ( [ appres2 ] ) match the @xmath52 and @xmath12 spectra generated in 10 tev @xmath65 events simulated by the herwig monte carlo code @xcite to within a factor of 2 in the range @xmath66 @xcite .
there are other analytical approximations of the fragmentation functions used in the literature .
another form is @xcite @xmath67 which can be well approximated by @xcite @xmath68 for values of @xmath13 between @xmath69 .
a third formula is based on the modified leading logarithmic approximation ( mlla ) @xcite of qcd . at small @xmath13 ,
the expression becomes @xcite @xmath70 \ , , \ ] ] where @xmath71 @xmath72 and @xmath73 gev .
conservation of total energy in the jet @xmath74 gives the normalization constant @xmath75 . in all approximations , extrapolations and monte carlo simulations of the fragmentation functions at high energies ,
the fraction of the jet energy carried by nucleons ( or equivalently the nucleon multiplicity ) lies in the range 310% but the precise value and dependence on @xmath13 and @xmath1 is unknown @xcite .
[ figfragfunc ] shows a comparison of the three fragmentation functions ( [ primdecay ] ) , ( [ ff2 ] ) and ( [ ff3 ] ) .
the difference between these functions is small in the energy range @xmath76 , and will be unimportant for order of magnitude considerations .
hence , in the following we will use mostly approximation ( [ primdecay ] ) .
we stress however that no extrapolated fragmentation function should be regarded as accurate to more than an order of magnitude at highest and lowest @xmath13 .
investment in precision in those regions of the predicted spectra is unjustified and misleading .
for the first time in cosmic string emission analysis , we also include the distribution for the tau neutrinos produced in the jet decays , which had erroneously been assumed to be negligible in all previous work .
while the number of @xmath10 produced per @xmath77 tev jet is less than @xmath78 of the number of @xmath11 and @xmath12 , the high energy @xmath10 are predominantly produced by the initial decays of the heavier quarks with shorter lifetimes and the @xmath11 and @xmath12 are produced by the final state cluster decays of the much lighter pions .
this leads to significantly greater relative contribution from @xmath10 at high @xmath13 than previously assumed .
the fragmentation distribution ( [ primdecay ] ) is no longer relevant for the tau neutrino .
instead macgibbon , wichoski , and webber @xcite find that the fragmentation function for @xmath10 production in 300 gev - 100 tev @xmath65 events , as simulated by herwig @xcite can be parametrized as @xmath79 \ ; .\ ] ] consistent results are obtained from jet decay events simulated by pythia / jetset @xcite .
the expressions for the number density of jets ( [ jetdensity ] ) , and for the energy distribution of the jet decay products , ( [ secdecay ] ) , ( [ appres1 ] ) , ( [ appres1b ] ) , ( [ appres2 ] ) , ( [ protonflux ] ) or ( [ nutaufrag ] ) , can be convolved to obtain the expected flux @xmath80 of high energy photons , neutrinos , and cosmic rays with energy @xmath37 from a vhs string distribution @xmath81}{d e } \ ; .\ ] ] in eq .
( [ flux ] ) , particles observed today ( @xmath82 ) with energy @xmath37 were produced at a time @xmath83 with energy @xmath84 $ ] , where @xmath85 denotes the redshift at time @xmath83 .
the factor @xmath86 expresses the dilution of the particle number density in an expanding universe .
the lower cutoff time @xmath87 corresponds to the maximal redshift from which particles of present - day energy @xmath37 can reach us .
this cutoff can be due either to interactions with the ambient extragalactic and galactic media during propagation , or to the constraint that the initial energy @xmath88 must be less than the initial jet energy @xmath1 .
the upper cutoff time @xmath89 corresponds to the latest time that the emission from the vhs cosmic string network can be considered isotropic ( see section [ sec : obs ] ) .
thus our upper cutoff time is that from which emission will have just isotropized by today . for protons , in order to calculate the jacobian @xmath90}{d e}$ ]
we apply the continuous energy loss approximation ( cel ) and solve eq .
( [ flux ] ) numerically .
the use of the cel approximation is acceptable in this case where the distance to the sources is much larger than the attenuation length for the particle in the cosmic radiation backgrounds ( see section [ sec : obs ] ) . for photons ,
if electromagnetic cascades are not taken yet into account ( see below ) , we obtain ( see eq .
( [ secdecay ] ) with @xmath49 ) @xmath91 \ , , \end{aligned}\ ] ] for neutrinos , if cascade off the relic neutrino background is not included , we have ( see eq .
( [ appres2 ] ) ) @xmath92 \ , , \end{aligned}\ ] ] for electron neutrinos and ( see eqs .
( [ appres1 ] ) and ( [ appres1b ] ) ) @xmath93 \\ \mbox{for } \quad e \le \frac{m_j}{\alpha } \ ; , \\ \\
\frac{1}{4 \pi } \ , \frac{5}{2 } \nu \mu t_0^{-2 } m_j \bigl(\frac{e}{m_j}\bigr)^{3 } \biggl [ { 2.96}\ , ( ( z_{max } + 1 ) - ( z_{min } + 1 ) ) \\ - \ , { 0.24}\ , ( ( z_{max } + 1)^{- 1/2 } - ( z_{min } + 1)^{- 1/2 } ) \bigl(\frac{e}{m_j}\bigr)^{- 3/2 } \\ - \ , { 2.90}\ , ( ( z_{max } + 1)^{1/2 } - ( z_{min } + 1)^{1/2 } ) \bigl(\frac{e}{m_j}\bigr)^{- 1/2 } \\ - \ , { 1.22}\ , ( ( z_{max } + 1)^{3/2 } - ( z_{min } + 1)^{3/2 } ) \bigl(\frac{e}{m_j}\bigr)^{1/2 } \\
+ \ , { 0.08}\ , ( ( z_{max } + 1)^{3 } - ( z_{min } + 1)^{3 } ) \bigl(\frac{e}{m_j}\bigr)^{2 } \\ - \ , { 0.01}\ , ( ( z_{max } + 1)^{4 } - ( z_{min } + 1)^{4 } ) \bigl(\frac{e}{m_j}\bigr)^{3 } \biggr ] \\ \mbox{for } \quad \frac{m_j}{\alpha } < e \le m_j \ ; , \end{array } \right.\ ] ] for muon neutrinos . similarly for the tau neutrinos ,
the flux from the string distribution neglecting cascading is ( see eq .
( [ nutaufrag ] ) ) @xmath94 \ , , \end{aligned}\ ] ] in eqs .
( [ fluxsec][nutauflux ] ) , @xmath95 is the redshift corresponding to @xmath89 and @xmath96 is the redshift corresponding to @xmath97 : @xmath98 where @xmath99 denotes the redshift cutoff due to astrophysical interactions . for convenience ,
we have presented eqs .
( [ fluxsec ] [ nutauflux ] ) for the case @xmath100 . in our figures ,
we have calculated the more general result for @xmath101 where appropriate . the first important conclusion to draw from eqs .
( [ fluxsec][numuflux ] ) is that for energies @xmath37 substantially smaller than the effective initial jet energy @xmath1 , the fragmentation term proportional to @xmath102 similarly dominates the flux of photons , electron and muon neutrinos giving @xmath103 in the case of the tau neutrinos the fragmentation term proportional to @xmath104 dominates the flux so that eq .
( [ nutauflux ] ) gives , @xmath105 hence , for fixed cosmic string mass per unit length @xmath5 , the spectrum of high energy photons and neutrinos increases as the jet energy @xmath1 decreases .
this occurs because for smaller values of @xmath1 there are more jets emitted from the string and the fragmentation function as a function of @xmath37 is more compressed .
these two effects overcome the @xmath106 or @xmath107 factor of the dominant term of the fragmentation function .
we also note that the fluxes in the vhs scenario increase as @xmath9 increases , or equivalently the symmetry breaking scale increases , in contrast to the general decrease seen in the standard cosmic string scenario . to complete the calculation of the photon , neutrino and cosmic rays we need to know the values of @xmath96 .
for protons and antiprotons with energy between @xmath108 ev and @xmath109 ev , the dominant interaction is @xmath110 pair production off the cosmic microwave background . at higher energies ,
photopion production becomes the dominant energy loss mechanism .
the onset of photopion production corresponds to the gzk cutoff energy @xcite . above this energy ,
the rate of energy loss increases dramatically , and nucleons do not reach us from cosmological distances . applying the results of @xcite
, we find , in the case of protons and antiprotons , @xmath111 for various values of @xmath112 ( where @xmath0 is the scale of the symmetry - breaking ) corresponding to @xmath113 and @xmath114 as shown in fig .
[ figpzmax ] . in fig .
[ figavdist ] we plot the average maximum distance that a ultra - high energy proton or antiproton can travel in the cmb . after propagating @xmath115
mpc the proton or antiproton energy decreases below @xmath116 ev irrespective of the initial jet energy .
for an expanded treatment see e.g. @xcite . for photons of energies greater than about @xmath117
ev , @xmath118 is determined by pair production off the cosmic photon background .
the resulting function @xmath118 is shown in fig .
[ figphotabs ] ( taken from ref .
@xcite ) . for lower energies ,
pair production off nuclei determines the form of @xmath119
. for neutrinos , the dominant process is scattering off the relic @xmath120k cosmic background neutrinos . in this case , @xmath121 is given by @xcite @xmath122 and @xmath123 the decay of lepton pairs produced in the neutrino scattering process further enhances the spectrum of all neutrino species and allows neutrino emission , corresponding to present day energies @xmath124 ev , to be detected from significantly higher redshifts than those given in ( [ zn1 ] ) and ( [ zn2 ] ) @xcite .
the cascade enhancement due to decay is only relevant at high redshift where the neutrino mean free path is sufficiently small . in the case of electron and muon neutrinos
, we will neglect the cascade enhancement of @xmath125 because the dominant term in the neutrino flux from vhs strings , ( [ nueflux ] ) and ( [ numuflux ] ) , is proportional to @xmath126 and the modifications to the @xmath52 , @xmath12 , and @xmath10 spectra expected at earth are negligible above @xmath127 @xcite .
the cascade enhancement at low energies increases as @xmath25 increases due to the greater energy in the initial jets .
as we mentioned above , the diffuse fluxes of particles in the vhs scenario are produced directly by the decay of x - particles .
the x - particles decay immediately after they are radiated from the long string segments .
the result is that the particle fluxes are produced along the strings .
the number of long strings per hubble volume @xmath7 we noted in section [ sec : intro ] is approximately 13 .
as the universe expands the comoving inter - string distance grows as @xmath128 fig .
[ figavdlgstr ] depicts the evolution of the average distance between long string segments .
the distance between the observer and a long string segment , @xmath129 is estimated to be @xmath130{2 } } \ ; .\ ] ] if the distance today between the observer and the long string segment is larger than the maximum average propagation distance for particles arriving with an energy @xmath37 then the fluxes at earth are exponentially suppressed @xcite . based on eq .
( [ intersdist ] ) we see that presently the average distance to a long string segment is @xmath131 which corresponds to a redshift @xmath132 . the present average redshift to a long string segment @xmath133 can be taken as a rough estimate of @xmath95 , i.e. , of the redshift corresponding to the latest time ( @xmath89 ) that the emission from the vhs string network can be considered isotropic .
this sets an upper limit of the integral ( [ flux ] ) and consequently in eqs .
( [ fluxsec][nutauflux ] ) we have @xmath134 hence only the particle fluxes emitted from the vhs network before @xmath89 ( @xmath135 ) are relevant for the diffuse flux
. a segment of long string could be substantially closer than @xmath136 , the expected average distance to the nearest string . in this case , a flux of cosmic rays protons ( antiprotons ) and photons above the gzk cutoff might be observed .
however , these fluxes would be highly anisotropic in contradiction to the present observations of uhe events above the gzk cutoff @xcite . in what follows
, we discuss the proton flux . because the fragmentation functions and relevant interactions with the cosmic backgrounds are to first order charge symmetric , the expected antiproton flux from the vhs strings is the same as the proton flux .
the average maximum distance at the present time to the source of a proton arriving with energy @xmath137 ev ( the onset of the gzk cutoff ) is @xmath138 mpc ( see fig .
[ figavdist ] ) .
this is equivalent to the source being at a redshift @xmath139 , much smaller than the distance from which the flux can be considered isotropic ( @xmath140 mpc ) .
it implies that no isotropic flux of protons from vhs strings is expected at earth above the gzk cutoff . on the other hand , a diffuse proton flux from the strings
should reach the earth at arrival energies @xmath141 ev , well below the uhe event energies .
although no isotropic flux of protons at uhe energies is expected at earth , the uhe protons emitted at high redshifts will have interacted with the universal radio background ( urb ) , cmb , and the infra - red / optical ( ir / o ) backgrounds to produce secondary fluxes of @xmath8-rays , electrons , and neutrinos .
these secondary fluxes could in principle be used to constrain the vhs model . however , in the context of the td scenarios including the vhs scenario , these nucleon - induced secondary fluxes are negligible compared to the corresponding fluxes directly produced by the decay of the x - particles .
this is because in the x - particle decay , far more direct @xmath8-rays and neutrinos are produced than nucleons .
the maximum redshift at which a photon arriving today with an energy @xmath37 could have been emitted is given in fig .
[ figphotabs ] .
if we compare this redshift to the present average distance to a long string ( @xmath132 ) we would deduce that the isotropic uhe @xmath8-ray flux should also be exponentially suppressed . in the case of @xmath8-rays , this is not the full story .
[ figphotabs ] describes the interaction length of the photons , i.e. , the distance traveled by a uhe photon before it is absorbed in the cosmic backgrounds .
the uhe photons initiate electromagnetic cascades as they travel in the cosmic backgrounds .
an important consequence is that the effective penetration length of the cascade is considerably larger than the interaction length @xcite . in the vhs scenario , both the interaction length and the effective penetration length
are much smaller than the distance to the nearest expected string ( cf .
( [ intersdist ] ) ) .
the result is that the @xmath142 ev @xmath8-ray are essentially subtracted from the arriving flux : the flux above @xmath143 ev is depleted due to interactions with the urb ; the flux in the range @xmath144 ev is depleted by the cmb ; and the flux in the range @xmath145 ev is depleted by the ir / o backgrounds . the energy lost by the photons emitted with energies above @xmath146 ev at high redshifts is recycled into energies below @xmath142 ev .
analytic calculations @xcite have shown that the electromagnetic cascade spectrum has a generic shape below @xmath147 ev , i.e. , the secondary cascade radiation spectrum is insensitive to the injection spectrum .
this implies that , even in the absence of the detection of direct uhe photons , the @xmath8-ray flux below @xmath142 ev can be used to constrain the vhs scenario .
the charged leptons created in the x - particle decay also initiate electromagnetic cascades on the cosmic backgrounds . as in the case of the uhe photons ,
the energy of the uhe charged leptons is recycled and enhances the photon flux below @xmath148 ev . the constraints on the electromagnetic emission in td models with a spectral index @xmath149 , which is applicable to the vhs scenario , stem from the limits on the present diffuse @xmath8-ray background between @xmath150 ev and @xmath117 ev @xcite .
these require that @xcite @xmath151 for @xmath152 where @xmath153 is the present rate of total energy injection into the electromagnetic channels . in the vhs scenario , we have from eq .
( [ jetdensity ] ) that the present rate of total energy injection into the electromagnetic channel ( photon and charged lepton emission ) is @xmath154 \ ; \left[\frac{\nu}{13}\right ] \left[g \mu \right ] \left[\frac{h}{0.65}\right]^{3 } \ ; \mbox{ev } \ ; \mbox{cm}^{-3 } \ ; \mbox{sec}^{-1 } \ ; .\ ] ] comparing to the @xmath153 limit in ( [ qem ] ) , we obtain a constraint on the energy density of a string in the vhs scenario of @xmath155 here we have assumed an extragalactic magnetic field of @xmath156 g. stronger values of the extragalactic magnetic field generate greater initial synchrotron losses at ultra - high energies , decreasing the uhe spectra , enhancing the cascade reprocessing into low energies , and thus lowering the upper limits on @xmath153 and @xmath9 .
because of the large distances to the sources , we have also neglected the uncertainties in the cosmic background urb and ir / o intensities @xcite .
the predicted flux of high energy neutrinos in the vhs scenario can be obtained from eqs.([nueflux ] - [ nutauflux ] ) using ( [ zcutoff ] ) , ( [ zn1 ] ) , and ( [ zn2 ] ) to determine @xmath157 .
the resulting @xmath11 and @xmath12 fluxes are shown in fig .
[ figemuflux ] and the @xmath10 flux in fig .
[ figtauflux ] , for the case when @xmath158 where @xmath0 is the scale of symmetry breaking .
the cascading of the tau neutrinos off the relic neutrino backgrounds will have little effect on the tau neutrino spectrum shown in fig .
[ figtauflux ] .
however , a secondary @xmath10 component peaking at lower energies will be created in collisions of the string - produced @xmath11 and @xmath12 with the relic neutrino backgrounds .
this @xmath11 and @xmath12 cascading should increase the net @xmath10 flux below @xmath159 . the @xmath10 signal expected at earth from vhs strings will be the sum of these two components , as shown in fig .
[ fignuscasc ] for @xmath160 @xcite . in fig .
[ fignuscasc ] we have plotted the cascade component calculated in ref .
@xcite for a @xmath161 td scenario .
this is only an estimate because the distance distribution of strings varies from model to model .
we will address full modeling of the @xmath10 flux including cascades in a later paper .
note however that it is the high @xmath13 region , where the cascades have little effect , which is the most observationally interesting .
our new result is that the @xmath11 , @xmath12 and @xmath10 fluxes are of similar magnitude as @xmath162 , unlike all previous string emission calculations which predicted a negligible @xmath10 component at high @xmath13 . in other string emission or x - decay scenarios ,
the @xmath11 , @xmath12 and @xmath10 flux should also be of comparable magnitude as @xmath163 .
because of the uncertainty in extrapolating to high emission energies , the precise ratio of the neutrino fluxes at high @xmath13 is not known ( also , as plotted in fig . [ fignuscasc ] , the extrapolation for @xmath10 is derived from herwig monte carlo simulations while the @xmath11 and @xmath12 functions are derived from the less rigorous approximation eq .
( [ primdecay ] ) . ) the actual ratio will be determined by the first step in the x - decay chain which is unknown and may be influenced by , for example , leptonic and susy decay channels ( see section [ sec : ext ] ) .
the most recent observational limits come from the eas - top @xcite , the frjus @xcite , and the fly s eye @xcite experiments .
these experiments give upper limits on the cosmic ray neutrino flux in the energy range between @xmath164 ev and @xmath165 ev . in figs .
[ figemuflux ] and [ figtauflux ] we have plotted the observational limits using the updated @xmath166 cross - section @xcite . by inspection , we see that the most stringent constraint comes from the highest energy @xmath12 flux in fig .
[ figemuflux ] which gives an upper bound on @xmath9 of @xmath167 this bound is derived at high @xmath13 so it should be little affected by the cascades which are neglected in fig .
[ figemuflux ] .
we have also plotted on figs .
[ figemuflux ] and [ figtauflux ] the upper limits stemming from the planned sensitivities of the pierre auger project as estimated in ref .
@xcite , and the owl / airwatch project as estimated in ref .
it is assumed that both experiments will collect data over a period of a few years .
their sensitivities span the energy range between @xmath168 ev and @xmath169 ev and would allow more stringent upper limits on @xmath9 up to @xmath170 in the case of null detection by the pierre auger project ; and @xmath171 in the owl / airwatch case .
unlike the present constraint ( [ presconst ] ) , the auger and owl / airwatch projects will constrain the @xmath12 flux in fig .
[ figemuflux ] at energies @xmath172 ev ( i.e. the lower end of the energy range of these experiments ) .
the @xmath10 flux is below the present observational upper limits .
the @xmath10 sensitivities of future experiments are expected to constrain @xmath9 , but these constraints are expected to be an order of magnitude weaker than those from the @xmath11 and @xmath12 fluxes . the slope of the @xmath10 spectrum is steeper than the slope of the observational upper limits for both present and future data ( cf . fig .
[ figtauflux ] ) . on the other hand , the @xmath10 component of a td signature maybe easier to distinguish from other sources than the @xmath11 and @xmath12 components . as is obvious from ( [ asymptemu ] ) , ( [ asympttau ] ) and fig . [ fignumj ]
the predicted flux of cosmic ray neutrinos increases as @xmath1 decreases .
thus , the vhs scenario with gut - scale strings can not be made compatible with observational constraints by lowering the energy scale @xmath1 of the initial jets ( which likely happens if new physics evolves between the electroweak symmetry breaking scale and the gut scale ) .
raising @xmath1 above @xmath0 would alleviate the conflict , but this appears quite unnatural .
the initial decay channels and branching ratios for the x - particle are unknown .
the application of the extrapolated qcd - derived fragmentation functions to x - decay assumes implicitly that the branching ratios for @xmath173 are the same as those for @xmath174 .
this may not be so for even hadronic decays at very high energies .
the initial decay channel @xmath175 , where @xmath176 is a charged lepton , should produce spectra which are decreased by no more than a factor of 2 , compared with the @xmath177 spectra ( assuming equi - partition of the x - particle energy between the initial @xmath178 and @xmath176 ) , with the exception of an @xmath10 spectrum enhancement as @xmath179 due to decay of the initial @xmath180 @xcite .
the initial charged lepton @xmath176 would also generate em cascades which are effectively attenuated at high energies in the vhs scenario , as we remarked in section [ sec : jet ] for the other em decay products , but which would contribute to the spectra at low energies .
this should increase the total energy going into the em cascade channel by less than @xmath181 ( see eq .
( [ q0const ] ) ) .
pure leptonic decay channels @xmath182 , @xmath183 or @xmath184 would produce substantially suppressed spectra , because of the low multiplicity per decay compared with hadronic jets , except for a possible neutrino spectra increase as @xmath179 due to the initial step in the decay chain . in the case of purely leptonic decays ,
the nucleon and @xmath8-ray spectra would be wholly created by the collisions of the emitted leptons with the cosmic ambient media .
it has been shown that in td models generally , pure leptonic decays can only be made interesting if the cosmic relic neutrinos are massive , non - relativistic and locally clustered thus enhancing the relative cascade - generated component ( ref .
however , it is highly unlikely that the x - particles ( being high energy gauge and scalar particles ) would decay purely into leptons in the initial step . because of the much higher multiplicity of hadronic jets , any scenario containing reasonable initial branching ratios for the x - particle ( say , to within a factor of 23 : @xmath185 pure hadronic , @xmath185 pure leptonic and @xmath185 hadronic and leptonic ) would produce spectra dominated by the particles produced via the hadronic decay channels
hence , even in the case of locally clustered relic neutrinos , the component produced in pure leptonic decays would be essentially dominated by the hadronic decays , with the possible exception of a uhe neutrino feature as @xmath179 .
in the analysis so far , we have assumed @xmath186 .
if neutrinos do possess a small mass , as present neutrino oscillation experiments hint , the predicted spectra from vhs strings are modified in two ways .
firstly , the interaction of the string - produced neutrinos with the @xmath120k relic background neutrinos is strongly enhanced at the resonance for @xmath187 production @xcite , @xmath188 @xmath189 ev .
( if neutrinos are massless , the @xmath187 resonance occurs at @xmath190 ev @xcite . )
the present upper limits from particle physics experiments on the neutrino masses are @xmath191 ev , @xmath192 mev , and @xmath193 mev @xcite . the cosmological requirement that @xmath194 places a stricter upper limit on the sum of the masses of the three neutrino species of @xmath195 ev @xcite . taken together
these limits imply that at least one @xmath187 resonance must occur for neutrinos today at some energy in the range @xmath165 ev @xmath196 ev . the exact value(s ) of @xmath197 will depend on whether or not the neutrinos have mass , and the value(s ) of the neutrino mass(es ) if non - zero .
if the neutrino masses are non - degenerate , the @xmath187 resonance will occur at two or three distinct @xmath197 in this range . in the case that @xmath198 , the @xmath187 resonance in the vhs scenario will produce a slight decrease in the predicted neutrino fluxes around @xmath197 and in the energy decade below @xmath197 , and an accompanying slight enhancement on all spectra below @xmath199 due to secondary @xmath187 decay products .
the amplitudes of all spectra should be modified by less than a factor of 2 .
( see for example ref .
@xcite for the case of degenerate @xmath200 ev neutrinos in other cosmic string models . )
if the relic neutrinos are non - relativistic and clustered locally on a scale less than the attenuation length for photons and nucleons , the effect on the photon , proton and antiproton spectra may be enhanced @xcite .
secondly , a non - zero neutrino mass may lead to oscillations between neutrino species as the flux propagates to earth .
in this case , the total number of neutrinos is preserved but the ratio of the flux in each species may depend on the neutrino energy , the distance traveled and neutrino masses .
recent observations by the superkamiokande , kamiokande , macro and soudan experiments of @xmath11 and @xmath12 neutrinos produced by cosmic rays colliding with the earth s atmosphere strongly suggest that neutrino oscillations ( at least between the @xmath12 and @xmath10 flavors ) occur . [ for a review of the current status of neutrino oscillation experiments and measurement implications see @xcite and references therein ] .
the evidence for zenith - angle dependence in the superkamiokande data further strengthens this interpretation .
similarly , the long - standing observed deficit of solar neutrino flux may be resolved by @xmath11 oscillation into @xmath12 or @xmath10 although the solar neutrino experiments are not yet sufficiently consistent to favor a particular solution .
the third set of evidence for neutrino oscillations comes from the lsnd collaboration accelerator experiment which has seen evidence for @xmath12 into @xmath11 oscillations . when combined with the atmospheric and solar data
, this would require the existence of a fourth `` sterile '' neutrino species .
the lsnd measurements have yet to be independently confirmed .
a number of new experiments are presently being commissioned with the objective of resolving these matters .
the k2k , num - minos and cngs experiments are expected to provide the definitive answer and exploration of the relevant parameter space . in these experiments , neutrino beams will be sent from accelerators to detectors over baselines of hundreds of kilometers .
k2k is already taking data and the num - minos and cngs will be operational in 2002 and 2005 , respectively . ignoring for now cosmological evolution
, the probability of a relativistic neutrino undergoing vacuum flavor transition when propagating a distance @xmath201 is @xmath202\end{aligned}\ ] ] in the 3-flavor solution .
the atmospheric data to date are best fit by an amplitude of @xmath203 and a mass difference of @xmath204 @xcite . for models with distant cosmological sources ( e.g. tds and agns ) and a neutrino ratio at emission of @xmath205 ( in our case
this occurs at @xmath206 ) , the ratio of the spectra reaching the earth will be @xmath207 in the 3-flavor model ; similarly , in the 4-flavor model consistent with the lsnd data , an initial emission ratio of @xmath208 will become @xmath209 at earth @xcite . since we showed earlier that the current average distance to the nearest long string is 2000 mpc , even the highest energy neutrinos emitted in the present era will have undergone transition before reaching earth . in this case , at @xmath206 , both the @xmath12 and @xmath10 flux reaching earth should approximately equal half the @xmath12 flux expected without oscillations . in the case of the flux emitted in previous epochs ,
this will only be so provided the wavelength of the oscillation is much less than the mean free path of the @xmath12 and @xmath10 in the universe .
if the oscillation wavelength is greater than the mean free path then the cascade structure will be determined by the original neutrino flavors at emission .
the mean free path decreases sharply as @xmath210 increases @xcite .
for example , a @xmath211 ev neutrino at @xmath212 has a mean free path of @xmath115 kpc @xcite .
since these too are the redshifts and energies at which cascading becomes relevant , cascade development should be predominantly determined by the neutrino flavors at emission . in the vhs scenario
, however , the neutrino cascades have little influence on our regions of interest .
hence the predicted observable @xmath12 and @xmath10 vhs fluxes in the case with neutrino oscillations should be approximately half the unoscillated @xmath12 flux at @xmath206 .
as @xmath213 , i.e. in the region where the initial @xmath11 , @xmath12 , and @xmath10 vhs fluxes are of comparable magnitude , the ratio should also approach @xmath214 in the case with neutrino oscillations .
the precise value of the modulated neutrino flux @xmath215 as @xmath213 will depend on the initial absolute and relative string - produced fluxes which we regard as uncertain by more than a factor of 2 .
additionally , one possible solution to the solar neutrino deficit problem is vacuum @xmath216 or @xmath10 oscillations with @xmath217 and @xmath218 . in this case
, the phase of the transition probability , @xmath219 , is @xmath220 for @xmath221 ev neutrinos emitted today from a long string at 2000 mpc .
thus the @xmath11 spectra at earth would be determined by @xmath216 or @xmath10 oscillation below @xmath222 ev and exhibit negligible @xmath11 oscillation above @xmath222 ev .
if this particular solar deficit solution is combined with the solution implied by the atmospheric data @xmath223 , an initial ratio at emission of @xmath205 or @xmath214 would be modulated to @xmath214 reaching earth because the wavelength for @xmath14 oscillation is much smaller than that of the oscillation solution associated with the atmospheric solar data .
a detailed analysis of the detector signatures for various oscillation scenarios is presented in refs . @xcite and @xcite .
other extensions to the standard model could influence the predicted spectra .
the existence of a supersymmetric ( susy ) sector would contribute extra decay channels in the fragmentation of the x - particle and jets .
( see @xcite and references therein for possible approaches to modeling susy channels for application to uhe decays ) .
the net effect would be a skewing of the primary flux spectra to lower energies and a decrease in the energy reprocessed in high energy cascades .
for example , in the p=1 td model studied in ref .
@xcite with @xmath224 gev , including a susy sector decreases the spectra of all species expected at earth by an order of magnitude at @xmath225 , shifts the uhe peaks in the spectra to lower energies by at least an order of magnitude and increases the amplitude of the @xmath11 , @xmath12 , @xmath226 and @xmath227 spectra by about an order of magnitude at energies below the peaks
. the predicted @xmath228 ev photon spectrum , predominantly generated by em cascades , though is unchanged because the same fraction of initial x - particle energy is reprocessed into low energies in the em cascades with or without a susy sector in that model . in the case of the vhs scenario
, these remarks must be convolved with the suppression of the uhe @xmath8-ray and nucleon spectra due to the nearest string being farther than the relevant uhe attenuation lengths . to date
there is no experimental evidence for a susy sector and no unique theoretical predictions for the architecture of the susy sector and superparticle properties such as mass .
other extensions to the standard model , for example technicolor or leptoquarks , which also increase the number of degrees of particle freedom would be expected to similarly skew the spectra to lower energies .
as well , extensions to the standard model may modify the high - energy cross - sections governing the interactions of the emission with the cosmic ambient media .
we have computed the ultra - high energy @xmath3ray , neutrino and cosmic ray fluxes in the vhs cosmic string scenario , in which the long string network loses its energy directly by particle emission .
this is in contrast to the standard cosmic string model with a scale - invariant distribution of string loops which lose energy predominantly by gravitational radiation .
the predicted particle fluxes are much larger in the vhs scenario .
the predictions for the particle fluxes depend on extrapolating the qcd fragmentation function to energies much higher than current accelerator energies .
new physics between the qcd scale and the scale @xmath229 at which the defects are formed could significantly affect the final particle distributions .
we parametrize this uncertainty by introducing a scale @xmath1 as the energy scale to which the qcd fragmentation functions can be extrapolated and regard @xmath1 as the initial jet energy . the smaller the value of @xmath1 , the larger the number of jets which are generated by the initial emission from the cosmic string .
we calculate the resulting particle fluxes as a function of both @xmath1 and @xmath5 .
a new aspect of our work is the computation of the significant tau neutrino fluxes directly produced by decay in the particle jets .
our calculations show that the predicted flux @xmath80 of @xmath3rays , electron and muon neutrinos and cosmic rays scales as @xmath230 , as in other defect models , whereas the present observations show a much weaker increase with @xmath37 .
hence , the most stringent constraints on topological defect models come from the highest energies for which data exist .
setting @xmath231 , the vhs strings with @xmath232 produce electron and muon neutrinos in excess of the uhe observations .
thus , assuming that the strings evolve as in the vhs scenario , models with @xmath232 or , equivalently a symmetry breaking scale of @xmath233 ev , are ruled out by the uhe data .
( this result is in accordance with the limit on @xmath9 in ref .
@xcite ) . at a given particle energy , the predicted fluxes reaching the earth scale as @xmath234 and
increase with @xmath9 .
lowering @xmath1 increases the disagreement between predictions and observations , although the upper cutoff on the predicted spectra diminishes . a consistent but slightly stronger constraint on the vhs string scenario comes from the cascading of the electromagnetic emission off the cosmic radiation backgrounds and the potential conflict with the observed diffuse @xmath235 ev @xmath236 @xmath117 ev gamma ray background as probed by egret .
this requires @xmath237 .
we conclude that , generically , gut - scale strings are ruled out in the vhs scenario .
however , vhs models with lighter @xmath9 , corresponding to lower symmetry - breaking scales , may be contributing interestingly to the observed uhe events .
we reiterate that the vhs scenario is not universally accepted as giving the correct dynamics of cosmic strings .
finally , we note that the uncertainties inherent in the calculation of the expected fluxes in any td model mean that the absolute and relative flux in each particle species should not be regarded as accurate to more than an order of magnitude in any model .
the list of uncertainties includes : the mass and other properties of the x - particle(s ) ; the form of the potential of the scalar field at string formation ; the branching ratios and decay channels of the x - particle(s ) ; the validity of the extrapolated fragmentation functions ; the possibility of new particles ( e.g. a susy sector ) and other extensions to the standard model which may influence particle production , decay and propagation ; the strength of the extragalactic magnetic field ; the intensity of the cosmic ir background ; and the hubble constant . taken together ,
these factors have the capability to increase or decrease the absolute and relative fluxes expected at earth .
we would like to thank v. berezinsky for his comments and suggestions . this work has been supported ( at brown ) in part by the us department of energy under contract de - fg0291er40688 , task a , and was performed in part while jhm held a nrc - nasa / jsc senior research associateship .
ufw has been supported at centra - ist by `` fundao para a cincia e a tecnologia '' ( fct ) under the program `` praxis xxi '' and in part by lip - lisbon .
in the main text we computed the flux of particles resulting from two - body decay of charged pions in a jet . the muons which are produced in this decay are unstable and in turn decay via a three body decay process producing electrons , muon neutrinos and anti - electron neutrinos . in this appendix
we compute the full spectra neutrinos from the @xmath238 decay chain . because of the finite muon rest mass , the muon and muon neutrinos created in the first stage of charged pion decay can not carry an arbitrary fraction of the pion energy . in the limit @xmath239 , the energies of the decay products lie between @xmath240\ ] ] for muon neutrinos and @xmath241\ ] ] for muons @xcite . here , @xmath242 where @xmath47 is given by ( [ alph ] ) .
the muon distribution is then @xmath243 in order to obtain the @xmath12-flux , we integrate ( [ primdecay ] ) over the range ( [ range1 ] ) , resulting in ( [ secdecay ] ) . to obtain the flux of muon decay products ,
we follow the method of ref . @xcite . in the laboratory frame the distribution of leptons with energy @xmath244 produced by the decay of a muon of energy @xmath245
is given by [ gaisser @xcite p. 92 ] @xmath246 in the limit @xmath247 , where @xmath248 @xmath249 for @xmath12 and @xmath250 @xmath251 for @xmath52 .
in ( [ dn / dy ] ) , @xmath252 is the projection of the muon spin in the muon rest frame along the direction of the muon velocity in the laboratory frame @xmath253 convolving the distribution ( [ dn / dy ] ) with the distribution of muons from pion decay ( [ muondistr ] ) and integrating over muon and pion energies , the neutrino spectrum in the laboratory frame is @xmath254 @xmath255 \ ; .\ ] ] for simplicity we will first evaluate ( [ ninte ] ) for an initial pion distribution of the form @xmath256 changing the order of integration , @xmath257 the result is @xmath258 \biggr]\ ] ] where @xmath259 @xmath260 for @xmath12 and @xmath261 @xmath262 for @xmath52 .
because muons and antimuons are created with opposite spin polarization in charged pion decays , this spectrum also applies to the antiparticle decay chain .
the distribution of pions produced in qcd jet decay can be approximated by the polynomial fragmentation function ( [ secdecay ] ) .
evaluating ( [ ndistr ] ) for the appropriate values of @xmath75 and @xmath263 and extending the analysis to @xmath264 , we find the final distribution of neutrinos produced by jet decay to be approximately @xmath265 \ ; , \qquad \quad e_{\nu_{\mu } } \le \frac{m_j}{\alpha } \ ; , \nonumber\end{aligned}\ ] ] where @xmath47 is given in ( [ alph ] ) , for the muon neutrinos produced in the first stage of the pion decay , @xmath266 \ ; , \qquad \quad e_{\nu_{\mu } } \le m_j \ ; , \nonumber\end{aligned}\ ] ] for the muon neutrinos produced by the decay of muons in the pion decay , and @xmath267 \ ; , \qquad \quad e_{\nu_{e } } \le m_j \ ; , \nonumber\end{aligned}\ ] ] for the final electron neutrinos .
p. bhattacharjee and g. sigl , phys .
327 * , 109 ( 2000 ) .
g. sigl , astro - ph/9611190 .
v. berezinsky , p. blasi and a. vilenkin , phys .
d * 58 * , 103515 ( 1998 ) . j. macgibbon and r. brandenberger , nucl .
b331 * , 153 ( 1990 ) .
p. bhattacharjee , phys .
d * 40 * , 3968 ( 1989 ) ; + p. bhattacharjee and n. rana , phys .
b * 246 * , 365 ( 1990 ) .
macgibbon and r.h .
brandenberger , phys .
d * 47 * , 2283 ( 1993 ) . ya.b .
zeldovich , mon . not . r. astron
. soc . * 192 * , 663 ( 1980 ) .
a. vilenkin , phys .
lett . * 46 * , 1169 ( 1981 ) .
a. vilenkin , phys . rep . * 121 * , 263 ( 1985 ) .
a. vilenkin and e.p.s .
shellard , _ cosmic strings and other topological defects _ ( cambridge univ . press ,
cambridge , 1994 ) .
m. hindmarsh and t.w.b .
kibble , rept .
. phys . * 58 * , 477 ( 1995 ) .
r. brandenberger , int . j. mod
. phys . a * 9 * , 2117 ( 1994 ) .
n. turok and r.h .
brandenberger , phys .
d * 33 * , 2175 ( 1986 ) .
h. sato , prog .
. phys . * 75 * , 1342 ( 1986 ) .
a. stebbins , astrophys
. j. lett . *
303 * , l21 ( 1986 ) .
vincent , m. hindmarsh , and m. sakellariadou , phys .
d * 56 * , 637 ( 1997 ) .
a. albrecht and n. turok , phys .
d * 40 * , 973 ( 1989 ) .
d.p . bennett and f.r .
bouchet , phys .
lett . * 60 * , 257 ( 1988 ) .
b. allen and e.p.s .
shellard , phys .
lett . * 64 * , 119 ( 1990 ) .
d. austin , e.j .
copeland , and t.w.b .
kibble , phys .
d * 48 * , 5594 ( 1993 ) .
g. vincent , n.d .
antunes , and m. hindmarsh , phys .
lett . * 80 * , 2277 ( 1998 ) .
shellard , private communication ( 1998 ) .
k.d . olum and j.j .
blanco - pillado , phys .
84 * , 4288 ( 2000 ) .
r. brandenberger , nucl . phys . *
b293 * , 812 ( 1987 ) .
s. hawking , phys .
b * 231 * , 237 ( 1989 ) ; + a. polnarev and r. zembowicz , phys .
d * 43 * , 1106 ( 1991 ) .
macgibbon , r.h .
brandenberger , and u.f .
wichoski , phys .
d * 57 * , 2158 ( 1998 ) .
p. bhattacharjee , q. shafi , and f.w .
stecker , phys .
lett . * 80 * , 3698 ( 1998 ) .
m. takeda _
_ , astrophys
. j. * 522 * , 225 ( 1999 ) .
macgibbon , u.f .
wichoski , and b.r .
webber , hep - ph/0106337 .
j.h . macgibbon and u.f .
wichoski , ( in preparation ) .
b. anderson et al . , phys . rep .
* 97 * , 31 ( 1983 ) .
corcella , g. _ et al .
_ , jhep * 0101 * , 010 ( 2001 ) ; + g. marchesini and b. webber , nucl .
b310 * , 461 ( 1988 ) ; + g. marchesini et al . ,
* 67 * , 465 ( 1992 ) . c. hill , nucl
b224 * , 469 ( 1983 ) .
hill , d.n .
schramm , and t.p .
walker , phys .
d * 36 * , 1007 ( 1987 ) .
f. stecker , astrophys .
j. * 228 * , 919 ( 1979 ) .
f. halzen , b. keszthelyi , and e. zas , phys .
d * 52 * , 3239 ( 1995 ) .
p. bhattacharjee , in _ astrophysical aspects of the most energetic cosmic rays _ , edited by m. nagano and f. takahara ( world scientific , singapore , 1991 ) , p. xx .
a. mueller , nucl .
b * 213 * , 85 ( 1983 ) . v. berezinsky , m. kachelriess , and a. vilenkin , phys .
lett . * 79 * , 4302 ( 1997 ) .
dokshitzer , v. khoze , a. mueller , and s. troyan , _ basics of perturbative qcd _ ( editions frontires , saclay , 1991 ) ; + r. ellis , w. stirling and b. webber , _ qcd and collider physics _ ( cambridge univ . press ,
cambridge , 1996 ) .
s. sarkar , hep - ph/0005256 .
k. greisen , phys .
* 16 * , 748 ( 1966 ) ; + g. zatsepin and v. kuzmin , jetp lett .
* 4 * , 78 ( 1966 ) .
v. berezinsky and s. grigoreva , astron .
astrophys . *
199 * , 1 ( 1988 ) ; + s. yoshida and m. teshima , prog .
. phys . * 89 * , 833 ( 1993 ) ; + j. rachen and p. bierman , astron .
* 272 * , 161 ( 1993 ) .
r. protheroe and p. johnson , astropart .
phys . * 4 * , 253 ( 1996 ) , astro - ph/9506119 .
bhattacharjee , c.t . hill , and d.n .
schramm , phys .
* 69 * , 567 ( 1992 ) .
s. yoshida , astropart .
* 2 * , 187 ( 1994 ) ; + s. yoshida et al . ,
j. * 479 * , 547 ( 1997 ) .
bonometto , il nuovo cim . lett . * 1 * , 677 ( 1971 ) .
j. wdowczyk , w. tkaczyk , and a.w .
wolfendale , j. phys .
a * 5 * , 1419 ( 1972 ) .
a. chen , j. dwyer , and p. kaaret , astrophys .
j. * 463 * , 169 ( 1996 ) ; + p. sreekumar et al . , astrophys .
j. * 494 * , 523 ( 1998 ) .
g. sigl , k. jedamzik , d.n .
schramm , and v.s .
berezinsky , phys .
d * 52 * , 6682 ( 1995 ) .
r. j. protheroe and t. stanev , phys .
lett . * 77 * , 3708 ( 1996 ) ; _ idem _ * 78 * , 3420 ( 1997 ) .
g. sigl , s. lee , p. bhattacharjee , and s. yoshida , phys .
d * 59 * , 043504 ( 1999 ) .
s. coppi and f. a. aharonian , astrophys . j. lett .
* 487 * , l9 ( 1997 ) .
m. agliettta _
et al . _ , in _ proceedings of the 24th international cosmic ray conference _ , rome , 1995 , vol.1 , p.638 .
w. rhode et al .
, astropart . phys . * 4 * , 217 ( 1996 ) .
r. baltrusaitis et al . , astrophys .
* 281 * , l9 ( 1984 ) ; + r. baltrusaitis et al . ,
d * 31 * , 2192 ( 1985 ) . c. quigg , _
neutrino interaction cross sections _ , fermilab - conf-97/158-t ( 1997 ) .
k. s. capelle , j. w. cronin , g. parente , and e. zas , astropart . phys .
* 8 * ( 1998 ) 321 .
j. f. ormes _ et al .
_ , _ proceedings of the 25th international cosmic ray conference _ , durban , 1997 , vol . 5 , p. 273 ; + y. takahashi _ et al .
_ , _ international symposium on `` extremely high energy cosmic rays : astrophysics and future observatories '' _ , edited by m. nagano , ( institute for cosmic ray research , tokyo , 1996 ) , p.310 .
t. weiler , phys .
49 * , 234 ( 1982 ) ; + e. roulet , phys .
d * d47 * , 5247 ( 1993 ) .
d. e. groom et al .
j. c * 15 * , 1 ( 2000 ) .
s. a. bludman , phys .
d * 45 * , 4720 ( 1992 ) .
s. yoshida , g. sigl , and s. lee , phys .
rev . lett . * 81 * , 5505 ( 1998 ) ; + j. j. blanco - pillado , r. a. vazquez , and e. zas , ) , phys .
d * 61 * , 123003 ( 2000 ) .
akhmedov , hep - ph/0001264v2 .
a. hussein , hep - ph/0004083 .
s. i. dutta , m. h. reno , and i. sarcevic , hep - ph/0104275 .
s. i. dutta , m. h. reno , and i. sarcevic , phys .
d * 62 * , 123001 ( 2000 ) .
v. berezinsky and m. kachelrei@xmath268 , hep - ph/0009053 .
t. k. gaisser , _ cosmic ray and particle physics _ ( cambridge univ . press , cambridge , 1990 ) . | decaying topological defects , in particular cosmic strings , can produce a significant flux of high energy neutrinos , photons and cosmic rays . according to the prevailing understanding of cosmic string dynamics in an expanding universe
, the network of long strings loses its energy first into string loops , which in turn give off most of their energy as gravitational radiation .
however , it has been suggested by vincent _
et al . _ ( vhs ) that particle emission may be the dominant energy loss channel for the long string network .
in this case , the predicted flux of high energy particles would be much larger . here
we calculate the predicted flux of high energy gamma rays , neutrinos and cosmic ray antiprotons and protons as a function of the scale of symmetry breaking @xmath0 at which the strings are produced and as a function of the initial energy @xmath1 of the particle jets which result from the string decay . assuming the validity of the vhs scenario
, we find that due to the interactions with the cosmic radiation backgrounds all fluxes but the neutrino flux are suppressed at the highest energies .
this indicates that the observed events above the gzk cutoff can only be accounted for in this scenario if the primary particle is a neutrino and @xmath0 is somewhat less than the gut scale , i.e. @xmath2 ev .
the domain of parameter space corresponding to gut - scale symmetry breaking is excluded also by the current observations below the gzk cutoff .
a new aspect of this work is the calculation of the spectrum of the tau neutrinos directly produced in the decay of the x - particles .
this significantly increases the tau neutrino signal at high energies in all cosmic string scenarios . |
for @xmath4 fixed consider the family of polynomials @xmath0 , @xmath1 .
a _ multiplicity vector ( mv ) _ is a vector whose components are the multiplicities of the real roots of @xmath3 ( for @xmath5 fixed ) listed in increasing order .
e.g. for @xmath6 the mv @xmath7 $ ] means that for the real roots @xmath8 one has @xmath9 and there is a complex conjugate couple about whose real and imaginary part the mv gives no information .
define the _ length _
( resp . the _ multiplicity surplus _ ) of a mv @xmath10 $ ] as the integer @xmath11 ( resp .
@xmath12 ) .
the number of complex conjugate couples of roots of @xmath3 ( counted with the multiplicities ) equals @xmath13 .
stratify the space @xmath2 each mv with multiplicity surplus @xmath14 defines a stratum of codimension @xmath14 . [ dimension ] 1 ) it is clear that if @xmath15 , @xmath16 , are two strata defined by the mvs @xmath17 and if @xmath18 belongs to the closure of @xmath19 , then for the lengths @xmath20 of the mvs @xmath17 one has @xmath21 . indeed , to pass from @xmath18 to @xmath19 one or several real roots must bifurcate . if all new roots are real , then @xmath22 . if there is at least one complex conjugate couple , then @xmath23 .
\2 ) if the mvs @xmath17 are of the same length , then @xmath18 belongs to the closure of @xmath19 if and only if @xmath24 is obtained from @xmath25 by replacing groups of consecutive multiplicities by their sums ( call this an _ operation of type a _ ; it corresponds to confluence of real roots ) .
\3 ) an _ operation of type b _ consists either in adding a component equal to @xmath26 at an arbitrary place of the mv ( a complex conjugate couple becomes a double real root different from the other real roots of @xmath3 ) or in increasing of one of the components of the mv by @xmath26 ( a complex conjugate couple becomes a double real root which coincides with one of the other real roots ) .
if for the lengths @xmath20 of @xmath15 one has @xmath27 , @xmath28 , then @xmath18 belongs to the closure of @xmath19 if and only if @xmath24 is obtained from @xmath25 either by an operation of type a followed by @xmath29 operations of type b or just by @xmath29 operations of type b. \4 ) given a stratum @xmath30 of dimension @xmath31 , with mv @xmath32 , one can obtain all mvs of strata of dimension @xmath33 adjacent to @xmath30 either by replacing some component @xmath34 of @xmath32 by two consecutive components @xmath35 , @xmath36 or by deleting from @xmath32 a component equal to @xmath26 ( which means that a double real root becomes a complex conjugate couple ) .
indeed , to pass to a stratum of next dimension adjacent to the given one a root must bifurcate into two roots .
if these two roots are complex conjugate , then they can not be multiple because the stratification does not take into account the multiplicities of the complex roots .
the first aim of the present paper is to prove the following [ trans ] a stratum of codimension @xmath14 is a smooth real contractible algebraic variety of dimension @xmath37 .
it is the graph of a smooth @xmath14-dimensional vector - function defined on the projection of the stratum in @xmath38 .
the field of tangent spaces to the stratum is continuously extended to the strata belonging to its closure .
the extension is everywhere transversal to the space @xmath39 .
the theorem is proved in section [ prtrans ] .
it is illustrated by an example in section [ example ] .
we discuss in section [ discuss ] ( in view of the theorem and of previous results from @xcite ) the mutual disposition of adjacent strata .
\1 ) the theorem generalizes theorem 1.8 from @xcite .
the latter treats the case when @xmath3 is _ hyperbolic _ , i.e. all roots are real .
\2 ) the above stratification ( or a similar one ) has been considered ( at least in some aspects ) by other authors as well , see for instance @xcite , @xcite and their bibliographies , @xcite , @xcite and @xcite ; the list is anything but exhaustive .
1 . we show for @xmath40 , @xmath41 , the well - known picture of the swallowtail , i.e. the surface @xmath42 .
the three strata of dimension @xmath43 and their respective mvs are the open subset of @xmath44 `` above '' @xmath45 , with empty mv ( no real roots ) , its open subset `` below '' @xmath45 , with mv @xmath46 $ ] , and the interior of the curvilinear pyramid @xmath47 , with mv @xmath48 $ ] .
the two - dimensional stratum @xmath49 defined by the mv @xmath50 $ ] is the swallowtail without the boundary of @xmath51 .
the limits of the tangent spaces to @xmath49 are different on the different sides of the self - intersection curve @xmath52 ( excepting @xmath53 where the limit is unique ) , see remark [ remim ] .
the strata @xmath54 and @xmath55 ( without the boundaries ) are defined respectively by the mvs @xmath56 $ ] and @xmath57 $ ] ; the stratum @xmath58 ( without the boundary ) is defined by the mv @xmath59 $ ] .
the one - dimensional strata @xmath60 and @xmath61 ( without the point @xmath53 ) are defined respectively by the mvs @xmath62 $ ] and @xmath63 $ ] while @xmath52 ( without the point @xmath53 ) is defined by the mv @xmath64 $ ] . finally , the point @xmath53 is defined by the mv @xmath65 $ ] .
the self - intersection curve @xmath52 has an analytic continuation @xmath66 ( represented by a dashed line on fig .
1 ) which consists of polynomials of the form @xmath67 , @xmath68 , i.e. having a conjugate complex couple of roots of multiplicity @xmath26 .
this curve is not a stratum of the stratification ( although it belongs to @xmath45 ) because the multiplicities of the complex roots are not taken into account , see part 4 ) of remarks [ dimension ] .
[ remim ] the following proposition is proved in @xcite , p. 52 - 53 .
[ propim ] if @xmath3 , @xmath69 , @xmath70 , @xmath71 are monic polynomials where @xmath72 and @xmath73 have two by two no root in common , then there exist neighbourhoods @xmath74 , @xmath75 , @xmath70 , @xmath76 of @xmath3 , @xmath69 , @xmath70 , @xmath71 such that the product map @xmath77 , @xmath78 , is a diffeomorphism .
this result implies that at any point of the open arc @xmath52 the swallowtail is locally diffeomorphic to two intersecting planes in @xmath44 .
indeed , on @xmath52 the polynomial @xmath3 has two double real roots , hence , is of the form @xmath79 , @xmath80 .
one can set @xmath81 , @xmath82 and @xmath83 , @xmath84 .
the two planes are defined respectively by the variables @xmath85 and @xmath86 .
proposition [ propim ] can be applied in the general case as well ( i.e. when @xmath3 has any mv ) to understand what the set @xmath87 is locally like up to a diffeomorphism .
consider a point @xmath88 of a stratum @xmath74 of dimension @xmath89 . intersect @xmath74 by the affine space @xmath90 of dimension
@xmath26 containing @xmath88 and parallel to @xmath91 . by theorem [ trans ]
the intersection is the point @xmath88 .
the intersections with @xmath90 of the strata of dimension @xmath92 are curves containing @xmath88 in their closures and having non - vertical limits at @xmath88 of their tangent lines ( theorem [ trans ] ) .
each such curve ( considered locally , at @xmath88 ) projects only `` to the left '' or `` to the right '' of @xmath88 on @xmath93 . the intersections with @xmath90 of the strata of dimension @xmath94 are sectors delimited by these curves .
it is explained in @xcite , subsection 1.2 , how the above curves are situated near @xmath88 in the case when @xmath74 is a stratum of hyperbolic polynomials .
we generalize here these results for the case of arbitrary stratum @xmath74 .
denote by @xmath95 $ ] the mv of the stratum @xmath74 .
denote by @xmath96 the stratum with mv obtained from @xmath32 by replacing the component @xmath97 by two components @xmath98 , @xmath99 where @xmath100 .
if @xmath101 , denote by @xmath17 the stratum whose mv is obtained from @xmath32 by deleting the component @xmath97 . by part 4 ) of remarks
[ dimension ] , these are all strata of dimension @xmath92 adjacent to @xmath74 .
we use the notation @xmath96 , @xmath17 also for the intersections of these strata with @xmath90 . in what follows
we assume that the equations of the limits at @xmath88 of the tangent lines to the curves @xmath96 are given in the form @xmath102 .
[ slope ] the slopes of the limits at @xmath88 of the tangent lines to the curves @xmath103 , @xmath104 , are the same for @xmath105 and different for @xmath106 .
the lemma is proved by full analogy with lemma 16 from @xcite ( using proposition [ propim ] ) . in what follows
we assume that the equations of the limits at @xmath88 of the tangent lines to the curves @xmath96 are given in the form @xmath102 . here
@xmath107 is the slope of the limit of the tangent line .
[ uv ] if @xmath101 , then the tangent lines to @xmath108 and @xmath17 are the same . the curves @xmath108 , @xmath17 and the point @xmath88 are parts of one and the same curve smooth at @xmath88 . _ proof : _ in the particular case when @xmath109 the strata @xmath108 and @xmath17 are the half - lines @xmath110 and @xmath111 . in the general case
the lemma is proved by analogy with lemma 16 from @xcite ( using proposition [ propim ] ) .
@xmath112 [ slopebis ] for the slopes @xmath107 of the limits at @xmath88 of the tangent lines to the curves @xmath96 one has @xmath113 .
_ proof : _ @xmath114 . in the case
when the stratum @xmath74 consists of hyperbolic polynomials the lemma is proved in @xcite , see lemma 22 there .
suppose that @xmath74 does not consist of hyperbolic polynomials .
denote by @xmath115 the stratum whose mv is obtained from @xmath32 by adding to the right @xmath116 components equal to @xmath117 .
hence , @xmath115 consists of hyperbolic polynomials .
choose a point @xmath119 .
connect it with @xmath88 by a continuous curve ( parametrized by @xmath120 $ ] ) passing only through strata with mvs whose first @xmath121 components are the same as the ones of @xmath32 and such that for @xmath13 distinct values @xmath122 of @xmath123 the greatest two of the real roots become equal after what they form a complex conjugate couple .
the slopes @xmath107 can be defined for any @xmath120 $ ] . for different @xmath124
they remain different throughout the deformation ( even for @xmath125 which can be proved like lemma [ slope ] ) . for all @xmath123
they are finite .
hence , their order is the same at @xmath88 and at @xmath126
. @xmath112 [ leftright ] if @xmath124 is even , then the projection on @xmath93 of @xmath96 is `` to the right '' of the one of @xmath88 ; if @xmath124 is odd , then `` it is on its left '' .
_ proof : _ in the case when @xmath74 consists of hyperbolic polynomials this is lemma 18 from @xcite . in the general case
the lemma is proved like the previous one being `` to the left or to the right '' does not change throughout the deformation , because it depends continuously on @xmath123 , i.e. in fact it does not depend on @xmath123 .
@xmath112 it follows from lemmas [ uv ] and [ leftright ] that if @xmath101 and if the projection of @xmath108 on @xmath93 is `` to the right '' ( resp .
`` to the left '' ) of the one of @xmath88 , then the projection of @xmath17 on @xmath93 is `` to the left '' ( resp .
`` to the right '' ) of the one of @xmath88 .
[ updown ] for @xmath124 fixed the curve @xmath127 is `` above '' the curve @xmath128 if and only if either @xmath124 is odd and @xmath129 or @xmath124 is even and @xmath130 . _
proof : _
if the stratum @xmath74 consists of hyperbolic polynomials , then this is lemma 20 from @xcite .
if not , then use the same deformation as in the proof of lemma [ slopebis ] . for any value of @xmath123 and for any
@xmath124 fixed the curves @xmath96 have the same mutual disposition .
indeed , one can apply proposition [ propim ] the curves @xmath96 with one and the same @xmath124 correspond to one and the same neighbourhood @xmath131 and to the respective curves constructed after the polynomial @xmath132 .
the latter s mutual disposition does not depend on @xmath123 or on @xmath5 .
@xmath112 on fig .
2 we show the curves @xmath96 and @xmath17 in the case when @xmath133 , @xmath134 , @xmath135 .
the reader can check the above lemmas on this example . if for some @xmath124 one has @xmath136 , then the mvs of the strata @xmath17 and @xmath137 are the same .
hence , this is one and the same stratum but it gives rise to two ( or more ) different curves @xmath17 at @xmath88 , with different slopes of their tangent lines at @xmath88 lemma [ slope ] ) .
the simplest example of such a situation is the one of the previous section ( the stratum defined by the mv @xmath50 $ ] admitting two different limits of the tangent space along the curve @xmath52 ) .
contractibility follows from the contractibility of the parameter space ( the roots play the role of parameters and define the coefficients @xmath138 via the vieta formulas ; these formulas define a homeomorphism ) .
further we prove some statements of the theorem not for the stratification of @xmath139 , the space of the coefficients @xmath140 ( where @xmath141 is the @xmath124-th symmetric function of the roots of @xmath3 counted with the multiplicities ) , but for the space of the newton functions @xmath142 which are the sums of the @xmath124-th powers of the roots .
the statements formulated for the two spaces ( of the quantities @xmath138 or @xmath142 ) are equivalent because there exist polynomials @xmath143 , @xmath144 such that @xmath118
. denote by @xmath146 the roots of @xmath3 where the real roots are distinct while the complex ones might not be all distinct .
the jacobian matrix corresponding to the mapping @xmath147 is obtained from the vandermonde matrix @xmath148 by multiplying the columns of the real roots by their respective multiplicities . hence , if all complex roots are distinct , then the determinant @xmath149 of the matrix @xmath150 is nonzero and at such a point the mapping is a local diffeomorphism which means that the stratum is locally of dimension @xmath37 . at such a point
one can express the roots @xmath8 as smooth functions of @xmath151 , and then express @xmath152 as smooth functions of @xmath8 ; hence , as smooth functions of @xmath151 .
prove the smoothness of the stratum regardless of whether the complex roots are all distinct or not .
( we use the same ideas here as in @xcite , p. 52 - 53 . ) to this end set @xmath154 , @xmath155 , @xmath156 where all roots of @xmath157 are complex and all roots of @xmath158 are real . the mapping @xmath159 ( where @xmath160 , @xmath161 are regarded as free parameters ) is a local diffeomorphism . indeed
, its jacobian matrix is the sylvester matrix of @xmath157 and @xmath158 ; the latter s determinant equals res@xmath162 which is nonzero because @xmath157 and @xmath158 have no root in common .
@xmath163 . _ the field of tangent spaces to the given stratum is continuously extended to the strata of lower dimension belonging to its closure and to the points where some complex roots coincide .
the extension is everywhere transversal to the space @xmath39 . _
the rest of the theorem follows from the statement .
the latter implies in particular that even in a neighbourhood of a point of the stratum where some complex roots coincide , the stratum is locally the graph of a smooth @xmath14-dimensional function defined on the projection of the stratum in @xmath38 , see @xmath164 .
@xmath165 . to prove the statement from @xmath163
compute the partial derivatives @xmath166 , @xmath167 , @xmath168 bearing in mind that @xmath169 is the @xmath98-th newton function of the roots @xmath8 .
( we follow here the same ideas as the ones used in the proof of theorem 1.8 from @xcite . )
denote by @xmath170 the quantity equal to the multiplicity of @xmath8 if @xmath8 is a real root and to @xmath117 if it is a complex one .
one has @xmath178 . put @xmath179 in ( [ partial ] ) . then for @xmath180 one has @xmath181 ( two proportional columns ) .
suppose first that the roots @xmath182 , @xmath183 are complex .
then one has @xmath184 and @xmath185 ( to be checked directly ) .
this means that the numerator of the right hand - side of ( [ partial ] ) is @xmath186 when @xmath179 , i.e. it is representable in the form @xmath187 for some polynomial @xmath188 .
hence , @xmath191 .
the closure of the stratum can be defined by a continuous parametrization of the roots @xmath8 by some parameters @xmath192 ( one can choose as such parameters part of the variables @xmath169 ; in general , these variables are more than the parameters needed ) . by ( [ partial1 ] ) ,
the partial derivatives @xmath166 are bounded continuous functions of the parameters @xmath192 .
hence , the limits of these partial derivatives exist on the closure of the stratum .
this proves the statement .
dillo 83 v.i .
arnold , a.n .
varchenko , s.m .
guseyn - zade , singularits des applications differentiables .
partie 1 et 2 .
moscou , ed . mir , 1986 .
arnold , v.a .
vasilev , v.v .
goryunov , o.v .
lyashko , singularities . local and global theory . in : v.i .
arnold ( ed . ) , dynamical systems vi , singularity theory i , springer - verlag berlin - heidelberg 1993 .
g. katz , how the tangents solve algebraic equations , or a remarkable ruled geometry of the discriminant varieties .
manuscript 31 p. v. p. kostov , on the hyperbolicity domain of the polynomial @xmath193 .
serdica math .
j. 25 , no . 1 ( 1999 ) , 47 - 70 . v. p. kostov , on the geometric properties of vandermonde s mapping and on the problem of moments .
edinb . 112a ( 1989 ) , 203 - 211 .
i. meguerditchian , gomtrie du discriminant rel et des polynmes hyperboliques .
thse soutenue le 24.01.1991 luniversit de rennes . b. z. shapiro , b. a. khesin , swallowtails and whitney umbrellas are homeomorphic
geom . 1 ( 1992 ) , 549 - 560
. b. z. shapiro , v. welker , combinatorics and topology of stratifications of the space of monic polynomials with real coefficients .
results math .
33 ( 1998 ) , no . 3 - 4 , 338 - 355 | we consider the family of polynomials @xmath0 , @xmath1 , and the stratification of @xmath2 defined by the multiplicity vector of the real roots of @xmath3 . we prove smoothness of the strata and a transversality property of their tangent spaces .
* key words : * multiplicity vector ; multiplicity surplus * ams classification : * 12d10 [ tm]definition [ tm]remark [ tm]remarks [ tm]lemma [ tm]example [ tm]corollary [ tm]proposition |
let @xmath6 be a connected nonorientable surface of genus @xmath5 with @xmath7 holes ( boundary components ) and let @xmath8 denote the mapping class group of @xmath6 , the group of isotopy classes of all diffeomorphisms @xmath9 .
our work is motivated by the following theorem of n. v. ivanov .
( ivanov)[ivanov ] let @xmath10 , @xmath11 , @xmath12 or @xmath13 .
if @xmath14 is neither a sphere with at most @xmath15 holes , nor a torus with at most @xmath16 holes , nor a closed surface of genus @xmath16 , then every automorphism of @xmath17 is given by the restriction of an inner automorphism of @xmath12 .
in particular , @xmath18 is a finite group and moreover , @xmath19 , @xmath20 . in the next section we study
the orientation double cover of a nonorientable surface and we investigate relation between mapping class groups of the nonorientable surface and its orientation double cover .
furthermore , we describe maximal abelian subgroups of the mapping class group of nonorientable surfaces and construct some special type of abelian subgroup of the mapping class group of nonorientable surface . in section [ dehn ] we characterize the dehn twists about circles on a closed nonorientable surface using the ideas of ivanov s analogous work on the orientable surface @xcite . in section [ outaut ] , we prove that the outer automorphism of @xmath21 injects into @xmath4 if @xmath22 , on the other hand , the outer automorphism of @xmath23 injects into the mapping class group of sphere with 4 holes if @xmath24
. finally , we compute the outer automorphism of mapping class groups of some sporadic cases using the presentations of mapping class groups of some nonorientable surfaces not covered in our theorems .
let @xmath25 and @xmath6 denote the connected orientable surface of genus @xmath5 with @xmath7 holes and the connected nonorientable surface of genus @xmath5 with @xmath7 holes , respectively . recall that the genus of a nonorientable surface is the maximum number of projective planes in a connected sum decomposition .
let @xmath26 be the orientation double cover of the nonorientable surface @xmath6 with the deck transformation @xmath27 .
then @xmath28 is a normal subgroup of index @xmath16 in @xmath29 and the deck transformation group is @xmath30 . if @xmath31 is a diffeomorphism of @xmath32 such that @xmath33 , then @xmath31 induces a diffeomorphism @xmath34 on @xmath6 .
let @xmath35 .
so , we have a homomorphism @xmath36 .
moreover , this homomorphism is onto .
indeed , we have an exact sequence @xmath37 on the other hand , any diffeomorphism @xmath34 of the nonorientable surface @xmath6 has a unique lift to an orientation preserving diffeomorphism @xmath38 .
hence , the above exact sequence splits and therefore we can identify @xmath39 with the subgroup @xmath40 , the subgroup of orientation preserving diffeomorphisms of @xmath32 commuting with the deck transformation @xmath41 . indeed , any isotopy of the nonorientable surface @xmath6 lifts to the surface @xmath32 .
moreover , if the euler characteristic of the surface is negative ( hence it carries a complete hyperbolic metric ) then any isotopy of orientation preserving diffeomorphisms of @xmath32 , whose end maps commute with the deck transformation @xmath41 , can be deformed into one which commutes with @xmath41 , and hence descends to an isotopy of @xmath39 ( c.f . lemma 5 in @xcite ) .
therefore , we can identify the mapping class group @xmath42 as the subgroup @xmath43 , the subgroup of orientation preserving mapping classes of @xmath32 commuting with the mapping class of the deck transformation @xmath41 . if @xmath44 is a circle on @xmath6 , by which we mean a simple closed curve , then according to whether a regular neighborhood of @xmath44 is an annulus or a mbius strip , we call @xmath44 two - sided or one - sided simple closed curve , respectively .
we say that a circle is nontrivial if it bounds neither a disc with at most one puncture nor annulus together with a boundary component , nor a mbius band on @xmath6 .
if @xmath44 is a circle , then we denote the complement of @xmath44 by @xmath45 the surface obtained by cutting @xmath6 along @xmath44 .
a circle @xmath44 is called nonseparating if @xmath45 is connected and otherwise , it is called separating . following ivanov @xcite
, we will call a one dimensional submanifold @xmath46 of a surface @xmath14 a system of circles if @xmath46 consists of several pairwise nonisotopic circles .
by @xmath47 we will denote the surface @xmath14 cut along all the circles in the system @xmath46 .
a finite sequence of embedded two - sided circles @xmath48 on the any surface is called a chain
if the geometric intersection number of the isotopy classes @xmath49 and @xmath50 , satisfies @xmath51 , for all @xmath52 , where the geometric intersection number @xmath53 defined to be the infimum of the cardinality of @xmath54 with @xmath55 $ ] , @xmath56 $ ] . for an orientable surface @xmath25 we have the classical definition below .
let @xmath57 be any diffeomorphism .
@xmath58 @xmath34 is called periodic if it is isotopic to some diffeomorphism @xmath59 with @xmath60 for some @xmath61 , @xmath58 @xmath34 is called reducible if there is a system of circles @xmath62 on @xmath25 and a diffeomorphism @xmath59 , isotopic to @xmath34 , with @xmath63 , @xmath58 otherwise , @xmath34 is called pseudo - anosov .
if @xmath34 is periodic , or reducible or pseudo - anosov , its isotopy class @xmath64 is called periodic element , reducible element , pseudo - anosov element , respectively .
it has been shown in @xcite that one can define periodic , reducible and pseudo - anosov diffeomorphisms for nonorientable surfaces exactly in the same way . moreover , if @xmath34 is a diffeomorphism of the nonorientable surface @xmath6 with its unique orientation preserving lift @xmath31 to the orientation double cover @xmath32 , then @xmath34 is periodic , reducible or pseudo - anosov if and only if @xmath31 is periodic , reducible or pseudo - anosov , respectively .
a diffeomorphism @xmath65 is called pure if for some system of circles @xmath46 the following condition is satisfied : all points of @xmath46 and @xmath66 are fixed by @xmath34 , @xmath34 does not interchange components of @xmath67 and it induces on every component of @xmath67 a diffeomorphism , which is either isotopic to a pseudo - anosov one , or to the identity .
let @xmath69 , where @xmath70 , @xmath71 , be the kernel of the natural homomorphism @xmath72 then @xmath69 is a subgroup of finite index in @xmath73 .
let @xmath74 , regarding @xmath8 as a subgroup of @xmath73 as described above .
it is well known that @xmath69 and hence @xmath68 consist of pure elements only provided that @xmath75 , ( @xcite ) . throughout the paper
, we will consider @xmath69 for only @xmath75 .
following ivanov , suppose that @xmath46 is a system of circles on @xmath6 and a diffeomorphism @xmath76 is fixed on each component @xmath77 of @xmath67 .
moreover , suppose that @xmath78 is fixed at all points of the boundary of @xmath77 and is either isotopic to the identity or to a pseudo - anosov map on @xmath77 .
we extend @xmath78 to @xmath79 by the identity .
let the subgroup @xmath80 of @xmath81 be generated by @xmath82 s , which are the isotopy classes of such @xmath83 s , and all dehn twists about two - sided circles in @xmath46 .
the subgroup @xmath80 is abelian .
indeed , in the orientable surface case , ivanov showed that any maximal abelian subgroup of @xmath69 is of this type ( @xcite ) .
the analogous result holds in the nonorientable case also : [ lambdasubgroup1 ] every abelian subgroup of @xmath68 is contained in some abelian subgroup constructed as in the above paragraph .
consider an abelian subgroup @xmath84 .
regarding @xmath85 let @xmath46 be a reduction system for @xmath86 described by ivanov for orientable surfaces @xcite .
for any @xmath87 and @xmath88 , we have @xmath89 , because each @xmath90 is invariant under the involution @xmath91 .
so we have @xmath92 .
let @xmath93 be a circle in the system with @xmath94 .
since @xmath41 has no fixed points , the restriction of @xmath41 to the circle @xmath95 is an orientation preserving involution on that circle .
so each circle of the system @xmath46 maps onto a circle in @xmath6 and hence the system @xmath46 gives a system @xmath96 on @xmath6 .
now @xmath97 , where @xmath80 is an abelian subgroup corresponding to the system of circles @xmath96 , as described in the above paragraph .
the system of circles @xmath96 constructed in the above proof will be called a reduction system for the abelian subgroup @xmath86 .
let @xmath98 be a subgroup of finite index of @xmath69 .
then , the its @xmath41 invariant part , @xmath99 , is also finite index subgroup of @xmath68 .
[ lambdasubgroup2 ] the center of every subgroup of finite index of @xmath68 is trivial .
the centralizer in @xmath69 of a pseudo - anosov element is isomorphic to @xmath4 , ( @xcite ) .
therefore , the centralizer in @xmath68 of a pseudo - anosov element is also isomorphic to @xmath4 . from this
, we obtain that the center of every subgroup of finite index of @xmath68 is trivial .
stukow s result @xcite on dehn twists on nonorientable surfaces shows that dehn twists on nonorientable surfaces enjoy the same properties with those on orientable surfaces .
for instance , two dehn twists commute if and only if their circles can be chosen disjoint ; they satisfy the braid relation if and only if their circles meet transversally at one point . before we go
further we give the following remark . *
a ) * the pure mapping class group of real projective plane with less than or equal to two holes is finite .
* b ) * the pure mapping class group of the klein bottle is finite and that of the klein bottle with one hole has no pseudo - anosov element ( see proposition [ sporadic cases ] ) . +
* c ) * the pure mapping class group of klein bottle with two holes contains @xmath100 .
+ * d ) * the pure mapping class group of @xmath101 ( @xmath102 ) contains a copy of @xmath100 , though that of @xmath103 does not support any pseudo - anosov diffeomorphism ( @xcite ) . using the above remark we deduce that the smallest surfaces which support a pseudo - anosov diffeomorphism in their pure mapping class group which does not have an abelian subgroup of rank two are @xmath104 , @xmath105 , @xmath106 or @xmath107 .
examples of pseudo - anosov mappings on the surfaces @xmath104 , @xmath105 and @xmath106 are well known . for the nonorientable surface @xmath107 ,
the map @xmath108 , described in the figure [ rp2 - 3bc ] , is indeed pseudo - anosov ( theorem 4.1 in @xcite ) .
we will finish this section with the following result , whose proof will be omitted .
let @xmath109 be a maximal abelian subgroup of @xmath110 of highest rank , where @xmath5 is at least three .
then the @xmath111 and each component of @xmath112 is either sphere with three holes , sphere with four holes , torus with one hole , or real projective plane with two or three holes , where @xmath46 is a minimal reduction system for @xmath86 and @xmath113 is @xmath114 or @xmath115 depending on whether @xmath5 is odd or even .
moreover , in both even and odd genus cases , one can choose maximal abelian subgroups of highest rank so that they are generated by dehn twists about nonseparating circles .
throughout this section @xmath116 is a finite index subgroup of @xmath68 , @xmath75 .
[ chr-1 ] let @xmath110 be a compact closed connected nonorientable surface of odd ( respectively , even ) genus @xmath117 ( respectively , @xmath118 ) .
an element @xmath119 is a dehn twist about a nonseparating circle ( respectively , nonseparating circle whose complement is nonorientable ) if and only if the following conditions are satisfied : + * @xmath120 is isomorphic to @xmath4 , for any integer @xmath121 such that @xmath122 .
* there is an abelian subgroup @xmath86 of rank @xmath123 ( respectively , @xmath124 ) generated by @xmath64 and its conjugates freely .
* @xmath64 is a primitive element of @xmath125 .
we will prove the odd genus case mainly following the proof of ivanov in @xcite : the even genus case is basically the same with minor changes , which we will emphasize in the when needed .
if the above conditions are satisfied , then we need to show that @xmath64 is a dehn twist about a nonseparating circle .
let us denote the number @xmath123 ( @xmath124 , in the even genus case ) by @xmath126 .
if @xmath127 is equal to the identity , then @xmath128 and hence by the first condition , @xmath129 is isomorphic to @xmath4 .
however , this is a contradiction by lemma [ lambdasubgroup2 ] .
therefore , @xmath127 is not equal to the identity .
we will denote a minimal reduction system of @xmath127 by @xmath130 , and the subgroup generated by twists about two - sided circles in @xmath130 by @xmath131 .
set @xmath132 . @xmath131 and @xmath133 are free abelian groups
. first we will show that @xmath134 .
let @xmath135 .
this implies that @xmath127 and @xmath5 commute and hence @xmath127 and @xmath5 belong to some abelian subgroup @xmath80 constructed as in the previous section .
let @xmath46 denote the system of circles used in the construction of @xmath80 .
therefore , @xmath130 is isotopic to a subsystem of @xmath46 , because @xmath136 .
thus , we have @xmath137 .
so , each element of @xmath131 commutes with all @xmath138 .
therefore , we obtain that @xmath139 and we obtain that @xmath140 . by the assumption that @xmath141 , we have @xmath130 has no or just one two - sided circle .
suppose that @xmath130 has no two - sided circle .
say @xmath142 , where @xmath143 is a one - sided circle for all @xmath144 . then @xmath145 is connected and the restriction @xmath146 is either the identity or a pseudo - anosov . if @xmath147 is the identity , then @xmath64 must be a product of dehn twists about some circles in @xmath130 which is not possible since each @xmath143 is one - sided . therefore , @xmath146 is a pseudo - anosov .
however , the maximal abelian group containing @xmath64 has rank one . by the assumption @xmath1 and
thus the rank is @xmath148 .
this is a contradiction .
( similarly , in the even genus case , we have @xmath149 . )
so , @xmath130 has exactly one two - sided circle and so @xmath150 , where @xmath44 is a two - sided circle and each @xmath151 is one - sided .
let @xmath152 be the group generated by @xmath127 and the twist about @xmath44 and set @xmath153 .
then , we have @xmath154 and hence @xmath155 is isomorphic to @xmath4 .
it follows that there are no pseudo - anosov diffeomorphisms in the thurston normal form of @xmath127 and thus @xmath127 is a power of the dehn twist @xmath156 about @xmath44 .
so , @xmath157 for some integer @xmath158 .
now we need to show that @xmath44 is nonseparating .
we assume that @xmath44 is a separating circle .
so , @xmath159 such that @xmath160 and @xmath161 . since @xmath44 is nontrivial we have @xmath162 for @xmath163 . without loss of generality ,
assume @xmath164 .
let @xmath165 be the elements conjugate to @xmath64 generating the abelian subgroup @xmath86 in the statement of the theorem .
then @xmath166 , where @xmath167 is a circle .
similarly , @xmath167 separates the surface into two components , say @xmath168 and @xmath169 , which are diffeomorphic to @xmath170 and @xmath171 , respectively . by the structure of abelian groups , the circles
@xmath167 can be chosen to be disjoint and pairwise nonisotopic .
moreover , @xmath168 is diffeomorphic to @xmath172 and @xmath167 is not isotopic to @xmath173 for any @xmath174 .
so , @xmath175 can not be contained inside @xmath172 .
therefore , if @xmath176 , then @xmath177 . since @xmath167 is not isotopic to @xmath173 , we see that @xmath178 .
hence , we see that @xmath179 which is a contradiction .
so , @xmath180 , in other words , the surfaces @xmath168 are all disjoint .
let @xmath181 .
the fact that @xmath167 are being pairwise nonisotopic implies that @xmath182 can not an annulus . also , @xmath182 is not a sphere , a torus , a klein bottle , a projective plane , a disk or a mbius band .
thus , @xmath183 .
now , @xmath184 .
so , @xmath185 . then , @xmath186 . so we get @xmath187 , which implies @xmath188 .
this is a contradiction .
thus , @xmath44 is a nonseparating circle .
( for even genus , we would have @xmath189 , which would yield @xmath190 , a contradiction to the assumption that @xmath191 . )
we take an element @xmath192 .
@xmath193 denotes a diffeomorphism in the isotopy class @xmath194 .
the group @xmath86 is generated by @xmath165 and @xmath195 as before . since @xmath196 , @xmath197 and thus @xmath198 . in other words , @xmath199 and so @xmath200 .
so , replacing @xmath201 by an isotopy we can assume that @xmath202 .
let @xmath203 .
since @xmath126 is the maximal number of pairwise nonisotopic disjoint two - sided circles on @xmath110 , every component of @xmath204 is a disc with two holes or a one holed mbius band .
total number of the boundary components of the components of @xmath204 is @xmath205 and hence all the components of @xmath204 except one must be pair of pants .
( in the even genus case , @xmath206 is the maximal number of nonisotopic two - sided circles on @xmath110 .
now , similar arguments prove that all the components @xmath204 but one are discs with two holes , and the remaining component is a sphere with four holes . ) suppose that @xmath201 sends one of these components to another one .
since @xmath201 fixes the circles(boundary components ) of @xmath207 , we see that @xmath208 where @xmath209 .
this is a contradiction to the assumption that @xmath110 is a nonorientable surface of odd genus .
so @xmath201 does not permute the components .
hence , @xmath201 is a composition of twists about @xmath167 .
( again , in the even genus case , since there is only one sphere with four holes component of @xmath204 , we see that @xmath201 can not interchange the components . )
now note that , @xmath210 because @xmath90 , and @xmath86 is abelian . without loss of generality ,
suppose that @xmath211 .
so , @xmath212 .
so by the above paragraph , we obtain that @xmath64 is a power of @xmath213 ; @xmath214 for some @xmath215 , because @xmath86 is free abelian .
therefore , @xmath216 and @xmath86 is generated by @xmath217 .
thus , @xmath156 is in @xmath125 . by the last condition
, @xmath64 is primitive and hence , @xmath218 . in the even genus case
, one has to prove that the complement of @xmath44 is nonorientable .
assume that the complement of @xmath44 is orientable .
since each @xmath219 is conjugate to @xmath220 the complement of each @xmath221 is also orientable .
it is easy to see that each @xmath221 is a @xmath222 homology cycle poincre dual to the first stiefel - whitney class @xmath223 of the surface @xmath110 . in particular , for any @xmath224 , @xmath225 , @xmath221 and @xmath226 form a boundary of a subsurface @xmath227 of @xmath110 .
moreover , since the complement of each @xmath221 is orientable , @xmath227 is an orientable surface of genus at least one with two boundary components . by relabeling @xmath221
s we may assume that the interior of each @xmath228 ( where we set @xmath229 ) does not intersect any @xmath230 .
hence , @xmath231 , if and only if @xmath232 .
now , @xmath233 , and therefore we get @xmath234 , a contradiction since @xmath191 .
hence , the complement of @xmath44 must be nonorientable .
the other direction of the theorem is straight forward and left to the reader ( see also @xcite ) .
similar arguments prove the following theorems ( compare with the corresponding result in @xcite ) : [ chr-2.1 ] let @xmath1 be an odd integer .
an element @xmath235 is a dehn twist if and only if the following conditions are satisfied : * @xmath120 is isomorphic to @xmath4 , for any integer @xmath121 such that @xmath122 .
* there exists a free abelian subgroup @xmath86 of @xmath236 generated by @xmath64 and @xmath237 twists about two - sided nonseparating circles such that @xmath238 .
* @xmath64 is a primitive element of @xmath125 .
for dehn twists on even genus surfaces we have , [ chr-2.2 ] let @xmath191 be an even integer .
an element @xmath235 is a dehn twist if and only if the following conditions are satisfied : * @xmath120 is isomorphic to @xmath4 , for any integer @xmath121 such that @xmath122 .
* there exists a free abelian subgroup @xmath86 of @xmath236 generated by @xmath64 and @xmath239 dehn twists about nonseparating circles whose complements are nonorientable , such that @xmath240 .
* @xmath64 is a primitive element of @xmath125 .
[ ch-3.2 ] * 1 ) * let @xmath241 be an automorphism , where the genus of @xmath242 is at least five . since @xmath116 is a subgroup of @xmath68 of finite index the subgroup @xmath243 is still a finite index subgroup of @xmath68 .
hence the above theorems give an algebraic characterization of a dehn twist .
therefore , we can also characterize algebraically the subgroup @xmath244 of @xmath245 generated by all dehn twists .
+ * 2 ) * a two - sided embedded circle on a closed nonorientable surface of odd genus is either separating or nonseparating with nonorientable complement .
therefore , by theorem [ chr-1 ] and theorem [ chr-2.1 ] an automorphism of the mapping class group respects the topological type of the circles of the dehn twists .
however , in case of nonorientable surfaces of even genus , theorem [ chr-1 ] and theorem [ chr-2.2 ] are not strong enough to distinguish a dehn twist about a separating circle from a dehn twist about a nonseparating circle with orientable complement . to make sure that automorphisms respect the topological type of the circles the dehn twists are about , we need to study chains of dehn twists .
now we would like to discuss an application of the above theorems to the chains of dehn twists , which we will use in the study outer automorphisms of mapping class groups of nonorientable surfaces . whenever we consider a chain or a tree of dehn twists we will choose the circles the dehn twists are about so that they intersect minimally ( e.g. if the surface has negative euler characteristic then we put an hyperbolic metric on it and choose the unique geodesics from each isotopy class ) .
stukow s results ( @xcite ) mentioned in section @xmath16 together with the above theorems imply that any chain of dehn twists is mapped to a chain of dehn twists under an automorphism of the mapping class group . on an orientable surface ,
a chain of circles is maximal if and only if it is separating . moreover , in this case the chain has odd number of circles .
this is no longer true on nonorientable surfaces of even genera . for example , a two sided circle with orientable complement forms a maximal chain of length one ; however it is not separating .
nevertheless , we have the following result . [ chain ]
let @xmath110 be the nonorientable surface of genus @xmath5 .
* if a chain of two sided nontrivial circles @xmath246 on the closed surface @xmath110 is maximal then @xmath215 is an odd integer .
moreover , if @xmath5 is an odd integer , then , the chain is maximal if and only if the chain separates the surface into two components . + in case of surfaces of even genus @xmath5 , the chain @xmath246 is maximal if and only if it is ether separating , or nonseparating with orientable complement . *
let @xmath247 be a separating maximal chain of two - sided nontrivial circles on @xmath110 .
then it separates a disc if and only if there is no dehn twist which commutes with all of the dehn twists about the circles of the chain , but @xmath248 .
the proof of the part ( i ) is left to the reader . for the statement in part ( ii ) ,
figure [ 2o ] , which establishes the proof in two special cases , indeed provides the proof of the if " part of the statement .
for the other direction assume that there is a smooth curve @xmath95 which intersects the chain transversally at finitely points of only @xmath249 and does not meet any other circle of the chain .
let @xmath14 be any one of the two subsurfaces obtained by cutting the surface @xmath110 along the chain .
since @xmath249 is the first element of the chain , the @xmath249 part of the boundary of @xmath14 is connected .
note that @xmath250 is a disjoint union of arcs .
now , if @xmath14 is a disc we can isotope these arcs outside the surface @xmath14 so that @xmath95 does not intersect the chain anymore .
as mentioned in remark [ ch-3.2 ] ( 1 ) we can use chains of dehn twists to distinguish algebraically dehn twists about separating circles from dehn twists about nonseparating circles with orientable complements on the nonorientable surface of even genus @xmath110 : indeed , the longest chain on @xmath110 , which lies in the complement of a nonseparating circle with orientable complement is @xmath251 ( see also lemma [ lemma - y - hom ] ) . on the other hand , in the complement of a nontrivial separating circle the maximal length of a chain is again @xmath251 but in this case the circle separates the surfaces into two components , a klein bottle with one boundary component and an orientable surface of genus @xmath252 with one boundary component .
hence , in the case of nonseparating circle with orientable complement one can find one maximal chain of length @xmath251 , whereas in the case of separating circles one can find two maximal chains , which lie in different components ( and thus any element of one chain commutes with any element of the other chain ) of lengths @xmath251 and @xmath114 .
these arguments prove the following : [ ch3.3 ] let @xmath191 be an even integer .
an automorphism of the mapping class group of @xmath110 maps a dehn twist about a separating circle to a dehn twist about a separating circle .
hence , an automorphism respects the topological types of the circles of the dehn twists .
we will end this section with a result which will be useful in the latter part of the paper .
[ chains - bd ] let @xmath1 be an integer .
an automorphism of the mapping class group of @xmath110 maps any maximal chain of dehn twists on @xmath110 , which separates a disc , to a maximal chain of dehn twists , which separates a disc . by lemma [ chain ]
it is enough to prove that the image of a separating chain must be again separating . in the case of surfaces of odd genus
a chain is separating if and only if it is maximal , which is a property preserved by algebraic automorphisms .
so we may assume that @xmath247 is a separating chain on the nonorientable surface @xmath110 of even genus and the chain separates a disc from the surface .
hence , @xmath110 cut along the chain is the disjoint union of a disc and a nonorientable surface with one boundary component of genus @xmath253 , which is a positive even integer because @xmath5 is even and the complement must be nonorientable .
therefore , there is an element , in the centralizer of the chain ( i.e. , it is a @xmath254 homeomorphism which commutes with all elements of the chain ) which is not a product of dehn twists .
now assume that the image @xmath255 , of this chain under an automorphism of the mapping class group , is nonseparating .
so by the above lemma , the complement of the chain is an orientable surface , say @xmath14 , with two boundary components of genus @xmath256 .
now any mapping class , which is in the centralizer of the chain @xmath255 , has a representative @xmath34 , a diffeomorphism of @xmath110 , which maps a tubular neighborhood @xmath257 , of the chain onto itself .
so , @xmath34 induces a diffeomorphism of both @xmath257 and @xmath258 . since both pieces are orientable @xmath34 is isotopic to a product of dehn twists .
however , this contradicts to the fact that dehn twists and hence their products are respected by automorphisms of the mapping class group .
indeed , similar arguments can used to prove that the topological type of other separating dehn twists are preserved under automorphisms of the mapping class group . as an application of the previous results we will an algebraic characterization of y - homeomorphisms . by definition a y - homeomorphism on a nonorientable surface
@xmath6 is a diffeomorphism whose isotopy class is not a product of dehn twists but its square ( composition with itself ) is a dehn twist about a separating circle which bounds a one holed klein bottle ( c.f . see @xcite ) .
first we give some easy observations : [ lemma - y - hom ] the longest maximal chain in @xmath259 ( respectively , in @xmath260 ) has length @xmath261 ( respectively , @xmath262 ) . the longest chain in the centralizer of a @xmath254-homeomorphism in @xmath259 ( respectively , in @xmath260 ) has length @xmath262 ( respectively , @xmath263 ) .
we will give the proof of the odd genus case only .
the proof of the first statement is easy . for the second statement suppose that there is a maximal chain in the centralizer of a @xmath254-homeomorphism in @xmath259 of length of @xmath264 .
let @xmath257 be a tubular neighborhood of the chain of circles the dehn twists of the chain are about .
@xmath265 . a maximal chain should be of odd length and thus @xmath266 is odd
. then @xmath257 has two boundary components .
since @xmath257 is the orientable surface of genus @xmath267 with two boundary components , @xmath268 , we have @xmath269 .
this is a contradiction since @xmath270 must contain a one holed klein bottle .
hence , we get @xmath271 . finally , it is easy to see that chains of claimed length exist in each case .
let @xmath272 denote the normal subgroup generated by dehn twists .
an element @xmath273 is a @xmath254-homeomorphism if and only if + i ) @xmath274 , @xmath275 , where @xmath276 is a dehn twist about separating circle @xmath277 .
+ ii ) there is a chain of dehn twist about nonseparating circles @xmath278 each of which has nonorientable complement , and such that @xmath279 , for all @xmath280 , @xmath281 , @xmath282 .
the tubular neighborhood of the union of the circles corresponding to the chain @xmath283 is the orientable surface of genus @xmath284 with two boundary components and hence , by assumption , @xmath64 is supported in @xmath285 , which is a copy of @xmath286 with boundary components , say @xmath249 and @xmath287 .
since @xmath277 is separating in @xmath288 and is nontrivial , @xmath277 must separate a klein bottle .
also @xmath289 and @xmath277 do not meet but they are both contained in @xmath286 and hence we see that @xmath289 is the unique two - sided circle in the klein bottle bounded by @xmath277 . by assumption the complement of @xmath289 is nonorientable and hence complement of klein bottle must be nonorientable .
now , @xmath64 must be a product of @xmath290 , @xmath291 and some diffeomorphism supported in the klein bottle bounded by @xmath277 .
however , @xmath292 implies that @xmath64 has no @xmath290 and @xmath291 factors and hence must be supported in the klein bottle bounded by @xmath277 .
hence , it is a @xmath254-homeomorphism .
the other part is easy and left to the reader .
an element @xmath293 is a @xmath254-homeomorphism if and only if + i ) @xmath274 , @xmath275 , where @xmath276 is a dehn twist about separating circle @xmath277 .
+ ii ) @xmath294 contains two chains of dehn twists @xmath295 and @xmath296 such that all but @xmath297 are about nonseparating circles , satisfying for all @xmath144 , @xmath298=1 $ ] and @xmath299 .
+ iii ) the chain @xmath296 is separating but does not separate a disc and the chain @xmath297 can not be chosen of bigger length .
let @xmath257 be the tubular neighborhood of @xmath300 .
then @xmath301 and hence , @xmath302 also , @xmath257 and hence @xmath303 has two boundary components ( the chain @xmath296 has odd number of elements ) . since it is also a separating chain we see that @xmath304 or @xmath305 .
so , @xmath277 must be the boundary of @xmath306 or @xmath106 .
the second case is not realizable by the condition ( iii ) because otherwise @xmath277 would bound a one holed torus and we could replace the chain @xmath297 with a longer chain @xmath307 in the one holed torus . since @xmath308=1 $ ]
, we can isotope @xmath64 to fix @xmath309 pointwisely . since @xmath310=1 $ ] and @xmath311 , we can isotope @xmath64 to fix @xmath312 pointwisely .
similarly , we can assume that @xmath64 fixes @xmath313 pointwisely .
hence , @xmath64 is supported outside of @xmath257 , because @xmath64 can not interchange the two sides of the tubular neighborhood . moreover , since @xmath314 and @xmath315 does not support any nontrivial diffeomorphism which is identity on the boundary we see that @xmath64 is supported in @xmath306 .
so , @xmath64 is an element of @xmath316 which is not a dehn twist so that @xmath275 , where @xmath317 . in particular
, @xmath297 is the dehn twist about the unique ( up to isotopy ) two - sided circle , say @xmath44 , in the one holed klein bottle bounded by @xmath277 .
hence @xmath64 is a y - homeomorphism .
the other direction is left to the reader .
[ y - hom ] using the notation of the above proof , let us continue our investigation of y - homeomorphisms in the mapping class group of the klein bottle with one boundary component .
we have the presentation ( @xcite ) @xmath318 where @xmath220 is the dehn twist @xmath297 .
one can easily see that @xmath319 where @xmath320 is @xmath321 .
it is not hard to see that , any word in @xmath322 has the form @xmath323 for @xmath324 , and this form is unique .
let @xmath325 .
then @xmath326 .
so , @xmath327 equals @xmath328 if @xmath158 is even , or @xmath329 if @xmath158 is odd .
hence , any @xmath330 satisfies the equation @xmath331 , @xmath332 .
since @xmath333 up to conjugation @xmath334 and @xmath335 are the only @xmath254-homeomorphisms supported inside a punctured klein bottle .
[ mainthm1 ] @xmath337 injects into @xmath4 if @xmath22 .
it is well known that the dehn twists @xmath219 , @xmath338 and a y - homeomorphism , say @xmath339 , supported in the tubular neighborhood of the union of the circles @xmath340 and @xmath277 ( which is a copy of one holed klein bottle ) generate the mapping class group @xmath341 ( c.f . see @xcite ) .
let @xmath342 be an automorphism
. by the results of the previous section the image of this tree , which we denote by @xmath244 , will be again a tree , where each maximal chain in @xmath244 bounding a disc will be sent to a maximal chain in @xmath343 bounding a disc . indeed , to get the maximal tree of circles corresponding to the maximal tree of dehn twists @xmath344
, @xmath338 , we can choose a hyperbolic metric on @xmath345 , and then let @xmath346 be the unique geodesic in the homotopy class of a circle corresponding to the dehn twist @xmath344 .
in particular , @xmath347 , for each @xmath144 ( see also @xcite ) .
the tubular neighborhood of the tree , say @xmath257 , of @xmath244 is an orientable surface .
choose an orientation on @xmath257 and on each @xmath167 so that the intersections @xmath348 ( @xmath349 ) is compatible with the orientation .
choose an orientation of a tubular neighborhood of the tree @xmath350 and a diffeomorphism of @xmath351 onto @xmath352 .
keeping track of orientations we can extend this diffeomorphism to whole of @xmath244 and hence to a neighborhood of @xmath244 , call it again @xmath257 .
this diffeomorphism extends to a diffeomorphism of the surface since we need to extend through discs and a one holed real projective plane .
say , @xmath353 is one such diffeomorphism .
composing @xmath354 with @xmath355 , we may assume that @xmath354 fixes all @xmath356 .
let @xmath357 , where @xmath339 is as above .
since @xmath339 commutes with all @xmath356 for @xmath358 , so is @xmath359 .
since a power of @xmath360 is a composition of @xmath356 s , @xmath361 , @xmath359 commutes with @xmath360 also ( figure [ grp2c ] ) .
we can isotope @xmath359 to fix @xmath95 pointwisely .
note that @xmath359 can not interchange the two sides of a tubular neighborhood of @xmath95 since the genus of @xmath345 is odd .
so , we may assume that @xmath359 is identity on a tubular neighborhood of @xmath95 .
so , @xmath359 induces a diffeomorphism on @xmath362 which is the disjoint union of @xmath363 and @xmath364 . since @xmath359 commutes with all @xmath356 s , @xmath358 , with @xmath360 , and the center of @xmath365 is @xmath366 we may assume that after an isotopy , @xmath359 is identity on @xmath363 .
hence , it is supported on @xmath364 .
now , we need to study @xmath367 .
let us rename the circles @xmath340 , @xmath368 , @xmath369 and @xmath370 , and the dehn twists about them by the letters @xmath371 , @xmath372 , @xmath373 and @xmath374 , respectively .
it is known that @xmath367 has the following presentation @xcite : generators:@xmath371 , @xmath373 , @xmath374 and @xmath339 .
+ relations : + @xmath375 @xmath376 + @xmath377 @xmath378 , and @xmath379 + @xmath380 @xmath381 + @xmath382 @xmath383 + @xmath384 @xmath385(@xmath386 ) .
generators : @xmath371 , @xmath373 , @xmath374 and @xmath388 .
+ relations : + @xmath389 @xmath376 + @xmath390 @xmath378 , and @xmath379 + @xmath391 @xmath392 + @xmath382 @xmath393 + @xmath394 @xmath395 + @xmath396 @xmath397(@xmath386 ) .
let @xmath402 which is @xmath403 .
note that @xmath404 and @xmath405 , for all @xmath144 . also , @xmath406 and @xmath407 .
set @xmath408 .
it is easy to see that this is a normal subgroup of @xmath367 generated by dehn twists and hence @xmath409 .
nevertheless , @xmath410 and hence @xmath401 has index two in @xmath367 .
so , @xmath401 must be the subgroup generated by all dehn twists .
now , any element of @xmath367 which is not a product of a dehn twist has the form @xmath411 for some @xmath412 . in particular , @xmath413 for some @xmath152 . since @xmath414 , @xmath415
, we see that @xmath416 , where @xmath417 .
so , @xmath418 .
apply @xmath354 to @xmath422 to get @xmath423 .
it follows that @xmath424 .
similarly , apply @xmath354 to @xmath425 to get @xmath426 .
so , @xmath152 maps a tubular neighborhood of @xmath371 and @xmath374 diffeomorphically onto itself . by composing @xmath152 with @xmath427
if necessary we may assume that @xmath152 is identity on a tubular neighborhood of @xmath371 and @xmath374 , where @xmath428 is the boundary of the tubular neighborhood of the union of @xmath371 and @xmath374 .
the complement of the tubular neighborhood of the union of @xmath371 and @xmath374 is a projective plane with two boundary components ; let us call it @xmath429 . since @xmath430 , we get @xmath431 for some @xmath158 and @xmath7 @xmath432 . now , use the relation @xmath433 to get @xmath434 .
it follows that @xmath435 .
so , @xmath436 .
hence , we associate an integer , namely @xmath158 to each element of @xmath437 .
indeed , this is a group homomorphism @xmath438 .
to see this first note the following .
let @xmath439 and @xmath440 @xmath441 , so that @xmath439 and @xmath440 are identity on all @xmath356 , @xmath442 , @xmath374 and @xmath443 and @xmath444 .
so , @xmath445 .
to finish the proof of the theorem we need to show that such an automorphism ( if it exist ) can not be an inner automorphism . to see that this is not an inner automorphism note the following : any diffeomorphism which commutes with the all @xmath356 is isotopic to the identity on the orientable component of @xmath446 .
so , it must be isotopic to the identity on the whole surface @xmath345 . however , the automorphism induced by the identity diffeomorphism is the identity .
so , it should send @xmath388 to @xmath388 also .
it is easy to see in this case that the exponent @xmath158 of @xmath447 must be zero . in particular , the homomorphism @xmath438 is injective .
the idea of this proof is the same as the odd genus case .
however , there are some technical differences .
consider the maximal tree of dehn twists @xmath219 , @xmath451 , in figure [ 2grp2 ] , together which a @xmath254-homeomorphism supported in the one holed klein bottle bounded by the circle @xmath277 generate the mapping class group @xmath23 ( c.f .
see @xcite ) .
the maximal tree above has tubular neighborhood diffeomorphic to @xmath452 .
so , if @xmath354 is an automorphism of @xmath453 , let @xmath454 , @xmath455 , then @xmath456 form a tree as above .
similar to the odd genus we can find a diffeomorphism of a tubular neighborhood of the tree formed by @xmath167 s ( @xmath455 ) , call @xmath244 onto that by @xmath457 s , call @xmath458 .
each maximal subchain bounding a disc is sent to a maximal subchain bounding a disc .
therefore , we can find a diffeomorphism @xmath459 , which are both a copies of the orientable surface @xmath460 .
then @xmath461 .
so , @xmath462 .
since , @xmath460 has two boundary components @xmath463 must be a cylinder and the ends are glued to @xmath460 so that the surface becomes nonorientable . in particular , we see that @xmath288 is obtained by gluing the two boundary components of @xmath460 .
now , it is easy to extend the diffeomorphism @xmath464 to @xmath242 , say @xmath465 .
note that by construction @xmath466 , for all @xmath467 , where @xmath464 is a diffeomorphism of orientable subsurfaces @xmath468 onto @xmath469 .
so , the automorphism @xmath470 from @xmath453 to @xmath453 sends @xmath356 to @xmath471 , @xmath467 .
composing @xmath354 with @xmath355 we may assume that @xmath472 @xmath467 .
since @xmath473 commutes with all other @xmath356 it should be supported in the one holed klein bottle obtained by cutting the surface @xmath288 along @xmath474 , @xmath475 and @xmath476 .
however , there is a unique nontrivial two - sided circle inside this one holed klein bottle , which is @xmath340 .
so , @xmath477 . since @xmath354 fixes all @xmath478 any diffeomorphism supported in @xmath286 will be sent by @xmath354 to a diffeomorphism supported in @xmath286
so , @xmath354 induces an automorphism of the mapping class group of the surface with boundary @xmath286 ( klein bottle with two boundary components ) , obtained by cutting the surface along the circles @xmath474 and @xmath476 .
we will rename some of the circles on @xmath288 so that the presentation in @xcite would not be confusing : the circles @xmath474 , @xmath475 , @xmath368 , @xmath340 and the dehn twists about them will be denoted @xmath484 , @xmath485 , @xmath486 and @xmath374 , respectively . in @xcite , it is given that the mapping class group of @xmath479 is generated by the dehn twists @xmath484 , @xmath485 , @xmath486 , @xmath374 and a y - homeomorphism called @xmath487 .
the relation @xmath488 of @xmath481 gives @xmath489 in @xmath480 , for some @xmath490 and @xmath491 , where @xmath492 and @xmath493 are the boundary components of @xmath479 cut along @xmath484 .
moreover in the mapping class group of @xmath288 , we get @xmath494 .
the relation @xmath488 holds in the mapping class group of @xmath479 ( theorem 2.1 ( 11 ) of @xcite ) .
now lemma 4.1 of @xcite finishes the proof of the first statement .
the second statement holds trivially .
let @xmath401 be normal closure of @xmath356 s and @xmath495 .
hence , we have @xmath496 .
since @xmath354 fixes all @xmath356 s and @xmath495 , @xmath354 induces an automorphism @xmath497 of @xmath498 and so that @xmath499 .
so , @xmath500 for some @xmath501 .
apply @xmath354 to @xmath502 and @xmath503 to get @xmath504 .
this implies that @xmath505 and similarly @xmath506 .
also , @xmath507 yields @xmath508 .
it follows that @xmath509 .
so , @xmath510 commutes with @xmath492 , @xmath493 and @xmath373 . indeed , by similar arguments @xmath510 commutes with all @xmath356 , @xmath511 .
clearly , @xmath510 can not change the two sides of tubular neighborhoods of @xmath492 and @xmath493 and hence @xmath510 is identity on a neighborhood of @xmath492 and @xmath493 .
the mapping class @xmath510 can not change the two sides of a tubular neighborhood of @xmath373 also , since otherwise we would have @xmath512 .
so , @xmath510 is an element of the mapping class group of the four holed sphere component of the surface @xmath288 cut along the curves @xmath492 , @xmath493 and @xmath373 : [ sporadic cases ] * 1 ) * @xmath519 and hence @xmath520 , the symmetric group on three letters .
+ * 2 ) * @xmath521 . + * 3 ) * @xmath522 contains a copy of @xmath523 .
+ * 4 ) * @xmath524 contains a copy of the infinite cyclic group @xmath525 .
generators : @xmath371 , @xmath373 , @xmath374 and @xmath388 . relations : + @xmath389 @xmath376 + @xmath390 @xmath378 , and @xmath379 + @xmath391 @xmath392 + @xmath382 @xmath393 + @xmath394 @xmath395 + @xmath396 @xmath397(@xmath386 ) , and the automorphism is given by @xmath398 , @xmath399 , @xmath400 and @xmath526 .
it is easy to check that this map is really an automorphism of the mapping class group .
the fact that any power of it is not an inner automorphism follows from the fact that any mapping class commuting with the dehn twists @xmath484 , @xmath486 and @xmath374 should be a power of the dehn twist about @xmath95 .
however , @xmath527 for any @xmath528 .
this finishes the proof .
this work has been initiated and partially completed during my stay at the michigan state university mathematics department in the spring of 2007 .
i would like to thank the mathematics department and specially to prof .
n. ivanov and prof .
s. akbulut for invitation . | let @xmath0 be the connected closed nonorientable surface of genus @xmath1 and @xmath2 denote the mapping class group of @xmath0 .
we prove that the outer automorphism group of @xmath3 is either trivial or @xmath4 if @xmath5 is odd , and injects into the mapping class group of sphere with four holes if @xmath5 is even . |
high precision electroweak measurements have been used extensively to test the standard model ( sm ) predictions .
the sm successfully describes the results of most particle and nuclear physics experiments and it is believed that any deviation from its predictions would indicate the existence of new physics .
lepton universality being a key characteristics of the sm , the @xmath6 @xmath7 decay , which is the simplest leptonic decay among the @xmath8 decay channels , is therefore one of the best channels to perform such tests .
any violation of lepton universality clearly indicates the existence of new physics beyond the sm . by measuring the ratio of leptonic decay widths of the charged kaon , @xmath1 , defined as
@xmath9 the @xmath10 hadronic form factor cancels out and the sensitivity to new physics is greatly enhanced by the helicity suppression in @xmath11 .
the sm prediction can be determined with very high precision ( @xmath12 ) @xcite , giving a value of @xmath13 where @xmath14 is a radiative correction due the internal bremsstrahlung process ( ib ) which , by definition , contributes to @xmath1 ( unlike the structure dependent process ( sd ) ) . a minimal super - symmetric ( susy )
extension of the sm ( mssm ) with r - parity conservation has recently been considered as a candidate for new physics to be tested by @xmath1 @xcite . in the case of @xmath10 , in addition to the @xmath15 exchange , a charged higgs - mediated susy lepton flavor violating ( lfv ) contribution can be enhanced if accompanied by the emission of a @xmath16 neutrino .
this effect can be described as @xmath17 where @xmath18 is the mass of the charged higgs and @xmath19 ( @xmath20 ) is the term induced by the exchange of a bino and a slepton .
taking @xmath21 , @xmath22 and @xmath23 gev , would yield to a value of @xmath24 .
it is therefore possible to reach a contribution at the percent level due to possible lfv enhancements arising in susy models . + from ( a ) sm and ( b ) lfv susy .
a charged higgs - mediated lfv susy contribution can be strongly enhanced by the emission of a @xmath16 neutrino.,width=529 ] recent in - flight decay experiments , such as kloe @xcite and na62 @xcite , have measured the @xmath1 ratio leading to a current world average of @xmath25 .
the e36 experiment at the japan proton accelerator complex ( j - parc ) by the trek ( time reversal experiment with kaons ) collaboration aims to provide a competitive measurement of @xmath1 with different systematics .
the e36 experiment , which is part of the trek program at j - parc , was set up and fully commissioned at the k1.1br kaon beam line between fall 2014 and spring 2015 .
the data collection was completed by the end of 2015 using an upgraded version of the kek - ps e246 12-sector superconducting toroidal spectrometer @xcite used in a previous t - violation experiment at kek @xcite .
the incoming @xmath4 is tagged by the fitch cherenkov counter before stopping and decaying in the active target , which consists of 256 scintillating fibers ( @xmath26 mm@xmath27 ) oriented in the direction of the beam , providing a precise kaon stop location in the transverse plane . wrapped around the fiber target is a spiral fiber tracker ( sft ) , which consists of two pairs of scintillating fiber ribbons of opposite helicity bundled together around the target , providing the longitudinal coordinate of the outgoing decay particles @xcite .
this target+sft system is surrounded by 12 time - of - flight counters ( tof1 ) and 12 aerogel @xcite counters ( ac ) aligned with the 12 sectors of the toroidal spectrometer and forms the `` central detector '' .
+ a highy segmented ( 768 crystals ) large acceptance csi(tl ) photon calorimeter barrel covering about 75% of the total solid angle is used to identify and correct for structure dependent ( sd ) background events .
the calorimeter features 12 holes ( known as `` muon holes '' ) aligned with the sectors of the spectrometer allowing charged particles , such as @xmath28 , @xmath29 and @xmath30 , to be momentum analyzed by tracking with the multi - wire proportional chambers ( mwpc ) c2 , c3 and c4 . at the exit of each magnet sector ,
another set of fast scintillator detectors ( ttc and tof2 ) and lead glass counters ( pgc ) @xcite provides both trigger signals and @xmath31 particle identification .
three particle identification systems allow to redundantly distinguish between positrons , muons and pions .
the threshold aerogel cherenkov ( ac ) counters sensitive to positrons surround the target bundle , the time - of - flight ( tof ) is measured between scintillators near the target ( tof1 ) and in each gap ( tof2 ) , and lead glass counters ( pgc ) are located at the end of each gap to identify positrons by their shower .
+ the @xmath30 detection efficiency is estimated to be greater than 98% for the aerogel ( ac ) and the lead glass ( pgc ) counters and the mis - identification probability to be about 3% and 4% respectively .
the typical path length between tof1 and tof2 is about 250 cm and the time of flight is determined with a resolution of 200 ps ; the time difference between @xmath30 and @xmath32 at @xmath33 mev/@xmath34 along this path being approximately 500 ps .
+ figure 3 shows the particle momentum from tracking as a function of the adc pulse height for ac ( top left panel ) and for pgc ( top right panel ) .
the bottom left panel shows the momentum versus the squared mass from the time - of - flight analysis . after applying threshold cuts in ac and pgc adcs ,
the remaining events are shown in the bottom right panel where @xmath35 is suppressed , @xmath36 strongly reduced , and a clear separation between @xmath36 and @xmath37 events is observed .
the expected @xmath11 region is circled . + it is important to note that the analysis of the e36 experiment data is at an early stage and that these results are very preliminary .
refinements of the particle identification method are being implemented and further improvements are expected .
this work is supported in part by nserc and nrc in canada , the kaken - hi grants - in - aid for scientific research in japan , the doe and nsf in the us , and in russia by the ministry of science and technology .
9 cirigliano v and rosell i 2007 _ phys . rev . lett . _ * 99 * 231801 masiero a , paradisi p and petronzio r 2006 _ phys
_ d * 74 * 011701 masiero a , paradisi p and petronzio r 2008 jhep * 0811 * 042 girrbach j , nierste u 2012 [ arxiv:1202.4906 ] ambrosino f _ et al . _ ( kloe collaboration ) 2009 _ eur .
c * 64 * 627 lazzeroni c _ et al . _
( na62 collaboration ) 2013 _ phys . lett .
_ b * 719 * 326 macdonald j a _ et al . _ 2003 nucl .
instr .. and meth . a * 506 * 60 abe m _
( e246 collaboration ) 1999 _ phys .
lett . _ * 83 * 4253 abe m _
( e246 collaboration ) 2004 _ phys .
lett . _ * 93 * 131601 abe m _ et al . _
( e246 collaboration ) 2006 _ phys .
_ d * 73 * 072005 mineev o _
( trek collaboration ) the design and performance of a spiral fiber tracker for the j - parc e36 experiment [ arxiv:1607.00219 ] tabata m _
( trek collaboration ) 2015 _ nucl .
instr . and meth .
_ a * 795 * 206 miyazaki y _ et al . _
( trek collaboration ) 2015 _ nucl .
instr . and meth .
_ a * 779 * 206 | the e36 experiment recently conducted at j - parc by the trek collaboration will provide a precise mesurement of the decay ratio @xmath0 with the aim of testing lepton universality , a basic assumption of the standard model ( sm ) .
the sm prediction for the decay ratio @xmath1 is highly precise ( @xmath2 ) and any observed deviation from this prediction would clearly indicate the existence of new physics .
the e36 experimental apparatus , designed to offer small systematic errors , was installed at the j - parc k1.1br kaon beam line in the fall of 2014 , fully commissioned in the spring of 2015 , and the production data were collected in the fall of 2015 .
a scintillating fiber target was used to stop a beam of @xmath3 per spill .
the @xmath4 decay products were then momentum analyzed by a 12-sector superconducting toroidal spectrometer and charged particles were tracked with high precision by multi - wire proportional chambers ( mwpc ) combined with a photon calorimeter with a large solid angle ( 75% of @xmath5 ) and 3 different particle identification systems .
details of the e36 experimental apparatus , as well as the status of the data analysis will be presented . |
among the magnetic materials in which geometrical frustration plays an important role , the most intensely studied are the systems with triangular motif lattices such as stacked triangular lattices @xcite , two - dimensional ( 2d ) and three - dimensional ( 3d ) kagom lattices @xcite , and 3d networks of corner - sharing ( mostly found in pyrochlore @xcite and spinel @xcite compounds ) or edge - sharing tetrahedra @xcite .
the honeycomb lattice , where magnetically active ions form a 2d network of edge - sharing hexagons , is not a geometrically frustrated lattice if only the first neighbour exchange interactions are considered .
however , since it has the smallest possible 2d coordination number ( @xmath5 ) , quantum fluctuations are expected to be larger than for a square lattice , which justifies the considerable theoretical interest in the honeycomb lattice .
frustration can still originate from competing exchange interactions due to next - nearest neighbour interactions or additional coupling out of the plane .
the honeycomb lattice @xcite can be viewed as composed of two interlacing triangular sublattices with each site having its three nearest neighbours on the other sublattice and its six second - nearest neighbours on its own sublattice @xcite .
recently , karunadasa _ et al . _
@xcite reported on the crystal structure and magnetic properties of the family of materials with general formula sr@xmath6o@xmath7 , where @xmath8 is gd , dy , ho , er , tm and yb . these materials crystallise in the cafe@xmath9o@xmath7 ( calcium ferrite)-type structure @xcite and have been studied for a number of years now .
what had not been realised prior to karunadasa s publication , however , is the fact that in these compounds ( and in several other isostructural compounds , such as ba@xmath6o@xmath7 @xcite , pr@xmath9bao@xmath7 @xcite ) the magnetic @xmath8 ions are linked through a network of triangles and hexagons , where geometrical frustration can arise , provided that some exchange interactions are antiferromagnetic . in these systems ,
the honeycomb lattice is distorted in that the hexagons of magnetic rare - earth sites , octahedrally coordinated by oxygen ions , adopt a buckled configuration .
the honeycomb layers are stacked along the @xmath10 axis and each of the magnetic sites in one layer is connected to one of the sites in an adjacent honeycomb layer , forming a triangular network perpendicular to the honeycomb planes .
the magnetic susceptibility measurements reported by karunadasa for @xmath0 ( seo ) present no sharp features or pronounced changes in slope down to at least 1.8 k. yet the value of the curie - weiss constant @xmath11 is -13.5 k , which suggests the presence of relatively strong antiferromagnetic interactions in this system @xcite .
this disparity between the ordering temperature and the scale of the exchange interactions is typical for frustrated systems , although a significant reduction of the transition temperature may also arise from the low - dimensionality of the magnetic lattice . in the present work we report on the low - temperature properties of seo , investigated by powder neutron diffraction and single crystal specific heat and magnetisation measurements .
the system is found to order magnetically at @xmath2 k with a collinear magnetic structure .
however , only one of the two er sites possesses a sizeable magnetic moment of 4.5 @xmath4 at @xmath12 k. the magnetic moments point along the @xmath10 axis and along this direction form ferromagnetic chains , which are arranged antiferromagnetically with respect to one of the neighbouring chains .
these ordered
double " chains are separated by chains which contain almost no ordered magnetic moments ( less than 0.5 @xmath4 at @xmath12 k ) .
it is very likely that long - range and short - range magnetic order coexist at all temperatures below @xmath13 .
magnetisation measurements show that the material is highly anisotropic above @xmath13 , with an anisotropy energy not totally overcome by magnetic fields as high as 280 koe .
temperature dependence of the specific heat of @xmath0 ( solid symbols ) and of the non - magnetic isostructural compounds srlu@xmath9o@xmath7 and sry@xmath9o@xmath7 ( open symbols ) .
solid and dashed lines are calculations of the specific heat schottky anomaly expected for crystal field effects between a doublet ground state and a first excited doublet separated by energies indicated in the figure ( see text for details ) . ]
polycrystalline samples of @xmath0 were prepared from the high - purity starting materials srco@xmath14 and er@xmath9o@xmath14 following ref . .
x - ray diffraction measurements have confirmed the absence of any significant amount of impurities in the samples prepared .
high - resolution neutron powder diffraction experiments were conducted using a wavelength of 2.4 on the d2b diffractometer at the institut laue - langevin , france .
diffraction patterns were collected in zero - field at various temperatures between 0.45 and 15 k using a @xmath15he insert in a standard cryostat .
rietveld refinements were performed with the fullprof program @xcite .
possible symmetry arrangements of the low - temperature magnetic structure were determined using representation analysis .
the single crystals of seo were grown by the floating zone technique using an infra - red image furnace ; the details of crystal growth of the entire sr@xmath6o@xmath7 family of compounds will be reported separately @xcite .
the principal axes of the samples were determined using x - ray diffraction laue photographs ; for the magnetisation measurements the crystals were aligned to within an accuracy of better than 3@xmath16 .
low temperature specific heat measurements were performed using a quantum design ppms calorimeter equipped with a @xmath15he option . in order to estimate the lattice contribution to the specific heat we have measured the heat capacity of a single crystal of srlu@xmath9o@xmath7 and sry@xmath9o@xmath7 , nonmagnetic compounds isostructural to seo .
taking advantage of the high steady - state fields at the grenoble high magnetic field laboratory , we have taken magnetisation measurements for @xmath17 $ ] , @xmath18 $ ] and @xmath19 $ ] at various temperatures above 1.6 k in applied fields of up to 280 koe .
these measurements were performed using a standard extraction technique and a resistive magnet .
the same samples were used for more detailed investigations of the temperature dependence of magnetisation down to 1.4 k for different directions of an applied field using an oxford instruments vibrating sample magnetometer ( vsm ) in an applied field up to 120 koe .
temperature dependence of the specific heat divided by temperature of @xmath0 .
the inset shows the temperature dependence of entropy , calculated as an area under the @xmath20 curve , which has been extended linearly down to @xmath21 k. the solid line indicates the position of @xmath22 , which corresponds to a magnetic contribution for a system with an effective @xmath23 . ]
the temperature dependence of the specific heat of @xmath0 , shown in figure [ cvst ] , present two main features at low temperature .
first , a sharp lambda - type anomaly at about 0.75 k corresponding to a magnetic phase transition ( as will be seen from the neutron diffraction results ) and a much broader peak centred at around 1.3 k. the broad peak resembles a feature that could be caused by a low - lying crystal - field ( cf ) excitation . in order to test the latter possibility , we plot on the same graph the theoretical curves ( solid lines ) for a simple model , which contains a single excitation level separated from the ground state by 3 k , 3.5 k or 4 k. the ground multiplet @xmath24i@xmath25 of er@xmath26 is split into eight kramers doublets in low symmetry crystals , such as @xmath0 .
we therefore have to presume that the degeneracy of both the ground state and the excited level is two .
as long as this is the case , a cf level can not be solely responsible for the broad peak in specific heat
the peak is about 50% more intense than the doublets could generate regardless of their energy .
( colour online ) low scattering angle fragment of the powder neutron diffraction data taken on @xmath0 using the d2b instrument at the ill at 0.45 k , 0.8 k and 15 k. the 0.8 k and 15 k data are shifted by 40 and 80 counts respectively for clarity .
the inset on the left shows the temperature dependence of the integrated intensity for the two magnetic peaks labelled with arrows on the main panel .
the inset on the right is a difference plot between the combined low temperature data ( @xmath27 from 0.8 k to 1.5 k ) and the @xmath28 k data . ] above 5 k the heat capacity begins to increase with temperature .
it is tempting to associate the observed rise with the lattice contribution ( similar to what has been suggested for the ba@xmath6o@xmath7 compounds @xcite ) .
however , the heat capacities of srlu@xmath9o@xmath7 and sry@xmath9o@xmath7 , two nonmagnetic compounds isostructural with seo , suggest that this is not the case .
in fact , the lattice contribution to the specific heat remains negligibly small even at @xmath29 k ( see fig . [ cvst ] ) . unlike the lower temperature broad peak , the additional contribution to @xmath30 at @xmath31 k is probably caused by a cf level .
the dashed line in figure [ cvst ] was calculated for a cf level separated from the ground state by 38 k , which corresponds to the first peak observed at @xmath32 mev in our recent inelastic neutron scattering experiment @xcite .
a detailed analysis of the cf scheme in seo goes beyond the scope of this paper .
it is still helpful , however , to consider the results of the specific heat measurements in terms of their implications for the magnetic properties of seo . for this purpose
we plot in figure [ cbyt ] the temperature dependence of the specific heat divided by temperature .
the area under the @xmath20 curve represents the entropy , @xmath33 , which is shown in the inset of figure [ cbyt ] . at @xmath34
k ( above which temperature the input from the broad peak centred at 1.3 k is practically negligible ) the entropy reaches the value of @xmath22 , corresponding to a purely magnetic contribution for an effective spin 1/2 system .
the entropy recovered by the ordering temperature @xmath13=0.75 k amounts to less than a quarter of the @xmath22 value . from the neutron diffraction measurements in the paramagnetic phase ,
the space group of @xmath0 was confirmed as pnam and the lattice parameters were found to be consistent with those previously reported @xcite .
the refined values of @xmath35 , @xmath36 and @xmath10 at 15 k were found to be 10.018(2 ) , 11.852(2 ) and 3.384(2 ) respectively .
examples of the neutron diffraction patterns collected at various temperatures are shown in figure [ neutr1 ] .
broad diffuse scattering peaks associated with the short - range magnetic order developing in seo are observed at temperatures below 3 k ( see inset b in fig .
[ neutr1 ] ) .
this observation corroborates the suggestion that a broad peak at 1.3 k in the @xmath30 curve is magnetic in origin .
the sharp peaks corresponding to long - range order appear in all the diffraction patterns taken below 0.8 k. the intensity of these peaks has a pronounced temperature dependence ( see inset a in fig .
[ neutr1 ] ) , their positions , however , remained unchanged in the entire temperature range measured .
it was found that these peaks could be indexed with a propagation vector @xmath3 .
refined magnetic powder neutron diffraction data of @xmath0 taken as a difference between @xmath12 k and @xmath37 k. ] the best fit to the data was found when the symmetry used was restricted to that of the basis vectors @xmath387 for both er ion sites . at @xmath12 k the fit has a good agreement factor for the magnetic phase of @xmath39% and is shown in figure [ refinement ] . at the lowest experimentally available temperature of 0.45 k
the agreement becomes less satisfactory , as the diffuse intensity observed at around 21 degrees and 34 degrees in scattering angle can not be accounted for in a @xmath3 model . in seo ,
the presence of significant temperature dependant diffuse scattering with a pronounced structure factor allows a portion of the sample to support short range magnetic correlations , while the rest becomes long range ordered .
the proposed structure of the magnetic moments derived from the refinements and the observed magnetic extinction rules is shown in figure [ structure ] .
the two panels give the view parallel and perpendicular to the @xmath10 axis .
since the @xmath40 and @xmath41 sites in the magnetic space group pnam have mirror symmetry , the magnetic moments are either in the @xmath42 plane or along the @xmath10 axis .
it should be concluded that the moments in seo align along the @xmath10 axis , as no @xmath43 reflections were observed .
they align ferromagnetically with their nearest neighbour , and adjacent chains align antiferromagnetically .
the er@xmath44 site has a magnetic moment of 4.5 @xmath4 , whereas the second er@xmath9 site has a much reduced moment of less than 0.5 @xmath4 . within our model the magnetic moments on er@xmath44 and er@xmath9 sites
can be swapped without changing the calculated powder diffraction pattern .
( colour online ) magnetic structure of @xmath0 as determined from rietveld refinements of the neutron diffraction pattern at @xmath12 k. the same structure is shown twice to emphasise different arrangement of the magnetic moments along the @xmath10 axis and with respect to the hexagons in the plane .
two different er sites and their surroundings are shown in different colour .
only one of the sites carries a significant magnetic moment . ]
although the reasons for a large difference in the magnetic moment value on the two er sites are likely to be related to the low lying cf levels , they can not be solely responsible for the observed effects , as they can not split further the kramers doublet .
the involvement of exchange interactions and possibly dipolar forces must be presumed .
it is interesting to note that the proposed magnetic structure is favourable in terms of the dipolar interactions .
if only dipole - dipole interactions between the er ions are considered ( presuming equal moments on both sites ) then a collinear model with the moments along the @xmath10 axis has one of the lowest possible energies .
the results of the vsm magnetisation measurements in lower - field are shown in figure [ vsm ] .
the magnetisation is highly anisotropic at all measured temperatures and it does not reach the values predicted from the high - temperature susceptibility measurements of about 9 @xmath45 per er ion @xcite .
the magnetisation curves for different directions of the applied field are highly non - linear and have multiple crossing points ( see fig .
[ vsm ] ) . in this respect
the definition of the easy - axis is field dependent : while the @xmath36 axis remains a hard direction in any field , the @xmath10 axis seems to be an easy direction in a weak field ( up to 15 koe ) , but becomes less favourable in higher fields up to about 90 koe where the @xmath46}(h)$ ] curve again crosses the @xmath47}(h)$ ] curve .
overall the magnetic anisotropy in seo is best described as an easy - plane type , which is not uncommon for the er@xmath26 ions .
it has to be borne in mind that all of this unusual behaviour is observed at 1.5 k , a temperature at which seo remains magnetically disordered according to neutron powder diffraction data , which implies a strong cf influence .
( colour online ) field dependence of magnetisation @xmath48 ( top panels ) and magnetic susceptibility @xmath49 ( bottom panels ) measured for @xmath17 $ ] , @xmath18 $ ] and @xmath19 $ ] at 5 k ( left panels ) and 1.5 k ( right panels ) .
@xmath50 denotes the demagnetisation factor used for calculation of the internal field .
the data for a powder sample are also shown for comparison . ] at 1.5 k , the susceptibility curve dm / dh for a powder sample contains two broad maxima at 7 koe and 17 koe , while the single crystal curves for @xmath51 and @xmath52 only have a single , well pronounced maximum at 6 koe and 14 koe respectively .
it has been suggested @xcite that a double peak behaviour of the dm / dh curve for a powder sample is indicative of the presence of metamagnetism .
it is more likely , however , that a double peak feature in powder curves is simply a superposition of highly anisotropic curves for different directions of an applied magnetic field .
usually broad maxima in susceptibility are the precursors of sharp peaks , which correspond to phase transitions at lower temperatures
. it would be very useful to extend the susceptibility measurements down to the magnetic ordering temperature in order to check if these maxima develop into field - induced phase transitions . in order to check
if the full magnetic moment values expected for er@xmath26 ion can be achieved in higher fields we have extended the magnetisation measurements up to 280 koe .
the results shown in figure [ ghmfl ] clearly illustrate that even a magnetic field of 280 koe does not overcome the anisotropic magnetisation and fails to fully saturate the system .
the magnetisation , however , does not saturate in higher fields , it continues to increase with an applied field contrary to the conclusions made from the low - field measurements @xcite .
a gradient of approximately 0.005 @xmath53 and 0.008 @xmath53 is observed in a field above 200 koe for the @xmath54 curves with @xmath19 $ ] and @xmath18 $ ] respectively .
the fact that average magnetic moment per er ion in high fields is well above half of the value predicted from the high - temperature susceptibility measurements suggests that both of the two different er sites carry a significant magnetic moment .
( colour online ) field dependence of magnetisation @xmath48 measured at the grenoble high magnetic field laboratory for @xmath17 $ ] , @xmath18 $ ] and @xmath19 $ ] at different temperatures .
the data have been corrected for the demagnetisation effect . ]
to summarise , neutron diffraction and specific heat measurements extended to lower temperatures prove that @xmath0 is a magnetically frustrated system which orders at 0.75 k in zero field .
the moments are pointing along the @xmath10 axis , but the order is incomplete : only one of the two er sites carries a sizeable magnetic moment .
strong diffuse scattering is always present at sufficiently low temperature in zero field .
further evidence of partial ordering and low - dimensionality is gathered from the heat capacity measurements , which show that three quarters of magnetic entropy are released only above the @xmath13 .
the behaviour above the ordering temperature is characterised by a significant degree of magnetic anisotropy of the easy - plane type , with the @xmath36 axis being a hard - magnetisation axis .
an application of a high magnetic field ( 280 koe ) fails to completely polarise the magnetic moments .
these results call for an introduction of a theoretical model which considers a competition between single ion anisotropy , dipole - dipole interactions and exchange interactions . a combination of the three energies and of an external magnetic field is very likely to result in a complex @xmath55 phase diagram of @xmath0 at low temperatures .
the authors are indebted to m. r. lees and m. e. zhitomirsky for valuable discussions and a. richard for technical assistance .
we also acknowledge a collaboration with j. mercer and m. l. plumer on a closely related project .
this work is supported by the epsrc research grant on single crystal growth at warwick .
the experimental time at the ill was obtained through the epsrc super - d2b grant .
the work at grenoble high magnetic field laboratory is supported by the
transnational access to infrastructures - specific support action " program - contract .
rita - ct-2003 - 505474 of the european commission .
99 h. kawamura , j. phys .
matter * 10 * , 4707 ( 1998 ) ; m. f. collins and o. a. petrenko , can . j. phys . * 75 * , 605 ( 1997 ) .
a. p. ramirez , g. p. espinosa and a. s. cooper , phys .
lett . * 64 * , 2070 ( 1990 ) ; c. broholm , g. aeppli , g. p. espinosa , and a. s. cooper , phys . rev . lett . * 65 * , 3173 ( 1990 ) .
p. schiffer , a. p. ramirez , d. a. huse , p. l. gammel , u. yaron , d. j. bishop and a. j. valentino , phys .
lett . * 74 * , 2379 ( 1995 ) ; p. schiffer , a. p. ramirez , d. a. huse and a. j. valentino , phys .
lett . * 73 * , 2500 ( 1994 ) ; o. a. petrenko , c. ritter , m. yethiraj and d. @xmath56 paul , phys .
* 80 * , 4570 ( 1998 ) . for reviews on frustrated pyrochlore magnets
see s. t. bramwell and m. j. p. gingras , science * 294 * , 1495 ( 2001 ) ; j. e. greedan , j. mater .
chem . * 11 * , 37 ( 2001 ) and j. alloys and compounds , * 408 - 412 * , 444 ( 2006 ) .
a. p. ramirez , handbook on magnetic materials , edited by k. j. h. busch ( elsevier science , amsterdam , 2001 ) , vol .
13 , p. 423
. k. binder , phys .
* 45 * , 811 ( 1980 ) ; c. l. henley , j. applied phys . *
61 * , 3962 ( 1987 ) . to avoid confusion
, it should be stated that in some cases the term honeycomb structure " is used to describe the magnetic structure of ising - like spins on a hexagonal lattice , where one third of the magnetic moments are aligned antiparallel to another third , while the residual third of the moments remain disordered @xcite . in this article
we use the term honeycomb structure " in a strictly crystallographic sense .
a. mattsson , p. frojdh and t. einarsson , phys .
b * 49 * , 3997 ( 1994 ) .
h. karunadasa , q. huang , b. g. ueland , j. w. lynn , p. schiffer , k. a. regan and r.j .
cava , phys .
b * 71 * , 144414 ( 2005 ) . b. f. decker and j. s. kasper , acta cryst . *
10 * , 332 ( 1957 ) .
y. doi , w .
nakamori and y. hinatsu , j. phys . condens .
matter * 18 * , 333 ( 2006 ) .
i. felner , y. yeshurun , g. hilscher , t. holubar , g. schaudy , u. yaron , o. cohen , y. wolfus , e. r. yacoby , l. klein , f. h. potter , c. s. rastomjee and r. g. egdell , phys .
b * 46 * , 9132 ( 1992 )
. j. rodriguez - carvajal , physica b , * 192 * , 55 ( 1993 ) . g. balakrishnan , t. j. hayes , o. a. petrenko and d. @xmath56 paul , in preparation ( 2008 ) . t. j. hayes , d.t . adroja and o. a.
petrenko , in preparation ( 2008 ) . l. m. lopato and a. e. kushchevskii , ukrainskii khimicheskii zhurnal * 39 * , 7 ( 1973 ) .
m. mekata , j. phys .
japan * 42 * , 76 ( 1977 ) ; h. shiba , prog .
phys . * 64 * , 466 ( 1980 ) ; f. matsubara and s. inawashiro , j. phys .
japan * 56 * , 2666 ( 1987 ) . | @xmath0 has been characterised by low - temperature powder neutron diffraction , as well as single crystal specific heat and magnetisation measurements .
magnetisation measurements show that the magnetic system is highly anisotropic at temperature above ordering
. a magnetic field of 280 koe applied at @xmath1 k does not overcome the anisotropic magnetisation and fails to fully saturate the system .
long range antiferromagnetic ordering develops below @xmath2 k , identified by magnetic bragg reflections with propagation vector @xmath3 and a lambda anomaly in the specific heat .
the magnetic structure consists of ferromagnetic chains running along the _ c _ axis , two adjacent chains being stacked antiferromagnetically .
the moments point along the _ c _ direction , but only one of the two crystallographically in - equivalent er sites has a sizeable ordered magnetic moment , 4.5 @xmath4 at 0.55 k. the magnetic properties of @xmath0 are discussed in terms of the interplay between the low - dimensionality , competing exchange interactions , dipolar interactions and low lying crystal field levels . |
in a seminal paper , haldane showed that integer spin chains display a gap in the spectrum , where as half - integer spin chains are gapless @xcite .
this was confirmed both by experiments @xcite and numerical calculations @xcite .
currently , the best estimates of the gap for the isotropic @xmath0 chain are @xmath7 , using the density matrix renormalization group ( dmrg ) method @xcite , and @xmath8 using exact diagonalization @xcite .
the dynamical structure factor has also been determined in great detail , by numerical studies of finite systems @xcite .
most of the experimental effort has in recent years focused on ni(c@xmath5h@xmath9n@xmath5)@xmath5no@xmath5(clo@xmath10 ) ( nenp ) which show clear evidence of the haldane gap @xcite .
nenp has not been observed to undergo a three - dimensional ordering at any accessible temperature as opposed to csnicl@xmath11 which yielded the first experimental confirmation of the haldane gap @xcite .
nenp has a fairly large single ion anisotropy estimated to be @xmath12 @xcite where as csnicl@xmath11 is almost isotropic .
recently , experiments have been performed on y@xmath5banio@xmath6 @xcite .
this charge - transfer insulator has long one - dimensional chains of ni atoms at the center of compressed corner sharing oxygen octahedra .
thus , along the chains , the ni@xmath13 ions , with @xmath0 , are separated by nonmagnetic o@xmath14 ions .
the antiferromagnetic ( af ) superexchange interaction between the ni@xmath13 ions is then mediated through the oxygen sites .
y@xmath5banio@xmath6 has a fairly large exchange coupling of @xmath15=285k @xcite-322k @xcite , estimated from high temperature susceptibility measurements . from inelastic neutron scattering experiments on powders
@xcite it is known that the inter chain coupling is small , @xmath16 .
thus , the compound is highly one - dimensional and no three - dimensional ordering has been seen . in powder
averaged inelastic neutron scattering experiments @xcite two gaps have been observed ; for fluctuations parallel to the chain , @xmath17 , and for fluctuations perpendicular to the chain , @xmath18 .
this is in good agreement with single crystal measurements of @xmath19 @xcite .
the two gaps are explained in terms of a single ion anisotropy @xmath20 @xcite , a value close to what is found for nenp .
the one - dimensional haldane gap systems are by now fairly well understood .
however , the important effect of impurities on these systems have only recently been considered .
y@xmath5banio@xmath6 offers several possibilities for doping and is thus well suited for such studies .
non - magnetic zn@xmath13 ions can be substituted for ni@xmath13 , thus effectively severing the ni chains .
experiments have been performed on zn doped samples , y@xmath5bani@xmath21zn@xmath22o@xmath6 @xcite , confirming the picture of chain cutting , with a possibility for an _ increase _ in the gap with doping @xcite due to the fact that the average length of the chain segments is small enough for finite - size effects to be observable .
severing of the chains is predicted to have rather dramatic effects on the low energy excitations .
it can be shown that the af heisenberg model is fairly well approximated by a valence bond solid ( vbs ) model @xcite .
the breaking of the valence bonds by the introduction of impurities should produce free @xmath3 spins at the sites next to the impurity .
the effect of such free @xmath3 at the chain - ends has been observed in electron - spin - resonance ( esr ) experiments on nenp by doping with cu @xcite , thereby introducing a @xmath3 at the impurity site , and by doping with non - magnetic zn , cd and hg @xcite .
these findings have recently been contrasted by specific heat measurements @xcite on y@xmath5bani@xmath21zn@xmath22o@xmath6 which show a schottky anomaly consistent with free @xmath0 . another possibility for doping exists in y@xmath5banio@xmath6 .
carriers , in the form of holes , can be added to the system by replacing the _ off - chain _
y@xmath23 with ca@xmath13 . ditusa et .
al @xcite show , from polarized x - ray absorption spectra , that the added holes are localized on the oxygen _ between _ the ni atoms .
the addition of a hole on the oxygen sites could effectively turn these into @xmath3 spins , if strongly localized .
ca doping should then have significant effects on the spectrum .
powder averaged inelastic neutron scattering have been performed on y@xmath24ca@xmath22banio@xmath6 @xcite showing that ca doping produces states in the haldane gap . from analysis of the spectral weight per impurity , ditusa et al .
@xcite have proposed that the magnetic disturbance caused by the ca doping introduces either a triplet or a quartet of subgap bound states . in order to address the experimental results on doped @xmath0 spin chains
we therefore consider the simple model hamiltonian @xmath25 here the single ion anisotropy is neglected for simplicity , and * s * represents @xmath0 spins .
the impurity hamiltonian @xmath26 is taken to be either a perturbed bond @xmath27 or a @xmath3 impurity , in which case we have @xmath28 where @xmath29 is the @xmath3 impurity spin .
schematically the two kinds of impurities are shown in fig .
[ fig : chains]a , [ fig : chains]b , respectively . for the carrier doped compounds
( [ eq : h ] ) may not be a good description and a hubbard like model may be more appropriate .
however , in themselves the hamiltonians , eq . ( [ eq : h ] ) , deserves attention and they should form a good starting point also for the carrier doped compounds if the carriers are strongly localized . our approach is to study the low - lying excitations in these systems by comparing large - scale numerical results , obtained using the density matrix renormalization group ( dmrg ) method , to perturbative results , valid at strong or large coupling , based on effective @xmath3 chain - end excitations generated by broken valence bonds . in section [ sec : methods ] we discuss the numerical approach .
section [ sec : bonds ] addresses the effects of a single perturbed bond at the center of a chain , while we in section [ shalf ] present our results for a @xmath3 impurity .
the bulk of our results concerns antiferromagnetically coupled impurities , however , in sec .
[ sec : ferro.coupling ] we briefly discuss scenarios for ferromagnetically coupled impurities .
we use the density matrix renormalization group ( dmrg ) method proposed by white @xcite . for a detailed discussion of the dmrg method
we refer the reader to refs . .
in all our runs we truncate each half - space to @xmath30 states , and we work in a subspace defined by the total z - component of the magnetization , @xmath31 , and by the parity with respect to reflection around the central bond , or central spin , for even and odd length chains , respectively . for the dmrg method
we therefore denote the different states by @xmath32 .
note that , this is the z - component of the spin and _ not _ the total spin of the state .
the idea behind the dmrg method revolves around the notion of a subsystem , in this case half of the chain , that is strongly coupled to the universe , here the complete chain . as usual in quantum mechanics ,
the subsystem is then best described by its density matrix , whose eigenvalues give the probability for the subsystem to be in a given eigenstate ( of the subsystem density matrix ) , under the constraint that the universe ( the complete chain ) is in a given pure state .
here we mean , by pure state , an eigenstate of the hamiltonian for the complete system .
it is now clear that if we start to effectively decouple the subsystem from the universe , by changing the relevant coupling at the center of the chain , the standard dmrg method , where the system is split in half at the impurity site , will become progressively inaccurate .
this can be seen in the limit where the coupling to the universe , i. e. the central bond in the chain , is zero , in which case the eigenstates of the density matrix and the hamiltonian for the subsystem are the same and the dmrg method reduces to standard real space renormalization which is not comparable to the strongly coupled dmrg in precision . we therefore expect to have problems whenever the perturbations that we consider at the center of the chain behave in a way that effectively breaks the system in two . however , as we shall see , this still leaves a workable region where we can obtain useful results .
we assume that these difficulties could be overcome by considering more complicated dmrg schemes where the system is not split at the impurity site .
however , most other schemes will reduce the symmetry of the problem and we therefore prefer to work with the standard method .
in the case where the ni based @xmath0 heisenberg antiferromagnets are doped by impurities that form divalent non - magnetic ions , such as zn , cd , and hg , we expect the impurities to effectively cut the ni chains . in the valence bond picture @xcite
this leads to free @xmath3 at the neighboring sites . in this section
we consider the more general case where the central bond in an open chain is different from the rest of the chain , although still antiferromagnetic .
the hamiltonian describing the system is eq .
( [ eq : h ] ) and ( [ eq : himpa ] ) . for the experiments performed with non - magnetic impurities
the relevant situation is that of an extremely weak bond .
this can be seen in the experiments of glarum et al .
@xcite who interpret their results in terms of _ free _ @xmath3 chain - end spins .
if the bond at the center of the chain is weak enough we can view the problem as two weakly coupled open chains of half the length . at the end of an open @xmath0 chain
we expect to find an effective @xmath3 excitation , this we shall refer to as a chain - end spin .
thus , for sufficiently long chains , the low - lying excitations can be described by an effective hamiltonian of two weakly coupled chain - end spins .
this should be valid once the coupling of the chain - end spins across the impurity bond is larger than the coupling _ along _ the chain segment .
the latter is known to decrease exponential with the length of the chain segment @xcite , and should thus quickly become negligible .
experimentally this is the relevant situation for nenp , since it is difficult to obtain high zn doping levels @xcite , and the average chain segment is therefore likely to be long .
y@xmath5banio@xmath6 offer the possibility of higher doping levels .
if we denote by @xmath33 the projection onto the @xmath3 subspace , we can write @xmath34 . we need to determine @xmath35 .
fortunately , an estimate can easily be obtained by looking at the expectation value one of the spins at the _ end _ of a _
free _ chain .
if we by @xmath36 denote the state that has both the chain - end spins polarized in the up "- state we find @xmath37 previous work @xcite have established that @xmath38 at the end of a free chain , and thus @xmath39 .
the effective hamiltonian can then be written as @xmath40 where @xmath41 denotes a @xmath3 spin .
we then immediately see that the low energy spectrum will be given by a ground - state singlet and a higher lying triplet , the splitting between the two given by @xmath42 this corresponds to a bound state in the haldane gap , @xmath43 . for the isotropic antiferromagnetic heisenberg model that we are studying here we have @xmath44 in the absence of any impurities . in fig .
[ fig : gapj.1 ] we show the energy gap between the two lowest lying states for a chain that has a weak bond of @xmath45 in the middle of an otherwise isotropic chain .
since we have two essentially free @xmath3 chain - end spins at each end of the open chain , the ground - state becomes four - fold degenerate . for finite chain lengths we expect these two chain - end spins to combine into a ground - state singlet , @xmath46 , with an exponentially low - lying triplet , @xmath47 @xcite .
in the thermodynamic limit the triplet and the singlet become degenerate .
thus , if we want to calculate the spectrum caused by the impurity we need to look at higher excited states . the resulting gap , which remain finite in the thermodynamic limit , can be calculated between the @xmath48 and @xmath47 states , where we for computational convenience take the @xmath47 state to represent the ground - state since they become degenerate in the thermodynamic limit .
as is clearly evident in fig .
[ fig : gapj.1 ] , the gap converges quickly to a value of @xmath49 . with @xmath50
we find that eq .
( [ eq : deweak ] ) gives @xmath51 in excellent agreement .
the good agreement between the perturbative results and the dmrg results lends further support to the picture of effective @xmath3 chain - end spins .
the dmrg method is convenient for studying correlations in real - space . in fig .
[ fig : szj.1 ] we show the expectation value of the z - component of the spin along a 100 site chain for @xmath50 . in the @xmath48 state
we have effectively polarized all four @xmath3 chain - end spins at sites 1 , 50 , 51 , and 100 in the up " direction . in order to focus attention on the effects of the impurity
we have subtracted off @xmath52 . in the @xmath47 state
the chain - end spins at site 50 , 51 are combined into a singlet , by subtracting off @xmath52 we therefore cancel out the effect of the chain - end spins at site 1 and 100 .
one sees that the two chain - end spins surrounding the perturbed bond have effectively formed a localized state . using the dmrg method we can now map out the energy of this bound state as a function of the weak coupling , @xmath2 . due to the problems mentioned in section [ sec : methods ] it is not possible to treat values of @xmath2 below @xmath53 within the standard dmrg
however , since the impurity state is localized , the convergence to the thermodynamic limit is very fast as can be seen in fig .
[ fig : gapj.1 ] .
our results are shown in fig .
[ fig : weakgap ] .
the filled squares are the numerical results , the solid line the haldane gap for the isotropic chain , @xmath54 , and the dashed line indicates the result from eq .
( [ eq : deweak ] ) . from fig .
[ fig : weakgap ] it is seen that only for @xmath55 does a bound state below the isotropic haldane gap exist .
around @xmath56 the bound state seen in fig .
[ fig : szj.1 ] delocalizes and spreads out over the entire chain .
this is shown in fig .
[ fig : szj.7 ] where we plot @xmath57 as a function of chain index @xmath58 for a 100 site chain with @xmath56 .
clearly this state is not localized and it resembles what one finds for the isotropic chain @xcite . we therefore conclude that it is necessary to weaken the central bond below a critical value @xmath59 in order to create a bound state .
we now turn to a discussion of the opposite limit where the central bond is very strong , @xmath60 . for the antiferromagnetic coupling that we consider here
, the two neighboring spins will be contracted into a singlet . in the limit @xmath61
the chain will be severed as was the case for a weak bond .
we now want to derive an effective hamiltonian valid for @xmath60 . as a starting point
it is convenient to consider the case where we only have four spins described by the hamiltonian : @xmath62 we do perturbation theory in @xmath63 .
note that when @xmath64 there is a degneracy since the spins , @xmath65 and @xmath66 can be in arbitrary states .
this is lifted in second order perturbation theory in @xmath67 .
the result of second order perturbation theory can be expressed in terms of an effective hamiltonian acting on the low - energy subspace , in which @xmath68 and @xmath69 are projected onto the singlet state : @xmath70 here @xmath71 and @xmath72 projects onto the singlet ground - state of @xmath73 , with energy @xmath74 .
acting with @xmath75 on the ground - state of @xmath73 produces the eigenstate of @xmath73 with spin 1 and energy , @xmath76 .
thus : @xmath77 where @xmath78 denotes the spin-0 ground - state of @xmath73 .
using : @xmath79 we obtain : @xmath80 the same calculation applies to a chain of arbitrary length .
@xmath81 now contains all the additional couplings which do nt involve the pair of strongly coupled spins , @xmath82 and @xmath83 ie .
: @xmath84 this is exactly the same as the original hamiltonian of eq .
( [ eq : h ] ) with two sites removed from the ends of the chain and @xmath2 replaced by @xmath85 . assuming this impurity coupling is much smaller than @xmath67 we write an effective low - energy hamiltonian by proceeding along the lines of section [ sec : weak.link ] . by introducing spin-1/2 variables , @xmath86 and @xmath87
, we obtain : @xmath88 from this effective hamiltonian we can then derive a simple expression for the bound state : @xmath89 we have calculated the gap @xmath90 as a function of the coupling , @xmath63 , to the @xmath3 impurity spin .
our results , extrapolated to the thermodynamic limit , are shown in fig .
[ fig : destrong ] .
the solid line indicates the haldane gap , @xmath54 , for the pure chain .
the solid squares are the numerical results obtained using the dmrg method .
the dashed line denotes the perturbative result eq .
( [ eq : destrong ] ) .
the open circles indicate a direct diagonalization of eq .
( [ eq : hstrong ] ) with @xmath65 and @xmath66 spin one - half . as explained in section [ sec : methods ] we are limited to study @xmath91 . for @xmath92
the limiting value of the gap is very close to the value of the isotropic chain , @xmath43 , and for @xmath93 there appears to be no bound state in the haldane gap . for smaller @xmath63 a bound
state emerges , and we see good agreement between the numerical results and the effective hamiltonian eq .
( [ eq : heff.strong ] ) .
as was the case for a weak bond we find a critical coupling , @xmath94 , but now with an approximate value of @xmath95 above which no bound state occurs .
the picture of a quenched singlet in the middle of the chain weakly coupled to two chain - end spins on either side , can be substantiated by calculating the expectation value of the z - component of the spin along the chain .
this is shown in fig .
[ fig : szj4.0 ] for @xmath96 , @xmath97 .
again we have subtracted off the @xmath98 state configuration .
the quenched singlet is clearly visible at the center of the chain and we see a large signal from the two polarized chain - end spins at @xmath99 and @xmath100 . since one site has effectively been removed from each of the two half - chains fig .
[ fig : szj4.0 ] appears inverted with respect to fig .
[ fig : szj.1 ] . at the point where the bound state disappear
we again find that the expectation value @xmath57 spreads out over the entire chain .
we may think of a magnon as behaving , in many respects , like an ordinary particle of mass @xmath101 , where @xmath102 is the spin - wave velocity .
thus it may , at first , seem surprising that no bound state exists for a weak perturbation of the coupling , @xmath103 since there is a well - known theorem of one - dimensional quantum mechanics which states that an attractive potential always produces a bound state , no matter how weak it is @xcite .
our result is not surprising if one looks in a bit more detail at the mapping of the spin problem into an effective one - particle quantum mechanics problem .
we represent the spin operators at long length - scales in terms of the non - linear @xmath104 ( nl@xmath104 ) model field , @xmath105 ( obeying the constraint @xmath106 ) and the associated canonical momentum , @xmath107 , as : @xmath108 the hamiltonian density becomes : @xmath109 acting on the vacuum , @xmath105 creates the triplet of single - magnon states .
a local perturbation in the exchange couplings leads to additional local terms in the field theory hamiltonian density of the form : @xmath110 where @xmath111 and @xmath112 are some short - range functions . note that we _ do not _ get simply @xmath113 , since this term is trivial due to the constraint , @xmath114 .
we sometimes approximate the nl@xmath104 model by a free massive field theory model with hamiltonian density @xcite : @xmath115 where the constraint on @xmath105 is now relaxed and @xmath116 is the associated cononical momentum .
this gives the dispersion relation : @xmath117 ( we have shifted the wave - vector by @xmath118 . )
we may easily treat the additional impurity term , @xmath119 .
this gives the schrdinger equation : @xmath120 ' = me\psi .\ ] ] we obtain this equation for each of the three magnon polarizations .
note that this _ does not _ correspond to a standard potential scattering term .
we could obtain a potential scattering term from a term in the nl@xmath104 model hamiltonian of the form @xmath121 or @xmath122 .
however , these terms break the @xmath123 symmetry and therefore ca nt appear in the heisenberg model .
the latter term corresponds to a local magnetic field .
the former could arise if we have a crystal field term in addition to the heisenberg exchange .
the peculiar derivative potential term does not necessarily lead to a bound state if it is weak . to demonstrate this we consider the case of a step function : @xmath124 solutions to the schrdinger equation will have discontinuous @xmath125 at @xmath126 . defining @xmath127 to be the limit of @xmath128 as @xmath129 from above or below respectively and similarly for @xmath130
we obtain : @xmath131\nonumber \\ \psi ' ( -a^- ) & = & \psi ' ( -a^+)[1 - 2v_0].\end{aligned}\ ] ] the important thing to notice is that if @xmath132 , the derivative does not change sign at the discontinuity .
there can then be no solution which vanishes at @xmath133 .
this follows since , assuming an even wave - function , it must have the form @xmath134 for @xmath135 which is increasing as we move away from @xmath136 .
it has to turn into @xmath137 $ ] for @xmath138 .
this ca nt happen without a change in sign of the derivative .
[ @xmath139 itself must be continuous .
] alternatively , we could assume an odd wave - function but we have the same problem .
if @xmath140 , there is a solution .
it is an odd function of @xmath141 ; a sine function for @xmath135 .
the binding energy is determined by : @xmath142 this has a solution for all @xmath140 . as @xmath143 , we get : @xmath144 this provides a counter - example to the statement that there is always a bound state for a weak derivative potential term . on the other hand , there is a rigorous proof of the fact ( see , for example ref . ) for an ordinary potential term .
this argument is admittedly quite incomplete since we havent analyzed the other , @xmath145 term , nor dealt carefully with the passage from the nl@xmath104 model to the free model .
( this additional term seems to give a type of non - linear schrdinger equation . ) ignoring these complications , and assuming that the step function result is universal , the above argument suggests that as @xmath146 the critical coupling where the bound state disappears , the binding energy behaves as : @xmath147 this is not inconsistent with our numerical results , but substantially more numerical work for @xmath2 close to @xmath148 would be needed in order to verify this dependence .
our findings in this section are in disagreement with a recent schwinger - boson calculation @xcite .
we now turn to a discussion of @xmath3 impurities in a @xmath0 af heisenberg chain . experimentally two scenarios have been studied .
first , cu@xmath13 ( @xmath3 ) was substituted for ni@xmath13 ( @xmath0 ) in nenp @xcite . here
, the introduced @xmath3 impurity structurally takes the same place as the original ni ion .
experimentally one finds a weak _ ferromagnetic _ coupling to the neighboring chain - end spins @xcite .
secondly , carriers were added to y@xmath5banio@xmath6 by substituting ca for y. ditusa et al .
@xcite argue that the holes become localized on the o sites between the ni ions .
here the situation is different since the introduced @xmath3 impurity does _ not _ take the place of a @xmath0 , instead , it is placed _ between _ two ni ions .
thus , the superexchange between the two ni ions is broken and presumably replaced by a _ direct _ exchange to the @xmath3 , @xmath2 , which is likely to be large compared to @xmath67 .
aharony et al .
@xcite have argued that for la@xmath24(sr , ba)@xmath22cuo@xmath149 the analogous coupling should be large , @xmath150 .
however , the appropriate sign and value of @xmath2 for a simple linear chain heisenberg model capable of describing the experiments on y@xmath24ca@xmath22banio@xmath6 @xcite is an open question . in the following
we shall limit our discussion to the case of an antiferromagnetic coupling , @xmath151 . following the approach in section [ sec : weak.link ] , for a weak bond at the center of the chain , we can try to find an effective hamiltonian for a weakly coupled @xmath3 impurity at the center of the chain . in this case
we again view the sites @xmath152 and @xmath153 of the @xmath0 chain as being occupied by chain - end spins that effectively behave as if they were spin one - half . by projecting onto the @xmath3 subspace
we represent the chain - end spins at these two sites by @xmath154 , where @xmath41 is now a real spin one - half .
we then obtain an effective low energy hamiltonian . with @xmath155 representing the @xmath3 impurity spin
we have : @xmath156 .
\label{eq : heff3}\ ] ] the spectrum for this effective hamiltonian consists of a ground - state doublet and an excited doublet and quadruplet .
the ground - state doublet has @xmath157 , and energy @xmath158 .
the excited doublet , @xmath159 , has energy @xmath160 . finally , the quadruplet , @xmath161 , has energy @xmath162 . from this effective hamiltonian
we see that we should have _ two _ low - lying levels with energies in the haldane gap with energies given by @xmath163 and @xmath164 for a weak _ ferromagnetic _ coupling we would have obtained @xmath165 . in order to observe these two levels with the dmrg method we need to determine the relevant quantum numbers for the complete system .
we begin by considering the case where the two half - chains surrounding the @xmath3 impurity are of _ even _ length .
we consider the situation where the chain - end spins at sites @xmath152 and @xmath153 are coupled weakly with a coupling @xmath166 , @xmath167 , to the chain - end spins at the sites @xmath168 and @xmath169 , respectively . since the chain segments have an _ even length _ this coupling , @xmath166 , is _
antiferromagnetic _
it is now straight forward to obtain the level diagram shown in fig .
[ fig : levels]a . if the two half - chains have an _ odd length _ the coupling , @xmath166 , is _ ferromagnetic _ @xcite and we obtain the level diagram shown in fig .
[ fig : levels]b .
note that , since we in either case have an even number of valence bonds we do not get a change of parity from the rest of the chain .
thus , we see that in the thermodynamic limit the ground - state is eight - fold degenerate with an eight - fold and sixteen - fold degenerate level above it . in our implementation of the dmrg method we can not use the total spin as a good quantum number but only the total z - component .
in order to determine the two levels given by eqs .
( [ eq : de.shalf.weak.doublet ] ) and ( [ eq : de.shalf.weak.quadruplet ] ) we then see from fig .
[ fig : levels ] that we need to calculate the energy differences @xmath170 and @xmath171 . only for @xmath152
odd is @xmath172 the true ground - state .
the haldane gap to bulk excitations ( i. e. excitations not localized at the center and/or ends of the chain ) will be between the @xmath173 and @xmath172 states . in fig .
[ fig : gap.shalf.j0.2 ] we show the energy gap , @xmath174 , as a function of chain length @xmath169 , for a @xmath0 chain with a @xmath3 impurity .
the coupling to the @xmath3 impurity is @xmath175 .
the symbols denote the gaps : @xmath176 ( triangles ) to the state @xmath173 ( haldane gap ) , @xmath177 ( squares ) to the state @xmath178 , and @xmath179 ( circles ) to the state @xmath180 . clearly , we have two low - lying excitations well separated from the haldane gap .
for the @xmath180 state we find the gap to be @xmath181 , compared to @xmath182 from eq .
( [ eq : de.shalf.weak.doublet ] ) .
although at large @xmath169 we have minor problems with the numerical results for the @xmath178 state we can also in this case estimate the gap to @xmath183 which compares favorably to the result @xmath184 from eq .
( [ eq : de.shalf.weak.quadruplet ] ) .
it is also possible to study the spatial extension of the bound states .
however , as opposed to section [ sec : bonds ] , @xmath185 for the state @xmath172 is non - zero across the impurity site .
it is therefore difficult to give a quantitative interpretation of the results since the subtraction of @xmath186 will produce more complicated effects than just the cancellation of the contribution from the free chain - end spins at the sites 1 and l. however , one can still differ between extended and localized states .
we found that the two states @xmath180 and @xmath178 clearly were localized around the impurity site for @xmath175 whereas the state @xmath173 was extended .
we can now calculate the low - lying excitations as a function of coupling to the @xmath3 impurity .
our results are shown in fig .
[ fig : shalf.weak ] .
the symbols denote the gaps : @xmath187 ( triangles ) to the state @xmath173 ( haldane gap ) , @xmath171 ( squares ) to the state @xmath178 , and @xmath170 ( filled squares ) to the state @xmath180 .
the solid line is the haldane gap , @xmath188 .
the dashed line is the perturbative result eq .
( [ eq : de.shalf.weak.doublet ] ) , and the dot - dashed line eq .
( [ eq : de.shalf.weak.quadruplet ] ) . for small @xmath2
we see excellent agreement with the perturbative results . for @xmath189
a bound state , calculated from the @xmath180 level , appears , and for @xmath190 also bound state , calculated from the @xmath178 level , is stable . as before
, we try to arrive at an effective hamiltonian capable of describing the low energy excitations .
again we shall assume that the coupling to the @xmath3 impurity spin , @xmath2 , is always positive corresponding to an antiferromagnetic coupling .
as a starting point we consider the three central spins @xmath191 and @xmath155 , the @xmath3 impurity .
disregarding the rest of the chain , these three spins will be described by a hamiltonian @xmath192 \label{eq : h0.shalf}\ ] ] the spectrum of @xmath26 is as follows : the ground - state is a quadruplet @xmath161 with @xmath193 , then a doublet @xmath159 with @xmath194 , another doublet @xmath157 with @xmath160 , a quadruplet @xmath195 with @xmath196 and finally a sextuplet @xmath197 with energy @xmath198 .
since the ground - state of @xmath26 is a multiplet with @xmath199 , the effective hamiltonian becomes that of an effective @xmath199 impurity in the otherwise uniform af @xmath0 chain . if we by @xmath200 denote the projection onto the @xmath199 ground - state subspace of @xmath26 , we can write @xmath201
. the effective hamiltonian can then be written : @xmath202 where the term describing the @xmath199 impurity is given by : @xmath203\nonumber\\ & = & \beta j[{\bf s}_{l/2 - 1}\cdot { \bf s}^{\prime\prime}_{s=3/2}+ { \bf s}^{\prime\prime}_{s=3/2}\cdot{\bf s}_{l/2 + 2 } ] .
\label{eq : heff3.2}\end{aligned}\ ] ] here @xmath204 denotes the effective @xmath199 spin .
we need to determine @xmath205 .
we begin by writing down the matrices for @xmath206 and @xmath207 projected onto the @xmath172 ground - state subspace of @xmath26 .
these have to be proportional to a spin-3/2 representation .
we find : @xmath208 clearly , this is just @xmath209 , as expected .
thus , if we project @xmath82 and @xmath210 onto the @xmath172 subspace we get in both cases a result proportional to a spin-3/2 by a factor 3/5 , i.e. @xmath211 . thus , @xmath212 the coupling to the @xmath199 impurity is therefore fixed at @xmath213 .
let us consider the more general case where we instead of eq .
( [ eq : heff3.2 ] ) have @xmath214 , \label{eq : heff3.2.gen}\ ] ] i. e. we regard the coupling to the @xmath199 impurity as a free parameter , @xmath215 .
we would expect bound states to appear for @xmath216 .
for very small @xmath215 , @xmath217 , we could again assume that effective @xmath3 chain - end spins on the sites @xmath218 and @xmath219 would be a valid description , and we would arrive at an effective low - energy hamiltonian consisting only of the two chain - end spins , and the @xmath199 impurity : @xmath220,\ \
k\ll k_c^{(3/2)}. \label{eq : kurt}\ ] ] here , @xmath221 , and @xmath222 are @xmath3 spins .
the low - energy hamiltonian , eq . ( [ eq : kurt ] ) , has the following spectrum .
a ground - state doublet with energy @xmath223 , an excited quadruplet at @xmath224 , another quadruplet at @xmath160 , and finally a sextuplet at @xmath225 . for sufficiently small @xmath215
all three bound states should be stable giving rise to energy levels in the haldane gap given by the corresponding energy differences , @xmath226 , @xmath227 , and @xmath228 .
returning to the full effective hamiltonian , eq .
( [ eq : heff4 ] ) , the question of whether bound states appear comes down to the question of whether @xmath213 is below or above the threshold , @xmath229 . as we shall see below
, the numerical results indicate that @xmath213 is too large for any bound states to be stable .
it is then _ not _ possible to write down an effective low - energy hamiltonian based on the @xmath3 chain - end spins , as we did in eq .
( [ eq : kurt ] ) , and it is necessary to work with the full effective hamiltonian , eq .
( [ eq : heff4 ] ) , which , due to the size of the corresponding hilbert space , is not very useful .
we have calculated the gap @xmath171 for @xmath230 , 0.2 , and 0.1 using the dmrg method . for large @xmath2
this is the lowest - lying excitation .
the results are shown in fig .
[ fig : shalf.strong ] .
in this case the convergence to the thermodynamic limit is rather slow and we have difficulty extrapolating the results to a high accuracy , we therefore show the unextrapolated results for a 120 site chain . here
the solid squares are the dmrg results , and the solid line denotes the isotropic haldane gap of @xmath231 . in this case
the dmrg method can be used for even very large @xmath2 . as is clearly evident from fig .
[ fig : shalf.strong ] we never observe any states in the haldane gap .
a calculation of @xmath170 also did nt show any bound states in the haldane gap .
we therefore conclude that the coupling to the effective @xmath199 impurity , @xmath213 , is too large for any bound states to appear .
we have also calculated the expectation value of the z - component of the spin for the values of @xmath230 , @xmath232 , and @xmath233 .
our results are shown in fig .
[ fig : sz.j1510 ] .
the circles are the results for @xmath230 , the squares for @xmath234 and the triangles for @xmath235 . in order to differ between extended and localized states
we have subtracted @xmath236 . as opposed to section [ sec : bonds ] the spin configuration that we subtract , here @xmath236 ,
is non - zero across the impurity making it difficult to give a quantitative interpretation of these results .
however , clearly the state is in all cases extended over the whole length of the chain .
thus , we conclude that for a strong antiferromagnetic coupling to the @xmath3 impurity no bound states exists below the haldane gap .
the fact that ditusa et al .
@xcite observes such states , indicates that , the appropriate ( af ) @xmath2 , for a simple linear chain heisenberg model description of their results , should be rather small .
we have only performed the numerical dmrg calculations for antiferromagnetically coupled impurities .
it would be interesting to repeat the calculation using ferromagnetic couplings . due to the rather large amount of cpu time required
we have not done this . since we in our implementation of the dmrg method do not work with the total spin as a good quantum number , but only with it s z - component , ferromagnetic interactions are generally also more difficult to treat , due to the fact that the ground - state belongs to a different multiplet for even and odd @xmath152 .
however , the effective hamiltonians we have obtained in the previous sections can be derived for a ferromagnetic coupled impurity in a straightforward way .
below we briefly discuss what one finds .
for a central bond that is _ weak _ , @xmath237 , and _
ferromagnetic _ in an otherwise antiferromagnetically coupled @xmath0 chain , we expect the results of sec .
[ sec : weak.link ] to still apply .
in particular eq .
( [ eq : deweak ] ) should still be correct .
however , in this case the ground state of eq .
( [ eq : heff.weak ] ) is the triplet . for a central bond that is _ strong _ , @xmath238 , and _ ferromagnetic _ , the two central spins , @xmath82 and @xmath210 , will have a @xmath239 ground - state .
we then obtain an effective hamiltonian in terms of an effective @xmath239 impurity : @xmath240 where the term describing the @xmath239 impurity is given by : @xmath241\nonumber\\ & = & \gamma j[{\bf s}_{l/2 - 1}\cdot{\bf s}^{\prime\prime}_{s=2}+ { \bf s}^{\prime\prime}_{s=2}\cdot{\bf s}_{l/2 + 2 } ] .
\label{eq : heffs2}\end{aligned}\ ] ] here @xmath242 is determined by the projection of @xmath82 and @xmath210 onto the @xmath239 subspace , @xmath243 .
it is easy to see that @xmath244 , thus , the effective coupling to the @xmath239 impurity is @xmath245 . if the coupling to the @xmath239 impurity is regarded as a free parameter we would expect bound states to appear below a threshold , @xmath246 .
however , in the absence of any numerical data , we have no way of telling whether @xmath245 is above or below such a threshold and we can not draw any conclusions . let us briefly consider the case where the coupling to the @xmath239 impurity is assumed to be a free variable @xmath215 . for @xmath247 we can again use the picture of effective @xmath3 chain - end spins to write an effective low - energy hamiltonian : @xmath248 $ ] .
as in previous sections , @xmath249 and @xmath250 are now effective @xmath3 spins .
this low - energy effective hamiltonian has the following spectrum : a ground - state triplet with energy @xmath251 , followed by a quintuplet at @xmath224 , another quintuplet at @xmath160 , and finally a septuplet at @xmath252 . for sufficiently small @xmath215 , @xmath247
, we then expect bound states in the haldane gap to occur at energies , @xmath253 , @xmath254 , and @xmath255 .
the effective hamiltonian , eq .
( [ eq : heff5 ] ) , corresponds to @xmath256 , since this is not a small coupling , a low - energy hamiltonian based on the @xmath3 chain - end spins is presumably not valid .
since we do not know @xmath246 , it is not possible to tell if bound states occur .
if a @xmath3 impurity is coupled with a _ weak ferromagnetic _
bond to the otherwise antiferromagnetically coupled @xmath0 chain , we expect bound states to occur . as already mentioned in sec . [
sec : shalf.weak ] , we expect these to have energies @xmath257 , with @xmath258 . for a @xmath3 impurity coupled with a
_ strong ferromagnetic _ bond to an otherwise uniform @xmath0 chain , we know from sec .
[ sec : shalf.strong ] that , the ground state of @xmath26 , eq .
( [ eq : h0.shalf ] ) should be a sextuplet , @xmath197 .
if we by @xmath259 denote projection onto the @xmath260 subspace , we can define a constant , @xmath261 , by @xmath262 .
we can then write an effective hamiltonian in terms of an effective @xmath263 impurity in the following way : @xmath264 where the term describing the @xmath263 impurity is given by : @xmath265\nonumber\\ & = & \delta j[{\bf s}_{l/2 - 1}\cdot{\bf s}^{\prime\prime}_{s=5/2}+ { \bf s}^{\prime\prime}_{s=5/2}\cdot{\bf s}_{l/2 + 2 } ] .
\label{eq : heff5.2}\end{aligned}\ ] ] in this case @xmath266 .
the effective coupling to the @xmath263 impurity is then @xmath267 .
if we regard this coupling as a free parameter , @xmath215 , we expect bound states to occur for @xmath215 below a threshold , @xmath268 .
we are not able to determine if @xmath267 is below such a threshold .
as above , let us consider the case where the coupling to the @xmath263 impurity is a free variable @xmath215 . for @xmath269 we can again use the picture of effective @xmath3 chain - end spins to write an effective low - energy hamiltonian : @xmath270 $ ] .
as above , @xmath249 and @xmath250 are now effective @xmath3 spins .
this low - energy effective hamiltonian has the following spectrum : a ground - state quadruplet with energy @xmath271 , followed by a sextuplet at @xmath224 , another sextuplet at @xmath160 , and finally an octuplet at @xmath272 .
for sufficeintly small @xmath215 we then expect bound states in the haldane gap to occur at energies , @xmath227 , @xmath273 , and @xmath274 . the effective hamiltonian , eq .
( [ eq : heff6 ] ) , corresponds to @xmath275 , this coupling can not be regarded as small and we are not in the regime @xmath269 .
thus , the description based on the @xmath3 chain - end spins is of limited use .
however , whether @xmath267 is below or above @xmath268 is an open question .
we see that only in specific cases is it straight forward to derive a useful _ low - energy _ effective hamiltonian . for a strong ferromagnetic bond , and for a @xmath3 impurity coupled with a _ strong ferromagnetic _ bond , further numerical work will be needed in order to determine whether bound states in the haldane gap exists .
we have shown that , by the introduction of an af bond with @xmath276 in an otherwise isotropic af @xmath0 chain , a triplet bound state below the haldane gap occurs for sufficiently large and small @xmath2 .
the energy of the bound state is in very good agreement with perturbative results based on effective @xmath3 chain - end excitations created by broken valence bonds . in a region
@xmath277 no bound state is found to within numerical precision .
if a @xmath3 impurity is introduced with an af coupling @xmath276 in an isotropic af @xmath0 chain we find that , only for @xmath2 below @xmath278 do bound states occur .
for @xmath279 a bound state appears , and for @xmath280 a second bound state is stable . at large impurity couplings , @xmath281 , we find no bound states . in experiments on y@xmath24ca@xmath22banio@xmath6
@xcite , states below the haldane gap have been found . in a simple linear chain spin model of y@xmath24ca@xmath22banio@xmath6
, our results would imply that the coupling to a @xmath3 impurity , introduced by the ca doping , should be rather small , if it is antiferromagnetic .
this may seem surprising since we would expect @xmath282 to be the more likely scenario in y@xmath24ca@xmath22banio@xmath6 .
however , the coupling could be large and _ ferromagnetic _ in which case we can not rule out the existence of bound states . secondly , the non - zero single ion anisotropy term , @xmath283 , relevant for y@xmath24ca@xmath22banio@xmath6 could create bound states in the gap even for a large antiferromagnetic coupling to the @xmath3 impurity
. also , the substitution of y@xmath23 by ca@xmath13 will lower the crystal field symmetry on close by ni sites , justifying new _ local _ anisotropy terms in a spin hamiltonian .
possibly such terms could create bound states .
ditusa et al .
@xcite argue that the magnetic disturbance caused by the ca doping introduces either a triplet or a quartet of subgap bound states .
our results indicate that several different multiplets of bound states are likely to be present , some of which have spin larger than @xmath284 , even for the simple scenario of ca doping giving rise to @xmath3 impurities , thus justifying the large contribution to the spectral weight from the impurities seen in the experiments .
our finding that a critical coupling is needed to form bound states in this problem may seem surprising at first , since one would expect that even an infinitely small impurity potential would form a bound state in one dimension .
however , in a nl@xmath104 model description one can argue that the introduction of impurities do not give rise to simple local potentials due to the constraint @xmath106 .
this might change in a model where the rotational symmetry is broken by the introduction of a single ion anisotropy term , @xmath283 .
such a term is relevant for the description of nenp and y@xmath5banio@xmath6 .
it would therefore be very interesting to repeat the above dmrg calculations with @xmath285 .
we gratefully acknowledge many enlightning discussions with w. j. l. buyers .
one of us ( ess ) would like to thank c. broholm , d. reich , r. hyman , s. m. girvin and a. macdonald for useful discussions and comments .
ess is supported by nsf grant number nsf dmr-9416906 .
this research was supported in part by nserc of canada .
r. botet and r. julien , phys .
b * 27 * , 613 , ( 1983 ) ; r. botet , r. julien , and m. kolb , phys .
b * 28 * , 3914 ( 1983 ) ; m. kolb , r. botet , and r. julien , j. phys .
a * 16 * l673 ( 1983 ) ; m. p. nightingale and h. w. j. bl " ote , phys .
b * 33 * , 659 ( 1986 ) ; h. j. schulz and t. a. l. ziman , phys .
b * 33 * , 6545 ( 1986 ) .
o. golinelli , th .
jolicur , and r. lacaze , phys . rev .
b * 46 * , 10854 ( 1992 ) ; j. phys .
matt . * 5 * , 1399 ( 1994 ) .
s. haas , j. riera , and e. dagotto , phys .
b * 48 * , 3281 ( 1993 ) ; m. takahashi , phys .
rev.b * 48 * , 311 ( 1993 ) ; * 50 * , 3045 ( 1994 )
. j. p. renard , m. verdaguer , l.p .
regnault , w. a. c. erkelens , j. rossat - mignod , j. ribas , w. g. stirling , and c. vettier , j. appl .
phys . * 63 * , 3538 ( 1988 ) ; w. lu , j. tuchendler , m. von ortenberg , and j. p. renard , phys . rev.lett .
* 67 * , 3716 ( 1991 ) ; l. p. regnault , c. vettier , j. rossat - mignod , and j. p. renard , physica b * 180 - 181 * , 188 ( 1992 ) ; l. c. brunel , t. m. brill , i. zaliznyak , j. p. boucher , and j. p. renard , phys .
* 69 * , 1699 ( 1992 ) ; s. ma , c. broholm , d. h. reich , b. j. sternlieb , and r. w. erwin , phys .
* 69 * , 3571 ( 1992 ) ; w. palme , h. kriegelstein , b. lthi , t. m. brill , t. yosida , m. date , int .
j. mod .
b * 7 * , 1016 ( 1993 ) .
m. hagiwara , k. katsumata , i. affleck , b. i. halperin , and j. p. renard , phys .
65 * , 3181 ( 1990 ) ; m. hagiwara , k. katsumata , h. hori , t. takeuchi , m. date , a. yamagishi , j. p. renard , and i. affleck , physica b * 177 * , 386 ( 1992 ) ; m. hagiwara , k. katsumata , j. p. renard , i. affleck , and b. i. halperin , j. mag .
mag . mat .
* 104 * , 839 ( 1992 ) . | the @xmath0 heisenberg antiferromagnet is studied in the presence of two kinds of local impurities .
first , a perturbed antiferromagnetic bond with @xmath1 at the center of an even - length open chain is considered . using the density matrix renormalization group method we find that , for sufficiently strong or weak @xmath2 , a bound state is localized at the impurity site , giving rise to an energy level in the haldane gap .
the energy of the bound state is in agreement with perturbative results , based on @xmath3 chain - end excitations , both in the weak and strong coupling limit . in a region around the uniform limit , @xmath4 ,
no states are found with energy below the haldane gap .
secondly , a @xmath3 impurity at the center of an otherwise even - length open chain is considered .
the coupling to the @xmath3 impurity is varied .
bound states in the haldane gap are found _ only _ for sufficiently weak ( antiferromagnetic ) coupling . for a @xmath3 impurity coupled with a strong ( antiferromagnetic ) bond , _ no _
states are found in the haldane .
our results are in good qualitative agreement with recent experiments on doped nenp and y@xmath5banio@xmath6 . |
here , we describe the iterative procedure aiming at constructing the set of stochastic matrices @xmath93 .
let us define two matrices @xmath94 and @xmath95 whose entries are initially set to zero . at the @xmath20th iteration ,
the @xmath20th row of @xmath94 and the @xmath20th column of @xmath95 are filled by setting @xmath96 and @xmath97 , where the pivot entry @xmath98 is the negative of the escape probability from state @xmath20 and @xmath30 is the probability of a round - trip from the same state .
given stochastic matrix @xmath39 , the probability of an indirect transition from @xmath4 to @xmath23 via @xmath20 is @xmath99^{f}p_{n\gamma}^{(n-1)}=\mathcal{l}_{\beta n}u_{n\gamma},\ ] ] where the sum accounts for the probabilities of all possible round - trips from @xmath20 .
since @xmath100 for @xmath101 , matrix @xmath94 remains upper triangular after the @xmath20th row addition and the equality @xmath102 remains satisfied for @xmath21 .
the @xmath20th iteration is completed by constructing @xmath31 as follows : @xmath103 where @xmath104 for @xmath21 , as required .
this property holds by induction on @xmath20 up to @xmath105 . after adding @xmath106 in , the probabilities of transitions from @xmath4 to @xmath2 ( @xmath23 )
subsume the cancelled probability of all possible transitions from @xmath4 to @xmath20 .
this ensures that the transformed matrices @xmath31 remain stochastic .
the present path factorization is connected to gauss - jordan elimination method .
summing @xmath107 from @xmath108 to @xmath109 when @xmath110 or from @xmath111 to @xmath112 when @xmath113 yields the relation @xmath114 let @xmath115 be the lower triangular matrix defined by @xmath116 for @xmath117 and @xmath118 otherwise .
let also denote the upper triangular matrix @xmath119 by @xmath120 . then substituting @xmath121 for @xmath122 and identifying yields the equivalent relation ( @xmath123 ) @xmath124 choosing @xmath125 in eq
. entails @xmath126 and @xmath127 , while setting @xmath128 yields @xmath129 .
factorizations of @xmath130 and @xmath131 result from transformations above and below the eliminated pivot , respectively .
this corresponds to the gauss - jordan pivot elimination method .
the conditional probabilities used in the randomization procedure write @xmath132 by resorting to and and by using the stored entries of @xmath94 and @xmath133 , the information necessary for evaluating the conditional probabilities used in the iterative reverse randomization ( see sec .
[ algorithm ] below ) can be easily retrieved . from @xmath36 with @xmath134 : dashed and solid arrows point to eliminated and non - eliminated states , respectively .
arrow labels indicate typical numbers of transitions .
a transition out of an eliminated state means that an exiting path starts from this state : here , two paths start from @xmath4 and three from @xmath20.,scaledwidth=80.0% ]
to show how the space - time randomization procedure is implemented in practice , the following preliminary definitions are required .
the binomial law of trial number @xmath135 and success probability @xmath136 is denoted by @xmath137 .
the probability of @xmath138 successes is @xmath139 .
the negative binomial law of success number @xmath140 and success ( escape ) probability @xmath141 is denoted by @xmath142 .
the probability of @xmath143 failures before the @xmath140-th success is @xmath144 where @xmath145 is the failure or flicker probability ( flickers will correspond to round - trips from a given state ) .
the gamma law of shape parameter @xmath140 and time - scale @xmath10 is denoted by @xmath146 .
@xmath147 denotes the categorical laws whose probability vector is the @xmath34-th line of @xmath33 if @xmath148 or of the stochastic matrix obtained from @xmath15 by eliminating the single state @xmath149 .
the symbol @xmath150 means `` is a random variate distributed according to the law that follows '' .
let @xmath151 and @xmath152 denote the set of absorbing states and the set of noneliminated transient states , respectively . the absorbing boundary @xmath153 constains the state of @xmath151 that can be reached directly from @xmath154 .
let @xmath155 be a vector and @xmath34 denote the current state of the system .
the cyclic structure of the algorithm , referred to as kinetic path sampling ( kps ) , is a. compute @xmath33 by iterating on @xmath20 from @xmath45 to @xmath17 , and label the states connected to @xmath35 in ascending order from @xmath156 to @xmath157 through appropriate permutations ; define @xmath151 and @xmath158 ; set the entries of @xmath47 and @xmath155 to zero ; b. [ item : run ] draw @xmath159 ; increment @xmath160 or @xmath161 by one depending on whether @xmath162 or @xmath163 ; move current state @xmath34 to @xmath2 ; if new @xmath164 goto ( [ item : randomize ] ) otherwise repeat ( [ item : run ] ) ; c. [ item : randomize ] iterate in reverse order from @xmath105 to 1 : a. for @xmath165 and @xmath166 draw @xmath167 b. for @xmath165 count the new hops from @xmath4 to @xmath20 @xmath168 c. for @xmath166 count the hops from @xmath20 to @xmath2 @xmath169 d. compute @xmath170 , the number of hops from @xmath20 , and draw the flicker number @xmath171 e. store @xmath172 , the number of transitions from @xmath20 , and deallocate @xmath36 , @xmath31 and @xmath44 ; d. for @xmath173 , store @xmath174 where @xmath175 ; e. evaluate @xmath176 , the total number of flickers and hops associated with the path generated in ; increment the physical time @xmath5 by @xmath177 .
[ item : gamma ] after this cycle , the system has moved to the absorbing state @xmath34 reached in . the gamma law @xmath178 in simulates the time elapsed after performing @xmath179 consecutive poisson processes of rate @xmath180 . indeed , after any hop or flicker performed with @xmath15 , the physical time must be incremented by a residence time drawn in the exponential distribution of decaying rate @xmath180 . the way @xmath35 and @xmath151
are constructed at each cycle is specific to the application . in simulations of anomalous diffusion on a disordered substrate ,
@xmath181 implying @xmath182 .
any transition from @xmath35 reaches @xmath153 . in constrast ,
when @xmath158 is not empty , the generated path may return to set @xmath35 several times prior to reaching @xmath153 . with this more general set - up ,
several transitions exiting @xmath35 are typically recorded in the hopping matrix , as illustrated in fig .
[ fig : diagram ] .
this amounts to storing the path factorization and using it as many times as necessary . as a result , the elapsed physical time is generated for several escaping trajectories simultaneously .
let @xmath183 denote the subset of states @xmath184 whose distances to @xmath4 along ( 1,0 ) and ( 0,1 ) directions are shorter than 48@xmath185 , @xmath185 being the lattice parameter .
then , the energy landscape is constructed using @xmath186 where @xmath187 and @xmath188 are independent random variables taking values -1 or 1 with equal probabilities .
in fe - cu , a kps cycle starts by factoring the evolution operator when the vacancy binds to a cu atom and forms a new @xmath189cu cluster shape .
set @xmath35 contains the configurations corresponding to the initial vacancy position and to the possible vacancy positions inside the neighboring cu cluster ( wherein the vacancy can exchange without moving any fe atom ) .
cu and fe atoms are unlabeled , hence the size @xmath17 of @xmath35 is the number of cu atoms in the cu cluster plus one .
note that labeling the atoms would entail @xmath190 , making the simulations impractical for @xmath191 .
the @xmath20th eliminated pivot in @xmath35 corresponds to the least connected entry of @xmath31 .
the stochastic matrix @xmath33 is then used to evolve the system a first time and then each time the system returns to @xmath35 , as illustrated by the vacancy path of fig .
[ figs2].a .
a kmc algorithm is implemented when the vacancy is embedded in the iron bulk .
the set of non - eliminated transient states , @xmath158 , encompasses all states corresponding to the vacancy embedded in the iron bulk . in practice
, the vacancy tends to return to the cu cluster from which it just exited , meaning that the factorization is used many times during a kps cycle , up to an average of 60 for a cluster of size 40 at @xmath192 k. one stops generating the path whenever the vacancy returns to the initial cu cluster but with a different @xmath189cu@xmath193 cluster shape ( see fig .
[ figs2].b ) or reaches another cu cluster ( see fig . [ figs2].c ) .
the possible ending states define the absorbing boundary @xmath153 .
then , kps completes its cycle with space - time randomization . ) on a square lattice with cu - like atoms colored in apricot .
dashed lines delimit the @xmath189cu@xmath193 cluster shapes . ,
scaledwidth=40.0% ] the copper solubility limits in iron for the three simulated temperatures ( @xmath192 k , @xmath194 k and @xmath195 k ) are @xmath196 , @xmath197 and @xmath198 , respectively .
these limits are very small compared with @xmath199 , the cu concentration in the simulations .
supersaturations in copper are thus very important , meaning that the solid solution is unstable at these temperatures and that the cu clusters that form during subsequent ageing unlikely dissolve during the initial incubation at @xmath200 .
we in fact observe that the cross - over from the slow initial `` incubation '' to faster growth occurs concomittantly with a vacancy - containing copper cluster growing to 15 - 16 cu atoms , as shown in fig .
[ figs3 ] .
additional `` magic numbers '' around @xmath82 , 27 , 35 are evidenced at @xmath200 by visualizing the 41 simulated trajectories displayed in fig .
[ figs4 ] .
these numbers correspond to the numbers of site in compact clusters with fully filled nearest - neighbor shells . of the migrating cu@xmath193v cluster : single ( non - averaged ) ageing kinetics at @xmath201 k.,scaledwidth=45.0% ] associated with the migrating cu@xmath193v clusters for the 41 trajectories simulated at @xmath201 k.,scaledwidth=45.0% ]
we now prove that the algorithm generates the correct distribution of first - passage times to @xmath151 for @xmath90 , i.e @xmath202 . in step ( e ) of the kps algorithm , we resort to the distributivity of the gamma law with respect to its shape parameter and decompose the first - passage time as follows : @xmath203 where @xmath204 and @xmath205 are the generated numbers of hops and flickers from @xmath20 . for any state @xmath173 that is visited @xmath140 times , the probability of having @xmath143 flickers before the @xmath140-th escape from @xmath206 is @xmath207 , which corresponds to the probability mass of the negative binomial law of success number @xmath140 and success probability @xmath208 . in the following ,
we introduce the effective rate @xmath209 the residence time associated with @xmath210 hops and flickers from state @xmath206 is distributed according to @xmath211 , the gamma law of shape parameter @xmath210 and time - scale @xmath10 .
the probability mass of @xmath211 law at @xmath5 being @xmath212 $ ] , the overall probability to draw residence time @xmath5 after @xmath140 visits of state @xmath206 is obtained by summing the compound probabilities associated with all possible occurrences of the number @xmath143 of flickers as follows @xmath213^{f } \exp \left[-\frac{t } { \tau } \right ] = \nonumber \frac { ( k_\ell t ) ^{h-1 } } { ( h-1 ) ! } k_\ell \exp\left [ - k_\ell t \right ] .
\nonumber\end{aligned}\ ] ] remarkably , the summation removes the dependence on @xmath10 and yields the distribution of the gamma law of shape parameter @xmath140 and time - scale @xmath214 , which corresponds to the convolution of @xmath140 decaying exponentials of rate @xmath215 .
this is the expected distribution for the time elapsed after @xmath140 consecutive poisson processes of rate @xmath215 .
note that the standard kmc algorithm simply draws an escape time according to @xmath216 , the exponential law of rate @xmath215 for each visit of @xmath206 , which is statistically equivalent .
this amounts to prescribing a success probability of 1 , which results in @xmath217 and no failures .
the proof that the algorithm is correct for @xmath218 follows by induction on @xmath17 resorting to the iterative structure of the algorithm . finally , it is worth noticing that trajectories escaping a trapping basin can in principle be analysed to provide information about the kinetic pathways since the numbers of transitions involving each eliminated state are randomly generated . obtaining information about time correlation functions and commitor probabilities
would require testing the involved bernoulli trials sequentially instead of drawing a binomial or negative binomial deviates given a number of transitions , in order to construct the path .
however , the computational cost of these additional operations scales linearly with the mean first - passage time for escaping a trapping basin . | the computational efficiency of stochastic simulation algorithms is notoriously limited by the kinetic trapping of the simulated trajectories within low energy basins . here
we present a new method that overcomes kinetic trapping while still preserving exact statistics of escape paths from the trapping basins .
the method is based on path factorization of the evolution operator and requires no prior knowledge of the underlying energy landscape .
the efficiency of the new method is demonstrated in simulations of anomalous diffusion and phase separation in a binary alloy , two stochastic models presenting severe kinetic trapping .
time evolution of many natural and engineering systems is described by a master equation ( me ) , i.e. a set of ordinary differential equations for the time - dependent vector of state probabilities @xcite . for models with large ( or infinite but countable ) number of states , direct solution of the me
is prohibitive and kinetic monte carlo ( kmc ) is used instead to simulate the time evolution by generating sequences of stochastic transitions from one state to the next @xcite .
statistically equivalent to the ( most often unknown ) solution of the me , kmc finds growing number of applications in natural and engineering sciences
. however still wider applicability of kmc is severely limited by the notorious kinetic trapping where the stochastic trajectory repeatedly visits a subset of states , a trapping basin , connected to each other by high - rate transitions while transitions out of the trapping basin are infrequent and take great many kmc steps to observe . in this letter , we present an efficient method for sampling stochastic trajectories escaping from the trapping basins . unlike recent methods that focus on short portions of the full kinetic path directly leading to the escapes and/or require equilibration over a path ensemble @xcite , our method constructs an entire stochastic trajectory within the trapping basin including the typically large numbers of repeated visits to each trapping state as well as the eventual escape . referred hereafter as kinetic path sampling ( kps ) ,
the new algorithm is statistically equivalent to the standard kmc simulation and entails ( i ) iterative factorization of paths inside a trapping basin , ( ii ) sampling a single exit state within the basin s perimeter and ( iii ) generating a first - passage path and an exit time to the selected perimeter state through an exact randomization procedure .
we demonstrate the accuracy and efficiency of kps on two models : ( 1 ) diffusion on a random energy landscape specifically designed to yield a wide and continuous spectrum of time scales and ( 2 ) kinetics of phase separation in super - saturated solid solutions of copper in iron .
the proposed method is immune to kinetic trapping and performs well under simulation conditions where the standard kmc simulations slows down to a crawl .
in particular , it reaches later stages of phase separation in the fe - cu system and captures a qualitatively new kinetics and mechanism of copper precipitation .
the evolution operator , obtained formally from solutions of the me , can be expressed as an exponential of the time - independent transition rate matrix @xcite @xmath0 where @xmath1 is the probability to find the system in state @xmath2 at @xmath3 given that it was in state @xmath4 at time @xmath5 , @xmath6 is the rate of transitions from state @xmath4 to state @xmath2 ( off - diagonal elements only ) and the standard convention is used to define the diagonal elements as @xmath7 .
as defined , the evolution operator belongs to the class of stochastic matrices such that @xmath8 and @xmath9 for any @xmath4 , @xmath2 , @xmath5 and @xmath10 .
if known , the evolution operator can be used to sample transitions between any two states and over arbitrary time intervals @xmath10 @xcite . in particular ,
substantial simulation speed - ups can be achieved by sampling transitions to distant states on an absorbing perimeter of a trapping basin .
two main defficiencies of the existing implementations of this idea @xcite is that states within the trapping basin are expected to be known _ a priori _ and that computing the evolution operator requires a partial eigenvalue decomposition of @xmath11 entailing high computational cost @xcite .
in contrast , the kps algorithm does not require any advance knowledge of the trapping basin nor does it entail matrix diagonalization .
instead , kps detects kinetic trapping and charts the trapping basin iteratively , state by state , and achieves high computational efficiency by sequentially eliminating all the trapping states through path factorization @xcite . here , wales s formulation @xcite of path factorization is adopted for its clarity .
consider the linearized evolution operator @xmath12 where @xmath13 is the identity matrix .
assuming that @xmath14 , @xmath15 is a proper stochastic matrix that can be used to generate stochastic sequences of states from the ensemble of paths defined by matrix @xmath11 .
the diagonal elements of @xmath15 define the probabilities of round - trip transitions after which the system remains in the same state . to correct for the linearization of the evolution operator in ,
the time elapsed before any transition takes place is regarded as a stochastic variable and sampled from an exponential distribution @xmath16 @xcite .
this simple time randomization obviates the need for exponentiating the transition rate matrix in .
following @xcite , consider a bi - directional connectivity graph defined by @xmath15 in which @xmath17 states in the trapping basin are numbered in order of their elimination , @xmath18 .
an iterative path factorization procedure then constructs a set of stochastic matrices @xmath19 such that , after the @xmath20-th iteration , all states @xmath21 are eliminated in the sense that the probability of a transition from any state @xmath4 to state @xmath21 is zero .
specifically , at @xmath20-step of factorization the transition probability @xmath22 ( @xmath23 , @xmath24 ) is computed as the sum of the probability of a direct transition @xmath25 and the probabilities of all possible indirect paths involving round - trips in @xmath20 after having initially transitioned from @xmath4 to @xmath20 and before finally transitioning from @xmath20 to @xmath2 , e.g. @xmath26 , @xmath27 , @xmath28 , @xmath29 , and so on . with the round - trip probability being @xmath30
, it is a simple matter to sum the geometric series corresponding to the round - trip paths .
although any intermediate @xmath31 can be used to generate stochastic escapes from any state @xmath32 , a trajectory generated using @xmath33 is the simplest containing a single transition from @xmath34 that effectively subsumes all possible transitions involving the deleted states in the trapping basin @xmath35 . on the other end , a detailed escape trajectory can be generated using @xmath15 that accounts for all transitions within @xmath35 , reverting back to the standard kmc simulation .
remarkably , it is possible to construct a detailed escape trajectory statistically equivalent to the standard kmc without ever performing a detailed ( and inefficient ) kmc simulation .
consider matrix @xmath36 whose elements @xmath37 store the number of @xmath26 transitions observed in a stochastic simulation with @xmath31 .
given @xmath36 , one can randomly generate @xmath38 , the matrix similarly used to count the transition numbers observed in a stochastic process based on @xmath39 without actually performing the simulation using @xmath39 .
the ratio of transition probabilities ( @xmath23 ) @xmath40 defines the conditional probability that a trajectory generated using @xmath39 contains a direct transition from @xmath4 to @xmath2 given that the trajectory generated with @xmath31 contains the same transition .
for @xmath41 , @xmath42 is independent of @xmath2 and is equal to @xmath43 , the probability of escape from @xmath20 .
it is thus possible to generate @xmath38 by performing a stochastic simulation with @xmath31 , harvesting @xmath36 and drawing random variates from ( standard and negative ) binomial distributions whose exponents and coefficients are given by the elements of @xmath36 and @xmath44 , respectively .
this randomization procedure can be used iteratively on @xmath20 in the reverse order from @xmath17 to @xmath45 to generate @xmath46 containing a detailed count of transitions involving all states in @xmath35 . finally , the time of exit out of @xmath35 is sampled by drawing a random variate from the gamma distribution whose scale and shape parameters are defined by @xmath10 and the total number of transitions contained in @xmath46 , respectively .
thus , in its simplest form the kps algorithm proceeds by first deleting all states in @xmath35 through iterative forward path factorization , then using @xmath33 to sample a single transition from @xmath32 and to generate @xmath47 followed by a backward randomization to reconstruct a detailed stochastic path within @xmath35 and to sample an escape time to the selected exit state .
a detailed description of the kps algorithm is given in the supplemental material @xcite .
we first apply kps to simulations of a random walker on a disordered energy landscape ( substrate ) @xcite .
the substrate is a periodically replicated 256@xmath48256 fragment of the square lattice on which the walker hops to its four nearest - neigbour ( nn ) sites with transition rates @xmath49,\ ] ] where @xmath50 is the temperature , @xmath51 the site energy and @xmath52 the saddle energy between sites @xmath4 and @xmath2 .
the energy landscape is purposefully constructed to contain trapping basins of widely distributed sizes and depths ( see the supplemental material @xcite for details ) and is centered around the walker s initial position next to the lowest energy saddle ( fig .
[ fig : landscape ] ) . .
artificial smoothing is used for better visualization.,scaledwidth=60.0% ] when performed at temperature t@xmath53 , standard kmc simulations ( with hops only to the nn sites ) are efficient enabling the walker to explore the entire substrate .
however at t@xmath54 , the walker remains trapped near its initial position repeatedly visiting states within a trapping basin . to chart a basin set @xmath35 for subsequent kps simulations , the initial state 1
is eliminated at the very first iteration followed in sequence by the `` most absorbing states '' for which @xmath55 is found to be largest at the @xmath20-th iteration ( @xmath56 ) .
the expanding contours shown in fig . [ fig : landscape ] depict the absorbing boundary @xmath57 ( perimeter of the basin ) obtained after eliminating @xmath58 , @xmath59 , @xmath60 , @xmath61 , @xmath62 and @xmath63 states .
the perimeter contour @xmath57 consists of all states @xmath2 for which @xmath64 is nonzero . for the walker to remain within a trapping basin containing @xmath65 states ( solid line ) and the distribution of times of escape out of the same basin ( dashed lines ) using a log scale for the bins ; ( b ) the mean first - passage time @xmath66 as a function of the number of states @xmath17 included in the trapping basin ; ( c ) computational cost of kmc and kps simulations as a function of @xmath17 at two different temperatures ( in the units of a single kmc hop ) . , scaledwidth=45.0% ] to demonstrate correctness of kps , we generated @xmath67 paths starting from state @xmath68 and ending at the absorbing boundary @xmath57 of the basin containing @xmath65 states , using both kps and kmc at t=2.5 .
the perfect match between the two estimated distributions of exit times is shown in fig .
[ fig2].a .
the mean times of exit to @xmath57 are plotted as a function of the number of eliminated states at t@xmath53 and t=1.0 in fig .
[ fig2].b , while the costs of both methods are compared in fig .
[ fig2].c . at t=1.0 ,
kmc trajectories are trapped and never reach @xmath57 : in this case we plot the expectation value for the number of kmc hops required to exit @xmath35 which is always available after path factorization @xcite .
we observe that the kps cost scales as @xmath69 , as expected for this factorization , and exceeds that of kmc for @xmath70 at t=2.5 .
however , at t=1 trapping becomes severe rendering the standard kmc inefficient and the wall clock speedup achieved by kps is four orders of magnitude for @xmath71 .
we observe that in kps the net cost of generating an exit trajectory is nearly independent of the temperature but grows exponentially with the decreasing temperature in kmc . at the same time , an accurate measure of the relative efficiency of kps and kmc is always available in path factorization , allowing one to revert to the standard kmc whenever it is relatively more efficient .
thus , when performed correctly , a stochastic simulation combining kps and kmc should always be more efficient than kmc alone . as a second illustration
, we apply kps to simulate the kinetics of copper precipitation in iron within a lattice model parameterized using electronic structure calculations @xcite .
the simulation volume is a periodically replicated fragment of the body centered cubic lattice with 128@xmath48128@xmath48128 sites on which 28,163 cu atoms are initially randomly dispersed .
fe atoms occupy all the remaining lattice sites except one that is left vacant allowing atom diffusion to occur by vacancy ( v ) hopping to one of its nn lattice sites . its formation energy being substantially lower in cu than in fe , the vacancy is readily trapped in cu precipitates rendering kmc grossly inefficient below @xmath72 k @xcite . whenever the vacancy is observed to attach to a cu cluster ,
we perform kps over a pre - charted set @xmath35 containing @xmath17 trapping states that correspond to all possible vacancy positions inside the @xmath73 cluster containing @xmath74 cu atoms : the shape of the trapping cluster is fixed at the instant when the vacancy first attaches . the fully factored matrix @xmath33 is then used to propagate the vacancy to a lattice site just outside the fixed cluster shape which is often followed by vacancy returning to the same cluster . if the newly formed trapping cluster has the same shape as before , the factorized matrix is used again to sample yet another escape . however a new path factorization ( kps cycle ) is performed whenever the vacancy re - attaches to the same cu cluster but in a different cluster shape or attaches to another cu cluster ( see the supplementary material for additional simulation details @xcite ) .
we simulated copper precipitation in iron at three different temperatures t@xmath75 k , t@xmath76 k and t@xmath77 k for which the atomic fraction of cu atoms used in our simulations significantly exceeds copper solubility limits in iron . defined as the ratio of physical time simulated by kps to that reached in kmc simulations over the same wall clock time , the integrated speed - up is plotted in fig .
[ fig3].a . as a function of the physical time simulated by kps ( averaged over 41 simulations for each method and at each temperature ) .
the precipitation kinetics are monitored through the evolution of the volume - averaged warren - cowley short - range order ( sro ) parameter @xcite shown in fig .
[ fig3].b both for kps and kmc simulations . at t@xmath78 and t@xmath79
the kinetics proceed through a distinct incubation stage reminescent to a time lag associated with repeated re - dissolution of subcritical nuclei prior to reaching the critical size in the classical nucleation theory @xcite .
however , `` incubation '' observed here is of a distinctly different nature since all our simulated solid solutions are thermodynamically unstable and even the smallest of cu clusters , once formed , never dissolve .
at all three temperatures the growth of @xmath73 clusters is observed to proceed not through the attachment of mobile v - cu dimers but primarily through the cluster s own diffusion and sweeping of neighboring immobile cu monomers @xcite .
this is consistent with an earlier study that also suggested that , rather counter - intuitively , the diffusivity of @xmath73 clusters should increase with the increasing @xmath17 before tapering off at @xmath80 ( see fig .
9 of ref . ) .
we futher observe that at t@xmath78 the cross - over from the slow initial `` incubation '' to faster `` agglomeration '' growth seen on [ fig3].b occurs concomittantly with the largest cluster growing to 15 - 16 cu atoms @xcite .
individual realizations of the stochastic precipitation kinetics reveal that , in addition to @xmath81 , cluster growth slows down again once the cluster reaches @xmath82 , 27 , 35 and so on ( see figure s4 in the supplementary materials ) . leaving precise characterization of these transitions to future work , we speculate that the observed `` magic numbers '' correspond to compact clusters with fully filled nearest - neighbor shells in which vacancy trapping is particularly strong reducing the rate of shape - modifying vacancy escapes required for cluster diffusion .
numerically , as expected , the integrated speed - up rapidly increases with the decreasing temperature as vacancy trapping becomes more severe .
two line segments of unit slope and two pairs of vertical arrows are drawn in fig .
[ fig3 ] to compare evolution stages achievable within kps and kmc over the same wall clock time . as marked by the pair of two solid vertical arrows on the right , the integrated speed - up exceeds seven orders of magnitude at t@xmath78 .
subsequent reduction in the speed - up concides with the transition into the agglomeration regime where increasingly large vcu clusters repeatedly visit increasingly large number of distinct shapes @xcite .
unquestionably , the efficiency of kps simulations for this particular model can be improved by indexing distinct cluster shapes for each cluster size and storing the path factorizations to allow for their repeated use during the simulations @xcite . in any case , given its built - in awareness of the relative cost measured in kmc hops , kps is certain to enable more efficient simulations of diffusive phase transformations in various technologically important materials . in particular
, it is tempting to relate an anomalously long incubation stage observed in aluminium alloys with mg , si and se additions @xcite to possible trapping of vacancies on se , similar to the retarding effect of cu on the ageing kinetics reported here for the fe - cu alloys . in summary
, we developed a kinetic path sampling algorithm suitable for simulating the evolution of systems prone to kinetic trapping .
unlike most other algorithms dealing with this numerical bottleneck @xcite , kps does not require any _ a priori _ knowledge of the properties of the trapping basin .
it relies on an iterative path factorization of the evolution operator to chart possible escapes , measures its own relative cost and reverts to standard kmc if the added efficiency no longer offsets its computational overhead . at the same time
, the kps algorithm is exact and samples stochastic trajectories from the same statistical ensemble as the standard kmc algorithms .
being immune to kinetic trapping , kps is well positioned to extend the range of applicability of stochastic simulations beyond their current limits .
furthermore , kps can be combined with spatial protection @xcite and synchronous or asynchronous algorithms to enable efficient parallel simulations of a still wider class of large - scale stochastic models @xcite .
this work was supported by defi needs ( project mathdef ) and lawrence livermore national laboratory s ldrd office ( project 09-erd-005 ) and utilized hpc resources from genci-[ccrt / cines ] ( grant x2013096973 ) .
this work was performed under the auspices of the u.s .
department of energy by lawrence livermore national laboratory under contract de - ac52 - 07na27344 .
the authors wish to express their gratitude to t. opplestrup , f. soisson , e. clouet , j .- l .
bocquet , g. adjanor and a. donev for fruitful discussions .
10 n. g. van kampen , stochastic processes in physics and chemistry ( elsevier science , 2007 ) .
s. redner , a guide to first - passage processes , cambridge university press ( 2001 ) .
j .- m .
lanore , rad .
effects * 22 * , 153 ( 1974 ) .
a. bortz , m. kalos , j. lebowitz , j. comp .
phys .
* 17 * , 10 ( 1975 ) .
d. gillespie , j. chem .
phys . * 81 * , 2340 ( 1977 ) .
p. bolhuis , d. chandler , c. dellago and p. geissler , ann .
rev .
phys . chem . *
53 * , 291 ( 2002 ) .
s. x. sun , phys .
rev .
lett . * 96 * 210602 ( 2006 ) and * 97 * , 178902 ( 2006 ) .
b. harland and x. sun , j. chem .
phys .
* 127 * 104103 ( 2007 ) . c. d. van siclen , j. phys . : condens .
matter 19 , 072201 ( 2007 ) .
n. eidelson and b. peters j. chem .
phys . * 137 * , 094106 ( 2012 ) .
t. mora , a. m. walczak and f. zamponi , phys .
rev .
e * 85 * , 036710 ( 2012 ) .
m. manhart and a. v. morozov , phys .
rev .
lett .
* 111 * 088102 ( 2013 ) .
the evolution operator is obtained by integrating the me @xmath83 from @xmath5 to @xmath3 and identifying the formal solution with @xmath84 where @xmath85 is the state - probability ( row ) vector at @xmath5 .
if the system is in @xmath4 at a given time , then the state - probability vector at a later time @xmath10 is @xmath86 where @xmath87 denotes the row vector whose @xmath4th entry is one and the other ones are zero .
the entries of @xmath88 are @xmath89 and can be used as transition probabilities for kmc moves from @xmath4 .
m. a. novotny , phys .
rev .
lett . * 74 * , 1 ( 1995 ) .
g. boulougouris and d. theodorou , j. chem . phys . *
127 * , 084903 ( 2007 ) .
m. barrio , a. leier and t. marquez - lago , j. chem phys . * 138 * , 104114 ( 2013 ) .
g. nandipati , y. shim and j. g. amar , phys .
rev .
b * 81 * 235415 ( 2010 ) . c. moler and c. van loan , siam rev . *
45 * , 3 ( 2003 ) .
m. athnes , p. bellon and g. martin , phil . mag . a , * 76 * , 565 ( 1997 ) .
s. trygubenko and d. wales , j. chem . phys . * 124 * , 234110 ( 2006 ) .
d. wales , j. chem .
phys . * 130 * , 204111 ( 2009 ) .
s. a. serebrinsky , phys .
rev .
e * 83 * , 037701 ( 2011 ) .
see supplemental material below for the connection between path factorization and gauss - jordan elimination method and for a proof that the algorithm is correct for @xmath90 .
y. limoge and j .- l .
bocquet , phys .
rev .
lett .
* 65 * , 60 ( 1990 ) .
f. soisson , c. c. fu , phys . rev .
b * 76 * , 214102 ( 2007 ) .
m. athnes , p. bellon and g. martin , acta mat .
* 48 * , 2675 , ( 2000 ) .
e. clouet , asm handbook vol .
22a , fundamentals of modeling for metals processing d. u. furrer and s. l. semiatin ( eds . ) , pp . 203 - 219 ( 2010 ) . to understand the origin of the efficiency decrease
, we have monitored the number of distinct shapes of the vacancy - copper cluster . for a v - cu@xmath91 cluster
, we found that , over the @xmath92 last factorizations that have been performed , there are only 21 different cluster shapes and that the 5 most frequent shapes occur with a frequency of about 60% .
l. k. bland , p. brommer , f. el - mellouhi , j. f. joly and n. mousseau , phys .
rev .
e * 84 * , 046704 ( 2011 ) .
s. pogatscher , h. antrekowitsch , m. werinos , f. moszner , s. s. a. gerstl , m. f. francis , w. a. curtin , j. f. lffler and p. j. uggowitzer , phys .
rev .
lett . * 112 * , 225701 ( 2014 ) .
d. mason , r. rudd and a. sutton , comp .
. comm . *
160 * , 140 ( 2004 ) .
b. puchala , m. falk and k. garikipati , j. chem . phys .
* 132 * , 134104 ( 2010 ) .
t. opplestrup , v. v. bulatov , g. h. gilmer , m. h. kalos , and b. sadigh , phys .
rev .
lett . * 97 * , 230602 ( 2006 ) .
y. shim and j. g. amar , phys .
rev .
b * 71 * , 115436 ( 2005 ) . m. merrick and k. fichthorn , * 75 * , 011606 ( 2007 ) .
e. martnez and p. r. monasterio , j. marian , j. comput . phys . *
230 * , 1359 ( 2011 ) .
f. wieland , and d. jefferson , proc .
1989 intl conf .
parallel processing , vol.iii , f. ris , and m. kogge , eds .
, pp .
255 - 258 .
* supplemental materials : path factorization approach to stochastic simulations * |
models based on coupled harmonic oscillators have long attracted attention in several different fields due to their wide range of applications @xcite . in particular , the case of two harmonic modes coupled through their angular momentum , which describes the motion of a charged particle within a general harmonic trap in a uniform magnetic field or , equivalently , the motion in a rotating anisotropic harmonic potential @xcite , has been employed in distinct scenarios , such as rotating nuclei @xcite , quantum dots in a magnetic field @xcite and fast rotating bose - einstein condensates @xcite within the lowest landau level approximation @xcite . since its hamiltonian is quadratic in the field operators , the model is also suitable for simulation with optical techniques @xcite . in a previous work @xcite its dynamics in coordinate representation
was analyzed in detail , showing that it exhibits a complex dynamical phase diagram , with stable as well as distinct types of unstable ( i.e. , unbounded ) dynamics .
we have also examined the generated entanglement between the modes , both in ground and thermal states @xcite ( vacuum and thermal entanglement ) as well as that obtained after starting from a separable vacuum state @xcite .
it was shown , in particular , that this system is able to mimic typical entanglement growth regimes arising after a quantum quench in complex many body scenarios @xcite .
entanglement is of course essential for quantum information applications @xcite , and a large entanglement growth with time after starting from a separable state in a many - body system , is indicative of a system dynamics which can not be efficiently simulated by classical means . on the other hand
, quantum squeezing constitutes another topic of great current interest @xcite .
its relation with entanglement has been investigated in different systems @xcite , with entanglement normally inducing squeezing in certain observables .
in particular , in @xcite the exact dynamics of entanglement and squeezing in a two - mode bose - einstein condensate interacting through a josephson - like coupling was determined .
squeezing is important for quantum metrology , i.e. , for improving the accuracy in quantum measurements @xcite , and it has been shown that in some cases spin squeezing can be employed to detect entanglement @xcite . nonetheless , in some regimes ( like the linear case of the model considered in @xcite ) squeezing may also arise without substantial entanglement . in this work
we first derive the exact analytic expressions for the temporal evolution of the heisenberg field operators of two harmonic modes coupled through their angular momentum .
the obtained result is valid for all values of the system parameters , i.e. , in stable as well as unstable dynamical regimes , and allows to determine the exact evolution of an arbitrary observable of the system .
we then apply this result to determine and examine the dynamics of squeezing , which has so far not been investigated in this model , and its relation with the generated entanglement , when starting from a separable coherent initial state .
we will show that different regimes can arise depending on the value of the rotational frequency .
close to the instability point , appreciable entanglement is generated , accompanied with simultaneous squeezing in one of the variables of each mode , while for small couplings , the generated entanglement is weak , with small squeezing appearing in both variables of the mode at alternating times .
approximate simple expressions describing these two distinct regimes are also provided .
the hamiltonian of the system under study can be written as @xmath0 where the subscripts @xmath1 and @xmath2 refer obviously to the first and second oscillator .
the oscillator frequencies @xmath3 , @xmath4 and the rotation frequency @xmath5 will be assumed real and satisfying , without loss of generality , @xmath6 in the case of a particle of charge @xmath7 in a magnetic field @xmath8 along the @xmath9 axis within a harmonic trap with spring constants @xmath10 , @xmath11 in the @xmath12 plane , we have @xmath13 , with @xmath14 ( @xmath15 ) @xcite ( motion along @xmath9 is obviously decoupled from that in @xmath12 plane , which is that described by ( [ 1 ] ) ) .
( [ 1 ] ) also represents the intrinsic hamiltonian describing the motion ( in the @xmath12 plane ) of a particle in a harmonic trap with constants @xmath16 rotating with frequency @xmath5 around the @xmath9 axis @xcite . by expressing the position ( @xmath17 ) and momentum ( @xmath18 ) operators in terms of the dimensionless annihilation ( @xmath19 ) and creation @xmath20 boson operators , @xmath21 , @xmath22 , @xmath15 , we can rewrite ( [ 1 ] ) as @xmath23 where @xmath24 in the isotropic case @xmath25 , @xmath26 and both @xmath27 and the angular momentum , which becomes just @xmath28 , commute with the total boson number @xmath29 and between themselves .
however , in the anisotropic case @xmath30 , the angular momentum term does not commute with @xmath27 and no longer conserves the total boson number .
this entails , in particular , that in contrast with the isotropic case , the system vacuum will become entangled as @xmath5 increases , and that an initially separable state @xmath31 ( product of the vacuum of each oscillator ) will become entangled as time increases if @xmath32 . in addition
, the system may become unstable if @xmath5 increases sufficiently , as discussed below @xcite .
note that the exact dynamics of the special case @xmath33 with a @xmath34 coupling , which was studied in detail in @xcite , corresponds formally to the case @xmath35 , after replacing @xmath36 .
also , the two - mode bose einstein condensate model of @xcite ( which contains as well non - linear terms ) would formally correspond in the linear case to @xmath26 and a time - dependent @xmath37 .
let us now derive the explicit solution of the heisenberg equations of motion for the field operators , @xmath38\,,\,\,\,\ , j=1,2\,.\label{6}\ ] ] eq .
( [ 6 ] ) leads to the linear system @xmath39 which can be written in matrix form as @xmath40 where @xmath41 , @xmath42 and @xmath43 is the @xmath44 non - hermitian matrix @xmath45 the exact solution of eq .
( [ 8 m ] ) can be expressed as @xmath46 where @xmath47=\left(\begin{array}{cc}u(t)&v(t)\\ v^*(t)&u^*(t)\end{array}\right)\,,\label{exp}\ ] ] is a @xmath44 matrix satisfying ( @xmath48 denotes the @xmath49 identity matrix ) @xmath50 since @xmath51 .
( [ un ] ) ensures the preservation of the equal time commutation relations @xmath52 @xmath53 : @xmath54_{ij}(t)=\delta_{ij}\ , , \;\;[a_i(t),a_j(t)]=[uv^{\rm t}-vu^t]_{ij}(t)=0\,,\label{com}\ ] ] implying that eq .
( [ m ] ) represents a proper time - dependent bogoliubov transformation for the field operators . setting in what follows @xmath55 , @xmath56 , eq .
( [ m ] ) leads explicitly to @xmath57 and the corresponding adjoint equations for @xmath58 , where the elements @xmath59 , @xmath60 can be obtained from eq .
( [ exp ] ) through the diagonalization of @xmath43 : @xmath61\}\,,\label{sol0}\\ v_{jj}(t)&=&i(-1)^{j+1}\frac{\omega_1\omega_2\lambda_+\lambda_-}{\omega_j\,\delta } ( \frac{\sin\omega_+ t}{\omega_+}-\frac{\sin\omega_- t}{\omega_-})\,,\\ \left(\begin{array}{c}u_{12}(t)\\v_{12}(t)\end{array}\right)&= & \frac{\lambda_{\pm}}{2}[(1+\frac{(\omega_1\pm \omega_2)^2}{2\delta } ) \frac{\sin\omega_+t}{\omega_+}+(1-\frac{(\omega_1\pm\omega_2)^2}{2\delta } ) \frac{\sin\omega_-t}{\omega_-}\nonumber\\&&+i\frac{\omega_1\pm\omega_2}{\delta } ( \cos\omega_+ t-\cos\omega_- t)]\,,\\ u_{21}(t)&=&-u_{12}(t)\,,\;\;v_{21}(t)=v_{12}^*(t)\,,\label{solf}\end{aligned}\ ] ] with @xmath62 here @xmath63 are the system _ eigenfrequencies _ ,
i.e. , the eigenvalues of the matrix @xmath43 ( which are @xmath64 , @xmath65 ) , given by @xmath66 where @xmath67}\\ & = & \sqrt{{\textstyle\frac{(\omega_1 ^ 2-\omega_2 ^ 2)^2}{4}}+2\omega^2(\omega_1 ^ 2+\omega_2 ^ 2)}\,.\end{aligned}\ ] ] it is verified that for @xmath68 ( @xmath69 ) , @xmath70 and hence @xmath71 , leading to @xmath72 , @xmath73 . the free evolution @xmath74 is then recovered .
also , in the special case @xmath75 and @xmath76 we recover the eigenfrequencies @xmath77 of @xcite . on the other hand , in the _ isotropic _ case @xmath25 , we have @xmath26 , implying @xmath78 and @xmath79 which leads finally to @xmath80 this is equivalent to a beam - splitter type transformation of angle @xmath81 of the field operators . in this case
the angular momentum term commutes with @xmath27 and just rotates the field operators with angular frequency @xmath5 .
the general solution ( [ 8])([solf ] ) can be derived by many other methods .
for instance , we may write the solution of eq .
( [ m ] ) for each operator as @xmath82 + \frac{1}{2!}\left(\frac{it}{\hbar}\right)^{2}\left[h,\left[h , a_j\right]\right]+\ldots \label{9}\ ] ] which leads immediately to the form ( [ 8 ] ) for @xmath83 . and
insertion of a trial solution of the form ( [ 8 ] ) in ( [ 7 ] ) ( i.e. , the so called sen - mandal approach @xcite ) leads to a linear system of first order differential equations for the coefficients @xmath59 , @xmath60 , namely @xmath84 , i.e. , @xmath85 where @xmath86 , @xmath87 are the @xmath88 column of @xmath89 , @xmath90 .
( [ 8me ] ) and the initial conditions @xmath91 , @xmath92 lead again to the solution ( [ sol0])([solf ] ) .
we finally notice that the system ( [ 7 ] ) leads , after successive derivation , to the fully decoupled quartic equations @xmath93 for @xmath15 , which lead at once to the eigenfrequencies ( [ wpm ] ) ( after inserting a trial solution @xmath94 ) and again to the solution ( [ sol0])([solf ] ) after inserting the initial conditions for the operators and their derivatives .
the matrix @xmath95 , and hence all coefficients @xmath59 , @xmath60 , also satisfy eq .
( [ o4 ] ) . in the general case
, a close inspection of the eigenvalues ( [ wpm ] ) reveals that @xmath63 are both _
real _ and _ non - zero _ only if @xmath96 or @xmath97 , i.e. , if @xmath98 or @xmath99 , which is equivalent to @xmath100 dynamical stability ( bounded quasiperiodic dynamics ) is then ensured if eq.([st ] ) is satisfied . on the other hand ,
if @xmath101 , i.e. , @xmath102 or in general @xmath103 @xmath104 remains real but @xmath105 becomes _ imaginary _
( @xmath106 ) implying that the dynamics becomes _ unbounded_. in this case we should just replace @xmath107 in eqs .
( [ sol0])([solf ] ) , entailing that all operators `` increase '' ( i.e. , deviate from their initial values ) exponentially with time .
nevertheless , eq .
( [ un ] ) and hence the commutation relations ( [ com ] ) remain satisfied . finally , if @xmath108 or @xmath109 , i.e. , if @xmath110 we obtain a _ critical _ regime where @xmath111 but @xmath112 , in which case the matrix @xmath43 becomes _ non - diagonalizable _ unless @xmath33 ( landau case ) .
if @xmath112 , we should just replace the corresponding expressions by their natural limits , i.e. , @xmath113 in eqs .
( [ sol0])([solf ] ) ( the ensuing solution also follows from ( [ exp ] ) after using the jordan canonical form of @xmath43 for @xmath112 @xcite ) , entailing that the dynamics is again unbounded if @xmath30 , with the deviation from the initial values increasing now linearly with time .
( [ un ] ) are again preserved . in the landau case @xmath114 ,
the coefficients of the linearly increasing terms vanish and the dynamics is again bounded , given by eq .
( [ iso ] ) for @xmath115 .
a standard normal mode decomposition of @xmath27 becomes feasible in the dynamically stable phases ( @xmath116 ) @xcite . in terms of the standard normal mode boson operators
@xmath117 , @xmath118 given in the appendix , we may express ( [ 1 ] ) in the first dynamically stable sector @xmath98 as @xmath119 with @xmath120 , whereas in the second dynamically stable sector @xmath99 we have @xmath121 with @xmath122 ( and @xmath123 ) .
( [ r2 ] ) entails that in this region , the system is no longer energetically stable .
we have now all the elements for investigating the evolution of distinct quantum properties of the system , such as entanglement and squeezing .
we start by noting that the number operators for each mode are given by ( here @xmath124 ) @xmath125\ , , \label{34}\end{aligned}\ ] ] indicating that they will acquire a non - zero average even if there are initially no bosons : if the system starts at the separable vacuum @xmath126 , where @xmath127 , from eqs .
( [ 34 ] ) and ( [ sol0])([solf ] ) we obtain , setting @xmath128 and @xmath129 , @xmath130 \,,\label{v2a}\end{aligned}\ ] ] which will be normally non - zero for @xmath131 unless @xmath33 .
@xmath132 is then proportional to @xmath133 , and is larger in the mode with the lowest frequency @xmath134 ( due to the last term in ( [ v2a ] ) ) .
if the initial state is instead a product @xmath135 of _ coherent _ states @xmath136 for each oscillator , with @xmath137 , the same expressions ( [ v2])([v2a ] ) remain valid for the corresponding _ covariance _ of the operators @xmath138 , @xmath83 : @xmath139 where @xmath140 .
( [ v3 ] ) is then _ independent _ of @xmath141 and @xmath142 .
similarly , we may evaluate the coordinates and momenta fluctuations and their dimensionless ratios to their initial values , @xmath143 which for a coherent initial state satisfy @xmath144 , due to the uncertainty principle and the fact that a coherent initial state has minimum uncertainty .
squeezing in @xmath17 or @xmath18 occurs whenever @xmath145 or @xmath146 ) becomes smaller than @xmath1 .
we obtain , explicitly , @xmath147\nonumber\\ & = & 1 + 2[\langle n_j(t)\rangle_0\pm{\rm re}(\langle a_j^2(t)\rangle_0)]\ , , \label{rqps}\end{aligned}\ ] ] where @xmath148 denotes real part , @xmath149 ( @xmath150 ) corresponds to @xmath17 ( @xmath18 ) and @xmath151 these ratios are then also independent of @xmath152 , and deviate from @xmath1 unless @xmath153 .
an initial coherent state is a pure separable gaussian state , which under the present hamiltonian will remain gaussian ( but no longer separable ) @xmath154 .
its entanglement entropy @xmath155 can then be evaluated through the gaussian state formalism @xcite and can be written as @xmath156\ln [ 1+f(t)]\,,\label{s}\ ] ] where @xmath157 denotes the reduced state of one of the modes and @xmath158 is the symplectic eigenvalue of the single mode covariance matrix : @xmath159 which is non - negative and the same for @xmath160 or @xmath161 if the state is pure and gaussian .
it represents the effective occupation number of the mode @xcite .
it is obviously also independent of @xmath152 , i.e. , the same for _ any _ coherent initial state .
the entanglement entropy ( [ s ] ) is just an increasing concave function of @xmath158 .
again , in the isotropic case @xmath33 , @xmath153 ( eq . ( [ iso ] ) ) , entailing no generated entanglement when starting from @xmath162 .
results for the previous quantities are depicted in figs .
[ f1][f2 ] for @xmath163 .
we concentrate on the first dynamically stable sector @xmath98 . for improved visualization of squeezing ,
we use there the quantities @xmath164 with squeezing in @xmath17 @xmath165 indicated by a _ negative _ value of @xmath166 ( @xmath167 ) . in fig .
[ f1 ] we consider the anisotropic case @xmath168 with @xmath5 close to @xmath4 , i.e. , to the first instability , such that @xmath169 , @xmath170 .
the evolution then exhibits large amplitude low frequency oscillations governed by @xmath105 , together with small amplitude high frequency oscillations governed by @xmath104 .
the picture clearly shows that in this regime squeezing and entanglement oscillate _ in phase _ : maximum entanglement occurs at times @xmath171 , @xmath172 odd ( see eqs .
( [ ap1])([ap2 ] ) ) , simultaneously with maximum squeezing in the operators @xmath173 and @xmath174 , and maximum average boson number in the oscillators ( or maximum covariance ( [ v3 ] ) in the case of an initially coherent state ) .
these results can be approximately described by conserving just the main terms in @xmath60 and @xmath59 ( eqs .
( [ sol0])([solf ] ) ) for small @xmath105 , which are those proportional to @xmath175 .
we obtain @xmath176 \frac{\sin^2\omega_- t}{\omega_-^2 } } \,,\label{ap1}\\ \langle a_j^2(t)\rangle_0&\approx&(-1)^j { \textstyle\frac{\omega^2(\omega_1 ^ 2-\omega_2 ^ 2 ) } { 16\omega_1\omega_2 } [ \frac{2\omega_1\omega_2(1-\delta_j)}{\delta}-(1-\frac{(\omega_1-\omega_2)^2}{2\delta } ) ( 1-\frac{(\omega_1+\omega_2)^2}{2\delta})]\frac{\sin^2\omega_- t}{\omega_-^2 } } \ , , \nonumber\\&&\label{ap2}\end{aligned}\ ] ] where @xmath177 is given in ( [ delj ] ) . hence ,
both @xmath178 and @xmath179 become proportional to @xmath180 , entailing that all quantities plotted in fig .
[ f1 ] will be governed by this term , thus oscillating in phase .
the presence of a factor @xmath181 in ( [ ap1 ] ) also entails @xmath182 , i.e. , the boson number will be larger in the mode with the smallest frequency , as verified in the bottom panel of fig .
[ f1 ] . moreover , for @xmath5 below but close to @xmath4 , the bracket in ( [ ap2 ] ) will be positive for @xmath15 and larger for @xmath161 ( since @xmath183 ) , leading to @xmath184 and @xmath185 , with latexmath:[$|\langle a_1 ^ 2(t)\rangle_0|\leq squeezing will occur in @xmath173 and @xmath174 , as seen in fig .
[ f1 ] , being more pronounced in @xmath174 .
( [ ap1])([ap2 ] ) also show that squeezing and entanglement are driven by the anisotropy , i.e. , they vanish for @xmath33 and at fixed @xmath5 become larger as @xmath187 increases ( or the ratio @xmath188 decreases ) .
they also increase , of course , as @xmath5 approaches @xmath4 , i.e. , as the first instability is reached ( @xmath112 for @xmath108 ) .
we also mention that for @xmath5 above but close to @xmath3 ( i.e. , in the second stable region but close to instability ) , the behavior is similar although squeezing will occur for @xmath189 and @xmath190 , since the bracket in ( [ ap2 ] ) becomes negative ( with @xmath191 ) , implying @xmath192 , @xmath193 .
in contrast , away from instability ( @xmath5 well below @xmath4 ) , quantum effects become much smaller even though they remain non - zero , as seen in fig.[f2 ] for @xmath194 ( where @xmath195 , @xmath196 ) . alternating squeezing in both @xmath18 and @xmath17
is now observed ( the behavior of @xmath197 and @xmath198 is analogous ) and the correspondence with the evolution of entanglement ( i.e. with @xmath158 ) is less direct , with the maxima of @xmath158 reflecting essentially the largest squeezing ( that of @xmath173 or @xmath174 ) .
nonetheless , the average boson numbers @xmath199 follow approximately @xmath158 .
we can easily understand these results by considering the expansion of the exact expressions ( [ v2])([aj2 ] ) for small @xmath5 ( @xmath200 $ ] ) .
we obtain , neglecting terms of order @xmath201 , @xmath202\label{aj2a2}\,,\end{aligned}\ ] ] which show that _ both _ @xmath203 and @xmath204 are of order @xmath205 , with @xmath206 at this order .
this implies that @xmath207 will be of order @xmath201 , so that eq .
( [ ft ] ) leads to @xmath208 up to @xmath209 .
hence , the @xmath203 and @xmath158 will be close for small @xmath5 .
besides , @xmath210 $ ] will change its sign as @xmath53 evolves , indicating that squeezing will alternate between @xmath17 and @xmath18 , being again larger for the oscillator with the lowest frequency due to the factor @xmath211 in ( [ aj2a2 ] ) . note that for small @xmath178 and @xmath212 , @xmath213 $ ] , which will not strictly follow @xmath203 , as @xmath179 is of the same order as @xmath214 but not proportional to it . finally , it is verified from ( [ nj2a])([aj2a2 ] ) that @xmath178 and @xmath179 are again proportional to @xmath133 and @xmath215 respectively , hence vanishing for @xmath33 and leading to a larger entanglement and squeezing as the anisotropy @xmath187 increases .
we finally mention that for short times @xmath53 such that @xmath216 , we obtain , after an expansion up to @xmath217 of the exact expressions ( [ v2])([aj2 ] ) , @xmath218 which are in agreement with the @xmath219 limits of eqs.([nj2a])([aj2a2 ] ) . hence , these quantities increase initially _ quadratically _ with time @xmath53 , with @xmath220 in this limit .
moreover , in this regime @xmath221 is real , and positive if @xmath222 .
this entails that squeezing will initially start in @xmath189 ( as seen in fig .
[ f2 ] ) and @xmath190 , with @xmath223
we have derived the exact analytical closed form solution for the field operators of two linear oscillators coupled through angular momentum .
we then applied the solution to investigate the relation between squeezing and entanglement generation in this model when starting from a separable coherent state . in the vicinity of instability
, the generated entanglement between the modes shows a large amplitude low frequency behavior ( almost periodic ) , which is reflected in a similar behavior of the squeezing in the coordinate of one of modes and the momentum of the other mode .
a different behavior occurs in the weak coupling regime , away from instability , where the generated entanglement is small and the squeezing is weak , exhibiting an essentially alternating behavior for the coordinate and momentum of each oscillator .
approximate analytical expressions describing these two regimes have also been derived from the general exact solution .
the present solution has of course potential applications for studies of other quantum statistical properties such as higher order squeezing , antibunching of photons and other nonclassical photon statistics .
the solution is also of interest for quantum information applications .
as stated in the introduction , the present model admits distinct physical realizations , so that results could in principle be tested in quite different scenarios ( optical simulations , particles in anisotropic harmonic traps , condensates , etc . ) .
we remark , finally , that expressions similar to ( [ m])([exp ] ) and ( [ v2 ] ) , ( [ aj2 ] ) remain formally valid for general systems of @xmath172 harmonic modes interacting through quadratic ( in @xmath224 , @xmath225 ) couplings , replacing @xmath43 by the corresponding @xmath226 matrix .
* acknowledgements : * _ we are thankful to the third world academy of sciences ( twas ) , trieste , italy and conicet of argentina , for financial support through twas - unesco fellowship program .
nc and rr acknowledge support from conicet and cic of argentina , while sm thanks the university grants commission , government of india , for support through the research project ( f.no.42-852/2013(sr ) ) . _
by means of the canonical transformation ( @xmath15 ) @xmath227 where @xmath228 , @xmath229 , we may rewrite ( [ 1 ] ) in the decoupled form ( we set here @xmath230 ) @xmath231 where @xmath232 , @xmath233 . here
@xmath234 , @xmath235 for @xmath15 in the fully stable region @xmath98 , whereas @xmath236 , @xmath237 , @xmath238 , @xmath239 in the second dynamically stable sector @xmath99 , with @xmath240 , @xmath239 in the unstable sector @xmath102 and @xmath241 at the borders @xmath108 or @xmath109 .
( [ nm1 ] ) then leads , in the dynamically stable regions @xmath98 or @xmath99 , to @xmath242 where @xmath243 ( real ) , @xmath244{\alpha_{^1_2}/\hbar\beta_{^1_2}}\,p_{^1_2}$ ] , @xmath245{\beta_{^1_2}/\hbar\alpha_{^1_2}}\,q_{^1_2}$ ] , and the minus sign in ( [ hd ] ) applies for @xmath99 .
( [ r1])([r2 ] ) are then obviously obtained for @xmath246 , @xmath247 .
the possible normal representations in the unstable regime are discussed in detail in @xcite .
999 louisell w h 1990 _ quantum statistical properties of radiation _ , wiley ( ny ) walls d f and milburn g j 1994 _ quantum optics _ , springer ( berlin ) mollow b r 1967 * 162 * 1256 estes l e , keil t h and narducci l m 1968 286 iafrate g j and croft m 1975 a * 12 * 1525 holzwarth g and chabay i 1972 _ j. chem .
phys . _ * 57 * 1632 belkin m a , shen y r and flytzanis c 2002 _ chem .
lett . _ * 363 * 479 cochrane p t , milburn g j and munro w j 2000 a * 62 * 062307 fan h , li c and jiang z 2004 a * 327 * 416 ng k m , lo c f 1997 a * 230 * 144 valatin j g 1956 _ proc .
_ london * 238 * 132 feldman a and kahn a h 1970 b * 1 * 4584 ring p and schuck p 1980 _ _ t__he nuclear many - body problem , springer ( ny ) ; + blaizot j p ripka g 1986 _ _ q__uantum theory of finite systems , mit press ( ma ) madhav a v and chakraborty t 1994 b * 49 * ( 1994 ) 8163 linn m , niemeyer m and fetter a l 2001 a * 64 * 023602 oktel m 2004 a * 69 * 023618 fetter a l 2007 a * 75 * 013620 aftalion a , blanc x and lerner n 2009 a * 79 * 011603(r ) aftalion a , blanc x and dalibard j 2005 a * 71 * 023611 bloch i , dalibard j and zwerger w 2008 885 fetter a l 2009 647 prina j , hradil z and juro b 1994 _ quantum optics and fundamentals of physics _ , kluwer , dordrecht rossignoli r and kowalski a m 2009 a * 79 * 062103 rebn l and rossignoli r 2011 a * 84 * 052320 rebn l ,
canosa n and rossignoli r 2014 a * 89 * 042312 schachenmayer j et al 2013 x * 3 * 031015 ; daley a j et al 2012 020505 ; bardarson j h , pollmann f and moore j e 2012 017202
nielsen m and chuang i l 2000 _ quantum computation and quantum information _ ( cambridge univ .
press , cambridge , uk ) srensen a et al 2001 _ nature _ * 409 * 63 ; orzel c et al 2001 _ science _ * 291 * 2386 ; bigelow n 2001 _ nature _ * 409 * 27 esteve j et al 2008 _ nature _ * 455 * 1216 ghne o and toth g 2009 _ phys . rep . _ * 474 * 1 ma j , wang x , sun c p , nori f 2011 _ phys .
rep . _ * 509 * 89 choi s and bigelow n p 2005 a * 72 * 033612 toth g , knapp c , ghne o and briegel h j , 2009 a * 79 * 042334 .
n et al 2010 a * 82 * 014101 ; chew l c and chung n n 2014 _ symmetry _ * 6 * 295 gross c 2012 103001 sen b and mandal s 2005 _ j. mod
. opt _ * 52 * 1789 ; sen b , mandal s and prina j 2007 1417 sen b et al 2013 a * 87 * 022325 audenaert k et al 2006 a * 66 * 042327 adesso g , serafini a and illuminati f 2004 a * 70 * , 022318 ; serafini a , adesso g and illuminati f 2005 a * 71 * 032349 braunstein s l and van loock p 2005 513 ; weedbrook c et al 2012 621 rossignoli r and kowalski a m 2005 a * 72 * 032101 | we investigate the exact dynamics of a system of two independent harmonic oscillators coupled through their angular momentum .
the exact analytic solution of the equations of motion for the field operators is derived , and the conditions for dynamical stability are obtained . as application , we examine the emergence of squeezing and mode entanglement for an arbitrary separable coherent initial state .
it is shown that close to instability , the system develops considerable entanglement , which is accompanied with simultaneous squeezing in the coordinate of one oscillator and the momentum of the other oscillator .
in contrast , for weak coupling away from instability , the generated entanglement is small , with weak alternating squeezing in the coordinate and momentum of each oscillator .
approximate expressions describing these regimes are also provided . |
the instabilities presented by a system are directly related with the possible phase transitions it can undertake . at subsaturation densities a liquid - gas phase transition in nuclear matter is predicted and it is normally tested in nuclear reactions .
the formation of highly excited composed nuclei in equilibrium with a gas of evaporated particles can be interpreted in the framework of hydrodynamics as two coexisting phases of nuclear matter , a liquid and a gas phase . during these reactions
, phase transitions may occur depending on the temperature and densities involved .
the liquid - gas phase transition also plays an important role in the description of the crust of compact star matter at densities between @xmath1 @xmath2 and saturation density ( @xmath3 @xmath2 ) .
it essentially consists of neutron rich nuclei immersed in a gas of neutrons .
it has been shown that this phase transition leads to an isospin distillation phenomenon : the isospin content of each phase is different , most of the gas being composed of neutrons and the liquid being closer to symmetric matter @xcite in the present paper , we employ the quark - meson coupling model ( qmc ) @xcite to investigate the thermodynamical instabilities of asymmetric nuclear matter . in the qmc model ,
nuclear matter is described as a system of non - overlapping mit bags which interact through the exchange of scalar and vector mean fields .
an earlier study of anm within this model has been focused on the effect of isospin asymmetry and temperature on the equation of state and on the coexistence surface @xcite .
we here consider an extension of the model that includes the scalar isovector virtual @xmath4 field @xcite .
its presence introduces in the isovector channel the structure of relativistic interactions , where a balance between a scalar ( attractive ) and a vector ( repulsive ) potential exists .
the @xmath0- and @xmath5-mesons give rise to the corresponding attractive and repulsive potentials in the isovector channel .
the introduction of the @xmath0-meson will affect both the behaviour of the system at low and high densities .
in the last case due to lorentz contraction , its contribution is reduced , leading to a harder equation of state ( eos ) at densities larger than @xmath6 @xcite . at low densities
a reduction of the symmetry energy will occur which will allow for more asymmetric matter .
in @xcite the instabilities in anm have been investigated within relativistic mean field hadron models , both with constant and density - dependent couplings at zero and finite temperatures .
it was shown that the main differences occur at large isospin asymmetry and at finite temperature .
in particular it has been shown that the predicted density at the inner edge of the crust of a compact star , from the crossing of the @xmath7-equilibrium equation of state ( eos ) , is model dependent @xcite . in the present work we investigate thermodynamical instabilities within the qmc model with and without the isovector - scalar @xmath0-meson .
although in this model the isoscalar vector channel described by the @xmath5 meson is included in a similar way to the non - linear walecka model ( nlwm ) , the non linearities in the @xmath8 and @xmath0-fields arise from the minimization of the bag energy .
in particular , the nlwm used in @xcite does not include non - linearities in the @xmath0-meson .
we may therefore expect a different behavior of asymmetric matter .
the paper is organized as follows : in section ii an extension of the qmc model to include the @xmath0-meson is discussed , in section iii we make a short review of the calculation of the spinodal surface , in section iv results are presented and discussed and some conclusions are drawn in the last section .
in what follows we present a review of the qmc model and its generalization to include the iso - vector - scalar @xmath0-meson . in the qmc model , the nucleon in nuclear medium
is assumed to be a static spherical mit bag in which quarks interact with the scalar ( @xmath8 , @xmath0 ) and vector ( @xmath9 , @xmath5 ) fields , and those are treated as classical fields in the mean field approximation ( mfa ) @xcite .
the quark field , @xmath10 , inside the bag then satisfies the equation of motion : @xmath11 \,\psi_{q_{i}}(x)=0\ , \quad q = u , d \label{eq - motion}\end{aligned}\ ] ] where @xmath12 is the current quark mass , and @xmath13 , @xmath14 , @xmath15 and @xmath16 denote the quark - meson coupling constants .
the normalized ground state for a quark in the bag is given by @xmath17 where @xmath18 with the normalization factor given by @xmath19 \big/ x_q^2 ~,\ ] ] where @xmath20 , @xmath21 , @xmath22 is the bag radius of nucleon @xmath23 and @xmath24 is the quark spinor .
the bag eigenvalue for nucleon @xmath23 , @xmath25 , is determined by the boundary condition at the bag surface @xmath26 the energy of a static bag describing nucleon @xmath23 consisting of three quarks in ground state is expressed as @xmath27 where @xmath28 is a parameter which accounts for zero - point motion of nucleon @xmath23 and @xmath29 is the bag constant .
the set of parameters used in the present work is given in ref .
the effective mass of a nucleon bag at rest is taken to be @xmath30 the equilibrium condition for the bag is obtained by minimizing the effective mass , @xmath31 with respect to the bag radius @xmath32 the total energy density of the nuclear matter reads @xmath33^{1/2}\end{aligned}\ ] ] and the free energy density is given by @xmath34 where the chemical potentials are given by @xmath35 @xmath36 the vector mean field @xmath37 and @xmath38 are determined through @xmath39 where @xmath40 and @xmath41 . finally , the mean fields @xmath42 and @xmath43 are fixed by @xmath44 in order to set the model parameters , we start by fixing the free space bag properties . they are obtained by fitting the nucleon mass and enforcing the stability condition for the bag in free space .
we consider two sets of free space parameters , taking an equal proton and neutron mass value in a first moment , and then proceeding by considering different proton and neutron masses after . in the first case
, we consider the bare nucleon mass @xmath45 mev and the bag radius , @xmath46 fm . the unknowns @xmath47 and @xmath48 mev are then obtained by setting the nucleon bag energies to that ( single ) bare nucleon mass value .
in the next step , we take the physical nucleon mass values as @xmath49 mev and @xmath50 mev and the bag radius for protons as @xmath51 fm .
the unknowns @xmath52 , @xmath53 , @xmath54 mev and the neutron radius @xmath55 are then obtained .
note that for fixed proton bag radius @xmath51 we observe a decrease on @xmath56 and on the bag parameter @xmath57 for the nucleons .
next , we fit the quark - meson coupling constants @xmath13 , @xmath14 , @xmath58 and @xmath59 for the nucleons , so as to obtain the correct saturation properties of nuclear matter , @xmath60 mev at @xmath61 @xmath2 , @xmath62 mev . for the couplings , we have @xmath63 , @xmath64 . in our first case
( when no effective mass difference between @xmath65 is considered ) , @xmath66 . [ cols="<,^,<,^,^,^,^,^,^,^ " , ] for large isospin asymmetries
the presence of @xmath0 reduces the instability region both in qmc and nl3 .
nl3 has a higher symmetry energy than nl3@xmath0 for densities ranging from 0 up to @xmath67 , and the same occurs for qmc although the differences are smaller .
this means that highly asymmetric matter is less bound in nl3 and qmc than in nl3@xmath0 and qmc@xmath0 .
a mere inspection of the symmetry energy is not enough to account for the differences in the instability region if different models are considered @xcite . with the introduction of the @xmath0-meson the scalar channel
is not affected , and therefore we are essentially changing the isovector channel .
we also notice that at @xmath68 the curvature of the qmc spinodal is intermediate between nl3 and tw .
as shown in @xcite this curvature is defined by the symmetry energy and its first and second derivatives .
the nuclear liquid - gas coexistence phase is characterized by different isospin contents for each phase , i.e. , the clusterized regions are more isospin symmetric than the surrounding nuclear gas , the so - called isospin distillation @xcite .
the extension of the distillation effect is model dependent and it has been shown that nl3 and other nlwm parametrizations lead to larger distillation effects than the density dependent hadron models @xcite . on the other hand ,
the distillation effect was also studied within a bhf calculation @xcite and a smaller distillation effect was generally obtained .
= 0.06 @xmath2 .
the bhf results were obtained by vidaa and polls @xcite .
, width=302 ] in fig .
[ fig6 ] we show the ratio of the proton versus the neutron density fluctuations corresponding to the unstable mode .
this ratio defines the direction of instability of the system .
we show the results for different proton fractions ( including rather small values ) , for the sake of studying the effectiveness of the models in restoring the symmetry in the liquid phase .
we first compare the three models not including the @xmath0-meson : qmc , nl3 and tw .
we see that qmc has a behaviour which is intermediate between nl3 and tw .
the distillation effect for densities above 0.02 @xmath2 is larger than the prediction of tw , but for the larger densities it also shows a tendency to decrease , contrary to nl3 .
the presence of @xmath0-meson makes the distillation effect more efficient . in fig .
[ fig7 ] we plot the proton - neutron density fluctuation ratio as a function of the isospin asymmetry for a fixed nuclear density , @xmath5=0.06 @xmath2 .
we compare all models under study , nl3 , nl3@xmath0 , tw , qmc and qmc@xmath0 , and include also the results derived from the bhf approach @xcite referred above .
all the relativistic models predict larger fluctuation ratios than the corresponding value of @xmath69 , dotted line .
the behaviour gives rise to a distillation effect , which , as referred before , is larger for nl3 and smaller for tw , with qmc presenting intermediate values .
the @xmath0-meson stresses the distillation effect , clearly seen both in qmc@xmath0 and nl3@xmath0 .
however , except for the very asymmetric matter ( @xmath70 ) , both qmc and the other relativistic models predict fluctuations with larger proton fractions than bhf .
the instability properties within the bfh calculation of @xcite show also differences for the spinodal surface : the unstable region is larger , extending to larger densities both in the isoscalar an isovector directions and the curvature of the spinodal at @xmath68 is much larger than the one of all the relativistic models considered . a larger extension of the unstable region is justified because the saturation density is larger , 0.182 @xmath2 .
the shape of the spinodal itself depends on the density dependence of the symmetry energy and its derivatives .
it would be important to identify the properties that define the shape of the spinodal for the more asymmetric matter .
we have studied subsaturation nuclear instabilities for both symmetric and asymmetric matte within the qmc model , with and without the inclusion of the @xmath0-meson . in this model
the nucleons are described as non - overlapping bags .
we propose a parametrization for qmc@xmath0 with a the symmetry energy slope value @xmath71 mev within the interval @xmath72 mev proposed in @xcite and which was determined from nuclear laboratory data .
it was interesting to notice that for qmc and qmc@xmath0 the quantity @xmath73 defined in @xcite , and which can be directly extracted from measurements of the isotopic dependence of the giant monopole resonances , falls inside the interval predicted by experiments .
the bhf calculation predicts a value of @xmath73 which is not very far , though lower than the interval obtained from gmr , @xmath74 mev .
a comparison was done with the results obtained from a nlwm parametrization ( nl3 ) , one density dependent relativistic model ( tw ) , along with bhf with the argonne v18 potential calculation .
it was shown that the restoration of isospin symmetry , obtained by a distillation effect , was more efficient in qmc with respect to tw but less efficient when compared with nl3 .
the spinodal surface within qmc is closer to tw although with a larger curvature at @xmath68 and a slightly smaller instability extension at larger asymmetries , while the inclusion of the @xmath0-meson ( qmc@xmath0 ) shrinks the asymmetric parts of the instability envelope .
the bhf results of @xcite , although with similar general properties , differ in the extension and shape of the spinodal and in the amount of distillation predicted .
a study of the qmc instability properties at finite temperatures is now under preparation .
it is also important to identify how the density dependence of the symmetry energy determines the shape of the spinodal for large isospin asymmetries .
this work was partially supported by feder and fct ( portugal ) under the projects poci / fp/81923/2007 , cern / fp/83505/2008 and sfrh / bpd/29057/2006 .
the authors acknowledge i. vidaa for kindly providing us with some data for comparison in this work .
k. saito and a.w .
thomas , phys . lett .
b * 327 * , 9 ( 1994 ) ; * 335 * , 17 ( 1994 ) ; * 363 * , 157 ( 1995 ) ; phys . rev .
c * 52 * , 2789 ( 1995 ) ; p.a.m .
guichon , k. saito , e. rodionov , and a.w .
thomas , nucl . phys . *
a601 * 349 ( 1996 ) ; k. saito , k. tsushima , and a.w .
thomas , nucl . phys . *
a609 * , 339 ( 1996 ) ; phys .
c * 55 * , 2637 ( 1997 ) ; phys .
b * 406 * , 287 ( 1997 ) ; p. k. panda , a. mishra , j. m. eisenberg , w. greiner , phys .
c * 56 * , 3134 ( 1997 ) ; p.k .
panda , d.p .
menezes , c. providncia .
c * 69 * , 025207 ( 2004 ) .
panda , g. krein , d.p .
menezes and c. providncia , phys .
rev c * 68 * , 015201 ( 2003 ) .
b. liu , v. greco , v. baran , m. colonna and m. di toro , phys .
c * 65 * , 045201 ( 2002 ) .
constana providncia , int .
j. mod phys .
e * 16 * , 2680 ( 2007 ) .
avancini , l.brito , d. p. menezes and c. providncia , phys . rev .
* c 70 * , 015203 ( 2004 ) .
s. s. avancini , l. brito , ph .
chomaz , d. p. menezes , c. providncia , _ phys .
* c74 * ( 2006 ) 024317 .
s. s. avancini et al . , phys .
c * 78 * , 015802 ( 2008 ) .
panda , d.p .
menezes , c. providncia .
phys . rev . * c69 * , 025207 ( 2004 ) . j. xu , l. w. chen , b. a. li and h. r. ma , phys .
014302 ( 2008 ) . j. xu , l. w. chen , b. a. li and h. r. ma , arxiv:0807.4477v1 [ nucl - th ] . v. baran , m. colonna , m. di toro , and a. b. larionov , nucl .
a632 * , 287 ( 1998 ) .
j. margueron and p. chomaz , phys .
c * 67 * , 041602(r ) ( 2003 ) ; p. chomaz , m. colonna and j. randrup , phys . rep . * 389 * , 263 ( 2004 ) h. mller and b.d .
serot , phys .
c * 52 * , 2072 ( 1995 ) .
g. a. lalazissis , j. knig and p. ring , phys .
c * 55 * , 540 ( 1997 ) .
s. typel and h. h. wolter , nucl . phys . *
a656 * , 331 ( 1999 ) .
a. santos , c. providncia and l. brito , unpublished .
i. vidaa and a. polls , _ phys . lett .
b _ * 666 * 232 ( 2008 ) . bao - an li , lie - wen chen and che ming ko , phys . rep . * 464 * , 113 ( 2008 ) . c. fuchs and h. h. wolter , eur .
j. a * 30 * , 5 ( 2006 ) .
w. zuo , a. lejeune , u. lombardo , j.f .
mathiot , nucl . phys . a 706 , 418 ( 2002);t .
frick , kh .
gad , h. muther , p. czerski , phys . rev .
c * 65 * , 034321 ( 2002 ) . c. ducoin , c. providncia , a. m. santos , l. brito and ph .
chomaz , phys .
c * 78 * , 055801 ( 2008 ) .
chomaz , nucl . phys . * a 681 * , 199c ( 2001 ) . | in the present work we include the isovector - scalar @xmath0-meson in the quark - meson coupling model ( qmc ) and study the properties of asymmetric nuclear within qmc without and with the @xmath0-meson .
recent constraints set by isospin diffusion on the slope parameter of the nuclear symmetry energy at saturation density are used to adjust the model parameters .
the thermodynamical spinodal surfaces are obtained and the instability region at subsaturation densities within qmc and qmc@xmath0 models are compared with mean - field relativistic models .
the distillation effect in the qmc model is discussed .
pacs number(s ) : 21.65.-f , 21.30.-x , 95.30.tg |
construction will begin soon on the american laser interferometer gravitational wave observatory ( ligo ) @xcite , and on its french / italian counterpart , virgo @xcite .
early in the next century there will likely be in operation a worldwide network of detectors , with sites in america , europe , and possibly japan and australia @xcite .
it is important at this stage for physicists to look ahead and identify the types of science that the community might focus on using this network when it reaches a mature stage , perhaps a decade after the first gravitational waves are detected .
one of the reasons for doing so is that some properties and parameters of the network , which ultimately will constrain what it can accomplish in the future , are being finalized today .
the orientation of the detector arms is one example .
one of the long term aims of this detector network will be to place upper limits on ( or perhaps detect ) the energy density of a stochastic background ( sb ) of gravitational waves . this background would be analogous to the relic @xmath8 electromagnetic background , except that its spectrum is not expected to be thermal .
the spectrum is usually characterized by a quantity @xmath9 which is the gravitational - wave energy density per unit logarithmic frequency , divided by the critical energy density @xmath10 to close the universe : @xmath11 some possible sources of a sb include : ( i ) random superposition of many weak signals from binary - star systems @xcite , ( ii ) decaying cosmic strings@xcite and first order phase transitions@xcite in the early universe , and ( iii ) parametric amplification of quantum mechanical zero - point fluctuations in the metric tensor during inflation @xcite .
see refs .
@xcite for reviews .
the predicted wave strengths from all of these stochastic sources are highly uncertain , reflecting our relative ignorance of the relevant physics and/or astrophysics .
hence , detecting or placing upper limits on the sb can bring us valuable information , particularly about the very early universe .
relic gravitational waves produced during inflation are particularly interesting , because , as grischuk has shown @xcite , the energy spectrum for these waves contains a unique imprint of the time evolution of the universe s scale factor @xmath12 .
we now discuss what is known about the magnitude of the contribution to @xmath9 from these waves , at frequencies relevant to ligo / virgo .
the predictions for @xmath13 from cosmological models are not very firm : they can vary between @xmath14 and @xmath15 or less
. however , observational upper bounds on @xmath9 in various frequency bands give interesting constraints on the models @xcite . in turn
these constraints can be used , within the context of particular inflation models , to place upper bounds on the contribution of relic gravitons to the value of @xmath9 in the ligo / virgo waveband , see , e.g. , ref .
the strongest such upper bound comes from matching the normalization of the scalar and tensor fluctuations produced during inflation to the recent cobe measurement of the microwave background anisotropy @xcite .
the result of this matching is somewhat discouraging : standard exponential inflation models predict that @xmath16 at the @xmath17 confidence level @xcite , far too small to be detected [ cf .
. @xmath18 below ] .
while it is far from certain that exponential inflation is correct , it seems unlikely that the expansion during the inflationary era was so much faster than exponential as to tilt the gravitational wave spectrum enough to give a detectable signal at high , ligo / virgo frequencies .
it is also possible , of course , that the observed microwave anisotropy was caused by physical processes of structure formation other than inflation . despite these pessimistic prospects for the detection of relic gravitons by ligo / virgo ,
it is certainly possible that there will be a detectable signal from other sources such as cosmic strings @xcite .
hence , it is important to determine how the detector arm orientations , which will not be changeable in the future , affect the sensitivity of the detector network to the sb .
to do so is the first of the two principal purposes of this paper .
this issue was first considered by michelson @xcite , and has been extensively discussed by christensen @xcite .
essentially we build on and extend slightly their analyses .
our conclusions are also slightly different from those of christensen .
the second principal purpose of this paper is to spell out the optimal data - processing procedure for searching for the sb with a network of detectors .
the algorithm for two detectors is implicit in michelson @xcite ( and is incorrectly treated in ref .
@xcite ) ; we give a more detailed description and a generalization to a network of detectors , taking into account the possible effect of correlated sources of noise .
we now turn to a description of our results and an explanation of how they relate to earlier work .
the effect of the sb on a gravitational - wave detector is essentially to produce a small contribution to the random , gaussian noise in its output . for one detector
this contribution will be swamped by the detector s own sources of noise , unless the sb strength is implausibly large ( @xmath19 ; see sec .
[ correlations ] ) . for two detectors which have no common sources of noise ,
however , the only contribution to the correlated fluctuations in the detector outputs will be the sb . by cross correlating the outputs of the detectors , the sb can in principle
be measured .
if one had identical , ligo type detectors at the same site , oriented in same way so that they respond in exactly the same way to the sb , and with levels of intrinsic noise corresponding to the `` advanced detectors '' of ref .
@xcite , then cross correlating would give a sensitivity to @xmath13 of the order of @xmath20 in the frequency band @xmath21 @xcite . for a pair of separated ,
non - aligned detectors , two new physical effects complicate the analysis @xcite .
first , if the detectors are not aligned the same way , they will respond to different polarization components of the sb .
orthogonal polarization components of the sb are expected to be statistically independent , and so the cross correlation will be reduced .
second , for each mode there will be a time lag between exciting the first detector and the second detector , and hence phase lags in the cross correlation . for the ligo detectors separated by @xmath22 and having maximum sensitivity at a frequency of @xmath23 , a typical phase lag will be of order unity
hence , there will be some destructive interference between parts of the cross - correlation that are due to modes which propagate in different directions .
thus we expect a reduction in the sensitivity of the detector pair to the sb . to analyze this reduction , it is necessary to ( i ) determine how to _ optimally process _ the output from the detectors , and ( ii ) find the signal to noise ratio ( snr ) that results from this method of filtering .
we also want to ( iii ) determine how the optimal snr depends on the detector orientations , and ( iv ) find those orientations that maximize the snr . steps ( i ) and ( ii ) were analyzed in refs .
@xcite . they found that , when optimal signal processing is used , the square of the signal to noise ratio for a broadband measurement of the sb is@xcite @xmath24 here @xmath25 is the duration of the measurement , and @xmath26 is the spectral noise density in either detector .
the key quantity appearing in this equation is the dimensionless function @xmath27 , which we call the _ overlap reduction function_. it characterizes the reduction in sensitivity to the sb of the detector pair at frequency @xmath28 that is due to their separation and non - optimal orientations , and its value is unity for coincident , aligned detectors . in refs .
@xcite a formula for the overlap reduction function was derived , which expresses it as an integral over all solid angles of the complex phase lag between the detectors , weighted by combinations of the detector beam pattern functions [ cf .
. @xmath29 below ] .
christensen@xcite numerically calculated this function for various detector configurations and discussed some of its properties .
however its dependence on the detector orientations was not apparent .
in this paper we derive an analytic formula for the overlap reduction function . using this formula we are able to carry through steps ( iii ) and ( iv ) outlined above .
we also determine how good are the choices that have been made for the orientations of the detectors in ligo , virgo and geo ( an as - yet - unfunded british / german collaboration ) ; i.e. , we determine how their sensitivity to the sb compares to the sensitivity they would have if they were optimally oriented .
our results are as follows .
call @xmath30 the angle between the bisector of the arms of the first detector and the arc of the great circle that joins the detectors , and similarly define @xmath31 for the second detector .
let @xmath32 and @xmath33 , so that @xmath34 describes the relative rotation of the detector pair , and @xmath35 describes their average rotation with respect to the line joining them .
then in sec .
[ optimization ] below we show that the optimum snr @xmath36 is given by @xmath37 where the quantities @xmath38 , @xmath39 , and @xmath40 are independent of @xmath34 and @xmath35 , and @xmath38 and @xmath40 are positive .
there are thus two possibilities for the optimum detector orientation , depending on the sign of @xmath39 : ( i ) @xmath41 , corresponding to each detector having an arm along the line joining them , and ( ii ) @xmath42 , corresponding to each detector arm being at an angle of @xmath1 ( mod @xmath2 ) to this line . in fig .
[ beta ] we plot the snr for both of these choices of orientation , as a function of the angle @xmath43 subtended between the detectors at the center of the earth .
the configuration ii is optimal for @xmath44 , while configuration i is optimal for detectors which are further apart .
[ beta ] shows that detectors which are close together are the most sensitive ; the sensitivity of ligo is roughly @xmath45 of what it would be if its detectors were coincident .
it also shows that detectors whose planes are roughly perpendicular ( @xmath46 ) have poor sensitivity , as we would expect .
we also show in sec .
[ optimization ] that the narrowband sensitivity of the detector pair near a given frequency @xmath28 , which is proportional to @xmath47 , is also always optimized at either configuration i or ii .
for example , at very low frequencies , @xmath47 becomes essentially the overlap of the polarization tensors of the two detectors [ cf .
( [ general_result ] ) below with @xmath48 , which is maximized in configuration i. this low frequency limit was previously derived by christensen @xcite .
motivated by this , he suggested that configuration i was always the best orientation to choose .
[ beta ] shows that , though this is not true for some values of the separation angle @xmath43 , the amount lost by choosing i rather than ii is never more than a few percent in snr .
we now discuss the orientations that have been chosen for the detector systems that are under construction or that have been proposed .
the relative rotation angle @xmath34 for a pair of detectors essentially determines whether the detectors respond to different polarization components , or to the same polarization component , of the gravitational wave field .
the advantage in responding to different components ( @xmath49 ) is that more information can be extracted from incoming burst signals . on the other hand ,
if the detectors respond to the same component ( @xmath50 ) then the detection signal to noise threshold is reduced , i.e. the fact that the same waveform is seen in both detectors means that one can be more confident that a candidate event is not due to detector noise .
these considerations guided the choices of the presently planned values of @xmath34 for the detector pairs ligo / ligo ( there will be two ligo detectors ) and virgo / geo , which are @xmath51 and @xmath52 respectively @xcite .
within the context of these constraints , a key issue that we wanted to understand was the following : given the above values of @xmath34 , how much does the broadband sensitivity of the detector pair depend on @xmath35 , i.e. by how much can the snr be reduced if @xmath35 is chosen arbitrarily instead of being optimally chosen . the answer we obtain [ secs .
[ gamma_sec ] and [ optimization ] below ] is that for relatively close detectors with @xmath53 , the dependence on @xmath35 is very weak ; but the dependence is strong for detectors on different continents .
hence for ligo s parallel detectors , the sensitivity will be close to optimal irrespective of the value of @xmath35 ; the present choice of @xmath54 implies that the snr for ligo is @xmath55 of the sensitivity at optimal orientation .
the virgo / geo orthogonal detector pair , however , will have a snr of less than @xmath56 times the optimal value , irrespective of the value of @xmath35 , if @xmath57 is chosen to be @xmath1 ( as is planned , so as to optimize the information obtainable from burst sources ) . finally , we estimate the @xmath58 upper confidence limit that can be placed on @xmath9 by the so called `` advanced detectors '' in ligo @xcite , in a broadband measurement using a one third of a year integration time , to be @xmath59 in the frequency band @xmath60 [ cf .
. @xmath61 below ] .
this is a little worse than earlier estimates which assume that the detectors are coincident and aligned @xcite .
the layout of this paper is as follows . in sec .
[ correlations ] we define the cross correlation matrix for a network of detectors and give the formula for the contribution to this quantity from the sb . in sec .
[ processing ] we describe the general optimal data - processing strategy , the justification of which is given in appendix [ stats ] . in this appendix we also derive the generalization of the signal to noise formula for optimal signal processing @xmath36 , to a network of more than two detectors , but more importantly to a network that has more than one interferometer per site : we show how to take into account the effect of correlated noise in detectors at the same site by introducing the concept of the _ effective spectral noise density _ of a detector site . in appendix [ gamma_calc ]
we derive the formula for the overlap reduction function , and we describe some of its properties in sec .
[ gamma_sec ] . next , in sec .
[ optimization ] , we show how to optimize the orientations of a pair of detectors , both for narrowband and for broadband measurements of the sb .
section [ implications ] describes the implications of our results for the ligo , virgo and geo detector facilities . finally in sec .
[ conclusion ] we summarize our main results .
we use units throughout in which the speed of light @xmath62 and newton s gravitational constant @xmath63 are unity .
the effect of a stochastic background on a detector network will be essentially to produce statistical correlations between the outputs of the various detectors . a key result which we will need
is an expression for these correlations in terms of the spectrum @xmath9 of the gravitational background .
this was first given by christensen@xcite , although it is implicit in the work of michelson@xcite .
we briefly describe the derivation in this section , and we lay the foundations for our analysis of appendix [ stats ] and sec . [ processing ] .
a detector network with @xmath64 detectors will have outputs @xmath65 for @xmath66 . here
@xmath67 is the strain amplitude that we read out from the @xmath68 detector ; it consists of an intrinsic detector noise @xmath69 , a contribution from the sb @xmath70 , and possibly a contribution @xmath71 from non - stochastic gravitational waves ( bursts and periodic waves ) .
the noise @xmath69 and the sb induced strain @xmath70 are independent random processes , which we assume to be gaussian and stationary .
the detector correlations can be described by the correlation matrix @xmath72 where @xmath73 means an ensemble average or a time average .
the fourier transform of the correlation matrix , multiplied by two , is the power spectral density matrix : @xmath74 this is a positive definite hermitean matrix which satisfies the equations @xmath75 and @xmath76 for any functions @xmath77 . here
tildes denote fourier transforms , according to the convention that @xmath78 since the random processes @xmath79 and @xmath80 are uncorrelated , the spectral density matrix of the detector outputs is just the sum of those for the detector noise and for the background : @xmath81 to derive an expression for @xmath82 , two key ingredients are needed .
the first is a mode expansion for the metric perturbation for an isotropic , stationary sb .
expressed in frequency space , this is @xmath83 for @xmath84 , and @xmath85 for @xmath86 . here
the tensors @xmath87 are the usual transverse traceless polarization tensors , normalized according to @xmath88 , and @xmath89 denotes the integral over solid angles parameterized by the unit vector @xmath90 . the coefficients @xmath91 are random processes which satisfy @xcite @xmath92 and @xmath93 for @xmath94 @xcite . here
@xmath95 is the delta function on the unit sphere .
the second ingredient is the expression for the response of the @xmath96th detector to the background .
this is @xmath97 where @xmath98 is the position of the detector , the @xmath99 denotes a double contraction , and @xmath100 is a symmetric tensor that characterizes the detector s orientation ( its polarization tensor ) . if the arms of the detector are in the directions of the unit vectors @xmath101 and @xmath102 , then @xmath103 @xcite . from eqs .
@xmath104 and @xmath105 we obtain that @xmath106 where @xmath107 are the detector beam pattern functions .
inserting this response function into an equation analogous to eq .
@xmath108 , and using eqs .
@xmath109 and @xmath110 , we obtain @xmath111 where @xcite @xmath112.\end{aligned}\ ] ] the functions @xmath113 are the overlap reduction functions discussed in sec .
[ intro ] .
it can be seen that their value is unity for coincident , aligned detectors .
below when considering a single pair of detectors we shall write @xmath114 simply as @xmath27 .
in this section we describe the optimal method for filtering the detector outputs when searching for the sb .
we discuss the two detector case in subsection [ two ] ; the filtering method in this subsection is implicit in the formulae of ref .
@xcite . a detailed proof that the method is optimal is given in appendix [ stats ] . in subsection [ cnoise ]
we discuss the effects of correlated sources of noise , and explain why correlation measurements between detectors at widely separated sites yield much better upper bounds on @xmath9 than correlation measurements between detectors at one site .
next we describe the modifications to the filtering method necessitated by correlated noise , in subsection [ extra ] . to measure the stochastic background one really needs to measure the spectral density matrix of the detector outputs @xmath115 .
now there is no way in principle to separate out the portions of @xmath116 due to detector noise @xmath117 , and due to the sb @xmath118 .
if we had only one detector , we could only conclude that @xmath119 . from eq .
@xmath120 and using @xmath121 , this would give an upper bound on @xmath13 of @xmath122 here @xmath123 , which is projected to be @xmath124 for ligo , even at an advanced stage @xcite .
the quantity @xmath125 is the hubble constant scaled to the value of @xmath126 .
it is unlikely that @xmath13 will be as large as the value in eq .
however , if it does happen that @xmath128 in the ligo / virgo waveband , then the sb induced noise may dominate over the other sources of detector noise at some frequencies , and may ultimately constrain the amount of information that we can extract from burst gravitational waves . in this paper we shall from now on assume that @xmath13 is small , and restrict attention to measurements made using two or more detectors . with two or more detectors ,
one takes advantage of the fact that the sources of noise in each detector will be independent
. this will be the case for detectors at widely separated sites , because sources of noise that are correlated between the detectors on timescales of the order of the light travel time between them are expected to be insignificant , or if not they can be monitored and compensated for @xcite .
when correlated noise is unimportant , then the off - diagonal elements of @xmath117 will be very small , so that @xmath129 by measuring these components we can gain information about the sb .
one does this by cross correlating the two output streams @xcite .
one takes each detector output @xmath130 , @xmath131 , and constructs using an optimizing linear filter @xmath132 the quantity @xmath133 .
the purpose of this filter is essentially to suppress the signal at those frequencies at which the detector noise is strong , and it is given by [ cf .
( [ hfilter ] ) below ] @xmath134 for coincident , aligned detectors the next step is simply to integrate @xmath135 against @xmath136 , see , e.g. , ref .
@xcite . for noncoincident detectors ,
however , a different strategy is necessary .
one first constructs the correlation with time delay @xmath137 , @xmath138 where @xmath139 is the observation time , typically of the order of a year .
then one calculates the weighted average @xmath140 where @xmath141 is the light travel time between the detectors , and @xmath142 is a weighting function which must be carefully chosen for each detector pair in order to maximize the sensitivity . roughly speaking
, this smearing of the cross - correlation compensates in some measure for the phase lags between the detectors which were discussed in sec . [ intro ] .
the quantity @xmath143 will then have a truncated gaussian distribution ( i.e. gaussian but restricted to positive values ) with signal to noise ratio given by eq .
@xmath36 @xcite .
in fact the sliding delay function @xmath142 is just the fourier transform of the overlap reduction function [ see eqs .
( [ omega_est ] ) and ( [ hfilter ] ) ] . in appendix [ gamma_calc ]
we give an analytic formula for @xmath142 , and we show in fig .
[ delay ] the sliding delay function that will need to be used for the ligo pair of detectors . in order of magnitude
, the @xmath58 confidence upper limit that can be placed on @xmath13 by cross - correlating between the detectors is @xcite @xmath144 here @xmath145 is the upper bound ( [ est ] ) obtainable from one detector , and @xmath146 is the bandwidth of the measurement . [ if no bandpass filtering of the data is carried out , @xmath146 will be roughly the width of the peak of the function @xmath147 ; and if filtering with a bandwidth @xmath146 is used , then the domain of integration in eq .
( [ snr ] ) must be suitably restricted ] . up to this point
we have assumed that the intrinsic detector noise is uncorrelated between different detectors , i.e. that the matrix @xmath148 is diagonal .
we now relax this assumption and consider the effects of correlated noise .
it is planned for ligo / virgo to ultimately have two or three detectors per detector site , perhaps optimized for different types of gravitational wave sources .
thus , there will be two possible types of correlation measurements : _ intrasite _ measurements at one site , and _
measurements between detectors at different sites .
as previously mentioned , it seems very unlikely that there will be any significant correlated noise in intersite measurements , and so it will only be important for intrasite measurements .
we now consider what information can be extracted from intrasite correlations .
the quantity we can measure is the sum @xmath149 , the two terms of which are in principle indistinguishable @xcite .
hence , we can draw inferences about @xmath82 _ only _ if we have some information about the correlated noise .
the most obvious such a priori information is the fact that @xmath150 is positive hermitean matrix .
it follows that if we measure the total power spectral density matrix to be @xmath151 , then @xmath152 .
however , this only tells us that @xmath153 , [ cf .
( [ est ] ) above ] , since the inequality applies to the total matrix , and the off - diagonal elements are small compared to the diagonal elements . in particular , it is not true without further assumptions that @xmath154 ( as is implicitly assumed in ref .
@xcite ) , since @xmath155 may be negative ( or complex ) .
a complex value of @xmath155 corresponds physically to sources of noise that excite two detectors with a certain preferred phase lag between them . in appendix
[ stats ] , we derive the @xmath58 confidence upper limit that one can place on @xmath13 in a bandwidth @xmath146 using intrasite correlations , which generalizes eq .
( [ win_factor ] ) .
if we define the noise correlation coefficients by @xmath156 then the result depends on ( i ) the measured value @xmath157 of @xmath158 ( determined from @xmath159 ) , and ( ii ) the assumed a priori maximum value @xmath160 of @xmath161 . in order of magnitude
we find [ cf . eqs .
( [ beta2 ] ) and ( [ pc2 ] ) below ] @xmath162.\ ] ] since there will be various unknown sources of weak correlated noise , it is not appropriate to choose a very small value of @xmath163 .
hence , the upper bound ( [ win_factor1 ] ) will be much worse than the bound ( [ win_factor ] ) obtained from intersite correlations .
we also consider the assumption that the correlated detector noise always excite different detectors in phase , but can be arbitrarily large in magnitude , so that @xmath164 in this case intrasite correlation measurements can not be used to _ detect _ the sb , but can only be used to place upper bounds on its magnitude . the resulting upper limit [ cf .
( [ pc1 ] ) ] is given by eq .
( [ win_factor1 ] ) with @xmath165 .
thus , the intersite correlations will still give better bounds unless the actual amount of correlated noise in the detectors satisfies @xmath166 which is @xmath167 for a year - long measurement with a bandwidth of @xmath168 .
since it is not clear that either of the conditions ( [ your_kidding ] ) or ( [ condt ] ) will be appropriate , we henceforth consider only intersite correlations . in appendix [ stats ]
we show how to optimally filter the intersite data when there is a detector network with several interferometers per detector site . in this case
the detector noise matrix @xmath117 will have a block diagonal form , with each block corresponding to a site .
let @xmath169 be the subblock corresponding to the @xmath38th site .
the strategy is essentially to measure the off - diagonal blocks of @xmath115 ( i.e. the intersite correlations ) , and to use these to obtain information about the sb .
the resulting snr squared is then given [ cf .
( [ snr3 ] ) below ] by a simple generalization of eq .
@xmath36 : @xmath170 here the sum is over the pairs of sites , and @xmath171 is the overlap reduction function for any detector at site @xmath38 together with any detector at site @xmath39 .
the quantities @xmath172 , which we call the effective site spectral noise densities for correlations between sites , are given by @xmath173^{-1}.\ ] ] they will be real since the matrices @xmath169 are hermitean . in the case where the off - diagonal elements of @xmath169 are all equal to @xmath174 , and the common value of the diagonal elements is @xmath26 , then we obtain that @xmath175 where @xmath64 is the number of detectors .
this formula shows that the use of @xmath64 detectors instead of a single detector at one site reduces that sites effective noise density by a factor of @xmath176 , _ if _ the detectors noise sources are independent ( @xmath177 ) .
if the noise sources are strongly correlated , however , so that @xmath178 , then there is no significant reduction in the effective noise .
in the preceding sections we have seen that for a pair of interferometric detectors , the overlap reduction function @xmath27 characterizes the dependence on the detector separation and orientations of both the correlation matrix @xmath120 , and the broadband sensitivity to the sb @xmath36 .
we now give an analytic formula for this function and discuss some of its properties .
we first define the variables which describe the orientations and separation of a pair of detectors .
let @xmath179 be the line joining the two detectors and @xmath180 ( @xmath181 ) be the plane formed by the arms of the first ( second ) detector ; see fig .
[ angles ] .
the variables we will use are : ( i ) the distance @xmath182 between the detectors ; ( ii ) the acute angle @xmath183 between @xmath179 and @xmath180 ; ( iii ) the angle @xmath30 between the projection of @xmath179 onto @xmath180 and the bisector of the two arms of the first detector ; ( iv ) corresponding angles @xmath184 and @xmath31 ; ( v ) the angle @xmath185 between @xmath179 and and the intersection of @xmath180 and @xmath181 .
the directions ( clockwise or anticlockwise ) in which @xmath30 and @xmath31 are chosen to be positive are unimportant , as long as the conventions for @xmath30 and for @xmath31 coincide as @xmath186 .
let @xmath187 and @xmath188 . in sec .
[ intro ] we have called these the _ relative rotation _ and _ total rotation _
angles respectively , since for @xmath189 , the angle @xmath34 is half of the relative rotation of the detectors while @xmath35 measures the average rotation with respect to the line joining them @xcite . in general
the overlap reduction function as given in eq .
@xmath29 will depend on all of the variables @xmath183 , @xmath184 , @xmath34 , @xmath35 , @xmath185 , and on the phase lag @xmath190 between the detectors , where @xmath28 is the frequency .
now for terrestrial detectors @xmath191 and @xmath192 ( @xmath193 say ) , which is also the angle subtended at the center of the earth by the detector pair . for this case
we derive in appendix [ gamma_calc ] the following formula : @xmath194 where the functions @xmath195 and @xmath196 are @xmath197 and @xmath198.\end{aligned}\ ] ] the functions @xmath199 are given in eqs .
( [ g1 ] ) , ( [ g2 ] ) and ( [ g3 ] ) of the appendix , and are all linear combinations of the functions @xmath200 and @xmath201 where @xmath202 .
several properties of the overlap reduction function @xmath27 , which were discovered by christensen and others from numerical calculations@xcite , can be read off the formula @xmath203 .
first , there will always be frequencies @xmath28 for which @xmath27 vanishes , and correspondingly near which the narrow - band sensitivity of the detector pair to the sb is very poor . for detectors that are less than a few thousand kilometers apart ,
the first null frequency is at @xmath204 , irrespective of the detector orientations ( see below ) .
second , the reduction function falls off like @xmath205 when @xmath206 , or equivalently when @xmath207 .
hence the 90% confidence limit that we can place on @xmath9 scales like @xmath208 for large @xmath182 .
however for detectors which are @xmath22 apart , the phase lag @xmath209 is of order unity for typical detector frequencies , and hence the variation of the sensitivity with distance is more complex than simply scaling like @xmath210 ( see sec .
[ implications ] below and also fig . [ beta ] above ) . to understand the behavior of @xmath27 as a function of @xmath34 , @xmath35 and @xmath43 , it is useful to consider some limiting cases . for coplanar , coincident detectors , @xmath211 , and using @xmath212 and @xmath213
we find that @xmath214 .
this is just what we would expect physically .
consider , for instance , detectors which are rotated with respect to one another by @xmath1 .
vertically incident gravitational waves will couple to the two detectors via polarization components that are orthogonal and hence statistically independent ; thus these waves give no contribution to the cross - correlation .
waves which are incident from non - vertical directions do give rise to correlations between the detectors , but when we average over all incident directions the total correlation vanishes @xcite .
it is also clear from rotational symmetry that there should be no dependence on the total rotation angle @xmath35 .
now suppose that the detectors are still coincident but no longer coplanar , so that @xmath215 .
equivalently , suppose the detectors are separated by a distance @xmath182 but evaluate @xmath27 at frequencies @xmath216 .
the result is @xmath217 this equation agrees with eq .
( 3.9 ) of ref .
@xcite after making an appropriate change of variables .
as we would expect , the dependence on @xmath35 is small when @xmath43 is small ; it is smaller than the dependence on the relative rotation angle @xmath34 by a factor of @xmath218 ( @xmath219 for the planned ligo detectors ; see below ) .
a third limiting case is when @xmath220 but @xmath221 , so that the detectors are coplanar but effectively separated . in this case ,
@xmath222 plots of the functions @xmath223 and @xmath224 are shown in fig .
[ planar ] .
we see that for @xmath225 , the @xmath226 term dominates , just as above for coincident detectors for small @xmath43 . when the phase lag @xmath209 is large , the functions in eq .
@xmath227 can be approximated as @xmath228 and @xmath229 , so we obtain @xmath230 what is happening physically here is that the cross - correlation is dominated by modes whose propagation vectors are nearly parallel to the line joining the detectors .
this can be seen by applying a stationary phase argument to the integral @xmath29 .
hence it is reasonable that the overlap reduction function should have separate factors for each detector that depend on how they are oriented with respect to the line joining the detectors . in this case
the dependence of the overlap reduction function @xmath231 on the total rotation angle @xmath35 is strong . in summary ,
the dependence on the relative rotation angle @xmath34 will dominate over the dependence on the total rotation angle @xmath35 unless two conditions are satisfied : ( i ) @xmath232 , and ( ii ) @xmath233 , i.e. @xmath234 , where @xmath235 is a typical frequency at which the detector is sensitive . in reality of course @xmath43 and @xmath182 are not independent ; they are related by @xmath236 where @xmath237 is the radius of the earth .
due to the coincidence that the quantity @xmath238 is of order unity , conditions ( i ) and ( ii ) above become satisfied at approximately the same value of @xmath43 . for a pair of terrestrial detectors , as @xmath43 and @xmath182 are increased from zero , the following three effects occur : ( i ) @xmath27 decreases because @xmath182 is increasing , as discussed above ( ii ) @xmath27 decreases because @xmath43 is increasing . from eq .
@xmath239 we see that the maximum over all orientations of the value of @xmath240 is @xmath241 .
hence the effect of increasing @xmath43 contributes typically a factor of at most @xmath242 to the decrease in @xmath27 .
( iii ) the dependence of @xmath27 on @xmath35 increases to become as strong as the dependence on @xmath34 .
we now turn to the issue of how to _ optimize _ the orientations of a single pair of detectors so as to detect or measure the sb .
we assume that the detector s noise curves are the same , so that @xmath243 .
we consider separately the cases of maximizing the narrowband sensitivity to the sb in the vicinity of some frequency @xmath28 , and of maximizing the overall broadband sensitivity . now
if we rotate a detector through @xmath2 , then its polarization tensor @xmath244 [ cf .
@xmath105 ] will have its sign flipped .
hence @xmath47 is a periodic function of both of the angles @xmath30 , @xmath30 ( or @xmath34 , @xmath35 ) with period @xmath2 . in the following discussion , all values of these angles and all equations involving them
should be treated modulo @xmath2 . from eq .
@xmath36 we see that the @xmath58 confidence limit that we can place on @xmath9 in a small interval of frequency of width @xmath146 centered about @xmath28 , using a measurement of duration @xmath246 , is @xcite @xmath247 to minimize this we need to maximize the quantity @xmath47 . from eq .
@xmath203 it follows that the maximum value over all orientations of @xmath47 is latexmath:[$|\theta_1(f)| + now if the signs of @xmath195 and @xmath196 are different , then this maximum will be achieved at @xmath249 ( i.e. either @xmath250 , @xmath251 or @xmath252 , @xmath253 ) .
this corresponds to each detector having one arm parallel to the line which joins them .
we will call this type of orientation _
configuration i_. this will be the optimal orientation at zero frequency , since @xmath254 , and @xmath255 .
it will also be optimal at high frequencies ( @xmath256 ) , since from from eqs .
@xmath203 and ( [ g1 ] ) ( [ fdef ] ) we see that @xmath257 for @xmath258 .
if , on the other hand , the signs of @xmath195 and @xmath196 are the same , then we see from eq .
@xmath203 that @xmath47 will be maximized at @xmath259 ( i.e. either @xmath260 or @xmath261 ) . in this case , which we will call _ configuration ii _
, each detector arm makes an angle of @xmath1 ( mod @xmath2 ) to the line joining the detectors .
this will be optimal , e.g. , for @xmath262 when @xmath263 ( the ligo value ; see below ) .
now suppose that the detector pair is in either configuration i or configuration ii .
then for some frequencies the orientations will be optimal , while for some other frequencies they will not be .
it is useful to calculate , for a given frequency @xmath28 , the factor @xmath264 by which @xmath47 is reduced if the wrong configuration out of i and ii is chosen .
note that choosing between i and ii is equivalent to fixing the relative rotation angle @xmath34 to be zero , and then choosing @xmath35 to be either @xmath265 or @xmath1 .
hence @xmath266 measures the sensitivity of @xmath27 to the total rotation angle @xmath35 when the detectors are parallel . from eq .
@xmath203 , we obtain @xmath267 at low frequencies ( @xmath268 ) this is approximately @xmath269 where @xmath270 .
thus for small values of @xmath43 we have @xmath271 , and whether configuration i or ii is chosen is relatively unimportant . at high frequencies ( @xmath258 ) , however , we find that @xmath264 asymptotes to a constant @xmath272 independent of frequency , where @xmath273 which can be quite small .
[ @xmath274 for the ligo detectors ; see below .
] hence the choice of @xmath35 will have a large effect on the high frequency sensitivity to the sb , especially for detectors which are close together for which @xmath275 is small .
it is quite likely that the sb will be so weak that it will only be detectable ( if at all ) by integrating over a broad range of frequencies , e.g. from @xmath276 to @xmath277 .
hence it is more important to optimize the overall broadband response @xmath36 of the pair of detectors than to optimize the narrow band sensitivity at some frequency .
we now determine how to do this if we assume that the spectrum @xmath9 is approximately constant over the relevant frequency range - the so - called `` zeldovich spectrum '' . by inserting eq .
@xmath203 into eq .
@xmath36 , we find that the snr squared after optimal filtering is @xmath278 where @xmath279 and @xmath38 and @xmath40 are given by similar formulae with @xmath280 replaced by @xmath281 and @xmath282 respectively .
note that @xmath38 , @xmath39 and @xmath40 depend on @xmath43 in two distinct ways , since from eq .
@xmath283 , @xmath284 the explicit formula @xmath285 makes it easy to determine how to optimally orient the detectors at a fixed value of @xmath43 , and also how the snr at the optimal orientation varies with @xmath43 .
the snr @xmath285 is maximized at @xmath286 , and the optimum orientations are just configurations i and ii again , applying when @xmath287 and @xmath288 respectively .
the planned ligo detectors in hanford , washington and livingston , louisiana will have @xmath289 , @xmath290 and @xmath54 ( the last two mod @xmath2 ) @xcite . a graph of @xmath27 for this configuration
is shown in fig .
[ ligo_gamma ] . in fig .
[ ligo_thetas ] we show what the functions @xmath195 and @xmath196 look like at the value @xmath291 appropriate for ligo . for any possible ligo orientations , @xmath27 will be a linear combination of these two functions as in eq .
@xmath203 . from fig .
[ ligo_thetas ] we can see that there will be a zero of @xmath27 at @xmath23 , whose position is relatively insensitive to the orientations chosen .
the noise power spectral density in the ligo detectors after some years of operation might be roughly @xmath292,\ ] ] where @xmath293 and @xmath294 . for frequencies less than @xmath295
the noise will be effectively infinite .
@xmath296 is a crude analytic fit to the noise curve estimate for the `` advanced detectors '' given by the ligo team in ref
. @xcite .
a more detailed model of the noise is unnecessary because of the uncertainty as to what the actual noise levels might be .
we now estimate what the effective site noise level @xmath297 [ cf .
. @xmath298 might be if - after an upgrade of the ligo facility - there are eventually two or three interferometers at each site , as is planned .
the noise spectral density @xmath26 in each detector will be a sum of the form@xcite @xmath299 including contributions due to seismic noise , thermal noise , photon shot noise and residual gas noise , among others .
by contrast , the typical value @xmath300 of the off - diagonal elements of the spectral density matrix will be approximately a sum of terms that are due to sources of noise which are strongly correlated between different interferometers at the same site : @xmath301 since @xmath302 when @xmath303 for the advanced detectors of ref .
@xcite , we find from eq .
( [ site_noise0 ] ) that to a good approximation for ligo , @xmath304 here the sum is over the different detectors , and @xmath305 corresponds to the primary , broadband interferometer with noise curve ( [ v_noise ] ) .
now the second and/or third detectors at each site will most likely be specialized ones such as dual - recycled detectors@xcite , which have high sensitivity only in some narrow frequency band .
thus in the relevant frequency band of @xmath60 [ cf .
[ lastplot ] below ] , it is likely that @xmath306 .
if , instead , three identical broadband detectors are operated at each site ( which is unlikely initially ) , then @xmath307 , and the @xmath58 confidence upper limit estimated below should be divided by @xmath308 . if we now make the plausible assumption that the sb has approximately constant @xmath9 over the above waveband , then by inserting eq .
@xmath296 into eq .
@xmath309 we can calculate the @xmath58 confidence upper limit that we can place on @xmath13 . using the planned ligo positions and orientations we thus obtain @xmath310^{-1/2 } \nonumber \\ & = & 5.1 \times 10^{-10 } \ , h_{75}^{-2 } \ ,
n \left ( { { \hat \tau } \over 10 ^ 7 \ , { \rm s } } \right)^{-1/2},\end{aligned}\ ] ] where @xmath125 is the hubble constant scaled to the value @xmath311 , @xmath25 is the observation time , the prefix @xmath312 indicates broadband , and @xmath313 .
note that this is the upper bound obtained from intersite cross - correlations , since as argued in sec .
[ cnoise ] , it will not help much to include the information from intrasite correlations .
this bound ( [ expected ] ) is a reasonably conservative estimate of what the ligo sensitivity might be when the advanced detectors are operating .
however the ultimate noise levels are largely unknown .
this is because the experimental techniques and technology , which are presently somewhat far from the advanced level , will be steadily improving . to obtain an upper limit to @xmath314 we take @xmath315 , which is a source of correlated noise that can not be changed once the beam tube is built .
using @xmath316 @xcite gives as a rather firm upper limit for the ligo sensitivity @xmath317 we now consider how @xmath318 would change if we were to vary the angles @xmath43 , @xmath34 and @xmath35 .
the integral @xmath319 is positive for @xmath289 with the noise power spectral density @xmath296 .
hence the optimal configuration of the detectors is configuration ii , i.e. at @xmath320 or @xmath1 .
however ligo already has @xmath321 ( corresponding to parallel detectors because @xmath322 ) .
also , as we have discussed in sec .
[ gamma_sec ] , the dependence of @xmath27 on @xmath35 is very weak for @xmath323 and at @xmath324 , see eq .
@xmath203 and fig .
[ ligo_thetas ] .
because of these two facts the expected ligo sensitivity at its planned configuration of @xmath290 and @xmath54 is only @xmath325 less than the optimal sensitivity which would be attained at @xmath261 .
this follows from eq .
@xmath285 where we find that @xmath326 and @xmath327 .
the european virgo and geo detectors , however , are currently planned to have @xmath328 .
such a configuration will allow virgo / geo to extract information from both polarization components of burst and periodic gravitational waves , but will severely reduce the sensitivity to the sb .
for these detectors @xmath329 , and we find from eq .
@xmath319 that @xmath330 and @xmath331 . in fig .
[ virgo - geo ] we plot the broadband signal to noise ratio @xmath285 as a function of @xmath34 ( assuming that @xmath35 is optimally chosen ) ; it can be seen that at @xmath250 , the sensitivity to the sb is reduced by about three orders of magnitude .
an intermediate choice of @xmath332 would give a broadband sensitivity to the sb comparable to that of ligo , while still allowing the detectors to access two polarization components of the gravitational - wave field that are largely independent .
the total rotation angle @xmath35 is unimportant for the broadband sensitivity to the sb in the case of detectors which are sufficiently close to each other [ but _ not _ unimportant for the narrow - band sensitivity , cf .
( [ loose ] ) above ] .
however , for detectors which are on separate continents , @xmath35 will become important .
for example in fig .
[ long_baseline ] we show the functions @xmath333 and @xmath334 for @xmath335 , the value appropriate for one of the two ligo / virgo cross correlations .
since @xmath196 is typically large compared to @xmath195 in this case , varying @xmath35 will have a large effect on the sensitivity .
this is confirmed by eq .
@xmath285 , where we find @xmath336 and @xmath337 .
next we consider how the broadband sensitivity to the sb varies with the angle @xmath43 , assuming that the orientations are optimally chosen .
recall [ cf .
[ optimization ] ] that this means either configuration i or ii is chosen . in fig . [ beta ]
we plot the snr @xmath285 as a function of @xmath43 , for both configurations i and ii , again assuming the noise curve @xmath296 . it can be seen that the sensitivity falls off rapidly when the detectors become more that a few thousand kilometers apart .
the sensitivity for the ligo separation is already a factor of @xmath6 worse than that for coincident detectors .
hence if there is ultimately a worldwide network of detectors in america , europe , japan and australia , only the cross - correlations between proximate detectors like ligo / ligo or virgo / geo will be important .
the various conclusions that we have drawn are relatively insensitive to the the detailed properties of the detector noise curve @xmath296 that we have assumed ( except for the value of the minimum frequency ) .
this is because the integral @xmath36 is dominated by contributions in the narrow frequency band @xmath338 ; see fig .
[ lastplot ] . at these frequencies thermal noise dominates over photon shot noise@xcite , so using sophisticated optical configurations of the interferometer such as dual recycling@xcite to reduce the shot noise will not help much .
it thus will be important for the detectors to have good low frequency sensitivity for the purposes of placing upper limits on the sb .
to place upper bounds on the strength of the gravitational stochastic background , or perhaps to detect it , is but one of the aims of ligo and other detector systems .
moreover , this background may well be very weak compared to waves from astrophysical sources ; consider by analogy the weakness of the electromagnetic stochastic background in the visible region of the spectrum . however , these waves if detected would be amongst the most interesting that the detectors would see .
the sensitivity to these waves is therefore just one of various factors that need to be taken into account when choosing the fixed orientations of interferometric detectors .
nevertheless , in order to be able to make a wise choice , it is important to know the effect for the sensitivity of choosing this or that orientation .
one of the main results of this paper is the simple expression ( [ general_result ] ) describing this dependence . in particular , for frequencies such that the phase lag between the detectors is @xmath339 , the sensitivity is determined solely by the overlap of the polarization tensors of the two detectors . at higher frequencies the direction of the vector which joins the detectors also becomes important .
we also determine the orientation of a pair of interferometers that will optimize the sensitivity of the pair to the sb , and show that the orientations which have been chosen for the ligo detectors are close to optimal .
the author is grateful to kip thorne for an introduction to the issue of detector orientations , for many helpful discussions and comments on the presentation , and also for the foresight of guessing , roughly , what the optimal filtering method and implications for ligo / virgo might be , which to a large extent motivated and guided this papers analysis .
thanks are also due to eric poisson for some helpful and detailed comments .
this work was supported in part by nsf grant phy-9213508 . in this appendix
we justify the method of optimal filtering described in sec . [ processing ] and
the resulting signal to noise ratio @xmath309 .
we do this by deriving the probability distribution for the spectrum @xmath13 given the detectors output @xmath340 , @xmath341 $ ] .
first we transform from a continuous to a discrete description of the measured data . in the body of this paper
we have treated the output of the detector network as a continuous vector random process @xmath342 . however , a real , discrete measurement will be of a finite duration @xmath139 , and will contain frequencies only up to some maximum frequency determined by the time resolution of the sampling .
in other words the output of the detectors will consist of the numbers @xmath343 for @xmath344 and @xmath345 , where @xmath346 is the number of detectors , @xmath347 is the starting time , and @xmath35 is the time resolution , of the order of @xmath348 .
the number of samples per interferometer is @xmath349 .
we denote by @xmath350 the vector of numbers @xmath351 , which we assume to have a multivariate gaussian distribution with some variance - covariance matrix @xmath352 .
this matrix is essentially the correlation matrix @xmath353 : @xmath354 however if we take a finite fourier transform , which amounts to making a change of basis in the space of vectors @xmath350 , then @xmath352 corresponds instead to the spectral density matrix @xmath355 .
now from eq .
@xmath356 , we have @xmath357 , where the contribution @xmath358 from the sb depends on the spectrum @xmath9 as in eq
. @xmath120 .
[ because of the finite frequency resolution of order @xmath359 , we represent the function @xmath9 as a finite vector @xmath360 . ]
what we want to do is to extract information about @xmath360 from a measurement of @xmath350 .
there are two general approaches to the task of quantifying the information obtained , in situations of this sort .
first , one can in principle compute the probability distribution function ( pdf ) for @xmath361 given the measurement @xmath350 , @xmath362 .
this pdf then contains complete information about the measurement .
however , the calculation of the pdf @xmath363 is frequently difficult in practice ; and so one has to resort to instead calculating estimators ( statistics ) @xmath364 which are functions of @xmath350 , chosen so that the pdf for their values given some value of @xmath361 , @xmath365 , is peaked near @xmath366
. there are standard criteria for choosing such estimators , see for example refs .
@xcite .
now suppose that , instead of one measurement of @xmath350 , we have @xmath367 measurements @xmath368 .
a standard statistical result is that in the cramer - frechet - rao limit of @xmath369 @xcite , the two approaches discussed above yield the same unique result .
more precisely , the pdf @xmath370 becomes a gaussian centered at @xmath371 , where @xmath372 is the so - called maximum likelihood estimator of @xmath361 .
the variance - covariance matrix @xmath373 of this probability distribution depends on the @xmath374 only through @xmath375 , @xmath376 .
conversely , the pdf @xmath377 of the estimator @xmath372 given some value of @xmath361 , becomes a gaussian centered at @xmath361 with width @xmath378 .
thus the two probability distributions , which are conceptually very different objects , become effectively the same , and all one really needs to calculate is the variance - covariance matrix @xmath379 .
however , this simplifying cramer - frechet - rao limit does not apply in a straightforward manner to the calculation of @xmath362 for measurements of the sb ; there are a number of subtleties and differences from the usual situations discussed in refs .
first , we have only one measurement of @xmath350 instead of a large number @xmath367 of measurements .
nevertheless , something like a `` large number of measurements limit '' does apply , which we discuss farther below .
second , one needs to address the issue of distinguishing between the contributions to @xmath352 from the detector noise and from the sb , since both are unknown a priori .
third , the cfr limit only applies to the extent that our _ a priori _ knowledge of the variables being measured is unimportant .
however , our _ a priori _ knowledge about the correlated detector noise in intrasite correlations _ is _ important in the calculation of @xmath380 , and consequently this pdf is not approximately gaussian [ cf .
( [ prob ] ) and ( [ pians ] ) below ] . to resolve these complications ,
we now derive from first principles an approximate expression for @xmath380 .
we also identify the conditions under which the approximations we make are valid , and show that they will all be satisfied in the ligo / virgo context .
we assume throughout that the effect of the sb on the detectors is small compared to the detectors intrinsic noise , so that from eq .
( [ est ] ) , @xmath381 .
we start by considering the _ frequency resolution _ of the measurement of @xmath9 . in a finite measurement of length @xmath139
there are roughly @xmath382 independent frequencies @xmath383 , and we get what amounts to one measurement of each @xmath384 . now
if @xmath139 is doubled , we double the number of measurements , but we also double the number of variables measured . clearly , the cramer - frechet - rao limit of repeated measurements of the same variables can not be attained in this context without further assumptions . below we shall argue that in the ligo / virgo context , to a good approximation , @xmath385 for some correlation time @xmath386 @xcite . in this case
the unknown quantities to be measured are @xmath387 for @xmath388 , or equivalently @xmath116 at @xmath367 different frequencies ; and the number of measurements of each of these variables is @xmath389 . in order to explain eq .
( [ assume ] ) , we fix a value of @xmath390 , and decompose @xmath391 into @xmath392 where @xmath393 is the step function .
let @xmath394 be the corresponding decomposition in frequency space , so that @xmath395 is @xmath115 averaged over frequency scales of the order of @xmath396 . then from eq .
( [ sn_sum ] ) there will be two contributions to @xmath397 : @xmath398 the first term here will be small if @xmath9 is smooth over frequency scales @xmath399 [ cf .
( [ correlation ] ) ] , which will be the case for currently conceived of sb sources when @xmath400 , for example . the second term in eq .
( [ twoterms ] ) will also be small for this value of @xmath390 , except near isolated frequencies corresponding to high - q resonances @xcite , the effect of which can be neglected .
we now calculate the pdf for @xmath361 using bayes rule ; by virtue of the condition ( [ assume ] ) , something like the cfr limit will apply . let @xmath401 , which is the maximum likelihood estimator of @xmath352 .
this is a @xmath402 matrix , where @xmath346 is the number of detectors .
similarly the matrices @xmath403 and @xmath358 are of dimension @xmath404 , and are given by formulae analogous to ( [ sigmap ] ) . however because of the condition ( [ assume ] ) the independent variables in @xmath405 can be formed into a smaller matrix @xmath406 , which is given by @xmath407_{ab}\ ] ] for @xmath408 .
note that since @xmath409 , the matrix @xmath410 contains only @xmath411 independent variables .
we define the matrices @xmath412 , @xmath413 and @xmath414 analogously in terms of @xmath415 , @xmath416 and the estimator @xmath417 which is defined for @xmath418 .
the estimator @xmath419 is the quantity that will be measured ; it is a sufficient statistic for @xmath361 , and moreover its discrete counterpart @xmath420 is the maximum likelihood estimator of @xmath410 .
the joint pdf for @xmath421 and @xmath413 given @xmath422 is @xmath423},\ ] ] where @xmath424 and @xmath425 and @xmath426 are the pdfs that represent our _ a priori _ knowledge . from the distribution ( [ e1 ] ) we obtain the pdf @xmath380 for @xmath360 given @xmath350 by integrating over @xmath421 : @xmath427 where @xmath428 is a normalization constant .
the above formulae involve primed , @xmath402 matrices .
we now express them in terms of the corresponding , unprimed , @xmath429 matrices , by using the fact that all the matrices will be approximately diagonal in frequency space .
that is , they will be approximately diagonal in the indices @xmath430 after a finite fourier transform change of basis , but not in the indices @xmath431 .
for example , on a suitable basis , @xmath432 where @xmath433 for @xmath434 .
it is straightforward to verify using eqs .
( [ assume ] ) and ( [ c_h_estimator ] ) that @xcite @xmath435 where @xmath436 is the effective number of measurements , and @xmath437 for example , the second term in @xmath438 can be written as @xmath439.\end{aligned}\ ] ] here we have used eq .
( [ p2 ] ) , switched into the time domain , used eq .
( [ c_h_estimator ] ) and switched back @xcite .
the result is just @xmath440 , and the second term in eq .
( [ likelihood ] ) transforms analogously .
now eqs .
( [ e1 ] ) , ( [ e2 ] ) , ( [ reduce_lambda ] ) and ( [ likelihood1 ] ) together yield a formal expression for @xmath441 . to proceed further we need to invoke some approximations .
the first of these is the quadratic approximation : we expand the likelihood function @xmath442 to give @xmath443 + { \cal o}(\delta { \bf \sigma}^3),\end{aligned}\ ] ] where @xmath444 .
clearly this approximation will be good only for certain values of @xmath421 and of @xmath361 ; we discuss below the implications of this restriction .
first we derive the conditions under which most of the probability of the pdf @xmath445\ ] ] will be concentrated in the region @xmath446 where the quadratic approximation ( [ quad ] ) is valid , so that the normalization of the pdf ( [ pdf2 ] ) can be calculated correctly using eq .
( [ quad ] ) .
if we let @xmath447 be the eigenvalues of @xmath448 , then up to an additive constant , @xmath449.\ ] ] from this formula it can be seen that the quadratic regime @xmath446 is given by @xmath450 where @xmath451 is some small number .
if we demand that the total probability in @xmath452 be @xmath453 for some @xmath454 , and assume that the a priori distribution @xmath425 is not sharply peaked , then we find the condition @xmath455 this condition will be just barely satisfied for , e.g , @xmath456 , @xmath457 , @xmath458 , and @xmath459 , which are values that are appropriate for ligo / virgo . now
if @xmath460 , then the intgeral over @xmath421 in eq .
( [ e2 ] ) will be dominated by contributions from the region @xmath446 , and so the approximation is valid . for @xmath461 ,
the exact probability @xmath462 is very small , and so the quadratic approximation will still give a qualitatively correct result .
in particular , the normalization of the pdf @xmath463 resulting from eq .
( [ quad ] ) will be approximately correct .
the next approximation involves consideration of the relative magnitudes of various components of the measured autocorrelation matrix @xmath414 [ cf .
( [ c_h_estimator ] ) above ] .
we introduce the following notation : for any matrix @xmath464 , @xmath465 denotes the matrix consisting of the diagonal subblocks ( in the indices @xmath431 ) of @xmath464 corresponding to intrasite correlations , and @xmath466 consists of the off - diagonal subblocks
. we also define @xmath467 to be the matrix of diagonal elements ( in the indices @xmath431 ) of @xmath464 and @xmath468 .
thus , the estimator @xmath414 decomposes into @xmath469 , where @xmath470 contains the detector noises @xmath471 , @xmath472 contains the measured intrasite correlations , and @xmath473 the measured intersite correlations .
now we expect the contribution of correlated detector noise to @xmath473 to be very small , cf . the discussion in sec .
[ extra ] above .
hence from eq .
( [ est ] ) , @xmath474 , where @xmath475 for example , if the sb is just barely detectable , then @xmath476 . we similarly define in order of magnitude the small parameter @xmath477 by @xmath478 which we expect to be of the order of @xmath479 or smaller if @xmath480 ; see eq .
( [ correl ] ) . in the expression ( [ quad ] ) for the likelihood function , to leading order in @xmath451 we can replace the factors of @xmath481 by @xmath482 , so that the cross terms between @xmath483 and @xmath484 vanish .
note that to this order , @xmath485 .
hence , the pdf ( [ e1 ] ) splits into a product of two factors which incorporate the measured intersite and the intrasite correlations .
the _ a priori _ pdf for the detector noise in eq .
( [ e1 ] ) can be written as @xmath486 since we expect @xmath487 .
using eqs .
( [ e1 ] ) , ( [ e2 ] ) , ( [ reduce_lambda ] ) , ( [ quad ] ) and ( [ e3 ] ) we obtain @xmath488 ^ 2 \right\ } } , \end{aligned}\ ] ] where @xmath489 is a normalization constant . here
^ 2 \right\}},\end{aligned}\ ] ] is a function representing the information from intrasite correlations , and @xmath491 . now if @xmath492 as estimated above , then most of the information we obtain from the sb will come from the intersite correlation measurements . in this case
the factor @xmath493 in eq .
( [ prob ] ) can be approximated to be constant [ see eq .
( [ pians ] ) below ] .
we discuss the implications of the resulting pdf for data - processing in subsection [ implic ] .
however it may turn out that @xmath494 , which from eqs .
( [ small_1 ] ) and ( [ small_2 ] ) will be the case if @xmath495 .
even if @xmath496 , it may happen that the correlation coefficient between different detectors for residual gas noise in the beam tube will be small compared to unity , so that in the notation of eq .
( [ correl ] ) , @xmath497 and again @xmath498 .
alternatively , it may be possible to detect and take account of bursts of outgassing from the beam tube walls .
more detailed discussions of the possible magnitude of intrasite correlated noise can be found in ref . @xcite .
because of the possibility that @xmath499 , we now derive an approximate expression for @xmath500 . to evaluate the expression ( [ intrasite ] )
we first make a change of variables .
let @xmath501 be the matrix of correlation coefficients @xmath502 and similarly define matrices @xmath503 and @xmath504 by replacing @xmath472 in eq .
( [ corr_def1 ] ) by @xmath505 and @xmath506 respectively .
also let @xmath507 we assume that the volume element in eq .
( [ intrasite ] ) can be written as @xmath508 to leading order in @xmath477 the cross term between @xmath509 and @xmath510 in the argument of the exponential in eq .
( [ intrasite ] ) vanishes , and so it becomes proportional to @xmath511.\ ] ] using the relations @xmath512 , @xmath513 , @xmath514 , and @xmath515 , and assuming that the pdf @xmath516 is slowly varying , we can approximately carry out the integral over @xmath503 . from eqs .
( [ intrasite ] ) , ( [ volume_el ] ) and ( [ arg1 ] ) , the result is @xmath517 ^ 2 \right\}},\end{aligned}\ ] ] where @xmath518 is a constant .
this expression is actually only a good approximation when @xmath519 , but the integral @xmath520 is dominated by values of @xmath521 close to @xmath522 which is @xmath523 .
now the behavior of the function ( [ intrasite2 ] ) depends strongly on our _ a priori _ information about the correlation matrix @xmath524 . as in eq .
( [ diagonalf ] ) , this matrix will be approximately diagonal on a frequency basis : @xmath525 where from eq .
( [ corr_def2 ] ) , @xmath526 since @xmath115 is positive definite , we have @xmath527 .
moreover , the variables @xmath528 will be real whenever @xmath529 for all @xmath530 and @xmath137 , and thus in particular they will be real if there are no sources of noise that preferentially excite one detector later than another one .
if we assume that such sources of noise are negligible , then by inserting into eq .
( [ intrasite2 ] ) the pdf @xmath531 we obtain an equation again of the form ( [ intrasite2 ] ) , but where @xmath521 and @xmath501 are now understood to be real .
[ a factor of @xmath532 $ ] is absorbed into @xmath518 . ]
next we assume that our _ a priori _ information about the correlation coefficients @xmath528 is of the form @xmath533 , so that we can take @xcite @xmath534 where @xmath393 is the step function .
using eqs .
( [ intrasite2 ] ) and ( [ assumed_pdf ] ) and the replacement @xmath535 [ cf .
( [ expand3 ] ) ] , gives @xmath536,\end{aligned}\ ] ] where @xmath537 if we define @xmath538 to be twice the fourier transform of the estimator @xmath539 as in eq .
@xmath355 , and put @xmath540 then we find from eq .
( [ est ] ) that @xmath541 where @xmath542 .
we now discuss the content of eqs .
( [ pians ] ) ( [ beta2 ] ) .
if we make no assumption about the amount of correlated noise present , and so take @xmath543 and @xmath544 , then we see using @xmath545 that @xmath500 is roughly constant for @xmath546 .
thus we obtain little information .
more information about the correlated noise needs to be input in order to constrain the strength of the sb .
one assumption which may be valid is that @xmath547 which can be enforced by setting @xmath548 in eq .
( [ assumed_pdf ] ) . equation ( [ intra_assume2 ] )
will hold if there are no sources of noise @xmath549 which contribute an amount @xmath550 to the output of one detector , and an amount @xmath551 to the output of another .
as example of such a source of noise could be a mode of vibration of a suspension system that couples in a suitable way the vibrations of mirrors in two different interferometers .
it may be a valid assumption that all such sources of noise will be negligible . in this case
we can use in eq .
( [ pians ] ) the values @xmath548 and @xmath552 . [
the value chosen for @xmath160 is unimportant as long as @xmath553 . then from eq .
( [ beta2 ] ) we see that @xmath554 as @xmath555 , and so to a good approximation , @xmath556.\ ] ] essentially this pdf gives an upper bound on each @xmath557 of @xmath558 , which is of the order of @xmath559 if @xmath480 .
as already mentioned , this upper bound will be much worse than that obtained from intersite correlations , unless the dimensionless correlation coefficient @xmath477 is @xmath560 .
moreover , the bound is only obtained by making the specific assumptions ( [ intra_assume1 ] ) and ( [ intra_assume2 ] ) about sources of correlated noise at one site . for these reasons ,
in the following subsection on data - analysis we consider only intersite correlations and take @xmath561 .
we now show in more detail that very little information about the sb is obtained if the assumptions ( [ intra_assume1 ] ) and ( [ intra_assume2 ] ) are dropped . as a simple model , we consider the opposite extreme of assuming equal probability for all relative phases in the contributions to the outputs of different detectors from correlated sources of noise .
thus , we take the _ a priori _ pdf to be of the form @xcite
@xmath562 d ( { \rm re } \ , { \cal c}_{abi } ) \ , d ( { \rm i m } \ , { \cal c}_{abi}),\end{aligned}\ ] ] where @xmath160 is the a priori maximum correlation .
then from eq .
( [ intrasite2 ] ) we obtain @xmath563.\ ] ] this function is of the form of eq .
( [ simpler ] ) below , where , from eq .
( [ beta2 ] ) , @xmath564 this is roughly the minimum detectable value of @xmath13 that can be detected in a bandwidth of @xmath399 .
thus , the function ( [ pc2 ] ) is qualitatively similar to the pdf ( [ simpler ] ) obtained from intersite correlations , with the simple change that the minimum detectable value of @xmath13 in any frequency band is increased by a factor of @xmath565 .
if we take @xmath160 to be of order unity and so make no assumption about the amount of correlated noise present , then the upper bounds on @xmath13 from intrasite correlations will be worse by a factor of @xmath566 than those obtained from intersite correlations . only if @xmath567 will the two be comparable .
however , because of the possibility of weak , unknown sources of noise , it is probably inappropriate to make such a strong assumption .
the distribution ( [ prob ] ) with @xmath568 is of the form @xmath569 where @xmath570 , because the matrices are all approximately diagonal in frequency . using eq .
@xmath120 , we find that the argument of the exponential in eq .
( [ prob ] ) becomes @xmath571 ^ 2 \right\},\ ] ] where @xmath572 .
hence we find , using @xmath573 , that @xmath574,\ ] ] and @xmath575.\ ] ] one would like to summarize the information contained in eq .
( [ simpler ] ) by calculating some kind of signal to noise ratio .
there are various , inequivalent ways of doing this .
for example , one could calculate probability distributions for the quantities @xmath576 or @xmath577 ; it is clear that the snr for @xmath578 would be worse than that for @xmath579 .
what we shall in fact do is calculate the probability distribution @xmath580 of @xmath581 assuming that @xmath582 , as this is the easiest to calculate .
now if we ignore the _ a priori _ information represented by @xmath426 in eq .
( [ simpler ] ) [ so that @xmath583 is a multivariate gaussian ] , then this is equivalent to calculating the pdf for the weighted average @xmath584 however when we include the information contained in @xmath426 , the main effect is to truncate and renormalize@xcite the pdfs for each @xmath585 , since all the @xmath581 s must be positive .
hence a simple interpretation of the pdf @xmath586 and the corresponding snr in terms of the average ( [ mean2 ] ) not possible .
nevertheless we suspect that @xmath586 approximately represents the probability distribution of some type of average of @xmath9 .
we now insert the assumption @xmath587 ( constant ) into eq .
( [ simpler ] ) , which gives a pdf of the form @xmath588.\ ] ] here the quantity @xmath589 is the statistic that should be calculated to estimate the value of @xmath581 .
we obtain @xmath590.\ ] ] this can be written in the time domain as @xmath591 where @xmath592 , and @xmath593 the functions @xmath594 are the sliding delay functions discussed in sec .
[ processing ] .
note that @xmath595 is constructed from the measured correlations in the following way : the intrasite correlations are used only to estimate the detector noise matrix @xmath596 , and then this noise matrix is used together with the measured inter - site correlations to estimate @xmath595 .
however , the matrix @xmath597 will typically be approximately diagonal , and so neglecting the intrasite correlations will give only a negligible error .
finally the intersite snr @xmath598 can be obtained from eq .
( [ cts ] ) @xcite .
if we assume that @xmath599 , so that all the intersite correlations we measure come from the sb , then the snr takes the form @xmath600.\end{aligned}\ ] ] using the fact that the matrix @xmath601 will be constant on each subblock corresponding to 2 detector sites , one can derive eqs .
@xmath309 and @xmath602 from eq .
( [ snr3 ] ) .
the @xmath96th detector is characterized by its position @xmath98 and by the tensor @xmath603 , where @xmath604 and @xmath605 are unit vectors in the direction of its arms . in terms of these quantities ,
the overlap reduction function is given , from eq .
@xmath29 , by @xmath606.\end{aligned}\ ] ] if we write @xmath607 where @xmath102 is a unit vector , and put @xmath608 ( in units in which @xmath609 ) , then we obtain @xmath610 where @xmath611 this integral can be evaluated by the standard method of writing down the most general possible answer : @xmath612 \nonumber \\ \mbox { } & + & c(\alpha ) \left [ \delta_{ij } m_k m_l + \delta_{kl } m_i m_j \right ] + d(\alpha ) m_i m_j m_k m_l \nonumber \\ \mbox { } & + & e(\alpha ) \left [ \delta_{ik } m_j m_l + \dots + \delta_{jl } m_i m_k \right].\end{aligned}\ ] ] one might expect to have to include a term proportional to @xmath613 where @xmath614 , but in fact this tensor is a linear combination of the five tensors included above .
contracting eq .
( [ gamma_ans ] ) with the five different tensorial expressions that appear on its right hand side yields a system of linear equations for @xmath615 which involves scalar integrals that are straightforward to evaluate . solving these equations and substituting the results back into eqs .
( [ trick ] ) and ( [ gamma_ans ] ) gives @xmath616 the functions @xmath617 are linear combinations of the spherical bessel functions : @xmath618 @xmath619 and @xmath620 the result ( [ general_result ] ) applies to any gravitational wave antennas , such as interferometers with non - perpendicular arms and arbitrary orientations , or resonant bar antennas .
the special case of two resonant bar antennas , for which each @xmath621 for some vector @xmath622 , has been previously derived by michelson @xcite . as can be seen from fig .
[ rho_fns ] , the first term in eq .
( [ general_result ] ) dominates for @xmath623 ( unless @xmath624 ) .
a similar simplification applies for @xmath258 : we have @xmath625 where @xmath626 denotes the tracefree part of the projection @xmath627 of @xmath628 perpendicular to @xmath102 .
the fact that @xmath27 does not depend on the components of @xmath100 and @xmath629 parallel to @xmath102 is due to the fact that the cross correlation at @xmath206 is dominated by modes whose wavevectors are nearly parallel to @xmath102 , as discussed in sec .
[ special_cases ] . to apply eq .
( [ general_result ] ) to terrestrial detectors , we need to choose a coordinate system and express the tensors @xmath630 , @xmath631 and @xmath102 in terms of the angles @xmath34 , @xmath35 and @xmath43 defined in sec . [ gamma_sec ] .
a convenient choice is to take the detectors to be located at @xmath632 and @xmath633 , where @xmath634 are spherical polar coordinates with origin at the earth s center , so that the unit vector in the direction joining the detectors is @xmath635 .
if we let @xmath636 , @xmath637 and @xmath638 be the usual basis of orthonormal vectors , and define @xmath639 then we can take @xmath640 and @xmath641 . inserting these expressions into the formula ( [ general_result ] ) yields , after some manipulation , the result @xmath203
, where the functions @xmath199 are @xmath642 @xmath643 @xmath644 and @xmath645 from eq .
@xmath203 it is straightforward to evaluate the fourier transform of the overlap reduction function , which gives the sliding delay function discussed in sec . [ processing ] , @xmath646 the function @xmath594 vanishes for @xmath647 . for @xmath648 , @xmath594 is given by eqs .
@xmath203 to @xmath649 , where the functions @xmath650 are now @xmath651 @xmath652 @xmath653 and @xmath654 . a. abrmovici , w.e .
althouse , r.w.p .
drever , y. g " ursel , s. kawamura , f.j .
raab , d. shoemaker , l. sievers , r.e .
spero , k.s .
thorne , r.e .
vogt , r. weiss , s.e .
whitcomb , and m.e .
zucker , science * 256 * , 325 ( 1992 ) .
the phrase _ signal to noise ratio _ usually means the following : from the measured data one can compute a number @xmath143 which has to a good approximation a gaussian statistical distribution , and snr is just the expected value of @xmath143 divided by the square root of its variance . in this paper
we frequently encounter quantities @xmath143 whose distributions are _ truncated _ gaussians : @xmath655 \theta(y),\ ] ] where @xmath393 is the step function , and the normalization factor @xmath64 depends on @xmath656 . in this case we still use snr to mean just @xmath656 .
the use of truncated gaussian distributions affects the constants appearing in eqs .
@xmath657 and @xmath61 ( where we assume @xmath658 ) . for relic gravitational waves from the big bang
, grischuk@xcite has shown that eq .
@xmath110 must be replaced by @xmath662 for @xmath663 , where @xmath664 is a complex function satisfying @xmath665 the corresponding term in eq .
this is because the relic background is phase - squeezed .
however , eq . @xmath110 remains an adequate approximation for our purposes , since @xmath664 is a rapidly oscillating function of @xmath28 with a period of order @xmath666 , the hubble constant @xcite .
observations of the sb by the methods discussed in this paper will have a frequency resolution @xmath667 , and consequently will `` average out '' the effects of the phase squeezing .
@xmath120 and @xmath29 are essentially equivalent to eqs .
( 4.5 ) and ( 4.6 ) of ref .
@xcite , and can be obtained from those by ( i ) taking a fourier transform , ( ii ) dividing by the square of the transfer function @xmath668 to convert from correlations of light intensities to correlations of strain amplitudes , ( iii ) replacing the cosine with a complex exponential , and ( iv ) correcting for the difference in normalization convention for @xmath27 between ref .
@xcite and this paper .
it is not quite true that the contributions @xmath148 and @xmath82 are indistinguishable .
the structure of the matrix @xmath82 at a fixed frequency is tightly constrained by eq .
@xmath120 , so the whole ( off - diagonal ) matrix can be determined from just one element .
moreover all the off - diagonal elements must be real .
if we measure the off - diagonal part of @xmath116 to be nonzero and it does not have the form ( [ correlation ] ) , then we know that we are seeing noise that is not due to the sb
. this fact will be helpful in identifying noise in inter - site correlations ( if there is any inter - site correlated noise ) .
it will not be as useful for identifying noise in intra - site correlations , because in this case the constraint is just that all the off - diagonal elements of @xmath669 be equal and positive , and this is likely to be satisfied by some sources of correlated noise .
using eq .
( [ general_result ] ) and the fact that @xmath671 , we see that for _ any _ symmetric traceless tensors @xmath672 and @xmath673 , the metric perturbation @xmath674 due to the sb satisfies @xmath675 this is a precise expression of the fact that orthogonal components of the strain tensor at a fixed point in space are statistically independent .
if the condition ( [ assume ] ) were not valid , one could still in principle integrate over @xmath676 to derive the reduced pdf for @xmath677 [ in the notation of eq .
( [ twoterms ] ) ] . however , in this case the cfr limit need not apply . | we analyze the sensitivity of a network of interferometer gravitational - wave detectors to the gravitational - wave stochastic background , and derive the dependence of this sensitivity on the orientations of the detector arms .
we build on and extend the recent work of christensen , but our conclusion for the optimal choice of orientations of a pair of detectors differs from his . for a pair of detectors ( such as ligo ) that subtends an angle at the center of the earth of @xmath0 , we find that the optimal configuration is for each detector to have its arms make an angle of @xmath1 ( modulo @xmath2 ) with the arc of the great circle that joins them . for detectors that are farther separated , each detector should instead have one arm aligned with this arc .
we show that the broadband sensitivity to the stochastic background of a detector pair which are @xmath3 apart is essentially determined by their relative rotation .
their average rotation with respect to the arc joining them is unimportant .
we also describe in detail the optimal data - analysis algorithm for searching for the stochastic background with a detector network , which is implicit in earlier work of michelson
. the ligo pair of detectors will be separated by @xmath4 .
the minimum detectable stochastic energy - density for these detectors with their currently planned orientations is @xmath5 greater than what it would be if the orientations were optimal , and @xmath6 times what it would be if their separation were @xmath7 a few kilometers . [
the detectors are chosen to be far apart so that their sources of noise will be uncorrelated , and in order to improve the angular resolution of the determinations of positions of burst sources . ]
= 10000 the sensitivity of the laser interferometer gravitational + wave observatory ( ligo ) to a stochastic background , + and its dependence on the detector orientations .
theoretical astrophysics , california institute of technology , pasadena , california 91125 |
in a recent paper @xcite , the @xmath1 distributions for dimuon production at the lhc were calculated in the framework of the randall - sundrum - like scenario with a _ small curvature _
( rssc model , in what follows ) .
the lhc discovery limits on 5-dimensional gravity scale @xmath4 were obtained for both 7 tev and 14 tev . in the present paper ,
the combined analysis of _ both dielectron and dimuon _ events at 7 , 8 , and 13 tev will be done . before presenting results of calculations for the drell - yan ( dy ) process at the lhc
, it is useful to compare the rssc model with the standard randall - sundrum model with two 3d branes ( rs1 model @xcite ) .
the classical action of this scenario is given by and @xmath9 are redefined . ]
@xmath10 where @xmath11 is the 5-dimensional metric , with @xmath12 ; @xmath13 ; @xmath14 is the 5-th dimension coordinate ; and @xmath15 is the size of the ed .
the quantities @xmath16 are induced metrics on the branes , and @xmath17 and @xmath18 are brane lagrangians .
it is assumed that the cosmological constant @xmath19 is negative .
thus , we have a slice of the ads@xmath20 space - time . in ref .
@xcite the background warped metric was found to be @xmath21 where @xmath22 is the minkowski tensor with the signature @xmath23 .
the periodicity @xmath24 is imposed and the points @xmath25 and @xmath26 are identified .
so , one gets the orbifold @xmath27 . in the rs1 model
the 3d branes are located at the fixed points @xmath28 ( plank brane ) and @xmath29 ( tev brane ) .
the sm fields are constrained to the tev brane , while the gravity propagates in all spatial dimensions .
the hierarchy relation between the 5-dimensional _ reduced _ gravity scale @xmath30 and _ reduced _ planck mass @xmath31 looks like @xcite , @xmath32 the reduced scales in are defined as follows @xmath33 @xmath34 the warp factor @xmath35 has the following values on the branes @xmath36 for @xmath37 , we get for boundary cosmological terms @xmath38 thus , the planck brane has a positive tension , while the tev brane has a negative tension . in order for the hierarchy relation to be satisfied , one has to put @xmath39 the rs1 model predicts a series of _ massive _ kaluza - klein ( kk ) graviton resonances with the lightest graviton about 1 tev . after the replacement @xmath40 the rs1 metric becomes @xcite @xmath41 the hierarchy relation is modified as follows @xmath42 the warp factor @xmath43 and brane cosmological terms acquire the meanings @xmath44 @xmath45 thus , after replacement the branes _ are interchanged_. the brane @xmath46 becomes the tev brane on which all the sm fields live .
the ratio of the warp factors remains _ the same _
( namely , @xmath47 ) , but the very value of the warp factor on the tev brane becomes equal to _
unity_. in such a scheme , @xmath30 can be one to tens tev , while the curvature @xmath0 is allowed to vary from hundred mev to tens gev @xcite .
the masses of the kk gravitons are proportional to the curvature @xmath0 ( see below eq . ) .
note that after changing variables in @xmath48 one obtains the metric @xcite @xmath49 which also leads to the modified hierarchy relation .
the metric with the small curvature , which takes into account the orbifold structure of the space - time , looks like ( for details , see appendix a ) @xmath50 with @xmath51 note that the metrics and coincide for @xmath52 .
the warp factor @xmath53 $ ] with the rssc function @xmath54 has the following values at the fixed points @xmath55 the boundary cosmological terms are @xmath56 the hierarchy relation , @xmath57 is the same as eq . .
in order for relation to be satisfied , it is enough to take @xmath58 .. in particular , for @xmath59 tev , @xmath60 gev ( 100 mev ) , one has @xmath61 ( 9.8 ) and @xmath62 fm ( 19.4 fm ) . ] in between the branes , the 5-dimensional scalar curvature is negative , @xmath63 while the radius of the curvature is equal to @xmath64 ( see appendix a ) .
let us use a linear expansion of the metric about its minkowski value ( see , for instance , @xcite ) @xmath65 after redefinition @xmath66 , where @xmath67 @xcite , and imposing transverse - traceless gauge , @xmath68 the metric fluctuation @xmath69 describes the massive spin-2 field with 5 degrees of freedom .
it can be decomposed into kk graviton excitations has only 2 degrees of freedom and describes the massless graviton @xcite . ]
@xmath70 the wave functions of the kk excitations obey the equation @xcite @xmath71 with @xmath72 being the mass of four - dimensional gravitons , @xmath73 equation follows from 4-dimensional components of the einstein - hilbert equation , if one keeps only linear terms in @xmath69 .
note that eq .
means that @xmath74 where @xmath75 is the 5-dimensional dalembertian in the background metric .
the eigenfunctions @xmath76 satisfy the boundary conditions @xmath77 as well as the orthonormality condition @xmath78 thus , we have the sturm - liouville problem , - . to solve it ,
let us put for @xmath79 @xmath80 where @xmath81 then we get from @xmath82 \left ( \frac{\sigma'}{\kappa } \right)^2 \nonumber \\ + & \bigg [ z_n ( \phi^{(n)}(z_n ) ) ' + 2 \phi^{(n)}(z_n ) \bigg ] \frac{\sigma''}{\kappa^2 } = 0 \;.\end{aligned}\ ] ] here and in what follows the _ prime _ denotes the derivative with respect to variable @xmath14 .
the solution of eq . which satisfies the _ right _ boundary condition is given in terms of bessel functions @xmath83 \;,\ ] ] where @xmath84 and @xmath85 is a constant .
the _ left _ boundary condition defines the masses of the kk gravitons with respect to the tev brane .
taking into account that @xmath86 , we get @xmath87 where @xmath88 let us demonstrate that the second term in is equal to zero .
indeed , by using relation @xmath89 , where @xmath90 or @xmath91 , it can be presented in the form @xmath92 [ \delta(y ) - \delta(y - \pi r_c ) ] \;.\ ] ] note that @xmath93 thus , the expression in vanishes due to the boundary condition . as a result
, we obtain for @xmath79 @xmath94 \;,\ ] ] where the normalization constant @xmath95 is defined from the orthonormality condition .
an explicit form of @xmath95 is derived in appendix b. as for the zero mode excitation , its wave function looks like @xmath96 the orthogonality of the zero mode and kk modes comes from the equation ( @xmath97 ) @xmath98 \nonumber \\ & = \frac{n_0 n_n}{a_n \kappa } [ j_1 ( a_n ) y_1(b_n ) - y_1 ( a_n ) j_1(b_n ) ] = 0 \;.\end{aligned}\ ] ] the interactions of massless gravitons on the tev brane are given by the lagrangian @xmath99 where @xmath100 is the energy - momentum tensor of the sm fields .
let us consider eq . in more detail .
since @xmath101 @xmath72 are defined by the equation @xmath102 as a result , the graviton masses have appeared to be proportional to @xmath0 , @xmath103 where @xmath104 are zeros of the bessel function @xmath105 . in the limit of a very small curvature ,
@xmath106 one can use asymptotic values of the bessel functions @xmath107 \ ; , \nonumber \\
y_1 ( z ) & = - \sqrt{\frac{2}{\pi z } } \left [ \cos \left ( z - \frac{\pi}{4 } \right ) + \mathrm{o } ( |z|^{-1 } ) \right ] \;.\end{aligned}\ ] ] then we get from @xmath108 that results in the well - known graviton spectrum in the model with one _ flat _
ed @xcite-@xcite @xmath109 as one can see , @xmath110 .
thus , using asymptotic expressions was fully justified .
let us stress , however , that the ads@xmath111 space becomes indistinguishable from a five - dimensional flat space only for _ negligible _ values of the curvature @xmath0 .
indeed , in the limit , eq .
transforms into the hierarchy relation for the flat ed @xmath112 where @xmath113 is the volume of the compact ed .
then the inequality @xmath114 means @xmath115 the newton potential between two test masses in the rssc model was estimated in @xcite @xmath116 where @xmath117 is the mass of the lightest kk graviton , @xmath118 being the first zero of the bessel function @xmath105 .
thus , relative corrections to the newton law appear to be negligible @xcite .
the interaction of the massive kk gravitons with the the sm fields on the tev brane is described by the lagrangian @xcite-@xcite ( see also @xcite ) @xmath119 the parameter @xmath120 has the meaning of the physical scale on the tev brane . in a number of papers ( see , for instance , refs .
@xcite , @xcite ) , the linear expansion about the _ background _ metric , @xmath121 is used instead of expansion . in such a case ,
@xmath122 , and the eigenvalue functions are equal to @xmath123 \ ; , \label{nonzero_egenfunctions}\end{aligned}\ ] ] with the constants @xmath124 and @xmath95 defined above .
the eigenvalue functions obey the following equations @xmath125 \tilde{\psi}^{(n)}(y ) = - m_n^2 e^{2\sigma(y ) } \tilde{\psi}^{(n)}(y)\ ] ] and boundary conditions @xmath126 \tilde{\psi}^{(n)}(y ) = 0 \ ; , \quad \mathrm{for \ } y = 0 , \
, \pi r_c \;.\ ] ] correspondingly , the orthonormality condition looks like @xmath127 it is clear that the kk graviton masses are defined by eq . as before . since @xmath128
, the lagrangian on the tev brane also remains the same for zero mode and massive modes .
the goal of this section is to estimate gravity effects in the dilepton production ( @xmath129 or @xmath130 ) , @xmath131 at the lhc in the rssc model .
the formulas for the @xmath1 distribution of the leptons are presented in appendix c. at fixed values of the dimensionless variable @xmath132 , the gravity cross section has the following dependence on fundamental gravity scale @xmath30 @xmath133 for numerical calculations , we impose the cut on the lepton pseudorapidity used by the cms collaboration . for the dimuon events it looks like @xmath134 @xmath135 is assumed for the dilepton events @xcite .
we use the mstw nnlo parton distributions @xcite , and convolute them with the partonic cross sections .
the pdf scale is taken to be equal to the invariant mass of the lepton pair , @xmath136 . in order to take into account sm higher order corrections , the @xmath137 factor 1.5
is used for the sm background , while a conservative value of @xmath138 is taken for the signal .
the differential cross section of the process under consideration has three terms @xmath139 where the last one comes from the interference between the sm and graviton interactions . since the sm amplitude is pure real ,
while the real part of each graviton resonance is antisymmetric with respect to its central point , the interference term has appeared to be negligible in comparison with the pure gravity and sm terms after integration in partonic momenta @xcite .
the account of the _ graviton widths _ is a crucial point for both analytical calculations and numerical estimations .
as it was shown in our previous papers @xcite , @xcite , an ignorance of the graviton widths is a _ rough _ approximation , since it results in very large suppression of the cross sections .
the reason lies partially in the fact that @xmath140 while in zero width approximation one gets @xmath141 let us stress that in the rssc model the gravity cross sections _ do not depend _ on the curvature @xmath0 ( up to small power corrections ) , provided @xmath142 , in contrast to the standard rs1 model in which all bounds on @xmath30 depend on the ratio @xmath143 @xcite . in figs .
[ fig : cs_1_8tev ] and [ fig : cs_3_8tev ] we present the gravity cross sections for the dielectron events at 8 tev lhc .
the differential cross sections at 13 tev are shown in figs .
[ fig : cs_1_13tev ] and [ fig : cs_3_13tev ] .
note that the gravity mediated contributions to the cross sections do not include the sm contribution ( i.e. solid lines in all figures correspond to pure gravity contributions ) .
the figures for dimuon cross sections look similar to figs .
[ fig : cs_1_8tev]-[fig : cs_3_13tev ] .
let @xmath144(@xmath145 ) be a number of signal ( background ) dilepton events with @xmath146 , @xmath147 then we define the statistical significance @xmath148 , and require a @xmath149 effect . in fig .
[ fig : s_7_8tev ] the statistical significance is shown for total number of `` events '' with @xmath150 tev and @xmath151 tev as a function of the transverse momentum cut @xmath152 and _ reduced _ 5-dimensional gravity scale @xmath30 .
the integrated luminosity was taken to be 5 fb@xmath7 and 20 fb@xmath7 for @xmath150 tev and @xmath153 tev , respectively .
figure [ fig : s_13tev ] represents the significance @xmath154 for the dilepton events with @xmath6 tev and 30 fb@xmath7 . previously , calculations of dilepton cross sections were done in @xcite _ without _ taking into account finite widths of the kk gravitons . as was shown in @xcite
( see also @xcite ) , in zero width approximation the gravity cross sections are very small in comparison with the background cross section at low and moderate values of @xmath1 .
that is why , a high cut @xmath152 is needed in order to get @xmath144 comparable with @xmath145 .
correspondingly , lhc search limits have appeared to be significantly smaller than in our case .
in the present paper the rssc model @xcite-@xcite is considered in detail in which the reduced 5-dimensional planck scale @xmath30 can vary from few tev to tens tev , while the curvature @xmath0 is allowed to vary from hundred mev to few gev .
in fact , the only condition @xmath142 should be satisfied .
the mass spectrum and experimental signature of the model with the small curvature are similar to those in the add model @xcite with one flat extra dimension .
the @xmath1-distributions for the lepton pairs with high @xmath1 at the lhc are calculated for the collision energies 7 , 8 , and 13 tev ( for the dielectron case , the results of our calculations are shown in figs .
[ fig : cs_1_8tev]-[fig : cs_3_8tev ] and figs .
[ fig : cs_1_13tev]-[fig : cs_3_13tev ] ) .
tev and cross sections for the dimuon events , since they look very similar to the dielectron cross sections . ] the account of the kk graviton widths is the crucial point for the numerical calculations , since the zero width approximation significantly underestimates the gravity cross sections even at large @xmath1 .
the statistical significance as a function of the _ reduced _ 5-dimensional planck scale @xmath30 and cut on the lepton transverse momentum @xmath152 is calculated ( see figs .
[ fig : s_7_8tev]-[fig : s_13tev ] ) .
let us underline that _ both the dielectron and dimuon _ events are taken into account .
recently , a search for large extra spatial dimensions in the dimuon ( dielectron ) channel was presented using a data sample of 20.6 fb@xmath7 ( 19.6 fb@xmath7 ) at the center - of - mass energy of 8 tev collected by the cms detector @xcite-@xcite .
no significant deviations from sm were observed .
previously , dilepton spectra were found to be consistent with sm expectations at the energy of 7 tev @xcite . by using our calculations for 7 and 8 tev
, we conclude that in the rssc framework the region @xmath155 is excluded at 95@xmath5 c.l .
note that for the add scenario the exclusion limits on the model parameter @xmath156 are 4.49 and 4.77 tev for the dimuon and dielectron events , respectively @xcite-@xcite .
correspondingly , we obtain the discovery limit for the 13 tev lhc with the integrated luminosity 30 fb@xmath7 : @xmath157 in deriving eqs .
, , we used the relation @xmath158 ( [ grav_scale_reduced ] ) in order to present the bounds on fundamental gravity scale @xmath4 .
it is important that these bounds on @xmath4 do not depend on the curvature @xmath0 ( up to small powerlike corrections ) , contrary to the original rs1 model @xcite in which estimated bounds on @xmath4 depend on the ratio @xmath143 .
previously , analogous bounds were obtained for the diphoton production @xcite .
recently , the lhc search limits were estimated for dimuon events @xcite .
dilepton production at very high luminosities ( hl - lhc ) was studied in @xcite .
the author is indebted to v.a .
petrov for useful discussions .
from the action , 5-dimensional einstein - hilbert s equations follow @xmath159 \;.\end{aligned}\ ] ] in order to solve them , let us assume that the background metric respects 4-dimensional poincare invariance ( @xmath160 ) @xmath161 after orbifolding ( see section [ sec:1 ] ) , the coordinate of the extra compact dimension varies within the limits @xmath162 .
let us define @xmath163 then the 5-dimensional background metric tensor looks like @xmath164 correspondingly , @xmath165 with @xmath166 non - trivial elements of the christoffel symbols of the second kind looks like ( there is no summation in @xmath130 ) @xmath167 as a result , non - zero elements of the curvature tensor @xmath168 are the following ( @xmath169 ) @xmath170 \ , g_{\mu\mu } \ ; , \nonumber \\
\mathcal{r}_{\mu \nu , \nu \mu } & = \mathcal{r}_{\nu \mu , \mu \nu } = - \mathcal{r}_{\mu \nu , \mu \nu } = - \mathcal{r}_{\nu \mu , \nu \mu } = - \sigma'^2(y ) \ , g_{\mu\mu } \ , g_{\nu\nu}\;.\end{aligned}\ ] ] the non - zero elements of the ricci tensor @xmath171 are @xmath172 \ ; , \nonumber \\ \mathcal{r}_{\mu\mu } & = [ - 4\sigma'^2(y ) + \sigma''(y ) ] \ , g_{\mu\mu } \;.\end{aligned}\ ] ] the 5-dimensional scalar curvature @xmath173 is equal to @xmath174 then the einstein - hilbert s equations are reduced to @xmath175 \;. \label{rs_sigma_eq_2}\end{aligned}\ ] ] the first solution of this set of equations was presented in ref .
@xcite @xmath176 with the cosmological constant @xmath177 and boundary cosmological terms @xmath178 here @xmath0 is a scale with the dimension of mass .
however , we get from that @xmath179 instead of @xmath180 $ ] . moreover , eqs . , say us that the cosmological constant @xmath19 _ should depend _ on coordinate @xmath14 . in the bulk
the set of equations looks like @xmath181 with the evident solution @xmath182 then we have for the full interval @xmath183 @xmath184 \ ; , \label{rssc_sigma_eq_2}\end{aligned}\ ] ] where @xmath185 for @xmath52 .
one obtains from eq .
@xmath186 + a \;,\ ] ] where @xmath187 is a constant , and @xmath188 in fact , @xmath189 is the function @xmath190 supplemented by its value at @xmath191 . will be _ uncertain _ at the boundary points @xmath192 and @xmath29 .
] let us note that @xmath193 ^ 2 = 1 $ ] for all @xmath194 , and @xmath195 = \left\ { \begin{array}{ll } 0 \ ; , & y < 0 \ ; , \\ 1 \ ; , & 0 \leqslant y \leqslant \pi r_c \ ; , \\ 0 \ ; , & y > \pi r_c \;.
\end{array } \right.\ ] ] since the choice of @xmath187 is equivalent to a redefinition of @xmath196 and @xmath197 , in what follows , we can put @xmath198 . as for the boundary cosmological terms @xmath9 and function @xmath199
, we get from the equation @xmath200 \;.\ ] ] thus , we obtain valid for all @xmath14 . ]
@xmath201 \ ; , \label{lambda_solution } \\
\lambda_1 & = - \lambda_2 = 6 \kappa \;. \label{brane_tensions_solutions}\end{aligned}\ ] ] as one can see from eqs . and , the cosmological constant @xmath19 is equal to @xmath202 at @xmath183 , and it is zero outside this region . finally , we find @xmath203 where @xmath204 is a constant .
note that eqs . , differ from rs1 fine tuning solutions , .
to get the rssc scenario , we take is equivalent to changing variables @xmath205 . ]
@xmath206 then @xmath207 and we come to the metric .
the function @xmath54 is @xmath208symmetric due to the periodicity condition ( points @xmath209 and @xmath210 are identified ) .
the @xmath14-dependent part of @xmath54 is symmetric under substitutions @xmath211 it means that the branes are interchanged if we take an opposite sign for @xmath0 ( see comments to eqs . - in the main text ) .
note that the the ricci tensor is proportional to the metric tensor _ only _ in between the branes , @xmath212 with @xmath19 being cosmological constant . for a space - time with the constant curvature @xmath137 , the following equation holds @xmath213 and we find @xmath214 correspondingly , the radius of the curvature _ in the bulk _
is @xmath215 at the boundaries , both the tensors @xmath168 , @xmath216 and scalar curvature @xmath217 have singular terms , as one can see from eqs . - and .
the normalization constants are obtained from the following relation ( @xmath79 ) @xmath218 ^ 2 = \frac{1}{\pi\kappa r_c b_n^2 } \ , n_n^2 \
, i_n \;,\ ] ] where @xmath219 ^ 2 \;.\ ] ] the parameters @xmath220 and @xmath221 are defined in the main text ( see eqs . , ) . to calculate @xmath222
, we use the table integral @xcite , @xmath223 + \mathrm{constant } \;,\end{aligned}\ ] ] where @xmath224 or @xmath91 , as well as relations between bessel functions , @xmath225 then the integral is equal to @xmath226 [ j_2 ( a_n ) y_1(b_n ) - y_2 ( a_n ) j_1(b_n ) ] \ ; , \nonumber \\ & - \frac{a_n^2}{2 } [ j_1 ( a_n ) y_1(b_n ) - y_1 ( a_n ) j_1(b_n ) ] ^2 \nonumber \\ & - \frac{a_n^2}{2 } [ j_2 ( a_n ) y_1(b_n ) - y_2 ( a_n ) j_1(b_n ) ] ^2 + \frac{2}{\pi^2 } \;.\end{aligned}\ ] ] due to the left boundary condition in the form , two first terms in the r.h.s of eq . vanish , and we get @xmath227 = \frac{2}{\pi^2 } \left [ 1 - \frac{j_1 ^ 2(b_n)}{j_1 ^ 2(a_n ) } \right ] \;.\ ] ] as a result
, we find the expression for the normalization constant @xmath228 \;.\ ] ] our formula is in agreement with that from ref .
@xcite . for @xmath229
, we have @xmath230 where @xmath231 \nonumber \\ \times & [ j_2 ( b_m z ) y_1(b_m ) - y_2 ( b_m z ) j_1(b_m ) ] \;.\end{aligned}\ ] ] by using table integral @xcite , @xmath232 \nonumber \\ & + \mathrm{constant } \;,\end{aligned}\ ] ] and eq .
, one can be easily show that @xmath233 .
the differential cross section of the dy process is given by ( @xmath129 or @xmath130 ) @xmath234 with the transverse energy of the lepton pair equals to @xmath235 . in two dimensionless quantities
are introduced @xmath236 where @xmath237 is the momentum fraction of the parton @xmath238 in . without cuts , integration variables in vary within the following limits
@xmath239 after imposing kinematical cut on lepton rapidity , the integration region becomes more complicated ( see appendix a in ref .
@xcite ) .
the sm contribution to the @xmath1-distribution looks like @xmath240 \;,\ ] ] with @xmath241 here @xmath242 , and @xmath243 with @xmath244 being third component of the quark isospin , @xmath245 being quark electric charge ( in units of @xmath246 ) .
the graviton contribution comes from both quark - antiquark annihilation and gluon - gluon fusion subprocesses ( see , for instance , @xcite ) @xmath247 where @xmath248 is the invariant part of the partonic matrix elements , with @xmath249 being _ total width of the graviton _ with the kk number @xmath250 and mass @xmath72 @xcite : @xmath251 let us note that the function @xmath252 is the same for all processes mediated by @xmath253-channel virtual gravitons .
it is important to underline that the magnitude of @xmath252 is defined by the fundamental gravity scale @xmath30 , not by the scale @xmath257 presented in the lagrangian . in general , this property is valid in the rssc model for _ both real and virtual _ production of the kk gravitons @xcite . 11 a.v .
kisselev , jhep * 04 * ( 2013 ) 025 [ arxiv:1210.3238 ] .
l. randall and r. sundrum , phys .
* 83 * ( 1999 ) 3370 [ hep - ph/9905221 ] .
g. f. giudice , t. plehn and a. strumia , nucl .
* b 706 * ( 2005 ) 455 [ hep - ph/0408320 ] .
kisselev and v.a .
petrov , phys . rev .
* d 71 * ( 2005 ) 124032 [ hep - ph/0504203 ] .
kisselev , phys .
d 73 * ( 2006 ) 024007 [ hep - th/0507145 ] .
h. davoudiasl , j.l . hewett and t.g .
rizzo , phys .
* 84 * ( 2000 ) 2080 [ hep - ph/9909255 ] .
m. toharia , mod .
. lett . * a 19 * ( 2004 ) 37 [ arxiv:0212036 ] .
boos , yu.a .
kubyshin , m.n .
smolyakov and i.p .
volobuev , class .
* 19 * ( 2002 ) 4591 [ hep - th/0202009 ] .
n. arkani - hamed , s. dimopoulos and g. dvali , phys . lett .
* b 429 * ( 1998 ) 263 [ hep - ph/9803315 ] . i. antoniadis , n. arkani - hamed , s. dimopoulos and g. dvali , phys . lett . * b 436 * ( 1998 ) 257 [ hep - ph/9804398 ] . n. arkani - hamed , s. dimopoulos and g. dvali , phys . rev . *
d 59 * ( 1999 ) 086004 [ hep - ph/9807344 ] .
kisselev , jhep * 09 * ( 2008 ) 039 [ arxiv:0804.3941 ] .
rubakov , phys .
* 44 * ( 2001 ) 871 [ hep - ph/0104152 ] .
cms collaboration , phys .
b 711 * ( 2012 ) 15 [ arxiv:1202.3827 ] .
a. martin , w. stirling , r. thorne and g. watt , eur .
j. * c 63 * ( 2009 ) 189 [ arxiv:0901.0002 ] .
cms collaboration , cms - pas - exo-12 - 027 .
cms collaboration , cms - pas - exo-12 - 031 .
cms collaboration , phys .
* b 711 * ( 2012 ) 15 [ arxiv:1202.3827 ] .
kisselev , _ rs model with a small curvature and drell - yan process at the lhc _ , in proceedings of cms workshop on perspectives on physics and on cms at very high luminosity , alushta , ukraine , 2012 .
dubna : jinr , e1 - 2013 - 96 , p. 155
[ arxiv:1208.3844 ] . a.p .
prudnikov , yu.a .
brychkov and o.i .
marichev , _ integrals and series _ , vol . 2 : _ special functions _ , gordon and breach sci . | the randall - sundrum - like scenario with the small curvature @xmath0 ( rssc model ) is studied in detail in comparison with the original rs1 model . in the framework of the rssc model ,
the @xmath1 distributions for the dilepton production at the lhc are calculated .
both dielectron and dimuon events are taken into account .
the important feature of calculations is the account of the widths of massive graviton excitations . for the summary statistics taken at 7 tev ( @xmath2 ) and 8 tev ( @xmath3 ) , the exclusion limit on the 5-dimensional gravity scale @xmath4
is set to be 6.84 tev at 95@xmath5 c.l . for @xmath6 tev and integrated luminosity 30 fb@xmath7
, the lhc search limit is found to be 10.16 tev .
these bounds on @xmath4 are independent of @xmath0 ( up to powerlike corrections ) , provided @xmath8 . |
the low - energy optical phonons in graphene can be represented by the relative displacement vector of two sub - lattice atoms @xmath82 and @xmath83 as @xcite , ( ) = , [ 4.1 ] which depends on spatial momentum @xmath84 .
the second quantized version of the above in configuration space is , ( ) = _ g,(a_g,+a^_g,-)_g()(i.).[4.2 ] with number of unit cells @xmath85 , carbon atom mass @xmath86 and @xmath87 labeling longitudinal or transverse modes .
the phonon creation @xmath88 and annihilation @xmath89 operators correspond to states with definite polarization @xmath90 .
the electron - phonon interaction hamiltonian , for a low - energy nearest - neighbor tight - binding model , can now be expressed as @xcite , _
int=-(u_y()_xu_x()_y),[4.3 ] corresponding to @xmath10 valleys respectively . here , @xmath91 is the equilibrium bond length , @xmath92 with @xmath93 being the resonance integral between nearest neighbor carbon atoms , and = -.the appearance of pauli matrices @xmath94 in eq .
[ 4.3 ] and comparison to the effective dirac hamiltonian @xmath95 identifies a @xmath0 gauge field with spatial components , \{a_x(),a_y()}=-\{u_y(),-u_x()},[4.4 ] that couples minimally to the dirac fermions .
the normalization and re - definitions of the emergent gauge field is done accordingly .
while explicating the 1-loop contribution to the gauge propagator , and subsequent emergence of the exciton , we consider standard qed@xmath44 parametrization for brevity .
this amounts to the replacements : @xmath96 , @xmath97 and @xmath98 .
the role of the tree - level gauge lagrangian , _
g =- f^f_+^a__a_,[a1 ] is to determine the tree - level propagator , g_f 0^=--,[a2 ] with covariant @xmath99 gauge depicted by the last term .
it modifies the 1-loop schwinger - dyson equation ( sde ) @xcite : ^-1=^-1+^(q)-q^q^.[5 ] thereby effecting the pole structure of the full propagator @xmath100 .
the full propagator gets contribution from a series of infinite vacuum polarization @xmath101 $ ] terms , making the result non - perturbative @xcite . on integrating out the fermion field from the full lagrangian , & = & |(x)(i^_-m)(x ) + & -&e|(x)a^(x)(x)_+_g , ( = c=1),[a3 ] the vacuum polarization contribution to the effective gauge action
is obtained as , with the schwinger regularization @xcite adopted in the second term in the integrand that removes the uv divergence for zero gauge momentum ( @xmath103 ) . the rest of the integral is standard and can be evaluated by the derivative expansion method @xcite , to obtain @xcite , ^(q)&&^_e(q)+^_o(q ) , + ^_e(q)&=&-_e(q)q^ , ^_o(q)=_o(q)^q _ , + q^&=&^q^2-q^q^ , ^=(1,-1,-1 ) , + _ e(q)&= & , + _ o(q)&=&-i().[4 ] the parity - odd contribution contain the valley index @xmath104 , which appear multiplied to @xmath105 , due to complete antisymmetry of @xmath106.]@xmath107 , unique to 2 + 1 dimensions @xcite arising due to non - zero trace of three dirac matrices , is the induced cs contribution .
the parity even contribution @xmath108 is responsible for wave - function renormalization @xcite .
the plots of both the form factors @xmath109 $ ] have been shown in fig .
[ f1 ] , with the well - known singularities at the two - particle threshold : @xmath110 .
the above results are valid below the same , and requires the replacement : @xmath111 above it , owing to the corresponding branch - cut . for dynamic cs qed , including the tree level propagator from eq .
[ a2 ] , the full 1-loop propagator is obtained through the sde as @xcite , . here
@xmath112 is assumed to be @xmath37 , for simplicity , in natural units.,width=3 ] @xmath113q^2}\nonumber\\ & \times\left[\left(q^{\mu}q^{\nu}-\eta^{\mu\nu}q^2\right)\{1+\pi_e(q)\}\right.\nonumber\\ & \qquad-\left.\epsilon^{\mu\nu\rho}q_{\rho}\left\{\pi_o(q)+i\mu\right\}\right]-\xi\frac{q^{\mu}q^{\nu}}{q^4}.\label{12}\end{aligned}\ ] ] this leads to the non - trivial pole governed by , q^2\{1+_e(q)}^2+\{_o(q)+i}^2=0.[13 ] the near two fermion threshold is parametrized as @xcite , @xmath114 , where @xmath40 is small and positive , and to be interpreted as the _ binding energy _ of the exciton .
on expanding the form factors to lowest power of @xmath40 and noticing that the logarithm diverges much faster than any other term in their expressions @xcite , one obtains , _
e(q ) & & + & & + _ o(q)&&-i().[29 ] then , from eq .
[ 13 ] , the binding energy of the exciton is , 4m\{(2m-)}.[14 ] as expected , the above expression of binding energy yields hagen s result as a special case , without the first term in the exponent .
the extension of above system to finite temperature has been carried out in this section . in 2 + 1 dimensions ,
results were known for mass - less fermion case @xcite and induced cs term @xcite at low energies .
we here obtain the general results , near two - particle threshold . adopting the real - time formalism of qft
, the fermionic propagator at @xmath78 is obtained from that at @xmath77 as @xcite , we first proceed to obtain the contribution to the even part of vacuum polarization , for which we follow the formalism by weldon @xcite , originally carried - out in 3 + 1 dimensions .
introduction of finite temperature includes the notion of a thermal bath as a physical reference frame , thus breaking the manifest lorenz co - variance .
however , it is possible to obtain a lorentz covariant formulation by projecting onto and out of the bath coordinate vector @xmath117 @xcite .
then the temperature dependent part of the even component of vacuum polarization can be expressed as , ^_e(q , t)&=&_t(q,)p^+_l(q,)r^ , + p^&=&^-u^u^+ , + r_&=&-(q^2u_+q_)(q^2u_+q _ ) , + & = & q.u , q^2=^2-q^2 , + q^&=&q^-u^.[16 ] here @xmath118 is transverse and @xmath119 is longitudinal in nature and are orthogonal to each - other . the _
thermal _ form - factors are expressed as : _ l(q,)&=&-u_u_^ , + _ t(q,)&=&-_l(q,)+_^ , + re(_^)&=&e^2g_f(q , ) , + re(u_u_^)&=&e^2h_f(q,).[17 ] the real part of @xmath73 is of interest here , as the imaginary part of the same corresponds to the induced cs term , which was obtained by babu _
et al _ @xcite , along with the @xmath77 contribution , in 2 + 1 in the high temperature limit @xmath75 suitable for present discussion .
we will obtain the general result in the imaginary - time formalism shortly , and show the mentioned result as a limiting case . where @xmath121
. the finite temperature contribution does not introduce additional divergences @xcite .
it is well - known that the above @xmath122-integrals can not be solved exactly @xcite . as mentioned before
, we ought to take the high temperature ( t ) limit with @xmath123 , finally yielding , _
l(q , t)&&-e^2(2 ) + _ t(q , t)&&e^2(2).[18 ] these contributions are linear in t , as in the hard thermal loop ( htl ) approximation for 2 + 1 .
this is expected from dimensional arguments as the 3 + 1 counterparts are quadratic in t @xcite .
t)}^-1&&\{g^f 0_(q , t)}^-1+^_e(q,0)+^_e(q , t ) + & + & ^_0(q , t)+q^q^ , + ^_o(q , t)&=&-i_0 ^ 1dx_m(x+is ) + & & ^q_,[20 ] with @xmath127 and @xmath128 , where @xmath129 here , @xmath130 is the complete finite temperature odd 1-loop contribution at suitable high temperature and low momentum limits that falls back to the low - energy expression , given in ref .
@xcite , for @xmath131 . in that domain ,
up to @xmath132 , the form factor @xmath133 is independent of the external momentum @xmath134 as has been checked explicitly .
the full tree level gauge propagator at finite temperature ( including zero temperature contributions ) , @xmath135 , is obtained through the replacement @xcite , ---(q^2-m_g^2),in the over - all factor .
the finite - temperature contribution vanishes for @xmath136 , where @xmath137 is the mass of the gauge particle . since @xmath134 is the external momentum , and we eventually are interested in the region just _ below
_ the two - particle threshold , this contribution can be neglected from the onset as either @xmath138 or @xmath139 , with quantum corrections included .
thus , one can work with the zero - temperature tree - level propagator , which has been verified directly too .
g^(q , t ) & & a(q , t)^+b(q , t)q^q^+c(q , t)u^u^ + & + & d(q , t)(q^u^+u^q^)+e(q , t)^q _ ; + a(q , t)&= & , + b(q , t)&=&a(q , t ) + & + & , + c(q , t)&=&-a(q , t ) , + d(q , t)&=&-c(q , t ) , + e(q , t)&=&-a(q , t),[21 ] the gauge dependence retains its covariant @xmath141 form in @xmath142 .
also , @xmath143 and @xmath144 vanish at zero temperature and the corresponding full propagator is retained .
_ l - q^2(_e+1)&=&0 + ^2+q^2(_o+i)^2&=&0,[22 ] without gauge dependence , ensuring physicality . the conventional high - temperature approximation ( @xmath75 ) , along with the @xmath40-expansion near two - fermion threshold yields , _
o(q , t)&&m , + _ e(q,0)&&[()-1].[23 ] consequently , the first pole equation unacceptably shows _ increase _ of binding energy with temperature , in the physical domain @xmath145 , as illustrated in fig .
[ f3 ] by plotting temperature - independent terms , @xmath146 . on the other hand , the _ second _ pole equation in eqs .
[ 22 ] correctly represents evaporation of the exciton at sufficiently high temperature . near two - fermion threshold , on substituting the expressions in eqs .
[ 18 ] and [ 23 ] in this pole equation leads to the exciton binding energy , 99 l. baulieu , v. kazakov , m. picco , and p. windey , _ low - dimensional applications of quantum field theory _ , ( springer science and business media , new york , 2013 ) and references therein .
s. nadj - perge , i. k. drozdov , j. li , h. chen , s. jeon , j. seo , a. h. macdonald , b. a. bernevig , and ali yazdani , science * 346 * , 602 ( 2014 ) .
m. okamoto , y. takane , and k .-
imura , phys .
b * 89 * , 125425 ( 2014 ) .
p. moneau , n. p. ong , a. m. portis , a. meerschaut , and j. rouxe , phys .
lett . * 37 * , 602 ( 1976 ) ; g. grner , rev . mod .
phys . * 60 * , 1129 ( 1988 ) and references therein . s. s. hong , y. zhang , j. j. cha , x .-
qi , and y. cui , nano lett . , * 14 * , 2815 ( 2014 ) .
b. j. kim , h. koh , e. rotenberg , s .- j .
oh , h. eisaki , n. motoyama , s. uchida , t. tohyama , s. maekawa , z .- x .
shen , and c. kim , nature physics * 2 * , 397 ( 2006 ) ; y. jompol , c. j. b. ford , j. p. griffiths , i. farrer , g. a. c. jones , d. anderson , d. a. ritchie , t. w. silk , and a. j. schofield , science * 325 * 597 ( 2009 ) and references therein .
g. a. steele , f. pei , e. a. laird , j. m. jol , h. b. meerwaldt , and l. p. kouwenhoven , nature communications * 4 * , 1573 ( 2013 ) .
l. fu and c. l. kane , phys .
lett . * 100 * , 096407 ( 2008 ) .
k. novoselov , a. k. geim , s. v. morozov , d. jiang , y. zhang , s. v. dubonos , i. v. grigorieva , and a. a. firsov , science * 306 * , 666 ( 2004 ) .
thouless , m. kohmoto , m.p .
nightingale , and m. den nijs , phys .
lett . * 49 * , 405 ( 1982 ) .
s. c. zhang , t. h. hansson , and s. kivelson , phys .
* 62 * , 82 ( 1989 ) ; d. h. lee and s. c. zhang , phys .
lett . * 66 * , 1220 ( 1991 ) ; s. c. zhang , int . j. mod .
b * 6 * , 25 ( 1992 ) .
wen , int .
b * 6 * , 1711 ( 1992 )
. m. stone , _ quantum hall effect _
( world scientific , singapore , 1992 ) and references therein .
c. l. kane and e. j. mele , phys .
lett . * 95 * , 146802 ( 2005 ) .
b. a. bernevig , t. l. hughes , and s .- c .
zhang , science * 314 * , 1757 ( 2006 ) .
m. z. hasan and c. l. kane , rev .
phys . * 82 * , 3045 ( 2010 ) and references therein .
x. chen , z .- x .
liu , and x .-
wen , phys .
b * 84 * , 235141 ( 2011 ) .
j. f. schonfeld , nucl .
b * 185 * , 157 ( 1981 ) .
s. deser , r. jackiw , and s. templeton , phys .
lett . * 48 * , 975 ( 1982 ) ; ann .
( n. y. ) * 140 * , 3372 ( 1982 ) .
g. v. dunne , _ aspects topologiques de la physique en basse dimension .
topological aspects of low dimensional systems _ , ( springer berlin , heidelberg , 1999 ) and references therein .
f. wilczek , phys .
lett . * 49 * , 957 ( 1982 ) . c. r. hagen , ann .
( n. y. ) * 157 * , 342 ( 1984 ) .
p. k. panigrahi , s. roy , and w. scherer , phys .
lett . * 61 * , 2827 ( 1988 ) .
j. m. leinaas and j. myrheim , nuovo cimento b * 37 * , 1 ( 1977 ) .
semenoff and l.c.r .
wijewardhana , phys .
b * 184 * , 397 ( 1987 ) .
f. wilczek , _ fractional statistics and anyon superconductivity _ , ( world scientific , singapore , 1990 ) and references therein . c. r. hagen , phys .
lett . * 68 * , 3821 ( 1992 ) .
a. das and s. panda , j. phys .
a * 25 * , l245 ( 1992 ) .
a. n. redlich , phys .
* 52 * , 18 ( 1984 ) ; phys .
d * 29 * , 2366 ( 1984 ) .
t. w. appelquist , m. bowick , d. karabali , and l. c. r. wijewardhana , phys .
d * 33 * , 3704 ( 1986 ) .
d. boyanovsky , r. blankenbecler , and r. yahalom , nucl .
b * 270 * , 483 ( 1986 ) . c. r. hagen , p. panigrahi , and s. ramaswamy , phys .
lett . * 61 * , 389 ( 1988 ) .
p. r. wallace , phys . rev . * 71 * , 622 ( 1947 ) .
g. w. semenoff , phys .
53 * , 2449 ( 1984 ) .
t. ohta , a. bostwick , j. l. mcchesney , t. seyller , k. horn , and e. rotenberg , phys .
. lett . * 98 * , 206802 ( 2007 ) ; s. y. zhou , g .- h .
gweon , a. v. fedorov , p. n. first , w. a. de heer , d .- h .
lee , f. guinea , a. h. c. neto , and a. lanzara , nature materials * 6 * , 770 ( 2007 ) ; c. r. woods , l. britnell , a. eckmann , r. s. ma , j. c. lu , h. m. guo , x. lin , g. l. yu , y. cao , r. v. gorbachev _ et al .
_ , nature physics * 10 * , 451 ( 2013 ) .
e. mccann and v. i. falko , phys .
96 * , 086805 ( 2006 ) ; e. mccann , phys .
b * 74 * , 161403 ( 2006 ) ; t. ohta , a. bostwick , t. seyller , k. horn , and e. rotenberg , science * 313 * , 951 ( 2006 )
. b. hunt , j. d. sanchez - yamagishi , a. f. young , m. yankowitz , b. j. leroy , k. watanabe , t. taniguchi , p. moon , m. koshino , p. jarillo - herrero , and r. c. ashoori , science * 340 * , 1427 ( 2013 ) and references therein .
lu , w .- y .
shan , w. yao , q. niu , and s .- q .
shen , phys .
b * 81 * , 115407 ( 2010 ) .
d. hsieh , d. qian , l. wray , y. xia , y. s. hor , r. j. cava , and m. z. hasan , _ nature _ * 452 * , 970 ( 2008 ) .
y. l. chen , j .- h .
chu , j. g. analytis , z. k. liu , k. igarashi , h .- h .
kuo , x. l. qi , s. k. mo , r. g. moore , d. h. lu _
_ , _ science _ * 329 * , 659 ( 2010 ) ; m. i. katsnelson , f. guinea , and m. a. h. vozmediano , epl * 104 * , 17001 ( 2013 ) ; o. vafek and a. vishwanath , annu . rev . of condens .
matter phys .
* 5 * , 83 ( 2014 ) .
r. denk , m. hohage , p. zeppenfeld , j. cai , c. a. pignedoli , h. sde , r. fasel , x. feng , k. mllen , s. wang _
_ , _ nature communications _ * 5 * , 4253 ( 2014 ) . s .-
xu , y. xia , l. a. wray , s. jia , f. meier , j. h. dil , j. osterwalder , b. slomski , a. bansil , h. lin _ et al .
_ , _ science _ * 332 * , 560 ( 2011 ) . i. i. kogan , phys .
b * 262 * , 83 ( 1991 ) .
n. tombros , c. jozsa , m. popinciuc , h. t. jonkman , and b. j. van wees , nature ( london ) * 448 * , 571 ( 2007 ) ; h. haugen , d. huertas - hernando , and a. brataas , phys .
b * 77 * , 115406 ( 2008 ) ; m. popinciuc , c. jzsa , p. j. zomer , n. tombros , a. veligura , h. t. jonkman , and b. j. van wees , phys . rev .
b * 80 * , 214427 ( 2009 ) .
o. v. yazyev , j. e. moore , and s. g. louie , phys .
lett . * 105 * , 266806 ( 2010 ) .
y. nambu and g. jona - lasinio , phys . rev . *
122 * , 345 ( 1961 ) ; * 124 * , 246 ( 1961 ) .
j. wess , b. zumino , phys .
b * 37 * , 95 ( 1971 ) ; s. p. novikov , sov . math . , dokl . * 24 * , 222 ( 1981 ) ; _ russian mathematical surveys _ * 37 * , 1 , ( 1982 ) ; e. witten , nucl . phys .
b * 223 * , 422 ( 1983 ) ; commun .
phys . * 92 * , 455 ( 1984 ) .
k. ishikawa and t. ando , j. phys .
. jpn . * 75 * , 84713 ( 2006 ) .
a. h. c. neto , f. guinea , n. m. r. peres , k. s. novoselov , and a. k. geim , rev .
phys . * 81 * , 109 ( 2009 ) and references therein .
k. sasaki and r. saito , prog .
. suppl . *
176 * , 253 ( 2008 ) ; m. a. h. vozmediano , m.i .
katsnelson , and f. guinea , phys . rep . *
496 * , 109 ( 2010 ) and references therein .
a. h. c. neto , nature materials * 6 * , 176 ( 2007 ) .
a. m. j. schakel , _ boulevard of broken symmetries : effective field theories of condensed matter _
( world scientific , singapore , 2008 ) and references therein . w. kohn , phys .
* 2 * , 393 ( 1959 ) .
a. das , b. chakraborty , s. piscanec , s. pisana , a. k. sood , and a. c. ferrari , phys .
b * 79 * , 155417 ( 2009 ) .
s. piscanec , m. lazzeri , f. mauri , a. c. ferrari , and j. robertson , phys .
lett . * 93 * , 185503 ( 2004 ) . c. itzykson and j .- b .
zuber , _ quantum field theory _
( courier dover , new york , 2012 ) and references therein . r. jackiw and s. templeton ,
d * 23 * , 2291 ( 1981 ) ; t. appelquist and u. heinz , phys .
d * 24 * , 2169 ( 1981 ) .
s. coleman and b. hill , phys .
b * 159 * , 184 ( 1985 )
. k. s. babu , a. das , and p. k. panigrahi , phys . rev.d * 36 * , 3725 ( 1987 ) .
a. kovner and b. rosenstein , phys .
lett . * 59 * , 857 ( 1987 ) .
see supplemental material at [ electronic link for the supplemental material ] for detailed derivation .
r. d. pisarski and s. rao , phys .
d * 32 * , 2081 ( 1985 ) .
s. rao and r. yahalom , phys .
b * 172 * , 227 ( 1986 ) ; p. maris , phys .
d * 52 * , 6087 ( 1995 ) .
m. mecklenburg and b. c. regan , phys .
lett . * 106 * , 116803 ( 2011 ) .
f. dyson , phys .
rev . * 75 * , 1736 ( 1949 ) ; j. schwinger , proc .
u. s. a. * 37 * , 452 ( 1951 ) .
s. zhang , j. zhouc , q. wanga , x. chen , y. kawazoe , and p. jena , proc .
u. s. a. * 112 * , 2372 ( 2015 ) .
j. bardeen , l. n. cooper , and j. r. schrieffer , phys . rev . * 106 * , 162 ( 1957 ) ; * 108 * , 1175 ( 1957 ) .
h. frhlich , h. pelzer and s. zienau , philos . mag . * 41 * , 221 ( 1950 ) ; h. frhlich , adv .
phys . * 3 * , 325 ( 1954 ) .
v. l. ginzburg and l. d. landau , zh .
phys .- jetp ] * 20 * , 1064 ( 1950 ) .
w. siegel , nucl .
b * 156 * , 135 ( 1979 ) .
a. das , _ finite temperature field theory _
( world scientific , singapore , 1999 ) and references therein . n. dorey and n.e . mavromatos , nucl . phys .
b * 386 * , 614 ( 1992 ) .
l. dolan and r. jackiw , phys .
d * 9 * , 3320 ( 1974 ) ; h. umezawa , h. matsumoto and m. tachiki , _ thermo field dynamics and condensed states _ " , north - holland , amsterdam ( 1982 ) and references therein .
h. a. weldon , phys .
d * 26 * , 1394 ( 1982 ) .
m. le bellac , _ thermal field theory _ " , cambridge monographs on mathematical physics , cambridge university press , cambridge ( 2000 ) and references therein .
d. a. dicus _
et al . _ ,
d * 26 * , 2694 ( 1982 ) ; d. a. dicus , p. down and e. w. kolb , nucl .
b * 223 * , 525 ( 1983 ) ; r. l. kobes , g. w. semenoff and n. weiss , z. phys .
c - particles and fields * 29 * , 371 ( 1985 ) . | it is shown that a charge - neutral spin-1 exciton , realizable only in planar systems like graphene , can effectively be used for controlled spin transport in such media .
the excitonic bound state is destabilized by quantum fluctuations , characterized by a threshold for excitation and melts in a smooth manner under thermal fluctuations .
this planar exciton differs from the conventional ones , as it owes its existence to the topological chern - simons ( cs ) term .
the parity and time - reversal violating cs term can arise from quantum effects in systems with parity - breaking mass - gap .
the spinning exciton naturally couples to magnetic field , leading to the possibility of controlled spin transport . being neutral , it is immune to adverse effects , afflicting spin transport by charged fermions .
# 1#2#3#4#1 * # 2 * , # 3 ( # 4 ) lower dimensional systems exhibit a host of unique features @xcite , absent in their higher - dimensional counterparts .
realization of majorana fermion @xcite , thin film topological insulators ( tis ) @xcite , charge density wave @xcite , helical transport @xcite and charge - spin separation @xcite characterize one - dimensional systems .
planar ( 2 + 1 dimensional ) materials have led to physical realization of large spin - orbit coupling @xcite , spin - momentum locking @xcite and low - energy dirac fermions @xcite etc .
2 + 1 dimensional systems are endowed with non - trivial inherent topology , which manifests in quantum hall effect @xcite and boundary conductivity in tis @xcite , having symmetry protected topological order @xcite .
[ [ section ] ] the characteristic non - trivial feature of the planar world , absent in their even - dimensional counterparts , is the manifestation of the topological chern - simons ( cs ) term , _
cs=^a__a_.[e1 ] in the gauge sector , due to which , propagating photons can have gauge - invariant mass @xcite .
interacting particles can acquire additional spin @xcite , leading to a change in their statistics @xcite .
the cs term arises due to quantum effects , in theories with parity - breaking massive fermions @xcite and bosons @xcite .
the advent of planar materials like graphene @xcite and tis @xcite have led to physical realization of low - energy ` relativistic ' planar systems .
massive fermions have been experimentally realized by inducing sub - lattice density asymmetry in bi - layer graphene under transverse electric field @xcite and in graphene mono - layer , misaligned with the substrate @xcite .
the hexagonal structure of tis leads to a single dirac point , with locally inducible gap through magnetic induction @xcite .
the importance of mass term lies in the fact that in 2 + 1 dimensions , mass - less particles are spin - less .
furthermore , parity - breaking mass term is essential for generating the cs term , not present in natural electromagnetism .
[ [ section-1 ] ] in this letter , we demonstrate controlled spin transport in graphene - like materials , through charge - less spin-1 excitons , owing their existence to this topological cs term .
the fact that , the in - plane optical phonon and lattice curvature couple as @xmath0 gauge fields to the emergent dirac fermions of graphene , leads to possible realization of the topological cs term , along with a dynamic gauge part , through quantum corrections .
the quantum fluctuations destabilize the bound state , at zero temperature .
this leads to a parametric threshold for excitation of this topological quasi - particle .
the bound - state smoothly melts at high temperature .
[ [ section-2 ] ] quite some time back , hagen demonstrated the existence of this spin-1 _ exciton _ , a weakly bound fermion - antifermion pair , arising due to the cs term in 2 + 1 dimensional gauge theory @xcite .
this is unlike the non - topological excitons , recently seen in graphene nano - ribbons @xcite and in tis @xcite .
being a spin-1 object , this exciton naturally couples to magnetic field @xcite , leading to possible controlled _ spin - transport _ in planar systems like graphene , which is fundamentally different from the spin - transport of spin-@xmath1 charged excitations , already realized in graphene @xcite and in tis @xcite .
similar fermionic bound states , possessing different spins , naturally manifest in various branches of physics .
the nambu jona - lasinio model @xcite realizes pseudoscalar pions as mass - less bound state of protons and neutrons .
the space - time parity - breaking low - energy topological wess - zumino - novikov - witten term @xcite correctly describes the interactions of these particles in 3 + 1 dimensions .
[ [ section-3 ] ] as mentioned earlier , in graphene , in - plane optical phonons @xcite and defect - induced lattice curvature @xcite couple to the low - energy relativistic fermions as @xmath0 gauge fields , the former having _ spatial _ components @xmath2 @xcite , with @xmath3 depicting relative displacement of the sub - lattices ( appendix a ) .
this has led to ` non - adiabatic ' effects , leading to ` phonons behaving badly ' @xcite .
a temporal gauge component can also be introduced through the chemical potential @xcite .
the structure of gauge self - energy in the plane has a characteristic logarithmic singularity , causing ` kohn anomaly ' @xcite .
this has been experimentally observed @xcite , and realized in graphene through electron - phonon interaction @xcite .
[ [ section-4 ] ] though lattice excitations in graphene couple to dirac fermions as effective gauge fields , they do _ not _ a priory possess kinetic energy or a parity - breaking cs term .
both of these can arise through the induced effects of the underlying fermions to which the optical phonon couples . in case of graphene
, the fermion mass arises through broken valley degeneracy , resulting in a dispersion of the form : @xmath4 @xcite , with scaling of mass ( @xmath5 ) and coupling strength @xmath6 by fermi velocity @xmath7 , as compared to standard planar gauge theory @xcite . at low energies , where massive fermions are not excited , the gauge field dynamics is obtained by integrating out one of the minimally coupled fermionic fields , @xmath8 , from the primary lagrangian : & = & |_s(i^_s_-m_sv_f^2-gv_f^a_)_s + & + & |_s(i^_s_-m_sv_f^2-gv_f^a_)_s,[n01 ] with @xmath9 denoting the two valleys @xmath10 of graphene , with respective dirac matrices @xmath11 , in momentum - space @xcite .
the covariant momentum operator is defined as @xmath12 .
the corresponding lowest order quantum contribution to the remaining tree - level lagrangian , is expressed through the vacuum polarization tensor @xcite , split into parity even and odd parts as , ^(q)&=&_e(q)\{q^q^-^q^2}+_o(q)^q _ , + q^&=&cq^0+v_f , with corresponding form factors @xmath13 respectively .
this further represents the linear response of the system @xcite , which is of interest here .
the topological part , responsible for the induced cs term , is unaffected by contributions from fermion sector beyond 1-loop @xcite .
[ [ section-5 ] ] following proper regularization ( appendix b ) , the vacuum polarization form factors are found to be @xcite , @xmath14,\nonumber\\ \pi_o(q)&=-is\frac{m_sv_f^2}{4\pi}\frac{g^2v_f^2}{|q|}\log\left(\frac{2\vert m_s\vert v_f^2+\vert q\vert}{2\vert m_s\vert v_f^2-\vert q\vert}\right).\label{4}\end{aligned}\ ] ] evidently , both the form factors possess logarithmic singularities at the fermionic two - particle threshold , @xmath15 , and these expressions are valid _ below _ the same , which is the domain of bound - state formation .
the even form factor @xmath16 influences wave - function renormalization , while the odd one , @xmath17 , is the 1-loop cs contribution .
[ [ section-6 ] ] the effective lagrangian now takes the form , _
e&=&|_s(i^_-m_sv_f^2-gv_f^a_)_s+l_g ; + _ g&=&-f_f^+^a__a _ , + f_&=&_a_-_a _ , = c=1,[n02 ] with dynamic cs gauge lagrangian @xmath18 and field - strength tensor @xmath19 for the emergent gauge field @xmath20 .
it is worth mentioning that @xmath21 is a compact way to write the kinetic energy of the lattice vibrations .
@xmath18 is formed by the vacuum polarization tensor , and @xmath22 is normalized modulo @xmath23 , leading to the expression , = -s=-2sm_sv_f^2.[n03 ] the remaining fermionic field @xmath24 can now be integrated out to obtain 1-loop quantum corrections , through vacuum polarization , having form - factors exactly like those in eq .
[ 4 ] , except @xmath25 replaced by @xmath26 .
[ [ section-7 ] ] both species of dirac fermion are of _ same _ mass in graphene @xcite , following the electron hamiltonian @xcite , h=_i - t_i , j.the individual spin densities @xmath27 of @xmath28 sub - lattices yield the mass @xmath29 for both , akin to a finite chemical potential , as the sub - lattices have fixed opposite spins .
the corresponding dirac hamiltonians for the two valleys are @xmath30 at low energies .
this equality of mass cancels out the individual 1-loop topological contributions @xcite , unlike the even parts of vacuum polarization which add up , yielding the dynamic part of the effective gauge lagrangian .
the topological exciton manifests only if @xmath31 , with the ` lighter ' dirac fermion having sufficiently low two - particle threshold @xmath32 to validate dirac dispersion in graphene .
it can be integrated - out to yield a cs contribution that does not cancel - out with @xmath33 in eq .
[ n03 ] , as the latter will be logarithmically growing near the threshold ( eq . [ 4 ] ) .
this will finally yield the exciton through 1-loop schwinger - dyson equation ( sde ) @xcite .
recently proposed ` penta - graphene ' displays intrinsic quasi - direct band gap @xcite , which is a superposition of _ two _ hexagonal arrays , wherein inducing a second independent mass gap may be possible .
[ [ section-8 ] ] it is worth pointing - out that , a pure cs term , dominant one in low energy limit , is devoid of dynamics .
it represents a parity - breaking four - fermi interaction between the light fermions , akin to the bcs heory of superconductivity @xcite , where the phonon interactions in the frhlich hamiltonian @xcite , leads to an effective four - fermi term .
further , the expression for binding energy resembles that of the landau - ginzburg gap equation @xcite
. however , here it is a physical bound state formation , unlike the spontaneous symmetry breaking in superconductivity . as function of coupling strength @xmath34 with effective gauge field and the cs coefficient @xmath35 , for @xmath36 being @xmath37 ( maroon ) , @xmath38 ( green ) and @xmath39 ( blue ) respectively ( in natural units ) .
all three show suitable regions for bound state formation ( small @xmath40 ) marked in red color , signifying the importance of the condition @xmath41.,width=3 ] [ [ section-9 ] ] the modification of dynamics at tree - level due to quantum corrections can be represented through the sde . incorporating covariant gauge - fixing , the non - perturbative 1-loop propagator , containing contributions from an infinite number of bubble diagrams ,
is obtained as ( appendix b ) , @xmath42q^2}\nonumber\\ & \times\left[\left(q^{\mu}q^{\nu}-\eta^{\mu\nu}q^2\right)\{1+\pi_e(q)\}\right.\nonumber\\ & \qquad-\left.\epsilon^{\mu\nu\rho}q_{\rho}\left\{\pi_o(q)+i\bar{\mu}\right\}\right]-\xi\frac{q^{\mu}q^{\nu}}{q^4},\label{12}\end{aligned}\ ] ] with a non - trivial , gauge - invariant pole corresponding to @xmath43 , which can represent a physical state @xcite .
here the over - all scaling , introduced in eq .
[ n03 ] , is factored - out .
such a pole was obtained by hagen @xcite , in ` pure ' cs qed@xmath44 ( @xmath45 replaced with @xmath46 in eq .
[ n01 ] , with parameters suitably redefined ) , just below the fermionic two - particle threshold : @xmath47 , interpreted as a fermion - antifermion _
spin_-1 bound state ( exciton ) , as planar particle and antiparticle have the same spin projection @xcite .
it has a ` binding energy ' that translates as @xmath48 in graphene parameters . in presence of the dynamic term
, we obtain a significantly modified expression : 4m_sv_f^2\{(2m_sv_f^2-|)}.[14 ] in both these cases , presence of @xmath46 at tree - level in the effective theory , is necessary for obtaining a self - consistent ( small magnitude ) value of the exciton binding energy , that disappears for vanishing @xmath0 coupling @xmath49 .
as can be checked with different gauge lagrangians , for @xmath50 , such a value can not exist , thereby highlighting the intrinsic topological nature of this exciton .
[ [ section-10 ] ] the smallness of the binding energy ( @xmath40 ) necessitates a _ negative _ exponent , thereby fixing a threshold value @xmath51 for bound - state formation , highlighting the destabilizing effect of vacuum fluctuation on the excition due to the kinetic energy term .
it further explicates the fact that as the theory is topologically massive at the tree level ( appendix b ) , to be realizable in the gauge sector , the exciton can not be lighter than the rest mass of the gauge particle .
this is one of the main results of this letter .
this ` non - perturbative ' scenario , even with relatively large coupling strength @xmath52 , is justified in terms of the large - n suppression of higher - order contributions @xcite . for a negative exponent ,
a shallow bound state is physically meaningful for _ small _ coupling @xmath52 , and deep otherwise , maintaining the self - consistency regarding @xmath53 .
the parametric regions for attaining this exciton is shown in fig .
[ f2 ] . [ [ section-11 ] ] interestingly , the appearance of @xmath54 in the denominator of the exponent in eq .
[ 14 ] , makes the transition about @xmath55 more prominent , than what it would have been in vacuum , as @xmath56 in natural units . hence , though both exciton binding energy and formation threshold are small in graphene , due to smallness of @xmath57 , the sensitivity to the bound state formation will be considerably higher , making experimental verification more likely .
[ [ section-12 ] ] the physical properties of this exciton are most evident from the renormalization coefficients , large - n protected beyond 1-loop , that can be read - off from eq . [ 12 ] .
the lehmann weight @xmath58^{-1}$ ] vanishes near the two - particle threshold , owing to the logarithmic singularity , marking emergence of bound - state as per klln - lehmann spectral representation @xcite .
this further shows vanishing of the renormalized charge @xmath59 , representing a charge - less state .
the renormalized topological mass : @xmath60z_3 $ ] leads to @xmath61 in the same limit , marking the threshold .
[ [ section-13 ] ] this topological excitation has unit spin @xmath62 @xcite , with , & : = & -2sm_sv_f^2-sm_sv_f^2 ( ) + & & -sm_sv_f^2(),being the net cs coefficient , near the two - particle threshold of the lighter fermion .
it is being proposed as an ideal candidate for controlled spin - transport , unaffected by local electric fields for being charge - neutral .
its dynamics can be controlled by an _
external _ magnetic field @xmath63 , allowing for spintronics . the corresponding ` external ' gauge field ( @xmath64 ) couples to the dirac fermions with the same strength @xmath34 .
the resultant effective gauge action , at low energies , contains an interaction term @xmath65 between the two gauge fields , resolvable as , ^m_i=&-&f_f^ + & + & ^a_f_,[st1 ] with external field tensor @xmath66 .
@xmath67 is the component of @xmath63 normal to the surface of graphene , that couples to the exciton spin pseudo - scalar in the same direction .
this is clearly evident in the exciton rest - frame , where the first term of eq .
[ st1 ] , representing orbital motion , vanishes ( @xmath68 ) .
however , the gauge - invariant _ mixed _ cs term @xcite in the second , with corresponding valley contributions adding - up @xcite , survives .
this represents exiton spin coupling with @xmath67 , with magnetic moment @xmath69 , which is logarithmically large near the two lighter fermion threshold .
the _ temporal _ gauge field component @xmath70 , physically equivalent to a chemical potential @xcite , is unlike the ` spatial ' ones @xmath71 , which are resultants of relative sub - lattice displacements @xcite and thus , are geometric in nature .
the non - dynamic mixed cs term inseparably associates quantized magnetic flux to an electric charge @xcite , rendering a non - zero @xmath70 that represents electrostatic potential due to charge polarization induced by @xmath67 @xcite .
additionally , the planar spin being a pseudo - scalar , @xmath72 equivalently represents interaction of exiton spin with external gauge momentum , akin to a reduced pauli - lubanski ` pseudo - scalar ' .
[ [ section-14 ] ] the effect of thermal fluctuations on this topological exciton is demonstrated through a finite temperature treatment @xcite , yielding the temperature dependent extension to @xmath73 ( appendix c ) .
interestingly , the corresponding 1-loop propagator has _ two _ non - trivial poles , with only one being physically acceptable . near two - fermion threshold , the corresponding expression for exciton binding energy obtained as , @xmath74,\label{24}\end{aligned}\ ] ] correctly depicting smooth evaporation of the exciton at sufficiently high temperature . in the conventional high - temperature approximation ( @xmath75 ) ,
the finite temperature form factor is @xmath76 , where , _
t(q)=(2),[23 ] is temperature independent , as per dimensional arguments @xcite . in the @xmath77 limit
, @xmath40 goes back to the zero - temperature expression , modulo a constant , owing to non - analytic continuation from non - covariant @xmath78 sector to covariant @xmath77 one @xcite .
the condition for exciton melting can be estimated as @xmath79 from this , which may be physically verifiable , provided the corresponding temperature is still within the low - energy domain where dirac dispersion is valid .
[ [ section-15 ] ] near two - fermion threshold in graphene , one has @xmath80 , as @xmath81 .
the presence of fermi velocity @xmath57 enhances the temperature effect , making the exciton more vulnerable to melting in grahphene , than in vacuum .
this increases the possibility of experimental observation , given that the crystal does not melt first .
[ [ section-16 ] ] in summary , we have demonstrated that emergent dirac fermions coupled with phonon - induced gauge fields , effecting a dynamic cs gauge lagrangian , yields intrinsically topological novel exciton in planar systems . appearing in the gauge sector , they are charge - neutral spin-1 excitations , ideal for spin - transport in graphene - like systems , where massive dirac modes have been realized
. this exciton is characterized by a parametric threshold , reflecting competition of quantum fluctuations with topological stability .
further , it smoothly disappears at high temperatures .
the dynamics of this exciton can be controlled by an external magnetic field , leading to spin - transport .
the observation of the topological planar exciton will not only demonstrate the gauge - invariant mass of a propagating spin-1 bound state , but will also substantiate the corresponding topological spin .
the charge - neutrality and the topological origin of spin ensures its stability against charge - dependent forces and both quantum and thermal fluctuations .
[ [ acknowledgments ] ] acknowledgments : + + + + + + + + + + + + + + + + the authors would like to thank dr .
vivek m. vyas for valuable inputs .
ka is grateful to prof .
ashok das for many useful discussions and suggestions . |
we have now an overwhelming astrophysical evidence of dark matter . to explain this experimental evidence many theoretical models @xcite introducing new dark forces mediated by new gauge boson have been introduced .
these models propose weakly interacting massive particles ( wimps ) that can annihilate into pairs of dark bosons , which subsequently annihilate to lepton pairs ( protons are kinematically forbidden ) .
one of this model @xcite introduce a new dark sector that couples to the sm with a dark boson ( i.e. the dark photon @xmath1 ) through a small kinetic mixing term . thanks to their large luminosities and low - background environment
the @xmath2-factories offer an ideal place to probe for mev - gev dark matter , complementing searches from lhc .
the dark boson mass is generated via the higgs mechanism , adding a dark higgs boson ( @xmath3 ) to the theory .
the dark photon and the higgs bosons could have a comparable mass ( gev - scale ) .
a very minimal scenario has a single dark photon and a single dark higgs boson .
+ in the analysis @xcite we make use of the @xmath4 process @xmath5 using 521 @xmath6 .
the signal is either fully reconstructed into lepton or pion pair ( exclusively mode ) , or partially reconstructed ( inclusive mode ) .
only two of the three hidden photons are reconstructed in the latter case , and the four - momentum of the third one is identified to that of the recoiling system .
+ in these searches no significant signal is observed . using uniform priors in the cross section upper limits on the @xmath5 cross section
are obtained as a function of the hidden higgs and hidden photon masses . these limits on the cross section
are translated into 90 % upper limit on @xmath7 shown in fig .
[ fig:1 ] , where @xmath8 , @xmath9 is the dark sector gauge coupling , @xmath10 is the mixing strength .
values down to @xmath11 - @xmath12 are excluded for a large range of hidden photon and hidden higgs masses , assuming prompt decay .
assuming @xmath13 , limits on the mixing strength in the range @xmath14 - @xmath15 are derived as shown in fig .
[ fig:2 ] , these limits are an order of magnitude smaller than the current experimental bounds extracted from direct photon production in this mass range . for the dark photon ( left ) and
dark higgs ( right).,title="fig:",scaledwidth=50.0% ] for the dark photon ( left ) and dark higgs ( right).,title="fig:",scaledwidth=47.0% ] , for various higgs masses.,scaledwidth=50.0% ]
in recent years a number of theoretical models @xcite predicted for the existence of a light cp - odd higgs boson @xmath16 related to the next - to - minimal supersymmetric model ( nmssm ) .
direct searches constrain the mass of @xmath17 , where @xmath18 is the b quark and the decay of @xmath19 is forbidden .
of particular interest is to search the lightest cp - odd higgs boson in @xmath20 decays such as @xmath21 , where @xmath22 particles . in these @xmath23 transitions
low mass dark matter candidate ( @xmath24 ) can be also directly produced ( i.e. @xmath25 ) .
the large data sets available at experiment with more than 500 @xmath6 of data at the @xmath26 resonances and just below the @xmath27 resonance allow us to place stringent constraints on such theoretical models .
+ we searched for light higgs in using two - body - radiative decay on @xmath20 states .
the key experimental signature is the monochromatic photon in the cm frame @xmath28 . with this analysis technique
we searched for the following transitions : * @xmath29 @xcite * @xmath30@xcite * @xmath31@xcite in all the searches we did not find any evidence of signals but upper limits versus hypothesis mass have been extracted the results are shown in fig .
[ fig:3 ] for the @xmath32 , in fig .
[ fig:4 ] for the @xmath33 ( where @xmath34 is reconstructed both in @xmath35 and @xmath36 ) and in fig .
[ fig:5 ] for the @xmath37 . , ( b ) @xmath38,scaledwidth=47.0% ] and ( c ) effective coupling @xmath39 as a function of @xmath40 .
the shaded areas show the regions around the @xmath41 and the @xmath42 resonances excluded from the search .
[ fig:3 ] and ( right axis ) @xmath43 for ( a ) cp - all analysis and ( b ) cp - odd analysis.,scaledwidth=47.0% ]
we have searched for evidence of dark sector candidates and evidence of cp - odd light higgs in the @xmath44 and @xmath45 data sample at .
we did not observe any significant signal but more stringent limit have been set on space parameters of np model .
99 b. batell , m. pospelov and a. ritz , phys .
d * 79 * , 115008 ( 2009 ) n. arkani - hamed , d. p. finkbeiner , t. r. slatyer and n. weiner , phys .
d * 79 * , 015014 ( 2009 ) r. essig , p. schuster and n. toro , phys .
d * 80 * , 015003 ( 2009 ) arxiv:1202.1313 [ hep - ex ] ] r. dermisek , j. f. gunion and b. mcelrath , phys .
d * 76 * , 051105 ( 2007 ) b. aubert _ et al .
_ , phys .
lett . * 103 * , 081803 ( 2009 ) b. aubert _ et al .
_ , phys .
lett . * 103 * , 181801 ( 2009 ) j. p. lees _ et al .
_ , phys .
lett . * 107 * , 221803 ( 2011 ) | several types of new - physics models predict the existence of light dark matter candidates and low - mass higgs states .
previous searches for invisible light - higgs decays have excluded large regions of model parameter space .
we present searches for a dark - sector higgs produced in association with a dark gauge boson and searches for a light higgs in @xmath0 decays . |
the old problem of magnetic structure of rare - earth metals and their compounds is still a subject of experimental and theoretical investigations .
these substances have complicated phase diagrams and demonstrate a number of orientational phase transitions . in particular , such transitions take place in the orthoferrites and practically important intermetallic compounds rco@xmath0 ( r = pr , nd , tb , dy , ho ) , see , e.g. , ref .
. qualitative explanation of these transitions has been obtained many years ago within the heisenberg model with inclusion of magnetic anisotropy @xcite . in a number of systems , lattice ( magnetoelastic )
effects are important .
the standard description is usually performed within mean - field approaches .
however , quantitative comparison with experimental data requires a more detailed treatment .
provided that the orientational transition temperature is low ( in comparison with the magnetic ordering point ) , spin - wave theory is applicable @xcite . in the simplest case of the second - order anisotropy
the magnetization lies either along the easy axis or in the easy plane .
inclusion of higher - order anisotropy constants can lead to cone phases where magnetization makes the angle @xmath1 with the @xmath2-axis .
the case @xmath3 was considered in refs . @xcite where the temperature renormalization of anisotropy constants and the spin - wave spectrum in tb and dy within the standard spin - wave theory were calculated . in the present paper
we consider the cone phase with arbitrary @xmath4 .
the situation here is analogous to that for the field - induced orientational phase transitions , e.g. , in the transverse - field ising model ( see ref.@xcite and references therein ) . unlike the latter model
, one can expect that at low enough temperatures and small value of anisotropy the spin - wave theory is applicable for an arbitrary relation between anisotropy parameters , not only close to the orientational phase transition . even for the second - order easy - plane anisotropy , the holstein - primakoff representation for spin operators used in refs .
@xcite leads to so - called kinematical inconsistencies because of incorrect treating on - site kinematical relations . to avoid this difficulty
, we use the technique of spin coherent states .
our approach is to some extent similar to the operator approach used in ref .
@xcite , but gives a possibility to treat more simply higher - order anisotropy constants , as well as to calculate higher - order terms of @xmath5-expansion .
the anisotropic heisenberg model used is formulated in sect 2 . in sect 3
we develop a special form of the @xmath5-expansion which gives a possibility to take into account exactly on - site kinematical relations . in sect 4
we treat the cone - plane transition owing to the temperature dependence of the cone angle @xmath1 and discuss experimental data on the rare earth metals .
we start from the hamiltonian of the heisenberg model with inclusion of single - site anisotropy @xmath6 where @xmath7 is the exchange parameter , @xmath8 are the irreducible tensor operators of second and fourth orders , @xmath9 are the corresponding anisotropy constants
. up to unimportant constant we can rewrite the hamiltonian ( [ h ] ) in the form @xmath10 where @xmath11b_4 ^ 0 \label{rel } \\ d^{\prime } & = & 35b_4 ^ 0 \nonumber\end{aligned}\ ] ] for @xmath12 spins of magnetic ions lie in the easy plane @xmath13 , while for @xmath14 we have the easy axis @xmath15 for @xmath16 a first - order transition takes place between the easy plane ( which is favored by second - order anisotropy ) and easy axis ( which is favoured by large @xmath17 ) .
we consider only the case @xmath18 where the cone phase occurs at intermediate values of @xmath19 , so that spin orientation direction makes the angle @xmath1 with the @xmath2-axis and the orientational phase transitions are of the second order .
this is the case for gd and also for ho , er in low - temperature phases . in the phenomenological approach
it is supposed ( see , e.g. , refs .
@xcite ) @xmath20 where @xmath21 is the isotropic ( @xmath1-independent ) part of the free energy
. then we obtain by minimization of @xmath22@xmath23 so that at the point where @xmath24 the spins become directed in the @xmath13 plane while at @xmath25 they are aligned along the @xmath2-axis .
the temperature dependence of @xmath26 is supposed to have the form @xmath27 with @xmath28 thus at @xmath29 the transition from the easy - plane to cone structure takes place , while at @xmath30 the transition from the cone to easy - axis structure occurs . at the same time
, zener s @xcite resut for the temperature dependence of anisotropy constants in an axially symmetric state with @xmath31 has the form @xmath32 where @xmath33 is the relative magnetization , and @xmath34 @xmath35 are determined by the same relations ( [ rel ] ) with @xmath36 . as pointed in refs .
@xcite , the temperature dependences of anisotropy constants have a more complicated form for the cone structures with @xmath37 ( in fact , only the case @xmath38 is discussed in refs .
@xcite ) .
a systematic way of calculating temperature dependences of anisotropy constants is the @xmath5-expansion which is considered in the next section .
the @xmath5-expansion developed here is slightly different from the standard scheme of @xmath5-expansion @xcite since it gives a possibility to take into account exactly the kinematical relations between powers of spin operators on each site .
we use the coherent state approach ( see , e.g. , ref .
@xcite ) to write down the partition function in the form @xmath39 where @xmath40 is the unit - length vector , @xmath41 and @xmath42 are its polar and azimuthal angles respectively , @xmath43 are the coherent states ( @xmath44 ) .
to construct the @xmath5-expansion we rotate the coordinate system around the @xmath45-axis through the angle @xmath46 the hamiltonian ( [ h ] ) takes the form @xmath47 where @xmath48 are the wigner matrices of the rotation group irreducible representation , @xmath49 - 1)!!}\frac 1{k_l^m}\ ] ] ( @xmath50=m$ ] for @xmath51 even and @xmath50=m+1 $ ] for @xmath51 odd ) , for @xmath52 we have @xmath53 and the tilde sign here and hereafter is referred to the rotated coordinate system . since the partition function ( [ zp ] ) is invariant under rotation of the states @xmath54 it is convenient to use the coherent states defined in the same coordinate system , i.e. , @xmath55with @xmath56 the advantage of using the coherent states is the simple form of the averages of the tensor operators ( [ tens ] ) over @xmath57 by direct calculation we obtain @xmath58 where @xmath59 are the associated legendre polynomials , the factors @xmath60 $ ] take into account properly the kinematical relations on each site . in particular
, the second - order anisotropy term vanishes for @xmath61 and the fourth - order for @xmath62 as it should be ( unlike the results of boson representations in refs.@xcite ) . using ( [ av ] )
we obtain for the case @xmath63 under consideration the result @xmath64 further calculations are performed in the same line as in ref .
representing @xmath65 and expanding in @xmath66 we obtain the @xmath5-expansion of the partition function .
it should be stressed that we retain the factors @xmath67 as well as @xmath68-dependences in ( [ rel ] ) , non - expanded . by performing decouplings , terms of third order
are reduced to linear ones , and terms of fourth order to quadratic ones .
the requirement of absence of @xmath66-linear terms leads to the result for the cone angle @xmath1 @xmath69 \label{teta}\ ] ] where @xmath70 and the `` bare '' magnon spectrum reads @xmath71 , \nonumber\end{aligned}\ ] ] @xmath72 being the energy gap .
the corrections in ( [ teta ] ) can be collected into powers in the same way as in refs.@xcite to obtain the correct description of thermodynamics at not too low temperatures .
( in the presence of higher - order anisotropy this is essential since the coefficients at @xmath73 increase as @xmath74 with anisotropy order . ) then we have @xmath75 \label{teta1}\ ] ] where @xmath76 are the temperature - renormalized anisotropy constants , @xmath77 the relations ( [ anisren ] ) extend the results of refs .
( @xcite ) to the case where spins make a non - zero angle with the @xmath2-axis .
the renormalized gap has the form @xmath78
-196\sin ^2\theta \cos ^2\theta \nonumber \\ & & \ \ \ \ \ \ \ \times \sum_{k,\omega _ n}\left [ \frac{3b_2 ^ 0s_2(j_0-j_{{\bf k% } } ) -10b_4 ^ 0(s_4/s)\delta
_ 0\cos ^2\theta } { \omega _ n^2+s^2(j_0-j_{{\bf k}% } ) ( j_0-j_{{\bf k}}+\delta _ 0/s)}\right ] ^2 \label{spren}\end{aligned}\ ] ] where @xmath79 @xmath80 . after introducing the temperature - renormalized second- and fourth order anisotropy parameters @xmath81 and
@xmath82 @xmath83 the expression for @xmath84 coincides with that of the phenomenological theory ( [ ct ] ) .
collecting again corrections in ( [ spren ] ) into powers , we obtain for the renormalized gap in the notations ( [ rel1 ] ) the expression @xmath85 ^2\end{aligned}\ ] ] which also coincides with that obtained in the phenomenological theory except for the last term . note that at @xmath37 the renormalizations ( [ rel1 ] ) are present even at @xmath86 which should be taken into account when treating experimental data .
now we can pass to description of possible orientational phase transitions .
consider first the case of a small enough constant @xmath87 ( or , equivalently , @xmath88 ) , so that @xmath89 is close to unity .
then @xmath90 increases with temperature and there occurs a transition to the easy - axis phase at the point determined by @xmath91 = 1 \label{pht1}\ ] ] in the opposite case of a large enough @xmath92 @xmath93 is small , and @xmath90 decreases with temperature , so that at the point where @xmath94 a phase transition to the easy - plane phase occurs . thus one can expect that there exists the critical value @xmath95 : for @xmath96 we have a decrease of @xmath97 with @xmath98 and the phase transition from cone to easy - axis phase , while for @xmath99 we have an increase of @xmath97 with @xmath98 and the phase transition from cone to easy - plane phase .
the numerical computations for the simple cubic lattice ( see fig.1 ) yield @xmath100 .
fig.2 shows the corresponding temperature dependences of the anisotropy constants @xmath34 @xmath35 . for simplicity , @xmath101 is taken for the simple cubic lattice .
the phase transitions described by eqs .
( [ pht1 ] ) and ( [ pht2 ] ) are analogous to those in the phenomenological theory of ref .
@xcite that occur at @xmath24 and @xmath102 respectively .
however , unlike the phenomenological approach , microscopical consideration leads to either cone to easy - axis or cone to easy - plane transition with increasing temperature , depending on the zero - temperature value of @xmath46 at the same time , the transition from the easy - plane to easy - axis structure ( through the intermediate cone phase ) can not be explained by purely magnetic renormalizations of anisotropy constants .
the result ( [ teta ] ) gives the mean - field values of the critical exponents ( e.g. , @xmath103 ) for both the ground - state and temperature orientational phase transitions . unlike the systems discussed in ref .
@xcite , the system under consideration has the dynamical critical exponent @xmath104 ( i.e. , single excitation mode with nearly quadratic dispersion is present ) .
thus the upper critical dimensionality for the ground - state qpt is @xmath105 in this respect , the system is analogous to @xmath106 model in the transverse magnetic field @xcite .
a characteristic feature of such systems is the mean - field behavior of critical exponents both above and below the critical dimensionality . for ( hypothetical )
systems with @xmath107 logarithmic corrections to ground - state properties near qpt are present ( see , e.g. , ref .
@xcite ) . at the same time ,
the upper critical dimensionality for the temperature phase transition is @xmath108 and at @xmath109 the temperature - transition critical exponents differ from their mean - field values .
now we discuss the experimental situation . in gd ( see , e.g. , refs .
@xcite ) the orientational phase transition from cone - phase to easy - axis phase is observed at @xmath110k .
the temperature dependence of the cone angle at @xmath111 ( and also of magnetic anisotropy constants ) is non - monotonous , unlike the results obtained in sect 2 .
this complicated situation is connected with the absence of orbital momentum and smallness of anisotropy in gadolinium . in holmium
the low - temperature phase is conical spiral one , the angle of the cone changing from @xmath112 to @xmath113 in the temperature interval @xmath114 the spiral angle makes up about @xmath115 since the sixth - order anisotropy is important , we use the hamiltonian @xcite @xmath116 the hcp lattice is not of a bravais type . however ,
if we neglect the optical mode ( which is possible at @xmath117 k ) one can put ( see , e.g. , ref .
@xcite ) @xmath118 \nonumber \\ & & \ \ \ \ + 2j^{\prime } \cos \frac{q_z}2\left| \exp ( iq_y/\sqrt{3})+2\cos \frac{q_x}2\exp ( -iq_y/2\sqrt{3})\right|\end{aligned}\ ] ] the parameters of the hamiltonian were taken from ref .
@xcite : @xmath119k@xmath120 @xmath121k , @xmath122k , @xmath123k ( note that our value of @xmath124 includes also renormalization due to dipolar anisotropy ) . for simplicity , we restrict ourselves to a collinear magnetic structure ( this is justified by that the spiral angle in the rare earths is small , especially at low temperatures ) .
the calculations with the hamiltonian ( [ hho ] ) are completely analogous to those in the previous section
. calculated dependence of the cone angle is compared with the result of the mean - field approximation and experimental data in fig.3 .
one can see that our results improve somewhat those of the mean - field theory where the temperature dependence of the anisotropy constants is given by ( [ zener ] ) . to conclude
, we have formulated a consistent spin - wave approach to description of thermodynamic properties of anisotropic magnets at low temperatures .
the renormalizations of the anisotropy constants and spin - wave spectrum for an arbitrary cone angle are calculated .
this gives a possibility to describe the orientational phase transition between the cone and plane phases .
the theoretical temperature dependences of the cone angle @xmath97 for @xmath125 and different values of second - order anisotropy : @xmath126 @xmath127 @xmath128 from upper to lower curve .
the value of @xmath129 is @xmath130 fig.3 .
calculated dependences of the cone angle in the mean - field approximation ( short - dashed line ) and renormalized spin - wave theory ( rswt , long - dashed line ) as compared with experimental points for holmium ( refs . | orientational phase transitions are investigated within the heisenberg model with single - site anisotropy .
the temperature dependence of the cone angle is calculated within the spin - wave theory .
the role of the quantum renormalizations of anisotropy constants is discussed .
a comparison with the experimental data on the cone - plane orientational transition in holmium is performed . |
the continuum fermion approach ( cfa ) to regularizing chiral fermions appears to be a promising method . for a summary of results obtained
so far see @xcite .
the basic idea of the approach is the following .
we start from a lattice with spacing @xmath0 .
we call this the original lattice . on this lattice
the simulations are done .
next we construct a finer lattice with lattice spacing @xmath1 , using a suitable interpolation of the gauge field @xcite . on this lattice
we formulate the fermions .
the action for a single fermion of chirality @xmath2 is taken to be @xmath3 where , restricting ourselves to gauge group u(1 ) , @xmath4 \nonumber \\ & \!\!\ !
\!\!\!&\times \;\psi(n \pm \hat{\mu } ) - \psi(n)\},\end{aligned}\ ] ] with @xmath5 being the link variable on the fine lattice , @xmath6 the fractional charge , and @xmath7 .
the effective action is then computed from ( [ la ] ) in the limit @xmath8 , where @xmath0 is kept fixed .
hence the name continuum fermions .
a particular feature of this action or better of the wilson term we have chosen is that the wilson - dirac operator @xmath9 fulfills the ginsparg - wilson relation @xcite : @xmath10 this is not a great surprise .
the motivation behind our construction was to find an action which obeys the index theorem
what , in fact , it does as we shall see below .
the ginsparg - wilson relation ( [ gw ] ) does , however , not guarantee that the theory is invariant under chiral gauge transformations . indeed , the resulting effective action , @xmath11 , is not gauge invariant , even in the limit @xmath8 .
but there exists a local , purely bosonic counterterm @xmath12 , so that @xmath13 is invariant under chiral gauge transformations . for the imaginary part
this is only true in the anomaly - free model .
the counterterm can be and has been computed in perturbation theory . a further important feature of the action ( [ la ] )
is that it has a shift symmetry which makes sure that the ungauged fermion with chirality @xmath14 decouples . in the chiral schwinger model we found for the topologically trivial sector of the theory @xmath15 where @xmath16 and @xmath17 are the effective actions of the corresponding vector model and the free theory , respectively .
the imaginary part of the effective action turned out to be given by the harmonic part of the gauge field , i.e the toron field , alone , and it could be computed analytically from the gauge field we started with .
thus we arrived at an action which can be simulated relatively easily on the original lattice .
we believe to find a similar result in more realistic models in four dimensions .
after so much of introduction , let us now come to our new results .
our first results concern the chiral schwinger model . here
we want to test whether the index theorem is fulfilled .
this is a non - trivial task because any topologically non - trivial gauge field involves at least one singular plaquette .
some authors have argued that the cfa would fail this test .
this would be true had we employed the standard gauged or ungauged wilson terms . furthermore , we will investigate whether our approach reproduces the correct anomaly . in two dimensions
the index theorem says that in a background gauge field configuration of topological charge @xmath18 we should find exactly @xmath19 zero modes of chirality @xmath2 . for a configuration of ,
e.g. , charge @xmath20 this would mean @xmath21 and @xmath22 . to leading order in @xmath1
the wilson - dirac operator can be written @xmath23 where @xmath24 is the average of forward and backward derivatives .
being a finite matix , the dirac operator @xmath25 has the same number of zero modes as the corresponding vector operator , namely @xmath26 , thus violating the index theorem .
but the situation is different for the wilson term @xmath27 .
it has @xmath28 zero modes of chirality @xmath2 and none of chirality @xmath14 , exactly as required by the index theorem . for small , but
finite @xmath1 the zero modes approach the value @xmath29 where @xmath30 is the size of the original lattice .
we consider two configurations of charge @xmath18 .
we denote the link variables on the original lattice by @xmath31 , where @xmath32 are the lattice points on the original lattice .
the first configuration is @xcite ( @xmath33 ) @xmath34 where @xmath35 and @xmath36 .
the second configuration is @xmath37 where @xmath38 is the inverse lattice laplacian .
both configurations have constant field strength @xmath39 , zero toron field , and for @xmath40 they have one singular plaquette at @xmath41 .
the two configurations are related by a periodic ( topologically trivial ) gauge transformation . in the following
we shall take @xmath42 , and we shall use the interpolation given in @xcite .
both configurations give the same value for the effective action ( [ eff ] ) . as the anomaly - free model we took @xmath43 .
this indicates that we have gauge invariance in the background of singular gauge fields as well .
for @xmath44 we find exactly one zero mode with chirality @xmath45 , and none with chirality @xmath46 . for @xmath47
we find two zero modes with chirality @xmath45 , and none with chirality @xmath46 . for negative charges we find the same result but with @xmath45 and @xmath46 interchanged , in agreement with the index theorem .
the values of the zero modes are in good agreement with the analytical result ( [ zero ] ) .
a further requirement of the method is that it reproduces the correct anomaly . in the chiral schwinger model
the anomaly condition reads @xmath48 where @xmath49 is a gauge transformation .
we consider configuration ( [ vm ] ) with @xmath50 and @xmath51 now .
the r.h.s . of ( [ anomaly ] ) is known analytically .
the l.h.s .
is computed for @xmath52 , i.e. on the @xmath53 lattice . in fig .
1 we show the result for both sides separately for @xmath54 and @xmath55 , as well as for the anomaly - free model .
we find excellent agreement between our numerical results and the theoretical expectations ( i.e. the l.h.s . versus the r.h.s . ) .
moreover , we see that the anomaly cancels in the anomaly - free model .
we find similar results for configuration ( [ sv ] ) .
our other new results are on the chiral u(1 ) model in four dimensions . this model is simpler than the chiral schwinger model because it has no topological charge and zero modes to worry about .
the first question is again whether the theory exists at all , i.e. can be made finite by the addition of local counterterms .
we find three possible counterterms : @xmath56 as before , the coefficients , @xmath57 and @xmath58 , can be computed in lattice perturbation theory .
the result so far is @xmath59 we can not quote any number for @xmath58 yet .
numerically we find that gauge invariance is , within our present accuracy , already restored by the counterterms @xmath60 and @xmath61 , if we use the perturbative values ( [ ct ] ) for the coefficients .
this indicates that @xmath58 is very small , or zero . as an upper bound we can quote @xmath62 .
what distinguishes the chiral theory from the corresponding vector theory is the imaginary part , @xmath63 , of the effective action . in two dimensions this
was entirely determined by the toron field contribution . in four dimensions , however , we find that it is zero for toron field configurations .
this leaves the interesting possibility that the imaginary part of the effective action is generally zero in this model .
this work is supported in part by intas grant intas-96 - 370 .
discussions with m. gckeler about the ginsparg - wilson relation are gratefully acknowledged .
v. bornyakov , g. schierholz and a. thimm , in _ theory of elementary particles _ , proceedings of the 31st international symposium ahrenshoop , buckow , 1997 , p.265 ( wiley - vch , berlin , 1998 ) ( hep - lat/9805007 ) ; + v. bornyakov , g. schierholz and a. thimm , desy 98 - 065 ( 1998 ) , to appear in prog .
( hep - lat/9806010 ) . | we report new results on the lattice regularization of the chiral schwinger model and the chiral u(1 ) model in four dimensions in the cfa . |
orion kl ( kleinmann - low object ) is one of the nearest massive star - forming regions at a distance of 420 pc @xcite and hence , has been the most extensively studied source in the context of star - formation study @xcite .
there are powerful outflows and jets possibly associated with young stellar objects ( ysos ) in this region at a wide range of spatial scales from @xmath2100 au to @xmath310000 au .
they are traced by various molecular lines such as vibrationally excited sio masers at 43 ghz @xcite , 22 ghz h@xmath0o maser @xcite , millimeter / submillimeter thermal sio @xcite and co lines @xcite , and near - infrared h@xmath0 line @xcite .
these observational studies have revealed physical properties of the outflows , jets , and surrounding shocked materials interacting with ambient clouds . among a number of tracers ,
the 22 ghz h@xmath0o maser is known to be a unique tracer to investigate a three - dimensional velocity structure .
thanks to its extremely high brightness , it can be detected by high - resolution interferometers and very long baseline interferometers ( vlbi ) providing a capability of proper motion measurements . as a result
, the 22 ghz h@xmath0o maser in orion kl is found to be tracing a low - velocity outflow along a northeast - southwest axis @xcite centered at a strong radio continuum source labeled i @xcite .
hereafter we call this source i. one of these h@xmath0o maser features in orion kl has exhibited an enormous outburst , which is sometimes called supermaser .
the h@xmath0o maser burst in orion kl was reported in 1979 - 1985 for the first time @xcite followed by 1998 - 1999 @xcite .
based on vlbi observations , the supermaser was revealed to have east - west or northwest - southeast elongations in 1979 - 1985 and 1998 - 1999 , respectively .
the line - of - sight velocity with respect to the local standard of rest ( lsr ) was stable at @xmath4 8 km s@xmath5 , suggesting a common origin .
although the maser burst could be related to star - formation activities in orion kl , either a circumstellar disk @xcite or a jet / outflow driven by a yso @xcite , its origin and mechanism are still under debate .
the third burst of the h@xmath0o maser in orion kl has started since february 2011 @xcite .
the burst appears to be possibly periodic with an interval of about 13 years ( 1985 , 1998 , and 2011 ) .
it could support the hypothesis of a common origin @xcite .
we have carried out astrometric monitoring observations with vlbi exploration of radio astrometry ( vera ) since 2011 march @xcite and studied basic properties of the supermaser .
main conclusions by @xcite are ( 1 ) the burst occurs at two spatially nearby but distinct components ( separation of @xmath410 mas ) at the lsr velocities of 6.95 and 7.58 km s@xmath5 ( hereafter we call them `` 6.9 km s@xmath5 '' and `` 7.5 km s@xmath5 '' features , respectively , using representative velocities ) as reported in previous bursts @xcite , ( 2 ) these features are elongated along a northwest - southeast direction perpendicular to the low - velocity outflow and their proper motions are parallel to the outflow axis , and ( 3 ) the absolute position of the supermaser is coincident with a shocked molecular gas called the orion compact ridge .
all of these characteristics suggest that the supermaser could be excited in an interacting region between the low - velocity outflow and the ambient gas in the compact ridge @xcite . nevertheless , the maximum flux density for the current burst phase is reported to be 44000 jy on may 2011 @xcite .
it is much lower than those in previous bursts in which the flux densities of the supermaser reached 10@xmath6 jy @xcite .
although the maser bursts could have possible periodicity of 13 years , 1985 , 1998 , and 2011 , it is still unclear whether all of three events are caused by a common origin . in order to understand their origin and to reveal their relationship between star - formation activities in orion kl in relation to outflows , jets , and interaction with ambient clouds ,
it is important to continue further monitoring observations of the supermaser at a current valuable opportunity . in this paper , we report results of our continued monitoring observations of the supermaser with vera to investigate time variations of their properties .
in addition , we present follow - up observations with atacama large millimeter / submillimeter array ( alma ) to identify possible powering source(s ) , as well as to study pumping mechanisms based on multi - transition h@xmath0o lines .
although we find that the current burst activity is lower by a factor of 10 than those in previous bursts in terms of the maser flux density as discussed later , we call the current bursting maser feature `` supermaser '' throughout the present paper .
observations of the @xmath7-@xmath8 transition of h@xmath0o at 22.235080 ghz @xcite in orion kl were carried out with vera .
details of the observations have been described in separate papers @xcite , and we briefly summarize key parameters in this section .
vlbi monitoring observations have been conducted regularly since 2011 march with a typical interval of 1 - 4 weeks .
the maximum baseline length of vera is 2270 km and the resultant fringe - spacing is 1.2 mas .
we employed the dual beam observation mode , in which orion kl and an icrf source j054138.0 - 054149 @xcite were observed simultaneously , to measure the absolute position of the target h@xmath0o maser in orion kl with respect to the positional reference source .
the astrometric accuracy is estimated to be 0.4 mas according to post - fit residuals of proper motion measurements .
these two sources were observed for 8 hours being tracked from horizon to horizon at the lowest elevation angle of @xmath415 degrees .
a typical on - source time was 6 hours for each epoch . in this paper , we report the astrometric and imaging results of the first 18 epochs of observations from 2011/068 to 2012/269 ( hereafter we indicate an epoch of observation as year / day of the year ) . a spectral resolution was set to be 15.625 khz , corresponding to a velocity resolution of 0.21 km s@xmath5 .
calibration and synthesis imaging were performed using the nrao astronomical image processing system ( aips ) software package in the same manner employed previously @xcite .
we also used other data of vera observations at 22 ghz including orion kl as a fringe - finder to investigate a time variation in spectral profiles .
a total number of spectral line monitoring data is 53 epochs from 2008/329 to 2014/025 . on - source
integration times of each spectral line monitoring data for orion kl was 5 - 10 minutes .
in addition , we used alma data taken in one of the early science projects in cycle 0 ( ads / jao.alma#2011.0.00199.s ) to compare images of submillimeter continuum emissions and molecular lines .
observations were carried out at the alma band 7 ( in the range of 320 - 340 ghz ) on 2012 july 16 , august 25 , and october 21 .
they were almost coincident with the most active burst phase as discussed in the next section . in this paper
, we mainly focus on the @xmath9-@xmath10 h@xmath0o line at 321.225656 ghz @xcite and methyl formate ( hcooch@xmath1 ) lines close to the h@xmath0o line .
the maximum baseline length in our alma observations was 400 m , corresponding to a spatial resolution of @xmath40.5 , depending on the uv weighting in the imaging .
synthesis imaging and self - calibration were done with the common astronomy software applications ( casa ) package .
further details of observations are reported elsewhere @xcite .
figure [ fig - sp ] shows spectra of the supermaser observed in 18 epochs of the regular astrometry observations from 2011 to 2012 , and figure [ fig - time ] shows time variations of the flux density , velocity width , and peak velocity for all the observed epochs including both 18 epochs of the regular astrometry and 53 epochs of the spectral line monitoring observations . first , these parameters are determined by fitting a single gaussian to the observed profile .
however , there are two distinct velocity components at @xmath46.9 km s@xmath5 and @xmath47.5 km s@xmath5 as reported in @xcite .
these components can be clearly seen in spectral profiles such as the double - peaked structure on 2011/121 and 2011/137 and red - shifted shoulder on 2011/208 ( figure [ fig - sp ] ) .
this results in a significant deviation from the observed line profile in the fitting .
thus , we next fit two gaussians to the observed profiles for these epochs . although these line parameters show slightly larger error bars as shown in figure [ fig - time ] , we can see time variations of flux density , velocity width , and peak velocity for the @xmath47.5 km s@xmath5 component continuously changing even during this phase .
in contrast , the velocity component that appeared later at @xmath46.9 km s@xmath5 shows a different time variation . according to characteristics of the time variation in figure [ fig - time ]
, we can divide the monitoring period of the current h@xmath0o maser burst into four phases ; a moderately active phase in 2009 followed by a relatively quiescent phase in 2010 , a start of the burst with rapid variation in 2011 , the maximum phase in 2012 , and a quiescent / possible initial phase of the next burst in 2013 . in 2009 , there is a moderately active phase slowly increasing and decreasing in the flux density both in total power and cross power spectra .
timescales of both rising and falling phases , which are defined as periods from minimum to maximum and maximum to minimum flux densities , respectively , are almost equal , 10 months and 11 months , respectively ( figure [ fig - flux](a ) ) . during this phase , the peak velocity is also slightly changing from 7.4 to 7.3 km s@xmath5 while the velocity width , @xmath40.8 km s@xmath5 , is almost stable with time .
after this moderately active phase in 2009 , the maser activity becomes a quiescent phase , whereas the total and correlated flux densities are as high as 2400 jy and 500 jy , respectively , at the minimum phase on 2010/227 . toward this minimum phase ,
the velocity width of the total - power spectrum gradually increases although that of the cross - power spectrum shows a decreasing trend . at the minimum phase on 2010/227 , the velocity width and peak velocity change drastically .
furthermore , these parameters show significant differences between total - power and cross - power spectra but with larger uncertainties than other epochs .
this could be caused by other velocity components as their contributions become more and more significant relative to that of the 7.5 km s@xmath5 component . before and after this minimum phase , there were breaks of monitoring observations each for 100 days due to the maintenance of the vera antennas , and hence , we could not trace time variations in this quiescent phase .
the h@xmath0o maser burst has started in the end of 2010 or the beginning of 2011 @xcite .
the flux density of the total - power spectrum at @xmath47.5 km s@xmath5 has started to increase since 2010/350 while that of the cross power spectrum is temporarily halted on 2010/360 followed by the brightening from 2011/023 . although there is a lack of maser images , this might reflect a variation in the spatial structure of the emitting region . at the beginning of the current burst phase in 2011
, there are two maximum phases on 2011/60 - 2011/121 and 2011/296 - 2011/311 with the total flux density of about 50000 jy ( figure [ fig - time ] ) .
the timescales of rising and falling phases are 7 months and 3 months , respectively , for the maximum phase of 2011/60 - 2011/121 ( figure [ fig - flux](b ) ) , although the flux density in this phase shows a variability rather than a monotonic increase or decrease .
the timescales for the subsequent maximum phase of 2011/296 - 2011/311 ( figure [ fig - flux](c ) ) are 2 months for both rising and falling phases .
apart from the different timescales , the behaviors in rising and falling phases seem to be symmetric as seen in the active phase in 2009 ( figure [ fig - flux ] ) . as discussed above , the burst appears to occur in two different velocity components at 6.9 km s@xmath5 and 7.5 km s@xmath5 @xcite as seen in figures [ fig - time](b ) and ( c ) .
the burst activity is temporarily stopped at the end of 2011 , but more active burst has started in early 2012 .
the maximum total flux density of 135000 jy is recorded on 2012/154 at the peak velocity of 7.4 km s@xmath5 . in this most active phase , the flux density monotonically increases with the timescale of 5 months .
the velocity width is about 0.60 - 0.75 km s@xmath5 , which is the narrowest value observed in our monitoring as shown in figure [ fig - time](b ) .
on 2012/123 , the derived velocity widths show the minimum values of [email protected] km s@xmath5 and [email protected] km s@xmath5 for the total - power and cross - power spectra , respectively . only in this epoch ,
the velocity widths are significantly narrower than those observed just before and after this epoch ( 2012/107 and 2012/145 ) . due to the lack of maser images on 2012/123 , we do not discuss what is the true reason for these narrower velocity widths only at this epoch .
the flux density of the maser decreases with the timescale of 4 months after this maximum phase .
it should be noted that the time variations in rising and falling phases again seem to be symmetric as seen in figure [ fig - flux](d ) . at the end of the burst phase in 2012/265 and 2012/269 ,
the peak velocities of cross - power spectra show significant differences from those of the total - power spectra .
they are thought to be the results of significant contributions from other velocity components and/or effects of structure changes in maser features .
this effect is seen particularly in less active phases such as in 2010 .
finally , the burst activity seems to end around 2012/269 while it again appears to be brightening from 2013/138 .
it is most likely an ignition of a newly occurred burst event .
we will continue monitoring observations with vera of the time variation of the supermaser in orion kl .
still , the activity of the 22 ghz h@xmath0o maser burst in 2011 - 2012 is relatively lower by a factor of 10 in comparison with the previous bursts in which the peak flux densities reached 10@xmath6 jy .
we note that the fluxes reported in figures [ fig - time ] and [ fig - flux ] are those integrated over the whole region including both the supermaser and other features .
we investigate possible contribution from other maser features at almost the same velocity with the supermaser feature in the 18 epochs of our vlbi imaging .
although it is difficult to make full channel maps with a wide field of view of an order of 10 , we search for positions of the maser features detected by @xcite . as a result , we detect one maser feature around the position offset of ( @xmath43.175 , @xmath4 - 1.850 ) in right ascension and declination , respectively , with respect to the supermaser feature .
@xcite detected two maser features near this position at the velocities of 6.3 km s@xmath5 and 7.6 km s@xmath5 with the flux densities of 21.0 jy and 897.7 jy , respectively .
figure [ fig - se ] shows the flux density of this maser feature ( we call this feature `` se feature '' ) and the flux density ratio relative to the supermaser feature .
the flux density of the se feature increases with time and the maximum phase is observed on 2012/244 .
it is about 2 months after the maximum phase of the supermaser . even around the maximum phase of the se feature ,
a possible contribution to the total flux density is as large as 8% as shown in figure [ fig - se ] ( b ) .
therefore , we conclude that the contamination from other maser features including the se features are not significant in figures [ fig - time ] and [ fig - flux ] . the time variation of the linewidth of the supermaser implies an anti - correlation with the flux density as shown in figure [ fig - dv ] .
according to model calculations , the line profile of the maser first tends to be narrower as its intensity increases as long as the maser is unsaturated @xcite .
then the maser line starts to rebroaden when it becomes saturated @xcite . under the condition of an unsaturated maser
, the logarithm of the flux density of the maser line , @xmath12 , is predicted to be proportional to the inverse square of the linewidth , @xmath13 . in figure [ fig - dv ] , we also show best - fit results ; @xmath14 our results may suggest that the supermaser in the current burst phase is still unsaturated .
it is consistent with the observed narrow linewidths , 0.40 - 0.48 km s@xmath5 , of the supermaser in 1998 - 1999 possibly attributed to the line narrowing for an unsaturated maser @xcite .
in contrast , no clear relationship between the flux density and linewidth is found in the burst in 1979 - 1985 @xcite , suggesting a saturated maser emission .
the difference in the degree of saturation could be attributed to the lower flux density , @xmath410@xmath15 jy in the current burst phase than those in the previous bursts , @xmath310@xmath6 jy @xcite .
we note that the line narrowing is predicted to occur even for a saturated maser if a stimulated emission rate is not large compared with the cross - relaxation rate @xcite .
this means that the above relation holds when the maser flux density is not enough large regardless of saturated or unsaturated cases .
thus , we can not firmly estimate the degree of saturation of the supermaser by judging from the linewidth - flux density relation only .
figure [ fig - map ] represents channel maps of the phase - referenced images of the supermaser in 2011/137 .
three spatially distinct components are found in the initial phase of the burst from 2011/068 to 2011/137 as reported previously @xcite .
it is well known that the 22 ghz h@xmath0o line ( 6@xmath16 - 5@xmath17 ) has a hyperfine structure with 3 main components and that the @xmath18=7 - 6 and 5 - 4 components appear with the separation of + 0.45 km s@xmath5 and -0.58 km s@xmath5 with respect to the central component , @xmath18=6 - 5 @xcite .
the velocity separation of -0.58 km s@xmath5 is consistent with that observed for two bursting features .
however , our results suggest that the two velocity components observed at 6.9 km s@xmath5 and 7.5 km s@xmath5 in the supermaser are not emitted from the same masing gas clump . thus , these two velocity components are unlikely due to the hyperfine structure but they corresponds to distinct masing gas clumps with different positions and velocities .
we conduct two - dimensional gaussian fittings on the phase - referenced images to measure the absolute position of each maser spot .
hereafter , we define a `` spot '' as a maser emission identified in a single velocity channel and a `` feature '' as a group of spots that are spatially coincident in two or more consecutive velocity channels . by using these results , we derive the absolute proper motions of the detected spots as summarized in table [ tab - proper ] and figure [ fig - position ] .
we find that the northern feature , feature 1 ( or spots 1 , 3 , and 5 ) , is detectable only until 2011/296 ( table [ tab - proper ] ) .
the lifetime of the northern feature 1 is about 8 months . given the typical monitoring interval and detection criterion , in which a spot is regarded as real when it is detected for more than three contiguous epochs , we might miss a short - term variation of a maser spot on phase - referenced images within a lifetime of less than @xmath42 months . according to the absolute proper motion measurements
, the supermaser features are apparently moving toward south and southwest directions as shown in the gray arrows in figure [ fig - map ] .
the derived proper motions are consistent but are more reliable than those previously obtained @xcite because we have continued the proper motion measurements in a longer observing period . when we subtract a modulation of an annual parallax of 2.39 mas @xcite and the absolute proper motion of a possible powering source of the low - velocity outflow , source i ( @xmath19 , @xmath20)=(6.3 mas yr@xmath5 , -4.2 mas yr@xmath5 ) , measured by astrometric observations with the very large array ( vla ) @xcite , the derived proper motion vectors with respect to source i point toward west and southwest directions as drawn in the black arrows in figure [ fig - map ] . the magnitude of the proper motions range from 2.7 to 11.4 mas yr@xmath5 or from 5 to 23 km s@xmath5 ( table [ tab - proper ] ) .
similar to the previous reports @xcite the supermaser features are elongated along the northwest - southeast direction perpendicular to the low - velocity outflow @xcite . observed proper motions with respect to source i are almost perpendicular to the major axis of each feature .
the size of the maser spot at the peak channel on the maximum epoch , 2012/164 , is [email protected] [email protected]@xmath110.006 mas
( 0.95 [email protected] au ) obtained from a gaussian fitting on the self - calibrated image .
the peak flux density is ( [email protected])@xmath22 jy , corresponding to the brightness temperature of 1.2@xmath23 k. the isotropic luminosity is estimated to be 2.4@xmath2110@xmath24 by assuming a linewidth of 0.6 km s@xmath5 . to study physical and chemical properties of the supermaser , we utilize alma cycle 0 data to make images of submillimeter continuum emissions and molecular lines at band 7 .
first , we have searched for any 321.225656 ghz h@xmath0o line signal toward the compact ridge where the supermaser resides for the 321.20 - 321.25 ghz range , as shown in figure [ fig - alma ] .
several spectral lines are seen , but they are identified as the hcooch@xmath1 lines . if the 321 ghz h@xmath0o line is excited by the maser action like that at 22 ghz , it should be much more compact than those of other molecular lines . with this in mind , we then investigate spectra of the h@xmath0o lines by using the visibility data with the baseline length longer than 100 k@xmath25 , corresponding to the angular size of the emission region of @xmath22 . as a result
, we can resolve out all of the lines in the compact ridge while a significant fraction of the 321 ghz h@xmath0o line .
in contrast , the broad 321 ghz h@xmath0o line is clearly seen in source i with a double peaked profile @xcite .
the 321 ghz h@xmath0o line remains in source i even with the baselines longer than 200 k@xmath25 suggesting a compact structure @xcite .
thus , we conclude that no clear evidence of the 321 ghz h@xmath0o maser is detected at the position of the supermaser in spite of the extraordinary strong emission of the 22 ghz h@xmath0o maser .
we note that in the observed spectral range a vibrationally excited ( @xmath26=1 , 5@xmath27 - 6@xmath28 ) h@xmath0o line at 336.227931 ghz is not detected in the compact ridge .
non - detection of the 321 ghz and 336 ghz h@xmath0o lines implies that the physical conditions around the compact ridge are certainly different from those in source i @xcite .
we next compare the distribution of the molecular gas in the compact ridge which hosts the supermaser . for this purpose
, we employ the hcooch@xmath1 line at 321.22338 ghz ( @xmath29=0 , 28@xmath30 - 27@xmath31 a ) , which is close to the frequency of the 321 ghz h@xmath0o line ( figure [ fig - alma ] ) , as a tracer of the dense molecular gas in the compact ridge @xcite . the 321.223 ghz hcooch@xmath1 line
is not blended with other hcooch@xmath1 transitions unlike those at 321.229 ghz ( @xmath29=1 ) lines , and hence , it is more suitable to analyze the velocity structure of the molecular gas .
figure [ fig - moment](a ) shows the moment 0 ( integrated intensity ) map of the 321.223 ghz hcooch@xmath1 line superposed on the alma band 7 continuum emission .
the supermaser is located at 8 southwest of source i and is coincident with the compact ridge .
the position of the supermaser is close to the hcooch@xmath1 peak but is rather located at the edge of the hcooch@xmath1 condensation . as can be seen in the moment 1 ( peak velocity ) map in figure [ fig - moment ] ( b ) ,
the velocity of the supermaser @xmath46.9 - 7.5 km s@xmath5 is slightly blue - shifted compared with that of the hcooch@xmath1 line .
furthermore , the moment 2 ( velocity width ) map in figure [ fig - moment ] ( c ) implies that the linewidth of the hcooch@xmath1 line tends to be broader around the supermaser .
we stress that the velocity structure seems to change abruptly around the supermaser .
these apparent velocity structures are caused by two different velocity components of the molecular gas traced by the hcooch@xmath1 line . in figure [ fig - pv ] , we can see detailed velocity structures in the compact ridge by the position - velocity diagram along the x and y axes as indicated in figure [ fig - moment ] ( a ) .
it is clear that there exist two different velocity components at 7.5 km s@xmath5 and 9.0 km s@xmath5 .
these two components were identified in the previous observations @xcite , but we can resolve both the velocity and spatial structures by higher spectral and spatial resolutions with alma .
the supermaser is thought to be associated with one of them at 7.5 km s@xmath5 .
interestingly , the velocity of the weak emission from the hcooch@xmath1 line in figure [ fig - pv ] ranges from @xmath46 to 12 km s@xmath5 , although the linewidths of the hcooch@xmath1 lines are only about 1 km s@xmath5 as shown in figure [ fig - alma ] .
the velocity range of the weak hcooch@xmath1 emission is significantly broader at x=0 and y=0 , corresponding to the position of the supermaser ( see figure [ fig - pv ] ) .
this is probably an evidence of a shock front at the position close to the supermaser . indeed ,
possible effects of shocks in the compact ridge have been suggested by the enhancement of organic molecules such as hcooh @xcite , hcooch@xmath1 @xcite , and ch@xmath1och@xmath1 @xcite .
figure [ fig - cont ] shows a close - up view of distributions of the alma band 7 continuum emission superposed on the integrated intensity map of the 321.223 ghz hcooch@xmath1 line ( figure [ fig - moment](a ) ) .
positions of the 22 ghz h@xmath0o masers @xcite including that of the supermaser are also indicated .
we detect the compact continuum emission both at alma bands 6 and 7 @xcite .
the accurate position of the supermaser is significantly shifted from the continuum peak by 0.3 southeast .
we also indicate positions of other possible counterparts observed previously .
their relationship will be discussed in the next section .
in this paper , we report results of continued monitoring observations of the supermaser with vera for more than 5 years .
we also acquire follow - up observational data with alma of submillimeter continuum emissions and h@xmath0o lines at 321 ghz to reveal characteristics of the supermaser and their environments . in this section
, we will address following issues related to the supermaser : what is the powering source of the supermaser and what makes the 6.9 and 7.5 km s@xmath5 features in the compact ridge only such a special burst event ? why there is no bursting h@xmath0o maser feature at 321 ghz despite the extraordinary high flux density in the 22 ghz h@xmath0o maser ? how the drastic burst of the h@xmath0o maser can be explained ?
is there a periodicity of the burst event ? according to the absolute position of the supermaser derived from our vera observations , we propose that the supermaser would be excited in the shocked molecular gas interacting between the low - velocity outflow driven by source i and ambient dense gas in the compact ridge @xcite .
proper motions and elongations of the supermaser features ( figure [ fig - map ] ) and the broader hcooch@xmath1 line at the position of the supermaser ( figure [ fig - pv ] ) support this scenario .
it has also been suggested by the fact that the h@xmath0o maser tends to be distributed in the interface of the shocked molecular gas in the compact ridge traced by hcooh @xcite and hcooch@xmath1 @xcite lines .
furthermore , @xcite point out a possible relation between the compact ridge and an explosive outflow traced by the co lines @xcite .
indeed , the supermaser is found to be located at the edge of the low - velocity bipolar molecular outflow traced by the thermal sio lines @xcite as well as the explosive outflow @xcite .
the latter one is thought to be the result of a recent close encounter of source i and bn object @xcite , although a relationship between these two outflows is still unclear .
note that an infrared source irc 5 is located at 1 southeast of the supermaser .
the position of irc 5 corresponds to a nearby 3 mm continuum peak , c29 @xcite , and an hcooch@xmath1 peak , mf14 @xcite and hence , it is more closely related to these sources . since irc 5 is thought to be a shocked molecular gas heated externally @xcite , it would not be directly related to the excitation of the supermaser .
alternatively , both of them are thought to be heated by the common outflow . as discussed previously @xcite , however , it has been still unknown why only the 6.9 km s@xmath5 and 7.5 km s@xmath5 features show such an extraordinary burst phenomenon .
one of the possible explanations is a special physical or geometrical condition of the maser excitation in this region .
an anomalous amplification of the h@xmath0o maser burst might need a long path length .
such a condition could be achieved only in the systemic velocity component . because it can supply plenty of ambient molecular gas within a narrow velocity range
, maser emission would be efficiently amplified only close to the systemic velocity @xcite .
millimeter and submillimeter continuum emissions are detected by @xcite , smm1 in @xcite and cb1 in @xcite with spatial resolution of @xmath40.8 - 3 .
the mass of the submillimeter core in the compact ridge is estimated to be 4@xmath32 @xcite , and hence , they could provide the dense gas required to form new stars .
a similar conclusion is obtained from the alma science verification continuum data that have been analyzed by @xcite . to account for a necessary condition for the supermaser
, it is also suggested that the maser burst could occur as a result of an interaction with outflows and a pre - existing yso in the compact ridge @xcite . as discussed in detail by @xcite
, there are possible candidate sources physically associated with the supermaser ; a radio continuum source labeled r detected with vla @xcite , 3 mm continuum emission sources detected with the combined array for research in millimeter - wave astronomy ( carma ) @xcite , molecular peaks traced by hcooch@xmath1 lines with the iram plateau de bure interferometer ( pdbi ) @xcite , and optical source parenago 1822 , which is also identified from millimeter , near infrared , optical and x - ray ( see simbad ) , in the compact ridge ( figure [ fig - cont ] ) .
@xcite detect a compact 1.3 mm continuum source named hc 438 in the compact ridge with a flux density of [email protected] mjy by using carma .
we also detect a compact continuum source at alma band 7 ( see contours in the middle of figure [ fig - cont ] ) which is also detected at band 6 @xcite .
these observations achieve much higher spatial resolution of @xmath40.5 than those reported above @xcite .
the flux density at alma bands 6 and 7 are 57@xmath115 mjy and 129@xmath1114 mjy , respectively @xcite .
the alma band 6 result is consistent with that of the carma 230 ghz observation @xcite .
the supermaser is located within the submillimeter core detected by alma but is shifted by 0.3 from the peak position . taking into account positional uncertainties , they are identical to hc 438 @xcite , the 3 mm continuum source c32 @xcite and hcooch@xmath1 peak mf1 @xcite .
the circumstellar mass of hc 438 is derived to be 0.085 @xmath32 with the assumed dust temperature of 20 k @xcite . if we assume a higher temperature of 100 k derived from the hcooch@xmath1 data @xcite , the smaller mass of 0.01@xmath32 is obtained .
these mass estimates would be a lower limit of the dust continuum source possibly associated with the supermaser because most of the flux is filtered out by alma and carma with extended configurations .
we speculate that an embedded yso associated with jet / outflow / disk @xcite would play an important role to realize special conditions required for the supermaser .
we notice that the h@xmath0o maser features near the supermaser detected with vla @xcite are distributed around the submillimeter continuum peak as seen in figure [ fig - cont ] .
it may imply an expanding shell ejected by an embedded yso , which are sometimes seen in massive ysos @xcite . as discussed in section [ sec - vera ] , we tried to search for other maser features in orion kl region previously detected with vla @xcite .
we made channel maps of the h@xmath0o maser features with vera within an 1@xmath211 field of view around the supermaser in the compact ridge .
however , we found no maser emission because of the limited dynamic range seriously affected by the extremely strong supermaser .
further observations of continuum emission and h@xmath0o masers with higher dynamic range will unveil the exact location of the embedded yso in the compact ridge .
it is proposed that the pumping mechanism of the h@xmath0o maser is dominated by collisional excitation @xcite .
the burst of the h@xmath0o maser in orion kl is proposed as the result of collisional pumping in the magnetohydrodynamic shock wave caused by the outflow @xcite . on the other hand ,
alternative mechanisms are also considered to account for the extraordinary h@xmath0o maser burst , such as overlapping of two maser gas clumps @xcite and accidental beaming of radiation toward the earth rather than a change in excitation @xcite . to constrain the physical properties of the supermaser ,
we have carried out observations of the submillimeter h@xmath0o lines which could be masing @xcite by using alma . among a number of h@xmath0o lines ,
we employ the 10@xmath33 - 9@xmath34 transition at 321 ghz which is proposed to be a good indicator of some key parameters including temperature , density , and h@xmath0o abundance in the masing gas clump @xcite . however , there is no significant emission of the 321 ghz h@xmath0o line at the position around the supermaser .
the upper limit of the flux density of the 321 ghz h@xmath0o line is @xmath40.1 jy .
if we assume physical properties of the masing gas are common for both the 22 ghz and 321 ghz lines , then a photon luminosity ratio , @xmath35(22 ghz)/@xmath35(321 ghz ) , is identical to the observed flux density ratio , @xmath18(22 ghz)/@xmath18(321 ghz ) .
then , an upper limit of @xmath35(22 ghz)/@xmath35(321 ghz)@xmath310@xmath6 is obtained from the flux density of the 22 ghz maser of @xmath310@xmath15 jy . when the beaming angle of the 22 ghz maser is also the same as that of 321 ghz , an emissivity ratio @xmath36(22 ghz)/@xmath37(321 ghz ) @xcite is also equal to the photon luminosity ratio and flux density ratio , @xmath38 . as demonstrated in the model calculations @xcite , the lower limit of @xmath38
can be achieved when the temperature is @xmath4200 - 400 k or the parameter @xmath39 is @xmath40.4 - 1.8 depending on the temperature within the calculated range below 2000 k ( see their figure 3 ) .
the parameter @xmath40 is defined as @xmath41(h@xmath0o ) @xmath42/@xmath43 , in which @xmath44(h@xmath0o ) , @xmath45 , and @xmath43 are a fractional abundance of h@xmath0o relative to hydrogen nuclei in units of 10@xmath46 , a number density of hydrogen nuclei in units of 10@xmath47 @xmath48 , and a velocity gradient in units of 10@xmath49 cm s@xmath5 cm@xmath5 @xcite .
we note that the model employed here is calculated under the assumption of a saturated maser @xcite and hence , the estimated parameters may be different from an unsaturated case as adopted for the supermaser .
furthermore , the model calculation only considers collisional pumping of h@xmath0o molecules ignoring infrared pumping .
in addition , the collision rates used in these works are now superseded by newer calculations @xcite . despite these uncertainties ,
we use the model of @xcite to roughly interpret the supermaser burst at 22 ghz and the non - detection of the 321 ghz line . as a result , we find the supermaser burst at 22 ghz and the non - detection at 321 ghz could still be explained by the theoretical model @xcite without extraordinary physical conditions .
thus , under the condition of the unsaturated 22 ghz h@xmath0o maser , the lower temperature and larger value of @xmath40 would also be expected according to a qualitative consideration .
to explain the non - detection of the 321 ghz maser line ( high value of @xmath51 ) , either high h@xmath0o abundance , high h@xmath0 density , small velocity gradient , or low temperature of 200 - 400 k would be required .
the h@xmath0o abundance in the compact ridge is estimated from the herschel data to be 2.6@xmath2110@xmath52 with the source size of 6 @xcite .
if a local abundance enhancement occurs around the supermaser , the higher h@xmath0o abundance up to 6.5@xmath2110@xmath46 as observed in the hot core region @xcite may be possible .
the average h@xmath0 number density of the dust continuum source in the compact ridge is estimated to be ( 0.5 - 5)@xmath2110@xmath47 @xmath48 assuming a diameter ( fwhm ) of 0.43 or 90 au @xcite and a total mass of 0.01 - 0.1 @xmath32 @xcite .
if we assume a velocity width of 0.6 km s@xmath5 and a path length of 1 au as obtained from a gaussian fitting for a bursting maser spot , the velocity gradient @xmath43 is calculated to be 0.4 .
although the above results contain large uncertainties by a factor of 10 or more , these parameters seem reasonable to explain estimated range of @xmath39 of 0.4 - 1.8 unless the h@xmath0o abundance is the lowest estimate of @xmath44(h@xmath0o)=0.026 @xcite .
it is also conceivable that the supermaser is the result of strong beaming effects @xcite .
if only the 22 ghz maser emission has a smaller beaming angle compared with that of the 321 ghz line , the emissivity ratio @xmath53 is proportional to the flux density ratio , @xmath18(22 ghz)/@xmath18(321 ghz ) , and the ratio of beaming angles , @xmath54(22 ghz)/@xmath54(321 ghz ) , which reduces the upper limit of @xmath53 estimated from the observed flux density ratio .
the smaller beaming angle of the 22 ghz maser line than that of 321 ghz by a factor of @xmath4100 would enable a large @xmath54(22 ghz)/@xmath54(321 ghz ) value of @xmath410000 .
this may result in the lower @xmath53 value by a factor of 10000 which can be achieved under wider range of physical conditions as seen in figure 3 of @xcite .
in spite of the presence or absence of a beaming effect , the condition for the h@xmath0o maser burst only at the 22 ghz transition , @xmath38 , does not require extreme physical properties but rather those typical in hot molecular cores @xcite .
we suggest that the symmetric time variation in the flux density ( figure [ fig - flux ] ) could be explained by a change in the beaming angle of the maser rather than changes in the physical properties such as h@xmath0o abundance , h@xmath0 density , and/or temperature . on the other hand ,
a change in a velocity gradient caused by a length of the maser cloud and/or velocity structure might happen such as those in the overlapping maser model @xcite .
nevertheless , it is still uncertain how such an accidental overlapping and/or a beaming effect could occur in two spatially distinct clouds for several times during our monitoring period of @xmath45 years . comparing the basic properties such as the peak lsr velocity , the spatial structure of each feature , and the location coincident with the compact ridge , all of the burst events in 1979 - 1985 , 1998 - 1999 , and 2011 - 2012 could have a common origin .
it is also suggested that the h@xmath0o maser burst may have periodicity of 13 years @xcite .
however , the current burst shows less activities in terms of the maximum flux density .
furthermore , no such burst event is reported in 1973 @xcite , in which the burst event is expected if 13 year - periodicity is valid ( i.e. 1972 - 1973 , 1979 - 1985 , 1998 - 1999 , and 2011 - 2012 ) . in 1973
july , the observed flux density of the h@xmath0o maser at 7.4 km s@xmath5 was only about 100 jy and was weaker than the 11.0 km s@xmath5 component @xcite .
this component was probably identical to the second brightest component in the current burst phase ( figure [ fig - sp ] ) .
the position of the 7.4 km s@xmath5 component was measured to be r.a.=05h35m14.22s and [email protected] ( j2000 ) and was consistent with that of the current bursting feature located in the compact ridge . in 1975 - 1976 ,
the h@xmath0o maser feature at 7.6 km s@xmath5 was detected at the position of r.a.=05h35m14.07s and [email protected] ( j2000 ) .
the position of this feature was again in good agreement with that of the current h@xmath0o maser burst , whereas it was rather weak ( much less than 1000 jy ) compared with another strong feature at 10.8 km s@xmath5 .
although the h@xmath0o maser feature at @xmath47.5 km s@xmath5 could be excited by a common origin in the compact ridge , it is not very likely that the maser burst occurs regularly at a period of 13 years .
more archive data of past observations will be helpful to investigate a long - term variability of the h@xmath0o maser activities since early 1970s .
the flux increase started in 2013 is now on - going and hence , further monitoring will also provide some hints to understand a possible relation with previous bursts .
along with the 22 ghz h@xmath0o maser lines , it could be useful to continue monitoring observations of the 321 ghz h@xmath0o lines in orion kl with alma in order to see whether the compact ridge and source i , which is a powering source of the outflow interacting with the compact ridge , behave in the same way .
figures [ fig - sp]-[fig - flux ] imply that the current burst in 2011 and 2012 sometimes shows variations in the flux density and velocity structure with the timescales of 2 - 7 months .
it is comparable to those of typical h@xmath0o masers in orion kl @xcite .
although the spectral component of the supermaser seems to survive for more than 5 - 10 years @xcite , a spatial structure of each bursting feature is found to be changing within a timescale of a few months to @xmath22 years at the longest lifetime ( table [ tab - proper ] ) .
thus , it is unlikely that the bulk of the bursting maser gas is identical for longer than @xmath42 years .
we have carried out astrometric monitoring observations of the bursting 22 ghz h@xmath0o maser feature , called supermaser , in orion kl using vera since 2011 march . combined with spectral line data obtained as a fringe - finder ,
we reveal a time variation of the supermaser for more than 5 years .
in addition , we utilize alma cycle 0 data at band 7 ( 320 - 330 ghz ) for both spectral lines and continuum emissions to investigate physical and chemical properties of the supermaser . the main findings are summarized as follows : 1 . within the monitoring period in 2011 and 2012
, we detect three active phases with rising and falling timescales ranging from 2 to 7 months .
we also find a smaller burst in 2009 with a slightly longer rising and falling timescales of @xmath410 months .
the flux variations of the supermaser seem to be symmetric in time for all of the active phases . 2 .
the maximum phase of the current burst is observed on 2012/154 with the total flux density of 135000 jy , which is one order of magnitude lower than those in previous bursts .
although the burst events may occur with a possible periodicity of 13 years , ( i.e. 1979 - 1985 , 1998 - 1999 , and 2011 - 2012 ) , a lower flux density by a factor of 10 in the current burst , as well as no significant burst event in 1973 @xcite would make the claimed periodicity of 13 years still unconvincing .
the outburst of the maser emission can be seen in two velocity components at @xmath46.9 km s@xmath5 and @xmath47.5 km s@xmath5 as reported in @xcite and @xcite .
the 6.9 km s@xmath5 feature is alive only in 2011 with the lifetime of @xmath48 months .
note that we detect two spatially distinct features at 6.9 km s@xmath5 and 7.5 km s@xmath5 , and hence , these two velocity components are unlikely due to the hyperfine structure of the h@xmath0o molecule .
4 . the flux density and linewidth show an anti - correlation as expected for an unsaturated maser emission @xcite , in which the logarithm of the flux density of the maser line , @xmath12 , is proportional to the inverse square of the linewidth , @xmath13 .
however , we can not rule out a possibility of a saturated maser case because such a line narrowing is also predicted to occur even for a saturated maser if a stimulated emission rate is not large compared with that of the cross - relaxation @xcite .
as reported in previous papers @xcite the bursting maser features are elongated along the northwest - southeast direction perpendicular to the low - velocity outflow driven by source i. absolute proper motions of the bursting maser features are measured based on a longer monitoring period than that in the previous paper @xcite .
their proper motions with respect to a powering source of the low - velocity outflow , source i , are almost perpendicular to the elongation of the maser features , suggesting that they are related to the low - velocity outflow .
the size of the maser spot at the peak channel on the maximum epoch , 2012/164 , is 2.265 [email protected] mas , and the total flux density is 1.029@xmath2110@xmath15 jy .
these values correspond to the brightness temperature and the isotropic luminosity of 1.2@xmath23 k and 2.4@xmath2110@xmath24 , respectively .
the supermaser is located at the compact ridge region in orion kl .
continuum peaks detected by alma bands 6 and 7 are located 0.3 northwest of the supermaser . as suggested by @xcite
, there are a radio continuum source labeled r @xcite , 3 mm continuum source c32 @xcite , hcooch@xmath1 core mf1 @xcite , and optical source parenago 1822 , which are probably identical or closely related to the compact continuum sources detected by alma and the supermaser .
the position of the supermaser is coincident with the dense gas traced by the hcooch@xmath1 line .
the velocity of the supermaser agrees well with one of the two velocity components of the hcooch@xmath1 line detected by alma and pdbi @xcite .
a slightly broader velocity range of the weak hcooch@xmath1 emission would imply a shock front at the position of the supermaser .
the 321 ghz h@xmath0o line is not detected around the 22 ghz supermaser features , although the strong maser emission is found to be associated with source i @xcite .
given the high emissivity ratio , the maser burst at 22 ghz could occur under conditions of the temperature of @xmath2200 - 400 k and/or the parameter @xmath56 , suggesting a high abundance of h@xmath0o , a high number density of h@xmath0 , and/or a small velocity gradient .
meanwhile , these physical properties are not extraordinary values but are very similar to typical hot molecular cloud cores .
thus , the supermaser burst at 22 ghz with non - detection of the 321 ghz maser emission could be explained by the theoretical model qualitatively @xcite .
further detailed calculations would be required to interpret the observed flux ratios of the masers @xcite .
alternatively , the supermaser may be caused by an accidental beaming only for the 22 ghz h@xmath0o maser @xcite .
the symmetric time variation in the flux density is consistent with a result of the beaming rather than changes in physical properties of the masing gas .
however , it still remains a matter of debate how accidental overlapping and/or beaming occur repetitively for more than @xmath430 years since the discovery of the supermaser .
we are grateful to alain baudry for useful comments , which is essential to improve our scientific discussion .
we thank the staff of the vera stations and alma observatory to support our project .
this paper makes use of the following alma data : ads / jao.alma#2011.0.00199.s .
alma is a partnership of eso ( representing its member states ) , nsf ( usa ) and nins ( japan ) , together with nrc ( canada ) and nsc and asiaa ( taiwan ) , in cooperation with the republic of chile .
the joint alma observatory is operated by eso , aui / nrao and naoj .
this research has made use of the simbad database , operated at cds , strasbourg , france .
t.h . is supported by the mext / jsps kakenhi grant numbers 21224002 , 24684011 , and 25108005 , and the alma japan research grant of naoj chile observatory , naoj - alma-0006 .
is supported by the mext / jsps kakenhi grant numbers 24540242 and 25120007 .
allen , d. a. & burton , m. g. 1993 , nature , 363 , 54 bally , j. 2008 , in handbook of star forming regions , vol .
1 , ed . b. reipurth ( asp monograph publications , vol .
4 ; san francisco , ca : asp ) , 459 bally , j. , cunningham , n. j. , moeckel , n. , burton , m. g. , smith , n. , frank , a. , & nordlund , a. 2011 , apj , 727 , 113 baudry , a. , forster , j. r. , & welch , w. j. 1974 , a&a , 36 , 217 beuther , h. , & nissen , h. d. 2008 , apjl , 679 , l121 brouillet , n. , despois , d. , baudry , a. , et al . 2013 , a&a , 550 , 46 brouillet , n. et al .
2014 , in preparation chen , p. , pearson , j. c. , pickett , h. m. , matsuura , s. , & blake , g. a. 2000 , apjs , 128 , 371 churchwell , e. , felli , m. , wood , d. o. s. , & massi , m. 1987 , apj , 321 , 516 daniel , f. & cernicharo , j. 2013 , a&a , 553 , a70 deguchi , s. & watson , w. d. 1989 , apjl , 340 , l17 eisner , j. a. , plambeck , r. l. , carpenter , j. m. , corder , s. a. , qi , c. , & wilner , d. 2008 , apj , 683 , 304 elitzur , m. , hollenbach , d. j. , mckee , c. f. 1989 , apj , 346 , 983 favre , c. , despois , d. , brouillet , n .. , baudry , a. , combes , f. , gulin , m. , wootten , a. , & wlodarczak , g. 2011 , a&a , 532 , a32 felli , m. , churchwell , e. , wilson , t. l. , taylor , g. b. 1993 , a&as , 98 , 137 forster , j. r. , welch , w. j. , wright , m. c. h. , & baudry , a. , 1978 , apj , 221 , 137 friedel , d. n. & widicus weaver , s. l. 2011 , apj , 742 , 64 garay , g. , moran , j. m. , & haschick , a. d. 1989 , 338 , 244 gaume , r. a. , wilson , t. l. , vrba , f. j. , johnston , k. j. , schmid - burgk , j. 1998 , apj , 493 , 940 genzel , r. , reid , m. j. , moran , j. m. , & downes , d. 1981 , apj , 244 , 884 genzel , r. , & stutzki , j. 1989 , araa , 27 , 41 goddi , c. , humphreys , e. m. l. , greenhill , l. j. , chandler , c. j. , & matthews , l. d. 2011 , apj , 728 , 15 goldreich , p. & kwan , j. 1974 , apj , 190 , 27 greenhill , l. j. , goddi , c. , chandler , c. j. , matthews , l. d. , humphreys , e. m. l. 2013 , apj , 770 , 32 hirota , t. , tsuboi , m. , fujisawa , k. et al .
2011 , apj , 739 , l59 hirota , t. , bushimata , t. , choi , y. k. et al .
2007 , pajs , 59 , 897 hirota , t. , kim , m. k. , kurono , y. , & honma , m. 2014a , apjl , 782 , l28 hirota , t. , kim , m. k. , kurono , y. , & honma , m. 2014b , in preparation kaifu , n. , usuda , t. , hayashi , s. et al .
2000 , pasj , 52 , 1 kim , m. k. , hirota , t. , honma , m. , et al .
2008 , pasj , 60 , 991 kim , m. k. , hirota , t. , honma , m. , & kobayashi , h. 2014 , in preparation kukolich , s.g .
1969 , j. chem .
phys . , 50 , 3751 liu , s. -y . , girart , j. m. , remijan , a. , & snyder , l. e. 2002 , apj , 576 , 255 ma , c. , arias , e. f. , eubanks , t. m. , et al .
1998 , aj , 116 , 516 matthews , l. d. , greenhill , l. j. , goddi , c. , chandler , c. j. , humphreys , e. m. l. , & kunz , m. w. 2010 , apj , 708 , 80 matveenko , l. i. , graham , d. a. , & diamond , p. j. 1988 , sov .
, 14 , 468 matveyenko , l. i. , zhakharin , k. m. , diamond , p. j. , & graham , d. a. 2004 , astron .
lett . , 30 , 100 matveyenko , l. i. , demichev , v. a. , ipatov , a. v. , melnikov , a. e. , & surkis , i. f. 2012a , astl , 38 , 575 matveyenko , l. i. , demichev , v. a. , ipatov , a. v. , melnikov , a. e. , & surkis , i. f. 2012b , astl , 38 , 764 menten , k. m. , & reid , m. j. 1995 , apj , 445 , l157 menten , k.m . , reid , m. j. , forbrich , j. , & brunthaler , a. 2007 , a&a , 474 , 515 neill , j. , l. , wang , s. , bergin , e. a. , crockett , n. r. , favre , c. , plume , r. , melnick , g. j. 2013 , apj , 770 , 142 neufeld , d. a. & melnick , g. j. 1990 , apjl , 352 , l9 neufeld , d. a. & melnick , g. j. 1991 , apj , 368 , 215 niederhofer , f. , humphreys , e. m. l. , & goddi , c. 2012 , a&a , 548 , a69 okumura , s. , yamashita , t. , sako , s. , miyata , t. , honda , m. , kataza , h. , okamoto , y. k. 2011 , pasj , 63 , 823 omodaka , t. , maeda , t. , miyoshi , m. , et al . 1999 ,
pasj , 51 , 333 otto , s. & gaylard , m. j. 2012 , in iau symp . , 287 , cosmic masers from oh to h@xmath57 , ed .
r. s. booth , e. m. l. humphreys , & w. h. t. vlemmings ( cambridge : cambridge univ .
press ) , 110 plambeck , r. l. , wright , m. c. h. , friedel , d. n. , et al .
2009 , apj , 704 , l25 shimoikura , t. , kobayashi , h. , omodaka , t. , diamond , p. j. , matveyenko , l. i. , & fujisawa , k. 2005 , apj , 634 , 459 tang , y. w. , ho , p. t. p. , koch , p. m. , & rao , r. 2010 , apj , 717 , 1262 torrelles , j. m. et al . 2003 , apjl , 598 , l115 tolmachev , a. 2011 , atel , 3177 zapata , l. a. , rodrguez , l. f. , schmid - burgk , j. , loinard , l. , menten , k. m. , & curiel , s. 2012 , apj , 754 , l17 zapata , l. a.,schmid - burgk , j. , ho , p. t. p. , rodrguez , l. f. , menten , k. m. 2009 , apj , 704 , l45 zapata , l. a. , schmid - burgk , j. , & menten , k. m. 2011 , a&a , 529 , a24 cccccc feature & spot & @xmath58 & @xmath19 & @xmath20 & + id@xmath59 & id@xmath60 & [ km s@xmath5 ] & [ mas yr@xmath5 ] & [ mas yr@xmath5 ] & epoch + 1 & 1 & 6.74 & -2.0(0.4 ) & -1.8(0.4 ) & 2011/068 - 2011/296 + 1 & 3 & 6.95 & -2.2(0.5 ) & -1.7(0.6 ) & 2011/068 - 2011/296 + 1 & 5 & 7.16 & -2.7(1.2 ) & -1.0(1.4 ) & 2011/068 - 2011/250 + 2 & 2 & 6.74 & 5.2(0.9 ) & -10.2(0.8 ) & 2011/296 - 2012/164 + 2 & 4 & 6.95 & 4.8(0.6 ) & -9.7(0.5 ) & 2011/250 - 2012/207 + 2 & 6 & 7.16 & 4.0(0.3 ) & -9.0(0.2 ) & 2011/068 - 2012/207 + 2 & 7 & 7.37 & 4.0(0.2 ) & -9.1(0.2 ) & 2011/068 - 2012/207 + 2 & 8 & 7.58 & 4.0(0.2 ) & -9.1(0.2 ) & 2011/068 - 2012/269 + 2 & 10 & 7.79 & 4.0(0.2 ) & -9.0(0.2 ) & 2011/068 - 2012/269 + 3 & 9 & 7.58 & -2.6(9.2 ) & -4.1(8.3 ) & 2011/121 - 2011/208 + 3 & 11 & 7.79 & 0.6(3.7 ) & -5.5(3.7 ) &
2011/121 - 2011/208 + + + | the 22 ghz h@xmath0o maser in orion kl has shown extraordinary burst events in 1979 - 1985 and 1998 - 1999 , sometimes called supermaser .
we have conducted monitoring observations of the supermaser in orion kl using vera , vlbi exploration of radio astrometry , in the current third burst since 2011 march .
three flux maxima are detected in 2011 and 2012 with rising and falling timescales of 2 - 7 months .
time variations of the supermaser seem symmetric for all of the active phases .
the maximum total flux density of 135000 jy is observed in 2012 june while it is still one order of magnitude lower than those in previous bursts .
the supermaser consists of two spatially different components at different velocities .
they are elongated along a northwest - southeast direction perpendicular to the low - velocity outflow driven by source i. proper motions of the supermaser features with respect to source i are measured toward west and southwest directions , almost parallel to the low - velocity outflow .
the flux density and linewidth show an anti - correlation as expected for an unsaturated maser emission .
the supermaser is located close to the methylformate ( hcooch@xmath1 ) line and continuum emission peaks in the orion compact ridge detected by alma , atacama large millimeter / submillimeter array .
the broader velocity range of the weak hcooch@xmath1 emission at the supermaser position would be an evidence of a shock front .
on the other hand , the 321 ghz h@xmath0o line is not detected at the position of the supermaser .
it can be explained qualitatively by one of the theoretical h@xmath0o excitation models without extraordinary conditions .
our results support a scenario that the supermaser is excited in the dense gas interacting with the low - velocity outflow in the compact ridge .
the extremely high flux density and its symmetric time variation for rising and falling phases could be explained by a beaming effect during the amplification process rather than changes in physical conditions . |
nonlocality , _ i.e. _ the existence of correlations which can not be explained by any local hidden variable model , is perhaps the most debated implication of quantum mechanics . during the last decade other aspects of nonlocality , in addition to generating nonlocal correlations , have been discovered .
for example , the possibility of teleporting and effectively encoding information , as well as the ability to perform certain computations exponentially faster than any classical device .
realistic implementations of quantum information protocols require the investigation of nonlocality properties of quantum states in a noisy environment .
in particular , the robustness of nonlocality should be addressed , as well as the design of protocols to preserve and possibly enhance nonlocality in the presence of noise .
the evolution of nonlocality for a twin - beam state of radiation ( twb ) in a thermal environment was studied in ref .
@xcite by means of the displaced parity test @xcite , whereas in ref .
@xcite its nonlocality was investigated using the pseudospin operators @xcite when only dissipation occurs . in ref .
@xcite we have suggested a conditional measurement scheme on twb leading to a non - gaussian entangled mixed state , which improves fidelity of teleportation of coherent states .
this process , termed inconclusive photon subtraction ( ips ) , is based on mixing each mode of the twb with the vacuum in an unbalanced beam splitter and then performing inconclusive photodetection on both modes , _ i.e. _ revealing the reflected beams without discriminating the number of the detected photons .
ips states have the following properties : they improve the teleportation fidelity for coherent states @xcite and show enhanced nonlocal correlations in the phase space @xcite in ideal conditions , namely in the absence of noise .
motivated by these results and by the recent experimental generation of ips states @xcite , in this paper we extend the previous studies on the twb and consider the nonlocality of the ips state in the presence of noise .
the paper is structured as follows . in sec .
[ s : lossy ] we address the evolution of the twb in a noisy channel where both dissipation and thermal noise are present , whereas in sec . [ s : ips ] we briefly review the ips process . in secs .
[ s : dp ] , [ s : hd ] and [ s : ps ] we investigate the nonlocality of twb and ips by means of three different tests : displaced parity , homodyne detection , and pseudospin test , respectively .
finally , sec .
[ s : remarks ] closes the paper with some concluding remarks .
the so called twin - beam state of radiation ( twb ) , _ i.e. _ @xmath0 with @xmath1 , @xmath2 being the twb squeezing parameter .
@xmath3 is obtained by parametric down - conversion of the vacuum , @xmath4 , @xmath5 and @xmath6 being field operators , and it is described by the gaussian wigner function @xmath7}}\ , , \label{twb : wig}\ ] ] with @xmath8}}\,,\qquad \widetilde{b}_0 = \frac{b_0}{16 \sqrt{{\rm det}[{\boldsymbol \sigma}_0]}}\,,\ ] ] where @xmath9 , @xmath10 and @xmath11 is the covariance matrix @xmath12 b_0\ , { \boldsymbol \sigma}_3 & a_0\ , \mathbbm{1}_2 \end{array}\right)\:,\ ] ] @xmath13 being the @xmath14 identity matrix and @xmath15 . using a more compact form , eq .
( [ twb : wig ] ) can also be rewritten as @xmath16}}\,,\ ] ] with @xmath17 , @xmath18 and @xmath19 , and @xmath20 denoting the transposition operation .
when the two modes of the twb interact with a noisy environment , namely in the presence of dissipation and thermal noise , the evolution of the wigner function ( [ twb : wig ] ) is described by the following fokker - planck equation @xcite @xmath21 with @xmath22 .
the damping matrix is given by @xmath23 , whereas @xmath24 \boldsymbol{0 } & { \boldsymbol \sigma}_{\infty}^{(2 ) } \end{array } \right)\,,\end{aligned}\ ] ] where @xmath25 is the @xmath14 null matrix and @xmath26 0 & 1 + 2 n_k \end{array } \right)\,.\ ] ] @xmath27 , @xmath28 denotes the damping rate and the average number of thermal photons of the channel @xmath29 , respectively .
@xmath30 represents the covariance matrix of the environment and , in turn , the asymptotic covariance matrix of the evolved twb .
since the environment is itself excited in a gaussian state , the evolution induced by ( [ fp : eq : cmp ] ) preserves the gaussian form ( [ gauss : form ] ) . the covariance matrix at time @xmath31 reads as follows @xcite @xmath32 where @xmath33 .
the covariance matrix @xmath34 can be also written as @xmath35 b_t(\gamma_2)\ , { \boldsymbol \sigma}_3 & a_t(\gamma_2,n_2)\ , \mathbbm{1}_2 \end{array } \right)\ ] ] with @xmath36 let us now consider channels with the same damping rate @xmath37 but different number of thermal photons , @xmath38 and @xmath39 : using the density matrix formalism , the state corresponding to the covariance matrix ( [ evol : cvm:12 ] ) has the following form @xmath40 where @xmath41 is the thermal state @xmath42 @xmath43 , @xmath44 being the mode operators .
the average number of photons are given by @xmath45\label{m1:evol}\,,\\ m_2 & = \frac14 \left [ \sqrt{a_{+}^2 - 16 b_t } - ( 2 + a_{- } ) \right]\label{m2:evol}\,,\end{aligned}\ ] ] @xmath46 , @xmath47 and @xmath48 .
( [ rho : t : evol ] ) @xmath49 denotes the two - mode squeezing operator , with parameter @xmath50 @xmath51 = \pi/2\label{arg : xi : evol}\,.\end{aligned}\ ] ] eq .
( [ rho : t : evol ] ) says that the quantum state of a twb , after propagating in a noisy channel , is the same of a state obtained by parametric down - conversion from a noisy background @xcite .
their properties , and in particular entanglement and nonlocality , can be addressed in an unified way using eq .
( [ rho : t : evol ] ) or , equivalently , eqs .
( [ evol : cvm:12 ] ) and ( [ atbt ] ) . finally , if we assume @xmath52 and @xmath53 , then the covariance matrix ( [ evol : cvm:12 ] ) becomes formally identical to ( [ cvm : twb ] ) and the corresponding wigner function reads @xmath54}}\ , , \label{twb : wig : noise}\ ] ] with @xmath55}}\ , , \qquad \widetilde{b}_t = \frac{b_t(\gamma)}{16\sqrt{{\rm det}[{\boldsymbol \sigma}_t]}}\,,\ ] ] whereas the density matrix , _ mutatis mutandis _
, is still given by eq .
( [ rho : t : evol ] ) .
scheme of the ips process . ] when thermal noise and dissipation affect the propagation of an entangled state , its nonlocal properties are reduced and , finally , destroyed @xcite .
therefore it is of interest to look for some technique in order to preserve , at least in part , such correlations , or to enhance the nonlocality of the state which will face the lossy transmission line . since it has been shown that the de - gaussification of a twb can enhance its entanglement in the ideal case and since non - gaussian states can be produced using the current technology @xcite , in this and the following sections we will investigate whether or not this process can be useful also in the presence of noise .
the de - gaussification of a twb can be achieved by subtracting photons from both modes @xcite . in ref .
@xcite we referred to this process as to inconclusive photon subtraction ( ips ) and showed that the resulting state , the ips state , can be used to enhance the teleportation fidelity of coherent states for a wide range of the experimental parameters . moreover , in ref .
@xcite , we have shown that , in the absence of any noise during the transmission stage , the ips state has nonlocal correlations larger than those of the twb irrespective of the ips quantum efficiency ( see also refs .
@xcite ) .
first of all we briefly recall the ips process , whose scheme is sketched in fig .
[ f : ips : scheme ] .
the two modes , @xmath5 and @xmath6 , of the twb are mixed with the vacuum ( modes @xmath56 and @xmath57 , respectively ) at two unbalanced beam splitters ( bs ) with equal transmissivity ; the modes @xmath56 and @xmath57 are then revealed by avalanche photodetectors ( apds ) with equal efficiency , which can only discriminate the presence of radiation from the vacuum : the ips state is obtained when the two detectors jointly click .
the mixing with the vacuum at a beam splitter with transmissivity @xmath58 followed by the on / off detection with quantum efficiency @xmath59 is equivalent to mixing with an effective transmissivity @xmath60 @xcite @xmath61 followed by an ideal ( _ i.e. _ efficiency equal to @xmath62 ) on / off detection . using the wigner formalism , when the input state arriving at the two beam splitters is the twb @xmath63 of eq .
( [ twb : wig ] ) , the state produced by the ips process reads as follows ( see ref .
@xcite for the details about the calculation and about the de - gaussification map for the density matrix and wigner function in the case of a twb ) @xmath64 where @xmath65 is the probability of a click in both the apds . in eqs .
( [ ips : wigner ] ) and ( [ ips : probability ] ) we introduced @xmath66 and defined @xmath67}\,[x_k y_k - 4 { \widetilde{b}}_0 ^ 2 ( 1-\tau)^2]}\,,\ ] ] where @xmath68 , @xmath69 , @xmath70 ; @xmath71 , and @xmath72 are @xmath73 with @xmath74 $ ] , @xmath75 $ ] ; finally , @xmath76 , @xmath77 , and @xmath78 depend on @xmath2 and @xmath60 and are given by @xmath79\,,\\ g_k & = { { \cal n}}_k \ , [ x_k ( 1-{\widetilde{a}}_0)^2 + 4 { \widetilde{b}}_0 ^ 2 ( 1-{\widetilde{a}}_0 ) ( 1-\tau ) + y_k { \widetilde{b}}_0 ^ 2]\,,\\ h_k & = { { \cal n}}_k \ , \{(x_k + y_k ) { \widetilde{b}}_0 ( 1-{\widetilde{a}}_0 ) + 2 { \widetilde{b}}_0 [ { \widetilde{b}}_0 ^ 2 + ( 1-{\widetilde{a}}_0)^2 ] ( 1-\tau)\}\,,\\ { { \cal n}}_k & \equiv { { \cal n}}_k(r,\tau ) = { \displaystyle \frac{4 \tau\ , ( 1-\tau)}{x_k y_k - 4 { \widetilde{b}}_0 ^ 2 ( 1-\tau)^2}\ , . } \end{aligned}\ ] ] the state corresponding to eq .
( [ ips : wigner ] ) is no longer a gaussian state and its nonlocal properties , in ideal conditions , were studied in ref .
@xcite .
here we are interested in the case when the ips process is performed on a twb evolved in a noisy environment with both the channels having the same damping rate and thermal noise .
the wigner function of the state arriving at the beam splitters is now given by eq .
( [ twb : wig : noise ] ) , and the output state is still described by eq .
( [ ips : wigner ] ) , but with the following substitutions @xmath80 we will denote with @xmath81 the wigner function of this degraded ips state . in the next sections
we will analyze the nonlocality of the ips state in the presence of noise by means of bell s inequalities .
parity is a dichotomic variable and thus can be used to establish bell - like inequalities @xcite .
the displaced parity operator on two modes is defined as @xcite @xmath82 where @xmath83 , @xmath5 and @xmath6 are mode operators and @xmath84 and @xmath85 are single - mode displacement operators . since the two - mode wigner function @xmath86 can be expressed as @xcite @xmath87 @xmath88 being the expectation value of @xmath89 , the violation of these inequalities is also known as nonlocality in the phase - space . the quantity involved in such inequalities
can be written as follows @xmath90 which , for local theories , satisfies @xmath91 .
following ref .
@xcite , one can choose a particular set of displaced parity operators , arriving at the following combination @xcite @xmath92 which , for the twb , gives a maximum @xmath93 , greater than the value @xmath94 obtained in ref .
notice that , even in the infinite squeezing limit , the violation is never maximal , _
i.e. _ @xmath95 @xcite . in ref .
@xcite we studied eq .
( [ bell : ale ] ) for both the twb and the ips state in an ideal scenario , namely in the absence of dissipation and noise ; we showed that , using ips , the maximum violation is achieved for @xmath96 and for values of @xmath2 smaller than for the twb .
( 70,100)(0,0 ) ( 10,0 ) for the twb ( top ) and ips ( bottom ) ; we set @xmath97 and @xmath98 .
the dashed lines refer to the absence of noise ( @xmath99 ) , whereas , for both the plot , the solid lines are @xmath100 with @xmath101 and , from top to bottom , @xmath102 and @xmath103 . in the ideal case
the maxima are @xmath104 and @xmath105 , respectively.,title="fig:",width=226 ] ( 42,42)@xmath2 ( -2,65)@xmath106 ( 42,0)@xmath2 ( -2,23)@xmath107 now , by means of the eq .
( [ ips : wigner ] ) and the substitutions ( [ sostituzioni ] ) , we can study how noise affects @xmath108 .
the results are showed in fig .
[ f : dp ] : as one may expect , the overall effect of noise is to reduce the violation of the bell s inequality .
when dissipation alone is present ( @xmath109 ) , the maximum of violation is achieved using the ips for values of @xmath2 smaller than for the twb , as in the ideal case . on the other hand
, one can see that the presence of thermal noise mainly affects the ips results .
in fact , for @xmath110 and @xmath111 , one has @xmath112 for a range of @xmath2 values , whereas @xmath113 falls below the threshold for violation .
we conclude that , considering the displaced parity test in the presence of noise , the ips is quite robust if the thermal noise is below a threshold value ( depending on the environmental parameters ) and for small values of the twb parameter @xmath2 .
in principle there are two approaches how to test the bell s inequalities for bipartite state : either one can employ some test for continuous variable systems , such as that described in sec .
[ s : dp ] , or one can convert the problem to bell s inequalities tests on two qubits by mapping the two modes into two - qubit systems . in this and the following section we will consider this latter case . the wigner function @xmath114 given in eq .
( [ ips : wigner ] ) is no longer positive - definite and thus it can be used to test the violation of bell s inequalities by means of homodyne detection , _
i.e. _ measuring the quadratures @xmath115 and @xmath116 of the two ips modes @xmath5 and @xmath6 , respectively , as proposed in refs . @xcite . in this case
, one can dichotomize the measured quadratures assuming as outcome @xmath117 when @xmath118 , and @xmath119 otherwise .
the nonlocality of @xmath114 in ideal conditions has been studied in ref .
@xcite where we also discussed the effect of the homodyne detection efficiency @xmath120 .
let us now we focus our attention on @xmath81 , namely the state produced when the ips process is applied to the twb evolved through the noisy channel .
after the dichotomization of the homodyne outputs , one obtains the following bell parameter @xmath121 where @xmath122 and @xmath123 are the phases of the two homodyne measurements at the modes @xmath5 and @xmath6 , respectively , and @xmath124\ , p(x_{\vartheta_h } , x_{\varphi_k})\,,\ ] ] @xmath125 being the joint probability of obtaining the two outcomes @xmath126 and @xmath127 @xcite . as usual , violation of bell s inequality
is achieved when @xmath128 .
( 70,100)(0,0 ) ( 4,0 ) for the ips states for two different values of the homodyne detection efficiency : @xmath129 ( top ) , and @xmath130 ( bottom ) .
we set @xmath131 .
the dashed lines refer to the absence of noise ( @xmath99 ) , whereas , for both the plots , the solid lines are @xmath132 with @xmath133 and , from top to bottom , @xmath134 and @xmath103.,title="fig:",width=245 ] ( 37,45)@xmath2 ( -2,70)@xmath132 ( 37,-1)@xmath2 ( -2,24.5)@xmath132 in fig .
[ f : hd ] we plot @xmath132 for @xmath135 , @xmath136 , @xmath137 and @xmath138 : as for the ideal case @xcite , the bell s inequality is violated for a suitable choice of the squeezing parameter @xmath2 .
obviously , the presence of noise reduces the violation , but we can see that the effect of thermal noise is not so large as in the case of the displaced parity test addressed in sec . [
s : dp ] ( see fig . [
f : dp ] ) .
notice that the high efficiencies of this kind of detectors allow a loophole - free test of hidden variable theories @xcite , though the violations obtained are quite small .
this is due to the intrinsic information loss of the binning process , which is used to convert the continuous homodyne data in dichotomic results @xcite .
another way to map a two - mode continuous variable system into a two - qubit system is by means of the pseudospin test : this consists in measuring three single - mode hermitian operator @xmath139 satisfying the pauli matrix algebra @xmath140=2i\varepsilon_{hkl}\,s_l$ ] , @xmath141 , @xmath142 , and @xmath143 is the totally antisymmetric tensor with @xmath144 @xcite . for the sake of clarity
, we will refer to @xmath145 , @xmath146 and @xmath147 as @xmath148 , @xmath149 and @xmath150 , respectively . in this way one can write the following correlation function @xmath151 where @xmath152 and @xmath153 are unit vectors such that @xmath154 with @xmath155 . in the following , without loss of generality , we set @xmath156
. finally , the bell parameter reads @xmath157 corresponding to the chsh bell s inequality @xmath158 . in order to study eq .
( [ bell : ps ] ) we should choose a specific representation of the pseudospin operators ; note that , as pointed out in refs .
@xcite , the violation of bell inequalities for continuous variable systems depends , besides on the orientational parameters , on the chosen representation , since different @xmath139 leads to different expectation values of @xmath159 . here
we consider the pseudospin operators corresponding to the wigner functions @xcite @xmath160\big]\,,\quad w_z(\alpha)= -\frac{1}{2}\,\delta^{(2)}(\alpha)\,,\label{ps : w : xz}\\ & w_y(\alpha)=-\frac{1}{2\pi}\ , \delta\big(\re{\rm e}[\alpha ] \big)\ , { \cal p } \frac{1}{\im{\rm m}[\alpha]}\,,\end{aligned}\ ] ] where @xmath161 denotes the cauchy s principal value .
thanks to ( [ ps : w : xz ] ) one obtains @xmath162\,,\ ] ] for the twb , and , for the ips , @xmath163\ ] ] where @xmath164 , and all the other quantities have been defined in sec .
[ s : ips ] .
( 70,50)(0,0 ) ( 4,0 ) in ideal case ( @xmath99 ) : the dashed line refers to the twb , whereas the solid lines refer to the ips with , from top to bottom , @xmath165 , and @xmath166 .
there is a threshold value for @xmath2 below which ips gives a higher violation than twb .
note that there is also a region of small values of @xmath2 for which the ips state violates the bell s inequality while the twb does not .
the dash dotted line is the maximal violation value @xmath167.,title="fig:",width=245 ] ( 40,-3)@xmath2 ( -2,24.5)@xmath159 in fig .
[ f : ps : id ] we plot @xmath159 for the twb and ips in the ideal case , namely in the absence of dissipation and thermal noise . for all the figures we set @xmath168 , @xmath169 , and @xmath170 .
as usual the ips leads to better results for small values of @xmath2 .
whereas @xmath171 as @xmath172 , @xmath173 has a maximum and , then , falls below the threshold @xmath174 as @xmath2 increases .
it is interesting to note that there is a region of small values of @xmath2 for which @xmath175 , _ i.e. _ the ips process can increases the nonlocal properties of a twb which does not violates the bell s inequality for the pseudospin test , in such a way that the resulting state violates it .
this fact is also present in the case of the displaced parity test described in sec .
[ s : dp ] , but using the pseudospin test the effect is enhanced .
notice that the maximum violations for the ips occur for a range of values @xmath2 experimentally achievable .
( 70,50)(0,0 ) ( 4,0 ) for @xmath110 : the dashed line refers to the twb , whereas the solid lines refer to the ips with , from top to bottom , @xmath176 , and @xmath166 .
the same comments as in fig .
[ f : ps : id ] still hold.,title="fig:",width=245 ] ( 40,-3)@xmath2 ( -2,24.5)@xmath159 in fig .
[ f : ps : tau ] we consider the presence of the dissipation alone and vary @xmath60 .
we can see that ips is effective also when the effective transmissivity @xmath60 is not very high .
we take into account the effect of dissipation and thermal noise in figs .
[ f : ps : gamma ] , and [ f : ps : th ] : we can conclude that ips is quite robust with respect to this sources of noise and , moreover , one can think of employing ips as a useful resource in order to reduce the effect of noise .
( 70,50)(0,0 ) ( 4,0 ) for different values of @xmath177 and in the absence of thermal noise ( @xmath178 ) : the dashed lines refer to the twb , whereas the solid ones refer to the ips with @xmath98 ; for both the twb and ips we set , from top to bottom , @xmath179 , and @xmath180 .
the dash dotted line is the maximal violation value @xmath167.,title="fig:",width=245 ] ( 40,-3)@xmath2 ( -2,24.5)@xmath159 ( 70,50)(0,0 ) ( 4,0 ) for @xmath110 and different values @xmath178 : the dashed lines refer to the twb , whereas the solid ones refer to the ips with @xmath98 ; for both the twb and ips we set , from top to bottom , @xmath181 , and @xmath103.,title="fig:",width=245 ] ( 40,-3)@xmath2 ( -2,24.5)@xmath159
we have addressed three different nonlocality tests , namely , displaced parity , homodyne detection and pseudospin test , on twb and ips in the presence of noise .
we have shown that the ips process on twb enhances nonlocality not only in ideal cases , but also when noise ( dissipation and thermal noise ) affects the propagation .
as in the ideal situation , the enhancement is achieved when the twb energy is not too high ( small squeezing parameter @xmath2 ) , depending on the environmental parameters . moreover , in the case of the pseudospin test , we have seen that there is a region of small @xmath2 for which the twb itself does not violates the bell s inequality , where s after the ips process it does . finally , we mention that the enhanced nonlocality also in the presence of noise makes the ips states useful resources for continuous variable quantum information processing . stimulating and useful discussions with m. s. kim , a. ferraro and a. r. rossi are gratefully acknowledged .
99 h. jeong , j. lee , and m. s. kim , phys .
a * 61 * , 052101 ( 2000 ) .
k. banaszek , and k. wdkiewicz , phys .
a * 58 * , 4345 ( 1998 ) .
r. filip , and l. mista , phys .
a * 66 * , 044309 ( 2002 ) . z .-
b . chen , j .- w .
pan , g. hou , and y .- d .
zhang , phys .
lett . * 88 * , 040406 ( 2002 ) .
s. olivares , m. g. a. paris , and r. bonifacio , phys .
rev . a * 67*,032314 ( 2003 ) .
s. olivares , and m. g. a. paris , phys .
a * 70 * , 032112 ( 2004 ) .
j. wenger , r. tualle - bouri , and p. grangier , phys .
rev . lett . * 92 * , 153601 ( 2004 ) .
d. walls , and g. milburn , _ quantum optics _
( springer , berlin , 1994 ) .
s. olivares , and m. g. a. paris , j. opt .
b : quantum semiclass . opt .
* 6 * , 69 ( 2004 ) .
a. serafini , f. illuminati , m. g. a. paris , and s. de siena , phys .
a * 69 * , 023318 ( 2004 ) .
a. ferraro , s. olivares , and m. g. a. paris , _ gaussian states in c.v . quantum information _ ( bibliopolis , napoli , 2005 ) .
a. r. rossi , s. olivares , and m. g. a. paris , j. mod . opt . *
51 * , 1057 ( 2004 ) .
t. opatrn , g. kurizki , and d .-
welsch , phys .
a * 61 * , 032302 ( 2000 ) .
p. t. cochrane , t. c. ralph , and g. j. milburn , phys .
rev . a * 65 * , 062306 ( 2002 ) .
h. nha , and h. j. carmichael , phys . rev . lett . * 93 * , 020401 ( 2004 ) .
r. garca - patrn snchez et al . , phys .
lett . * 93 * , 120409 ( 2004 ) ; r. garca - patrn , j. fiur ' aek , and n. j cerf , phys .
a * 71 * , 022105 ( 2005 ) .
j. f. clauser , m. a. horne , a. shimony , and r. a. holt , phys .
* 23 * , 880 ( 1969 ) .
h. jeong et al . , phys .
a * 67 * , 012106 ( 2003 ) .
a. gilchrist , p. deuar , and m. d. reid , phys .
80 * , 3169 ( 1998 ) ; a. gilchrist , p. deuar , and m. d. reid , phys .
a * 60 * , 4259 ( 1999 ) .
w. j. munro , phys .
a * 59 * , 4197 ( 1999 ) .
g. gour et al . , phys .
a * 324 * , 415 ( 2003 ) ; m. revzen et al .
a * 71 * , 022103 ( 2005 ) .
a. ferraro , and m. g. a. paris , quant - ph/0410194 . | we address nonlocality of continuous variable systems in the presence of dissipation and noise .
three nonlocality tests have been considered , based on the measurement of displaced - parity , field - quadrature and pseudospin - operator , respectively .
nonlocality of twin - beam has been investigated , as well as that of its non - gaussian counterparts obtained by inconclusive subtraction of photons .
our results indicate that : i ) nonlocality of twin - beam is degraded but not destroyed by noise ; ii ) photon subtraction enhances nonlocality in the presence of noise , especially in the low - energy regime . |
since the discovery of the correlation between the stellar bulge mass in galaxies and the mass of their supermassive black holes @xcite , it is now widely acknowledged that these central supermassive black holes play an important role in governing the formation and evolution of their host galaxies . in our current picture of galaxy formation , galaxies and their supermassive black - holes co - evolve .
molecular gas and the dusty interstellar medium ( ism ) in galaxies , acts as the fuel for both star formation and black hole accretion .
therefore , studying the distribution of dust and gas in star forming galaxies with actively accreting supermassive black holes can help shed light on the exact physical processes that are driving co evolution . both star formation and black hole accretion activity in the universe peak at @xmath17 ( e.g. @xcite ) and this represents the key epoch in the universe s history at which to study the feeding and feedback processes in galaxies and quasars that are the main drivers of co - evolution . over the last two decades , surveys starting with the scuba bolometer on the james clerk maxwell telescope ( jcmt ) and , more recently
, using the _ herschel _ space satellite , have led to the discovery of large populations of far infrared ( fir ) and submillimetre bright starburst galaxies ( smgs ) at the main epoch of galaxy formation ( e.g. @xcite ) .
molecular gas detections are now being assembled for many tens to hundreds of such galaxies ( e.g. @xcite ) .
several studies of molecular gas in @xmath18 galaxies have also taken advantage of gravitational lensing to amplify the molecular line signal ( e.g. @xcite ) . while clearly an advantage for probing intrinsically fainter galaxies , differential magnification between the different dust and gas components can make interpretations regarding their actual spatial distribution complicated in such lensed systems .
quasars , on account of being among the brightest extragalactic sources in our universe , were some of the first sources to be observed in molecular gas ( e.g. @xcite ) . using the compilations of extragalactic sources with co detections in @xcite and @xcite , we find there are a total of 89 quasars that have been detected in co all the way from the local universe out to the highest redshifts of @xmath19 .
many of the recent efforts have focussed on @xmath20 quasars ( e.g. @xcite ) and , as a consequence , the numbers of quasars with co detections at the peak epoch of galaxy formation ( 1.5@xmath21 ) is considerably smaller - @xmath2220 quasars .
many of the quasars that have been observed in molecular gas at these epochs are gravitationally lensed systems ( e.g. @xcite ) .
excluding these lensed systems and focussing only on unlensed quasars where spatially mapping the gas and dust distributions is considerably easier , there are only 13 quasars at @xmath23 with co detections @xcite . detecting molecular gas in even a small number of unlensed quasars at these redshifts , to synchronously study the fuelling of star formation and black hole accretion at the epoch when both are at their peak , is therefore valuable .
most searches for molecular gas in quasars have started from samples of ultraviolet ( uv ) luminous quasars ( e.g. @xcite ) that were already known to be fir / millimetre bright @xcite .
the studies by @xcite and @xcite find that only 9 out of 53 uv luminous quasars at @xmath24 ( i.e. 17 per cent ) are detected at 850@xmath25 m down to 3@xmath26 flux density limits of @xmath107 - 9 mjy and 9 out of 26 uv luminous quasars at @xmath24 ( i.e. 35 per cent ) are detected at 1.2 mm down to 3@xmath26 flux density limits of @xmath101.8 - 4 mjy . in the galaxy formation scenario first advocated by @xcite and
now commonly adopted in galaxy formation models ( e.g. @xcite ) , the most highly star forming galaxies at high redshift e.g. the smgs , will eventually evolve into uv luminous quasars .
the relatively small fraction of fir / mm - bright quasars argues for a quick transition between the two populations and recent studies suggest that the transition phase could be as short lived as @xmath101 myr @xcite . in such a picture of galaxy formation , transition populations of hybrid smg - quasars should exist and a larger fraction of these transition quasars should be fir bright , gas rich systems compared to the uv luminous quasars . molecular gas observations of smgs and uv luminous quasars have also revealed some differences between the two populations .
while smgs appear to have substantial amounts of low excitation gas @xcite , this does not seem to be the case in quasars @xcite which could therefore represent a later evolutionary stage when the extended gas reservoirs in the galaxy have been significantly depleted . once again , if transition objects between these two populations do exist , their gas fractions might be expected to be intermediate between the smgs and uv luminous quasars .
searches for these so called transition " galaxies have often focussed on obscured agn populations with different selection methods picking out agn and quasars with a wide variety of dust extinctions and luminosities ( e.g. @xcite ) . while in principle , any or all of these populations could represent the missing evolutionary link between star forming galaxies and optical quasars , in practice further multiwavelength observations are necessary to establish that these obscured quasars are indeed distinct in terms of their physical properties from matched control samples of unobscured quasars .
our searches for obscured quasars have been focussed at the highest luminosities and at the peak epoch of galaxy formation at @xmath27 with colour cuts deliberately chosen to isolate high luminosity quasars with the same levels of dust extinction as seen in smgs at similar redshifts . to date
, we have spectroscopically confirmed a new sample of almost 60 heavily reddened , luminous broad line quasars with a median a@xmath28 2.5 mags ( c.f .
a@xmath29 for smgs ; @xcite ) .
we have already found evidence that a higher fraction of our reddened quasars are actively star forming relative to optical quasars @xcite consistent with these being transition objects seen as they are both rapidly forming stars and rapidly growing their black holes .
our reddened quasars are also among the most luminous quasars known at these epochs with recent results suggesting that their space density actually exceeds that of unobscured , uv luminous quasars at the highest luminosities @xcite .
the bolometric luminosities are @xmath1010@xmath30 erg s@xmath9 and black hole masses are 10@xmath31 - 10@xmath32 m@xmath3 inferred from the broad h@xmath33 emission lines in the nir spectra , consistent with the most luminous quasars found in optical surveys such as the sloan digital sky survey ( sdss ) . in order to determine whether the reddened broad line quasars are indeed a transition population between smgs and optical quasars , direct comparison of their molecular gas and ism properties to both smgs and optical quasars is now necessary . here
we discuss the first alma observations of four heavily reddened quasars from @xcite and @xcite ( b12 and b15 hereafter ) .
our new data is supplemented with two more detections of the @xmath4co(3 - 2 ) line in two quasars from our b12 parent sample @xcite to give a sample of six heavily reddened quasars with molecular gas measurements . in section [ sec : data ] we present the alma observations of the four new quasars . section [ sec : results ] discusses both the dust continuum and @xmath4co(3 - 2 ) line properties of the quasars and presents physical properties of the quasar host galaxies including star formation rates , dust and molecular gas masses . in section [ sec : discussion ]
we attempt to put our results in context with previous observations of molecular gas in both high redshift smgs and optical quasars to determine whether the dust and gas properties of the reddened quasars are in fact intermediate between the two . throughout this paper
we assume a flat @xmath34cdm cosmology with h@xmath35=0.7 , @xmath36 , @xmath37 .
in this work we make use of alma data obtained as part of the cycle 3 project 2015.1.01247.s ( pi : banerji ) . the aim is to detect , for the first time , the molecular gas reservoirs in four heavily reddened quasars from b12 and b15 via the @xmath4co(3 - 2 ) emission line .
we refer readers to b12 , b15 and @xcite for details of the parent sample of reddened ( a@xmath38 ) broad emission line quasars from which our targets are drawn .
three of the four quasars selected for alma observations - ulasj0123 + 1525 , ulasj1234 + 0907 and ulasj2315 + 0143 - correspond to three of the reddest quasars in our sample with inferred dust extinctions towards the quasar continuum of a@xmath39=4.0 , 6.0 and 3.4 mags respectively .
assuming the extinction is connected to star formation in the quasar host galaxy as discussed extensively in b12 and b15 , these broad line quasars would be the best candidates for highly star forming gas rich hosts .
the fourth quasar in this paper - vhsj2101 - 5943 - represents a quasar with average extinction ( a@xmath39=2.5 mags ) in our sample . as such
, it can be taken to be representative of the gas and dust properties of the larger sample of heavily reddened broad line quasars .
we searched for radio counterparts to these quasars in the vla - first , nvss and sumss radio catalogues and did not find any matches within 10@xmath40 .
the h@xmath33 derived redshifts , agn luminosities ( at rest - frame 6@xmath25 m ) and black hole masses for these quasars can be seen in table [ tab : properties ] .
we emphasise that apart from ulasj1234 + 0907 , which was detected in the _ herschel _ pacs and spire bands @xcite , none of the other three targets were known a - priori to be fir or millimetre bright .
this is in contrast to the targeting strategies for detecting molecular gas in uv luminous , unobscured quasars where only uv luminous quasars already known to be fir bright have typically been observed in molecular gas .
as discussed in b12 and @xcite , it is unlikely that our reddened quasars are lensed .
optical and near infrared @xmath41 band images of our quasars demonstrate that the reddened quasars are extremely faint or completely invisible at these wavelengths and , given the depths of these imaging data , a lensing galaxy should have been visible out to @xmath42 . [ cols="<,^,^,^,^",options="header " , ] we fitted a tilted ring disk model to the velocity field shown in fig .
[ fig:2315_spatial ] . in such an analysis ,
the gas emission is modelled as series of concentric rings with inclination , @xmath43 , position angle , p.a .
, systemic velocity @xmath44 , centre @xmath45 $ ] and rotational velocity , @xmath46 .
the line of sight velocity at any position @xmath47 $ ] can then be described on a ring with radius , @xmath48 as : @xmath49 where the angle , @xmath50 is related to p.a . in the plane of the sky by @xmath51
@xmath52 we used the _ rotcur _ task within gipsy in iterative mode for the fitting .
the expansion velocity was set to zero and the dynamical centre of all the rings was set to be identical .
the rings have widths ranging from 0.21 - 3@xmath40 .
the source has a maximum radial extent of 2.9@xmath40 or @xmath1023 kpc .
we fixed the p.a . to be that measured from the velocity field by looking at the axis connecting the maximum and minimum velocities .
is measured to be 352.3@xmath53 .
the _ rotcur _
task was run 15 times with one to fifteen rings covering the entire radius with only the systemic velocity and then the inclination angle , @xmath43 , as free parameters .
the systemic velocity and @xmath43 were then determined from the average of all successful rings where a successful ring is defined to be one with errors on @xmath43 of @xmath2220@xmath53 .
we have checked that using an error weighted average instead makes very little difference to the model fit parameters . finally , we ran the task with @xmath46 as the only free parameter and used the results from the most successful rings ( defined as those with uncertainties on @xmath54 ) in a single run to produce a model velocity field and a rotation curve for the source .
the best fit rotation curve can be seen in the bottom panel of fig .
[ fig:2315_spatial ] .
the best fit tilted ring model has @xmath43=57@xmath558@xmath53 .
given the resolution of our data , the model fails to constrain the internal dynamics of this source at @xmath56 kpc and the rotation curve shown in fig .
[ fig:2315_spatial ] is noisy and should be considered as illustrative .
however , given that these data were taken at the lowest spatial resolution available in alma band 3 , the fact that dynamical modelling is even possible with this data highlights the enormous promise of gaining detailed insight into the dynamics of such high redshift quasar host galaxies with higher resolution alma observations .
we defer a more detailed discussion of the dynamical properties of this source to future papers .
we can estimate the dynamical mass from the rotation curve shown in fig .
[ fig:2315_spatial ] .
the maximum rotational velocity , @xmath57=125@xmath556 km s@xmath9 at r=21.5 kpc . using @xmath58 , we get a dynamical mass of ( [email protected])@xmath510@xmath32m@xmath3 which is a factor of @xmath102 lower than that derived using the spectroastrometry method in section [ sec : dynamics ] . correcting for the full range in inclination angles allowed by the fit , m@xmath59=([email protected])@xmath510@xmath61m@xmath3 .
ulasj0123 + 1525 shows two distinct peaks in its co line profile , which could indicate the presence of kinematically distinct components in the gas emission from this galaxy .
a double gaussian fit yields a significantly better reduced @xmath62 value compared to a single gaussian ( 1.2 versus 2.6 for the single gaussian fit ) .
the two gaussians have fwhm=240@xmath5580 km s@xmath9 and 570@xmath5570 km s@xmath9 respectively and are separated in velocity by 260@xmath5560 km s@xmath9 . at the spatial resolution of our data , there is no evidence for a spatial offset between these velocity components .
similar double peaked line profiles have been observed in smgs ( e.g. @xcite ) and could arise from a galaxy merger , a rotating gas disk with a large inclination angle relative to the sky plane or large scale inflows and outflows affecting the molecular gas .
further high resolution observations of the molecular gas would help discriminate between these scenarios .
in order to understand whether our reddened quasars are indeed a distinct population in terms of their molecular gas and dust continuum properties , we now draw direct comparisons to observations of the molecular gas and dust continuum in samples of blue , unobscured quasars with comparable luminosities and redshifts . to date ,
the largest surveys for cold dust in high luminosity quasars at @xmath17 , are still those conducted using the scuba bolometer ( at 850@xmath25 m ) on the james clerk maxwell telescope and using the mambo bolometer ( at 1200@xmath25 m ) on the iram 30-m telescope ( e.g. @xcite ) .
the scuba surveys have typically probed down to 3@xmath26 flux limits of @xmath107 mjy at 850@xmath25 m while the mambo survey reaches sensitivities of @xmath102 - 4 mjy ( 3@xmath26 ) . in the scuba studies , only @xmath1020 per cent of the optically unobscured , blue quasars are submillimetre loud " with 850@xmath25 m flux densities of @xmath87 mjy .
we can estimate the 850@xmath25 m fluxes for the reddened quasars using the assumptions regarding the dust sed set out in section [ sec : dust ] .
these 850@xmath25 m flux densities are compared to other samples from the literature in fig .
[ fig : z_850 ] and the errorbars reflect the difference in 850@xmath25 m flux between assuming t@xmath63=47k and @xmath7=1.6 @xcite and t@xmath63=41k and @xmath7=1.95 @xcite .
three of the four reddened quasars lie near or below the 2@xmath26 limit for the scuba surveys . from this initial sample of reddened quasars , and with the assumptions regarding their dust sed set out above , @xmath1025 per cent of the reddened quasars appear to be
submillimetre loud " , similar to the submillimetre loud fraction of blue quasars .
we can also compare the submillimetre and millimetre flux densities for our reddened quasars with the stacked results from the scuba surveys , which suggest that the average 850@xmath25 m flux density of a flux limited sample of blue quasars is [email protected] mjy at @xmath24 @xcite and [email protected] mjy at @xmath64 @xcite . based on our alma detections ,
our reddened quasars would all have 850@xmath25 m fluxes that are higher than these average properties for the blue quasars as can be seen in fig .
[ fig : z_850 ] . the larger 850@xmath25 m fluxes could suggest that luminous reddened quasars have higher dust luminosities and are therefore , on average , more likely to be hosted in starburst galaxies relative to their unobscured blue counterparts .
however , larger samples of reddened quasars with cold dust continuum measurements and better constraints on their dust sed are clearly required before any robust conclusions can be drawn .
we also compare the submillimetre fluxes of our reddened quasars to those of x - ray selected agn .
x - ray surveys of high redshift agn have generally covered smaller areas on the sky compared to optical surveys and therefore probe quasars and agn at lower luminosities . considering only the most x - ray luminous , high redshift quasars in the cosmos field with scuba-2 850@xmath25 m observations from the scuba-2 cosmology legacy survey , the stacked 850@xmath25 m fluxes are [email protected] mjy for type 1 agn and [email protected] mjy for type 2 agn @xcite , once again lower than those measured for the reddened quasars in this work .
there are however a number of x - ray agn that are individually detected at 850@xmath25 m in @xcite with 850@xmath25 m flux densities of @xmath104 - 10 mjy , comparable with the inferred submillimetre fluxes of our reddened quasars .
previous scuba surveys of x - ray absorbed and x - ray unabsorbed quasars have suggested a link between cold dust emission and x - ray obscuration @xcite and we note that for those reddened quasars where we have x - ray spectra ( e.g. @xcite ) , moderate column densities of @xmath1010@xmath65 @xmath66 are present .
surveys for molecular gas in blue quasars have focussed on the subset of the population that are ( sub)millimetre loud , in contrast to our blind " search for co in the reddened quasar population .
so far six of our near infrared selected , heavily reddened quasars from the b12 and b15 samples have been observed in co - four in this paper and one each in @xcite and @xcite . despite not being selected a - priori to be fir / mm bright ,
all six quasars have been detected in co. as stated earlier , most optical quasars that have been observed in co were known a - priori to be fir luminous therefore increasing the probability of a co detection . @xcite
have looked for co in a sample of 10 quasars of comparable redshift to our reddened quasar sample .
three quasars in their sample were detected in blank field submm surveys over much smaller areas of sky than the wide - field near infrared surveys that have been used to select our reddened quasars ( b12,b15 ) , and are therefore lower luminosity .
the remaining seven quasars in the @xcite sample , are brighter than the @xmath67-band flux limit imposed on our reddened quasar searches in b12 and b15 and can therefore be directly compared to our sample . from these seven fir bright optical quasars , five
have been detected in co with the limits on the co line luminosity for the remaining two placing them below the co line luminosities of all six of our reddened quasars .
presumably , starting from a complete sample of unobscured quasars ( i.e. not just those known a - priori to be millimetre bright ) would result in a smaller fraction being detected in co. the black hole masses of our reddened quasars and the uv - luminous @xcite quasars are comparable given the typical uncertainties in these black hole mass estimates .
hence , the co - detection fraction in our sample suggests that a larger fraction of luminous reddened quasars have significant molecular gas reservoirs compared to uv luminous blue quasars of comparable luminosity and mass . in b15 we demonstrated that our reddened quasars outnumber blue quasars at the highest luminosities whereas the reddened quasar population is subdominant as we approach more typical luminosities around l@xmath68 .
one interpretation of these trends put forward in b15 is that the most luminous , massive black holes have a longer duty cycle associated with their assembly or growth phase and therefore are seen as dust obscured for a longer period of time .
as all our alma targets are drawn from the high luminosity , massive end of the reddened quasar population , the detection of significant amounts of cold dust and molecular gas in these systems fits in with this overall picture .
in such an evolutionary sequence , ( sub)millimetre - loud blue quasars are presumably a later , more evolved phase when significant gas reservoirs and star formation are still detectable but some of the dust has been cleared out .
the ( sub)millimetre - faint blue quasars may then correspond to an even more advanced phase when significant star formation has ceased either because the gas supply has been depleted , or as a result of agn feedback . under the premise that dust reddened quasars are a phase in massive galaxy evolution preceding the unobscured quasar phase , and that this phase is associated with high levels of star formation and high rates of accretion onto the supermassive black hole , we would predict that populations of high luminosity agn that are even more obscured than our reddened quasars may have similar or higher levels of star formation in their host galaxies .
these most highly obscured agn might therefore be expected to be characterised by high 850@xmath25 m fluxes .
the most suitable sample of mid - infrared luminous , heavily obscured agn with comparable agn luminosities and redshifts as our reddened quasars are the so called agn dominated hot dust obscured galaxies or hotdogs that have emerged from the _ wise _ all sky survey ( e.g. @xcite ) .
we use as the basis for comparison , the scuba-2 850@xmath25 m and alma 870@xmath25 m flux densities of the spectroscopically confirmed _ wise _ hotdogs at @xmath69 from @xcite and @xcite .
these studies provide a sample of 42 hotdogs with secure spectroscopic redshifts overlapping the redshift range of both our reddened quasar sample and the @xcite optical quasars .
once again we extrapolate our reddened quasar 3 mm flux densities using the sed assumptions above to predict their 850@xmath25 m fluxes .
a comparison of the 850@xmath25 m flux densities can be seen in fig .
[ fig : z_850 ] .
m flux density for blue quasars @xcite , reddened quasars ( this work , @xcite ) and _ wise _ hotdogs @xcite . in all cases
downward triangles denote upper limits and the large blue open circle shows the stacked 850@xmath25 m flux of the blue quasars from @xcite .
the dashed lines mark the range in 2@xmath26 flux limits for the scuba survey @xcite and the alma survey @xcite.,title="fig : " ] + out of the 42 wise hotdogs , 19 are detected at high significance at 850/870@xmath25 m - a detection fraction of @xmath1045 per cent .
if we consider the fraction detected to the scuba limits of @xmath107 mjy , the number drops to only 3 - i.e. just 7 per cent of the population . as seen in fig .
[ fig : z_850 ] , in many cases the limits on the submillimetre fluxes obtained for the _ wise _ hotdogs , place them well below the flux densities of our reddened quasars .
thus , as concluded in @xcite and several other recent _ wise _ papers , these hotdogs on average do not show evidence for the dust luminosities being dominated by cold dust heated by star formation and instead , it is likely that most of the dust heating in these galaxies comes from the agn .
the _ wise _ hotdogs therefore do not seem to fit our simple evolutionary picture linking obscured and unobscured quasars . as suggested in @xcite
, these _ wise _ agn could perhaps represent a class of objects where the peak of agn activity precedes that of star formation .
alternatively , the higher line - of - sight extinction seen in these sources could simply reflect a difference in viewing angle between these _ wise _ agn and both the reddened and unobscured broad line quasar populations .
there have been relatively few studies of co molecular gas in the _ wise _ population so far but @xcite have looked for @xmath4co(3 - 2 ) and @xmath4co(4 - 3 ) in two of these galaxies .
neither source is detected in their observations but the limits on the gas mass from that study are consistent with the low end of the gas masses we have derived for the reddened quasar population in this paper .
+ it is clear from fig .
[ fig : z_850 ] that the number of dust continuum observations for optical and _ wise_-selected agn in much larger than the number available for reddened quasars . with a large spectroscopic sample of the reddened quasars now in place ,
it is imperative that we extend the investigations carried out in this work to larger samples in order to gain a coherent picture of the cold dust and molecular gas emission and star formation in these different quasar populations covering a very wide range in line - of - sight extinction .
having detected both cold dust and molecular gas in dust reddened quasars , we can attempt to compare the star formation efficiencies and gas depletion timescales to those measured in high redshift starburst galaxies and quasars .
we use as the basis for our comparisons , the compilations of co detections in high redshift galaxies and quasars published by @xcite and @xcite .
while these compilations are not necessarily exhaustive and complete , they can be taken to be representative of current observational samples . throughout this analysis we work with the full sample of six heavily reddened quasars with
co detections - i.e. the four in this paper plus the two from @xcite and @xcite .
for the last two quasars we have re - derived the co line luminosities using the same assumptions regarding the co excitation as in this work ( see table [ tab : derived ] ) . in fig .
[ fig : lir_lco ] we show the co line luminosity , l@xmath70 , versus the infrared luminosity for all extragalactic sources from @xcite and the six heavily reddened quasars in table [ tab : derived ] .
also shown is the best - fit relation from @xcite : @xmath71 the reddened quasars in general have higher infrared luminosities or star formation rates for a given gas mass , than predicted by this relation , which has been derived from the entire distribution of high redshift galaxies in the literature .
this is consistent with a high star formation efficiency in these reddened quasar hosts compared to other extragalactic sources the infrared luminosities are however broadly consistent with those of other high redshift quasar host galaxies and the scatter in the observed relation shown in fig .
[ fig : lir_lco ] is large .
we caution also that the infrared luminosities in the literature have been calculated using heterogenous methods both in terms of the wavelength range used to compute these luminosities and the assumptions made regarding the form of the dust sed .
to gain further insight into the similarities and differences in the star formation efficiencies between smgs , quasars and reddened quasars , we now compare both the fir luminosity distributions and co luminosity distributions over the same redshift range , using a ks - test to determine the probability that they are drawn from the same parent population .
there is no evidence for a difference in the fir luminosity distributions of the three samples when the same assumptions are made regarding the dust temperature and emissivity index of the fir sed .
comparing the l@xmath72 luminosities for the three samples however , we find that quasars and smgs have a 3 per cent probability of being drawn from the same distribution , highlighting that the molecular gas reservoirs may indeed be different in the two samples . reddened quasars on the other hand have a 32 per cent probability of having co luminosities drawn from the same luminosity distribution as smgs and a 54 per cent probability of being drawn from the same luminosity distribution as optical quasars so their gas properties are consistent with both smgs and optical quasars .
we calculate typical gas depletion timescales using the gas masses and star formation rates in table [ tab : derived ] . regardless of whether the co based or dust based gas masses are used , the gas depletion timescales are very short - @xmath13 myr .
the black hole mass accretion rates of @xmath10100 - 500 m@xmath3yr@xmath9 ( b12,b15 ) imply black hole gas consumption timescales that are at least an order of magnitude longer - @xmath10100 - 1000 myr .
the black holes are already very massive - @xmath73m@xmath3 - suggesting we are witnessing a phase when the galaxies are running out of gas ( to further fuel the black hole ) following the main period of black - hole mass assembly .
the reddened quasars could thus be plausible progenitors of the largest supermassive black holes seen in the universe today , which have measured black hole masses of @xmath7410@xmath32m@xmath3 ( e.g. ngc 3842 and ngc 4899 ; @xcite ) . .
the lines show the best - fit relation from @xcite given in eq .
[ eq : lir_lco ] together with the dispersion in this relation . ]
a detailed discussion of the sizes of the co emitting regions in reddened type 1 quasars is beyond the scope of this paper and will have to wait for higher resolution observations of the gas distribution in these sources .
however , we have seen that the co emission is already resolved in one of our reddened quasars - ulasj2315@xmath140143 ( section [ sec:2315 ] ) - with an implied source diameter of @xmath7420 kpc .
there are hints also that the gas emission is marginally resolved in ulasj1234 + 0907 and vhsj2101@xmath605943 .
we used casa to fit a 2-d gaussian ( in ra , dec ) to the co maps constructed by summing all the channels covering the full extent of the line .
the deconvolved source sizes along the major axis are ( [email protected] ) arcsec @xmath75(20@xmath554 kpc ) and ( [email protected] ) arcsec @xmath75(11@xmath552 ) kpc for ulasj1234 + 0907 and vhsj2101@xmath605943 respectively .
we note that the images used for the size estimates were generated using _
natural _ weighting , which results in strong sidelobes in the image plane .
we have therefore checked that the sizes quoted here are consistent within the error bars with the deconvolved sizes estimated from images generated using a _ briggs _ weighting ( robust=0.5 ) , where the effects of the sidelobes should be minimised . finally , for our most extended and highest s / n source , ulasj2315@xmath140143 , we have also checked that the major axis derived from directly fitting to the visibility data using the casa task _ uvmodelfit _ , is once again consistent with the deconvolved sizes from the image plane quoted here .
the large implied sizes for the warm gas reservoirs in these quasars , while still highly uncertain , are nevertheless at odds with previous observations of gas in quasar host galaxies . in the few cases where the gas emission has been resolved in high redshift quasar host galaxies ,
it is distributed in compact regions of @xmath22 few kpc in size ( e.g. @xcite ) .
the conventional interpretation is that the molecular gas in quasar hosts resides in a nuclear region close to the accreting black hole , and its excitation properties can therefore be explained by a single temperature and density gas component .
smgs on the other hand often show evidence for substantially larger reservoirs of low excitation gas @xmath810 kpc in size ( e.g. @xcite ) .
although the excitation properties of the gas in reddened quasars remains to be determined , we tentatively conclude that the co sizes appear to more closely resemble smgs rather than high redshift quasar hosts . with the co detections providing us with molecular gas masses and the co velocity widths providing a crude estimate of the dynamical masses , we can estimate the molecular gas fractions and compare directly to gas fractions in both smgs and optical quasars . a ks - test comparing the observed fwhm of the co lines in smgs , optical quasars and reddened quasars reveals no evidence for a difference in the distributions between the three populations .
if the three populations indeed form an evolutionary sequence with the optical quasars representing more evolved systems and smgs representing the earlier gas rich stages of massive galaxy formation , we would expect the reddened quasars to have gas fractions that are intermediate between smgs and optical quasars . in fig .
[ fig : fwhm_lco ] we show the fwhm of the co line - a proxy for the dynamical mass - versus the co(1 - 0 ) luminosities - a proxy for gas mass .
we compare our sample to optical quasars from @xcite and @xcite and to the smgs from @xcite , selecting quasars and smgs that are unlensed only .
we also show the scaling relation between these two quantities derived for unlensed sources by @xcite , as well as a simple relation that equates the gas mass to the dynamical mass as follows : @xmath76 where the constant , @xmath77 depends on the geometry and inclination of the source , @xmath26 represents the line dispersion @xmath10fwhm/2.35 for a gaussian line , @xmath78 is the source size and @xmath79 is the gravitational constant .
we consider two separate cases : ( i ) @xmath80 , @xmath81 kpc , @xmath82=4.6 m@xmath3 ( k km s@xmath9 pc@xmath83)@xmath9 , which is appropriate for disk galaxies and ( ii ) @xmath84 , @xmath85 kpc , @xmath82=1 m@xmath3 ( k km s@xmath9 pc@xmath83)@xmath9 , representing a virialised , spherical source . for both disk like and spherical geometries . ]
we find that the reddened quasars occupy a region roughly between the majority of the optical quasars and the smgs in fig .
[ fig : fwhm_lco ] .
however , the observed fwhm is also sensitive to inclination with thin rotating disks viewed face - on having narrower fwhm . @xcite
have argued based on the fact that smgs show a correlation between co linewidth and line luminosity , that they are primarily thick disks / turbulent ellipsoids where the fwhm is not as dependent on inclination .
however , optical quasars , where broad emission lines can be seen in the quasar spectra , may have an inclination bias relative to smgs although the opening angles needed for detection of broad lines in quasars likely span a large range . if the molecular gas sits in a thin disk in these quasars , the optical quasars with very narrow fwhm that sit above the various lines in fig .
[ fig : fwhm_lco ] , could be seen more face - on . on the other hand , if the gas morphologies are more turbulent and inhomogenous - e.g. as would be expected for merging systems - the fwhm is unlikely to be as dependent on inclination .
further observations are required to established whether the reddened quasars have any inclination bias relative to optical quasars and smgs .
we note that there is a general trend among our objects for the more highly obscured quasars to have larger observed co fwhm - e.g. ulasj1234 + 0907 , the dustiest quasar in our sample ( a@xmath86 mag ; b12 ) has the broadest co line whereas vhsj2101@xmath605943 , the least dusty quasar observed ( a@xmath87 mag ; b15 ) , has the narrowest co line , suggesting that some of the reddening in these quasars could be linked to orientation .
however , in section [ sec:2315 ] we have seen that in at least one of our quasars the gas appears to be distributed in a disk at an inclination angle of 57@xmath53 .
this is consistent with the inclination angles inferred for smgs .
if smgs , reddened quasars and optical quasars have similar source sizes and inclinations on average , for a fixed dynamical mass or co fwhm , reddened quasars have gas masses that are intermediate between smgs and optical quasars , and are therefore on their way to exhausting their gas supply and evolving into optical quasars .
further spatially resolved observations of the reddened quasars to get rotational dynamical parameters as well as internal velocity dispersions , is clearly the next step forward in order to get robust gas fractions in these sources .
we have presented alma band 3 observations of a sample of four heavily reddened ( a@xmath88 mag ) , high luminosity broad line quasars at @xmath1 selected from wide field near infrared surveys . the aim of these observations is to test whether such reddened quasars could represent the evolutionary link between starburst galaxies and unobscured , optically bright quasars at high redshifts .
all four quasars are detected in both the dust continuum ( at observed frame wavelengths of @xmath103 mm ) and in the @xmath4co(3 - 2 ) line .
we infer star formation rates , dust and gas masses from these observations and attempt to place these measurements in context with corresponding values for other high redshift galaxy populations : * the dust masses range from 2.5@xmath510@xmath6m@xmath3 to 1.2@xmath510@xmath31m@xmath3 .
the fir luminosities are @xmath810@xmath89l@xmath3 and the implied star formation rates in the quasar host galaxies is @xmath90 m@xmath3 yr@xmath9 , suggesting that these reddened quasars reside in prodigiously star forming host galaxies with very high dust luminosities .
* we estimate gas masses from the measured co(3 - 2 ) line luminosities , assuming an excitation ratio @xmath91=0.8 and a co - to - h@xmath92 conversion factor of @xmath93 ( k km s@xmath9 pc@xmath83)@xmath9 .
these gas masses are in the range @xmath9410@xmath32m@xmath3 and the gas - to - dust ratios are in the range @xmath1030 - 110 . * instead , adopting a gas - to - dust ratio of 91 , as found in nearby galaxies , dust- and co - based gas masses agree within a factor of @xmath102 for three of the quasars . for ulasj2315 + 0143
, the dust - based gas mass is @xmath103 times larger .
the co- and dust - based masses could be made consistent for ulasj2315 + 0143 if a co - to - h@xmath92 conversion factor of @xmath95 ( k km s@xmath9 pc@xmath83 ) applies .
such a value is intermediate between that found in nearby galaxies / high - redshift star - forming disks and in nuclear starbursts / quasar host galaxies . *
the co linewidths span @xmath96 kms@xmath9 , comparable to linewidths in other high - redshift smgs and quasars . using several different estimators , dynamical masses of m@xmath59sin@xmath83(i ) @xmath103@xmath510@xmath32m@xmath3 to @xmath103@xmath510@xmath61m@xmath3 are found .
* ulasj2315 + 0143 ( @xmath97 ) shows evidence for spatially resolved co emission over @xmath1020 kpc , with a strong velocity gradient across the source .
we demonstrate that the velocity field is consistent with the presence of a large , rotating gas disk although other scenarios such as mergers can not be ruled out at the resolution of the current data .
ulasj2315@xmath140143 is the quasar with the largest discrepancy between its dust and co based gas masses , and where there is evidence for @xmath98 ( k km s@xmath9 pc@xmath83)@xmath9 . *
the sizes of the @xmath4co(3 - 2 ) emitting regions appear to be large ( @xmath8 10 kpc ) hinting at the presence of extended reservoirs of warm gas .
gas emission in high - redshift , unobscured quasars is generally compact ( @xmath22few kpc ) . if extended emission is confirmed via higher resolution observations , the reddened quasars would instead be more similar to the properties of smgs .
* ulasj1234 + 0907 ( @xmath16 ) has two other millimetre bright galaxies at the same redshift in the field - of - view , with implied gas masses of a few times 10@xmath32m@xmath3 .
the two galaxies are located @xmath1090 kpc and @xmath10170 kpc from the quasar .
ulasj1234 + 0907 may reside in a significant over - density at @xmath1 . *
a similar fraction of reddened quasars are submillimetre loud " compared to unobscured uv luminous quasars .
however the average dust luminosities of reddened quasars would appear to be higher than for the uv luminous quasar population at similar redshifts .
reddened quasars also appear to have higher cold dust luminosities compared to the recently discovered population of hot dust obscured galaxies from _
wise_. so far , 100 per cent of our reddened quasars ( 6/6 ) have been observed to have very large molecular gas reservoirs .
a direct comparison to the unobscured quasar population is not possible as surveys of molecular gas in uv luminous quasars have focussed on the subset of these that were already known to be millimetre bright , therefore increasing the probability of detecting molecular gas .
nevertheless , several uv luminous , millimetre bright quasars with comparable black hole masses and agn luminosities to our sample , have limits on their molecular gas masses that place them below the gas masses seen so far in the reddened quasar population , indicating that these unobscured quasars could indeed correspond to a later evolutionary phase .
further observations of molecular gas in larger samples of both reddened and unobscured quasars matched in luminosity and black hole mass would help confirm this trend .
* using the co linewidths in our sample as a proxy for the dynamical mass , and the co line luminosities as a proxy for gas mass , we find that the reddened quasars have molecular gas fractions that are intermediate between smgs and optical quasars assuming that all three populations have similar source sizes and inclinations
. it therefore seems plausible that reddened quasars are being seen as they are transitioning to optical quasars from starburst galaxies .
their supermassive black holes already appear to be fully assembled and , given the rate of ongoing star formation , they will very quickly exhaust their gas reservoirs and presumably blow out the surrounding gas and dust on their way to evolving into some of the most luminous , unobscured quasars - progenitors of the largest supermassive black holes seen in the universe today .
overall , our observations demonstrate the high scientific return possible from alma observations of luminous quasars at the main epoch of galaxy formation .
studies of the host galaxy properties of intrinsically luminous , unlensed quasars at @xmath17 are still scarce and further observations with alma both in terms of assembling larger samples of such sources and obtaining higher spatial resolution detections , will help shed light on the exact nature of the host galaxies of the largest supermassive black holes , seen as they are being assembled in the high redshift universe .
the authors thank the referee for a useful and detailed report which helped improve the paper .
mb would like to thank matt bothwell and stefano carniani for useful discussions .
mb acknowledges funding from the uk science and technology facilities council ( stfc ) via an ernest rutherford fellowship .
gj is grateful for support from nrao through the grote reber doctoral fellowship program .
rgm and pch acknowledge funding from stfc via the institute of astronomy , cambridge consolidated grant .
sa - z acknowledges support from peterhouse , cambridge .
this paper makes use of the following alma data : ads / jao.alma#2015.1.01247.s .
alma is a partnership of eso ( representing its member states ) , nsf ( usa ) and nins ( japan ) , together with nrc ( canada ) , nsc and asiaa ( taiwan ) , and kasi ( republic of korea ) , in cooperation with the republic of chile .
the joint alma observatory is operated by eso , aui / nrao and naoj . | we present alma observations of cold dust and molecular gas in four high - luminosity , heavily reddened ( a@xmath0 mag ) type 1 quasars at @xmath1 with virial m@xmath2m@xmath3 , to test whether dusty , massive quasars represent the evolutionary link between submillimetre bright galaxies ( smgs ) and unobscured quasars .
all four quasars are detected in both the dust continuum and in the @xmath4co(3 - 2 ) line .
the mean dust mass is 6@xmath510@xmath6m@xmath3 assuming a typical high redshift quasar spectral energy distribution ( t=41k , @xmath7=1.95 or t=47k , @xmath7=1.6 ) .
the implied star formation rates are very high - @xmath81000 m@xmath3 yr@xmath9 in all cases .
gas masses estimated from the co line luminosities cover @xmath101 - 5@xmath11(@xmath12)m@xmath3 and the gas depletion timescales are very short - @xmath13myr .
a range of gas - to - dust ratios is observed in the sample .
we resolve the molecular gas in one quasar - ulasj2315@xmath140143 ( @xmath15 ) - which shows a strong velocity gradient over @xmath1020 kpc .
the velocity field is consistent with a rotationally supported gas disk but other scenarios , e.g. mergers , can not be ruled out at the current resolution of these data . in another quasar - ulasj1234 +
0907 ( @xmath16 ) - we detected molecular line emission from two millimetre bright galaxies within 200 kpc of the quasar , suggesting that this quasar resides in a significant over - density .
the high detection rate of both cold dust and molecular gas in these sources , suggests that reddened quasars could correspond to an early phase in massive galaxy formation associated with large gas reservoirs and significant star formation .
galaxies : evolution galaxies : formation galaxies : high - redshift galaxies : starburst |
for @xmath0 , let @xmath1 be a vector of holomorphic germs at infinity . given a multi - index @xmath2 , the _ type ii hermite - pad approximant _ to @xmath3 corresponding to @xmath4 , @xmath5 is a vector of rational functions with common denominator @xmath6 satisfying @xmath7 for each @xmath8 .
we shall not deal with type i hermite - pad approximants in this work and therefore henceforth we will drop the `` type ii '' modifier . such approximants were introduced by hermite @xcite for the vector of exponentials @xmath9 , with the interpolation taking place at the origin rather than at infinity , as a tool in proving the transcendence of @xmath10 .
later , his student pad systematically studied the scalar case @xmath11 @xcite and such approximants are now called _
pad approximants_. from our perspective interpolating at infinity is more convenient than interpolating at the origin in the following sense .
any holomorphic function can be written as a cauchy integral of its boundary values on any curve encircling a domain of analyticity . for a holomorphic function on a domain
we will use the terminology _ trace _ to mean the boundary values of the function on the boundary of the domain .
when the function is holomorphic at infinity , such an integral representation can in some important cases be deformed analytically into an integral over a `` one dimensional '' set .
a particular fruitful example of this principle is when we deal with markov functions .
let @xmath12 be a positive measure and take @xmath13\subset{\mathbb{r}}.\ ] ] the vector @xmath3 is called an _ angelesco system _ if @xmath14\cap[a_i , b_i]=\varnothing$ ] for @xmath15 ( such systems were initially considered by angelesco @xcite and later rediscovered by nikishin @xcite ) .
angelesco has shown that @xmath6 has exactly @xmath16 zeros on @xmath17 $ ] .
this means that an angelesco system is an example of a _ perfect _ system , i.e. , a system for which every multi - index is _ normal _ , that is , @xmath18 .
as far as the asymptotic behavior of @xmath19 is concerned , convergence properties are not as straightforward as one could hope . given an angelesco system and a sequence of multi - indices such that @xmath20 , @xmath21 , gonchar and rakhmanov @xcite have shown that for each @xmath22 the complement of @xmath23 $ ] is separated by a system of analytic arcs into two domains , say @xmath24 , containing the point at infinity , and @xmath25 , possible empty , such that @xmath19 converges to @xmath26 in @xmath24 and diverges to infinity in @xmath25 .
moreover , the polynomials @xmath6 can have an asymptotic zero distribution ( in the sense of weak@xmath27 convergence ) on a strict subset of @xmath28 ( _ pushing effect _ ) .
the pushing effect always implies existence of a divergence region but the reverse implication is not true .
more detailed ( strong ) asymptotics for hermite - pad approximants to angelesco systems when @xmath29 and the weights @xmath30 satisfy the so - called _ szeg condition _ , was obtained by the first author in @xcite .
another class of markov functions for which positive convergence results were obtained is now known as _
_ @xcite . the functions @xmath31 in form a nikishin system if they all are supported on the same interval @xmath32 $ ] and the radon - nikodym derivatives @xmath33 , @xmath34 , form a nikishin system of order @xmath35 with respect to some interval @xmath36 $ ] such that @xmath36\cap[a , b]=\varnothing$ ] .
nikishin himself @xcite has shown that such systems are perfect when @xmath29 and the hermite - pad approximants converge uniformly outside of the interval @xmath32 $ ] in this case .
this puts nikishin systems more in line with the pad case @xmath11 ( markov theorem @xcite ) as neither the pushing effect nor the possibility of non - empty divergence regions appears for them .
however , nikishin systems do possess one new phenomenon , namely overinterpolation .
it turns out that @xmath37 has zeros on @xmath36 $ ] that are dense on this interval .
it took 30 years to prove that nikishin systems are perfect for any @xmath38 @xcite . in @xcite ,
fidalgo prieto and lpez lagomasino also proved uniform convergence for multi - indices close to the diagonal .
strong asymptotics in the case of diagonal multi - indices and szeg weights was derived by the first author in @xcite .
it is interesting to observe that the first result on strong asymptotics of hermite - pad approximants was obtained by kalaygin @xcite for the case of two touching symmetric intervals ( the limiting case of an angelesco system ) .
as often happens in mathematics , the treatment of angelesco and nikishin systems can be unified under the umbrella of _ generalized nikishin systems _
( gn - systems ) as introduced in @xcite , where gonchar , rakhmanov , and sorokin defined a system of markov functions with the help of a rooted tree graph and considered the question of uniform convergence of hermite - pad approximants to such a system . in such a set - up
, an angelesco system corresponds to a tree where the root is connected by @xmath38 edges to @xmath38 leaves and a nikishin system corresponds to a tree in which every node except for the final leaf has exactly one child .
strong asymptotics of hermite - pad approximants to gn - systems of markov functions generated by more general ( than rooted tree ) graphs ( admitting cycles ) was derived by the first author and lysov in @xcite .
an example of gn - system from @xcite is a pair of two markov functions in where the support of one of them is strictly included in the support of the other , i.e. , @xmath39 \subset { \mathsf{supp}}(\mu_1)=[a_1,b_1 ] $ ] and the radon - nikodym derivative @xmath40 along @xmath41 $ ] is a markov function with support @xmath42 $ ] where @xmath43\cap[a_1,b_1]=\varnothing$ ] .
weak asymptotics of hermite - pad approximants to this example of markov functions was derived by rakhmanov in @xcite .
we emphasize that in all the results we listed above the _ geometry _ of the problem is _ real _
, i.e. , the supports of the limiting distributions of the poles of the approximants and the overinterpolation points belong to @xmath44 . in this paper
we consider hermite - pad approximants to a pair of cauchy transforms ( markov functions ) of generally speaking complex measures with overlapping supports and aim at strong asymptotics .
the set up does not fall into the framework of gn - systems , even when the measures are positive , as their supports overlap ( for gn - systems the supports are either disjoint or one coincides with or contains the other ) .
it turns out that both phenomena , the pushing effect and overinterpolation , appear in this case .
hermite - pad approximants to a pair of markov functions with overlapping supports were first considered by stahl with the goal of proving weak asymptotics @xcite .
he had the important insight that the geometry of this problem is complex , i.e. , the overinterpolation points are distributed on analytic arcs in the complex plane ( later , a similar effect was observed in @xcite ) .
this discovery was very unusual at the time because the input geometry ( i.e. , the supports of the measures generating the markov functions ) is completely real .
unfortunately , stahl s results have never been published .
this work was strongly motivated by the desire to provide a detailed proof of his findings ( in an even more delicate setting of strong asymptotics ) .
opting here for complex measures is natural from the point of view of complex analysis .
however , many techniques , like those in @xcite , do not apply as they use positivity in an essential way .
an approach that does not rely on positivity was outlined by nuttall in his seminal paper @xcite .
there nuttall conjectured that the main term of the asymptotics of hermite - pad approximants is a function solving of a certain _ explicit _ boundary value problem on some _ unknown _ riemann surface .
he identified this surface only in a handful of special cases .
elaborating on nuttall s approach , the first two authors and kuijlaars @xcite pinpointed the algebraic equation which defines the appropriate riemann surface in the case of two cauchy transforms of complex measures supported on two arcs joining pairs of branch points in the complex plane ( the simplest example is a complexified angelesco system ) and derived formulae of strong asymptotics in the case when the riemann surface has genus zero .
below , we build upon the ideas developed in @xcite and extend the results of @xcite to the cases when the appropriated riemann surface has positive genus ( elliptic and ultra- elliptic case ) . in section [ sec : rs ] we identify the riemann surface in nuttall s program by an algebraic equation , discuss its realization as a ramified cover of @xmath45 , and construct a certain function on this surface whose level lines will geometrically describe convergence and divergence domains of the approximants . in section [ sec : ns ]
, we construct the nuttall - szeg functions that will provide the leading term of the asymptotics of the hermite - pad approximants .
finally , in section [ sec : hp ] we state the main result of this work .
the remaining part of the paper is devoted to the proofs of all the stated results .
let @xmath46 be given . our goal is to investigate hermite - pad approximants to a pair of markov - type functions generated by measures with supports @xmath47 $ ] and @xmath48 $ ] . to this end
we consider the algebraic equation @xmath49 where the polynomials @xmath50 , @xmath51 , and @xmath52 are defined by @xmath53 for some parameter @xmath54 .
denote by @xmath55 , @xmath56 , the three distinct branches of the algebraic function @xmath57 determined by .
naturally , these branches satisfy @xmath58 hence , it necessarily holds that @xmath59 and therefore @xmath60 as @xmath61 .
it readily follows from the above equations that the constants @xmath62 are the solutions of @xmath63 .
thus , we put @xmath64 for @xmath65 .
it can easily be checked that all three solutions of are real for positive large @xmath66 .
hence , we can label the branches so that for all @xmath67 large enough @xmath68 denote by @xmath69 the riemann surface of @xmath57 .
it is a three - sheeted ramified cover of @xmath45 .
we shall denote by @xmath70 , @xmath56 , the sheet on which @xmath57 coincides with @xmath55 ( a particular realization of this surface is specified in theorem [ thm : rs ] below ) .
we denote by @xmath71 a generic point on @xmath69 with _ natural projection _
@xmath72 . if we want to specify the sheet , we write @xmath73 for @xmath74 so that @xmath75 .
thus , @xmath76 .
consider the differential @xmath77 where @xmath71 is a generic point on @xmath69 .
the choice of the parameter @xmath38 is driven by the following condition : @xmath78 when it exists , @xmath79 is defined up to an additive constant . if we denote by @xmath80 the restriction of @xmath81 to @xmath70 , then it is easy to see that @xmath82 is a well defined harmonic function in @xmath45 and therefore it is constant . thus , we normalize @xmath81 so that @xmath83 notice also that is equivalent to requiring that all the periods of the differential @xmath84 are purely imaginary . [ thm : rs ]
consider the algebraic equation with @xmath46 . * if @xmath85 , then there exists @xmath86 such that condition is fulfilled . in this case
@xmath69 has eight ramification points whose projections are @xmath87 and @xmath88 for some uniquely determined @xmath89 and @xmath90 .
moreover , the surface can be realized as on figure [ rs](a ) ; * if @xmath91 , then condition is fulfilled for @xmath92 . in this case
@xmath69 has four ramification points whose projections are @xmath93 and it can be realized as on figure [ rs](b ) ; * if @xmath94 , then condition is fulfilled for for @xmath95 . in this case @xmath69
has six ramification points whose projections are @xmath87 and @xmath96 for some uniquely determined @xmath97 .
moreover , the surface can be realized as on figure [ rs](c ) .
the points @xmath98 and @xmath99 in case i and the point @xmath98 in case iii can be explicitly computed as they are solutions of a certain explicit quadratic or linear ( in @xmath100 ) equation whose parameters depend on @xmath101 and @xmath38 .
it follows from theorem [ thm : rs ] that @xmath69 has genus @xmath102 when @xmath103 and genus @xmath104 otherwise
. moreover , in cases i and iii , all the ramification points have order 2 while in case ii the points @xmath105 have order 2 while @xmath106 have order 3 . in the light of theorem [ thm : rs ]
, it will be convenient to fix the following notation : @xmath107 clearly , @xmath108 is defined only in case i. let @xmath84 be defined by .
the function @xmath109 is analytic , except for a double pole at @xmath110 , and multiplicatively multi - valued on @xmath69 .
moreover , it is single - valued in @xmath111 , where @xmath112 is a homology basis on @xmath69 . later on , see figure [ fig : hb ] , we shall specify the basis in more detail , but right now it is sufficient to assume that each cycle @xmath113 possesses a _ projective involution _ : @xmath114 for any @xmath115 which is not a branch point of @xmath69 ( the involution is then defined by mapping a point on @xmath116 to the other one with the same projection ) .
we normalize @xmath117 so that @xmath118 where @xmath119 is the pullback to @xmath120 of the restriction of @xmath117 to @xmath70 .
let us show that such a normalization is indeed possible . since @xmath121 is simply connected and @xmath84 has integer residues , the restriction of @xmath117 to @xmath121 is single - valued .
it satisfies @xmath122 @xmath123 , where the constants @xmath124 and @xmath125 are real ( this is guaranteed by ) and given by @xmath126 at times , it will be convenient to use vector notation @xmath127 and @xmath128 if @xmath102 and @xmath129 and @xmath130 if @xmath104 .
since @xmath117 has a double pole at @xmath110 and simple zeros at @xmath131 , we can write @xmath132 it follows from that @xmath133 on @xmath45 .
hence , if we choose the normalization @xmath134 , then is fulfilled due to our choice of the homology basis : since the cycles possess projective involutions , exactly two pullbacks of @xmath117 have jumps at each point belonging to the projection of a homology cycle , moreover , the jumps are reciprocal as the projected parts of the cycle coincide as sets but have opposite orientations . due to the symmetries of the riemann surface , the vectors @xmath135 and @xmath136 take special forms . to be more specific ,
let us fix a homology basis .
we choose @xmath137 and , in case i , @xmath138 , see , while the @xmath139-cycles are as on figure [ fig : hb](a ) in case i and the @xmath140-cycle should be chosen as on figure [ fig : hb](b ) in cases ii and iii .
[ prop : periods ] let @xmath112 be the homology basis which we just fixed . in case i , one has @xmath141 for some real constants @xmath142 and @xmath143 . in cases
ii and iii , one has @xmath144 recall that @xmath145 is a single - valued function on @xmath69 , see .
the asymptotics of the hermite - pad approximants will depend on the relative sizes of the different branches of @xmath146 ( in other words , on the size of the branches of @xmath81 ) . to this end , define @xmath147 to be open subset of @xmath45 such that @xmath148 we also define the closed set @xmath149 by @xmath150 clearly , @xmath151
. then the following theorem holds .
[ thm : omegas ] in case i , the regions @xmath147 are distributed as on figure [ oms](a ) ; in case ii , the domains are distributed as on figure [ oms](b ) ; in case iii , the domains are distributed as on figure [ oms](c , d ) .
+ + we prove theorems [ thm : rs ] and [ thm : omegas ] as well as proposition [ prop : periods ] in section [ sec : geom ] .
to define the nuttall - szeg functions , we first need to formulate a certain _ jacobi inversion problem_. to this end , denote by @xmath152 the vector of @xmath153 normalized holomorphic integrals on @xmath69 .
that is , @xmath154 is a vector of analytic and additively multi - valued functions on @xmath69 that are single - valued on @xmath121 and satisfy @xmath155_{j , i=1}^g,\ ] ] where @xmath156 is the @xmath157-th vector of the standard basis in @xmath158 .
it is known that @xmath159 is a symmetric matrix with positive definite imaginary part when @xmath102 and is a complex number with @xmath160 when @xmath104 .
the jacobi inversion problem we need consists of finding an integral divisor is a formal expression @xmath161 , where @xmath162 and @xmath163 .
a _ principal divisor _ is an expression @xmath164 such that there exists a rational function on @xmath69 with a zero of multiplicity @xmath16 at @xmath165 for each @xmath166 , a pole of order @xmath167 at @xmath168 for each @xmath169 , and is otherwise non - vanishing and finite ( in particular , one has that @xmath170 ) . ] of order @xmath153 , say @xmath171 , such that @xmath172 where @xmath173 is a vector of constants that depends only on the approximated functions and is defined further below in , @xmath136 and @xmath135 are the vectors defined after , and the equivalence of two vectors @xmath174 is defined by @xmath175 it is known that the jacobi inversion problem has a unique solution when @xmath104 .
hence , @xmath171 is well defined for all @xmath176 in this case . moreover
, the following proposition holds .
[ prop : genus1 ] assuming @xmath104 , let @xmath171 be the unique solution of
. then @xmath177 and @xmath178 for all @xmath179 .
moreover , either @xmath180 or @xmath181 and therefore @xmath182 is infinite .
the unique solvability of a jacobi inversion problem is no longer guaranteed when @xmath102 . however , it is known that solutions are either unique or given by any _ _ special divisor _ _ with simple poles at the elements of the divisor ( a double pole if they coincide ) and otherwise regular . ] .
notice that integral divisors of order 2 can be considered as elements of @xmath183 , where @xmath184 is the symmetric group of two elements .
this is a compact topological space .
hence , we can talk about limit points of sequences of integral divisors .
[ prop : genus2 ] let @xmath102 , then has a unique solution for at least one of the indices @xmath185 for any @xmath186
. moreover , there always exists an infinite subsequence @xmath187 such that no limit point of @xmath188 is a special divisor or is of the form @xmath189 for some @xmath190 .
the following theorem is the main result of this section .
[ thm : n - s ] let @xmath191 and @xmath192 be functions holomorphic and non - vanishing in a neighborhood of @xmath193 $ ] .
in case i , assume in addition that the ratio @xmath194 holomorphically extends to a non - vanishing function in a neighborhood of @xmath108 .
denote by @xmath195 the chain that separates @xmath196 , @xmath197 , and @xmath198 ( @xmath199 is a union of sets defined in ) .
for each @xmath176 for which is uniquely solvable , there exists a function , say @xmath200 , meromorphic in @xmath201 such that the zero / pole multi - set of @xmath200 is given by @xmath202 its traces are bounded except at the branch points of @xmath69 where @xmath200 behaves like @xmath203 for every branch point @xmath204 of @xmath69 except when @xmath205 ( case ii only ) where the exponent @xmath206 should be replaced by @xmath207 .
furthermore @xmath208 where @xmath209 is the interior of the arc @xmath210 . moreover , if @xmath211 is a function meromorphic in @xmath201 satisfying , , and with @xmath171 replaced by any other integral divisor @xmath212 , then @xmath211 is a constant multiple of @xmath200 .
it might seem , especially after looking at , that the construction of the nuttall - szeg functions @xmath200 , particularly , the divisors @xmath171 , depends on the choice of the homology basis .
however , this is precisely the point of the uniqueness part of theorem [ thm : n - s ] that the functions @xmath200 are independent of the intermediate steps of their construction .
the nuttall - szeg functions @xmath200 are certain multiples of the functions @xmath213 .
in fact , their ratios form a normal family on each @xmath214 obtained from @xmath69 by excising circular neighborhoods of radius @xmath215 around the branch points .
[ thm : n - sasymp ] let @xmath216 be a subsequence as in propositions [ prop : genus1 ] or [ prop : genus2 ] .
for each @xmath217 , there exists a constant @xmath218 such that @xmath219 we prove propositions [ prop : genus1 ] and [ prop : genus2 ] as well as theorems [ thm : n - s ] and [ thm : n - sasymp ] in section [ sec : n - s ] .
this section is devoted to the construction of the nuttall - szeg functions of theorems [ thm : n - s ] and [ thm : n - sasymp ] . along
the way we prove propositions [ prop : genus1 ] and [ prop : genus2 ] . for brevity
, we shall use the following notation : @xmath610 a holomorphic differential on @xmath69 is a differential of the form @xmath611 where @xmath612 is a rational function on @xmath69 whose _ principal divisor _ is of the form @xmath613 for some integral divisor @xmath614 of order @xmath615 .
it is known that such rational functions ( integrands of holomorphic differentials ) form a subspace of dimension @xmath153 .
hence , there exists precisely one such integrand ( up to a multiplicative constant ) when @xmath104 , and there exist @xmath616 linearly independent ones when @xmath102 .
it is further known that if @xmath335 and @xmath617 are distinct integrands of holomorphic differentials , then @xmath618 and @xmath619 have no element in common . moreover
, any @xmath620 has a unique complementary point , say @xmath621 , such that @xmath622 for some holomorphic integrand @xmath623 . when @xmath102 , the integral divisors @xmath614 in are exactly the special divisors mentioned before proposition [ prop : genus2 ] . clearly , if @xmath618 and @xmath619 are any two special divisors , then @xmath624 is principal .
a point @xmath71 is a weierstrass point of a genus 2 surface @xmath69 if there exists a rational function on @xmath69 with a double pole at @xmath71 and no other poles .
hence , @xmath71 is a weierstrass point if and only if @xmath625 . to find weierstrass points
it is enough to find a two - sheeted riemann surface conformally equivalent to @xmath69 .
indeed , weierstrass points are mapped into weierstrass points and the weierstrass points of a two sheeted surface of genus 2 are precisely the branch points .
it was shown in ( * ? ? ?
* theorem 1.1 ) that @xmath69 is conformally equivalent to @xmath626 where @xmath627 is the same polynomial as in and the correspondence between the surfaces is given by by @xmath628 in to get the correct correspondence with ( * ? ? ? * theorem 1.1 ) . ]
@xmath629 for the uniformizing parameter @xmath630 .
it follows from and that @xmath631 as @xmath632 , it is not a branch point of and hence @xmath633 is not a weierstrass point of @xmath69 , which is a conclusion we shall need later .
as just explained , @xmath69 has @xmath153 linearly independent holomorphic differentials .
denote by @xmath634 the column vector of such differentials or the unique differential normalized so @xmath635 where @xmath636 is the usual kronecker symbol .
the differentials @xmath637 form a basis for the space of holomorphic differentials on @xmath69 .
the vector of analytic integrals @xmath638 defined before is given by @xmath639 where @xmath640 is the branch point of @xmath69 such that @xmath641 ( we could have chosen any other point to be the initial point for integration ) .
the significance of this vector lies in abel s theorem : if @xmath642 and @xmath643 are integral divisors , the divisor @xmath644 is principal if and only if the orders of @xmath645 coincide and @xmath646 where @xmath647 for @xmath161 . in what follows we shall also use differentials @xmath648 that are the normalized ( that is , @xmath649 , @xmath123 ) abelian differentials of the third kind with simple poles at @xmath71 and @xmath650 of respective residues @xmath651 and @xmath359 , and holomorphic otherwise . to solve the boundary value problem stated in theorem [ thm : n - s ] , we need a ( discontinuous ) cauchy kernel suited for our purposes . define @xmath652 to be the normalized third kind differential with three simple poles , say @xmath653 , assuming @xmath654 , which have respective residues @xmath655 .
it is quite simple to see that for any fixed @xmath650 we have @xmath656 let @xmath116 be a chain on @xmath69 possessing projective involution .
for each @xmath657 which is not a branch point of @xmath69 , we shall denote by @xmath658 another point on @xmath116 having the same canonical projection , i.e. , @xmath659 .
when @xmath657 is a branch point of the surface , we simply set @xmath660 .
henceforth , fix @xmath116 as above and let @xmath661 be a hlder continuous function on @xmath116 . for simplicity , we shall also assume that the set @xmath662 consists of finitely many points .
define @xmath663 the function @xmath664 is holomorphic in the domain of its definition .
moreover , when @xmath665 , one of the points @xmath666 , say @xmath667 , approaches @xmath668 , while @xmath669 does not approach any point on @xmath670 ( projective involution property ) .
hence , @xmath671 \\ & = & \displaystyle -\oint_{{{\boldsymbol\gamma}}}\lambda\,\mathrm{d}\omega_k,\end{aligned}\ ] ] since , according to ( * ? ?
* eq . ( 1.9 ) ) , one has @xmath672 on the other hand , if @xmath673 , @xmath674 , and @xmath675 , then @xmath676 = 0.\ ] ] furthermore , if @xmath677 , @xmath678 , while @xmath669 does not approach any point on @xmath679 , then @xmath680 according to ( * ? ? ?
* eq . ( 2.8 ) ) . finally ,
if @xmath673 , @xmath681 , and @xmath682 , then @xmath683 thus , if we additionally require that @xmath684 , then @xmath664 is a holomorphic function in @xmath685 such that @xmath686 it also can readily be verified that @xmath687 since the cycles @xmath116 and @xmath688 possess projective involutions .
it is known that the continuity of @xmath689 , in fact , hlder continuity , depends on the hlder continuity of @xmath661 only locally .
that is , if @xmath661 is hlder continuous on some open subarc of @xmath116 , so are the traces @xmath689 on this subarc irrespective of the smoothness of @xmath661 on the remaining part of @xmath116 .
of particular importance for us are the logarithmic discontinuities at the branch points of @xmath69 .
more precisely , let @xmath204 be a branch point of @xmath69 and @xmath690 be its circular neighborhood . given a cycle on @xmath69 that possesses a projective involution and passes through @xmath204 , denote by @xmath116 its part that belongs to @xmath690 . in this case
@xmath691 is a jordan arc ending at @xmath692 .
recall ( * ? ?
i.8.6 ) that @xmath693 as @xmath694 , where @xmath695 , @xmath696 , and @xmath697 , @xmath698 , are holomorphic outside of @xmath691 , @xmath699 is any continuous determination of the logarithm on @xmath691 , @xmath700 is the trace of @xmath697 on the positive side of @xmath691 , and the choice of the sign in @xmath701 is determined by the orientation of @xmath691 : @xmath702 if @xmath691 is oriented away from @xmath10 and @xmath703 if it is oriented towards @xmath10 . assuming that @xmath69 has double branching at @xmath204 , define @xmath704 , where the choice of the sign depends on whether @xmath71 lies to the left ( @xmath702 ) or to the right ( @xmath703 ) of @xmath116 .
set @xmath705 the differential @xmath706 is holomorphic in @xmath707 with residues @xmath708 at @xmath71 and @xmath359 at @xmath709 , where @xmath709 is the unique point in @xmath690 which is different from @xmath71 and has the same natural projection .
clearly , the difference @xmath710 is a holomorphic differential in @xmath690 .
therefore , if @xmath711 is of the form @xmath712 satisfying @xmath684 , then @xmath713 as @xmath714 by applied with @xmath715 ( observe that the second equality above is valid independently of the orientation of @xmath691 ) .
assume now that @xmath69 has triple branching at @xmath204 .
since @xmath116 possesses projective involution there is exactly one component of @xmath716 that lies schlicht over @xmath717 .
so that this component lies to the right when @xmath116 is traversed in the positive direction ( it is bordered by @xmath718 ) .
define @xmath719 where @xmath720 is equal to the lift of a fixed determination of @xmath721 to the `` schlicht '' component of @xmath690 and then is continued holomorphically to the whole @xmath690 . since the values of @xmath722 at the points with the same natural projection differ by @xmath723 , one has that @xmath706 is a holomorphic differential in @xmath724 , with 3 simple poles at the elements of @xmath725 having residues @xmath708 at @xmath71 and @xmath359 at the two elements of @xmath726 .
again , @xmath710 is a holomorphic differential in @xmath690 and therefore @xmath727 where the sign @xmath702 is used when @xmath691 is oriented towards @xmath10 and the sign @xmath703 is used if it is oriented away from @xmath10 , and we used the fact that @xmath728 with the same choice of signs . then applied with @xmath729 and @xmath730 gives that @xmath731 where the determination of @xmath721 is precisely the one of @xmath720 within the component of @xmath690 that lies to the right of @xmath116 .
let @xmath195 be the chain introduced in theorem [ thm : n - s ] .
we assume that the cycles forming @xmath195 and separating @xmath196 and @xmath733 , @xmath65 , are oriented so that @xmath196 lies on the left ( positive ) side as each of the cycles is traversed in the positive direction , while the cycle separating @xmath197 and @xmath198 is oriented so that @xmath197 remains on the left when it is traversed in the positive direction . to define a density whose cauchy transform on @xmath195
we shall need , let us introduce auxiliary `` trigonometric '' weights .
namely , @xmath734 in case i with branch cuts along @xmath236 , @xmath735 , respectively ( all are positive for positive reals large enough ) , and @xmath736 in cases ii and iii with branch cuts along @xmath236 and @xmath237 , respectively ( all are positive for positive reals large enough ) .
we set @xmath737 where we choose continuous branches of the logarithms , which is possible as the functions @xmath247 and the trigonometric weights are non - vanishing except at the endpoints .
put @xmath738 then @xmath732 is a holomorphic and non - vanishing function on @xmath739 with continuous traces except at the branch points that satisfy @xmath740 moreover , @xmath732 satisfies with @xmath117 replaced by @xmath732 as follows from , and , excluding the case @xmath741 , one has @xmath742 by , where @xmath743 and @xmath744 in case i while @xmath745 and @xmath746 in cases ii and iii , and the points @xmath319 appear only in case i. when @xmath747 , we need to use instead of .
now there are two cycles that pass through a branch point : one that separates @xmath196 and @xmath197 ( @xmath197 lies to the right of this cycle ) , and one that separates @xmath196 and @xmath198 ( @xmath198 lies to the right of this cycle ) .
combing the contributions from the integrals over both cycles we get that @xmath748 the theta function associated with @xmath749 is an entire transcendental function of @xmath153 complex variables defined by @xmath750 as shown by riemann , the symmetry of @xmath749 and positive definiteness of its imaginary part ensures the convergence of the series for any @xmath751 .
it is known that @xmath752 for some @xmath190 , where the equivalence @xmath753 was defined in and @xmath754 is the vector of riemann constants defined by @xmath755_{kk}-1\right)/2 -\oint_{{{\boldsymbol\alpha}}_{3-k}}\omega_k^-\mathrm{d}\omega_{3-k } , & g=2 , \smallskip \\
\vec k & : = & \left(\mathbf b-1\right)/2 , & g=1 .
\end{array } \right.\ ] ] let @xmath212 be an integral divisor of order @xmath153 . then @xmath756 is a multi - valued holomorphic function on @xmath69 if @xmath212 is not special and is identically zero otherwise . indeed , in the later case @xmath757 , @xmath758 and therefore @xmath759 by , where we used the fact @xmath760 and the periodicity property of theta functions : @xmath761 set @xmath762 , @xmath56 , and define @xmath763 ( recall that @xmath764 is not a special divisor , see and after , and therefore @xmath765 is always well defined ) , where @xmath766 are such that @xmath767 if the numerator is not identically zero , @xmath765 is a multiplicatively multi - valued meromorphic function on @xmath69 with @xmath153 simple poles at @xmath633 , @xmath153 zeros that we shall describe by an integral divisor @xmath768 , and otherwise non - vanishing and finite ( it is also possible that zeros could cancel poles ) .
moreover , it follows from and that @xmath765 is meromorphic and single - valued in @xmath769 and @xmath770 together with the functions @xmath765 we shall need two more auxiliary theta functions .
we set @xmath771 for any fixed @xmath650 such that @xmath772 for all @xmath56 .
clearly , @xmath773 and @xmath774 , @xmath65 , is a meromorphic function in @xmath769 , with a simple zero at @xmath110 , a simple pole at @xmath775 , and otherwise non - vanishing and finite .
moreover , its traces on @xmath776 satisfy @xmath777 it follows from and that the divisor @xmath768 is a solution of the jacobi inversion problem @xmath778 clearly , @xmath779 are precisely the divisors @xmath171 defined in .
as we already mentioned after , any jacobi inversion problem is uniquely solvable on surfaces of genus 1 .
thus , @xmath768 are well defined in this case . since @xmath780 proposition [ prop : periods ] and
the unique solvability property imply that @xmath781 for all @xmath179 .
this finishes the proof of proposition [ prop : genus1 ] .
observe also that @xmath782 the latter equivalence together with the unique solvability property imply that @xmath783 hence , if @xmath784 , then @xmath785 , @xmath65 , which will be important in what follows . as to proposition [ prop : genus2 ] , observe that its first claim follows from and proposition [ prop : periods ] as @xmath638 assumes equivalent values at special divisors by abel s theorem . to prove the second claim , assume to the contrary that all the limit points of @xmath786 are either special or of the form @xmath189 for some @xmath190 .
using compactness , we always can select a subsequence @xmath787 such that @xmath788 for some divisors @xmath789 simultaneously for all @xmath790 . according to our assumption
, one has @xmath791 where @xmath792 when @xmath789 is special .
observe also that continuity of @xmath793 and imply that @xmath794 for @xmath795 . by combining for different @xmath22 and then recalling
, we deduce that @xmath796 @xmath797 . the latter equivalence and abel s theorem
imply that @xmath798 and @xmath799 are special divisors , and therefore necessarily @xmath800 . the above conclusion and
imply that @xmath801 which is impossible since @xmath802 and @xmath803 by proposition [ prop : periods ] .
this contradiction proves the second claim of proposition [ prop : genus2 ] .
finally , observe that the subsequence @xmath187 is such that no limit point of @xmath804 is special or of the form @xmath805 for @xmath65 . indeed , otherwise there would exist a limit point @xmath212 of one those sequences such that @xmath806 where @xmath807 if @xmath212 is special .
choose a limit point @xmath808 of @xmath809 along the same subsequence .
the continuity of @xmath638 and then would imply that @xmath810 which means that either @xmath808 is special or contains @xmath110 .
as both options are impossible by the choice of @xmath187 , such a limit point @xmath212 does not exist . finally , we are ready to prove theorem [ thm : n - s ] .
let @xmath811 .
then it follows from that @xmath812 is a holomorphic function on @xmath121 with continuous traces that satisfy @xmath813 moreover , implies that holds with @xmath117 replaced by @xmath814 .
finally , set @xmath815 where the functions @xmath816 and @xmath817 are defined by or , depending on the considered case .
the nuttall - szeg functions we are looking for are defined by @xmath818 where @xmath819 was defined in and we assume @xmath176 is such that the corresponding jacobi inversion problem is uniquely solvable .
indeed , it can readily be checked , using , , , , , and , that each @xmath820 is holomorphic in @xmath201 and its traces satisfy @xmath821 using the definition of @xmath822 and , we see that the pullbacks of @xmath820 solve the boundary value problem .
further , the functions @xmath820 have the local behavior around the branch points of @xmath69 as stated in by the very definition of @xmath822 and or . finally ,
the zero / pole multi - set of @xmath820 in @xmath201 is described by the divisor @xmath823 as follows from the properties of @xmath213 , @xmath765 , @xmath824 , and @xmath822 . in particular , the functions satisfying conditions of theorem [ thm : n - s ] are given by @xmath825 .
it only remains to show the uniqueness of @xmath820 . to this end , consider @xmath826 , where @xmath211 satisfies , , and with @xmath768 replaced by any other integral divisor , say @xmath212 .
then the analytic continuation property implies that @xmath826 is a rational function on @xmath69 whose divisor is given by @xmath827 .
as @xmath768 is not special , @xmath828 , that is @xmath826 is a constant . in section [ sec : caseii ]
, we shall need the following observation .
assuming we are in case ii , let @xmath829 be the solution of for the weights @xmath191 and @xmath192 and @xmath830 be the solution of for the weights @xmath831 and @xmath832 .
then it can easily be seen from , , and that @xmath833 moreover , it also follows from and that @xmath834 is a rational function on @xmath69 with the divisor equal to the sum of the divisors and @xmath835 .
as the divisor is a divisor of a rational function , we get from abel s theorem that @xmath836 in particular , it follows from the unique solvability of the jacobi inversion problem that @xmath837 that is , @xmath838 if and only if @xmath839 . below we prove theorem [ thm : n - sasymp ] .
in fact , we shall show that a more general statement is true . given @xmath187 and @xmath217 small
, there exists a constant @xmath840 such that @xmath841 for all @xmath842 large , where @xmath843 .
it follows from the continuity of the boundary values of @xmath732 as well as from that @xmath844 is bounded from above and away from zero in @xmath214 ( that is , including the boundary values on @xmath195 and the @xmath139-cycles ) .
moreover , since the image of @xmath121 under @xmath638 is bounded and @xmath845 by the very definition , see , the moduli @xmath846 are uniformly bounded from above and away from zero .
thus , we only need to estimate the function @xmath847 .
it follows from that the zero / pole multi - set of @xmath847 in @xmath848 is equal to @xmath849 and therefore these functions are holomorphic there as well as non - vanishing in @xmath850 for all @xmath217 small and @xmath176 large by the very choice of @xmath187 .
hence , @xmath851 thus , if we show that @xmath852 is a normal family and every limit point satisfies the estimates as in , a standard compactness argument will finish the proof of . the only varying part of this family is given by @xmath853 boundedness of its image follows from the continuity of @xmath854 , boundedness of the image of @xmath154 for @xmath855 , and the fact that @xmath856 .
thus , @xmath852 is a normal family and its elements are indexed by pairs @xmath857 .
the continuity of @xmath854 and imply that any limit point of the functions in has the same form with @xmath858 replaced by a limit point @xmath859^{2g}$ ] .
denote by @xmath212 the integral divisor of order @xmath153 describing the zeros of this limit point .
then @xmath860 the continuity of @xmath638 implies that the right - hand side of the equivalence above is a limit point of the right - hand sides in . by the definition of @xmath187
, it determines @xmath212 uniquely as it is a limit point of @xmath861 .
moreover , all such limit points are uniformly bounded away from containing @xmath862 and therefore from containing an element from @xmath850 for all @xmath217 small enough , which is sufficient to get the required lower estimate .
below , we consider vector functions @xmath220 of the form @xmath221 where @xmath222 $ ] and @xmath223 $ ] , @xmath46 , and @xmath224 are holomorphic and non - vanishing in a neighborhood of @xmath193 $ ] .
additionally , we impose the following condition on the functions @xmath224 : [ cond ] the ratio @xmath194 extends from @xmath225 to a holomorphic and non - vanishing function * in a domain that contains in its interior the closure of all the bounded components of the regions @xmath147 in case i , see figure [ oms](a ) ; * in a domain whose complement is compact and belongs to the right - hand component of @xmath226 in cases ii and iiia , see figures [ oms](b , c ) ; * in the extended complex plane , i.e. , the ratio is a non - zero constant , in case iiib .
condition [ cond ] needs to be regarded in the following context .
as conjectured in @xcite and supported by the results for nikishin systems , given @xmath3 of the form , the appropriate riemann surface @xmath69 must depend on the analytic continuation of @xmath194 from @xmath227 . in this work , on the other hand , we fixed the surface by considering , which necessitates a condition on the continuation of @xmath194 of the above type .
in what follows , we assume that @xmath46 is fixed and the vector @xmath228 is given by , where @xmath191 and @xmath192 satisfy condition [ cond ] ; @xmath200 are the nuttall - szeg functions of theorem [ thm : n - s ] corresponding to @xmath101 and the functions @xmath191 and @xmath192 , constructed on the riemann surface @xmath69 of @xmath57 realized as in theorem [ rs ] ; the constants @xmath229 are defined by @xmath230 @xmath6 are the monic denominators of the hermite - pad approximants @xmath231 to @xmath3 , which are also multiple orthogonal polynomials with respect to the weights @xmath191 and @xmath192 , see ) , and @xmath232 , @xmath65 , are the linearized errors of approximation corresponding to the multi - indices @xmath233 , @xmath234 , see . to describe the asymptotics of @xmath232 , it will be convenient to put @xmath235 where @xmath236 and @xmath237 were introduced in , @xmath238 , @xmath239 , and @xmath240 is present only in case i and is equal to the unbounded component of @xmath241 , see figures [ fig : results ] and [ oms ] .
observe that @xmath242 in case iiib .
define @xmath243 in case i , it follows from that @xmath244 is the analytic continuation of @xmath245 across @xmath108 until @xmath246 , which always exists by the analyticity of @xmath247 s . in cases
ii and iii it still holds by that @xmath244 is an analytic continuation of @xmath245 .
however , this time one needs to continue @xmath248 through @xmath249 into @xmath196 and then through @xmath250 into @xmath198 while @xmath251 needs to be continued through @xmath252 into @xmath196 and then through @xmath253 into @xmath197 .
for any @xmath254 , we further define @xmath255 and @xmath256 finally , let us introduce the following notation . given a sequence of functions @xmath257 ,
a sequence of finite multi - sets @xmath258 , and positive numbers @xmath259 , we write @xmath260 then the following theorem holds .
[ thm : main ] let @xmath187 be a subsequence from either proposition [ prop : genus1 ] or proposition [ prop : genus2 ] ( depending on whether @xmath261 or @xmath262 ) , and @xmath65 . then @xmath263 locally uniformly in @xmath264 and @xmath265 , respectively , where @xmath258 is the multi - set of zeros of @xmath266 in @xmath264 , @xmath267 is the multi - set of zeros of @xmath244 in @xmath265 , and and @xmath268 in case ii . ]
moreover , there exists @xmath270 such that for all @xmath271 one has @xmath272 locally uniformly in @xmath273 and @xmath274 , respectively , where @xmath275 are functions holomorphic in @xmath273 that coincide with @xmath276 on @xmath277 and @xmath278 are their multi - sets of zeros in @xmath273 , and @xmath279 are functions holomorphic in @xmath280 that coincide with @xmath281 on @xmath282 and @xmath283 are their multi - sets of zeros in @xmath284 .
it follows from theorems [ thm : omegas ] and [ thm : n - sasymp ] that in cases i and iii we observe the pushing effect , i.e , @xmath285 , and the presence of divergence regions ( both phenomena are observed in angelesco systems ) . indeed , according to and , one has locally uniformly in @xmath286\cup \gamma_i\big)$ ] that @xmath287 hence , the error of approximation @xmath288 is geometrically small in @xmath289 for all @xmath215 small enough . on the other hand , in case i ,
the error @xmath290 is at least geometrically big on compact subsets of @xmath291 , see figure [ oms](a ) , ( it could be infinite if the elements of @xmath171 belong to @xmath196 and project onto this component ) and the error @xmath292 is at least geometrically big on compact subsets of @xmath293 . in case iii ,
both components of the approximant diverge in @xmath294 and in case ii there are no divergence domains . as to the zeroes of the functions
, it can be deduced from theorem [ thm : omegas ] and that @xmath6 must vanish in @xmath273 and @xmath232 must vanish in @xmath274 , which is precisely the phenomenon of overinterpolation first observed in nikishin systems .
the basis of our approach to asymptotics of hermite - pad approximants lies in their connection with multiple orthogonal polynomials .
it is quite simple to verify that if the functions @xmath31 are of the form , then is fulfilled if and only if @xmath295 moreover , the linearized error functions @xmath232 admit the following integral representation : @xmath296 the analysis of the system then proceeds via its reformulation as a matrix riemann - hilbert problem .
this fundamental fact in the theory of orthogonal polynomials was first revealed by fokas , its , and kitaev @xcite and the extension to multiple orthogonal polynomials was given in @xcite . set @xmath297 and @xmath298 , and assume that the index @xmath176 is such that @xmath299 for some constants @xmath300 . under condition
, the matrix @xmath301 solves the following riemann - hilbert problem ( rhp-@xmath302 ) : * @xmath302 is analytic in @xmath303 $ ] and @xmath304 where @xmath305 is the diagonal matrix and @xmath306 is the identity matrix ; * @xmath302 has continuous traces on @xmath307 that satisfy @xmath308 , where @xmath309 and @xmath310 is the indicator function of @xmath311 ; * @xmath312 as @xmath193\not\ni z\to e$ ] for @xmath313
. vice versa , if is solvable , then the solution is necessarily of the form and holds . to prove theorem [ thm : main ] , we then follow the framework of the non - linear steepest descent method for matrix riemann - hilbert problems , first introduced in the @xmath314 case by deift and zhou @xcite .
the proof of theorem [ thm : main ] is carried out in sections [ sec : casei][sec : caseiii ] .
this section is devoted to proving theorems [ thm : rs ] and [ thm : omegas ] together with proposition [ prop : periods ] . in section [ ssec : real - rs ] we establish that @xmath69 can indeed be realized as in figure [ rs ] . in section [ ssec : param ]
we justify the the choice of the parameter @xmath38 , thus , finishing the proof of theorem [ thm : rs ] , while simultaneously proving proposition [ prop : periods ] . finally , we prove theorem [ thm : omegas ] in section [ ssec : domains ] .
it can be readily computed that the discriminant of is equal to @xmath315 \medskip \\ & = & 108 a(z)\big[(1+a^2 - 3p^2)z^4+(3p^4-a^2)z^2-p^6\big ] .
\end{array}\ ] ] assume first that we are in case i , i.e. , @xmath316 since the polynomial @xmath317 , which is symmetric and of degree 4 , is negative at the origin and has positive leading coefficient , it follows that it has four zeros , which we denote by @xmath318 and @xmath319 , where @xmath320 .
furthermore , @xmath89 since @xmath317 is positive at @xmath38 and negative at @xmath101 .
observe that if a point is a branch point of @xmath57 of order 3 ( all three branches coincide ) and @xmath57 is finite at this point , then necessarily all three branches are equal to zero there .
hence , @xmath321 are branch points of order 2 .
furthermore , the first equation in implies that neither of the points @xmath322 can be a pole of one the branches while the second equation implies that the branches can not have a cubic root singularity there .
hence , the points @xmath322 are branch points of order 2 , two branches are infinite at them and one is finite .
it follows from the above discussion that we can analytically continue the branches @xmath55 so that the inequalities in hold for @xmath323 . at 1 two branches
blow up and , of course , all three add up to zero .
this is possible only if @xmath324 tends to @xmath325 , @xmath326 tends to @xmath327 , and @xmath328 remains bounded .
thus , 1 is a branch point joining @xmath196 and @xmath198 .
it can easily be seen from that all the branches must satisfy @xmath329 therefore , all the branch cuts must be conjugate - symmetric .
thus , the branch cut starting at 1 must end at @xmath98 .
that is , @xmath330 on @xmath331 , which immediately implies @xmath332 on @xmath331 .
moreover , implies that the traces above are purely imaginary .
as @xmath333 are unbounded near 1 and bounded near @xmath98 , one has @xmath334 where @xmath335 is holomorphic , non - vanishing , and real on @xmath336 . as @xmath335
must be positive for @xmath323 and is non - vanishing , it is , in fact , positive for @xmath337 . therefore , @xmath338 , @xmath339 .
moreover , since @xmath328 is real and non - vanishing for @xmath337 and is negative for @xmath66 large enough by , @xmath328 is negative for all @xmath337 .
thus , @xmath340 as in the case of 1 , two branches are infinite at @xmath101 and one is finite .
the inequalities in imply that the unbounded branches are @xmath324 and @xmath328 and therefore @xmath101 is a branch point between @xmath196 and @xmath197 .
the branch cut is @xmath225 since it must be along the real axis .
that is , @xmath341 on @xmath225 and @xmath342 on @xmath225 where the traces are purely imaginary .
as @xmath343 are unbounded near @xmath344 , one has @xmath345 where @xmath346 is holomorphic , non - vanishing , and real on @xmath347 .
the function @xmath346 is positive on @xmath348 according to and therefore is positive on @xmath347 .
hence , @xmath349 for @xmath350 .
since both branches blow up at @xmath351 , one has @xmath352 furthermore , it is quite simple to deduce from that the branches @xmath55 must satisfy @xmath353 for all @xmath56 , where @xmath354 depends on @xmath355 . according to and
, the considered continuations of @xmath55 satisfy @xmath356 for @xmath357 .
therefore , @xmath358 is a branch point for @xmath324 and @xmath326 as well as @xmath359 . in particular , remains valid for @xmath360 .
hence , @xmath361 where @xmath362 is holomorphic , non - vanishing , and real on @xmath363 for some @xmath364 . from , we know that @xmath362 is negative on @xmath365 and therefore it is negative on @xmath363 . thus , @xmath366 for @xmath367 .
this and the blowing up of @xmath324 and @xmath326 at @xmath359 imply that @xmath368 on the other hand , write @xmath369 for @xmath65 . plugging this expansion into and considering the @xmath370 term on the left - hand side
, we get that @xmath371 this means that the inequality @xmath372 holds for all @xmath373 large enough .
that is , @xmath374 where necessarily @xmath375 .
as there are no branch points between @xmath376 and @xmath377 , there should be a branch cut passing between them and this cut should necessarily be between @xmath197 and @xmath198 .
in other words , @xmath319 are branch points of @xmath328 and @xmath326 .
this finishes the proof of the claim that @xmath69 can be realized as in figure [ rs](a ) in case i. assume now that we are in case ii , i.e. , @xmath378 then the discriminant of is equal to @xmath379 . as before , @xmath105 are branch points of order 2 .
furthermore , we know from the third identity in that some of the branches are unbounded near @xmath380
. however , if the branching were of order 2 , the left - hand side of the second relation in would be unbounded near @xmath380 , but it is bounded there .
hence , @xmath380 are branch points of order 3 . plugging a power expansion for @xmath381 , @xmath65 , into
, one can compute that @xmath382 where the @xmath383 s are solutions of @xmath384 .
in particular , @xmath385 for all @xmath386 large . then , by repeating the initial steps of the analysis for case i , we see that extends for all @xmath323 and that @xmath387 therefore , @xmath388 and @xmath389 are the branch cuts for @xmath324 and @xmath326 . around the point
@xmath390 it can be directly verified that is solved by the following puiseux series : @xmath391,\ ] ] where @xmath392 is any solution of @xmath393 , @xmath394 , and all the roots are principal . since @xmath328 is negative and holomorphic for @xmath395 , one finds that @xmath396 locally around @xmath390 . on the other hand , @xmath397 where @xmath398 is non - vanishing and holomorphic for @xmath399 . since @xmath400 for @xmath323
, we get that @xmath400 for @xmath395 and @xmath401 thus , @xmath402 has values in the fourth quadrant and @xmath403 has values in the first quadrant . as the first summand in is dominant around @xmath390 , we can conclude that @xmath404 locally around @xmath390 . from this
it is easy to see that @xmath326 is holomorphic across the interval @xmath405 and this interval is the branch cut for @xmath328 and @xmath324 .
finally , assume that we are in case iii , that is , @xmath406 in this case @xmath317 is a polynomial of degree 2 which has two roots @xmath318 satisfying @xmath97 . exactly as in case i , we see that all the branch points , namely @xmath407 , are of order 2 .
furthermore , the same reasoning as in case i gives that holds on @xmath408 . since @xmath324 and @xmath326 are unbounded at both @xmath409 and @xmath101 , we get that @xmath410 where @xmath335 is holomorphic , non - vanishing , and real on @xmath411 .
since @xmath335 is positive for @xmath323 and hence for @xmath412 , we can conclude that @xmath413 for @xmath414 .
the blowing up of @xmath324 and @xmath326 at @xmath101 implies that @xmath415 it further follows from the third equation in that two branches of @xmath57 are negative and one branch is positive on @xmath416 .
the inequalities in show that the negative branches are @xmath324 and @xmath328 .
this , in turn , implies that @xmath98 is a branch point of @xmath324 and @xmath328 and so is @xmath358 .
hence , holds on @xmath347 .
as all the branches are bounded at @xmath318 , we have that @xmath417 for @xmath418 , where @xmath419 is holomorphic , non - vanishing , and is negative on @xmath416 .
therefore , it is negative on the whole interval @xmath225 and we get that @xmath420 for @xmath421 . as before , the third equation in implies that two branches of @xmath57 are positive and
one branch is negative on @xmath422 .
moreover , it also implies that @xmath423 . as @xmath424 for @xmath67 and it has no other zeros , it must be the negative branch .
thus , @xmath425 now , as the branches that meet at @xmath351 are unbounded , they must be @xmath324 and @xmath326 by .
hence , holds on @xmath426 and we have that @xmath427 where @xmath362 is holomorphic , non - vanishing , and positive on @xmath422 . as before , this means that @xmath362 is positive for all @xmath428 and holds in this case as well .
this finishes the proof of the claim that @xmath69 can be realized as in figure [ rs](c ) in case iii .
we start with cases ii and iii as they are much simpler .
here we show that for the choice of the parameter @xmath38 as in theorem [ thm : rs ] the condition is fulfilled .
that is , the period of the nuttall differential @xmath84 over any given chain on @xmath69 is purely imaginary .
the latter simplifies to showing that the periods of @xmath84 on the cycles of a homology basis are purely imaginary .
in fact , implies that both periods of @xmath84 are equal to @xmath429 . in proving we shall rely on the following observation
: one has that @xmath430 for some @xmath354 , which can be deduced immediately from .
in fact , we see from and that @xmath431 for all @xmath56 .
this implies that @xmath432 the sum of the last two integrals is equal to 1 by cauchy s residue theorem applied to @xmath326 ( recall that @xmath330 on @xmath210 and @xmath433 ) .
this gives the desired conclusion about @xmath434 .
furthermore , from the choice of the @xmath139-cycle , see figure [ fig : hb](b ) , one has for case iii that @xmath435 as @xmath324 is an odd function , the sum of the first and the third integrals is equal to zero .
since @xmath326 is odd as well , the fourth integral is zero too . using the symmetry considerations
once more , we can get that @xmath436 then by applying cauchy s residue theorem to @xmath328 ( recall that @xmath437 ) , we see that @xmath438 in case ii , we have that @xmath439 and the conclusion @xmath440 follows from an analogous symmetry argument .
this finishes the proof for cases ii and iii of theorem [ thm : rs ] . in case
i we again start with .
the symmetry of the surface implies in this case that @xmath324 is an odd function , i.e. , @xmath441 , the functions @xmath381 , @xmath65 , are odd within the bounded domain delimited by @xmath108 and @xmath442 , and @xmath443 for @xmath386 within the unbounded domain delimited by @xmath108 and @xmath444 .
then it follows from and the cauchy residue theorem that @xmath445 where we used the notation from , , and .
the analysis right after , , and implies that each of the last three integrals is positive . from the definition of our homology basis
, it can easily be seen that the first integral is equal to @xmath434 , the second one is equal to @xmath446 , and the third is equal to the first as @xmath324 is an odd function .
that is , @xmath447 and @xmath448 .
the latter clearly implies the first relation in and the fact that @xmath449 .
applying now the cauchy residue theorem to @xmath326 , we get that @xmath450 where @xmath108 is oriented from @xmath451 to @xmath452 .
hence , we can conclude that @xmath453 , if we show that the last integral is negative .
since @xmath454 is a trace of a holomorphic function , we can deform the path of integration to get @xmath455 where we used the symmetry @xmath456 to get the last equality .
notice also that the branches satisfy @xmath457 as follows directly from .
hence , @xmath458 for @xmath459 .
hence , this difference is real there .
moreover , @xmath460 for all @xmath66 large as follows from and . since the difference @xmath461 can be equal to zero only at the branch points of @xmath69 , @xmath460 for all @xmath459 , which shows that the integral in is negative as desired . to prove the second relation in , observe that our choice of the homology basis can be made so that the @xmath139-cycles are contained within the bounded domain delimited by @xmath108 and @xmath444 , see figure [ fig : hb](a )
moreover , they can freely be deformed within the domain of holomorphy of @xmath84 .
thus , it follows from and that @xmath462 where we used the fact @xmath431 with the bounded domain delimited by @xmath108 and @xmath444 , which , in particular , implies that @xmath463 for @xmath418 .
this finishes the proof of .
it only remains to show that there exists a choice of the parameter @xmath38 in so that @xmath464 from is real .
showing that there is a choice of the parameter @xmath465 such that @xmath464 in is real , is equivalent to proving that @xmath466 , where @xmath467 in order to prove we consider the limiting values of @xmath468 as @xmath469 and @xmath470 .
if they have opposite signs , then clearly such a choice of @xmath38 is indeed possible .
we start with the case @xmath469 .
define @xmath471 then it obviously holds that @xmath472 . in section [ ssec
: domains ] further below we shall argue that @xmath473 is a well defined harmonic function on the riemann surface of @xmath474 and that @xmath475 as @xmath476 depends continuously on the parameter @xmath38 , we can conclude that @xmath477 .
let now @xmath478
. in this case equation becomes @xmath479 the branch points with projections @xmath101 and @xmath98 of the curve merge together into a triple branch point with projection @xmath101 ( this can be observed directly from ) . to verify that @xmath480
, we deform the cycle @xmath481 into a cycle @xmath139 which is involution - symmetric and whose projection from @xmath197 , say @xmath482 , is as on figure [ fig.gamma3 ] .
that is , @xmath482 emanates from @xmath101 into the first quadrant along some special arc @xmath483 , @xmath484 $ ] , and then proceeds along the imaginary axis from @xmath485 down to @xmath486 . as in one has @xmath487 and @xmath443 for all @xmath386 large ( and hence for all @xmath488 ) , which leads to the same conclusion that @xmath489 is real for @xmath488 and therefore the real part of the integral of this difference on @xmath490 is equal to zero .
hence , we get that @xmath491 therefore , in order to complete the proof of theorem [ thm : rs ] , it remains to show that there exists an arc @xmath483 , such that @xmath492 to construct this arc we use a parametrization of the algebraic curve .
this parametrization was suggested in @xcite and it has the form @xmath493 where @xmath494 since the expression for @xmath57 in has the cardano form @xmath495 the equivalence of and can be written ( using the vieta relations ) as @xmath496 with some work one can verify that plugging in the expression for @xmath497 from and the expression for @xmath483 from into the above equations produces an identity , which proves that is indeed a parametrization of , ( for details see @xcite ) . observe that @xmath498 , \quad t_{-}:=\left(\frac{a^{2}}{1-a^{2}}\right)^{1/6},\ ] ] and that the parametrization defines a jordan arc with end points @xmath499 as @xmath500 , the first line of is satisfied .
the local analysis around the point @xmath101 shows that @xmath328 is given by with @xmath501 and for @xmath326 one needs to select @xmath502 . then , after a tedious computation , we get @xmath503 hence , the second line of follows from as @xmath504 .
this finishes the proof of theorem [ thm : rs ] granted we can prove that @xmath505 in is a well defined harmonic function , which we do at the end of the upcoming subsection .
we start with two general observations that are consequences of the single - valuedness of @xmath79 on @xmath69 .
first , the regions @xmath147 could be equivalently defined by @xmath506 and this definition does not depend on the initial point of integration chosen in as changing the initial point results in adding the same constant to all @xmath80 simultaneously .
secondly , let @xmath507 be the analytic continuation of @xmath508 from the point at infinity .
then @xmath507 is an algebraic function and @xmath509 is a well defined harmonic function on the riemann surface of @xmath507 .
hence , the set @xmath510 is a subset of a projection of the zero level line of @xmath509 to @xmath45 . as such it can not be dense in an open set .
we start with case i. it will be convenient to consider a slightly different realization of @xmath69 , namely , we shall suppose that @xmath511\cup[\mathrm ic,\mathrm i\infty].\ ] ] our first goal is to show that @xmath510 in this case has the form as shown in figure [ fig.3a ] . - solid lines , @xmath512 - dashed lines , and @xmath513 - dotted dashed line . ]
it follows from our first observation that lines @xmath514 emanating from a projection of a branch point @xmath204 of @xmath69 can be described by @xmath515 for @xmath516 locally around @xmath10 .
that is , @xmath514 is a trajectory of a quadratic differential @xmath517 .
the local behavior of the trajectories is well known , @xcite .
this implies that exactly one line of the set @xmath510 emanates from the points @xmath518 and @xmath344 ( hard edges " ) and exactly three lines ( with angle @xmath519 between them ) emanate from the points @xmath520 ( soft edges " ) . in order to proceed ,
recall the inequalities and as well as the decompositions and .
furthermore , it follows from and the fact that @xmath441 that @xmath324 is purely imaginary on @xmath521 .
analogously , we can conclude that @xmath381 , @xmath65 , are purely imaginary on @xmath522 .
it follows immediately from and , by choosing the initial point of integration to be @xmath101 in , that @xmath523 similarly , by choosing the initial point of integration in to be @xmath98 , we deduce from and that @xmath524 hence , we get from the symmetry of the surface that @xmath525 moreover we have @xmath526 to conclude our qualitative analysis based on figure [ fig.3b-3c ] , we continue the integration beyond the points @xmath409 ( to the right ) and @xmath101 ( to the left ) .
we get that there exists @xmath527 and @xmath528 such that @xmath529.\ ] ] indeed , if we start integrating from @xmath409 in the positive direction , then @xmath530 increases and @xmath531 decreases .
thus , follows from and by continuity .
we summarize the order of the branches of @xmath81 along @xmath532 in figure [ fig.3d ] .
( 450,20)(0,0 ) ( 0,7)(1,0)380 ( 0,-5)@xmath533 ( 370,-5)@xmath325 ( 0,5)@xmath534 ( 15,5)@xmath534 ( 40,5)@xmath534 ( 105,5)@xmath534 ( 170,5)@xmath534 ( 235,5)@xmath534 ( 300,5)@xmath534 ( 14,-5)@xmath535 ( 40,-5)@xmath536 ( 105,-5)@xmath101 ( 170,-5)@xmath98 ( 235,-5)@xmath409 ( 300,-5)@xmath537 ( 45,15)@xmath538 ( 110,15)@xmath539 ( 175,15)@xmath540 ( 240,15)@xmath541 ( 305,15)@xmath542 we note that @xmath543 decreases if its argument moves from @xmath536 to the left while @xmath530 and @xmath544 increase , see figure [ fig.3b-3c ] .
thus , it is possible that there exists @xmath545 such that @xmath546 moreover , if such a point exists , it is unique . in order to prove the last claim , set @xmath547 where the function @xmath346 is positive and
was defined in , @xmath548 where the function @xmath335 is positive and was defined in while and @xmath362 is negative and was defined in , and finally @xmath549 the measures @xmath550 , @xmath551 , and @xmath552 are positive ( for @xmath553 this claim follows from the discussion after where one needs to recall that we deformed @xmath108 to be as in ) .
it further follows from , , , and privalov s lemma ( * ? ? ?
iii.2 ) that @xmath554 since @xmath555 , we deduce that the branches @xmath556 , @xmath65 , have the following global representation @xmath557 where @xmath558 , @xmath56 .
notice that the symmetry of @xmath69 implies that @xmath551 is an even measure .
therefore , in a complex neighborhood of @xmath559 $ ] ( which is the gap between two connected components of @xmath560 ) , the potential @xmath561 has the form of a saddle such that on the imaginary axis @xmath521 it is an even , concave and decreasing function .
hence , @xmath562 , \quad \mbox{and } \quad \frac{\partial^{2}v_{2}(iy)}{\partial y^{2}}<0,\quad y\in{\mathbb{r}}.\ ] ] the same is true for @xmath563 , that is , @xmath564.\ ] ] thus , @xmath543 is a convex function on @xmath559 $ ] . as @xmath565 on @xmath566 $ ]
, @xmath567 is a concave function on @xmath566 $ ] .
thus , inside @xmath566 $ ] , the inequality @xmath568 can be true on the connected set only , i.e. , on the interval @xmath569 $ ] , so and the uniqueness claim are proved and we obtain @xmath570 now , consider the trajectory @xmath571 ( i.e. , the set where @xmath572 ) emanating from the point @xmath486 into @xmath573 .
it can not cross the set @xmath574\cup[-\infty,-1]$ ] as it would contradict the maximum principle for harmonic functions .
hence , the only possibility for @xmath571 to cross @xmath575 $ ] is to cross it at @xmath535 .
analogous considerations lead to the conclusion that the trajectory of @xmath512 emanating from the point @xmath98 to the lower half - plane arrives at @xmath535 and coincides with the continuation of the considered trajectory @xmath571 .
thus , we have that the three subarcs of @xmath571 emanating from the point @xmath486 ( resp .
@xmath576 ) terminate at the points @xmath577 ( resp .
@xmath578 ) and @xmath579 ; three subarcs of @xmath512 emanating from the point @xmath98 ( resp .
@xmath358 ) terminate at the points @xmath409 ( resp .
@xmath359 ) and @xmath579 ; the trajectory @xmath513 joins the points @xmath351 and @xmath580 .
that is , we have shown that the set @xmath510 has indeed the form as in figure [ fig.3a ] .
the structure of @xmath510 and the order of the branches @xmath581 on @xmath44 ( see figure [ fig.3d ] and ) allow us to identify the decomposition of @xmath582 into @xmath583 like in figure [ fig.3aa ] . when @xmath584\cup[\mathrm ic,\mathrm i\infty]$ ] . ]
it remains to deform the cut to the one in figure [ oms](a ) while simultaneously interchanging indices 1 and 2 in the subscripts of @xmath147 bounded by @xmath585 and @xmath586\cup[\mathrm ic,\mathrm i\infty]$ ] .
this finishes the proof of the case i of theorem [ thm : omegas ] .
now we prove cases ii and iii of theorem [ thm : omegas ] .
recall that we put @xmath95 in and therefore the imaginary branch points @xmath587 annihilate at infinity . similar to case i , we can use , , , and , to show that @xmath588 where @xmath589 is a positive measure supported inside of @xmath590 , @xmath65 . then the choice of the additive constant in and the ordering of the branches @xmath81 along @xmath44 give us @xmath591 , \end{array}\right.\\ \\ ( 2v_{2}+v_{1})(t)-\gamma_{2}\left\{\begin{array}{ll } = 0 , & t\in{\mathsf{supp}}(\lambda_{2 } ) \medskip \\ \geqslant 0 , & t\in[-1,-a]\cup[a,1 ] , \end{array}\right . \end{array}\right .
\qquad \left\{\begin{array}{l } \gamma_{1}=2c_{1}+c_{2},\medskip \\ \gamma_{2}=2c_{2}+c_{1}.\\ \end{array}\right.\ ] ] this type of vector - potential equilibrium problem is well - known in the theory of hermite - pad approximants and systems of measures @xmath592 which satisfy such equilibrium conditions are called angelesco systems , @xcite . using the symmetry of the problem ( [ a.10 ] ) with respect to the imaginary axis we can transform it to the following one . set @xmath593 where we used the change of variables @xmath594 and @xmath595 .
analogously , we put @xmath596 thus , ( with the same constants @xmath597 as in ) we have @xmath598 , \medskip \\
\geqslant 0 & x\in[0,a^{2 } ] , \\
\end{array}\right . \\
\\ ( 2\widetilde{v}_{2}+\widetilde{v}_{1})(x)-\gamma_{2 } \left\{\begin{array}{ll } = 0 & x\in{\mathsf{supp}}(\lambda_{2}(\sqrt{x}))\cap[0,1 ] , \medskip \\
\geqslant 0 & x\in[a^{2},1].\\ \end{array}\right .
\\ \end{array}\right.\ ] ] therefore , @xmath599 angelesco systems for two touching intervals @xmath600 $ ] and @xmath601 $ ] ( for any @xmath46 ) are well studied , see @xcite .
in particular , one knows that the regions @xmath147 constructed for @xmath602 decompose @xmath45 as in figure [ fig.3e ] .
thus , if we plot images of the sets @xmath147 from figure [ fig.3e ] ( cases ii and iii ) under the transformation @xmath603 , then ( due to ) we obtain the corresponding sets from figure [ oms ] ( b , c , d ) .
this finishes the proof theorem [ thm : omegas ] .
it remains to show that defines single - valued harmonic function on the riemann surface of @xmath604 , @xmath605 , which satisfies .
this follows from the fact we can obtain @xmath606 from @xmath607 through the transformation @xmath608 and the latter can be computed via and .
the partition of @xmath120 by the domains @xmath147 for @xmath609 is presented in figure [ fig.3e ] [ case ( i ) ] .
hence , the corresponding domains for @xmath81 are distributed as in figure [ fig.3f ] , which finishes the proof of theorem [ thm : rs ] and justifies the proof of theorem [ thm : omegas ] .
set @xmath863 observe that @xmath864 , @xmath65 .
moreover , with @xmath865 as in , @xmath866 that is , conjugating @xmath867 by @xmath868 allows one to modify the @xmath157-th variable of the matrix @xmath867 using the other variable .
let , as before , @xmath869 , @xmath90 , be the projections of the branch points of @xmath69 and @xmath510 be the cut along which the sheets @xmath197 and @xmath198 are glued to each other .
pick arcs @xmath870 and @xmath871 , all oriented from @xmath451 to @xmath452 , as in figure [ fig : global - i ] .
these arcs delimit three bounded domains that we label from left to right as @xmath872 , @xmath873 , and @xmath874 . and the domains @xmath875 , @xmath873 , and @xmath874 . ]
set @xmath876 outside of the closure of @xmath877 and inside put @xmath878 where the @xmath247 are the weight functions in . this way @xmath879 has jumps on @xmath870 , @xmath871 , and @xmath108 . in particular , the jump on @xmath108 is equal to @xmath880 for brevity , let us write @xmath881 put @xmath882\cup\delta_{01+}\cup\delta_{01-}\cup\delta_{02}\cup\delta_0 $ ] .
then @xmath879 solves rhp-@xmath883 : * @xmath879 is analytic in @xmath884 and @xmath885 ; * @xmath883 has continuous traces on @xmath886 that satisfy @xmath887 * @xmath879 satisfies ( c ) ( see section [ mop ] ) with @xmath193 $ ] replaced by @xmath888 .
if is solvable , then so is , and the solutions are connected via . for the next step
we introduce additional arcs @xmath889 and systems of two arcs @xmath890 as on figure [ fig : local - i ] , all oriented from left to right . : specifically , the arcs @xmath889 and systems of arcs @xmath890 .
the black curves constitute the system @xmath891 . ]
denote further by @xmath892 the domains bounded by @xmath225 and the arcs @xmath889 , respectively , and by @xmath893 the open sets bounded by @xmath890 and @xmath894 .
set @xmath895 @xmath65 , where @xmath896 it can readily be checked that @xmath897 on @xmath590 , @xmath65 , see , where @xmath898 put @xmath899 .
then @xmath900 solves rhp-@xmath900 : * @xmath900 is analytic in @xmath901 and @xmath902 ; * @xmath900 has continuous traces on each side of @xmath903 that satisfy @xmath904 where @xmath905 * @xmath906 satisfies ( c ) with @xmath888 replaced by @xmath907 . as before , if is solvable , then so is , and the solutions are connected via .
let @xmath908 , see figure [ fig : local - i ] .
in this section we are seeking the solution of the following riemann - hilbert problem ( rhp-@xmath909 ) : * @xmath909 is analytic in @xmath910 and @xmath911 ; * @xmath909 has continuous traces on each side of @xmath912 that satisfy @xmath913 on @xmath914 , @xmath56 .
we solve only along the subsequence @xmath187 defined in proposition [ prop : genus2 ] .
let @xmath820 be the @xmath176-th nuttall - szeg functions constructed in .
as the functions @xmath820 satisfy with the zero / pole sets described by , it readily follows that a solution of is given by @xmath915 where @xmath842 , @xmath916 is a diagonal matrix of constants chosen to fulfill the normalization condition in ( a ) , @xmath917 , see , and @xmath918 .
observe that @xmath919 is an entire function in @xmath920 since the determinants of the jump matrices in ( b ) are all equal to 1 .
moreover , the normalization at infinity implies that @xmath921 and therefore @xmath919 is a rational function .
it also follows from that near any of the points in @xmath922 the entries of one of the columns of @xmath923 are bounded and the other entries behave like @xmath924 .
hence , @xmath925 and therefore @xmath926 .
denote by @xmath927 , @xmath928 , a disk centered at @xmath10 of small enough radius so that @xmath929 consists of disjoint , except at @xmath10 , analytic arcs .
we are seeking a solution of the following rhp-@xmath930@xmath931 : * @xmath932 is analytic in @xmath933 ; * @xmath932 has continuous traces on each side of @xmath934 that satisfy ( b ) within @xmath927 ; * @xmath932 is either bounded or has the behavior near @xmath10 within @xmath927 described by ( c ) ; * @xmath935 uniformly on @xmath936 , where @xmath937 and @xmath938 are defined by .
we solve each only for @xmath842 .
for these indices the above problem is well - posed as @xmath926 and therefore @xmath939 is an analytic matrix function in @xmath910 .
in fact , the solution does not depend on the actual choice of @xmath187 , however , the term @xmath940 in ( d ) may depend on the choice of this subsequence .
in solving , it will be convenient to use the notation @xmath941 and introduce the following matrix transformations @xmath942 , @xmath943 , and @xmath944 given by @xmath945_{11 } & [ \boldsymbol a]_{12 } \medskip \\ 0 & [ \boldsymbol a]_{21 } & [ \boldsymbol a]_{22 } \end{array } \right ) , \quad \left ( \begin{array}{ccc } ~[\boldsymbol a]_{11 } & 0 & [ \boldsymbol a]_{12 } \medskip \\ 0 & 1 & 0 \medskip \\
~[\boldsymbol a]_{21 } & 0 & [ \boldsymbol a]_{22 } \end{array } \right ) , \quad \text{and } \quad \left ( \begin{array}{ccc } ~[\boldsymbol a]_{11 } & [ \boldsymbol a]_{12 } & 0 \medskip \\ ~[\boldsymbol a]_{21 } & [ \boldsymbol a]_{22 } & 0 \medskip \\ 0 & 0 & 1 \end{array } \right),\ ] ] respectively , where @xmath946_{ik}$ ] is the @xmath947-th entry of @xmath948 .
observe that @xmath949 for all @xmath314 matrices @xmath950 .
[ [ section ] ] in @xcite , a @xmath314 matrix function was constructed out of bessel and hankel functions that solves rhp-@xmath951 : * @xmath952 is holomorphic in @xmath953)$ ] , where @xmath954 , all the rays are oriented towards the origin ; * @xmath952 has continuous traces on @xmath955 that satisfy @xmath956 * @xmath957 as @xmath958 ; * @xmath952 has the following behavior near @xmath325 : @xmath959 uniformly in @xmath953)$ ] . furthermore , @xmath960 solves the same r - h problem only with the reversed orientation of all the rays .
notice also that ( d ) should be replaced in this case by @xmath961 to carry @xmath952 from the @xmath962-plane to @xmath927 , @xmath963 , we need to introduce local conformal maps . to this end ,
set @xmath964 @xmath965 $ ] , where the second equality follows from and .
as mentioned after , @xmath966 has purely imaginary traces on @xmath967 that differ by a sign .
moreover , since @xmath966 vanishes at @xmath10 as a square root , @xmath968 is conformal in @xmath927 .
furthermore , @xmath969 indeed , the first property follows from and while the second is a consequence of the fact that @xmath966 has purely imaginary traces there .
the last property is the requirement we impose on the arcs @xmath890 . choosing the branch of @xmath970 which is positive on @xmath971 , one has on @xmath967 that @xmath972 now , it can readily be verified that the matrix function @xmath973 satisfies ( a , b , c ) for any holomorphic matrix function @xmath974 , where @xmath975 , @xmath976 , and @xmath977 moreover , one has on @xmath978 that @xmath979 thus
, it only remains to choose @xmath974 so that ( d ) is fulfilled .
direct computation using ( b ) and shows that the matrix @xmath980 is holomorphic in @xmath981 . as the entries of @xmath937 behave like @xmath982 as @xmath694 by , the entries of @xmath974 can have at most a square root singularity there .
thus , @xmath974 is holomorphic throughout @xmath927 as desired .
[ [ section-1 ] ] to solve , we again use the matrix @xmath952 . in this case
we have an additional complication coming from the jumps on @xmath983 . to circumvent it , we shall need the following fact about the matrix @xmath952 : @xmath984_{11}(\zeta ) = i_0(2\zeta^{1/2 } ) \quad \text{and } \quad [ \boldsymbol\psi]_{21}(\zeta ) = 2\pi\mathrm{i}\zeta^{1/2}i_0^\prime(2\zeta^{1/2})\ ] ] within @xmath985 , where @xmath986 is the modified bessel function of order @xmath533 .
observe that both functions above are in fact entire in the whole complex plane .
define @xmath987 @xmath988 $ ] . as mentioned after
, @xmath966 has purely imaginary traces on @xmath989 that differ by a sign .
moreover , since @xmath966 vanishes at @xmath10 as a square root , @xmath968 is conformal in @xmath927 .
furthermore , @xmath990 where the first property follows from and , the second is a consequence of the fact that @xmath966 has purely imaginary traces there , and the third is a requirement we impose on the arcs @xmath889 . choosing the branch of @xmath970 which is positive on @xmath991 , we see that holds on @xmath989 .
we further define @xmath992 where @xmath986 is the modified bessel function of order 0 , see .
the above functions are holomorphic in @xmath993 $ ] and @xmath994_{i1}\big(n^2g_a^2/4\big)\rho_2\big(\phi^{(2)}\big)^{-3n/2}\ ] ] on @xmath348 . according to ( *
10.40.5 ) , @xmath995 uniformly on @xmath996 by and since @xmath997 by theorem [ thm : omegas ] and the choice of the radius of @xmath998 .
clearly , @xmath999 in the above equality is geometric .
moreover , a completely analogous estimate holds for @xmath1000 . given these two functions
, we can set @xmath1001_{11}\big(n^2g_a^2/4\big ) & [ \boldsymbol\psi]_{12}\big(n^2g_a^2/4\big ) & f_{n1 } \medskip \\ ~[\boldsymbol\psi]_{21}\big(n^2g_a^2/4\big ) & [ \boldsymbol\psi]_{22}\big(n^2g_a^2/4\big ) & f_{n2 } \medskip \\ 0 & 0 & 1 \end{matrix } \right).\ ] ] then the solution of is given by @xmath1002 , where @xmath1003 to verify ( d ) , observe that @xmath1004 by the asymptotic properties of @xmath1005 .
the matrix @xmath1006 solving can be constructed analogously .
[ [ section-2 ] ] in @xcite , a @xmath314 matrix was constructed out of airy functions that solves rhp-@xmath1007 : * @xmath1008 is holomorphic in @xmath1009 , where the real line is oriented from left to right ; * @xmath1008 has continuous traces on @xmath1010 that satisfy @xmath1011 * @xmath1008 is bounded around the origin ; * @xmath1008 has the following behavior near @xmath325 : @xmath1012 uniformly in @xmath1009 .
again , @xmath1013 solves the same r - h problem only with the reversed orientation of all the rays . as in the case of @xmath952 , @xmath1014 to map @xmath1008 into @xmath927 , @xmath1015 , define @xmath1016 @xmath1017\cup[b,1]\big)$ ] . as before
, @xmath966 has purely imaginary traces on @xmath1018 that differ by a sign .
moreover , since @xmath966 vanishes at @xmath10 as @xmath1019 , @xmath1020 is conformal in @xmath927 .
furthermore , @xmath1021 where the first property follows from and , the second is a consequence of the fact that @xmath966 has purely imaginary traces there , and the third is a requirement we impose on the arcs @xmath890 . choosing the branch of @xmath1022 which is positive on @xmath1023 , one has on @xmath1018 that @xmath1024 as in the previous cases , one can verify that the solution of is given by @xmath1025 where @xmath1026 and @xmath1027 , @xmath1028 is defined by , and @xmath1029 whose holomorphy can be checked as in the previous cases using .
[ [ section-3 ] ] to map @xmath1008 into @xmath927 , @xmath1030 , set @xmath1031 and define @xmath1032 @xmath1033 .
since @xmath510 is the branch cut for @xmath328 and @xmath326 , the traces of @xmath966 on @xmath1034 differ by a sign .
moreover , since @xmath966 vanishes at @xmath10 as @xmath1019 , @xmath1020 is conformal in @xmath927 .
the following are conditions we impose on the arcs @xmath510 , @xmath1035 , and @xmath1036 : @xmath1037 choosing the branch of @xmath1022 which is positive on @xmath1038 , we see that holds on @xmath1034 with @xmath1039 replaced by @xmath1040 .
then the solution of is given by @xmath1041 where @xmath1042 and @xmath1043 , and @xmath1044 consider the following riemann - hilbert problem ( rhp-@xmath1045 ) : * @xmath1046 is analytic in @xmath1047 , where @xmath1048 is the contour shown in figure [ fig : sigma_r ] , and @xmath1049 ; * @xmath1046 has continuous traces on each side of @xmath1050 that satisfy @xmath1051 where @xmath1052 was defined in , while @xmath937 and @xmath938 were introduced after . :
solid lines .
the dashed lines represent the relevant borders of the domains @xmath147 , see figure [ oms](a ) . ]
let us prove that the jumps of @xmath1046 are uniformly close to @xmath1053 for @xmath842 . in section [ ssec : gp - i ]
we have shown that @xmath1054 , while @xmath1055 by and @xmath1056 by .
hence , @xmath1057 and therefore @xmath1058 , where @xmath1059 is the adjoint matrix of @xmath937 . thus , @xmath1060
uniformly away from @xmath922 by .
therefore , it follows from ( d ) that @xmath1061 uniformly on each @xmath978 .
furthermore , we can write @xmath1062 where @xmath1063 is the matrix with all zero entries except for the @xmath1064-st one , which is 1 , and @xmath1065 is always a combination of @xmath191 and @xmath192 ( particular values of @xmath1066 and the value of the entry depend on the arc in question ) .
thus , @xmath1067 uniformly on @xmath1068 for some constant @xmath1069 by theorem [ thm : omegas ] ( it is a simple examination of the five relevant cases ) .
therefore , we get from that @xmath1070 uniformly on @xmath1068 . the relations and together with ( * ?
* corollary 7.108 ) imply that is solvable for all @xmath842 large enough and satisfies @xmath1071 uniformly in @xmath45 , that is , including the boundary values on @xmath1048 .
inverting and , we get from that @xmath1072_{11}(z ) + \frac1{\rho_i^*(z)}\left\ { \begin{array}{rl } \pm[\boldsymbol z]_{1i+1}(z ) , & z\in o_{i\pm } , \medskip \\ 0 , & \text{otherwise } , \end{array } \right.\ ] ] where @xmath1073 ; @xmath1074_{12}(z ) - \frac{\rho_1(z)}{\rho_2(z)}\left\ { \begin{array}{rl } -[\boldsymbol z]_{13}(z ) , & z\in o_0 , \medskip \\ ~[\boldsymbol z]_{13}(z ) , & z\in o_1 , \medskip \\ 0 , & \text{otherwise } , \end{array } \right.\ ] ] for @xmath1075 ; and @xmath1076_{13}(z ) , & \text{otherwise } \end{array } \right\ } + \frac{\rho_2(z)}{\rho_1(z)}\left\ { \begin{array}{rl } ~[\boldsymbol z]_{12}(z ) , & o_1\cup o_2 \medskip \\ 0 , & \text{otherwise } \end{array } \right\}\ ] ] for @xmath1077 .
let @xmath1046 be the solution of .
then @xmath1078 solves for all @xmath842 large enough .
denote the first row of @xmath1046 by @xmath1079 .
it follows from that @xmath1080 is replaced by @xmath1081 when @xmath1082 .
if @xmath1083 , @xmath928 , then @xmath1084_{1k+1}(z ) = c_n\psi_{n,0}^{(k)}(z ) + c_n\sum_{j=0}^2\upsilon_{n , j}(z)\psi_{n , j}^{(k)}(z ) , \quad k\in\{0,1,2\},\ ] ] according to , where the functions are replaced by their traces when necessary .
formula is valid for @xmath1085 when @xmath1086 and for @xmath1087 , @xmath1088 , when @xmath1089 .
indeed , in these cases @xmath1090 , see and , and , and . as the @xmath1091-st column of @xmath1028 is
the same as the @xmath1091-st column of @xmath1092 , the @xmath1091-st column of @xmath932 is the same as the @xmath1091-st column of @xmath1093 from which the claim follows . to estimate the sum on the right - hand side of , recall that @xmath1094 see , where @xmath1095 is a rational function on @xmath69 with the divisor @xmath1096 .
then , the maximum modulus principle , and the same normal family argument as in and imply that @xmath1097 uniformly in @xmath45 , where the function on the left - hand side of needs to be replaced by its traces when @xmath1098 . by combining with we get that @xmath1099 where @xmath1100 is uniform in @xmath45 . the first relation in follows immediately from the first line of , , and .
moreover , implies that the second line of can be rewritten as @xmath1101 for @xmath1102 and @xmath1103 , @xmath1104 , where @xmath1105 clearly , each @xmath1106 extends to a holomorphic function in @xmath1107 by ( recall that @xmath1108 and @xmath1109 also satisfy these relations ) .
the first part of now follows from .
furthermore , , , and imply the second line of outside of @xmath1110 for @xmath1111 . in the spirit of ,
define @xmath1112 which is holomorphic in @xmath1113 by , and @xmath1114 which is holomorphic in @xmath1115 , again , by . then and imply that @xmath1116 for @xmath1117 and @xmath1083 , @xmath1118 .
clearly , formula is an immediate consequence of .
to finish the proof of for @xmath1111 , let us show that @xmath1119 locally uniformly in @xmath1120 .
indeed , we have that @xmath1121 see .
the claim now follows from theorem [ thm : omegas ] and the normal family argument along the lines of and .
this proves outside of @xmath1122 .
analogously we can argue that @xmath1123 locally uniformly in @xmath1124 , which finishes the proof of for @xmath1125 .
the proof of and for @xmath37 can be completed analogously starting with .
let @xmath1126 and @xmath1127 be defined by .
further , let @xmath1128 be the left half - plane .
we orient the boundary of @xmath873 , say @xmath1129 ( the imaginary axis ) , so that @xmath873 lies to the left of @xmath1129 when the latter is traversed in the positive direction .
denote by @xmath871 a simple jordan curve lying within the right component of @xmath226 containing all the singularities of @xmath194 , see figure [ fig : global - ii ] without the circle @xmath1130 , the domains @xmath873 ( shaded regions on the left ) and @xmath874 ( unshaded region ) , local lenses @xmath889 and @xmath890 . ]
( this is always possible because of condition [ cond ] ) .
we orient @xmath871 counter - clockwise and set @xmath874 to be the intersection of the exterior domain of @xmath871 and the right half - plane @xmath1131 . put @xmath1132 where @xmath1133 .
put @xmath882\cup\delta_{01}\cup\delta_{02}$ ] .
then @xmath879 solves rhp-@xmath883 : * @xmath879 is analytic in @xmath884 and @xmath1134 ; * @xmath883 has continuous traces on @xmath1135 that satisfy @xmath1136 where @xmath1137 is defined by and @xmath590 in ; * @xmath879 satisfies ( c ) ( see section [ mop ] ) with @xmath193 $ ] replaced by @xmath888 . if is solvable , then so is , and the solutions are connected via
as in case i , we introduce additional arcs @xmath889 and systems of two arcs @xmath890 as in figure [ fig : global - ii ] , all oriented from left to right .
we further denote by @xmath1138 the open sets bounded by @xmath590 and the arcs @xmath1139 , @xmath65 .
set @xmath1140 where the matrices @xmath1141 are defined by . put @xmath899 .
then it can readily be checked that @xmath900 solves rhp-@xmath900 : * @xmath900 is analytic in @xmath901 and @xmath1142 ; * @xmath900 has continuous traces on each side of @xmath1143 that satisfy @xmath1144 where @xmath1145 are defined by and @xmath1146 * @xmath906 satisfies ( c ) with @xmath888 replaced by @xmath907 . as before , if is solvable , then so is , and the solutions are connected via .
let @xmath1147 be given by .
then it is a solution of the following riemann - hilbert problem ( rhp-@xmath909 ) : * @xmath909 is analytic in @xmath303 $ ] and @xmath911 ; * @xmath909 has continuous traces on each side of @xmath1148 , @xmath65 , that satisfy @xmath1149 .
we can not argue that @xmath1150 as in section [ ssec : gp - i ] since every entry of @xmath937 behaves like @xmath1151 as @xmath1152 .
however , we can construct @xmath1153 explicitly .
denote by @xmath1154 the matrix that satisfies as above with @xmath191 and @xmath1137 replaced by @xmath831 and @xmath1155 .
it follows from that the construction of the matrices @xmath923 and @xmath1156 is simultaneously applicable or not applicable for each index @xmath176 .
observe that @xmath1157 on @xmath1158 and @xmath1159 by and .
moreover , @xmath1160 on @xmath1148 since @xmath1161 .
it also follows from that the entries of @xmath937 and @xmath1162 have at most @xmath1163 and @xmath1164-root singularities at @xmath105 and @xmath1165 , respectively ( the functions @xmath1108 and @xmath1109 possess exactly the same behavior around those points as @xmath1166 ) .
hence , the product @xmath1167 is holomorphic in the entire complex plane .
since @xmath1168 , we deduce that @xmath1169 in particular , it follows from that @xmath1170 for some constant matrices @xmath1171 and @xmath1172 .
then it follows immediately from that @xmath1173 again , we need to solve locally , this time around @xmath105 , @xmath1165 , and @xmath325 .
the local problems are exactly the same as in case i and therefore their solutions are given by . [ [ section-4 ] ] let @xmath1174 be large enough so that @xmath1175 . in this section , we are looking for a solution to the following riemann - hilbert problem ( rhp-@xmath930@xmath1176 ) : * @xmath1177 is holomorphic in @xmath1178 and @xmath1179 * @xmath1177 has continuous traces on each side of @xmath1180 that satisfy @xmath1181 ; * @xmath1182 uniformly on @xmath1130 .
let us show that is solved by @xmath1183 where the constants @xmath1184 are defined in , the function @xmath722 is given by @xmath1185 and @xmath962 is defined by @xmath1186 indeed , @xmath1187 is a conformal function in @xmath1188 that vanishes at infinity by ( make @xmath1189 larger if necessary ) . here , we choose the branch of the square root so that @xmath1190 tends to @xmath1191 when @xmath61 .
hence , we can deform @xmath1129 in @xmath1188 so that @xmath1192 .
thus , the right - hand side of is holomorphic in @xmath1178 . as it follows from (
* eq . ( 7.2.2 ) ) that @xmath1193 ( a ) is indeed satisfied . to verify ( b ) , notice that @xmath1194 and that @xmath1195 for @xmath1196 . since @xmath1197 ( b ) follows .
finally , ( * ? ? ?
( 7.12.1 ) ) implies that @xmath1198 uniformly in the left and right half - planes .
hence , @xmath1199 uniformly on @xmath1200 . as @xmath1201 by theorem [ thm : omegas ] , we see that ( c ) holds as well .
[ [ section-5 ] ] denote by @xmath1202 . in this section
, we construct a solution to the following riemann - hilbert problem : * @xmath1203 is analytic in @xmath1204 ; * @xmath1203 has continuous traces on each side of @xmath1205 that satisfy ( b ) ; * @xmath1203 has the behavior near @xmath1206 within @xmath1207 described by ( c ) ; * @xmath1208 uniformly on @xmath1209 . in @xcite ( an alternative approach to asymptotics of multiple orthogonal polynomials around cubic branch point was developed in @xcite ) , a @xmath1210 matrix function was constructed out of solutions to @xmath1211 that solves rhp-@xmath1212 : * @xmath1213 is holomorphic in @xmath1214 , where @xmath1215 and the positive direction on all the lines goes from the left half - plane to the right half - plane ; * @xmath1213 has continuous traces on @xmath1216 that satisfy @xmath1217 and @xmath1218 * @xmath1219 as @xmath958 ; * @xmath1213 has the following behavior near @xmath325 : @xmath1220 uniformly for @xmath1221 and @xmath464 on bounded sets with @xmath1222 where @xmath1223 @xmath1224 and @xmath1225 below , we explain how to solve using @xmath1213 , a solution of can be constructed analogously . to carry into a neighborhood of @xmath390 , define @xmath1226 where the function @xmath507 was introduced in and we choose the principle branch of the square root .
then @xmath962 is conformal in some neighborhood of @xmath390 , @xmath1227 , and @xmath1228 , @xmath395 .
further , define @xmath1229 it readily follows that @xmath1230 is also conformal in some neighborhood of @xmath390 , @xmath1231 , and that @xmath1232 by and , and since @xmath117 has value 1 at the point of @xmath69 whose natural projection is @xmath390 .
set @xmath1233 it can easily be verified that @xmath1234 as the entries of @xmath937 as well as the entries of @xmath1235 can have at most cubic root singularity at @xmath390 , the matrix @xmath1236 is holomorphic in @xmath1237 . then @xmath1238 satisfies ( a c ) ( we always can adjust @xmath1139 so that @xmath962 maps them into @xmath1239 )
. it also follows from ( d ) and that @xmath1240 as @xmath1241 uniformly on @xmath1242 ( recall that @xmath1230 is conformal in @xmath1237 and vanishes at @xmath390 , which implies that @xmath1243 remains bounded as @xmath1241 and therefore ( d ) is applicable ) , where @xmath1244 since @xmath1245 as @xmath1246 or @xmath1241 , we can write @xmath1247 by and , where @xmath1248 is a matrix of constants ( more precisely , it is one matrix of constants in @xmath1249 and another one in @xmath1250 , see the definition of @xmath1251 ) with entries whose moduli are uniformly bounded as @xmath1241 . as the fraction in the above equality is holomorphic around @xmath390 and has value @xmath1252 at @xmath390 , we get that @xmath1253 since @xmath1254 on @xmath1242 , we have that @xmath1255 uniformly on @xmath1242 . thus , according to , is solved by @xmath1256 the final riemann - hilbert problem is from section [ ssec : fr - h ] with @xmath1052 defined in . without the circle @xmath1130 ( solid lines ) and the borders of the domains @xmath147 ( dashed lines ) , see figure [ oms](b ) . ]
the same analysis shows that the jump matrices in are of order @xmath1257 for @xmath1258 and therefore is solvable for all @xmath842 large enough and satisfies with @xmath1259 replaced by @xmath1260 .
it can readily be verified that formulae , , and remain valid .
let @xmath1046 be the solution of .
then the solution of is given by , where @xmath1261 , for all @xmath842 large enough .
the same argument as before shows that holds in this case as well .
since the first column of @xmath1177 is the same as the first column of @xmath1093 , remains valid when @xmath1086 .
thus , the proof of the first parts of and is exactly the same .
the second parts of and are claimed to hold uniformly on each compact subset of the respective domains ( because @xmath246 contains the point at infinity and @xmath274 is bounded ) .
hence , given a compact subset , we always can enlarge @xmath1189 in so that this set is contained in @xmath1262 .
this way is valid on this compact and the remaining part of the proof is the same as in case i.
let @xmath1126 and @xmath1127 be as in . fix a domain , say @xmath873 , that contains @xmath1263 $ ] and whose boundary @xmath1264 is smooth , lies entirely in @xmath1265 ( except for the point where it crosses @xmath347 ) , see figure [ oms](c ) , while crossing the real line at the origin , see figure [ fig : global - iii ] . in case iiia ,
denote by @xmath871 a smooth jordan curve lying within the right component of @xmath226 and containing within its interior all the singularities of @xmath194 ( recall that @xmath1266 in case iiib by condition [ cond ] ) , see figure [ fig : global - iii ] .
we also denote by @xmath874 the intersection of the exterior domains of @xmath1129 and @xmath871 .
* @xmath879 is analytic in @xmath884 and @xmath885 ; * @xmath883 has continuous traces on @xmath1268 that satisfy @xmath1269 where @xmath1137 is defined by and @xmath590 in ; * @xmath879 satisfies ( c ) ( see section [ mop ] ) with @xmath193 $ ] replaced by @xmath888 . as usual
, we introduce additional arcs @xmath889 and systems of two arcs @xmath890 as in figure [ fig : global - iii ] , all oriented from left to right .
we further denote by @xmath1138 the open sets bounded by @xmath590 and the arcs @xmath1139 , @xmath65 .
set @xmath1270 where the matrices @xmath1141 are defined by . put @xmath899 .
then it can readily be checked that @xmath900 solves rhp-@xmath900 : * @xmath900 is analytic in @xmath901 and @xmath902 ; * @xmath900 has continuous traces on each side of @xmath1271 that satisfy @xmath1272 where @xmath1145 are defined by and @xmath1273 * @xmath906 satisfies ( c ) with @xmath888 replaced by @xmath907 .
aptekarev , a.b.j .
kuijlaars , and w. van assche .
asymptotics of hermite - pad rational approximants for two analytic functions with separated pairs of branch points ( case of genus 0 ) .
, 2008 , rpm007 , 128 pp .
aptekarev and v.g .
systems of markov functions generated by graphs and the asymptotics of their hermite - pad approximants . , 201(2):2978 , 2010 .
english transl . in _ sb .
_ , 201:183234 , 2010 .
k. deschout and a.b.j .
double scaling limit for modified jacobi - angelesco polynomials . in p.
brndn , m. passare , and m. putinar , editors , _ notions of positivity and the geometry of polynomials _ , trends in mathematics , pages 115161 , basel , 2011 .
springer .
h. stahl .
asymptotics of hermite - pad polynomials and related convergence results a summary of results . in _ nonlinear numerical methods and rational approximation _ , volume 43 of _ math .
_ , pages 2353 .
reidel , dordrecht , 1987 .
w. van assche , j.s .
geronimo , and a.b.j .
riemann - hilbert problems for multiple orthogonal polynomials . in _ special functions 2000 : current perspective and future directions _ , number 30 in nato sci . ser .
, pages 2359 , dordrecht , 2001 .
kluwer acad . publ . | hermite - pad approximants of type ii are vectors of rational functions with common denominator that interpolate a given vector of power series at infinity with maximal order .
we are interested in the situation when the approximated vector is given by a pair of cauchy transforms of smooth complex measures supported on the real line .
the convergence properties of the approximants are rather well understood when the supports consist of two disjoint intervals ( angelesco systems ) or two intervals that coincide under the condition that the ratio of the measures is a restriction of the cauchy transform of a third measure ( nikishin systems ) . in this work
we consider the case where the supports form two overlapping intervals ( in a symmetric way ) and the ratio of the measures extends to a holomorphic function in a region that depends on the size of the overlap .
we derive szeg - type formulae for the asymptotics of the approximants , identify the convergence and divergence domains ( the divergence domains appear for angelesco systems but are not present for nikishin systems ) , and show the presence of overinterpolation ( a feature peculiar for nikishin systems but not for angelesco systems ) . our analysis is based on a riemann - hilbert problem for multiple orthogonal polynomials ( the common denominator ) . |
simulations suggest that galaxy mergers can produce remnants with two or more massive black holes @xcite . when these black holes accrete , they can appear as two or more active galactic nuclei ( agns ) on kiloparsec scales @xcite .
systematic surveys for such multiple agns could provide observational constraints on agn activation and tidally enhanced star formation ( e.g. , * ? ? ?
* ; * ? ? ?
* ) , and on the black hole merger rate , with consequences for the signals expected for pulsar timing arrays and gravity - wave detectors @xcite . guided by early serendipitous discoveries of dual agns like 3c75 @xcite and ngc6240 @xcite
, recent systematic surveys are now yielding spectroscopic samples of dual agn candidates ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
candidate dual agns can be difficult to confirm due to obscuration and resolution issues @xcite .
moreover , some candidates may turn out to be imposters because they harbor a recoiling agn , a situation that both complicates and enriches matters ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
this paper focuses on one such case , cosmosj100043.15 + 020637.2 ( j1000 + 0206 hereafter ) .
early reports on j1000 + 0206 @xcite were prompted by its unusual optical nature , involving two apparent nuclei in a merger remnant with a tidal tail @xcite .
the integrated optical emission - line redshift of j1000 + 0206 is @xmath3 @xcite . adopting the nomenclature and cosmology = 70 km s@xmath4 mpc@xmath4 , @xmath5 = 0.3 , @xmath6 = 0.7 ] of @xcite , the two nuclei , se and nw , have a projected separation of 500 mas ( 2.5 kpc ) .
two scenarios have emerged for these optical nuclei : either nw and se each contain an agn with its own narrow emission - line region , or se is an agn recoiling from nw due to the asymmetric emission of gravitational waves during black - hole coalescence .
hydrodynamic simulations , with radiative transfer calculations , of a gas - rich major merger show that each scenario is consistent with available data @xcite .
@xcite assumed that the emission from j1000 + 0206 at a frequency @xmath1 = 1.4 ghz @xcite is driven by its spectroscopically - identified agn .
but this is far from certain .
figure 1 shows that the radio emission is localized to the inner portions of the merger remnant @xcite , where simulations indicate that star formation can occur @xcite .
also , @xcite recently discovered an extended x - ray feature offset by about 500 mas ( 2.5 kpc ) to the southwest ( sw ) of the se optical nucleus , further complicating the observational picture .
if the radio emission from j1000 + 0206 arises from the agns and star formation , figure 1 makes it clear that disentangling these contributions requires imaging at subarcsecond resolution .
[ imaging ] presents new images from the karl g. jansky very large array ( vla ; * ? ? ?
* ) at @xmath1 = 9.0 ghz and a fwhm resolution @xmath2 = 320 mas ( 1.6 kpc ) , and from the very long baseline array ( vlba ; * ? ? ?
* ) at @xmath1 = 1.5 ghz and @xmath2 = 15 mas ( 75 pc ) .
[ implications ] explores the implications of the new imaging , while a summary and conclusions appear in [ summary ] .
j1000 + 0206 was observed with the vla in its a configuration on 2012 october 12 ut under proposal code 12b-072 .
a coordinate equinox of 2000 was used .
j1024 - 0052 , with a one - dimensional position error of 10 mas ( 1 @xmath7 ) , was employed as a phase calibrator .
the switching time between it and j1000 + 0206 was 480 s , and involved a switching angle of 6.6 .
the _ a priori _ pointing position for j1000 + 0206 was shifted 0.6 south of the sdss position @xcite to avoid the risk of phase - center artifacts .
every 1 s the correlator produced 1024 contiguous 2-mhz channels , yielding a total bandwidth of 2.048 ghz per circular polarization centered at @xmath1 = 9.0 ghz .
observations of 3c138 were used to set the amplitude scale to an accuracy of about 1% @xcite .
the net exposure time on j1000 + 0206 was about 2150 s. release 4.1.0 of the common astronomy software applications ( casa ) package @xcite was used to calibrate and edit the data in an automated fashion . after further minor edits , the casa task imagr was used to form and deconvolve a naturally - weighted image of the stokes @xmath8 emission from j1000 + 0206 .
this imaged spans 2561 @xmath9 60 mas in each coordinate , has an rms noise level @xmath10 mjy beam@xmath4 and a geometric resolution at fwhm @xmath2 = 320 mas .
its inner portions are shown in figure 2 . with adequate signal - to - noise
, structures as large as about 10 @xmath2 @xmath11 3.2 could be represented in figure 2 .
no emission was detected above a threshold of @xmath12 mjy beam@xmath4 ( 3 @xmath7 ) .
j1000 + 0206 was observed with the vlba during four 6-hour sessions spanning 2012 june 24 to 2013 january 11 under proposal code bm360 .
a coordinate equinox of 2000 was used .
j1011 + 0106 , with a one - dimensional position error of 2 mas ( 1 @xmath7 ) , was used as a phase calibrator .
the switching time between it and j1000 + 0206 was 360 s , and involved a switching angle of 2.7 .
every 4.1 s the correlator @xcite produced 128 contiguous 2-mhz channels , yielding a total bandwidth of 0.256 ghz per circular polarization centered at @xmath1 = 1.52 ghz .
vlba system temperatures and gains were used to set the amplitude scale to an accuracy of about 5% .
a total of about 450 baseline hours were accrued on j1000 + 0206 . for each 6-hour session , nrao s astronomical image processing system ( aips ;
* ) was used to calibrate and edit the data following the approach described by @xcite .
the aips task imagr was used to form and deconvolve a naturally - weighted image of the stokes @xmath8 emission from j1000 + 0206 .
this image spans 2048 @xmath9 1 mas in each coordinate , has a geometric resolution at fwhm @xmath2 = 15 mas , and was corrected for primary beam attenuation . combining the u - v data from the four sessions resulted in an rms noise level @xmath13 mjy beam@xmath4 . with adequate signal - to - noise , structures as large as about 10 @xmath2 @xmath11 150 mas could be imaged .
no emission was detected above a threshold of @xmath14 mjy beam@xmath4 ( 6 @xmath7 ) .
the new vlba image at @xmath1 = 1.52 ghz and @xmath2 = 15 mas ( 75 pc ) filters for emission with a rest - frame brightness temperature @xmath15 k , not achieved by even the most compact starbursts @xcite . moreover ,
the vlba image is too shallow to detect even the most luminous radio supernova beyond @xmath16 ( e.g. , * ? ? ?
* ; * ? ? ?
thus the vlba image is insensitive to emission driven by star formation .
in contrast , the new vla image ( figure 2 ) at @xmath1 = 9.0 ghz and @xmath2 = 320 mas ( 1.6 kpc ) is sensitive to emission driven by star formation , and to agn - driven emission in the se or nw optical nuclei or associated with the x - ray feature offset to the sw of the se optical nucleus @xcite . applying a k - correction for a spectral index @xmath17 ( @xmath18 ) , the vla nondetections in figure 2 imply that on scales of 1 @xmath2 = 0.32 ( 1.6 kpc ) to 10 @xmath2 = 3.2 ( 16 kpc ) the merger remnant has a spectral luminosity @xmath19 w hz@xmath4 . given these sub - galactic scales , we adopt local indicators of star formation rates ( sfrs ) ( e.g. , * ? ? ? * ) .
we apply equation ( 15 ) of @xcite at a rest - frame frequency @xmath20 ghz = 12.2 ghz , include a thermal contribution for an electron temperature @xmath21 k and a nonthermal contribution with a spectral index @xmath22 , and find a sfr @xmath23 17 m@xmath24 yr@xmath4 for a kroupa initial mass function ( imf ) .
this is an interesting regime to be exploring , given that the host galaxies of optically - selected agns at @xmath25 could have sfr @xmath11 1 - 20 m@xmath24 yr@xmath4 , spanning values for normal spiral galaxies to those for moderately luminous starbursts @xcite .
prior vla imaging at @xmath1 = 1.4 ghz @xcite localized the emission to the circular and elliptical regions , outlined in figures 1 and 2 , that encompass the inner merger remnant , the optical nuclei se and nw , and the x - ray feature sw of the se nucleus .
the circular region marks a point - like source with an integrated flux density @xmath26 mjy and a k - corrected spectral luminosity @xmath27 w hz@xmath4 . if driven by star formation only , the corresponding sfr = 34 m@xmath24 yr@xmath4 ( * ? ? ?
this is at least a factor of two higher than the new vla limit derived above , meaning that up to half of the integrated flux density could arise from star formation .
conversely , any agn - driven emission could have an integrated @xmath28 mjy and a k - corrected luminosity @xmath29 erg s@xmath4 , the regime of low - luminosity agn ( llagn ) at @xmath30 @xcite .
analyses of the _ hst _ f814w acs image and the keck / deimos spectrum of the galaxy were reported in @xcite , and can be summarized as follows .
using source extractor on the _ hst _ image , the se and nw optical nuclei were found to be aligned along a pa = 139.6 , with a projected separation of @xmath31 mas ( @xmath32 kpc ) .
the deimos spectrum was taken with a slit pa = 143.3 and spanning 47309840 in wavelength .
this spectrum revealed double narrow emission - line components , with line flux ratios of both components falling within the agn region of the baldwin - phillips - terlevich diagram , within the 1 @xmath7 errors on the flux measurements .
projected spatial separations between the double emission - line components of @xmath33 kpc , @xmath34 kpc , and @xmath35 kpc were measured for , , and , respectively .
the mean of the three spatial separations measurements , weighted by their inverse variances , is @xmath36 kpc .
however , the spatial separation measurement is not consistent with the separations measured in and and is skewed low because of partial obscuration by an imperfectly subtracted night - sky line .
using only the spatial separations measured in and ( which are consistent with one another , to within the 1 @xmath7 errors ) , the inverse - variance - weighted mean spatial separation is @xmath37 kpc . in projection ,
the spatial separation between the double emission - line components ( @xmath37 kpc ) is consistent with the spatial separation between the double nuclei in the _ hst _ image ( @xmath38 kpc ) , to within their errors .
there is a 3.7 difference between the pa of the deimos slit and the pa of the two nuclei in the _ hst _ image , but this results in a negligible ( 0.2% ) difference between the spectrum - based and image - based measurements of spatial separations .
these similar spatial separations , combined with the excitation state of the narrow emission - line components , are taken as strong indicators for the presence of two agns , likely associated with se and nw .
the exposure time of the deimos spectrum was too short to allow us to localize , with confidence , each narrow emission - line component with each optical nucleus .
for now we make the plausible assumption that luminosities @xmath39 } ) = 4.8 - 9.9 \times 10^{42}$ ] erg s@xmath4 and excitation ratios [ oiii]@xmath405007/h@xmath41 @xmath42 for the two agns @xcite . in the near future
we will obtain keck / osiris data to test this important assumption ( f. muller - sanchez , priv .
comm . ) .
@xcite provide strong evidence that the se optical nucleus is an unobscured agn of type 1 , thus featuring a broad emission - line region .
se has an x - ray counterpart with a luminosity @xmath43 erg s@xmath4 @xcite .
its eddington ratio is 0.04 @xcite , typical for a type 1 agn in cosmos @xcite .
the new vla and vlba nondetections of se imply that any agn - driven emission has k - corrected luminosities @xmath44 erg s@xmath4 ( 3 @xmath7 ) on scales below 1.6 kpc and @xmath45 ergs s@xmath4 ( 6 @xmath7 ) on scales below 75 pc . from [ overview ] , the luminosity of any agn - driven emission from se can not exceed @xmath46 erg s@xmath4 . following @xcite , the new vla photometry for se yields log @xmath47 .
this value is consistent with se s modest eddington ratio @xcite . from
[ overview ] , the luminosity of se is @xmath39 } ) = 4.8 - 9.9 \times 10^{42}$ ] erg s@xmath4 @xcite .
the vlba luminosity , free from contamination from star formation , thus implies that se is radio quiet as defined by @xcite or @xcite , a trait that is also consistent with its modest eddington ratio . for context ,
several key properties of se resemble or are consistent with those of the prototypical type 1 agn ngc4151 . adopting a distance of 13.3 mpc for ngc4151 , it has ( a ) @xmath48 erg s@xmath4 @xcite ; ( b ) when integrated within a radius of 2 ( 130 pc ) , @xmath49 erg s@xmath4 , @xmath50 erg s@xmath4 and @xmath51 @xcite ; and ( c ) log @xmath52 . @xcite and @xcite report evidence that the radio source dynamically interacts with the interstellar medium ( ism ) at these radii in ngc4151 .
@xcite noted that the x - ray feature offset by about 500 mas ( 2.5 kpc ) to the sw of the se optical nucleus could be driven by star formation .
indeed , @xcite suggest that the impulse from the passage of a recoiling black hole could could trigger star formation in its wake . in the scenario where se has recoiled from nw ,
the sw x - ray feature might be its star formation wake . that feature is about 18% as luminous as se in the passband of the _ chandra _ high resolution camera ( hrc ) , indicating that it has @xmath53 erg s@xmath4 @xcite . from [ overview ] , the new vla upper limit to the sfr at the location of the sw x - ray feature is sfr @xmath23 17 m@xmath24 yr@xmath4 for a kroupa imf .
this corresponds to @xmath54 erg s@xmath4 according to @xcite , who use a salpeter imf .
even with this caveat of different imfs , the x - ray emission from the sw feature is much too luminous to be driven by star formation .
this points to the sw x - ray feature being agn - driven , perhaps , as @xcite suggest , because se photoionizes the ambient ism on kiloparsec scales and produces optical and x - ray emission lines analogous to the situation at radii beyond 2 ( 130 pc ) in ngc4151 @xcite .
notably , se s [ oiii]@xmath405007 luminosity is about 37 - 76 times that of ngc4151 on kpc scales @xcite , perhaps because of se s location in a merger remnant rather than in a typical llagn host .
the new vla and vlba nondetections of nw imply that any agn - driven emission has k - corrected luminosities @xmath55 erg s@xmath4 ( 3 @xmath7 ) on scales below 1.6 kpc and @xmath56 ergs s@xmath4 ( 6 @xmath7 ) on scales below 75 pc . from
[ overview ] , the luminosity of any agn - driven emission from nw can not exceed @xmath46 erg s@xmath4 . from
[ overview ] , the luminosity of nw is @xmath39 } ) = 4.8 - 9.9 \times 10^{42}$ ] erg s@xmath4 @xcite .
the vlba luminosity , free from contamination from star formation , thus implies that nw is radio quiet following the @xcite definition .
the x - ray emission from the nw optical nucleus is less than 4.5% as luminous as se in the hrc passband , suggesting that nw has @xmath57 erg s@xmath4 @xcite .
the nw optical nucleus could harbor an intrinsically faint agn , with a narrow emission - line region only .
however , evidence for a very obscured type 2 agn in nw is beginning to emerge . recent prescriptions for identifying obscured agns
@xcite can be applied . from
[ overview ] , the excitation ratio of nw is [ oiii]@xmath405007/h@xmath41 @xmath42 @xcite .
such a ratio , when combined with estimates for the host galaxy s stellar mass @xmath58 @xcite , occupy the agn locus of the mass - excitation diagram ( fig . 4 ; * ? ? ?
moreover , as @xcite suggest , nw could lack an x - ray counterpart due to atomic absorption . comparing nw s compton - thickness parameter @xmath59})$ ] to its luminosity @xcite
identifies it as a compton - thick candidate ( fig .
10 ; * ? ? ?
nw could also lack a counterpart at @xmath1 = 9.0 ghz due to free - free absorption by its narrow emission - line gas , a concept applied to account for the flat radio spectra of some type 2 agn with @xmath60 } ) = 10^{42}$ ] erg s@xmath4 @xcite . for a characteristic electron temperature @xmath21
k and density @xmath61 @xmath62 , a path length of about 380 pc gives a free - free optical depth of unity at @xmath1 = 10 ghz @xcite .
intriguingly , this free - free path length resembles the scale length of about 100 mas ( 500 pc ) reported for nw s starlight @xcite .
this path length can be cross - checked in several ways .
first , the h@xmath41 luminosity for either the blueshifted or redshifted narrow emission - line gas @xcite should nt be exceeded ; this can be achieved by invoking a plausible volume - filling factor @xmath63 .
second , if the narrow emission - line gas has a spherical geometry with diameter 150 mas ( 2 @xmath9 380 pc ) , the source at @xmath1 = 9.0 ghz should be much smaller ; this can be checked with sensitive vlba imaging .
third , at vlba resolutions the spectrum of the source at frequencies @xmath64 9.0 ghz should be exponentially suppressed ; this is also testable with sensitive vlba imaging .
@xcite assembled the spectral energy distribution at far - infrared and shorter wavelengths , corrected it for the contributions of the type 1 agn in se , and inferred an infrared ( ir ) luminosity @xmath65 erg s@xmath4 .
that luminosity corresponds to a kroupa - based sfr @xmath11 23 m@xmath24 yr@xmath4 ( * ? ? ?
* eqn . 4 ) , although this value would be overestimated if dust warmed by an obscured type 2 agn in nw contributed to the ir luminosity ( e.g. , * ? ? ?
* ; * ? ? ?
* ) . from
[ overview ] , the new vla upper limit to the sfr at the location of the nw optical nucleus is sfr @xmath23 17 m@xmath24 yr@xmath4 .
this lower radio - based sfr suggests that the ir photometry could suffer some contamination from dust warmed by an obscured agn .
we presented new radio imaging of j1000 + 0206 , a merger remnant at @xmath66 with two optical nuclei , nw and se , offset by 500 mas ( 2.5 kpc ) .
the new vla imaging at @xmath1 = 9.0 ghz is sensitive to emission driven by agns and/or star formation , and the new vlba imaging at @xmath1 = 1.52 ghz is sensitive only to emission driven by agns .
no radio emission was detected at these frequencies . combining new and prior results ,
the following self - consistent picture emerged : * the new vla photometry at @xmath67 ghz implies a sfr @xmath23 17 m@xmath24 yr@xmath4 for the inner merger remnant , the optical nuclei se and nw , and the x - ray feature adjacent to the se nucleus . after correcting the prior vla photometry at @xmath68 ghz for this sfr
, the agn - driven emission has an integrated luminosity @xmath69 erg s@xmath4 , in the realm of llagn in the local universe . *
the properties of se and its adjacent x - ray feature match those of the unobscured prototype agn in ngc4151 . however , the luminosity of the narrow emission - line gas associated with se is about forty to eighty times higher than that associated with the agn in ngc4151 .
deeper radio , optical and x - ray studies of se will test its resemblance to , and difference from , the type 1 agn prototype in ngc4151 .
for example , the difference in narrow emission - line luminosities could reflect atypical , merger - induced conditions near se . *
the properties of nw are consistent with it hosting a compton - thick agn that warms the ambient dust and photoionizes the narrow emission - line gas , and is free - free absorbed by that gas at @xmath1 = 9.0 ghz .
these findings enhance the prospects that agn - driven emission could emerge from radio and x - ray studies of nw at deeper levels and x - ray studies of nw at harder wavelengths . for example , deeper vlba imaging could reveal a radio spectrum that is exponentially suppressed by free - free absorption .
these findings also underscore the urgency of localizing narrow emission - line gas , whether blueshifted or redshifted , to nw . in the scenario where se is an agn that has recoiled from nw due to the asymmetric emission of gravitational waves during black - hole coalescence , nw can not host an obscured agn .
the emerging evidence for a very obscured type 2 agn in nw thus both weakens the case for se being a gravitational - wave recoil and strengthens the case that each optical nucleus contains it own agn .
two interpretations are consistent with the latter scenario : ( i ) se and nw mark inspiralling dual agns like those sought in systematic spectroscopic surveys for dual agn candidates @xcite , or ( ii ) se is a gravitational - slingshot recoil moving away from nw and the type 2 agn that it harbors @xcite . it must be noted that se s broad h@xmath41 emission line is redshifted by more than 1000 km s@xmath4 relative to the narrow h@xmath41 emission line @xcite .
such large velocity offsets occur in less than 1 % of normal type 1 agn @xcite , making it very unlikely that they will occur by chance in an inspiralling type 1 agn .
in contrast , such large velocity offsets are predicted for a gravitational - slingshot recoil @xcite , a point in favor of that interpretation for j1000 + 0206 .
we thank the referee for a timely and helpful report .
we also acknowledge using ned wright s cosmology calculator @xcite and benefiting from helpful discussions with michael cooper , francisco muller - sanchez , jonathan trump and craig walker .
em acknowledges financial support from the deutsche forschungsgemeinschaft through project for1254 .
mcmullin , j. p. , waters , b. , schiebel , d. , young , w. , & golap , k. 2007 , astronomical data analysis software and systems xvi ( asp conf .
376 ) , ed .
r. a. shaw , f. hill , & d. j. bell ( san francisco , ca : asp ) , 127 | cosmosj100043.15 + 020637.2 is a merger remnant at @xmath0 with two optical nuclei , nw and se , offset by 500 mas ( 2.5 kpc ) .
prior studies suggest two competing scenarios for these nuclei : ( 1 ) se is an active galactic nucleus ( agn ) lost from nw due to a gravitational - wave recoil . ( 2 ) nw and se each contain an agn , signaling a gravitational - slingshot recoil or inspiralling agns .
we present new images from the very large array ( vla ) at a frequency @xmath1 = 9.0 ghz and a fwhm resolution @xmath2 = 320 mas ( 1.6 kpc ) , and the very long baseline array ( vlba ) at @xmath1 = 1.52 ghz and @xmath2 = 15 mas ( 75 pc ) .
the vla imaging is sensitive to emission driven by agns and/or star formation , while the vlba imaging is sensitive only to agn - driven emission .
no radio emission is detected at these frequencies .
folding in prior results , we find : ( a ) the properties of se and its adjacent x - ray feature resemble those of the unobscured agn in ngc4151 , albeit with a much higher narrow emission - line luminosity .
( b ) the properties of nw are consistent with it hosting a compton - thick agn that warms ambient dust , photoionizes narrow emission - line gas and is free - free absorbed by that gas . finding ( a ) is consistent with scenarios ( 1 ) and ( 2 )
. finding ( b ) weakens the case for scenario ( 1 ) and strengthens the case for scenario ( 2 ) .
follow - up observations are suggested . |
let @xmath4 be the fundamental group of a closed oriented surface @xmath5 of genus @xmath6 . for @xmath1 or @xmath2 , the set
@xmath7 is naturally a @xmath8algebraic set , called the _
@xmath9 representation variety _ of @xmath4 .
the representation variety inherits a topology from its ambient affine space ; call this the _ classical topology_. we prove the following [ main ] for @xmath10 or @xmath2 , the set of faithful representations is dense in @xmath7 equipped with its classical topology .
this answers a question of w. goldman @xcite .
given a representation @xmath11 , there is an associated flat principal @xmath9bundle over @xmath5 , see @xcite .
the only obstruction to building a cross - section of this bundle is a class @xmath12 in @xmath13 . when @xmath14 , @xmath12 is the _ euler number _ @xmath15so called as it is the euler number of the associated @xmath16bundle over @xmath5and when @xmath1 , @xmath12 is the _ second stiefel - whitney class _ @xmath17as it is the second stiefel - whitney class of the associated @xmath18bundle over @xmath5 .
the second stiefel - whitney class of a real representation is the euler number modulo two . in @xcite
, goldman proves that the topological components of @xmath19 are @xmath20 for @xmath21 and those of @xmath22 are @xmath23 for @xmath24 .
furthermore , each representation variety has two irreducible components , corresponding to the two possible stiefel - whitney classes @xcite .
we prove that in each of these irreducible components , the set of faithful representations is a countable intersection of open dense sets , which by the baire category theorem is dense . to do this
, we use the fact that fuchsian representations have even euler numbers and the following [ tool ] for any nontrivial @xmath25 , there is a representation @xmath26 such that @xmath27 and @xmath28 is nontrivial . in section 4
we discuss the effect upon the obstruction classes when a representation is altered by choosing a new embedding of the entry field of its image . in the final section we use our techniques to prove a result about representations into @xmath3 .
fix a hyperbolic structure on @xmath5 . if @xmath29 is a geodesic in @xmath5 , let @xmath30 denote its length with respect to the chosen hyperbolic structure .
let @xmath31 be a closed geodesic in @xmath5 .
since the closed geodesics are in 1 - 1 correspondence with the conjugacy classes in @xmath32 for any choice of basepoint @xmath33 , it is enough to show that for such a geodesic , there is some basepoint @xmath33 and a representation of @xmath32 sending some element @xmath34 in the conjugacy class corresponding to @xmath31 to a nontrivial isometry of @xmath35 .
choose a simple closed geodesic @xmath36 on @xmath5 with @xmath37 , @xmath38 , and so that @xmath36 cuts @xmath5 into a surface @xmath39 of genus @xmath40 and a punctured torus @xmath41 .
this can be done by applying high powers of a pseudo - anosov homeomorphism to any simple closed curve cutting off a punctured torus .
let @xmath42 , and let @xmath43 and @xmath44 .
then @xmath45 , where @xmath46 is the subgroup generated by @xmath47 $ ] , the homotopy class of @xmath36 relative to @xmath33 .
given an element @xmath48 in @xmath32 , let @xmath49 denote its normal closure in @xmath32 .
[ noncomm ] there exists @xmath50 in the free homotopy class of @xmath31 such that @xmath51 where , for each @xmath52 , @xmath53 \rangle \rangle}$ ] and @xmath54^k \rangle \rangle}$ ] for all @xmath55 and is not conjugate to a power of @xmath47 $ ] .
orient @xmath36 , and write @xmath31 as a concatenation of oriented geodesic arcs , @xmath56 where @xmath57 is contained in @xmath39 and @xmath58 is contained in @xmath41 for all @xmath52 .
for each @xmath52 let @xmath59 be the arc of @xmath36 from @xmath33 to the terminal point of @xmath58 in the positive direction , and let @xmath60 be the shortest arc of @xmath36 from the terminal point of @xmath58 to its initial point . for any arc @xmath29 , let @xmath61 denote @xmath29 with the opposite orientation .
now for @xmath62 , define @xmath63 \ \
\ b_i=[\epsilon_i
.\delta_i .\beta_i .\overline{\epsilon_i}],\ ] ] where all indices are taken modulo @xmath64 .
then transparently , the word @xmath65 is in the free homotopy class of @xmath31 .
note that @xmath66 is freely homotopic to the closed curve @xmath67 with length less than @xmath68 .
this curve is essential and not homotopic into @xmath36 , since transverse geodesics intersect minimally , and so @xmath66 is not conjugate to a power of @xmath47 $ ] .
similarly , note that @xmath69 is freely homotopic to @xmath70 .
this curve has length less than @xmath71 and is essential for the same reason as above .
the lemma follows immediately from the following ( compare the proofs of lemma 1 and theorem 6 of @xcite ) : if @xmath72^k \rangle \rangle}$ ] , then @xmath73 . by theorem 1 of @xcite ,
the cover @xmath74 corresponding to @xmath75^k \rangle \rangle}$ ] is planar .
let @xmath76 have the shortest geodesic representative among all nontrivial elements of @xmath75^k \rangle \rangle}$ ] .
then every lift of @xmath76 to @xmath74 has a simple geodesic representative , for otherwise one could find a shorter element in @xmath77^k \rangle \rangle}$ ] .
suppose the geodesic representatives of two lifts @xmath78 and @xmath79 intersect transversely . by the planarity of @xmath74 ,
the number of points of intersection is even .
choose two such points , and let @xmath29 be the shortest of the arcs of @xmath78 and @xmath80 the shortest of the arcs of @xmath79 in the complement of these two points .
then @xmath81 has a shorter geodesic representative than @xmath78 , a contradiction .
this shows that @xmath76 is some power of a simple closed curve in @xmath5 .
but by corollary 4.2 of @xcite , the only such powers in @xmath75^k \rangle \rangle}$ ] are of the form @xmath47^{ik}$ ] , @xmath82 .
we now define a family of representations @xmath83 , @xmath84 , with euler number @xmath85 , using a construction of goldman @xcite .
let @xmath86 denote the projective class of @xmath87 pick two nonparallel simple closed curves on @xmath41 intersecting in a single point , and let @xmath76 and @xmath88 represent them in @xmath89 .
then @xmath90 , and we define a solvable representation @xmath91 by @xmath92 where @xmath93 and @xmath94 .
we record for future reference that @xmath95 ) = \phi_b([x , y ] ) = \begin{bmatrix } 1 & \beta(\alpha^2 - 1 ) \\ 0 & 1 \end{bmatrix},\ ] ] is parabolic . choose a fuchsian representation @xmath96 so that @xmath97)=\phi_b([\gamma])$ ] .
note that since @xmath98 is discrete and faithful , and @xmath97)$ ] fixes @xmath99 , for no element @xmath100 can @xmath101 fix @xmath99
. therefore , for every @xmath100 , @xmath101 has a nonzero @xmath102 entry . to complete the construction , for @xmath84 define @xmath103 $ ] , and
note that since @xmath104 commutes with elements of @xmath105 , the representations @xmath98 and @xmath106 agree on @xmath46 .
the universal property of free products with amalgamation yields a representation @xmath107 given by @xmath108 the euler number of each @xmath109 is @xmath85 , just as in the proof of lemma 8.2 of @xcite .
the following lemma , which uses techniques of the proof of proposition 1.3 of @xmath110 , gives a criterion for an element to survive @xmath109 .
[ trans ] suppose @xmath111 is of the form @xmath112
where for each @xmath52 , @xmath113 has a nonzero @xmath102 entry and @xmath114 is hyperbolic .
if @xmath115 is transcendental over the entry field of @xmath116 , then @xmath117 is not the identity .
let @xmath115 be transcendental over the entry field of @xmath116 .
the lemma follows from the following the entries of @xmath117 are polynomials in @xmath115 , where the degree of the @xmath118 entry is @xmath119 , the degree of the @xmath120 entry is at most @xmath119 , and the other entries have degree at most @xmath121 .
we prove the claim by induction on @xmath119 .
suppose @xmath122 .
let @xmath123= \phi_a(a_1)$ ] and @xmath124=\phi_b(b_1)$ ] . then
@xmath125 since by assumption @xmath126 and @xmath127 ( because @xmath128 is hyperbolic ) , the claim visibly holds .
now suppose @xmath129 .
write @xmath130 by the base case , @xmath131 and @xmath132 are both constant in @xmath115 , and @xmath133 and @xmath134 are at most linear in @xmath115 ( where @xmath134 has a nonzero @xmath115-coefficient ) .
then by the inductive hypothesis , @xmath135 has degree @xmath119 and @xmath136 has degree at most @xmath121 , and hence the @xmath118 entry of @xmath28 has degree @xmath119 in @xmath115 .
the argument for the other entries is similar .
it is easy to see that the set of elements of @xmath89 taken by @xmath137 to parabolics is precisely the commutator subgroup of @xmath89 , which is contained in @xmath75 \rangle \rangle}$ ] .
therefore , by lemma [ noncomm ] , there is a word @xmath34 in the conjugacy class corresponding to @xmath31 satisfying the hypotheses of lemma [ trans ] .
this proves the theorem .
write @xmath138 where @xmath139 is the irreducible component consisting of representations with stiefel - whitney class @xmath52 for @xmath21 .
similarly write @xmath140 where @xmath141 is the irreducible component consisting of representations with euler number equal to @xmath52 modulo 2 for @xmath21 .
note that @xmath142 under the natural inclusion @xmath143 . for
an element @xmath34 of @xmath4 , let @xmath144 ( respectively , @xmath145 ) be the algebraic subset of @xmath146 ( @xmath147 ) consisting of representations killing @xmath34 . by theorem [ tool ] ,
if @xmath25 is nontrivial then @xmath148 and @xmath149 are proper algebraic subsets .
it is a standard fact about irreducible complex varieties ( see page 124 of @xcite ) that the complement of any proper subvariety is dense in the classical topology .
therefore @xmath150 is an open dense subset of @xmath146 . for arbitrary irreducible real algebraic varieties
it is not true that the complements of proper subvarieties are dense in the classical topology .
however it is true given the additional hypothesis that smooth points are dense ( see the discussion below ) , and this follows from proposition 3.7 of @xcite : the set of smooth points of @xmath151 is dense in @xmath151 .
so @xmath152 is an open dense set of @xmath147 .
but the set of faithful representations in @xmath147 is precisely @xmath153 and similarly for faithful representations in @xmath146 .
this is an intersection of open dense subsets , which by the baire category theorem is dense .
this proves that the set of faithful representations is dense in the irreducible component of each representation variety corresponding to a nonzero stiefel - whitney class . in the other component
this is immediately evident by the above argument , since fuchsian representations , for instance , are faithful with stiefel - whitney class 0 .
here is an example that shows that the complement of a proper subvariety of an irreducible real algebraic variety is not always dense .
the two - variable polynomial @xmath154 is irreducible , and so the set @xmath155 is an irreducible real algebraic variety .
the points @xmath156 and @xmath157 are the only elements of @xmath158 with @xmath88-coordinate equal to @xmath159 , and if @xmath160 with @xmath161 , then @xmath162 , which implies that @xmath163 .
hence @xmath156 is an isolated point of @xmath158 in the classical topology , even though @xmath158 is one dimensional .
it is also a proper subvariety , which , by the above , is not approached by points in its complement . in this example , the origin is not a smooth point of the variety .
the following fact shows that the situation at smooth points is analogous to the complex case .
we give a proof , adapted from the proof of the complex case in @xcite .
[ realalg ] let @xmath164 be a real algebraic variety of dimension @xmath165 , and let @xmath166 be a smooth point .
let @xmath167 be a subvariety of dimension @xmath168 with @xmath169 .
then @xmath76 is approached by a sequence in @xmath170 .
the proof is by induction on @xmath165 . if @xmath171 , then @xmath172 .
hence @xmath173 consists of a finite collection of points .
since @xmath76 is a smooth point of @xmath174 , there is a chart @xmath175 around @xmath76 .
now pick any sequence in @xmath176 that approaches @xmath177 and misses @xmath178 .
now suppose @xmath179 .
for each irreducible component @xmath141 of @xmath173 , let @xmath180 , and choose an affine hyperplane @xmath181 so that @xmath182 , @xmath183 for all @xmath52 , and @xmath181 is transverse to @xmath174 at @xmath76 .
then @xmath184 is @xmath185dimensional and no irreducible component of @xmath173 is contained in @xmath184 .
so for each @xmath52 , @xmath186 furthermore , @xmath76 is a smooth point of @xmath187 since the tangent space at @xmath76 is of the proper dimension , and the claim follows by induction .
now for a variety @xmath174 in which smooth points are dense , any point @xmath76 in a proper subvariety @xmath173 is approached by smooth points of @xmath174 , each of which is either in @xmath170 or approached by a sequence in @xmath170 .
a diagonal argument gives a sequence in @xmath170 approaching @xmath76 .
given a representation @xmath188 with image @xmath189 , a presentation @xmath190 \cdots [ x_{2g-1},x_{2 g } ] = 1\,\rangle,\ ] ] and lifts @xmath191 to the universal covering @xmath192 , we have @xmath193 \cdots [ { \widetilde{\phi}}(x_{2g-1}),{\widetilde{\phi}}(x_{2 g } ) ] \in \ker\thinspace p \cong \pi_1 ( \mathrm{psl}_2(\mathbb{k})),\ ] ] see @xcite .
we may choose a new embedding @xmath194 of the entry field of @xmath189 to obtain a new representation @xmath195 with image @xmath196 .
computing the stiefel - whitney class of @xmath195 , we have @xmath197 \cdots [ { \widetilde{\phi^{\sigma}}}(x_{2g-1}),{\widetilde{\phi^{\sigma}}}(x_{2 g } ) ] \\ & = & ( \pm i)^{\sigma } \\ & = & \pm i , \end{aligned}\ ] ] since @xmath198 has a unique complex embedding
. we therefore have @xmath199 .
this implies that the euler numbers of two such real representations have the same parity .
they may not be equal , as the following example , from @xcite pp .
161 - 162 ( originally in @xcite ) , shows . consider the saccheri quadrilateral above , where the angle at a is @xmath200 , embedded in the upper half plane so that c is at @xmath52 and the side dc lies along the geodesic between @xmath159 and @xmath99 .
the fuchsian subgroup @xmath201 of the group generated by reflections in the sides of this quadrilateral has presentation @xmath202 where @xmath76 , @xmath88 , and @xmath203 are rotations by @xmath4 around b , c , and d , respectively .
it surjects @xmath204 ( where @xmath205 acts on @xmath206 by @xmath207 ) by @xmath208 the kernel of this map , @xmath209 , is torsion - free with coarea @xmath210 and so it is the deck group of a genus 2 surface .
let @xmath211 be the length of the side bc , @xmath212 .
as in @xcite , the invariant trace field of @xmath201 is @xmath213 .
note that @xmath209 is contained in the commutator subgroup of @xmath201 , and so its trace field is contained in @xmath213 .
the element @xmath214 is in @xmath209 .
conjugate @xmath209 so that @xmath215 is @xmath216 , where @xmath217 .
we may conjugate the result to obtain a group @xmath189 whose entries lie in @xmath218see the discussion on page 115 of @xcite and so the entries of @xmath189 lie in @xmath219 .
choose @xmath211 so that @xmath220 .
then the entries of @xmath189 lie in @xmath221 .
let @xmath194 be the embedding given by @xmath222 .
consider the element @xmath223 of @xmath209 , which has trace @xmath224 .
note that @xmath225 and so the trace of @xmath226 ( where @xmath227 ) is @xmath228 .
now , by the main theorem of @xcite , any representation of the genus 2 surface group with euler number @xmath229 is fuchsian .
since @xmath196 contains the elliptic element @xmath226 , the associated representation must have euler number 0 .
it is natural to ask to what extent these techniques prove informative about representations of surface groups into other lie groups .
one example is the following [ toledo ] for @xmath230 and any even @xmath165 , @xmath231 , there are faithful representations @xmath232 with toledo invariants @xmath233 .
theorem [ toledo ] follows from [ blonk ] for @xmath230 and any finite collection of nonidentity elements @xmath234 , there is a representation @xmath235 with toledo invariant @xmath236 such that @xmath237 is nontrivial for all @xmath238 .
we recall the construction of proposition 5.1 of @xcite .
let @xmath36 be a simple closed geodesic on @xmath5 which cuts off a punctured torus , and let @xmath239 and @xmath240 be the fundamental groups of the punctured genus @xmath40 surface and torus , respectively , cut off by @xmath36 , where @xmath33 is a point on @xmath36
. then @xmath181 and @xmath89 are free and @xmath241 , as before .
choose generators @xmath242 for @xmath181 represented by simple closed curves on @xmath39 so that @xmath243 \cdots [ x_{2g-3},x_{2g-2}]= [ \gamma]$ ] and generators @xmath244 for @xmath89 represented by simple closed curves on @xmath41 so that @xmath245=[\gamma]^{-1}$ ] .
this determines an isomorphism @xmath246 .
now let @xmath247 be the regular @xmath248-gon in @xmath249 with vertex angles @xmath250 , and let @xmath251 be the regular quadrilateral in @xmath35 with vertex angles @xmath252 .
define @xmath253 by sending the @xmath254 to the appropriate transvections pairing sides of @xmath247 so that the quotient of @xmath255 by @xmath256 is a genus @xmath40 surface with one cone point of order 3 .
then @xmath257)$ ] is an elliptic of order 3 fixing some vertex @xmath258 of @xmath247 . as in @xcite
, this rotates by @xmath259 in @xmath260 and by @xmath261 in the normal complex geodesic .
similarly define @xmath262 so that the quotient is a punctured torus with one cone point of order @xmath263 , then we may assume without loss of generality that @xmath264)=\phi_a([\gamma])$ ] .
let @xmath265 fix @xmath258 and interchange @xmath266 with its normal complex geodesic at that point ; then @xmath267 ) \theta^{-1 } = \phi_b([\gamma])^{-1}$ ] .
we thus obtain a representation @xmath235 given by @xmath268 and @xmath269 .
this representation has toledo invariant @xmath236 ( see @xcite ) . given a nontrivial element @xmath25 , lemma [ noncomm ] allows us to choose @xmath36 so that @xmath270 where the @xmath271 and @xmath272 are hyperbolic .
in fact , it is easy to see that @xmath36 may be chosen so that the above holds for any finite collection of elements @xmath273 .
half the side length of the regular @xmath64-gon in @xmath260 with vertex angles @xmath274 is @xmath275 given by @xmath276 . the largest ball around a vertex of the polygon @xmath247 which is either disjoint from or identical to each of its translates under @xmath256 has radius @xmath277 .
the largest such ball around any vertex of @xmath251 has radius @xmath278 .
let @xmath279 and @xmath280 be the orthogonal projections .
an elementary argument , similar to the proof of proposition 3.1 of @xcite , shows that @xmath281 does not intersect @xmath282this is where the hypothesis that @xmath230 is used , as @xmath283 and @xmath284 do indeed intersect .
therefore any hyperbolic element of @xmath256 takes @xmath285 inside @xmath286 and vice - versa . by a ping - pong argument
, @xmath28 is a nontrivial isometry of @xmath287 . by the main theorem of @xcite
, there is a unique topological component @xmath288 of the @xmath3 representation variety of @xmath4 consisting of representations with toledo invariant @xmath236 .
let @xmath289 be the irreducible components that contain smooth points in @xmath288 .
suppose that for each @xmath52 there is an element @xmath290 such that all representations in @xmath291 kill @xmath292 .
theorem [ blonk ] produces a representation in which all of the @xmath292 survive ; furthermore , this representation is a smooth point of @xmath288 ( this follows from @xcite , proposition 3.7 ) , a contradiction .
hence there is some irreducible component @xmath293 for which each @xmath294 is a proper subvariety , and a representation @xmath295 with toledo invariant @xmath236 which is a smooth point of @xmath293 .
let @xmath296 be an open neighborhood of @xmath297 in @xmath293 consisting entirely of smooth points .
then for each @xmath298 , @xmath299 is nowhere dense in @xmath296 by lemma [ realalg ] .
the union of such sets is nowhere dense in @xmath296 , and so faithful representations exist in @xmath296 .
the more general statement follows from the fact that , given a surface @xmath5 of genus @xmath230 and a finite set of nontrivial elements in @xmath300 , there is a degree one map from @xmath5 to a surface of genus @xmath40 so that each element of the finite set survives the induced map between fundamental groups .
a stronger theorem concerning @xmath3 has been announced by n. gusevskii .
the authors thank peter storm and ben mcreynolds for making us aware of this question and william goldman for providing a copy of @xcite .
we also thank alan reid for useful conversations , in particular for making us aware of @xcite . | we show the set of faithful representations of a closed orientable hyperbolic surface group is dense in both irreducible components of the @xmath0 representation variety , where @xmath1 or @xmath2 , answering a question of w. goldman .
we also prove the existence of faithful representations into @xmath3 with certain nonintegral toledo invariants . |
a variety of surface and near - surface thermodynamic and lattice dynamical phenomena are governed by the phonon density of states ( dos ) , the number of vibrational modes per unit frequency , per unit volume of solid . in an inhomogeneous system , however , it is possible to distinguish between the `` global '' dos , which makes reference to only to the vibrational - mode frequencies of the entire system , and a local , eigenfunction - weighted dos . in systems with a small _ mechanical _ surface - to - volume ratio ,
the volume including all parts of the system and environment participating in the vibrations , the global dos is essentially unchanged from that of a bulk system .
the eigenfunction - weighted dos is interesting both because it is relevant to a variety of physical phenomena , and because it contains local , position - dependent mechanical and vibrational information about the system . in this paper
, we calculate the local phonon dos tensor for a model substrate consisting of a semi - infinite isotropic elastic continuum with a stress - free surface , appropriate for low - frequency acoustic phonons in real solids with an exposed clean surface .
the elastic continuum is also assumed to be linear and nonpolar .
the free surface is taken to lie in the @xmath3 plane , with the substrate occupying the @xmath4 half - space .
periodic boundary conditions are applied in the lateral directions , on an area @xmath5 .
normalized vibrational eigenfunctions have been obtained for this geometry in a classic paper by ezawa.@xcite
we begin by defining the local dos tensor @xmath6 , the quantity to be calculated in this paper . the lagrangian density for a linear , nonpolar , isotropic elastic continuum is @xmath7 where @xmath8 is the displacement field , and where @xmath9 is the strain tensor .
@xmath10 is the mass density , and @xmath11 and @xmath12 are the standard lam coefficients , related to the bulk longitudinal and transverse sound velocities @xmath2 and @xmath13 according to @xmath14 and @xmath15 the equation of motion following from eq .
( [ lagrangian ] ) is @xmath16 it will also be useful to define an elastic stress tensor @xmath17 from the continuity equation @xmath18 for momentum density @xmath19 . in an isotropic elastic medium
, it follows from eq .
( [ elasticity equation ] ) that @xmath20 where @xmath21 is the corresponding elastic tensor .
the boundary condition at the @xmath22 exposed planar surface @xmath23 is @xmath24 to quantize the elastic waves we require the displacement field and momentum density to satisfy the canonical commutation relation @xmath25 = i \hbar \delta_{ij } \delta({\bf r}-{\bf r } ' ) .
\label{commutation relation}\ ] ] therefore , we can expand the displacement field in a basis of bosonic creation and annihilation operators according to @xmath26 , \label{displacement field expansion}\ ] ] where the @xmath27 are vibrational eigenfunctions , defined to be time - periodic solutions of the elasticity equation ( [ elasticity equation ] ) , in the presence of the stated boundary conditions , and normalized according to @xmath28 here the integration is over the volume @xmath29 of the system .
the vibrational eigenfunctions are assumed to be complete , so that @xmath30 the quantity we shall calculate in this paper is@xcite @xmath31^ * \ ! , \label{g definition}\ ] ] where @xmath32 and @xmath33 are cartesian tensor indices .
this quantity weights each vibrational mode @xmath34 by the square of the amplitude of the eigenfunction at position @xmath35 .
the trace of @xmath6 also contains physically interesting information , namely @xmath36 now , averaging the trace of @xmath6 over the system volume @xmath29 leads to @xmath37 which is the ordinary intensive global dos , i.e. , the number of vibrational modes per unit frequency per volume . in an isotropic system with translational invariance ,
the local dos is independent of position , and eq .
( [ g definition ] ) simplifies to @xmath38 where a three - dimensional system is assumed .
because there is no position dependence here , the trace of this quantity yields the global dos [ the right - hand - side of eq .
( [ average trace expression ] ) ] . is the local dos defined eq .
( [ g definition ] ) a quantum mechanical quantity ? one the one hand , it does not involve @xmath39 , and in that sense it is not quantum mechanical . on the other hand , however , the normalization assumed for the @xmath27 is of quantum - mechanical origin : assuming the conventional bosonic commutation relations for the @xmath40 and @xmath41 appearing in the expansion of eq .
( [ displacement field expansion ] ) , the normalization condition on the vibrational eigenfunctions is required for eq .
( [ commutation relation ] ) to be satisfied.@xcite therefore , the local dos defined eq . ( [ g definition ] ) naturally shows up in quantum mechanical calculations of the vibrational properties of inhomogeneous systems .
in what follows it will be useful to recall the well - known debye dos formula .
the spectrum and normalized vibrational eigenfunctions for a three - dimensional isotropic elastic continuum , with periodic boundary conditions applied to the surface of a three - dimensional volume @xmath29 , are @xmath42 where @xmath43 labels the longitudinal and two transverse branches , the sound velocities are the bulk values quoted in sec .
[ definition section ] , and the @xmath44 are unit polarization vectors forming an orthogonal basis .
@xmath45 is orietnted in the @xmath46 direction , and @xmath47 and @xmath48 are perpendicular to @xmath45 and to each other . in this case , eq . ( [ g definition ] ) yields @xmath49 the fact that @xmath50 is independent of the direction of @xmath46 allows the product of polarization vectors to be averaged over the unit sphere , leading to @xmath51 in @xmath52 dimensions .
we then obtain @xmath53 an example of the form given in eq .
( [ translationally invariant g ] ) .
therefore , in the @xmath54 limit , @xmath55 eq . ( [ bulk g ] ) is the conventional three - dimensional debye dos tensor for acoustic phonons .
of course , in a real solid the continuum approximation breaks down at short wavelengths and the spectrum is cut off at high frequencies .
the evaluation of the local dos for our substrate model requires the vibrational eigenfunctions of a semi - infinite isotropic elastic continuum with a stress - free surface , which have been obtained by ezawa.@xcite the modes are labeled by a branch index @xmath56 , taking the five values sh , @xmath57 , @xmath58 , 0 , and r , by a two - dimensional wavevector @xmath59 in the plane defined by the surface , and by a parameter @xmath60 with the dimensions of velocity that is continuous for all branches except the rayleigh branch @xmath61 . the range of the parameter @xmath60 depends on the branch @xmath56 , and is given in table [ c table ] .
the frequency of mode @xmath62 is @xmath63 where @xmath64 is the magnitude of the two - dimensional wavevector . as stated above
, we shall assume periodic boundary conditions in the @xmath65 and @xmath66 direction applied to a finite area @xmath5.@xcite
.values of the parameter @xmath60 for the five branches of vibrational modes of a semi - infinite substrate .
the rayleigh branch @xmath67 has a single value of @xmath60 , which we denote by @xmath68 . [
cols="^,^",options="header " , ] the definition of the local dos given in eq .
( [ g definition ] ) applies to a system with a discrete spectrum .
the generalization of eq .
( [ g definition ] ) to our model , which has a mixed discrete - continuous spectrum , is @xmath69^ * + \sum_{\bf k } \sum_{m \neq { \rm r } } \int_{\gamma_{\ ! m } } \ ! \ ! \ ! dc \
\delta(\omega - ck ) \ , f_{m{\bf k}c}^i({\bf r } ) \ , [ f_{m{\bf k}c}^j({\bf r})]^ * \ ! .
\label{substrate g definition}\ ] ] the first term accounts for the rayleigh surface wave branch @xmath70 , and these vibrational mode eigenfunctions are normalized as in sec .
[ definition section ] .
the value of @xmath60 for the rayleigh branch , @xmath71 is calculated below . the second term in eq .
( [ substrate g definition ] ) accounts for the four remaining branches @xmath72 and @xmath73 having continuous values of @xmath60 .
their eigenfunctions are instead normalized according to @xmath74 furthermore , @xmath75 is the integration domain given in table [ c table ] .
we turn now to an evaluation of eq .
( [ substrate g definition ] ) . the contribution from each branch in table [ c table ] will be discussed separately in the sections to follow .
first we consider the @xmath76 branch , for which@xcite @xmath77 where @xmath78 and where @xmath79 is a unit vector in the @xmath59 direction .
this mode is polarized in the @xmath3 plane , perpendicular to @xmath59 .
the contribution made by this branch to the local dos is @xmath80 taking the @xmath81 limit , and making use of the _ two_-dimensional dos formula @xmath82 leads to @xmath83 where @xmath84 denotes a two - dimensional average over the directions of @xmath59 .
finally , using @xmath85 leads to @xmath86 where @xmath87 below we will make use of the large @xmath88 limit of @xmath89 , which is @xmath90 for the @xmath92 and @xmath58 branches the normalized vibrational eigenmodes are@xcite @xmath93 { \bf e}_{\rm k } \nonumber \\ & + & \bigg [ \pm \alpha^{1 \over 2 } \big ( e^{-i \alpha k z } + \zeta_{\pm } \ , e^{i \alpha k z } \big ) \nonumber \\ & + & i \beta^{-{1 \over 2 } } \big ( e^{-i \beta k z } - \zeta_{\pm } \ , e^{i \beta k z } \big ) \bigg ] { \bf e}_z \bigg\ } e^{i { \bf k } \cdot { \bf r } } , \label{pm modes}\end{aligned}\ ] ] where @xmath94 here @xmath95 is a phase factor @xmath96 ^ 2 \over ( \beta^2 -1)^2 + 4 \alpha \beta},\ ] ] with @xmath97 we also note that @xmath98 the @xmath99 branches therefore contribute an amount @xmath100 to the local dos , where @xmath101.\ ] ] here @xmath102 ^ 2\over ( { \textstyle{s^2 \over \nu^2 } } - 2)^2 + 4 ( s^2 - 1)^{1 \over 2}({\textstyle{s^2 \over \nu^2 } } - 2)^{1 \over 2 } } \bigg ) e^{i z\sqrt{s^2 - 1}/s } \bigg ] \nonumber \\ & \mp & i \ , \big({\textstyle{s^2 \over \nu^2}}-1\big)^{{1 \over 4 } } \bigg [ e^{-i z \sqrt{(s/\nu)^2 - 1 } /s } + \bigg ( { [ { \textstyle{s^2 \over \nu^2 } } - 2 \pm 2 i ( s^2 - 1)^{1 \over 4}({\textstyle{s^2 \over \nu^2 } } - 2)^{1 \over 4}]^2 \over ( { \textstyle{s^2 \over \nu^2 } } - 2)^2
+ 4 ( s^2 - 1)^{1 \over 2}({\textstyle{s^2 \over \nu^2 } } - 2)^{1 \over 2 } } \bigg ) e^{i z\sqrt{(s/\nu)^2 - 1}/s}\bigg ] \bigg|^2 \ ! , \label{ipm definition}\end{aligned}\ ] ] and @xmath103 ^ 2 \over ( { \textstyle{s^2 \over \nu^2 } } - 2)^2 + 4 ( s^2 - 1)^{1 \over 2}({\textstyle{s^2 \over \nu^2 } } - 2)^{1 \over 2 } } \bigg ) e^{i z\sqrt{s^2 - 1}/s } \bigg ] \nonumber \\ & \pm & i \ , \big({\textstyle{s^2 \over \nu^2}}-1\big)^{-{1 \over 4 } } \bigg [ e^{-i z \sqrt{(s/\nu)^2 - 1 } /s } - \bigg ( { [ { \textstyle{s^2 \over \nu^2 } } - 2 \pm 2 i ( s^2 - 1)^{1 \over 4}({\textstyle{s^2 \over \nu^2 } } - 2)^{1 \over 4}]^2 \over ( { \textstyle{s^2 \over \nu^2 } } - 2)^2 + 4 ( s^2 - 1)^{1 \over 2}({\textstyle{s^2 \over \nu^2 } } - 2)^{1 \over 2 } } \bigg ) e^{i z\sqrt{(s/\nu)^2 - 1}/s}\bigg ] \bigg|^2 \ ! , \label{jpm definition } \end{aligned}\ ] ] where @xmath104 is the ratio of transverse and longitudinal bulk sound velocities . it can be shown that @xmath105 and @xmath106 next we consider the @xmath107 branch , for which@xcite @xmath108 { \bf e}_{\rm k } + \bigg [ - \gamma { \cal c } \ , e^{- \gamma k z } \nonumber \\ & + & i e^{-i \beta k z } - i { \cal a } \
, e^{i \beta k z } \bigg ] { \bf e}_z \bigg\ } e^{i { \bf k } \cdot { \bf r } } , \label{0 modes}\end{aligned}\ ] ] where @xmath109 @xmath110 and @xmath111 note that @xmath112 .
the @xmath113 branch therefore contributes to eq .
( [ substrate g definition ] ) the quantity @xmath114,\ ] ] where @xmath115 and @xmath116 furthermore , we note the asymptotic results @xmath117 and @xmath118 the normalized vibrational eigenmodes for the rayleigh branch @xmath61
are@xcite @xmath119 { \bf e}_{\rm k } \nonumber \\ & - & \bigg[\varphi e^{- \varphi k z } - \big({\textstyle{2 \varphi \over 1 + \eta^2 } } \big ) \ , e^{- \eta k z } \bigg ] { \bf e}_z \bigg\ } e^{i { \bf k } \cdot { \bf r } } , \label{r modes}\end{aligned}\ ] ] where @xmath120 furthermore , @xmath121 in eq .
( [ varphi and eta definitions ] ) , @xmath122 is the velocity of the rayleigh ( surface ) waves , given by @xmath123 where @xmath124 is the root between 0 and 1 of the polynomial @xmath125 with @xmath126 the velocity ratio defined above in eq .
( [ nu definition ] ) . in terms of @xmath124
we can write @xmath127 the rayleigh branch contributes an amount @xmath128 , \end{aligned}\ ] ] where @xmath129 ^ 2 \label{ir definition}\ ] ] and @xmath130 ^ 2 .
\label{jr definition}\ ] ] both @xmath131 and @xmath131 vanish in the large @xmath88 limit . by symmetry ,
the local dos tensor is diagonal , with the in - plane components @xmath132 generally different than the perpendicular component @xmath133 . combining the results derived above ,
we obtain our principal result , @xmath134 , \label{tensor decomposition}\ ] ] where @xmath135 + { 3 \over 4(2+\nu^3 ) } \ , i_{0}\big(\theta/ \nu\big ) + { 3 \pi \over 2(2+\nu^3 ) { \cal k } \xi^3 } \ , i_{\rm r}\big(\theta/ \xi \nu\big ) \label{g par}\ ] ] and @xmath136 + { 3 \over 2(2+\nu^3 ) } \ , j_{0}\big(\theta/ \nu\big ) + { 3 \pi \over ( 2+\nu^3 ) { \cal k } \xi^3 } \ , j_{\rm r}\big(\theta/ \xi \nu\big ) . \label{g per}\ ] ] the prefactor @xmath137 in eq . ( [ tensor decomposition ] ) is the bulk debye dos . in the large @xmath138 limit ,
corresponding to high frequencies @xmath0 , large perpendicular distances @xmath139 , or both , the asymptotic results given above lead to @xmath140 in which case the local dos of eq .
( [ tensor decomposition ] ) reduces to the debye value given in eq .
( [ bulk g ] ) .
dimensionless local dos functions @xmath141 ( solid curve ) and @xmath142 ( dashed curve ) , defined in eqs .
( [ g par ] ) and ( [ g per ] ) , for an isotropic elastic continuum model of a si substrate .
here @xmath143 the dos functions are essentially equal to unity for @xmath144.,width=302 ] to apply these results , we shall approximate si as an isotropic elastic continuum with longitudinal and transverse sound velocities @xmath145 and mass density @xmath146 the velocities above follow from using the measured elastic constants @xmath147 and @xmath148 at @xmath149 reported in ref . [ ] . a more precise method of determining these velocities would be to use the actual anisotropic elastic tensor to construct a best fit to an isotropic one.@xcite however this only makes small changes in the final results .
the rayleigh sound speed is determined from eq .
( [ rayleigh velocity equation ] ) . the velocity ratio defined in eq .
( [ nu definition ] ) is @xmath150 leading to @xmath151 and hence @xmath152 in addition , @xmath153 @xmath154 and @xmath155 in fig .
[ gtensor figure ] , we present plots of the dimensionless dos functions @xmath156 and @xmath157 for si . here
@xmath158 where @xmath0 is the angular frequency , @xmath139 is the perpendicular distance from the surface , or depth , and @xmath2 is the bulk longitudinal sound velocity .
the dimensionless dos has an interesting oscillatory structure when @xmath138 is small , and approaches unity in the large @xmath138 limit .
when @xmath159 , @xmath160 and @xmath161 on the substrate surface , @xmath162 , the local dos is considerably modified from the bulk debye value , as a consequence of eqs .
( [ gpar(0 ) ] ) and ( [ gper(0 ) ] ) .
in particular , @xmath163 and @xmath164 are quadratic in frequency with a constant of proportionality @xmath165 times that of the debye model .
similarly , @xmath166 is quadratic in @xmath0 , with a constant of proportionality a factor of @xmath167 larger than in the debye case .
the oscillatory structure in @xmath156 and @xmath157 also modifies the frequency dependence of the local dos inside the substrate .
this is illustrated in fig .
[ frequency figure ] , where @xmath168 is plotted in units of @xmath169 . in these units ,
the debye dos is simply @xmath170 , shown as the dotted line in the figure .
frequency dependence of the local dos @xmath171 ( solid curve ) and @xmath172 ( dashed curve ) for a fixed @xmath173 , compared to the purely quadratic frequency dependence ( dotted curve ) of the debye model . in all cases
the dos is given in units of @xmath169 .
the frequency is plotted in units of @xmath174 , where @xmath2 is the bulk longitudinal sound velocity .
elastic parameters corresponding to si are assumed.,width=302 ]
we have calculated the local , eigenfunction - weighted acoustic phonon dos tensor for a model substrate consisting of a semi - infinite isotropic elastic continuum with a stress - free surface , and applied the results to si .
the local dos on the surface is quadratic in @xmath0 , with a proportionality constant that is enhanced compared to the three - dimensional bulk debye model .
the enhancement factors for si are given in eqs .
( [ gpar(0 ) ] ) and ( [ gper(0 ) ] ) .
inside the substrate the frequency dependence is non - quadratic , as illustrated in fig .
[ frequency figure ] . at depths @xmath139 large
compared with @xmath1 , where @xmath2 is the bulk longitudinal sound velocity , or at frequencies @xmath0 large compared with @xmath174 , the local dos approaches the isotropic debye value .
these regimes corresponds to the large @xmath138 limits of @xmath156 and @xmath157 , defined in eqs .
( [ g par ] ) and ( [ g per ] ) .
as expected , the presence of the free surface significantly changes the vibrational properties of the system when the condition @xmath175 is met , where @xmath11 is a characteristic bulk phonon wavelength with frequency @xmath0 . at a fixed frequency
, this condition reduces to a requirement on distances from the surface , whereas at fixed depths it becomes a condition on frequency . the normalization condition on the @xmath27 is often said to be required to make the elastic energy associated with the @xmath34th term of @xmath176 in eq .
( [ displacement field expansion ] ) equal to one quanta , @xmath177 .
this energy requirement " actually follows from the more fundamental requirement of eq .
( [ commutation relation ] ) . | the local , eigenfunction - weighted acoustic phonon density of states ( dos ) tensor is calculated for a model substrate consisting of a semi - infinite isotropic elastic continuum with a stress - free surface . on the surface ,
the local dos is proportional to the square of the frequency @xmath0 , as for the three - dimensional debye model , but with a constant of proportionality that is considerably enhanced compared to the debye value , a consequence of the rayleigh surface modes .
the local dos tensor at the surface is also anisotropic , as expected .
inside the substrate the local dos is both spatially anisotropic and non - quadratic in frequency .
however , at depths large compared with @xmath1 , where @xmath2 is the bulk longitudinal sound velocity , the local dos approaches the isotropic debye value .
the results are applied to a si substrate . |
since the first @xmath7 radiative penguin decays of b mesons have been reported by cleo in 1993 @xcite , the measurements of radiative penguins at the b factory experiments babar and belle have expanded into a rich field of physics .
as many as 15 individual exclusive @xmath1 modes have been observed and some of the corresponding branching fractions have been measured with a precision of about @xmath8 .
this makes it possible to test sm predictions of these rare decays , e.g. qcd calculations of form factors .
additionally , the inclusive measurement of @xmath9 allows strong constraints on new physics models @xcite and precision measurements of heavy quark effective theory predictions . with the success of the @xmath1 penguin measurements in hand ,
the next step is to explore @xmath0 penguins , which due to cabbibo suppression are not only about one order of magnitude more rare , but also face a much more severe background due to the abundance of pions in the background .
this paper gives an overview of the recent results from babar for the exclusive modes @xmath10 and @xmath3 .
section [ sec : newphysics ] will briefly discuss the interest in these measurements from the point of view of the search for new physics , sections [ sec : btodll ] and [ sec : btodg ] will summarize the measurements , an discuss the measurement of @xmath5 from @xmath11 .
( 20,15 ) ( 20,15 ) ( 20,15 ) the beauty of the search for new physics in penguin decays lies in the fact that the dominating diagram in the sm is a loop or box diagram with heavy particles in the loop .
this is exemplified in figure [ fig : newphysicsgraphs ] .
possible new physics , supersymmetry ( susy ) or the two - higgs doublet model ( 2hdm ) in this case , contribute with particles in the same mass range of 100 to several hundred gev and often also with similar couplings .
this allows strong constraints on new physics models .
however , while the measurement of exclusive decays is often experimentally clean , the interpretation of the experimental results for very rare exclusive decays often suffers from theoretical uncertainties in the prediction of form factors , radiative corrections , and other suppressed graphs .
still , strong constraints and precise sm measurements can be gained from ratios of branching fractions .
( 40,20 ) ( 40,20 ) the smallest @xmath13 branching fractions measured up to now are the @xmath14 modes @xmath15 and @xmath16 .
@xcite in comparison to that , @xmath17 is expected to be suppressed by @xmath18 , with the most recent prediction of @xmath19 . @xcite any deviation from this order of magnitude would be a striking sign of new physics . the two dominating diagrams for this transition are shown in figure [ fig : btodllpenguins ] , showing one penguin and one @xmath20 box diagram . + ( a )
+ ( b ) the babar analysis @xcite on a dataset of @xmath21 faces another experimental challenge with respect to @xmath22 in addition to the reduced branching fraction , namely the much higher rate of @xmath23 in the background than @xmath24 .
the analysis strategy is to first select clean @xmath23 , @xmath25 and @xmath26 candidates and then veto resonances decaying into @xmath27 .
the @xmath28 events are strongly reduced by requiring two @xmath29 gev leptons in the event . the charmonium veto against @xmath30 events
is shown in figure [ fig : btodllpeakingbackground ] ( a ) .
the tilt in the mass veto stems from the fact that events with bremsstrahlung of one of the leptons are off both in mass and in reconstructed energy , visible in @xmath31 .
event shape variables against continuum events are grouped in a fisher discriminant , and event shape variables against @xmath32 background events are grouped into a likelihood .
after these cuts , only combinatoric background from @xmath33 and @xmath32 events are left .
backgrounds are controlled with various control samples ( @xmath34 , @xmath35 ) .
hadronic mistags are measured in a separate study by inverting the hadron vetos and then imposing the measured mistag rates as event weights .
the resulting small peaking background fraction is shown in figure [ fig : btodllpeakingbackground ] ( b ) with @xmath36 on the left and @xmath37 on the right side .
finally , the background in the @xmath38 and @xmath39 sidebands is extrapolated into the signal region ( see figure [ fig : btodllresults ] ) to assess the background level independent of the mc simulation .
.__results of the @xmath10 selection and limits on the branching fraction inferred from the absence of an excess over the sm background . _ _ [ cols="<,^,^,^ " , ] no excess over the background is observed , hence limits are set using a frequentist cut - and - count method in the signal boxes . the resulting limit of @xmath19 assuming isospin symmetry is within a factor of 3 of the sm prediction . a detailed summary of the results can be found in table [ tab : btopillresults ] .
@xmath41 penguins have been first observed by belle with @xmath42 with evidence for the @xmath43 mode @xcite .
in addition to the possibility of finding new physics if the measured branching fractions exceed @xmath44 for the neutral mode or @xmath45 for the charged mode significantly , it offers the important possibility to measure @xmath5 via @xmath46 with @xmath47 @xcite containing radiative corrections and sub - dominant helicity suppressed @xmath20-fusion diagrams ( hence depending on @xmath48 and the ckm fit itself and thus not uncorrelated from @xmath5 ) , and a form - factor ratio of @xmath49 @xcite .
the graphs involved here are expected to have completely independent possible new physics contributions than @xmath50 , hence the comparison of the two results is very important . here
, the experimental challenge lies not only in the particle i d requirements to suppress @xmath24 background , but also in the @xmath23 combinatorics coming from the wide resonance states ( @xmath51 ) and the high photon background from continuum events with @xmath52 .
the babar analysis @xcite on @xmath53 of data tackles this challenge by a likelihood veto against @xmath54 based on the invariant mass of the photon pair and the energy of the lower energetic photon , improving the veto significantly over a simple cut on @xmath55 .
additionally , a neural net ( nn ) based continuum suppression is applied . to extract the result ,
a simultaneous maximum - likelihood fit is performed to @xmath56 , the nn output and @xmath57 , where @xmath58 is the helicity angle of the vector meson @xmath59 or @xmath60 , which are transversely polarized in signal events . for @xmath60 , the dalitz angle of the @xmath60 decay is used additionally .
an example fit is shown for the @xmath4 mode in figure [ fig : rhofit ] .
many control sample checks are performed . with off - peak data used to control the continuum simulation , and
@xmath61 and @xmath62 used to control resolutions and efficiencies .
lcc mode & result & significance + @xmath63 & @xmath64 & @xmath65 + @xmath66 & @xmath67 & @xmath68 + @xmath69 & @xmath70 & @xmath71 + + @xmath72 & @xmath73 & @xmath74 + the result for the individual modes and different combinations is shown in table [ tab : btorhogresults ] . while the @xmath75 signal is not yet significant on its own , this result represents the first evidence for @xmath76 .
figure [ fig : babarbellecomparison ] shows the good agreement of the results with the sm predictions and the agreement with the earlier belle results .
the result for the isospin combination @xmath77 is to be taken with care , however , since the @xmath60 does not belong to the isospin triplet .
the results of @xmath72 can be used to extract @xmath5 using eq .
[ eq : vtdvts ] . using the world average of @xmath78 ,
a value of @xmath79 @xcite can be extracted , in very good agreement with the result obtained from @xmath80 of @xmath81 from the tevatron @xcite .
while the precision from @xmath77 is not sufficient to compete with @xmath82 , it provides an important independent check of possible new physics , due to the different nature of the possible new physics contributions in the two processes .
a comparison of the two results with the ckm fit and the imposed constrained in the @xmath83 plane of the ckm parameterization can be found in fig .
[ fig : deltamvsbtorho ] .
with the extraordinary luminosities of the b - factories belle and babar , the field of strange radiative penguin decays of the b meson has evolved into one of the most important areas of precision measurements in the search of new physics .
while it is still unclear whether @xmath2 transitions will be seen in the lifetime of the present b - factories , the rare decay @xmath41 has now been seen by belle and babar , and these measurements will possibly evolve towards precision measurements in the same way as the @xmath9 decays before . with the anticipated @xmath84 of luminosity per b - factory at the end of 2008 , a @xmath85 measurement of the cp asymmetry in @xmath77 should be feasible .
another interesting measurement would be the isospin asymmetry @xmath86 , which presents a completely new way of obtaining a measurement of the ckm - angle @xmath87 , as outlined in @xcite and shown in fig .
[ fig : gammaai ] .
the author would like to thank the pep - ii accelerator and its crew for the excellent luminosity delivered to babar , and the radiative penguin analysis group in babar for lots of support and very lively discussions .
99 r. ammar _ et al . _ [ cleo collaboration ] , phys .
* 71 * ( 1993 ) 674 .
see e.g. m. misiak _ et al .
_ , phys .
* 98 * ( 2007 ) 022002 and references therein .
b. aubert _ et al . _ [ babar collaboration ] , phys .
d * 73 * ( 2006 ) 092001 .
t. m. aliev and m. savci , phys .
d * 60 * , 014005 ( 1999 ) .
b. aubert _ et al . _ [ babar collaboration ] , arxiv : hep - ex/0703018 , submitted to prl
. k. abe _ et al . _ [ belle collaboration ] , phys .
* 96 * ( 2006 ) 221601 .
a. ali , e. lunghi and a. y. parkhomenko , phys .
b * 595 * , 323 ( 2004 ) .
p. ball and r. zwicky , jhep * 0604 * , 046 ( 2006 ) .
b. aubert _ et al . _ [ babar collaboration ] , phys .
* 98 * ( 2007 ) 151802 .
p. ball , g. w. jones and r. zwicky , phys .
d * 75 * ( 2007 ) 054004 .
a. abulencia _ et al . _
[ cdf collaboration ] , phys .
lett . * 97 * , 242003 ( 2006 ) . | this article gives an overview of the recent searches and measurements of @xmath0 penguin transitions with the babar experiment .
the branching fraction of these decays in the standard model ( sm ) is expected to be a factor of 10 or more lower than the corresponding @xmath1 penguin transitions , but a deviation from the sm prediction would be an equally striking sign of new physics .
the exclusive decay @xmath2 is searched by babar with no excess over the background found .
the babar measurement of @xmath3 provides the first evidence of @xmath4 , is in good agreement with the previous belle results and provides a measurement of @xmath5 independent of the one from @xmath6 mixing .
no deviation from the sm is found . |
brown dwarfs were predicted theoretically by kumar ( 1962 ) and hayashi & nakano ( 1963 ) .
they were first detected more than thirty years later ( rebolo et al .
1995 ; nakajima et al .
1995 ; oppenheimer et al .
since then hundreds have been observed , both in nearby young star clusters and in the field ( see luhman et al .
2007 and references therein ) .
@ll & meaning + star & pressure - supported object forming on a + & dynamical timescale , by gravitational instability .
+ planet & pressure - supported object forming on a much + & longer than dynamical timescale , by core + & accretion .
+ hb & hydrogen - burning star , @xmath2 .
+ bd & brown - dwarf star , @xmath3 .
+ pm & planetary - mass star , @xmath4 . + ur - star & star at the centre of the ur - disc ; present from + & the start of the simulation , and provides the + & background gravitational and radiation fields .
+ ur - disc & disc around the ur - star ; the material whose + & dynamics is to be simulated . + ur - system & ur - star plus ur - disc .
+ a distinction is often made between brown dwarfs below , and stars above , the hydrogen - burning limit at @xmath5 ( where @xmath6 is the mass of jupiter ) .
similarly , a distinction is often made between planetary - mass objects below , and brown dwarfs above , the deuterium - burning limit at @xmath7 . however , in the context of star formation , these distinctions are at best unhelpful .
we will therefore refer collectively to all objects forming by gravitational instability , on a dynamical timescale , as _
stars_. within this definition , we will label as hydrogen - burning stars ( hbs ) those stars with masses above the h - burning limit , as brown dwarf stars ( bds ) those stars with masses between the h - burning limit and the d - burning limit , and as planetary - mass stars ( pms ) those stars with masses below the d - burning limit .
this nomenclature is summarised in table [ tab : nomen ] . since there are only 3 pms formed in our simulations , we will , for statistical purposes , sometimes lump the pms in with the bds .
we note that for very low mass stars the dynamical timescale is essentially the freefall timescale at rather high density .
therefore it is very short , specifically @xmath8 where we have estimated the freefall time at the centre of a marginally unstable isothermal sphere , and @xmath9 is the isothermal sound speed for molecular gas at @xmath10 .
therefore the formation of pms , bds and low - mass hbs ( say below @xmath11 ) is extremely rapid ( @xmath12 ) .
in contrast to our definition of stars , we will refer to all objects which form by core accretion , on a much longer timescale ( @xmath13 ) , as planets . with these definitions
, we can discuss a single coherent stellar initial mass function ( imf ) which extends below the h- and d - burning limits .
this stellar imf is bounded by a minimum stellar mass , which is determined by the requirement that a star - forming condensation is able to radiate away , on a dynamical timescale , the @xmath14 work done by compression .
this requirement is usually called the opacity limit , but see masunaga & inutsuka ( 1999 ) . for contemporary star formation , in the solar vicinity
, the minimum stellar mass is estimated to lie in the range from @xmath15 ( rees 1976 ; low & lynden - bell 1976 ; boyd & whitworth 2005 ; whitworth & stamatellos 2006 ) . given this low minimum stellar mass , it is very likely that the stellar imf overlaps the planetary mass function in the interval between 1 and @xmath16 .
however , with the definitions above , the two mass functions are distinct in their origin . during the last decade
there has been much speculation about the origin of bds ( see review by whitworth et al .
currently there are three main theories proposed : ( i ) magneto - turbulent or gravo - turbulent fragmentation , ( ii ) premature ejection of protostellar embryos from their natal cores , and ( iii ) disc fragmentation .
padoan & nordlund ( 2004 ) suggest that magneto - turbulent fragmentation of molecular clouds produces prestellar cores of brown - dwarf mass with densities high enough for them to be gravitationally unstable and collapse .
their model follows the standard paradigm of low - mass star formation , i.e. a prestellar core collapses to form a single star .
the model does not attempt to explain the formation of bds in clusters and binaries . additionally , the highly co - ordinated supersonic velocities needed for these dense low - mass prestellar cores to form have yet to be observed
. gravo - turbulent fragmentation of collapsing prestellar cores may also produce bds .
simulations ( e.g. bate et al . 2003 ; goodwin et al .
2004a , b , 2006 ) of collapsing turbulent cores produce both hbs and bds , with numbers in good agreement with observation .
however , these simulations have been performed with a barotropic equation of state , and therefore the effects of radiative transfer ( e.g. boss et al .
2000 ; whitehouse & bate 2006 ; stamatellos et al .
2007a ) are not properly captured .
the ejection scenario is closely associated with the gravo - turbulent fragmentation mechanism . in numerical simulations of collapsing turbulent cores ,
bds frequently form when protostellar embryos ( i.e. very low - mass protostars which have just formed ) are ejected from their natal cores due to dynamical interactions with other protostars . as a consequence of ejection they stop accreting , and
so they never acquire sufficient mass to sustain hydrogen burning ( reipurth & clarke 2001 ) .
bds can also form by disc fragmentation ( e.g. whitworth & stamatellos 2006 ; stamatellos et al .
there are two conditions which must be met for a disc to fragment .
first , the disc must be massive enough so that locally gravity can overcome thermal and centrifugal support ( toomre 1964 ) , i.e. @xmath17 here @xmath18 is the toomre parameter , @xmath19 is the sound speed , @xmath20 is the epicyclic frequency , and @xmath21 is the surface density .
second , the disc must cool efficiently so that proto - condensations forming in the disc do not simply undergo an adiabatic bounce and dissolve . both theory and simulations ( gammie 2001 ; johnson & gammie 2003 ; rice et al .
2003 , 2005 ; mayer et al . 2004 ; mejia et al . 2005 ; stamatellos et al .
2007b ) indicate that for disc fragmentation to occur , the cooling time must satisfy @xmath22 here @xmath23 is the local orbital period , and @xmath24 the adiabatic exponent .
numerical simulations suggest that disc fragmentation produces bds , pms and also low - mass hbs ( stamatellos et al .
these objects may either remain bound to the primary star , or be ejected into the field .
the main concern with this model is whether massive extended discs actually form in the first place .
observations of such discs are difficult , since they are short lived and heavily embedded in their parental clouds .
we return to this issue below .
two other mechanisms that might form bds are the photo - erosion of massive prestellar cores by nearby ionising stars ( whitworth & zinnecker 2004 ) , and fragmentation during the second collapse , i.e. when h@xmath25 dissociates ( whitworth & stamatellos 2006 ) .
the former mechanism can only operate in the vicinity of a massive star , and therefore it can not be a major source of bds . the latter mechanism has been advocated on theoretical grounds , but it has not yet been shown to work by numerical simulations ( e.g. bate 1998 ; stamatellos et al .
2007a ) .
the above bd formation mechanisms are not exclusive of each other and can work in conjunction in star forming regions .
for example , gravo - turbulent fragmentation can generate cores which collapse to form discs , the discs may then fragment to produce bds , and the bds may avoid accreting additional mass by being ejected .
since all these mechanisms should produce both bds and low - mass hbs , we expect there to be a continuity in their statistical properties across the h - burning limit , and this is indeed what seems to be observed . like low - mass hbs , young bds have discs ( and all the consequences thereof , i.e. ir excesses , signatures of magnetospheric accretion , x - rays , outflows , etc . ) , they are sometimes found in binary and higher multiple systems , and they coexist with stars ( e.g. luhman et al .
2007 and references therein ) .
nonetheless , some authors ( e.g. thies & kroupa 2007 ) find observational evidence that bds form in a different way to hbs . in this regard
the most critical factor is the _ brown dwarf desert _
i.e. the lack of close bd companions to sun - like stars .
radial velocity surveys have revealed a large number of planets orbiting close to sun - like stars , as well as a large number of low - mass hbs , but very few bds ( marcy & butler 2000 ; see burgasser et a. 2007 for review ) . as we shall show below
, the brown dwarf desert finds a natural explanation in the mechanics of disc fragmentation and subsequent protostellar dynamics . in this paper
we explore the fragmentation of a disc having comparable mass to the central primary star .
in section 2 we discuss the numerical methods we use , and in section 3 we describe the initial disc configuration . in section 4
we present in detail one of the simulations performed , and then describe an ensemble of simulations . in section 5
we discuss the statistical properties of the stars produced by disc fragmentation and compare them with the observed properties of bds and low - mass hbs .
in section 6 we describe what observers might expect to see if they observe a fragmented disc after a few hundred years , and in section 7 we discuss the implications of the disc fragmentation mechanism for planet formation .
finally , in section 8 we summarise our results .
the evolution of the disc is initially followed using smoothed particle hydrodynamics ( sph ) , until @xmath26 of the disc mass has been accreted , either onto the stars condensing out of the disc , or onto the central primary star ; this typically takes 10 to 20 kyr .
then the residual gas is ignored and the long term dynamical evolution of the resulting ensemble of stars is followed up to 200 kyr , using an n - body code . for the hydrodynamics we use the sph code dragon ( e.g. goodwin et al .
2004a , b ) , which invokes an octal tree ( to compute gravity and find neighbours ) , adaptive smoothing lengths , multiple particle timesteps , and a second - order runge - kutta integration scheme .
the code uses time - dependent viscosity with parameters @xmath27 , @xmath28 ( morris & monaghan 1997 ) and a balsara switch ( balsara 1995 ) , so as to reduce artificial shear viscosity ( artymowicz & lubow 1994 ; lodato & rice 2004 ; rice , lodato , & armitage 2005 ) .
the smoothing lengths are adapted so that each particle has exactly @xmath29 neighbours ; this reduces numerical diffusion to a minimum ( attwood , goodwin & whitworth 2007 ) .
sinks are created wherever a bound condensation forms and its density exceeds @xmath30 .
sph particles are accreted onto a sink if they are within @xmath31 of the sink and bound to it .
sinks are identified with stars .
we note that our high value of @xmath32 makes the use of sinks relatively safe .
the energy equation and associated radiative transfer are treated with the method of stamatellos et al .
( 2007a ) .
this method takes into account compressive heating or expansive cooling , viscous heating , radiative heating by the background , and radiative cooling .
the method does not solve the radiation transfer equation ( which is unfeasible in hydrodynamic simulations with the current computing resources ) , but it is an appoximate method that performs well , in the optically - thin , intermediate and optically - thick regimes , and has been extensively tested ( stamatellos et al .
in particular it reproduces the detailed 3d results of masunaga & inutsuka ( 2000 ) , boss & bodenheimer ( 1979 ) , boss & myhill ( 1992 ) , whitehouse & bate ( 2006 ) , and also the analytic test of spiegel ( 1957 ) .
it also performs well in the analytic test of hubeny ( 1990 ) for disc geometries .
it is easy to implement in particle- and grid - based codes , and highly efficient ( incurring only a @xmath33 overhead , in comparison with simulations performed using a barotropic equation of state ) .
the gas is assumed to be a mixture of hydrogen and helium .
we use an equation of state ( black & bodenheimer 1975 ; masunaga et al .
1998 ; boley et al . 2007 ) that accounts for the rotational and vibrational degrees of freedom of molecular hydrogen , and for the different chemical states of hydrogen and helium .
we assume that ortho- and para - hydrogen are in equilibrium . for the dust and gas opacity we use the parameterization by bell & lin ( 1994 ) , @xmath34 , where @xmath35 , @xmath36 , @xmath37 are constants that depend on the species and the physical processes contributing to the opacity at each @xmath38 and @xmath39 .
the opacity changes due to ice mantle evaporation and the sublimation of dust are taken into account , along with the opacity contributions from molecules and h@xmath40 ions .
for the n - body part of the simulation , exploring the long term dynamical evolution of the stars formed by fragmentation of the outer disc ( from @xmath41 ) , we use a 4th - order hermite integration scheme ( makino & aarseth 1992 ) , with a conservative timestep criterion so that energy is conserved to better than one part in @xmath42 ( hubber & whitworth 2005 ) .
only a few extended massive discs have been observed around sun - like stars ( e.g. eisner et al .
2005 , 2008 ; eisner & carpenter 2006 ; rodriguez et al .
2005 ) . however , we suggest that this is because the outer parts of such discs are rapidly dissipated by gravitational fragmentation rather than because they seldom form in the first place .
indeed , it seems that the formation of such discs is inevitable .
for example , a @xmath43 prestellar core with ratio of rotational to gravitational energy @xmath44 will if it collapses monolithically forms a protostellar disc with outer radius @xmath45 . since the observations of goodman et al .
( 1993 ) indicate that many prestellar cores have @xmath46 , the formation of extended discs should be rather common .
alternatively , if an existing @xmath47 protostar attempts to assimilate matter with specific angular momentum @xmath48 , this matter is initially parked in an orbit at @xmath49 .
the quantum of specific angular momentum used to normalise this relation is rather modest by protostellar standards , as can be seen by expressing it in terms of a lever - arm in parsecs and a velocity in kilometres per second , viz .
@xmath50 .
previous simulations ( stamatellos et al .
2007b ) suggest that the outer parts of massive extended discs fragment on a dynamical timescale ( @xmath51 ) , so they are indeed very short - lived .
we shall assume a star - disc system ( hereafter the _ ur - system _ ) , in which the central primary star ( hereafter the _ ur - star _ ) has initial mass @xmath52 . initially the disc ( hereafter the _ ur - disc _ ) has mass @xmath53 , inner radius @xmath54 , outer radius @xmath55 , surface density @xmath56 temperature @xmath57 and hence approximately uniform initial toomre parameter @xmath58 .
thus the disc is at the outset marginally gravitationally unstable .
( the _ ur- _ prefix is used solely because it otherwise becomes difficult later on to discuss unambiguously the smaller discs that attend newly - formed bds , the orbital parameters of bd / bd binaries , and so forth ) .
we have designed a disc with @xmath59 over a wide range of radii , because we are wanting to establish which parts of the disc can , and which can not , fragment because they obey , or violate , the gammie criterion .
the radiation of the ur - star is taken into account by invoking a background blackbody radiation field with temperature @xmath60 . in effect
, this means that , if the material in the disc is heated by compression and/or viscous dissipation , it can only cool radiatively if it is warmer than @xmath60 given by eqn .
( [ eqn : tbg ] ) .
we perform twelve simulations : ten using 150,000 particles , one with 250,000 and one with 400,000 ( the last two to check convergence ) .
the minimum resolvable mass ( corresponding to 100 sph particles ) is @xmath61 , and therefore the jeans condition is obeyed at all times .
all of the simulations start with the same ur - system ( i.e. the same ur - disc and ur - star parameters ) .
the only difference ( apart from the number of particles in the last two simulations ) is the random seed used to construct each ur - disc .
thus the poisson density fluctuations are different in each simulation .
this allows us to study the statistical properties of the stars produced .
because the ur - discs are toomre unstable , spiral arms form and sweep up material into dense condensations . within a few thousand years , between 5 and 11 stars have condensed out of the disc . in all 12 simulations a total of 96 stars are formed .
3% are pms , 67% are bds , and the rest are low - mass hbs ( with masses up to @xmath62 ) .
( hereforth bd statistics will include pms , unless otherwise stated ) .
stars initially condense out of the ur - disc where @xmath59 and @xmath63 , i.e. typically at distances @xmath64 au from the central star .
hbs generally form closest to the ur - star , and pms form furthest from the ur - star .
there are three factors producing this trend of lower - mass stars forming further out in the ur - disc : ( i ) stars condensing out at small @xmath65 condense out earlier , so they then have longer to accrete ; ( ii ) the ambient density at small @xmath65 is higher , so there is more matter to accrete ; and ( iii ) the most massive stars tend to migrate inwards , so they get to mop up the inner parts of the ur - disc where no stars have condensed out . with between 5 and 11 stars formed in each ur - disc , initially between @xmath66 and @xmath67 , interactions between the stars are frequent . during 2-body interactions
, binary systems can be formed if there is sufficient dissipation ( e.g. due to the small accretion discs which normally attend the individual stars ) . in 3-body interactions ,
the lowest - mass star is often ejected into the field , and therefore stops accreting altogether .
@xmath68 of the stars formed end up being ejected into the field .
most of the stars below the h - burning limit , and all the stars below the d - burning limit are ejected into the field .
most are ejected as single stars , but some are ejected as components of binary systems .
these binary systems comprise bd / bd pairings , bd / hb pairings , and even one bd / pm pairing ; there are both wide and close systems . ejected
bds frequently retain accretion discs with masses on the order of a few jupiter masses .
hence they can sustain accretion and outflows as observed ( jayawardhana et al .
2003 ; muzerolle et al . 2005 ; mohanty et al . 2005 ; whelan et al . 2005 ) .
the final outcome typically has a low - mass hb orbiting the ur - star at close distance ( @xmath69 ) , a low - mass hb in a wider orbit , and a couple of bds or a bd / bd binary orbiting at large distance ( @xmath70 ) .
all the other stars have been ejected into the field . in fig .
[ fig : rdm1 ] we present column - density images of the ur - disc , every 500 yr , for one of the simulations . in this simulation
7 stars form : 4 bds and 3 low - mass hbs . at the end of the simulation ( i.e. after the hydrodynamic simulation _ and _ the n - body simulation , at @xmath71 ) ,
only 3 stars remain bound to the ur - star : a low - mass hb orbiting close to the ur - star ( @xmath72 ) , and a binary comprising a bd and a low - mass hb , orbiting at @xmath73 . fig .
[ fig : rdm1.qtoomcool ] shows the toomre parameter @xmath18 and the ratio of the cooling time to the orbital time @xmath74 ( both azimuthally averaged ) , every 1 kyr from 0 to 6 kyr . initially the ur - disc is marginally unstable ( blue line ) , and outside @xmath75 the cooling time is favorable for fragmentation ( i.e. @xmath76 ) .
hence the ur - disc fragments here . as time progresses the region where @xmath59 moves farther away from the ur - star as the gas in the inner regions is used up . and
the ratio of the cooling time to the orbital period @xmath74 ( both azimuthaly averaged ) are plotted against distance from the ur - star , every 1 kyr from 0 to @xmath77 ( blue , red , cyan , magenta , black , green).,height=415 ] in the two simulations with higher resolution , the growth of gravitational instabilities , and the properties of the stars formed as a result of these instabilities , follow the same patterns as in the simulations with lower resolution .
the variance is the same as between two simulations performed with the same resolution , and is attributable to the different seed noise in the different simulations ( due to random positioning of the particles ) and the chaotic , non - linear nature of gravitational instability .
thus , the simulations appear to be converged , in a statistical sense . in the following sections we describe in detail the results of the simulations , focusing on the statistical properties of the stars formed , and compare them with the observed properties of bds and low - mass hbs .
the ur - discs start fragmenting after @xmath78 , which corresponds to about one quarter of an outer rotation period ( @xmath79 ) ( see fig . [
fig : tformation ] ) .
stars condense out on a dynamical timescale .
hence star formation occurs faster in the inner ur - disc , but even at the edge of the ur - disc it is virtually over by @xmath80 .
the stars which form first , in the inner ur - disc , have more time to accrete , and more material to accrete , and hence they tend to end up as hbs , with masses up to @xmath81 ( see fig . [
fig : tformation ] ) .
lower mass stars ( i.e. bds and pms ) tend to form at later times and at larger distances from the ur - star . the mass distribution of the objects produced by disc fragmentation is shown in fig .
[ fig : mspectrum ] .
@xmath82 of the stars produced have masses below the h - burning limit , and @xmath83 of these have masses below the d - burning limit . the mass distribution peaks around @xmath84 .
it decreases smoothly towards higher masses , roughly as @xmath85 , with @xmath86 .
there is no significant discontinuity across the h - burning limit at @xmath87 .
the mass distribution drops precipitously towards lower masses , reflecting that the minimum mass for star formation is @xmath88 .
the small number of stars that end up below the d - burning limit are the ones that are ejected from the ur - disc by a 3-body interaction , almost as soon as they have formed .
we emphasise that the mass function presented in fig.[fig : mspectrum ] represents stars formed by fragmentation of a particular ur - system ( i.e. an ur - disc with mass @xmath89 , radius @xmath90 , etc . in orbit
round an ur - star with mass @xmath91 ) .
therefore it can not be compared meaningfully with the overall stellar imf .
rather it will be necessary first to repeat the numerical experiment reported here for many different ur - systems ( i.e. different combinations of @xmath92 , @xmath93 , @xmath94 , etc . ) , and then to convolve the results with the appropriate distributions of @xmath92 , @xmath93 and @xmath94 , before comparing with the observed stellar imf .
a monte carlo experiment addressing this aspect of the problem is in progress ( attwood et al .
, in prep . )
[ fig : rspectrum ] shows the radial distribution of the stars formed by disc fragmentation , i.e. the distance @xmath95 from the ur - star .
radial distributions are shown for all stars , and then separately for hbs and bds , ( i ) at the moment of formation ( @xmath96 ; left ) , ( ii ) at the end of the sph simulation ( @xmath97 ; centre ) , and ( iii ) at the end of the n - body simulation ( @xmath98 ; right ) .
stars are categorised as hbs or bds on the basis of their final masses ; here the bd category includes the pms .
the most likely location at which stars form is between @xmath99 from the ur - star ( fig .
[ fig : rspectrum ] ; left ) .
this is the combined effect of the two criteria for disc fragmentation discussed in section [ sec : intro ] .
close to the central primary star ( @xmath100 ) , there is ample mass for fragmentation , so the toomre criterion is satisfied ( @xmath101 ) .
however , the cooling time is high compared with the orbital period , and hence stars can not condense out fast enough and they are quickly sheared apart .
this minimum radius for fragmentation agrees well both with analytical predictions ( rafikov 2005 ; whitworth & stamatellos 2006 ) , and with previous numerical simulations ( see also stamatellos et al . 2007b , stamatellos & whitworth 2008 ) .
conversely , far from the ur - star ( @xmath102 ) , the cooling time is quite low , but there is only just enough mass for the disc to be toomre unstable ( @xmath103 ) , and so fragmentation is sluggish . at locations between @xmath104 the conditions are just right for vigorous fragmentation to occur . in the lower panels of fig .
[ fig : rspectrum ] ( left ) , we show separately the formation locations of hbs and bds .
stars forming close to the ur - star generally accrete more gas and so their masses tend to increase above the h - burning limit .
the initial radial distribution quickly changes due to dynamical interactions between the newly - formed stars . in each disc between 5 and 11 stars
form between 60 and 400 au .
therefore the probability for interactions is high . in fig .
[ fig : rspectrum ] ( centre ) we present the radial distribution of the stars at the end of the sph simulation at @xmath97 .
the hbs are mostly within 100 au of the ur - star , whereas the bds are mostly outside 100 au .
the radial distribution has changed even more by the end of the @xmath105-body evolution , after @xmath98 ( fig . [ fig : rspectrum ] ; right ) .
most hbs have remained bound to the ur - star ( half at close distances , @xmath106 and the rest on wider orbits of a few hundred au ) . on the other hand ,
most of the bds have been ejected , and those that remain bound to the ur - star have wide orbits of a few hundred au .
this is also seen on fig .
[ fig : rmasslg ] , where the object final mass is plotted against the radius at formation ( top ) , the radius at the end of the sph simulation ( @xmath107 ; middle ) , and the radius at the end of the n - body simulation ( @xmath98 ; bottom ) . both hbs and bds form over a wide range of radii , but after the system has evolved for 200 kyr there are 13 hbs orbiting within 60 au of the ur - star , and only 1 bd . the hbs , being more massive , tend to form at smaller radii , _ and _ during dynamical 3-body interactions the more massive stars tend to become more tightly bound .
this phenomenology provides a natural explanation for the brown dwarf desert , i.e. the lack of bds as close companions to sun - like stars ( @xmath108 ; marcy & butler 2000 ) in contrast with the relatively high frequency of both hbs and planets at these radii .
first , stars with masses below the h - burning limit only condense out of the ur - disc at large radii ( @xmath109 ) as predicted analytically by whitworth & stamatellos ( 2006 ) .
second , such stars are only likely to stay below the h - burning limit if they remain at large radius or are ejected from the ur - disc .
third , the interactions which occur between stars formed in the ur - disc tend to scatter the more massive ones inwards and the less massive ones outwards . consequently the more massive stars tend to end up in the inner parts of the ur - disc where there is still gas to be accreted , and hence they tend to grow even more massive until they exceed the h - burning limit . at the end of our simulations , only one of the 12 ur - stars has a close ( @xmath110 au ) brown - dwarf companion , whereas there are 12 low - mass hbs in the same radial interval .
hence , it is @xmath111 times more likely for the close companion of a sun - like star to be an hb , rather than a bd .
this is comparable with observations , which suggest that @xmath112 of sun - like stars have bd companions within @xmath113 ( e.g. marcy & butler 2000 ; udry et al .
2003 ) , or @xmath114 within @xmath115 ( e.g in the hyades cluster , guenther et al .
this fraction rises to @xmath116 for hb companions ( duquennoy & mayor 1991 ) , i.e. hb companions are between 7 and 25 times more frequent than bd companions .
kyr ( middle ) , and the radius at @xmath117 kyr , are plotted against the final stellar mass .
hbs are scattered towards the ur - star , whereas bds are ejected or migrate towards wider orbits .
hence , the region close to the ur - star is exclusively populated by low - mass hbs .
the blue triangles correspond to objects that are ultimately not bound to the ur - star ( ejected or about to be ejected from the system).,scaledwidth=50.0% ] in fact , at the end of our simulations only 4 out of 24 ( @xmath118 ) of the stars within @xmath67 of the ur - star are bds , suggesting that the brown dwarf desert extends to a wider region around sun - like stars .
this is consistent with the results of the gemini deep planet survey that found a very low probability of bds ( with @xmath119 @xmath120 in the region 25 - 250 au ( @xmath121 1.9% ; lafrenire et al .
hence , systems like 1rxs j160929.1 - 210524 , a pm with mass @xmath122 m@xmath123 orbiting a star with @xmath124 m@xmath125 at @xmath126 au ( lafrenire et al . 2008 ) , can be explained with this model ; however , such systems should be rare .
in contrast , bds dominate the region from @xmath127 to @xmath128 ; 19 out of 22 ( @xmath129 ) of the stars at these large radii are bds .
gizis et al .
( 2001 ) report that at separations @xmath130 sun - like stars have comparable numbers of brown dwarf and m dwarf companions . once again
, we should not necessarily expect a detailed correspondence with the observations , because we have only studied a single ur - system ( i.e. a single combination of @xmath92 , @xmath93 , @xmath94 , etc . ) . ;
top ) , at @xmath131 kyr ( middle ) and at @xmath117 kyr ( bottom ) .
the stars form very close to the ur - disc midplane ( within @xmath132 ) but due to 3-body interactions many of them acquire inclined orbits .
coplanarity is not a characteristic of disc fragmentation.,height=415 ] initially ( at formation time , @xmath96 ) the orbits of the stars formed in the ur - disc are almost coplanar with the ur - disc ; their orbital planes all lie within @xmath133 of the ur - disc plane ( fig .
[ fig : anglespec ] , top ) .
however , due to 3-body interactions the stars start to acquire inclined orbits , . by @xmath134
only @xmath135 of the orbital planes lie within @xmath133 of the ur - disc plane , and by @xmath136 this percentage has dropped to @xmath137 ( fig .
[ fig : anglespec ] , middle and bottom ) .
we expect that in a realistic asymmetric disc , still accreting from its natal cloud , these percentages will be even smaller .
the inclination of the orbital plane does not appear to depend on the mass of the star , or its distance from the ur - star .
consequently , coplanarity in observed multiple systems , or the absence of it , can not confirm or rule out formation by disc fragmentation .
when stars condense out of the ur - disc , they are normally attended by their own individual accretion discs , but some of these individual discs are subsequently stripped by dynamical interactions . at the end of the sph simulation ( @xmath97 ) , @xmath135 of the stars still have individual accretion discs .
in addition , some of the binary systems have circumbinary discs .
these percentages are comparable with , but somewhat higher than , those reported in the literature .
jayawardhana et al .
( 2003 ) found a disc fraction of @xmath138 in a mixed sample from chamaeleon i , ic 348 , taurus , and u sco , using @xmath139 photometry .
luhman et al .
( 2005c ) , using spitzer nir observations , found a disc fraction of @xmath140 for ic348 and @xmath141 for chamaeleon i. however , these observed statistics correspond to stars in the fields of these clusters ( i.e. stars which , if they have been formed by fragmentation of an ur - disc , have since been ejected ) and so we expect their disc frequency to be somewhat reduced .
of the 13 systems ( 11 singles and 2 binaries ) ejected by the end of our sph simulations at @xmath97 , only one of the singles , but both the binaries , retain discs .
thereafter , the simulation switches to an @xmath105-body code , and so we can not say anything quantitative about how the disc statistics subsequently evolve .
however , since there are many stars with discs orbiting at very large distances from the ur - star for example , at the end of the sph simulations , there are 16 systems , including 2 binaries , orbiting at @xmath142 we should expect many of these to be liberated from the ur - star by the tides of passing stars , and to retain their discs ( e.g. whitworth & stamatellos 2006 ; goodwin & whitworth 2007 ) .
thus we infer that young bds in the field should have a significant disc fraction , and that young bds orbiting sun - like stars at large radius should have an even higher disc fraction .
the masses and radii of individual discs are plotted against the masses of their host stars on fig .
[ fig : mrdiscmstar ] .
the disc masses are generally @xmath143 , but there are a couple of discs with higher masses ( a few tens of m@xmath123 )
. there seems to be no correlation between the disc masses and the star masses . the disc radii range up to @xmath75 , but most of them are @xmath144 .
again there seems to be no correlation between disc radius and star mass .
the individual discs around bds and hbs are statistically indistinguishable .
the distributions of mass and radius for discs around bds are shown in fig .
[ fig : mrdiscspecbd ] . most bd discs ( @xmath145 )
have masses below @xmath146 , but the mass can be as high as @xmath147 ( fig .
[ fig : mrdiscspecbd ] , top ) .
most bd discs have radii @xmath148 , but the radius can be as high as @xmath75 ( fig .
[ fig : mrdiscspecbd ] , bottom ) . these disc masses and radii are consistent with observations , which show discs to be very common around bds ( klein et al .
2003 ; luhman 2004 ; luhman et al .
2005a ; scholz et al .
2006 ; guieu et al .
2007 ; riaz & gizis 2007 , 2008 ) , even around very low - mass ones ( e.g. luhman et al .
the bd discs are massive enough to make the formation of rocky planets with masses up to 5 m@xmath149 around bds a plausible scenario , at least around some of them ( payne & lodato 2007 ) .
@cccccc @xmath150m@xmath123 & @xmath151m@xmath123 & q & @xmath152/au & @xmath153 & @xmath36/au + 96 & 86 & 0.90 & 0.3 & 0.4 & 20 + 98 & 89 & 0.91 & 1.4 & 0.5 & 140 + 176 & 82 & 0.47 & 1.5 & 0.8 & 230 + 88 & 83 & 0.94 & 0.3 & 0.7 & - + 89 & 59 & 0.66 & 1.0 & 0.9 & 7700 + 102 & 44 & 0.43 & 0.3 & 0.8 & - + 109 & 35 & 0.33 & 0.6 & 0.8 & 800 + 105 & 73 & 0.70 & 0.5 & 0.3 & 200 + 50 & 46 & 0.92 & 0.6 & 0.5 & 1500 + 28 & 24 & 0.86 & 235 & 0.7 & 4000 + 62 & 59 & 0.95 & 0.6 & 0.9 & - + 73 & 43 & 0.59 & 1.3 & 0.7 & 1350 + 53 & 11 & 0.21 & 112 & 0.6 & - + here we discuss the statistical properties of the binary systems that exist at the end of the simulations .
we confine the discussion to binaries in which both components are stars that have condensed out of the disc , and we refer to these as low - mass binaries .
there are also binary systems and higher multiples in which one of the components is the ur - star ; we do not discuss these here .
our simulations produce 13 low - mass binaries , of which 4 comprise two low - mass hbs , 4 comprise a low - mass hb and a bd , 4 comprise two bds , and one comprises a bd and a pm .
the main properties of these binaries are recorded in table [ tab : binaries ] . in total ,
@xmath154 of the stars that form end up in binary systems , corresponding to a binary fraction of @xmath155 .
this is comparable with the low - mass binary fraction in taurus - auriga ( @xmath156 , kraus et al .
2006 ) , chamaeleon i ( @xmath157 , ahmic et al .
2007 ) , and the field ( e.g. @xmath158 , gizis et al . 2003 ) .
however , these are optical surveys and they are not able to detect tight binaries with separation of a few au . if we consider only the 2 wide binaries then the binary fraction we estimate is @xmath159 , which is smaller than observed .
very tight binaries can be probed by radial velocity surveys ( joergens 2006a ; kurosawa et al .
2006 ; maxted et al . 2008 ; joergens 2008 ) .
these surveys find a low - mass tight - binary fraction of @xmath160 which is consistent with our model .
11 of the 13 low - mass binaries have semi - major axes @xmath161 .
thus we predict that close low - mass binaries should outnumber wide ones , and this seems to be what is observed ( e.g. burgasser et al . 2007 ) . in fig .
[ fig : vlmbinaries ] ( top ) we plot the projected separation distribution for the very low - mass binaries from the http://www.vlmbinaries.org dataset . in this dataset
only @xmath162 of the binaries have separations @xmath163 au , which is similar to the predictions of our model ( 2/13 binaries , i.e. 15% ) .
we note that in table [ tab : binaries ] we quote the semi - major axis of the binary , whereas fig .
[ fig : vlmbinaries ] refers to the projected separation , which for eccentric binaries is most often larger than semi - major axis ) .
the two wide low - mass binaries with @xmath164 are the systems with the lowest total mass . 4 low - mass binaries , including one of the very wide systems , have become unbound from the ur - star .
therefore the ejection mechanism does not militate against delivering wide low - mass binaries to the field .
one of the unbound low - mass binaries comprises a bd and a pm ; it is therefore quite similar to 2mass 1207 - 3932 ( chauvin et al .
2004 , 2005 ) . amongst the 9 low - mass binaries that remain bound to the ur - star , the semi - major axis @xmath36 of the orbit around the ur - star tends to increase as the total mass of the low - mass binary decreases .
thus most of the low - mass binaries with an hb primary orbit the ur - star within @xmath165 , whereas most of the low - mass binaries with a bd primary orbit the ur - star outside @xmath165 .
all of the low - mass binaries have high eccentricities , as a result of the dynamical interactions which form them , and/or subsequent dynamical interactions with other stars that have condensed out of the ur - disc .
the high eccentricities of the wide binaries should persist , but the high eccentricities of the closer binaries may be damped somewhat by tidal interaction between their attendant discs ; these tidal interactions are not accounted for in our n - body simulations .
most of the low - mass binaries ( @xmath68 ) have components with similar masses ( @xmath166 ) , in agreement with the observed properties of low - mass binaries ( e.g. burgasser et al . 2007
; see fig .
[ fig : vlmbinaries ] ) . 25 ( out of 67 , i.e. 37% ) of the bds formed in the ur - disc remain bound to the ur - star . the remaining 42 ( out of 67 , i.e. 63% ) are ejected .
of the 25 bds that remain bound to the ur - star , 10 ( @xmath167 ) are in low - mass binaries .
in contrast , of the 42 bds that become unbound from the ur - star , only 5 ( @xmath168 ) are in low - mass binaries ( one with an hb primary ) .
hence , bds that are companions to sun - like stars are more likely to be in binaries ( binary frequency 25% ) than bds in the field ( binary frequency @xmath169 ) .
this trend is comparable to what is observed , although the observed binary frequencies are somewhat higher .
burgasser et al .
( 2005 ) report a binary fraction of @xmath170 for bd companions to sun - like stars , and a binary fraction of only @xmath171 for bds in the field . .
the first plot shows the accretion rate onto the ur - star ( black line ) and the accretion rate onto a star which is ejected ( red line ) .
the second plot shows the accretion rates onto two stars which eventually become a binary with a circumbinary disc .
stars which form earlier and/or closer to the ur - star ( black lines on the last 3 plots ) tend to accrete more than stars which form later and/or further away from the ur - star ( red lines on the last three plots ) .
the accretion rates shows signs of periodic modulation ( due to passing close to the ur - star and/or accretion from a circumbinary disc ) and non - periodic modulation ( when a star passes through a spiral arm or another star s disc).,height=415 ] newly formed stars continue to accrete material from the ur - disc .
stars forming closer to the ur - star tend to accrete more material than stars forming further away , firstly because there is more material in the inner parts of the ur - disc to accrete ( particularly in the innermost part of the ur - disc , which is unable to fragment ) , and secondly because the more massive stars tend to migrate inwards . in our simulations , typical accretion rates are @xmath172 , which is high compared with the observed accretion rates onto brown dwarfs ( typically @xmath173 ; e.g. natta et al .
2004 ; muzerolle et al . 2005 ; mohanty et al . 2005 ; herczeg & hillenbrand 2008 ) .
however , the high accretion rates we record refer to an early , short - lived phase ( @xmath174 ) , and they are expected to decline significantly at later times .
the accretion rates often exhibit periodic modulations , during which the rate may increase by more than an order of magnitude .
these modulations are partly due to high orbital eccentricity , and correspond to periastron passages within @xmath175 , occuring every few thousand years .
there are also smaller modulations involving the components of close binaries as material is accreted onto them from a circumbinary disc .
finally , there are occasional episodic events when the accretion rate increases by up to an order of magnitude as a star passes through a spiral arm in the ur - disc or through the disc of another star . in fig .
[ fig : rdm1.accretion ] we plot for the simulation presented in fig [ fig : rdm1 ] the accretion rates onto the ur - star and onto the 7 stars formed from the ur - disc .
these plots are representative of the accretion rates onto stars in the other simulations . on the first plot
, we present the accretion rate onto the ur - star ( black line ) .
the accretion rate initially increases , as the inner gap is filled in , and then remains approximately constant around @xmath176 m@xmath177 .
the other star presented on the first plot ( red line ) is one which is ejected and so accretion is effectively terminated .
there is actually a very low persistent accretion rate onto this star , because it retains a small accretion disc , but the rate is well below @xmath178 . on the second plot we present the accretion rates onto two stars which form at different times but eventually become a binary system with a circumbinary disc .
the star which forms later , and at larger distance from the ur - star , initially accretes less material than the star which forms earlier , and closer to the ur - star . after the binary forms , the accretion rates onto the two stars become very similar . on the last two plots ,
the accretion rates onto the other four stars are presented .
again it can be seen that the stars which form later tend to accrete less than the stars which form earlier .
in some instances the difference exceeds an order of magnitude .
the stars which experience persistently higher accrertion rates tend to be the ones that form , and/or end up , closer to the ur - star .
the ur - discs dissipate on a dynamical timescale , i.e. within a few thousand years . in fig .
[ fig : endstardisc ] the mass of the remnant ur - disc is plotted against its radius . by this stage ,
most of the ur - discs have masses @xmath179 , although one has @xmath180 .
the ur - disc radii are all @xmath181 .
these ur - disc masses and radii are typical of the discs observed around sun - like stars ( e.g. mundy et al . 1995 ; bally et al . 1998 ; eisner et al . 2005 , 2008 ; williams et al . 2005 ; eisner & carpenter 2006 ) . in our simulations
the ur - star ends up with one or two h - burning companions and one or two brown - dwarf companions .
the masses of the companions are almost equally distributed across the low - mass regime ( see fig . [
fig : mspecbound ] ) .
one of the companions is in a relativel close orbit ( @xmath182 ) , and this one is almost always an h - burning star .
the other two or three companions orbit much further out ( @xmath64 ) , and often two of them are in a binary system .
the eccentricities of these wide companions are high , due to dynamical interactions with other stars ( fig .
[ fig : nbfinal.ecc ] ) .
thus , although their semi - major axes are greater than @xmath183 , due to their high eccentricity they pass within @xmath184 of the ur - star with a periodicity between @xmath185 , disrupting the remaining ur - disc and/or the orbits of any planets formed from the inner ur - disc ( see sect . [
sec : planets ] ) .
more than half ( @xmath68 ) of all the stars are ejected into the field .
9 ( out of 29 , i.e. @xmath186 ) of the hbs are ejected , and 45 ( out of 67 , i.e. @xmath187 ) of the bds . the mass distribution of the ejected stars is shown in fig .
[ fig : mspec.ejected ] .
most of the objects ejected are low - mass bds ; there are also some pms .
4 binaries are ejected , of which one comprises two hbs , one comprises an hb and a bd , one comprises two bds , and one comprises a bd and a pm .
the velocity distribution of the ejected stars is shown in fig .
[ fig : ejectedvel ] ( top ) .
most stars are ejected with low velocities ( @xmath188 ) .
hbs tend to be ejected with somewhat lower velocities , but otherwise there is no strong dependence of the ejection velocity on the mass of the star ( fig .
[ fig : ejectedvel ] , bottom ) .
a small percentage of the ejected stars in our simulations have very high velocities ( cf .
umbreit et al .
2005 ) , and the lack of such high velocities in observational surveys of bd kinematics , and the apparent absence of a diaspora of brown dwarfs around young star clusters , has often been used as an argument against the ejection mechanism . however , in our simulations both bds and hbs are ejected , and hence they should be roughly co - extensive ( cf .
luhman 2004 ; goodwin et al .
2005 ; joergens 2006b )
. moreover the ejection velocities would be significantly lower if the ejection were treated fully with hydrodynamics , rather than invoking sinks and then switching to an n - body code .
[ fig : warped ]
there are four types of system formed in our simulations that are rare but nevertheless have similarities with intriguing observed systems .
free floating planetary - mass stars
i.e. single field stars with masses @xmath189 have been observed in orion ( lucas & roche 2000 ) , in the @xmath190 orionis cluster ( zapatero osorio et al .
2000 ) and in ic 348 ( najita et al . 2000 ) .
the simulations presented here show that such low - mass stars can form by disc fragmentation , and this is confirmed analytically by whitworth & stamatellos ( 2006 ) .
it is uncertain whether they can also form from the collapse of low - mass cores ( e.g. greaves et al . 2005 ) . 3 pms form in our simulations , and all three of them are ejected into the field .
their masses remain below the d - burning limit because they are ejected from the disc very soon after formation . according to our simulations
bds should outnumber pms by a factor of @xmath191 . allowing for the fact that 35% of the bds remain bound to the ur - star but all pms
are ejected , bds should outnumber pms in the field by a factor of @xmath192 .
however , we re - iterate that these numbers are for a very specific ur - system .
one of the 13 low - mass binaries formed in our simulations has a @xmath193 bd primary and an @xmath194 pm secondary .
this system has a semi - major axis of @xmath195 and an eccentricity of 0.6 .
the two components of the system formed independently by disc fragmentation , paired up in the disc , and then were ejected as a binary into the field .
this system is qualitatively similar to 2mass 1207 - 3932 ( chauvin et al .
2004 , 2005 ) , a system with a @xmath196 bd primary orbited by a @xmath197 pm secondary , with a projected separation of @xmath198 .
pms are unlikely to be able to form by core accretion in bd discs ( lodato et al . 2005 ; payne & lodato 2007 ) .
lodato et al .
( 2005 ) suggest that 2mass 1207 - 3932b may have formed by fragmentation of a disc around 2mass 1207 - 3932a .
we suggest that such systems form by pairing up of two components that form indepedently in a fragmenting disc , capture one another , and then are ejected into the field as a binary .
four of the 58 circumstellar or circumbinary discs formed in our simulations are poorly aligned with the ur - disc ( fig .
[ fig : nonaligned ] ) .
this is because ( a ) stars do not form exactly on the midplane of the ur - disc , and ( b ) subsequently they experience impulsive perturbations due to passing stars that have formed in the disc .
we presume that poorly aligned discs are even more common in real systems , i.e. asymmetric ur - discs forming in collapsing , fragmenting cores with lumpy material continuing to infall onto the ur - disc .
the existence of poorly aligned discs is inferred from polarimetric observations ( jensen et al .
2004 ; monin et al .
2006 ) and from non - parallel jets ( e.g. davis et al . 1994 ) .
one of the discs formed in our simulations is noticeably warped ( see fig . [
fig : warped ] ) .
the formation of such discs is a rare event , requiring a violent interaction between a star which is about to be ejected and a dense spiral arm .
warped discs have been observed ( e.g. heap et al .
2000 ; quillen 2006 ) .
our simulations show that disc fragmentation can produce pms , but the probability is low ; for every pm formed , there are more than 30 bds and/or hbs formed . the pms form at large distances from the ur - star , and are then ejected before they have time to grow by accretion .
if they were not ejected , they would grow in mass by accretion , and end up above the d - burning limit .
thus , it is very hard , if not impossible , for planetary - mass stars to form in the outer reaches of massive extended ur - discs and then migrate inwards to a tight orbit around the ur - star .
fragmentation is also unlikely to happen close to the ur - star ( e.g. matzner & levin 2005 ; rafikov 2005 ; whitworth & stamatellos 2006 ; boley et al .
2006 ; durisen et al .
2007 ; stamatellos et al . 2007b ;
stamatellos & whitworth 2008 ; cai et al .
. therefore exoplanets ( e.g. udry & santos 2007 ; _ extrasolar planets encyclopaedia _ at http://exoplanet.eu ) probably have to form by the core accretion mechanism ( e.g. safronov 1969 ; goldreich & ward 1973 ; pollack et al .
1996 ) , at later stages during disc evolution .
in fact even this may be rather hard in the ur - discs we have simulated here , although not necessarily in other discs .
this is because each ur - disc we have simulated ends up with a low - mass star ( usually an hb ) on a close ( @xmath199 ) eccentric orbit , and this is likely to inhibit planet formation in the inner disc ( thbault et al .
the fragmentation of massive , extended ur - discs produces stars which populate the low - mass end of the stellar initial mass function : low - mass h - burning stars ( hbs ) , brown - dwarf stars ( bds ) and planetary - mass stars ( pms ) .
despite the fact that we have considered only a single ur - system , the predictions of the model compare well with observation .
simulations of different ur - systems ( different ur - disc masses , radii , surface - density profiles , etc . and different ur - star masses ) yield similar results ( stamatellos et al .
, in prep ) , as do simulations in which we follow the formation of the ur - system from a collapsing turbulent prestellar core ( attwood et al . , in prep ) . for the particular ur - system
that we simulate ( a @xmath47 @xmath200 disc in orbit about an @xmath47 star ) , we predict the following . *
each ur - disc fragments to form between 5 and 11 stars ; 30% are hbs , 67% are bds and 3% are pms . * 55% of the stars
are ejected into the field ( 30% of the hbs , 65% of the bds and 100% of the pms ) . *
the ur - star ends up with a close companion ( usually an hb ) and two or three wide companions ( typically an hb and one or two bds ) .
* these wide companions may include a low - mass binary system .
* within a few thousand years the ur - disc is reduced to @xmath179 and @xmath201 .
* we hypothesise that there is a range of masses ( say between @xmath202 ) where _ proceeding to lower masses _ the balance gradually shifts _ from _ the majority of stars forming by turbulent fragmentation ( as the ur - star presumably did ) , _ to _ the majority of stars forming as the result of fragmentation of discs around such stars ( like the ur - disc ) .
* only a small fraction ( say @xmath203 ) of the higher - mass sun - like stars ( g , k and early m dwarfs ) need to support disc fragmentation of the type we have modelled here to supply most of the low - mass hbs , bds and pms that are observed . * most stars
initially condense out of the ur - disc between @xmath184 and @xmath204 . *
the stars forming closer to the ur - star tend to grow to become hbs , and may migrate inwards to much closer final orbits . *
the stars forming further out tend to migrate outwards and/or be ejected .
* as a result there is a brown dwarf desert , i.e. a lack of brown dwarfs orbiting close to the ur - star ; those brown dwarfs which are not ejected tend to end up orbiting out beyond @xmath204 . * due to dynamical interactions between stars , stellar orbits can end up quite eccentric and strongly inclined to the plane of the ur - disc .
* the majority of stars formed from the ur - disc have their own accretion discs , and some of these discs are retained during ejection ; however , we expect bds that remain bound to the ur - star to have a higher disc frequency than those which are ejected .
* stars formed in the ur - disc frequently pair up to form low - mass binary systems .
the low - mass binary frequency is @xmath205 . *
these binary systems are usually close , but can be wide ; they tend to have eccentric orbits and large mass ratios ( i.e. components of comparable mass ) ; they include hb / hb pairs , hb / bd pairs , bd / bd pairs and bd / pm pairs . * these binary systems can survive ejection ; however , a bd that remains bound to the ur - star is three to five times more likely to be in a bd / bd binary system than a bd in the field . *
ejection velocities are @xmath206 , but these would probably be reduced if the hydrodynamics of ejection were modelled properly ( i.e. not invoking sinks and not switching to an n - body code ) .
* hot jupiters are very unlikely to form by disc fragmentation . moreover in discs which fragment like the ones modelled here , formation of hot jupiters by core accretion may also be inhibited by the stellar companions that form on close eccentric orbits . we conclude that disc fragmentation is a robust mechanism for the formation of brown dwarf stars , as well as planetary - mass stars and low - mass h - burning stars .
it explains successfully properties that are not satisfactorily explained by other formation mechanisms , for example the brown dwarf desert and the observed statistics of low mass binary systems .
we suggest that a large proportion of brown dwarf stars and planetary - mass stars may be formed by disc fragmentation
we would like to thank the referee for his helpful comments .
the computations reported here were performed using the uk astrophysical fluids facility ( ukaff ) .
n - body simulations were done by a code originally developed by d. hubber , who we thank for the support provided .
we also thank ken rice , aaron boley , & viki joergens for useful discussions / suggestions .
we also gratefully acknowledge the support of an stfc rolling grant ( pp / e000967/1 ) and a marie curie research training network ( mrtn - ct2006 - 035890 ) . | we suggest that a high proportion of brown dwarf stars are formed by gravitational fragmentation of massive extended discs around sun - like primary stars .
we argue that such discs should arise frequently , but should be observed infrequently , precisely because they fragment rapidly . by performing an ensemble of radiation - hydrodynamic simulations ,
we show that such discs typically fragment within a few thousand years , and produce mainly brown dwarf ( bd ) stars , but also planetary - mass ( pm ) stars and very low - mass hydrogen - burning ( hb ) stars .
subsequently most of the lower - mass stars ( i.e. the pm and bd stars ) are ejected by mutual interactions .
we analyse the statistical properties of these stars , and compare them with observations .
after a few hundred thousand years the sun - like primary is typically left with a close low - mass hb companion , and two much wider companions : a low - mass hb star and a bd star , or a bd - bd binary .
the orbits of these companions are highly eccentric , and not necessarily coplanar , either with one another , or with the original disc .
there is a bd desert extending out to at least @xmath0 ; this is because bds tend to be formed further out than low - mass hb stars , and then they tend to be scattered even further out , or even into the field .
bds form with discs of a few jupiter masses and radii of a few tens of au , and they are more likely to retain these discs if they remain bound to the primary star .
binaries form by pairing of the newly - formed stars in the disc , giving a low - mass binary fraction of @xmath1 .
these binaries include close and wide bd / bd binaries and bd / pm binaries .
binaries can be ejected into the field and survive , even if they have quite wide separations .
bds that remain as companions to sun - like stars are more likely to be in bd / bd binaries than are bds ejected into the field .
the presence of close and distant companions around sun - like stars may inhibit planet formation by core accretion .
we conclude that disc fragmentation is a robust mechanism for bd formation . even if only a small fraction of sun - like stars host the required massive extended discs , this mechanism can produce all the pm stars observed , most of the bd stars , and a significant proportion of the very low - mass hb stars .
[ firstpage ] stars : formation stars : low - mass , brown dwarfs accretion , accretion disks methods : numerical , radiative transfer , hydrodynamics |
this paper is an enlarged version of the three lectures we gave in july 2007 during the summer school _ higher dimensional geometry over finite fields _ , june 25 - july 06 , 2007 , mathematisches
institut , georg - august - universitt gttingen .
the goal of the paper is to provide to non - specialists an _ efficient , accessible , and in depth _ introduction to the theory of _ integral models _ of _ shimura varieties of hodge type_. accordingly , the paper will put a strong accent on defining the main objects of interest , on listing the main problems , on presenting the main techniques used in approaching the main problems , and on stating very explicitly the main results obtained so far .
this is not an easy task , as only to be able to list the main problems one requires a good comprehension of the _ language of schemes _ , of _ reductive groups _ , of _ abelian varieties _ , of _ hodge cycles _ on abelian varieties , of
_ cohomology theories _ ( including _ tale and crystalline _ ones ) , of _ deformation theories _ , of _ @xmath0-divisible groups _ , and of _
@xmath1-crystals_. whenever possible , proofs are included .
we begin with a motivation for the study of shimura varieties of hodge type .
let @xmath2 be a connected , smooth , projective variety over @xmath3 .
we recall that the _ albanese variety _ of @xmath2 is an abelian variety @xmath4 over @xmath3 equipped with a morphism @xmath5 that has the following universal property .
if @xmath6 is another morphism from @xmath2 to an abelian variety @xmath7 over @xmath3 , then there exists a unique morphism @xmath8 such that the following identity @xmath9 holds .
this universal property determines @xmath4 uniquely up to isomorphisms .
not only @xmath4 is uniquely determined by @xmath2 , but also the image @xmath10 is uniquely determined by @xmath2 up to isomorphisms . thus to @xmath2 one associates an abelian variety @xmath4 over @xmath3 as well as a closed subvariety @xmath10 of it . if @xmath2 belongs to a good class @xmath11 of connected , smooth , projective varieties over @xmath3 , then deformations of @xmath2 would naturally give birth to deformations of the closed embedding @xmath12 .
thus the study of moduli spaces of objects of the class @xmath11 is very much related to the study of moduli spaces of abelian schemes endowed with certain closed subschemes ( which naturally give birth to some polarizations ) .
for instance , if @xmath2 is a curve of positive genus , then @xmath13 and the morphism @xmath14 is a closed embedding ; to this embedding one associates naturally a principal polarization of @xmath15 .
this implies that different moduli spaces of geometrically connected , smooth , projective curves are subspaces of different moduli spaces of principally polarized abelian schemes . for the sake of generality and flexibility
, it does not suffice to study moduli spaces of abelian schemes endowed with polarizations and with certain closed subschemes .
more precisely , one is naturally led to study moduli spaces of polarized abelian schemes endowed with _
families _ of _ hodge cycles_. they are called _ shimura varieties of hodge type _ ( see subsection 3.4 ) . the classical _ hodge conjecture _
predicts that each hodge cycle is an _ algebraic cycle_.
thus we refer to subsection 2.5 for a quick introduction to hodge cycles on abelian schemes over reduced @xmath16schemes .
subsections 2.1 to 2.5 review basic properties of algebraic groups , of hodge structures , and of families of tensors . _ shimura varieties _ can be defined abstractly via few axioms due to deligne ( see subsection 3 ) .
they are in natural bijection to _ shimura pairs _ @xmath17 .
here @xmath18 is a reductive group over @xmath16 and @xmath2 is a hermitian symmetric domain whose points form a @xmath19-conjugacy class of homomorphisms @xmath20 of real groups , that are subject to few axioms .
initially one gets a complex shimura variety @xmath21 defined over @xmath3 ( see subsection 3.1 ) .
the totally discontinuous , locally compact group @xmath22 acts naturally on @xmath21 from the right .
cumulative works of shimura , taniyama , deligne , borovoi , milne , etc . , have proved that @xmath21 has a _ canonical model _ @xmath23 over a number field @xmath24 which is intrinsically associated to the shimura pair @xmath17 and which is called the
_ reflex field _ of @xmath17 ( see subsection 3.2 ) .
one calls @xmath23 together with the natural right action of @xmath22 on it , as the shimura variety defined by the shimura pair @xmath17 . for instance , if @xmath25 and @xmath26 is isomorphic to two copies of the upper half - plane , then @xmath23 is the _
elliptic modular variety _ over @xmath16 and is the projective limit indexed by @xmath27 of the _ affine modular curves _
@xmath28 .
let @xmath29 be a compact , open subgroup of @xmath22 .
the quotient scheme @xmath30 exists and is a normal , quasi - projective scheme over @xmath24 . if @xmath31 is a prime of @xmath24 of residue field @xmath32 and if @xmath33 is a good integral model of @xmath30 over the local ring @xmath34 of @xmath31 , then one gets a shimura variety @xmath35 over the finite field @xmath32 .
the classical example of a good integral model is _ mumford moduli scheme _ @xmath36 .
here @xmath37 , the @xmath38-scheme @xmath36 is the _ course moduli scheme _ of principally polarized abelian scheme of relative dimension @xmath39 , and the @xmath16scheme @xmath40 is of the form @xmath30 for @xmath17 a shimura pair that defines ( see example 3.1.2 ) a _ siegel modular variety_. in this paper , we are mainly interested in shimura varieties of hodge type .
roughly speaking , they are those shimura varieties for which one can naturally choose @xmath33 to be a finite scheme over @xmath41 . in this paper
we study @xmath33 and its special fibre @xmath35 .
see subsections 4.1 and 4.2 for some moduli interpretations of @xmath33 .
see section 5 for different results pertaining to the uniqueness of @xmath33 . see section 6 for basic results that pertain to the smooth locus of @xmath33 .
see section 7 for the list of cases in which @xmath33 is known to be ( or it is expected to be ) a projective @xmath34-scheme .
section 8 presents four main stratifications of the ( smooth locus of the ) special fibre @xmath35 and their basic properties .
these four stratifications are defined by ( see subsections 8.3 , 8.4 , 8.6 , and 8.7 respectively ) : newton polygons of @xmath0-divisible groups ; isomorphism classes of principally quasi - polarized @xmath1-isocrystals with tensors ; inner isomorphism classes of the reductions modulo integral powers of @xmath0 of principally quasi - polarized @xmath1-crystals with tensors ; isomorphism classes of principally quasi - polarized @xmath1-crystals with tensors . the principally quasi - polarized @xmath1-crystals with tensors attached naturally to points of the smooth locus of @xmath35 with values in algebraically closed fields are introduced in subsection 8.1 . generalities on stratifications of reduced schemes over fields are presented in subsection 8.2 .
subsection 8.5 shows that the smooth locus of @xmath35 is a _
quasi shimura @xmath0-variety of hodge type _ in the sense of [ va5 , def .
subsection 8.5 is used in subsections 8.6 and 8.7 to define the last two stratifications , called the _
level @xmath42 _ and _ traverso stratifications_.
in this section we review basic properties of algebraic groups , of hodge structure , of families of tensors , and of hodge cycles on abelian schemes over reduced @xmath16schemes .
we denote by @xmath43 an algebraic closure of a field @xmath44 .
we denote by @xmath45 and @xmath46 the affine , smooth groups over @xmath44 with the property that for each commutative @xmath44-algebra @xmath47 , the groups @xmath48 and @xmath49 are the additive group of @xmath47 and the multiplicative group of units of @xmath47 ( respectively ) . as schemes , we have @xmath50)$ ] and @xmath51[{1\over x}])$ ]
. thus the dimension of either @xmath45 or @xmath46 is @xmath52 . for @xmath53 ,
let @xmath54 be the kernel of the @xmath55-power endomorphism of @xmath46 .
an algebraic group scheme over @xmath44 is called _ linear _ , if it is isomorphic to a subgroup scheme of @xmath56 for some @xmath57 .
let @xmath18 be a smooth group over @xmath44 which is of finite type .
let @xmath58 be the identity component of @xmath18 .
we have a short exact sequence @xmath59 where the quotient group @xmath60 is finite and tale .
a classical theorem of chevalley shows that , if @xmath44 is either perfect or of characteristic @xmath61 , then there exists a short exact sequence @xmath62 where @xmath63 is an abelian variety over @xmath44 and where @xmath64 is a connected , smooth , linear group over @xmath44 . in what follows
we assume that ( 2 ) exists .
let @xmath65 be the _ unipotent radical _ of @xmath64 .
it is the maximal connected , smooth , normal subgroup of @xmath64 which is _ unipotent _
( i.e. , which over @xmath43 has a composition series whose factors are @xmath45 groups ) .
we have a short exact sequence @xmath66 where @xmath67 is a _ reductive group _ over @xmath44 ( i.e. , it is a smooth , connected , linear group over @xmath44 whose unipotent radical is trivial ) . by the _ @xmath44-rank _ of @xmath68 we mean the greatest non - negative integer @xmath69 such that @xmath70 is a subgroup of @xmath68 .
if the @xmath44-rank of @xmath68 is equal to the @xmath43-rank of @xmath71 , then we say that @xmath68 is _
split_. let @xmath72 be the ( scheme - theoretical ) center of @xmath68 . it is a group scheme of _ multiplicative type _ ( i.e. , over @xmath43 it is the extension of a finite product of @xmath54 group schemes by a torus @xmath73 ; here @xmath74 and @xmath53 ) .
the quotient group @xmath75 is called the _ adjoint group _ of @xmath68 ; it is a reductive group over @xmath44 whose ( scheme - theoretical ) center is trivial .
let @xmath76 be the _ derived group _ of @xmath68 ; it is the minimal , normal subgroup of @xmath68 with the property that the quotient group @xmath77 is abelian .
the group @xmath78 is a _ torus _
( i.e. , over @xmath43 it is isomorphic to @xmath73 ) .
the groups @xmath79 and @xmath76 are _
semisimple_. we have two short exact sequences @xmath80 and @xmath81 the short exact sequences ( 1 ) to ( 5 ) are intrinsically associated to @xmath18 . if @xmath82 , then @xmath83 and @xmath84 are isomorphic to @xmath46 , @xmath85 , and @xmath86 . if @xmath87 , then @xmath83 and @xmath84 are isomorphic to @xmath46 , @xmath88 , and @xmath89 . if @xmath90 , then @xmath83 and @xmath84 are trivial and therefore from ( 4 ) and ( 5 ) we get that @xmath91 .
let @xmath92 , @xmath93 with @xmath94 .
let @xmath95 be the simply connected semisimple group over @xmath16 whose @xmath16valued points are the @xmath96valued points of @xmath97 that leave invariant the hermitian form @xmath98 over @xmath96 .
let @xmath99 be the semisimple group over @xmath16 of @xmath100 by @xmath100 matrices of determinant 1 that leave invariant the quadratic form @xmath101 on @xmath102 .
let @xmath103 .
let @xmath104 be the semisimple group over @xmath16 whose group of @xmath16valued points is the subgroup of @xmath105 that leaves invariant the skew hermitian form @xmath106 over @xmath96 ( @xmath107 s and @xmath108 s are related here over @xmath96 via @xmath109 ) .
[ df1 ] by a _ reductive group scheme _ @xmath110 over a scheme @xmath111 , we mean a smooth group scheme over @xmath111 which is an affine @xmath111-scheme and whose fibres are reductive groups over fields . as above , one defines group schemes @xmath112 , @xmath113 , @xmath114 , and @xmath115 over @xmath111 which are affine @xmath111-schemes .
the group scheme @xmath112 is of multiplicative type .
the group schemes @xmath113 and @xmath114 are semisimple .
the group scheme @xmath115 is a torus .
let @xmath116 be a separable finite field extension .
let @xmath18 be a group scheme over @xmath44 which is of finite type .
let @xmath117 be the group scheme over @xmath118 obtained from @xmath18 through the weil restriction of scalars .
thus @xmath119 is defined by the functorial group identification @xmath120 where @xmath121 is an arbitrary @xmath118-scheme .
we have @xmath122 from ( 7 ) we easily get that : if @xmath18 is a reductive ( resp .
connected , smooth , affine , linear , unipotent , torus , semisimple , or abelian variety ) group over @xmath44 , then @xmath119 is a reductive ( resp .
connected , smooth , affine , linear , unipotent , torus , semisimple , or abelian variety ) group over @xmath118 .
if @xmath123 is another separable finite field extension , then we have a canonical and functorial identification @xmath124 as one can easily check starting from formula ( 6 ) .
if @xmath29 is a group scheme over @xmath118 , then we have a natural closed embedding homomorphism @xmath125 over @xmath118 which at the level of @xmath118-valued points induces the standard monomorphism @xmath126 .
let @xmath127 be the two dimensional torus over @xmath128 whose group of @xmath128-valued points is the multiplicative group @xmath129 of @xmath3 .
as schemes , we have @xmath130[{1\over { x^2+y^2}}])$ ] . by applying
( 8) we get that we have a short exact sequence @xmath131 the group @xmath132 is isomorphic to the unit circle and thus to @xmath133 . the short exact sequence ( 9 ) does not split ;
this is so as @xmath134 is isomorphic to @xmath135 , where @xmath136 is embedded diagonally into the product .
we have @xmath137 .
we identify @xmath138 in such a way that the natural monomorphism @xmath139 induces the map @xmath140 , where @xmath141 .
let @xmath142 be a @xmath38-subalgebra of @xmath128 ( in most applications , we have @xmath143 ) .
let @xmath144 be a free @xmath142-module of finite rank .
let @xmath145 . by a _ hodge @xmath142-structure _ on @xmath144 we mean a homomorphism @xmath146 we have a direct sum decomposition @xmath147 with the property that @xmath148 acts via @xmath149 on @xmath150 as the scalar multiplication with @xmath151 .
thus the element @xmath152 acts via @xmath153 on @xmath150 as the scalar multiplication with @xmath154 .
therefore @xmath155 acts on @xmath156 as the scalar multiplication with @xmath157 .
this implies that for all @xmath158 we have an identity @xmath159 conversely , each direct sum decomposition ( 11 ) that satisfies the identities ( 12 ) , is uniquely associated to a homomorphism as in ( 10 ) . by the _ type _ of the hodge @xmath142-structure on @xmath144 , we mean any symmetric subset @xmath160 of @xmath161 with the property that we have a direct sum decomposition @xmath162 here symmetric refers to the fact that if @xmath163 , then we also have @xmath164 .
if we can choose @xmath160 such that the sum @xmath165 does not depend on @xmath163 , one says that the hodge @xmath142-structure on @xmath144 has _ weight _ @xmath166 . for @xmath167 ,
let @xmath168 be the hodge @xmath142-structure on @xmath169 which has type @xmath170 .
suppose that the hodge @xmath142-structure on @xmath144 has weight @xmath166 . by a polarization of the hodge @xmath142-structure on @xmath144 we mean a morphism @xmath171 of hodge @xmath142-structures such that the bilinear form @xmath172 defined for @xmath173 , is symmetric and positive definite .
here we identify @xmath174 with its scalar extension to @xmath128 .
let @xmath63 be an abelian variety over @xmath3 .
we take @xmath175 .
let @xmath176 be the first cohomology group of the analytic manifold @xmath177 with coefficients in @xmath38 .
then the classical hodge theory provides us with a direct sum decomposition @xmath178 where @xmath179 and @xmath180 ( see [ mu , ch .
i , 1 ] ) . here
@xmath181 is the structured ring sheaf on @xmath63 and @xmath182 is the @xmath181-module of @xmath52-forms on @xmath63 .
we have @xmath183 and therefore ( 13a ) defines a hodge @xmath38-structure on @xmath144 .
let @xmath184 ; it is called the _
hodge filtration _ of @xmath185 .
let @xmath186 .
let @xmath187 . taking the dual of ( 13a )
, we get a hodge @xmath38-structure on @xmath188 of the form @xmath189 one can identify naturally @xmath190 .
each @xmath152 acts on the complex vector space @xmath191 as the multiplication with @xmath155 and this explains the convention on negative power signs used in the paragraph after formula ( 11 ) .
we have canonical identifications @xmath192 if @xmath193 is a polarization of @xmath63 , then the non - degenerate form @xmath194 defined naturally by @xmath193 , is a polarization of the hodge @xmath38-structure on @xmath188 .
we have @xmath195 .
due to the identities ( 13a ) and ( 13b ) , the hodge @xmath38-structure on @xmath196 is of type @xmath197 [ df2 ] we use the notations of example 2.3.1 .
let @xmath198 . by the _
tate group _ of the complex abelian variety @xmath63 , we mean the smallest subgroup @xmath199 of @xmath200 with the property that the homomorphism @xmath201 that defines the hodge @xmath38-structure on @xmath188 , factors through @xmath202 .
[ p0 ] the group @xmath199 is a reductive group over @xmath16 .
* proof : * from its very definition , the group @xmath199 is connected . to prove the proposition it suffices to show that the unipotent radical @xmath203 of @xmath199 is trivial .
let @xmath204 be the largest rational subspace of @xmath205 on which @xmath203 acts trivially .
as @xmath203 is a normal subgroup of @xmath199 , @xmath204 is an @xmath199-module .
thus @xmath206 normalizes @xmath207 and therefore we have a direct sum decomposition @xmath208\oplus [ ( w_1\otimes_{\bbb q } \bbb c)\cap w^{0,-1}].\ ] ] thus @xmath209 $ ] is a closed analytic submanifold @xmath210 of @xmath211 .
a classical theorem of serre asserts that @xmath210 is algebraizable i.e. , it is the analytic submanifold associated to an abelian subvariety @xmath212 of @xmath63 . the short exact sequence @xmath213 splits up to isogenies ( i.e. , @xmath63 is isogeneous to @xmath214 , where @xmath215 )
. let @xmath216 .
we have a direct sum decomposition @xmath217 whose extension to @xmath128 is normalized by @xmath206 .
thus the direct sum decomposition @xmath217 is normalized by @xmath199 . in particular
, @xmath218 is an @xmath203-module . if @xmath219 , then the unipotent group @xmath203 acts trivially on a non - zero subspace of @xmath218 and this represents a contradiction with the largest property of @xmath204 .
thus @xmath220 i.e. , @xmath203 acts trivially on @xmath205 .
therefore @xmath203 is the trivial group .
let @xmath221 be a free module of finite rank over a commutative @xmath38-algebra @xmath47 .
let @xmath222 . by the _
essential tensor algebra _ of @xmath223 we mean the @xmath47-module @xmath224 let @xmath225 be a direct summand of @xmath221 .
let @xmath226 and @xmath227 .
let @xmath228 , @xmath229 , and @xmath230 .
let @xmath231 be the tensor product filtration of @xmath232 defined by the exhaustive , separated filtrations @xmath233 and @xmath234 of @xmath221 and @xmath235 ( respectively ) .
we refer to @xmath231 as the filtration of @xmath232 defined by @xmath225 and to each @xmath236 as the @xmath237-filtration of @xmath232 defined by @xmath225 .
we identify naturally @xmath238 and @xmath239 .
let @xmath240 be a non - divisor of @xmath61 .
a family of tensors of @xmath241)=\mathcal{t}(m)[{1\over x}]$ ] is denoted @xmath242 , with @xmath243 as the set of indexes .
let @xmath244 be another free @xmath47-module of finite rank .
let @xmath245 be a family of tensors of @xmath246)$ ] indexed also by the set @xmath243 . by an isomorphism @xmath247 we mean a @xmath47-linear isomorphism @xmath248 that extends naturally to a @xmath47-linear isomorphism @xmath241)\tilde\to\mathcal{t}(m_1[{1\over x}])$ ] which takes @xmath249 to @xmath250 for all @xmath251 .
we emphasize that we will denote two tensors or bilinear forms in the same way , provided they are obtained one from another via either a reduction modulo some ideal or a scalar extension .
we will use the terminology of [ de3 ] on hodge cycles on an abelian scheme @xmath252 over a reduced @xmath16scheme @xmath2 .
thus we write each hodge cycle @xmath31 on @xmath252 as a pair @xmath253 , where @xmath254 and @xmath255 are the _ de rham _ and the _ tale component _ of @xmath31 ( respectively ) .
the tale component @xmath255 at its turn has an @xmath118-component @xmath256 , for each rational prime @xmath118 . in
what follows we will be interested only in hodge cycles on @xmath252 that involve no tate twists and that are tensors of different essential tensor algebras .
accordingly , if @xmath2 is the spectrum of a field @xmath257 , then in applications @xmath256 will be a suitable @xmath258-invariant tensor of @xmath259 , where @xmath260 .
if @xmath261 is a subfield of @xmath3 , then we will also use the betti realization @xmath262 of @xmath31 .
the tensor @xmath262 has the following two properties ( that define hodge cycles on @xmath252 which involve no tate twist ; see [ de3 , sect .
2 ] ) : it is a tensor of @xmath263 that corresponds to @xmath254 ( resp . to @xmath256 ) via the canonical isomorphism that relates the betti cohomology of @xmath264 with @xmath16coefficients with the de rham ( resp .
the @xmath265 tale ) cohomology of @xmath266 ; it is also a tensor of the @xmath267-filtration of the filtration of @xmath268 defined by the hodge filtration @xmath269 of @xmath270 .
we have the following particular example : if @xmath271 , then from riemann theorem we get that @xmath262 is the betti realization of a @xmath16endomorphism of @xmath266 and therefore the hodge cycle @xmath253 on @xmath252 is defined uniquely by a @xmath16endomorphism of @xmath252 .
the class of hodge cycles is stable under pull backs . in particular ,
if @xmath2 is a reduced @xmath16scheme of finite type , then the pull back of @xmath253 via a complex point @xmath272 , is a hodge cycle on the complex abelian variety @xmath266 .
let @xmath63 be an abelian variety over @xmath3 .
let @xmath142 be an irreducible , closed subvariety of @xmath63 .
let @xmath166 be the codimension of @xmath142 in @xmath63 .
to @xmath142 one associates classes @xmath273_{\rm dr}\in h_{\rm dr}^{2n}(a/\bbb c)$ ] , @xmath273_l\in h_{\acute et}^{2n}(a,\bbb q_l)(n)$ ] , and @xmath273_b\in h^{2n}(a^{\operatorname{an}},\bbb q)(n)$ ] .
if @xmath273_{\acute et}:=([s]_l)_{l \;{\rm a}\;{\rm prime}}$ ] , then the pair @xmath274_{\rm dr},[s]_{\acute et})$ ] is a hodge cycle on @xmath63 which involves tate twists and whose betti realization is @xmath273_b$ ] .
one can identify @xmath275 with a @xmath265-subspace of @xmath276^{\otimes n}$ ] and @xmath277 with a @xmath3-subspace of @xmath278^{\otimes n}$ ] ; thus one can naturally view @xmath274_{\rm dr},[s]_{\acute et})$ ] as a hodge cycle on @xmath63 which involves no tate twists .
the @xmath16linear combinations of such cycles @xmath274_{\rm dr},[s]_{\acute et})$ ] are called algebraic cycles on @xmath63 .
in this section we introduce shimura varieties and their basic properties and main types .
all continuous actions are in the sense of [ de2 , subsubsect .
2.7.1 ] and are right actions .
thus if a totally discontinuous , locally compact group @xmath279 acts continuously ( from the right ) on a scheme @xmath121 , then for each compact , open subgroup @xmath280 of @xmath279 the geometric quotient scheme @xmath281 exists and the epimorphism @xmath282 is pro - finite ; moreover , we have an identity @xmath283 .
a _ shimura pair _
@xmath17 consists of a reductive group @xmath18 over @xmath16 and a @xmath19-conjugacy class @xmath284 of homomorphisms @xmath285 that satisfy deligne s axioms of [ de2 , subsubsect .
2.1.1 ] : the hodge @xmath16structure on @xmath286 defined by each element @xmath287 is of type @xmath288 ; no simple factor of the adjoint group @xmath289 of @xmath18 becomes compact over @xmath128 ; @xmath290 is a cartan involution of @xmath291 , where @xmath292 is the adjoint representation . axiom ( iii )
is equivalent to the fact that the adjoint group @xmath293 has a faithful representation @xmath294 with the property that there exists a polarization of the hodge @xmath128-structure on @xmath295 defined naturally by any @xmath287 which is fixed by @xmath293 .
these axioms imply that @xmath284 has a natural structure of a hermitian symmetric domain , cf .
[ de2 , cor .
1.1.17 ] . for @xmath287
we consider the _
hodge cocharacter _
@xmath296 defined on complex points by the rule : @xmath297 is mapped to @xmath298 .
let @xmath299 be the number subfield of @xmath3 that is the field of definition of the @xmath300-conjugacy class @xmath301 $ ] of the cocharacters @xmath302 s of @xmath303 , cf .
[ mi2 , p. 163 ] .
more precisely @xmath301 $ ] is defined naturally by a @xmath304-conjugacy class @xmath305 $ ] of cocharacters @xmath306 ; the galois group @xmath307 acts naturally on the set of such @xmath304-conjugacy classes and @xmath24 is the number field which is the fixed field of the stabilizer subgroup of @xmath305 $ ] in @xmath307 .
one calls @xmath24 the _ reflex field _ of @xmath17 .
we define the _ complex shimura space _
@xmath308 where @xmath309 is the set of compact , open subgroups of @xmath22 endowed with the inclusion relation ( see [ de1 ] , [ de2 ] , and [ mi1 ] to [ mi4 ] ) .
thus @xmath310 is a normal complex space on which @xmath22 acts .
we have an identity @xmath311,\leqno ( 16)\ ] ] where @xmath312 is the topological closure of @xmath313 in @xmath22 ( cf .
[ de2 , prop .
2.1.10 ] ) .
let @xmath287 and @xmath314 .
let @xmath315\in \operatorname{sh}(g,\mathcal{x})_{\bbb c}(\bbb c)$ ] be the point defined naturally by the equivalence class of @xmath316 , cf .
the @xmath22-action on @xmath310 is defined by the rule @xmath315\cdot g:=[x , ag]$ ] .
for @xmath317 a compact subgroup of @xmath22 let @xmath318 .
let @xmath319 .
we can write @xmath320 as a disjoint union of normal complex spaces of the form @xmath321 , where @xmath322 is a connected component of @xmath284 and @xmath323 is an _
arithmetic subgroup _ of @xmath324 ( i.e. , is the intersection of @xmath324 with a compact , open subgroup of @xmath22 ) .
a classical result of baily and borel allows us to view naturally @xmath320 as the complex space associated to a finite , disjoint union @xmath325 of normal , quasi - projective , connected varieties over @xmath3 ( see [ bb , thm .
10.11 ] ) .
thus @xmath325 is a normal , quasi - projective @xmath3-scheme and @xmath326 is a normal @xmath3-scheme on which @xmath22 acts .
we have a canonical identification @xmath327 .
if @xmath328 is small enough , then @xmath328 acts freely on @xmath21 and thus @xmath325 is in fact a smooth , quasi - projective @xmath3-scheme .
let @xmath63 be an abelian variety over @xmath3 .
let @xmath199 be its mumford tate group .
let @xmath329 be the homomorphism that defines the hodge @xmath38-structure on @xmath330 , cf .
definition [ df2 ] .
let @xmath331 be the @xmath332-conjugacy class of @xmath206 .
we check that the pair @xmath333 is a shimura pair .
the fact that the axiom 3.1 ( i ) holds for @xmath333 is implied by ( 15 ) . if @xmath334 has a ( non - trivial ) simple factor @xmath335 which over @xmath128 is compact , then the fact that @xmath331 is a hermitian symmetric domain implies that the image of @xmath206 in @xmath336 is trivial and this contradicts the smallest property ( see definition [ df2 ] ) of the mumford
tate group @xmath199 .
thus the axioms 3.1 ( ii ) holds for @xmath333 .
the fact that the axioms 3.1 ( iii ) holds is implied by the fact that @xmath7 has a polarization and thus by the fact that ( 14 ) holds .
we emphasize that the reflex field @xmath337 can be any cm number field .
the most studied shimura pairs are constructed as follows .
let @xmath205 be a vector space over @xmath16 of even dimension @xmath338 .
let @xmath174 be a non - degenerate alternative form on @xmath205 .
let @xmath339 be the set of all monomorphisms @xmath340 that define hodge @xmath16structures on @xmath205 of type @xmath341 and that have either @xmath342 or @xmath343 as polarizations .
thus @xmath339 is two copies of the siegel domain of genus @xmath39 ( the two copies correspond to either @xmath342 or @xmath343 being a polarization of the resulting hodge @xmath16structures on @xmath205 ) .
it is easy to see that @xmath339 is a @xmath344-conjugacy class of homomorphisms @xmath345 .
one can choose an abelian variety @xmath63 over @xmath3 such that in fact we have @xmath346 and therefore @xmath347 is a shimura pair , cf .
example 3.1.1 .
we call @xmath347 a shimura pair that defines a siegel modular variety @xmath348 ( to be defined in subsection 3.2 below ) . as @xmath349 is a split group , the @xmath350-conjugacy class @xmath305 $ ]
is defined naturally by a cocharacter of @xmath349 and therefore we have @xmath351 .
let @xmath166 be a positive integer .
let @xmath352 ; it is the identity component of the group that fixes the quadratic from @xmath353 on @xmath354 .
the group @xmath18 has a subgroup @xmath355 which normalizes the rational vector subspaces of @xmath354 generated by the first two and by the last @xmath166 vectors of the standard @xmath16basis for @xmath354 .
let @xmath356 be a homomorphism whose image is the subgroup @xmath357 of @xmath358 and whose kernel is the split torus @xmath46 of @xmath134 .
let @xmath284 be the @xmath19-conjugacy class of @xmath359 .
then the pair @xmath17 is a shimura pair .
the group @xmath360 is split ( i.e. , @xmath361}$ ] is a subgroup of it ) and thus the @xmath304-conjugacy class @xmath305 $ ] is defined naturally by a cocharacter @xmath362 .
we can choose @xmath363 such that the non - trivial element of @xmath364 takes @xmath363 under galois conjugation to @xmath365 .
it is easy to see that the two cocharacters @xmath363 and @xmath365 are @xmath366-conjugate .
therefore @xmath367 .
if @xmath368 , then @xmath17 is the shimura pair associated to moduli spaces of polarized k3 surfaces . let @xmath369 be a torus over @xmath16 . let @xmath370 be an arbitrary homomorphism .
then the pair @xmath371 is a shimura pair .
its reflex field @xmath372 is the field of definition of the cocharacter @xmath373 .
we denote also by @xmath374 the homomorphism whose extension to @xmath3 is @xmath302 . from the homomorphism @xmath374 we get naturally a new one @xmath375
thus for each commutative @xmath16algebra @xmath47 we get a homomorphism @xmath376 .
let @xmath377 be the maximal abelian extension of @xmath257 .
the reciprocity map @xmath378 is defined as follows : if @xmath379 and if @xmath380 is an idle ( of @xmath257 ) such that @xmath381 , then @xmath382 , where @xmath383 is the finite part of @xmath69 . here
the artin reciprocity map @xmath384 is such that a uniformizing parameter is mapped to the geometric frobenius element .
[ df3 ] by a _ map
_ @xmath385 of shimura pairs we mean a homomorphism @xmath386 of groups over @xmath16 such that for each @xmath387 we have @xmath388 .
if @xmath386 is a monomorphism , then we say @xmath385 is an _
injective map_. if @xmath389 is a torus and if @xmath390 is an injective map , then @xmath391 is called a _ special pair _ in @xmath392 . by a _ model _ of @xmath21 over a subfield @xmath44 of @xmath3 , we mean a scheme @xmath142 over @xmath44 endowed with a continuous right action of @xmath22 ( defined over @xmath44 ) , such that there exists a @xmath22-equivariant isomorphism @xmath393 the _ canonical model _ of @xmath21 ( or of @xmath17 itself ) is the model @xmath23 of @xmath21 over @xmath24 which satisfies the following property : the canonical model of @xmath23 exists and is uniquely determined by the property ( * ) up to a unique isomorphism ( see [ de1 ] , [ de2 ] , [ mi2 ] , and [ mi4 ] ) . by the _ dimension _
@xmath394 of @xmath23 ( or @xmath17 or @xmath21 ) we mean the dimension of @xmath284 as a complex manifold .
one computes @xmath394 as follows . for @xmath395 ,
let @xmath396 be the hodge decomposition defined by @xmath359 .
let @xmath397 be the centralizer of @xmath359 in @xmath358 ; it is a reductive group over @xmath128 ( cf .
[ bo , ch .
iv , 13.17 , cor .
2 ] ) . we have @xmath398 and ( as analytic real manifolds ) @xmath399/[k_{\infty}(\bbb r)]$ ] .
thus as @xmath400 , we get that @xmath401 for @xmath317 a compact subgroup of @xmath22 let @xmath402 .
if @xmath319 , then @xmath403 is a normal , quasi - projective @xmath24-scheme which is equidimensional of dimension @xmath394 and whose extension to @xmath3 is ( canonically identified with ) the @xmath3-scheme @xmath325 we have introduced in subsection 3.1 .
if @xmath385 is a map between two shimura pairs , then @xmath404 is a subfield of @xmath405 and there exists a unique @xmath406-equivariant morphism ( still denoted by @xmath391 ) @xmath407 which at the level of complex points is the map @xmath315\to [ f(x),f(a)]$ ] ( [ de1 , cor . 5.4 ] ) .
we get as well a @xmath22-equivariant morphism ( denoted in the same way ) @xmath408 of @xmath404-schemes .
if @xmath391 is an injective map , then based on ( 16 ) one gets that ( 18 ) is in fact a closed embedding .
let @xmath17 be a shimura pair .
if @xmath287 , let @xmath409 and @xmath410 be the homomorphisms defined naturally by @xmath356 .
the homomorphism @xmath411 does not depend on @xmath287 and the shimura pair @xmath412 has dimension @xmath61 .
let @xmath413 be the @xmath414-conjugacy class of @xmath415 .
the shimura pairs @xmath412 and @xmath416 are called the _
toric _ and the _ adjoint _ ( respectively ) shimura pairs of @xmath17 .
the centralizer @xmath417 of @xmath415 in @xmath293 is a reductive group over @xmath128 which is a maximal compact subgroup of @xmath293 .
the hermitian symmetric domain structure on @xmath413 is obtained via the natural identification @xmath418/[k_{\infty,\operatorname{ad}}(\bbb r)]$ ] .
the hermitian symmetric domain @xmath284 is a finite union of connected components of @xmath413 . in particular
, we have @xmath419 .
we have a product decomposition @xmath420 into _ simple adjoint _
shimura pairs , where each @xmath421 is a simple group over @xmath16 . for each @xmath422 there exists a number field @xmath423 such that we have an isomorphism @xmath424 , where @xmath425 is an absolutely simple adjoint group over @xmath423 ( see [ ti , subsubsect .
3.1.2 ] ) .
the number field @xmath423 is uniquely determined up to @xmath307-conjugation ( i.e. , up to isomorphism ) .
axiom 3.2 ( iii ) is equivalent to the fact that @xmath293 is an inner form of its compact form @xmath426 , cf .
[ de2 , p. 255 ] .
thus @xmath427 is a product of absolutely simple , adjoint groups over @xmath128 .
but for each @xmath422 we have @xmath428 $ ] . from the last two sentences , we get that for each @xmath422 the @xmath128-algebra @xmath429 is isomorphic to a finite number of copies of @xmath128 . in other words , for each @xmath422 the number field @xmath423 is totally real .
we have the following conclusions of the last three paragraphs : let @xmath18 be a reductive group over @xmath16 . to give a shimura pair @xmath17
is the same thing as to give a shimura pair @xmath412 of dimension @xmath61 ( i.e. , a homomorphism @xmath409 ) and an adjoint shimura pair @xmath416 , with the properties that for an ( any ) element @xmath430 the homomorphism @xmath431 lifts to a homomorphism @xmath356 , where @xmath432 is the standard isogeny .
one takes @xmath284 to be the @xmath19-conjugacy class of @xmath359 .
we emphasize that the shimura pair @xmath17 can depend on the choice of @xmath430 ( though its isomorphism class does not ) . to give an adjoint shimura pair @xmath416 is the same thing as to give a finite set @xmath433 of simple adjoint shimura pairs , cf .
( 19 ) . to give a simple adjoint shimura pair @xmath433 , one has to first give a totally real number field @xmath423 and an absolutely simple , adjoint group @xmath425 over @xmath423 that satisfies the following property : for each embedding @xmath434 , the group @xmath435 is either compact ( and then one defines @xmath436 to be a set with one element ) or is not compact and associated naturally to a connected hermitian symmetric domain @xmath436 .
the product @xmath437 is a connected hermitian symmetric domain isomorphic to the connected components of @xmath438 . if @xmath439 is of classical lie type and if @xmath435 is not compact , then @xmath435 is isomorphic to either @xmath440 with @xmath441 , or @xmath442 with @xmath443 , or @xmath444 with @xmath443 , or @xmath445 with @xmath446 .
the last think one has to give is a family of homomorphisms @xmath447 , where @xmath448 @xmath449 is trivial if @xmath435 is compact , and @xmath448 @xmath449 identifies @xmath450 with the identity component of the center of a maximal compact subgroup of @xmath435 if @xmath435 is not compact .
one takes @xmath438 to be the @xmath451-conjugacy class of the composite of the natural epimorphism @xmath452 with @xmath453 .
once @xmath423 and @xmath425 are given , there exist a finite number of possibilities for @xmath438 ( they correspond to possible replacements of some of the @xmath449 s by their inverses ) . a shimura variety @xmath454 is called _ unitary _ if the adjoint group @xmath455 is non - trivial and all simple factors of @xmath456 are @xmath457 groups over @xmath3 .
let @xmath17 be a simple , adjoint shimura pair .
let @xmath458 be the lie type of anyone of the simple factors of @xmath303 .
if @xmath458 is either @xmath459 , @xmath460 , @xmath461 , @xmath462 , or @xmath463 , then one say that @xmath17 is of @xmath458 shimura type .
if @xmath458 is @xmath464 with @xmath446 , then there exist three disjoint possibilities for the type of @xmath17 : they are @xmath465 , @xmath466 , and @xmath467 .
if @xmath468 , then @xmath17 is of @xmath465 ( resp . of @xmath466 ) shimura type if and only if each simple , non - compact factor of @xmath358 is isomorphic to @xmath445 ( resp . to @xmath469 ) .
the only if part of the previous sentence holds even if @xmath470 .
we will not detail here the precise difference between the shimura types @xmath471 , @xmath472 , and @xmath473 ( see [ de2 , p. 272 ] ) .
let @xmath17 be a shimura pair .
we say that @xmath23 ( or @xmath17 ) is of _ hodge type _ , if there exists an injective map @xmath474 into a shimura pair that defines a siegel modular variety .
the hodge @xmath16structure on @xmath205 defined by any @xmath287 is of type @xmath341 , cf .
this implies that @xmath475 is the group of scalar automorphisms of @xmath476 .
therefore @xmath83 contains the group @xmath477 of scalar automorphisms of @xmath205 . the image of @xmath478 in @xmath479 is contained in the centralizer of the image of @xmath359 in @xmath480 and thus it is contained in a compact group . from the last two sentences we get that we have a short exact sequence @xmath481 where @xmath482 is a compact group of multiplicative type .
in this way we get the only if part of the following proposition ( see [ de2 , prop .
2.3.2 or cor .
2.3.4 ] ) .
[ p1 ] a shimura pair @xmath17 is of hodge type if and only if the following two properties hold : there exists a faithful representation @xmath483 with the property that the hodge @xmath16structure on @xmath205 defined by a ( any ) @xmath287 is of type @xmath341 ; we have a short exact sequence as in ( 20 ) .
if @xmath17 is of hodge type , then ( 16 ) becomes ( cf .
[ de2 , cor . 2.1.11 ] ) @xmath484 let @xmath474 be an injective map .
we fix a family @xmath485 of tensors of @xmath486 such that @xmath18 is the subgroup of @xmath349 that fixes @xmath487 for all @xmath251 , cf .
[ de3 , prop . 3.1 ( c ) ] .
let @xmath64 be a @xmath38-lattice of @xmath205 such that we have a perfect form @xmath488 .
we follow [ va1 , subsect .
4.1 ] to present the standard moduli interpretation of the complex shimura variety @xmath21 with respect to the @xmath38-lattice @xmath64 of @xmath205 and the family of tensors @xmath485 .
we consider quadruples of the form @xmath489 $ ] where : @xmath490 is a principally polarized abelian variety over @xmath3 ; @xmath491 is a family of hodge cycles on @xmath63 ; @xmath44 is an isomorphism @xmath492 whose tensorization with @xmath16 ( denoted also by @xmath44 ) takes the betti realization of @xmath493 into @xmath494 for all @xmath251 and which induces a symplectic similitude isomorphism between @xmath495 and @xmath496 .
we define @xmath497 to be the set of isomorphism classes of quadruples of the above form that satisfy the following two conditions : there exists a similitude isomorphism @xmath498 that takes the betti realization of @xmath493 into @xmath494 for all @xmath251 ; by composing the homomorphism @xmath499 that defines the hodge @xmath128-structure on @xmath500 with an isomorphism of real groups @xmath501 induced naturally by an isomorphism as in ( i ) , we get an element of @xmath284 .
we have a right action of @xmath22 on @xmath497 defined by the rule : @xmath502\cdot g:=[a',\lambda_{a'},(v_\alpha)_{\alpha\in\mathcal{j}},g^{-1}k].\ ] ] here @xmath503 is the abelian variety which is isogeneous to @xmath63 and which is defined by the @xmath38-lattice @xmath504 of @xmath505 whose tensorization with @xmath506 is @xmath507 , while @xmath508 is the only rational multiple of @xmath509 which produces a principal polarization of @xmath503 ( see [ de1 , thm . 4.7 ] for the theorem of riemann used here ) . here
as well as in ( e ) below , we will identify a polarization with its betti realization .
there exists a @xmath22-equivariant bijection @xmath510 defined as follows . to @xmath511\in\operatorname{sh}(g,\mathcal{x})(\bbb c)=g(\bbb q)\setminus ( \mathcal{x}\times g(\bbb a_f))$ ] we associate the quadruple @xmath512):=[a,\lambda_a,(v_\alpha)_{\alpha\in\mathcal{j}},k]$ ] where : @xmath63 is associated to the hodge @xmath16structure on @xmath205 defined by @xmath359 and to the unique @xmath38-lattice @xmath513 of @xmath514 for which we have an isomorphism
@xmath515 induced naturally by the automorphism @xmath516 of @xmath517 ; thus we have @xmath518 , where @xmath519 is the hodge decomposition defined by @xmath359 ; @xmath509 is the only rational multiple of @xmath174 which gives birth to a principal polarization of @xmath63 ; for each @xmath251 , the betti realization of @xmath493 is @xmath494 .
the inverse @xmath520 of @xmath521 is defined as follows .
we consider a quadruple @xmath489\in \mathcal{a}(g,\mathcal{x},w,\psi)$ ] .
we choose a symplectic similitude isomorphism @xmath522 as in ( i ) .
it gives birth naturally to an isomorphism @xmath523 of groups over @xmath16 .
we define @xmath287 to be @xmath524 ( with @xmath206 as in definition [ df2 ] ) and @xmath525 to be the composite isomorphism @xmath526 .
then @xmath527):=[x , g].\ ] ] taking @xmath528 and @xmath529 , we get a bijection between the set @xmath530 and the set of isomorphism classes of principally polarized abelian varieties over @xmath3 of dimension @xmath531 that have ( compatibly ) level-@xmath532 symplectic similitude structures for all positive integers @xmath532 . thus to give a @xmath3-valued point of @xmath348
is the same thing as to give a triple @xmath533 $ ] , where @xmath490 is a principally polarized abelian variety over @xmath3 of dimension @xmath531 and where @xmath534 s are forming a compatible system of symplectic similitude isomorphisms .
the compatibility means here that if @xmath535 and @xmath536 are positive integers such that @xmath537 , then @xmath538 is obtained from @xmath539 by tensoring with @xmath540 .
let @xmath541 .
let @xmath542 be a positive integer .
let @xmath543 be the mumford
moduli scheme over @xmath544 $ ] that parametrizes isomorphism classes of principally polarized abelian schemes over @xmath544$]-schemes that have level-@xmath532 symplectic similitude structure and that have relative dimension @xmath39 , cf .
[ mfk , thms . 7.9 and 7.10 ] .
we consider the @xmath16scheme @xmath545 we have a natural identification @xmath546 of sets , cf .
end of subsubsection 3.4.1 .
one can easily check that this identification is in fact an isomorphism of complex manifolds . from the very definition of the algebraic structure on @xmath547 ( obtained based on [ bb , thm .
10.11 ] ) , one gets that there exists a natural identification @xmath548 of @xmath3-schemes .
classical works of shimura , taniyama , etc .
, show that the last identification is the extension to @xmath3 of an identification @xmath549 of @xmath16schemes . the reflex field @xmath24 is the _
smallest _ number field such that the closed subscheme @xmath21 of @xmath547 is defined over @xmath24 .
in other words , we have a natural closed embedding ( cf .
end of subsection 3.2 ) @xmath550 the pull back @xmath551 to @xmath23 of the universal principally polarized abelian scheme over @xmath552 is such that there exists naturally a family of hodge cycles @xmath553 on the abelian scheme @xmath554 .
if @xmath555\in\operatorname{sh}(g,\mathcal{x})(\bbb c)$ ] and if @xmath512)=[a,\lambda_a,(v_\alpha)_{\alpha\in\mathcal{j}},k]$ ] , then each @xmath556 is the hodge cycle @xmath557 on @xmath558 .
[ df4 ] let @xmath559 be a shimura pair .
we say that @xmath559 is of preabelian type , if there exists a shimura pair @xmath17 of hodge type such that we have an isomorphism @xmath560 of adjoint shimura pairs .
if moreover this isomorphism @xmath560 is induced naturally by an isogeny @xmath561 , then we say that @xmath559 is of abelian type .
[ rmk1 ] let @xmath559 be an arbitrary shimura variety .
let @xmath562 be a faithful representation .
as in subsubsection 3.4.1 , one checks that @xmath563 is a moduli space of hodge @xmath16structures on @xmath204 equipped with extra structures . if moreover @xmath559 is of abelian type , then @xmath563 is in fact a moduli scheme of polarized _ abelian motives _ endowed with hodge cycles and certain compatible systems of level structures ( cf .
[ mi3 ] ) .
let @xmath559 be a simple , adjoint shimura pair .
then @xmath559 is of abelian type if and only if @xmath559 is of @xmath459 , @xmath460 , @xmath461 , @xmath465 , or @xmath466 shimura type . for this classical result due
to satake and deligne we refer to [ sa1 ] , [ sa2 , part iii ] , and [ de2 , table 2.3.8 ] .
there exists a shimura pair @xmath17 of hodge type whose adjoint is isomorphic to @xmath559 and whose derived group @xmath564 is simply connected if and only if @xmath559 is of @xmath459 , @xmath460 , @xmath461 , or @xmath466 shimura type ( cf .
[ de2 , table 2.3.8 ] ) .
in this section we follow [ mi2 ] and [ va1 ] to define different integral models of shimura varieties .
let @xmath565 be a prime .
let @xmath566 be the location of @xmath38 at its prime ideal @xmath567 .
let @xmath568 be the ring of finite adles with the @xmath0-component omitted ; we have @xmath569 .
let @xmath17 be a shimura pair .
let @xmath31 be a prime of @xmath24 that divides @xmath0 .
let @xmath34 be the local ring of @xmath31 . *
( a ) * let @xmath29 be a compact , open subgroup of @xmath570 . by an _
integral model _ of @xmath571 over @xmath34 we mean a faithfully flat scheme @xmath33 over @xmath34 together with a @xmath572-continuous action on it and a @xmath572-equivariant isomorphism @xmath573 when the @xmath572-action on @xmath33 is obvious , by abuse of language , we say that the @xmath34-scheme @xmath33 is an integral model .
the integral model @xmath33 is said to be _ smooth _
( resp . _ normal _ ) if there exists a compact , open subgroup @xmath574 of @xmath572 such that for every inclusion @xmath575 of compact , open subgroups of @xmath574 , the natural morphism @xmath576 is a finite tale morphism between smooth schemes ( resp . between normal schemes ) of finite type over @xmath34 . in other words
, there exists a compact open subgroup @xmath574 of @xmath572 such that @xmath33 is a pro - tale cover of the smooth ( resp . the normal ) scheme @xmath577 of finite type over @xmath34 . a regular , faithfully flat @xmath34-scheme @xmath121 is called _ @xmath0-healthy _ ( resp .
_ healthy _ ) regular , if for each open subscheme @xmath578 of @xmath121 which contains @xmath579 and all points of @xmath121 of codimension @xmath52 , every @xmath0-divisible group ( resp .
every abelian scheme ) over @xmath578 extends uniquely to a @xmath0-divisible group ( resp .
extends to an abelian scheme ) over @xmath121 . a scheme @xmath111 over @xmath34 is said to have the _ extension property _
if for each healthy regular scheme @xmath121 over @xmath34 , every @xmath24-morphism @xmath580 extends uniquely to an @xmath34-morphism @xmath581 . a smooth integral model of @xmath571 over @xmath34 that has the extension property is called an _ integral canonical model _ of @xmath30 over @xmath34 .
let @xmath582 be a dedekind domain .
let @xmath328 be the field of fractions of @xmath582 .
let @xmath583 be a smooth scheme of finite type over @xmath328 . by a
_ nron model _ of @xmath583 over @xmath582 we mean a smooth scheme of finite type @xmath111 over @xmath582 whose generic fibre is @xmath583 and which is uniquely determined by the following universal property : for each smooth scheme @xmath121 over @xmath582 , every @xmath328-morphism @xmath584 extends uniquely to a @xmath582-morphism @xmath581 .
the group @xmath585 is called _ unramified _ if and only if extends to a reductive group scheme @xmath586 over @xmath587 .
in such a case , each compact , open subgroup of @xmath588 of the form @xmath589 is called a _
hyperspecial _ subgroup of @xmath588 .
let @xmath111 be a flat @xmath34-scheme and let @xmath121 be a closed subscheme of @xmath590 .
the _ dilatation _
@xmath205 of @xmath111 centered on @xmath121 is an affine @xmath111-scheme defined as follows . to define @xmath205
, we can work locally in the zariski topology of @xmath111 and therefore we van assume that @xmath591 is an affine scheme .
let @xmath592 be the ideal of @xmath47 that defines @xmath121 and let @xmath593 be a uniformizer of @xmath34 . then @xmath205 is the spectrum of the @xmath47-subalgebra of @xmath594 $ ] generated by @xmath595 with @xmath422 .
the affine morphism @xmath596 of @xmath34-schemes enjoys the following universal property .
let @xmath597 be a morphism of flat @xmath34-schemes .
then @xmath598 factors uniquely through a morphism @xmath599 of @xmath111-schemes if and only if @xmath600 factors through @xmath121 ( i.e. , @xmath601 is a composite morphism @xmath602 ) .
let @xmath474 be an injective map .
let @xmath64 be a @xmath38-lattice of @xmath205 such that @xmath174 induces a perfect form @xmath603 .
let @xmath542 be a natural number which is prime to @xmath0 .
let @xmath604 we have an identity @xmath605 .
let @xmath606 it is a @xmath566-scheme that parametrizes isomorphism classes of principally polarized abelian schemes over @xmath566-schemes that have compatible level-@xmath532 symplectic similitude structures for all @xmath27 prime to @xmath0 and that have relative dimension @xmath39 .
the totally discontinuous , locally compact group @xmath607 acts continuously on @xmath608 and moreover @xmath608 is a pro - tale cover of @xmath609 for all @xmath27 prime to @xmath0 . from ( 21 ) we get that we can identify @xmath610 and @xmath611 . from the last two sentences we get that @xmath608 is a smooth integral model of @xmath612 over @xmath566 .
let @xmath613 be the zariski closure of @xmath18 in @xmath614 ; it is an affine , flat group scheme over @xmath566 whose generic fibre is @xmath18 .
let @xmath615 and @xmath616 .
from ( 21 ) we easily get that we have finite morphisms @xmath617 and @xmath618 as @xmath542 , a principally polarized abelian scheme with level-@xmath532 structure has no automorphism ( see [ mu , ch .
iv , 21 , thm .
5 ] for this result of serre ) .
this implies that @xmath619 acts freely on @xmath620 . from this and ( 21 )
we get that @xmath621 acts freely on @xmath23 .
therefore the @xmath24-scheme @xmath622 is smooth and thus @xmath623 is a regular scheme which is formally smooth over @xmath24 .
let @xmath624 be the normalization of @xmath543 in the ring of fractions of @xmath622 and let @xmath625 be the normalization of @xmath608 in the ring of fractions of @xmath623 .
[ comment : the role of the integral model @xmath33 used in section 1 , will be played in what follows by @xmath624 .
] let @xmath626 be the ring of integers of @xmath24 .
let @xmath627 ; it is the normalization of @xmath566 in @xmath24 .
the scheme @xmath624 is a faithfully flat @xmath628$]-scheme which is normal and of finite type and whose generic fibre is @xmath622 ( the finite type part is implied by the fact that @xmath34 is an excellent ring ) .
the scheme @xmath625 is a faithfully flat @xmath629-scheme which is normal and whose generic fibre is @xmath623 .
one gets the existence of a finite map @xmath630 and of a pro - finite map @xmath631 that extends naturally ( 22a ) and ( 22b ) ( respectively ) .
moreover , the totally discontinuous , locally compact group @xmath572 acts continuously on @xmath625 .
let @xmath632 [ p2 ] * ( a ) * the @xmath34-scheme @xmath633 is a normal integral model of @xmath23 over @xmath34 .
moreover , @xmath633 is a pro - tale cover of @xmath634 .
the morphism @xmath635 is finite .
* proof : * let @xmath574 be a compact , open subgroup of @xmath572 such that @xmath636 is a compact , open subgroup of @xmath621 . as @xmath621 acts freely on @xmath608 , it also acts freely on @xmath625 .
this implies that @xmath633 is a pro - tale cover of both @xmath634 and @xmath637 .
therefore for all open subgroups @xmath638 and @xmath639 of @xmath574 with @xmath640 , the morphism @xmath641 is a finite morphism between tale covers of @xmath634 and therefore it is an tale cover .
based on this , one easily checks that the right action of @xmath572 on @xmath633 is continuous . thus ( a ) holds . part ( b )
is an easy consequence of the fact that @xmath642 is a finite scheme over @xmath643 .
let @xmath644 be the @xmath16subalgebra of @xmath645 formed by elements fixed by @xmath18 .
we consider two axioms : the group @xmath18 is the identity component of the centralizer of @xmath644 in @xmath349 ; the group @xmath18 is the centralizer of @xmath644 in @xmath349 .
if the axiom ( * ) holds , then one calls @xmath23 a shimura variety of _ pel type_. if the axiom ( * * ) holds , then one calls @xmath23 a shimura variety of pel type of either _ a or c type_. here pel stands for polarizations , endomorphisms , and level structures while the a and c types refer to the fact that all simple factors of @xmath646 are ( under the axiom ( * * ) ) of some @xmath459 or @xmath461 lie type ( and not of @xmath464 lie type with @xmath446 ) .
if the axiom ( * * ) holds , then we can choose the family @xmath647 to be exactly the family of all elements of @xmath644 .
in such a case , all hodge cycles mentioned in subsection 3.4.2 are defined by endomorphisms .
let @xmath648 ; it is a @xmath566-order of @xmath644 .
suppose that @xmath649 is a semisimple @xmath566-algebra and that @xmath613 is the centralizer of @xmath649 in the group scheme @xmath650 .
then @xmath613 is a reductive group scheme and moreover @xmath625 ( resp .
@xmath633 ) is a moduli scheme of principally polarized abelian schemes which are over @xmath629-schemes ( resp . over @xmath34-schemes ) , which have relative dimension @xmath39 , which have compatible level-@xmath532 symplectic similitude structures for all @xmath27 prime to @xmath0 , which are endowed with a @xmath566-algebra @xmath649 of @xmath566-endomorphisms , and which satisfy certain axioms that are related to the properties ( d ) to ( f ) of subsubsection 3.4.2 .
unfortunately , presently this is the _ only case _ when @xmath625 ( resp .
when @xmath633 ) has a good moduli interpretation .
this explains the difficulties one encounters in getting as well as of stating results pertaining to either @xmath625 or @xmath633 .
here is a list of six main problems in the study of @xmath624 , @xmath625 , and @xmath633 . for simplicity
, these problems will be stated here only in terms of @xmath624 .
determine when @xmath624 is uniquely determined up to isomorphism by its generic fibre @xmath622 and by a suitable universal property .
determine when @xmath624 is a smooth @xmath628$]-scheme .
determine when @xmath624 is a projective @xmath628$]-scheme .
identify and study different stratifications of the special fibres of @xmath624 .
describe the points of @xmath624 with values in finite fields .
describe the points of @xmath624 with values in @xmath651 $ ] , where @xmath652 is the ring of integers of some finite field extension of @xmath24 . in the next four sections we will study the first four problems one by one , in a way that could be useful towards the partial solutions of the problem ( e ) .
any approach to the problem ( f ) would require a very good understanding of the first five problems and this is the reason ( as well as the main motivation ) for why the six problems are listed together .
until the end we will use the following notations introduced in section 4 : @xmath653 @xmath654 let @xmath655 be the index of ramification of @xmath31 . let @xmath32 be the residue field of @xmath31 .
let @xmath656 } k(v)\;\;\ ; \text{and}\;\;\;\mathcal{l}_v:=\mathcal{n}_v\otimes_{o_{(v ) } } k(v).\ ] ] in this section we study when the @xmath32-scheme @xmath657 ( resp .
@xmath658 ) is uniquely determined in some sensible way by @xmath622 ( resp . by @xmath623 ) and by the prime @xmath31 of @xmath24 .
whenever one gets such a uniqueness property , one can call @xmath657 ( resp .
@xmath658 ) as the _ canonical fibre model _ of @xmath622 ( resp .
@xmath623 ) at @xmath31 ( or over @xmath32 ) .
milne s original insight ( see [ mi2 ] and [ va1 ] ) was to prove in many cases the uniqueness of @xmath633 and @xmath658 by showing first that : the @xmath34-scheme @xmath633 has the extension property , and @xmath633 is a healthy regular scheme in the sense of definition 4.1 ( b ) .
while ( i ) always holds ( see proposition [ p3 ] below ) , it is very hard in general to decide if ( ii ) holds .
however , results of [ va1 ] , [ va2 ] , and [ va11 ] allow us to get that ( ii ) holds in many cases of interest ( see subsection 5.1 ) .
subsection 5.2 shows how one gets the uniqueness of @xmath624 ( and therefore also of @xmath657 ) via ( the uniqueness of ) nron models .
[ p3 ] the @xmath34-scheme @xmath633 has the extension property . *
proof : * let @xmath121 be a healthy regular scheme over @xmath34 .
let @xmath659 be a morphism of @xmath24-schemes .
let @xmath660 be the pull back to @xmath661 of the universal principally polarized abelian scheme over @xmath608 ( via the composite morphism @xmath662 ) .
as the universal principally polarized abelian scheme over @xmath608 has a level-@xmath532 symplectic similitude structure for all @xmath27 prime to @xmath0 , the same holds for @xmath660 . from this and the nron
ogg shafarevich criterion of good reduction ( see [ blr , ch
. 7 , 7.4 , thm . 5 ] ) we get that @xmath663 extends to an abelian scheme @xmath664 over an open subscheme @xmath578 of @xmath121 which contains @xmath665 and for which we have @xmath666 .
thus @xmath664 extends to an abelian scheme @xmath667 over @xmath121 , cf . the very definition of a healthy regular scheme .
the polarization @xmath668 extends as well to a polarization @xmath669 of @xmath667 , cf .
[ mi2 , prop . 2.14 ] . moreover , each level-@xmath532 symplectic similitude structure of @xmath670 extends to a level-@xmath532 symplectic similitude structure of @xmath671 .
this implies that the composite of @xmath598 with the finite morphism @xmath672 extends uniquely to a morphism @xmath673 .
as @xmath121 is a regular scheme and thus also normal and as @xmath674 is a finite morphism , we get that the morphism @xmath673 factors uniquely through @xmath633 ( as it does so generically ) .
this implies that @xmath659 extends uniquely to a morphism @xmath675 of @xmath34-schemes . from this
the proposition follows .
[ p3.5 ] let @xmath121 be a regular scheme which is faithfully flat over @xmath566 .
then the following two properties hold : let @xmath578 be an open subscheme of @xmath121 which contains @xmath579 and the generic points of @xmath676 .
let @xmath677 be an abelian scheme over @xmath578 with the property that its @xmath0-divisible group @xmath678 extends to a @xmath0-divisible group @xmath582 over @xmath121 .
then @xmath677 extends to an abelian scheme @xmath63 over @xmath578 .
if @xmath121 is a @xmath0-healthy regular scheme , then it is also a healthy regular scheme .
* proof : * part ( b ) follows from ( a ) and the very definitions . to prove ( a ) we follow [ va2 , prop .
let @xmath542 be a positive integer relatively prime to @xmath0 . to show that @xmath63 exists
, we can assume that @xmath121 is local , complete , and strictly henselian , that @xmath578 is the complement of the maximal point @xmath679 of @xmath121 , that @xmath677 has a principal polarization @xmath680 , and that @xmath681 has a level-@xmath532 symplectic similitude structure @xmath682 ( see [ fc , ( i)-(iii ) of pp . 185 , 186 ] ) .
we write @xmath683 .
let @xmath684 be the principal quasi - polarization of @xmath678 defined naturally by @xmath680 ; it extends to a principal quasi - polarization @xmath685 of @xmath582 ( cf .
tate s theorem [ ta , thm .
4 ] ) . let @xmath39 be the relative dimension of @xmath677 .
let @xmath686 be the universal principally polarized abelian scheme over @xmath543 .
let @xmath687 be the morphism defined by @xmath688 .
we show that @xmath689 extends to a morphism @xmath690 .
let @xmath691 be prime to @xmath0 . from the classical purity theorem we get that the tale cover @xmath692\to u$ ] extends to an tale cover @xmath693 .
but as @xmath121 is strictly henselian , @xmath121 has no connected tale cover different from @xmath121 .
thus each @xmath694 is a disjoint union of @xmath695-copies of @xmath121 . from this
we get that @xmath681 has a level-@xmath696 symplectic similitude structure @xmath697 for every @xmath691 prime to @xmath0 .
let @xmath698 be a projective , toroidal compactification of @xmath543 such that ( cf .
[ fc , chap .
iv , thm . 6.7 ] ) : the complement of @xmath543 in @xmath698 has pure codimension 1 in @xmath698 and there exists a semi - abelian scheme over @xmath698 that extends @xmath699 . let @xmath700 be the normalization of the zariski closure of @xmath578 in @xmath701 .
it is a projective , normal , integral @xmath121-scheme which has @xmath578 as an open subscheme .
let @xmath47 be the complement of @xmath578 in @xmath700 endowed with the reduced structure ; it is a reduced , projective scheme over the residue field @xmath44 of @xmath679 .
the @xmath38-algebras of global functions of @xmath121 , @xmath578 , and @xmath700 are all equal to @xmath68 ( cf .
[ ma , thm .
38 ] for @xmath578 ) .
thus @xmath47 is a connected @xmath44-scheme , cf .
[ ha , ch .
iii , cor .
11.3 ] applied to @xmath702 .
let @xmath703 be the semi - abelian scheme over @xmath700 that extends @xmath704 ( it is unique , cf .
[ fc , chap .
due to the existence of the @xmath697 s , the nron
shafarevich criterion implies that @xmath703 is an abelian scheme in codimension at most 1 .
therefore , since the complement of @xmath543 in @xmath698 has pure codimension 1 in @xmath698 , it follows that @xmath703 is an abelian scheme .
thus @xmath689 extends to a morphism @xmath705 .
let @xmath706 .
tate s theorem implies that the principally quasi - polarized @xmath0-divisible group of @xmath707 is the pull - back @xmath708 of @xmath709 to @xmath700 .
hence the pull back @xmath710 of @xmath708 to @xmath47 is constant i.e. , it is the pull back to @xmath47 of a principally quasi - polarized @xmath0-divisible group over @xmath44 .
we check that the image @xmath711 of @xmath47 through @xmath712 is a point @xmath713 of @xmath543 .
since @xmath47 is connected , to check this it suffices to show that , if @xmath714 is the completion of the local ring @xmath715 of @xmath47 at an arbitrary point @xmath716 of @xmath47 , then the morphism @xmath717 defined naturally by @xmath712 is constant .
but as @xmath710 is constant , this follows from serre
tate deformation theory ( see [ me , chaps . 4 , 5 ] ) .
thus @xmath711 is a point @xmath713 of @xmath543 .
let @xmath718 be the local ring of @xmath543 at @xmath719 . because @xmath121 is local and @xmath700 is a projective @xmath121-scheme , each point of @xmath700 specializes to a point of @xmath47 .
hence each point of the image of @xmath712 specializes to @xmath719 and thus @xmath712 factors through the natural morphism @xmath720 . since @xmath68 is the ring of global functions of @xmath700 , the resulting morphism @xmath721 factors through a morphism @xmath722 .
therefore @xmath712 factors through a morphism @xmath723 that extends @xmath689 .
this ends the argument for the existence of @xmath42 .
we conclude that @xmath724 is an abelian scheme over @xmath121 which extends @xmath677 .
thus ( a ) holds . in ( the proofs of ) [ fc , ch .
iv , thms . 6.4 , 6.4 , and 6.8 ] was claimed that every regular scheme which is faithfully flat over @xmath566 is @xmath0-healthy regular as well as healthy regular . in
turns out that this claim is far from being true .
for instance , an example of raynaud gabber ogus ( see [ djo1 , sect . 6 ] ) shows that the regular scheme @xmath725/((xy)^{p-1}-p))$ ] is neither @xmath0-healthy nor healthy regular . here
@xmath726 is the ring of witt vectors with coefficients in a perfect field @xmath44 of characteristic @xmath0 .
based on proposition [ p3.5 ] ( b ) and a theorem of raynaud ( see [ ra , thm .
3.3.3 ] ) , one easily checks that if @xmath727 , then each regular scheme which is formally smooth over @xmath34 is a healthy regular scheme ( see [ va1 , subsubsect .
3.2.17 ] ) . in [ va2 , thm .
1.3 ] it is proved that the same holds provided @xmath728 . in [ va11 ]
it is proved that the same holds provided @xmath729 .
even more , in [ va11 , thm . 1.3 and cor .
1.5 ] it is proved that : [ thm1 ] * ( a ) * suppose that @xmath730 . each regular scheme which is formally smooth over @xmath34 is healthy regular if and only if the following inequality holds @xmath731 .
suppose that @xmath732 and @xmath728 .
then each regular scheme which is formally smooth over @xmath34 is healthy regular .
part ( a ) also holds for @xmath732 but this is not checked loc .
this is why above for @xmath732 we stated only one implication in the form of ( b ) . from theorem [ thm1 ] and propositions
[ p2 ] ( a ) and [ p3 ] we get the following answer to the problem 4.3 ( a ) : [ c1 ] suppose that @xmath731 and that @xmath633 is a regular scheme which is formally smooth over @xmath34 ( i.e. , and that @xmath634 is a smooth @xmath34-scheme ) .
then @xmath633 is the integral canonical model of @xmath623 over @xmath34 and it is uniquely determined up to unique isomorphism . thus also @xmath657 and @xmath658 are uniquely determined by @xmath622 and @xmath623 ( respectively ) and @xmath31 .
the integral canonical model of @xmath612 over @xmath566 is @xmath608 . in [ ne ]
it is showed that each abelian variety over the field of fractions @xmath328 of a dedekind domain @xmath582 has a nron model over @xmath582 . in [ blr ]
it is checked that many other closed subschemes of torsors of certain commutative group varieties over @xmath328 , have nron models over @xmath582 .
but most often , for @xmath733 the @xmath24-scheme @xmath622 can not be embedded in such torsors ; we include one basic example .
suppose that @xmath427 is isomorphic to @xmath734 for some positive integers @xmath735 and @xmath736 .
one has @xmath737 for each connected component @xmath738 of @xmath739 , cf .
[ pa , thm . 2 , 2.8 ( i ) ] .
the analytic lie group @xmath740 associated to the albanese variety @xmath741 is isomorphic to @xmath742/h_1(\mathcal{c},\bbb z)$ ] and therefore it is @xmath61 .
the @xmath16rank of @xmath289 is @xmath61 and this implies that @xmath622 is a projective @xmath24-scheme , cf .
[ bhc , thm . 12.3 and cor
. 12.4 ] . from the last two sentences one gets that @xmath738 is a connected
, projective variety over @xmath3 whose albanese variety @xmath741 is trivial .
thus @xmath738 can not be embedded into commutative group varieties over @xmath3 .
therefore the connected components of the @xmath24-scheme @xmath622 can not be embedded into torsors of commutative group varieties over @xmath24 .
based on the previous example we get that the class of nron models introduced below is new ( cf .
[ va4 , prop .
4.4.1 ] and [ va11 , thm .
4.3.1 ] ) . [ thm2 ]
suppose that for each prime @xmath0 that does not divide @xmath532 and for every prime @xmath31 of @xmath24 that divides @xmath0 , we have @xmath731 .
suppose that @xmath624 is a smooth , projective @xmath628$]-scheme .
then @xmath624 is the nron model of its generic fibre @xmath622 over @xmath628 $ ] ( and thus it is uniquely determined by @xmath622 and @xmath532 ) .
theorem [ thm2 ] provides a better answer to problem 4.3 ( a ) than corollary 1 , _ provided _ in addition we know that @xmath624 is a projective @xmath628$]-scheme .
we will use the notations listed at the beginning of section 5 . in this section
we study the smoothness of @xmath633 and @xmath743 .
let @xmath559 be a shimura pair such that the group @xmath744 is unramified .
let @xmath638 be a hyperspecial subgroup of @xmath745 , cf .
definition 4.1 ( f ) . in 1976
langlands conjectured the existence of a good integral model of @xmath746 over each local ring @xmath747 of @xmath405 at a prime @xmath748 of @xmath405 that divides @xmath0 ( see [ la , p. 411 ] ) ; unfortunately , langlands did not explain what good is supposed to stand for .
we emphasize that the assumption that @xmath744 is unramified implies that @xmath405 is unramified above @xmath0 ( see [ mi3 , cor .
4.7 ( a ) ] ) ; thus the index of ramification @xmath749 of @xmath748 is @xmath52 . in 1992
milne made the following conjecture ( slight reformulation made by us , as in [ va1 , conj .
3.2.5 ] ; strictly speaking , both langlands and milne stated their conjectures over the completion of @xmath747 ) .
[ conj1 ] there exists an integral canonical model of @xmath746 over @xmath747 . from the classical works of zink ,
langlands , and kottwitz one gets ( see [ zi1 ] , [ lr ] , and [ ko2 ] ) : [ thm3 ] * ( a ) * the milne conjecture holds if @xmath750 and @xmath454 is a shimura variety of pel type .
the milne conjecture holds if @xmath454 is a shimura variety of pel type of either a or c type .
the main results of [ va1 ] and [ va6 ] say ( see [ va1 , 1.4 , thm . 2 , and thm .
6.4.1 ] and [ va6 , thm .
1.3 ] ) : [ thm4 ] suppose one of the following two conditions holds : @xmath751 and @xmath746 is of abelian type ; @xmath0 is arbitrary and @xmath746 is a unitary shimura variety
. then the milne conjecture holds .
moreover , for each prime @xmath748 of @xmath405 that divides @xmath0 , the integral canonical model of @xmath746 over @xmath747 is a pro - tale cover of a smooth , quasi - projective @xmath747-scheme .
[ rmk2 ] see [ va2 , thm .
1.3 ] and [ va6 , thm .
1.3 ] for two corrections to the proof of theorem [ thm4 ] under the assumption that condition ( a ) holds .
more precisely : @xmath448 the original argument of faltings for the proof of proposition [ p3.5 ] was incorrect and it has been corrected in [ va2 ] ( cf .
proof of proposition [ p3.5 ] ) ; @xmath448 the proof of theorem [ thm4 ] for the cases when @xmath456 has simple factors isomorphic to @xmath752 for some @xmath753 was partially incorrect and it has been corrected by [ va6 , thm .
1.3 ] ( cf .
[ va6 , appendix , e.3 ] ) . to explain the four main steps of the proof of the ( a ) part of theorem 4
, we will sketch the argument why the assumptions that @xmath754 and that @xmath613 is a reductive group scheme over @xmath566 imply that @xmath625 is a formally smooth scheme over @xmath566 .
let @xmath755 be the ring of witt vectors with coefficients in an algebraic closure @xmath756 of @xmath757 .
let @xmath758 $ ] .
let @xmath759 and let @xmath760 be a lift of @xmath679 , where @xmath761 is a finite , discrete valuation ring extension of @xmath755 .
let @xmath762 be the index of ramification of @xmath761 .
let @xmath763 be the @xmath0-adic completion of the @xmath764$]-subalgebra of @xmath765 $ ] generated by @xmath766 with @xmath57 .
let @xmath767 be the frobenius endomorphism of @xmath763 which is compatible with the frobenius automorphism of @xmath755 and which takes @xmath359 to @xmath768 .
we have a natural @xmath755-epimorphism @xmath769 which maps @xmath359 to a fixed uniformizer @xmath770 of @xmath761 .
the kernel @xmath771 of @xmath772 has divided power structures and thus we can speak about the evaluation of @xmath1-crystals at the thickening @xmath773 defined naturally by @xmath772 .
we now consider the principally quasi - polarized , filtered @xmath1-crystal of the pull back @xmath774 to @xmath775 of the universal principally abelian scheme over @xmath608 ( via @xmath776 ) .
its evaluation at the thickening @xmath773 is of the form @xmath777 where @xmath778 is a free @xmath763-module of rank @xmath338 , @xmath779 is a direct summand of @xmath780 of rank @xmath39 , @xmath781 is a @xmath767-linear endomorphism of @xmath778 , @xmath782 is an integrable , nilpotent modulo @xmath0 connection on @xmath778 , and @xmath783 is a perfect alternating form on @xmath778 .
the generic fibre of @xmath784 is equipped with a family of hodge cycles whose de rham realizations belong to @xmath785/j_{\pi}m_{r_e}[{1\over p}])$ ] and lift naturally to define a family of tensors @xmath786 of @xmath785)$ ] .
the * first main step * is to show that , under some conditions on the closed embedding homomorphism @xmath787 and under the assumption that @xmath754 , the zariski closure in @xmath788 of the subgroup of @xmath789,\psi_{m_{r_e}})$ ] that fixes @xmath790 for all @xmath251 , is a reductive group scheme @xmath791 over @xmath763 .
see [ va1 , subsect .
5.2 ] for more details and see [ va1 , ( 5.2.12 ) ] for the fact that the reductive group scheme @xmath791 is isomorphic to @xmath792 .
the * second main step * is to show that we can lift @xmath779 to a direct summand @xmath793 of @xmath778 in such a way that @xmath794 and that for each element @xmath251 the tensor @xmath790 belongs to the @xmath267-filtration of @xmath785)$ ] defined by @xmath795 $ ] .
the essence of this second main step is the classical theory of infinitesimal liftings of cocharacters of smooth group schemes ( see [ dg , exp .
ix ] ) . due to the existence of @xmath793 ,
the morphism @xmath760 lifts to a morphism @xmath796 .
the reduction of @xmath797 modulo the ideal @xmath798 $ ] of @xmath763 is a lift @xmath799 of @xmath679 .
thus , by replacing @xmath155 with @xmath800 we can assume that @xmath801 .
see [ va1 , subsect .
5.3 ] for more details .
the * third main step * uses the lift @xmath799 of @xmath679 and faltings deformation theory ( see [ fa , sect . 7 ] ) to show that @xmath625 is formally smooth over @xmath566 at its @xmath756-valued point defined by @xmath679 .
see [ va1 , subsect .
5.4 ] for more details .
the * fourth main step * shows that for @xmath754 the mentioned conditions on the closed embedding homomorphism @xmath787 always hold , _ provided _ we replace @xmath474 by a suitable other injective map @xmath802 with the property that @xmath803 .
see [ va1 , subsects .
6.5 and 6.6 ] for more details .
[ rmk3 ] in [ va7 ] it is shown that theorem [ thm4 ] holds even if @xmath804 and @xmath454 is of abelian type . in [ ki ]
it is claimed that theorem [ thm4 ] holds for @xmath750 .
the work [ ki ] does not bring any new conceptual ideas to [ va1 ] , [ va6 ] , and [ va7 ] .
in fact , the note [ ki ] is only a variation of [ va1 ] , [ va6 ] , and [ va7 ] .
this variation is made possible due to recent advances in the theory of crystalline representations achieved by fontaine , breuil , and kisin .
we emphasize that [ ki ] does not work for @xmath732 while [ va7 ] works as well for @xmath732 .
to explain the three main steps of the proof of the ( b ) part of theorem 4 , in this subsection we will assume that @xmath559 is a simple , adjoint , unitary shimura pair of isotypic @xmath459 dynkin type . in [ de2 , prop .
2.3.10 ] it is proved the existence of an injective map @xmath474 of shimura pairs such that we have @xmath805 .
the * first step * uses a modification of the proof of [ de2 , prop .
2.3.10 ] to show that we can choose @xmath391 such that @xmath613 is the subgroup of @xmath650 that fixes a semisimple @xmath566subalgebra @xmath649 of @xmath645 ( see [ va6 , prop .
3.2 ] ) . let @xmath806 ; it is a hypersecial subgroup of @xmath807 .
the * second step * only applies theorem 3 to conclude that @xmath625 is a formally smooth @xmath629-scheme .
the * third step * uses the standard moduli interpretation of @xmath625 to show that the analogue @xmath808 of @xmath625 but for @xmath809 instead of for @xmath810 exists as well ( see [ va6 , thm . 4.3 and cor . 4.4 ] ) .
if we fix a @xmath566-monomorphism @xmath811 , then every connected component @xmath812 of @xmath813 will be isomorphic to the quotient of a connected component @xmath738 of @xmath814 by a suitable group action @xmath815 whose generic fibre is free and which involves a torsion group . the key point is to show that the action @xmath815 itself is free ( i.e. , @xmath812 is a formally smooth @xmath755-scheme ) .
if @xmath730 and @xmath0 does not divide @xmath816 , then the torsion group of the action @xmath815 has no elements of order @xmath0 and thus the action @xmath815 is free ( cf .
proof of [ va1 , thm .
6.2.2 b ) ] ) . in [ va6 ]
it is checked that the action @xmath815 is always free i.e. , it is free even for the harder cases when either @xmath732 or @xmath0 divides @xmath816 .
the proof relies on the moduli interpretation of @xmath625 which makes this group action quite explicit .
the cases @xmath732 and @xmath0 divides @xmath816 are the hardest due to the following two reasons . if @xmath732 and
if @xmath63 is an abelian variety over @xmath756 whose @xmath817-rank @xmath92 is positive , then the group @xmath818 is naturally a subgroup of the group of automorphisms of the formal deformation space @xmath819 of @xmath63 in such a way that the filtered dieudonn module of a lift @xmath820 of @xmath63 to @xmath821 depends only on the orbit under this action of the @xmath821-valued point of @xmath819 defined by @xmath820 . for a positive integer @xmath42 divisible by @xmath822 there
exist actions of @xmath823 on @xmath824 $ ] such that the induced actions on @xmath824[{1\over p}]$ ] are free .
[ thm5 ] we assume that either @xmath825 divides @xmath532 or @xmath23 is a unitary shimura variety . we also assume that the zariski closure of @xmath18 in @xmath826}$ ] is a reductive group scheme over @xmath544 $ ] . then @xmath624 is a smooth scheme over either @xmath628 $ ] or @xmath544 $ ] . *
proof : * let @xmath0 be an arbitrary prime that does not divide @xmath532 and let @xmath31 be a prime of @xmath24 that divides @xmath0 .
the group scheme @xmath613 is reductive .
thus the group @xmath585 is unramified .
this implies that @xmath24 is unramified over @xmath0 and that @xmath827 is a hyperspecial subgroup of @xmath588 .
therefore @xmath628 $ ] is an tale @xmath544$]-algebra . from this and proposition
[ p2 ] ( a ) we get that to prove the theorem it suffices to show that each scheme @xmath633 is regular and formally smooth over @xmath34 .
let @xmath828 be the integral canonical model of @xmath623 over @xmath34 , cf .
theorem [ thm4 ] . as @xmath828 is a healthy regular scheme ( cf .
theorem 1 ) , from proposition [ p3 ] we get that we have an @xmath34-morphism @xmath829 whose generic fibre is the identity automorphism of @xmath623 .
the morphism @xmath92 is a pro - tale cover of a morphism @xmath830 of normal @xmath34-schemes of finite type , where @xmath574 is a small enough compact , open subgroup of @xmath572 . from theorem [ thm4 ]
we get that @xmath831 is a quasi - projective @xmath34-scheme .
thus @xmath832 is a quasi - projective morphism between flat @xmath34-schemes .
as each discrete valuation ring of mixed characteristic @xmath833 is a healthy regular scheme , the morphism @xmath92 satisfies the valuative criterion of properness with respect to such discrete valuation rings . from the last two sentences we get that @xmath832 is in fact a projective morphism .
we consider an open subscheme @xmath834 of @xmath633 which contains @xmath623 and for which the morphism @xmath835 is an isomorphism . as @xmath828 has the extension property ( cf .
definition 4.1 ( d ) ) , from theorem [ thm4 ] we easily get that we can assume that @xmath834 contains the formally smooth locus of @xmath633 over @xmath34 . as @xmath832 is projective , from proposition 3 ( a )
we get that we can also assume that we have an inequality @xmath836 .
obviously we can assume that @xmath834 is @xmath574-invariant .
thus the projective morphism @xmath830 is an isomorphism above @xmath837 . to check that @xmath633 is a regular scheme which is formally smooth over @xmath34 it suffices to show that @xmath832 is an isomorphism . to check that @xmath832 is an isomorphism , it suffices to show that @xmath838 contains all points of @xmath831 of codimension @xmath52 ( this is so as the projective morphism @xmath832 is a blowing up of a closed subscheme of @xmath637 ; the proof of this is similar to [ ha , ch .
ii , thm . 7.17 ] ) .
let @xmath839 be the set of points of @xmath831 which are of codimension @xmath52 and which do not belong to @xmath840 .
we show that the assumption that the set @xmath839 is non - empty leads to a contradiction .
let @xmath738 be the open subscheme of @xmath831 that contains : ( i ) the generic fibre of @xmath831 and ( ii ) the union @xmath841 of those connected components of the special fibre of @xmath831 whose generic points are in @xmath839 .
the image @xmath842 has dimension less than @xmath841 and is contained in the non - smooth locus of @xmath637 . the morphism @xmath843 factors through the dilatation @xmath844 of @xmath637 centered on the reduced scheme of the non - smooth locus of @xmath637 , cf .
the universal property of dilatations ( see definition 4.1 ( g ) or [ blr , ch . 3 , 3.2 , prop . 3.1 ( b ) ] ) . but
@xmath844 is an affine @xmath637-scheme and thus the image of the projective @xmath637-scheme @xmath841 in @xmath844 has the same dimension as @xmath845 . by repeating the process
we get that the image of @xmath841 in a smoothening @xmath846 of @xmath637 obtaining via a sequence of blows up centered on non - smooth loci ( see [ blr , ch .
3 , thm . 3 of 3.1 and thm . 2 of 3.4 ] ) , has dimension @xmath847 and thus it has dimension less than @xmath841 .
but each discrete valuation ring of @xmath846 dominates a local ring of @xmath831 ( as @xmath832 is a projective morphism ) and therefore it is also a local ring of @xmath831 . as @xmath846 has at least one discrete valuation ring which is not a local ring of @xmath837 , we get that this discrete valuation ring is the local ring of some point in @xmath839 .
thus the image of @xmath841 in @xmath846 has the same dimension as @xmath841 . contradiction .
the @xmath3-scheme @xmath739 is projective if and only if the @xmath16rank of @xmath289 is @xmath61 , cf .
[ bhc , thm .
12.3 and cor . 12.4 ] . based on this morita
conjectured in 1975 that ( see [ mo ] ) : [ conj2 ] suppose that the @xmath16rank of @xmath289 is @xmath61
. then for each @xmath27 with @xmath542 , the @xmath628$]-scheme @xmath624 is projective .
[ conj3 ] let @xmath848 be an abelian variety over a number field @xmath257 .
let @xmath199 be the mumford
tate group of some extension @xmath63 of @xmath848 to @xmath3 .
if the @xmath16rank of @xmath334 is @xmath61 , then @xmath848 has potentially good reduction everywhere ( i.e. , there exists a finite field extension @xmath849 of @xmath257 such that @xmath850 extends to an abelian scheme over the ring of integers of @xmath849 ) . in [ mo ]
it is shown that conjectures 2 and 3 are equivalent .
we recall the argument for this .
suppose that conjecture 2 holds . to check that conjecture 3 holds
, we can replace @xmath257 by a finite field extension of it and we can replace @xmath848 by an abelian variety over @xmath257 which is isogeneous to it . based on this and [ mu , ch .
iv , 23 , cor . 1 ]
, we can assume that @xmath848 has a principal polarization @xmath851 . by enlarging @xmath257
, we can also assume that all hodge cycles on @xmath63 are pull backs of hodge cycles on @xmath848 ( cf .
[ de3 , prop . 2.9 and thm . 2.11 ] ) and that @xmath852 has a level-@xmath853 symplectic similitude structure . here
@xmath854 and @xmath855 are two distinct odd primes . by taking @xmath856 and @xmath206 to belong to @xmath331
, we can assume that @xmath852 is the pulls back of the universal principally polarized abelian schemes over @xmath857 and @xmath858 .
as @xmath857 and @xmath858 are projective schemes over @xmath859 $ ] and @xmath860 $ ] ( respectively ) , we get that @xmath852 extends to a principally polarized abelian scheme over the ring of integers of @xmath257 .
thus conjecture 2 implies conjecture 3 .
the arguments of the previous paragraph can be reversed to show that conjecture 3 implies conjecture 2 .
[ df5 ] we say @xmath848 ( resp .
@xmath17 ) has _ compact factors _ , if for each simple factor @xmath861 of @xmath334 ( resp . of @xmath289 )
there exists a simple factor of @xmath862 which is compact . in [ va4 ,
1.2 and cor .
4.3 ] it is proved that : [ thm6 ] suppose that @xmath17 ( resp .
@xmath848 ) has compact factors .
then conjecture [ conj2 ] ( resp .
[ conj3 ] ) holds .
let @xmath863 and @xmath864 .
we present different approaches to prove conjectures [ conj2 ] and [ conj3 ] developed by grothendieck , morita , paugam , and us .
suppose that there exists a prime @xmath0 such that the group @xmath865 is anisotropic ( i.e. , its @xmath866-rank is @xmath61 )
. then conjectures [ conj2 ] and [ conj3 ] are true ( see [ mo ] for the potentially good reduction outside of those primes dividing @xmath0 ; see [ pau ] for the potentially good reduction even at the primes dividing @xmath0 ) .
let @xmath644 be as in subsubsection 4.2.1 ( resp .
be the centralizer of @xmath199 in @xmath867 ) .
we assume that the centralizer of @xmath644 in @xmath645 ( resp . in @xmath867 )
is a central division algebra over @xmath16 . then conjecture [ conj2 ] ( resp .
[ conj3 ] ) holds ( see [ mo ] ) . by replacing @xmath257 with a finite field extension of it
, we can assume that @xmath848 has everywhere semi - abelian reduction .
let @xmath868 be a prime different from @xmath0 .
let @xmath869 be the @xmath118-adic tate
module of @xmath848 . as @xmath870-modules
we can identify @xmath871 . by replacing @xmath257 with a finite field extension of it
, we can assume that for each prime @xmath868 the @xmath118-adic representation @xmath872 factors through @xmath873 .
let @xmath797 be a prime of @xmath257 that divides @xmath0 .
if @xmath848 does not have good reduction at @xmath797 , then there exists a @xmath870- submodule @xmath369 of @xmath869 such that the inertia group of @xmath797 acts trivially on @xmath369 and @xmath874 and non - trivially on @xmath869 ( see [ sga7 , vol .
ix , thm . 3.5 ] ) .
this implies that @xmath873 has unipotent elements of unipotent class @xmath817 . in [ pau ]
is is shown that if @xmath873 has no unipotent element of unipotent class @xmath817 , then conjecture [ conj3 ] holds for @xmath63 . using this ,
conjectures [ conj2 ] and [ conj3 ] are proved in [ pau ] in many cases .
these cases are particular cases of either theorem [ thm6 ] or ( a ) .
we explain the approach used in [ va4 ] to prove theorem [ thm6 ] .
let @xmath875 be another abelian variety over @xmath257 .
we say that @xmath848 and @xmath875 are _ adjoint - isogeneous _ , if the adjoint groups of the mumford
tate groups @xmath199 , @xmath876 , and @xmath877 are isomorphic ( more precisely , the standard monomorphism @xmath878 induces naturally isomorphisms @xmath879 and @xmath880 ) . to prove conjecture [ conj3 ] for @xmath848 it is the same thing as to prove conjecture [ conj3 ] for @xmath875 .
based on this , to prove conjecture [ conj3 ] , one can replace the monomorphism @xmath881 by another one @xmath882 which is simpler .
based on this and subsection 7.1 , to prove conjectures [ conj2 ] and [ conj3 ] it suffices to prove conjecture [ conj2 ] in the cases when : the adjoint group @xmath289 is a simple @xmath16group ; if @xmath1 is a totally real number field such that @xmath883 , with @xmath884 an absolutely simple adjoint group over @xmath1 , then @xmath1 is naturally a @xmath16subalgebra of the semisimple @xmath16algebra @xmath644 we introduced in subsubsection 4.2.1 ; the monomorphism @xmath483 is simple enough .
suppose that @xmath17 has compact factors . by considering a large field that contains both @xmath128 and @xmath866 ,
one obtains naturally an identification @xmath885 .
thus we can speak about the @xmath0-adic field @xmath886 which is the factor of @xmath887 that corresponds ( via the mentioned identification ) to a simple , compact factor of @xmath888 .
the existence of such a simple , compact factor is guaranteed by definition [ df5 ] . to prove theorem [ thm6 ]
, it suffices to show that each morphism @xmath889 , with @xmath44 an algebraically closed field of prime characteristic @xmath0 that does not divide @xmath532 , extends to a morphism @xmath890)\to\mathcal{n}(n)$ ] .
we outline the argument for why the assumption that there exists such a morphism @xmath889 which does not extend , leads to a contradiction .
the morphism @xmath716 gives birth naturally to an abelian variety @xmath257 of dimension @xmath39 over @xmath891 .
we can assume that @xmath257 extends to a semi - abelian scheme @xmath892}$ ] over @xmath893 $ ] whose special fibre @xmath894 is not an abelian variety .
let @xmath895 be the maximal torus of @xmath896 .
the field @xmath1 acts naturally on @xmath897 , where @xmath898 is the abelian group of characters of @xmath895 .
let @xmath899 be an algebraic closure of @xmath891 .
let @xmath900 be the contravariant dieudonn module of @xmath901 .
due to ( ii ) , one has a natural decomposition of @xmath902-modules @xmath903,\phi)=\oplus_{j\in j } ( m_j,\phi).\leqno ( 23b)\ ] ] for each @xmath753 , the composite monomorphism @xmath904\hookrightarrow e_k[p^m]\hookrightarrow e_k$ ] lifts uniquely to a homomorphism @xmath905)_{k[[x]]}\to e_{k[[x]]}$ ] ( see [ dg , exp .
ix , thms .
3.6 and 7.1 ] ) which due to nakayama s lemma is a closed embedding .
this implies that we have a monomorphism @xmath905)_{k((x))}\hookrightarrow e[p^m]$ ] .
taking @xmath906 , at the level of dieudonn modules over @xmath899 we get an epimorphism @xmath907,\phi)\twoheadrightarrow ( x^*(t_k)\otimes_{\bbb z } b(k_1),1_{x^*(t_k)}\otimes p\sigma_{k_1})\leqno ( 23c)\ ] ] which ( due to the uniqueness part of this paragraph ) is compatible with the natural @xmath1-actions . here
@xmath908 is the frobenius automorphism of the field of fractions @xmath909 of the ring @xmath910 of witt vectors with coefficients in @xmath899 . from ( 23b ) and
( 23c ) we get that each @xmath911 has newton polygon slope @xmath52 . but based on ( iii ) one can assume that the @xmath1-isocrystal @xmath912 has no integral newton polygon slope . contradiction .
we will use the notations listed in the beginning of section 5 . let @xmath913 be the smooth locus of @xmath743 over @xmath34 ; its generic fibre is @xmath622 ( cf .
subsection 4.2 ) . in this section
we will study different stratifications of the special fibre @xmath914 of @xmath915 .
we begin with few extra notations .
let @xmath916 be the perfect alternating form on @xmath917 induced naturally by @xmath174 .
let @xmath918 be the flat , closed subgroup scheme of @xmath613 which fixes @xmath916 ; its generic fibre is a connected group @xmath919 .
let @xmath920 be a family of tensors as in subsection 3.4.2 .
we denote also by @xmath551 the pull back to @xmath624 of the universal principally polarized abelian scheme over @xmath543 ( to be compared with subsubsection 3.4.2 ) . by replacing @xmath532 with an integral power of itself , we can speak about a family @xmath921 of hodge cycles on @xmath922 obtained as in subsubsection 3.4.2 .
such a replacement is irrelevant for this section as we are interested in points of @xmath624 with values in @xmath44 , @xmath726 , or @xmath923 . here
@xmath44 is an algebraically closed field of characteristic @xmath0 , @xmath726 is the ring of witt vectors with coefficients in @xmath44 , and @xmath924 $ ] is the field of fractions of @xmath726 . let @xmath925 be the frobenius automorphism of @xmath44 , @xmath726 , or @xmath923 .
let @xmath926 .
let @xmath927 be a lift of @xmath679 , cf .
let @xmath928 .
let @xmath929 be the principally quasi - polarized dieudonn module of @xmath930 .
thus @xmath931 is a perfect alternating form on @xmath221 such that we have @xmath932 for all @xmath933 .
the @xmath925-linear automorphism @xmath934\tilde\to m[{1\over p}]$ ] extends naturally to a @xmath925-linear automorphism @xmath935)\tilde\to \mathcal{t}(m[{1\over p}])$ ] . the abelian variety @xmath936 is endowed naturally with a family @xmath647 of hodge cycles ( it is obtained from the family @xmath921 of hodge cycles on @xmath922 via a natural pull back process ) .
let @xmath937)$ ] be the de rham component of @xmath557 .
let @xmath938 be the hodge filtration of @xmath221 defined by the lift @xmath63 of @xmath939 .
we have @xmath940 .
let @xmath941 be the inverse of the canonical split cocharacter of @xmath942 defined in [ we , p. 512 ] .
it gives birth to a direct sum decomposition @xmath943 such that @xmath46 acts via @xmath944 trivially on @xmath267 and via the inverse of the identical character of @xmath46 on @xmath938 .
it is known that the element @xmath945 of @xmath946)$ ] is a de rham and thus also crystalline cycle .
if the abelian variety @xmath936 is definable over a number subfield of @xmath923 , then this result was known since long time ( for instance , see [ bl , thm .
( 0.3 ) ] ) .
the general case follows from loc .
cit . and [ va1 , principle b of 5.2.16 ] ( in the part of [ va1 , subsect . 5.2 ]
preceding the principle b an odd prime is used ; however the proof of loc .
applies to all primes ) .
the fact that @xmath945 is a crystalline cycle means that : let @xmath949 be the subgroup of @xmath950,\psi_m)$ ] that fixes @xmath945 for all @xmath251 .
let @xmath951 be the zariski closure of @xmath949 in @xmath952 ( or @xmath953 ) ; it is an affine , flat group scheme over @xmath726 .
we refer to the quadruple @xmath954 as the _ principally quasi - polarized @xmath1-crystal with tensors _ attached to @xmath955 .
it is easy to see that this terminology makes sense ( i.e. , @xmath945 depends only on @xmath926 and not on the choice of the lift @xmath927 of @xmath679 ) .
we note down that @xmath951 is uniquely determined by @xmath956 .
we refer to the quadruple @xmath957,\phi,(t_{\alpha})_{\alpha\in\mathcal{j}},\psi_m)\ ] ] as the _ principally quasi - polarized @xmath1-isocrystal with tensors _ attached to @xmath955 . from ( i ) and the functorial aspects of [ wi , p. 513
] we get that each tensor @xmath945 is fixed by @xmath944 .
this implies that : if @xmath960 is a point indexed by the elements @xmath961 of some set , then we will use the index @xmath961 to write down @xmath962 as well as @xmath963,\phi_i,(t_{i,\alpha})_{\alpha\in\mathcal{j}},\psi_{m_i})$ ] .
if @xmath964 does not factor through @xmath914 , then we define @xmath965 to be the principally quasi - polarized dieudonn module of @xmath966 .
similarly we define @xmath967,\phi_i,\psi_{m_i})$].@xmath968)$. we do not know if this family of tensors : ( i ) does not depend on the choice of the lift and ( ii ) can be used in subsections 8.4 and 8.5 in the same way as the families of tensors attached to $ k$-valued points of $ \mathcal{l}(n)_v^{\rm s}$.}$ ] before studying different stratifications of @xmath969 defined naturally by basic properties of the @xmath956 s , we will first present basic definitions on stratifications of reduced schemes over fields . for each field
@xmath118 that is either @xmath328 or an algebraically closed field which contains @xmath328 and that has countable transcendental degree over @xmath328 , a set @xmath972 of disjoint reduced , locally closed subschemes of @xmath973 is given such that each point of @xmath973 with values in an algebraic closure of @xmath118 factors through some element of @xmath972 ; if @xmath974 is an embedding between two fields as in ( a ) , then the reduced scheme of the pull back to @xmath855 of every member of @xmath975 , is an element of @xmath976 ; thus we have a natural pull back injective map @xmath977 . each element
@xmath978 of some set @xmath972 is referred as a _ stratum _ of @xmath970 ; we denote by @xmath979 the zariski closure of @xmath978 in @xmath973 .
if all maps @xmath980 s are bijections , then we identify @xmath970 with @xmath981 and we say @xmath970 is _ of finite type_. the _ strong purity property _ if for each field @xmath118 as in ( i ) above and for every stratum @xmath978 of @xmath972 , locally in the zariski topology of @xmath979 we have @xmath982 , where @xmath92 is some global function of @xmath979 and where @xmath983 is the largest open subscheme of @xmath979 over
which @xmath92 is an invertible function ; the _ purity property _ if for each field
@xmath118 as in ( i ) above , every element of @xmath972 is an affine @xmath973-scheme ( thus @xmath970 has the purity property if and only if each stratum of it is an affine @xmath2-scheme ) ; the _ weak purity property _
if for each field @xmath118 as in ( i ) above and for every stratum @xmath978 of @xmath972 , the scheme @xmath979 is noetherian and the complement of @xmath978 in @xmath979 is either empty or has pure codimension @xmath52 in @xmath979 . as the terminology suggests , the strong purity property implies the purity property and the purity property implies the weak purity property .
the converses of these two statements do not hold .
for instance , there exist affine , integral , noetherian schemes @xmath121 which have open subschemes whose complements in @xmath121 have pure codimension @xmath52 in @xmath121 but are not affine ( see [ va3 , rm . 6.3 ( a ) ] ) .
suppose that @xmath984 , that @xmath2 is an integral @xmath44-scheme , and that there exists a barsotti tate group @xmath582 of level @xmath52 over @xmath2 which generically is ordinary .
let @xmath985 be the stratification of @xmath2 of finite type which has two strata : the ordinary locus @xmath986 of @xmath582 and the non - ordinary locus @xmath987 of @xmath582 .
we have @xmath988 and @xmath989 .
moreover locally in the zariski topology of @xmath2 we have an identity @xmath990 , where @xmath92 is the global function on @xmath2 which is the determinant of the hasse witt map of @xmath582 .
thus the stratification @xmath985 has the strong purity property .
let @xmath991 be the stratification of @xmath657 of finite type with the property that two geometric points @xmath992 factor through the same stratum if and only if the newton polygons of @xmath993 and @xmath994 coincide .
in [ djo2 ] it is shown that @xmath991 has the weak purity property ( see [ zi2 ] for a more recent and nice proof of this ) .
* proof : * the stratification @xmath991 is the newton polygon stratification of @xmath657 defined by the @xmath1-crystal over @xmath657 associated to the @xmath0-divisible group of @xmath995 .
thus the theorem is a particular case of [ va3 , main thm .
let @xmath996 be the stratification of @xmath969 with the property that two geometric points @xmath997 factor through the same stratum if and only if there exists an isomorphism @xmath998 to be called a _
rational isomorphism_. * proof : * we use left lower indices to denote pulls back of @xmath1-crystals .
let @xmath118 be either @xmath32 or an algebraically closed field that contains @xmath32 and that has countable transcendental degree over @xmath32 .
let @xmath1000 be a stratum of @xmath999 contained in @xmath1001 .
let @xmath1002 be an irreducible component of @xmath1000 . to prove the part ( a ) it suffices to show that for each two geometric points @xmath1003 and @xmath1004 of @xmath1002 with values in the same algebraically closed field @xmath44 , there exists a rational isomorphism @xmath998 .
we can assume that @xmath44 is an algebraic closure of @xmath1005 and that @xmath1003 and @xmath1004 factor through the generic point and the special point ( respectively ) of a morphism @xmath1006)\to\mathcal{l}(n)^{\rm s}_{v , l}$ ] of @xmath118-schemes . here
@xmath359 is an independent variable .
we denote also by @xmath1003 and @xmath1004 , the @xmath44-valued points of @xmath1007)$ ] or of its perfection @xmath1007^{\text{perf}})$ ] defined naturally by the factorizations of @xmath1003 and @xmath1004 through @xmath42 .
let @xmath1008 be the frobenius lift of @xmath1009 $ ] that is compatible with @xmath1010 and that takes @xmath359 to @xmath768 .
let @xmath1011 be the principally quasi - polarized @xmath1-crystal over @xmath1012 $ ] of @xmath1013 .
thus @xmath761 is a free @xmath1009$]-module of rank @xmath338 equipped with a perfect alternating form @xmath1014 , @xmath1015 is a @xmath1008-linear endomorphism , and @xmath1016 is a connection .
let @xmath1017)$ ] be the de rham realization of the hodge cycle @xmath1018 on @xmath1019 , where @xmath1020)\to\mathcal{n}(n)_{w(l)}^{\rm s}$ ] is a lift of @xmath42 .
fontaine s comparison theory ( see [ fo ] ) assures us that there exists an isomorphism @xmath1021,(t_{1,\alpha})_{\alpha\in\mathcal{j}},\psi_{m_1})\tilde\to ( w^*\otimes_{\bbb q } b(k),(s_{\alpha})_{\alpha\in\mathcal{j}},\psi^*)$ ] .
based on this and [ ko1 ] we get that @xmath1022 is isomorphic to the pull back to @xmath44 of the principally quasi - polarized @xmath1-isocrystal @xmath1023 with tensors defined naturally by a principally quasi - polarized @xmath1-crystal @xmath812 with tensors over an algebraic closure @xmath1024 of @xmath32 . strictly speaking [ ko1 ]
uses a language of @xmath925-conjugacy classes of sets of the form @xmath1025 or @xmath1026 and not a language which involves polarizations and tensors ( and thus which involves @xmath925-conjugacy classes of sets of the form @xmath1027 , where @xmath1028 is an element whose image in @xmath1029 belongs to @xmath1030 ) ; but the arguments of [ ko1 ] apply entirely in the present principally quasi - polarized context which involves sets of the form @xmath1031 . here
@xmath1032 is @xmath1033 $ ] , where @xmath1034 is an arbitrary cocharacter whose extension to @xmath3 via an @xmath34-monomorphism @xmath1035 is @xmath300-conjugate to the cocharacters @xmath1036 with @xmath395 .
let @xmath1037 be @xmath812 but viewed only as an @xmath1-crystal over @xmath1024 .
let @xmath1038 be the @xmath726-lattice of @xmath1039 $ ] that corresponds naturally to @xmath1040 via an isomorphism @xmath1041 . from [ ka , thm .
2.7.4 ] we get the existence of an isogeny @xmath1042 , where @xmath1043 is an @xmath1-crystal over @xmath1012 $ ] whose extension to the @xmath1012$]-subalgebra @xmath1012^{\text{perf}}$ ] of @xmath44 is constant ( i.e. , it is the pull back of an @xmath1-crystal over @xmath1044 ) .
let @xmath1045 be the @xmath726-linear monomorphism that defines @xmath1046 .
we can assume that @xmath1047 is contained in @xmath1038 .
the inclusion @xmath1048 gives birth to a morphism @xmath1049 of @xmath1-crystals over @xmath44 .
it is the extension to @xmath44 of a morphism @xmath1050^{\text{perf}}}\to\mathcal{c}^-_{1,\bar l[[x]]^{\text{perf}}}$ ] , cf .
[ rr , lem .
3.9 ] and the fact that @xmath1051^{\text{perf}}}$ ] and @xmath1052^{\text{perf}}}$ ] are constant @xmath1-crystals over @xmath893^{\text{perf}}$ ] .
let @xmath1053^{\text{perf}}}\to\mathfrak{v}_{0,\bar l[[x]]^{\text{perf}}}$ ] be a morphism of @xmath1-crystal such that @xmath1054^{\text{perf}}}}$ ] for some @xmath1055 . by composing @xmath1056 with @xmath1057^{\text{perf}}}$ ]
we get an isogeny @xmath1058^{\text{perf}}}\to\mathfrak{v}_{\bar l[[x]]^{\text{perf}}}$ ] whose extension to @xmath44 is defined by the inclusion @xmath1059 . the isomorphism of @xmath1-isocrystals over @xmath890^{\text{perf}})$ ] defined by @xmath1060 times @xmath1061 takes @xmath1062 to @xmath1063 for all @xmath251 , as this is so generically .
thus @xmath1064 is an isogeny which when viewed as an isomorphism of @xmath1-isocrystals is @xmath1065 times an isomorphism @xmath1066 .
thus there exists a rational isomorphism @xmath1067 .
thus ( a ) holds .
[ rmk4 ] the proof of theorem [ thm8 ] ( a ) and ( b ) is in essence only a concrete variant of a slight refinement of [ rr , thm .
the only new thing it brings to loc .
, is that it weakens the hypotheses of loc .
( i.e. , it considers the newton point " of only one faithful representation which is @xmath1068 ) .
let @xmath1069 , where @xmath918 is as in the beginning of this section .
the group @xmath1070 is a connected group and we have a short exact sequence @xmath1071 we fix an @xmath34-embedding @xmath1072 .
let @xmath1073 be the set of cocharacters of @xmath1074 whose extension to @xmath3 are @xmath300-conjugate to any one of the cocharacters @xmath1036 with @xmath287 .
let @xmath1075 be the cocharacter introduced in the property ( ii ) of subsection 8.1 .
the set @xmath1077 of cocharacters of @xmath1074 formed by all cocharacters of the form @xmath1078 , with @xmath155 running through all @xmath726-valued points of @xmath1079 and with @xmath1080 running through all isomorphisms as in ( * * ) , is a @xmath1081-conjugacy class of cocharacters of @xmath1074 .
let @xmath1087\in\operatorname{sh}_{h(n)}(g,\mathcal{x})(\bbb c)$ ] be the complex point defined by the composite of the morphism @xmath1088 with @xmath1089 . under the fixed @xmath34-embedding @xmath1072
, we can identify : based on this we easily get that @xmath1096 is @xmath1097-conjugate to @xmath1098 . from this
we get that the cocharacter @xmath1078 belongs to @xmath1073 .
thus we have : @xmath1099 by composing @xmath1080 with an automorphism of @xmath1100 defined by an element of @xmath1081 , we can assume that in fact we have @xmath1101 ( cf .
( * * * * ) ) .
this implies that @xmath1080 gives birth to an isomorphism of the form @xmath1102 for some element @xmath1103 . for @xmath1104 ,
let @xmath1105 therefore @xmath1106 and moreover @xmath1107 let @xmath1108 it is a _ family _ of principally quasi - polarized @xmath1-crystals with tensors .
we emphasize that , due to ( * * * ) and ( * * * * ) , the isomorphism class of the family @xmath1109 depends only on @xmath969 and not on the choice of the element @xmath1082 of @xmath1077 .
[ df7 ] let @xmath753 . by the _ @xmath582-truncation of level @xmath42 ( or mod @xmath1110 ) _ of @xmath1111
we mean the reduction @xmath1112 $ ] of @xmath1113 modulo @xmath1110 ( here it is more convenient to use @xmath1114 instead of @xmath1115 ) . for @xmath1116 , by an inner isomorphism between @xmath1117 $ ] and @xmath1117 $ ] we mean an isomorphism @xmath1117\tilde\to\mathcal{e}_{g_1}[p^m]$ ] defined by an element of @xmath1118 .
[ rmk5 ] * ( a ) * statement ( * * * ) is a more general form of a _ conjecture of milne _ ( made in 1995 ) . in [ va8 ]
it is shown that ( * * * ) holds if either @xmath730 or @xmath732 and @xmath613 is a torus .
the particular case of this result when moreover @xmath613 is a reductive group scheme over @xmath566 , is also claimed in [ ki ] . in many cases
one can choose the cocharacter @xmath1082 in such a way that the quadruple @xmath845 is a canonical object of the family @xmath1109 .
for instance , if @xmath613 is a reductive group scheme over @xmath566 , then we have @xmath1119 and one can choose @xmath363 as follows .
let @xmath1120 be a borel subgroup scheme of @xmath1121 .
let @xmath1122 be a maximal torus of @xmath1120 .
let @xmath1123 .
let @xmath1124 be the unique cocharacter whose extension @xmath1082 to @xmath726 belongs to the set @xmath1073 , which factors through @xmath1125 , and through which @xmath46 acts on @xmath1126 via the trivial and the identical characters of @xmath46 ( cf .
[ mi3 , cor . 4.7 ( b ) ] ) . as pairs of the form @xmath1127
are @xmath589-conjugate , the isomorphism class of @xmath845 constructed via such a cocharacter @xmath363 does not depend on the choice of @xmath1127 .
thus @xmath845 is a canonical object of the family @xmath1109 .
* proof : * as [ va5 , def .
4.2.1 ] is a very long definition , the essence of its parts will be pointed out at the right time in this proof .
we emphasize that due to ( * * ) and ( 24 ) , the group scheme @xmath1114 is smooth over @xmath726 and therefore the statement of the theorem makes sense . let @xmath1129 be the direct sum decomposition such that @xmath46 acts through @xmath1082 trivially on @xmath1130 and via the inverse of the identical character of @xmath46 on @xmath1131 . to ( 25a ) corresponds a direct sum decomposition @xmath1132 of @xmath726-modules .
let @xmath1133 be the direct sum decomposition such that @xmath46 acts trough @xmath363 on @xmath1134 via the @xmath1135-th power of the identity character of @xmath46 .
thus we have an identity @xmath1136 the intersection being taken inside @xmath1137 .
let @xmath663 be the connected , smooth , unipotent subgroup scheme of @xmath1114 defined by the following rule : if @xmath47 is a commutative @xmath726-algebra , then @xmath1138 .
the smooth @xmath32-scheme @xmath914 is equidimensional of dimension @xmath394 . as @xmath363 belongs to @xmath1077 and thus to @xmath1073 , from formula
( 17 ) we get that the rank @xmath1139 of @xmath1140 is precisely @xmath394 .
thus the smooth @xmath32-scheme @xmath914 is equidimensional of dimension @xmath1139 . in other words , the axiom ( i ) of [ va5 , def .
4.2.1 ] holds .
let @xmath1141 be the completion of the local ring of @xmath1142 at its @xmath44-valued point defined by @xmath679 .
we fix an identification @xmath1143 $ ] .
let @xmath1008 be the frobenius lift of @xmath1141 which is compatible with @xmath925 and which takes @xmath108 to @xmath1144 for all @xmath1145 .
we have a natural morphism @xmath1146 which is formally tale .
the principally quasi - polarized filtered @xmath1-crystal over @xmath1147 of the pull back to @xmath1148 of @xmath551 is isomorphic to @xmath1149 where @xmath1150 is such that modulo the ideal @xmath1151 of @xmath1141 is the identity element of @xmath1081 and where @xmath1152 is an integrable , nilpotent modulo @xmath0 connection on @xmath1153 .
we have : for each element @xmath251 , the tensor @xmath1154)\otimes_{b(k ) } r_y[{1\over p}]=\mathcal{t}(m_0\otimes_{w(k ) } r_y[{1\over p}])$ ] is the de rham realization of the pull back to @xmath1155)$ ] of the hodge cycle @xmath1156 on @xmath922 and therefore it is annihilated by @xmath1152 ; the two properties ( i ) and ( ii ) hold as , up to @xmath726-automorphisms of @xmath1141 that leave invariant its ideal @xmath1151 , we can choose the morphism @xmath1157 to factor through a formally tale morphism @xmath1158 ( i.e. , we can choose @xmath1159 to be the universal element of the completion of @xmath663 ) . if @xmath613 is a reductive group scheme , then the fact that such a choice of @xmath1159 is possible follows from [ va1 , subsect .
the general case is entirely the same ( for instance , cf .
[ va7 , subsects .
3.3 and 3.4 ] ) .
we recall the standard argument that @xmath1152 annihilates each @xmath945 with @xmath251 .
we view @xmath1160 as a module over the lie algebra ( associated to ) @xmath1137 and accordingly we denote also by @xmath1152 the connection on @xmath1161)$ ] which extends naturally the connection @xmath1152 on @xmath1153 .
the @xmath1008-linear action of @xmath1162 on @xmath1153 extends to a @xmath1008-linear action of @xmath1162 on @xmath1161)$ ] .
for instance , if @xmath1163 and if @xmath1164 , then @xmath1165(a)\in m_0^*\otimes_{w(k ) } r_y[{1\over p}]$ ] maps @xmath1165(b)$ ] to @xmath1166 .
as @xmath1085 , @xmath1167 , and @xmath1159 fix @xmath945 , the tensor @xmath1168)$ ] is also fixed by @xmath1162 .
the connection @xmath1152 is the unique connection on @xmath1153 such that we have an identity @xmath1169=[h_y(g_y\phi_0\otimes\phi)]\otimes d\phi)\circ\nabla_y,\ ] ] cf .
[ fa2 , thm .
10 ] . from the last two sentences we get that @xmath1170(\nabla_y(t_{\alpha})).\ ] ] as we have @xmath1171 for all @xmath1172 , by induction on @xmath1055 we get that @xmath1173 $ ] . here
@xmath1174 is the @xmath0-adic completion of the sheaf of relative @xmath52-differential forms .
as @xmath1141 is complete with respect to the @xmath1151-topology , we have @xmath1175 . due to the property ( ii ) , the morphism @xmath1176 induces @xmath44-epimorphisms at the level of complete , local rings of residue field @xmath44 i.e. , it is a formal closed embedding at all @xmath44-valued points ( this is precisely the statement of [ va7 , part i , thm . 1.5 ( b ) ] ) . thus the axiom ( ii ) of [ va5 , def .
4.2.1 ] holds .
based on the property ( i ) and a standard application of artin s approximation theorem , we get that there exists an tale map @xmath1177 whose image contains the @xmath44-valued point of @xmath1178 defined naturally by @xmath679 and for which the following three properties hold : the principally quasi - polarized filtered @xmath1-crystal over @xmath1182 of the pull back to @xmath1183 of @xmath551 is isomorphic to @xmath1184 where @xmath1185 and where @xmath1186 is an integrable , nilpotent modulo @xmath0 connection on @xmath1153 which is versal at each @xmath44-valued point of @xmath1187 ; for each element @xmath251 , the tensor @xmath1154\otimes_{b(k ) } e_y^\wedge[{1\over p}])=\mathcal{t}(m_0\otimes_{w(k ) } e_y^\wedge[{1\over p}])$ ] is the de rham realization of the pull back to @xmath1188)$ ] of the hodge cycle @xmath1156 on @xmath922 and therefore it is annihilated by @xmath1189 .
let @xmath1190 be the tale map defined naturally by @xmath1191 .
let @xmath1192 be a finite set of @xmath44-valued points of @xmath969 such that we have an identity @xmath1193 this means that the axiom ( iii.a ) of [ va5 , def .
4.2.1 ] holds for the family of tale maps @xmath1194 .
let @xmath1195 be the maximal parabolic subgroup scheme of @xmath1196 that normalizes @xmath1131 .
let @xmath1197 ; it is a smooth subgroup scheme of @xmath1114 ( cf .
[ va5 , lem . 4.1.2 ] ) .
as @xmath1186 is versal at each @xmath44-valued point of @xmath1187 , we have : the fact that the axiom ( iii.d ) of [ va5 , def .
4.2.1 ] holds as well for @xmath1194 is only a particular case of faltings deformation theory [ fa , 7 , thm .
10 and rm .
i ) to iii ) after it ] , cf .
the versality part of the property ( iv ) .
more precisely , if @xmath1201 is such that the composite of @xmath1202 modulo @xmath0 with the quotient morphism @xmath1200 is formally tale , then there exists an @xmath726-automorphism @xmath1203 that leaves invariant the ideal @xmath1151 and for which the extension of ( 26a ) via @xmath1204 is isomorphic to @xmath1205 under an isomorphism defined by an element of @xmath1206 .
thus axioms ( i ) to ( iii ) of [ va5 , def .
4.2.1 ] hold i.e. , @xmath969 is a quasi shimura @xmath0-variety of hodge type relative to @xmath1109 in the sense of [ va5 , def .
4.2.1 ] .
we assume that properties ( * ) to ( * * * * ) of this section hold .
let @xmath42 be a positive integer . from theorem [ thm9 ] and [ va5 , cor .
4.3 ] we get that there exists a stratification @xmath1207 of @xmath969 with the property that two geometric points @xmath997 factor through the same stratum if and only @xmath1208 $ ] is inner isomorphic to @xmath1209 $ ] .
we call @xmath1207 as the level @xmath42 stratification of @xmath969 . among its many properties we list here
only three : [ p4 ] let @xmath118 be either @xmath32 or an algebraically closed field of countable transcendental degree over @xmath32 . let @xmath1210 be a stratum of @xmath999 which is a locally closed subscheme of @xmath1211 .
then we have : * proof : * the newton polygon of a @xmath0-divisible group @xmath582 over @xmath44 of codimension @xmath716 and dimension @xmath394 is uniquely determined by @xmath1215 $ ] , cf .
[ nv2 , thm .
thus the newton polygon of @xmath900 is uniquely determined by the inner isomorphism class of @xmath1216 $ ] . from this
the part ( a ) follows .
[ rmk6 ] * ( a ) * for pel type shimura varieties , the idea of level @xmath42 stratifications shows up first in [ we ] . the level @xmath52 stratifications generalize the ekedahl oort stratifications studied extensively by kraft , ekedahl , oort , wedhorn , moonen , and van der geer .
suppose that @xmath613 is a reductive group scheme and that the properties ( * ) and ( * * * ) of this section hold . as @xmath613 is a reductive group scheme , it is easy to see that the properties ( * * ) and ( * * * * ) hold as well . thus the level @xmath42 stratification @xmath1207 exists .
it is known that @xmath1217 has a finite number of strata ( see [ va10 , sect .
12 ] ) .
we continue to assume that properties ( * ) to ( * * * * ) of this section hold .
let @xmath1218 be the smallest positive integer such that for all elements @xmath1104 and @xmath1219 , the quadruples @xmath1111 and @xmath1220 are isomorphic .
the existence of @xmath1221 is implied by [ va3 , main thm .
[ lem1 ] we assume that the assumptions ( * ) to ( * * * * ) of this section hold .
let @xmath1116 .
the @xmath582-truncations of level @xmath42 of @xmath1222 and @xmath1223 are inner isomorphic if and only if there exists @xmath1224 such that we have @xmath1225 for some element @xmath1226 . due to the property ( i )
, for @xmath1229 we have an identity @xmath1230 we refer to @xmath1231 as the traverso stratification of @xmath969 . such stratifications were studied in [ tr1 ] to [ tr2 ] ( using the language of group actions ) , in [ oo ] ( using the language of foliations ) , and in [ va3 ] and [ va5 ] ( using the language of ultimate or traverso stratifications ) . based on theorem 9 ,
the next theorem is only a particular case of [ va5 , cor .
4.3.1 ( b ) ] .
we assume that @xmath391 is an isomorphism i.e. , we have an identification @xmath528 .
we have @xmath1232 and thus we will denote @xmath1221 by @xmath1233 .
we also assume that @xmath679 is a supersingular point i.e. , all newton polygon slopes of @xmath900 are @xmath1234 . the isomorphism class of @xmath1235 is uniquely determined by @xmath1236 $ ] , cf . [ nv1 , thm .
moreover , in general we can not replace in the previous sentence @xmath1236 $ ] by @xmath1237 $ ] ( cf .
[ nv1 , example 3.3 ] and the result [ va3 , prop . 5.3.3 ] which says that each principally quasi - polarized dieudonn module over @xmath44 is the one attached to a principally polarized abelian variety over @xmath44 ) .
therefore , the restrictions of @xmath815 and @xmath1238 to the ( reduced ) supersingular locus of @xmath1239 coincide and we have an inequality @xmath1240 based on traverso s isomorphism conjecture ( cf . [ tr3 , 40 , conj .
4 ] or [ nv1 , conj . 1.1 ] ) , one would be inclined to expect that @xmath1233 is in fact exactly @xmath39 . however , we are not at all at the point where we could state this as a solid expectation .
p. deligne , _ varits de shimura : interprtation modulaire , et techniques de construction de modles canoniques , automorphic forms , representations and @xmath64-functions ( oregon state univ . , corvallis , or , 1977 ) ,
part 2 , pp .
247289 , proc .
pure math .
, vol . * 33 * , amer .
soc . , providence , ri , 1979 .
r. langlands , _ some contemporary problems with origin in the jugendtraum , mathematical developments arising from hilbert problems ( northern illinois univ . ,
de kalb , il , 1974 ) , pp . 401418 , proc .
pure math .
* 28 * , amer .
soc . , providence , ri , 1976 .
_ j. s. milne , _ canonical models of ( mixed ) shimura varieties and automorphic vector bundles , automorphic forms , shimura varieties and l - functions , vol .
i ( ann arbor , mi , 1988 ) , pp . 283414 , perspectives in math .
, vol . * 10 * , acad . press , boston ,
ma , 1990 .
_ d. mumford , _
abelian varieties , tata inst . of fund .
research studies in math .
, no . * 5 * , published for the tata institute of fundamental research , bombay ; oxford univ .
press , london , 1970 ( reprinted 1988 ) .
_ a. grothendieck , _ groupes de monodromie en gomtrie algbrique .
i , sminaire de gomtrie algbrique du bois - marie 19671969 ( sga 7 i ) , lecture notes in math .
, vol . * 288 * , springer - verlag , berlin - new york , 1972 . _ j. tits , _ classification of algebraic semisimple groups , algebraic groups and discontinuous subgroups ( boulder , co , 1965 ) , pp . 3362 , proc .
pure math .
* 9 * , amer .
soc . , providence , ri , 1966 .
_ a. vasiu , _ good reductions of shimura varieties of hodge type in arbitrary unramified mixed characteristic , parts i and ii , available at http://xxx.arxiv.org/abs/0707.1668 and at http://arxiv.org/abs/0712.1572._ | we present a general and comprehensive overview of recent developments in the theory of integral models of shimura varieties of hodge type .
the paper covers the following topics : construction of integral models , their possible moduli interpretations , their uniqueness , their smoothness , their properness , and basic stratifications of their special fibres . #
1tempa#1tempaempty firstpage#1lastpage#1 @page=#1 firstpage1 # 1tempa#1tempaempty lastpage#1 lastpage0 pagerange10 october 26 , 2007 abelian and semiabelian schemes , mumford
tate groups , shimura varieties , hodge cycles , integral models , nron models , @xmath0-divisible groups , @xmath1-crystals , and stratifications . |
we are currently in an exciting era of high - precision cosmology , with most of the `` standard '' cosmological model parameters now known to within a few percent . in particular , the recent analysis of the cosmic microwave background ( cmb ) temperature fluctuations recorded by the _ planck _
satellite has found that the present day density of baryons , @xmath5 , contributes just @xmath19@xmath6per cent of the critical density @xcite , where @xmath20 is the hubble constant in units of 100 km s@xmath21 mpc@xmath21 . _
planck _ s determination of @xmath5@xmath6 , which is now limited by cosmic variance , is the most precise measure of this fundamental physical quantity for the foreseeable future ( when derived from the cmb ) . for a long time
( e.g. @xcite ) , it has been appreciated that a complementary measurement of @xmath5@xmath6 can be deduced from the relative abundances of the light elements that were created during big bang nucleosynthesis ( bbn ) . aside from protons , the only stable nuclei that were produced in astrophysically accessible abundances are @xmath22h ( a.k.a .
deuterium , d ) , @xmath23he , @xmath13he , and trace amounts of @xmath24li ( see @xcite for a comprehensive review ) .
in particular , considerable effort has been devoted to measuring the abundance of deuterium ( d / h ) , the mass fraction of @xmath13he ( @xmath14 ) , and the abundance of @xmath24li . of these
, the primordial abundance of deuterium is generally accepted as the best ` baryometer ' , owing to its sensitivity and monotonic relationship to @xmath5@xmath6 . the mass fraction of @xmath13he , on the other hand , is relatively insensitive to @xmath5@xmath6 , but depends quite strongly on the expansion rate of the early universe ( typically parameterized in non - standard models of bbn as an effective number of neutrino species , @xmath8 ) . at present
, this measurement is primarily limited by systematic uncertainties in converting the he and h emission lines of metal - poor ii regions into an estimate of @xmath14 ( see e.g. @xcite ) .
unlike @xmath14 , the @xmath24li / h ratio depends modestly on both @xmath5@xmath6 and @xmath8 .
however , the observationally inferred value for the `` primordial '' level , derived from metal - poor galactic halo stars @xcite , is discrepant by a factor of @xmath25 compared with the standard model predictions , becoming even more discrepant at the lowest metallicities @xcite . as pointed out recently by @xcite
( see also @xcite ) , _ precise _ measures of the primordial d / h ratio can provide interesting bounds on @xmath8 , when combined with a measure of @xmath5@xmath6 from the cmb .
the promise of this method was recently demonstrated by @xcite , who obtained the most precise measure of the d / h ratio to date , using a metal - poor ( [ fe / h]@xmath26 ) damped lyman @xmath1 system ( dla ) , seen in absorption against a bright background quasar .
the prospect of measuring the d / h ratio in gas clouds seen in absorption along the line - of - sight to a high - redshift quasar was first pointed out by @xcite .
this vision was only realized much later , with the advent of 8 - 10 m class telescopes equipped with high resolution echelle spectrographs @xcite . since these first discoveries ,
a handful of additional cases have been identified ( see @xcite for the ` _ _ top - ten _ _ ' list ) , including one system that appears to be chemically pristine @xcite .
one lingering concern with this prime set of measurements is that the dispersion in the reported d / h measures is significantly larger than the quoted errors ( first noted by @xcite ) .
indeed , a simple @xmath27 test reveals that the errors for all d / h measurements would need to be scaled upwards by @xmath28 , if the observed dispersion were due to chance alone .
an alternative possibility , if the quoted random errors are in fact realistic , is that the analyses may suffer from small systematic biases that are now important to recognize and correct for .
this concern has prompted us to identify the rare handful of systems which afford _ precise _ measurements of the primordial abundance of deuterium together with realistic error estimates .
this effort is part of our ongoing survey to study the chemistry of the most metal - poor dlas , described in more detail in @xcite and @xcite . in this paper , we present the _ precision sample _ of primordial deuterium abundance measurements , consisting of the current best estimates of d / h in qso absorption line systems .
we also report a dla where a new , precise measurement of the primordial deuterium abundance could be obtained .
all of these systems have been reanalyzed in a self - consistent manner , taking into account the dominant sources of systematic uncertainty . in the following section , we provide the details of the observations and data reduction procedures for the newly observed dla , and
discuss the selection criteria we have adopted to define the precision sample . in section [ sec : profanalysis ] , we describe the analysis of the new system with deuterium absorption , and estimate its d / h ratio . in section [ sec :
cosmology ] we derive the value of ( d / h)@xmath3 from the precision sample , discuss its cosmological implications , and consider current limitations in using d / h abundance measurements for cosmology .
section [ sec : newphysics ] deals with the implications of our results for physics beyond the standard model , considering in particular bounds on the effective number of neutrino families and on the lepton asymmetry .
finally , in section [ sec : conc ] , we summarize the main findings from this work .
most newly discovered dlas are now identified using the low - resolution spectra provided by the sloan digital sky survey ( sdss ) .
recent searches for dlas have yielded several thousand systems @xcite . at the modest resolution ( @xmath30 ) and
signal - to - noise ratio ( s / n@xmath31 ) of the sdss spectra , the metal absorption lines of the most metal - poor dlas are unresolved and undetected .
follow - up , high spectral resolution observations are thus required to pin - down their chemical abundances ( see @xcite ) .
@xcite were the first to recognize the very metal - poor dla at redshift @xmath32 towards the @xmath33 quasar sdss j1358@xmath296522 .
on the basis of their intermediate resolution observations ( 60 km s@xmath21 full - width at half maximum ; fwhm ) with the keck echellette spectrograph and imager , these authors concluded that this dla has a metallicity of [ fe / h ] @xmath34 , and is thus among the most metal - poor known @xcite .
we re - observed this qso with the keck high resolution echelle spectrometer ( hires ; @xcite ) using the red cross - disperser to perform an accurate chemical abundance analysis @xcite .
the hires data pinned down the metallicity of the dla ( [ fe / h]@xmath350.03 ) and provided the first indication for the presence of a handful of resolved i absorption lines .
taking advantage of the high efficiency of the hires uv cross - disperser at blue wavelengths , we then performed dedicated hires observations with the goal of obtaining a precise measure of the deuterium abundance .
our observations included a total of @xmath36s on 2012 , may 11 , ( divided equally into nine exposures ) and an additional @xmath37s on 2013 , may 3 and 5 .
an identical instrument setup was employed for both runs ; we used the c1 decker ( @xmath38 , well - matched to the seeing conditions of @xmath39 ) , which delivers a nominal spectral resolution of @xmath40 ( @xmath41 km s@xmath21 fwhm ) for a uniformly illuminated slit .
our setup covers the wavelength range @xmath42 , with small gaps near 4550 and 5550due to the gaps between the hires detector chips .
the exposures were binned on - chip @xmath43 .
the data were reduced following the standard procedures of bias subtraction , flat - fielding , order definition and extraction .
the data were mapped onto a vacuum , heliocentric wavelength scale from observations of a thar lamp which bracketed the science exposures . to test the quality of the data reduction
, we reduced the data using two pipelines : makee which is maintained by t. barlow , and hiredux which is maintained by j. x. prochaska .
we found that the skyline subtraction performed by hiredux was superior to that of makee , and substantially improved the s / n of the spectra at blue wavelengths .
this was very important for the removal of telluric features near the redshifted i and i transitions of the dla .
all data analyzed and presented herein for j1358@xmath296522 were reduced with hiredux .
the individual orders were combined with uves_popler , which is maintained by m. t. murphy . as a final step , we flux - calibrated the echelle data by comparison to the sdss data .
the combined s / n ratio of the final spectrum per 2.5 km s@xmath21 pixel is @xmath44 near 4000 , @xmath45 at 5000 and @xmath46 near 6000 .
the reduced , flux - calibrated spectrum is presented in figure [ fig : fluxqso ] , where the red tick marks indicate the wavelengths of the redshifted i lyman series transitions . in this section ,
we outline the strict set of rules that we have used to define the precision sample .
our goal is to identify the small handful of systems currently known where the most accurate and precise measures of d / h can potentially be obtained .
the set of restrictions we applied are as follows : * we require that the i ly@xmath1 absorption line must exhibit lorentzian damping wings .
this criterion is satisfied for absorption line systems with log @xmath47/@xmath48 .
such absorption lines lie on the damping regime of the curve - of - growth , where @xmath47 can be derived independently of the cloud model ; the damping wings uniquely determine the total i column density . * due to the power of the i ly@xmath1 transition in establishing the total i column density , we require that the wings of this transition are not strongly blended with nearby , unrelated , strong i absorption systems . if there exist additional strong i absorption systems nearby , we impose that the edges of their ly@xmath1 absorption troughs should be separated by at least 500 km s@xmath21 , and that all such absorption systems should be modelled simultaneously . * at least two , resolved and apparently unblended , optically thin transitions of i must be available .
this ensures that the total i column density can be measured accurately , and independently of the cloud model . *
the data must have been obtained with a high resolution echelle spectrograph , with @xmath49 ( i.e. @xmath50 km s@xmath21 ) , to resolve the broadening of the i lines , and recorded at s / n @xmath51 per pixel ( @xmath25 km s@xmath21 ) at both ly@xmath1 and the weakest i absorption line used in the analysis . * several unblended metal lines with a range of oscillator strengths for a given species ( ideally oi and siii ) must be present if there is only one optically thin i transition in order to determine the velocity structure of the absorbing cloud , and ensure that the presence of any partly ionized gas ( which may contribute to the i column density ) is accurately modelled . however , if there exist at least two optically thin i transitions , metal lines are not strictly required .
of course , we are also limited by data that we , and others , have access to .
data that are not publicly available on archives could not be used in this study . in total , there are four systems in the literature that meet the above criteria .
we now give a brief description of these systems .
keck+hires observations of the qso hs0105@xmath291619 were obtained as follows : 1800s in 1999 ( program i d : u05h , pi : a. wolfe ) , 85000s in 19992000 by omeara et al .
( 2001 ) , and 21500s in 2005 ( program i d : g10h , pi : d. tytler ) .
the data obtained by @xcite are not available from the keck observatory archive .
thus , we have reanalysed only the last dataset which , in any case , was obtained with a much improved hires detector .
as reported by @xcite , the spectrum shows nine clean i transitions in the lyman series together with four optically thin i absorption lines ; the absorption is confined to a single velocity component at @xmath53 . the first ions exhibit slightly asymmetric profiles , arising from ionized gas blueshifted by @xmath54 km s@xmath21 relative to the main component ( see figure [ fig : hs0105p1619 ] ) .
the i column density of this ionized component contributes just 0.01 dex to the total column density @xmath55(hi)/@xmath56 .
the data and best - fitting model are shown in figure [ fig : hs0105p1619 ] .
the quasar q0913@xmath29072 intersects a dla with @xmath55(hi)/@xmath58 at @xmath57 .
q0913@xmath29072 was observed with the ultraviolet and visual echelle spectrograph ( uves ) on the european southern observatory s ( eso ) very large telescope ( vlt ) facility in 2002 ( program i d : 68.b-0115(a ) , pi : p. molaro ) , and then again in 2007 to obtain uv coverage down to the lyman limit ( program i d : 078.a-0185(a ) , pi : m. pettini ) , resulting in a total exposure time of 77500s .
the details of this system s chemical composition were reported by @xcite ( see also @xcite ) , and the analysis of the di lines was presented by @xcite ; with [ o / h]@xmath59 this is currently the most metal - poor dla known with resolved i lines .
there is a small column density of ionized gas at velocities near that of the dla ; we have carefully modelled its contribution to the overall absorption in the metal lines and the i lyman series ( see figure [ fig : q0913p072 ] ) .
since q0913@xmath29072 does not have an sdss spectrum from which we could perform a flux calibration , we obtained a set of vlt+fors2 ( focal reducer / low dispersion spectrograph 2 ) exposures from the vlt archive to flux calibrate the uves spectrum .
( program i d : 70.a-0425(c ) , pi : l. wisotzki ) .
the fors2 data were reduced following standard procedures , using the software routines from the eso fors2 data reduction pipeline .
the uves data and the corresponding model fits are shown in figure [ fig : q0913p072 ] .
the chemical properties of this dla were discussed by @xcite , while @xcite reported the analysis of the deuterium lines .
this system currently holds the record for the most precise determination of the d / h ratio .
the analysis technique developed in @xcite is the same as that employed here ( see section [ sec : analysis ] ) , except for two aspects : ( 1 ) in the present work all qso spectra were analysed `` blind , '' i.e. , without knowledge of the value of d / h until the line fitting procedure was concluded ; and ( 2 ) we now allow for an arbitrary continuum fitting with legendre polynomials , rather than the combination of power - law continuum and gaussian emission lines used by @xcite . to maintain consistency with the other dlas investigated here
, we have reanalyzed this system with the new prescription for the continuum definition , and the analysis was performed blind .
we found that the final value of d / h from this new analysis deviated by just 0.005 dex from that reported by @xcite , with an identical estimate for the error .
this small shift is due to our new approach for dealing with the systematic uncertainty in the continuum level .
the data for j1558@xmath610031 consist of a 11300s keck+hires spectrum obtained in 2006 ( program i d : u152hb , pi : j. x. prochaska ) , first reported by @xcite . with @xmath55(hi)/@xmath63 , this @xmath62 dla is the highest i column density system currently known with at least one optically thin iabsorption feature
. there exists another metal - poor dla along this sightline , at @xmath64 , that also exhibits damped ly@xmath1 absorption ( log@xmath47/@xmath65 = 19.726@xmath66 ) .
this second system contains a single absorption component , as evidenced by several siii absorption lines ( see bottom - right panel of figure [ fig : j1558m0031 ] ) .
although i absorption is only seen in the higher redshift dla , we include the i and metal lines of both dlas in the fitting procedure , since the damping wings of the two dlas overlap slightly .
portions of the hires spectrum , together with the best - fitting model , are reproduced in figure [ fig : j1558m0031 ] .
unlike the @xcite analysis , which uses data acquired with both magellan+mike and keck+hires , our analysis exclusively uses the keck+hires data described above .
[ cols="^,>,>,>,^,^,^,^ " , ] @xmath67we adopt the solar value @xmath68 @xcite
. + @xmath69references ( 1 ) this work , ( 2 ) @xcite , ( 3 ) @xcite , ( 4 ) @xcite , ( 5 ) @xcite , ( 6 ) @xcite , ( 7 ) @xcite .
+ [ tab : precision ] including the new d / h case reported herein , there are just five metal - poor absorption line systems in which a _ precise _ measure of the primordial abundance of deuterium can be obtained , based on the criteria outlined in section [ sec : litsyst ] .
relevant details of these systems are collected in table [ tab : precision ] .
the four dlas from the literature were all reanalyzed in an identical manner to that described above ( section 3 ) for j1358@xmath296522 .
in particular , we adopted the same blind analysis strategy and marginalized over the important systematic uncertainties . we refer to this sample of five high quality measurements as the _ precision sample_. in table [ tab : precision ]
, we provide a measure of the _ total _
i column density , along with the associated error .
many of our systems contain more than one component in i and the column density estimates for these multiple components are correlated with one another .
to calculate the error on the total i column density , we have drawn 10000 realizations of the component column densities from the covariance matrix .
we then calculated the total column density for each realization ; in table [ tab : precision ] , we provide the mean and @xmath70 error derived from this monte carlo analysis .
we consider the five measures of i / i in these dlas as five independent determinations of the primordial abundance of deuterium , ( d / h)@xmath3 , for the following reasons : ( 1 ) we are not aware of any physical mechanism that would alter the ionization balance of d compared to h. thus , to our knowledge , i / i@xmath71d / h ; ( 2 ) the degree of astration of d ( i.e. its destruction when gas is turned into stars ) is expected to be negligible at the low metallicities ( [ o / h]@xmath72 ) of the dlas considered here ( e.g. , see figure 2 of romano et al .
2006 ) ; thus ( d / h)@xmath73 = ( d / h)@xmath3 .
( 3 ) the lack of dust in metal - poor dlas makes it extremely unlikely that selective depletion of d onto grains occurs in the cases considered here [ it has been proposed that such a mechanism may be responsible for the local variations in ( d / h)@xmath74see linsky et al .
2006 ] ; ( 4 ) the five dlas sample entirely independent sites in the distant universe .
as can be seen from table [ tab : precision ] and figure [ fig : dhsample ] , the five measures of d / h in the precision sample are in very good mutual agreement and the dispersion of the measurements is consistent with the errors estimated with our improved analysis .
a @xmath27 test indeed confirms that the five measurements are consistent within @xmath17 of being drawn from a single value of d / h .
we can therefore combine the five independent determinations of ( d / h)@xmath73 to deduce the weighted mean value of the primordial abundance of deuterium : @xmath75 @xmath76 this value of ( d / h)@xmath3 is not markedly different from other recent estimates @xcite , but its precision is significantly better than achieved in earlier papers that considered a more heterogeneous set of @xmath77 determinations . for completeness , we have recalculated the weighted mean for all the known d / h measurements listed in table 2 of @xcite , after updating the d / h values of the systems we have reanalyzed here .
the resulting weighted mean value of the primordial deuterium abundance is ( d / h)@xmath3@xmath78 .
this compares very well with the value derived from the precision sample ( eqs .
[ eqn : primdh_log ] and [ eqn : primdh ] ) .
perhaps this is not surprising , since the literature systems that did not meet our selection criteria ( see section [ sec : criteria ] ) have larger uncertainties , and thus their contribution to the weighted mean value of d / h is relatively low . using the most up - to - date calculations of the network of nuclear reactions involved in bbn
, the primordial abundance of deuterium is related to the cosmic density of baryons ( in units of the critical density ) , @xmath5 , via the following relations ( @xcite ; g. steigman 2013 , private communication ) : @xmath79 @xmath80 where @xmath81 , @xmath82^{1/2}$ ] is the expansion factor and @xmath83 is the neutrino degeneracy parameter ( related to the lepton asymmetry by equation 14 from @xcite ) . the rightmost term in eq .
[ eqn : dhptoetad ] represents the current 3% uncertainty in the conversion of ( d / h)@xmath3 to @xmath84 due to the uncertainties in the relevant nuclear reactions rates ( see section [ sec : limitation ] ) . for the standard model , @xmath8@xmath85 and @xmath86 . in this case
, the precision sample of d / h measurements implies a cosmic density of baryons : @xmath87 where we have decoupled the error terms from our measurement ( i.e. the random error term ) and the systematic uncertainty in converting the d abundance into the baryon density parameter .
as can be seen from figure [ fig : dhsample ] , this value of @xmath5@xmath6 is in excellent agreement with that derived from the analysis of the cmb temperature fluctuations measured by the _
satellite @xcite : @xmath88 in the era of high - precision cosmology , we feel that it is important to highlight the main limitations affecting the use of ( d / h)@xmath3 in the estimation of cosmological parameters . as can be seen from eq .
[ eq : dhobary ] , the main source of error is in the conversion of ( d / h)@xmath3 to the baryon density parameter ( @xmath84 , and hence @xmath5@xmath6 ) . in large part , this systematic uncertainty is due to the relative paucity of experimental measures for several nuclear cross - sections that are important in the network of bbn reactions , particularly deuteron deuteron reactions and the @xmath89(@xmath90)@xmath23he reaction rate at the relevant energies @xcite . since these studies , estimates for the deuteron
deuteron reaction cross - sections @xcite have improved and their contribution to the error budget has been reduced . the main lingering concern involves the reaction rate @xmath89(@xmath90)@xmath23he , for which only a single reliable dataset of the _ s_-factor is currently available in the relevant energy range @xcite .. it has nothing to do with the expansion factor @xmath91 in eq . [ eqn : etad ] . ]
the concern for @xmath89(@xmath90)@xmath23he is made worse by the fact that the theoretical and experimental values of the _ s_-factor do not agree .
this paucity of data , in addition to the difficulties of obtaining an accurate and precise measure of @xmath89(@xmath90)@xmath23he at bbn energies , led @xcite to adopt a theoretical curve for the _ s_-factor between 50500kev .
this has resulted in the improved conversion given in eq .
[ eqn : dhptoetad ] .
it is worth pointing out here that a further reduction by a modest factor of two in the uncertainty of the conversion from ( d / h)@xmath3 to @xmath84 would make the precision of @xmath5@xmath6(bbn ) from the dla precision sample comparable to that of @xmath5@xmath6(cmb ) achieved by the _
planck _ mission ( cf .
[ eq : dhobary ] and [ eq : dhobary_cmb ] ) .
in this section we combine the results presented here with those of the cmb analysis by the _ planck _ collaboration to place bounds on some parameters denoting new physics beyond the standard model of cosmology and particle physics .
in particular , we use the markov - chain monte carlo chains from the combined analysis of the _ planck _ temperature fluctuations @xcite , the low - multipole _ wmap _ polarization data ( @xmath92 ; @xcite ) , and high - resolution cmb data ( see @xcite for an appropriate list of references ) . throughout
, we refer to this combined dataset as _ _
planck__+wp+highl , for consistency with the work by the @xcite . in
what follows , we have assumed that @xmath8 and the baryon - to - photon ratio remained unchanged from the epoch of bbn to the epoch of recombination .
it has long been known that the mass fraction of @xmath13he , @xmath14 , is potentially a very sensitive probe of additional light degrees of freedom @xcite .
unfortunately , systematic uncertainties in the determination of @xmath14 have limited its use as an effective probe of @xmath8 ( see figure 8 of @xcite ) .
however , @xcite ( see also @xcite ) have recently highlighted the potential of using precise measurements of the primordial deuterium abundance in conjunction with observations of the cmb to place a strong , independent bound on @xmath8 . in the left panel of figure [ fig : neff ] , we show the current @xmath70 and @xmath17 confidence contours for @xmath8 and @xmath5@xmath6 derived from the _ _ planck__+wp+highl cmb analysis added as a free parameter ( see section 6.4.4 of @xcite ) . ] ( green ) , and from the bbn - derived ( d / h)@xmath3 reported here ( blue ) .
the combined confidence bounds are overlaid as red contours . in
what follows , it is instructive to recall that the cmb - only bounds are @xcite : @xmath93 @xmath94 ( note that solving simultaneously for @xmath5@xmath6and @xmath8 leads to a slightly different best - fitting value of @xmath5@xmath6 than that obtained for the standard model ; cf .
[ eq : dhobary_cmb ] and [ eq : dhobary_cmb_ns ] ) . for comparison , from the joint bbn+cmb analysis
we deduce : @xmath95 @xmath96 thus , combining ( d / h)@xmath3 with the cmb does not significantly change the uncertainty in @xmath5@xmath6 , but does reduce the error on @xmath8 by @xmath46 per cent .
the uncertainty on @xmath8 could be reduced further by an improvement in the cross - section of the @xmath89(@xmath90)@xmath23he(see right panel of figure [ fig : neff ] , and section [ sec : limitation ] ) .
based on the current bound on @xmath8 from cmb+(d / h)@xmath3 , we can nevertheless rule out the existence of an additional ( sterile ) neutrino ( i.e. @xmath8 = 4.046 ) at @xmath97 confidence ( i.e. @xmath98 ) , provided that @xmath8 and @xmath12 remained unchanged between bbn and recombination .
however , as noted recently by @xcite , if the cmb photons are heated after the neutrinos have decoupled [ for example , by a weakly interacting massive particle that annihilates to photons ] , @xmath8 will be less than @xmath99 for three standard model neutrinos ; a sterile neutrino can in principle exist even when @xmath8 @xmath100 .
looking to the future , @xmath14 has contours that are almost orthogonal to those of the cmb and ( d / h)@xmath3 ( see e.g. @xcite ) .
thus , measures of @xmath14 that are not limited by systematic uncertainties could potentially provide a very strong bound , when combined with ( d / h)@xmath3 , on the number of equivalent neutrinos during the epoch of bbn , independently of cmb observations .
following improvements in the hei emissivity calculations ( porter et al .
2012 , 2013 ) , there have been two recent reassessments of @xmath101 , by @xcite and by @xcite respectively .
@xcite find @xmath102 ; this value includes a small correction of @xmath103 to reflect the fact that their _ cloudy _ modelling overpredicts @xmath101 by this amount at the lowest metallicities see their figure 7(a ) .
@xcite find @xmath104 from the average of all the low metallicity iiregions in their sample .
thus , the values deduced by these two teams are in good mutual agreement ; in the analysis that follows we adopt @xmath102 from @xcite . using the following conversion relation for @xmath14 (
steigman 2012 ; g. steigman 2013 , private communication ) : @xmath105 @xmath106 combined with the @xcite measure of @xmath14 , we derive the following bbn - only bound on the baryon density and the effective number of neutrino species : @xmath107 @xmath108 the corresponding contours are shown in figure [ fig : ypdhcont ] .
thus , it appears that even with the most recent reappraisals of the primordial abundance of @xmath13he by @xcite and @xcite , there is better agreement ( within the standard model ) between ( d / h)@xmath3 and the cmb , than between ( d / h)@xmath3 and @xmath14 . and @xmath17 confidence contours ( dark and light shades respectively ) for @xmath8 and @xmath5@xmath6 derived from the primordial deuterium abundance ( blue ) , the primordial he mass fraction ( green ) , and the combined confidence contours ( red ) .
dashed and dotted lines indicate the hidden contour lines for ( d / h)@xmath3 and @xmath14 bounds respectively . , width=332 ] in the past , the primordial deuterium abundance has been commonly used as a tool for measuring the present - day universal density of baryons ( see e.g. @xcite ) , and more recently as a probe of the effective number of neutrino families ( * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* see also section [ sec : neffobary ] ) . here , we demonstrate that _ precise _ measures of the primordial deuterium abundance ( in combination with the cmb ) can also be used to estimate the neutrino degeneracy parameter , @xmath83 , which is related to the lepton asymmetry by equation 14 from @xcite .
@xcite recently suggested that combined estimates for ( d / h)@xmath3 , @xmath14 , and a measure of @xmath8 from the cmb , can provide interesting limits on the neutrino degeneracy parameter ( @xmath109 , @xmath17 ; see also , @xcite ; @xcite ; and @xcite ) . by combining (
d / h)@xmath3 and @xmath14 , this approach effectively removes the dependence on @xmath5@xmath6 . using the conversion relations for ( d / h)@xmath3 and @xmath14 ( eqs .
[ eqn : dhptoetad][eqn : etad ] and [ eqn : yptoetad][eqn : etahe ] ) and the current best determination of @xmath14 ( @xmath110 ; @xcite ) , in addition to the _ _
planck__+wp+highl and @xmath14 added as free parameters ( see section 6.4.5 of @xcite ) . ]
constraint on @xmath8 and the precise determination of ( d / h)@xmath3 reported here , we derive a @xmath17 upper limit on the neutrino degeneracy parameter , @xmath111 , based on the approach by @xcite . we propose that an equally powerful technique for estimating @xmath83 does _ not _ involve removing the dependence on @xmath5@xmath6 by combining ( d / h)@xmath3 and @xmath14 , as in @xcite . instead
, one can obtain a measure of both @xmath5@xmath6 and @xmath8 from the cmb , and use _ either _ ( d / h)@xmath3 or @xmath14 to obtain two separate measures of @xmath83 .
this has the clear advantage of decoupling ( d / h)@xmath3 and @xmath14 ; any systematic biases in either of these two values could potentially bias the measure of @xmath83 . separating ( d / h)@xmath3 and @xmath14 also allows one to check that the two estimates agree with one another .
our calculation involved a monte carlo technique , whereby we generated random values from the gaussian - distributed primordial d / h abundance measurements , whilst simultaneously drawing random values from the ( correlated ) distribution between @xmath5@xmath6 and @xmath8 from the _
_ planck__+wp+highl cmb data @xcite@xmath6 and @xmath8from the appropriate distribution , we instead used the markov - chain monte carlo chains provided by the _ planck _
science team , which are available at : + ` http://www.sciops.esa.int/wikisi/planckpla/index.php ?
title = cosmological_parameters&instance = planck_public_pla ` ] . using equation 19 from (
* equivalent to eq .
[ eqn : etad ] here ) , we find @xmath15 for ( d / h)@xmath3 , leading to a @xmath17 upper limit of @xmath112 . with the technique outlined above , we have also computed the neutrino degeneracy parameter from the current observational bound on @xmath14 . for this calculation , we have used the mcmc chains from the _ _ planck__+wp+highl cmb base cosmology with @xmath8 and @xmath14 added as free parameters . in this case , the cmb distribution was weighted by the observational bound on @xmath14(@xmath14=@xmath110 ; @xcite ) . using equations 1920 from (
* equivalent to eqs .
[ eqn : etad ] and [ eqn : etahe ] here ) , we find @xmath113 for ( d / h)@xmath3 and @xmath114 for @xmath14 .
these values translate into corresponding @xmath17 upper limits @xmath115 and @xmath116 . combining these two constraints
then gives @xmath117 , or @xmath118 ( @xmath17 ) .
alternatively , if we assume that the effective number of neutrino species is consistent with three standard model neutrinos ( i.e. @xmath8 @xmath119 ) , we obtain the following bbn - only bound on the neutrino degeneracy parameter by combining ( d / h)@xmath3 and @xmath14 , @xmath120 , or @xmath121 ( @xmath17 ) .
we therefore conclude that all current estimates of the neutrino degeneracy parameter , and hence the lepton asymmetry , are consistent with the standard model value , @xmath86 . from the above calculations ,
it is clear that @xmath14 is the more sensitive probe of any lepton asymmetry , as is already appreciated .
however , ( d / h)@xmath3 offers an additional bound on @xmath83 that is complementary to that obtained from @xmath14 .
we note further that , if the uncertainty in the conversion of ( d / h)@xmath3 to @xmath84 could be reduced by a factor of two , the bound on @xmath16 would be reduced by 15% , corresponding to a @xmath70 uncertainty on @xmath83 of @xmath122 from the cmb and deuterium measurements alone .
we have reported a new precise measurement of the primordial abundance of deuterium in a very metal - poor damped lyman@xmath1 system at @xmath2 towards the qso sdss j1358 + 6522 .
furthermore , we have reanalyzed self - consistently all literature systems that meet a set of strict criteria ( four cases ) .
these criteria were imposed to identify the systems where both accurate and precise measures of d / h could potentially be obtained .
the new system , plus the four from the literature , constitute the _ precision sample _ of dlas that are best suited for a precise determination of ( d / h)@xmath3 .
our reanalysis was performed blind ( to remove human bias ) , and took advantage of a new software package that we have specifically developed for the measurement of d / h in qso absorption line systems . from the analysis of this sample , we draw the following conclusions .
( i ) the very metal - poor dla ( [ fe / h]@xmath123 ) towards sdss j1358 + 6522 provides a strong bound on the primordial abundance of deuterium , ( d / h)@xmath3 @xmath124 .
a weighted mean of the five systems in the precision sample gives the most stringent limit on ( d / h)@xmath3 to date : ( d / h)@xmath3 @xmath125 .
the corresponding baryon density is 100@xmath5@xmath6 @xmath126 , assuming the standard model of particle physics with three families of neutrinos and no lepton asymmetry .
this value is in excellent agreement with 100@xmath5@xmath6(cmb)@xmath127 deduced from the analysis of _ planck _ observations of the cmb .
( ii ) the main limitation in using ( d / h)@xmath3 for cosmological parameter estimation is the conversion of ( d / h)@xmath3 into the ratio of baryons - to - photons .
in particular , modern measurements of the cross - section for the reaction @xmath89(@xmath90)@xmath23heat energies between 50500 kev , where there currently exists a paucity of reliable data , are highly desirable .
we estimate that a factor of two improvement in the conversion from ( d / h)@xmath3 to @xmath5@xmath6would provide a measurement of the universal baryon density from d / h as precise as that derived from the published _ planck
_ data .
( iii ) by combining our d / h determination with _ planck _
observations of the cmb , we can place a tight bound on both @xmath5@xmath6 and @xmath8 .
the best - fitting parameters so derived are 100@xmath5@xmath6=@xmath128 and @xmath8=@xmath129 .
we therefore rule out the existence of an additional ( sterile ) neutrino at @xmath130% confidence , provided that @xmath8 and @xmath12 remained unchanged from bbn to recombination .
the uncertainty on @xmath8 could be reduced further by a more accurate set of reaction cross - sections for @xmath89(@xmath90)@xmath23he and , to a lesser degree , @xmath131 .
( iv ) for the first time , we have combined the _ planck _
cmb results with our measure of ( d / h)@xmath3 to place a bound on the neutrino degeneracy parameter , @xmath15 . when including the current best - estimate of the @xmath13he mass fraction , @xmath14 , derived from metal - poor hii regions , the combined bound on the neutrino degeneracy parameter is @xmath117 , or @xmath132 at @xmath17 .
we conclude by re - emphasizing that the most metal - poor dlas are potentially the best environments where both accurate and precise measures of the primordial abundance of deuterium can be obtained
. a combined effort to measure anew several important nuclear reaction cross - sections , in addition to dedicated searches for the rare metal - poor dlas that exhibit resolved deuterium absorption , are the next steps necessary to pin down further the value of ( d / h)@xmath3 .
this approach offers a promising and exciting avenue to test for departures from the standard model of cosmology and particle physics .
we are grateful to the staff astronomers at the eso vlt and keck observatory for their assistance with the observations .
we are indebted to gary steigman for kindly communicating ahead of publication the latest fitting formulae used in this work , and for providing valuable comments on the manuscript .
we also thank an anonymous referee who provided valuable comments that improved the presentation of this work .
discussions , advice and help with various aspects of the work described in this paper were provided by antony lewis , jordi miralda - escud , paolo molaro , ken nollett , pasquier noterdaeme , john omeara , jason x. prochaska , signe riemer - srenson , donatella romano , and john webb .
we thank the hawaiian people for the opportunity to observe from mauna kea ; without their hospitality , this work would not have been possible .
r. j. c. is partially supported by nsf grant ast-1109447 .
r. a. j. is supported by an nsf astronomy and astrophysics postdoctoral fellowship under award ast-1102683 .
m. t. m. thanks the australian research council for a qeii research fellowship ( dp0877998 ) and a discovery project grant ( dp130100568 ) .
in this appendix we show portions of the spectra of the metal - poor dlas from the literature that were reanalyzed in the present work .
further details are given in section [ sec : litsyst ] . | we report the discovery of deuterium absorption in the very metal - poor ( [ fe / h]=@xmath0 ) damped lyman-@xmath1 system at @xmath2 toward the qso sdss j1358 + 6522 . on the basis of
13 resolved i absorption lines and the damping wings of the i lyman @xmath1 transition , we have obtained a new , precise measure of the primordial abundance of deuterium .
furthermore , to bolster the present statistics of precision d / h measures , we have reanalyzed all of the known deuterium absorption - line systems that satisfy a set of strict criteria .
we have adopted a blind analysis strategy ( to remove human bias ) , and developed a software package that is specifically designed for precision d / h abundance measurements .
for this reanalyzed sample of systems , we obtain a weighted mean of ( d / h)@xmath3 @xmath4 , corresponding to a universal baryon density 100@xmath5@xmath6@xmath7 for the standard model of big bang nucleosynthesis ( bbn ) . by combining our measure of ( d / h)@xmath3 with observations of the cosmic microwave background ( cmb ) , we derive the effective number of light fermion species , @xmath8@xmath9 .
we therefore rule out the existence of an additional ( sterile ) neutrino ( i.e. @xmath8@xmath10 ) at 99.3 per cent confidence ( @xmath11 ) , provided that the values of @xmath8 and of the baryon - to - photon ratio ( @xmath12 ) did not change between bbn and recombination .
we also place a strong bound on the neutrino degeneracy parameter , independent of the @xmath13he primordial mass fraction , @xmath14 : @xmath15 based only on the cmb+(d / h)@xmath3 observations . combining this value of @xmath16 with the current best literature measure of @xmath14
, we find a @xmath17 upper bound on the neutrino degeneracy parameter , @xmath18 . |
given a link @xmath5 , let @xmath2 denote the double - cover of @xmath1 branched along @xmath0 .
this is a closed , oriented @xmath6-manifold , to which we can associate its heegaard floer homology group @xmath3 , defined by ozsvth and szab @xcite .
when the determinant of @xmath0 is non - zero , @xmath2 is a rational homology sphere , and this invariant takes the form of an abelian group , graded by rational numbers , which decomposes according to @xmath7 structures on @xmath2 .
closely related to this construction are two other invariants : the knot floer homology @xmath8 , defined by ozsvth and szab @xcite and by rasmussen @xcite ; and the khovanov homology @xmath9 @xcite .
both of these knot homology theories are abelian groups with an integer bigrading , one of which is commonly denoted by a @xmath4 in both theories .
these invariants exhibit many similarities , and understanding their precise relationship is an area of active interest .
for example , ozsvth and szab have constructed a spectral sequence from the reduced khovanov homology @xmath9 to @xmath10 , using @xmath11 coefficients @xcite .
rasmussen observed an inequality of ranks @xmath12 for many knot types , and conjectured the existence of a spectral sequence between these groups to explain this phenomenon ( @xcite , section 5 ) .
such a result would be particularly interesting since @xmath13 is known to distinguish the unknot @xcite , whereas the corresponding fact is unknown for @xmath14 . both of the knot homology theories @xmath13 and @xmath14 possess _ spanning tree models _ as well .
given a planar projection @xmath15 of a link @xmath0 , we can color its regions black and white in checkerboard fashion , and define a graph @xmath16 whose vertices correspond to the black regions and whose edges correspond to incidences between black regions at the crossings in @xmath15 .
there is a standard construction of a doubly - pointed heegaard diagram for @xmath5 out of @xmath15 , and its corresponding floer chain complex @xmath17 is freely generated by spanning trees of @xmath16 @xcite . in a different direction , by performing reductions on the chain complex appearing in the definition of khovanov homology , champanerkar and kofman @xcite and wehrli @xcite have shown that the same is true for khovanov homology .
these models are particularly economical in many cases .
for example , when @xmath15 is a connected alternating diagram , the differentials on @xmath17 and @xmath18 vanish .
moreover , this model remains useful in making calculations for small non - alternating knots , and for understanding pieces of @xmath13 for other families of knots @xcite .
however , the differential on @xmath17 and @xmath18 in the spanning tree models remains a mystery in general .
we note that the original construction of khovanov homology is algorithmically computable , and the same is now known for knot floer homology as well @xcite , although the complexes involved in these constructions typically have much larger rank than their homology groups .
it remains a puzzle to describe a simple , algorithmic model for either theory @xmath13 or @xmath14 based on a chain complex generated by spanning trees .
for the unknot @xmath0 .
( b ) the doubly - pointed heegaard diagram presenting @xmath5 .
( c ) the heegaard diagram presenting @xmath2 .
the heegaard surfaces shown have genus five , and the five red @xmath19 curves encircle the holes in the surface .
the blue @xmath20 curves appear thinner when they pass onto the bottom of the transparent heegaard surface . , title="fig:",width=480 ] ( -270,320)(a ) ( -370,160)(b ) ( -170,160)(c ) the main result of this paper is a spanning tree model for @xmath3 which is similar in spirit to the spanning tree model for @xmath8 .
figure [ f intro ] gives an example of the heegaard diagram involved .
shown there is a diagram @xmath15 for the unknot @xmath0 , the associated doubly - pointed heegaard diagram for @xmath5 ( following @xcite ) , and a heegaard diagram for @xmath2 . comparing the two pictures , observe that the @xmath19 and @xmath20 curves and their intersection points can be put into one - to - one correspondence between the two heegaard diagrams , and that the top half of both diagrams are identical .
the first observation indicates that both of the associated floer chain complexes have the same generating set , while the second is a bit more subtle , leading to a pair of spectral sequences which we examine in @xmath21[s specseq ] . this model for @xmath3
suffers the same drawback as does the model for @xmath8 , in that it does not lead to an algorithmic calculation of the homology group .
however , we can use it to make computations for many manifolds of interest , and it suggests the existence of a @xmath4-grading on @xmath3 analogous to the ones on @xmath13 and @xmath14 . + this paper is organized as follows . in @xmath21[s prelim ] we discuss some of the basic combinatorics of knot diagrams .
in particular , we define the goeritz matrices and coloring matrix associated to a knot diagram , and we introduce kauffman states to encode spanning trees of the black graph . in @xmath21[s hds ] we construct two heegaard diagrams for @xmath2 out of a planar diagram @xmath15 of @xmath0 .
the first of these , @xmath22 , arises out of the standard cut - and - paste description of @xmath2 .
the second , @xmath23 , is the one alluded to already .
it is constructed by resolving @xmath15 into a single unknotted curve with a collection of resolving arcs .
the branched double - cover of the unknotted curve is @xmath1 , and the preimages of the resolving arcs gives rise to a simple framed link @xmath24 .
surgery on @xmath25 produces @xmath2 , and by writing down a heegaard diagram subordinate to @xmath25 we obtain @xmath23 .
this procedure is depicted in figure [ f hdderivation ] . using this diagram
, we give a quick proof that the branched double - cover of a non - split alternating link is an l - space . in @xmath21[s absgradings ] we write down a simple formula for the grading and @xmath7 structure of a kauffman state @xmath26 ( theorem [ t grading ] ) , interpreted as a generator of @xmath27 .
one of the terms in the grading formula is a quantity @xmath28 akin to the @xmath4-grading on @xmath13 and @xmath14 . after proving these formulas ,
we show in @xmath21[ss alternating ] a simple way to compute the correction terms for @xmath2 when @xmath0 is a non - split alternating link ( cf .
@xcite , theorem 3.4 ) . in @xmath21[s domain ] we use the goeritz matrices , coloring matrix , and the results of @xmath21[ss regions ] to write down the domain of a homotopy class connecting a pair of generators ( theorem [ t domain ] ) . in @xmath21[s specseq ] we revisit the similarity between the diagram @xmath23 and the standard doubly - pointed heegaard diagram for @xmath5 gotten from @xmath15 .
we write down a pair of spectral sequences whose @xmath29 terms are freely generated by kauffman states , and which converge to @xmath3 and @xmath8 . moreover , the @xmath30 differential in both spectral sequences counts holomorphic disks whose domain is supported on top of the heegaard diagram , so the @xmath31 term in both sequences are the same . theorem [
t e1 ] asserts that this group is freely generated by so - called _ solitary _ kauffman states of @xmath15 ( definition [ d solitary ] ) .
we note that by construction the @xmath30 differential preserves the alexander grading in @xmath17 , and respects the decomposition of @xmath32 into @xmath7 structures . in @xmath21[s calculations ] we apply the spanning tree model of @xmath3 to make calculations for some knots with @xmath33 crossings .
specifically , we apply it to those knots which are neither alternating nor montesinos knots , since the invariant for these knots can be computed by other means ( as in @xmath21[ss alternating ] and @xcite ) .
we conclude in @xmath21[s speculation ] with some speculation .
we focus on the term @xmath28 appearing in the grading formula theorem [ t grading ] and conjecture the existence of a natural @xmath4-grading on @xmath3 , analogous to the one on @xmath8 and @xmath9 . in support of this conjecture
are some small calculations of @xmath34 , where @xmath35 is a rational homology sphere arising as the boundary of a negative definite plumbing on a tree . in these examples ,
the differential in the spanning tree model behaves surprisingly nicely , and suggests an algorithmically computable model for @xmath34 .
we also point out a strong similarity with work by nmethi @xcite , who has proposed an alternative algorithmically computable model for @xmath36 , which comes equipped with an additional integer grading reminiscent of our @xmath4 .
+ we point out that heegaard diagrams presenting @xmath2 , and more general @xmath37-fold cyclic branched covers @xmath38 , have appeared in the work of grigsby @xcite and levine @xcite . in their work ,
the approach is to begin with a doubly- or multiply - pointed heegaard diagram presenting @xmath5 , form the appropriate cyclic branched cover of the heegaard surface , and thereby obtain a heegaard diagram presenting the preimage of the link @xmath39 .
this approach is particularly well - suited to the calculation of the knot floer homology @xmath40 . by contrast ,
our constructions are specific to the case @xmath41 , and do not immediately present the preimage @xmath42 .
another heegaard diagram for @xmath2 was described by manolescu ( @xcite , section 7 ) . in that setting ,
the link @xmath0 is first presented as a plat closure .
manolescu then shows how the set of generators used by bigelow @xcite to compute the jones polynomial @xmath43 can also be used as a generating set for both the seidel - smith `` symplectic khovanov cohomology '' @xmath44 @xcite and the floer homology @xmath45 .
the triple use of the bigelow generators here bears a surface similarity to that of the spanning trees described above .
it is a pleasure to thank my advisor , zoltn szab , for his patience and guidance during this project .
i am also grateful to john baldwin , eli grigsby , adam levine , and zhong tao wu for conversations about this work .
thanks lastly to jake rasmussen for drawing my attention to the work @xcite in his excellent course at princeton , spring 2008 .
consider a connected diagram @xmath15 of a link @xmath0 with a marked point @xmath46 on one of its edges .
the diagram @xmath15 splits the plane into connected regions , which we color white and black in checkerboard fashion
. with respect to this coloration , each crossing @xmath47 in @xmath15 has an incidence number @xmath48 given as in figure [ f crossingincidence ] . when the link @xmath0 is oriented , each crossing additionally has a sign and a type ( i or ii ) given as in figure [ f crossingsignandtype ] .
( -190,-20)@xmath49 ( -58,-20)@xmath50 ( -410,-20)positive ( -300,-20)negative ( -168,-20)type i ( -65,-20)type ii we form a planar graph @xmath16 by drawing a vertex in every black region and an edge for every crossing that joins two black regions .
dually , we obtain a graph @xmath51 on the white regions . in both @xmath16 and @xmath51 , we associate the label @xmath52 to the edge @xmath53 corresponding to the crossing @xmath47 , and mark the vertex in the region adjacent to the marked point @xmath46 in @xmath15 .
we call these vertices _ roots _ and denote them by @xmath54 and @xmath55 .
we refer to these decorated plane drawings @xmath16 and @xmath51 as the _ black _ and _ white _ graphs corresponding to @xmath15 . deleting the roots and their incident edges result in the _ reduced _ black and white graphs @xmath56 and @xmath57 . for the unknot @xmath0 with a checkerboard coloring of its regions , along with its corresponding black graph @xmath16 and white graph @xmath51.,title="fig:",width=576 ] ( -355,-5)@xmath46 ( -450,120)@xmath15 ( -280,120)@xmath16 ( -115,120)@xmath51 ( -195,5)@xmath54 ( -45,130)@xmath55 ( -220,130)@xmath58 ( -220,95)@xmath58 ( -258,65)@xmath59 ( -185,65)@xmath58 ( -75,95)@xmath58 ( -75,40)@xmath58 ( -137,65)@xmath59 ( 5,65)@xmath58 the _ goeritz matrix _ @xmath60 of the black graph @xmath16 is defined as follows . enumerate the vertices of @xmath56 by @xmath61 , and set @xmath62 for example , if we label the top left vertex of the graph @xmath16 in figure [ f blackwhitegraph ] by @xmath63 and the top right vertex @xmath64 , then @xmath65 the goeritz matrix @xmath66 of the white graph @xmath51 is defined analogously .
we mention a few important properties of the goeritz matrix @xmath67 .
it is a symmetric @xmath68 matrix , and @xmath69 .
orienting @xmath0 somehow , we can define @xmath70 then gordon and litherland @xcite have shown that the signature of @xmath0 can be computed as @xmath71 ( this conforms with the somewhat backwards convention that a positive link has _ negative _ signature . ) the goeritz matrix is also a presentation matrix for @xmath72 , a fact that we will deduce in @xmath21[ss h_1 1 ] .
when a crossing of @xmath15 is marked , we also get an associated _ coloring matrix _ @xmath73 . enumerate the unmarked crossings of @xmath15 by @xmath74 and the unmarked arcs by @xmath75 .
then we set @xmath76 for example , orient the diagram @xmath15 in figure [ f blackwhitegraph ] out of the point @xmath46 and to the left .
following this orientation out of @xmath46 , let @xmath77 denote the first three crossings of @xmath15 , encountered in that order ; and @xmath78 the three arcs of @xmath15 , traversed in that order , after the marked one . marking the remaining unused crossing
, we obtain the coloring matrix @xmath79 the coloring matrix gets its name because of its relationship with @xmath80-colorings of the knot diagram ( not to be confused with the checkerboard coloring ! ) .
recall that an _
@xmath80-coloring _ is a mapping @xmath81 with the property that at every crossing , the identity @xmath82 holds , where @xmath83 denotes the value on the overstrand , and @xmath84 and @xmath47 the values on the two understrands .
the @xmath80-colorings of the diagram which take the value @xmath85 on the marked arc are precisely the elements in @xmath86 .
the matrix @xmath87 is typically _ not _ symmetric , but like the goeritz matrix , it has the property that @xmath88 , and it gives a presentation for @xmath89 .
every crossing in @xmath15 is incident a corner of four ( not necessarily distinct ) regions .
a _ kauffman state _
@xmath26 of @xmath15 is a matching between its crossings and corners of _ unmarked _ regions , so that each crossing gets paired with one of its incident corners .
the kauffman state @xmath26 can be visualized by placing a small marker nearby each crossing @xmath47 in the corner that @xmath47 gets paired with .
see figure [ f kauffman ] .
we write @xmath90 to indicate the corner of the region with which @xmath47 gets paired and @xmath91 to denote the crossing with which the region @xmath92 gets paired .
there is a 1 - 1 correspondence between kauffman states of @xmath15 and spanning trees of @xmath16 ( and @xmath51 ) . given a kauffman state @xmath26 ,
consider the edge @xmath93 for each crossing @xmath47 for which @xmath90 is black ; the collection of these edges is a spanning tree @xmath94 of the black graph .
moreover , @xmath94 inherits a natural orientation from @xmath26 , gotten by directing each edge @xmath53 to point towards the region containing @xmath90 .
this is the same as orienting @xmath94 out of its root @xmath54 that is , directing every edge of @xmath94 to point towards its endpoint which is further away from @xmath54 inside @xmath94 .
in the same way , @xmath26 gives rise to the spanning tree @xmath95 in the white graph , which is planar dual to @xmath94 .
conversely , starting with a spanning tree @xmath96 of @xmath16 , we may construct its planar dual @xmath97 , and orient each of these trees out of their respective roots .
we obtain a kauffman state @xmath98 as follows : at each crossing @xmath47 , there is a corresponding edge @xmath99 , and we take @xmath90 to lie in the corner of the region to which @xmath53 is directed . of the diagram @xmath15 pictured in figure [ f blackwhitegraph ] , along with the oriented trees @xmath94 and @xmath95.,title="fig:",width=576 ] ( -440,120)@xmath26 ( -293,120)@xmath94 ( -125,120)@xmath95
in this section we describe two heegaard diagrams for the branched double - cover of a link @xmath5 , beginning with a connected planar projection @xmath15 . in @xmath21[ss h_1 1 ]
we define the heegaard diagram @xmath22 based on a surgery diagram for @xmath2 written down by ozsvth and szab @xcite . using it we confirm that the goeritz matrix for @xmath15 presents @xmath100 , and we obtain a simple presentation for @xmath101 . in @xmath21[ss h_1 2 ] we recover @xmath22 from the standard cut - and - paste description of @xmath2 , and fit it into a heegaard triple presenting the double - cover of @xmath102 , branched along a spanning surface for @xmath0 . in @xmath21[ss h_2 ] we turn to our main construction of the heegaard diagram @xmath103 .
we show that @xmath104 is generated by kauffman states of @xmath15 , and examine what happens when @xmath15 is alternating . in @xmath21[ss regions
] we examine the regions of @xmath103 , work which is needed when we determine the domain of a homotopy class in @xmath21[s domain ] . in @xmath21[ss h_2 indep ] we show that @xmath103 is independent of the choice of spanning tree @xmath96 used in its definition .
finally , in @xmath21[ss heegaardmoves ] we examine a sequence of heegaard moves relating the diagrams @xmath22 and @xmath23 .
this relationship enables us to identify a subset of the generators of @xmath105 with kauffman states of @xmath15 , and use them in the determination of the grading on @xmath3 in @xmath21[s absgradings ] .
ozsvth and szab @xcite explain how to use the graph @xmath57 to produce a surgery diagram for the branched double - cover @xmath2 ( they actually use the black graph @xmath56 , but this difference is cosmetic ) .
specifically , label each vertex @xmath106 with the value @xmath107 and center a round unknot at it . for every edge @xmath53 between two vertices
, we add a right-/left - handed clasp between the corresponding unknots according as @xmath108 .
finally , we frame each unknotted component by the label on its vertex . denote the resulting framed link by @xmath25 ; then @xmath109 .
this dual description neatly leads to a heegaard triple subordinate to @xmath25 ( in the sense of @xcite , section 4 ) , and in particular to a heegaard diagram for @xmath2 .
first , let @xmath110 denote the boundary of a closed regular neighborhood @xmath111 of @xmath51 , let @xmath112 denote the collection of curves obtained on intersecting @xmath110 with the unmarked faces of @xmath51 , and let @xmath113 denote a collection of curves on @xmath110 chosen so that each @xmath114 meets @xmath115 geometrically once and avoids all other @xmath116 .
thus the @xmath19-handlebody @xmath117 describes the exterior of @xmath111 , while @xmath118 describes @xmath111 itself .
next , let @xmath119 denote the collection of curves on @xmath110 obtained by pushing the @xmath19 curves slightly into the upper half - space determined by the plane of the diagram .
finally , choose a meridian on @xmath110 around each edge @xmath53 of @xmath51 , and perform a @xmath120-dehn twist along it .
let @xmath20 and @xmath121 denote the images of @xmath122 and @xmath119 following these dehn twists .
then @xmath123 is a heegaard triple subordinate to @xmath25 , and @xmath124 .
verification of this fact is straightforward ; the key points are that @xmath111 is equal to a regular neighborhood of a bouquet of the link @xmath25 , and that the collection of curves @xmath121 is an isotopic copy of the framed link @xmath25 drawn on the surface @xmath110 .
we denote this heegaard diagram for @xmath2 so derived from @xmath15 by @xmath125 ( figure [ f hd1 ] ) . for the diagram @xmath15 in figure [ f blackwhitegraph ] .
omitting the green @xmath20 curves produces a heegaard diagram presenting @xmath2.,height=226 ] the diagram @xmath126 permits an easy calculation of the first homology and fundamental group of @xmath2 .
we begin by reviewing the general procedure of how this works . given a heegaard diagram @xmath127 for @xmath128 , enumerate its @xmath19 curves @xmath129 and @xmath20 curves @xmath130 .
orient these curves somehow , and fix an orientation on the surface @xmath110 .
[ d intersectionmatrix ] given these orientations , the matrix of intersection numbers @xmath131 is called the _ intersection matrix of @xmath132 _ , denoted @xmath133 .
the importance of the definition is that @xmath133 is a presentation matrix for @xmath134 .
a presentation for @xmath135 is given as follows .
we associate a generator @xmath136 to each @xmath115 and a relation @xmath137 to each @xmath138 .
the relation @xmath137 is gotten by traversing one full circuit around the curve @xmath138 : each time @xmath138 encounters a curve @xmath115 , we write down a generator @xmath139 , the sign of the exponent chosen according to the intersection number between @xmath115 and @xmath138 at that point .
the product of these generators in order is the word @xmath137 .
of course , this word depends on the initial point from which we traverse @xmath138 , but two different points give rise to conjugate words , and we will blur this small indeterminacy . in total , we obtain the presentation @xmath140 . returning to the case at hand of @xmath126 , observe that @xmath141 for all @xmath142 .
it follows that @xmath143 agrees with the goeritz matrix @xmath144 , and we recover the well - known fact that @xmath144 is a presentation matrix for @xmath100 . for the fundamental group ,
we write down a generator @xmath136 for each unmarked vertex @xmath145 , and for every edge @xmath53 incident the vertex @xmath146 , we write down the word @xmath147 , where @xmath148 denotes the other endpoint of @xmath53 .
when the neighbor @xmath148 is the marked vertex , we understand this word by taking @xmath149 in place of @xmath150 .
traverse a small clockwise circuit around the vertex @xmath146 and multiply these words together in the order the edges incident @xmath146 are encountered .
the result is a relation @xmath137 , and @xmath126 induces the presentation @xmath151 . in the example of figure [ f blackwhitegraph ] , we obtain the presentation @xmath152 for @xmath101 .
it is easy to check that this is a presentation of the trivial group , and this is consistent with the fact that @xmath153 .
there is a conceptually clearer derivation of @xmath126 which follows the standard cut - and - paste description of @xmath2 , which we presently recall ( see also @xcite , p. 85 and
@xcite , p. 56 ) . from the black regions of the diagram @xmath15
, we obtain a spanning surface @xmath154 for the link @xmath0 : it consists of the planar black regions away from the crossings , along with a half - twisted band nearby each crossing in @xmath15 . form a closed regular neighborhood @xmath155 of @xmath156 with spine equal to the link @xmath0 .
its boundary is the orientable double - cover of @xmath157 branched along @xmath0 , and so comes equipped with an involution @xmath158 .
we take two copies of @xmath159 , identified by some homeomorphism @xmath160 , and glue them together along their boundaries by means of the homeomorphism @xmath161 , and the result is @xmath2 .
we spell out this construction further in order to clarify how it gives rise to a heegaard diagram for @xmath2 .
resolve every crossing of @xmath15 so that its two incident black regions merge , yielding a collection @xmath162 of black regions in the plane .
form the product @xmath163 \times c$ ] and quotient it by collapsing the segment @xmath163 \times p$ ] to the point @xmath164 for each point @xmath165 .
the result is the handlebody @xmath155 , which is naturally embedded in @xmath1 so that @xmath166 is identified with @xmath162 .
notice that , away from its crossings , @xmath15 coincides with the intersection of @xmath155 with the plane of the diagram .
nearby a crossing , we can push the overstrand of @xmath15 onto the top half of @xmath167 and the understrand onto its lower half . in this way we obtain an embedding of the link @xmath0 in @xmath110 ( figure [ f h1 derivation](a ) ) , which extends to an embedding of @xmath157 in @xmath155 .
next we describe an involution @xmath158 on @xmath155 which fixes @xmath157 . away from the crossings of @xmath15 ,
@xmath158 interchanges points @xmath168 for @xmath169 $ ] , @xmath170 .
nearby a crossing , the pair @xmath171 is locally modelled by the pair @xmath172 \times i)$ ] ( viewing @xmath173 ) . with respect to this model , @xmath158
is given by identifying points @xmath174 and @xmath175 .
let @xmath19 denote the collection of curves obtained on intersecting @xmath110 with the unmarked white regions of the link diagram .
thus the pair @xmath176 describes the handlebody which is the complement of @xmath155 in @xmath1 . in order to form @xmath2 , we take two of these handlebodies and identify their boundaries by means of the homeomorphism @xmath158 .
let @xmath121 denote the image @xmath177 , perturbed so as to meet @xmath19 transversally .
then @xmath178 is a heegaard diagram for @xmath2 . in order to describe the @xmath121 curves explicitly , observe that @xmath158 fixes the @xmath19 curves away from the crossings of @xmath15 .
nearby a crossing , the effect of @xmath158 is as pictured in figure [ f h1 derivation](c ) .
we get a collection of curves @xmath121 , well - defined up to isotopy , by pushing @xmath177 off of @xmath19 into the top half of @xmath110 away from the crossings of @xmath15 , and extending in the obvious way nearby the crossings of @xmath15 .
one sees at once that @xmath178 agrees with the diagram @xmath126 described above .
drawn on @xmath110 .
( b ) nearby a crossing .
the red arcs are pieces of @xmath19 curves .
( c ) the effect of the involution @xmath158 on the @xmath19 curves , perturbed to give ( blue ) @xmath121 curves .
( d ) the local picture of the heegaard triple nearby the crossing.,title="fig:",height=226 ] ( -350,150)(a ) ( -150,170)(b ) ( -150,110)(c ) ( -150,50)(d ) the @xmath6-manifold @xmath2 naturally bounds @xmath179 , the double - cover of the four - ball @xmath180 branched along a pushed - in copy of the surface @xmath157 , and we can adapt the preceding construction to a heegaard triple which describes this @xmath181-manifold . this triple will ultimately enable us to calculate the absolute gradings on the floer chain complex , as well as its decomposition into @xmath182 structures ( @xmath21[s absgradings ] ) .
we begin with a careful description of @xmath179 .
form the product @xmath163 \times s^3 $ ] and remove @xmath183 \times int(u)$ ] from @xmath163 \times s^3 $ ] .
the result is homeomorphic to the original @xmath163 \times s^3 $ ] , but now the new @xmath184 boundary component has been decomposed into @xmath155 , a collar neighborhood @xmath185 , and the complement .
we extend the involution @xmath158 on @xmath155 to one on @xmath186 in the obvious way .
now form two copies of the modified @xmath163 \times s^3 $ ] and identify the distinguished @xmath186 contained within the boundaries of each by means of the homeomorphism @xmath158 .
the result is the @xmath181-manifold @xmath179 with two small balls removed .
now we show how to extract a heegaard triple for @xmath179 from this description .
decompose @xmath187 , now writing @xmath118 in place of @xmath155 .
let @xmath188 denote a triangle with edges @xmath189 in clockwise order , and vertices @xmath190 at which the corresponding edges meet .
form the product @xmath191 , and glue @xmath192 onto @xmath193 and @xmath194 onto @xmath195 .
smoothing its corners , this space is diffeomorphic to the modified @xmath163 \times s^3 $ ] from the previous paragraph . to aid in seeing this , note that its boundary consists of @xmath196 and @xmath197 .
the first of these boundary components gets identified with @xmath198 , and the second gets identified with @xmath184 decomposed into @xmath155 , @xmath185 , and the complement .
finally , take two copies of this space and identify them along the subsets @xmath199 sitting within the boundary of each by means of the homeomorphism @xmath158 .
once again we obtain @xmath179 minus two small balls .
presented in this fashion , we see that @xmath200 can be described by a heegaard triple @xmath123 , where the @xmath20 curves are chosen so that @xmath201 specifies the heegaard decomposition @xmath202 .
there is a great deal of flexibility in choosing @xmath20 curves that serve this purpose , and we focus on one way of doing so which depends on a choice of kauffman state @xmath26 in the diagram @xmath15 .
[ d edgecurve ] at each crossing of @xmath15 , there is a corresponding edge @xmath203 , and the handlebody @xmath118 is locally modelled by the cylinder @xmath204 nearby it .
we let @xmath205 denote the _ edge curve _ @xmath206 drawn there , and @xmath207 the collection of edge curves @xmath205 .
now , the kauffman state @xmath26 gives rise to a spanning tree @xmath96 in the white graph @xmath51 .
each @xmath205 clearly bounds a disk @xmath208 , and the curves in @xmath207 are homologically independent and of the right number because @xmath96 is a spanning tree .
we denote the resulting heegaard triple by @xmath209 .
observe that @xmath210 possesses an involution which exchanges the @xmath19 and @xmath121 curves and preserves @xmath207 ; thus @xmath211 . in total , we have shown : [ p htriple ] let @xmath15 be a connected , marked diagram of a link @xmath212 , and @xmath157 the spanning surface for @xmath212 induced its black graph @xmath16 . for any choice of kauffman state @xmath26 , the heegaard triple @xmath210 presents the space @xmath179 with two small balls deleted .
we will ultimately be interested in calculating the heegaard floer homology of @xmath2 , and there is a heegaard diagram @xmath213 presenting @xmath2 somewhat better suited to this purpose than @xmath126 .
its description closely resembles the standard doubly - pointed heegaard diagram for @xmath5 gotten from a diagram of @xmath0 @xcite ( cf . also @xmath21[s specseq ] ) ; in particular , the resulting floer chain complex is generated by the kauffman states of @xmath15 ( proposition [ p kauffman ] ) .
we begin with a description of @xmath213 , making a choice of spanning tree @xmath214 in its construction , and prove that it presents @xmath2 .
then we prove proposition [ p kauffman ] and use this result to show that the branched double - cover of a connected alternating link is an l - space . starting with the given diagram @xmath15 ,
choose a spanning tree @xmath96 of the black graph @xmath16 , and let @xmath97 denote the dual tree in the white graph @xmath51 .
we will imagine the diagram @xmath15 drawn simultaneously with the graphs @xmath16 and @xmath51 in the plane , so that edges of @xmath16 and @xmath51 meet in pairs at the crossings of @xmath15 , and in particular every crossing lies on a unique edge in @xmath215 . let @xmath155 denote a closed regular neighborhood of @xmath15 and @xmath110 its boundary . just as before
, let @xmath19 denote the collection of curves obtained on intersecting @xmath110 with every region of @xmath15 except the one marked white region . draw a meridional curve @xmath216 chosen to encircle the marked edge of @xmath15 nearby @xmath46 , and at every crossing of @xmath15 , introduce a pair of @xmath20 arcs on @xmath110 drawn on the top half of @xmath110 as in figure [ f intro ] .
we will complete each pair of @xmath20 arcs to a full @xmath20 curve by connecting their endpoints with an additional pair of @xmath20 arcs on the bottom half of @xmath110 . at a particular crossing in @xmath15 , let @xmath53 denote the edge of @xmath215 which passes through it .
project that portion of the edge @xmath53 which meets @xmath155 down onto the bottom half of @xmath110 . in this way
, we obtain a @xmath20 arc at every crossing of the diagram @xmath15 . to complete each to a @xmath20 curve , observe that on deleting @xmath53 from the spanning tree to which it belongs , we obtain two subtrees , one of which contains a marked vertex .
the other subtree @xmath217 has a silhouette in the bottom half of @xmath110 which we can connect up with the ends of the @xmath20 arc at the crossing through which @xmath53 passes , and this completes the @xmath20 curve at that crossing .
observe that we can so complete each @xmath20 curve simultaneously so that no two meet .
for example , if edge @xmath218 lies on the subtree @xmath217 , then the arc we draw for @xmath218 will appear nested inside the one drawn for @xmath53 . in short
, there is a unique way , up to isotopy , to extend each @xmath20 arc to a @xmath20 curve in the bottom half of @xmath110 in such a way that no two meet and remain disjoint from the meridional @xmath20 curve drawn at the outset .
the result @xmath219 is a heegaard diagram , and we denote it @xmath103 .
[ p h2 ] the heegaard diagram @xmath103 presents the space @xmath2 .
as an example , figure [ f intro](c ) depicts @xmath103 for the marked diagram @xmath15 pictured in figure [ f blackwhitegraph ] and the spanning tree @xmath96 pictured in figure [ f kauffman ] .
the approach is to fully resolve the diagram @xmath15 into a single unknotted component @xmath162 ; then @xmath2 and @xmath220 will be related by surgery along a particular framed link @xmath24 .
we write down a heegaard triple subordinate to @xmath25 and in this way produce the desired heegaard diagram . .
( b ) the full resolution @xmath162 specified by the pair of dual trees in figure [ f kauffman ] , along with the resolving arcs .
( c ) straightening the full resolution @xmath162 into a line .
( d ) the framed link @xmath25 obtained as the preimage of the resolving arcs in the double - cover of @xmath1 branched along @xmath162 .
( e ) an ambient isotopy of @xmath1 which restores @xmath162 to its original form and carries the link @xmath212 with it .
( f ) a heegaard triple subordinate to the link @xmath212 with @xmath221-framing on each component .
the blue @xmath121 curves away from the marked point correspond to components of @xmath212 , and the green @xmath20 curves are meridians to these components .
the @xmath20-handlebody describes the regular neighborhood of a bouquet of @xmath212 , while the @xmath19-handlebody describes its complement . by performing a @xmath222 dehn twist to each curve @xmath223 along the meridian @xmath224
, we obtain a heegaard triple subordinate to the framed link @xmath25 . omitting the @xmath20 curves from this triple results in the heegaard diagram pictured in figure [ f intro](c).,title="fig:",height=576 ] ( -400,420)(a ) ( -200,420)(b ) ( -400,260)(c ) ( -200,260)(d ) ( -130,265)@xmath59 ( -110,250)@xmath59 ( -170,180)@xmath59 ( -30,185)@xmath58 ( -400,130)(e ) ( -200,130)(f ) enumerate the crossings of @xmath15 by @xmath225 , and center a small closed ball @xmath226 around each one .
the choice of the pair @xmath227 specifies a full resolution of the diagram @xmath15 into a single connected component @xmath162 in the plane .
observe that all the black regions in @xmath15 merge to become the interior of @xmath162 , and similarly the white regions merge to become its exterior . at each crossing @xmath228 , connect the two strands in @xmath229 by a small arc @xmath230 ( figure [ f hdderivation](b ) ) .
notice that @xmath230 lies in the interior of @xmath162 precisely when the edge @xmath231 of @xmath215 which passes through @xmath228 belongs to the tree @xmath214 ; with this notation , we may take as @xmath230 a sub - arc of the dual edge @xmath232 , a technical observation that will be of use soon .
form the double - cover of @xmath1 , branched along @xmath162 , and let @xmath233 denote the preimage of @xmath230 . since @xmath162 is unknotted , this double - cover is simply @xmath1 itself , and @xmath234 defines a link inside it .
we can describe the link @xmath212 concretely as follows .
isotope @xmath162 to a line in the plane so that the marked point @xmath46 gets sent to @xmath221 .
the interior and exterior of @xmath162 become complementary half - planes , naturally distinguished black and white , and the disjoint curves @xmath230 sit to either side of @xmath162 . now reflect each @xmath230 into its complementary half - plane to obtain an immersed collection of curves in the plane .
we obtain the link @xmath212 by pushing the reflected image of @xmath230 down slightly from the plane ( figure [ f hdderivation](c)-(e ) ) .
( -250,50)0 ( -115,50)1 frame @xmath212 by placing the coefficient @xmath222 on component @xmath233 , where @xmath235 corresponds to the type of resolution that takes place at @xmath228 ( figure [ f resolutions ] ) .
we claim that @xmath109 . to see why this is , begin by observing that @xmath236 , and
the double - cover of @xmath237 branched along @xmath87 is the exterior of the link @xmath212 .
the double - cover of @xmath226 branched along either of @xmath238 or @xmath229 is a solid torus , and the preimage of @xmath226 inside @xmath220 is a regular neighborhood of the link component @xmath233 .
therefore , @xmath2 is obtained from @xmath220 by some surgery along @xmath212 .
determination of the framing follows as in @xcite , proposition 2.1 .
next we find a heegaard triple subordinate to @xmath25 , but proceed slightly differently from the construction of the heegaard triple in @xmath21[ss h_1 1 ] ( which would apply to any link which admits a diagram in which each component is an embedded , unknotted curve ) .
form a regular neighborhood of the graph @xmath239 and let @xmath110 denote its boundary .
we obtain the @xmath19 curves as before , intersecting @xmath110 with all the unmarked faces of @xmath240 , we obtain a curve @xmath224 chosen to encircle the arc @xmath230 for every @xmath158 , and we obtain one exceptional meridian @xmath216 near the marked point @xmath46 .
next , for each @xmath158 , push the arc @xmath230 to the top half of @xmath110 , and extend its ends slightly to meet the intersection of @xmath110 with the dual edge @xmath232 ( recall the definition of @xmath230 above ) .
complete this to a curve on the surface by an arc on the bottom of @xmath110 chosen so as to run parallel to @xmath162 and avoid the exceptional @xmath20 curve .
let @xmath241 denote the resulting curve .
observe that we can draw all these @xmath241 on @xmath110 simultaneously so as to remain disjoint , and the resulting collection of curves constitute an isotopic copy of the link @xmath212 .
lastly , let @xmath242 denote a small isotopic translate of @xmath216 , chosen to meet it in two transverse intersection points .
it is easy to see that @xmath243 is a heegaard triple subordinate to the link @xmath212 with @xmath221-framings on each of its components ( figure [ f hdderivation](f ) ) .
the main observation here is that each @xmath241 is a seifert - framed longitude for the corresponding link component . to obtain a heegaard triple subordinate to @xmath25
, we perform a @xmath222 dehn twist to @xmath223 along the curve @xmath224 for every @xmath158 .
direct inspection shows that @xmath244 concurs with @xmath103 ( figure [ f intro](c ) ) .
although the @xmath20 curves may appear somewhat meandering in @xmath103 , their intersections with the @xmath19 curves are very controlled . specifically , we obtain one intersection point for every incidence between an unmarked region and a crossing in the diagram @xmath15 , as well as one special intersection point between the meridional curve @xmath216 and the unique @xmath19 curve it meets . therefore , a collection of intersection points between these curves , one point on each curve , will consist of this special intersection point , along with one point for each unmarked region of @xmath15 ( which lies on its associated @xmath19 curve ) and one point for each crossing of @xmath15 ( which lies on its associated @xmath20 curve ) .
we have established the following result .
[ p kauffman ] the generators of the floer chain complex @xmath104 are in one - to - one correspondence with kauffman states of @xmath15 .
recall that for a rational homology sphere @xmath35 , there is an inequality @xmath245 ( @xcite , proposition 5.1 ) , and @xmath35 is an _
l - space _ precisely when equality holds .
proposition [ p kauffman ] leads to a quick proof of the following result , which is a special case of @xcite , proposition 3.3 .
[ c lspace ] let @xmath0 denote a non - split alternating link .
then @xmath2 is an l - space .
the link @xmath0 admits a connected alternating diagram , and its number of kauffman states equals the determinant @xmath246 .
this value , when finite ( as it is here ) , is well - known to equal @xmath247 .
the resulting sequence of inequalities @xmath248 implies the statement of the corollary .
now we turn to understanding the regions of @xmath103 .
in addition to the curve collections @xmath19 and @xmath20 already defined , we define two other special curves in @xmath110 . the first @xmath249 is the intersection of @xmath110 with the marked white region of @xmath15 .
the second @xmath250 is the result of pushing the boundary of a regular neighborhood of @xmath96 down onto the bottom of @xmath103 .
we arrange so that @xmath4 avoids all @xmath20 curves except for @xmath216 , which it meets in a single point .
let @xmath251 .
although @xmath252 is not an actual heegaard diagram , we will think of it as a modified version of one .
in particular , there is now an intersection point between @xmath253 and @xmath20 corresponding to every incidence between a crossing and a region in @xmath15 .
the curves in @xmath253 naturally split @xmath252 into its _ top _ and _ bottom_. recall that for the marked diagram @xmath15 , its _ arcs _ are the connected line segments in @xmath15 with each endpoint at an undercrossing or the marked point @xmath46 .
[ p h2regions ] the regions on top of @xmath252 correspond to arcs in the marked diagram @xmath15 , and the regions on bottom correspond to regions of @xmath15 .
all regions of @xmath252 are simply - connected , and the same is true of @xmath103 .
it is instructive to verify this proposition in the example pictured in figure [ f intro ] .
the assertion about the regions on top of @xmath252 is immediate from inspection .
note also the analogy to the case of the standard doubly - pointed heegaard diagram of @xmath5 gotten from @xmath15 @xcite ; we pursue this relationship further in @xmath21[s specseq ] . for the second assertion , select an unmarked region @xmath254 of @xmath15 .
the region @xmath254 abuts regions @xmath255 and crossings @xmath256 in counterclockwise order , where @xmath254 and the @xmath257 have opposite colors , and for every @xmath158 , the crossing @xmath228 is incident @xmath257 and @xmath258 ( subscripts @xmath259 ) .
let @xmath260 denote the intersection point corresponding to @xmath228 and @xmath257 , and @xmath261 the intersection point corresponding to @xmath228 and @xmath258 .
then the vertices @xmath262 induce a cycle @xmath263 , whose edges alternate between @xmath253 arcs ( between @xmath261 and @xmath264 ) and @xmath20 arcs ( between @xmath260 and @xmath261 ) .
we claim that there is a region @xmath92 on the bottom of @xmath252 corresponding to @xmath254 which is simply - connected and has boundary @xmath265 .
this is apparent from the construction of @xmath252 .
the case of a marked region @xmath254 is similar . in this case
, the only change is that the region @xmath266 includes the curve @xmath250 and a portion of the curve @xmath216 .
finally , every region on the bottom of @xmath252 must abut an arc of a curve in @xmath253 , and each such arc in turn abuts a unique region @xmath92 constructed in the above way ; it follows that we have enumerated all the regions on the bottom of @xmath252 .
next we analyze the regions of @xmath103 .
observe first that eliminating the curve @xmath250 merges the two regions of @xmath252 which correspond to the marked regions of @xmath15 .
the resulting region is simply - connected . eliminating the curve @xmath249 causes some regions on the bottom and top of @xmath110 to merge . specifically , every region atop @xmath110 and incident @xmath249 merges with a _ unique _ region on the bottom of @xmath110 along the common arc of @xmath249 where they abut .
it follows that every region following the merger is simply - connected , establishing the claim about @xmath103 .
the union of @xmath19 and @xmath20 curves in @xmath103 is an embedded graph @xmath267 , whose vertices are the intersection points between the @xmath19 and @xmath20 curves . at a vertex in @xmath268 , there is a natural cyclic ordering of the pieces of edges incident it , gotten by taking its incident arcs in counterclockwise order as they are embedded in @xmath110 .
thus @xmath268 has the structure of a _ ribbon graph _ : it is a graph with a cyclic ordering on the pieces of edges incident each vertex .
the reason for the phrase `` pieces of edges '' is that an edge may be incident a given vertex twice , and we will want to distinguish these two incidences .
[ l ribbon ] the ribbon graph @xmath268 so obtained is independent of the choice of spanning tree @xmath96 .
we describe the graph @xmath268 by constructing its vertex set , edge set , and collection of cyclic orderings without reference to @xmath96 .
the vertex set consists of one vertex for each incidence of an unmarked region of @xmath15 with a crossing , as well as one for the marked point @xmath46 .
traverse the boundary of an unmarked region of @xmath15 distinct from the one bound by @xmath269 , and for every pair of consecutive crossings , put in an edge between the corresponding vertices .
do the same for the region bound by @xmath269 , but in this case treat the marked point @xmath46 just like a crossing .
next consider a crossing @xmath47 , and enumerate its incident regions @xmath270 in counterclockwise order , so that @xmath271 and @xmath272 abut along the overstrand at @xmath47 . letting @xmath273 denote the corresponding vertices ,
put in edges @xmath274 .
if , say , @xmath275 is the marked region , then the vertex @xmath276 does not exist , and we instead take edges @xmath277 . this collection of vertices and edges specifies the underlying graph @xmath268 . orienting the @xmath19 and @xmath20 curves induces an orientation on the edges of @xmath268 , and additionally orienting @xmath110 specifies an intersection number @xmath278 at each vertex of @xmath268 . fix a vertex @xmath279 in @xmath268 , and let @xmath280 denote the pieces of edges incident @xmath279 , so that @xmath84 edges correspond to @xmath19 arcs and @xmath83 edges to @xmath20 arcs , and the symbols @xmath281 indicate whether the piece of edge is directed out of or into @xmath279 .
if the intersection number at @xmath279 is @xmath59 , we take the cyclic ordering @xmath282 , and if it is @xmath58 we take @xmath283 instead . it is straightforward to check that the resulting ribbon graph @xmath268 agrees with the one given by @xmath103 , regardless of the choice of @xmath96 .
[ p h2indep ] the diagram @xmath103 is independent , up to diffeomorphism , of the choice of spanning tree @xmath96 . here
diffeomorphism of heegaard diagrams _
@xmath284 is a diffeomorphism @xmath285 carrying each @xmath115 onto @xmath286 and @xmath138 onto @xmath287 .
a ribbon graph @xmath268 uniquely specifies an embedding of its underlying graph into a closed surface @xmath288 , up to diffeomorphism of @xmath288 , in the following standard way .
each vertex gives rise to a @xmath85-handle , and each edge gives rise to a @xmath149-handle , which we attach to the @xmath85-handles according to the cyclic ordering of the edges at each vertex .
the graph @xmath268 is embedded in the union of these @xmath85- and @xmath149- handles in the obvious manner ( the ribbon - like surface obtained so far is responsible for the terminology `` ribbon graph '' ) . finally , attach a @xmath289-handle by the identity map to each boundary component of the ribbon surface to produce the desired embedding @xmath290 .
now let @xmath268 denote the ribbon graph obtained from the heegaard diagram @xmath103 , with each edge decorated by the label of the curve in @xmath291 of which it is an arc . by lemma [ l ribbon ]
, @xmath268 is independent of @xmath96 .
the embedding @xmath290 from the preceding paragraph agrees with the original embedding @xmath267 , up to diffeomorphism , because every region of @xmath103 is simply - connected ( proposition [ p h2regions ] ) .
now from the embedding @xmath267 and the edge decorations we can recover the @xmath19 and @xmath20 curves , and so reconstruct @xmath103 up to diffeomorphism .
following proposition [ p h2indep ] , we denote the heegaard diagram @xmath103 by @xmath23 , or just @xmath213 when @xmath15 is understood .
we additionally make @xmath213 into a pointed heegaard diagram by placing a basespoint @xmath292 nearby the marked point @xmath46 in the knot diagram .
thus @xmath292 lies to one side of the meridional curve @xmath216 ; by proposition [ p h2regions ] , the regions just to either side of @xmath216 are the same , so the region of @xmath213 in which @xmath292 is chosen is well - defined .
we make @xmath126 into a pointed diagram in an analogous manner . according to a theorem of singer ,
the two diagrams @xmath126 and @xmath213 are related by a sequence of heegaard moves : isotopies , handleslides , and ( de-)stabilizations . in this section ,
we explicitly describe such a sequence from @xmath213 to @xmath126 .
this description will enable us to compare the absolute gradings of the generators of @xmath293 and @xmath294 , a key to establishing the absolute gradings formula in @xmath21[s absgradings ] .
begin by choosing a spanning tree @xmath96 of the black graph , let @xmath295 denote the corresponding kauffman state , and construct the diagram @xmath103 ( of course , proposition [ p h2indep ] removes the dependence of @xmath213 on @xmath96 , but making this choice is useful for descibing what follows ) .
orient @xmath96 out of the root @xmath54
( @xmath21[ss prelim_3 ] ) .
now enumerate its vertices @xmath296 in such a way that for every directed edge @xmath297 , we have @xmath298 .
to each vertex @xmath260 there corresponds a curve @xmath224 which is associated to the unique edge @xmath231 in @xmath96 directed into @xmath260 .
this is pictured in figure [ f heegaardmoves](c ) .
for @xmath299 in turn , handleslide the curve @xmath224 over all @xmath138 for which @xmath297 .
we can then perform an isotopy to the transformed @xmath224 so that it is supported on the bottom of the heegaard surface away from crossings in the diagram .
having done so , each @xmath224 curve is supported nearby the curve @xmath115 corresponding to @xmath260 ( figure [ f heegaardmoves](d ) ) .
next , handleslide @xmath224 over every @xmath20 curve incident @xmath115 which corresponds to an edge of the dual tree @xmath97 .
following an isotopy , the resulting @xmath224 is pictured as in figure [ f heegaardmoves](e ) . , with part of a kauffman state @xmath26 .
( b ) the portion of the black and white graphs nearby the pictured region .
the edges in @xmath300 and @xmath95 are thickened and directed accordingly .
( c ) the corresponding local picture of @xmath23 .
all the @xmath19 curves pictured are black .
the curve @xmath224 is shown in red , and the curves @xmath138 where @xmath146 neighbors @xmath260 are shown in blue .
the arc shown in green represents a number of nested parallel arcs of the @xmath301 , where @xmath302 is an edge along the path from @xmath254 to @xmath260 in @xmath96 .
( d ) the effect of handlesliding all the curves @xmath138 with @xmath303
. now the green curve corresponds to a unique @xmath138 .
( e ) the effect of another sequence of handleslides .
( f ) following the destabilizations , we obtain the local picture of @xmath22 shown .
the kauffman generator @xmath304 gets carried onto the corresponding generator in @xmath105 through this sequence of moves .
local pieces of both are pictured by the heavy black dots in ( c)-(f).,title="fig:",width=528 ] ( -370,450)@xmath305 ( -170,450)@xmath306 ( -400,350)@xmath307 ( -200,350)@xmath308 ( -400,150)@xmath309 ( -200,150)@xmath310 ( -123,430)@xmath260 ( -85,390)@xmath231 ( -272,212 ) observe that at this stage , each @xmath224 only meets @xmath19 curves which correspond to black regions , and meets them with sign and order exactly as in the diagram @xmath126 . to actually obtain @xmath126 , we perform a sequence of destabilizations . thus , enumerate the vertices of @xmath97 by @xmath311 in such a way that @xmath312 is a leaf of the tree @xmath313 for every @xmath158 .
adopt notation @xmath314 and @xmath315 analogous to @xmath115 and @xmath224 above .
now observe that for every @xmath158 , @xmath314 and @xmath315 meet in a single intersection point , and that any other @xmath20 curve which meets @xmath314 takes the form @xmath316 with @xmath317 .
it follows that for @xmath318 in turn , we can perform handleslides and isotopies to @xmath315 so as to eliminate all its intersection points except the single one with @xmath314 , and then destabilize the pair @xmath319 . in effect , this amounts to performing a @xmath289-surgery along each @xmath314 in @xmath110 and erasing all the curves @xmath314 and @xmath315 from the resulting surface .
the end result is the desired heegaard diagram @xmath126 ( figure [ f heegaardmoves](f ) ) .
( a rigorous argument as to why would go as follows : first , the @xmath19 and @xmath20 curves in both this diagram and @xmath126 give rise to isomorphic ribbon graphs ; second , the regions in @xmath126 are simply - connected ; and third , the proof of proposition [ p h2indep ] goes through to show that the two heegaard diagrams are diffeomorphic .
we leave further details to the reader . )
having described the transformation from @xmath213 into @xmath126 , we take a closer look at the generators of @xmath293 .
the @xmath19 and @xmath20 curves in @xmath126 come in pairs @xmath320 , @xmath321 .
intersection points in @xmath322 correspond to edges incident the vertex @xmath323 in @xmath51 .
let @xmath26 denote a kauffman state for @xmath324 , @xmath95 the corresponding spanning tree of @xmath51 , and orient @xmath95 out of its root @xmath55 . for each vertex @xmath321
, there is a unique edge in @xmath95 directed into it , and we let @xmath325 denote the intersection point in @xmath322 which corresponds to this edge . the tuple @xmath326 is an intersection point in @xmath327 , and we call it the _ kauffman generator of @xmath293 _ corresponding to @xmath26 . here is an alternative description of this generator .
regard @xmath26 as a choice of corners in @xmath15 , so that exactly one appears at each crossing , and exactly one appears in each region except for the two regions incident the marked point @xmath46 . in the heegaard diagram @xmath126 , at each corner that appears in a white region @xmath323 , there is an intersection point between @xmath328 and @xmath329 .
the collection of these intersection points is the kauffman generator corresponding to @xmath26 in @xmath293 .
for example , there are seven generators for the heegaard diagram @xmath22 pictured in figure [ f hd1 ] , and five of these are kauffman generators .
now let @xmath295 denote the kauffman state corresponding to @xmath96 , identified as an collection of intersection points between @xmath19 and @xmath20 curves in @xmath213 .
every heegaard move in the above transformation from @xmath213 to @xmath126 takes place in the complement of @xmath26 and the basepoint @xmath292 , with the exception of the destabilizations , each of which deletes a point from @xmath26 .
it follows that the absolute grading of @xmath330 equals the absolute grading of the resulting generator in @xmath293 , which in turn is precisely the kauffman generator corresponding to @xmath26 .
since the choice of spanning tree @xmath96 and hence the kauffman state @xmath26 here was arbitrary , we have established the following fact .
[ p gradingpreserving ] there is a sequence of heegaard moves taking @xmath213 to @xmath126 which carries the generator of @xmath294 corresponding to the kauffman state @xmath26 onto the corresponding generator of @xmath293 .
the absolute gradings of these two generators are equal in their respective complexes .
notice that the sequence of moves from @xmath213 to @xmath126 given above is adapted to the choice of @xmath26 .
presumably it is the case that any one such sequence matches up the kauffman generators between @xmath293 and @xmath294 , and could be used to show that matched - up generators have the same absolute grading .
instead we have opted for a somewhat indirect approach , leaning on proposition [ p h2indep ] .
let @xmath26 denote a kauffman state of the marked diagram @xmath15 , identified as a genenerator of @xmath27 .
our goal in this section is to give an algebraic - combinatorial formula for the absolute grading @xmath331 as well as its associated @xmath7 structure @xmath332 .
we first state the expression for @xmath331 , then give an outline of its derivation , and finally provide a detailed proof in @xmath333[ss triangle]-[ss chern ] . in @xmath21[ss spinc ] we deduce the formula for @xmath332 , and in @xmath21[ss alternating ] we investigate the absolute gradings formula in the case of a non - split alternating link .
we begin by positing an important assumption that will remain in place in the remainder of the paper . _
the link @xmath0 has non - zero determinant : @xmath334 .
_ this assumption guarantees that @xmath2 is a rational homology sphere , and that the goeritz forms and coloring matrix ( @xmath21[s prelim ] ) are invertible . given a kauffman state @xmath26
, we induce an orientation on the white graph @xmath51 in the following way . given an edge @xmath203 ,
consider the crossing @xmath47 to which it corresponds , as well as the white region which abuts @xmath47 to the same side of the overstrand as @xmath90 .
we direct @xmath53 to point towards the endpoint corresponding to this white region . at a vertex @xmath335 ,
we compute the _ signed degree _ @xmath336 as the number of edges directed out of @xmath323 minus the number of edges directed into @xmath323 , with respect to this orientation on @xmath51 .
number the unmarked vertices @xmath337 , and define the _ degrees vector _ @xmath338 as well as the goeritz matrix @xmath66 .
let @xmath339 denote the quadratic form @xmath340 .
finally , define @xmath28 as the number of edges @xmath53 in @xmath95 with @xmath341 .
[ t grading ] for a kauffman state @xmath26 in a marked diagram @xmath15 , the absolute grading of the corresponding generator in @xmath27 is given by the formula : @xmath342 the proof of theorem [ t grading ] goes as follows . by proposition [ p gradingpreserving ]
, we may calculate @xmath331 by regarding @xmath26 as a kauffman generator of the heegaard diagram @xmath126 .
this diagram fits into the heegaard triple @xmath210 subordinate to the link @xmath25 ( defined at the end of @xmath21[ss h_1 2 ] ) , where @xmath343 .
proposition [ p htriple ] identifies the @xmath181-manifold @xmath344 presented by @xmath210 , and in particular @xmath344 is independent of @xmath26 .
nevertheless , in order to calculate @xmath331 , we use the specific @xmath210 adapted to both @xmath26 and @xmath344 .
we invoke the following formula ( @xcite , p. 385
, equation ( 12 ) ) : @xmath345 here @xmath346 is a specific homotopy class of whitney triangles whose domain in @xmath210 consists of a disjoint union of embedded triangles . now the task is to identify the terms in equation ( [ e gradingformula ] ) with the corresponding terms in equation ( [ e absgrading ] ) .
we easily compute the maslov index and intersection number : @xmath347 for the remaining terms , we must first understand the intersection pairing on @xmath348 . a basis for @xmath348 is given in the following standard way .
each vertex @xmath321 corresponds to an unknotted link component @xmath349 .
take a disk spanning @xmath350 , push its interior into the @xmath181-ball bound by @xmath1 , and cap off with the core of the handle attachment along @xmath351 .
the result is a sphere @xmath352 , and the collection of the classes @xmath353 $ ] freely generate @xmath348 . with respect to the poincar duals of these classes , the intersection form on @xmath354
is given by the matrix @xmath355 .
in particular , @xmath356 it remains to calculate @xmath357 .
this amounts to computing @xmath358 \rangle$ ] for each vertex @xmath323 .
we write down a triply - periodic domain @xmath359 in @xmath210 representing the homology class @xmath353 $ ] and apply the first chern class formula ( @xcite , prop .
6.3 ) : @xmath360 \rangle = -2n_z(\p_w ) + \#(\del \p_w ) + \widehat{\chi}(\p_w ) + 2 \sigma(\psi_x,\p_w).\ ] ] the bulk of the proof involves showing that the right - hand side of equation ( [ e chernformula ] ) reduces to @xmath361 .
thus we obtain the expression @xmath362 with respect to the basis of poincar duals to the @xmath353 $ ] , and so @xmath363 .
this completes the identification of terms between equations ( [ e absgrading ] ) and ( [ e gradingformula ] ) and hence the proof of theorem [ t grading ] .
we make two remarks regarding theorem [ t grading ] .
first , @xmath331 has an equivalent expression to that in equation ( [ e absgrading ] ) with reference to the black graph @xmath16 in place of @xmath51 : one defines the degrees vector @xmath364 analogously to @xmath365 , and sets @xmath28 instead equal to the number of edges @xmath53 in @xmath94 with @xmath366 .
second , it is possible to obtain an expression for @xmath331 by working solely with the diagram @xmath213 , proceeding along the lines sketched above , without ever making reference to @xmath126 .
however , in order to obtain the given expression involving the goeritz form , the passage to @xmath126 seems the easiest route to take .
write @xmath368 , where @xmath124 and @xmath207 is as in definition [ d edgecurve ] .
enumerate the vertices @xmath369 of @xmath51 so that @xmath370 and each @xmath371 neighbors some @xmath261 , @xmath372 in @xmath95 , for all @xmath373 .
for @xmath373 , let @xmath116 and @xmath374 denote the curves corresponding to vertex @xmath371 .
recall ( @xmath21[ss h_1 2 ] ) that each curve in @xmath207 corresponds to an edge of @xmath95 ; we let @xmath138 denote the one corresponding to the edge of @xmath95 that directs into the vertex @xmath371 , when @xmath95 is oriented out of its root . since @xmath95 is a tree , an easy induction shows that there is a unique intersection point between @xmath375 and @xmath376 for every @xmath160 , and it lies in @xmath377 .
the same reasoning applies to the @xmath121 curves in place of the @xmath19 .
let @xmath378 and @xmath379 denote the resulting unique intersection points
. now fix a vertex @xmath371 and let @xmath380 , @xmath381 , and @xmath382 ( @xmath21[ss heegaardmoves ] ) .
there is a small triangular domain @xmath383 connecting @xmath384 , and @xmath385 in @xmath210 which will appear as shown in figure [ f smalltriangle ] if @xmath386 , and will appeared reflected from this picture if @xmath387 .
we let @xmath388 denote the homotopy class whose domain is the union of these small triangles .
clearly , @xmath389 . ( shaded ) at a crossing @xmath47 with @xmath390 = -1 .
a dual spider is pictured , and @xmath116 and @xmath374 have been oriented as the boundary of @xmath391 ( @xmath21[ss periodic ] ) . the orientation on @xmath138 will depend on the relative values of @xmath392 and @xmath393 .
this local picture appears on the bottom of the heegaard surface , viewed from below .
( contrast this viewpoint with that of figures [ f periodic ] and [ f periodicmodify ] , in which the top half of the heegaard surface has been sliced away.),title="fig:",width=192 ] ( -60,5)@xmath385 ( -95,72)@xmath394 ( -95,5)@xmath395 ( -37,10 ) ( -110,105 ) ( -72,105 ) .,title="fig:",width=96 ] ( -38,0)@xmath84 ( -38,100)@xmath83 ( -16,55)@xmath47 ( -62,55)@xmath396 ( -18,30 ) ( -64,30 ) ( -80,50 ) next we determine the maslov index @xmath397 .
this is calculated as the sum of contributions from the individual @xmath383 . for a domain @xmath383 which appears reflected from figure [ f smalltriangle ] ,
the maslov index contribution is @xmath398 ( @xcite , prop 9.5 , or @xcite , section 2 and theorem 4.1 ) .
the contribution @xmath399 for the kind pictured in figure [ f smalltriangle ] follows as well from @xcite .
alternatively , its contribution can be gleaned from figure [ f maslovtriangle ] . if @xmath400 has domain equal to the bigon in this picture , and @xmath401 , @xmath402 have domains equal to the two triangles , then @xmath403 , and @xmath404 on the one hand , while additivity implies @xmath405 .
the result @xmath399 in this case follows .
( thanks to zoltn szab for providing this argument . )
now the sum of the @xmath406 results in @xmath407 , which is the expression ( [ e maslov ] ) .
we revisit the description of the homology class @xmath353 \in h_2(x)$ ] ( @xmath21[ss absgradings1 ] ) .
express @xmath344 as the union @xmath408 , where we decompose @xmath409 and @xmath410 .
this decomposition distinguishes the surface @xmath411 .
now fix a vertex @xmath321 and consider the curve @xmath412 .
it bounds a disk inside @xmath413 on the one hand , as well as one inside @xmath414 since @xmath415 = [ \alpha_w ] + \sum_e c_e [ \beta_e]$ ] , for suitable @xmath416 .
the union of these two disks represents the homology class @xmath353 $ ] .
it follows that a periodic class @xmath417 in @xmath210 will represent @xmath353 $ ] provided @xmath418 , for suitable @xmath416 . by way of this latter property
, we can easily construct a periodic class @xmath359 which represents @xmath353 \in h_2(x)$ ] and also satisfies @xmath419 .
any point @xmath420 ( cf .
definition [ d edgecurve ] ) is connected to the basepoint @xmath292 via some path @xmath421 inside this subsurface which meets the @xmath149-cycle @xmath422 with some algebraic multiplicity , having oriented the path from @xmath292 to @xmath423 and @xmath328 and @xmath424 counterclockwise .
this multiplicity is the coefficient of @xmath359 on the region containing @xmath423 .
it follows at once that @xmath425 takes the required form and so @xmath359 is the desired periodic class .
notice , in particular , that @xmath359 does not depend on the other @xmath19 and @xmath121 curves in its construction .
it is convenient to encode the periodic class @xmath359 by a planar graph @xmath426 with labels on its faces .
the graph consists of a copy of the tree @xmath427 along with the collection of edges @xmath428 incident the vertex @xmath323 .
there are two copies of an edge @xmath429 , and we distinguish the copy in @xmath428 by pushing its interior clockwise off of the copy in @xmath96 when @xmath366 , and counterclockwise when @xmath341 .
we orient an edge in @xmath428 out of @xmath323 when @xmath366 and into it when @xmath341 .
the resulting planar graph has some faces , which we label by analogy to the construction of @xmath359 : for a point @xmath423 inside a face , we orient a path from the basepoint @xmath292 to it that avoids the edges of @xmath96 , and record the oriented intersection number with the oriented edges in @xmath428 . identified , and spanning tree highlighted .
( b ) the heegaard surface and the top of the periodic domain @xmath359 .
( c ) the bottom of the periodic domain @xmath359 .
the top half of the heegaard surface has been sliced off in this view .
( d ) the affiliated face - labeled planar graph @xmath426 . in ( b)-(d ) , a light gray region has coefficient @xmath59 ; medium gray , @xmath430 ; dark gray , @xmath431 ; and pink , @xmath58.,title="fig:",width=384 ] ( -320,320)(a ) ( -140,320)(b ) ( -320,130)(c ) ( -140,130)(d ) ( -290,260)@xmath323 ( -15,205)@xmath292 ( -105,95 ) ( -80,75 ) ( -110,35 ) we extract a couple of pieces of information from the graph @xmath426 .
for every edge @xmath432 , consider the subgraph of @xmath426 obtained on deleting all the other edges in @xmath428 .
the resulting graph has a unique cycle @xmath433 made up of @xmath218 and some edges of @xmath96 , and it gets an orientation from the one on @xmath218 .
we define a label @xmath434 observe that for every @xmath435 , we have @xmath436 . as an example , consider the edge @xmath437 pictured in figure [ f periodic](d ) , and also consider how that picture would change if the crossing corresponding to @xmath437 were reversed . in general , all but one such @xmath218 is oriented into @xmath323 in the orientation on @xmath51 induced by @xmath96 ( @xmath21[ss absgradings1 ] ) . similarly , for @xmath438 , one checks that @xmath439 according as @xmath218 is directed out of or into @xmath323 in that orientation .
consequently , @xmath440 here the @xmath289 " corrects for the unique edge in @xmath441 which is directed out of @xmath323 in @xmath51 .
we also remark that the label on a face @xmath157 of @xmath426 is equal to the sum of @xmath442 , over all edges @xmath432 for which @xmath157 is contained inside @xmath433 .
lastly , for every edge @xmath443 , let @xmath444 denote the pair of labels on the faces abutting along @xmath53 . in this section
we calculate the terms appearing in the first chern class formula ( [ e chernformula ] ) to deduce the expression ( [ e chernvector ] ) .
[ p chernprop ] for the triangle @xmath346 and periodic domain @xmath359 above , we compute 1 .
@xmath419 , 2 .
@xmath445 , 3 .
@xmath446 , and 4 . @xmath447 . combining the terms in proposition [ p chernprop ] into ( [ e chernformula ] ) , and simplifying the result using ( [ e dx ] ) and the miraculous cancelation " @xmath448 , yields the desired identity ( [ e chernvector ] ) .
now we establish proposition [ p chernprop ] piece by piece .
note that [ p chernprop].1 is immediate from the construction of @xmath359 .
+ _ proof of [ p chernprop].2 .
_ we have @xmath449 , for suitable @xmath416 , so @xmath450 . the value @xmath451 is computed as the absolute value in the difference between the coefficients on the pair of regions atop @xmath110 that abut along @xmath205 , which is @xmath452 .
+ _ proof of [ p chernprop].3 .
we first modify @xmath359 nearby the curve @xmath328 to get a @xmath289-chain @xmath417 with the same euler measure @xmath453 .
this is depicted in figure [ f periodicmodify](b ) .
the boundary @xmath454 is supported on edge curves @xmath205 ( @xmath21[ss h_1 2 ] ) , @xmath455 .
enumerate the edges in @xmath428 in counterclockwise order .
the difference between the two - chains @xmath456 atop @xmath110 is a collection of bigons , squares , and hexagons , each with coefficient @xmath59 ( figure [ f periodicmodify](c ) ) .
there is a square between each consecutive pair of edges @xmath457 with @xmath458 , a bigon between each pair with @xmath459 , @xmath460 , and a hexagon between each pair with @xmath461 , @xmath462 .
the euler measures of a bigon , square , and hexagon are @xmath463 , and @xmath464 , respectively , and the number of bigons equals the number of hexagons , so the difference @xmath289-chain @xmath456 restricted to the top of @xmath110 has euler measure zero .
the difference @xmath289-chain restricted to the bottom of @xmath110 consists of a pair of triangles of cancelling euler measure for each edge @xmath465 ( figure [ f periodicmodify](d ) ) , so it has euler measure zero as well . in total ,
@xmath466 , as desired .
-chain @xmath359 nearby @xmath328 .
the coefficients on the top of @xmath417 have been labeled .
( b ) the @xmath289-chain @xmath417 nearby @xmath328 .
( c ) the top of the difference @xmath289-chain @xmath467 , nearby @xmath328 .
( d ) the bottom of @xmath456 , nearby @xmath328 , with the top half of the heegaard surface sliced away .
the gray regions have coefficient @xmath59 , and the pink ones have coefficient @xmath58.,title="fig:",width=384 ] ( -300,260)(a ) ( -180,220)@xmath468 ( -185,235)@xmath84 ( -238,250)@xmath469 ( -285,215)@xmath470 ( -275,235)@xmath83 ( -275,170)@xmath47 ( -238,150)@xmath471 ( -190,185)@xmath471 ( -195,170)@xmath472 ( -150,260)(b ) ( -20,230)@xmath84 ( -73,250)@xmath469 ( -110,230)@xmath83 ( -110,170)@xmath47 ( -73,150)@xmath471 ( -30,170)@xmath472 ( -300,100)(c ) ( -150,100)(d ) next , @xmath417 decomposes into a sum of @xmath289-chains , one for each @xmath438 , which we now describe . given such an edge @xmath218 , the planar graph @xmath473 contains a unique cycle @xmath433 , which may have some vertices in its interior
. there are as many of these vertices as there are edges interior to @xmath433 , since @xmath96 is a tree .
the cycle @xmath433 gives rise to a subsurface @xmath474 whose boundary consists of edge curves @xmath205 , @xmath475 , and its genus is the number of edges interior to @xmath433 .
( an instructive example is the case of the edge @xmath476 in figure [ f periodic](d ) . ) therefore , @xmath477 .
lastly , we sign @xmath288 by @xmath442 and denote by @xmath478 the resulting @xmath289-chain
. then the sum of @xmath478 over all @xmath438 equals @xmath359 ( compare the remark following ( [ e dx ] ) ) , and we can calculate @xmath479 in order to recover the expression in [ p chernprop].3 , we reconsider the calculation of @xmath480 . write the value @xmath442 to each side of a curve in @xmath207 that appears in @xmath478 .
so @xmath442 appears once next to each boundary component of @xmath478 and twice next to each @xmath205 , @xmath53 interior to @xmath433 .
the sum of these labels is @xmath481 , hence the sum over all @xmath218 is @xmath482 . on the other hand ,
summing up the labels @xmath442 that appear next to the curve @xmath205 gives @xmath483 , for each @xmath484 , and @xmath485 , for each @xmath486 .
it follows that @xmath487 substituting @xmath488 into this equation results in [ p chernprop].3 .
+ _ proof of [ p chernprop].4 .
_ for the definition and use of the _ dual spider number _
@xmath489 , see @xcite , section 6 , especially the discussion between lemma 6.2 and the proof of lemma 6.1 . in the case at hand , @xmath490 decomposes as a sum of terms @xmath491 . for each , we choose an interior point @xmath46 and arcs @xmath84 , @xmath83 , @xmath47 to @xmath492 as shown in figure [ f smalltriangle ] .
we note that the mirror image of this picture will appear in the case of a crossing @xmath47 with @xmath493 , but this will have no effect on what follows . if @xmath484 , then @xmath494 is disjoint from @xmath495 , and so @xmath496 .
the value of @xmath497 is given by @xmath498 or @xmath499 depending on the labeling , so we may assume it is @xmath498 .
examining the cases @xmath500 , and @xmath501 in turn , we deduce that @xmath502 . in total , @xmath503 if @xmath429 , it is still true that @xmath494 is disjoint from the shifted boundary components @xmath504 , giving @xmath496 .
the value of @xmath497 in this case is given by either @xmath505 or @xmath506 depending on the labeling , so we may assume it is @xmath505 .
exactly as before , @xmath502 , and we obtain @xmath507 summing the value of @xmath491 over all @xmath508 results in [ p chernprop].4 .
the complex @xmath294 decomposes into subcomplexes @xmath509 indexed by @xmath510 , and each kauffman generator @xmath26 is contained in @xmath511 for some @xmath512 .
we describe how to compute @xmath332 .
in general , if @xmath513 is the result of attaching @xmath289-handles to @xmath102 , and @xmath514 is a rational homology sphere , then the long exact sequence in cohomology of the pair @xmath515 splits off a short exact sequence @xmath516 the first chern class @xmath517 sets up a 1 - 1 correspondence between @xmath182 structures on @xmath344 and characteristic covectors in @xmath354 for the intersection pairing on @xmath344 , and @xmath518 is identified as the set of @xmath519-orbits of characteristic covectors . in the case at hand , we have @xmath520 , an identification @xmath521 , and the intersection form given by the matrix @xmath66 .
a characteristic covector @xmath522 is defined by the condition that @xmath523 for all @xmath158 , and @xmath524 is the set of @xmath525-orbits of such vectors in @xmath526 .
the @xmath182 structure @xmath332 lifts to @xmath527 , and equation ( [ e chernvector ] ) expresses @xmath528 .
thus , @xmath332 corresponds to the orbit of @xmath529 .
when @xmath246 is odd , the story simplifies somewhat . in this case , there is a canonical identification @xmath530 and under this identification , we have @xmath531 \in \textup{coker}(g_w).\ ] ] we specialize now to the case that @xmath0 is a non - split alternating link , and let @xmath15 denote an alternating diagram of @xmath0 .
corollary [ c lspace ] implies that each kauffman generator @xmath26 of @xmath27 is unique in the @xmath182 structure @xmath532 it occupies ; consequently , the correction term in a given @xmath182 structure @xmath533 is calculated as the absolute grading @xmath331 .
thus theorem [ t grading ] can be used to calculate the correction terms @xmath534 .
on the other hand , theorem 3.4 of @xcite provides such a formula as well .
let us compare the two .
we begin by recalling the result of @xcite ( cf .
@xcite , corollary 1.5 ) .
let @xmath535 denote the goeritz matrix corresponding to the white graph of @xmath15 , @xmath536 its associated quadratic form , and @xmath37 its rank .
as in @xmath21[ss spinc ] , we interpret @xmath510 as a @xmath537-orbit of characteristic covectors for @xmath144 . with this notation ,
theorem 3.4 of @xcite reads as follows : @xmath538 on the other hand , the matrix @xmath144 is negative definite and @xmath539 for all @xmath26 , so the expression ( [ e absgrading ] ) reduces to @xmath540 comparing equations ( [ e d-1 ] ) and ( [ e d-2 ] ) gives the following result .
[ p d-3 ] let @xmath26 denote a kauffman state in a connected alternating diagram .
the vector @xmath529 attains the maximum value of @xmath536 in the @xmath541-orbit @xmath332 .
it is interesting to give a direct algebraic proof of proposition [ p d-3 ] .
thus , we must identify @xmath542 as the maximum value of @xmath536 over the @xmath543-orbit of @xmath529 .
in other words , we seek the inequality @xmath544 for all integer vectors @xmath423 ; or , what is the same , that @xmath545 in fact , we claim that inequality ( [ e ineq ] ) holds if we replace @xmath529 by any vector @xmath279 induced by an orientation @xmath546 of @xmath51 .
recall that this means that the @xmath261-entry of @xmath279 is equal to the number of edges directed out of @xmath261 , minus the number directed into it , in @xmath546 .
thus , given an orientation @xmath546 , form a matrix @xmath128 whose rows are indexed by @xmath547 , whose columns are indexed by @xmath548 , and whose @xmath549 entry is @xmath550 or @xmath58 according as @xmath53 is not incident @xmath323 , is directed away from it , or is directed into it in @xmath546 .
one checks that @xmath551 where @xmath552 denotes the all @xmath149 s vector of length @xmath37 .
therefore , @xmath553 and @xmath554 where @xmath555
. as the @xmath556 are integers , the inequality @xmath557 follows and completes the proof of proposition [ p d-3 ] .
we briefly review how to compute the domain of a homotopy class connecting two generators in @xmath558 in the case that @xmath559 presents a rational homology sphere @xmath35 .
orient all the @xmath19 and @xmath20 curves and @xmath110 , let @xmath560 denote the collection of oriented segments of @xmath19 and @xmath20 curves that run between two consecutive intersection points , and fix generators @xmath561 . we write the 1-chain @xmath562 which in turn bounds a _ rational _
2-chain @xmath563 fix a segment @xmath230 and let @xmath564 and @xmath565 denote the regions that abut along it to the left and right , respectively .
we obtain the relation @xmath566 and the collection of these relations for all @xmath158 determines the 2-chain up to a rational multiple of @xmath567 . enforcing @xmath568 for the region @xmath564 containing the basepoint @xmath292 removes this indeterminacy , and we denote the resulting 2-chain by @xmath324 .
the 2-chain @xmath324 constructed in this way is _ the domain connecting @xmath26 to @xmath423_. notice that we make no assumption on @xmath569 \in h_1(y)$ ] in making this definition .
however , in the case that @xmath569=0 $ ] , the domain @xmath324 equals @xmath570 for the unique homotopy class @xmath571 satisfying @xmath572 .
in proposition [ p h2regions ] we identified the regions in the heegaard diagram @xmath213 . in effect , they come in two types : those atop the diagram , which are in 1 - 1 correspondence with arcs of @xmath15 , and those on the bottom , which are in 1 - 1 correspondence with regions of @xmath15 . for an arc @xmath19 of @xmath15 , we let @xmath573 denote the corresponding region of @xmath213 . similarly , for a region @xmath279 of @xmath15
, we let @xmath574 denote the corresponding region of @xmath213 . the choice of marked point @xmath46 on @xmath15 will identify some of these regions , so we understand that @xmath573 and @xmath574 may coincide for some choices of @xmath19 and @xmath279 . under these identifications
, we split the determination of @xmath324 into three parts . to begin with , the assumption that @xmath575 allows us to restrict attention to unmarked arcs and regions of the diagram @xmath15 .
enumerate these arcs @xmath576 , black vertices @xmath577 , and white vertices @xmath337 .
we define @xmath578 thus , may we regard @xmath579 as a rational vector @xmath580 indexed by unmarked arcs of @xmath15 , and similar statements hold for @xmath581 and @xmath582 .
next , mark a crossing in @xmath15 and enumerate the others @xmath256 .
given a kauffman state @xmath26 and an unmarked crossing @xmath47 , define @xmath583 and collect these values into the vector @xmath584 let @xmath87 denote the corresponding coloring matrix ( @xmath21[ss prelim_2 ] ) , @xmath585 and @xmath66 the goeritz matrices ( @xmath21[ss prelim_1 ] ) , and @xmath364 and @xmath365 the degrees vectors ( @xmath21[ss absgradings1 ] ) .
[ t domain ] for a pair of kauffman states @xmath26 and @xmath423 in a marked diagram @xmath15 , the domain @xmath324 connecting @xmath26 to @xmath423 in the heegaard diagram @xmath23 is given by the formulas @xmath586 the reason for the difference in signs in the formulas for @xmath581 and @xmath582 is due in effect to the asymmetric dependence of the incidence number @xmath587 on the black - white coloration : switching the roles of black and white reverses the value of @xmath587 at a crossing . we remark on one peculiar feature of theorem [ t domain ] .
recall that equation ( [ e chernvector ] ) in @xmath21[s absgradings ] identified the degrees vector @xmath365 with minus the first chern class of a specific @xmath182 structure on @xmath179 .
its reincarnation in this context seems more than accidental , but a good explanation is lacking . regarding @xmath66 as the intersection pairing on @xmath179 and @xmath588 , the expression for @xmath582 enables us to identify it as a class in @xmath589 ; when @xmath569=0 $ ] , this class is integral . still , the significance of this class remains mystery .
of course , these same remarks apply with the roles of the black and white graphs switched . in @xmath21[ss dalpha ]
we settle the formula for @xmath579 .
its proof is a straightforward application of the procedure described at the beginning of this section . in this situtation
, we need only consider the segments @xmath230 which occur as portions of the @xmath20 curves atop @xmath213 . in @xmath21[ss db&w ] we settle the formulas for @xmath581 and @xmath582 .
we apply the same method , using only segments of the @xmath20 curves on the bottom of @xmath213 .
however , in order to put the resulting answer into the form given in theorem [ t domain ] , more work is needed , and in particular we must revisit the passage from @xmath213 to @xmath126 .
choose a crossing @xmath47 and let @xmath590 denote the overstrand and @xmath591 , @xmath592 the other two arcs that meet at @xmath47 .
orient the curve @xmath593 in the heegaard diagram @xmath213 .
it has two segments @xmath594 and @xmath595 on top of @xmath213 which appear oriented in the same direction , as in figure [ f dtoplocal ] .
notice that this is opposite the corresponding local picture in knot floer homology @xcite .
one of these segments will continue onto the bottom of @xmath110 in case the marked region is incident @xmath47 , but this will not affect what follows . also shown in the picture are the regions @xmath596 that appear at @xmath47 ( @xmath21[ss db&w ] ) . at a crossing @xmath47.,title="fig:",height=192 ] ( -82,120)@xmath597 ( -135,75)@xmath598 ( -20,75)@xmath599 ( -92,60 ) ( -45,60 ) ( -15,120)@xmath63 ( -135,120)@xmath64 ( -135,20)@xmath600 ( -15,20)@xmath276 now fix a pair of kauffman states @xmath26 and @xmath423 .
let @xmath601 denote the coefficient on @xmath602 in the domain connecting @xmath26 to @xmath423 , and @xmath603 the coefficients on segments @xmath604 in @xmath605 .
thus @xmath606 implying the single relation @xmath607 depending on the states @xmath90 and @xmath608 , the difference @xmath609 takes on one of the values @xmath85 or @xmath278 .
there are a few cases to consider . in considering each , it is helpful to imagine a choice of spanning tree @xmath96 in constructing @xmath213 and continuing the curve @xmath593 to the bottom of @xmath110 accordingly .
ultimately , of course , this choice is immaterial .
* _ the states @xmath90 and @xmath608 agree or are diagonally opposite at @xmath47 . _ in either case @xmath610 , which implies the relation @xmath611 . * _ the states @xmath90 and @xmath608 lie to the same side of the overstrand , but differ . _
assume that @xmath90 lies in the bottom - right quadrant of figure [ f dtoplocal ] and @xmath608 lies in the upper - right .
then @xmath612 and we obtain the relation @xmath613 .
a different placement of @xmath90 and @xmath608 can be transformed into this one by ( i ) rotating the local picture , ( ii ) swapping the roles of @xmath26 and @xmath423 , or ( iii ) both .
the effect of ( i ) is to reverse the roles of @xmath614 and @xmath615 as well as @xmath616 and @xmath617 , giving the relation @xmath618 instead .
the effect of ( ii ) is to negate @xmath619 and @xmath617 , which also results in the relation @xmath618 .
these effects are canceled in ( iii ) and result in the original relation . * _ the states @xmath90 and @xmath608 lie to the same side of the understrand , but differ . _
assume that @xmath90 lies in the bottom - right quadrant of figure [ f dtoplocal ] and @xmath608 lies in the bottom - left .
we obtain @xmath620 and hence the relation @xmath621 .
the three other possible placements of @xmath90 and @xmath608 can be treated now exactly as in the previous case .
now consider the black - white coloring of the regions of @xmath15 .
it is easy to check that the expressions for @xmath622 in each of the preceding cases condense into the following unified form : @xmath623 the collection of these relations , one for each crossing @xmath47 , is expressed in the matrix equation @xmath624 where @xmath87 denotes the coloring matrix .
this gives the formula for @xmath579 expressed in theorem [ t domain ] .
we proceed as in the previous subsection , obtaining a relation on the coefficients of @xmath324 on the bottom of @xmath213 , one for each crossing @xmath47 of the diagram .
let @xmath625 and @xmath626 denote the two segments of @xmath593 appearing on the bottom of @xmath213 , where @xmath625 runs between regions @xmath63 and @xmath600 at @xmath47 and @xmath626 runs between @xmath64 and @xmath276 .
let @xmath627 and @xmath628 denote the coefficients on these segments in @xmath605 and @xmath601 the coefficient on region @xmath629 , @xmath630 .
we obtain the relations @xmath631 whence the single relation @xmath632 again , this is most easily seen by making a choice of spanning tree in constructing @xmath213 and extending the curve @xmath593 onto the bottom of @xmath213 . as in the case of @xmath579 , depending on the states @xmath90 and @xmath608 , the difference @xmath633 takes on one of the values @xmath85 or @xmath278 . treating each case in turn , we obtain the following concise result . define @xmath634 and collect these values into a vector @xmath635 then @xmath636 verification is routine , and the collection of these relations , one for each crossing @xmath47 , is expressed in a matrix equation @xmath637 equation ( [ e db&w1 ] ) is the most direct way of expressing the defining relation for @xmath581 and @xmath582 .
however , in order to convert ( [ e db&w1 ] ) into the form given in theorem [ t domain ] , we first modify it into the intermediate form ( [ e db&w2 ] ) .
first , make a choice of spanning tree @xmath214 .
this induces orientations of @xmath16 and @xmath51 and gives rise to a kauffman state @xmath638 ( @xmath21[ss absgradings1 ] ) .
every edge @xmath99 gives rise to a segment of a @xmath20 curve on the bottom of @xmath213 running between the same regions as @xmath53 , and we orient this segment according to the direction on @xmath53 .
notice that at a given crossing @xmath47 , the orientations on the two segments of @xmath593 on the bottom of @xmath213 due to @xmath53 and @xmath639 are compatible in that they induce the same orientation of @xmath593 .
so the choice of @xmath96 ( or equivalently , the kauffman state @xmath26 ) defines an orientation of the curves @xmath593 .
now define @xmath640 in the same way we define @xmath641 for a region @xmath279 incident @xmath47 , and set @xmath642 in case @xmath279 is not incident @xmath47 at all or it meets it twice .
this assignment is depicted in figure [ f d(v , c ) ] . .
pictured is the crossing @xmath47 , where the black dot represents the position of @xmath643 . assuming the region @xmath279 meets the crossing @xmath47
once , the value of @xmath641 is given by the number written in it ; otherwise it takes the value @xmath85.,title="fig:",height=144 ] ( -235,50)@xmath58 ( -315,50)@xmath149 ( -275,90)@xmath390 ( -285,10)@xmath644 ( -65,90)@xmath58 ( -55,10)@xmath149 ( -25,50)@xmath390 ( -102,50)@xmath644 also , set @xmath645 and @xmath646 with these definitions in place , a simple check shows that ( [ e localrelation ] ) can be put into the following form : @xmath647 and the collection of these relations , one for each crossing @xmath47 , condense into a matrix equation @xmath648 here @xmath649 denotes the vector of values @xmath650 , while @xmath128 is the matrix of values @xmath651 , its rows indexed by all crossings @xmath47 and its columns by unmarked regions @xmath279 . in spite of the additional data required to express ( [ e db&w2 ] ) as compared with the equivalent ( [ e db&w1 ] ) , it is easier to convert it into the desired form .
we claim that @xmath652 together , ( [ e db&w2 ] ) and ( [ e alphad ] ) imply the formulas for @xmath581 and @xmath582 in theorem [ t domain ] .
observe that while both @xmath128 and @xmath19 depend on the choice of spanning tree @xmath214 , the left - hand side of ( [ e alphad ] ) is independent of this choice : indeed , a different choice of @xmath96 results in multiplying some columns of @xmath128 and the corresponding entries of @xmath649 by @xmath58 , which get canceled in the product @xmath653 .
consequently , it suffices to prove ( [ e alphad ] ) in the special case that @xmath638 .
notice that @xmath654 for all crossings @xmath47 , which simplifies the expression for @xmath655 , a convenient fact we use below .
we realize ( [ e alphad ] ) by examining the passage from @xmath213 to @xmath126 ( @xmath21[ss heegaardmoves ] ) at the homological level .
let @xmath656 denote the intermediate heegaard diagram obtained prior to the destabilizations .
the matrix @xmath657 is the intersection matrix @xmath658 ( definition [ d intersectionmatrix ] ) , having oriented the @xmath19 curves of @xmath213 to run counterclockwise and the @xmath20 curves as above .
the matrix @xmath659 is the submatrix of @xmath660 induced on the columns @xmath224 corresponding to edges in the tree @xmath214 .
now , if two heegaard diagram @xmath661 and @xmath662 are related by a handleslide of a @xmath20 curve , then @xmath663 is related to @xmath664 by an elementary column operation . if they are related instead by an isotopy , then @xmath665 .
thus , the passage from @xmath213 to @xmath656 entails a sequence of _ row _ operations which transform @xmath128 into @xmath666 . in short , there exists a matrix @xmath667 with the property that @xmath668 similarly there is a matrix @xmath669 so that @xmath670 .
it corresponds to the passage from @xmath213 to the heegaard diagram @xmath671 , where @xmath672 agrees with @xmath15 but with the opposite coloration of its regions . in what follows we concentrate on the case of @xmath667 , as the case of @xmath669 follows analogously .
it stands to actually write down an expression for @xmath667 obeying ( [ e e ] ) and also establish @xmath673 together , ( [ e e ] ) and ( [ e e0 ] ) prove the part of ( [ e alphad ] ) concerning @xmath16 , and the part concerning @xmath51 follows in exactly the same way . in order to write down @xmath667
, we define one final type of incidence number @xmath674 between an unmarked black vertex @xmath279 and crossing @xmath47 , which depends on the kauffman state @xmath675 .
its value is given as in figure [ f sigma ] .
we set @xmath667 equal to the matrix of values @xmath674 , indexed by unmarked black vertices @xmath279 and all crossings @xmath47 .
now we proceed to establish ( [ e e ] ) .
fix @xmath676 and @xmath677 .
the @xmath678-entry of the product @xmath679 is @xmath680 the general term @xmath681 in the summation is non - zero unless the crossing @xmath47 touches both regions @xmath279 and @xmath682 . assuming this occurs
, there are a few cases to consider .
+ . pictured is the crossing @xmath47 , where the black dot represents the position of @xmath643 .
assuming the region @xmath279 meets the crossing @xmath47 once , the value of @xmath674 is given by the number written in it ; otherwise it takes the value @xmath85.,title="fig:",height=144 ] ( -247,50)@xmath644 ( -315,50)@xmath390 ( -20,50)@xmath59 ( -100,50)@xmath58 \1 .
if @xmath90 lies in @xmath279 , then @xmath684 and @xmath685 , so their product is @xmath390 .
if @xmath90 instead lies in @xmath682 , then the signs on @xmath674 and @xmath686 negate and their product remains the same : @xmath390 . if @xmath90 instead lies in one of the white regions ,
there are four possibilities to consider , depending on which white region and the type of crossing at @xmath47 . in any event
, we always obtain @xmath687 it follows then that @xmath688 + 2 .
note that @xmath690 for @xmath691 the black region which touches @xmath47 opposite to the corner of @xmath279 under consideration .
in particular , it is possible that @xmath692 , in which case both sides of this equation are @xmath85 .
the crossings @xmath47 incident @xmath279 are partitioned according to the other black region to which they are incident . using the formula worked out in the preceding case , we conclude that @xmath693 + 3 .
first suppose that no crossing @xmath47 meets the same region @xmath682 twice .
the value of the product @xmath681 can be gotten from figure [ f d(v , c)sigma ] .
the regions @xmath682 and @xmath279 abut along disjoint edges in the knot diagram ; here we view the knot diagram as a planar graph whose vertices are the crossings .
each edge ends at a pair of crossings @xmath695 and @xmath696 .
a simple check shows that the values @xmath697 and @xmath698 cancel one another regardless of the placement of @xmath699 and @xmath700 , using only the property that @xmath699 and @xmath700 lie in different regions .
it follows that the @xmath678-entry of the product @xmath679 is equal to @xmath85 in this case . in the event that some crossing meets the same region @xmath682 twice , essentially the same argument carries through . in this case the regions @xmath682 and @xmath279 abut along disjoint paths in the knot diagram , where vertices internal to a path correspond to crossings which the region @xmath682 meets twice . for each internal crossing
we have @xmath701 , and for the endpoints of the paths we get canceling contributions just as in the previous case .
consequently , @xmath702 in any event .
+ . pictured is the crossing @xmath47 , where the black dot represents the position of @xmath90 . assuming the region @xmath682 meets @xmath47 once ,
the value of @xmath703 is gotten by multiplying together the values in the regions @xmath682 and @xmath279.,title="fig:",height=144 ] ( -235,50)@xmath58 ( -315,50)@xmath59 ( -275,90)@xmath59 ( -275,10)@xmath58 ( -60,90)@xmath58 ( -60,10)@xmath59 ( -20,50)@xmath59 ( -100,50)@xmath58 cases 1 - 3 together establish ( [ e e ] ) .
now we settle ( [ e e0 ] ) .
we seek the identity @xmath704 recall that for @xmath705 , @xmath706 for all crossings @xmath47 , so the expression for @xmath650 reduces to @xmath707 .
suppose that @xmath90 is black .
then @xmath708 according as @xmath90 is not or is contained in the region @xmath279 , while @xmath709 regardless .
hence the product @xmath710 according as @xmath90 is not or is contained in the region @xmath279 .
recalling the orientation on @xmath16 induced by @xmath96 ( @xmath21[ss absgradings1 ] ) , this is to say that the edge @xmath711 is oriented into or out of the vertex @xmath279 .
next suppose that @xmath90 is white ; thus @xmath712 .
examining the few possibilities , we see that the value of the product @xmath713 is also @xmath278 according as the edge @xmath714 is directed into or out of the vertex @xmath279 . adding up all these contributions results in the identity ( [ e e2 ] ) and hence ( [ e e0 ] ) .
tracing backwards , we have at last settled ( [ e alphad ] ) , which completes the proof of theorem [ t domain ] .
as pointed out in @xmath21[ss h_2 ] , there is a strong similarity between the standard doubly - pointed heegaard diagram @xmath715 presenting the knot @xmath5 and the heegaard diagram @xmath716 presenting @xmath2 , starting with a marked diagram @xmath15 of @xmath0 . in this section
, we exploit this similarity to construct a pair of spectral sequences @xmath717 and @xmath718 which converge to @xmath719 and @xmath720 , respectively . both terms @xmath29 and @xmath721
are free abelian , generated by kauffman states of the diagram @xmath15 , and moreover the complexes @xmath722 and @xmath723 are isomorphic .
it follows that @xmath724 .
although the term @xmath31 is not a knot invariant , it admits a simple combinatorial description ( theorem [ t e1 ] ) , it is independent of the choice of complex structure on @xmath110 , and it is useful in making some of the calculations in @xmath21[s calculations ] .
the two spectral sequences arise out of a standard construction in heegaard floer homology that we describe in the next paragraph . after setting this up ,
the remainder of the section is directed towards establishing theorem [ t e1 ] . to begin with ,
arrange so that the restriction of each pair @xmath224 , @xmath114 to the top of @xmath110 is the same ; then the tops of both @xmath132 and @xmath213 are the same , and the intersection points in @xmath327 are in 1 - 1 correspondence with kauffman states of @xmath15 in both cases . choose a complex structure @xmath725 on @xmath110 and a generic complex structure on @xmath726 close to the one induced by @xmath725 .
the complex @xmath558 decomposes into a direct sum of subcomplexes , where a pair of generators @xmath561 belong to the same subcomplex iff there is a domain @xmath324 connecting them which satisfies @xmath727 .
place a basepoint in each region that meets the bottom of @xmath132 .
we define a relative filtration on each subcomplex by setting @xmath728 the sum over all basepoints @xmath323 distinct from @xmath292 and @xmath729 .
the domain @xmath324 is uniquely determined by the condition @xmath730 , so @xmath731 is well - defined . in the same way
, we obtain a relative filtration @xmath732 on each subcomplex @xmath733 , @xmath734 .
the relatively filtered complexes @xmath735 and @xmath736 give rise to a pair of affiliated spectral sequences @xmath717 and @xmath718 , respectively . both terms @xmath29 and @xmath721
are free abelian , generated by kauffman states of @xmath15 , and the sequences converge to @xmath719 and @xmath720 , respectively . both differentials @xmath30 and @xmath737 count pseudo - holomorphic disks @xmath738 with @xmath739 ; as @xmath740 for any pseudo - holomorphic disk @xmath738 connecting @xmath26 to @xmath423 , it follows that these differentials count disks @xmath738 for which the support of @xmath570 is contained in the top of its respective heegaard diagram .
moreover , the closure of the union of regions not containing a basepoint @xmath323 , @xmath292 , or @xmath729 is the same for both @xmath132 and @xmath213 .
it follows that @xmath741 , and consequently @xmath724 . in order to describe @xmath31
, we first examine this question : given a pair of generators @xmath26 and @xmath423 , when does there exist a domain @xmath324 connecting @xmath26 to @xmath423 which is supported on top of @xmath110 ? we can ask this question with reference to either @xmath132 or @xmath213 ; of course , the answer is the same in either case .
recall that the kauffman state @xmath26 gives rise to a spanning tree @xmath742 , which in turn induces an orientation of @xmath51 ( @xmath21[ss absgradings1 ] ) ; we refer to this as the orientation on @xmath51 _ induced _ by @xmath26 .
[ l support ] the pair of generators @xmath26 and @xmath423 are connected by a domain @xmath324 supported on top of @xmath110 iff @xmath90 and @xmath608 lie to the same side of the overstrand at @xmath47 for every crossing @xmath47 , which happens iff @xmath26 and @xmath423 induce the same orientation on @xmath51 ( equivalently , on @xmath16 ) .
it is easy to check that the generators @xmath26 and @xmath423 induce the same orientation on @xmath51 ( equivalently , on @xmath16 ) iff @xmath90 and @xmath608 lie to the same side of the overstrand at @xmath47 , for every crossing @xmath47 .
this is to say that @xmath743 ( @xmath21[ss db&w ] , equation ( [ e psivector ] ) ) . by equation ( [ e db&w1 ] ) , this is the case iff @xmath324 is supported on top of @xmath110 , taking note that the intersection matrix @xmath744 for @xmath2 is invertible ( @xmath334 ) .
following lemma [ l support ] , the complex @xmath722 decomposes into a direct sum of subcomplexes @xmath745 , where @xmath746 is generated by kauffman states @xmath26 satisfying @xmath747 .
we now describe how the generators of a given subcomplex @xmath746 are related .
thus , fix a pair of kauffman generators @xmath748 in @xmath746 .
let @xmath749 denote the bipartite graph of incidences between regions and crossings in the diagram @xmath15 .
the kauffman state @xmath26 has a natural interpretation as a matching @xmath750 in @xmath749 between the _ unmarked _ regions and crossings , and conversely every such matching corresponds to a kauffman state .
we construct another maximal matching @xmath751 in @xmath749 by pairing each crossing @xmath47 with the other region that meets @xmath47 to the same side of the overstrand as @xmath90 .
a kauffman state @xmath423 satisfies @xmath752 exactly when @xmath753 is a subset of the edges in @xmath754 .
notice that the subgraph on the edges in @xmath754 consists of some number of ( even ) cycles @xmath755 , as well as two star - shaped components with star vertex at either marked region . therefore , there are @xmath756 such kauffman states @xmath423 .
[ d solitary ] the kauffman state @xmath26 is _ solitary _ if there are no cycles in @xmath754 .
thus a solitary kauffman state @xmath26 is uniquely determined by the vector @xmath757 , and it alone generates @xmath746 .
examining @xmath746 more closely in the general case , two - color the edges of @xmath755 red and blue .
the edges of @xmath749 can be identified with the intersection points between the @xmath19 and @xmath20 curves in @xmath132 ( aside from the single intersection point on @xmath216 ) .
let @xmath758 denote the tuple of intersection points corresponding to the red edges and @xmath257 the tuple corresponding to the blue edges of @xmath759 for every @xmath158 , and let @xmath760 denote the tuple of intersection points corresponding to the edges of @xmath750 not in any @xmath759 . then the kauffman generators of @xmath746 are the elements of the set @xmath761 .
next , we describe the domain connecting a pair of generators @xmath762 ( cf .
@xcite , section 2.2 for an alternative approach ) .
we begin by constructing a @xmath149-cycle @xmath121 in @xmath110 connecting the components of @xmath26 and @xmath423 , as in @xcite , section 2.4 . for every bounded region
@xmath92 with @xmath763 , traverse the portion of the curve @xmath764 that runs clockwise from @xmath91 to @xmath765 ; if @xmath92 is unbounded , traverse it in the counterclockwise direction .
for every crossing @xmath47 for which @xmath766 , traverse the portion of the curve @xmath593 that runs between @xmath608 and @xmath90 atop @xmath110 .
the union of these oriented portions of curves is the @xmath149-cycle @xmath121 .
this @xmath149-cycle decomposes into a union of disjoint , oriented curves @xmath223 which connect the points of @xmath758 and @xmath257 , one for each @xmath158 for which @xmath767 .
we regard the @xmath223 as a collection of disjoint , oriented curves in the plane of the knot diagram @xmath15 .
every arc @xmath84 in the diagram @xmath15 is separated from the unbounded region by some subset of the @xmath223 , and we set @xmath768 equal to the signed count of the curves in this subset , where @xmath223 gets weight @xmath278 according to whether it is oriented counterclockwise or clockwise .
identify an arc @xmath84 with the corresponding region in the heegaard diagram @xmath132 ( equivalently , @xmath213 ) .
see figure [ f e1domain ] .
( corresponding to the open circles ) and @xmath423 ( corresponding to the filled - in circles ) .
( b ) the three oriented curves @xmath223 connecting @xmath26 to @xmath423 , pictured in green .
( c ) the domain @xmath324 connecting @xmath26 to @xmath423 , interpreted as kauffman generators .
the lightly shaded regions have coefficient @xmath59 and the darkly shaded region has coefficient @xmath430.,title="fig:",width=480 ] ( -250,300)(a ) ( -360,140)(b ) ( -160,140)(c ) [ l e1domain ] the domain connecting @xmath26 to @xmath423 is @xmath769 we remark that @xmath770 does not equal @xmath121 on the nose in general , but the following proof implies that the two always differ by some number of copies of full @xmath19 curves .
let @xmath771 and let @xmath324 denote the domain connecting @xmath26 to @xmath423 .
the domain @xmath324 is supported on top of @xmath110 , hence so is @xmath770 .
it follows that @xmath772 consists of the oriented arc of @xmath593 from @xmath608 to @xmath90 on top of @xmath110 for every crossing @xmath47 ( which is void in case @xmath773 ) . on the other hand
, this is precisely @xmath774 restricted to the top of @xmath110 . according to @xmath21[ss dalpha ] , this is sufficient to conclude that the restrictions of @xmath324 and @xmath775 to the top of @xmath110 agree .
since @xmath324 is supported on top of @xmath110 , every region which meets both the top and bottom of @xmath110 appears with @xmath85 coefficient in @xmath324 and hence in @xmath775 .
lastly , every region supported on the bottom of @xmath110 appears with coefficient @xmath85 in @xmath775 .
it follows that @xmath776 , as desired . in the case that @xmath777 contains a single cycle @xmath759 ,
the @xmath149-cycle @xmath121 connecting @xmath26 to @xmath423 is the single curve @xmath223 , which we may take to be oriented counter - clockwise by switching the roles of @xmath26 and @xmath423 if necessary . in this case
, the domain @xmath324 in this case is a _ punctured polygon _ according to the following definition .
recall that a domain has an _ interior corner _ at an intersection point @xmath778 if it has coefficient @xmath149 on one of the four corners of regions incident @xmath536 and @xmath85 on the three others .
[ d polygon ] a domain @xmath324 from @xmath26 to @xmath423 is a _ punctured polygon _ if ( a ) it is a planar region with every coefficient @xmath85 or @xmath149 , ( b ) there are no components of @xmath26 and @xmath423 interior to @xmath324 , ( c ) all but one component of @xmath770 is a full @xmath19 curve , and ( d ) all corners of @xmath324 are interior .
given a punctured polygon @xmath324 from @xmath26 to @xmath423 , fix a component @xmath779 of @xmath423 . if @xmath780 lies on @xmath324 , let @xmath781 denote the maximal arc of @xmath224 which lies in @xmath324 and has both its endpoints on @xmath770 ; by definition [ d polygon ] ( b ) , @xmath780 is one of these endpoints . if @xmath780 does not lie on @xmath324 , let @xmath782 . now define a directed graph @xmath783 on the @xmath19 curves by declaring @xmath784 to be an edge if @xmath781 runs from @xmath780 to a point on @xmath116 .
the graph @xmath783 contains a directed cycle corresponding to its unique polygonal boundary component , as well as a directed loop at each @xmath115 for which @xmath782 .
[ d arborescent ] the punctured polygon @xmath324 is _ arborescent _ if there are no other cycles in the graph @xmath783 .
it is easy to see that the property of arborescence is equivalent to the property that @xmath785 is connected .
the importance of definition [ d arborescent ] is the following fact , which is in essence lemma 3.11 of @xcite .
[ l polygonlemma ] if the domain of a homotopy class @xmath571 is a punctured polygon , then @xmath786 ; if in addition it is arborescent , then @xmath787 , for any choice of complex structure on @xmath726 sufficiently close to one induced by a complex structure on @xmath110 . of a map @xmath738
for which @xmath788 will depend on the choice of complex structure.,title="fig:",width=192 ] ( -77,100)@xmath789 ( -77,70)@xmath790 ( -150,0)@xmath84 ( 0,140)@xmath84 ( 0,0)@xmath83 ( -150,140)@xmath83 ( -100,62)@xmath26 ( -50,77)@xmath26 ( -100,77)@xmath423 ( -50,62)@xmath423 we point out that the condition ( b ) in definition [ d polygon ] is absent from the original statement of this lemma in @xcite , which is necessary for its conclusion to hold ( @xmath791 otherwise ) . reassuringly , this additional condition is always met in applications of that lemma in @xcite .
observe also that the assumption on arborescence is essential in lemma [ l polygonlemma ] .
for consider the punctured polygon @xmath324 depicted in figure [ f poly ] . if @xmath792 for some homotopy class @xmath738 , then @xmath786 , but the value of @xmath788 will depend on the choice of complex structure .
thanks to zoltn szab for indicating this example .
the assertion that @xmath786 follows from lipshitz s formula for the maslov index ( @xcite , section 4 , esp .
corollary 4.3 ) . now suppose that @xmath324 is arborescent .
endow @xmath110 with an arbitrary complex structure , and let @xmath775 denote the decorated riemann surface obtained by erasing @xmath793 from @xmath324 , retaining its complex structure . in the event that @xmath793 is disjoint from @xmath324 , we have @xmath776 , and @xmath787 according to @xcite , lemma 3.11 .
moreover , the proof of @xcite , lemma 3.11 shows how to construct a heegaard diagram for which @xmath794 ( as riemann surfaces ) for some homotopy class @xmath795 , and that @xmath796 in this case .
let us compare @xmath797 and @xmath798 .
thus , let @xmath799 and @xmath800 .
the maximum principle implies that the image of @xmath682 is contained in @xmath801 , where @xmath802 are those components for which @xmath803 .
similarly , the image of @xmath682 is contained in @xmath804 .
since @xmath324 and @xmath775 are isomorphic as riemann surfaces , it follows that we have an identification of moduli spaces @xmath805 .
now @xmath787 follows .
although the complex @xmath722 will in general depend on the choice of complex structure on @xmath110 , lemmas [ l e1domain ] and [ l polygonlemma ] provide a sufficient understanding to enable the computation of its homology @xmath31 . as a motivation ,
let us return to the example described in figure [ f poly ] .
suppose the domain @xmath324 pictured there appears in a particular heegaard diagram , place a basepoint in each region besides @xmath789 and @xmath790 , and consider the floer chain complex whose differential counts holomorphic disks which avoid these basepoints .
choosing the heegaard diagram suitably , we can find a quadruple of generators which coincide away from the pictured region and meet it in one of four ways : @xmath806 .
this quadruple induces a subcomplex which looks as shown : @xmath807_{\d_2 } \ar[rrd]^{\d } & & \\ ( a , y ) \ar[rrd]_{\d } & & & & ( b , x ) \ar[lld]^{\d_2 } \\ & & ( b , y ) & & } \ ] ] both of the maps labeled by @xmath790 support a unique holomorphic representative for every choice of complex structure , thanks to the riemann mapping theorem ( a special case of lemma [ l polygonlemma ] ) . from this alone
we can conclude that the subcomplex pictured is acyclic , regardless of the value @xmath808 , @xmath809 .
the proof of the following lemma is in essence a generalization of this particular example .
[ l acyclic ] the subcomplex @xmath746 is acyclic if it has rank @xmath810 .
suppose that @xmath746 has rank @xmath811 .
number the ( unoriented ) curves @xmath812 in such a way that if @xmath223 is contained interior to @xmath374 , then @xmath298 .
select a pair @xmath762 so that @xmath759 is the unique cycle in @xmath777 and @xmath223 is oriented counterclockwise .
we pin down @xmath758 and @xmath257 ( which previously depended on an arbitrary two - coloring of @xmath759 ) by declaring that @xmath813 and @xmath814 . for @xmath815 and @xmath816 , let @xmath817 denote those generators @xmath818 for which @xmath819 , let @xmath820 , and for @xmath821 , let @xmath822 denote the subgroup of @xmath746 induced on @xmath817 .
we claim that @xmath822 is a subcomplex of @xmath746 . for suppose that @xmath823 and @xmath824 for some @xmath825 . in the @xmath149-cycle connecting @xmath26 to @xmath423
, @xmath826 is oriented clockwise and is not contained interior to any other curve .
it follows by lemma [ l e1domain ] that the domain @xmath570 connecting @xmath26 to @xmath423 has a negative coefficient , so that @xmath827 by the non - negativity principle .
hence @xmath828 , as desired .
consequently we obtain a filtration @xmath829 .
we prove by induction on @xmath160 that @xmath822 is acyclic ; the case @xmath830 is the statement of the lemma .
thus suppose @xmath831 , and let @xmath26 and @xmath423 denote the kauffman generators of @xmath832 with @xmath833 and @xmath834 .
we claim that the domain @xmath324 connecting @xmath26 to @xmath423 is arborescent .
for if it were not , let @xmath835 denote the vertices of a non - trivial directed cycle in @xmath783 .
we can decompose @xmath836 so that @xmath837 and @xmath789 is a punctured polygon .
let @xmath254 denote the corner points of @xmath789 that are in @xmath26 and @xmath838 the other corner points of @xmath789 .
then @xmath839 is another kauffman generator , and @xmath789 connects the pair @xmath840 .
it follows that @xmath841 as well , but the curve @xmath842 connecting the pair @xmath843 is contained interior to @xmath844 , a contradiction . hence @xmath324 is an arborescent punctured polygon ; now an application of lemma [ l polygonlemma ] shows that @xmath832 is acyclic .
for the induction step , suppose that @xmath822 is acyclic for some @xmath845 , and consider the short exact sequence @xmath846 the bijection @xmath847 via @xmath848 induces a group isomorphism @xmath849 .
the domain connecting a pair of generators @xmath850 is the same as the domain connecting the pair @xmath851 .
but if a pair of homotopy classes @xmath738 and @xmath852 have the same domain , then @xmath805 .
it follows that @xmath218 defines a chain isomorphism , so both @xmath822 and @xmath853 are acyclic .
now the long exact sequence in homology implies that @xmath854 is acyclic .
this completes the induction step and the proof of the lemma .
[ t e1 ] the group @xmath31 is free abelian , generated by solitary kauffman states .
the complex @xmath722 decomposes into a direct sum of subcomplexes @xmath745 .
those of rank @xmath149 are generated by solitary kauffman states , and those of rank @xmath855 are acyclic . as a passing remark , theorem [ t e1 ] and the fact that the differential on @xmath856 vanishes for an alternating diagram together imply that every kauffman state in an alternating diagram is solitary .
more substantially , we have the following result .
[ c e1collapse ] if the knot @xmath0 admits a diagram @xmath15 for which the number of solitary kauffman states is equal to @xmath246 , then @xmath857 and @xmath2 is an l - space .
note that it is unknown whether @xmath857 implies that @xmath0 is @xmath13-thin .
this is immediate from the construction of the spectral sequences , theorem [ t e1 ] , and the inequalities @xmath858 and @xmath859 . in the effort to find knots
whose branched double - covers are l - spaces , it is an interesting problem to find examples of non - alternating knots satisfying the hypothesis of corollary [ c e1collapse ] .
certainly there are non - alternating diagrams whose number of solitary kauffman states equals @xmath246 .
for example , changing one crossing in the alternating diagram of the trefoil knot gives a diagram of the unknot , and for a suitable placement of the marker @xmath46 there will be a unique solitary kauffman state .
however , a different placement of @xmath46 can result in a marked diagram with three solitary kauffman states .
in this section we explain how to use the foregoing results to make calculations of @xmath3 , where @xmath0 is a link with non - zero determinant , and in particular carry this out for some knots with @xmath33 crossings . in general
, suppose that @xmath132 is a heegaard diagram presenting a rational homology sphere @xmath35 , and we wish to compute @xmath34 as the homology of the chain complex @xmath558 .
one proceeds by enumerating the intersection points @xmath327 , calculating their associated @xmath7 structures , and ( in the case that @xmath132 fits into a heegaard triple subordinate to a link ) writing down their absolute gradings .
the hard work is in determining the differential on @xmath558 . in general
this depends on the choice of complex structure on the heegaard surface @xmath110 and a perturbation of the induced complex structure on @xmath726 .
however , in small examples , one can often circumvent the difficult analysis of moduli spaces en route to the answer by applying a few simple ideas which we discuss below . for now , simply note that the holomorphicity condition restricts our attention to pairs of generators connected by a domain with non - negative coefficients . for the case of interest , the above data can be determined in a completely algorithmic way .
we start with a marked diagram @xmath15 of @xmath0 with its regions two - colored . from this
we write down its coloring matrix and goeritz matrices , and enumerate its kauffman states .
each kauffman state corresponds to a generator of @xmath27 , whose absolute grading and associated @xmath7 structure are determined by theorem [ t grading ] and @xmath21[ss spinc ] , respectively .
theorem [ t domain ] enables us to determine the domain connecting a pair of generators , and in particular enables us to judge for which pairs the non - negativity condition is met .
we wrote code in mathematica to do this ( cf . a related program by ciprian manolescu @xcite ) .
to input a given example into the program , one begins with a marked diagram of a knot and enters in by hand the white and black graphs , the crossing incidence numbers @xmath587 , and the coloring matrix . in practice , this process is somewhat laborious , and ( at least in the hands of the author ) susceptible to careless errors which may take a while to detect and fix .
therefore , it is sometimes preferable to make the calculation by another method .
for example , for the case of an alternating knot , one knows that @xmath2 is an l - space ( cf .
corollary [ c lspace ] ) , so what remains to calculate are its correction terms .
this can be done as explained in @xmath860 ( indeed , calculations of this sort were carried out in @xcite and @xcite ) . for the case of a montesinos knot
, @xmath2 can be realized as the boundary of a negative definite plumbing on a tree with one bad vertex , and the algorithm of @xcite enables a calculation of @xmath861 .
all but twenty knots with @xmath33 crossings belong to one of these two types ( @xcite , table f.2 ) . for the remainder , we calculated @xmath3 using our program , making use of the minimum crossing diagram for @xmath0 from rolfsen s table ( @xcite , appendix c ) .
when performing a calculation , it is extremely helpful to know in advance something about the structure of the answer .
for example , if @xmath0 is _ quasi - alternating _ , then @xmath2 is an l - space ( @xcite , definition 3.1 and proposition 3.3 ) , and if @xmath0 has braid index @xmath6 , then this is nearly so : @xmath3 is torsion - free , and it is monic in all but at most one @xmath7 structure @xcite .
the following exceptional knots are quasi - alternating : @xmath862 , @xmath863 , and @xmath864 , @xmath865148 , 149 , 150 , 151 , 155 , 156 , 157 , 158 , 159 , 162 , 163 , 164 , 165@xmath866 . for @xmath862 and @xmath863
, pictures are shown in ( @xcite , figure 10 ) ; for those of braid index 3 , this follows from ( @xcite , theorem 5.6 ) . for the remainder ,
it is possible to find a crossing in the rolfsen diagram of @xmath0 for which the resolution @xmath867 is quasi - alternating , the resolution @xmath868 is a two - component alternating link , and @xmath869 .
furthermore , we can argue that @xmath870 is an l - space , although we could not determine whether this knot is quasi - alternating .
to see this , there is a crossing in its rolfsen diagram for which one resolution is @xmath871 and the other is @xmath872 .
the knot @xmath871 is not quasi - alternating , but @xmath873 is an l - space ( one can show this either using the spanning tree model or the algorithm of @xcite ) , @xmath872 is a non - split alternating link , and @xmath874 .
now the argument of @xcite , proposition 3.3 applies to show that @xmath870 is an l - space .
in addition to the knots listed here with l - space branched double - covers , the knots @xmath875 and @xmath876 have braid index @xmath6 .
thus , it is typically the case that we know in advance that @xmath877 is free and monic , and what remains to determine is the correction term @xmath534 in which this group is supported .
suppose we are in this position in a particular example . if @xmath878 is supported in a pair of gradings for some particular @xmath7 structure @xmath532 , then @xmath879 for some @xmath880 and @xmath881 .
in this case we can conclude from the euler characteristic that @xmath882 . alternatively , suppose that there is a unique solitary kauffman generator in a particular @xmath7 structure @xmath532 . in this case theorem [ t e1 ]
implies that @xmath534 equals the grading of this generator .
if @xmath878 does not have one of the two foregoing properties , it is possible that @xmath883 does , in which case we can invoke the conjugation symmetry @xmath884 .
occasionally , neither @xmath878 nor @xmath883 has either property . in this case
, we can still read off @xmath885 from the grading of any generator of one of these two groups .
if @xmath886 implies that @xmath887 or @xmath888 , then we can try to change the position of the marked point @xmath46 to get a new marked diagram @xmath672 .
having done so , we can locate the pair of subcomplexes @xmath889 and @xmath890 by the gradings of their generators @xmath891 , and see whether we can apply one of the preceding observations to one of the new complexes to determine @xmath534 . for the eighteen knots enumerated above , these techniques suffice to determine @xmath534 every time we know that @xmath877 is monic . indeed , for many of these knots
, @xmath246 is relatively close to the number of generators of @xmath27 , so that frequently @xmath878 consists of very few generators .
in addition , for a suitable choice of marked point @xmath46 in the diagrams of @xmath875 and @xmath892 , and @xmath893 the @xmath894 structure on @xmath2 , the spectral sequence of @xmath21[s specseq ] which computes @xmath895 collapses at the @xmath31 term . thus we can easily calculate @xmath895 in these two cases as well , even though these groups are not monic .
the two remaining knots @xmath896 and @xmath897 are more difficult to handle . for one thing ,
@xmath898 , so in this case we wind up considering a chain complex with @xmath899 generators in a single @xmath7 structure .
in addition , we have no a priori result asserting that @xmath3 is torsion - free for either of these knots , although this turns out to be the case .
instead , we must analyze @xmath27 more closely , applying the foregoing techniques with some additional observations .
for example , it is commonly the case that we have a pair of generators @xmath26 and @xmath423 for which @xmath900 and @xmath26 is not connected to any other generator by a non - negative domain .
then @xmath901 , and @xmath902 implies that @xmath903 .
this latter fact may not be so clear judging from domains of maps , as @xmath423 could be connected to some other generator @xmath292 by a non - negative domain @xmath324 . in any event
, we learn that @xmath904 for any @xmath738 with @xmath792 .
that is , if there is some other pair of generators @xmath905 and @xmath729 connected by this same domain @xmath324 , then we can conclude that @xmath906 does not involve any contribution from @xmath729 .
one can run this sort of argument and close in on the answer by comparing what the different complexes @xmath27 reveal by moving the marked point around . as a final resort ,
it is necessary on occasion to draw the domain connecting a pair of generators which is not counted by @xmath30 , recognize it as an arborescent punctured polygon , and apply lemma [ l polygonlemma ] . with a good deal of perseverance , these techniques enable the determination of @xmath3 , @xmath907 and @xmath897 .
the calculation of @xmath3 for the twenty non - alternating , non - montesinos knots @xmath0 with @xmath33 crossings is reported as follows .
we identify @xmath908 via @xmath909 , and when @xmath2 is an l - space , we record the element @xmath910.\ ] ] we express an element of @xmath911 as a monomial @xmath912 or @xmath913 , depending on the number of generators of this group . in the four instances in which @xmath2 is not an l - space ,
@xmath877 has rank one except in the @xmath894 structure @xmath893 , and @xmath895 takes the form @xmath914 , where @xmath915 . in this case
we write @xmath916 thus , for example , @xmath917 .
we briefly point out that these calculations can be used to determine @xmath861 from the algebraic fact that @xmath918 with @xmath919 implies that @xmath920 .
when @xmath911 is cyclic , the conjugation symmetry manifests itself as the equality between the coefficients on terms @xmath912 and @xmath921 .
when @xmath922 , we compare the coefficients on terms @xmath913 and @xmath923 instead .
the rational number appearing in front of each expression is @xmath924 or @xmath925 when @xmath911 is cyclic , and @xmath926 or @xmath927 when @xmath922 with @xmath928 .
1 . @xmath929 2 .
@xmath930 3 .
@xmath931 4 .
@xmath932 5 .
@xmath933 6 .
@xmath934 7 .
@xmath935 8 .
@xmath936 9 .
@xmath937 10 .
@xmath938 11 .
@xmath939 12 .
@xmath940 13 .
@xmath941 14 .
@xmath942 15 .
@xmath943 16 .
@xmath944 17 .
@xmath945 18 .
@xmath946 19 .
@xmath947 20 .
theorem [ t grading ] expresses the maslov grading of a kauffman generator @xmath26 in @xmath294 : @xmath949 notice that for a pair of generators @xmath26 and @xmath423 connected by the differential @xmath30 , we have @xmath950 , and the domain connecting @xmath26 to @xmath423 is supported on top of @xmath213 , implying that @xmath951 .
it follows in this case that @xmath952 , i.e. _ @xmath30 lowers @xmath4 by @xmath149_. in fact , in small examples we found that it is always the case that the full differential @xmath396 lowers @xmath4 by @xmath149 .
this phenomenon is particularly striking in examples when the reduced white graph @xmath57 is a tree and the goeritz matrix @xmath66 is negative definite . in other words
, @xmath2 is the boundary of a negative definite plumbing on a tree . as a representative case ,
consider the well - known plumbing on the @xmath953 diagram , in which every vertex receives weight @xmath954 .
this is the reduced white graph of a particular diagram of the @xmath955 torus knot , and the manifold it describes is the poincar dodecahedral space @xmath956 . in this case
, the spanning tree model has @xmath957 generators , each supported in the unique @xmath7 structure and ( notably ) in different gradings .
we represent each generator by a dot @xmath958 and draw an arrow between two dots @xmath959 if the generator to the left is connected to the one on the right by a non - negative domain .
each arrow therefore represents a potential differential connecting a pair of generators . in this way
, the complex @xmath960 decomposes as a union of linear chains @xmath961 , one of which terminates on the unique generator in grading @xmath954 .
now , it is known that @xmath962 ( cf .
@xcite , section 8 or @xcite , section 3.2 ) .
this enables us to conclude that in each linear chain , the maps alternate between isomorphisms and trivial maps . inspecting the complex more closely
, we notice that the trivial maps connect pairs of generators for which @xmath963 , while the isomorphisms connect pairs for which @xmath964 .
these facts are not at all obvious by considering the domains of maps . for example , one of the domains corresponding to a non - trivial map has a region with coefficient @xmath965 , which is remarkably large for such a small example . in the complex @xmath27 ,
the differential @xmath396 lowers @xmath4 by @xmath149 .
the induced @xmath4-grading on the homology group @xmath3 is independent of @xmath15 up to an overall shift , and can be pinned down to a well - defined integer absolute grading . in case conjecture [ conj deltagrading ] were true , we would naturally like to compare the @xmath4-grading on @xmath3 with that on the khovanov homology @xmath9 and knot floer homology @xmath8 . in particular , we might ask whether the inequality of ranks @xmath966 implied by the ozsvth - szab spectral sequence persists at the level of @xmath4-graded groups . as it stands , that spectral sequence is not known to reflect any of the finer features of the two theories involved : the integer bigrading on @xmath967 , or the rational grading on and decomposition into @xmath7 structures of @xmath3 .
note that the quantity @xmath28 is the same as the @xmath279-grading in the spanning tree complex for khovanov homology in @xcite .
we also note that in many small examples , such as knots on @xmath968 crossings and several pretzel and torus knots , we observed an inequality of ranks @xmath969 . for knots on @xmath968 crossings
this is no surprise , since @xmath970 in this range .
it is tempting to speculate on the relationship between @xmath3 and @xmath8 , due to the strong resemblance between the heegaard diagrams @xmath971 presenting @xmath5 and @xmath23 presenting @xmath2 .
the fact that the @xmath30 differential in the two spectral sequences of section [ s specseq ] is the same is particularly curious , since we know a priori that it respects the alexander filtration grading on @xmath13 ( absent in @xmath3 ) and the decomposition into @xmath7 structures of @xmath3 ( absent in @xmath13 ) . in a different direction ,
andrs nmethi has proposed a very interesting model for the floer homology @xmath972 of a negative - definite plumbed rational homology sphere ( @xcite , esp .
section 3 and conjecture 5.2.4 ) . in this model
, @xmath972 is enhanced by an additional non - negative integer grading .
any plumbed rational homology sphere takes the form @xmath2 for some link @xmath0 with @xmath334 .
it is very interesting to compare nmethi s model to our own , especially with a view towards our conjecture [ conj deltagrading ] and his conjecture 5.2.4 . | given a diagram of a link @xmath0 in @xmath1 , we write down a heegaard diagram for the branched - double cover @xmath2 .
the generators of the associated heegaard floer chain complex correspond to kauffman states of the link diagram . using this model we make some computations of the homology @xmath3 as a graded group .
we also conjecture the existence of a @xmath4-grading on @xmath3 analogous to the @xmath4-grading on knot floer and khovanov homology . |
in the work of lehmann , newton , and wu @xcite , the kobayashi - maskawa @xcite matrix is expressed in terms of the masses of the three generations of quarks : @xmath10 this is accomplished by introducing a new horizontal symmetry .
some of the earlier attempts in this direction are given in @xcite .
recent experiments at super - kamiokande indicate the presence of neutrino oscillations , which would imply that the neutrinos are not all massless .
if it is accepted that the neutrinos are not massless , then it is most natural in the standard model @xcite to introduce three right - handed neutrinos in addition to the three known left - handed ones . in this way
, there are six quarks and six leptons . in this paper , the consequences of a universal quark - lepton mixing are studied . in other words , the method of @xcite
is used to express the lepton km matrix and the neutrino mixing matrix in terms of the masses of the three generations of leptons : @xmath11 of course the masses of the three charged leptons are accurately known , leaving as unknown parameters the masses of the three neutrinos .
thus there are three parameters to be determined instead of seven , the three masses plus the four in the lepton km matrix .
it is the purpose of this paper to use the data from solar neutrinos @xcite and atmospheric neutrinos @xcite to determine the three neutrino masses separately , not only the differences of their squares .
the quark mixing matrix proposed by lehmann et al . @xcite relates the mixing to the actual quark masses .
these mass matrices are each given by four parameters @xmath12 , @xmath13 , @xmath14 , and @xmath15 , and , when applied to the leptons without modification , take the form @xmath16 \label{eq : mu - nondiag } m(\nu)&=&\begin{pmatrix } 0 & id(\nu ) & 0
\\ -id(\nu ) & c(\nu ) & b(\nu ) \\ 0 & b(\nu ) &
a(\nu ) \end{pmatrix},\end{aligned}\ ] ] with @xmath17 the diagonalization of these mass matrices @xcite is achieved by the orthogonal matrices @xmath18 and @xmath19 , where @xmath20 \label{eq : diagonal - u } m(\nu)&=&{{\rm diag}}(-i,1,1)r(\nu)m_{\rm diag}(\nu)r^{\rm t}(\nu){{\rm diag}}(i,1,1).\end{aligned}\ ] ]
if there is @xmath21 violation in the lepton sector , then the imaginary entries in ( [ eq : mu - nondiag ] ) and ( [ eq : diagonal - u ] ) are required .
if by any chance there is no @xmath21 violation in the lepton sector , then both @xmath22 and @xmath23 there should be replaced by 1 .
the diagonal mass matrices have the form @xmath24 where @xmath25 and @xmath26 we shall here only be concerned with @xmath27 and @xmath19 .
in dealing with quarks @xcite , the observed quark masses allow a much stronger form for the inequality ( [ eq : mass - order ] ) , namely @xmath28 the lack of direct experimental data on the neutrino masses implies that ( [ eq : mass - order ] ) can be used , but not ( [ eq : mass - order - quarks ] ) .
the first task is therefore to determine the allowed region in the space @xmath29 , which must be between those permitted by ( [ eq : mass - order ] ) and ( [ eq : mass - order - quarks ] ) .
the parameters @xmath12 , @xmath13 , @xmath14 , and @xmath15 are related to the masses by the following conditions , @xmath30 the cubic equation for the parameter @xmath12 is then @xmath31 any real cubic equation can have either one or three real solutions . where there is _ one _ real solution , that one is negative , and thus unphysical , as is seen from ( [ eq : m1m2m3 ] ) . where there are _ three _ real solutions , one of them is negative , while two are positive .
we shall refer to these two positive solutions as solution 1 ( larger @xmath12 ) and solution 2 ( smaller @xmath12 ) .
these considerations can be used to determine the allowed physical region in the @xmath32 plane , as shown in fig .
[ fig : disc ] .
this region is only slightly larger than the triangle given by the inequality ( [ eq : mass - order - quarks ] ) , with two additional regions , one where @xmath33 and the other a very small one with @xmath34 .
[ fig : disc ] ( 12,9.5 ) ( 2.0,1.0 )
the coupled equations satisfied by the three neutrino wave functions are @xcite @xmath35 \begin{pmatrix } \phi_1(r ) \\
\phi_2(r ) \\
\phi_3(r ) \end{pmatrix},\ ] ] where @xmath36 , with @xmath37 the fermi weak - interaction constant and @xmath38 the solar electron density at a distance @xmath39 from the center of the sun .
furthermore , we denote @xmath40 , @xmath41 , @xmath42 . the evolution of the neutrino wave functions is determined by the squared mass matrix , @xmath43 ^ 2 = \begin{pmatrix } d^2 & cd & bd \\ cd & b^2+c^2+d^2 & b(a+c ) \\ bd & b(a+c ) & a^2+b^2 \end{pmatrix } \equiv\begin{pmatrix } m^2_{11 } & m^2_{12 } & m^2_{13}\\ m^2_{21 } & m^2_{22 } & m^2_{23}\\ m^2_{31 } & m^2_{32 } & m^2_{33 } \end{pmatrix},\ ] ] the neutrino momentum , @xmath44 , and the solar electron density . here , @xmath45 .
the eigenvalues of the squared mass matrix ( multiplied by @xmath46 , with @xmath47 defined below ) are denoted @xmath48 , @xmath49 , and @xmath50 , and ordered such that @xmath51 it is actually a good approximation to take an exponential electron density , @xmath52 .
a fit to the solar density as given by @xcite leads to @xmath53 km .
for this case of an exponential solar density , the three - component wave equation can be solved in terms of generalized hypergeometric functions , @xmath54 @xcite .
we scale and shift the radial variable , @xmath55 , with @xmath56 determined such that @xmath57 the above equation ( [ eq : schr-1 ] ) may then be written as @xmath58 where @xmath59 and @xmath60 denote the eigenvalues of the @xmath61 mass matrix involving @xmath62 and @xmath63 . with @xmath64 ,
one finds for @xmath65 a differential equation of the form @xmath66 \psi_1=0\ ] ] the solutions of which , @xmath67 can be given in terms of generalized hypergeometric functions @xmath54 as @xmath68\ ] ] with @xmath69 , @xmath70 , @xmath71 .
for the other flavours , @xmath72 and @xmath73 are given by similar expressions , with shifted parameters .
there is not much information about these functions @xmath54 in the literature @xcite . for the case of two flavours , the products in ( [ eq : ode ] ) go only up to @xmath74 , and
a familiar confluent hypergeometric function @xmath75 ( also denoted whittaker function or parabolic cylinder function ) is obtained @xcite . in order to impose the boundary conditions that only _ electron _
neutrinos are produced in the sun , we have to determine these functions at large and negative values of @xmath76 .
the series expansion is in principle convergent , but it is not practical for large absolute values of both parameters and the argument .
instead , methods have been developed for the evaluation by a combination of relating them to @xmath77 and asymptotic methods @xcite .
since the only unknown parameters for the present theory of universal quark - lepton mixing are the three neutrino masses , it remains to determine these masses using experimental data on solar and atmospheric neutrinos .
for the atmospheric neutrinos we take the 8 data points for @xmath78 and the 8 data points for @xmath79 , as reported recently @xcite .
these sixteen data points are treated as separate inputs , but we allow two overall normalization constants for the two sets of data . for the solar - neutrino data , we use the total rates from the chlorine experiment @xcite , the gallium experiments @xcite ( we average the two results ) and the super - kamiokande experiment @xcite .
we adopt the neutrino energy spectra and detector efficiencies as given by bahcall et al . , and , for the latter detector , we also include the neutral - current cross section @xcite .
the determination of the three neutrino masses is to be carried out by a @xmath0 fit to these pieces of data .
since the number of degrees of freedom is 14 , a good @xmath0 fit would give support to this idea that the mass mixing is universal for quarks and leptons .
for this purpose , we have scanned the entire ( @xmath80 , @xmath81 , @xmath82 ) parameter space .
it is the necessity to cover this entire space that makes it essential to develop the theory described in sec . 3 here .
a few simple conclusions can be drawn from this set of mass values .
first , with 14 degrees of freedom , the three values of @xmath0 must be considered to be quite good .
_ this is evidence in favour of universal quark - lepton mixing_. in our mind the difference in these three values of @xmath0 is not significant .
secondly , the three values for @xmath81 are the same a surprise to us .
this presumably means that the relative accuracy in the present determination of @xmath81 is better than those of @xmath80 and @xmath82 .
this can be understood in the following way .
the relative accuracy for @xmath80 is low because @xmath80 itself is quite small , and the value of @xmath0 is not very sensitive to such small masses .
the average of these three @xmath80 is given in the abstract , and very roughly we guess the error to be about a factor of two .
on the other hand , the uncertainly in the value of @xmath82 is simply due to the three minima giving different values .
it is a matter of which one of the three minima is physically the correct one . to demonstrate the scanning of the entire region of allowed values of neutrino masses , we show in fig .
[ fig : sol2-lvls - fin ] an example of such a scan showing contours of constant @xmath0 . in this figure ,
the value of @xmath82 used is 0.058 ev , corresponding to the minimum ( b ) above .
it is from such scans that we know there are only three minima . in this figure , some minor irregularities along the edges of the allowed region are artifacts of the finite grid spacing . it is difficult to compare the present result with the previously given allowed regions in the parameter space .
the reason is that the allowed regions have typically been given on the basis of the mixing of two neutrino species , while there is significant mixing among all three neutrinos , or the studies concentrate on _ either _ solar _ or _ atmospheric neutrinos @xcite .
the following comparisons are nevertheless of interest .
\(a ) it is probably correct to compare the @xmath88 here with the @xmath89 from atmospheric - neutrino data . in this case
, the @xmath88 for the three minima covers roughly the allowed range of @xmath89 , with minimum ( b ) near the center , and minima ( a ) and ( c ) near the edges of the allowed region .
\(b ) it is probably not too far wrong to identify @xmath90 here with the @xmath89 from solar - neutrino data .
if so , the values are reasonably close , except with the so - called `` low - mass low - probability '' ( low ) solution @xcite .
\(c ) it is more difficult to discuss the strength of the coupling
. it can nevertheless be concluded , in the context of solar neutrinos , that the present solutions are significantly closer to the `` large - mixing - angle '' ( lma ) solution than to the `` small - mixing - angle '' ( sma ) solution . in connection with the problem of distinguishing between the three minima ,
more accurate data from super - kamiokande and related experiments are needed .
also , a better understanding of the high - energy @xmath91 neutrino flux @xcite could make the electron recoil energy spectrum useful for this purpose .
furthermore , important information may be forthcoming also from the long - baseline experiments @xcite and exotic atom experiments @xcite .
we are greatly indebted to professor harry lehmann and professor jack steinberger for the most helpful discussions .
we would also like to thank dr .
steve armstrong , professor alvaro de rujula , dr . conrad newton , professor gabriele veneziano , and especially geir vigdel , for useful discussions .
one of us ( ttw ) wishes to thank the theory division of cern for its kind hospitality .
h. fritzsch , nucl .
* b 155 * ( 1979 ) 189 ; h. fritzsch , nucl .
* b 194 * ( 1987 ) 391 ; h. fritzsch and d. holtmannsptter , phys .
* b 338 * ( 1994 ) 290 ; h. fritzsch and z .- z .
xing , phys .
* b 372 * ( 1996 ) 265 ; phys .
* d57 * ( 1998 ) 594 .
h. harari , h. hant and j. weyers , phys .
b 78 * ( 1978 ) 459 ; p. kaus and s. meshkov , mod .
lett . * a 3 * ( 1988 ) 1251 ; m. tanimoto , phys .
* d 41 * ( 1990 ) 1586 ; m. fukugita , m. tanimoto and t. yanagida , prog .
* 89 * ( 1993 ) 263 ; m. fukugita , m. tanimoto and t. yanagida , phys .
* d57 * ( 1998 ) ; y. nomura and t. yanagida , phys .
* d59 * ( 1999 ) 017303 ; s.f .
king , hep - ph/9912492 .
s. l. glashow , nucl .
phys . 22 ( 1961 ) 579 ; s. weinberg , phys .
lett . 19 ( 1967 ) 1264 ; a. salam , in _ elementary particle theory : relativistic groups and analyticity ( nobel symposium no .
8) _ , edited by n. svartholm ( almqvist and wiksell , stockholm , 1968 ) , p. 367 ; s. l. glashow , j. iliopoulos and l. maiani , phys . rev .
d2 , ( 1970 ) 1285 .
r. davis , jr .
, in _ frontiers of neutrino astrophysics _ , edited by y. suzuki and k. nakamura ( universal academy press , inc . ,
tokyo , 1993 ) , p. 47 ; b. t. cleveland _
et al . _ ,
( proc . suppl . )
* b 38 * , 47 ( 1995 ) ; astrophys .
j. * 496 * ( 1998 ) 505 .
l. wolfenstein , phys .
* d 17 * , 2369 ( 1978 ) , _ ibid .
_ * 20 * , 2634 ( 1979 ) ; s. p. mikheyev and a. yu .
smirnov , yad .
fiz . * 42 * , 1441 ( 1985 ) [ sov .
j. nucl .
* 42 * , 913 ( 1985 ) ] , nuovo cimento * 9c * , 17 ( 1986 ) .
t. kaneko , prog .
phys . * 78 * , 532 ( 1987 ) ; m. ito , t. kaneko and m. nakagawa , _ ibid . _ * 79 * , 13 ( 1988 ) ; s. toshev , phys .
b * 196 * , 170 ( 1987 ) ; s. t. petcov , phys .
b * 200 * , 373 ( 1988 ) ; nucl .
b ( proc . suppl . )
* 13 * , 527 ( 1990 ) .
j. n. bahcall _ et al .
c * 54 * ( 1996 ) 411 ; j. n. bahcall , phys .
c * 56 * ( 1997 ) 3391 ; j. n. bahcall and r. k. ulrich , rev .
phys . * 60 * ( 1988 ) 297 ; j. n. bahcall , m. kamionkowski and a. sirlin , phys .
d * 51 * ( 1995 ) 6146 ; see also j. n. bahcall s homepage , http://www.sns.ias.edu/@xmath92jnb .
see , for example , p. fisher , b. kayser and k. s. mcfarland , ann .
nucl . part .
* 49 * ( 1999 ) 481 ; g.l .
fogli , e. lisi , a. marrone and g. scioscia , hep - ph/9906450 ; j. ellis , talk given at 15th international conference on particle and nuclei ( panic 99 ) , uppsala , sweden , 10 - 16 june 1999 , hep - ph/9907458 ; e.k .
akhmedov , g.c . branco and m.n .
rebelo , hep - ph/9912205 ; g. l. fogli , e. lisi , d. montanino and a. palazzo , phys .
* d62 * ( 2000 ) 013002 ; g. altarelli , talk given at 6th international workshop on topics in astroparticle and underground physics ( taup 99 ) , paris , france , 6 - 10 sep 1999 , hep - ph/0001024 . the k2k collaboration ( y. oyama ) , kek - preprint-97 - 266 , jan 1998 , talk given at the yitp workshop on flavor physics , kyoto , japan , 2830 jan 1998 ,
hep - ex/9803014 ; the minos collaboration ( e. ables et al . ) .
fermilab - proposal - p-875 , feb 1995 ; the opera collaboration ( k. kodama et al . ) , cern / spsc 99 - 20 , aug 1999 .
s. boris et al . , phys .
* 58 * ( 1987 ) 2019 ; a. i. belesev et al .
* b 350 * ( 1995 ) 263 ; k .- h .
hiddelmann et al . , jour .
* g 21 * ( 1995 ) 539 ; w. stoeffl and d. j. decman , phys .
* 75 * ( 1995 ) 3237 . | if three right - handed neutrinos are added to the standard model , then , for the three known generations , there are six quarks and six leptons .
it is then natural to assume that the symmetry considerations that have been applied to the quark matrices are also valid for the lepton mass matrices . under this assumption
, the solar and atmospheric neutrino data can be used to determine the individual neutrino masses .
three minima have been found , using the @xmath0 fit , and , from these minima , it is determined that the mass of the lightest neutrino is @xmath1 ev , that of the next heavier neutrino is @xmath2 ev , while the mass of the heaviest neutrino is @xmath3 , @xmath4 or @xmath5 ev .
cern - th/99 - 285 + * universal quark - lepton mixing and determination of neutrino masses * per osland@xmath6 and tai tsun wu@xmath7 + @xmath8 _ department of physics , university of bergen , + allgaten 55 , n-5007 bergen , norway _ + @xmath9 _ gordon mckay laboratory , harvard university , + cambridge , massachusetts 02138 , u.s.a . _ + _ and _ + _ theoretical physics division , cern , ch-1211 geneva 23 , switzerland _ + cern - th/99 - 285 + march 2000 |
blazars are a type of active galactic nuclei with a jet oriented close to the line of sight .
their spectral energy distribution is broad and characterized by the presence of two peaks : a low energy one produced by the synchrotron mechanism , and a high energy one produced by inverse compton scattering of a photon field from the jet , accretion disk , the broad line region or the dust torus .
their emission is variable over a whole electromagnetic spectrum . due to these characteristics ,
a complete understanding of the emission mechanisms in blazars requires their study over a broad range of energies with simultaneous monitoring at different energy bands .
this monitoring needs to be independent of the state of the source to avoid biases .
moreover , understanding the statistical properties of different classes requires the study of large samples of objects .
even though the general picture of synchrotron emission at low energies and inverse compton at the high energy end is well established , many important aspects of these remarkable objects are still not well understood .
our work concerns one such question : where in the blazar structure is the gamma - ray emission region located ?
some theories locate it near the black hole / accretion disk @xcite , while others place parsecs away down the jet @xcite .
the _ fermi _ large area telescope ( lat ) and other gamma - ray telescopes lack the resolution needed to directly image the gamma - ray emission regions in blazars , but we can investigate their location by studying the connection between the radio and gamma - ray variability . to enable such studies , we are carrying out a blazar monitoring program with a sample size and cadence matched to _
fermi s _ capabilities .
this paper presents a summary of the state of this project as was presented at the time of the symposium .
a detailed discussion concerning the data analysis and results will be presented in max - moerbeck et al .
( in preparation ) .
we are currently monitoring about 1700 blazars including all the sources north of @xmath0 in declination in the candidate gamma - ray blazar sample @xcite and all _ fermi _ detected sources associated with blazars in the first and second lat agn catalogs @xcite with the same declination limit .
the sources are observed in total flux density twice per week at a frequency of 15 ghz with a 3 ghz bandwidth .
a typical observation has about 4 mjy of thermal noise with an additional @xmath1 uncertainty coming mostly from pointing errors .
a complete description of the program can be found in @xcite .
these observations complement the continuous monitoring of the gamma - ray sky in the band from 100 mev to 300 gev by _ fermi_-lat . in this study , we consider all the gamma - ray sources detected in at least 75% of the monthly time bins as reported in @xcite .
this sample contains 86 sources for which we obtain weekly binned gamma - ray light curves for integrated photon flux in the energy range from 100 mev to 200 gev covering the time period from 4 august 2008 through 8 august 2011 .
the radio light curves cover the period from 1 january 2008 through 26 february 2012 .
to study the physical significance of cross - correlation peaks we use monte carlo simulations that account for the uneven sampling and noise properties of the light curves .
the uneven sampling is incorporated using the discrete correlation function @xcite and the noise properties of the sources are characterized by fitting their power spectral density ( psd ) with a simple power law in both bands ( p@xmath2 ) using an adaptation of the method presented in @xcite . here
we present results of a study using populations values of the psd power law index at both bands . for the gamma - ray band
we adopt the average of the results obtained for the mean psd of the brightest bl lacs and fsrqs ( @xmath3 from @xcite ) . for the radio band we use the mean population value
we obtain in this work from fits to the radio light curves of the complete ovro monitored sample ( @xmath4 ) .
the significance of a cross - correlation peak is measured by obtaining an estimate of the distribution of random cross - correlation amplitudes at the given time lag .
this distribution is estimated by simulating a large number of independent time series ( we use 20,000 in this case ) with the appropriate power spectral densities and observational noise .
of the 86 sources with good cadence in radio and gamma - ray bands included in this study , only 63 show variability in both bands and are thus included in the cross - correlation analysis . among these 63
we find one case with @xmath5 significant correlations , j0238 + 1636 ( figure [ plots_j0238 + 1636 ] ; 0.2 cases are expected by chance ) .
there are 6 other cases with @xmath6 significant correlations ( 3 cases are expected by chance ) .
however , we consider five of these unreliable because of high noise or slow trends .
the remaining example , j1504 + 1029 , is shown in figure [ plots_j1504 + 1029 ] .
the well known blazar mrk 421 underwent a major gamma - ray flare in july 2012 @xcite which was followed by a major event at 15 ghz , the largest flare since at least 1980 as shown by a comparison to university of michigan radio astronomy observatory 14.5 ghz data @xcite .
when we extend the light curves to include the flaring period we are able to constrain the power spectral density in both bands and find correlated variability with @xmath6 significance .
preliminary light curves and results of the significance analysis including only the rising part of the radio flare are shown in figure [ plots_mrk421 ] .
this result has to be considered with care as in this case we are extending the period analyzed after observing the flare , thus doing a posteriori " statistics .
blazars produce their radio emission by the synchrotron mechanism which can result in polarized emission .
variations in the polarization can help us understand the evolution of the magnetic fields in the emission regions and its relation to high - energy emission . to study these magnetic fields
we are building a new receiver which will also measure linear polarization in the ku - band .
this new receiver , which we call kupol , will cover the band from 12 ghz to 18 ghz with 16 mhz spectral resolution , which can be used for examining spectral indices and removal of radio frequency interference .
this will be an important addition to our program and commissioning is planned for spring 2013 .
we study the connection between radio and gamma - ray emission in _ fermi _ detected blazars and find that significant correlations are present in a small number of these sources with the current data set .
the major flare on mrk 421 shows correlated radio / gamma - ray activity .
these cases add support to the idea that radio and gamma - rays are produced in related regions for some sources , but more data are needed to test whether this holds for the entire population . in particular longer radio and gamma - ray light curves will allow us to probe the variability on different time scales , and to incorporate information for multiple flaring periods .
this will improve the statistics and the significance of possible correlations .
we thank russ keeney for his efforts in support of observations at ovro .
the ovro 40 m monitoring program is supported in part by nasa grants nnx08aw31 g and nnx11a043 g , and nsf grants ast-0808050 and ast-1109911 .
t.h . was supported by the jenny and antti wihuri foundation .
support from mpifr for upgrading the ovro 40 m telescope receiver is also acknowledged . | blazars are powerful , variable emitters from radio to gamma - ray wavelengths .
even though the general picture of synchrotron emission at low energies and inverse compton at the high energy end is well established , many important aspects of these remarkable objects are still not well understood .
for example , even the location of the gamma - ray emission region is still not clearly established , with some theories locating it close to the black hole / accretion disk while others place it at parsec scales in the radio jet . since mid-2007 we have carried out a large scale monitoring program at 15 ghz using the ovro 40 m telescope .
we are currently observing about 1700 blazars twice per week .
the sample includes all the _
fermi_-lat detected blazars north of declination @xmath0 .
here , we study the existence of correlated variability between these two bands for 86 sources bright enough to be detected weekly by lat .
the existence of correlated variability can be interpreted as an indication of a related spatial locations for the radio and gamma - ray emission , making the evaluation of its statistical significance a key goal of our program .
a study of the statistical significance of these cross - correlations is presented along with a discussion of the monte carlo simulations used to evaluate them .
more information about the conditions on the radio emission zone can be obtained through polarization monitoring which tells us about the configuration of the magnetic fields in this region . to study radio polarization variability we are building kupol , a radio polarization receiver for the 12 to 18 ghz band that will replace the current total power receiver at the ovro 40 meter telescope . |
in a number of commonly considered supersymmetric ( susy ) models , strongly - interacting sparticles will be abundantly pair - produced at the large hadron collider ( lhc ) , and the resulting events containing jets and missing energy will stand out above the standard model background . the next challenge will be to make sense of the events , which will typically involve cascade decays through several intermediate sparticles
. it will be particularly desirable to obtain information on the susy masses and branching ratios .
it has been demonstrated that cascades involving lepton emission often provide relatively clean signals , which can be used to reconstruct several sparticle masses in some favourable benchmark scenarios , see e.g. @xcite .
however , these decay channels often suffer from branching ratios that are low compared to hadronic decay modes , reducing the available sample size and therefore restricting access to high - mass sparticles .
moreover , the charged leptons are often associated with neutrinos , for example in chargino ( @xmath6 ) decays .
the presence of the neutrinos would confuse the interpretation of the missing - energy signal provided by escaping neutralinos in scenarios with @xmath7-parity conservation , as we assume here , and make the reconstruction of chargino masses difficult .
some progress on techniques for measuring chargino masses has been reported for scenarios with one leptonic and one hadronic decay chain @xcite , but these require models with a favourable combination of the corresponding branching ratios .
the ability to reconstruct purely hadronic cascade decays would facilitate the discovery and measurement of charginos in more general cases , as well as enable heavier sparticle masses to be reconstructed . in this paper , we take a new approach to the reconstruction of fully - hadronic sparticle events , starting from the missing - energy signal provided by the neutralino in @xmath7-conserving models , which suppresses standard model ( sm ) backgrounds .
we use jet analysis techniques based on the algorithm to identify the hadronic decays of massive bosons such as the @xmath8 , @xmath9 or higgs boson ( @xmath10 ) . in general , the majority of sparticle cascade decays yield fully - hadronic final states containing such massive sm particles decaying to @xmath11 jet pairs @xcite . a certain fraction of these bosons , dependent on the sparticle masses in the scenario considered ,
are highly boosted and the jets from the decays are closely aligned , indeed overlapping , in pseudo - rapidity ( @xmath12 ) - azimuthal angle ( @xmath13 ) space .
such a situation presents both a challenge and an opportunity .
the challenge is that the choice of jet finder becomes crucial : the treatment of jet overlaps can be sensitive to rather soft and subtle qcd effects , and yet somehow one needs to retain the information that the jet arises from the two - body decay of a massive particle .
the opportunity is that , since the jets are close together , there is no combinatorial background ; no need , for example , to combine all pairs of jets to see if they reconstruct to the @xmath14 mass .
we address the reconstruction issue using the sub - jet technique that was proposed previously as a way to identify high - energy @xmath15 events @xcite .
this technique and its extension to hadronic sparticle decays are described in section [ sec : subjet ] .
then , in section [ sec : benchmarks ] some specific benchmark susy scenarios and the specific decay chains of interest are described : these involve typical decays of intermediate heavy charginos or neutralinos such as @xmath16 and @xmath17 .
the simulations and analysis method are described in section [ sec : analysis ] , and the subsequent section contains the results and conclusions .
the identification of jets which originate from a decaying massive particle begins by using a jet algorithm to define the jet . in this analysis the algorithm @xcite
is used , in the inclusive mode with the @xmath18 reconstruction scheme . for each particle @xmath19 and pair of particles @xmath20 ,
the algorithm calculates the quantities @xmath21 where @xmath22 is the transverse momentum of particle @xmath19 with respect to the beam axis and @xmath23 the parameter @xmath7 is a number of order one . if it is set below unity , it is less likely that a given particle will be merged with a jet , which in turn leads to narrower jets
. thus @xmath7 plays a somewhat similar role to the adjustable cone radius in cone algorithms . in this analysis
, initial studies indicated that @xmath24 provided a good compromise between efficiency and the mass resolution , and this value is used throughout .
returning to the algorithm : all the @xmath25 values are then ordered .
if @xmath26 or @xmath27 is the smallest , then particle @xmath19 or @xmath28 is labelled a jet and removed from the list . if @xmath29 is the smallest , particles @xmath19 and @xmath28 are merged by adding their four - momenta . the list is recalculated and the process is repeated until the list is empty . thus the algorithm clusters all particles into jets , and a cut on transverse momentum can then be used to select the hardest jets in the event .
the algorithm is infrared safe , and has the additional benefit that each particle is uniquely assigned to a single jet . recently
a fast implementation of the algorithm has been developed @xcite which makes it practical for use even in the very high multiplicity events expected at the lhc . in selecting a candidate for a hadronic decay of the @xmath0 or @xmath10 , first cuts on the @xmath30 and the pseudo - rapidity ( @xmath12 ) of the jet
are applied so as to ensure that they are energetic enough to contain a boosted heavy - particle decay and are in a region of good detector acceptance .
a cut is then applied on the mass of the jet ( calculated from the four - vectors of the constituents ) to ensure that it is in a window around the nominal mass of the desired particle .
the next step is to force the jet to decompose into two sub - jets .
the main extra piece of information gained from the sub - jet decomposition is the @xmath31 cut at which the sub - jets are defined : @xmath32 , where @xmath33 is the transverse momentum of the candidate jet containing the sub - jets @xmath19 and @xmath28 . in the case of a genuine @xmath0 or @xmath10 decay ,
the expectation is that the scale at which the jet is resolved into sub - jets ( i.e. , @xmath34 ) will be @xmath35 , where @xmath36 is the @xmath0 or @xmath10 mass . for qcd jets initiated by a single quark or gluon , the scale of the splitting is expected to be substantially below @xmath37 , i.e. , @xmath38 , since in the region around the jet strongly - ordered dglap - like @xcite qcd evolution dominates .
this distinction provides new information in addition to the jet mass itself , as is illustrated in figures [ fig : scale]a and [ fig : scale]b , where the correlation between the jet mass and the splitting scale is shown for @xmath8 jets and qcd jets respectively .
the events shown are @xmath8+jet events and susy events generated using pythia 6.408 @xcite , and @xmath8 + 3jet events generated using alpgen @xcite for the matrix element , herwig 6.510 @xcite for parton showering and jimmy @xcite for the underlying event .
the parameters for the underlying event and the parton showers were those of the atlas tune of pythia and the cdf tune a of herwig and jimmy , taken from @xcite . ] .
these models have been shown to give a good description of a wide variety of data .
in particular , the modelling of the internal jet structure by leading - logarithmic parton showers is known to be good for jets produced in @xmath39 collisions @xcite , @xmath40 collisions and photo - production @xcite , and in @xmath41 annihilation events and @xmath42 collisions @xcite . although there are some differences between the results of the pythia and herwig / alpgen simulations , the conclusions are similar . fig .
[ fig : scale]a confirms that the splitting scale in @xmath8 decays is relatively large , typically @xmath43 gev , whereas fig .
[ fig : scale]b shows that the splitting scale for qcd jets with masses @xmath44 gev is typically @xmath45 gev .
the distributions of of @xmath46 and of @xmath31 are shown in figures [ fig : scale]c and [ fig : scale]d for @xmath8 jets and qcd jets .
the distributions are qualitatively similar , whether they are generated in susy events or in conventional @xmath8+jets events , and whether ( in the latter case ) they are generated using pythia or herwig . the scale of the splitting is seen to be peaked close to the @xmath8 mass for genuine @xmath8 decays , whichever the environment in which they are produced , and to be softer for qcd jets which just happen to reconstruct to the @xmath8 mass , again in both environments . this distinction was noted and successfully used to identify @xmath8 decays in simulations of @xmath47 scattering at lhc energies in @xcite . here , this information is used in the analysis described in section [ sec : analysis ] to refine the selection of @xmath0 and @xmath10 candidates in sparticle decay cascades .
although no detector simulation is employed in this analysis , the cut on the scale has been shown previously to be robust against the effects of the underlying events @xcite , as well as effects due to the calorimeter granularity and resolution @xcite .
the pair - production of heavy sparticles in generic susy models yields on - shell electroweak gauge bosons or higgs bosons via the decays of heavy gauginos that appear as intermediate steps in cascades , e.g. , @xmath48 or @xmath49 , followed by @xmath50 or @xmath51 decays .
exceptions are cases with gaugino masses that are nearly degenerate , in which there may be large branching ratios for decays via off - shell heavy bosons . since the largest branching ratios for @xmath0 and @xmath10 decays are those into hadronic @xmath52 final states , purely hadronic final states dominate in cascade decays via on - shell bosons , and these are also potentially important in off - shell decays . for the decay chain @xmath53 shown in fig .
[ fig : qw ] , one can demonstrate that the invariant mass distribution of the quark-@xmath14 system has a minimum and maximum given by @xmath54 where @xmath55 is the @xmath14 momentum in the chargino rest frame .
if measurable , these endpoints give a model - independent relationship between the three susy masses , modulo the existence of the decay chain . if both endpoints can be determined experimentally , the squark mass can be eliminated , giving the chargino mass in terms of the lightest neutralino .
supplementary model - dependent assumptions , such as the relationship @xmath56 that holds approximately in the constrained mssm ( cmssm ) , can then be used to determine individual masses of all three sparticles in specific theoretical frameworks .
the decay chains @xmath57 and @xmath58 involving the second lightest neutralino , have endpoint formulae with the same structure , given by the substitutions @xmath59 and @xmath60 respectively , in eqs .
( [ eq : endqw ] ) and ( [ eq : pw ] ) . in @xcite benchmark models with such cascade decays were explored in a more generic setting than that allowed in the cmssm . relaxing the gut - scale universality of the higgs scalar masses or introducing a gravitino dark matter candidate admits values of the ( supposedly universal ) scalar squark and slepton mass @xmath61 outside the narrow range allowed by the cold dark matter density @xcite within the restrictive cmssm framework .
this , in turn , allows for a wide range of values for the branching ratios of the lightest chargino and the next - to - lightest neutralino into @xmath0 and @xmath10 . in our study
, we use the benchmark points @xmath62 , @xmath63 , @xmath64 and @xmath65 from @xcite to illustrate these possibilities .
the points @xmath62 , @xmath63 , @xmath64 are scenarios with relatively light susy masses , that should be easy to probe at the lhc , while @xmath65 represents a more challenging scenario with lower cross sections .
common to all four points is the large branching ratio for squark to chargino / neutralino decays , @xmath66% and @xmath67% .
the corresponding chargino and neutralino bosonic branching ratios are shown in table [ tab : brs ] .
.branching ratios for @xmath68 and @xmath69 in the selected susy benchmark models .
the decays are calculated using sdecay 1.1a @xcite . [ cols="^,^,^,^",options="header " , ] we see that the endpoint estimates have statistical errors which are in the @xmath70% region for benchmarks @xmath71 and slightly below @xmath72% for @xmath65 .
the statistical errors are in general smaller without the cut on the sub - jet separation scale , particularly in the case of benchmark @xmath65 , due to somewhat larger statistics . in the case of the @xmath62 benchmark point ,
the fitted values are both fairly close to the nominal value , but with indications of larger systematic errors when not using the scale cut . both fits overestimate the endpoint , which could be expected as the reconstructed jets from squark decays have not been calibrated for the algorithm s tendency to overestimate the jet . for @xmath63 both fits are very close to the nominal value , while for @xmath64 the fits again overestimate the edge position . from the three benchmarks considered here
this systematic error seems to increase for lower values of the endpoint , barring other more important hidden systematics . in the case of the @xmath65 benchmark
both fits underestimate the endpoint , but here the statistical errors are large .
considered as a whole , the results of table [ tab : qwmax ] indicate that the systematical errors on the endpoints in our fitting procedure are comparable to the statistical errors . for most of the distributions
it is impossible to estimate the lower endpoint , since any structure at low invariant masses is washed away by the hard kinematical cuts on jet energies used to isolate the signal .
the exception is the relatively high lying lower endpoint of the @xmath73 distribution for the @xmath64 benchmark , where a similar fitting technique to the one used for the upper endpoint yields a lower endpoint estimate of @xmath74 gev before and @xmath75 gev after the sub - jet cut is applied , compared to the nominal value of @xmath76 gev .
the difference in endpoint estimate originates from a worse fit of the sideband distribution at high invariant masses when the sub - jet cut is used .
we have in addition estimated both the lower and upper edge of the @xmath77 distribution for the @xmath62 benchmark point , where we reconstruct the @xmath78 from a lepton pair .
after a similar sideband subtraction we estimate @xmath79 gev and @xmath80 gev , to be compared to the nominal values of @xmath81 gev and @xmath82 gev , respectively . for the @xmath83 distributions ,
fitting is difficult with the low statistics available both at the @xmath63 and @xmath65 benchmarks . from sideband subtracted distributions we make estimates by visual inspection , giving values of @xmath84 gev for @xmath63 and @xmath85 gev for @xmath65 , compared to the nominal values of @xmath86 gev and @xmath87 gev , respectively . as has been noted earlier
, the feasibility of measuring these edges depends on the @xmath88-tagging achievable , and on the use of the sub - jet separation scale cut .
the positions of the edges of the invariant mass distributions may be used to extract information on the spectrum of the susy particles involved . whilst the four equations giving the edges of the @xmath73 and @xmath89 distributions may in principle be solved to obtain the four masses involved , extracting accurate values of the absolute masses - as opposed to mass differences
- will be difficult in practise because of the degeneracies between the masses of the bosons and of the gauginos , respectively , and thus in the upper edges , as shown in fig .
[ fig : imtruth ] .
however , for the four benchmarks considered here this is an academic problem , as neither of the benchmarks have four measurable edges . given both the lower and upper edge of a single distribution
the squark mass can be eliminated from eq .
[ eq : endqw ] , and one can solve for the mass of the chargino or the next - to - lightest neutralino in terms of the mass of the lightest neutralino .
this can in turn be used to arrive at the squark mass .
we show in fig .
[ fig : masses_alpha ] the squark and @xmath68 masses determined using the upper and lower @xmath77 edge measured for the @xmath62 benchmark point , as functions of the undetermined @xmath90 mass .
we also show the @xmath91 error bands on the masses resulting from the statistical uncertainty of the edge measurements , assuming that the errors on the two edges are independent . with the squark mass known in terms of the @xmath90 mass , the upper edge of the @xmath73 distribution
can also be used to give the @xmath69 mass . because of the quadratic nature of eq .
[ eq : endqw ] two solutions result , none of which can be rejected out of hand mass , but one can be rejected as unphysical since it is always less than the @xmath90 mass . ] .
we show both solutions in fig . [
fig : masses_alpha ] . for values of the @xmath90 mass below @xmath92 gev
there are no solutions . for the @xmath64 benchmark point the same procedure
can be followed for the two endpoints of the @xmath73 distribution , and we show the resulting squark and chargino masses as functions of the lightest neutralino mass in fig . [
fig : masses_gamma ] . for the remaining two benchmarks ,
extracting masses is more difficult since only one edge is well measured .
however , the lhc experiments have the potential for measuring other susy mass - dependent quantities , such as the effective mass @xcite , whose determination would give complementary relations between the involved masses .
given an accurate measurement of the @xmath90 mass , for example at a future linear collider , the heavier neutralino , chargino and squark masses could then be found with statistical errors in the range of @xmath93% , using lhc data , even if the particles themselves are too heavy to be produced at the linear collider .
we have shown in this paper , we believe for the first time , that it is possible to extract a susy signal solely from the hadronic decays of @xmath8 bosons produced in cascade decays involving charginos , and that the measurement of this signal may provide useful information about the sparticle mass spectrum . in each of the benchmarks studied , the upper edge of the @xmath73 mass distribution in the decay chain @xmath94 can be measured , with a statistical error of @xmath70% for the relatively light sparticle masses of benchmarks @xmath71 and @xmath72% for the heavier @xmath65 , with squark and gluino masses above @xmath95 tev .
systematic errors in the background and fitting are estimated to be of the same order as the statistical errors . in the case of benchmark @xmath64 , we are also able to extract the position of the lower edge of the @xmath73 mass distribution , and in the case of point @xmath62 we extract both the upper and the lower edges of the @xmath77 mass distribution , using the leptonic decays of @xmath78 bosons .
there are also clear indications in benchmarks @xmath63 and @xmath65 for the possibility of observing the upper edge of the @xmath83 mass distribution using @xmath96 decays with collimated @xmath88-jets , but conclusions on such a signal require further analysis and better understanding of the @xmath88-tagging efficiency for these jets in a full detector simulation .
the hadronic signals were extracted using the algorithm for jet reconstruction , including the single - jet mass and the proposal of @xcite to improve the identification of hadronic decays of heavy bosons via a cut on the sub - jet separation scale .
the algorithm is shown to be well suited for reconstructing edge features in the considered invariant mass distributions .
the sub - jet cut procedure improves quantitatively the signal - to - background ratio , and while the loss of statistics increases the statistical errors , fits to the edges of the @xmath73 distribution show improved results with respect to the nominal values when using the sub - jet cut . for the @xmath83 distributions with their small signal - to - background ratio ,
the sub - jet cut is crucial in reducing the background to a manageable level . a detailed exploration of the capabilities of the lhc experiments to measure and use all the information that could be gained from these benchmark points , and hence a more detailed display of the value added to the global fits by the measurements described here , lies beyond the scope of this study .
however , we do note that this technique provides novel information on chargino and neutralino spectroscopy , e.g. , the difference between the @xmath97 and @xmath69 mases , and in at least one case the @xmath98 mass difference .
these pieces of information are useful for potentially constraining susy models and perhaps foreseeing the locations of interesting thresholds in @xmath99 annihilation .
as pointed out in @xcite , similar techniques for analysing hadronic final states arising from the decays of heavy particles should be useful in other situations .
examples include the analysis of top physics , the search for @xmath7-violating hadronic decays of sparticles , or in the isolation of a generic susy signal from qcd sm backgrounds .
indeed , some recent studies have highlighted the potential of single - jet mass cuts in exotic searches @xcite , and we note that the @xmath31-scale cut should also be useful in these cases .
we believe this to be an area meriting much further experimental and phenomenological study .
the authors thank jeppe anderson for useful discussions .
arr acknowledges support from the european community through a marie curie fellowship for early stage researchers training , and from the norwegian research council .
i. hinchliffe , f. e. paige , m. d. shapiro , j. soderqvist and w. yao , phys .
d * 55 * ( 1997 ) 5520 [ arxiv : hep - ph/9610544 ] .
h. bachacou , i. hinchliffe and f. e. paige , phys . rev .
d * 62 * ( 2000 ) 015009 [ arxiv : hep - ph/9907518 ] ; + b. c. allanach , c. g. lester , m. a. parker and b. r. webber , jhep * 0009 * ( 2000 ) 004 [ arxiv : hep - ph/0007009 ] ; + c. g. lester , cern - thesis-2004 - 003 ; + b. k. gjelsten , d. j. miller and p. osland , jhep * 12 * ( 2004 ) 003 [ arxiv : hep - ph/0410303 ] ; + d. j. miller , p. osland and a. r. raklev , jhep * 03 * ( 2006 ) 034 [ arxiv : hep - ph/0510356 ] .
m. m. nojiri , g. polesello and d. r. tovey , arxiv : hep - ph/0312318 .
a. de roeck , j. r. ellis , f. gianotti , f. moortgat , k. a. olive and l. pape , [ arxiv : hep - ph/0508198 ]
. j. m. butterworth , b. e. cox and j. r. forshaw , phys .
d * 65 * ( 2002 ) 096014 [ arxiv : hep - ph/0201098 ] .
s. catani , y. l. dokshitzer , m. h. seymour and b. r. webber , nucl . phys .
b * 406 * ( 1993 ) 187 .
m. cacciari and g. p. salam , phys .
b * 641 * ( 2006 ) 57 [ arxiv : hep - ph/0512210 ] .
v. n. gribov and l. n. lipatov , sov .
j. nucl .
phys . * 15 * ( 1972 ) 438 [ yad
. fiz . * 15 * ( 1972 ) 781 ] .
+ g. altarelli and g. parisi , nucl
. phys .
b * 126 * ( 1977 ) 298 .
+ y. l. dokshitzer , sov .
jetp * 46 * ( 1977 ) 641 [ zh .
* 73 * ( 1977 ) 1216 ] .
t. sjostrand , s. mrenna and p. skands , jhep * 0605 * , 026 ( 2006 ) [ arxiv : hep - ph/0603175 ] ; note that the events were generated calling pyevnt , which uses the same underlying event and parton shower models as the pre-6.3 versions of pythia .
m. l. mangano et al , jhep * 0307 * ( 2003 ) 001 [ arxiv : hep - ph/0206293 ] ; m. l. mangano , m. moretti and r. pittau , nucl .
b * 632 * ( 2002 ) 343 [ arxiv : hep - ph/0108069 ] ; f. caravaglios et al , nucl .
b * 539 * ( 1999 ) 215 [ arxiv : hep - ph/9807570 ] .
j. m. butterworth , j. r. forshaw and m. h. seymour , z. phys .
c * 72 * ( 1996 ) 637 [ arxiv : hep - ph/9601371 ] . c. buttar _
_ in _ hera and the lhc - a workshop on the implications of hera for lhc physics : proceedings part a _ , pp.192 - 217 [ arxiv : hep - ph/0601012 ] . v. m. abazov _ et al .
_ [ d0 collaboration ] , phys . rev .
d * 65 * ( 2002 ) 052008 [ arxiv : hep - ex/0108054 ] ; + d. acosta _ et al .
_ [ cdf collaboration ] , [ arxiv : hep - ex/0505013 ] . s. chekanov _ et al . _
[ zeus collaboration ] , nucl .
b * 700 * ( 2004 ) 3 [ arxiv : hep - ex/0405065 ] .
g. abbiendi _ et al . _
[ opal collaboration ] , eur .
j. c * 37 * ( 2004 ) 25 [ arxiv : hep - ex/0404026 ] ; eur .
j. c * 31 * ( 2003 ) 307 [ arxiv : hep - ex/0301013 ] ; + d. buskulic _ et al .
_ [ aleph collaboration ] , phys . lett .
b * 384 * ( 1996 ) 353 .
s. allwood , manchester university phd thesis 2005 .
+ e. stefanidis , university college london phd thesis 2006 . c. l. bennett _ et al .
_ , astrophys .
j. suppl .
* 148 * ( 2003 ) 1 [ arxiv : astro - ph/0302207 ] .
d. n. spergel
_ et al . _ ,
astrophys .
j. suppl .
* 148 * ( 2003 ) 175 [ arxiv : astro - ph/0302209 ] .
m. muhlleitner , a. djouadi and y. mambrini , in comput .
commun . * 168 * ( 2005 ) 46 [ arxiv : hep - ph/0311167 ] . h. l. lai _ et al . _
[ cteq collaboration ] , eur .
j. c * 12 * ( 2000 ) 375 [ arxiv : hep - ph/9903282 ] .
see _ hera and the lhc - a workshop on the implications of hera for lhc physics : proceedings part b _
, pp.617 - 621 .
[ arxiv : hep - ph/0601013 ] .
+ http://projects.hepforge.org/hztool .
d. r. tovey , phys .
b * 498 * ( 2001 ) 1 [ arxiv : hep - ph/0006276 ] .
a. l. fitzpatrick et al , arxiv : hep - ph/0701150 ; + b. lillie , l. randall and l. t. wang , arxiv : hep - ph/0701166 ; + w. skiba and d. tucker - smith , arxiv : hep - ph/0701247 ; + b. holdom , arxiv : hep - ph/0702037 . | most sparticle decay cascades envisaged at the large hadron collider ( lhc ) involve hadronic decays of intermediate particles .
we use state - of - the art techniques based on the jet algorithm to reconstruct the resulting hadronic final states for simulated lhc events in a number of benchmark supersymmetric scenarios .
in particular , we show that a general method of selecting preferentially boosted massive particles such as @xmath0 or higgs bosons decaying to jets , using sub - jets found by the algorithm , suppresses qcd backgrounds and thereby enhances the observability of signals that would otherwise be indistinct .
consequently , measurements of the supersymmetric mass spectrum at the per - cent level can be obtained from cascades including the hadronic decays of such massive intermediate bosons . `
february ` ` 2007 ` + cern - ph - th/2007 - 019 j. m. butterworth@xmath1 , john ellis@xmath2 and a. r. raklev@xmath3 + + @xmath4 theory division , physics department , cern + @xmath5 department of physics & technology , university of bergen + _ keywords : monte carlo ; supersymmetry ; jets ; lhc _ |
due to the many well known successes of the standard theory of stellar evolution , rotation has generally been considered as a secondary effect .
this was justified in many cases . however , since some years a number of very significant discrepancies between model predictions and observations have been found for massive stars , and also for red giants of lower masses .
these difficulties have been listed and examined ( cf .
maeder @xcite ) .
recently , large excesses of [ n / h ] have been found in a
type supergiants , particularly in the smc ( venn et al .
@xcite ; venn @xcite ) , where the excesses may reach up to an order of magnitude with respect to the average [ n / h ] local value in the smc .
these excesses , not predicted by current models , are the signatures of mixing effects over the entire star .
this extensive mixing changes the size of the reservoir of nuclear fuel available for evolution , and thus the lifetimes , the tracks and all the model outputs ( cf . also langer 1992 ; langer 1997 ; heger et al .
2000 ) . also the chemical abundances , the yields and the final stages may be modified .
rotation appears as a natural driver for this mixing or , at least , as one of the first mechanisms whose consequences for mixing has to be explored .
this is especially true for massive stars which are known to be fast rotators ( see e.g. penny 1996 ; howarth et al .
1997 ) .
the physical effects of stellar rotation are numerous .
the basic equations of stellar structure need to be modified .
the meridional circulation and its interaction with the horizontal turbulence , the diffusion effects produced by shear turbulence , the inhibition of the transport mechanisms by the @xmath6gradients , the transport of the angular momentum and of the chemical elements , the loss of mass and angular momentum at the surface , the enhancement of mass loss by rotation , etc are some of the effects to be included in realistic models .
in addition to the model physics , rotation also brings a number of new numerical problems in the stellar code , such as the 4@xmath7 order equation for the transport of the angular momentum , which has to be coupled to the equations of stellar structure . in this work
, we apply the investments in the treatment of the physical and numerical effects of rotation made in the previous works and we explore their consequences for the outputs of stellar models . in sect .
2 , we describe the various effects considered .
the non rotating stellar models are briefly discussed in sect .
the evolution of the internal rotation during the evolution is examined in sect .
5 discusses the effects of rotation on the evolutionary tracks in the hr diagram , on the lifetimes and isochrones .
the evolution of the rotational velocities at the stellar surface is discussed in sect .
6 and the effects on the surface abundances are analysed in sect .
let us briefly summarize here the basic physical ingredients of the numerical models of rotating stars we are constructing here . the differential rotation which results from the evolution and transport of the angular momentum as described by eq .
( 3 ) below , makes the stellar interior highly turbulent .
the turbulence is very anisotropic , with a much stronger geostrophic like transport in the horizontal direction than in the vertical one ( zahn @xcite ) , where stabilisation is favoured by the stable temperature gradient .
this strong horizontal transport is characterized by a diffusion coefficient @xmath8 , which is quite large as will be shown below .
the horizontal turbulent coupling favours an essentially constant angular velocity @xmath9 on the isobars .
this rotation law , constant on shells , applies to fast as well as to slow rotators . as an approximation
, it is often represented by a law of the form @xmath10 ( zahn @xcite ; see also endal and sofia @xcite ) . in a rotating star ,
the equations of stellar structure need to be modified ( kippenhahn and thomas @xcite ) .
the usual spherical coordinates must be replaced by new coordinates characterizing the equipotentials .
the classical method applies when the effective gravity can be derived from a potential @xmath11 , i.e. when the problem is conservative .
there , @xmath12 is the gravitational potential which in the roche approximation is @xmath13 .
if the rotation law is shellular , the problem is non conservative .
most existing models of rotating stars apply , rather inconsistently , the classical scheme by kippenhahn and thomas .
however , as shown by meynet and maeder ( @xcite ) , the equations of stellar structure can still be written consistently , in term of a coordinate referring to the mass inside the isobaric surfaces .
thus , the problem of the stellar structure of a differentially rotating star with an angular velocity @xmath14 can be kept one dimensional .
the distribution of temperature at the surface of a rotating star is described by the von zeipel theorem ( @xcite ) .
usually , this theorem applies to the conservative case and states that the local radiative flux @xmath15 is proportional to the local effective gravity @xmath16 , which is the sum of the gravity and centrifugal force , @xmath17 with @xmath18 ; @xmath19 is the luminosity on an isobar and @xmath20 the mean internal density .
the local @xmath21 on the surface of a rotating star varies like @xmath22 .
we define the average stellar @xmath21 by @xmath23 , where @xmath24 is stefan s constant and @xmath25 the total actual stellar surface . of course , for different orientation angles @xmath26
, the emergent luminosity , colours and spectrum will be different ( maeder and peytremann @xcite ) . in the case of non conservative rotation law , the corrections to the von zeipel theorem
depend on the opacity law and on the degree of differential rotation , but they are small , i.e. @xmath27 in current cases of shellular rotation ( kippenhahn @xcite ; maeder @xcite ) .
wether or not the star is close to the eddington limit , the von zeipel theorem keeps the same form as the one given by maeder ( 1999 , see also maeder & meynet @xcite ) . in this last work ,
it is in particular shown that the expression for the eddington factor in a rotating star needs to be consistently written and that it contains a term depending on rotation .
observationally , a growth of the mass flux of ob stars with rotation , i.e. by 23 powers of 10 , was found by vardya ( @xcite ) , while nieuwenhuijzen and de jager ( @xcite ) concluded that the @xmath29rates seem to increase only slightly with rotation for o and b type stars . on the theoretical side , friend and abbott ( 1986 )
find an increase of the @xmath28rates which can be fitted by the relation ( langer @xcite ) @xmath30 with @xmath31 and @xmath32 the equatorial velocity at break up .
this expression is used in most evolutionary models and this is also what is done in the present work .
the critical rotation velocity of a star is often written as @xmath33 , where @xmath34 is the equatorial radius at break up velocity and @xmath35 is the ratio of the stellar luminosity to the eddington luminosity ( cf .
langer 1997 , 1998 ) .
glatzel ( 1998 ) has shown that when the effect of gravity darkening is taken into account , the above expression for @xmath36 does not apply .
glatzel ( 1998 ) gives @xmath37 , which we adopt here .
the problem is now being further examined by maeder & meynet ( 2000a ) , who critically discuss the eddington factors , their dependence on rotation , the expression of the critical velocity , the dependence of the mass loss rates on rotation .
these various new results will be applied in subsequent works , particularly for the study of the effects of rotation on the formation of w
further improvements , based either on the observations or on the theory , to account for the anisotropic winds which selectively remove the angular momentum need also to be performed . the above expression gives the change of the mass loss rates due to rotation . as reference mass loss rates in the case of no rotation
, we use the recent data by lamers and cassinelli ( @xcite ) ; for the domain not covered by these authors we use the results by de jager et al .
( @xcite ) . during the wolf
rayet phase we use the mass loss rates proposed by nugis et al ( 1998 ) for the wnl stars ( mean , clumping corrected rates from radio data @xmath38
m@xmath39 y@xmath40 ) . for the wne and wc stars we use the prescription devised by langer ( 1989 ) , modified according to schmutz ( 1997 ) for taking into account the clumping effects in wolf - rayet stellar winds ( @xmath41 m@xmath39 y@xmath40 ) .
these mass loss rates are smaller by a factor 23 than the mass loss rates used in our previous stellar grids ( schaller et al .
1992 ; meynet et al .
1994 ) . for shellular rotation ,
the equation of transport of angular momentum in the vertical direction is in lagrangian coordinates ( cf .
zahn @xcite ; maeder and zahn @xcite ) @xmath42 & & \frac{1}{5 r^2 } \frac{\partial}{\partial r } \left(\rho r^4 \omega u(r ) \right ) + \frac{1}{r^2 } \frac{\partial}{\partial r } \left(\rho d r^4 \frac{\partial \omega}{\partial r } \right ) .\end{aligned}\ ] ] @xmath1 is the mean angular velocity at level @xmath43 . the vertical component @xmath44 of the velocity of the meridional circulation at a distance @xmath43 to the center and at a colatitude @xmath45 can be written @xmath46 where @xmath47 is the second legendre polynomial
. only the radial term @xmath48 appears in eq .
the quantity @xmath49 is the total diffusion coefficient representing the various instabilities considered and which transport the angular momentum , namely convection , semiconvection and shear turbulence . as a matter of fact
, a very large diffusion coefficient as in convective regions implies a rotation law which is not far from solid body rotation . in this work
, we take @xmath50 in radiative zones , since as extra convective mixing we consider shear mixing and meridional circulation . in case the outward transport of the angular momentum by the shear
is compensated by an inward transport due to the meridional circulation , we obtain the local conservation of the angular momentum .
we call this solution the _ stationary solution_. in this case , @xmath48 is given by ( cf .
zahn @xcite ) @xmath51 the full solution of eq .
( 3 ) taking into account @xmath48 and @xmath49 gives the _ non stationary solution _ of the problem . in this case
, @xmath52 evolves as a result of the various transport processes , according to their appropriate timescales , and in turn differential rotation influences the various above processes .
this produces a feedback and , thus , a self consistent solution for the evolution of @xmath52 has to be found .
the transport of angular momentum by circulation has often been treated as a diffusion process ( endal and sofia @xcite ; pinsonneault et al .
@xcite ; heger et al .
@xcite ) . from eq .
( 3 ) , we see that the term with @xmath53 ( advection ) is functionally not the same as the term with @xmath49 ( diffusion ) .
physically advection and diffusion are quite different : diffusion brings a quantity from where there is a lot to other places where there is little .
this is not necessarily the case for advection .
a circulation with a positive value of @xmath48 , i.e. rising along the polar axis and descending at the equator , is as a matter of fact making an inward transport of angular momentum .
thus , we see that when this process is treated as a diffusion , like a function of @xmath54 , even the sign of the effect may be wrong .
the expression of @xmath48 given below ( eq . 12 ) involves derivatives up to the third order , thus eq .
( 3 ) is of the fourth order , which makes the system very difficult to solve numerically . in practice
, we have applied a henyey scheme to make the calculations .
( 3 ) also implies four boundary conditions . at the stellar surface ,
we take ( cf .
talon et al .
@xcite ; denissenkov et al .
@xcite ) @xmath55 and at the edge of the core we have @xmath56 we assume that the mass lost by stellar winds is just embarking its own angular momentum .
this means that we ignore any possible magnetic coupling , as it occurs in low mass stars .
this is not unreasonable in view of the negative results about the detection of magnetic fields in massive stars ( mathys @xcite ) .
it is interesting to mention here , that in case of no viscous , nor magnetic coupling at the stellar surface , _ i.e. _ with the boundary conditions ( 6 ) , the integration of eq .
( 3 ) gives for an external shell of mass @xmath57 ( maeder 1999 , paper iv ) @xmath58 this equation is valid provided the stellar winds are spherically symmetric ( see paper iv ) , an assumption we do in this work .
when the surface velocity approches the critical velocity , it is likely that there are anisotropies of the mass loss rates ( polar ejection or formation of an equatorial ring ) and thus the surface condition should be modified according to the prescriptions of maeder ( @xcite ) . for now , these effects are not included in these models .
their neglect should not affect too much the results presented here since the critical velocity is reached only in some rare circumstances .
a diffusion advection equation like eq .
( 3 ) should normally be used to express the transport of chemical elements . however ,
if the horizontal component of the turbulent diffusion @xmath8 is large , the vertical advection of the elements can be treated as a simple diffusion ( chaboyer and zahn @xcite ) with a diffusion coefficient @xmath59 .
as emphasized by chaboyer and zahn , this does not apply to the transport of the angular momentum .
@xmath59 is given by @xmath60 where @xmath8 is the coefficient of horizontal turbulence , for which the estimate is @xmath61 according to zahn ( 1992 ) .
( 8) expresses that the vertical advection of chemical elements is severely inhibited by the strong horizontal turbulence characterized by @xmath8 .
thus , the change of the mass fraction @xmath62 of the chemical species @xmath26 is simply @xmath63 + \left(\frac{d x_i}{dt } \right)_{\rm nucl } .\end{aligned}\ ] ] the second term on the right accounts for composition changes due to nuclear reactions .
the coefficient @xmath64 is the sum @xmath65 and @xmath59 is given by eq .
the characteristic time for the mixing of chemical elements is therefore @xmath66 and is not given by @xmath67 , as has been generally considered ( schwarzschild 1958 ) .
this makes the mixing of the chemical elements much slower , since @xmath59 is very much reduced .
in this context , we recall that several authors have reduced by large factors , up to 30 or 100 , the coefficient for the transport of the chemical elements , with respect to the transport of the angular momentum , in order to better fit the observed surface compositions ( cf .
heger et al .
this reduction of the diffusion of the chemical elements is no longer necessary with the more appropriate expression of @xmath59 given here . when the effects of the shear and of the meridional circulation compensate each other for the transport of the angular momentum ( _ stationary solution _ , see sect .
2.5 ) , the value of @xmath53 entering the expression for @xmath59 is given by eq .
( 5 ) .
meridional circulation is an essential mixing mechanism in rotating stars and there is a considerable litterature on the subject ( see ref . in tassoul @xcite ) .
the velocity of the meridional circulation in the case of shellular rotation was derived by zahn ( @xcite ) .
the effects of the vertical @xmath6gradient @xmath68 and of the horizontal turbulence on meridional circulation are very important and they were taken into account by maeder and zahn ( @xcite ) .
contrarily to the conclusions of previous works ( e.g. mestel @xcite ; kippenhahn and weigert @xcite ; vauclair @xcite ) , the @xmath6gradients were shown not to introduce a velocity threshold for the occurence of the meridional circulation , but to progressively reduce the circulation when @xmath68 increases .
the expression by maeder and zahn ( @xcite ) is @xmath69 } } \nonumber \\[2 mm ] & & \ \ \ \ \ \ \ \ \ \left\ { \frac{l}{m_\star } ( e_\omega + e_\mu ) \right\ } \ ; .\end{aligned}\ ] ] @xmath70 is the pressure , @xmath71 the specific heat , @xmath72 and @xmath73 are terms depending on the @xmath9 and @xmath6distributions respectively , up to the third order derivatives and on various thermodynamic quantities ( see details in maeder and zahn , @xcite ) .
the term @xmath68 in eq .
( 12 ) results from the vertical chemical gradient and from the coupling between the horizontal and vertical @xmath6gradients due to the horizontal turbulence .
this term @xmath68 may be one or two orders of magnitude larger than @xmath74 in some layers , so that @xmath75 may be reduced by the same ratio .
this is one of the important differences introduced by the work by maeder and zahn ( @xcite ) .
another difference is that the classical solution usually predicts an infinite velocity at the interface between a radiative and a semiconvective zone with an inverse circulation in the semiconvective zone .
expression ( 12 ) gives a continuity of the solution with no change of sign from semiconvective to radiative regions .
finally , we recall that in a stationary situation , @xmath48 is given by eq .
( 5 ) , as seen above . in a radiative zone ,
shear due to differential rotation is likely to be a most efficient mixing process .
indeed shear instability grows on a dynamical timescale that is of the order of the rotation period ( zahn @xcite ) .
the usual criterion for shear instability is the richardson criterion , which compares the balance between the restoring force of the density gradient and the excess energy present in the differentially rotating layers , @xmath76 where we have taken the average over an isobar , @xmath43 is the radius and @xmath77 the brunt - visl frequency given by @xmath78.\end{aligned}\ ] ] when thermal dissipation is significant , the restoring force of buoyancy is reduced and the instability occurs more easily , its timescale is however longer , being the thermal timescale .
this case is referred to as `` secular shear instability '' .
the criterion for low peclet numbers @xmath79 ( i.e. of large thermal dissipation , see below ) has been considered by zahn ( @xcite ) , while the cases of general peclet numbers @xmath79 have been considered by maeder ( @xcite ) , maeder and meynet ( @xcite ) , who give @xmath80 & & \ \ \ \ \ \ \ \
\left [ \frac{\gamma}{\gamma + 1 } ( \nabla_{ad } -\nabla ) + \frac{\varphi}{\delta } \nabla_{\mu } \right ] < \frac{1}{4}\end{aligned}\ ] ] the quantity @xmath81 , where the peclet number @xmath79 is the ratio of the thermal cooling time to the dynamical time , i.e. @xmath82 where @xmath83 and @xmath84 are the characteristic velocity and length scales , and @xmath85 is the thermal diffusivity .
a discussion of shear driven turbulence by canuto ( @xcite ) suggests that the limiting @xmath86 number may be larger than @xmath87 . to account for shear transport and diffusion in eqs .
( 3 ) and ( 11 ) , we need a diffusion coefficient .
amazingly , a great variety of coefficients @xmath88 have been derived and applied , for example : 1 . endal and sofia ( @xcite ) apply the reynolds and the richardson criterion by zahn ( @xcite ) .
they estimate @xmath89 from the product of the velocity scale height of the shear flow and of the turbulent velocities of cells at the edge of reynolds critical number .
2 . pinsonneault et al .
( @xcite ) notice that the amount of differential rotation permitted by the secular shears is proportional to a critical number , which they treat as an adjustable parameter . to account for the effects of the @xmath6gradient in mixing , of the loss of angular momentum , etc
they introduce several adjustable parameters in the equations for the transport of the chemical elements and of the angular momentum .
chaboyer et al .
( @xcite ) use a coefficient derived from the velocity and path length from zahn ( @xcite ) .
following pinsonneault et al .
( @xcite ) , they also introduce two adjustable parameters for adjusting the transports of chemical elements and angular momentum respectively .
we notice that thanks to the reduction of the diffusion produced by the horizontal turbulence ( cf .
eq . 9 above ) , it is no longer necessary to arbitrarily reduce the vertical transport of the chemical elements .
zahn ( @xcite ) defines the diffusion coefficient corresponding to the eddies which have the largest @xmath79 number so that the richardson criterion is just marginally satisfied .
however , the effects of the vertical @xmath6gradient are not accounted for and the expression only applies to low peclet numbers .
the same has been done by maeder and meynet ( @xcite ) , who considers also the effect of the vertical @xmath6gradient , the case of general peclet numbers and , in addition they account for the coupling due to the fact that the shear also modifies the local thermal gradient .
this coefficient has been used by meynet and maeder ( @xcite ) and by denissenkov et al .
( @xcite ) .
the comparisons of model results and observations of surface abundances have led many authors to conclude that the @xmath6gradients appear to inhibit the shear mixing too much with respect to what is required by the observations ( chaboyer et al .
@xciteb ; meynet and maeder @xcite ; heger et al .
namely , the observations of o type stars ( herrero et al .
@xcite , @xcite ) show much more he and n enrichments than predicted by the models which apply richardon s criterion with the @xmath68 term .
thus , instead of using a gradient @xmath68 in the criterion for shear mixing , chaboyer et al .
( @xcite ) and heger et al .
( @xcite ) write @xmath90 with a factor @xmath91 or even smaller .
this procedure is not satisfactory since it only accounts for a small fraction of the existing @xmath6gradients in stars .
the problem is that the models depend at least as much ( if not more ) on @xmath92 than on rotation , _
i.e. _ a change of @xmath92 in the allowed range ( between 0 and 1 ) produces as important effects as a change of the initial rotational velocity .
this situation has led to two other more physical approaches discussed below .
also heger et al .
( 2000 ) introduce another factor @xmath93 to adjust the ratio of the transport of the angular momentum and of the chemical elements like pinsonneault et al .
( 1989 ) .
talon and zahn ( @xcite ) account for the horizontal turbulence , which has a coefficient @xmath8 and which weakens the restoring force of the gradient of @xmath6 in the usual richardson criterion .
this allows some mixing of the chemical elements to occur as required by the observations ( talon et al .
@xcite ) .
indeed , around the convective core in the region where the @xmath6gradient inhibits mixing , there is anyway some turbulence due to both the horizontal turbulence and to the semiconvective instability , which is generally present in massive stars .
this situation has led to the hypothesis ( maeder @xcite ) that the excess energy in the shear , or a fraction @xmath94 of it of the order of unity , is degraded by turbulence on the local thermal timescale .
this progressively changes the entropy gradient and consequently the @xmath6gradient .
this hypothesis leads to a diffusion coefficient @xmath95 given by @xmath96.\end{aligned}\ ] ] the term @xmath97 in eq .
( 16 ) expresses either the stabilizing effect of the thermal gradients in radiative zones or its destabilizing effect in semiconvective zones ( if any ) .
when the shear is negligible , @xmath95 tends towards the diffusion coefficient for semiconvection by langer et al .
( @xcite ) in semiconvective zones .
when the thermal losses are large ( @xmath98 ) , it tends towards the value @xmath99 given by zahn ( @xcite ) .
( 16 ) is completed by the three following equations expressing the thermal effects ( maeder @xcite ) @xmath100 @xmath101 the system of 4 equations given by eqs . ( 16 ) , ( 18 ) and ( 19 ) form a coupled system with 4 unknown quantities @xmath49 , @xmath102 , @xmath103 and @xmath104 . the system is of the third degree in @xmath102 . when it is solved numerically , we find that as a matter of fact the thermal losses in the shears are rather large in massive stars and thus that the peclet number @xmath79 is very small ( of the order of 10@xmath105 to 10@xmath106 , see sect .
for very low peclet number @xmath107 , the differences @xmath108 are also very small as shown by eq .
thus , we conclude that eq .
( 16 ) is essentially equivalent , at least in massive stars , to the original eq .
( 17 ) above , as given by zahn ( @xcite ) .
we may suspect that this is not necessarily true in low and intermediate mass stars since there the @xmath79 number may be larger .
of course , the reynolds condition @xmath109 must be satisfied in order that the medium is turbulent .
the quantity @xmath110 is the total viscosity ( radiative + molecular ) and @xmath111 the critical reynolds number estimated to be around 10 ( cf .
denissenkov et al .
@xcite ; zahn @xcite ) .
the numerical results in sect .
4 will show the values of the various parameters and also indicate that the conditions for the occurence of turbulence are satisfied .
the physical treatment around the core also depends on the choice of the criterion for convection . in current literature , there are at least three basic sets of assumptions : a ) ledoux criterion , which leads to small cores , b ) schwarzschild s criterion , which gives `` medium size '' cores , c ) schwarzschild s criterion and overshooting , which gives large cores . here , we choose to apply the intermediate solution b ) , which means that the above equation 16 is only applied in fully radiative regions .
some consequences of this choice of convective criterion are discussed in sect . 3 below . for purpose of comparison
, we adopted here the same physical ingredients as for the solar metallicity models computed by meynet et al .
( @xcite ) .
the only exceptions , apart from the inclusion of the effects induced by rotation described above , are the following : we use the schwarzschild criterion for convection without overshooting and the mass loss rates are as indicated in sect . 2.4 .
for comparison purpose , we have computed non rotating stellar models with exactly the same physical ingredients as the rotating ones . the corresponding evolutionary tracks and lifetimes
are presented in fig . 8 and table 1 .
these stellar models have similar properties as older ones computed with the schwarzschild criterion for convection .
for instance , our non rotating 9 m@xmath39 stellar model has similar h and he burning lifetimes as the 9 m@xmath39 model computed with the schwarzschild criterion by bertelli et al .
our models also well agree with more recent computations .
indeed our ms lifetimes are similar within about two percents to the ones obtained by heger et al .
( 2000 ) for their non rotating stellar models .
these authors used the ledoux criterion with semiconvection .
however , since during the ms phase the convective core mass decreases , one expects , for this phase , only small differences between the models computed with the ledoux and the schwarzschild criterion .
these comparisons show that the present non rotating stellar models
well agree with results obtained with different stellar evolutionary codes .
many papers ( e.g. bertelli et al .
1985 ; maeder & meynet 1989 ; chin & stothers 1990 ; langer & maeder 1995 ; canuto 2000 ) have discussed the effects of different criteria for convection on massive star evolution . since our main aim in this paper is to emphasize the effects of rotation , we shall restrict ourselves to briefly mention some differences with previous grids of models computed by the geneva group . in the present
non rotating stellar models , computed without overshooting , the convective core masses are smaller than in the models by schaller et al .
( 1992 ) and meynet et al .
( 1994 ) . as a consequence ,
the stars evolve at smaller luminosities . the turn off point at the end of the ms occurs for higher t@xmath112 and lower luminosities , reducing the extent of the ms width .
the ms lifetimes are reduced . for initial masses below about 25 m@xmath39 , the he burning lifetime is increased implying an increase of the ratio @xmath113 of the lifetimes in the he and h burning phases . for the higher initial masses ,
the effects of the stellar winds become very important and dominate the effects due to a change of the criterion for convection . as a numerical example , in our non rotating 20 m@xmath39 stellar model , the convective core mass is reduced by 11.2 m@xmath39 during the whole h burning phase compared to its value given by meynet et al .
( 1994 ) . at the end of the ms phase ,
the mass of the helium core has decreased by @xmath2 20% with respect to its value in models with a moderate overshoot .
the position of the turn off point has a @xmath114/l@xmath39 decreased by 0.05 dex and a @xmath5 increased by 0.04 dex .
the ms lifetime is decreased by about 10% , the ratio @xmath113 passes from 10% in the models of meynet et al .
( 1994 ) to 14% in the present case .
let us first mention that even in some recent works the assumption of solid body rotation is often considered .
however , it is much more physical to examine the evolution of @xmath1 resulting from the transport of angular momentum by shears , meridional circulation and convection , from the effects of central contraction and envelope expansion and from the losses of angular momentum by stellar winds .
in particular , the large losses of angular momentum at the surface lead to a redistribution of the angular momentum in the interior by meridional circulation , shear turbulent diffusion and convection .
the whole problem is self consistent , since the meridional circulation and shear transport depend in turn on the degree of differential rotation.thus , the full system of equations has to be solved in order to provide the internal evolution of @xmath9 .
[ fconvo ] shows the results for the initial convergence on the zero age main sequence of @xmath1 in a 20 m@xmath0 model with solar composition . the initial model on the zero age main sequence
is supposed to rotate uniformly and with an initial surface velocity @xmath115 of 300 km s@xmath40 .
very short time steps of the order of 1920 yr are taken in the initial calculations for obtaining the internal equilibrium rotation .
the results are shown in fig .
[ fconvo ] for every 10@xmath116 model , i.e. with time intervals of 19200 yr .
we see the very fast initial changes of @xmath1 , mainly due to the meridional circulation . the circulation velocity , which is very large , is also positive everywhere at the beginning ( cf .
[ fconvu ] ) .
this means that the circulation rises along the polar axis and goes down at the equator , thus bringing angular momentum inwards . as
a consequence the angular velocity of the core increases , while that in the envelope goes down .
it is important to realize that here these fast changes of the rotation law are not due , as later in the evolution , to core contraction and envelope expansion .
we see that the changes are big at the beginning and then smaller .
the profiles of @xmath1 rapidly converges towards an equilibrium profile in a time of about 1 to 2% of the ms lifetime @xmath117 .
this is in agreement with the results for 10 and 30 m@xmath0 stars by denissenkov et al .
( @xcite ) . as noted by these authors , the timescale for the adjustment of the rotation law , i.e. @xmath118 , is short with respect to the ms lifetime for any significant rotation .
later on during the bulk of ms evolution , the profiles @xmath1 will evolve more slowly . as emphasized in sect .
2.6 , this timescale @xmath119 is not characteristic for the mixing of chemical elements .
we also notice that the degree of differential rotation in the equilibrium profile is rather modest ( cf .
zahn , @xcite ) .
[ fconvu ] shows the corresponding initial evolution of the vertical component @xmath48 of the velocity of the meridional circulation in the 20 m@xmath0 model . as seen above , initially
this velocity is large and positive everywhere , which explains the fast evolution of @xmath1 in fig .
[ fconvo ] .
then , @xmath48 decreases and becomes negative in the very external layers .
the physical reason for that is the so called gratton
pik term of the form @xmath120 which is contained in the expression of @xmath72 in eq .
when the density is very low , as in the outer regions , this negative term becomes important and produces an inverse circulation .
this means that the circulation has two big cells , an internal cell rising along the polar axis and an external one descending at the pole .
the velocity @xmath48 also converges towards an equilibrium distribution , characterized by small velocities @xmath48 , a result in agreement with denissenkov et al .
( @xcite ) .
it is interesting to compare the stationary solution of @xmath48 given by eq .
( 5 ) and the non stationary solution given by eq .
this is done in fig .
[ fdifu ] , where the curve in continuous line corresponds to the asymptotic distribution reached in the non stationary regime illustrated in fig .
[ fconvu ] .
this curve shows positive values of @xmath48 in the inner regions and negative values in the outer ones . on the contrary , the solution ( broken line ) obtained in the stationary approximation given by eq .
( 5 ) is always positive .
this is a logical consequence of the approximation made : as @xmath1 decreases outwards , only positive values of @xmath48 are possible .
said in other words , the outward transport by shears can only be compensated by the inward transport by meridional circulation , _
i.e. _ by positive values of @xmath48 . however , in non stationary models , this compensation is not achieved since the two curves in fig .
[ fdifu ] are very different .
one concludes that the stationary approximation which has been used in some works ( cf .
urpin et al .
@xcite ) is not satisfactory .
the stationary solution can also be viewed as giving a velocity which is the inverse of a velocity which could be associated to the diffusive transport by shears . from the comparison of the two curves in fig .
[ fdifu ] , one thus immediately concludes that the meridional circulation is much more efficient for the transport of the angular momentum than the shear instabilities .
the justification for this claim is the following one . in the stationary situation ,
shear transport and meridional circulation compensate each other and are thus of the same magnitude .
the fact that @xmath48 in the full non stationary case treatment is much larger than in the stationary case as shown in fig .
3 implies that the effects of meridional circulation on the transport of angular momentum are much larger than the effects of the shear .
this conclusion can also be deduced from the comparison of the timescales for the two processes . in general
one has that @xmath121 .
for the chemical elements , the transport by the shear instabilities is more efficient than the transport by the meridional circulation as discussed in sect .
2.6 where we have seen that @xmath59 is reduced by horizontal turbulence . the evolution of @xmath1 during the ms evolution of a 20 m@xmath0 star is shown in fig . [ fom ] .
we notice that initially ( i.e. for high values of the central h
content @xmath122 ) , the degree of differential rotation is small , then , it grows during the evolution .
the ratio between the central and surface value of @xmath9 remains however small during the ms evolution .
the rotation rate does not vary by more than about a factor two throughout the star . _
although this degree of differential rotation is not large , it plays an essential role in the shear effects and the related transports of angular momentum and of chemical elements .
_ the same remark also applies to the effects of differential rotation on the transport by meridional circulation , since @xmath48 critically depends on the derivatives of @xmath9 in the star ( cf .
zahn 1992 ; maeder & zahn 1998 ) .
we notice a general decrease of @xmath1 , even in the convective core which is contracting .
the main reason is mass loss at the stellar surface , which removes a substantial fraction of the total angular momentum .
this makes @xmath1 to decrease with time everywhere , because of the internal transport mechanisms , which ensure some coupling of rotation .
the same behaviour was obtained by langer ( 1998 ) in the case of rigid rotation ( _ i.e. _ in the case of maximum coupling ) . in the present models ,
shear transports the angular momentum outwards , which reduces the core rotation , while circulation makes an inward transport in the deep interior and an outwards transport in the external region .
this is responsible for the progressive flattening of the curves above @xmath123 in fig .
[ fom ] .
from the end of the ms evolution onwards , i.e. when @xmath124 , central contraction becomes faster and starts dominating the evolution of the angular momentum in the center .
the central @xmath9 grows quickly ( cf .
dotted curve in fig .
[ fom ] ) and this will in principle continue in the further evolutionary phases until core collapse .
the evolution of the angular momentum after the h burning phase ( @xmath125 ) is mainly dominated by the local conservation of the angular momentum , since the secular transport mechanisms have little time to operate .
this is the assumption we are making in the present models . during these phases ,
the chemical elements continue to be rotationally mixed in the radiative layers mainly through the effect of the shear . in the further evolutionary phases , the strong central contraction will lead to very large central rotation , unless some other processes as fast dynamical instabilities or magnetic braking occur in these stages . the evolution of @xmath48 during ms evolution is shown in fig . [ fur ] .
the outer zone with inverse circulation progressively deepens in radius during ms evolution due to the gratton
pik term , because a growing part of the outer layers has lower densities . also , we notice that the velocities @xmath48 are small in general ( cf .
urpin et al .
@xcite ) and tests have shown us that , contrary to the classical result of the eddington sweet circulation ( cf .
mestel 1965 ) , @xmath48 depends rather little on the initial rotation .
the deepening of the inverse circulation has the consequence that the stationary and non stationary solutions differ more and more as the evolution proceeds , since as said above no inverse circulation is predicted by the stationary solution .
thus we conclude that the stationary solutions are much too simplified . also , in fig .
[ fur ] , we see that the values of @xmath48 become more negative in the outer layers for the model at the end of the ms phase , when central contraction occurs . again , this effect is due to the gratton pik term .
the results of fig .
5 also shows that the application of a diffusive treatment to meridional circulation transport is unappropriate . paradoxically , the problem is less serious in regions which exhibit an inverse circulation , since there the signs of the effects are at least the same in the two treatments ( however even there the equations for advection and diffusion are different ) .
the various diffusion coefficients inside a 20 m@xmath0 star are shown in fig .
6 , when the hydrogen mass fraction @xmath126 at the center is equal to 0.20 .
we notice that in general @xmath127 . following the thermal diffusivity @xmath128 , the coefficient of horizontal diffusion @xmath129 is the second largest one , this is consistent and necessary for the validity of the assumption of shellular rotation ( zahn , @xcite ) .
since we have @xmath130 according to eq .
( 18 ) , fig .
[ fdiff ] shows that @xmath102 is typically of the order of 10@xmath105 to 10@xmath106 .
the coefficient @xmath131 , expressing the transport of the chemical elements by the meridional circulation with account of the reduction by the horizontal turbulence , is only slightly smaller ( by a factor of about 3 ) than @xmath132 in the interior
. one must be careful that @xmath131 concerns only the tranport of the elements and not that of angular momentum , for which meridional circulation is the largest effect .
6 also shows the values of the total viscosity @xmath110 , and we notice that consistently the prandtl number @xmath133 is of the order of 10@xmath134 to 10@xmath135 .
evolutionary tracks of 20 m@xmath39 models at solar metallicity for different initial velocities are plotted on fig . [ hrvingt ] .
the @xmath21 for a rotating star has been defined in sect .
2.3 . on and near the zams
, rotation produces a small shift of the tracks towards lower luminosities and @xmath21 .
this effect is due to both atmospheric distorsions and to the lowering of the effective gravity ( see e.g. kippenhahn and thomas @xcite ; maeder and peytremann @xcite ; collins and sonneborn @xcite ) . at this stage
the star is still nearly homogeneous .
when evolution proceeds , the tracks with rotation become more luminous than for no rotation .
this results from essentially two effects . on one side , rotational mixing brings fresh h
fuel into the convective core , slowing down its decrease in mass during the ms . for a given value of the central h
content , the mass of the convective core in the rotating model is therefore larger than in the non rotating one and thus the stellar luminosity is higher ( maeder @xcite ; talon et al .
@xcite ; heger et al .
@xcite ) . as a numerical example , in the @xmath136 300 km s@xmath40 models the he cores at the end of the ms are about 20% more massive than in their non rotating counterparts .
this is equal to the increase obtained by a moderate overshooting ( see sect .
3 ) . on the other side ,
rotational mixing transports helium and other h burning products ( essentially nitrogen ) into the radiative envelope .
the he enrichment lowers the opacity .
this contributes to the enhancement of the stellar luminosity and favours a blueward track .
indeed , in fig . [ hrvingt ]
, one sees that when the mean velocity on the ms becomes larger than about 130 km s@xmath40 , the @xmath21 at the end of the ms increases when rotation increases . for sufficiently low velocities
, rotation acts as a small overshoot , extending the ms tracks towards lower @xmath21 .
this results from the fact that , at sufficiently low rotation , the effect of rotation on the convective core mass overcomes the effect of helium diffusion in the envelope .
indeed , for small rotation , the time required for helium mixing in the whole radiative envelope is very long , while hydrogen just needs to diffuse through a small amount of mass to reach the convective core .
since not all the stars have the same initial rotational velocity , one expects a dispersion of the luminosities at the end of the ms . for the 20 m@xmath39 models shown on fig .
[ hrvingt ] one sees a difference of @xmath137 in @xmath138 between the luminosities of the low and fast rotators at the end of the ms .
rotation induces also a scatter of the effective temperatures at the end of the ms . in reality , the dispersion results from both different initial velocities and also , for a given initial velocity , from different angles between the axis of rotation and the line of sight .
indeed due to the von zeipel theorem ( @xcite ) the star appears bluer seen pole on than equator on . when integrated over the visible part of the star
, the effects due to orientation can reach a few tenths of a magnitude in luminosity and a few hundredths in @xmath5 ( cf . maeder and peytreman 1970 ) .
figure [ hrge ] shows the evolutionary tracks of non rotating and rotating stellar models for initial masses between 9 and 120 m@xmath39 . for the rotating stellar models , the initial velocity is @xmath115 = 300 km s@xmath40 .
there is little difference between tracks with @xmath115 = 200 or 300 km s@xmath40 ( see also talon et al .
if the effects behaved like @xmath139 , there would be larger differences .
the present saturation effect occurs because outward transport of angular momentum by shears are larger when rotation is larger , also a larger rotation produces more mass loss , which makes a larger reduction of rotation during the evolution .
let us note however that for some surface abundance ratios as n / c or n / o ( see table 1 ) , the increase from @xmath115 = 200 to 300 km
s@xmath40 produces significant changes .
thus , the similarity of the evolutionary tracks does not necessarily imply the similarity of the surface abundances for these elements .
rotation reduces the ms width in the high mass range ( m @xmath140 m@xmath39 ) .
let us recall that when the mass increases , the ratio of the diffusion timescale for the chemical elements to the ms lifetime decreases ( maeder @xcite )
. as a consequence , starting with the same @xmath115 on the zams , massive stars will be more mixed than low mass stars at an identical stage of their evolution .
this reduces the ms width since greater chemical homogeneity makes the star bluer .
moreover , due to both rotational mixing and mass loss , their surface will be rapidly enriched in h burning products .
these stars will therefore enter the wolf
rayet phase while they are still burning their hydrogen in their core .
this again reduces the ms width . for initial masses between 9 and 25 m@xmath39 ,
the ms shape is not much changed by rotation at least for @xmath141 km s@xmath40 . the post
ms evolution of the most massive stars ( m @xmath142 m@xmath39 ) which become w r stars will be discussed in a forthcoming paper .
we shall just mention one point of general interest here : for low or moderate rotation , the convective core shrinks as usual during ms evolution , while for high masses ( @xmath143 m@xmath0 ) and large initial rotations ( @xmath144 ) , the convective core grows in mass during evolution .
this latter situation occurs in the fast rotating 60 m@xmath39 model shown on fig .
these behaviours , i.e. reduction or growth of the core , determine whether the star will follow respectively the usual redwards ms tracks in the hr diagram , or whether it will bifurcate to the blue ( cf .
maeder @xcite ; langer @xcite ) towards the classical tracks of homogeneous evolution ( schwarzschild @xcite ) and likely produce w
r stars .
the stars with initial masses between 15 and 25 m@xmath39 become red supergiants ( rsg ) .
rotation does not change qualitatively this behaviour but accelerates the redwards evolution , especially for the 15 and 20 m@xmath39 models . as a numerical example , for an initial @xmath115 = 300 km s@xmath40 , the model stars burn all their helium as red supergiants at @xmath21 below 4000 k , while the non rotating models spend a significant part of the he burning phase in the blue part of the hr diagram : for the non rotating 15 and 20 m@xmath39 models , respectively 25 and 20% of the total he
burning lifetime is spent at @xmath145 .
the behaviour of the rotating models results mainly from the enhancement of the mass loss rates .
this effect prevents the formation of a big intermediate convective zone and therefore favours a rapid evolution toward the rsg phase ( stothers and chin @xcite ; maeder @xcite ) .
let us note that the dispersion of the initial rotational velocities produces a mixing of the above behaviours .
very interestingly , for the 12 m@xmath39 model a blue loop appears when rotation is included .
this results from the higher luminosity of the rotating model .
the higher luminosity implies that the outer envelope is more extended , and is thus characterized by lower temperatures and higher opacities at a given mass coordinate . as a consequence , in the rotating model during the first dredge up , the outer convective zone proceeds much more deeply in mass than in the non rotating star .
typically in the non rotating model the minimum mass coordinate reached by the outer convective zone is 6.6 m@xmath39 while in the rotating model it is 2.6 m@xmath39 .
this prevents temporarily the extension in mass of the he core and enables the apparition of a blue loop .
indeed the lower the mass of the he
core is , the lower its gravitational potential . according to lauterborn et al .
( @xcite , see also the discussion in maeder and meynet @xcite ) , a blue loop appears when the gravitational potential of the core @xmath146 is inferior to a critical potential @xmath147 depending only on the actual mass of the star which is about the same for the rotating and non rotating model .
this explains the appearance of a blue loop in the 12 m@xmath39 rotating model .
for the 9 m@xmath39 model , the minimum mass coordinate reached by the outer convective zone is not much affected by rotation and the models with and without rotation present very similar blue loops .
[ cols="^,^,^,^,^,^,^,^,^,^,^,^,^,^,^,^,^,^,^ " , ] from table [ tbl-1 ] one sees that for @xmath148 the lifetimes are increased by about 2030% when the mean rotational velocity on the ms increases from 0 to @xmath2200 km s@xmath40 .
this modest increase is explained by the fact that even if there is more fuel available in the core , the luminosity is also increased . from the data presented in table [ tbl-2 ]
, one can deduce a nearly linear relation between the relative enhancement of the ms lifetime , @xmath149 and @xmath150 , where @xmath151/t_{\rm h}(0).\ ] ] one obtains , @xmath152 where @xmath150 is in km s@xmath40 .
this relation reproduces the values of @xmath153 from table [ tbl-2 ] with an accuracy better than 5% .
it also applies with the same accuracy to the values listed in table [ tbl-1 ] for the masses between 15 and 40 m@xmath39 .
the he burning lifetimes are less affected by rotation than the ms lifetimes .
the changes are less than 10% .
the ratios @xmath154 of the he to h burning lifetimes are only slightly decreased by rotation and remain around 1015% . on fig .
[ isoc ] isochrones for ages between about 8 and 20 10@xmath155 yr , computed from non rotating and rotating stellar models are presented .
the `` rotating '' isochrones are computed from the models with an initial rotational velocity @xmath115 of 200 km s@xmath40 on the zams . at a given age ,
the upper part of the `` rotating '' isochrones are bluer and more luminous .
typically , for an age equal to about @xmath156 years ( @xmath157 age = 7.3 ) the reddest point on the ms is shifted by 0.015 dex in @xmath5 and by - 0.6 in @xmath138 when rotation is taken into account .
an isochrone with rotation is almost identical to an isochrone without rotation with log age smaller by 0.1 dex .
this has for consequence that rotation slightly increases the age associated to a given cluster . from fig .
9 , one sees that the `` rotating '' isochrone for an age equal to @xmath156 years has the same luminosity at the turn off than the `` non - rotating '' isochrone for an age equal to @xmath158 years .
thus rotation increases the age estimate by about 25% .
we may wonder whether this effect explains the age difference between the estimates based on the upper ms and the estimates based on the lithium content of the very low mass stars ( see e.g. martin et al .
1998 ; barrado y navascus et al .
one must also account for the dispersion of rotational velocities and possibly of the orientation angles .
these two effects introduce some dispersion in the way stars are distributed in the hr diagram and thus affect the interpretation of the clusters observed sequences ( cf .
maeder @xcite ) .
if a bluewards track occurs , as for very massive stars with fast rotation , the larger core and mixing lead to much longer lifetimes in the h burning phase . in this case , the fitting of time lines becomes hazardous .
figs . [ v / age ] and [ om / age ] show the evolution of @xmath83 and of @xmath159 as a function of the age for the present models .
firstly , we notice that for models without mass loss , as shown for the 20 m@xmath0 with @xmath28 = 0 , @xmath83 and @xmath159 go up fastly so that the critical velocity would be reached near the end of the ms phase .
the current model of 20 m@xmath0 with mass loss show a significant decrease of @xmath83 , while the critical ratio remains almost constant during most of the ms phase .
figs . [ v / age ] and [ om / age ] show how fastly rotation decreases at the surface of the most massive stars , which lose a lot of mass .
consistently we see that the reduction of the surface rotation is much larger for the more massive stars .
this is of course a consequence of the removal of large amounts of angular momentum by the stellar winds .
the effect is amplified by the increase of the mass loss in fast rotators ( eq . 2 ) .
we see that the decrease of @xmath159 is so strong that it will prevent a massive star to reach the critical velocity near the end of the ms phase .
if the star makes extended excursions in the hr diagram at the end of the ms like is the case for the 60 m@xmath0 model , then it may reach the critical velocity .
the specific case of stars close to the @xmath9limit will be examined in a future study , since this requires some further theoretical developments .
let us mention here that the present results differ from those obtained by sackmann and anand ( @xcite ) and langer ( @xcite , @xcite ) .
indeed , these authors find that the star reaches the break up limit during the ms phase . as an example , in a 60 m@xmath0 star , even a model with an initial @xmath160 of 100 km
s@xmath40 reaches the break up limit near the end of the ms phase ( langer 1998 ) .
this result leads langer to conclude that most massive stars may reach the break up limit or the so called @xmath9limit during their ms evolution .
let us however emphasize here that such a conclusion is based on a particular definition of @xmath32 still subject to discussion ( see sect .
2.4 ) , on the assumption of solid body rotation , and on models not accounting for the effects of rotationnally induced mixing .
we see here that modifying these hypothesis ( and also using other prescriptions for the mass loss rates ) lead to very different results .
one of the first step to clarify the situation is to determine which expression for @xmath32 ( cf .
glatzel 1998 ; langer 1998 ) is the correct one and how rotation affects the mass loss rates .
these developments , now in progress , will be particularly needed for the study of the evolution of the most massive stars , like a 120 m@xmath39 model , which have a high value of the eddington factor .
also this is important for the formation and evolution of w
r stars , which will be studied in a further work . fig .
[ v20/age ] shows the evolution of @xmath83 with age for models of a 20 m@xmath0 star with different initial velocities @xmath115 from 50 to 580 km s@xmath40 .
we see that the decrease in the surface @xmath83 is larger for larger initial rotation .
this is a consequence of the larger mass loss rates in fast rotators .
we notice some convergence ( cf .
langer 1998 ) of the curves for the large initial @xmath160 .
this convergence would be more pronounced and affect the models with lower @xmath115 if the dependence of @xmath29 vs @xmath83 would be stronger than given by eq .
this certainly reduces the scatter of @xmath83 near the end of the ms phase , but does not produce a full convergence . the models of 12 and 15 m@xmath0 show curves in figs .
[ v / age ] and [ om / age ] with little reduction of @xmath83 , while there are slight increases of the critical ratio @xmath159 during ms evolution .
a peak is reached during the overall contraction phase at the end of the ms phase .
then , @xmath83 and the critical ratio go down as the star moves to the red supergiant phase . during such a fast evolution
, we may say that the rotation evolves almost like the case of simplified models , where the angular momentum is conserved locally .
the large growth of the radius just implies a decrease of @xmath83 and of @xmath159 . later during the blue loops , where the cepheid instability strip is crossed , the rotation velocity becomes very large again and could easily become close to critical .
this behaviour , also found by heger & langer ( 1998 ) , results from the stellar contraction which concentrates a large fraction of the angular momentum of the star ( previously contained in the extended convective envelope of the rsg ) in the outer few hundredths of a solar mass .
this result suggests that rotation may also somehow influence the cepheid properties , in addition to the increase of stellar luminosity discussed above .
amazingly , we notice that the stars with initially low mass loss during the ms phase , like for stars with m @xmath161 m@xmath0 , have more chance to reach the break up velocities and thus huge mass loss than the more massive stars which lose a lot of mass on the ms .
it is somehow surprising that little mass loss during the ms may favour large mass loss rates at the end of the ms phase .
this may explain why stars close to break
up , like the be stars , do not form among the o type stars , but mainly among the b type stars , where we see that the ratio @xmath159 may increase during the ms phase .
another related observation is the fact that the relative number of be stars with respect to b type stars is much higher in the lmc and smc than in the galaxy ( cf .
maeder et al .
@xcite ) . in the lmc and smc , due to the lower metallicity ,
the average mass loss rates are lower and thus these stars may keep higher rotation in general and thus form more be stars .
this explanation does not exclude differences in the distribution of @xmath160 as well . the evolution of @xmath83 in the hr diagram is shown in fig .
[ v_hr ] . starting with models having a velocity of 300 km s@xmath40 on the zero age main sequence ,
we give some lines of constant @xmath83 over the hr diagram . on the ms , we notice in particular that the decrease is much faster for the most massive stars than for stars with m @xmath162 15 m@xmath0 .
this difference remains also present in the domain of b supergiants . during the crossing of the hr diagram ,
the rotational velocities decrease fastly , to become very small , i.e. of the order of a few km s@xmath40 , in the red supergiant phase .
it is beyond the scope of this paper to make detailed comparisons of the evolution of the distribution of the velocities over the hr diagram , however , we may notice a few points .
the fact that the average @xmath163 is lower for o type stars than for the early b type stars ( slettebak , @xcite ) may be the consequence of the higher losses of mass and angular momentum in the most massive stars . also , we remark that the increase of @xmath163 from o stars to b stars is larger for the stars of luminosity class iv than for class v ( fukuda , @xcite ) .
this is consistent with our models , which show ( cf . fig .
[ v_hr ] ) that the differences of @xmath163 beween o and b type stars are much larger at the end of the ms phase .
another fact in the observed data is the strong decrease of @xmath163 for the massive supergiants of ob types .
this is predicted by all stellar models ( cf . also langer @xcite ) due to the growth of the stellar radii .
further detailed comparisons may perhaps provide some new tests and constraints .
the chemical abundances offer a very powerful test of internal evolution and they give strong evidences in favour of some additional mixing processes in massive stars .
a review of the observations may be found in maeder and meynet ( 2000 ) .
here we shall concentrate on the discussion of the theoretical results and we shall compare them with some recent observations . the most striking feature appearing in tables [ tbl-1 ] and [ tbl-2 ] as well as on fig .
[ hrcn ] is the change of surface abundances in rotating stellar models ( cf .
langer 1992 ) .
the he , n enrichments and the related c and o depletions at the surface already occur on the ms .
the more massive the star is , the more pronounced are the enrichments , supporting the expectation that mixing becomes more and more efficient when the mass increases ( maeder @xcite ) .
the same is true when the initial rotational velocity increases . during the crossing of the hertzsprung
russel diagram , the evolution is sufficiently short for not allowing any important change of the surface abundances ( see fig .
[ hrcn ] ) .
further changes occur when the star becomes a red supergiant and undergoes the first dredge up .
in contrast , non rotating stellar models show no change of the surface abundances until the first dredge up in the red supergiant stage ( see table [ tbl-1 ] ) .
this implies that these models predict some enrichment neither during the ms nor in the blue supergiant phase unless a blue loop is formed .
moreover , as can be seen in fig .
[ hrcn ] , the ratios obtained at the end of the he burning phase are significantly lower than the ones obtained in rotating models . fig .
[ ysvit ] shows the evolution in the @xmath165 versus surface velocity plane where @xmath166 , @xmath167 and @xmath168 being the abundances in number of helium and hydrogen at the surface .
the theoretical tracks correspond to the equatorial velocities @xmath83 , while the observed points from herrero et al .
( @xcite , @xcite , @xcite ) are @xmath164 .
this implies that the observed values are smaller on the average by a factor @xmath169 with respect to the theoretical values . in fig .
15 , the tracks go generally from the bottom on the right to the top on the left . along a given track ,
the he enrichment increases when the velocity decreases . the shaded zone in fig .
[ ysvit ] corresponds to the ms for the models with @xmath115 = 300 km s@xmath40 .
the higher initial mass , the greater the he enrichments which can be reached during the ms .
we see also that during the ms phase , for initial masses inferior to about 60 m@xmath39 , all the tracks with @xmath115 = 300 km s@xmath40 follow more or less the same @xmath165 versus @xmath83 relation .
however , the relation is changed when the initial velocity is different ( see the fast rotating 20 and 60 m@xmath39 models ) .
the surface he enrichments on the ms generally depend on the following factors : the initial mass , the initial metallicity ( maeder & meynet 2000b ; meynet 2000a ) , the initial velocity and the age of the star .
a low surface velocity at a given evolutionary stage does not exclude that in the past the star was a fast rotator .
the slow rotation may result from the loss of angular momentum by stellar winds and/or from the increase of radius of the star . on the other hand ,
a high velocity in the past implies he and n enrichments of the surface .
let us compare in fig .
[ ysvit ] the theoretical tracks in the @xmath165 versus @xmath83 plane with the observations of ob stars performed by herrero et al .
( @xcite , @xcite , @xcite ) .
recent works ( mcerlean et al .
@xcite ; smith and howarth @xcite ) indicate that accounting for a microturbulent velocity line broadening in the model atmosphere reduces the derived he
abundances for supergiants later than o9 . however , according to villamariz and herrero ( @xcite ) this effect can not explain all the observed overabundances , especially for the earlier types . on fig .
[ fgeff ] the observed points are plotted in the @xmath4 versus @xmath5 plane where @xmath170 is the effective surface gravity .
we estimate @xmath171 for the average orientation angle .
let us note that for the models plotted in fig .
[ fgeff ] , there is little difference between the effective gravities at the pole and at the equator .
indeed the ratio between these two gravities never exceeds 1.3 which means a vertical dispersion of about 0.1 dex in fig .
rotating and rotating evolutionary tracks are superposed to the observed points in fig .
[ fgeff ] . since most of the enriched stars are in the vicinity of the 120 and 60 m@xmath39 tracks ( see fig . [ fgeff ] ) , we can wonder wether the changes of the surface abundances can be explained as an effect of mass loss only .
it does not seem to be the case , because the part of the track shown on fig .
[ fgeff ] for the non rotating 60 m@xmath39 model presents no surface he enrichment . in the case of the non rotating 120 m@xmath39 model
only the part of the track with @xmath4 inferior to 3.3 has @xmath172 .
thus the he enrichments can not be accounted for by current evolutionary models as was already pointed out by herrero et al .
( @xcite , cf .
also maeder 1987 ) . in the following
we shall suppose that these enhancements are due to rotation .
let us consider four groups of stars .
in the first group we place all the stars presenting no he enrichment at their surface ( @xmath173 ) and having @xmath174 km s@xmath40 ( empty squares on figs .
[ ysvit ] and [ fgeff ] ) , in the second one are the enriched stars ( @xmath175 0.12 , solid triangles ) having @xmath174 km s@xmath40 .
the third group ( which is empty at present ) consists of the fast rotators with no he enrichment .
these stars would occupy the bottom right corner in fig .
[ ysvit ] .
finally , the fourth group contains the he enriched stars with @xmath176 km s@xmath40 ( solid circles ) .
* group 1 * the non enriched stars with low rotation can be interpreted either as stars with small initial velocities or as young fast rotators whose surface has not yet been enriched in helium by rotational mixing ( see also herrero et al .
@xcite ) . in this respect
let us mention that all the stars having @xmath177 , and which therefore are probably not too evolved , present no he enrichment .
* group 2 * the most striking feature of the second group of stars ( @xmath178 and @xmath179 km s@xmath40 ) is the fact that they are distributed in a relatively narrow range of @xmath180 between 80 and 160 km s@xmath40 .
this may result from the following facts : firstly there exists a minimum value of the initial velocity for rotational mixing to be able to drive changes of the helium surface abundances during the h burning phase .
secondly the observed distribution also reflects the way the surface velocity declines during the h burning phase and the narrow range of observed velocities may result from some convergence effect as the one mentionned in sect .
6.1 ( see also fig .
we see that the high @xmath165 values reached by some of these stars ( superior to @xmath20.16 ) would be compatible with their high initial mass implied from their position in fig .
[ fgeff ] .
* group 3 * very interestingly no stars are observed with a @xmath181 km s@xmath40 and no he enrichment . keeping in mind that the observed stars do not represent a statistical complete sample ,
one can nevertheless wonder why no stars are observed in this zone .
a possibility would be that these very fast rotators do not exist , but this is not realistic .
indeed stars with velocities as high as 300 km s@xmath40 and more have been observed on the ms ( e.g. penny @xcite ; howarth et al .
@xcite ; see also fig .
[ ysvit ] ) . moreover if such stars were not formed how to explain the stars in group 4 , having a very high @xmath164 and an important he
enrichment ( the solid circles in fig .
[ ysvit ] ) ?
these stars are likely the descendants of very fast ms rotating stars ( see the discussion in the subsection group 4 below ) . the fact that no stars are observed in the group 3 may indicate that the change of the surface abundances occur within a small fraction of the visible ms life . for the fast rotating 60 m@xmath39 model plotted on fig .
[ ysvit ] , the surface retains its initial composition ( @xmath182 ) only during a third of its h burning lifetime .
more rapidly rotating models would still reduce the fraction of the ms time spent with no change of the surface abundances .
the lack of fast rotators with normal surface composition may be due to the fact that young massive fast rotators are still embedded in the cloud from which they formed .
the models of massive star formation with accretion ( bernasconi and maeder @xcite ) indicate that when the stars become visible , they have already burnt some fraction of their central hydrogen , thus some transport of he and n to the stellar surface may have occured .
* group 4 * could the stars belonging to the fourth group ( @xmath178 and @xmath181 km/ s ) be formed by stars previously in the low velocity range and which have been accelerated for instance by contraction on a blue loop ?
the answer is likely no .
indeed blue loops can not accelerate the surface beyond the critical velocity which , for a 12 m@xmath39 blue supergiant at the tip of a blue loop , is of the order of 250 km s@xmath40 . in addition
some stars in the group 4 are classified as ms stars ( see the stars labeled with a v on figs . [ ysvit ] and [ fgeff ] ) .
another possibility would be to consider the stars in group 4 as secondary stars in close binary systems which would have been accelerated through the process of mass accretion ? but these stars are not observed to belong to binary systems . therefore the most reasonable hypothesis is to consider the stars in group 4 as the natural descendants of very fast ms rotating stars .
their chemical enrichment at the surface is very fast .
their chemical structure as a result of the strong rotational mixing is probably near homogeneity .
this view is supported by the fact that high values of @xmath164 are observed for very high values of @xmath165 implying that the surface velocity does not decrease too much in the course of the evolution .
this can be accounted for if the star remains compact , _
i.e. _ in the blue part of the hr diagram as is the case for a strongly mixed star ( maeder @xcite ; see also the fast rotating 60 m@xmath39 track in fig .
[ ysvit ] ) .
another effect could also be important in that respect , _
i.e. _ the anisotropy of the stellar winds when stars are rotating near break up . for the hot stars
, the von zeipel ( @xcite ) theorem implies that most of the mass is ejected from the pole ( maeder @xcite ) .
this prevents the loss of important angular momentum and maintains a high surface velocity .
if the stars in the fourth group are nearly homogeneous objects , one would expect higher effective gravities than observed .
typically , the fast rotating well mixed 60 m@xmath39 model remains at a high value of @xmath4 , at least for the portion of the evolution computed here ( see fig .
[ fgeff ] ) .
moreover , as noted above , the maximum dispersion in @xmath4 due to orientation effects is around 0.1 dex .
could the @xmath4 be underestimated ?
even if it is difficult on the base of the present data to ascertain such a point of view , one can note that the @xmath183 for these fast rotating stars can not be too far from 1 , otherwise one would obtain surface velocities above the break up value .
this means that these stars are seen essentially equator on .
the @xmath184 in the equator band is inferior to the surface averaged @xmath184 and thus the observed @xmath184 might be underestimated .
of course a quantitative analysis requires a detailed study of how the variation of @xmath184 with the latitude affects the spectroscopically determined gravities .
most blue supergiants present surface enrichments . for instance , walborn ( @xcite , @xcite ) showed that ordinary ob supergiants have he and n enrichments as a result of cno processing .
only the small group of peculiar obc stars has the normal cosmic abundance ratios ( cf . also howarth and prinja @xcite ; herrero et al .
@xcite ; gies and lambert @xcite ; lennon @xcite ) . a possibility to explain the he and n enrichments in supergiants would be that blue supergiants are on a blue loop after a first red supergiant stage where dredge up has occurred producing the observed surface enrichments .
however as we shall see below , this does not appear as the good explanantion at least for some of the observed enrichments .
[ nh ] illustrates the changes of the nitrogen to carbon ratios
n / c from the zams to the red supergiant stage for stars in the mass range from 9 to 20 m@xmath0 .
the n / c ratio appears as the most sensitive observable parameter . for non rotating stars , the surface enrichment in nitrogen only occurs when the star reaches the red supergiant phase ; there , cno elements are dredged up by deep convection .
the behaviour is the same as for the he
enrichments discussed above . for rotating stars ,
n excesses already occur during the ms phase and they are larger for higher rotation and initial stellar masses . at the end of the ms phase of the 12 m@xmath39 model , the n / c ratio is enhanced by factors 2.4 and 4.3 for @xmath115 = 200 and 300 km s@xmath40 respectively .
these factors are increased up to 4 and 7.1 for the 20 m@xmath39 models with the same @xmath115 . on fig .
[ nh ] we plot also some observations of supergiants performed by gies and lambert ( @xcite ) , lennon ( @xcite ) and venn ( @xciteb ) .
the plotted values are @xmath185_*-[{\rm n / c}]_{b}$ ] , where the abundance ratios in number are measured at the surface of the star ( @xmath186 ) , and at the surface of main sequence b stars ( @xmath187 ) supposed to have retained their pristine cosmic abundances . gies and lambert ( @xcite )
provide `` lte '' and `` nlte '' n / c ratios .
we plot here the smaller `` lte '' ratios since , for the stars in common with the lennon sample , they are similar to lennon s results ( see the discussion in venn @xcite ) . from the positions of the observed stars in the @xmath4 versus @xmath5 diagram , one obtains that the range of initial masses for the a type supergiants shown on fig .
[ nh ] are between 5 and 20 m@xmath39 . from fig .
[ nh ] , one notes first that at the @xmath21 of the observed supergiants , non rotating stellar models predict no enrichment at all unless there is a blue loop .
however it is not likely that all the observed points are at the tip of a blue loop .
indeed some of the stars have initial masses above 15 m@xmath39 and , at solar metallicity , current grids of models ( arnett @xcite ; schaller et al .
@xcite ; alongi et al .
@xcite ; brocato and castellani @xcite ) only predict blue loops for masses equal or lower than 15 m@xmath39 .
moreover many of the observed stars have n / c ratios too low to result from a first dredge up episode ( see for instance the position of the blue loop of the non rotating 9 m@xmath39 model in fig .
[ nh ] ) . therefore venn ( @xcite ) suggested that at least those stars presenting the lowest n / c ratios are on their way from the ms to the red giant branch and have undergone some mixing in the early stage of their evolution .
if such stars are not at all accounted for by standard evolutionary tracks , rotating models can naturally reproduce their observed surface abundances as can be seen on fig.[nh ] .
moreover as already noted above , theory predicts larger excesses for higher masses , a result in agreement with the suggestion of takeda and takeda hidai ( @xcite ) recently confirmed by mcerlean et al .
( @xcite ) .
mass loss by stellar winds and rotational mixing in the stellar interior are certainly the two hydrodynamical phenomena which most deeply affect the evolution of massive stars . far from being a small refinement in the physics of stellar interior , rotation appears as an essential ingredient of future grids of stellar models . in particular among the important points
wich are not discussed here but which will be studied in more details in forthcoming papers are the effects of rotation on the population of red and blue supergiants at various metallicities , on the evolutionary scenarios leading to the formation of wolf rayet stars ( maeder 1987 ; fliegner and langer 1995 ; meynet 2000b ) and on the stellar yields ( heger et al .
these questions are important for a better understanding of starbursts regions and of the chemical evolution of galaxies .
kippenhahn r. , thomas h.c .
, 1970 , a simple method for the solution of the stellar structure equations including rotation and tidal forces . in : slettebak a. ( ed . ) proc .
4 , stellar rotation . gordon and breach science publishers , p. 20
kippenhahn r. , weigert a. , 1990 , stellar structure and evolution , springer verlag , berlin lamers h. , cassinelli j. p. , 1996
, mass loss from stars . in : leitherer c. , fritze - von - alvensleben u. , huchra j. ( eds . ) from stars to galaxies : the impact of stellar physics on galaxy evolution .
series 98 , p. 162
langer n. , 1992 , a&a 265 , l17 langer n. , 1997 , the eddington limit in rotating massive stars . in : nota a. , lamers h. ( eds . ) luminous blue variables : massive stars in transition .
series 120 , p. 83
maeder a. , 1995a , population i stellar structure and evolution : facing the lingering difficulties to make a step forward . in : stobie r.s .
, whitelock p.a .
( eds ) astrophysical applications of stellar pulsation .
series 83 , p. 1
maeder a. , 1997 , a&a 321 , 134 ( paper ii ) maeder a. , 1998 , the physics of rotational mixing in massive stars . in : howarth i. ( ed . )
boulder - munich ii : properties of hot , luminous stars .
131 , p. 85
mathys g. , 1999 , direct observational evidence for magnetic fields in hot stars . in : wolf b. , stahl o. , fullerton a.w .
169 , variable and non - spherical stellar winds in luminous hot stars , springer , berlin , p. 95
meynet g. , 1999 , rotation and wolf rayet star formation . in : wolf b. , stahl o. , fullerton a.w .
169 , variable and non - spherical stellar winds in luminous hot stars , springer , berlin , p. 377
penny l.r . , gies d.r . ,
bagnuolo w.g . ,
1999 , discrepancies between observational and theoretical parameters of three o - type binaries . in : van der hucht k.a . , koenigsberger g. , eenens p.r.j .
193 , wolf rayet phenomena in massive stars and starburst galaxies .
ser . , p. 86
, mccarthy j.h . ,
lennon d.j .
, kudritzki r.p . , 1998 , a - supergiant abundances in local - group galaxies .
in : howarth i. ( ed . ) boulder - munich ii : properties of hot , luminous stars .
131 , p. 177
villamariz m.r .
, herrero a. , 1999 , the effect of micro turbulence in o type star analyses . in : van der hucht k.a . , koenigsberger g. , eenens p.r.j .
193 , wolf rayet phenomena in massive stars and starburst galaxies .
, p. 265
walborn n.r . ,
1988 , evolutionary helium and cno anomalies in the atmospheres and winds of massive hot stars . in : nomoto k. ( eds . ) proc .
108 , atmospheric diagnostics of stellar evolution .
chemical peculiarity , mass loss , and explosion .
springer - verlag , berlin , p. 70 | grids of models for rotating stars are constructed in the range of 9 to 120 m@xmath0 at solar metallicity .
the following effects of rotation are included : shellular rotation , new structure equations for non conservative case , surface distorsions , increase of mass loss with rotation , meridional circulation and interaction with horizontal turbulence , shear instability and coupling with thermal effects , advection and diffusion of angular momentum treated in the non stationary regime , transport and diffusion of the chemical elements .
globally we find that for massive stars the effects of rotation have an importance comparable to those of mass loss .
due to meridional circulation the internal rotation law @xmath1 rapidly converges , in 12 % of the ms lifetime , towards a near equilibrium profile which then slowly evolves during the ms phase .
the circulation shows two main cells . in the deep interior ,
circulation rises along the polar axis and goes down at the equator , while due to the gratton pik term it is the inverse in outer layers .
this external inverse circulation grows in depth as evolution proceeds .
we emphasize that a stationary approximation and a diffusive treatment of meridional circulation would be unappropriate .
after the ms phase , the effects of core contraction and envelope expansion dominate the evolution of the angular momentum .
the surface velocities decrease very much during the ms evolution of the most massive stars , due to their high mass loss , which also removes a lot of angular momentum .
this produces some convergence of the velocities , but not necessarily towards the break up velocities .
however , stars with masses below @xmath212 m@xmath0 with initially high rotation may easily reach the break up velocities near the end of the ms phase , which may explain the occurence of be stars .
some other interesting properties of the rotational velocities are pointed out . for an average rotation ,
the tracks in the hr diagram are modified like a moderate overshoot would do . in general
, an average rotation may increase the ms lifetime up to about 30 % ; for the helium burning phase the effects are smaller and amount to at most 10 % . from plots of the isochrones , we find that rotation may increase the age estimate by about 25 % in general . however , for stars with m @xmath340 m@xmath0 and fast rotation , a bluewards `` homogeneous like '' track , with important he and n enrichments , may occur drastically affecting the age estimates for the youngest clusters .
rotation also introduces a large scatter in the mass luminosity relation : at the same @xmath4 and @xmath5 , differences of masses by 30 % may easily occur , thus explaining what still remains of the alleged mass discrepancy .
rotation also brings significant surface he and n enhancements , they are higher for higher masses and rotation . while it is not difficult to explain very fast rotators with he and n excesses , the present models also well account for the many ob stars exhibiting surface enrichments and moderate or low rotation , ( cf .
herrero et al .
@xcite , @xcite ) .
these stars likely result from initially fast rotators , which experienced mixing and lost a lot of angular momentum due to enhanced mass loss .
the comparison of the n
excesses for b and a type supergiants supports the conclusion by venn ( @xcite , @xcite ) , that these enrichments mostly result from mixing during the ms phase , which is also in agreement with the results of lyubimkov ( @xcite ) . |
neutrinos undergo the phenomenon of flavor oscillations ; a neutrino created in a particular flavor ( electron , mu , or tau ) will change its flavor , a phenomenon not accounted for in the standard model of the electroweak interaction .
a possible explanation of this behavior is neutrino oscillations .
the neutrinos are assigned masses .
the created neutrino has definite flavor , but these flavor eigenstates are not the physical free particles , the mass states .
this leads to a phenomenology that is parameterized by three mixing angles , a cp violating phase , and two mass - squared differences , @xmath10 .
@xmath11to appear in horizons of innovative theories , experiments , and supercomputing in nuclear physics ( new orleans , june 4 - 6 , 2012 ) experiments have found that neutrino oscillations approximately break into two @xmath12 oscillations , termed solar and atmospheric oscillations .
the solar oscillations are predominantly governed by the solar data and the kamland experiment , and they predominantly determine @xmath13 and @xmath14 , while the atmospheric oscillation parameters @xmath15 and @xmath16 are predominantly determined by the atmospheric data and the long - baseline muon disappearance data .
the last mixing angle @xmath0 and the cp - violating phase @xmath17 mix the two @xmath12 oscillations .
an outstanding question in neutrino phenomenology is that of the hierarchy , whether the small mass - squared difference lies below ( normal hierarchy ) or above ( inverse hierarchy ) the large mass - squared difference . a related question , as we will show , is whether the value of @xmath0 is positive or negative .
we utilize the convention @xcite that @xmath18 and @xmath19 .
this is the more convenient parameterization for discussing the symmetries of interest here .
recent long - baseline reactor experiments @xcite have accurately measured the value of @xmath0 and found it to be large , @xmath20 . before these measurements
, the world s data demonstrated very little preference for either the hierarchy or the sign of @xmath0 @xcite .
the goal of this work is to examine the effect that the recent knowledge of @xmath0 has on the hierarchy question and on the sign of @xmath0 .
this report is preliminary in that we will fix the mixing parameters except @xmath0 at values from a recent analysis .
since there is no strong evidence for cp - violation , we set the cp phase @xmath17 to zero . in vacuum versus neutrino energy for the t2k experiment .
the solid curve depicts the normal hierarchy , positive @xmath0 as well as the inverse hierarchy , negative @xmath0 results .
the dashed curve depicts the normal hierarchy , negative @xmath0 as well as the inverse hierarchy , positive @xmath0 results .
the horizontal bar represents the range where the muon interaction rate is greater than 50% of its maximum .
, width=288 ] [ fig1 ]
the task of extracting neutrino mixing parameters from data is complicated by the existence of symmetries that lead to degeneracies @xcite , where degeneracies mean different mixing parameters yield the same oscillation probability .
the minos @xcite , t2k @xcite , and the future no@xmath5a @xcite experiments are at an @xmath3 where such a symmetry nearly exists . in the limit of @xmath21 and
no matter effects , these experiments would be insensitive to the hierarchy .
the symmetry pertinent to these experiments , using our bounds on the mixing angles , is the combined interchange of hierarchy and a change of the sign of @xmath0 , a four - fold degeneracy . in fig .
[ fig1 ] we depict the vacuum oscillation probability @xmath22 for the t2k experiment for full three - neutrino mixing using the parameters @xcite @xmath23 ev@xmath24 , @xmath25 , @xmath26 ev@xmath24 , @xmath27 , and @xmath28 . versus neutrino energy for the t2k experiment , left , and the minos / no@xmath5a experiments , right .
the blue curves depict the normal hierarchy , the red curves the inverse hierarchy .
the solid curves depict positive @xmath0 , the dashed curves negative @xmath0 . for the left curve , the horizontal brown bar represents the range where the muon interaction rate is greater than 50% of its maximum , while for the right curve the brown bar represents the same quantity for the no@xmath5a experiment , and the turquoise bar represents the minos experiment . ,
title="fig:",width=288 ] versus neutrino energy for the t2k experiment , left , and the minos / no@xmath5a experiments , right .
the blue curves depict the normal hierarchy , the red curves the inverse hierarchy .
the solid curves depict positive @xmath0 , the dashed curves negative @xmath0 . for the left curve ,
the horizontal brown bar represents the range where the muon interaction rate is greater than 50% of its maximum , while for the right curve the brown bar represents the same quantity for the no@xmath5a experiment , and the turquoise bar represents the minos experiment .
, title="fig:",width=288 ] versus @xmath29 for t2k , left , and minos , right .
solid curves are normal hierarchy , dashed inverse hierarchy .
blue curves are positive @xmath0 , red curves negative @xmath0.,title="fig:",width=288 ] versus @xmath29 for t2k , left , and minos , right .
solid curves are normal hierarchy , dashed inverse hierarchy .
blue curves are positive @xmath0 , red curves negative @xmath0.,title="fig:",width=288 ] we see that the four - fold degeneracy is broken by non - zero @xmath0 .
however , the solid curve represents both the normal hierarchy , positive @xmath0 as well as the inverted hierarchy , negative @xmath0 results while the dashed curve represents the inverted hierarchy , positive @xmath0 and the normal hierarchy , negative @xmath0 results .
thus non - zero @xmath0 breaks the four fold degeneracy down to two two - fold degeneracies .
hierarchy and linear in @xmath0 effects require interference between the two mass - squared difference terms , and this is the smallest value of @xmath3 where such effects appear .
these effects are greatest at the peak of the first oscillation generated by the small solar mass - squared differences . for vacuum oscillations
this occurs @xcite when @xmath30 km / gev . including matter effects yields @xmath31 km and @xmath32 gev as typical ideal values . extrapolating from the results in ref .
@xcite to the now known value of @xmath0 gives a large @xmath22 with a 20% difference between @xmath22 for @xmath2 and @xmath33 for such an experiment .
we present in fig .
[ fig2 ] the @xmath22 verses neutrino energy including the msw matter effects for t2k @xcite and minos
@xcite/ no@xmath5a @xcite .
the effect of including the matter interaction of the neutrino is to alter the magnitude of the oscillations while leaving the position of the peaks nearly unchanged .
thus spectral information that measures the position of the peaks maintains approximately the two - fold degeneracy of the vacuum oscillations .
the magnitude of the signal then breaks this symmetry to distinguish the hierarchy while it alone does not distinguish well the sign of @xmath0 .
there is a complementarity between the t2k and no@xmath5a experiments .
t2k is situated to determine the peak of the oscillation probability , ideal for breaking the four - fold degeneracy .
no@xmath5a has the longer baseline and the larger matter effects thus increasing the difference in the magnitude of the oscillations , giving it the greater sensitivity to the hierarchy .
.results for t2k , minos , and the combination . for the four cases of hierarchy and the sign of @xmath0 , we present the best fit values of @xmath34 , its 90% error bars , the maximum confidence level at which @xmath35 , and the probability that each case is the correct choice . [ cols="<,^,^,^,^,^,^ " , ]
we present our results for an analysis that includes all the experiments in fig .
[ fig5 ] where we show @xmath36 for the four cases of hierarchy and sign of @xmath0 .
the location of the minima are found to be only slightly affected by the choice of hierarchy or sign of @xmath0 . for the now known value of @xmath37 , atmospheric data distinguish between hierarchy and the sign of @xmath0 at a not negligible level .
the normal hierarchy , positive @xmath0 case is disfavored , while the inverse hierarchy , negative @xmath0 case is favored with the other two cases lying in the middle .
this is made quantitative in table [ tab3 ] . by marginalizing over @xmath38
we calculated the relative probability of each of the four cases .
the probability that normal hierarchy , positive @xmath0 is correct is 0.2% , a fairly strong indication that it is not correct . the inverse hierarchy , negative @xmath0 case is 90.5% probable that it is correct , a not negligible preference , while the normal hierarchy , negative @xmath0 and the inverse hierarchy , positive @xmath0 cases are 2.2% and 7.1% probable , respectively .
we compare our best fit value @xmath39 for negative hierarchy , negative @xmath0 to @xmath40 found for normal hierarchy , positive @xmath0 or negative hierarchy , negative @xmath0 from ref .
@xcite and to @xmath41 found for positive or negative hierarchy and negative @xmath0 for ref .
there is disagreement on the sign of @xmath0 and the hierarchy question .
we find a larger hierarchy difference than do the other two .
since the present t2k and minos experiments have little preference for the hierarchy and the sign of @xmath0 , the differences between the analyses arise from the atmospheric analysis .
these differences may be resolved by future super - k atmospheric data .
although the super - k data is statistically significant on the whole , the role of @xmath0 in the analysis is a small effect . for example
, the bins that indicate the presence or absence of the msw resonance , are statistically not too significant .
further , t2k and no@xmath5a have not produced data of statistical significance . with the large value for @xmath0 their future data
has great promise to help resolve this situation .
the work of d.j.e . is supported , in part , by us department of energy grant no .
de - fg02 - 96er40975 , the work of b.k.c .
is supported , in part , by us department of education grant no .
p200a090275 , and the work of j.e - r . is supported , in part , by conacyt , mexico .
10 latimer d c and ernst d j 2005 _ phys . rev .
_ d * 71 * 017301 abe y _
[ double - chooz collaboration ] 2012 _ phys .
lett . _ * 108 * 131801 an f p _ et al . _ [ daya bay collaboration ] 2012 _ phys .
lett . _ * 108 * 171803 ahn j _
[ reno collaboration ] 2012 _ phys .
lett . _ * 108 * 191802 escamilla - roa j , latimer d c and ernst d j 2010 _ phys .
_ c 82 028501 ; roa j e , latimer d c and ernst d j 2009 phys .
* 103 * 061804 barger v , marfatia d and wisnant k 2002 _ phys .
_ d * 65 * 073023 ; latimer d c and ernst d j 2005 _ mod .
lett . _ a * 20 * 1663 adamson p _
[ minos collaboration ] 2012 _ .
lett . _ * 108 * 191801 abe k _
[ t2k collaboration ] 2011 _ phys .
lett . _ * 107 * 041801 nowak [ no@xmath5a collaboration ] 2012 _ aip conf .
proc . _ * 1441 * 423 forereo d v , tortla m and valle j w f 2012 arxiv:1205.4018 [ hep - ph ] latimer d c and ernst d j 2005 _ phys .
_ c * 71 * 062501 latimer d c , escamilla j and ernst d j 2007 _ phs .
_ c * 76 * 05502 hagiwara k , okamura n and sendai k - k 2011 _ jhep _ * 1109 * 082 hunter h r , cogswell b k , escamilla - roa j , latimer d c and ernst d j , 2012 _ phys .
_ c * 85 * 068501 apollonio m _
_ 2003 _ eur .
_ c * 27 * 331 mueller th a _ et al .
_ 2011 _ phys .
_ c * 83 * 054615 hosaka j _ et al . _
[ super - k collaboration ] 2006 _ phys .
_ d * 74 * 032002 fogli g l , lisi e , marrone a , palazzo a and rotunno a , 2006 _ prog . part .
. phys _ * 57 * 742 fogli g , lisi e , marrone a , montanino d and palazzo a 2012 arxiv:1205;5254 [ hep - ph ] | the daya bay , reno , and double chooz experiments have discovered a large non - zero value for @xmath0 .
we present a global analysis that includes these three experiments , chooz , the super - k atmospheric data , and the @xmath1 t2k and minos experiments that are sensitive to the hierarchy and the sign of @xmath0 .
we report preliminary results in which we fix the mixing parameters other than @xmath0 to those from a recent global analysis .
given there is no evidence for a non - zero cp violation , we assume @xmath2 .
t2k and minos lie in a region of @xmath3 where there is a hierarchy degeneracy in the limit of @xmath4 and no matter interaction . for non - zero @xmath0 , the symmetry is partially broken , but a degeneracy under the simultaneous exchange of both hierarchy and the sign of @xmath0 remains .
matter effects break this symmetry such that the positions of the peaks in the oscillation probabilities maintain the two - fold symmetry , while the magnitude of the oscillations is sensitive to the hierarchy .
this renders t2k and no@xmath5a , with different baselines and different matter effects , better able in combination to distinguish the hierarchy and the sign of @xmath0 .
the present t2k and minos data do not distinguish between hierarchies or the sign of @xmath0 , but the large value of @xmath0 yields effects from atmospheric data that do .
we find for normal hierarchy , positive @xmath0 , @xmath6 and is 0.2% probable it is the correct combination ; for normal hierarchy , negative @xmath0 , @xmath7 and is 2.2% probable ; for inverse hierarchy , positive @xmath0 , @xmath8 and is 7.1% probable ; for inverse hierarchy , negative @xmath0 , @xmath9 and is 90.5% probable , results that are inconsistent with two similar analyses . |
the galactic interstellar medium ( ism ) is characterized by complex spatial distributions of density , temperature , and magnetic fields , as well as a turbulent velocity field that animates the whole system .
the relative proportions of ism gas in different thermal / ionization phases , and their respective dynamical states , may reflect many contributing physical processes of varying importance throughout the milky way ( or external galaxies ) . even considering just the galaxy s atomic gas component , observable in hi emission and absorption ,
a wide variety of temperatures and pervasive high - amplitude turbulence is inferred @xcite , and a number of different physical processes may collude or compete in establishing these conditions . in the traditional picture of the ism , turbulence in atomic gas
is primarily attributed to the lingering effects of supernova blast waves that sweep through the ism @xcite .
densities and temperatures of atomic gas are expected to lie preferentially near either the warm or cold stable thermal equilibria available given heating primarily by the photoelectric effect on small grains @xcite .
thermal instability ( ti ) is believed to play an important role in maintaining gas near the stable equilibria @xcite . certain potential difficulties with this picture
motivate an effort to explore effects not emphasized in the traditional model . in particular , because energetic stellar inputs are intermittent in space and time , while turbulence is directly or indirectly inferred to pervade the whole atomic ism , it is valuable to assess alternative spatially / temporally _ distributed _ turbulent driving mechanisms .
candidate mechanisms recently proposed for driving turbulence include both ti @xcite and the magnetorotational instability ( mri ) @xcite .
in addition to uncertainties about the source of turbulence in hi gas , other puzzles surrounding hi temperatures ( e.g. @xcite ) have grown more pressing with recent observations @xcite .
namely , the heiles and troland observations suggest that significant hi gas ( @xmath4 ) could be in the thermally - unstable temperature regime between 500 - 5000 k. using observational evidence from various tracers , @xcite has also recently argued that very large pressures and other large departures from dynamical and thermal equilibrium are common in the ism , and indicate rapid changes likely driven by turbulence . to assess and interpret this evidence theoretically
, it is necessary to understand the nonlinear development of ti , the effects of independent dynamical ism processes on ti , and the ability in general of magnetohydrodyanmic ( mhd ) turbulence to heat and cool ism gas via shocks , compressions , and rarefactions . in recent years
, direct numerical simulation has become an increasingly important tool in theoretical investigation of the ism s structure and dynamics , and has played a key role in promoting the increasingly popular notion of the ism as a `` phase continuum '' . in mhd ( or hydrodynamic ) simulations , the evolution of gas in the computational domain is formalized in terms of time - dependent flow equations with appropriate source terms to describe externally - imposed effects .
fully realistic computational ism models will ultimately require numerical simulations with a comprehensive array of physics inputs .
recent work towards this goal that address turbulent driving and temperature / density probability distribution functions ( pdfs ) include the three - dimensional ( 3d ) simulations of @xcite , @xcite , @xcite , and @xcite ; and the two - dimensional ( 2d ) simulations of @xcite , @xcite , @xcite , and @xcite . among other physics inputs ,
all of these simulations include modeled effects of star formation , with either supernova - like or stellar - like localized heating events that lead to expanding flows .
for some of these models , the cooling functions also permit ti in certain density regimes .
since many of the individual processes affecting the ism s structure and dynamics are not well understood , in addition to comprehensive physical modeling , it is also valuable to perform numerical simulations that focus more narrowly on a single process , or on a few processes that potentially may interact strongly .
this controlled approach can yield significant insight into the relative importance of multiple effects in complex systems such as the ism . using models that omit supernova and stellar energy inputs , it is possible to sort out , for example ,
whether the appearance of phase continua in density / temperature pdfs requires localized thermal energy inputs , or can develop simply from the disruption of ti by moderate - amplitude turbulence such as that driven by mri .
recent simulations that have focused on the nonlinear development of ti under ism conditions include @xcite , @xcite , @xcite , @xcite , @xcite , @xcite .
previous simulations of mri in 2d and 3d have focused primarily on the situation in which the density is relatively uniform , for application to accretion disks ( e.g. @xcite , @xcite , @xcite ) . in recent work , @xcite began study of mri in the galactic context using isothermal simulations , focusing on dense cloud formation due to the action of self - gravity on turbulently - compressed regions . in this work ,
we initiate a computational study aimed at understanding how density , temperature , velocity , and magnetic field distributions would develop in the diffuse ism in the absence of localized stellar energy input .
of particular interest is the interaction between ti and mri .
ti tends to produce a cloudy medium , and this cloudy medium may affect both the growth rate of mri and its dissipation rate , and hence the saturated - state turbulent amplitude that is determined by balancing these rates . on the other hand , the turbulence produced by mri may suppress and/or enhance ti by disrupting and/or initiating the growth of dense condensations .
evaluation of quasi - steady - state properties such as the mean turbulent velocity amplitude and the distribution of temperatures will await 3d simulations . in the present work ,
which employs 2d simulations , we focus on evaluation of our code s performance for studies of thermally bistable media , and on analysis of nonlinear development in models of pure ti , ti together with mri , and mri in a medium of pre - existing clouds . in 2 , we describe our numerical methods and code tests . in 3 , we present results from simulations of thermally unstable gas without magnetic fields , and in 4 we present results of models
in which magnetic fields and sheared rotation have been added so that mri occurs .
finally , in 5 , we summarize our results , discuss their implications , and make comparisons to previous work .
we integrate the time - dependent equations of magnetohydrodynamics using a version of the zeus-2d code @xcite .
zeus uses a time - explicit , operator - split , finite difference method for solving the mhd equations on a staggered mesh , capturing shocks via an artificial viscosity .
velocities and magnetic field vectors are face - centered , while energy and mass density are volume - centered .
zeus employs the ct and moc algorithms @xcite to maintain @xmath5 and ensure accurate propagation of alfvn waves . for the present study ,
we have implemented volumetric heating and cooling terms , and a thermal conduction term .
we also model the differential rotation of the background flow and the variation of the stellar / dark matter gravitational potential in the local limit with @xmath6 , where @xmath7 is the galactocentric radius of the center of our computational domain .
the equations we solve are therefore : @xmath8 @xmath9 @xmath10 @xmath11 all symbols have their usual meanings .
the net cooling per unit mass is given by @xmath12 .
we adopt the simple atomic ism heating and cooling prescriptions of @xcite , in which the cooling function , @xmath13 , is a piecewise power - law fit to the detailed models of @xcite .
the heating rate , @xmath14 , is taken to be constant at 0.015 @xmath15 . in the tidal potential term of equation ( [ mom ] )
, @xmath16 is the local dimensionless shear parameter , equal to unity for a flat rotation curve in which the angular velocity @xmath17 .
the present set of simulations is 2d , with the computational domain representing a square sector in the radial - vertical ( @xmath18 ) plane . in the local frame ,
the azimuthal direction @xmath19 becomes the @xmath20 coordinate axis ; @xmath21-velocities and magnetic field components are present in our models , but @xmath22 for all variables . to reduce diffusion from advection in the presence of background shear ,
we apply the velocity decomposition method of @xcite .
we employ periodic boundary conditions in the @xmath23-direction , and shearing - periodic boundary conditions in the @xmath24-direction @xcite .
this framework allows us to incorporate realistic galactic shear , while avoiding numerical artifacts associated with simpler boundary conditions . because cooling times can be very short , the energy equation update from the net cooling terms
is solved implicitly using newton - raphson iteration . at the start of each iteration
the time step is initially computed from the cfl condition using the sound speed , alfvn speed , and conduction parameter .
this is followed by a call to the cooling subroutine .
the change in temperature within each zone is limited to ten percent of its initial value .
if this requirement is not met for all cells in the grid , the time step is reduced by a factor of two , and the implicit energy update is recalculated .
tests with our cooling function show that this time step restriction could in principle become quite prohibitive if zones were far from thermal equilibrium . in practice , though , for our model simulations this is typically not the case , and the time step is reduced once or twice at most .
the update from the conduction operator is solved explicitly , using a simple five point stencil for the spatial second derivative of temperature ( cf .
@xcite equation 19.2.4 ) . in two dimensions
the cfl condition is @xmath25/ ( 4\mathcal{k})$ ] .
as @xcite have recently pointed out , the importance of incorporating conduction in simulations which contain thermally unstable gas has been occasionally overlooked in past work . without conduction , the growth rates for thermal instability are largest at the smallest scales , and unresolved growth at the grid scale may occur .
the inclusion of conduction , however , has a stabilizing effect on ti at small scales , and the conduction parameter can be adjusted to allow spatial resolution of ti on the computational grid . here , we treat @xmath26 as a parameter that may be freely specified for numerical efficacy ; we discuss the physical level of conduction in the ism below .
the zeus mhd code has undergone extensive numerical testing and has been used in a wide variety of astrophysical investigations . in addition , we have tested the code without cooling and conduction and have found it can accurately reproduce the linear growth rates of the mri for an adiabatic medium ( see also @xcite ) . to test our implementation of the heating , cooling , and conduction terms , we performed 1d simulations to compare with the linear growth rates of the thermal instability @xcite .
the models were initialized with eigenmodes of the instability , and three levels of conduction were chosen : @xmath27 .
for these tests , the grid was 128 zones and the box size @xmath28 pc .
the initial density and pressure were set to @xmath29 and @xmath30 , implying corresponding cutoffs for thermal instability ( `` field length '' ) , @xmath31 pc for our adopted cooling function ( see 3.1).^{-1/2}$ ] when @xmath32 function of @xmath33 .
] in figure [ fig1 ] we plot the growth rates from the simulations on top of the analytic curves .
the numerical growth rates are obtained by measuring the logarithmic rate of change of the maximum density .
the agreement between the analytic and numerical growth rates is quite good .
this test confirms that the newly added cooling and conduction subroutines are working correctly , and is critical in assessing the performance of the code as applied to multi - phase ism simulations .
note that at small scales ti is essentially isobaric , so that this test demonstrates the ability of the code to maintain near - uniform pressure via hydrodynamic flow to compensate for changes in temperature driven by the cooling term ( see eq .
3 ) . by comparison with simulations in which we set @xmath34
, tests with non - zero @xmath35 also confirm that conduction provides a needed numerical stabilization .
when conduction is omitted , growth rates for ti will always be greatest at the largest available wavenumbers .
our 2d tests with @xmath34 have confirmed this is indeed the case : simulations in which ti is seeded from random perturbations form high density clouds which are the size of a single grid zone .
further 2d tests show that provided @xmath36 is resolved by at least 8 zones , this grid - scale growth is suppressed . for the models we shall present ,
the conduction parameter and grid resolution were chosen such that we can adequately resolve all modes for which ti is unstable . because we are modeling a medium containing very large density contrasts
, it is desirable to assess the evolution of contact discontinuities between the high and low density regions , representing cold and warm phases in pressure equilibrium .
the diffusive smearing of contact discontinuities is an inherent limitation of all finite difference codes , but the numerical problem can be magnified with the inclusion of a thermally bistable net cooling function . as these contact discontinuities are advected through the grid , upstream and downstream zones adjacent to the discontinuity are set to intermediate densities which may be thermally unstable .
thermally unstable gas adjoining the initial contact rapidly heats or cools to reach a pressure different from the initial equilibrium , and this can potentially introduce additional dynamics to the problem . to explore this numerical issue for the problem at hand
, we have performed 1d advection tests of relaxed profiles of high density clouds in a low density ambient medium .
the resolution is 512 zones for all runs .
we define @xmath37 , i.e. the number of zones in a field length at the mean density , giving a measure of the resolution at scales for which conduction is important .
we vary @xmath26 so that @xmath38 varies from 8 to 32 in powers of 2 .
the initial conditions consist of a top - hat function of high and low density set to be in approximate pressure equilibrium .
the exact equilibrium state is the solution of @xmath39 from equation ( [ energy ] ) , and is different for each value of @xmath35 .
initial oscillations in pressure , velocity , and density gradually decay , and we consider the profile to be relaxed when these oscillations reach approximately one percent of their value early in the simulation . after a relaxed `` cloud '' profile
is achieved the velocity of all cells is set to a constant value comparable to the sound speed , and the `` cloud '' is advected through the grid twice .
profiles at the end of these runs are compared to the initial relaxed equilibrium profile in figure [ fig2 ] .
notice that as @xmath40 increases so does the number of zones over which the contact discontinuity is spread ; the results clearly show that the profile is preserved more faithfully as @xmath40 increases from 8 to 32 . for comparison
we also show results from the original zeus-2d code , without conduction and cooling .
it is clear that running higher levels of conduction at a given resolution has the advantage of smearing contact discontinuities over an increasing number of zones , thus improving the performance of the code in the advection tests .
however , increasing @xmath26 has the disadvantage of inhibiting thermal instability at larger and larger spatial scales , such that only very large scale structures can develop from thermal instability .
as we are interested in how wavelengths of growing mri modes in a cloudy medium may be affected by the distances between condensations , it is undesirable to limit the available dynamic range for this exploration . for @xmath38 = 32 at a resolution of 256 zones , @xmath41 is 12.5pc , and the maximum ti growth rate occurs at 29 pc , about one third of our box .
a second disadvantage of increasing @xmath26 is that conduction quickly begins to set the time step for the simulations .
so , we make the practical choice of setting @xmath42 . at resolution of @xmath43 , and a box size of 100 pc , we then set @xmath44 , to yield @xmath45 pc .
this compromise choice allows us to resolve ti developing from the initial conditions , to maintain advection profiles within @xmath46 , and to have adequate numbers and resolution of the condensations that form to represent a cloudy medium in a meaningful way .
the true thermal conduction level in the ism is not well known observationally , and must be affected by the fractional ionization ( since electrons are highly mobile when present ) and magnetic field geometry .
a minimum level of conduction for the atomic ism is that of neutral hydrogen gas , @xmath47 @xcite . at 2000
k , @xmath48 , about a factor of one hundred smaller than our value , such that @xmath49 would be reduced by a factor @xmath50 .
the use of a smaller conduction parameter would not , however , significantly alter our main results . the most unstable wavelength for ti ( see figure [ fig1 ] )
would be significantly smaller , 3 pc compared to 12 pc for our adopted @xmath26 which would reduce the size scale of the clouds in the initial condensation phase .
however , the growth rate of ti for the most unstable wavelength would increase by only 15% compared to our models .
the evolution towards more massive clouds via agglomeration would proceed similarly to the results we have found .
in addition , the overall tendency to maintain a two - phase medium after ti has developed , as well as the characteristics we identify for mri growth in a cloudy medium , would not be affected by the initial sizes of clouds that form .
various heating and cooling processes in the ism define a thermal equilibrium pressure - density curve on which energy is radiated at the same rate it is absorbed .
perturbations from the equilibrium curve will either be stable or unstable depending on its local shape @xcite .
our adopted equilibrium curve @xcite is shown in figure [ fig3 ] ( together with contours of temperature corresponding to transitions in the cooling function , and with scatter plots from our first simulation ) .
gas in the region above the curve has net cooling ( @xmath51 ) , and gas below the curve net heating ( @xmath52 ) .
when gas on the equilibrium curve in the warm phase ( phase `` f '' , at t @xmath53 6102 k ) is perturbed to higher ( lower ) temperatures , there is net cooling ( heating ) and the gas returns to equilibrium .
the same situation applies to gas in the cold phase ( phase `` h '' , at t @xmath54 141 k ) . at intermediate temperatures ( phase ``
g '' , 313 k @xmath54 t @xmath54 6102 k ) , however , perturbations from equilibrium to higher ( lower ) temperatures results in net heating ( cooling ) , and the gas continues heating ( cooling ) until it reaches equilibrium in phase f ( h ) .
this is the physical basis for ti , which was first analyzed comprehensively by @xcite .
thermal instability has long been believed to play an important role in structuring the ism because at typical volume - averaged atomic densities estimated for the ism ( e.g. @xmath55 from @xcite ) gas in thermal equilibrium would lie on the unstable portion of the curve .
thus , much of the mass of the ism in the milky way and similar galaxies is believed to be in cold clouds or a warm intercloud medium , the two stable neutral atomic phases ( e.g. @xcite ) .
recent theoretical work has emphasized that dynamical processes may drive gas away from these two stable phases , potentially explaining observationally - inferred temperatures that depart from equilibrium expectations ; we shall discuss these issues in
[ discuss ] .
it is , nevertheless , of significant interest to study in detail how ti develops nonlinearly to establish a two - phase cloudy medium in the absence of other potential effects such as localized heating , impacts from large - scale shocks , or stresses associated with mhd turbulence .
results of these carefully - controlled `` pure ti '' simulations are valuable for characterizing the timescale to develop a two - phase medium as well as its structural and kinetic properties .
`` pure ti '' models also represent a baseline for comparison of models incorporating more complex physics .
once the particular form of the cooling curve has been chosen , the development of ti is primarily a function of four parameters : the cooling time , @xmath56 , the heating time , @xmath57 , the sound crossing time , @xmath58 , and the conduction length scale , @xmath49 .
the cooling time depends on the specific cooling function ; for the cooling curve we adopt @xcite , in varying temperature regimes we have @xmath59 the heating time , @xmath60 for @xmath61 .
we can also define an effective cooling / heating time as @xmath62 ; at thermal equilibria , @xmath63 becomes infinite .
the sound crossing time over distance @xmath64 is @xmath65 in figure [ times ] , we plot @xmath63 , and @xmath58 for @xmath66 as a function of density for @xmath67 . for a range of densities @xmath68 , the sound crossing
time is significantly shorter than the effective cooling time ; otherwise @xmath58 is typically more than an order of magnitude longer than @xmath63 . to simulate thermal instability under conditions representative of the general diffuse ism
, we adopt an initial density of @xmath69 and set the initial pressure to @xmath3 ; all velocities are initially set to zero . random pressure perturbations with an amplitude less than 0.1% are added to seed the ti .
the gas is initially in a weakly cooling state , since the equilibrium pressure corresponding to @xmath70 is @xmath71 .
the simulation proceeds for 470 myr , corresponding to two galactocentric orbits at the solar circle . the initial development of ti proceeds quickly . in figure [ fig3 ]
we show four snapshots at different times of the density distribution alongside scatter plots of pressure versus density overlayed on the equilibrium cooling curve .
structure begins to form at about @xmath72 , and density contrasts continue to increase at constant pressure until @xmath73 as shown in figure [ fig3 ] .
the fourier transform of the density distribution at this time , figure [ den_pow ] , shows that the majority of the power is concentrated at @xmath74 . allowing for geometrical factors ,
this is consistent with the one - dimensional linear theory prediction that the most unstable wavelength is @xmath75 for our chosen value of conduction .
a network of filaments briefly forms at about 18 myr connecting the regions of highest density , with flow moving along the filaments increasing the number density to as high as @xmath76 . by 23
myr the dense gas has collected in condensations and has relaxed to approximate thermal equilibrium .
the size scale of the cold clumps is initially only a few parsecs and is relatively uniform .
however , random motions of the newly formed clouds leads to merging and disassociation ; conductive evaporation at its boundaries can also alter the shape of a clump . by the end of the simulation , at 474 myr ,
the typical cloud size has increased significantly to @xmath77 , though a number of smaller clouds either remain from the initial ti development or have formed as a result of disassociation . in figure [ figxa ] ,
panels a , b , and c , we plot the volume - weighted and mass - weighted density probability distribution functions ( pdfs ) for the first three snapshots from figure [ fig3 ] . at t=14 myr , in panel a , all of the gas is in the unstable range .
already , though , in panel b , after 18 myr , the distribution has begun to separate into two distinct phases . at t=18 myr ,
69% , 28% , and 3% of the gas by volume is in the f ( warm ) , g ( unstable ) , and h ( cold ) phases respectively , which are defined by our cooling curve as the ranges @xmath78 , @xmath79 , and @xmath80 , respectively . by mass these proportions
are reversed to 29% , 23% , and 49% , respectively . from about 100 myr through the end of the simulation ,
the distribution of cloud sizes evolves , but the pdf remains relatively unchanged , with 12% , 2% , and 86% of the mass residing in the f , g , and h phases .
we also performed the same simulation , but increased the resolution to @xmath81 .
the conduction coefficient is not changed , so @xmath82 .
we find similar results overall .
in particular , the mass - weighted density pdfs are compared in panel d of figure [ figxa ] . these pdfs exhibit no significant differences , confirming the robustness of our results . alongside the images of density in figure [ fig3 ] we show scatter plots of @xmath83 against @xmath84 for all zones in the grid at the same times .
in the initial state , pressure is constant , and @xmath58 at 5 pc ( @xmath85 the length of the fastest - growing ti mode ) is shorter than @xmath56 ( see figure [ times ] ) . towards the low density warm phase , @xmath86 , and gas parcels in regions undergoing rarefaction
are able to heat nearly isobarically .
thus , all zones at densities lower than the mean are filled with an intercloud medium that maintains spatially nearly uniform pressure . for gas parcels undergoing compression and net cooling , as @xmath87 becomes large , @xmath88 , so the gas tends to cool towards the thermal equilibrium curve at a faster rate than the flow is able to readjust dynamically .
after gas parcels reach near thermal equilibrium in the cold phase , they continue to be compressed until pressure equilibrium with the warm medium is re - established . over time , the average pressure in the simulation box decreases due to radiation from the cold phase . because the scatter plots in figure [ fig3 ] contain a large number of points , many fall in the unstable range , although the actual amount of material there is small . to quantify this , in figure [ figxb ]
we plot mass - weighted and volume - weighted temperature pdfs at times corresponding to the snapshots in figure [ fig3 ] . as with the density pdf , the distribution is separates into two distinct phases in panel b at 18 myr , and remains so for the duration of the simulation .
thermal instability is a dynamic process , and a number of recent works have proposed that ti may help contribute to exciting turbulence in the ism . in figure [ fig3a ]
we plot the mass - weighted velocity dispersions for gas in the f , g , and h phases separately .
for all phases , the largest velocities occur during the condensation stage at early times ( 10 - 20 myr ) , corresponding to about 5 times the @xmath89-folding time of the dominant linear instability .
the peak velocity is about 0.45 @xmath90 for the unstable ( g ) phase , and is @xmath91 for the two stable phases . at later times
the velocity dispersion in each phase remains relatively constant , with the largest value ( 0.35 @xmath90 ) for phase g , next largest ( 0.25 @xmath92 ) for the warm phase , f , and smallest ( 0.15 @xmath93 ) for the cold phase , h. the standard deviation of the total velocity dispersion is typically 0.15 @xmath94 .
for comparison , typical sound speeds of the f , g , and h phases are 7 - 8 @xmath92 , 1 - 5 @xmath92 and 0.6 - 1 @xmath95 , respectively .
thus , the mean turbulent velocities are all subsonic .
in a series of four papers , balbus & hawley presented the first linear analysis and numerical simulations of mri in the context of an astrophysical disk @xcite .
the physical basis for the instability is relatively simple , and there are two requirements for the instability to be present : a differentially rotating system with decreasing angular velocity as one moves outward through the disk , and a weak magnetic field ( strong magnetic fields have a stabilizing effect ) .
as fluid elements are displaced outward ( inward ) the magnetic field resists shear and tries to keep the fluid moving at its original velocity . due to these magnetic stresses ,
fluid elements gain ( lose ) angular momentum , the centrifugal force becomes too large ( small ) to maintain equilibrium at the new position , and the fluid element moves farther outward ( inward ) .
this leads to the transport of angular momentum outward through the disk . for a complete linear analysis of the mri in 2d
we refer the reader to @xcite . here
we simply summarize the important formulae for axisymmetric modes with wavenumber @xmath96 and @xmath97 .
the growth rates are given by @xmath98^{1/2 } ) } \right]\ ] ] where @xmath99 in terms of the alfvn speed @xmath100 .
the maximum growth rate occurs when @xmath101 i.e. @xmath102 for @xmath103 ; here the growth rate is @xmath104 .
the highest wavenumber for which axisymmetric mri exists when @xmath105 is @xmath106 we have tested the code without cooling and conduction , and found that it can accurately reproduce the predicted linear growth rates of the mri .
we do not detail the results here ; instead we refer the reader to @xcite for a complete analysis of similar models . based on the linear dispersion relation , for sufficiently weak magnetic fields , modes with a range of @xmath107 ( and also @xmath108 ) may grow .
the smallest permissible wavenumber for a simulation is @xmath109 , where @xmath110 is the vertical dimension of the computational box . at late times in 2d axisymmetric simulations @xcite , mri
becomes dominated by a `` channel '' solution corresponding to the smallest permissible vertical wavenumber , i.e. with flow moving towards the inner regions of the disk on one ( vertical ) half of the grid , and flow moving towards the outer regions in the other ( vertical ) half of the grid .
this pure `` channel flow '' is unphysical ; for a 3d system it is subject to nonaxisymmetric parasitic instabilities @xcite . in 3d
non - axisymmetric simulations ( e.g. @xcite ) the channel solution forms at early times , but later develops into a fully turbulent flow . the mri has primarily been studied in the context of accretion disks , but can be important in any differentially rotating disk system provided the magnetic fields are not too strong .
in addition to axisymmetric modes , nonaxisymmetric mri modes can also grow directly ( see e.g. @xcite and @xcite eqs .
80 , 81 for instantaneous growth rates and instability threshold criteria in various limits ) .
the axisymmetric mode with wavelength @xmath111 is the most difficult to stabilize as @xmath112 increases ; from equation ( [ maxkmri ] ) , taking @xmath113 , @xmath103 , a disk scale height @xmath114=150 pc and a uniform density @xmath115 , mri will be present provided @xmath116 . for the milky way ,
this is consistent with the observed solar - neighborhood estimate @xmath117 @xcite .
if a multi - phase system behaves similarly to the corresponding uniform - density medium , then we may expect mri to be important in the galactic ism . here , we explore how mri development can be affected by strong non - uniformity in the density structure . to study nonlinear development of the mri in a nonuniform medium , we first perform a simulation identical to the ti model run described in [ ti ] , but now include magnetic fields and sheared rotation .
all hydrodynamical variables are initialized as described in [ ti ] .
the magnetic field is vertical with @xmath118 .
the rotation rate is set to 26 km @xmath119 representative of the local value near the sun , and we set the shear parameter @xmath120 to describe a flat rotation curve . with these parameters , from equation ( [ maxkmri ] ) ,
the smallest - scale uniform - density mri mode that would fit within our @xmath121 box has @xmath122 ( in units of @xmath123 ) .
on the left in figure [ fig3sf ] we show snapshots of number density overlayed with magnetic field lines at three representative times , and on the right we show the corresponding mass - weighted density pdf .
the time - scale for development of the mri is much longer than that of ti , so that the initial development is essentially the same as in the purely hydrodynamical case . during the ti condensation phase
the magnetic field becomes kinked as the filaments condense into small clouds .
the remaining random motions of the clouds leads to further distortion of the magnetic field , as can been seen at 237 myr .
the channel solution has clearly taken hold by 474 myr , and the @xmath124 mode ( in units of @xmath123 ) dominates . similarly to our analysis of kinetic evolution for the ti model , in figure [ fig3sb ] we plot the velocity dispersion for the f , g , and h phases as a function of time in the ti + mri model .
initially these are similar to the hydrodynamical case , with all velocities less than @xmath125 .
as the channel solution develops the velocity dispersion begins to increase at about 500 myr , and peaks at the end of the simulation in phase f at approximately @xmath126 .
the peak velocity dispersion is about @xmath127 for phase g , and 0.9 @xmath90 ( approximately the sound speed ) for phase h , all towards the end of the simulation . as in the hydrodynamical model , the pdfs for the ti + mri model are clearly two - phase , with small amounts of gas contained in the unstable regime . at very late times
the fully developed channel solution tends to increase the proportion of unstable gas . in figure [ mri_temper ]
we plot the mass - weighted temperature pdf of the ti + mri model 800 myr , and the same quantity for the ti run at 474 myr .
we do not expect that the pdf for the ti run would evolve significantly if the simulation had been continued to 800 myr .
evidently , the dynamical flows induced by the mri can significantly affect the temperature distribution .
the larger velocity dispersion and kinked magnetic fields due to the channel solution can compress portions of cold clouds , decreasing the temperature correspondingly .
although still dominated by distinct warm and cold phases , there is also a higher proportion of gas in the unstable regime .
for the same model snapshot , figure [ mri_scatter ] shows a @xmath128 vs. @xmath83 scatter plot , overlayed on the equilibrium cooling curve . in the high density regime ,
cooling times are short , and the gas is not far from equilibrium . at low densities cooling times are longer , and gas can be found out of thermal equilibrium .
since the 2d channel solution would break up in a real 3d disk due to parasitic instabilities @xcite , we do not expect the late - time effects seen in our models to have direct implications for the temperature distribution in the diffuse ism .
they illustrate , however , the more generic point that spatially - varying rotational shear coupled to magnetic fields can create stresses that force gas away from the stable equilibrium phases .
we shall discuss this further , highlighting differences that might be expected for 3d mri , in
[ discuss ] .
important questions for assessing mri development in a cloudy medium are how the spatial- and time - scales of the fastest - growing modes differ compared to those in single - phase counterpart systems .
we measure mri mode amplitudes in the simulations by taking the fourier transform of @xmath129 as a function of time , from which we can calculate the growth rates .
the @xmath130 and 3 mode amplitudes are plotted in figure [ fig_ti_mhd_mode ] .
there is not an obvious linear stage from which we can measure the growth rate , but between 284 and 470 myr the average @xmath131 growth rate is about 0.28 @xmath132 , compared to the predicted rate of 0.34 @xmath132 at the average density of the model . at the average density linear theory predicts that the most unstable mode is at @xmath133 , with a growth rate of 0.50 @xmath132 , and the @xmath134 mode is unstable as well , with a predicted growth rate of 0.41 @xmath132 . between 284 and 470 myr we measure a mean growth rate for the @xmath133 and @xmath134 modes to be 0.12 @xmath132 and 0.18 @xmath132 , respectively . at the density of the warm medium , only the @xmath131 mode is predicted to be unstable in our simulations , with a growth rate of 0.45 @xmath132 .
thus , growth rates of available modes appear somewhat lower than they would be for a medium at either the mean density or the density of the warm medium . in addition , initial growth does not show the clear dominance of a single fastest growing mode that is evident in comparison adiabatic test simulations for a single - phase medium .
however , at late times , the @xmath131 mode grows to exceed the other low - order modes , similarly to the findings of @xcite for a single - phase medium . because the initial mri growth in the previous model may be strongly affected by lingering dynamical effects of ti , it is of interest to consider mri development in a medium which contains two distinct phases from the outset . in our next simulation
we therefore begin with a two phase medium in approximate equilibrium , rather than developing a two phase medium from thermally unstable gas .
we embed 59 high density clouds in a low density ambient medium such that the average density is the same as that of previous ti simulations . to set up the initial conditions allowing for conduction at cloud / intercloud interfaces , we first create `` template '' cloud profiles by embedding a single high density cold cloud in a low density warm medium and evolving until a thermally- and dynamically - relaxed state
is reached .
this profile ( density , pressure and velocity ) is then copied to randomly chosen locations on the grid , with the condition that cloud centers must be at least 20 zones apart .
we initialize the magnetic field after this `` cloud embedding '' procedure .
the simulation is then evolved as the mri develops . in figure [ gmc_ave ]
we show three snapshots from the `` cloud + mri '' simulation , along with mass - weighted density pdfs .
the mri takes about 500 myr until its development begins to become apparent , as can be seen in the poloidal field lines at 464 myr in figure [ gmc_ave ] . at 701 myr
many of the clouds have merged and have significant velocities as the mri channel solution begins to take hold .
initially the mass weighted density pdf shows almost no gas in the unstable range , but as the mri begins to develop , this phase begins to become populated , and the pdf becomes very similar to those from the ti + mri simulations .
perhaps the most interesting results from this `` cloud + mri '' simulation are the behavior of the mode amplitudes and growth rates . initially the clouds contain negligible velocities , and are in what would be a steady state if magnetic fields were not present .
we might expect , then , to find `` cleaner '' mri growth rates than for the ti + mri runs .
the mode amplitudes for @xmath130 , and 3 are plotted in figure [ gmc_ave_mode ] .
initially the @xmath133 mode is dominant and shows approximate linear growth between 300 myr - 700 myr , with an average growth rate of 0.34 @xmath132 , compared to the predicted rate 0.50 @xmath132 at the average density .
the @xmath131 mode also shows approximate linear growth with an average rate of 0.47 @xmath132 measured from 450 myr - 900 myr , and becomes the dominant mode at around 800 myr . at the density of the warm medium
the theoretical growth rate of the @xmath131 mode is 0.45 @xmath132 , and at the average density it is 0.34 @xmath132 .
the @xmath134 growth rate , measured between 400 myr and 800 myr is 0.33 @xmath132 , which we can compare to the predicted value at the average density of 0.41 @xmath132 .
thus , similarly to the ti + mri model , growth rates are slightly lower than they would be in a medium with the same mean density .
there is , however , a longer period of dominance by the mode with the fastest expected growth rate . for comparison we also performed an mri simulation with cooling and conduction disabled , initially at uniform density and seeded with the velocity and pressure profiles from the initial state of the previous `` cloud + mri '' simulation ; we also performed zero - conduction , zero - cooling test simulations seeded with random perturbations .
the growth rates are in reasonable agreement between these simulations , although the case directly seeded with random perturbations had cleaner growth of mode amplitudes .
most importantly , we find that rather than @xmath133 predominating as we found in the `` cloud + mri '' simulation , the @xmath134 mode in the single - phase comparison model is dominant until late time .
this suggests that the initial perturbations are strongest at @xmath134 , but the presence of the multi - phase medium in the `` cloud + mri '' model significantly inhibits the growth of this mode because the inertial load varies strongly over a wavelength for @xmath134 .
there are a number of ways mri growth rates and preferred scales could be affected by the influence of a cloudy background density structure , and signs of these effects are evident in our simulations .
first , both growth rates and preferred scales are dependent on the alfvn speed , which is a function of density .
the cold , dense clouds will be mri unstable at small scales compared to the larger preferred scales of the warm , low density , ambient medium . for fixed magnetic field strength , mri wavelengths are inversely proportional to density .
thus , mri wavelengths in cold , dense gas will be a few parsecs , while mri wavelengths in warm , diffuse gas will be several tens of parsecs . although the mri wavelengths in dense gas may be smaller than individual cold clouds , permitting initial rapid growth , further development of the small - scale instability is limited by clouds small radial extent .
long term mri development must therefore have characteristic wavelengths representative of either the average density conditions or the pervasive low - density warm , intercloud gas .
this expectation is indeed consistent with our results .
as seen in figures [ fig3sf ] and [ gmc_ave ] , both the diffuse gas and the cold clouds participate together in an overall large - scale flow .
the cold clouds frequently are the sites of strong kinks in the magnetic field .
one might also expect the preferred scales for mri to be affected by cloud spacing .
if cloud spacing is small compared to a given wavelength , then the mri growth rate might be expected to be similar to that under average density conditions .
thus , if the fastest - growing wavelength at the mean density is large compared to cloud spacing along a field line then this might be expected to be the dominant wavelength .
this situation is indeed evident in the second frame ( 464 myr ) of figure [ gmc_ave ] , in which the @xmath133 mode dominates , even though , as described above , the input perturbation spectrum is such that the @xmath134 mode would dominate if the density were uniform . on the other hand , if fewer , more massive clouds exist , cloud separations will be larger , which could suppress mri growth at wavelengths shorter than cloud spacings and encourage mri development on the largest scales .
evidence of this effect can be seen in the third panel ( 474 myr ) of figure [ fig3sf ] .
finally , if total mass is distributed very unevenly with magnetic flux , then mri may develop more rapidly and at longer wavelengths in regions where there is a comparatively low inertial load . in simulations ( not shown ) we have performed which have alternating radial zones of high and low mass loading on field lines ( initiated with cold clouds at intersections in a cartesian grid ) , we indeed see this effect .
development of mri in the low - inertia `` pure warm '' phase is , however , checked when the radially - moving flow collides with the high - inertia cold clouds . to test
whether the growth rate of the smaller - scale k=2 mode ( essentially near the lower wavelength limit for behavior as at a single average density ) could be enhanced in ti+mri models , we also performed an additional random ti simulations in which a @xmath133 perturbation was added shortly after the initial condensation phase .
the perturbation was added at about the 20 percent level in @xmath135 .
for the first 500 myr the @xmath133 mode is strongest with a growth rate of 0.36 @xmath132 .
after 500 myr the k=1 mode , with a growth rate of 0.47 @xmath132 , is again dominant . taken together , the simulations of this section show that the development of the mri in the presence of a cloudy medium has modest differences compared to the corresponding development in a single phase medium at the average density .
the dominant wavelengths are similar to those predicted by linear theory at mean - density conditions , and growth rates are also similar , but slightly smaller .
the spacings between clouds affects which among the low-@xmath136 modes dominates the power during the exponential - growth phase . at very late times
the @xmath124 mode is dominant in all simulations , consistent with the ultimate dominance of this channel flow in the single - phase models of @xcite .
thermal and magnetorotational instabilities may play a major role in determining the physical properties of the diffuse ism . in regions far from active star formation or a recent supernova explosion , ti and mri may even be the primary processes driving structure and dynamics in the ism on scales @xmath137 pc .
the pdfs of gas density and temperature , the characteristic sizes , shapes , and spatial distributions of cloudy structures , and the amplitudes and spectral properties of turbulent velocities and magnetic fields may all be strongly influenced by ti and mri .
in addition , development and saturation of ti and mri may be strongly interdependent . in this paper
, we have initiated a study of these important processes using numerical mhd simulations .
the current work focuses on code tests and 2d models using a microphysics implementation appropriate for the atomic ism .
in addition to characterizing the properties of ti and mri modes in their nonlinear stages , this study lays the groundwork for future 3d simulations which will be used to investigate quasi - steady turbulence .
_ numerical methods : _ we have implemented atomic - ism heating / cooling and thermal - conduction source terms in the energy equation of the zeus code ( using implicit and explicit updates , respectively ) . for conditions representing the mean pressure and density in the ism
, we find excellent numerical agreement with the analytic growth rates of thermally - unstable modes for a large range of wavelengths and thermal conductivity coefficients .
based on these tests and confirmation of acceptable results for advection of high - contrast contact discontinuities ( warm / cold pressure equilibrium interfaces ) on the grid , we adopt a value of @xmath138 such that the field length is resolved by 8 ( 16 ) zones in ( 100 pc)@xmath139 simulations with 256@xmath139 ( 512@xmath139 ) cells
. explicit inclusion of conduction is important for suppressing numerically - unresolved ti - driven amplification of grid - scale noise ; @xcite have also recently highlighted the importance of implementing conduction for simulations of thermally bistable media .
in some previous simulations of ti @xcite under stongly cooling conditions , conduction was not included ; since those simulations began with relatively large - amplitude ( 5% ) perturbations on resolved scales , however , sub - dominant effects from unresolved growth at grid scales in the initial stages of ti would be less noticeable .
in other recent work @xcite simulations of ti using spectral algorithms ( with explicit diffusive terms in the equations of motions ) appear to have difficulty reproducing the analytic growth rates in some circumstances .
conceivably , this may be a sign of numerical diffusion that could tend to produce more gas in thermally - unstable regimes than is realistic , in simulations using these computational methods .
_ nonlinear development of ti : _ in `` pure ti '' simulations where we initialize gas at @xmath140 and @xmath70 in a ( 100 pc)@xmath139 box with 0.1% initial pressure perturbations , we find that ti develops at a characteristic length scale consistent with the predicted fastest - growing mode , @xmath141 pc for our adopted value of @xmath142 .
as seen in other 2d simulations ( e.g. @xcite ) , the structure initially resembles a `` honeycomb '' network of cells , and as nonlinear development proceeds , gas condenses into cold , compact clouds at the intersections of filaments .
gas undergoing rarefaction towards the warm phase heats nearly isobarically , because the sound crossing time is short compared to the net heating - cooling time .
gas undergoing compression towards the cold phase initially has isobaric evolution ( while density perturbations remain low - amplitude ) , but then tends first to cool toward the equilibrium curve very rapidly ( with an attendant pressure drop ) , and then dynamically readjusts its density and temperature until the pressure again matches ambient conditions . the time to establish a distinct two - phase structure of well - separated cold clouds within a warm ambient medium ( see third panel of fig .
4 ) is @xmath2 myr , or about 10 @xmath89-folding times in terms of the linear growth rate . in the subsequent evolution ,
the cold , dense clouds undergo successive mergers to produce larger structures .
the transition from nearly isobaric to more `` isochoric''-like evolution for cold gas during nonlinear stages of condensation was recently emphasized by @xcite , and snapshots of phase diagrams in @xcite show a similar dip in pressure for overdense gas as it cools toward thermal equilibrium .
@xcite found , similar to our results , that initial perturbations of similar or larger sizes to our dominant ti wavelength require times @xmath143 myr to complete the condensation process , even when a much larger ( 10% ) initial perturbations are used .
the real level of conduction in the atomic ism may be lower than the value we adopted ( for numerical efficacy ) , with the fastest - growing ti wavelength a factor @xmath144 smaller than our 12 pc value and the condensation time correspondingly shorter ; @xcite found that 3 pc - scale overdensities condense into clouds within 4 myr .
as the turbulent cascade is likely to maintain nonlinear - amplitude entropy perturbations down to sub - pc scales , we expect that the fastest - growing wavelength and the product of the sound speed and the cooling time .
] is likely to dominate when ti occurs under `` natural '' circumstances , with later mergers producing larger clouds ( see also @xcite ) .
_ gas phase distributions from ti : _ the bimodal density and temperature pdfs in our ti simulations mirror the distinct two - phase structure evident in late - time snapshots .
typical late - time warm- , cold- , and intermediate - temperature mass fractions are 12 , 86 , and 2% . for a two - phase medium with mean density @xmath145 and cold and warm densities @xmath146 and @xmath147 ,
the fraction of mass in the cold medium is @xmath148 .
provided @xmath149 , the mass fraction in the warm medium is thus @xmath150 . since the pressure at late stages of our evolution has dropped near the minimum value of @xmath84 at which two phases are present , and in thermal equilibrium at this pressure @xmath151 ( with @xmath152 ) , the relative proportions of gas @xmath153 in the cold and warm phases are just as expected ( with @xmath154 ) . findings on density and temperature pdfs from other recent ti simulations are varied . from the 3d simulations of @xcite , the late - time ( 1.5 myr )
mass fractions are @xmath155 , @xmath156 in the stable phases and @xmath157 in the intermediate , unstable regime .
kritsuk and norman use a somewhat different cooling curve from ours , with @xmath158 in thermal equilibrium at the minimum pressure at which two stable phases are available .
since they use @xmath159 , their result that @xmath150 is consistent with expectations for a two - phase medium , while the gas at intermediate temperatures appears to be due to mass exchange with the cold medium ( see their discussion ) . in the 1d simulations of @xcite ( and using the same cooling curve and mean density as ours ) , only a few percent of the gas in their `` multiple condensation '' runs remains at intermediate densities , similar to our results , and their @xmath160 at @xmath161 myr is similar to our results at comparable ( early ) times . in the 2d simulations of @xcite that also include `` stellar - like '' local heat sources , the late - time mass fractions are @xmath162 , @xmath163 @xmath164
it is not clear to what extent this large proportion of gas at intermediate temperatures is sustained by turbulence ( via adiabatic expansion / compression and/or shocks heating or cooling gas that would otherwise be in the warm or cold stable phases ) , versus being maintained by the localized heating turned on when @xmath165 . with 3d mri simulations in which turbulent
driving is `` cold '' , it will be possible to address this important issue .
_ turbulent driving by ti : _ we find that turbulence produced by `` pure ti '' has only modest amplitudes , when initiated from `` average ism '' pressure and density conditions . for the warm , unstable , and cold phases , respectively , we find typical mass - weighted velocity dispersions of @xmath166 , @xmath167 , and @xmath168 @xmath169 .
these velocities are all quite subsonic . in simulations starting from thermal equilibrium ,
@xcite similarly find subsonic turbulence ( @xmath170 at @xmath171 2myr ) , although when gas is initially very hot , supersonic turbulence can be produced .
when they include repeated episodes of strong uv heating @xcite find mach number variations @xmath172 between `` low '' and `` high '' states ; since their `` low state '' is dominated by cold gas with @xmath173 , this is consistent with our results for typical turbulent amplitudes . in the simulations of @xcite in which warm gas shocks on impact with a low - density , hot ( @xmath174 k ) layer , ti develops near the interface of shocked gas with the hot medium , leading to the formation of cold cloudlets with velocity dispersions of a few @xmath169 .
although koyama and inutsuka attribute this turbulence to the effects of ti , it is possible that other dynamical instabilities associated with the hot / warm interface contribute in driving these motions .
. _ nonlinear development of axisymmetric mri : _ we have studied the development of axisymmetric mri under atomic ism conditions , both with `` ti+mri '' models starting from uniform density and pressure ( @xmath175 and @xmath70 ) , and with `` cloud+mri '' models that are initiated with the same uniform pressure and total mass , but start with a population of cold clouds embedded in a warm ambient medium . the magnetic field in both types of models is vertical and initially uniform , with @xmath176 .
the peak growth rate of mri ( in a uniform medium ) is @xmath177 , where @xmath132 is the local angular velocity of the galaxy .
since this growth rate is a factor @xmath178 lower than typical ti growth rates , the early development of the ti+mri model is the same as in the `` pure ti '' model . by
the time mri begins to develop ( after a few @xmath179 myr ) , the ti+mri model has similar cloud / intercloud structure except with more variations in cloud size to the cloud+mri model . at early times ,
the density and temperature pdfs are essentially the same as those produced by ti ; at late times , however , while the pdfs remain bimodal , the dense gas is distributed over a somewhat larger range of densities and temperatures , due to the dynamics of the `` channel flow '' solution ( see below ) .
_ spatial scales of mri in a cloudy medium : _ in both our ti+mri and cloud+mri simulations , after a few galactic orbital times , the velocity and magnetic fields become dominated by large - scale structures .
since the smallest - scale mri mode that would fit in our @xmath180 box under _ uniform - density _ conditions has vertical wavenumber @xmath134 ( in units @xmath123 ; i.e. wavelength @xmath181 ) , and the fastest - growing mode would have @xmath133 , this implies , consistent with expectations , that cloudy density structure in the supporting medium does not grossly alter the character of mri .
we quantify mri structural development in terms of mode amplitudes of @xmath129 , the azimuthal magnetic field . for the ti+mri model ,
the amplitudes of the @xmath182 and @xmath183 modes are all similar and motions in the @xmath18 plane continue to be dominated by ti effects , with cloud agglomeration until @xmath184 , after which the clouds have become highly concentrated and the @xmath131 mri mode associated with the `` channel solution '' @xcite takes hold . for the cloud+mri model , on the other hand , the @xmath133 mode grows first ( with clouds remaining small and distributed ) and it dominates until @xmath185 , when the channel solution ( @xmath131 ) begins to take precedence .
these differences show that the spatial distribution of clouds can have a significant effect on selecting which mri modes are important .
if intercloud distances are small compared to its wavelength , the dominant mri mode is the same as that predicted for uniform - density conditions .
if , however , other turbulent processes acting on scales _ small _ compared to mri wavelengths ( and times small compared to @xmath186 ) collect the clouds and correspondingly increase their separations , then only mri modes at scales larger than twice the typical intercloud distance will be able to grow . as a consequence , for mri to play an important role in the ism , either the majority of the gas must remain in a warm , diffuse phase , or else if it collects in clouds their separations must not be too large .
it is interesting to relate these constraints to observational inferences of the hi spatial distribution . from the @xcite hi absorption observations that yielded 142 separate cold gas components on 47 lines of sight at @xmath187
, their mean separation would be @xmath178 pc ( taking the cold disk semi - thickness @xmath188 pc ) .
the distribution of warm gas is much harder to interpret , but in the limiting situation where it is mainly in overdense clouds at high latitudes is at lower temperatures than the @xmath189 required for approximate pressure equilibrium with the cold clouds ; since significant underpressures are difficult to achieve , this gas is likely to be in clouds denser than @xmath145 . ] , and using heiles and troland s finding that @xmath190 of emission components have no associated absorption , the mean distance between clouds would be @xmath191 pc .
intercloud separations similar to these estimates are small enough that vertical mri modes could be supported ; if cloud spacings are appreciably larger , however , they could not be .
_ growth rates and saturation amplitudes of mri : _ for the low-@xmath136 modes that are present in both our ti+mri and cloud+mri models , typical growth rates are generally comparable to those for modes of the same wavelength in a medium of the same mean density . for the ti+mri model , typical growth rates
are measured to be @xmath192 , @xmath193 , and @xmath194 for the @xmath131 , 2 , and 3 modes , respectively , compared to the rates @xmath195 , @xmath196 , and @xmath197 that would apply for a uniform medium . for the cloud+mri model ,
the exponential mri growth is `` cleaner '' ; rates are @xmath198 , @xmath199 , and @xmath200 for @xmath131 , 2 , and 3 modes , respectively
. the growth rates of smaller - scale ( @xmath201 ) modes are thus slightly more affected by the presence of cloudy structure than that of the largest - scale ( @xmath131 ) mode , consistent with expectations .
although definitive results await 3d simulations , these findings provide support for the possibility that mri may drive turbulence in the diffuse ism at amplitudes consistent with observations of hi emission and absorption . from previous 3d simulations under relatively _ uniform _ conditions ( accomplished by adopting an isothermal equation of state ) ,
the velocity dispersions driven by mri in steady - state were found to be smaller than observed values .
in particular , @xcite found that the typical 1d turbulent amplitudes are 3 - 4 @xmath169 , whereas the observed nonthermal contribution to the 1d velocity dispersion for both cold and warm gas amounts to @xmath202 @xcite .
thus , for a _ single phase medium _ , mri - driven turbulent velocity amplitudes in steady state
which are determined by a balance between excitation and dissipation fall a factor @xmath203 short of explaining observations .
since our present cloudy - medium models show growth rates quite comparable to those in a one - phase medium , the key question is therefore whether mri dissipation rates are reduced in a cloudy medium , and if so , whether the reduction can yield a factor two increase in @xmath204 . to see that a quantitative effect at this level is not unreasonable , consider the comparison to an idealized system of @xmath205 clouds per unit volume having individual radii @xmath206 , internal density relative to the mean value @xmath207 , and rms relative velocity dispersion @xmath204 . with turbulent energy driving and dissipation rates
@xmath208 and @xmath209 , where the collision time @xmath210 , @xmath204 in steady state is an order - unity factor times @xmath211 . for this idealized situation ,
concentrating material into clouds with @xmath212 ( similar to cold ism clouds ) would indeed increase @xmath204 by a factor two compared to the case with near - uniform conditions , @xmath213 . with 3d simulations
, it will be possible to test whether a similar scaling behavior holds for the saturated state of mri - driven turbulence in cloudy vs. single - phase ism models . | the structure and dynamics of diffuse gas in the milky way and other disk galaxies may be strongly influenced by thermal and magnetorotational instabilities ( ti and mri ) on scales @xmath0 pc .
we initiate a study of these processes , using two - dimensional numerical hydrodynamic and magnetohydrodynamic ( mhd ) simulations with conditions appropriate for the atomic interstellar medium ( ism ) .
our simulations incorporate thermal conduction , and adopt local `` shearing - periodic '' equations of motion and boundary conditions to study dynamics of a @xmath1 radial - vertical section of the disk .
we demonstrate , consistent with previous work , that nonlinear development of `` pure ti '' produces a network of filaments that condense into cold clouds at their intersections , yielding a distinct two - phase warm / cold medium within @xmath2 myr .
ti - driven turbulent motions of the clouds and warm intercloud medium are present , but saturate at quite subsonic amplitudes for uniform initial @xmath3 .
mri has previously been studied in near - uniform media ; our simulations include both ti+mri models , which begin from uniform - density conditions , and cloud+mri models , which begin with a two - phase cloudy medium .
both the ti+mri and cloud+mri models show that mri develops within a few galactic orbital times , just as for a uniform medium
. the mean separation between clouds can affect which mri mode dominates the evolution .
provided intercloud separations do not exceed half the mri wavelength , we find the mri growth rates are similar to those for the corresponding uniform medium .
this opens the possibility , if low cloud volume filling factors increase mri dissipation times compared to those in a uniform medium , that mri - driven motions in the ism could reach amplitudes comparable to observed hi turbulent linewidths . |
semiconductor superlattices ( ssls ) @xcite have attracted growing attention in view of their unique electronic properties , which can be used for generation , amplification and detection of a high - frequency electromagnetic radiation @xcite .
a dc - biased ssl , operating in the conditions of single miniband transport regime and demonstrating the static negative differential conductance , can be potentially used as an active element of thz field generator @xcite .
however , the static negative differential conductance makes ssl unstable against a formation of high - field electric domains @xcite .
the electric domains are believed to be destructive for the thz gain in ssls @xcite .
currently the main focus is on the possibilities to overcome this drawback within the scheme of dc - biased ssl @xcite .
however , schemes of thz superlattice devices with ac pump fields are also under discussion @xcite .
a strong microwave field @xmath0 pumps the ssl and a desirable signal field @xmath1 has a higher frequency @xmath2 .
because for typical ssls the characteristic scattering time @xmath3 at room temperature is of the order of @xmath4 fs , an interaction of the microwave fields with the miniband electrons is quasistatic ( @xmath5 ) .
importantly , we have showed recently that such quasistatic pump field can completely suppress domains in ssls @xcite .
two distinct possibilities exist for the signal field : it can be also quasistatic ( @xmath6 ) or it can not be described within the quasistatic approach if @xmath7 . the later situation , i.e. @xmath5 but
@xmath7 , can be called _
semiquasistatic interaction_. thus , the semiquasistatic approach is introduced to describe an amplification of thz field in ssl under the action of microwave field .
we should note that while in the experiments @xcite the interaction of electrons with both pump and signal fields is still quasistatic and , moreover , it is probably domains - mediated , the ultimate goal of this activity is to reach a parametric thz generation without electric domains @xcite .
therefore , we can say that the semiquasistatic approach describes the real experimental situations as well . in this paper
, we derive several beautiful formulas describing a high - frequency absorption ( gain ) of small - signal field in the presence of ac pump field , when the interaction of the fields with the miniband electrons in the superlattice is semiquasistatic . along with obvious potential applications in the calculations of gain in superlattice devices ,
our theoretical approach provides a powerful tool for finding the correspondence between quasistatic and dynamic regimes in ac - driven superlattices .
our main findings are following .
we consider the response of electrons , belonging to a single miniband of ssl , to an action of the total electric field @xmath8 where @xmath9 is the dc bias and @xmath10 is the weak signal ( probe ) field .
we should note that in the real devices @xmath11 is a mode of the resonator tuned to a desirable thz frequency . here
three distinct cases should be considered : ( i ) @xmath12 and @xmath13 are incommensurable , ( ii ) @xmath14 ( @xmath15 is an odd number ) and ( iii ) @xmath16 ( @xmath15 is an integer ) .
the later case of the parametric cascading is most interesting physically @xcite and we mainly focus on it in this paper .
we define the dimensionless absorption of the weak ac probe field in ssl as @xmath17 where @xmath18 is the scaled miniband electron velocity defined in the units of the maximal miniband velocity @xmath19 @xcite , averaging @xmath20 is performed over time and @xmath21 is the common period for the probe and the pump ac fields .
starting from the exact formal solution of the boltzmann transport equation , we represent the absorption @xmath22 as the sum of three terms @xmath23 here @xmath24 , @xmath25 and @xmath26 describe the absorption ( gain ) seeded by generation of harmonics , the parametric amplification of the probe field due to a coherent interaction of the pump and the probe fields , and the nonparametric absorption , correspondingly .
the term @xmath24 is just the expression for @xmath15th in - phase harmonic of the time - dependent current through ssl @xmath27 where @xmath28 is the dc voltage , @xmath29 is the length of ssl ( @xmath30 is the period of ssl and @xmath31 is the number spatial periods ) , @xmath32 is the amplitude of ac voltage created by the pump field across ssl , the current @xmath33 is normalized to the maximal current in ssl , @xmath34 , corresponding to the maximal miniband velocity @xmath19 .
the explicit expressions for the @xmath15th harmonic of the current , that determine the dependence @xmath35 , are well - known ( see @xcite for the case of a small pump amplitude , @xmath36 , and @xcite for the case of an arbitrary @xmath37 ) . note that @xmath24 does not depend on the probe ; it gives the main contribution to the absorption of a weak probe @xcite .
the expression for @xmath24 is not specific for the semiquasistic limit because of its independence on @xmath12 .
however , _ our main finding is that the absorption components @xmath25 and @xmath26 can be represented within the semiquasistatic approach using the specific terms of quantum derivatives _ as @xmath38 @xmath39 where @xmath40 is the amplitude of small - signal voltage and @xmath41 is the esaki - tsu voltage - current ( ui ) characteristic ( @xmath42 is the critical voltage and @xmath43 is normalized to the maximal current @xmath44 ) .
importantly , following eq .
( [ inc ] ) in order to find the incoherent absorption in ssl at arbitrary high frequency @xmath12 we need to know only @xmath45 that is , _ the time - averaged current induced by the quasistatic field ( voltage)_. for a given amplitude of the ac voltage @xmath37 , the dc current @xmath46 is a function of only dc bias @xmath47 .
it is easy to calculate or to measure the modifications of ui characteristic caused by the action of microwave ( quasistatic ) field @xcite . on the other hand , finding of the coherent component of absorption at the high frequency , which corresponds to the @xmath15th harmonic of the pump frequency ( @xmath16 ) ,
is reduced following eq .
( [ coh ] ) to _ the calculation of the @xmath48th harmonic of the current within the quasistatic approach_. this result is especially important , because the term @xmath25 can be responsible for the parametric gain in ac - driven ssls @xcite .
thus , one needs to know @xmath49 ( note that in comparison with the expression ( [ harm - gen ] ) we have in eq .
( [ harm - vm ] ) : @xmath50 ) .
for a given pump amplitude @xmath37 the amplitudes of harmonics of the quasistatic current @xmath51 are functions of only dc bias @xmath47 .
we underline that quasistatic calculations of harmonics @xmath51 are among the standard theoretical tools to explain experimental results on the frequency multiplications in the microwave frequency band @xcite . in this paper
we will present simple but instructive geometric interpretations of the formulas ( [ coh ] ) and ( [ inc ] ) allowing to find the high - frequency absorption components directly using only the ruler and the knowledge of the quasistatic curves @xmath52 and @xmath53 .
these geometric interpretations provide a powerful tool in a search of the optimum conditions for the gain which employs only a simple qualitative analysis . if the frequency of the probe @xmath12 incommensurates with the frequency of the pump @xmath13 , the expression for the absorption at @xmath12 consists of only one term the incoherent absorption @xmath54 with semiquasistatic limit ( [ inc ] ) .
additionally , we will also show that within the semiquasistatic approach the difference formulas similar to eqs .
( [ coh])-([inc ] ) can be written for the absorption at half - harmonics @xmath14 .
the organization of this paper is as follows .
we start with the calculations of absorption for arbitrary pump and probe frequencies solving the boltzmann transport equation in sec .
[ sec : general_case ] .
the derivation of the absorption formulas ( [ coh ] ) and ( [ inc ] ) within the semiquasistatic limit is presented in sec . [ sec : semiquasi ] . in the sec .
[ sec : quasi ] , an additional limitation on the frequency of the probe is imposed and the quasistatic formulas for the absorption are obtained . section [ sec : numerical ] is devoted to the presentation of useful geometric interpretations of the semiquasstatic results and their numerical verifications .
the summary of the work , as well as a brief discussion of the place of our findings among other works devoted to the use of quantum derivatives in solid - state systems , are presented in the final section .
appendixes [ app : coherent ] and [ app : harmonic ] are devoted respectively to the derivation of the semiquasistatic limit for the coherent component of absorption and to the derivation of the quasistatic limit for the harmonic component of absorption .
in this section we will represent the expressions for the components of absorption that follow from the exact formal solution of the transport boltzmann equation with a constant relaxation time @xcite . using this exact solution , we found @xcite that for general case of commensurate frequencies @xmath55 ( @xmath56 are integers and @xmath57 is an irreducible fraction ) the total absorption @xmath58 , eq .
( [ start ] ) has the form @xmath59\nonumber\\ & \times&\frac{(\omega_0+l_1\omega_1+l_2\omega_2)\tau } { 1+(\omega_0+l_1\omega_1+l_2\omega_2)^2\tau^2},\end{aligned}\ ] ] where @xmath60 , @xmath61 ( @xmath62 ) , and @xmath63 is the bloch frequency . to consider the absorption at @xmath15th harmonic of the pump frequency one should fix @xmath64 ; for the case of absorption at half - harmonics
@xmath65 should be equal to @xmath66 .
first , we consider the absorption at harmonics of the pump field . in the limit of the weak probe ( @xmath67 )
, we need to take only certain combinations of the bessel functions indexes . as a result
we represent the absorption as a sum of three terms @xcite @xmath68 @xmath69\frac{(\omega_0+l\omega_1)\tau } { 1+(\omega_0+l\omega_1)^2\tau^2}\end{aligned}\ ] ] @xmath70\frac{(\omega_0+l\omega_1)\tau } { 1+(\omega_0+l\omega_1)^2\tau^2},\end{aligned}\ ] ] @xmath71.\end{aligned}\ ] ] second , for the absorption at half - harmonics of the pump frequency we get from ( [ absorp ] ) the expression for @xmath58 as the sum of two terms @xmath72 where @xmath73\frac{(\omega_0+l\omega_1)\tau } { 1+(\omega_0+l\omega_1)^2\tau^2}\end{aligned}\ ] ] and @xmath26 is given by the formula ( [ aincoh ] ) .
finally , it can be shown that if @xmath74 is some irrational number , then the absorption is completely determined by the incoherent component : @xmath75 .
now we turn to the consideration of semiquasistatic limits of different absorption components .
the derivation of semiquasistatic formulas is based on the use of the following integral representation of the esaki - tsu characteristic @xcite @xmath76 and the asymptotic saddle - point method @xcite .
we start with the consideration of semiquasistatic limit for the term @xmath26 , eq .
( [ aincoh ] ) .
we rewrite @xmath26 in terms of @xmath43 ( [ et_integr ] ) as @xmath77 \nonumber\\&=&\frac{\beta_2}{2\tau}\int\limits_{0}^{\infty}\exp\left(-\frac{t}{\tau}\right ) \sum_{l=-\infty}^{\infty}j_{l}^2(\beta_1 ) \big\{\sin\big[(\omega_0+l\omega_1+\omega_2)t\big ] -\sin\big[(\omega_0+l\omega_1-\omega_2)t\big]\big\}\,dt.\nonumber\end{aligned}\ ] ] using the addition formula @xmath78=\sin\big[(\omega_0+\omega_2)t\big]\cos(l\omega_1t)+ \cos\big[(\omega_0+\omega_2)t\big]\sin(l\omega_1t)\ ] ] and taking into account that the sum with @xmath79 is equal to zero , we get @xmath80\nonumber\\&-&\sin\big[(\omega_0+l\omega_1-\omega_2)t\big]\big\}\,dt = \frac{\beta_2}{2\tau}\int\limits_{0}^{\infty}\exp\left(-\frac{t}{\tau}\right)j_{0}\left(2\beta_1\sin\frac{\omega_1t}{2}\right)\nonumber\\ & \times & \big\{\sin\big[(\omega_0+\omega_2)t\big ] -\sin\big[(\omega_0-\omega_2)t\big]\big\}\,dt.\end{aligned}\ ] ] in eq .
( [ promegut1 ] ) we also have used the equality @xcite @xmath81 as a next step , we substitute the integral representation for @xmath82 @xcite @xmath83 in ( [ promegut1 ] ) and get @xmath84\nonumber\\&- & \sin\big[(\omega_0-\omega_2)t\big]\big\}\,dt= \frac{\beta_2}{2\pi\tau}\int\limits_{0}^{\infty}\exp\left(-\frac{t}{\tau}\right ) \int\limits_{0}^{\pi}\left\{\sin\left[\left(\omega_1\frac{\sin(\omega_1t/2 ) } { \omega_1t/2}\cos\theta+\omega_0+\omega_2\right)t\right]\right.\nonumber\\ & -&\left.\sin\left[\left(\omega_1\frac{\sin(\omega_1t/2 ) } { \omega_1t/2}\cos\theta+\omega_0-\omega_2\right)t\right]\right\}\,d\theta\ , dt \stackrel{\omega_1\tau\ll1}{\longrightarrow } \nonumber\\ & & \frac{\beta_2}{4\pi\tau}\int\limits_{0}^{\infty}\exp\left(-\frac{t}{\tau}\right ) \int\limits_{0}^{2\pi}\big\{\sin\left[\left(\omega_1
\cos\theta+\omega_0+\omega_2\right)t\right]-\sin\left[\left(\omega_1 \cos\theta+\omega_0-\omega_2\right)t\right]\big\}\ , d\theta \ , dt .
\nonumber\\\end{aligned}\ ] ] in ( [ sumin ] ) we first used simple formula @xmath85 and then took into account that for @xmath5 the main contribution to the integrals with the exponential factor @xmath86 comes from the terms formally satisfying @xmath87 ( the saddle - point method ) .
finally , using eq .
( [ et_integr ] ) we make the inverse transformation to the esaki - tsu current and obtain @xmath88\,d\theta.\ ] ] now it is easy to see that after making the transformation from frequency to voltage variables and using the esaki - tsu characteristic in the form ( [ volt ] ) , we get for @xmath26 the formula ( [ inc ] ) .
applying the same method , one can obtain ( see appendix [ app : coherent ] ) in the semiquasistatic limit the following expression for the coherent component of absorption @xmath89\cos(2m\theta)\,d\theta,\ ] ] which is equivalent to the formula ( [ coh ] ) .
we turn to the discussion of the semiquasistatic limit for absorption in the cases of incommensurable frequencies and half - harmonics . as we have found in the previous section ,
the absorption for incommensurable frequencies is just @xmath26 defined by eq .
( [ aincoh ] ) .
therefore , its semiquasistatic limit is just the formula ( [ incohsemi ] ) .
next , in the general case of arbitrary values of @xmath90 and @xmath91 , the coherent component of absorption at half - harmonics @xmath92 has the same form as the coherent component of absorption at harmonics ( [ acoh ] ) , if one makes a formal change @xmath93 ( cf .
( [ acohhalf ] ) and ( [ acoh ] ) ) .
therefore , the quasistatic limit of @xmath92 coincides with eq .
( [ coh ] ) after the substitution @xmath93 .
however , we should note that the consistent proof of this statement is not trivial because the derivations of semiquasistatic formulas for odd and for even values @xmath15 have different ways ( in the case of half - harmonics @xmath15 is the odd number but in the case of harmonics @xmath48 is the even number ) .
to find the quasistatic limit of the components of absorption we have to impose the additional limitation on @xmath12 : @xmath94 .
for @xmath24 we obtain from eq .
( [ aharm ] ) in the quasistatic limit @xmath95 of course , this expression is identical to eq .
( [ harm - vm ] ) .
formally , eq . ( [ harmquasi ] ) follows from ( [ harm - gen ] ) after the substitution @xmath50 , but the correct and consistent derivation of this result still requires some algebra ( see appendix [ app : harmonic ] ) . to get the quasistatic limit for @xmath26 and @xmath25 we need to consider the limit @xmath96 in the eqs .
( [ incohsemi ] ) and ( [ cohsemi ] ) . in this
limit the finite difference @xmath97 goes into the derivative @xmath98 and hence @xmath25 and @xmath26 become @xmath99 these formulas in terms of voltage and current , eq .
( [ volt ] ) , have the following form @xmath100 @xmath101 thus , in the quasistatic limit the incoherent and the coherent components of absorption at @xmath15th harmonic are proportional respectively to the derivative of dc component ( [ vi ] ) and to the derivative of @xmath48th harmonic ( [ harm - vm ] ) . on the other hand , the formulas for the absorption components can be derived completely within the quasistatic approach .
we define the absorption of quasistatic field ( voltage ) as @xmath102 , where the esaki - tsu ui characteristic @xmath103 is given by eq .
( [ volt ] ) . for a weak signal
we expand the current in the taylor series @xmath104 ( here prime means the derivation with respect to @xmath105 , @xmath106 , @xmath107 ) . substituting this expansion to the definition of @xmath108 we have @xmath109 making comparison of eq .
( [ aesaki ] ) with other formulas from this section , we see that in the limit @xmath110 the semiquasistatic formulas for the components of absorption correctly reproduce the quasistatic results .
[ color online ] geometrical meaning of the incoherent component of absorption within semiquasistatic and quasistatic approaches .
time - averaged current @xmath46 under the action of quasistatic pump ( @xmath111 , @xmath112 ) vs dc voltage @xmath47 .
if we choose the working point at @xmath113 , then the dotted segment [ red online ] corresponds to the finite difference ( quantum derivative ) of a weak probe at @xmath114 and the dashed straight line [ blue online ] corresponds to the derivative .
the slopes of these strait lines determines @xmath26 respectively in semiquasistatic and quasistatic cases . ]
[ color online ] geometrical meaning of the coherent component of absorption @xmath25 within semiquasistatic and quasistatic approaches .
the @xmath48th harmonic of quasistatic current @xmath115 vs dc voltage @xmath47 for @xmath116 .
the current is induced by the quasistatic pump ( @xmath117 ) with the amplitude @xmath118 .
slope of dotted segment [ red online ] determines the ( negative ) finite difference for a weak probe with @xmath119 .
slope of dashed straight line [ blue online ] determines the derivative .
working point is chosen at @xmath113 . ]
[ color online ] geometric interpretation for the coherent component of absorption at half - harmonics @xmath120 within semiquasistatic ( dotted [ red online ] segment ) and quasistatic ( dashed [ blue online ] line ) .
the curve represents @xmath15th harmonic of quasistatic current @xmath121 vs dc voltage @xmath47 for @xmath122 , @xmath117 and @xmath123 .
the zero slopes of the straight lines show that @xmath120 calculated for @xmath113 is zero . ]
[ color online ] incoherent component of absorption for different strengths of ac pump . both difference and usual derivatives are simultaneously small and approach each other for the strong pump ( @xmath124 ) [ blue online ] , in contrast to the case of weak pump ( @xmath112 ) [ red online ] where they are quite different . here
@xmath117 and @xmath125 . ]
[ color online ] comparison of the coherent absorptions for different values of @xmath126 .
quantum derivative [ blue online ] approaches normal derivative ( the tangent line , not shown ) for @xmath127 ; they are quite different for @xmath116 [ red online ] .
quasistatic pump with @xmath128 has the amplitude @xmath118 . ] to discuss the geometric meaning of incoherent and coherent components of absorption it is convenient to present @xmath26 and @xmath25 explicitly in terms of zero and even harmonics of the quasistatic current @xmath43 . combining ( [ coh ] ) and ( [ inc ] ) with ( [ harm - vm ] ) and ( [ vi ] ) we have @xmath129\ ] ]
@xmath130\ ] ] alternatively , the expressions for the absorption components in terms of frequency ( [ incohsemi ] ) , ( [ cohsemi ] ) can be used . as is evident from fig .
[ fig:1 ] , the geometric representation of the quantum derivative ( finite difference ) of @xmath46 is the segment with the length defined by the frequency of probe field @xmath12 .
the ends of the segment belong to the curve @xmath53 and their locations are determined by the location of working point .
slope of the segment determines @xmath26 within the semiquasistatic approach . in the quasistatic limit , a finite difference becomes an usual derivative and therefore @xmath26 is determined just by the slope of the tangent line to @xmath53 curve at the working point . in a similar manner , the coherent component of absorption @xmath25 at the frequency @xmath16 is determined by the slope of the segment with the ends belonging to the curve @xmath131 ( fig .
[ fig:2 ] ) . respectively
, the slope of the tangent line to @xmath131 at the working point determines @xmath25 in the quasistatic limit . using these two geometrical pictures it is possible , in principle , to optimize a location of the working point ( i.e. the value of applied dc bias ) in order to obtain the maximal small - signal gain in ssl in the conditions of suppressed domains , when both @xmath126 and the pump amplitude @xmath37 are given .
however , such an analysis goes beyond the scope of the present paper . in the rest of this section
we will focus on the case of dc - unbaised ssl ( @xmath113 ) , which corresponds to the choice of working point at the origin of coordinates in figs .
[ fig:1 ] and [ fig:2 ] .
such a choice guarantees the suppression of electric domains @xcite : a slope of the dependence @xmath46 on @xmath47 is positive practically for all @xmath37 in the limit @xmath132 ( for examples , see figs .
[ fig:1 ] , [ fig:4 ] , [ fig:6 ] ) . to illustrate the usefulness of the geometric interpretation in finding of schemes allowing gain in the conditions of suppressed domains
, we will consider two characteristic examples ( earlier the solutions of these problems have been announced without proofs in @xcite ) .
_ problem i_. suppose that @xmath74 is an irrational number and therefore only the incoherent absorption @xmath26 contributes to the gain in ssl . employing the geometrical interpretation of @xmath26
, we immediately see that if the slope of the tangent line to @xmath46 curve is positive , then the quantum derivative is also positive for an arbitrary @xmath12 ( see , e. g. , figs . [ fig:1],[fig:4 ] ) . in physical terms , this means that a small - signal gain in the unbiased ssl is impossible in the conditions of suppressed domains , if the frequencies of pump and probe are incommensurable . _ problem ii_. consider the absorption at a probe frequency that is a half - harmonic of the pump , @xmath133 .
the total absorption is the sum of the incoherent component @xmath26 and the coherent absorption at half - harmonics @xmath120 ( see eq .
[ acohhalf ] ) .
the geometric interpretation of @xmath120 is similar to the interpretation of @xmath25 : one needs only to consider @xmath121 as a function of @xmath47 instead of a dependence @xmath131 . obviously , the dependence of any harmonic of the quasistatic current on the applied dc bias , i. e. @xmath134 , is always symmetric about the vertical line @xmath113 therefore , the choice of the working point at @xmath113 results in zero value of both the derivative and the quantum derivative in this point ( fig .
[ fig:3 ] ) .
thus , in the case of dc - unbiased ssl @xmath120 is always zero . on the other hand , as we have saw earlier , @xmath26 is always positive in the conditions of suppressed domains . summing up
we can conclude that a small - signal gain in unbiased ssl is impossible in the conditions of suppressed domains , if the frequency of probe is a half - harmonic of the pump frequency .
dependence of @xmath26 on the amplitude of ac pump calculated within semiquasistatic ( solid ) and quasistatic ( dashed ) approaches for @xmath135 and @xmath117 . ] same as in fig .
[ fig:6 ] but for @xmath25 and @xmath116 ] . these two conclusive findings allow to narrow the class of possible schemes for the superlattice thz emitters with ac - pump , which can operate in the domainless regimes .
the proper scheme should exploit a negative absorption at some particular harmonics of the probe , when negative values of harmonic ( @xmath24 ) and coherent contributions to the net gain can overcome always positive incoherent absorption component @xcite .
therefore , in our analysis we return to the case @xmath137integer and compare @xmath26 and @xmath25 in conditions of quasistatic , semiquasistatic and dynamic approaches .
semiquasistatic and quasistatic results have a visible difference for higher harmonics as is evident from figs .
[ fig:1 ] and [ fig:2 ] ( here the quasistatic condition @xmath6 is not satisfied for the probe field ) .
however , in the case of low - order harmonics the difference between semiquasistatic and quasistatic results is very small .
this fact is illustrated in fig .
[ fig:5 ] for @xmath25 , where the difference between the derivative ( tangent line ) and the quantum derivative is very small for @xmath127 , but it visible increases for @xmath116 .
the dependencies of @xmath26 and @xmath25 on the amplitude of pump field , calculated using semiquasistatic ( solid ) and quasistatic ( dashed ) formulas for large @xmath15 , are shown in figs .
[ fig:6 ] and [ fig:7 ] . here
@xmath26 is always positive and its value decreases with an increase of the pump amplitude .
[ fig:6 ] demonstrates that the quasistatic calculations significantly overestimate @xmath26 for a small pump @xmath138 ( see also fig .
[ fig:1 ] ) .
however , the agreement between semiquasistatic and quasistatic approaches in finding @xmath26 becomes good for a strong pump ( fig .
[ fig:6 ] ) .
such paradoxical behavior is explained by the effect of shift of the maximum of ui characteristic under the action of quasistatic ac field @xcite ( note : in our case the pump is still quasistatic ) . as we see in fig .
[ fig:4 ] , for a large pump ui characteristic is rather closed to the linear in the interval of dc voltages that is utilized for the calculation of quantum derivative , what is in a sharp contrast to the case of large ac pump where ui characteristic demonstrates a strongly nonlinear behavior in the same interval .
thus , for a strong pump the finite difference is close to the tangent line , providing a small difference between the results of semiquastatic and quasistatic approaches for @xmath26 . as for @xmath25 ,
the difference between semiquastatic and quasistatic calculations is quite visible for large @xmath15 ( fig .
[ fig:7 ] ) . moreover , the quasistatic approach not only overestimate gain in this case
, it can give even a wrong sign of @xmath25 in some interval of ac pump amplitudes ( fig .
[ fig:7 ] ) .
we also calculated numerically @xmath26 and @xmath26 using formulas ( [ acoh ] ) , ( [ aincoh ] ) valid for arbitrary values of @xmath90 and @xmath91 .
results of these calculations in comparison with semiquasistatic calculations are shown in figs .
[ fig:8 ] , [ fig:9 ] for the case of moderate pump strength and in figs
. [ fig:10 ] , [ fig:11 ] for a strong ac pump .
for all calculations , including those which are not shown in these figures , we always found an excellent agreement between results of dynamic ( dots ) and semiquasistatic ( solid ) approaches for small @xmath90 .
surprisingly , the semiquasistatic formulas work quite well even beyond their formal range of validity , i.e. not only for small values of @xmath90 . for the moderate pump ,
the semiquasistatic approach works well up to @xmath139 ( figures [ fig:8 ] , [ fig:9 ] ) . even for a strong enough pump
, the semiquasistatic approach still gives a correct smoothed - out curve , while it does not exactly reproduce small oscillations in the dependence of absorption on the pump frequency ( figures [ fig:10],[fig:11 ] ) .
dependence of @xmath26 on @xmath90 calculated within semiquasistatic ( solid ) , dynamic ( dotted ) and quasistatic ( dashed ) approaches for @xmath140 and pump amplitude @xmath141 . ]
dependence of @xmath25 on @xmath90 calculated within semiquasistatic ( solid ) , dynamic ( dotted ) and quasistatic ( dashed ) approaches for @xmath140 and @xmath141 . ]
same as in fig .
[ fig:8 ] but for @xmath142 . ]
same as in fig .
[ fig:9 ] but for @xmath142 . ]
in summary , we calculated a small - signal absorption ( gain ) of a high - frequency field in semiconductor superlattice driven by a quasistatic ac field . in particular , our theoretical approach well describes practically interesting cases of generation and amplification of thz signal in superlattice devices pumped by a microwave field .
we presented transparent geometrical interpretations for the coherent and the incoherent contributions to the net absorption at the different ratio of the pump ( @xmath13 ) and the signal ( @xmath12 ) frequencies .
we demonstrated that the use of these geometric pictures sufficiently simplifies analysis : it is easy to find gain for a signal with an arbitrary @xmath12 using the electric characteristics of superlattice formed under microwave fields .
in particular , we proved that for @xmath74 being some irrational number a small - signal gain at @xmath12 is possible only in the condition of static negative differential conductance .
next , we also found that an absorption at an arbitrary @xmath12 , that equals to some half - integer of the pump frequency , is always zero in an unbiased superlattice . at least but not last
, we presented the universal analytic procedure for finding the behavior of different physically important variables in the quasistatic limit starting from the exact solution of the boltzmann transport equations for superlattices .
numerical calculations confirmed our analytic findings and also demonstrated that the semiquasistatic formulas for absorption work reasonable well even beyond their formal range of validity .
our results can be generalized to other physical systems where the miniband transport regime exist , such as carbon nanotubes @xcite or dissipative optical lattices @xcite .
although in this work we used only a single scattering time , our results obtained within semiquasistatic approximation can be also applied to a more realistic case of two different relaxation times for electron energy and electron momentum . really , within the approximation of two relaxation times the static voltage - current characteristic of superlattice still has the esaki - tsu form @xcite . because in the semiquasistatic approach only the ( quasi)static voltage - current characteristic plays a role , we speculate that a variation of the ratio of relaxation times can lead only to a some change in absolute value of the absorption but not in its dependence on the amplitude of pump field .
preliminary numerical simulations of the superlattice balance equations @xcite confirm these speculations @xcite . the detailed consideration of thz absorption with the use of two different relaxation times will be published elsewhere .
it is instructive to consider our activity from the viewpoint of its place among other works devoted to the use of the quantum derivatives or , other words , to the use of the finite differences in expressions for currents . in the theory of semiconductor superlattices ,
the quantum derivative has been first introduced in @xcite for calculations of the absorption of weak monochromatic ac field in dc - biased superlattice .
later the formula has been used in the calculations of the response function of thz superlattice detector @xcite and in the finding of thz gain in the models of generalized bloch oscillator with suppressed domains @xcite .
interestingly , these formulas for superlattices appears to be same as the formulas describing the response of two - terminal structures to the action of dc and ac voltages @xcite .
the latter are widely used for calculations of signal mixing in the josephson junctions operating in the qusiparticle transport regime and in other single - barrier tunnel devices @xcite . to the best of our knowledge
this paper is the first work , where the quantum derivatives naturally appeared in a description of microstructure s response to a bichromatic field .
it is an exciting problem to understand whether our semiquasistatic approach , developed for superlattices operating in the miniband transport regime , can be generalized to other tunnelling structures .
this research was supported by academy of finland ( grants 109758 ) and aqdjj programme of european science foundation .
we thank timo hyart for useful discussions and critical reading of the manuscript .
in this appendix we consider the semiquasistatic limit for @xmath25 .
we rewrite @xmath25 ( [ acoh ] ) in terms of @xmath43 ( [ et_integr ] ) as @xmath143i^{et}(\omega_0+l\omega_1 ) \nonumber\\&=&\frac{\beta_2}{2\tau}\int\limits_{0}^{\infty}\exp\left(-\frac{t}{\tau}\right)\sum_{l=-\infty}^{\infty}j_{l}(\beta_1 ) \big[j_{l-2m}(\beta_1 ) -j_{l+2m}(\beta_1)\big]\sin\big[(\omega_0+l\omega_1)t\big]dt.\end{aligned}\ ] ] after changing indexes , @xmath144 , in the first term of eq .
( [ acohapp1 ] ) we obtain @xmath145-\sin\big[(\omega_0+l\omega_1)t\big]\big\}\,dt .
\nonumber\\ \label{acohapp}\end{aligned}\ ] ] using the addition formula for sines we get @xmath146 to make the summation over @xmath147 we use the identity @xcite @xmath148 and get @xmath149 now we can take into account that for @xmath150 the main contribution to the integrals with an exponential factor comes from the terms formally satisfying @xmath151 . in this case
, we can use the estimations @xmath152 @xmath153 because @xcite @xmath154\,dt.\end{aligned}\ ] ] using the integral representation for @xmath155 @xcite @xmath156 we have @xmath157-\sin\big[(\omega_0+\omega_2)t\big]\big\}\,dt\,d\theta .\end{aligned}\ ] ] in this expression we use the formula for a product of sines and cosines and as a result we get the sum of eight integrals .
these integrals are @xmath158\sin(2m\theta)\,d\theta = \int\limits_{0}^{\pi}\cos\big[\omega_1t\sin\theta\pm(\omega_0+\omega_2)t\big]\sin(2m\theta)\,d\theta=0.\ ] ] and @xmath159-\sin\big[\omega_1t\sin\theta -(\omega_0\mp\omega_2)t\big]\big\ } \cos(2m\theta)\,d\theta\nonumber\\ & = & 2(-1)^m\int\limits_{0}^{\pi}\sin\big[\omega_1t\cos\theta + ( \omega_0\mp\omega_2)t\big]\cos(2m\theta)\,d\theta.\end{aligned}\ ] ] now ( [ coh1 ] ) takes the form @xmath160- \sin\big[\omega_1t\cos\theta+(\omega_0+\omega_2)t\big]\big\}\nonumber\\ & \times&\cos(2m\theta)\,dt\,d\theta,\end{aligned}\ ] ] which finally allow us to make the inverse transformation by means of ( [ et_integr ] ) to the esaki - tsu current as @xmath161\nonumber\\ & \times&\cos(2m\theta)d\theta=\frac{\omega_1}{4\pi}\beta_2\int\limits_{0}^{2\pi/\omega_1 } \left[i^{et}(\omega_1\cos(\omega_1t)+\omega_0+\omega_2)- i^{et}(\omega_1\cos(\omega_1t)+\omega_0-\omega_2)\right]\nonumber\\ & \times&\cos(2m\omega_1t)dt .\end{aligned}\ ] ] that is the desirable semiquasistatic representation for @xmath25 .
here we consider the semiquasistatic limit for @xmath24 .
it is convenient for our purposes to change indexes @xmath144 in the first term of ( [ aharm ] ) and rewrite @xmath24 in the following form @xmath162\nonumber\\\end{aligned}\ ] ] now we need to rewrite @xmath24 in terms of @xmath43 ( [ et_integr ] ) as @xmath163\nonumber\\ & = & \frac{1}{\tau}\int\limits_{0}^{\infty}\exp\left(-\frac{t}{\tau}\right ) \sum_{l=-\infty}^{\infty}j_{l}(\beta_1)j_{l+m}(\beta_1 ) \big\{(-1)^m\sin\big[(\omega_0-l\omega_1)t\big]+\sin\big[(\omega_0+l\omega_1)t\big]\big\ } \,dt\nonumber\\ & = & \frac{1}{\tau}\int\limits_{0}^{\infty}\exp\left(-\frac{t}{\tau}\right ) \sum_{l=-\infty}^{\infty}j_{l}(\beta_1)j_{l+m}(\beta_1 ) \big\{\big[(-1)^m+1\big]\sin(\omega_0t)\cos(l\omega_1t)\nonumber\\&+&\big[(-1)^m-1\big]\cos(\omega_0 t ) \sin(l\omega_1t)\big\}\,dt\end{aligned}\ ] ] we can take the sum over @xmath147 in ( [ aharmapp ] ) using formula ( [ sum1 ] ) and the identity @xcite @xmath164 as a result we have @xmath165\sin(\omega_0t)\cos\left(m\arcsin\cos\frac{\omega_1t}{2}\right ) j_m\left(2\beta_1\sin\frac{\omega_1t}{2}\right ) \nonumber\\&+&\big[(-1)^m-1\big]\cos(\omega_0 t ) \sin\left(m\arcsin\cos\frac{\omega_1t}{2}\right ) j_m\left(2\beta_1\sin\frac{\omega_1t}{2}\right)\big\}\,dt .\end{aligned}\ ] ] for @xmath150 the main contribution to the integrals with an exponential factor comes from the terms formally satisfying @xmath151 .
therefore , the following limits @xmath166 are important . substituting ( [ est ] ) in ( [ aharm1 ] ) , we obtain @xmath167\sin(\omega_0t)\cos\frac{m\pi}{2}j_m(\omega_1 t ) \right.\nonumber\\&+&\left.\big[(-1)^m-1\big]\cos(\omega_0t)\sin\frac{m\pi}{2 } j_m(\omega_1t)\right\}\,dt .\end{aligned}\ ] ] now odd and even values of @xmath15 must be considered separately . to simplify the expression ( [ aharm3 ] ) we will need the following formula @xmath172\cos(m\theta)\,d\theta.\nonumber\\\end{aligned}\ ] ] to prove it we first convert the integral as @xmath173 and then use the formula for a product of sines and cosines @xmath174 + \sin[(\omega_0+\omega_1\cos\theta)t].\end{aligned}\ ] ] taking into account that @xmath175\sin(m\theta)\,d\theta- \int\limits_{0}^{\pi}\sin[(\omega_0+\omega_1\sin\theta)t]\sin(m\theta)\,d\theta\nonumber\\ & = & 2(-1)^{k+1}\int\limits_{0}^{\pi}\sin[(\omega_0+\omega_1\cos\theta)t]\cos(m\theta)\,d\theta,\end{aligned}\ ] ] we get ( [ int ] ) . now substituting ( [ int ] ) in ( [ aharm3 ] ) and making the inverse transformation to the esaki - tsu current , we obtain @xmath176\cos(m\theta)\,dt\,d\theta\nonumber\\ & = & \frac{1}{\pi\tau}\int\limits_{0}^{2\pi}i^{et}(\omega_0+\omega_1\cos\theta)\cos(m\theta)\ , d\theta=\frac{\omega_1}{\pi}\int\limits_{0}^{2\pi/\omega_1}i^{et } \big(\omega_0+\omega_1\cos(\omega_1t)\big)\cos(m\omega_1t)\,dt . \nonumber\\\end{aligned}\ ] ] we turn now to the consideration of even harmonics . combining ( [ aharm2 ] ) for @xmath177 @xmath178 and the integral representation ( [ bess ] ) , we have @xmath179 next stage of simplification includes several steps .
first , we take into account that @xmath180 second , we apply the formula for a product of sines and cosines
. finally we also use @xmath181+\sin\big[(\omega_0-\omega_1\sin\theta)t\big]\big\ } \cos(m\theta)\,d\theta\nonumber\\ & = & 2(-1)^k\int\limits_{0}^{\pi}\sin\big[(\omega_0+\omega_1\cos\theta)t\big ] \cos(m\theta)\,d\theta.\end{aligned}\ ] ] as a result , we can represent ( [ aharm4 ] ) in the form suitable for the inverse transformation to the esaki - tsu current and obtain @xmath176 \cos(m\theta)\,dt\,d\theta\nonumber\\&=&\frac{1}{\pi}\int\limits_{0}^{2\pi } i^{et}(\omega_0+\omega_1\cos\theta)\cos(m\theta ) d\theta\nonumber\\ & = & \frac{\omega_1}{\pi}\int\limits_{0}^{2\pi/\omega_1}i^{et } \big(\omega_0+\omega_1\cos(\omega_1t)\big)\cos(m\omega_1t)\,dt\nonumber\end{aligned}\ ] ] 99 l. esaki and r. tsu , ibm j. res .
( 1970 ) 61 .
a. wacker , phys .
357 ( 2002 ) 1 .
g. platero , r. aguado , phys .
395 ( 2004 ) 1 .
a. sibille , c. minot , f. laruelle , international journal of modern physics b 14 ( 2000 ) 909 .
s. a. ktitorov , g. s. simin , v. ya .
sindalovskii , fiz .
tela 13 ( 1971 ) 2230 [ engl
. tranlation : sov .
phys . solid state 13 ( 1972 ) 1872 ] .
m. bttiker , h. tomas , phys . rev . lett . 38 ( 1977 ) 78 .
a. a. ignatov , v. i. shashkin , zh .
( 1987 ) 935 [ engl .
tranlation : sov .
jetp 66 ( 1987 ) 526 ] .
p. g. savvidis , b. kolasa , g. lee , s. j. allen , phys .
92 ( 2004 ) 196802 .
h. kroemer , cond - mat/0009311 .
t. hyart , k. n. alekseev , a. leppnen , e. v. thuneberg , cond - mat/0511145 .
a. lisauskas et al . , appl .
lett . 86 ( 2005 ) 102103 .
t. feil , h .-
tranitz , m. reinwald , w. wegscheider , appl .
87 ( 2005 ) 212112 .
f. klappenberger et al .
84 ( 2004 ) 3924 .
k. f. renk et al . , phys .
95 ( 2005 ) 126801 .
k. n. alekseev , n. v. demarina , m. v. gorkunov , cond - mat/0503210 . k. n. alekseev et al .
, europhys .
lett . , 73 ( 2006 ) in press and cond - mat/0601056 .
l. esaki , r. tsu , appl .
( 1971 ) 246 .
a. romanov , optika spektroskopiya 33 ( 1972 ) 917 [ engl .
tranlation : optics and spectroscopy 33 ( 1972 ) 503 ] .
a. a. ignatov , yu . a. romanov , phys .
b 73 ( 1976 ) 327 . v. v. pavlovich , e. m. epshtein , fiz . tekh .
poluprovodn . 10
( 1976 ) 2001 [ engl .
tranlation : sov .
semicond . 10
( 1976 ) 1196 ] .
e. shomburg et al . ,
68 ( 1996 ) 1096 .
s. winnerl et al . , phys .
b 56 ( 1997 ) 10303 .
j. grenzer et al .
, annalen der physik ( leipzig ) 4 ( 1995 ) 184 .
e. shomburg et al . ,
ieee journal of selected topics in quantum electronics 2 ( 1996 ) 724 .
r. g. chambers , proc .
( london ) a 65 ( 1952 ) 458 .
h. budd , phys .
127 ( 1962 ) 4 .
detailed derivation of eqs .
( [ absorp])([aincoh ] ) will be published elsewhere : a. v. shorokhov , k. n. alekseev ( in preparation ) . m. v. fedoryuk , the saddle - point method , ( nauka , moscow , 1977 ) .
a. p. prudnikov , yu . a. brychkov ,
o. i marichev , integrals and series , vol . 1 , ( gordon and breach , new york , 1990 ) . s. y. mensah , j. phys . cond .
4 ( 1992 ) l325 .
f. klappenberger et al .
j. b 39 ( 2004 ) 483 .
o. v. kibis , d. g. parfitt , m. e. portnoi , phys .
b 71 ( 2005 ) 035411 .
g. grynberg , c. robilliard , phys .
355 ( 2001 ) 335 .
a. a. ignatov , e. p. dodin , v. i. shashkin , mod .
b 5 ( 1991 ) 1087 .
a. leppnen , t. hyart , k. n. alekseev , oulu university report , august 2005 ( unpublished ) .
a. a. ignatov et al . , superlatt .
microstr . 22
( 1997 ) 15 .
a. a. ignatov , a .- p .
jauho , j. appl .
85 ( 1999 ) 3643 .
a. wacker et al . , phys .
b 204 ( 1997 ) 95 .
t. hyart , m. sc .
thesis , university of oulu , 2005 .
j. r. tucker , ieee j. quantum electron .
qe-15 ( 1979 ) 1234 .
m. j. feldman , j. appl .
53 ( 1982 ) 584 . j. r. tucker , m. j. feldman , rev .
57 ( 1985 ) 1055 .
a. n. korotkov , d. v. averiev , k. k. likharev , phys .
b 49 ( 1994 ) 7548 .
m. abramowitz , i. a. stegun , handbook of mathematical functions with formulas , graphs , and mathematical table , ( dover , new york , 1972 ) . | we consider a generation and an amplification of thz radiation in semiconductor superlattices under the action of microwave pump field .
electrons belonging to a single miniband of the superlattice interact quasistatically with the pump field and dynamically with a signal thz field . within this semiquasistic approach
we derive elegant difference formulas describing absorption ( gain ) of the weak thz signal .
we present an instructive geometric interpretation of the absorption formulas which allows a search of optimum conditions for the gain employing only a simple qualitative analysis .
our theoretical findings contribute to the development of sources and detectors of thz radiation that are using nonlinear electric properties of semiconductor superlattices . |
a homogeneous polynomial @xmath0 $ ] of degree @xmath1 in variables @xmath2 is called * hyperbolic * with respect to a point @xmath3 if @xmath4 and for every @xmath5 , all roots of the univariate polynomial @xmath6 $ ] are real .
its * hyperbolicity cone * , denoted @xmath7 is the connected component of @xmath8 in @xmath9 and can also be defined as @xmath10 as shown in grding @xcite , @xmath7 is an open convex cone and @xmath11 is hyperbolic with respect to any point contained in it .
hyperbolicity is reflected in the topology of the real projective variety @xmath12 in @xmath13 . if @xmath14 is smooth , then @xmath11 is hyperbolic if and only if @xmath12 consists of @xmath15 nested ovaloids , and a pseudo - hyperplane if @xmath1 is odd ( see ( * ? ? ?
* thm . 5.2 ) ) .
a hyperbolic program , introduced and developed by gler @xcite , renegar @xcite and others , is the problem of maximizing a linear function over an affine section of the convex cone @xmath7 .
this provides a very general context in which interior point methods are effective .
for example , taking @xmath16 and @xmath17 , we see that @xmath7 is the positive orthant @xmath18 and the corresponding hyperbolic program is a linear program .
if instead we take @xmath11 as the determinant of a symmetric matrix of variables @xmath19 and @xmath8 is the identity matrix , then @xmath7 is the cone of positive definite matrices . [ cols="^,^,^,^",options="header " , ] it is a fundamental open question whether or not every hyperbolic program can be rewritten as a semidefinite program .
helton and vinnikov @xcite showed that if @xmath20 $ ] is hyperbolic with respect to a point @xmath8 , then @xmath11 has a _ definite determinantal representation _
@xmath21 where @xmath22 are real symmetric matrices and the matrix @xmath23 is positive definite .
thus every three dimensional hyperbolicity cone is a slice of the cone of positive semidefinite matrices . for a survey of these results and future perspectives , see also @xcite .
on the other hand , brndn @xcite has given an example of a hyperbolic polynomial @xmath11 ( see example [ ex : vamos ] ) such that no power of @xmath11 has a definite determinantal representation .
there is a close connection between definite determinantal representations of a hyperbolic polynomial @xmath11 and polynomials of degree one - less that _ interlace _ it , which has also been used in @xcite to study hermitian determinantal representations of hyperbolic curves .
let @xmath24 $ ] be univariate polynomials with only real zeros and with @xmath25 .
let @xmath26 be the roots of @xmath11 , and let @xmath27 be the roots of @xmath28 .
we say that * @xmath28 interlaces @xmath11 * if @xmath29 for all @xmath30 .
if all these inequalities are strict , we say that * @xmath28 strictly interlaces @xmath11*. if @xmath0 $ ] is hyperbolic with respect to @xmath8 and @xmath28 is homogeneous of degree @xmath31 , we say that * @xmath28 interlaces @xmath11 with respect to @xmath8 * if @xmath32 interlaces @xmath33 for every @xmath34 .
this implies that @xmath28 is also hyperbolic with respect to @xmath8 .
we say that @xmath28 * strictly interlaces @xmath11 * if @xmath32 strictly interlaces @xmath33 for @xmath35 in a nonempty zariski - open subset of @xmath36 .
the most natural example of an interlacing polynomial is the derivative .
if @xmath37 is a real polynomial with only real roots , then its derivative @xmath38 has only real roots , which interlace the roots of @xmath11 . extending this to multivariate polynomials , we see that the roots of @xmath39 interlace those of @xmath33 for all @xmath5 .
thus @xmath40 interlaces @xmath11 with respect to @xmath8 .
if @xmath11 is square - free , then @xmath41 strictly interlaces @xmath11 .
this was already noted by grding @xcite and has been used extensively , for example in @xcite and @xcite ; for general information on interlacing polynomials , see also @xcite and ( * ? ? ?
* ch . 6 ) . if @xmath11 is square - free and @xmath42 , then @xmath11 is hyperbolic with respect to @xmath8 if and only if @xmath33 has @xmath1 distinct real roots for @xmath35 in a zariski - open subset of @xmath43 . in this case , if @xmath28 interlaces @xmath11 and has no common factors with @xmath11 , then @xmath28 strictly interlaces @xmath11 . in this paper
, we examine the set of polynomials in @xmath44_{d-1}$ ] interlacing a fixed hyperbolic polynomial .
the main result is a description of a hyperbolicity cone @xmath7 as a linear slice of the cone of nonnegative polynomials .
using the cone of sums of squares instead gives an inner approximation of @xmath7 by a projection of a spectrahedron .
this is closely related to recent results due to netzer and sanyal @xcite and parrilo and saunderson @xcite .
we discuss both this theorem and the resulting approximation in section [ sec : hypcone ] . in section [ sec : hermdets ] we see that the relaxation we obtain is exact if some power of @xmath11 has a definite determinantal representation . a multiaffine example for which our relaxation is not exact
is discussed in section [ sec : multiaff ] .
here we also provide a criterion to test whether or not a hyperbolic multiaffine polynomial has a definite determinantal representation .
the full cone of interlacers has a nice structure , which we discuss in section [ sec : boundary ] .
first we need to build up some basic facts about interlacing polynomials .
+ * acknowledgements .
* we would like to thank alexander barvinok , petter brndn , tim netzer , rainer sinn , and victor vinnikov for helpful discussions on the subject of this paper .
daniel plaumann was partially supported by a feodor lynen return fellowship of the alexander von humboldt - foundation .
cynthia vinzant was partially supported by the national science foundation rtg grant dms-0943832 and award dms-1204447 .
let @xmath11 be a homogeneous polynomial of degree @xmath1 that is hyperbolic with respect to the point @xmath3 .
we will always assume that @xmath45 . define @xmath46 to be the set of real polynomials of degree @xmath47 that interlace @xmath11 with respect to @xmath8 and are positive at @xmath8 : @xmath48_{d-1 } \;\colon\;\text{$g$ interlaces $ f$ with respect to } e\text { and } g(e)>0\bigr\}.\ ] ]
as noted above , the hyperbolicity cone @xmath7 depends only on @xmath11 and the connected component of @xmath49 containing @xmath8 .
in other words , we have @xmath50 for all @xmath51 .
we will see shortly that @xmath46 does not depend on @xmath8 either , but only on @xmath7 .
[ thm : interlacers ] let @xmath52_d$ ] be square - free and hyperbolic with respect to @xmath3 , where @xmath45 . for @xmath53_{d-1}$ ] , the following are equivalent : 1 .
@xmath54 ; 2 .
@xmath55 for all @xmath56 ; 3 .
@xmath57 is nonnegative on @xmath58 ; 4 .
@xmath59 is nonnegative on @xmath43 .
the proof of this theorem and an important corollary are at the end of this section .
first , we need to build up some theory about the forms in @xmath46 .
[ lem : commonfactors ] suppose @xmath60 , @xmath61 , and @xmath62 are real homogeneous polynomials
the product @xmath63 is hyperbolic with respect to @xmath8 if and only if both @xmath60 and @xmath61 are hyperbolic with respect to @xmath8 . in this case ,
if @xmath60 and @xmath61 are hyperbolic with respect to @xmath8 , then @xmath65 interlaces @xmath63 if and only if @xmath62 interlaces @xmath61 .
if @xmath62 interlaces @xmath66 for @xmath67 , then @xmath68 divides @xmath62 .
these statements are checked directly after reducing to the one - dimensional case .
[ lem : samesign ] for any @xmath28 and @xmath62 in @xmath46 , the product @xmath69 is nonnegative on @xmath58 . to prove this statement , it suffices to restrict to any line @xmath70 where @xmath5 .
suppose that @xmath71 $ ] has roots @xmath72 and @xmath32 and @xmath73 have roots @xmath74 and @xmath75 , respectively . by the assumption that both @xmath28 and @xmath62 interlace @xmath11 , we know that @xmath76 $ ] for all @xmath77 .
thus , if @xmath78 and @xmath79 are not also roots of @xmath32 or @xmath73 , the polynomial @xmath80 has an even number of roots in the interval @xmath81 $ ] .
then the sign of @xmath82 is the same for all @xmath83 for which it is not zero . because @xmath84 , we see that that sign must be nonnegative .
[ lem : nonneginterlacers ] suppose that @xmath11 is square - free and that @xmath85 _ strictly _ interlaces @xmath11 . then a polynomial @xmath86_{d-1}$ ] belongs to @xmath46 if and only if @xmath69 is nonnegative on @xmath58 .
one direction follows from lemma [ lem : samesign ] . for the other ,
let @xmath86_{d-1}$ ] for which @xmath69 is nonnegative on @xmath87 .
first , let us consider the case where @xmath11 and @xmath62 have no common factor . then , for generic @xmath88 , the roots of @xmath33 are distinct from each other and from the roots of @xmath32 and @xmath73 .
the product @xmath80 is then positive on all of the roots of @xmath33 .
since @xmath32 changes sign on consecutive roots of @xmath33 , we see that @xmath73 must have a root between each pair of consecutive roots of @xmath33 , and thus @xmath62 interlaces @xmath11 with respect to @xmath8 .
now suppose @xmath89 and @xmath90 .
we will show that @xmath91 interlaces @xmath61 , and thus @xmath62 interlaces @xmath11 .
again , we can choose generic @xmath35 for which the roots of @xmath33 and @xmath32 are all distinct .
consider two consecutive roots @xmath92 of the polynomial @xmath93 .
let @xmath94 be the number of roots of @xmath95 in the interval @xmath96 . because @xmath28 strictly interlaces @xmath89
, its restriction @xmath32 must have @xmath97 roots in the interval @xmath98 .
thus the polynomial @xmath99 has an odd number of roots in this interval and must therefore have different signs in @xmath100 and @xmath101 . since @xmath102 on @xmath103 ,
the polynomial @xmath104 must have a root in this interval .
thus @xmath91 interlaces @xmath61 and @xmath62 interlaces @xmath11 . in the above lemma
, it is indeed necessary that @xmath11 and @xmath28 be without common factors .
for example , consider @xmath105 and @xmath106 .
both @xmath11 and @xmath28 are hyperbolic with respect to @xmath107 $ ] and @xmath28 interlaces @xmath11 with respect to this point . however
if @xmath108 , then @xmath69 vanishes identically on @xmath58 but @xmath62 does not interlace @xmath11 . for @xmath51 ,
the derivative @xmath109 obviously interlaces @xmath11 with respect to @xmath35 , since @xmath11 is hyperbolic with respect to @xmath35 .
we need to show that @xmath109 also interlaces @xmath11 with respect to @xmath8 .
[ lem : main ] for @xmath51 , the polynomial @xmath110 is nonnegative on @xmath58 . for any @xmath111 and @xmath112 , let @xmath113 denote the roots of @xmath114 .
because @xmath7 is convex , the line segment joining @xmath8 and @xmath35 belongs to this cone .
as we vary @xmath115 from @xmath8 to @xmath35 along this line segment , the roots @xmath116}$ ] , form @xmath1 _ non - crossing _ arcs in the plane @xmath117 , as shown in figure [ fig : arcs ] .
since @xmath118 , one of these arcs is just the point @xmath119 .
that is , there is some @xmath94 for which @xmath120 for all @xmath115 in the convex hull of @xmath8 and @xmath35 . now @xmath45 implies @xmath121 for all @xmath111 .
thus @xmath122 is positive for @xmath123 .
furthermore , the sign of this derivative on the @xmath83th root , @xmath124 depends only on @xmath83 .
specifically , for all @xmath125 , @xmath126 in particular , the sign of @xmath127 on the @xmath94th root , @xmath128 , is constant : @xmath129 then , regardless of @xmath94 , the product @xmath130 is non - negative . .
, width=226 ] now we are ready to prove theorem [ thm : interlacers ] .
@xmath131 ( @xmath132 ) clear .
( @xmath133 ) if @xmath11 is square free , then @xmath41 strictly interlaces @xmath11 .
this equivalence then follows from lemma [ lem : nonneginterlacers ] .
( @xmath134 ) here we need a useful fact about wronskians .
the wronskian of two univariate polynomials @xmath135 is the polynomial @xmath136 it is a classical fact that if the roots of @xmath137 and @xmath138 are all distinct and interlace , then @xmath139 is a nonnegative or nonpositive polynomial @xcite .
thus if @xmath140 is coprime to @xmath11 , then for generic @xmath119 , the roots of @xmath141 and @xmath142 interlace and are all distinct .
thus their wronskian @xmath143 is either nonnegative or nonpositive for all @xmath144 .
by @xmath145 , the product @xmath146 is nonnegative on the zeroes of @xmath11 , so we see that the wronskian is nonnegative .
setting @xmath147 gives us that @xmath148 is nonnegative for all @xmath149 , as desired .
if @xmath11 and @xmath62 share a factor , say @xmath89 , @xmath90 , we can use the identity @xmath150 to reduce to the coprime case .
( @xmath151 ) because @xmath11 is square free , both @xmath152 and @xmath109 share no factors with @xmath11 .
thus @xmath41 strictly interlaces @xmath11 with respect to @xmath8 and @xmath109 strictly interlaces @xmath11 with respect to @xmath35 .
suppose @xmath62 interlaces @xmath11 with respect to @xmath35 and @xmath153 .
by lemma [ lem : nonneginterlacers ] , @xmath154 is nonnegative on @xmath58 . using lemma [ lem : main ] ,
we see that @xmath155 is also nonnegative on @xmath58 .
taking the product , it follows that @xmath156 is nonnegative on @xmath58 . because @xmath157 and @xmath11 have no common factors
, we can conclude that @xmath158 is nonnegative on @xmath58 . using lemma [ lem : nonneginterlacers ] again
, we have @xmath140 . switching the roles of @xmath35 and @xmath8 in this argument
gives the reverse implication .
[ cor : interlacersnon ] the set @xmath46 is a closed convex cone . if @xmath11 is square - free , this cone is linearly isomorphic to a section of the cone of nonnegative polynomials of degree @xmath159 : @xmath160_{\deg(f)-1 } \;:\ ; d_ef \cdot h - f \cdot d_eh \;\geq 0 \;\text { on } { { \mathbb}{r}}^{n } \}.\ ] ] if @xmath89 where @xmath161 and @xmath61 is square - free , then @xmath162 and is isomorphic to a section of the cone of nonnegative polynomials of degree @xmath163 . for square - free @xmath11 ,
the description follows directly from theorem [ thm : interlacers ] .
the map @xmath164 is a linear map from @xmath165_{\deg(f)-1}$ ] to @xmath165_{(2\deg(f)-2)}$ ] .
we see that @xmath46 is the preimage of the cone of nonnegative polynomials in @xmath165_{(2\deg(f)-2)}$ ] under this map . we can also check that this map is injective . because @xmath11 is square free , @xmath41 and @xmath11 are coprime .
hence if @xmath11 were to divide @xmath158 , then @xmath11 would have to divide @xmath62 , which it can not .
thus @xmath166 can not be identically zero . if @xmath11 is not square - free , then @xmath11 factors as @xmath63 as above . by lemma [ lem : commonfactors](c ) ,
any polynomial that interlaces @xmath11 must be divisible by @xmath60 . by part ( b )
, the remainder must interlace @xmath61 .
thus @xmath167 . similarly ,
if @xmath62 interlaces @xmath61 , then @xmath65 interlaces @xmath89 .
thus @xmath46 is the image of the convex cone @xmath168 under a linear map , namely multiplication by @xmath60 .
this shows that it is linearly isomorphic to a section of the cone of nonnegative polynomials of degree @xmath163 .
an interesting consequence of the results in the preceding section is that we can recover the hyperbolicity cone @xmath7 as a linear section of @xmath46 , and thus as a linear section of the cone of nonnegative polynomials .
we show this by considering which partial derivatives @xmath169 interlace @xmath11 . using theorem [ thm : interlacers ]
, we often have to deal with the polynomials @xmath170 notice that @xmath171 is homogeneous of degree @xmath172 , symmetric in @xmath8 and @xmath35 , and linear in each .
[ thm : nicecones ] let @xmath0_d$ ] be square - free and hyperbolic with respect to the point @xmath3 .
the intersection of @xmath46 with the plane spanned by the partial derivatives of @xmath11 is the image of @xmath173 under the linear map @xmath174 .
that is , @xmath175 furthermore , @xmath176 is a section of the cone of nonnegative polynomials of degree @xmath172 : @xmath177 let @xmath178 be the set on the right hand side of . from theorem [ thm : interlacers ] , we see that @xmath179 interlaces @xmath11 with respect to @xmath8 for all @xmath51 .
this shows @xmath180 and hence the inclusion @xmath181 , since @xmath178 is closed .
if this inclusion were strict , there would exist a point @xmath182 with @xmath183 , since @xmath178 is also a convex cone by corollary [ cor : interlacersnon ] .
thus to show the reverse inclusion , it therefore suffices to show that for any point @xmath35 outside of @xmath173 with @xmath183 , the polynomial @xmath179 does not belong to @xmath46 .
if @xmath35 belongs to @xmath184 , then @xmath185 belongs to @xmath186 . in particular , @xmath187 and @xmath179
does not belong to @xmath46 .
thus we may assume @xmath188 .
since @xmath11 is hyperbolic with respect to @xmath8 , all of the roots @xmath189 of @xmath33 are real .
the reciprocals of these roots , @xmath190 , are roots of the polynomial @xmath191 . because @xmath35 is not in @xmath192
, there is some @xmath193 for which @xmath194 . since @xmath4 and @xmath183 , we can take reciprocals to find the roots of @xmath191 : @xmath195 by rolle s theorem , the roots of @xmath196 interlace those of @xmath191 .
note that the polynomial @xmath196 is precisely @xmath197 , so the roots of @xmath197 interlace those of @xmath191 .
in particular , there is some root @xmath101 of @xmath197 in the open interval @xmath198 , and thus @xmath199 $ ] is a zero of @xmath200 .
therefore @xmath200 has only @xmath201 roots in the interval @xmath202 $ ] and thus can not interlace @xmath11 with respect to @xmath8 . combining this with theorem [ thm : interlacers ] shows the equality in . outside of the hyperbolicity cone
, @xmath109 does not interlace @xmath11 .
, width=188 ] relaxing nonnegativity to sums - of - squares in gives an inner approximation to the hyperbolicity cone of @xmath11 : [ eq : sosrelax ] & & \{a r^n : _ e , af } .
& & if the relaxation is exact , then the hyperbolicity cone is a projection of a spectrahedron , namely of a section of the cone of positive semidefinite matrices in @xmath203 , where @xmath204_{d-1}$ ] .
a polynomial @xmath205 is a sum of squares if and only if there exists a positive semidefinite matrix @xmath206 such that @xmath207 , where @xmath208 is the vector of monomials of degree at most @xmath209 .
we call such a matrix @xmath206 a _ gram matrix _ of @xmath205 .
the linear equations giving the gram matrices of @xmath171 give the desired section of @xmath203 .
if the relaxation is not exact , one can allow for denominators in the sums of squares and successively improve the relaxation .
more precisely , for any integer @xmath210 consider @xmath211 as above , @xmath212 is seen to be a projection of a spectrahedron .
furthermore , by a result of reznick in @xcite , for any positive definite form @xmath213 $ ] there exists some positive integer @xmath214 such that @xmath215 is a sum of squares .
thus if @xmath12 is smooth , then @xmath216 contains a strictly positive polynomial , for example @xmath217 .
it follows that the hyperbolicity cone @xmath173 is the closure of the union of all the cones @xmath212 . in a recent paper @xcite , netzer and
sanyal showed that the hyperbolicity cone of a hyperbolic polynomial without real singularities is the projection of a spectrahedron .
their proof uses general results on projected spectrahedra due to helton and nie and is not fully constructive .
in particular , it does not imply anything about equality in or .
explicit representations of hyperbolicity cones as projected spectrahedra have recently been obtained by parrilo and saunderson in @xcite for elementary symmetric polynomials and for directional derivatives of polynomials possessing a definite determinantal representation .
we also have the relaxation @xmath218 it is unclear whether or not this relaxation is always equal to .
its exactness would also show @xmath173 to be the projection of a spectrahedron .
we will see below that if @xmath11 has a definite determinantal representation , then we get equality in and .
consider the quadratic form @xmath219 , which is hyperbolic with respect to the point @xmath220 .
the hyperbolicity cone @xmath7 is known as the lorentz cone . in this example , the relaxation is exact .
to see this , note that @xmath221 since every nonnegative quadratic form is a sum of squares , there is equality in .
in fact , taking the gram matrix of @xmath222 , we recover the lorentz cone as @xmath223 note also that this gram matrix gives a definite determinantal representation of @xmath224 . [ ex : cubic ] consider the hyperbolic cubic polynomial @xmath225 with @xmath226 $ ] . here
the polynomial @xmath171 has degree four in @xmath227 . in this case
, the relaxation is exact , as shown in corollary [ cor : good3var ] below .
( one can also see exactness from the fact that every nonnegative ternary quartic is a sum of squares )
. using the basis @xmath228 of @xmath229_2 $ ] , we can then write the cone @xmath176 as the set of @xmath230 in @xmath231 for which there exists @xmath232 to make the real symmetric matrix @xmath233 positive semidefinite . the sums of squares relaxation is not always exact .
a counterexample comes from a multilinear hyperbolic polynomial and will be discussed in example [ ex : vamos ] .
we consider @xmath234 as a polynomial in @xmath235 $ ] , where @xmath236 is a symmetric matrix of variables . since all eigenvalues of a real symmetric matrix are real , this polynomial is hyperbolic with respect to the identity matrix .
the hyperbolicity cone @xmath237 is the cone of positive definite matrices .
hence , for any positive semidefinite matrix @xmath238 , the polynomial @xmath239 interlaces @xmath234 , where @xmath240 denotes the adjugate matrix , whose entries are the signed @xmath241-minors of @xmath242 .
this holds true when we restrict to linear subspaces . for real symmetric @xmath243 matrices @xmath244 and variables @xmath245 , denote @xmath246 if @xmath247 is positive definite for some @xmath3 , then the polynomial @xmath248 is hyperbolic with respect to the point @xmath8 . [
prop : dethyperbolic ] if @xmath249 is a real symmetric matrix of linear forms such that @xmath250 for some @xmath3 , then for any positive semidefinite matrix @xmath251 , the polynomial @xmath252 interlaces @xmath253 with respect to @xmath8 . by the discussion above
, the polynomial @xmath254 interlaces @xmath234 with respect to @xmath251 .
( in fact these are all of the interlacers of @xmath234 .
see example [ ex : sdp ] below . ) by theorem [ thm : interlacers ] , @xmath255 interlaces @xmath234 with respect to any positive definite matrix , in particular @xmath247 .
restricting to the linear space @xmath256 shows that @xmath252 interlaces @xmath253 with respect to @xmath8 .
[ thm : goodsosrelax ] if @xmath0_d$ ] has a definite symmetric determinantal representation @xmath257 with @xmath258 and @xmath259 , then @xmath171 is a sum of squares . in particular
, there is equality in . because @xmath247 and @xmath260 are positive semidefinite , we can write them as sums of rank - one matrices : @xmath261 and @xmath262 , where @xmath263 .
then @xmath264 , so @xmath265 furthermore , by proposition [ prop : hessedet ] below , the second derivative @xmath266 is @xmath267 now , again using proposition [ prop : hessedet ] , we see that @xmath268 equals @xmath269 which is the desired sum of squares .
in fact , something stronger is true .
we can also consider the case where some power of @xmath11 has a definite determinantal representation .
this is particularly interesting because taking powers of a hyperbolic polynomial does not change the hyperbolicity cone .
[ cor : power ] if @xmath270_d$ ] and a power @xmath271 has a definite symmetric determinantal representation @xmath272 with @xmath273 , then @xmath274 is a sum of squares . in particular
, there is equality in .
let @xmath271 have a definite determinantal representation .
we have @xmath275 theorem [ thm : goodsosrelax ] states that @xmath276 is a sum of squares , @xmath277 for some @xmath278 $ ] .
let @xmath137 be an irreducible factor of @xmath279 .
then @xmath137 is hyperbolic with respect to @xmath8 and the right hand side vanishes on @xmath280 .
therefore , each @xmath281 vanishes on @xmath282 and thus on @xmath283 , since @xmath282 is zariski dense in @xmath283 .
thus we can divide the @xmath281 by @xmath137 . by iterating this argument
, we get the claim .
this result is closely related to ( but does not seem to follow from ) ( * ? ? ?
1.6 ) , which says that the parametrized hermite matrix of @xmath11 is a sum of matrix squares whenever a power of @xmath11 possesses a definite determinantal representation .
[ cor : good3var ] if @xmath284 $ ] , then there is equality in . by the helton - vinnikov theorem @xcite , every hyperbolic polynomial in three variables has a definite determinantal representation .
the claim then follows from theorem [ thm : goodsosrelax ] .
the following determinantal identities were needed in the proof of theorem [ thm : goodsosrelax ] above .
[ prop : hessedet ] let @xmath242 be a @xmath243 matrix of variables @xmath285 and let @xmath286 denote @xmath287 . then for any vectors
@xmath288 we have @xmath289 in @xmath290 $ ] .
furthermore , @xmath291 we will prove the first identity using schur complements .
see , for example , @xcite .
if @xmath292 is a @xmath293 submatrix of the @xmath294 matrix @xmath295 , then its determinant equals @xmath296 .
if @xmath297 is the zero matrix , this simplifies to @xmath298 to obtain the desired identity , we set @xmath299 , @xmath300 , and @xmath301 : @xmath302 multiplying both sides by @xmath234 finishes the proof of the determinantal identity . for the claim about derivatives of the determinant , by additivity , we only need to look at the case when @xmath303 are unit vectors , @xmath304 , respectively .
then @xmath305 is the derivative of @xmath306 with respect to the entry @xmath307 .
this is the signed minor of @xmath242 obtained by removing the @xmath308th row and @xmath83th column , which is precisely the determinant @xmath309 .
taking the derivative of this determinant with respect to @xmath310 the same way gives @xmath311 we conclude this section with a general result inspired by dixon s construction of determinantal representations of plane curves , which will be applied in the next section .
if @xmath312_d$ ] has a definite determinantal representation , @xmath257 with @xmath313 , then @xmath314 is a @xmath315 matrix with entries of degree @xmath47 .
this matrix has rank at most one on @xmath316 , as seen by the identity @xmath317 . by proposition [ prop : dethyperbolic ] , the top left entry @xmath318 interlaces @xmath11 with respect to @xmath8 .
in fact , these properties of @xmath314 are enough to reconstruct a definite determinantal representation @xmath249 .
[ thm : nonvanishingdet ] let @xmath319 be a symmetric @xmath315 matrix of real forms of degree @xmath47 .
suppose that @xmath0_d$ ] is irreducible and hyperbolic with respect to @xmath320 .
if @xmath292 has rank one modulo @xmath321 , then @xmath322 divides the entries of @xmath323 and the matrix @xmath324 has linear entries .
furthermore there exists @xmath325 such that @xmath326 if @xmath327 interlaces @xmath11 with respect to @xmath8 and @xmath292 has full rank , then @xmath328 and @xmath247 is definite . by assumption
, @xmath11 divides all the @xmath329 minors of @xmath292 .
therefore , @xmath322 divides all of the @xmath330 minors of @xmath292 and thus all of the entries of the adjugate matrix @xmath323 , see ( * ? ? ?
* lemma 4.7 ) .
we can then consider the matrix @xmath331 .
by similar arguments , @xmath332 divides @xmath333 . because these both have degree @xmath334
, we conclude that @xmath335 for some @xmath336 . putting all of this together
, we find that @xmath337 so we can take @xmath338 . now
, suppose that @xmath327 interlaces @xmath11 and that @xmath339 .
then @xmath333 is identically zero . in particular , the determinant of @xmath340 is zero , there is some nonzero vector @xmath341 in its kernel , and @xmath342 is also zero .
we will show that the polynomial @xmath343 is not identically zero and that it interlaces @xmath11 with respect to @xmath8 .
this will contradict the conclusion that @xmath344 . because @xmath292 has rank one on @xmath316 , for each @xmath125 we have that @xmath345 if @xmath343 is identically zero , then @xmath346 vanishes on @xmath316 .
since @xmath346 only has degree @xmath47 , it must vanish identically as well . as this holds for each @xmath83
, this implies that @xmath347 is zero , which contradicts our assumption .
thus @xmath343 can not be identically zero .
furthermore , shows that @xmath348 is nonnegative on @xmath14 . then lemma [ lem : nonneginterlacers ] shows that @xmath343 interlaces @xmath11 with respect to @xmath8 .
in particular , @xmath343 can not vanish at the point @xmath8 .
thus the determinant of @xmath292 and hence @xmath249 can not be identically zero . thus @xmath249 is a determinantal representation of @xmath11 . to show that @xmath247 is definite , it suffices to show that @xmath340 is definite . for any vector @xmath349 , we see from with @xmath350 that @xmath351 is nonnegative on @xmath14 .
thus @xmath352 belongs to @xmath46 by lemma [ lem : nonneginterlacers ] and in particular @xmath353 is positive for all @xmath341 .
hence the matrix @xmath340 is definite .
an interesting special case of a hyperbolic polynomial is a multiaffine polynomial whose hyperbolicity cone contains the positive orthant .
these polynomials are deeply connected to the theory of matroids @xcite .
a polynomial @xmath0 $ ] is called * affine * in @xmath354 if the degree of @xmath11 in @xmath354 is at most one .
if @xmath11 is affine in each variable @xmath355 , then @xmath11 is called * multiaffine*. much of the literature on these polynomials deals with complex polynomials , rather than real polynomials , and the property of _ stability _ in place of hyperbolicity .
a polynomial @xmath356 $ ] is called * stable * if @xmath357 is non - zero whenever the imaginary part of each coordinate @xmath358 is positive for all @xmath359 .
a real homogeneous polynomial @xmath52 $ ] is stable if and only if @xmath11 is hyperbolic with respect to every point in the positive orthant . after a linear change of variables ,
every hyperbolic polynomial is stable . in 2004 ,
choe , oxley , sokal , and wagner @xcite showed that if @xmath0_d$ ] is stable , homogeneous , and multiaffine , then its _ support _ ( the collection of @xmath360 for which the monomial @xmath361 appears in @xmath11 ) is the set of bases of a matroid .
they further show that any _ representable _ matroid is the support of some stable multiaffine polynomial . in 2010
, brndn @xcite used this deep connection to disprove the generalized lax conjecture by showing that the bases - generating polynomial of the vmos matroid ( see example [ ex : vamos ] ) is hyperbolic but none of its powers has a determinantal representation .
this example will also provide a counterexample to equality in our relaxation .
the wronskian polynomials @xmath171 also played a large part in the study of multiaffine stable polynomials .
they are particularly useful when the points @xmath8 and @xmath35 are unit vectors . in this case
, we will simplify our notation and write @xmath362 using these polynomials , brndn @xcite established a necessary and sufficient condition for multiaffine polynomials to be stable .
[ thm : brand ] for multiaffine @xmath0 $ ] , the following are equivalent : 1 .
@xmath363 is nonnegative on @xmath36 for all @xmath364 , 2 .
@xmath11 is stable .
brndn also notes that the implication ( 2)@xmath365(1 ) holds for polynomials that are not multiaffine , giving an alternative proof of a part of theorem [ thm : interlacers ] above .
the other implication however , does not hold in general , as the following example shows .
let @xmath366 , @xmath367 and @xmath368 .
clearly @xmath369 is not hyperbolic with respect to any @xmath370 , but for all @xmath371 we have @xmath372 now let @xmath373 be the minimal value that @xmath374 takes on the unit sphere and let @xmath375 .
then , since @xmath376 is homogeneous , we see that @xmath376 is nonnegative on @xmath377 .
because @xmath376 is a homogeneous polynomial in two variables , it is even a sum of squares .
thus @xmath378 is a sum of squares for all @xmath379 in the positive orthant .
this also shows that the converse of corollary [ cor : power ] is not true , i.e. there is some polynomial @xmath11 such that @xmath171 is a sum of squares for all @xmath380 in some full dimensional cone , but no power of @xmath11 has a definite determinantal representation . in an analogous statement
, the polynomials @xmath381 can also be used to determine whether or not a homogeneous multiaffine stable polynomial has a definite determinantal representation .
[ thm : muaff ] let @xmath0_d$ ] be homogeneous and stable .
suppose @xmath11 is affine in the variables @xmath382 and the coefficient of @xmath383 in @xmath11 is non - zero
. then the following are equivalent : 1 .
@xmath363 is a square in @xmath165 $ ] for all @xmath384 ; 2 .
@xmath385 is a square in @xmath165/(f)$ ] for all @xmath384 ; 3 .
@xmath11 has a definite determinantal representation .
[ lem : sqfac ] let @xmath0 $ ] be affine in @xmath354 and @xmath386 for some @xmath387 . if @xmath388 with @xmath389 $ ] , then @xmath363 is a square if and only if @xmath390 and @xmath391 are squares .
suppose @xmath363 is a square .
since @xmath11 is affine in @xmath392 , both @xmath28 and @xmath62 are affine in @xmath392 and either @xmath393 or @xmath394 .
it follows that either @xmath395 or @xmath396 . using the identity @xmath397
, we see that either @xmath395 or @xmath398 in both cases @xmath390 is a square .
the same holds true for @xmath391 . for the converse , suppose that @xmath390 and @xmath391 are squares . as we saw above ,
one of them is zero .
thus @xmath399 or @xmath400 . ( @xmath401 )
( @xmath402 ) for a start , suppose that @xmath11 is irreducible .
we will construct a matrix @xmath292 satisfying the hypotheses of theorem [ thm : nonvanishingdet ] . for every @xmath403 , the polynomial @xmath385 is equivalent to a square @xmath404 modulo @xmath321 . in the case
@xmath405 we can choose @xmath406 . then it is easy to check that @xmath407 equals @xmath408 modulo @xmath321 .
further , for every @xmath409 , the polynomials @xmath410 and @xmath411 are equivalent modulo @xmath11 . after changing the sign of @xmath412 if necessary ,
we see that @xmath413 equals @xmath414 modulo @xmath321 . because @xmath11 is irreducible
, it follows that the symmetric matrix @xmath415 has rank one on @xmath316 .
we now need to show that @xmath292 has full rank . for each @xmath416 , consider the point @xmath417\backslash\{k\}}e_j$ ] , which lies in the real variety of @xmath11 . for @xmath418
, we see that @xmath419 vanishes at @xmath420 , and therefore so must @xmath421 . on the other hand
, @xmath422 equals the nonzero coefficient of @xmath423 in @xmath11 .
now suppose that @xmath424 for some @xmath341 .
the @xmath94th row of this is @xmath425 . plugging in the point @xmath420
then shows that @xmath426 must be zero , and thus @xmath208 is the zero vector . since @xmath11 is stable , @xmath427 interlaces it , and so by theorem [ thm : nonvanishingdet ]
, @xmath11 has a definite determinantal representation .
if @xmath11 is reducible and @xmath28 is an irreducible factor of @xmath11 , then , by lemma [ lem : sqfac ] , @xmath390 is a square . since every irreducible factor of @xmath11 has a definite determinantal representation , so has @xmath11 .
( @xmath428 ) let @xmath429 where @xmath244 are real symmetric @xmath243 matrices where @xmath430 . because @xmath11 is affine in each of the variables @xmath431
, the matrices @xmath432 must have rank one .
furthermore , since @xmath11 is stable , these rank - one matrices must be positive semidefinite ( see @xcite , proof of theorem 2.2 ) .
thus we can write @xmath433 , with @xmath434 for each @xmath435 . then by and proposition [ prop : hessedet ] , we have @xmath436 for @xmath384 .
let @xmath0 $ ] be homogeneous , stable and multiaffine .
then the following are equivalent : 1 .
@xmath363 is a square for all @xmath364 ; 2 .
@xmath385 is a square in @xmath165/(f)$ ] for all @xmath364 ; 3 .
@xmath11 has a definite determinantal representation .
this is an immediate consequence of the preceding theorem .
let @xmath437 and let @xmath0 $ ] be a multiaffine stable polynomial . if @xmath11 has a definite determinantal representation , then @xmath438 and @xmath439 also have a definite determinantal representation .
let @xmath440 , @xmath441 and @xmath442 .
wagner and wei @xcite calculated @xmath443 where @xmath444 $ ] do not depend on @xmath445 .
since @xmath363 is a square , @xmath390 and @xmath391 are squares as well .
thus @xmath28 and @xmath62 have a definite determinantal representation .
let @xmath388 , where @xmath446 $ ] are multiaffine stable polynomials .
then @xmath11 has a definite determinantal representation if and only if both @xmath28 and @xmath62 do .
this follows directly from lemma [ lem : sqfac ] and theorem [ thm : muaff ] .
let @xmath447 $ ] be the elementary symmetric polynomial of degree @xmath1 .
we have @xmath448 , @xmath449 and @xmath450 for all @xmath451 .
it is a classical result that these are the only cases where @xmath452 has a definite determinantal representation @xcite .
indeed , for @xmath453 and @xmath454 the coefficients of the monomials @xmath455 , @xmath456 and @xmath457 in @xmath458 are all @xmath459 . specializing to @xmath460 for @xmath461 then shows that @xmath462 is not a square .
[ ex : vamos ] the relaxations and are not always exact .
an example of this comes from the multiaffine quartic polynomial in @xmath463_4 $ ] given as the bases - generating polynomial of the vmos matroid : @xmath464}{4 } \backslash c } \prod_{i\in i } x_i,\ ] ] where @xmath465 .
wagner and wei @xcite have shown that the polynomial @xmath62 is stable , using an improved version of theorem [ thm : brand ] and representing @xmath466 as a sum of squares .
but it turns out that @xmath467 is not a sum of squares .
because the cone of sums of squares is closed , it follows that for some @xmath468 in the hyperbolicity cone of @xmath62 , the polynomial @xmath469 is not a sum of squares . in order to show that @xmath467 is not a sum of squares , it suffices to restrict to the subspace @xmath470 and
show that the resulting polynomial @xmath471 is not a sum of squares .
this restriction is given by @xmath472 this polynomial vanishes at six points in @xmath473 , @xmath474 , \;[0:1:0],\ ; [ 0:0:1],\;[1:-1:0],\;[1:0:-1 ] , \text { and } \;[0:1:-1].\ ] ] thus if @xmath475 is written as a sum of squares @xmath476 , then each @xmath477 must vanish at each of these six points .
the subspace of @xmath229_3 $ ] of cubics vanishing in these six points is four dimensional and spanned by @xmath478 .
then @xmath475 is a sum of squares if and only if there exists a positive semidefinite @xmath479 matrix @xmath206 such that @xmath480 .
however , the resulting linear equations in the variables @xmath481 , have the unique solution @xmath482 one can see that @xmath206 is not positive semidefinite from its determinant , which is @xmath483 .
thus @xmath475 can not be written as a sum of squares .
this , along with corollary [ cor : power ] , provides another proof that no power of the vmos polynomial @xmath484 has a definite determinantal representation .
the polynomial @xmath467 is also not a sum of squares modulo the ideal @xmath485 . to see this ,
suppose @xmath486 for some @xmath487 $ ] and consider the terms with largest degree in @xmath488 and @xmath489 in this expression .
writing @xmath490 where @xmath491 lie in @xmath492 $ ] , we see that the leading form @xmath493 is _ real radical _ , meaning that whenever a sum of squares @xmath494 lies in the ideal @xmath495 , this ideal contains each polynomial @xmath281 .
since @xmath467 does not involve the variables @xmath488 and @xmath489 , we can then reduce the polynomials @xmath496 modulo the ideal @xmath485 so that they do not contain the variables @xmath497 . see ( * ? ? ?
* lemma 3.4 ) . because @xmath62 does involve the variable @xmath488 and @xmath489 , this results in a representation of @xmath467 as a sum of squares , which is impossible , as we have just seen .
here we investigate the convex cone @xmath46 of polynomials interlacing @xmath11 .
we compute this cone in two examples coming from optimization and discuss its _ algebraic boundary _ , the minimal polynomial vanishing on the boundary of @xmath46 , when this cone is full dimensional . for smooth polynomials , this algebraic boundary is irreducible .
if the real variety of a hyperbolic polynomial @xmath11 is smooth , then the cone @xmath46 of interlacers is full dimensional in @xmath165_{d-1}$ ] . on the other hand , if @xmath316 has a real singular point , then every polynomial that interlaces @xmath11 must pass through this point .
this has two interesting consequences for the hyperbolic polynomials coming from linear programming and semidefinite programming .
[ ex : lp ] consider @xmath498 .
the singular locus of @xmath316 consists of the set of vectors with two or more zero coordinates .
the subspace of polynomials in @xmath44_{n-1}$ ] vanishing in these points is spanned by the @xmath499 polynomials @xmath500 .
note that this is exactly the linear space spanned by the partial derivatives of @xmath11 .
theorem [ thm : nicecones ] then shows that the cone of interlacers is isomorphic to @xmath501 : @xmath502 [ ex : sdp ] let @xmath503 where @xmath242 is a @xmath243 symmetric matrix of variables .
the singular locus of @xmath316 is the locus of matrices with rank @xmath504 .
the corresponding ideal is generated by the @xmath330 minors of @xmath242 .
since these have degree @xmath47 , we see that the polynomials interlacing @xmath234 must lie in the linear span of the @xmath330 minors of @xmath242 . again , this is exactly the linear span of the directional derivatives @xmath505 .
thus theorem [ thm : nicecones ] identifies @xmath46 with the cone of positive semidefinite matrices : @xmath506 if @xmath14 is nonsingular , then the cone @xmath46 is full dimensional and its algebraic boundary is a hypersurface in @xmath165_{d-1}$ ] .
we see that any polynomial @xmath28 on the boundary of @xmath46 must have a non - transverse intersection point with @xmath11 .
as we see in the next theorem , this algebraic condition exactly characterizes the algebraic boundary of @xmath46 .
let @xmath52_d$ ] be hyperbolic with respect to @xmath507 and assume that the projective variety @xmath316 is smooth .
then the algebraic boundary of the convex cone @xmath46 is the irreducible hypersurface in @xmath508_{d-1}$ ] given by @xmath509_{d-1 } \;:\ ; \exists \ ; p\in{{\mathbb}{p}}^{n-1 } \text { such that } f(p)=g(p)=0 \text { and } \operatorname*{rank}\begin{pmatrix } \nabla f ( p ) \\
( p ) \end{pmatrix } \leq 1\right\}.\ ] ] first , we show that the set is irreducible .
consider the incidence variety @xmath242 of polynomials @xmath28 and points @xmath137 satisfying this condition , @xmath510_{d-1})\times{{\mathbb}{p}}^{n-1 } \;:\;f(p)=g(p)=0 \text { and } \operatorname*{rank}\begin{pmatrix } \nabla f ( p ) \\
\nabla g ( p ) \end{pmatrix } \leq 1\right\}.\ ] ] the projection @xmath511 onto the second factor is @xmath316 in @xmath512 . note that the fibres of @xmath511 are linear spaces in @xmath513_{d-1})$ ] of constant dimension .
in particular , all fibres of @xmath511 are irreducible of the same dimension .
since @xmath242 and @xmath316 are projective and the latter is irreducible , this implies that @xmath242 is irreducible ( see ( * ? ? ?
* , thm .
8) ) , so its projection @xmath514 onto the first factor , which is our desired set , is also irreducible .
if @xmath316 is smooth , then by ( * ? ? ?
* lemma 2.4 ) , @xmath11 and @xmath41 share no real roots .
this shows that the set of polynomials @xmath515_{d-1}$ ] for which @xmath516 is strictly positive on @xmath14 is nonempty , as it contains @xmath41 itself .
this set is open and contained in @xmath46 , so @xmath46 is full dimensional in @xmath165_{d-1}$ ] .
thus its algebraic boundary @xmath517 is a hypersurface in @xmath508_{d-1}$ ] .
to finish the proof , we just need to show that this hypersurface is contained in , since the latter is irreducible . to see this ,
suppose that @xmath518_{d-1}$ ] lies in the boundary of @xmath46 . by theorem [ thm : interlacers ]
, there is some point @xmath519 at which @xmath520 is zero .
as @xmath11 is nonsingular , @xmath521 can not be zero , again using ( * ? ? ?
* lemma 2.4 ) . thus @xmath522 .
moreover , the polynomial @xmath523 is globally nonnegative , so its gradient also vanishes at the point @xmath137 . as @xmath524 , this means that @xmath525 .
thus the pair @xmath526 belongs to @xmath242 above .
consider the two hyperbolic quartic polynomials @xmath527 whose real varieties are shown in figure [ fig : singularquartics ] in the plane @xmath528 .
both are hyperbolic with respect to the point @xmath529 $ ] and singular at @xmath107 $ ] .
every polynomial interlacing either of these must pass through the point @xmath107 $ ] . however ,
for a polynomial @xmath28 to interlace @xmath61 , its partial derivative @xmath530 must also vanish at @xmath107 $ ] .
thus @xmath531 has codimension one in @xmath229_3 $ ] whereas @xmath168 has codimension two .
theorem [ thm : nicecones ] states that @xmath176 is a linear slice of the cone @xmath46 . by taking boundaries of these cones
, we recover @xmath316 as a linear slice of the algebraic boundary of @xmath46 . for non - cylindrical @xmath11 ,
the map @xmath534_{d-1}$ ] given by @xmath535 is injective and maps the boundary of @xmath176 into the boundary of @xmath536 .
if @xmath11 is irreducible , this map identifies @xmath537 with a component of the zariski closure of the boundary of @xmath46 in the plane spanned by @xmath538 . since @xmath11 is not cylindrical , the @xmath499 partial derivatives @xmath539 are linearly independent , so that @xmath540 is injective
the claim now follows from taking the boundaries of the cones in .
if @xmath11 is irreducible , then the zariski closure of the boundary of @xmath176 is @xmath537 .
we take the cubic form @xmath541 , which is hyperbolic with respect to the point @xmath542 $ ] . using the computer algebra system
macaulay 2 , we can calculate the minimal polynomial in @xmath543 $ ] that vanishes on the boundary of the cone @xmath46 .
note that any conic @xmath544 on the boundary of @xmath46 must have a singular intersection point with @xmath316 .
saturating with the ideal @xmath545 and eliminating the variables @xmath227 from the ideal @xmath546 gives an irreducible polynomial of degree twelve in the six coefficients of @xmath138 .
this hypersurface is the algebraic boundary of @xmath46 .
when we restrict to the three - dimensional subspace given by @xmath547 , this polynomial of degree twelve factors as @xmath548 one might hope that @xmath46 is also a hyperbolicity cone of some hyperbolic polynomial , but we see that this is not the case . restricting to @xmath549 shows that the polynomial above , unlike @xmath11 , is not hyperbolic with respect to @xmath542 $ ] . | hyperbolic polynomials are real polynomials whose real hypersurfaces are maximally nested ovaloids , the innermost of which is convex .
these polynomials appear in many areas of mathematics , including optimization , combinatorics and differential equations . here
we investigate the special connection between a hyperbolic polynomial and the set of polynomials that interlace it .
this set of interlacers is a convex cone , which we write as a linear slice of the cone of nonnegative polynomials . in particular , this allows us to realize any hyperbolicity cone as a slice of the cone of nonnegative polynomials . using a sums of squares relaxation
, we then approximate a hyperbolicity cone by the projection of a spectrahedron .
a multiaffine example coming from the vmos matroid shows that this relaxation is not always exact .
using this theory , we characterize the real stable multiaffine polynomials that have a definite determinantal representation and construct one when it exists . |
the iras revised bright galaxy sample by sanders et al .
( 2003 ) includes all galaxies with total 60 @xmath6 flux density greater than 5.24 jy .
the sample is the result of a highly complete flux - limited survey conducted by iras covering the entire sky at galactic latitudes @xmath7 and was compiled after the final calibration of the iras level 1 archive .
it offers far more accurate and consistent measurements of the flux of objects with extended emission .
in addition , the infrared fluxes of over 100 sources from the sample were recalculated by the iras high resolution ( hires ) processing , which allowed the deconvolution of close galaxy pairs ( surace , sanders , mazzarela 2004 ) .
the latter provides a more - than - ever reliable database of the iras galaxies which can be proved crucial for statistical studies like this one . while the relation between ultra luminous iras galaxies ( ulirgs ) and strong interactions has been thoroughly studied ( e.g. sanders , surace & ishida 1999 , wang et al .
2006 ) , this is not the case for the environment of moderately and low luminous infrared galaxies . a 2-dimensional analysis of krongold et al .
( 2002 ) showed a trend for a bright iras galaxy ( birg ) sample on having neighbors in excess of normal galaxies and sy1 galaxies , but in relative agreement with sy2 galaxies .
however , the birg population consists of various types of active galaxies , including starbursts ( the majority ) , seyferts , liners and normal galaxies and thus it would be of great interest to clarify the connection between infrared emission , interactions and different types of active galaxies . during the last decade , many studies have investigated the relation among interacting galaxies , starbursting and nuclear activity ( eg .
hernandez - toledo et al .
2001 ; ho 2005 ) . despite the plethora of available information , searches for correlations between the above physical processes are inconclusive , the only exception being the coupling between interactions and starbursting
however , there is evidence that agn galaxies host a post - starburst stellar population ( eg .
boisson et al .
2000 , gonzlez delgado et al .
2001 ) while kauffmann et al .
( 2003 ) showed that the fraction of post - starburst stars increases with agn emission .
proving a relation of this type between starburst and agn galaxies would simultaneously solve also the problem of the agn triggering mechanism .
interactions would be the main cause of such activities , being starbursting and/or the feeding of a central black hole . however , this is not a trivial task .
the main difficulty arises from the fact that the star formation rate ( sfr ) estimation in agn host galaxies is still problematic .
all sfr estimation methods present complications and even those based on the fir continuum are doubtful , since the contribution of the active nuclei is unknown ( eg . ho 2005 ) . despite the difficulties , some studies , based on different diagnostics seem to conclude that there is indeed an evolutionary sequence from starburst to type 2 and then to type 1 agn galaxies ( e.g. oliva et al .
1999 , krongold et al .
in addition , kim , ho and i m ( 2006 ) , using the [ oii ] emission line as a sfr indicator , reach the conclusion that type 2 are the precursors of type 1 quasars supporting the previous claims .
these studies are based on the observed differences between different types of agns and resemblance of type 2 objects to starbursts .
this raises doubts about the simplest version of the unification scheme of agns .
it is true that the recent discovery of 10@xmath6 silicate emission in two luminous quasars implies the presence of dust , but it is not clear yet what is the spatial distribution of this material ( siebenmorgen et al .
in addition , silicate emission is not yet detected in other type 1 objects and thus more observations are needed to establish the existence of the dusty torus .
we can summarize all the previous in two statements : ( 1 ) the starburst - agn connection is still not well established and ( 2 ) the agn unification model , although successful in interpreting many observational facts , remains fragile . from our point of view , our birg sample offers a homogeneous and complete database , which is ideal for a statistical study on these issues
. we will discuss our galaxy samples in section 2 .
our data analysis and results will be presented in 3 , while in section 4 we will discuss our results and present our conclusions . due to the fact that all our samples are local , cosmological corrections of galaxy distances are negligible . throughout our paper
we use @xmath8 mpc .
the bright iras sample consists of 87 objects with redshifts between 0.008 and 0.018 and was compiled from the birg survey by soifer et al .
( 1989 ) for the northern hemisphere and by sanders et al .
( 1995 ) for the southern .
it includes only high galactic latitude objects ( @xmath9 ) in order to avoid extinction and confusion with galactic stars .
all objects lay in the luminosity range of @xmath10 .
this sample is volume limited and a @xmath11 test gives a value of @xmath12 . since the birg survey is highly complete , this sample is expected to be as well .
more details about the sample selection are given in krongold et al .
in addition we have refined the bright iras sample by correcting the infrared fluxes using the iras revised bright galaxy sample " by sanders et al .
furthermore for interacting galaxies we used the corrected fluxes given by surace , sanders and mazzarella ( 2004 ) .
we also use the control sample , compiled by krongold et al .
( 2002 ) in such a way as to reproduce the main characteristics , other than the infrared emission , of the bright iras sample .
specifically , the control sample was compiled from the original cfa catalog to reproduce closely the redshift , morphological type and diameter size distributions of the corresponding iras sample .
in other words , the selection of the iras sample and its corresponding control sample is exactly the same , the only difference being the infrared luminosity .
this is very important in order to validate that any possible environmental effect is related to the mechanisms that produce the observed high infrared luminosity and not to possible differences in the host galaxies or sample biases . in table 1
we present the names , celestial coordinates , zwicky magnitudes , redshifts nearest neighbor projected linear distance and spectral types of our final list of bright iras galaxies . in order to investigate the local and large scale environment around our birg and control sample galaxies we use the cfa2 and ssrs galaxy catalogues which cover a large solid angle of the sky .
although these galaxy catalogues date from the 80 s and 90 s they still provide an important database for studies of the properties of galaxies and their large - scale distribution in the nearby universe . we briefly present the main characteristics of these catalogues . the cfa2 redshift catalog contains approximately 18000 galaxy redshifts in the northern sky down to a magnitude limit of @xmath1315.5 ( huchra 1990 ) .
the magnitude system used is the merging of the original zwicky magnitudes and the more accurate rc1 @xmath14 magnitudes .
these exhibit a scatter of @xmath15 mags ( eg .
bothun & cornell 1990 ) .
following huchra ( 1990 ) , we do not attempt to translate these magnitudes to a standard photometric system since this requires accurate knowledge of the morphological type and size of each individual galaxy . the ssrs catalog ( da costa et al .
1998 ) contains redshifts , @xmath16 magnitudes and morphological classifications for @xmath175400 galaxies in two regions covering a total of 1.70 steradians in the southern celestial hemisphere and it is more than 99% complete down to @xmath18 = 15.5 .
the galaxies have positions accurate to about 1 arcsec and magnitudes with an rms scatter of about 0.3 mag .
the radial velocity precision is of @xmath17 40 km / s .
note that in the regions covered by the ssrs and cfa2 catalogues , only a subsample of the original birgs and their control samples can be found ( 76 bright iras galaxies and 61 control galaxies ) . in order to test whether these subsamples are statistically equivalent with their parent samples ( ie .
, their diameter , morphological type and redshift distributions ) we used the kolmogorov - smirnov two - sample test .
we verified that the null hypothesis , the subsamples being equivalent with their parent samples , can not be rejected at any significant statistical level . in order to cover a larger magnitude difference between the birgs and their nearest neighbor than that imposed by the cfa2/ssrs magnitude limit ( @xmath19 ) we have obtained our own spectroscopic observations of fainter neighbors around a subsample of our birgs , consisting of 24 galaxies ( selected randomly from their parent sample ) . around each birg we have obtained spectra of all neighboring galaxies within a projected radius of 100 @xmath2 kpc and a magnitude limit of @xmath20 .
our aim with this new fainter neighbor search is not to establish or not the existence of close neighbors around the birgs .
this will be done by using the brighter ssrs and cfa2 catalogues , at the magnitude limit of which we have well defined control samples .
what we seek with these observations is to facilitate a comparison with a similar analysis of seyfert galaxies by koulouridis et al .
( 2006 ) , in which sy2 s were found to have significantly higher fraction of neighbors with respect to sy1 s . in other words
we wish to establish whether the fractional differences in - between the sy1 , sy2 and birg samples of galaxies , already determined ( or not ) as significant with respect to their control samples , continue to fainter magnitudes .
optical spectroscopy was carried out using the faint object spectrograph and camera ( lfosc ) ( zickgraf et al . 1997 ) mounted on the 2.1 m guillermo haro telescope in cananea , operated by the national institute of astrophysics , optics and electronics ( inaoe ) of mexico .
a setup covering the spectral range @xmath21 with a dispersion of 8.2 / pix was adopted .
the effective instrumental spectral resolution was about 18 .
the data reduction was done using the iraf packages and included bias and flat field corrections , cosmic ray cleaning , wavelength linearization , and flux transformation . in table 2
we present the birg name , coordinates , redshift and magnitude for this subsample of birgs . below the row of each birg
we list the corresponding data for all its neighbors , within a projected separation of 100 @xmath2 kpc . since zwicky magnitudes were not available for the fainter neighbors and in order to provide a homogeneous magnitude system for all the galaxies we decided to list in table 2 the @xmath22 magnitudes ( blue ) poss i plate magnitudes of the minnesota automated plate scanner ( maps ) system . ] for all galaxies , being the central birg or their neighbors ( see http://aps.umn.edu/docs/photometry ) .
the neighbor measured redshift is presented in the fifth column ( while in some very few cases we list the redshift from the ned ) .
the uncertainties listed are estimated from the redshift differences which result from using different emission lines .
we search for the nearest neighbor around each birg and control galaxy in our samples with the aim of estimating the fraction of birg and normal galaxies that have a close neighbor .
to define the neighborhood search we use two parameters , the projected linear distance ( @xmath23 ) and the radial velocity separation ( @xmath24 ) between the central birg and the neighboring galaxies found in the cfa2 and ssrs catalogues or in our own spectroscopic observations .
we search for neighbors with @xmath25 600 km / s , which is roughly the mean galaxy pairwise velocity of the cfa2 and ssrs galaxies or about twice the mean pairwise galaxy velocity when clusters of galaxies are excluded ( marzke et al .
note however that our results remain robust even for @xmath25 1000 km / s .
we then define the fraction of birg and normal galaxies that have their nearest neighbor within the selected @xmath24 separation , as a function of increasing @xmath23 . in figure 1 ( upper panels ) we plot the fraction of birg and control galaxies as a function of the projected distance ( @xmath23 ) of the first companion and for two velocity separations ( @xmath27 km / s and @xmath28 km / s ) . for comparison
we also plot the results of a similar analysis by koulouridis et al .
( 2006 ) concerning seyfert galaxies and their control samples .
it is evident that a significantly higher fraction of birg galaxies have a near neighbor within @xmath29 h@xmath30 kpc with respect to their control sample . in koulouridis
et al . ( 2006 ) we found that there is a significantly higher fraction of sy2 galaxies ( @xmath17 27% ) having a near neighbor within @xmath31 h@xmath30 kpc with respect to both their control sample and the sy1 galaxies ( @xmath17 14% ) . adding here the birg sample , which includes mostly starburst and sy2 galaxies ( see table 1 ) , we can clearly see that an even higher fraction of birgs ( @xmath17 42% ) tend to have a close companion within @xmath31 h@xmath30 kpc .
the latter needs a further explanation since it is not consistent with most starburst - agn connection scenarios , which suggest a simultaneous creation of starburst and sy2 nuclei triggered by interactions . in order to investigate whether fainter neighbors , than those found in the relatively shallow cfa2 and
ssrs catalogues , exist around our birgs , we have performed a spectroscopic survey of all neighbors with @xmath32 ( @xmath333 magnitudes fainter than the cfa2 and ssrs limits ) .
this limit translates into an absolute magnitude limit of [email protected] for the most distant objects in our sample ( z@xmath35 ) .
this magnitude is fainter even than that of the small magellanic cloud .
here we present results of our spectroscopic observations of all the neighbors with @xmath36 kpc and @xmath37 for the subsample of 24 birg galaxies ( see table 2 ) .
we find that in total 13 out of the 24 birgs have at least one close neighbor within the above limits , with 9 of these having neighbors already detected in the ssrs and cfa2 catalogues , ie .
, only 4 birgs have fainter than @xmath38 neighbors .
this implies that the birgs having a close neighbor ( within @xmath36 kpc and for @xmath39 km / s ) increases only by @xmath1745% when going fainter .
koulouridis et al .
( 2006 ) showed that the percentage of both sy1 and sy2 galaxies that have a close neighbor ( within the above limits ) increases correspondingly by about 100% when we descent from @xmath40 to @xmath37 ( but remember that the host galaxies have magnitudes slightly closer to the cfa2 and ssrs limit ) . in detail , while the percentage of sy1 and sy2 galaxies having a close neighbor increases from 14% to 27% and from 27% to 55% respectively , for birgs it increases from 42% to 54% , reaching the equivalent sy2 levels .
we summarize that birgs , with respect to their control sample , show an excess of close neighbors which therefore should be responsible for their excess fir emission .
these results confirm a previous 2-dimensional analysis of krongold et al .
( 2002 ) of the same birg sample .
since the fractions of both birgs and sy2s that have a close neighbor is roughly the same , an interesting question is whether there are any magnitude differences between hosts and neighbors for the birgs and the sy2s . in figure 2
we present the distribution of such magnitude differences ( @xmath41 ) between hosts and nearest neighbor for the birgs and the sy2s . although there appears to be a slight preference for brighter neighbors of the birgs with respect to the sy2s , the two distributions are statistically equivalent , as quantified by a k - s test which gives a probability of them being drawn from the same parent population of @xmath42 . here
we investigate whether there are differences in the large scale environment of birgs with respect to their control galaxies and to the sy1 and sy2 samples of koulouridis et al .
( 2006 ) . to this end
we determine the galaxy overdensity , based on the cfa2 and ssrs catalogues , in a region around each birg or control sample galaxy .
we count all neighboring galaxies around each galaxy within a projected radius of 1 @xmath2 mpc , while to take into account the galaxy peculiar velocities , we use a radial velocity separation of @xmath43 km / s
. we estimate the expected cfa2 and ssrs field galaxy density , @xmath44 , at the distance of each galaxy by integrating the corresponding cfa2 or ssrs luminosity function ( marzke , huchra & geller 1994 ; da costa et al .
1994 ) using as a lower integration limit the minimum galaxy luminosity that corresponds to the galaxy catalogue magnitude limit ( ie . ,
@xmath45 ) at that distance .
we then compute the local overdensity around each agn , within the previously mentioned cylinder , which is given by : @xmath46 , where @xmath47 with @xmath48 the number of neighbors and @xmath49 the corresponding volume of the cylinder . in figure 3
we plot the overdensity frequency distribution for the birgs ( left panel ) and the corresponding distribution of their control galaxy sample . for comparison
, we also plot the distributions for the seyfert galaxies of koulouridis et al .
a kolmogorov - smirnov test shows that there is no statistically significant difference between any active galaxy sample ( birg or seyfert ) and their respective control sample distribution . however , there is a statistically significant difference , at a 0.03 and 0.09 level , between the overdensity distributions of birgs - sy1s and sy1s - sy2s , respectively .
similar , differences are also found between their respective control samples . on the other hand the corresponding distributions of birgs and sy2s ( and of their control samples )
are statistically equivalent at a 0.9 level .
this implies that the large scale environment of birgs and sy2s is similar but significantly different to that of sy1s , a difference which since it is also seen in their corresponding control samples should be attributed to differences of the host galaxies .
indeed , sy1 hosts are earlier type galaxies ( e.g. , hunt & malkan 1999 , koulouridis et al .
2006 ) which are known to be more clustered than late types ( e.g. , willmer et al . 1998 ) . in these section
we investigate whether there is relation between the strength of the interaction of birgs with their closest neighbor and their fir characteristics .
the strength of any interaction could be parametrized as a function of the distance between the birg and its first neighbor . at this first order analysis
we do not take into account the magnitude difference between birgs and their close neighbor , which as we have shown previously ( see figure 2 ) , does not appear to be significantly differ from that of sy2 galaxies .
we divided the interactions in our sample into three categories based on the proximity of the first neighbor .
we consider strong interactions when @xmath50 , weak interactions when @xmath51 and no interaction when @xmath52 . in figure 4 we present the color - color diagram of @xmath53 versus @xmath54 , where @xmath55 is the spectral index defined as @xmath56 with @xmath57 the flux in janskys at wavelength @xmath58 .
we can clearly see the differences between the fir characteristics among different types of galaxies and different interaction strengths .
the different interaction strengths are coded by different types of symbols while the different activity is color coded as indicated in the caption of the figure .
we cross - identified classifications of each birg combining various studies like : the pico dos dias survey ( coziol et al . 1998 ) , optical spectroscopy of luminous infrared galaxies ( veilleux et al .
1997 ; ho , filippenko , sargent 1995 ) , cola radio and spectroscopic diagnostics of nuclear activity in galaxies ( corbett et al .
2002 ) , warm iras sources ( de grijp et al . 1987 ) and , when available , sdss spectroscopic data .
it is evident that the fir characteristics of starburst galaxies in our birg sample differ significantly depending on the strength of the interaction .
the majority of highly interacting starburst have @xmath53 spectral indices greater than -2 , while all , except one , non - interacting starbursts have less .
we also find that normal galaxies and liners are at the lower end of this sequence .
the only highly - interacting starburst galaxy below @xmath59 is ngc7541 which happens to have more than two times the typical molecular gas mass of birgs ( mirabel & sanders 1988 ) .
the difference between highly interacting and non - interacting starburst galaxies , as quantified by a k - s test , is significant at a @xmath60 level when comparing their @xmath53 index distribution .
however , sy2 galaxies , interacting or no , seem to lay in the same area ( @xmath61 ) with non - interacting starburst galaxies , delineated in figure 4 by blue dashed lines .
the fir color analysis of our sample strengthens our previous results .
it clearly shows that the starburst activity is higher when interactions are stronger and ceases when the interacting neighboring galaxy moves away . while the starburst activity weakens ( if we link position on the plot with time ) sy2 nuclei appear , giving further evidence on the causal bridging between these objects .
we have compared the 3-dimensional environment of a sample of local birgs with that of a well defined control sample , selected in such a way as to reproduce the redshift , morphological type and diameter size distributions of the birg sample .
we searched for close neighbors around each birg and control sample galaxy using the distribution of cfa2 and ssrs galaxy catalogues as well as our own spectroscopic observations reaching a fainter magnitude limit ( but for a restricted birg subsample ) .
we also compared our results with those of a similar analysis of seyfert galaxies ( dultzin - hacyan et al .
1999 ; koulouridis et al .
2006 ) .
we find that the fraction of birg galaxies having a close neighbor , within a projected separation of 75 @xmath2 kpc and radial velocity difference of @xmath39 km / s , is significantly higher than the corresponding fraction of its control sample and that of sy1 galaxies while it is comparable to that of sy2 galaxies .
this result is in accordance with some previous two - dimensional studies ( eg .
krongold et al .
we reach similar results regarding the large scale environment of birgs ( within a projected radial separation of 1 @xmath2 mpc and a radial velocity difference @xmath43 km / s ) .
once more their behavior resembles that of sy2 s but not of sy1 s .
we also find a statistically significant difference between highly - interacting and non - interacting birgs , based on their fir color properties .
sy2s appear to display a similar behavior to that of non - interacting starburst galaxies , introducing new evidence for the starburst / agn connection scenarios .
our results can be accommodated in a simple evolutionary scenario , starting with an interaction , and ending in a sy1 phase .
first , close interactions would drive molecular clouds towards the central area , creating a circumnuclear starburst .
then , material could fall even further into the innermost regions of the galaxy , feeding the black hole , and giving birth to an agn which at first can not be observed due to obscuration . at this stage
only a starburst would be observed . as starburst activity relaxes and obscuration decreases , a sy2 nucleus
would be revealed ( still obscured by the molecular clouds from all viewing angles ) . as a final stage
, a sy1 phase could appear . in this case , the molecular clouds , initially in a spheroidal distribution , could flatten and form a `` torus '' ( as in the unification scheme for seyferts ) . as more material is accreted , it is possible that the agn strengthens driving away most of the obscuring clouds , and leaving a `` naked '' sy1 nuclei .
if indeed interactions play a role in triggering activity , as suggested by the above picture , then the lack of close companions among sy1 galaxies implies that the time needed for type 1 activity to appear should be larger than the timescale for an unbound companion to escape from the close environment , or comparable to the timescale needed for an evolved merger ( @xmath62 years , see krongold et al .
2002 ) .
it should be noted that the evolutionary scenario does not contradict the unification scheme .
it implies that sy1s and sy2s are the same objects ( as the unification model proposes ) but not necessarily at the same evolutionary phase .
however , there could be a phase where only orientation could define if an object appears as a sy1 or as a sy2 , which is the stage where molecular clouds form a torus but have not been swept away yet .
evidently , more detailed observations of large samples of galaxies are needed to resolve this important issue .
however , such a picture is consistent with the evolutionary scenario suggested by tran ( 2003 ) .
he studied a sample of sy2 in polarized light , and found that only 50% of them showed the presence of a hidden broad line region ( hblr ) .
he suggested that non - hblr sy2 galaxies could evolve into hblr sy2 galaxies . in this case , the appearance of the blr could be related to the accretion rate ( e.g. nicastro 2000 ) , and thus with the evolutionary stage of the object .
the trend is also consistent with the finding that 50% of sy2 galaxies also show the presence of a strong starburst ( gu et al .
2001 ; cid - fernandez et al .
2001 ) .
on the brightest end , there is also growing evidence showing ( a ) that ulirgs can be the precursors of quasars and ( b ) ulirgs are found in very strong interacting systems or in mergers ( e.g. sanders , surace and ishida 1999 ; wang et al .
therefore , the evolutionary sequence proposed above could be general to all nuclear activity and independent on luminosity ( we note that krongold et al .
2003 suggested a similar scheme for liners that can be considered as the very low luminosity extension of the evolutionary model suggested here ) .
further evidence comes from the fact that type 2 quasars also tend to be in interaction more often than type 1 quasars ( serber et al .
2006 ) . in order to better understand the role of interactions in driving starburst and nuclear activity ( and the validity of the evolutionary trend ) , we are in the process of studying agn and starburst manifestations in the nearest neighbors of the active galaxies in our samples , since the same physical processes should act on both members of the pair ( host and nearest neighbor ) .
ek thanks the ia unam for its warm hospitality were a major part of this work was completed .
mp acknowledges funding by the mexican government research grant no .
conacyt 39679 , vc by the conacyt research grants 39560-f , 42609 and d. d - h support from grant in100703 from dgapa , papiit , unam .
this research has made use of the maps catalog of poss i supported by the university of minnesota ( the aps databases can be accessed at http://aps.umn.edu/ ) and of the usnofs image and catalogue archive operated by the united states naval observatory , flagstaff station ( http://www.nofs.navy.mil/ data / fchpix/ ) .
ngc0023 & 00 09 53.1 & 25 55 25 & 12.50 & 0.0152 & isolated & starburst + eso079-g003 & 00 14 54.7 & @xmath6339 11 19 & 12.50 & 0.0090 & isolated & normal + ngc0174 & 00 36 59.0 & @xmath6329 28 40 & 13.62 & 0.0116 & isolated & starburst + ugc00556 & 00 54 49.6 & 29 14 43 & 15.30 & 0.0154 & isolated & liner + ugc00903 & 01 21 47.1 & 17 35 34 & 14.70 & 0.0084 & isolated & unclassified + eso353-g020 & 01 34 51.6 & @xmath6336 08 08 & 13.95 & 0.0161 & isolated & normal + ngc0716 & 01 52 59.3 & 12 42 31 & 14.00 & 0.0152 & isolated & normal + ugc01451 & 01 58 29.9 & 25 21 34 & 14.30 & 0.0164 & isolated & normal + ngc0835 & 02 09 24.5 & @xmath6310 08 06 & 13.14 & 0.0138 & 11.00 & starburst + ngc0838 & 02 09 38.3 & @xmath6310 08 45 & 14.22 & 0.0128 & 27.82 & starburst + ngc0839 & 02 09 42.8 & @xmath6310 10 59 & 14.20 & 0.0128 & 27.91 & starburst + ngc0873 & 02 16 32.2 & @xmath6311 20 52 & 12.83 & 0.0134 & isolated & starburst + ngc0877 & 02 17 58.5 & 14 32 53 & 12.50 & 0.0131 & isolated & starburst + ngc0922 & 02 25 04.4 & @xmath6324 47 15 & 12.63 & 0.0103 & isolated & starburst + ngc0992 & 02 37 25.2 & 21 06 06 & 13.50 & 0.0138 & 30.32 & starburst + ngc1083 & 02 45 40.5 & @xmath6315 21 24 & 15.19 & 0.0137 & isolated & starburst + ngc1134 &
02 53 40.9 & 13 00 58 & 13.20 & 0.0121 & 77.18 & normal + ugc02403 & 02 55 57.2 & 00 41 36 & 13.20 & 0.0139 & isolated & starburst + ngc1204 & 03 04 39.9 & @xmath6312 20 25 & 14.21 & 0.0143 & isolated & starburst + eso420-g013 & 04 13 49.5 & @xmath6332 00 23 & 13.52 & 0.0121 & isolated & sy2 + ngc1614 & 04 31 35.5 & @xmath6308 40 56 & 12.00 & 0.0160 & 2.00 & starburst + ngc2782 & 09 14 05.5 & 40 06 54 & 12.66 & 0.0085 & isolated & starburst + ngc1667 & 04 46 10.5 & @xmath6306 24 24 & 13.00 & 0.0150 & isolated & sy2 + ngc2785 & 09 15 15.8 & 40 55 08 & 14.90 & 0.0088 & 52.00 & starburst + ngc2856 & 09 24 16.9 & 49 14 58 & 13.90 & 0.0088 & 26.53 & starburst + ngc3147 & 10 16 55.8 & 73 24 07 & 11.52 & 0.0094 & isolated & sy2 + ngc3221 & 10 22 21.0 & 21 34 12 & 14.30 & 0.0137 & isolated & normal + ngc3367 & 10 46 34.5 & 13 45 10 & 12.22 & 0.0101 & isolated & unclassified + ngc3508 & 11 02 59.6 & @xmath6316 17 17 & 13.20 & 0.0130 & isolated & starburst + * ngc3690 & 11 28 32.9 & 58 33 19 & 13.20 & 0.0104 & 2.16 & starburst + ngc3735 & 11 36 01.0 & 70 32 06 & 12.60 & 0.0090 & isolated & sy2 + ngc3994 & 11 57 37.0 & 32 16
39 & 13.68 & 0.0105 & 18.17 & liner + ngc3995 & 11 57 44.9 & 32 17 42 & 12.96 & 0.0109 & 18.65 &
starburst + ngc4175 & 12 12 30.7 & 29 10 10 & 14.20 & 0.0131 & 17.39 & sy2 + ngc4194 & 12 14 10.1 & 54 31 34 & 13.00 & 0.0084 & isolated & starburst + ngc4332 & 12 22 47.8 & 65 50 36 & 13.20 & 0.0091 & isolated & starburst + ngc4388 & 12 25 46.7 & 12 39 40 & 12.20 & 0.0084 & isolated & sy2 + ngc4433 & 12 27 38.6 & @xmath6308 16 49 & 12.90 & 0.0099 & 63.09 & normal + mcg-02 - 33 - 098 & 13 02 19.8 & @xmath6315 46 07 & 15.38 & 0.0159 & isolated & starburst + mcg-03 - 34 - 014 & 13 12 34.5 & @xmath6317 32 28 & 13.02 & 0.0092 & isolated & normal + ngc5020 & 13 12 40.1 & 12 35 57 & 13.40 & 0.0112 & isolated & normal + ic0860 & 13 15 03.7 & 24 37 05 & 14.80 & 0.0129 & isolated & unclassified + ngc5073 & 13 19 21.4 & @xmath6314 51 46 & 13.50 & 0.0091 & isolated & starburst + ic4280 & 13 32 53.0 & @xmath6324 12 29 & 13.51 & 0.0163 & 93.52 & starburst + ngc5371 & 13 55 40.3 & 40 27 38 & 11.59 & 0.0085 & isolated & unclassified + ngc5394 & 13 58 33.7 & 37 27 10 & 13.85 & 0.0116 & 19.44 & starburst + ngc5395 & 13 58 37.9 & 37 25 27 & 12.47 & 0.0116 & 19.10 & liner + ngc5430 & 14 00 45.7 & 59 19 46 & 13.08 & 0.0099 & isolated & starburst + ngc5433 & 14 02 36.1 & 32 30 35 & 14.00 & 0.0145 & isolated & starburst + ngc5427 & 14 03 25.7 & @xmath6306 01 53 & 11.93 & 0.0087 & 22.81 & sy2 + ngc5595 & 14 24 13.2 & @xmath6316 43 28 & 13.12 & 0.0090 & 32.63 & normal + ngc5597 & 14 24 27.2 & @xmath6316 45 50 & 13.32 & 0.0089 & 32.68 & starburst + ngc5653 & 14 30 10.3 & 31 12 50 & 13.39 & 0.0119 & isolated & starburst + ngc5728 & 14 42 23.6 & @xmath6317 15 14 & 12.81 & 0.0095 & isolated & sy2 + ngc5757 & 14 47 46.2 & @xmath6319 04 45 & 13.50 & 0.0089 & isolated & starburst + ngc5793 & 14 59 24.6 & @xmath6316 41 38 & 14.17 & 0.0117 & 37.36 & sy2 + ugc09668 & 14 55 56.0 & 83 31 29 & 13.80 & 0.0131 & 63.55 & starburst + cgcg049 - 057 & 15 13 13.2 & 07 13 26 & 15.50 & 0.0130 & isolated & starburst + ngc5900 & 15 15 05.0 & 42 12 28 & 15.00 & 0.0084 & 70.48 & normal + ngc5930 & 15 26 07.8 & 41 40 30 & 13.00 & 0.0087 & 3.59 & starburst + ngc5936 & 15 30 01.1 & 12 59 21 & 13.41 & 0.0134 & isolated & starburst + ngc5937 & 15 30 46.0 & @xmath6302 49 49 & 13.35 & 0.0095 & isolated & starburst + ngc5990 & 15 46 16.0 & 02 24 49 & 13.10 & 0.0128 & 84.90 & sy2 + * ngc6052 & 16 05 13.1 & 20 32 27 & 14.70 & 0.0151 & isolated & starburst + eso402-g026 & 21 22 31.7 & @xmath6336 40 57 & 13.69 & 0.0093 & isolated & normal + ngc7130 & 21 48 19.5 & @xmath6334 57 10 & 13.33 & 0.0161 & isolated & unclassified + ngc7172 & 22 02 02.2 & @xmath6331 52 15 & 12.95 & 0.0086 & 46.75 & sy2 + ic5179 & 22 16 09.3 & @xmath6336 50 43 & 12.46 & 0.0114 & isolated & starburst + eso534-g009 & 22 38 41.7 & @xmath6325 51 02 & 13.55 & 0.0113 & 86.40 & liner + ngc7469 & 23 03 15.5 & 08 52 24 & 13.00 & 0.0162 & 18.76 & unclassified + ngc7541 & 23 14 43.0 & 04 32 03 & 12.70 & 0.0089 & 23.62 & starburst + ngc7591 & 23 18 16.0 & 06 35 08 & 13.80 & 0.0165 & isolated & sy2 + ngc7678 & 23 28 27.7 & 22 25 15 & 12.70 & 0.0116 & isolated & sy2 + ngc7714 & 23 36 14.0 & 02 09 17 & 13.10 & 0.0093 & 16.12 &
starburst + ngc7769 & 23 51 04.7 & 20 09 03 & 13.10 & 0.0141 & 2.64 & unclassified + ngc7771 & 23 51 24.7 & 20 06 41 & 13.39 & 0.0143 & 13.23 & unclassified + ugc12914 & 24 01 38.0 & 23 29 04 & 13.20 & 0.0146 & 14.82 & unclassified + ugc12195 & 24 01 42.2 & 23 29 41 & 13.90 & 0.0145 & 14.33 & unclassified + ngc0023 & 00 09 53.1 & 25 55 25 & 13.95 & 0.0152 & starburst + none + ugc00556 & 00 54 49.6 & 29 14 43 & 15.63&0.0154 & liner + neighbor 1 & 00 54 51.1 & 29 16 25&17.03 & [email protected] + ngc0716 & 01 52 59.3 & 12 42 31 & 14.51 & 0.0152 & normal + none + ugc01451 & 01 58 29.9 & 25 21 34 & 15.41 & 0.0164 & normal + none + ngc0835 & 02 09 24.5 & @xmath6310 08 06 & 13.67@xmath65 & 0.0138 & starburst + neighbor 1 & 02 09 20.9 & @xmath6310 08 00 & @xmath65 & [email protected] + neighbor 2 & 02 09 38.3 & @xmath6310 08 45 & 14.89 & [email protected] + neighbor 3 & 02 09 42.8 & @xmath6310 10 59 & 15.01 & [email protected] + ngc0877 & 02 17 58.5 & 14 32 53 & 13.07 & 0.0131 & starburst + neighbor 1 & 02 17 53.3 & 14 31 17 & 16.04 & 0.0136@xmath64 0.0007 + neighbor 2 & 02 17 26.3 & 14 34 49 & 16.77 & 0.013376@xmath66 + ngc0922 & 02 25 04.4 & @xmath6324 47 15 & 13.25 & 0.0103 & starburst + neighbor 1 & 02 24 30.0 & @xmath6324 44 44 & 16.73&0.1054@xmath66 + ngc0992 & 02 37 25.2 & 21 06 06 & 15.39 & 0.0138 & starburst + neighbor 1 & 02 37 28.2 & 21 08 31 & 16.99 & 0.0126@xmath64 0.0004 + ngc1614 & 04 31 35.5 & @xmath6308 40 56 & 14.55 & 0.0160 & sy2 + neighbor 1 & 04 31 35.5 & @xmath6308 40 56 & 16.44 & 0.0160 + ngc1667 & 04 46 10.5 & @xmath6306 24 24 & 13.00@xmath67 & 0.0160 & sy2 + none + ugc02403 & 02 55 57.2 & 00 41 36 & 15.33 & 0.0139 & starburst + neighbor 1 & 02 55 58.9 & 00 40 26 & 19.10 & 0.0749@xmath64 0.0002 + ngc2785 & 09 15 15.8 & 40 55 08 & 14.85 & 0.0088 & starburst + neighbor 1 & 09 15 33.8 & 40 55 27 & 14.54 & [email protected] + neighbor 2 & 09 14 43.1 & 40 52 47 & 14.54 & 0.008319@xmath66 + neighbor 3 & 09 14 35.6 & 40 55 24 & 17.58 & 0.008933@xmath66 + ngc2856 & 09 24 16.9 & 49 14 58 & 14.71 & 0.0088 & starburst + neighbor 1 & 09 24 03.1 & 49 12 16 & 14.52 & [email protected] + ngc3221 & 10 22 21.0 & 21 34 12 & 13.87 & 0.0137 & normal + neighbor 1 & 10 22 26.0 & 21 32 31 & 17.06 & [email protected] + neighbor 2 & 10 22 21.1 & 21 31 00 & 17.89 & [email protected] + neighbor 3 & 10 22 13.4 & 21 30 42 & 18.67 & [email protected] + ngc3690 & 11 28 32.9 & 58 33 19 & 13.76@xmath65 & 0.0104 & starburst + neighbor 1 & 11 28 33.5 & 58 33 47 & @xmath65 & 0.010411@xmath66 + neighbor 2 & 11 28 27.3 & 58 34 42 & @xmath65 & [email protected] + neighbor 3 & 11 28 45.8 & 58 35 36 & 16.14 & [email protected] + ngc4388 & 12 25 46.7 & 12 39 40 & 12.79&0.0084 & sy2 + neighbor 1 & 12 25 41.7 & 12 48 38 & 14.50 & [email protected] + neighbor 2 & 12 25 15.2 & 12 42 53 & 15.87 & @xmath680.001@xmath66 + ic0860 & 13 15 03.7 & 24 37 05 & 15.31 & 0.0129 & unclassified + none + ngc5073 & 13 19 21.4 & @xmath6314 51 46 & 14.03 & 0.0091 & starburst + neighbor 1 & 13 19 34.1 & @xmath6314 46 22 & 16.21 & [email protected] + neighbor 2 & 13 18 56.4 & @xmath6314 54 13 & 16.41 & [email protected] + ngc5433 & 14 02 36.1 & 32 30 35 & 14.68 & 0.0145 & starburst + neighbor 1 & 14 02 39.0 & 32 27 50 & 18.00 & [email protected] + neighbor 2 & 14 02 20.5 & 32 26 53 & 16.17 & [email protected] + ngc5653 & 14 30 10.3 & 31 12 50 & 14.10 & 0.0119 & starburst + none + ngc5990 & 15 46 16.0 & 02 24 49 & 14.29 & 0.0128 & starburst + neighbor 1 & 15 46 28.9 & 02 23 09 & 18.16 & [email protected] + neighbor 2 & 15 46 23.2 & 02 21 34 & 17.58 & [email protected] + neighbor 3 & 15 45 45.9 & 02 24 35 & 15.87 & [email protected] + ngc7541 & 23 14 43.0 & 04 32 03 & 13.22 & 0.0089 & starburst + neighbor 1 & 23 14 34.5 & 04 29 54 & 14.70 & [email protected] + ngc7714 & 23 36 14.0 & 02 09 17 & 13.10@xmath67 & 0.0093 & starburst + neighbor 1 & 23 36 22.1 & 02 09 24 & 14.90@xmath67 & [email protected] + ngc7771 & 23 51 24.7 & 20 06 41 & 13.81@xmath65 & 0.0143 & unclassified + neighbor 1 & 23 51 22.5 & 20 05 47 & @xmath65 & [email protected] + neighbor 2 & 23 51 13.1 & 20 06 12 & 17.13 & 0.013679@xmath66 + neighbor 3 & 23 51 04.0 & 20 09 02 & 14.05 & [email protected] + neighbor 4 & 23 51 13.9 & 20 13 46 & 17.02 & 0.043527@xmath66 | we present a 3-dimensional study of the local ( @xmath0 kpc ) and the large scale ( @xmath1 1 @xmath2 mpc ) environment of bright iras galaxies ( birgs ) . for this purpose
we use 87 birgs located at high galactic latitudes ( with [email protected] ) as well as a control sample of non - active galaxies having the same morphological , redshift and diameter size distributions as the corresponding birg sample . using the center for astrophysics ( cfa2 ) and southern sky redshift survey ( ssrs ) galaxy catalogues ( @xmath4 ) as well as our own spectroscopic observations ( @xmath5 ) for a subsample of the original birg sample , we find that the fraction of birgs with a close neighbor is significantly higher than that of their control sample . comparing with a related analysis of sy1 and sy2 galaxies of koulouridis et al .
( 2006 ) we find that birgs have a similar environment as sy2s , although the fraction of birgs with a bright close neighbor is even higher than that of sy2 galaxies .
an additional analysis of the relation between fir colors and the type of activity of each birg shows a significant difference between the colors of strongly - interacting and non - interacting starbursts and a resemblance between the colors of non - interacting starbursts and sy2s .
our results support the view where close interactions can drive molecular clouds towards the galactic center , triggering starburst activity and obscuring the nuclear activity .
when the close neighbor moves away , starburst activity is reduced with the simultaneous appearance of an obscured ( type 2 ) agn .
finally , the complete disentanglement of the pair gives birth to an unobscured ( type 1 ) agn . |
the idea of creating a coherent state for a quantum system was first proposed by schrdinger , in 1926 , in connection with the classical states of the quantum harmonic oscillator ; he referred to them as states of the minimum uncertainty product @xcite .
almost forty years later in 1963 , glauber @xcite introduced the field coherent states , these have been realized experimentally with the development of lasers , and can be obtained from any one of three mathematical definitions : ( i ) as the right - hand eigenstates of the boson annihilation operator @xmath0 , with @xmath1 being a complex number ; ( ii ) as the states obtained by application of the displacement operator @xmath2 upon the vacuum state ; and ( iii ) as those states with a minimum uncertainty relationship @xmath3 , with @xmath4 and @xmath5 the position and momentum operators respectively and @xmath6 .
the same states are obtained from any one of the three mathematical definitions when one makes use of the harmonic oscillator algebra . at the same time
, klauder @xcite developed a set of continuous states in which the basic ideas of coherent states for arbitrary lie groups were contained .
the complete construction of coherent states of lie groups was achieved by perelomov @xcite and gilmore @xcite almost ten years later ; the basic theme of this development was to connect the coherent states with the dynamical group for each physical problem thus allowing the generalization of the concept of coherent state . for systems
far from the ground state or with a finite number of bound states , the harmonic oscillator algebra is no longer adequate , therefore there has arisen an interest in generalizing the concept of coherent state for systems with different dynamical properties .
nieto and simmons @xcite generalized the notion of coherent states for potentials having unequally spaced energy levels .
their construction is such that the resultant states are localized , follow the classical motion and disperse as little as possible in time .
their coherent states are defined as those which minimize the uncertainty relation equation .
they constructed coherent states for the pschl - teller potential , the harmonic oscillator with a centripetal barrier and the morse potential .
gazeau and klauder @xcite proposed a generalization for systems with one degree of freedom possessing discrete and continuous spectra .
these states present continuity of labeling , resolution of unity , and temporal stability .
the key point is to parametrize such states by means of two real values : an amplitude @xmath7 and a phase @xmath8 , instead of using a complex value @xmath1 .
manko and collaborators @xcite introduced coherent states of a _ f - deformed _ algebra as eigenstates of a deformed annihilation operator @xmath9 , where @xmath10 is the usual annihilation operator of the harmonic oscillator algebra and @xmath11 is a number operator function that specifies the potential .
these states present nonclassical properties such as squeezing and anti - bunching .
more recently , they generalized the displacement operator method for the case of _ f - deformed _ oscillators introducing a deformed version of the displacement operator @xcite with the disadvantage that it is non unitary and does not displace the deformed operators in the usual form . on the basis of the f - deformed oscillator formalism , in previous works we put forward the construction of generalized coherent states for nonlinear systems by selecting a specific deformation function in a way such
that the energy spectrum of the hamiltonian it seeks to describe is similar to that of a f - deformed hamiltonian . in this regard , we examined the trigonometric and the modified pschl - teller potentials , as well as the morse potential , all of them containing linear and quadratic terms in the number operator @xcite .
in addition to these , a harmonic oscillator with an inverse square potential in two dimensions was also considered @xcite .
coherent states have been applied not only in optics but in many fields of physics , for instance , in the study of the forced harmonic oscillator @xcite , quantum oscillators @xcite , quantum nondemolition measurements @xcite , nonequilibrium statistical mechanics@xcite , and atomic and molecular physics @xcite .
excellent reviews can be found in @xcite . + the paper is organized as follows . in section [ harmonic ]
we briefly describe the properties of the usual coherent states . in section [ deformed ]
we present a methodology to construct deformed hamiltonians for specific systems . in section [ nlcs ]
we give two alternative forms to obtain the coherent states pertinent to a nonlinear hamiltonian . and finally , in section [ res ] we present some numerical results .
the harmonic oscillator coherent states , also called _ field coherent states _
@xcite , are quantum states of minimum uncertainty product which most closely resemble the classical ones in the sense that they remain well localized around their corresponding classical trajectory .
when one makes use of the harmonic oscillator algebra , the same coherent states are obtained from the three glauber s mathematical definitions mentioned above .
such states take the following form in terms of the number state basis : @xmath12 where @xmath1 is an arbitrary complex number . for every complex number @xmath13 the coherent state @xmath14 has a non zero projection on every fock state @xmath15 . from a statistical point of view , it follows from the above that the occupation number distribution of a coherent state , @xmath16 , is characterized by a poisson distribution : @xmath17 with an average occupation number @xmath18 and mean square root deviation @xmath19 such states are not orthogonal , and the overlap between two of them is given by @xmath20 moreover , they form an over - complete basis with the closure relationship , @xmath21 so that any state of the group generated by the operators @xmath10 , @xmath22 and @xmath23 can be expanded in terms of coherent states .
a deformed oscillator is a non - harmonic system characterized by a hamiltonian of the harmonic oscillator form @xmath24 with a specific frequency @xmath25 and deformed boson creation @xmath26 and annihilation @xmath27 operators defined as @xcite : @xmath28 with @xmath11 a number operator function that specifies the deformation .
the commutation relations between the deformed operators are : @xmath29 = ( \hat n+1)f^2(\hat n+1)-\hat n f^2(\hat n ) , \ \ [ \hat a , \hat n]=\hat a , \ \ [ \hat a^{\dagger},\hat n ] = -\hat a^{\dagger}.\ ] ] notice that the commutation relations between the deformed operators @xmath27 , @xmath26 and the number operator are the same as those among the usual operators @xmath10 , @xmath22 with the number operator . note that the eigenfunctions of the harmonic oscillator @xmath15 are also eigenfunctions of the deformed oscillator . in terms of the number operator , the deformed hamiltonian takes the form : @xmath30 the morse oscillator is a particularly useful anharmonic potential for the description of systems that deviate from the ideal harmonic oscillator conduct and has been used widely to model the vibrations of a diatomic molecule . if we choose the deformation function @xcite : @xmath31 with @xmath32 an anharmonicity parameter , and substitute it into eq .
[ eq : hdef ] , we obtain an algebraic hamiltonian of the form @xmath33,\ ] ] whose spectrum is in essence identical to that of the morse potential @xcite , i.e , @xmath34 provided that @xmath35 and @xmath36 , with @xmath37 being the number of bound states corresponding to the integers @xmath38 . for this particular choice of deformation function ,
the commutator between deformed operators is : @xmath39 = 1-\chi_a(2\hat n+1 ) = \frac{2(n-\hat{n})}{2n+1},\ ] ] from which one can see that the commutator equals a scalar plus a linear function of the number operator . + the action of the deformed operators upon the number states is : @xmath40 and @xmath41 the deformed operators change the number of quanta in @xmath42 and their corresponding matrix elements depend on the deformation function @xmath43 .
furthermore , from the commutation relations we see that the set @xmath44 is closed under the operation of commutation . the modified pschl - teller potential can be written as : @xmath45 where @xmath46 is the depth of the well , @xmath47 is the range of the potential and @xmath48 is the relative distance from the equilibrium position .
the eigenfunctions and eigenvalues are , respectively , @xcite @xmath49 and @xmath50 where @xmath51 is a normalization constant , @xmath52 , @xmath53 is the reduced mass of the molecule , @xmath54 is related with the depth of the well so that @xmath55 , @xmath56 and @xmath57 stands for the hypergeometric function @xcite
. if we write the eigenvalues in terms of the parameter @xmath54 , we obtain @xmath58 the number of bound states is determined by the dissociation limit @xmath59 for integer values of @xmath54 , the state associated with null energy is not normalizable , whereby the last bound state corresponds to @xmath60 @xcite .
the modified pschl - teller potential is a nonlinear potential symmetric in the displacement coordinate .
+ let us now consider a deformation function of the form @xcite @xmath61 by substituting it into eq .
[ eq : hdef ] , we obtain the deformed hamiltonian @xmath62 whose spectrum is identical to that of eq .
[ eq : modpt ] .
the harmonic limit is obtained by taking @xmath63 , @xmath64 with @xmath65 . for this choice of deformation function , deformed operators ,
together with the number operator , obey the following commutation relations : @xmath29 = \frac{\hbar a^2}{\mu\omega}(s-\hat n ) , \ \ [ \hat a , \hat n]=\hat a , \ \ [ \hat a^{\dagger},\hat n]=-\hat a^{\dagger},\ ] ] which have the correct harmonic limit and are similar to those of the generators of the @xmath66 group .
the trigonometric pschl - teller potential is given by : @xmath67 where @xmath46 is the strength of the potential and @xmath47 is its range .
this potential supports an infinite number of bound states , its eigenfunctions and eigenvalues are @xcite : @xmath68 @xmath69 where @xmath53 is the mass of the particle , @xmath70 and the parameter @xmath71 is related to the potential strength and range by @xmath72 . in the harmonic limit @xmath73 and @xmath64 with @xmath74 .
if we choose a deformation function @xcite @xmath75 the deformed hamiltonian becomes @xmath76 whose eigenvalues are identical to those given by eq .
[ eq : trigpt ] .
once we have given the deformation function , the deformed operators are specified . in this case
the commutation relations are : @xmath29 = \frac{\hbar a^2}{\mu\omega}(\hat n+\lambda ) , \ \ [ \hat a , \hat n]=\hat a , \ \ [ \hat a^{\dagger},\hat n ] = -\hat a^{\dagger},\ ] ] which are similar to those of the generators of the @xmath77 group .
we emphasize here that the set of operators @xmath78 is closed under the operation of commutation .
in this work we consider the generalization of two of the known definitions given to construct the field coherent states , namely , i ) as eigenstates of an annihilation operator and ii ) as the states obtained from the application of the displacement operator upon a maximal state .
following manko and collaborators , we construct deformed coherent states as eigenstates of the deformed annihilation operator @xcite : @xmath79 in order to solve eq .
( [ eq : aocs ] ) , let the state @xmath80 be written as a weighted superposition of the number eigenstates @xmath81 : @xmath82 on inserting this into eq .
[ eq : aocs ] , we get the following relation between the coefficients @xmath83 and @xmath84 : @xmath85 by applying @xmath86 times the annihilation operator , we get a relationship between @xmath83 and @xmath87 : @xmath88 where @xmath89
. substitution of eq .
[ eq : recurr ] into eq .
[ eq : aocs ] gives us the following expression for the annihilation operator coherent states ( aocs ) : @xmath90 as an example we replaced the explicit form of the deformation function for the trigonometric pschl - teller potential and obtain the aocs associated with this system @xmath91 with @xmath92^{-1/2}$ ] a normalization constant . in this subsection
we construct the displaced operator coherent states ( docs ) for the nonlinear potentials discussed in section [ deformed ] .
we propose to construct them by application of a deformed displacement operator on the fundamental state of the system , i.e , @xmath93|0\rangle \label{eq : docsdef}\ ] ] where @xmath1 is a complex parameter .
the deformed displacement operator is obtained from the usual one through the replacement of the harmonic oscillator creation and annihilation operators by their deformed counterparts .
since the commutator between the deformed operators can be a rather complicated function of the number operator , it is not possible , in general , to disentangle the exponential in ( [ eq : docsdef ] ) .
however , for the cases we are considering in this work , we have found that the commutator between the deformed operators @xmath27 , @xmath26 is equal to a scalar plus a linear function of the number operator , and therefore the dynamical group pertinent to these systems is composed by the operators @xmath94 which form a lie algebra .
so , the deformed displacement operator can be disentangled @xcite to get ( for the particular case of a morse potential ) @xcite @xmath95 , \nonumber \\ & = & \exp\left(\zeta\frac{\hat a^{\dagger}}{\sqrt{\chi_a}}\right ) \left(\frac{1}{1+|\zeta|^2}\right)^{g(\chi_a
\hat n)/2\chi_a } \exp\left(-\zeta^{*}\frac{\hat a}{\sqrt{\chi_a}}\right ) , \end{aligned}\ ] ] where , for a given value of @xmath96 , the complex parameter @xmath97 is introduced , and @xmath98 $ ] .
the docs for this particular case can be obtained as : @xmath99 where the explicit form of the deformation function has been used and the state is approximate due to the fact that the number of bound states supported by the potential is finite .
in this subsection we consider a hetero - nuclear diatomic molecule hf with 22 bound states . due to the asymmetry of the molecule we model it by means of a morse potential . according to carvajal _
et al _ @xcite , the position and momentum variables for this system can be expressed as a series expansion involving all powers of the deformed creation and annihilation operators .
keeping up to second - order terms we obtain : @xmath100 @xmath101 where the expansion coefficients are functions of the number operator and are given by @xcite : @xmath102 \\
f_{10}(\hat n)&=&f_{01}(\hat n)=\sqrt{\frac{k-1}{k}}\left(1+\frac{\hat n}{k-\hat n}\right ) \\
f_{20}(\hat n)&=&f_{02}(\hat n)=\frac{k-1}{2k\sqrt{k}}\left(\frac{-1}{(1-(\hat n-1)/k)(1-\hat n / k)}\right ) , \\ g_{10}(\hat n)&=&-g_{01}(\hat n)=\sqrt{\frac{k-1}{k}}\left(\frac{k-2\hat n}{k-\hat n}\right),\\ g_{20}(\hat n)&=&-g_{02}(\hat n)=-\frac{k-1}{k\sqrt{k}}\left(\frac{k-(2\hat n-1)}{k(1-(\hat n-1)/k)(1-\hat n / k)}\right ) , \end{aligned}\ ] ] where in turn , @xmath103 with @xmath104 being the euler constant and @xmath105 is child s parameter @xcite defined by @xmath106 . in what follows
we define the deformed position and momentum operators @xmath107 and @xmath108 taking @xmath109 in eqs .
[ eq : xdef ] and [ eq : pdef ] . when the deformation function is equal to 1 ( @xmath110 )
, these expressions take the harmonic values . in ref .
@xcite we compared the phase space trajectories obtained by averaging the deformed coordinate @xmath107 and momentum @xmath108 ( keeping up to second order terms ) with those obtained averaging the morse coordinate and momentum and we found a very good agreement between them when the averages were taken between nonlinear coherent states obtained as eigenstates of the deformed annihilation operator ( aocs ) . in order to evaluate the phase space trajectories
, here we will consider the temporal evolution of the nonlinear coherent states obtained by application of the deformed displacement operator on the vacuum state ( see eq .
[ eq : docsmorse ] ) .
that is , we apply the time evolution operator @xmath111 on the state @xmath112 .
@xmath113}|\zeta\rangle.\ ] ] the averages are : @xmath114 transformation of the deformed operators yield : @xmath115}\hat a e^{-i\omega t\left[\hat n-\chi_a\left(\hat n^2+\hat n\right)\right ] } \nonumber \\ & = & e^{-i\omega\chi_a t\hat n^2}e^{i\omega t(1-\chi_a)\hat n}\hat a e^{-i\omega t(1-\chi_a)\hat n}e^{i\omega\chi_a t\hat n^2}\nonumber \\ & = & e^{-i\omega ( 1-\chi_a)t}e^{-i\omega\chi_a t\hat n^2}\hat a e^{i\omega\chi_a t\hat n^2 } = e^{-i\omega t ( 1 - 2\chi_a)}e^{-2i\omega\chi_a t \hat n}\hat a\end{aligned}\ ] ] and @xmath116 notice the presence of the number operator in the exponentials . from these expressions
we can get @xmath117 and @xmath118 and obtain the deformed coordinate and momentum as a function of time . in figure [ fig : phasespace ]
we show the occupation number distributions @xmath119 ( left column ) and the corresponding phase space trajectories ( right column ) for @xmath120 , @xmath121 , and @xmath122 . in the calculation ,
a diatomic hf molecule is considered to be modeled by a deformed morse - like oscillator with @xmath123 bound states .
it can be seen from the figure that for the values used for the parameter @xmath1 , which fixes the average value of the number operator , the occupation number distributions are such that the states near dissociation are mostly unoccupied so it is to be expected that the contribution from the states in the continuum can be neglected . concerning the phase space trajectories we see that for a given value of the average number operator , that is , a given value of the average energy , there are several intersecting curves in contrast with the case of a field coherent state where one finds a single trajectory for a given energy .
the presence of several trajectories for a given energy is a signature of the nonlinear term in the deformed hamiltonian .
notice that for a small energy ( right column top ) the phase space trajectories fill an anular region of phase space with an unaccessible internal region .
the width of the anular region narrows when the parameter @xmath124 is decreased , that is , for smaller values of the average number operator .
this unaccessible region is lost as the energy is increased ( right column medium and bottom ) . due to the asymmetry of the morse potential
the deformed coordinate can attain small negative values and much larger positive values along the temporal evolution .
, @xmath121 , and @xmath122 . in the calculation ,
a diatomic hf molecule is considered to be modeled by a deformed morse - like oscillator with @xmath123 bound states.,title="fig:",width=264,height=188 ] , @xmath121 , and @xmath122 . in the calculation ,
a diatomic hf molecule is considered to be modeled by a deformed morse - like oscillator with @xmath123 bound states.,title="fig:",width=264,height=245 ] , @xmath121 , and @xmath122 . in the calculation ,
a diatomic hf molecule is considered to be modeled by a deformed morse - like oscillator with @xmath123 bound states.,title="fig:",width=264,height=188 ] , @xmath121 , and @xmath122 . in the calculation ,
a diatomic hf molecule is considered to be modeled by a deformed morse - like oscillator with @xmath123 bound states.,title="fig:",width=264,height=245 ] , @xmath121 , and @xmath122 . in the calculation ,
a diatomic hf molecule is considered to be modeled by a deformed morse - like oscillator with @xmath123 bound states.,title="fig:",width=264,height=188 ] , @xmath121 , and @xmath122 . in the calculation ,
a diatomic hf molecule is considered to be modeled by a deformed morse - like oscillator with @xmath123 bound states.,title="fig:",width=264,height=245 ] in figure [ fig : uncertainties ] we show the temporal evolution of the uncertainties in coordinate ( left column ) and momentum ( right column ) of deformed displacement operator coherent states ( docs ) for @xmath120 , @xmath121 , and @xmath122 .
the first row corresponds to the case with lowest energy considered @xmath125 , the dispersions in the deformed position and momentum are oscillating functions with amplitudes in the range @xmath126 with @xmath127 .
notice that the docs are not minimum uncertainty states and notice also that there is squeezing present whenever the dispersion in any coordinate is smaller than 0.5 .
the second row corresponds to @xmath128 , here the dispersions oscillate with larger amplitude @xmath129 . and finally , in the third row , @xmath130 , the dispersions oscillate with amplitude in the range @xmath131 . in the cases with @xmath132
the amplitude of the oscillations in the dispersion in the coordinate is larger than that in the momentum .
, @xmath121 , and @xmath122 .
in the calculation , a diatomic hf molecule is considered to be modeled by a deformed morse - like oscillator with @xmath123 bound states.,title="fig:",width=264,height=188 ] , @xmath121 , and @xmath122 . in the calculation ,
a diatomic hf molecule is considered to be modeled by a deformed morse - like oscillator with @xmath123 bound states.,title="fig:",width=264,height=188 ] , @xmath121 , and @xmath122 . in the calculation ,
a diatomic hf molecule is considered to be modeled by a deformed morse - like oscillator with @xmath123 bound states.,title="fig:",width=264,height=188 ] , @xmath121 , and @xmath122 . in the calculation ,
a diatomic hf molecule is considered to be modeled by a deformed morse - like oscillator with @xmath123 bound states.,title="fig:",width=264,height=188 ] , @xmath121 , and @xmath122 . in the calculation ,
a diatomic hf molecule is considered to be modeled by a deformed morse - like oscillator with @xmath123 bound states.,title="fig:",width=264,height=188 ] , @xmath121 , and @xmath122 . in the calculation ,
a diatomic hf molecule is considered to be modeled by a deformed morse - like oscillator with @xmath123 bound states.,title="fig:",width=264,height=188 ] in figure [ fig : puncertainties ] we show the temporal evolution of the normalized uncertainty product @xmath133 \rangle|^{2}$ ] of displacement operator coherent states for @xmath120 , @xmath121 , and @xmath122 ( frames ( a ) , ( b ) and ( c ) , respectively ) .
we see that there are some specific times at which the dispersion is minimal and these times do not depend on the value of @xmath134 , these correspond to the outermost trajectories shown in figure [ fig : phasespace ] . at those times when the trajectories evolve near the origin , so that @xmath135 and @xmath136 are small , the dispersions take their largest values @xcite .
most of the time the docs are not minimum uncertainty states , the product of the dispersions seems to be a periodic function of time and the amount of dispersion is an increasing function of @xmath134 .
\rangle|^{2}$ ] of displacement operator coherent states for @xmath120 , @xmath121 , and @xmath122 ( frames ( a ) , ( b ) and ( c ) , respectively ) . in the calculation ,
a diatomic hf molecule is considered to be modeled by a deformed morse - like oscillator with @xmath123 bound states.,width=415,height=566 ] in figure [ fig : uncertaintiesalpha ] we show the dispersions in coordinate and momentum at time @xmath137 as a function of the parameter @xmath124 . here
we see that the dispersion in the deformed coordinate @xmath107 starts as that of a minimum uncertainty state and is an increasing function of the parameter @xmath124 .
the dispersion in the momentum also starts as that of a minimum uncertainty state and is a decreasing function of the parameter @xmath124 so it is squeezed .
the normalized uncertainty product remains near a minimum uncertainty state for small values of @xmath124 .
however , for values of @xmath124 such that the average value of the number operator approaches @xmath37 the uncertainty product increases rapidly .
( a ) , in momentum @xmath138 ( b ) , and the corresponding normalized uncertainty product @xmath133 \rangle|^{2}$ ] ( c ) at time @xmath137 as functions of @xmath124 for the docs . in the calculation ,
a diatomic hf molecule is considered to be modeled by a deformed morse - like oscillator with @xmath123 bound states.,width=415,height=566 ] in this subsection we consider a homonuclear diatomic molecule @xmath139 supporting 10 bound states . due to the symmetry of the potential , the deformed coordinate and momentum
are written as an expansion involving odd powers of deformed operators @xcite as : @xmath140 @xmath141 where we have kept up to third order terms in the deformed operators and the coefficient functions are given by : @xmath142 @xmath143 @xmath144 @xmath145 here , the matrix elements @xmath146 and @xmath147 are evaluated by numerical integration using the eigenfunctions of the corresponding scrdinger equation .
the temporal evolution of the deformed coordinate and momentum is calculated taking the expectation values between the states @xmath148 with @xmath112 a nonlinear coherent state obtained by application of the deformed displacement operator on the vacuum state @xmath149 ( the state is approximate
since the sum contains a finite number of terms ) and @xmath150 with @xmath151 in figure [ fig : pspace ] we show the occupation number distributions and the phase space trajectories of docs for the particular case of @xmath152 and @xmath153 .
notice that the phase space trajectories are symmetric with respect to the origin , this is a reflection of the symmetry of the potential .
for a small value of the parameter @xmath124 we see a behavior similar to that found for the morse potential ( see figure [ fig : phasespace ] ) , that is , the phase space trajectories fill an anular region whose width depends upon the parameter @xmath124 ( for smaller values of @xmath124 the width decreases ) and there is an unaccessible internal region . for a larger value of the parameter @xmath124 the trajectories in phase space are intersecting curves occupying all the space consistent with the energy , the unaccessible region disappears .
the plots of the occupation number distribution manifest the fact that for the cases considered here the population is far from the dissociation and the approximation done keeping only the bounded part of the spectrum is justified . , @xmath153 .
in the calculation , a modified pschl - teller potential is modeled by a deformed oscillator with @xmath154 bound states.,title="fig:",width=264,height=207 ] , @xmath153 . in the calculation ,
a modified pschl - teller potential is modeled by a deformed oscillator with @xmath154 bound states.,title="fig:",width=207,height=283 ] , @xmath153 . in the calculation ,
a modified pschl - teller potential is modeled by a deformed oscillator with @xmath154 bound states.,title="fig:",width=264,height=207 ] , @xmath153 . in the calculation ,
a modified pschl - teller potential is modeled by a deformed oscillator with @xmath154 bound states.,title="fig:",width=207,height=283 ] in figure [ fig : uncertainties - rm ] we show the temporal evolution of the average value of the deformed coordinate @xmath155 , its dispersion @xmath156 and the uncertainty product @xmath157 for @xmath152 ( left column ) and @xmath153 ( right column ) . for these calculations we considered a system supporting 10 bound states . for a fixed , small value of the parameter @xmath124 ( thus a small @xmath158 )
the deformed coordinate is an oscillatory function with a slightly varying amplitude , the corresponding dispersion is an oscillatory function whose amplitude is largest when the amplitude of the oscillation is smallest .
notice the presence of squeezing .
the corresponding uncertainty product is shown in frame ( e ) , it can be seen that the docs we have constructed are minimum uncertainty states at the initial time and there are some specific instants of time when the states return to be of minimum uncertainty , this conduct is periodic .
most of the time the docs are not minimum uncertainty states .
+ for a larger value of the parameter @xmath124 ( thus a larger value of the average @xmath158 ) we see that the deformed coordinate presents oscillations with varying amplitudes .
the largest amplitude corresponds to that of a field coherent state with the same energy .
notice that when the amplitude is largest the dispersion is smallest and corresponds to that of a minimum uncertainty state , this conduct is present at the initial time and repeats itself periodically .
here we can see also the presence of squeezing .
when the amplitude of the oscillations in the deformed coordinate is small we see that the dispersion is large , this is a reflection of the nonlinear terms in the hamiltonian .
the uncertainty product also shows a periodic conduct with instants of time where the state is a minimum uncertainty state .
( left column ) and @xmath153 ( right column ) . in the calculation ,
a modified pschl - teller potential is modeled by a deformed oscillator with @xmath154 bound states.,width=529,height=566 ] finally , in figure [ fig : uncertainties - rm - alpha ] we see the dispersions in the deformed coordinate and momentum as a function of the parameter @xmath124 evaluated at time @xmath137 .
notice that for small @xmath124 the dispersion in the deformed coordinate is less than 1/2 meaning squeezing , in contrast , that of the deformed momentum is larger than 1/2 and the uncertainty product is near the minimum possible value .
as the parameter @xmath124 increases the dispersion in the deformed coordinate increases so that the state is less squeezed and that of the momentum decreases but is still far from the minimum allowed value ( in the range of values for the parameter @xmath124 shown in the figures ) so that the uncertainty product separates from the initial value . because we are dealing with a finite number of bound states @xmath54 and we are not taking into account the states from the continuum we must use a small enough value of @xmath124 so that the average occupation number is much smaller than @xmath54 .
( a ) , uncertainty in coordinate @xmath159 ( b ) , and the corresponding uncertainty product @xmath160 ( c ) at time @xmath137 as functions of @xmath124 for the docs . in the calculation ,
a modified pschl - teller potential is modeled by a deformed oscillator with @xmath154 bound states.,width=415,height=566 ] + to complete the examples , we just write the displaced operator coherent states associated we the trigonometric pschl - teller potential as @xmath161 with @xmath162 .
this expression should be compared with that of the aocs ( see eq .
[ aocs - trigo ] ) .
based on the f - deformed oscillator formalism we have introduced non linear coherent states by generalization of two definitions , as eigenstates of a deformed annihilation operator aocs and as the states that result by application of a deformed displacement operator on the vacuum state docs .
we have applied our method to hamiltonians that contain linear and quadratic terms in the number operator corresponding to the trigonometric and modified pschl - teller hamiltonians and the morse hamiltonian .
the aocs and the docs obtained for the trigonometric pschl - teller potential are exact because the number of bound states stupported by the potential is infinite . on the other hand ,
the aocs and the docs obtained for the morse and the modified pschl - teller potentials are approximate because the number of bound states is finite . for the numerical results we have considered only the states obtained by application of a deformed displacement operator on the vacuum state , in ref .
@xcite we discussed the nonlinear coherent states obtained as eigenstates of a deformed annihilation operator for the morse potential .
although from an algebraic - structure point of view the coherent states obtained from each generalization ( docs ) or ( aocs ) are different , as well as their respective statistical behavior @xcite , the average values and the phase space trajectories obtained with them are almost identical @xcite .
we must mentioned that the same form of nonlinear coherent states aocs and docs for the morse and pschl - teller potentials would be obtained , if we had worked with the true ladder operators associated with each potential @xcite .
this happen because the ladder operators and deformed operators share the same commutation relation for these kind of systems .
+ * acknowledgements : * we thank reyes garca for the maintenance of our computers and acknowledge partial support from conacyt through project 166961 and dgapa - unam project in 108413 .
9 mandel l. and wolf e , 1995 _ optical coherence and quantum optics _ ( cambridge university press ) p. 522
glauber r. j. , phys . rev . * 130 * , 2529 ( 1963 ) ; glauber r. j. , phys . rev . * 131 * , 2766 ( 1963 ) ; glauber r. j. , phys .
* 10 * , 84 ( 1963 ) .
klauder j. r. , j. math .
phys . , * 4 * , 1055 ( part i ) ( 1963 ) ; klauder j. r. , j. math . phys . * 4 * 1058 ( part ii ) ( 1963 ) .
perelomov a. m. , commun .
* 26 * , 222 ( 1972 ) .
gilmore r. , ann .
* 74 * , 391 ( 1972 ) ; gilmore r. , rev .
* 23 * , 142 ( 1974 ) ; gilmore r. , j. math .
phys . * 15 * , 2090 ( 1974 ) .
nieto , m. m. and simmons l. m. , phys . rev .
lett . * 41 * , 207 ( 1978 ) ; nieto , m. m. and simmons l. m. , phys . rev .
d * 20 * , 1321 ( 1979 ) ; nieto , m. m. and simmons l. m. , phys . rev .
d * 20 * , 1342 ( 1979 ) .
gazeau j. p. and klauder j. , j. phys .
gen . * 32 * , 123 ( 1999 ) . manko v. i. , marmo g. , sudarshan e.c.g . , zaccaria f. , physica scripta * 55 * , 528 ( 1997 ) .
aniello p. , manko v. i. , marmo g. , solimeno s. , zaccaria f. , j. opt .
b : quantum semiclass . opt .
* 2 * 718 ( 2000 ) ; manko v. i. , marmo g. , sudarshan e.c.g . , zaccaria f. , physica scripta * 55 * , 528 ( 1997 ) . o. de los santos - snchez and j. rcamier , j. phys .
a : math . theor . * 44 * 145307 ( 2011 ) .
r. romn - ancheyta , o. de los sntos - snchez and j. rcamier , j. phys .
a : math . theor . * 44 * , 435304 ( 2011 ) .
carruthers p. and nieto m m , am . j. phys .
* 7 * , 537 ( 1965 ) .
r j glauber , phys . lett . * 21 * , 650 ( 1966 ) .
w g unruh , phys .
d * 17 * , 1180 ( 1978 ) .
y takahashi and f shibata , j. phys .
japan * 38 * , 656 ( 1975 ) .
f t arrechi , e courtens , r gilmore and h thomas , phys .
rev . a * 6
* , 2211 ( 1972 ) .
y k wang and f t hioe , phys .
rev . a * 7 * , 831 ( 1973 ) .
c stopera , t v grimes , p m mclaurin , a privett , j a morales , advances in quantum chemistry * 66 * , 113 ( 2013 ) .
j r klauder and b s skagerstam , _ coherent states : applications in physics and mathematical physics _
( singapore ; world scientific ) 1985 .
s t ali , j p antoine and j p gazeau , _ coherent states , wavelets and their generalizations ( graduate texts in contemporary physics_(singapore ; world scientific ) 2000 .
wei - min zhang , da hsuan feng , r. gilmore , rev .
phys . * 62 * 867 ( 1990 ) .
rcamier j. , gorayeb m. , mochn w. l. , and paz j. l. , int .
j. quantum chem . * 106 * , 3160 ( 2006 ) .
landau l. d. and lifshitz e. m. , _ quantum mechanics ( non relativistic theory ) _ , ( oxford ; pergamon , 1977 ) .
m. abramowitz and i. stegun , _ handbook of mathematical functions _ ( new york ; dover , 1970 ) p. 555 .
r. lemus and r. bernal , chem .
* 283 * , 401 ( 2002 ) .
r. r. puri , _ mathematical methods of quantum optics _ , ( new york , springer , 2001 ) p. 48 .
r. gilmore , _ lie groups , lie algebras and some of their applications _ , ( new york , wiley , 1974 ) p. 460 .
o. de los santos - snchez and j. rcamier , j. phys . a : math .
* 45 * , 415310 ( 2012 ) .
m. carvajal , r. lemus , a. frank , c. jung and e. ziemniak , chem . phys . * 260 * , 105 ( 2000 ) . m. s. child , l. halonen , adv
* 57 * , 1 ( 1984 ) .
j. rcamier , w. l. mochn , m. gorayeb , j. l. paz and r. juregui , international journal of modern physics b * 20 * , 1851 ( 2006 ) .
o. de los santos - snchez and j. rcamier , j. phys .
a : math . theor . * 46 * , 375303 ( 2013 ) .
j. rcamier , m. gorayeb , w. l. mochn and j. l. paz , int .
j. theor . phys .
* 47 * 673 ( 2008 ) .
r. romn - ancheyta , c. gonzlez gutirrez and j. rcamier , j. opt .
b * 31*(1 ) 38 ( 2014 ) . | on the basis of the f - deformed oscillator formalism , we propose to construct nonlinear coherent states for hamiltonian systems having linear and quadratic terms in the the number operator by means of the two following definitions : i ) as deformed annihilation operator coherent states ( aocs ) and ii ) as deformed displacement operator coherent states ( docs ) .
for the particular cases of the morse and modified pschl - teller potentials , modeled as f - deformed oscillators ( both supporting a finite number of bound states ) , the properties of their corresponding nonlinear coherent states , viewed as docs , are analyzed in terms of their occupation number distribution , their evolution on phase space , and their uncertainty relations . |
in order to explain the quark gluon plasma ( qgp ) observables at relativistic heavy ion collider ( rhic ) having temperature ( @xmath8 ) @xmath9k , it has become apparent that a non - zero shear viscosity , @xmath0 , is needed .
one way of characterizing @xmath0 , which is a dimensionful number , is to take its ratio with the entropy density , @xmath1 .
it turns out that rhic plasma has [ ebs ] ~0.1 .
although @xmath0 itself is large ( in cgs units @xmath10 gm / cm / s ) , its ratio with @xmath1 is small . note that this ratio for a known fluid , say water , is about two orders of magnitude greater than the above value .
hydrodynamic simulations at rhic and large hadron collider ( lhc ) ( which provide the environments created around a few microseconds after the big bang ) as well suggest a small value for this ratio in the hot qgp state .
such a small value points towards strongly interacting matter .
string theory arguments and the famous gauge / gravity duality in the form of the ads / cft correspondence suggest a lower bound @xcite for @xmath6 for any matter given by ~0.08 , [ etabys ] where @xmath11 @xcite but non - zero @xcite ( see , e.g. , @xcite ) . for the present purpose
, we assume that there is a bound and that @xmath12 .
is there any natural site revealing an @xmath6 close to above lower bound ?
note that evenif @xmath13 , still the same question remains .
such a value may be possible to arise naturally , if the temperature and/or density of the systems are / is same as that in rhic / lhc . in the search of such an exotic system ,
our story starts which could serve as a natural verification of gauge / gravity duality in the form of the ads / cft correspondence .
observed data suggest that the accretion flows around certain black holes must be optically thin , geometrically thick and very hot
. such flows often exhibit outflows / jets in their hard spectral states .
the ion temperature ( @xmath14 ) in such flows / disks is as high as @xmath15k . the supermassive black hole system at the centre of our galaxy , sgr a@xmath16 @xcite , ( most of the temporal classes of ) the x - ray binary cyg x-1 @xcite are some of the examples of this kind of hot accretion disks , which exhibit radiatively inefficient flows .
they are particularly different from their radiatively efficient counter part , namely the keplerian accretion disk , which is geometrically thin , optically thick , cooler , and of temperature @xmath17k @xcite .
the flow in certain temporal classes ( and soft spectral states ) of micro - quasar grs 1915 + 105 @xcite is an example of keplerian disk . while the temperature of the latter cases could be same order in magnitude as that in the center of a star , due to much low density they practically do not exhibit any thermonuclear burning .
hence any source of energy therein must be due to magnetic and viscous effects .
on the other hand , the former cases may be favourable for thermonuclear reactions due to their very high temperature , evenif their density is low @xcite .
hence , the energy released / absorbed due to nuclear reactions in an optically thin , hot , accretion flow could be comparable to or even dominating over its viscous counter part @xcite .
being very hot , such flows are also highly ionized and hence expected to be strongly magnetized ( see e.g. @xcite ) rendering magnetic dissipation .
in fact , the celebrated blandford - payne mechanism @xcite ) is based on such a magnetized accretion disk in the keplerian regime , which argues for angular momentum transport due to outflows / jets through the outgoing magnetic field lines , in absence of shear ( turbulent ) viscosity .
now an obvious question arises : what is the value of @xmath6 in optically thin accretion flows ( e.g. advection dominated accretion flow ( adaf ) @xcite , general advective accretion flow ( gaaf ) @xcite ) when the temperature is close to that of the qgp matter in rhic ? primarily guess is that it is not too much different from its lower bound as in equation ( [ ebs ] ) .
is this that simple to anticipate ?
note that accretion flows with very small molecular viscosity ( and then very large reynolds number ) exhibit turbulence and then turbulent viscosity .
if the turbulent viscosity dominates , then the energy dissipated due to viscosity in disks is comparable / dominant to / over the magnetic / nuclear energy , as will be discussed in detail below .
note also that at low densities ( zero chemical potential ) , lattice qcd calculations @xcite suggest a crossover from the hadronic state to the qgp state at around @xmath18k @xcite . at such high temperatures ,
there is expected to be copious pion production in a strongly interacting system .
this qgp state is thought to have existed a few microseconds after the big bang and is being extensively studied at rhic and lhc .
of course the density in early universe is large as well , unlike that in accretion flows , a large temperature is well enough to reveal such a phase which has been mimicked in rhic / lhc .
however , at a very large density ( several orders of magnitude larger than that in accretion flows ) the lattice qcd results are not valid ( zoltan fodor , private communication ) .
recently , sinha & mukhopadhyay @xcite initiated to look at the above issue and argued that the shakura - sunyaev turbulent viscosity parameter @xmath19 @xcite should not be constant throughout the flow , and be decreasing with the increase of temperature and/or density of the flow .
based on the general relativistic model of a viscous adaf , they showed that @xmath6 for an astrophysical black hole , at finite @xmath19 , is always several orders of magnitude higher than its value in a qgp fluid .
then they argued for the flow to become very weekly viscous ( not exhibiting turbulence ) close to a black hole , rendering a smaller @xmath6 .
the results appear to be independent of the choice of equation of state ( eos ) ; whether of ideal gas or of qcd which is appropriate for flows with @xmath20k and low density ( close to zero chemical potential ) .
however , they could not clarify the natural circumstances when @xmath19 is small .
note that small @xmath19 would reveal negligible dissipation of energy in the flow .
hence , question arises what is the source of energy and entropy in the accretion flows , which is also very important in order to address @xmath6 ?
in addition , how is the accretion possible in a very small-@xmath19 flow ? moreover , as will be explained in the next section , above work did not model the entropy of the flow adequately , while generally interpreted the results correctly . instead , that was the first step forward in the direction of evaluating @xmath6 for any astrophysical flow , eventhough incomplete , which is a potential natural site for small @xmath6 . in the present work ,
we have removed the above uncertainties lying in the previous work
. then we plan to establish the underlying physics giving rise to the astrophysical values of @xmath6 .
more precisely , we plan to address : ( 1 ) regime of accretion flows giving rise to @xmath6 close to its theoretical lower bound , ( 2 ) observational implications of such regimes , ( 3 ) black hole sources presumably revealing lower bound of @xmath6 , ( 4 ) constraining physical parameters ( e.g. reynolds number ) of accretion flows based on @xmath6 bound . in the next section ,
we discuss the salient features of hot advective accretion flows and the importance of strong magnetic field and nuclear energy produced therein in order to obtain small @xmath6 .
subsequently in 3 we analyse the solutions of the equations and address the flow parameters making @xmath6 to be close to the lower limit . finally 4 summarizes our findings with discussion and implication .
in order to understand the underlying physics , let us consider the equations describing optically thin , viscous , magnetized , advective accretion disk @xcite .
we describe the model in the pseudo - newtonian framework with the paczyski - wiita potential @xcite .
this is particularly because , for the present purpose , the pure general relativistic results for a rotating black hole qualitatively do not reveal any new physics .
hence , the vertically averaged hydromagnetic equations of energy - momentum balance along with the equations of continuity , induction and divergence of magnetic field in the limit of very large conductivity are @xmath21 assuming that the variables do not vary significantly in the vertical direction such that @xmath22 , which is indeed true for the disk flows . here
@xmath23 is the conserved mass accretion rate , @xmath24 and @xmath25 are the vertically integrated density and density of the flow respectively , @xmath26 is the radial velocity , @xmath27 the total pressure , @xmath28 the specific angular momentum , @xmath29 the vertically integrated shearing stress , @xmath30 the half - thickness , @xmath1 the entropy per unit volume , @xmath8 the temperature of the flow , @xmath31 and @xmath32 are the vertically integrated net energy released and absorbed rates in / from the flow respectively , @xmath33 indicate the polytropic indices depending on the gas and radiation content in the flow ( see , e.g. , @xcite for exact expressions ) and @xmath34 , @xmath35 and @xmath36 are the components of magnetic field .
note that , the independent variable @xmath37 is the radial coordinate of the flow expressed in the units of @xmath38 , where @xmath39 is the gravitation constant , @xmath40 the mass of the black hole and @xmath41 the speed of light .
accordingly , all the above variables are expressed in dimensionless units . for any other details , see the existing literature @xcite .
let us now pay a special attention to the energy equation of the equation set ( [ eqnset ] ) .
we can rewrite it as t = q_vis+q_mag+q_nex - q_rad - q_nend , [ modion ] where @xmath42 , @xmath43 and @xmath44 respectively define the energies released per unit volume per unit time due to viscous dissipation , magnetic dissipation and thermonuclear reactions and @xmath45 and @xmath46 respectively indicate the energy radiated away per unit volume per unit time by various cooling processes like bremsstrahlung , synchrotron and inverse - comptonization of soft photons , and the energy absorbed per unit volume per unit time due to thermonuclear reactions . it was shown earlier that the nuclear energy released / absorbed in a hot accretion disk need not be negligible @xcite and the thermonuclear reactions therein could produce over - abundant metals
@xcite which are observed in galaxies .
however , the value of @xmath19 viscosity determines whether the energy due to nuclear reactions is comparable to that due to the viscous effects @xcite and ohmic dissipation or not . from equation ( [ modion ] )
we can write down the entropy per unit volume , namely the entropy density , of the flow as s = dx .
[ s ] now the turbulent kinematic viscosity @xmath47 , when @xmath48 and @xmath49 are respectively the turbulent eddy velocity and the size of eddy , which are respectively to be scaled linearly with sound speed ( @xmath50 ) of the disk and @xmath30 and hence = = c_s h , [ eta ] where @xmath51 , @xmath52 and hence @xmath19 are constants .
therefore , the shear viscosity is given by = c_s h. [ eta ] in an optically thin flow , often @xmath45 is approximately chosen to be proportional to @xmath53 , in order to understand the hydrodynamic properties of the flow without much loss of generality , such that @xmath54 ( see , e.g. , @xcite ) , when @xmath55 is a constant . for an adaf @xmath56 strictly ( and @xmath57 , when @xmath58 is the eddington accretion rate ) and for a gaaf @xmath59 ( and @xmath23 is larger than that in an adaf @xcite ) .
note that @xmath60 for an optically thick keplerian accretion disk .
hence , following previous work ( e.g. @xcite ) , we can write q_vis+q_mag - q_rad = f(q_vis+q_mag ) = f((i_n+1c_s^2+i_n^2)()+ ) , [ qvis ] where @xmath61 .
therefore , if @xmath62 , which we name as viscous regime , @xmath6 does not explicitly depend on @xmath19 ( and @xmath0 ) , , when @xmath63 .
] depends only on the hydrodynamic quantities whose numerical values , in a fixed range of radii , do not vary significantly in the optically thin regime with the change of @xmath19 . in the viscous regime , @xmath6 simply scales with @xmath40 .
note that viscosity depends on the size of the system , which in turn is determined by the mass of the black hole .
hence , only for a primordial black hole of @xmath64gm , @xmath6 is close to its theoretical lower bound , as was suggested earlier @xcite .
it was then also suggested that @xmath19 has to be very small in an adaf to obtain @xmath65 for an astrophysical black hole .
however , for a small @xmath19 , viscous dissipation ( and then @xmath1 ) in an adaf ( or generally in a gaaf ) should be small as well , as is apparent from equations ( [ modion ] ) and ( [ qvis ] ) .
hence , the ratio of @xmath0 to @xmath1 should not be invariably small , as described above , in the viscous regime .
therefore , while the earlier work @xcite is very interesting and the first attempt to evaluate @xmath6 in an astrophysical flow , the flow ( and hence any conclusion ) should be amended with extra physics . in other words
, the conclusion is incomplete .
indeed @xmath6 could be small at a small @xmath19 when @xmath1 does not decrease in that extent with the decrease of @xmath19 . however , this is possible only when @xmath66 or / and @xmath67 , which we name magnetic and nuclear regimes respectively .
the magnetic regime essentially corresponds to the blandford - payne mechanism in the hard spectral state , when accretion takes place via magnetic stress , independent of the presence of shear / turbulent viscosity @xcite .
this plausibly corresponds to the case of accretion flows with outflows / jets , e.g. sgr a@xmath16 , hard states of grs 1915 + 105 .
hence , even at a negligibly small @xmath19 accretion is possible with a negligible @xmath6 .
on the other hand , the nuclear reactions and corresponding energies depend on the density and temperature of the system .
as shown earlier @xcite that even in the inviscid limit , temperature and density of the optically thin flows remain as high as those of a viscous accretion flow for a fixed @xmath40 , which successfully explain many observation @xcite .
indeed , it was shown @xcite that significant nuclear energy ( @xmath68 ergs / cc / sec ) throughout the disk may be released / absorbed in an inviscid flow which could lead to disk instability .
hence , from equation ( [ s ] ) it is naturally possible that @xmath1 is large even at a small @xmath19 .
therefore , @xmath69 also plausibly in the nuclear regime of the accretion flow , when @xmath70 which could be the case in the hot accretion flows @xcite .
let us understand the above physics quantitatively .
figures [ m10b]a , b , c show hydrodynamic / hydromagnetic properties of a disk around a @xmath71 nonrotating black hole in the magnetic regime , accreting mass at the rate @xmath72 . in this case of weakly viscous hot flow with @xmath73 ( but not zero exactly ) , @xmath6 is shown to be close to the theoretical lower bound @xmath74 .
for comparison , @xmath6 is also shown for a viscous regime , which is quite large for @xmath75 and @xmath76 .
we also show the variations of mach number and magnetic field in the disk for the cases mentioned above for reference .
interestingly , increase of magnetic field in the magnetic regime increases rate of transport which further renders advancing the keplerian disk towards the black hole , making the sub - keplerian regime , which is shown here , smaller . figures [ m10b]d , e , f depict the properties for @xmath77 in the magnetic regime . a very small ( but nonzero ) @xmath19 ( @xmath78 ) reveals @xmath6 to exhibit the theoretical lower limit .
the impact of change of the magnetic field strength to the disk is same as that in the high accreting case discussed above .
interestingly , independent of the accretion rate , magnetic field @xmath79 kg is able to transport matter efficiently in the sub - keplerian flow in absence of viscosity , as was already argued by blandford & payne @xcite in the keplerian , self - similar framework .
a smaller field strength would also transport matter as well , except with the increase of size of the sub - keplerian regime .
[ ht ] in cgs unit when the horizontal line ( long - dashed ) refers to @xmath74 , as functions of flow radius for @xmath80 at @xmath81 .
solid line corresponds to the flow in viscous regime and dotted and dashed lines correspond to the flows in magnetic regime .
( d ) , ( e ) , ( f ) describe same as in ( a ) , ( b ) , ( c ) respectively , except for @xmath82 .
see table 1 for other details.,title="fig:",scaledwidth=82.0% ] + figure [ me7b ] is devoted for hydrodynamic / hydromagnetic properties around a @xmath83 nonrotating black hole in the magnetic regime . while figs .
[ me7b]a , b , c correspond to @xmath84 , figs .
[ me7b]d , e , f correspond to @xmath85 . in either of the cases ,
very weakly viscous ( @xmath86 for @xmath84 and @xmath87 for @xmath85 ) hot flows reveal small @xmath6 , close to its theoretical lower bound @xmath74 . for comparison
, @xmath6 is also shown for a viscous , high accreting regime which is quite large for @xmath75 and @xmath88 .
interestingly , magnetic field @xmath79 mg is able to transport matter efficiently in the sub - keplerian flow around a supermassive black hole in absence of viscosity , as was already argued by blandford & payne @xcite in the keplerian , self - similar framework . as is the case for a stellar mass black hole , a smaller field strength would also transport matter , except the fact that the keplerian to sub - keplerian transition would arise at a larger radius .
[ ht ] in cgs unit when the horizontal line ( long - dashed ) refers to @xmath74 , as functions of flow radius for @xmath89 at @xmath81 .
solid line corresponds to the flow in viscous regime and dotted line corresponds to the flow in magnetic regime .
( d ) , ( e ) , ( f ) describe same as in ( a ) , ( b ) , ( c ) respectively , except for @xmath90 . see table 2 for other details.,title="fig:",scaledwidth=82.0% ] + let us also look into the nuclear regime .
figure [ rad ] shows hydrodynamic properties of a disk around a @xmath71 nonrotating black with @xmath91 in the nuclear regime . in the cases of weakly viscous hot flow , @xmath6
is shown to be inversely proportional to the nuclear energy released per unit volume . when the ratio of nuclear energy per unit volume to viscous energy per unit volume is chosen to be @xmath92 ( which corresponds to @xmath93 erg / sec , depicted by the solid line ) , @xmath6 appears to be close to the theoretical lower bound @xmath74 .
note that the production of nuclear energy also depends on the residence time of matter in the disk which is determined by @xmath19 , polytropic constant @xmath94 , initial mass fraction of the inflowing isotopes etc . for comparison
, @xmath6 is also shown for a viscous regime , which is quite large for @xmath75 and @xmath76 .
we also show the variations of mach number and temperature in the disk for the cases mentioned above for reference .
interestingly , the latter profiles overlap each other in the nuclear regime , independent of @xmath95 , and are very similar to that in the viscous regime .
this also verifies that the mach number and temperature do not depend much on @xmath19 , in the optically thin flows . in fig .
[ mix ] , we show a couple of sets of solutions for highly sub - eddington hotter flows and a solution set for a low angular momentum , high mass accreting , hot flow , in the nuclear regimes . interestingly , a very low angular momentum ( @xmath96 ) , gas dominated , hot accretion flow with @xmath90 ( solid line ) violates the theoretical lower limit of @xmath6 .
this restricts @xmath28 of the flow .
if @xmath28 increases from @xmath97 to @xmath98 , the flow becomes relatively cooler due to decrease of radial velocity which restricts nuclear energy , rendering slightly larger @xmath99 ( dotted line ) .
a flow with similar @xmath28 ( @xmath100 ) and @xmath19 , but @xmath91 and hence revealing a radiation dominated accretion ( dashed line ) flow , exhibits however a similar @xmath101 .
note that while the nuclear energy among the three cases considered here is highest in the last flow , due to presence of radiation the cooling efficiency therein is maximum , which renders its @xmath6 above the lower limit throughout .
[ ht ] in cgs unit , as functions of flow radius for @xmath80 at high mass accretion rates .
solid and dotted lines correspond to the flows of very high and high nuclear regimes respectively and dashed line corresponds to the flow of viscous regime .
see table 3 for other details.,title="fig:",scaledwidth=62.0% ] + 0.2 cm table 1 parameters for the results shown in figure [ m10b ] when @xmath80 [ cols="^,^,^,^,^,^,^,^,^ " , ] [ ht ] in cgs unit , as functions of flow radius for @xmath80 .
solid and dotted lines correspond to the flows of very high and high nuclear regimes respectively at low mass accretion rates and dashed line corresponds to the flow of a nuclear regime at a high mass accretion rate .
see table 3 for other details.,title="fig:",scaledwidth=62.0% ] +
we have shown the astrophysical existence of small @xmath6 , close to its theoretical lower bound . as our physical inputs , governing the values of @xmath6 , are in accordance with observed data , this could be argued as a natural evidence for @xmath6 close to its lower bound .
this is possible in the cases of hot , optically thin , accretion flows .
more precisely , in order to have a small @xmath6 , flow should be magnetically dominated and/or producing huge energy / entropy other than that due to viscosity . the celebrated blandford - payne mechanism @xcite was already shown to produce accretion via magnetic stress ( which helps in producing outflows / jets and removing angular momentum ) in absence of viscosity in the framework of keplerian , self - similar flow .
we have employed this idea , but relaxing self - similar assumption , in order to obtain @xmath6 close to @xmath74 in our accretion flows .
we have found that in the presence of magnetic field of the order of mg to kg , the disk flows exhibit this value of @xmath6 for supermassive to stellar mass black holes . observed hard x - rays in cyg x-1 ,
some temporal classes ( e.g. @xmath102 ) of grs 1915 + 105 , sgr a@xmath16 are observed to be associated with jets which argues for the sources to exhibit strong magnetic fields .
hence , those sources could be considered to be natural sites for @xmath69 .
additionally , the flow exhibiting nuclear energy could be a natural site for small @xmath6 .
nuclear energy depends on the density and temperature of the flow .
the temperature in an optically thin , sub - keplerian accretion flow , producing observed hard x - rays , could be @xmath7k .
therefore , eventhough the density therein is small , very high temperature would suffice for nuclear burning .
we have recalled the previous work discussing possible nuclear burning in accretion flows and its observational evidences / consequences @xcite . based on them
, we have modeled the hot , optically thin accretion flows for the present purpose and shown how the nuclear energy in the flow determines the value of @xmath6 .
thus we argue that the flows producing hard x - rays in cyg x-1 , some temporal classes ( e.g. @xmath102 ) of grs 1915 + 105 could be hot enough to have @xmath69 , atleast close to the black hole for certain nuclear regimes , when flows are necessarily sub - keplerian due to the dominance of gravitational power of the black hole .
it will be now interesting to explore @xmath6 for other astrophysical flows , having , e.g. , strong magnetic field or / and high temperature and/or density .
finally , we end by enlightening maximum reynolds number ( @xmath103 ) to the flows , restricted by the lower bound of @xmath6 .
@xmath103 , which is the ratio of typical ( characteristic ) velocity times length scale to the kinematic viscosity of the flow , can be recast as @xmath104 .
now from figs .
[ m10b ] and [ rad ] , @xmath105 close to the black hole is @xmath106 .
however , the magnetic and nuclear regimes correspond to very low but nonzero viscosity revealing @xmath107 for @xmath6 lower bound to be satisfied . as typically @xmath108 for an optically thin , geometrically thick flow , for stellar mass black holes @xmath109 . if we look at fig .
[ me7b ] showing @xmath105 to be of the similar order in magnitude around supermassive black holes , however for smaller values of @xmath19 ( @xmath110 ) in larger accretion disks ( as size scales with the mass of the black hole ) , we obtain @xmath111 for those flows .
* acknowledgments * : the author would like to thank zoltan fodor for comments and suggestions along with bringing in his attention to important references . thanks are also due to upasana das and aninda sinha for discussion .
this work was partly supported by an isro grant isro / res/2/367/10 - 11 .
y. aoki . , z. fodor , s. d. katz , & k. k. szabo , phys .
b * 643 * , 46 ( 2006 ) ; y. aoki , s. borsanyi , s. durr , z. fodor , s. d. katz , s. krieg , & k. k. szabo , jhep * 0906 * , 088 ( 2009 ) ; s. borsanyi , z. fodor , c. hoelbling , s. d. katz , s. krieg , c. ratti , & k. k. szabo , jhep * 1009 * , 073 ( 2010 ) . | string theory and gauge / gravity duality suggest the lower bound of shear viscosity ( @xmath0 ) to entropy density ( @xmath1 ) for any matter to be @xmath2 , when @xmath3 and @xmath4 are reduced planck and boltzmann constants respectively and @xmath5 .
motivated by this , we explore @xmath6 in black hole accretion flows , in order to understand if such exotic flows could be a natural site for the lowest @xmath6 .
accretion flow plays an important role in black hole physics in identifying the existence of the underlying black hole .
this is a rotating shear flow with insignificant molecular viscosity , which could however have a significant turbulent viscosity , generating transport , heat and hence entropy in the flow .
however , in presence of strong magnetic field , magnetic stresses can help in transporting matter independent of viscosity , via celebrated blandford - payne mechanism .
in such cases , energy and then entropy produces via ohmic dissipation .
in addition , certain optically thin , hot , accretion flows , of temperature @xmath7k , may be favourable for nuclear burning which could generate / absorb huge energy , much higher than that in a star .
we find that @xmath6 in accretion flows appears to be close to the lower bound suggested by theory , if they are embedded by strong magnetic field or producing nuclear energy , when the source of energy is not viscous effects .
a lower bound on @xmath6 also leads to an upper bound on the reynolds number of the flow .
* keywords : * infall , accretion , and accretion disks ; gauge / string duality ; shear rate dependent viscosity ; magnetic field ; black holes ; relativistic plasmas + * pacs no .
: * 98.62.mw ; 11.25.tq ; 83.60.fg ; 97.60.lf ; 52.27.ny + |
galaxy evolution is a process driven largely by the lifecycle of gas and its accretion , conversion into stars , and eventual loss via supernova - driven winds . during the peak epoch of galaxy formation ( @xmath0 ; @xcite )
those systems that dominate the luminosity function are especially gas - rich , with gas frequently accounting for 50% or more of the total baryonic mass ( e.g. , @xcite ) .
this large supply of hydrogen fuels sustained star formation rates ( sfr ) commonly @xmath9 yr@xmath10 ( e.g. , @xcite ) and sfr surface densities peaking above @xmath11 yr@xmath10 kpc@xmath12 ( e.g. , @xcite ) that are comparable to those in transient starbursts in the modern universe ( @xcite ) . the mechanisms by which these galaxies acquire such large quantities of gas are uncertain ; although gas has traditionally been assumed to accrete primarily through mergers ( e.g. , @xcite ) and ` hot mode ' accretion via virial shocks ( e.g. , @xcite ) , some recent arguments suggest that semi - continuous cold accretion from cosmological filaments ( e.g. , @xcite ) may be required to sustain the observed sfr and produce the structures observed at @xmath13 . as discussed by reddy et al . (
2012 ) however , star formation at @xmath0 appears to have been an inefficient process , indicating that sfr is unlikely to have been limited primarily by the cold gas accretion rate .
in contrast to galaxies in the nearby universe , @xmath13 star forming galaxies do not exhibit regular spiral structure ( with some notable exceptions ; @xcite ) and instead tend to be compact , triaxial systems consisting of one or more irregularly - shaped clumps of emission ( e.g. , @xcite ; and references therein ) .
in many of the most massive galaxies ( @xmath14 ) such clumps may be embedded in low surface - brightness structures reminiscent of thick disks with typical scaleheights @xmath15 kpc ( e.g. , elmegreen & elmegreen 2006 ; genzel et al .
2006 , 2008 ; frster schreiber et al .
2011a , 2011b ) sustained by their large vertical velocity dispersions ( e.g. , genzel et al .
2006 ; law et al .
2007 , 2009 ; frster schreiber et al . 2009
; swinbank et al . 2011 ) .
the large cold gas fractions of these galaxies appears to play a significant role in their dynamical and morphological evolution .
simulations ( e.g. , noguchi 1999 ; immeli et al .
2004 ; bournaud et al .
2007 ; bournaud & elmegreen 2009 ) indicate that such gas - rich galaxies should be highly unstable to fragmentation , forming clumps similar to those observed which may either migrate to form a central bulge ( immeli et al .
2004 ; elmegreen et al . 2008 ; ceverino et al . 2010 ) or be rapidly disrupted by stellar feedback ( wuyts et al .
a reliable understanding of the relation between galactic stellar morphology and its gas - phase properties is therefore critical for constraining models of galaxy formation in the young universe . while observational signatures of _ infalling _ gas ( whether hot or cold - mode ) have proven subtle and difficult to detect ( e.g. , van de voort & schaye 2012 ; although c.f .
rauch et al .
( 2011 ) and @xcite ) , the intense star formation of the galaxies gives rise to ubiquitous and well - studied galactic - scale _ outflows _ evident in rest - frame uv spectra ( e.g. , shapley et al .
2003 ; weiner et al . 2009 ; steidel et al . 2010 ) .
these outflows are frequently powerful enough to drive enriched gas to the virial radius and beyond with mass loss rates comparable to or exceeding the star formation rate ( @xcite ) .
numerous authors ( e.g. , shapley et al . 2003 , penterrici et al .
2010 ; kornei et al . 2010 ; steidel et al . 2010 ; jones et al .
2012 ) have therefore sought to use the information about optical depth , covering fraction , geometry , and gas kinematics encoded within such spectra to learn about the mechanisms by which cold gas is coupled to feedback and physical observables such as stellar mass , star formation rate , and stellar population age .
efforts to understand the role of the gas in shaping the evolution of galaxy morphology however ( e.g. , law et al . 2007b ; penterrici et al .
2010 ) have historically been limited in their ability to test current galaxy formation models by the lack of availability of high resolution rest - frame optical imaging tracing the bulk of the stellar mass . in the present contribution , we combine rest - frame optical imaging of a large sample of @xmath13 star forming galaxies recently obtained with the wfc3 camera on board the _ hubble space telescope _ ( _ hst _ ) with ground - based rest - uv spectroscopy to investigate the relation between galaxy morphology and gas - phase kinematics . in section [ obs.sec ]
we describe the target galaxy sample and detail the methods by which morphological , photometric , and spectroscopic data were obtained .
we discuss the ly@xmath16 emission properties of the sample in
[ lya.sec ] , and the structure of major interstellar absorption lines arising in outflowing cold gas in [ metals.sec ] , highlighting the changing velocity substructure of the absorption lines as a function of galaxy morphology .
we discuss the implications of our findings for the evolution of typical @xmath0 star forming galaxies in
[ discussion.sec ] , combining our results with recent observations from ifu kinematic surveys .
we summarize our conclusions in [ summary.sec ] . throughout our analysis
, we adopt a standard @xmath17cdm cosmology based on the seven - year wmap results ( komatsu et al .
2011 ) in which @xmath18 km s@xmath10 mpc@xmath10 , @xmath19 , and @xmath20 .
our galaxy sample is drawn from the keck baryonic structure survey ( kbss ; trainor & steidel 2012 ) , a catalog of @xmath21 star forming galaxies selected according to optical @xmath22 color ( @xcite ) and confirmed using rest - uv spectroscopy .
these galaxies typically lie near sightlines to hyperluminous background qsos ( @xmath23 ) scattered widely across the sky .
we define two specific samples of galaxies used in our analysis : parent sample : : : a sample of 204 galaxies with _ hst_/wfc3 imaging data , magnitudes @xmath24 ab , and redshifts @xmath25 ( see fig . [ parmrange.fig ] ) derived from rest - uv spectroscopy .
hahq subsample : : : a subgroup of 35 galaxies from the parent sample for which the rest - uv spectroscopy is particularly high - quality ( ` uvqual ' @xmath26 ; see [ uvspectra.sec ] ) and secure systemic redshifts in the range @xmath27 ( see fig . [ parmrange.fig ] ) have been obtained from h@xmath16 nebular emission line spectroscopy . as part of _ hubble space telescope _ ( _ hst _ ) cycle 17 program go-11694 , we obtained wfc3/ir imaging of 306 @xmath22 color - selected star forming galaxies with spectroscopic redshifts in the range @xmath28 drawn from the kbss .
the details of these observations have been described at length by law et al .
( 2012a ; hereafter ` paper i ' ) . in brief
, we used the f160w ( @xmath29 ) filter to trace rest - frame optical emission ( @xmath30 for the parent sample , @xmath31 for the hahq sample ) .
our data reach a depth of 27.9 ab for a @xmath32 detection within a 0.2 arcsec radius aperture ( representing a total of 8100s integration in each pointing ) , and have a psf fwhm of 0.18 that corresponds to 1.5 kpc at @xmath33 .
the morphologies consist of three qualitative types .
as defined in paper i , these are type i : : : single nucleated component with little to no low surface brightness emission .
type 2 : : : two or more distinct nucleated components of comparable magnitude and little to no low surface brightness emission .
type 3 : : : extended objects with non - axisymmetric low surface - brightness features .
numerous morphological statistics have been defined in the literature that attempt to encapsulate this range of morphologies quantitatively . in paper
i we calculated 7 of the most widely used such statistics for each of the galaxies in our wfc3 imaging fields .
these statistics are : @xmath34 : : : semi - major axis radius derived from fitting sersic ( 1963 ) models convolved with the observational psf to the f160w light profile using the galfit ( peng et al .
2002 , 2010 ) analysis package .
we also define the circularized effective radius ( @xmath35 ) for consistency with previous analyses , where @xmath36 is the apparent minor / major axis ratio .
@xmath37 : : : radial index of the best - fit sersic model profile .
@xmath38 : : : gini statistic ( abraham et al .
2003 ; lotz et al .
2004 ) quantifying the cumulative flux distribution among the galaxy pixels .
@xmath39 : : : second order moment of the spatial distribution of the light profile ( lotz et al .
2004 , 2006 ) .
@xmath40 : : : concentration index of the flux distribution about the center of the galaxy ( bershady et al .
2000 ; conselice 2003 ) .
@xmath41 : : : asymmetry index ( schade et al .
1995 ; conselice et al .
2000 ) quantifying the @xmath42 rotational asymmetry of the galaxy .
@xmath43 : : : multiplicity index ( law et al .
2007b ) quantifying the ` work ' required to assemble the distribution of pixel fluxes within a galaxy ( distinguishing single - component from irregular and multiple - component sources ) .
although the first two morphological statistics ( @xmath44 ) assume a parametric model convolved with the observational psf , the latter five ( @xmath45 ) are non - parametrically derived from the spatial arrangement of individual pixel fluxes .
the values calculated for these five statistics depend systematically on the adopted segmentation map defining which pixels are considered to be part of a given galaxy . our approach is placed on the systematic reference frame of lotz et al .
( 2008a , 2008b , and references therein ) using the transformations described in appendix a.1 of paper i. all seven morphological statistics are robust for galaxies with @xmath24 ab , and show systematic bias at fainter magnitudes ( see appendix a.2 of paper i ) .
we therefore restrict our attention in the present contribution to those 204 galaxies with @xmath46 and no evidence of qso or agn in their uv spectra ( [ uvspectra.sec ] ) or broadband photometry ( [ sed.sec ] ) .
as derived in paper i , typical uncertainties for individual galaxies are @xmath47 2% in @xmath34 , 15% in @xmath37 , 3% in @xmath38 , 4% in @xmath39 , 11% in @xmath40 , 22% in @xmath41 , and 21% in @xmath43 .
rest - frame uv spectra for each of our galaxies have been obtained using the lris multislit spectrograph on keck i. details concerning the observational setup , data reduction , and calibration process of these spectra have been presented at length elsewhere in the literature ( e.g. , shapley et al . 2003 ; steidel et al . 2010 ; and references therein ) . in general , individual spectra represent integration times @xmath47 1.5 hours with characteristic velocity resolution @xmath48 @xmath6 .
these spectra vary considerably in quality according to both the intrinsic brightness of the target galaxy and the conditions under which particular slit masks were observed .
we sort the galaxies according to the quality of their spectra , assigning them quality indices ` uvqual ' ranging from 0 ( one or two spectroscopic features identified with relatively low confidence ) to 3 ( multiple strong spectroscopic features identified with extremely high confidence ) . the limited s / n ratio of even the highest - quality spectra however means that it is typically only possible to estimate the mean interstellar absorption ( emission ) line redshift @xmath49 ( @xmath50 ) for individual galaxies . in order to investigate the detailed structure of spectral lines
it is necessary to stack together multiple spectra to reach higher s / n ratios .
this stacking process is described in detail by steidel et al .
( 2010 ) , and includes doppler - correcting each spectrum to the rest frame , , otherwise we adopt the prescriptions given by steidel et al .
( 2010 ) . ]
resampling to a common 0.33 / pixel spectral dispersion , and rescaling to a common mode in the wavelength range @xmath51 before combining individual spectra using a sigma - clipped mean algorithm . in figure [ specstack.fig ] we show the composite spectrum for the 204 galaxies in our parent sample
. this composite spectrum exhibits a mixture of ly@xmath16 absorption and emission , deep blended interstellar absorption features due to various metal species in the galactic ism , and a host of weaker features due to stellar photospheric absorption and fine structure emission .
these absorption features are typically blueshifted with respect to the systemic redshift ( corresponding to absorption arising in the near side of an outflowing galactic wind ) , while ly@xmath16 emission is redshifted ( corresponding to off - resonance backscattering from the far side of the outflow ) . as described at length by steidel et al .
( 2010 ; see also kulas et al .
2012 ) , this outflow is likely composed of discrete gas clouds whose expansion velocities increase with galactocentric radius .
although the rest - uv spectra give an indication of the redshift of the interstellar absorption lines @xmath49 , it is not possible to perform a detailed analysis of the kinematic structure of the outflowing gas without knowing the systemic redshift @xmath52 of the galaxy as well .
we therefore define the ` h@xmath16 - high quality ' ( hahq ) subsample of galaxies with particularly robust measurements of @xmath49 and for which accurate ( uncertainty @xmath53 @xmath6 ; steidel et al .
2010 ) systemic redshifts have been derived from h@xmath16 nebular emission using a combination of long - slit keck / nirspec ( erb et al .
2006b ) , integral - field keck / osiris ( law et al .
2007a , 2009 ; wright et al . 2009 ) and vlt / sinfoni ( frster schreiber et al .
2009 ) spectroscopy .
the hahq sample represents the subset of 35 of the 89 galaxies considered by steidel et al .
( 2010 ) for which _
morphologies have been obtained .
extensive kbss ancillary data exist for our target galaxies , typically including deep ground - based optical and infrared @xmath54 photometry , and in some cases _ spitzer _ irac and/or mips photometry . in combination with @xmath55 magnitudes derived from the _
hst_/wfc3 imaging data , the long - wavelength photometry permits us to construct stellar population models by fitting the broadband spectral energy distribution of each galaxy as detailed in paper i ( see also shapley et al .
2001 , 2005 ; erb et al . 2006c ; reddy et al . 2006 , 2010 ) .
in brief , we fit charlot & bruzual ( 2012 , in prep . ) stellar population synthesis models in combination with a chabrier ( 2003 ) imf , calzetti et al .
( 2000 ) extinction law , and a constant ( @xmath56 ) star formation history using a customized idl code ( reddy et al .
monte carlo tests for the 204-galaxy parent sample indicate that the mean fractional uncertainties in derived values for @xmath57 , age , sfr , and @xmath58 are @xmath59 0.3 , 0.6 , 0.5 , and 0.4 respectively ( consistent with previous measurements for the larger kbss sample by shapley et al .
2005 , erb et al . 2006b , and reddy et al .
2012 ) . as illustrated in figure [ parmrange.fig ] ,
galaxies in the parent sample have stellar masses in the range @xmath60 , star formation rates @xmath9 yr@xmath10 , and moderate color excess @xmath61 .
the hahq subsample is biased relative to the parent sample by the requirement of both detecting h@xmath16 emission and observing a high - quality uv spectrum ( favoring uv - bright objects ) , and therefore has a slightly larger mean stellar mass ( @xmath62 vs. @xmath63 ) values represent standard deviation about the mean . ] .
ly@xmath16 emission originates deep within individual star forming regions and is resonantly scattered throughout the interstellar medium until either being absorbed on a dust grain or redshifted sufficiently far off - resonance that it can escape the galaxy .
although almost all star forming galaxies appear to have ly@xmath16 emitting haloes on large angular scales ( @xcite ) , the ly@xmath16 profile of the central few kpc traced by slit spectra is dependent upon the structure , kinematics , and ionization properties of each galaxy and its interstellar and circumgalactic medium .
substantial efforts have been made over the last decade to attempt to disentangle these various effects and connect the central ly@xmath16 line profile with global stellar population properties both for narrowband - selected ly@xmath16-emitters ( laes ; e.g. , gawiser et al .
2007 ) and for star forming galaxies selected via broadband optical color .
@xcite , for instance , found that @xmath64 lyman break galaxies ( lbgs ) with stronger @xmath2 emission tend to have lower sfr , bluer uv continua , and weaker low - ionization interstellar absorption lines .
similar results were found for @xmath13 star forming galaxies by @xcite and @xcite , who noted that galaxies with stellar masses @xmath65 on average have weaker @xmath2 emission than do galaxies with @xmath66 , and at @xmath67 by jones et al .
these results were borne out by the more detailed analysis of @xcite , which demonstrated that @xmath2 emission is both weaker and more strongly redshifted from the systemic velocity in galaxies with high baryonic mass , consistent with the picture that @xmath2 emission from the deepest gravitational potential wells must scatter further from resonance before being able to escape . using a similarly selected sample of lbgs at @xmath68 , @xcite found comparable results , showing that galaxies without @xmath2 emission tend to be more massive and dustier , although these authors observed that stellar population ages and sfrs did not depend strongly on @xmath2 emission characteristics .
the relation between @xmath2 emission and rest - frame uv galaxy morphology has previously been investigated by @xcite and @xcite ( see also bond et al .
2009 , malhotra et al .
@xcite found that @xmath2 emission was correlated with the gini coefficient in the sense that high-@xmath38 ( i.e. , strongly nucleated ) systems were more likely to exhibit @xmath2emission , which we previously ascribed to the lower quantity of dust in strongly nucleated galaxies .
although @xcite found no significant correlation of @xmath2 emission with any of the non - parametric morphological indices ( @xmath69 , or @xmath39 ) , these authors noted that while galaxies with weak or no @xmath2 emission spanned a range of projected angular sizes , galaxies with strong @xmath2 emission tended to be exclusively small . in [ lya_prob.sec ] we discuss the probability that galaxies within our sample exhibit @xmath2 emission as a function of various physical properties , and expand our discussion in [ lya_strength.sec ] to consider the trends in emission line strength within the @xmath2-emitting subset .
we divide the parent sample into two groups according to whether @xmath2 emission is present ( 71 galaxies ) or not ( 117 galaxies ) in the uv spectrum of each galaxy . from our sample since @xmath2 emission at these redshifts does not fall within the lris uv spectral range . ]
the first of these two groups includes galaxies with @xmath2 emission of any strength , including those with a small emission peak superimposed on a large @xmath2 absorption trough .
we later discuss the properties of a ` @xmath2-bright ' subgroup which meets the @xmath3 selection criterion used by dedicated surveys of @xmath2-emitters ( laes ; e.g. , gawiser et al . 2007 ) . in table
[ lyacorrel.tab ] we list probabilities calculated using the nonparametric kolmogorov - smirnov ( ks ) test to evaluate whether the two groups are statistically consistent with the null hypotheses that they are drawn from the same distribution of stellar mass , sfr , morphology , etc .
we find that sfr and the morphological parameters @xmath39 , @xmath40 , @xmath41 , @xmath43 , and @xmath37 are poor predictors of the likelihood of observing @xmath2 emission from a galaxy ; galaxies with and without @xmath2 in emission have statistically indistinguishable distributions .
in contrast , the null hypothesis is rejected at greater than 94 - 95% confidence when we examine the mass , age , color excess , @xmath38 , or @xmath34 distributions for galaxies both with and without @xmath2 in emission .
these correlations are in the sense that galaxies with @xmath2 emission on average are smaller , may have slightly lower mass ( @xmath70 vs @xmath71 ) , and slightly younger ages ( 500 vs 800 myr ) , and are less dusty ( @xmath72 vs 0.23 ) than galaxies with no @xmath2 emission .
these results are broadly consistent with relations that have already been established by @xcite , @xcite , @xcite , @xcite , and @xcite .
lc @xmath58 & * 6% * + age & * 8% * + sfr & 99 % + sigmasfr & * 1 % * + e(b - v ) & * 3% * + @xmath38 & * 8% * + @xmath39 & 83% + @xmath40 & 26% + @xmath41 & 99% + @xmath43 & 56 % + @xmath34 & * 3% * + @xmath73 & * 0.5% * + @xmath37 & 15 % + @xmath36 & 59 % + @xmath55 & 12 % + [ lyacorrel.tab ] we note , however , that @xcite studied the ly@xmath16 properties of @xmath64 lbgs and found a conflicting result that those with @xmath3 tended to have stellar populations slightly _ older _ than their non - emitting kin , which was hypothesized to indicate that @xmath2 emitters may represent a later stage in galaxy evolution in which supernova - driven outflows have reduced the dust covering fraction ( see also shapley et al .
to define their @xmath2-emitting population , in contrast to our simple criterion that there be any measurable @xmath2 emission . however , our results are qualitatively unchanged if we instead consider the @xmath2-bright subset of galaxies with @xmath3 . ]
the difference with our present findings may partly reflect the different mean redshift range of the samples , or may indicate that the relation between @xmath2 emission strength and stellar population age is simply too weak ( 96% rejection of the null hypothesis ) to measure consistently given typical uncertainties in the derived ages of individual galaxies .
indeed , while @xcite concurred that @xmath2 emitters tended to be slightly younger than similar - selected non - emitters on average , these authors too noted that the correlation was only marginally significant .
as indicated by table [ lyacorrel.tab ] , within our sample of star forming galaxies there is no statistically significant correlation between the presence / absence of @xmath2 emission and the morphological indices @xmath40 , @xmath41 , @xmath39 , @xmath43 , or @xmath37 .
similar to the result obtained by @xcite , there is a relation ( 92% confidence ) between @xmath2 emission and the gini parameter @xmath38 in the sense that galaxies with @xmath2 emission tend to have marginally more strongly nucleated light profiles in the rest - frame optical .
this trend does not appear to be of statistical significance however as the mean values of @xmath38 for galaxies with and without @xmath2 emission respectively differ by only @xmath74 ( @xmath75 vs @xmath76 ) .
uncertainty in the mean of each sample . ] in contrast , the presence of @xmath2 emission is strongly correlated with the morphological semi - major axis radius @xmath34 , circularized radius @xmath73 , and ( by construction ) the star formation rate surface density @xmath77 .
these correlations are significant at the 97 - 99% confidence level , comparable to or stronger than the well - known anti - correlation between @xmath2 emission and dust content @xmath57 ) .
similarly to the rest - uv angular size relation noted by @xcite , galaxies with @xmath2 in emission are more likely than non - emitters to have @xmath78 kpc ( 74% vs 56% ) , and the galaxies with @xmath78 kpc in our sample are twice as likely ( 41% vs 24% ) as those with @xmath79 kpc to have ly@xmath16 in emission ( see figure [ rlyahist.fig ] ) .
this is unlikely to be due to any intrinsic correlation between semi - major axis radius and dust content ; although @xmath34 and @xmath57 of the parent sample are marginally correlated at the @xmath80 level ( based on the spearman rank correlation test ) , this correlation is in the sense that larger galaxies tend to be marginally _ less _ dusty .
we illustrate the spectroscopic difference between large- and small-@xmath34 galaxies by stacking the halves of the parent and hahq samples with the smallest and largest semi - major axis radii ( figure [ fullspec.fig ] ) .
the @xmath2 emission equivalent width @xmath7 of the large and small-@xmath34 stacks is 1.3 and 2.6 respectively in the parent sample , representing a factor @xmath81 increase in the mean @xmath2 emission strength . of
the stacks typically vary by @xmath82% for the parent sample , and by a factor @xmath81 for the hahq subsample . ]
this difference is even more pronounced in the hahq subsample , increasing by a factor @xmath83 from the large-@xmath34 ( @xmath7 = 0.7 ) to small-@xmath34 ( @xmath7 = 6.7 ) stacks , corresponding to a k - s probability of @xmath84% that the radii of galaxies with and without @xmath2 in emission are drawn from the same distribution . within the group of galaxies with @xmath2 in emission similar correlations exist between @xmath2 emission _ strength _ and physical galaxy properties .
we compute the emission line equivalent width @xmath7 relative to the continuum level in our flux - calibrated spectra for each of the 71 galaxies with such emission in the parent sample . galaxies with @xmath2 emission in the kbss
, we estimate that uncertainties on @xmath7 for individual galaxies are typically @xmath85 . ] in table [ lyastrength.tab ] we tabulate the correlations between @xmath7 and physical galaxy properties in terms of the number of standard deviations from the null hypothesis ( that the two variables are uncorrelated ) using the spearman rank correlation test .
the factors most strongly influencing @xmath7 are the dust content @xmath57 of a galaxy ( @xmath86 ) , its total star formation rate ( @xmath86 ) , in emission , for the subsample of galaxies with @xmath2 in emission we find the sfr to be correlated with the strength @xmath7 of the emission . ] and its physical size ( @xmath87 ) , in the sense that larger , dustier galaxies with higher sfr have lower @xmath7 .
we illustrate these three trends in figure [ lyastrength.fig ] . similarly to previous studies ( e.g. , shapley et al .
2003 ; kornei et al .
2010 ; penterrici et al .
2010 ) , we find that galaxies with the strongest @xmath7 tend to contain lower than average quantities of dust and have lower star formation rates .
additionally , galaxies with large rest - optical sizes ( @xmath88 kpc ) tend to have @xmath89 , while the smallest galaxies ( @xmath1 kpc ) exhibit a much wider range in equivalent width from @xmath90 .
of the 19 star - forming galaxies in our optical - color selected sample with @xmath91 all but one have rest - optical radii @xmath4 kpc and type i and ii morphologies ( i.e. , one or two nucleated components to the light profile with little to no extended and/or diffuse low surface brightness emission in the wfc3 imaging data ; see figure [ lyamorphs.fig ] ) .
the correlation between @xmath2 equivalent width and optical semi - major axis radius is independent of the correlation with @xmath57 ; @xmath34 and @xmath57 are consistent with the null hypothesis of being uncorrelated ( @xmath92 rejection of the null hypothesis using the spearman rank correlation test ) , and confining our sample to a narrow range of values in @xmath57 results in a similarly strong trend of @xmath7 with @xmath34 after accounting for the reduced sample size .
lc @xmath58 & -1.7 + age & 0.5 + sfr & -2.4 + @xmath93 & -0.2 + e(b - v ) & * -2.4 * + @xmath38 & -0.2 + @xmath39 & 2.2 + @xmath40 & -0.7 + @xmath41 & 1.4 + @xmath43 & 1.6 + @xmath34 & * -2.1 * + @xmath73 & * -2.3 * + @xmath37 & -1.1 + @xmath36 & -1.9 + @xmath55 & 1.4 + [ lyastrength.tab ] combining our results with those previously published in the literature , the broad picture of @xmath94 @xmath2 emitters is therefore relatively clear .
compared to star forming galaxies at this epoch with negligible @xmath2 emission , galaxies with a measurable @xmath2 emission component tend to be physically smaller systems with lower stellar mass and less dust , lower star formation rates , and a slightly greater degree of nucleation in the rest - frame uv and optical light distribution .
most commonly , galaxies with visible @xmath2 emission also show some absorption features in their rest - uv spectra as well ( albeit frequently weak ; see discussion by shapley et al .
2003 ) . of the 71 galaxies in the parent sample with @xmath2 emission
, 7 do not have any identifiable absorption features ( even in a composite stack ) .
while the @xmath2 emission equivalent width for these sources range over a factor @xmath47 10 ( @xmath95 ) , all seven have type ii morphologies composed of two or more widely separated clumps ( see figure [ lyamorphs.fig ] ) , accounting for nearly half of the 17 total galaxies in the parent sample with @xmath2 emission and type ii morphology . by comparison ,
100% of the 28 ( 26 ) galaxies with type i ( iii ) morphology and ly@xmath16 emission have visible absorption features in their uv spectra , indicating that multi - component morphology is a strong factor in determining whether absorption line features will be observed in rest - uv spectra .
emission that do not have visible rest - uv absorption features is not possible to determine , since by definition such galaxies would not have spectroscopically confirmed redshifts . ]
morphological indices measuring the two - dimensional distribution of the light from a galaxy ( e.g. , @xmath41 , @xmath39 , @xmath43 ) therefore easily identify galaxies with exclusively @xmath2 emission in their rest - uv spectra and suggest that these multi - component galaxies are prime major - merger candidates .
if these galaxies are indeed mergers caught in the nearby - pair phase , the disturbed ism of the combined system may result in a reduced gas covering fraction , making it easier for resonantly - scattered @xmath2 emission to escape from the system .
however , type ii galaxies are no intrinsically brighter in @xmath2 emission than type i or iii galaxies ; instead , the only significant difference is that type ii galaxies are less likely to have absorption lines in their uv spectra .
a simpler explanation may be that when galaxies with multiple components separated by distances @xmath96 arcsec are observed with ground - based spectrographs the slit tends to be placed midway between the two components , thereby systematically reducing the s / n ratio of the continuum emission against which absorption features are measured .
as illustrated in figure [ lyamorphs.fig ] , in @xmath97% of cases the majority of the rest - uv continuum flux is expected to fall outside the slit based on the position angles used for the lris spectroscopy ; even in cases for which the slit was aligned approximately along the separation vector ( e.g. , q1623-bx547 ) the greater effective area over which the light is distributed may still increase the noise in the continuum such that the observations are systematically biased against the detection of absorption lines .
the deepest absorption lines in the rest - uv spectra are the , , , , and lines that arise in the metal - enriched neutral ism , which typically trace massive galactic - scale outflows driven by the intensely starbursting galaxy . as noted in many previous publications ( e.g. , shapley et al .
2003 ) there is generally an anti - correlation between @xmath2 emission and interstellar absorption line strength ; since the metal transitions are optically thick declining absorption line strength indicates a declining gas velocity dispersion and/or covering fraction , which also permits a greater @xmath2 escape fraction . in order to understand the physical mechanisms underlying variations in absorption line strength in greater detail we focus on the hahq sample for which accurate systemic redshifts have been obtained from rest - optical nebular emission .
the 35 galaxies in our hahq sample represent the subset of the 73 galaxies previously considered by @xcite for which _
hst_/wfc3 rest - optical imaging data has now been obtained . in their larger sample
, @xcite demonstrated that there is a correlation between baryonic galaxy mass @xmath98 and @xmath99 , the mean velocity offset of the interstellar absorption lines ( @xmath49 ) from the systemic redshift ( @xmath100 ) : @xmath101 lc @xmath102 & 0.6 + @xmath58 & 1.6 + @xmath103 & 1.5 + age & 1.9 + sfr & -0.9 + @xmath93 & -2.1 + @xmath57 & 0.7 + @xmath38 & -1.7 + @xmath39 & 1.6 + @xmath40 & -1.3 + @xmath41 & 1.0 + @xmath43 & 1.7 + @xmath34 & 2.6 + @xmath73 & 2.3 + @xmath37 & -1.6 + @xmath36 & -0.3 + @xmath55 & 0.3 + [ voutflow.tab ] using the spearman rank correlation test , @xcite found that the null hypothesis ( that @xmath99 and @xmath98 are uncorrelated ) could be ruled out at @xmath104 significance .
we repeat the calculations of @xcite using an estimate of @xmath105 , where @xmath58 is the stellar mass derived in paper i and @xmath102 the gas mass estimated using nirspec and/or osiris h@xmath16 fluxes in combination with the kennicutt - schmidt law ( @xcite ; see discussion by @xcite ) . in table
[ voutflow.tab ] we list the confidence at which the null hypothesis for zero correlation can be rejected for @xmath99 vs a variety of morphological and other physical properties . as expected we recover the sense of the previous correlation noted by @xcite between @xmath99 and total baryonic mass , although this correlation has a significance of only @xmath106 ; random resampling tests indicate that this lower significance is consistent with the lower total number of galaxies in our sample than in steidel et al .
( 2010 ) .
we find that @xmath99 correlates with the semi - major axis radius @xmath34 with a statistical significance of @xmath107 in the sense that the largest velocity differences ( i.e. , most negative @xmath99 ) are observed for the smallest ( type i ) galaxies , while the largest ( type iii ) galaxies have mean gas - phase velocities consistent with 0 @xmath6 .
combining measurements from the individual low - ionization absorption lines ( except @xmath108 , which is significantly weaker and noisier than the other transitions ) we find that @xmath109 @xmath6for the small radius half of the hahq sample ( @xmath110 kpc ) compared to @xmath111 @xmath6for the large radius half ( @xmath112 kpc ) .
uncertainties in the mean combining @xmath113 of individual galaxy spectra .
] we illustrate this trend visually in figure [ largersmallr.fig ] ; when sorted according to @xmath99 galaxies with the smallest ( blue boxes ) and largest ( red boxes ) sizes are clearly distinguished . adopting uncertainties of @xmath114 @xmath6 in @xmath99 and @xmath115
kpc in @xmath34 for each galaxy ( steidel et al . 2010 ; paper i ) , we use a least squares routine to fit a first - order polynomial to the data and obtain the average relation @xmath116 we stress , however , that @xmath99 does not characterize the outflow velocity of a particular galaxy , but rather represents the centroid of a single - component fit to interstellar absorption lines that are a complex blend of multiple components . in figure
[ abslines.fig ] we show the stacked , continuum - normalized spectra for the small- and large - radius subsamples ( solid blue / red lines respectively ) in zoomed - in regions around the major low - ionization gas absorption lines .
careful inspection of figure [ abslines.fig ] suggests that the low ionization absorption lines in both low-@xmath34 and high-@xmath34 stacks have similar optical depths in their blue wings , indicating that both have outflowing gas tracing a similar range of velocities .
rather , the major difference is the growth of an absorption component centered at @xmath117 @xmath6 in the large - radius sample that shifts the centroid of the absorption line profile closer to the systemic velocity ( see particularly @xmath118 ) and increases the depth of composite absorption line profiles even when precise systemic redshifts are unknown ( e.g. , figure [ fullspec.fig ] , top panel ) . this finding is similar to that described by steidel et al . ( 2010 ; see also weiner et al .
2009 ) who found that galaxies below a threshold @xmath119 lack the @xmath120 @xmath6component to the low - ionization interstellar absorption lines present in higher - mass galaxies . following the formalism adopted by steidel et al .
( 2010 ) , we define the additional optical depth @xmath121 that would need to be added to the low-@xmath34 composite galaxy spectrum in order to produce an absorption line profile identical to the high-@xmath34 composite spectrum : @xmath122 where @xmath123 and @xmath124 are the spectral intensity of the high-@xmath34 and low-@xmath34 halves of the hahq sample respectively . in figure [ tauplot.fig ]
we show @xmath121 for the three strongest unblended low - ionization absorption lines ( @xmath125 , @xmath126 , and @xmath127 ) along with the average composite for all three lines combined .
although there is some variation between the individual lines , the composite exhibits a significant broad absorption component from @xmath128 to @xmath129 @xmath6 in galaxies with the largest semi - major axis radii . as discussed at length by steidel et al .
( 2010 ) , this component could represent either infalling gas or stalled winds at large radii , although it most likely corresponds to cold gas located at small radii within the galaxy .
in addition to the low - ionization absorption lines , the high - ionization transitions @xmath130 , @xmath131 , and @xmath132 are also systematically less blueshifted for larger galaxies . combining measurements made from the two @xmath133 absorption lines , in this stack since it is complicated by the superposition of interstellar absorption on a stellar wind - induced p - cygni profile .
] we find that the mean blueshift of the high ionization lines ( his ) is @xmath134 @xmath6for the small - radius sample , but only @xmath135 @xmath6for the large - radius sample .
uncertainties in the mean combining @xmath136 of individual galaxy spectra . ] in contrast to the low - ionization lines however , this effect is not produced by an additional component of @xmath137 centered around @xmath120 @xmath6 , suggesting that the @xmath138 @xmath6 gas may be more strongly self - shielded than the majority of the outflowing gas .
instead , the large - radius subsample exhibits a weaker blue wing to the absorption profile ( figure [ tauplot.fig ] ) , suggesting that highly ionized ( @xmath139 k ) gas may have lower peak outflow velocities in galaxies with rest - optical continuum radii @xmath88 kpc .
in addition to exhibiting stronger @xmath2 emission and more systematically blueshifted low - ionization absorption features , galaxies with smaller rest - optical continuum radii also have stronger fine structure emission lines . as indicated by figure [ fullspec.fig ] ,
@xmath140 and @xmath141 emission features are more prominent in composite spectra of galaxies with effective radii @xmath1 kpc than in those with @xmath88 kpc ( most noticeably in the composite stack from the parent sample , and to a lesser extent in the hahq composite stack ) .
this result is consistent with the recent studies of berry et al .
( 2012 ) and jones et al .
( 2012 ) , who respectively noted that @xmath142 increases with @xmath7 and is stronger in star forming galaxies at @xmath67 than at lower redshifts ( which jones et al .
2012 ascribe to the increasing characteristic radius of low ionization gas with cosmic time ) .
as indicated in figure [ internalcorrel.fig ] there are significant underlying correlations between @xmath34 and the other morphological statistics . on average ,
galaxies with larger semi - major axis radii have larger @xmath58 , shallower radial density profiles ( lower @xmath37 ) , less concentrated light profiles ( i.e. , lower @xmath40 and @xmath38 ) , and tend to be more irregular ( whether measured using @xmath39 , @xmath41 , or @xmath43 ) .
such generic trends are apparent in figure [ largersmallr.fig ] ; the largest galaxies tend to be those that show most morphological irregularity to the light profiles , while the smallest tend to be single well - nucleated sources
. the physical mechanism underlying the change in @xmath99 is the buildup of neutral gas at rest with respect to the stars ( as discussed further in [ discussion.sec ] ) .
the internal correlation of semi - major axis radius with other parameters however gives rise to a variety of correlations of marginal significance ( see table [ voutflow.tab ] ) between @xmath99 and stellar / baryonic mass , the six morphological statistics @xmath38 , @xmath39 , @xmath40 , @xmath41 , @xmath43 , @xmath34 , and @xmath37 , and the sfr surface density @xmath93 .
given the known positive correlation between mass and effective radius for the galaxies in our sample ( nagy et al .
2011 ; paper i ) , it is perhaps unsurprising that similar qualitative trends in uv spectral features are observed when the galaxy sample is subdivided according to either mass or radius . likewise , although there is no significant correlation of @xmath99 with the total sed - derived star formation rate ( in part since all galaxies in our sample are selected on the basis of rest - uv color to be actively star forming with sfr @xmath143 yr@xmath10 ) , there is a correlation at the @xmath144 level with the sfr surface density in the sense that galaxies with higher @xmath93 tend to be missing the @xmath138 @xmath6absorption component ( see figure [ avelplot.fig ] , right - hand panel ) .
is more strongly correlated than galaxy size with @xmath99 for a sample of star forming galaxies at @xmath145 . ]
similarly , there is a @xmath80 correlation of @xmath99 with stellar population age in the sense that older populations have the strongest @xmath5 @xmath6absorption component ) , consistent with our finding that the @xmath110 and @xmath146 kpc samples have mean ages of 260 and 740 myr respectively .
@xcite suggested that the physical property governing both @xmath2 emission properties and interstellar absorption line kinematics was baryonic mass , which correlated with @xmath99 with a significance of @xmath104 in their sample of 73 galaxies .
since galaxies with higher baryonic masses tend to have older stellar populations and greater total quantities of interstellar gas , this additional gas may naturally lead to greater covering fractions and longer scattering paths , suppressing @xmath2 emission and making interstellar absorption lines deeper . in detail
though , mass may not be the most predictive indicator of the spectroscopic properties of a given galaxy . although the correlation that we find between @xmath99 and @xmath34 is of comparable statistical significance ( @xmath107 ) to the baryonic mass correlation described by steidel et al .
( 2010 ) , this may simply be because there are @xmath47 a factor of 2 fewer galaxies in the morphological hahq sample . replicating the hahq sample to a total sample size of 73 galaxies with random resampling using a monte carlo algorithm ( repeated 10,000 times ) , we estimate that the expected correlation significance between @xmath34 and @xmath99 in a sample of the size of steidel et al .
( 2010 ) is @xmath147 . indeed , while the galaxies with the strongest / weakest @xmath99 occupy an overlapping range of @xmath98 ( the galaxy with the _ weakest _ @xmath99 in our hahq sample has a baryonic mass smaller than the galaxy with the _ strongest _
@xmath99 ) , these galaxies have clearly distinct visual morphologies ( figure [ largersmallr.fig ] ) . likewise , although the galaxies with the highest @xmath93 are those with the strongest @xmath99(consistent with theoretical expectations that galaxies with a higher concentration of supernovae should be capable of more completely expelling the local ism ) , this correlation is driven entirely by the ( stronger ) relation between @xmath34 and @xmath99 and the definition @xmath148 ( although c.f .
kornei et al .
2012 for star forming galaxies at @xmath145 ) .
all of the correlations detailed herein may therefore simply be byproducts of a single underlying physical mechanism ; the growth of an extended disk .
star - forming @xmath0 galaxies with low to moderate stellar mass ( i.e. , @xmath149 ) tend to be compact , young , triaxial systems whose kinematics are dominated by velocity dispersion rather than rotation about a preferred kinematic axis .
high @xmath93 within these systems drives strong gaseous outflows into the surrounding igm and suppresses the buildup of a stable ism . in time , the increasing stellar mass formed by these galaxies is able to stabilize the formation of an extended ( albeit still thick ) gaseous disk ( as proposed by , e.g. , martig & bournaud 2010 ) in which the overall gas fraction decreases ( @xmath150 is anticorrelated with @xmath99 with a confidence of @xmath151 in the hahq sample ) and systemic rotation becomes an increasingly important means of dynamical support .
the lower @xmath93 in this extended disk is no longer able to expel gas as efficiently , a large component of which remains close to the galaxy and serves to both extinguish the ly@xmath16 emission ( or resonantly scatter it to larger radii ) and superimpose a zero - velocity absorption component atop the stellar continuum emission . while the growth of such extended structure is naturally associated with both evolving @xmath58 and @xmath93 , there is not a single well - defined value of either that demarcates the transition , and the mass at which extended structures may be supported can vary significantly based on the merger histories and other details of individual galaxies .
in contrast , rest - frame optical size @xmath34 characterizes where the transition to an extended disk has _ already occurred_. from a phenomenological standpoint , rest - optical morphology may therefore be the most robust and accessible means of identifying systems for which feedback is less efficient at evacuating gas from the galaxy .
this physical picture necessarily suggests that the zero - velocity gas feature should become more pronounced ( i.e. , @xmath99 is less strongly blueshifted ) as the contribution of rotational shear to the dynamical support of a galaxy increases .
such trends have already been noted in samples of galaxies observed with either long - slit ( see figure 2 of steidel et al .
2010 ) or ifu spectroscopy ( see figure 13 of law et al .
more recently , frster schreiber et al .
( 2011a ) presented _ hst_/nicmos morphologies of a sample of six galaxies observed previously with the sinfoni integral field spectrograph .
of these six galaxies , five are representative of the most disk - like known systems at @xmath13 based on their observed kinematics ( the other galaxy is identified as a major merger ) .
these five galaxies all lie at the upper edge of the distribution of effective radii for @xmath0 star forming galaxies ( see colored circles in figure 13 of paper i ) and have a mean @xmath152 @xmath6(excluding q1623-bx663 , whose rest - uv spectra are complicated by agn features ) .
these systems can therefore be identified from the overall @xmath8 galaxy sample by both their semi - major axis radii and spatially resolved velocity shear , but not by their stellar masses , which are widely distributed in the range @xmath153 .
steidel et al .
( 2010 ) suggested that the anticorrelation between @xmath99 and velocity shear may be due in part to projection effects .
if objects with the largest velocity gradients tend to be edge - on rotating disks , outflows from these galaxies might be expected to be collimated perpendicular to the disk so that there is simply little projection of the outflowing gas onto our line of sight ( see , e.g. , chen et al .
2010 , bordoloi et al .
2011 ) , while face - on systems would exhibit minimal rotation and maximum @xmath99 .
such a correlation between inclination and outflow velocity was recently found for a sample of star forming galaxies at @xmath145 by kornei et al .
however , in our @xmath13 sample we observe no correlation of @xmath99 with inclination as parametrized through the apparent minor / major axis ratio @xmath36 for the 35 galaxies in our hahq sample ( @xmath154 from the null hypothesis ; see table [ voutflow.tab ] ) . in part
, the lack of correlation may simple reflect the irregularity of @xmath13 star forming galaxies and the difficulty of determining inclination robustly for these intrinsically triaxial ( paper i ) systems .
however , no significant difference is observed in @xmath99 even for the two most disk - like galaxies in the sample .
morphology and ifu - derived kinematics suggest that q2343-bx389 is consistent with an edge - on disk ( frster schreiber et al .
2009 ) , while q2343-bx442 has both morphology and ifu kinematics consistent with a nearly face - on ( @xmath155 ) disk with spiral substructure ( see discussion by law et al .
despite their different orientations both galaxies have @xmath156 @xmath6 , suggesting either that @xmath99 is uncorrelated with inclination , or that the opening angle of the outflow is less than @xmath157 .
as illustrated by figure [ tauplot.fig ] much of the additional absorption in the largest galaxies has positive velocity @xmath158 @xmath6with respect to the systemic redshift , and formally @xmath159 @xmath6 for the three galaxies with largest semi - major axis radii in our hahq sample ( q2343-md59 , q2343-bx442 , q2343-bx389 ) .
it is therefore possible that absorption line spectroscopy of the largest galaxies is tracing _ infalling _ gas , whether from a recent merger , recycled wind accretion ( e.g. , oppenheimer et al .
2010 ) , or from cold cosmological flows ( e.g. , dekel et al .
2009 ; kere et al .
2009 ) . while such inflows are one plausible mechanism for solving the gas fueling probem , it is unclear why such accretion ( presumably with a small covering fraction ) would cause the near - complete absorption of ly@xmath16 ( see discussion by steidel et al .
similarly , simulations ( e.g. , dekel & birnboim 2006 ; kere et al .
2009 ) generally predict that such cold flows should be suppressed in the most massive galaxies at @xmath160 , in contrast to our observation that @xmath159 @xmath6 occurs more commonly for large and massive galaxies .
we have demonstrated that rest - optical morphology is correlated with the gas - phase properties of @xmath13 star forming galaxies as traced by their rest - uv spectra .
emission is most commonly observed in galaxies with small rest - optical half - light radii ( @xmath161 kpc ) , and the equivalent width @xmath7 of the emission is correlated with half light radius with @xmath162 confidence . although some large-@xmath34 galaxies with type iii morphology ( i.e. , extended and diffuse ) show @xmath2 in emission , this emission is relatively weak ; the twelve galaxies in our sample with @xmath163 all have small half - light radii ( @xmath161 kpc ) and type i or type ii morphologies ( i.e. , compact clumps ) .
this correlation is independent of the better - known relations between @xmath7 , dust content , and star formation rate .
additionally , all galaxies in our spectroscopic sample for which @xmath2 is observed in emission and no absorption lines are present in the rest - uv spectrum have morphological type ii ( i.e. , consist of two or more clumps of emission separated from each other by @xmath47 1 arcsec ) , suggesting that slit losses of continuum light from these galaxies may be significant .
combining our results with those previously published in the literature ( e.g. , @xcite , @xcite , @xcite ) , the broad picture of @xmath94 star forming galaxies is relatively clear .
galaxies with @xmath2 emission tend to be physically smaller than galaxies with negligible @xmath2 emission , with lower stellar mass , less dust , lower star formation rates , and a slightly greater degree of nucleation in the rest - frame uv and optical light distribution . using a subsample of 35 galaxies with both high - quality rest - uv spectra and h@xmath16-derived systemic redshifts
we have found that galaxy morphology is also correlated with the median blueshift of interstellar absorption line features @xmath99 with @xmath107 significance .
although similar blue wings in the low - ionization absorption line profiles indicate that galaxies of all morphological types drive comparably strong outflows at speeds up to @xmath164 @xmath6 into their surrounding environments , increasing optical half - light radius is accompanied by an increase in optical depth around the systemic redshift of the galaxy .
this finding is similar in statistical significance to trends with increasing baryonic mass described recently by @xcite , but may be of physically greater importance considering the smaller sample size . while galaxies with @xmath99 differing by almost 1000 @xmath6 can have baryonic masses identical to within observational uncertainty , such galaxies have clearly distinct morphologies spanning the range from compact clumps to extended disks .
the increased optical depth arises in gas nearly at rest with respect to the systemic redshift , in close physical proximity to the galaxy , and apparently corresponding to an extended disk of optical continuum light .
this stronger @xmath5 @xmath6 component is responsible both for the shifting absorption line profile and the attenuation of @xmath7 ( via a longer resonant scattering path ) in galaxies with larger rest - optical half light radii .
the @xmath5 @xmath6 absorption component is not present in the higher - ionization features @xmath130 and @xmath165 however , suggesting that this gas may be more strongly self - shielded than the majority of the outflowing gas .
in contrast to recent results at lower redshifts ( e.g. , @xcite ) , there is no obvious relation between @xmath99 and galaxy inclination as parameterized by the axis ratio @xmath36 for the 35 galaxies in our hahq sample . in part
this may be due to the difficulty determining robust inclinations for inherently triaxial star - forming systems . even for two
particularly well - resolved galaxies with visual morphology and ifu kinematics consistent with face - on and edge - on disks however , there are no obvious differences in the outflowing gas kinematics ( although the quality of individual spectra prohibits a detailed analysis of just these two galaxies ) .
there is therefore no evidence to support a classical bipolar outflow model for star forming galaxies at @xmath0 .
this may be because such galaxies are in the process of transforming from clumpy , irregular systems to more regular ( albeit thick ) disks , and outflowing gas at high velocities and large distances from the galaxy may predate establishment of the disk and its corresponding polar axis .
in general , our observations are consistent with inside - out growth of star forming galaxies in the young universe : ` typical ' @xmath13 star forming galaxies appear to be gas - rich , compact , triaxial systems systems that are dominated by velocity dispersion between individual star forming regions rather than systemic rotation , and whose high @xmath93 drives strong outflows into the surrounding igm .
as these galaxies mature they gain stellar mass , stabilizing the formation of extended ( albeit still thick ) gaseous disks in which rotational support is observed to play an increasing role .
as the star formation migrates from central regions into these extended disks the @xmath93 drops and is no longer capable of expelling gas from the galaxy as efficiently .
we caution , however , that it is challenging to generalize about the relation between rest - optical morphology and gas phase kinematics for the entire @xmath13 star forming sample on the basis of the relatively small number of galaxies in our sample .
a major limiting factor is systemic nebular redshifts , without which we can not quantify the outflow velocities .
although @xmath166 @xmath0 star forming galaxies with rest - uv spectra in the kbss have had systemic redshifts derived from nebular emission line spectroscopy ( e.g. , erb et al .
2006b ; law et al .
2009 ) , only a fraction of these galaxies fell within the footprint of our _ hst_/wfc3 imaging survey . in the future it will be possible to explore these trends in much greater detail with the aid of multi - object near - ir and optical spectrographs , using which it will be possible to obtain h@xmath16 redshifts and outflow kinematics efficiently for large and diverse samples of galaxies in fields that have already been imaged with _ hst_.
drl , ccs , and srn have been supported by grant go-11694 from the space telescope science institute , which is operated by the association of universities for research in astronomy , inc .
, for nasa , under contract nas 5 - 26555 .
ccs has been supported by the us national science foundation through grants ast-0606912 and ast-0908805 aes acknowledges support from the david and lucile packard foundation .
finally , we extend thanks to those of hawaiian ancestry on whose sacred mountain we are privileged to be guests . | we analyze rest - frame optical morphologies and gas - phase kinematics as traced by rest - frame far - uv and optical spectra for a sample of 204 star forming galaxies in the redshift range @xmath0 drawn from the keck baryonic structure survey ( kbss ) . we find that spectroscopic properties and gas - phase kinematics are closely linked to morphology : compact galaxies with semi - major axis radii @xmath1 kpc are substantially more likely than their larger counterparts to exhibit @xmath2 in emission .
although @xmath2 emission strength varies widely within galaxies of a given morphological type , all but one of 19 galaxies with @xmath2 equivalent width @xmath3 have compact and/or multiple - component morphologies with @xmath4 kpc .
the velocity structure of absorption lines in the galactic continuum spectra also varies as a function of morphology .
galaxies of all morphological types drive similarly strong outflows ( as traced by the blue wing of interstellar absorption line features ) , but the outflows of larger galaxies are less highly ionized and exhibit larger optical depth at the systemic redshift that may correspond to a decreasing efficiency of feedback in evacuating gas from the galaxy .
this @xmath5 @xmath6 gas is responsible both for shifting the mean absorption line redshift and attenuating @xmath7 ( via a longer resonant scattering path ) in galaxies with larger rest - optical half light radii .
in contrast to galaxies at lower redshifts , there is no evidence for a correlation between outflow velocity and inclination , suggesting that outflows from these puffy and irregular systems may be poorly collimated .
our observations are broadly consistent with theoretical models of inside - out growth of galaxies in the young universe , in which typical @xmath8 star forming galaxies are predominantly unstable , dispersion - dominated , systems fueled by rapid gas accretion that later form extended rotationally - supported disks when stabilized by a sufficiently massive stellar component . |
the continuous - time poisson channel models optical communication using direct detection .
the input to the channel @xmath1 is nonnegative @xmath2 and conditional on the input , the output @xmath3 is a conditional poisson process ( also known as a doubly - stochastic poisson process ) of intensity @xmath4 , where @xmath5 is a nonnegative constant called _
dark current_. thus , conditional on the input , the output @xmath3 is a nonhomogeneous poisson process and thus of independent increments with @xmath6 } = e^{-\lambda } \frac{\lambda^{\nu}}{\nu ! } , \quad \nu \in { \mathbb z},\ ] ] where @xmath7 to send a bit @xmath8 taking on the values @xmath9 and @xmath10 equiprobably over this channel _ without feedback _ we use two input waveforms @xmath11 and @xmath12 , and we send @xmath11 if @xmath13 and @xmath12 if @xmath14 .
we refer to @xmath15}\ ] ] as the transmitted energy in the time interval @xmath16 $ ] , although this is somewhat imprecise : this quantity is the expected number of transmitted photons in the interval , and one should technically multiply it by the energy in each photon ( which depends on the light frequency ) to obtain the transmitted energy in the interval .
we sometimes impose a _ peak - power constraint _ on the input , in which case we require that , with probability one , @xmath17 we then refer to @xmath18 as the maximal allowed power ( although , technically speaking , this needs to be normalized by the energy of each photon to have the sense of power . ) in the presence of feedback , the channel description is a bit more technical @xcite .
we require that conditional on @xmath13 , the channel output @xmath19 admit the @xmath20 intensity @xmath21 ( * ? ? ?
* chapter ii , section 3 , definition d7 ) .
that is , conditional on @xmath13 , @xmath19 is a point process adapted to some history @xmath22 ; @xmath23 is a nonnegative @xmath22-progressive process such that for all @xmath24 @xmath25 and for all nonnegative @xmath22-predictable processes
@xmath26 @xmath27 } = { \textnormal{\textsf{e}}_{}\!\left [ \int_{0}^{\infty } c(t ) \bigl(x_{0}(t ) + \lambda_{0}\bigr ) { \,\textnormal{d}}{t}\right]}.\ ] ] the conditional expected energy transmitted when @xmath13 over the time interval @xmath28 $ ] is @xmath29}.\ ] ] similarly , when @xmath14 the transmitted energy is @xmath30}.\ ] ] the average transmitted energy is thus @xmath31 a decoder is a mapping from the @xmath32-algebra generated by @xmath33 to the set @xmath34 .
we say that a bit can be transmitted reliably over our channel with average transmitted energy @xmath35 , if for any @xmath36 we can find some transmission interval @xmath37 and a coding / decoding rule of expected transmission energy @xmath35 and probability of error smaller than @xmath38 .
we denote by @xmath39 the least energy required to transmit a bit reliably over our channel .
the minimum energy required to send a single bit over the poisson channel with dark current @xmath5 , feedback , and no peak - power constraint is @xmath40 irrespective of the dark current .
if the dark current is zero , then this is achievable even under a peak - power constraint whenever @xmath41 .
the achievability when @xmath42 is straightforward . to send @xmath13
we transmit the all - zero input ; to send @xmath14 we transmit @xmath18 until , through the feedback link , we learn that a count was registered ; thereafter we send zero .
the decoder guesses `` @xmath13 '' if no counts were registered in the interval @xmath28 $ ] , and guesses `` @xmath14 '' otherwise .
since @xmath42 , the probability of error given @xmath13 is zero . also , the transmitted energy when @xmath13 is zero , so @xmath43 .
conditional on @xmath14 the time of the first count is exponential with mean @xmath44 .
consequently , @xmath45 .
conditional on @xmath14 , the probability of error is the probability that no counts are registered in the interval @xmath28 $ ] .
the probability of this event is the probability that the first count occurs after time @xmath37 , i.e. , the probability that a mean-@xmath44 exponential exceeds @xmath37 .
it thus tends to zero as @xmath46 . if @xmath47 and there is no peak - power constraint , we choose @xmath48 and @xmath49 and use the above scheme with @xmath50 .
we make sure that @xmath51 is small enough for the probability of a spurious count in the interval @xmath52 $ ] to be very small ( the probability of a spurious count in this interval is @xmath53 ) , and we choose @xmath18 large enough so that the probability that a mean-@xmath44 exponential exceeds @xmath51 is also very small .
to prove that @xmath39 can not be smaller than @xmath54 , it suffices to consider the case where @xmath42 .
we thus assume @xmath42 . in this case
there is no loss in optimality in assuming that to send @xmath13 we transmit the all - zero input .
indeed , given any general scheme consider the guess the decoder produces when faced with no counts .
call that false .
let true be its complement .
consider now a scheme with the same encoding rule for true , with the same decoding rule , but where we send the all - zero input to convey false .
the new scheme uses less ( or same ) energy ; has the same @xmath55 ; and has @xmath56 .
since the name we give to the hypotheses is immaterial , we can assume that false corresponds to @xmath13 .
next we argue that there is no loss in optimality in restricting ourselves to a detector that bases its decision on the presence of counts in the interval @xmath28 $ ] . since this enlarges the set of outcomes yielding the guess `` d=1 '' , this can not increase @xmath57 . to send @xmath13 we send the all - zero waveform , which results in no counts ( there is no dark current ) , so this does not change @xmath58 .
finally , we argue that there is no loss in optimality in stopping transmission once a count has been registered . indeed , this reduces the transmitted energy and does not change the performance of the above detector .
we next analyze the probability of error of such schemes .
conditional on @xmath13 , the probability of error is zero , because there is no dark current so sending zero input guarantees zero counts . as to the conditional probability of error given @xmath14 ,
let @xmath59 denote the random time at which the first count is registered .
substituting the stochastic process @xmath60 in yields @xmath61 } \\ & = { \textnormal{\textsf{e}}_{}\!\bigl[y(t_{1 } \wedge { { \textnormal{\usefont{u}{eur}{m}{n}\selectfont t } } } ) \kern-0.1em \bigm| \kern-0.1em d=1\bigr ] } \\ & = { \textnormal{\textsf{e}}_{}\!\biggl [ \int_{0}^{\infty } c(t ) { \,\textnormal{d}}y(t ) \kern-0.1em \biggm| \kern-0.1em d=1\biggr ] } \\ & = { \textnormal{\textsf{e}}_{}\!\left [ \int_{0}^{\infty } c(t ) x_{1}(t ) { \,\textnormal{d}}{t}\right ] } \\ & = { \textnormal{\textsf{e}}_{}\!\left [ \int_{0}^{t_{1 } \wedge { { \textnormal{\usefont{u}{eur}{m}{n}\selectfont t } } } } x_{1}(t ) { \,\textnormal{d}}{t}\right ] } \\ & = { \mathcal e}_{1}.\end{aligned}\ ] ] for the probability of error to tend to zero , the expected energy transmitted to convey @xmath14 must thus approach @xmath10 .
more generally , to send @xmath0 bits requires @xmath62 of the energy of a photon . when @xmath63 we recover the required energy to send one bit . to prove the converse that one can not accomplish the task with less energy requires a simple genie - aided argument .
once again we can assume no dark current , and we can show that there is no loss in optimality in conveying the zero message using the all - zero input and by limiting ourselves to detectors that guess that the transmitted message was the zero message if , and only if , no counts were registered .
we then consider a genie that , if a count is registered , tells the detector which message was sent . with the aid of the genie the detectors errs if , and only if , the transmitted message was not the zero message and no counts were registered .
thus , by our previous analysis , the required energy of each of the nonzero messages must be one .
averaging over the equally - likely messages demonstrates that @xmath64 of the energy of a photon is necessary for reliable transmission .
the direct part in the absence of dark current is based on a simple scheme where the transmission interval @xmath65 is divided into @xmath66 intervals .
we associate with the nonzero message @xmath67 the interval @xmath68 to send the nonzero message @xmath69 we transmit with a very high power starting at time @xmath70 until a count is registered or until the end of the interval at time @xmath71 . to send the zero message we send the all - zero signal .
the detector operates as follows . if no counts are registered , it guesses that the zeroth message was sent .
if a count is registered , it declares that the transmitted message was the one that corresponds to the interval in which the count was registered . in the presence of dark
current we use the same scheme except that we choose as our transmission time a very short interval @xmath52 $ ] in which the probability of a spurious count is negligible .
in the absence of a peak power constraint , the capacity of the poisson channel ( with or without feedback ) is infinite @xcite , @xcite .
thus , in sending a very large number of bits , reliable communication can be had with an arbitrarily small expected energy per bit .
the situation changes dramatically when sending a _ single _ bit . even in the presence of feedback ,
the required energy is finite ; it is , in fact , @xmath54 .
this should be contrasted with the infinite - bandwidth gaussian channel , where a single bit can be sent reliably with the same amount of energy that would be required per bit if one were sending a large number of bits @xcite , @xcite . | we consider the transmission of a single bit over the continuous - time poisson channel with noiseless feedback .
we show that to send the bit reliably requires , on the average , half the energy of a photon . in the absence of peak - power constraints this holds irrespective of the intensity of the dark current .
we also solve for the energy required to send @xmath0 bits . |
entanglement is an essential ingredient in quantum information and the central feature of quantum mechanics which distinguishes a quantum system from its classical counterpart . in recent years
, it has been regarded as an important physical resource , and widely applied to a lot of quantum information processing(qip ) : quantum computation [ 1 ] , quantum cryptography [ 2 ] , quantum teleportation [ 3 ] , quantum dense coding [ 4 ] and so on .
entanglement arises only if some subsystems ever interacted with the others among the whole multipartite system in physics , or only if the multipartite quantum state is not separable or factorable in mathematics .
the latter provides a direct way to tell whether or not a given quantum state is entangled . as to the separability of bipartite quantum states , partial entropy introduced by bennett _
_ al_. [ 5 ] provides a good criterion of separability for pure states .
later , wootters presents the remarkable concurrence for bipartite systems of qubits [ 6,7 ] . based on the motivation of generalizing the definition of concurrence to higher dimensional systems , many attempts have been made [ 8,9,10,11 ] , which provide good separability criteria for bipartite qubit systems under corresponding conditions , whilst ref.[8 ] also presents an alternative method to minimize the convex hull for mixed states . as to multipartite quantum systems ,
several separability criteria have been proposed [ 12,13,14,15,16,17 ] .
the most notable one is 3-tangle for three qubits
recently , the result has been generalized to the higher dimensional systems [ 18 ] . despite the enormous effort ,
the separability of quantum states especially in higher dimensional systems is still an open problem . in this paper
we construct the full separability criteria for arbitrary tripartite qubit system by a novel method , i.e. a tripartite pure state can be defined by a three - order tensor .
the definition provides an intuitionistic mathematical formulation for the full separability of pure states .
analogous to ref.[8 ] , we extend the definition to mixed states .
more importantly , our approach is easily generalized to higher dimensional systems . as applications ,
we discuss separability of two bound entangled states introduced in [ 19,20 ] , respectively .
we start with the separability definition of tripartite qubit pure state @xmath0 @xmath1 is fully separable iff @xmath2consider a general tripartite pure state written by@xmath3where @xmath4 , @xmath5 , @xmath6 , @xmath7 , the coefficients @xmath8s can be arranged as a three - order tensor ( tensor cube ) [ 21 ] as shown in figure 1 .
note that the subscripts of @xmath8 correspond to the basis @xmath9 .
every surface can be regarded as the tensor product of a single qubit and an unnormalized bipartite state .
hence , if the two vectors ( edges ) of a surface are linear relevant ( including one of the vectors is @xmath10 vector ) , then the bipartite state mentioned above can be factorized .
the conclusion for diagonal plane is analogous .
considering all the planes , one can easily find that the tripartite state is fully separable iff all the vectors which are parallel mutually shown in the cube , are linear relevant , according to the fundamental linear algebra .
i.e. the rank of every matrix composed of four coefficients on the corresponding surface and diagonal plane is @xmath11 .
equivalently , we can obtain the following lemma .
* * lemma1.-**a tripartite pure state @xmath12 with the form of eq.(2 ) in @xmath13 dimensional hilbert space is fully separable , iff the following six equations hold : @xmath14@xmath15@xmath16@xmath17@xmath18@xmath19where @xmath20 and @xmath21 . * proof . *
( sufficient condition ) if eq.(3)-eq.(5 ) hold , then the rank of every matrix that the cubic surface corresponds to is _
one_. if eq.(6)-eq.(8 ) hold , then the rank of every matrix that the cubic diagonal plane corresponds to is _
one_. hence , that eq.(3)-eq.(8 ) hold simultaneously shows that the tripartite pure state can be fully factorized , i.e. it is fully separable .
( necessary condition ) if a given tripartite state is fully separable , one can easily obtain that the rank of every corresponding matrix is @xmath11 .
namely , eq.(3)-eq.(8 ) hold .
consider that @xmath22 can be denoted by a vector in @xmath13 dimensional hilbert space,@xmath23with the superscript @xmath24 denoting transpose , we can write the above equations ( 3 - 8 ) in matrix notation by@xmath25where the star denotes complex conjugation , and @xmath26 , @xmath27 , @xmath28 , @xmath29 , @xmath30 , @xmath31 , @xmath32 , @xmath33 , @xmath34 , with @xmath35 , @xmath36 , @xmath37 and @xmath38 .
( alternatively , @xmath39 can be replaced by @xmath40 for tripartite pure states of qubit with @xmath41 , @xmath42 and @xmath43 , where @xmath44 , by which @xmath45 comlex optimal parameters can be reduced for the case of mixed states ) .
define a new vector @xmath46 by @xmath47 with @xmath48 @xmath49 , then the length of the vector can be given by @xmath50 .
therefore , the full separability criterion for a tripartite state can be expressed by a more rigorous form as follows . * * theorem1.-**a tripartite pure state @xmath51 is fully separable iff @xmath52
. * proof . * that * * * * @xmath53 is equivalent to that @xmath54 holds for any @xmath55 . according to lemma 1 , one can obtain that @xmath53 is the sufficient and necessary condition .
a tripartite mixed state @xmath56 is fully separable iff there exists a decomposition @xmath57 , @xmath58 such that @xmath59 is fully separable for every @xmath60 or equivalently iff the infimum of the average @xmath61 vanishes , namely,@xmath62among all possible decompositions .
therefore , for any a given decomposition @xmath63according to the minkowski inequality@xmath64one can get@xmath65consider the matrix notation [ 8 ] of equation ( 10 ) as @xmath66 , where @xmath67 is a diagonal matrix with @xmath68 , the columns of the matrix @xmath69 correspond to the vectors @xmath59 , and the eigenvalue decomposition , @xmath70 , where @xmath71 is a diagonal matrix whose diagonal elements are the eigenvalues of @xmath56 , and @xmath72 is a unitary matrix whose columns are the eigenvectors of @xmath56 , associated with the relation @xmath73 , where @xmath74 is a right - unitary matrix , inequality ( 12 ) can be rewritten as @xmath75 in terms of the cauchy - schwarz inequality @xmath76the inequality @xmath77is implied for any @xmath78 with @xmath79 and @xmath80 where @xmath81 .
the infimum of equation ( 15 ) is given by @xmath82 analogous to ref.[8 ] , with @xmath83s are the singular values , in decreasing order , of the matrix @xmath84 .
@xmath85 is as well expressed by @xmath86one can easily see that @xmath87 provides a necessary and even sufficient condition of full separability for tripartite mixed qubit systems , hence an effective separability criterion
. however , it is so unfortunate that @xmath85 can not serve as a good entanglement measure , but only an effective criterion to detect whether a state is fully separable , because @xmath46 for pure states is not invariant under local unitary transformations .
at first , consider the complementary states to shifts upb [ 19 ] .
shifts upb is the set of the following four product states@xmath88with @xmath89 .
the corresponding bound entangled ( complementary ) state is given by @xmath90with @xmath91 corresponding to the shifts upb . in ref.[19 ] , it is stated that this complementary state has the curious property that not only is it two - way ppt , it is also two - way separable . the numerical result based on our criterion
can show a _
non - zero _ ( @xmath92 ) entanglement for @xmath93 , which is consistent to [ 19 ] .
let s consider the second example , the dr - cirac -tarrach states [ 20 ] @xmath94we can also show the @xmath95 @xmath96 ( @xmath97 ) for @xmath98the conclusion is also implied in [ 20 ] .
the above numerical tests are operated as follows . in order to show the nonzero @xmath99 with @xmath100 or @xmath101 , we choose @xmath102 random vectors @xmath103 generated by _
matlab 6.5 _ for a given @xmath104 , then substitute these vectors into @xmath105 and obtain @xmath106 matrices .
we can get @xmath102 ( @xmath107)s by singular value decomposition for the matrices .
the maximal @xmath108 among the matrices is assigned to @xmath99 . due to the whole process
, it is obvious that our numerical approach is more effective to test the _
it can only provide a reference for the _ zero _ @xmath99 .
if a standard numerical process is needed , we suggest that the approach introduced in ref.[17 ] be preferred .
as a summary , we have shown effective criteria for tripartite qubit systems by the pioneering application of the approach to define a tripartite pure state as a three - order tensor . however , although our criteria can be reduced to wootters concurrence [ 7 ] for bipartite systems , as mentioned above , the criteria can not serve as a good entanglement measure . therefore , it is not necessary to find out the concrete value of @xmath85 , but whether @xmath110 are greater than @xmath10 , as can be found in our examples .
based on the tensor treatment for a tripartite pure state , if a more suitable @xmath111 that can serve as a good entanglement measure can be found , it will be interesting .
it deserves our attention that our approach can be easily extended to test the full separability of multipartite systems in arbitrary dimension , which will be given out in the forthcoming works .
we would like to thank x. x. yi for extensive and valuable advice .
we are grateful to the referees for their useful suggestion and comments .
this work was supported by ministry of science and technology , china , under grant no.2100cca00700 .
different from the previous definition of tensors , all the quantities with indices , such as @xmath112 , are called three - order tensors here . therefore , the set of all one - order tensor is the set of vectors , and the set of all two - order tensors is the set of matrices .
three - order tensors @xmath112 are matrices correpsonding to the planes in the cube ( vectors corresponding to the edges ) when any one ( two ) of their three indices is ( are ) fixed . | in this paper , we present a method to construct full separability criteria for tripartite systems of qubits .
the spirit of our approach is that a tripartite pure state can be regarded as a three - order tensor that provides an intuitionistic mathematical formulation for the full separability of pure states .
we extend the definition to mixed states and given out the corresponding full separability criterion . as applications ,
we discuss the separability of several bound entangled states , which shows that our criterion is feasible . |
the classification of riemannian manifolds with positive , and more generally , non - negative sectional curvature , is a long - standing open problem in riemannian geometry . as a step towards general classification results
one may consider manifolds whose isometry group is large .
this has been a fruitful avenue of research ( see , for example , the surveys @xcite ) .
it is well - known that the isometry group of a compact riemannian manifold is a compact lie group . in the context of this paper , the measure for the size " of an isometry group is its rank . in particular , we are interested in manifolds with non - negative curvature that have almost maximal symmetry rank , where the _ symmetry rank _ of a riemannian manifold @xmath5 is defined to be the rank of the isometry group of @xmath5 .
grove and searle @xcite showed that the symmetry rank of a closed , positively curved , riemannian @xmath6-manifold is bounded above by @xmath7 $ ] and classified closed , positively curved riemannian manifolds with maximal symmetry rank up to diffeomorphism . for a closed
, positively curved riemannian @xmath6-manifold of almost maximal symmetry rank , that is , one whose isometry group has rank @xmath8 $ ] , rong @xcite found topological restrictions for all dimensions ( distinguishing the cases for even and odd ) and showed that a closed , simply - connected , positively curved riemannian @xmath0-manifold with almost maximal symmetry rank , that is , with an effective isometric @xmath1 action , must be homeomorphic to the @xmath0-sphere ( in fact , it will be diffeomorphic , as a consequence of the generalized poincar conjecture ) .
later , wilking @xcite improved these results significantly for closed , positively curved , simply - connected @xmath6-manifolds of dimension @xmath9 , considering actions of rank approximately @xmath10 .
the maximal symmetry rank for closed , simply - connected @xmath6-manifolds with non - negative curvature and dimension @xmath11 is @xmath12 $ ] ( see @xcite ) .
kleiner @xcite and searle and yang @xcite independently classified up to homeomorphism closed , simply - connected @xmath13-manifolds of non - negative curvature with an effective isometric circle action , corresponding to the almost maximal symmetry rank case in dimension @xmath13 . in @xcite ,
the authors classified up to diffeomorphism closed , simply - connected , non - negatively curved riemannian manifolds of dimensions @xmath14 , @xmath13 , @xmath0 and @xmath15 with maximal symmetry rank . in this paper
we address the case of almost maximal symmetry rank for closed , simply - connected , non - negatively curved riemannian manifolds in dimensions @xmath14 , @xmath13 and @xmath0 .
our main result is the following : [ t : a ] let @xmath16 be a closed , simply - connected , non - negatively curved @xmath0-manifold . if @xmath17 acts isometrically and ( almost ) effectively on @xmath16 , then @xmath16 is diffeomorphic to one of @xmath18 , @xmath19 , @xmath3 ( the non - trivial @xmath20-bundle over @xmath21 ) or the wu manifold @xmath4 .
we remark that the @xmath0-manifolds listed in theorem a are all the known examples of closed , simply - connected @xmath0-manifolds with non - negative curvature and these manifolds are all the closed , simply - connected @xmath0-dimensional homogeneous spaces or biquotients of lie groups ( cf .
we also point out that the @xmath0-manifolds listed in theorem a coincide with the closed , simply - connected @xmath0-manifolds that are elliptic ( cf .
@xcite ) . further , each one of these @xmath0- manifolds @xmath5 is rationally elliptic , that is , @xmath22 , thus satisfying the ellipticity conjecture , which states that all closed , simply - connected manifolds of ( almost ) non - negative curvature are rationally elliptic ( cf .
it is also worth noting that these are exactly the @xmath0-dimensional topological manifolds @xmath5 for which @xmath23 , that is , @xmath5 can be covered by two open subsets @xmath24 , @xmath25 such that the inclusions @xmath26 factor homotopically through maps @xmath27 ( cf .
@xcite ) .
this paper is divided into seven sections .
the first two sections comprise the introduction and basic tools we will use throughout . in section 3 , using classification results for smooth circle actions on @xmath14- and @xmath13-manifolds , in combination with restrictions imposed by non - negative curvature , we classify closed , orientable manifolds with non - negative curvature and almost maximal symmetry rank in dimension @xmath14 and recall the classification of closed , simply connected manifolds with non - negative curvature and almost maximal symmetry rank in dimension @xmath13 . in section 4 we consider the problem of almost maximal symmetry rank in dimension @xmath0 from a purely topological perspective and in section 5 we find restrictions imposed by non - negative curvature . in section 6
we classify closed , simply - connected , non - negatively curved @xmath0-manifolds of almost maximal symmetry rank by applying the results of the previous three sections .
finally , in section 7 we give examples of actions of almost maximal symmetry rank on some of the manifolds listed in theorem a. the authors thank both karsten grove and burkhard wilking for pointing out an omission in an earlier version and for helpful conversations and suggestions for improvements .
both authors would also like to thank the american institute of mathematics ( aim ) for its support during a workshop where the work on this paper was initiated .
f. g. g. also thanks the mathematics department of the university of maryland , for its financial support , and the department of mathematics of the university of notre dame , as well as the imate - unam , cuernavaca , for their hospitality while part of this work was carried out . c. s. was supported in part by conacyt project # sep-106923 , conacyt project # sep-82471 .
in this section we gather several definitions and results that we will use in subsequent sections .
let @xmath28 be a lie group acting ( on the left ) on a smooth manifold @xmath5 .
we denote by @xmath29 the _ isotropy group _ at @xmath30 and by @xmath31 the _ orbit _ of @xmath32 .
the _ ineffective kernel _ of the action is the subgroup @xmath33 .
we say that @xmath28 acts _ effectively _ on @xmath5 if @xmath34 is trivial .
the action is called _ almost effective _ if @xmath34 is finite .
the action is _ free _ if every isotropy group is trivial and _ almost free _ if every isotropy group is finite .
we will denote the _ fixed point set _ @xmath35 of this action by @xmath36 and define its dimension as @xmath37 .
when convenient , we will denote the orbit space @xmath38 by @xmath39 .
we will denote by @xmath40 the image of a point @xmath41 under the orbit projection map @xmath42 . given a subset @xmath43
, we will denote its image in @xmath39 under the orbit projection map by @xmath44 and when convenient , we shall also denote the orbit space @xmath38 by @xmath45 .
one measurement for the size of a transformation group @xmath46 is the dimension of its orbit space @xmath38 , also called the _
cohomogeneity _ of the action .
this dimension is clearly constrained by the dimension of the fixed point set @xmath47 of @xmath28 in @xmath5 .
in fact , @xmath48 for any non - trivial action . in light of this , the _ fixed - point cohomogeneity _ of an action , denoted by @xmath49 , is defined by @xmath50 a manifold with fixed - point cohomogeneity @xmath51 is also called a _ fixed point homogeneous manifold_. we will use the latter term throughout this article .
we observe that the fixed point set of a fixed point homogeneous action has codimension 1 in the orbit space . throughout the rest of the paper we will assume all manifolds to be smooth
. we will only consider smooth ( almost ) effective actions and all homology and cohomology groups will have coefficients in @xmath52 , unless otherwise stated
. recall that a finite dimensional length space @xmath53 is an _ alexandrov space _ if it has curvature bounded from below ( cf .
when @xmath5 is a complete , connected riemannian manifold and @xmath28 is a compact lie group acting on @xmath5 by isometries , the orbit space @xmath54 is equipped with the orbital distance metric induced from @xmath5 , that is , the distance between @xmath40 and @xmath55 in @xmath39 is the distance between the orbits @xmath56 and @xmath57 as subsets of @xmath5 .
if , in addition , @xmath5 has sectional curvature bounded below , that is , @xmath58 , then the orbit space @xmath39 is an alexandrov space with @xmath59 .
the _ space of directions _ of a general alexandrov space at a point @xmath32 is by definition the completion of the space of geodesic directions at @xmath32 . in the case of orbit spaces @xmath54 , the space of directions @xmath60 at a point @xmath61 consists of geodesic directions and is isometric to @xmath62 where @xmath63 is the unit normal sphere to the orbit @xmath56 at @xmath41 .
we now state kleiner s isotropy lemma ( cf .
@xcite ) , which we will use to obtain information on the distribution of the isotropy groups along minimal geodesics joining two orbits and , in consequence , along minimal geodesics joining two points in the orbit space @xmath54 .
[ l : kleiner ] let @xmath64\rightarrow m$ ] be a minimal geodesic between the orbits @xmath65 and @xmath66 .
then , for any @xmath67 , @xmath68 is a subgroup of @xmath69 and of @xmath70 .
recall that the @xmath71-_extent _
@xmath72 , @xmath73 , of a compact metric space @xmath74 is the maximum average distance between @xmath71 points in @xmath39 : @xmath75 the extent lemma ( cf .
@xcite ) stated below provides an upper bound on the total number of isolated singular points in @xmath54 .
[ l : xt ] let @xmath76 be @xmath77 distinct points in @xmath54 . if @xmath78 , then @xmath79 we remark that in the case of strictly positive curvature , the inequality is also strict .
we will also use the following analogue for orbit spaces of the cheeger - gromoll soul theorem to obtain information on the geometry of the orbit space @xmath54 .
[ t : soulthm ] let @xmath54 .
if @xmath78 and @xmath80 , then there exists a totally convex compact subset @xmath81 with @xmath82 , which is a strong deformation retract of @xmath39 . if @xmath83 , then @xmath84 is a point , and @xmath85 is homeomorphic to @xmath86 .
when @xmath5 is a non - negatively curved , fixed point homogeneous riemannian @xmath28-manifold , the orbit space @xmath39 is a non - negatively curved alexandrov space and @xmath85 contains a component @xmath87 of @xmath88 .
let @xmath89 denote the set at maximal distance from @xmath90 and let @xmath91 .
the soul theorem [ t : soulthm ] implies that @xmath5 can be written as the union of neighborhoods @xmath92 and @xmath93 along their common boundary @xmath94 , that is , @xmath95 in particular , when @xmath96 and @xmath97 is another fixed point set component with maximal dimension , one has the following result from @xcite .
[ t : sy2 ] let @xmath5 be a complete , non - negatively curved riemannian manifold admitting an isometric @xmath98 action . if @xmath99 contains two codimension @xmath100 components @xmath101 and @xmath102 , with one of them being compact , then @xmath101 is isometric to @xmath102 , @xmath103 , and @xmath5 is diffeomorphic to an @xmath21-bundle over @xmath101 with @xmath98 as its structure group .
in other words , there is a principal @xmath98-bundle , @xmath104 , over @xmath101 such that @xmath5 is diffeomorphic to @xmath105 .
we recall the list of closed @xmath14-manifolds with a smooth cohomogeneity one @xmath1 action ( cf .
@xcite ) , as they will appear throughout the paper .
they are : @xmath20 , a lens space @xmath106 , @xmath107 , @xmath108 , @xmath109 , @xmath110 , @xmath111 and @xmath112 . here
@xmath113 denotes the @xmath100-dimensional klein bottle and @xmath110 the non - trivial @xmath21-bundle over @xmath114 .
the manifold @xmath112 is obtained by gluing @xmath115 and @xmath116 along their boundary torus , where @xmath117 denotes the mbius band .
in this section we classify closed , orientable @xmath14-manifolds and closed , simply - connected @xmath13-manifolds , assuming they have non - negative curvature and admit an isometric action of a circle @xmath98 . in the case of a @xmath98 action
, we have the following result , which follows from the orlik - raymond - seifert classification of @xmath14-manifolds with a smooth @xmath98 action @xcite .
[ t : n3.1 ] let @xmath98 act isometrically on @xmath118 , a closed , orientable @xmath14-manifold of non - negative curvature .
then @xmath118 is equivariantly diffeomorphic to a spherical @xmath14-manifold , @xmath107 , @xmath119 , @xmath109 or one of four @xmath1-bundles over @xmath114 .
we break the proof into three cases : case 1 , where the action is free , case 2 , where the action is almost free , and case 3 , where the action has non - trivial fixed point set .
.5 cm _ case 1 : @xmath120 acts freely _ .25 cm in this case @xmath121 is a closed , orientable @xmath100-manifold of non - negative curvature and thus @xmath122 or @xmath123 by the gauss - bonnet theorem . since the action is free , @xmath118 is a principal circle bundle over @xmath124 and therefore @xmath118 is diffeomorphic to one of @xmath20 , @xmath106 , @xmath107 or @xmath109 .
.5 cm _ case 2 : @xmath98 acts almost freely _ .25 cm here @xmath118 is a seifert manifold supporting a smooth circle action .
since we have assumed that @xmath118 has non - negative curvature , @xmath118 admits a geometric structure modeled on @xmath20 , @xmath125 or euclidean space @xmath126 ( cf .
closed , orientable seifert manifolds with @xmath20 , @xmath126 or @xmath125 geometry supporting a smooth @xmath98 action have been classified ( cf .
when @xmath118 has @xmath20 geometry , @xmath118 must be diffeomorphic to a spherical @xmath14-manifold , that is , a quotient of @xmath20 by a finite subgroup of @xmath127 acting freely on @xmath20 .
we denote these manifolds in the usual fashion by their @xmath100-dimensional orbit spaces .
the @xmath14-sphere @xmath20 is denoted by @xmath21 and @xmath106 by @xmath128 or @xmath129 .
the remaining manifolds with @xmath20 geometry are denoted by @xmath130 , @xmath131 , @xmath132 and @xmath133 , where @xmath134 is an integer .
when @xmath118 has @xmath125 geometry , @xmath118 must be @xmath107 and , when @xmath118 has @xmath126 geometry , it must be diffeomorphic to @xmath109 or to one of four of the remaining five possible orientable , closed flat manifolds covered by @xmath109 . the fifth possibility is excluded immediately since it does not admit a circle action .
these four flat manifolds covered by @xmath109 are @xmath1 bundles over @xmath114 , described in @xcite .
their orbit spaces are @xmath135 , @xmath136 , @xmath137 and @xmath138 .
further , all of these closed , orientable @xmath14-manifolds with @xmath126 , @xmath20 or @xmath125 geometry , with a seifert fibration induced by an almost free circle action , do admit isometric circle actions inducing the given seifert fibration ( cf .
@xcite ) .
.5 cm _ case 3 : @xmath98 has non - trivial fixed point set _ .25 cm by definition , the action is fixed point homogeneous .
closed fixed point homogeneous manifolds @xmath14-manifold with nonnegative curvature were classified in @xcite and we recall their classification in the orientable case .
observe first that the fixed point set is @xmath139-dimensional , with at most two components , and these components are circles .
if @xmath140 contains two components , then by the double soul theorem [ t : sy2 ] we see that @xmath118 is one of the two @xmath21 bundles over @xmath114 and since @xmath118 is assumed to be orientable , it must be @xmath107 . if @xmath140 consists of a single component @xmath141 , then @xmath142 is a @xmath100-dimensional alexandrov space of non - negative curvature with boundary @xmath143 .
thus @xmath124 is an orientable , non - negatively curved topological manifold with boundary and the only possibilities are @xmath144 and @xmath145 .
we may exclude @xmath145 since @xmath118 is assumed to be orientable .
thus @xmath144 is the only possible orbit space .
the non - negative curvature hypothesis implies that the interior of @xmath144 has either no points with non - trivial finite isotropy , one point with finite isotropy @xmath146 or two points with finite isotropy @xmath147 .
these correspond , respectively , to @xmath20 , a lens space @xmath106 and @xmath119 .
+ it follows from the three cases analyzed above that @xmath118 can only be @xmath148 , where @xmath149 is a finite subgroup of @xmath127 , @xmath150 , @xmath109 , one of the four flat @xmath1-bundles over @xmath114 covered by @xmath109 or , finally , @xmath151 .
each of these manifolds supports only one isometric @xmath98 action with non - negative curvature yielding the possible orbit space structures ( cf .
@xcite ) . given perelman s work on the poincar conjecture @xcite ,
the classification of closed , simply - connected , non - negatively curved @xmath13-manifolds admitting an isometric @xmath98 action follows from earlier classification results in a curvature - free setting and a restriction on the euler characteristic , which is a simple consequence of the non - negative curvature assumption .
the first theorem is due to work of fintushel @xcite in combination with work of pao @xcite and perelman s proof of the poincar conjecture @xcite .
[ t : fpp ] a closed , simply - connected smooth @xmath13-manifold with a @xmath98 action is equivariantly diffeomorphic to a connected sum of @xmath152 , @xmath153 and @xmath154 .
let @xmath155 be a closed , simply - connected , non - negatively curved @xmath13-manifold and let @xmath156 be its euler characteristic .
it follows from work done independently by kleiner @xcite and searle and yang @xcite that @xmath157 . combining this with theorem [ t : fpp ] yields the following result in the case of non - negative curvature ( cf .
@xcite ) .
[ t : sy1 ] a closed , simply - connected , non - negatively curved @xmath13-manifold with an isometric @xmath98 action is diffeomorphic to @xmath158 or @xmath159 .
here we gather general facts about smooth cohomogeneity three torus actions @xmath160 on simply - connected , smooth manifolds and then consider the specific case when @xmath5 is @xmath0-dimensional .
the main goal of this section is to understand the structure of the singular sets , that is , the set of points in the orbit space @xmath45 corresponding to orbits with non - principal isotropy groups .
we begin with the following theorem from bredon @xcite , which characterizes the orbit space of a cohomogeneity three action .
[ t : bredon ] let @xmath28 be a compact lie group acting by cohomogeneity three on @xmath5 , a compact , simply - connected smooth manifold . if all orbits are connected , then @xmath45 is a simply - connected topological @xmath14-manifold with or without boundary .
it follows from the resolution of the poincar conjecture ( cf .
@xcite ) that @xmath45 is homeomorphic to one of @xmath20 , @xmath161 , @xmath162 or , more generally , to @xmath20 with a finite number of disjoint open @xmath14-balls removed .
we will see in the next section that non - negative curvature implies that @xmath45 can only be one of the first three manifolds from this list .
we also recall the following general result of bredon @xcite about the fundamental group of the orbit space : [ t : bredonpi1 ] let @xmath28 be a compact lie group acting on a topological space @xmath39 . if either @xmath28 is connected or @xmath28 has a nonempty fixed point set , then the orbit projection map @xmath163 induces an onto map on fundamental groups .
the next theorem , again from bredon @xcite , implies the absence of special exceptional orbits and , in particular , allows us to conclude that no fixed point set of finite @xmath147-isotropy has codimension one in @xmath5 .
this result will be used in the proof of lemma [ l : nofreeoralmostfree ] .
[ t : bredon_nse ] let @xmath5 be a smooth , simply - connected manifold admitting an action by a compact lie group . if a principal orbit is connected ( and hence all orbits are connected ) then there are no special exceptional orbits , that is , the dimension of the set of points belonging to exceptional orbits is of codimension greater than or equal to @xmath100 .
[ l : nofreeoralmostfree ] let @xmath164 act on @xmath165 , a closed , simply - connected smooth manifold
. then some circle subgroup has non - trivial fixed point set .
if all circle subgroups were to act freely , this would imply a free circle action on a closed , simply - connected @xmath13-manifold @xmath166 , which is impossible .
likewise , if the action is almost free , then there are finitely many finite isotropy groups .
let @xmath149 be the finite group generated by all these finite groups and consider the action of @xmath167 on @xmath168 , a closed , simply - connected topological space by theorem [ t : bredonpi1 ] .
we claim that @xmath168 must be a topological manifold .
note that the fixed point set of any subgroup of finite isotropy must be at least @xmath6-dimensional since it is invariant under the @xmath164 action and it will be at most @xmath169-dimensional because there are no special exceptional orbits by theorem [ t : bredon_nse ] .
the space of directions to the projection of these fixed point sets in @xmath168 will be either topological @xmath100-spheres or circles in all but one case , that is , where @xmath147 fixes an @xmath6-dimensional submanifold and acts freely on its normal @xmath21 .
however , this latter case is impossible , because such a fixed point set would project to a point in @xmath170 with space of directions @xmath171 and , by theorem [ t : bredon ] , this is a contradiction .
hence @xmath168 must be a closed , simply - connected topological manifold .
now , @xmath167 must act freely on @xmath168 and we have just seen that this is impossible .
therefore @xmath164 can not act almost freely on @xmath165 either .
let @xmath165 be a closed , simply - connected @xmath172-manifold with a cohomogeneity three @xmath164 action .
by the previous lemma , there is a circle subgroup @xmath173 with non - trivial fixed point set . in the case
where @xmath174 , there is a unique codimension @xmath100 fixed point set component . in general ,
when @xmath45 is homeomorphic to @xmath20 with @xmath175 disjoint open @xmath14-balls removed , @xmath176 , there will be @xmath175 codimension @xmath100 components of the fixed point sets of possibly different circles , one for each boundary component of @xmath45 . in the case where @xmath45 is homeomorphic to @xmath20 , the components of @xmath177 are of codimension greater than or equal to @xmath13 .
in this case , the following proposition , generalizing a result of rong s in dimension @xmath0 ( cf .
@xcite ) , shows there must be a minumum number of codimension @xmath13 fixed point set components , corresponding to isolated singular orbits @xmath178 .
[ p : tn-1orbits ] let @xmath164 act on @xmath165 , a simply - connected , smooth manifold .
suppose that @xmath45 is homeomorphic to @xmath20 and that there are exactly two orbit types : principal orbits @xmath164 and isolated singular orbits @xmath178 .
then there are at least @xmath179 isolated singular orbits @xmath178 .
let @xmath180 denote the manifold with boundary obtained by removing a small tubular neighborhood around each isolated singular orbit @xmath178 .
let @xmath181 denote the quotient space @xmath182 . by a standard transversality argument
we know that _
1(m_0)_1(m ) = \{1 } and _ 2(m_0)_2(m ) .
since there is no isotropy group of finite order we obtain a fibration @xmath183 and therefore a long exact sequence in homotopy : 0_2(m_0)_2(m^*_0)_1(t^n)_1(m_0)_1(m^*_0)0 .
since @xmath184 , it follows that @xmath185 .
since @xmath45 is a 3-sphere , applying the mayer - vietoris sequence to the pair @xmath186 , we obtain that @xmath187 , where @xmath188 is the number of isolated singular orbits .
it follows from the hurewicz isomorphism that @xmath189 and the above exact sequence in homotopy becomes 0_2(m_0)^r^n0 .
we conclude that @xmath190 and thus there are at least @xmath179 isolated singular orbits .
[ c : finiteisotropy ] proposition [ p : tn-1orbits ] remains valid in the presence of finite isotropy .
let @xmath149 be the finite group generated by the finite isotropy groups of the action .
as we saw earlier in the proof of lemma [ l : nofreeoralmostfree ] , @xmath191 is a closed , topological manifold .
moreover , @xmath191 is simply - connected by theorem [ t : bredonpi1 ] .
finally , observe that @xmath167 acts without finite isotropy on @xmath168 and the isolated @xmath192 orbits in @xmath168 correspond to isolated @xmath178 orbits in @xmath165 .
we observe that in the case where we have a @xmath1 action on @xmath16 , proposition [ p : tn-1orbits ] implies that when @xmath193 , there are at least three isolated circle orbits . in this subsection , we use the isotropy representation of the possible isotropy groups to better understand fixed point components of finite isotropy and their corresponding images in the orbit space @xmath45 . by a theorem of chang and skjelbred @xcite , components of @xmath194
are smooth submanifolds . when @xmath195 these components are orientable and of even codimension .
if @xmath196 , components of @xmath197 may also be non - orientable and by theorem [ t : bredon_nse ] , of codimension at least @xmath100 . in the case of a smooth @xmath1 action on a closed , simply - connected smooth @xmath0-manifold @xmath16 , components of @xmath198 must be at least @xmath100-dimensional , as we saw in the proof of lemma [ l : nofreeoralmostfree ] .
an analysis of the isotropy representations will show that for all cases the components of @xmath198 must be @xmath14-dimensional .
[ p:3dimfixed ] let @xmath1 act smoothly on @xmath16 , a closed , simply - connected smooth @xmath0-manifold .
if @xmath193 , then the following hold . *
the singular orbits of the action are @xmath98 and @xmath199 , @xmath200 . + * the exceptional orbits are @xmath201 , @xmath202 , and @xmath203 .
+ * in all cases where there is finite cyclic isotropy , the corresponding fixed point set of finite isotropy is of dimension @xmath14 . since we have assumed that @xmath45 is homeomorphic to @xmath20 , there are no points with @xmath1 isotropy .
observe that the normal sphere at any point of an exceptional orbit will be of dimension two .
thus the finite isotropy group of an exceptional orbit must be a subgroup of @xmath204 and of @xmath1 .
hence the only possible finite isotropy groups are @xmath205 , @xmath202 , and @xmath206 .
this proves parts ( 1 ) and ( 2 ) .
now we prove part ( 3 ) .
we first consider the singular orbits , observing that if we have a singular orbit of the form @xmath199 , then we have a @xmath207 action on the normal @xmath14-sphere to any point of the orbit .
in particular , there will be a finite cyclic subgroup of order @xmath175 in @xmath208 fixing circles in this normal @xmath14-sphere and therefore this orbit is contained in a fixed point set of finite isotropy of dimension @xmath14 .
if the singular orbit is @xmath98 , then the action of the circle on the normal @xmath20 is either free or almost free . in the latter case ,
a finite cyclic subgroup fixes a @xmath14-dimensional submanifold which contains the singular orbit .
we now consider the exceptional orbits . for a @xmath201 orbit , @xmath195 ,
the @xmath205 action on @xmath21 is never free and thus this exceptional orbit will be contained in a @xmath14-dimensional submanifold fixed by @xmath205 , @xmath195 .
it remains to show that for the exceptional orbit @xmath209 , the @xmath147 isotropy group also does not act freely on its normal @xmath21 .
this follows from the fact that the antipodal map , which reverses orientation , generates the only free @xmath147 action on @xmath21 and it is not a subgroup of @xmath204 . finally , we consider the exceptional orbit @xmath203 . the action of the isotropy subgroup , @xmath206 , on the normal @xmath21 produces a quotient space equal to the double right - angled spherical triangle with three vertices , each of which is fixed by a different @xmath147 subgroup of @xmath206 .
each fixed vertex corresponds to a @xmath14-dimensional submanifold fixed by the corresponding @xmath147 subgroup . for each @xmath203 orbit we will have exactly three such fixed point sets intersecting in this orbit .
thus , we conclude that the fixed point set of a finite cyclic group is always of dimension @xmath14 .
we now determine the structure of the singular sets in the orbit space in the particular case when @xmath5 is @xmath0-dimensional and @xmath193 .
we first observe that the existence of simple closed curves in @xmath45 whose points correspond to orbits with non - trivial finite cyclic isotropy @xmath210 is ruled out by work of montgomery and yang ( cf .
@xcite lemma 2.3 ) . taking this into account
, the following proposition follows directly from the proof of proposition [ p:3dimfixed ] . [
p : singsets ] let @xmath1 act smoothly on @xmath16 , a closed , simply - connected smooth @xmath0-manifold .
if @xmath193 , then the set of points in @xmath45 with non - trivial isotropy corresponds to a graph and the following hold : * the vertices of the graph correspond to isolated singular points and the edges correspond to points with finite cyclic isotropy .
+ * the pre - image of the closure of an edge corresponds to a @xmath14-dimensional manifold fixed by a finite cyclic group admitting a @xmath1 action of cohomogeneity one .
+ * isolated singular points will correspond to isolated circle orbits .
+ * edges can belong to graphs with vertices of degree @xmath139 , @xmath100 or @xmath14 , but only the latter type corresponds to a @xmath203 exceptional orbit . , the external vertices have isotropy conjugate to @xmath114 and the edges have isotropy @xmath147 . ]
we will denote by _ arc _ the closure of an edge with finite cyclic isotropy in the set of orbits with non - trivial isotropy in @xmath45 .
since the graphs corresponding to the singular set in @xmath45 carry isotropy information , we will refer to them as _ weighted graphs_. we now begin the process of determining what @xmath14-manifolds may actually occur as fixed point set components of a finite cyclic isotropy group .
since these components admit an ( almost ) effective @xmath1 action , they must be one of the manifolds listed in subsection [ ss:3mt2 ] .
we will eventually show , in section [ s : dim5results ] , that the only such @xmath14-manifolds that can occur are @xmath20 , @xmath106 , @xmath107 and @xmath110 .
we first observe that we may immediately rule out @xmath109 , since its orbit space would correspond to a simple closed curve in @xmath45 with finite cyclic isotropy and , as mentioned above , simple closed curves with finite cyclic isotropy will not occur . of the possible @xmath14-manifolds on the list , the non - orientable ones are @xmath108 , @xmath110 , @xmath111 and @xmath112 , and as such , they may only be fixed point set components of @xmath147 isotropy .
all have at least one exceptional orbit and correspond to the possible pre - images of arcs containing a vertex of degree three .
all these @xmath14-manifolds are non - orientable and therefore can only be fixed point set components of @xmath147 . @xmath108 and @xmath110 each have one exceptional orbit and one singular orbit and @xmath111 and @xmath112 each have two exceptional orbits ( see @xcite ) .
if the singular set in @xmath45 contains a vertex of degree three , then it may contain different types of trees as subgraphs .
two types of trees may occur .
the first type occurs if either @xmath108 or @xmath110 is the pre - image of an arc of @xmath147 isotropy , in which case , the singular set contains a tree with one vertex of degree three joined to three vertices of degree one or two only .
the second type occurs if @xmath111 or @xmath112 is the pre - image of an arc of @xmath147 isotropy , in which case the singular set contains a tree with an edge terminating in two vertices of degree three , each of which is joined to two more vertices of degree one or two .
we will see that when we take into consideration the lower curvature bound this second type of tree can not occur , allowing us to exclude @xmath111 and @xmath112 as possible fixed point set components of @xmath147 isotropy .
the first type of tree is the bipartite graph @xmath211 , commonly known as a _ claw _ ( cf .
since vertices and edges carry isotropy information , we shall refer to this configuration as a _ weighted claw _ ( see figure [ f : triad ] ) .
an example of the second possible tree appears in figure [ f:2claws ] .
we will refer to such graphs as _
weighted trees_. these graphs will appear in our analysis of the finite isotropy case in section [ s:5_dim_1 ] .
finally , we point out that the weighted graph could also contain a cycle .
moreover , this cycle could potentially be knotted in @xmath193 .
we will see in section [ ss : unknot ] that when the orbit space is non - negatively curved the cycle can not be knotted .
, the external vertices have isotropy conjugate to @xmath114 and the edges have isotropy @xmath147 . ] as mentioned above , using the non - negative curvature hypothesis , we will show in the next section that neither @xmath111 nor @xmath112 may occur as a @xmath14-dimensional fixed point set of @xmath147 isotropy , thus leaving only five possibilities for such fixed point sets , namely @xmath20 , @xmath106 , @xmath107 , @xmath110 and @xmath108 . as mentioned earlier , the only time we will have @xmath110 or @xmath108 appearing as the fixed point set component of @xmath147 isotropy is when @xmath45 contains a weighted claw and then each arc in this weighted claw must correspond to either @xmath110 or @xmath108 .
we observe that each of these @xmath14-manifolds is fixed by one and only one of the three @xmath147 subgroups of @xmath212 .
we will eventually show in section [ s : dim5results ] that @xmath108 can not occur as a component of @xmath213 , once we have established the possible @xmath0-manifolds that can occur .
in this section we will see how non - negative curvature restricts the structure of the orbit space of an isometric @xmath1 action on a closed , simply - connected @xmath0-manifold . throughout this section
we will let @xmath16 be a closed , simply - connected @xmath0-manifold of non - negative curvature with an isometric @xmath1 action . as we noted earlier , the quotient space of a smooth @xmath1 action on a closed
, simply - connected smooth @xmath0-manifold is homeomorphic to one of @xmath20 or @xmath20 with a finite number of disjoint open @xmath14-balls removed . for every open @xmath14-ball we remove we obtain an @xmath21 boundary component . in the presence of non - negative curvature we have the following proposition .
[ p : m * ] let @xmath16 be a closed , simply - connected , non - negatively curved @xmath0-manifold .
if @xmath1 acts isometrically on @xmath16 , then @xmath45 is homeomorphic to one of the following : * @xmath20 , if for any @xmath214 for which @xmath215 , @xmath216 ; + * @xmath161 or @xmath162 , if @xmath217 for some @xmath214 .
part ( 1 ) follows easily since only points belonging to a codimension two fixed point set of a circle will correspond to boundary points in the orbit space @xmath45 .
note that part ( 1 ) is independent of the curvature assumption . part ( 2 )
follows from the double soul theorem [ t : sy2 ] . in the previous section , in proposition [ p : tn-1orbits ]
, we found a lower bound of three for the number of isolated circle orbits in @xmath16 for the case where @xmath193 .
we now propose to determine an upper bound on the number of isolated circle orbits when @xmath16 is non - negatively curved .
theorem [ t:4-circles ] below will show that there can be at most four such orbits .
a simple application of the extent lemma tells us that in @xmath218 , where @xmath28 acts isometrically on @xmath5 , a closed manifold of positive curvature , there are at most @xmath14 singular points with space of directions isometric to @xmath219 or a `` thin '' @xmath219 , that is , the quotient of @xmath220 by an almost free @xmath114 action . if @xmath5 is non - negatively curved , the extent lemma tells us that there will be at most @xmath0 such singular points . a closer analysis of the geometry will allow us to show that in the case where @xmath5 is @xmath0-dimensional , non - negatively curved and admits an isometric @xmath1 action , there will be at most @xmath13 isolated circle orbits .
this upper bound follows from the generalization of an argument used in kleiner s thesis @xcite showing that an isometric circle action on a closed , simply - connected , non - negatively curved @xmath13-manifold has at most four isolated fixed points .
we remark that the same result was obtained in @xcite , but the argument used to prove the result was specific to dimension @xmath13 and does not generalize to higher dimensions .
the key observation that allows us to apply the techniques in @xcite to our situation is that the normal sphere at a point to each one of the circles fixed by some @xmath221 is @xmath14-dimensional .
we include the proof of the theorem here for the sake of completeness since kleiner s result was never published .
[ t:4-circles ] let @xmath16 be a closed non - negatively curved @xmath0-manifold with an isometric @xmath1 action .
then there are at most 4 isolated circle orbits of the @xmath1 action .
the proof of theorem [ t:4-circles ] will occupy the remainder of this subsection .
we begin by fixing some notation and recasting several lemmas from @xcite to meet our needs .
we assume that all geodesics have unit speed unless stated otherwise .
let @xmath222 be four distinct points in @xmath16 .
for @xmath223 , let @xmath224 be the set of minimizing normal geodesics from @xmath225 to @xmath226 and , for each triple @xmath227 , let @xmath228 for each pair of points @xmath225 , @xmath226 , @xmath229 , let @xmath230 be the set of initial directions of a normal minimizing geodesic from @xmath225 to @xmath226 , that is , @xmath231 finally , let @xmath232 denote the ( possibly degenerate ) tetrahedron determined by the four points @xmath233 . + before proceeding with the proof of theorem [ t:4-circles ] , we recall the following fact ( cf .
@xcite ) .
[ l : lipschitz ] suppose @xmath114 acts isometrically and fixed point freely on @xmath220 . then @xmath234 is smaller than @xmath235 .
that is , there is a surjective @xmath139-lipschitz map @xmath236 .
we have the following lemma .
[ l : angle_bound ] if there are @xmath13 isolated circle orbits @xmath237 , then , for distinct points @xmath238 , @xmath239 , and every quadruple of distinct integers @xmath240 , the tetrahedron @xmath232 is rigid in the following sense : @xmath241 and @xmath242 that is , the sum of angles at each vertex and the sum of angles of each face of @xmath232 are both @xmath243 . in the orbit space @xmath244 , the @xmath13 circles @xmath245 correspond to 4 points @xmath246 . by toponogov
s theorem for alexandrov spaces ( cf .
@xcite ) , we know that the sum of the angles of a geodesic triangle in @xmath247 will be greater than or equal to @xmath243 . connecting each pair of distinct points in @xmath248 by a minimal geodesic we obtain a configuration of four triangles and the total sum of the angles in this configuration will be greater than or equal to @xmath249 . for each one of the four points @xmath222 the coresponding isotropy group acts freely or almost freely on the normal space @xmath250 and the quotient of the unit normal sphere
@xmath251 is @xmath252 , the round sphere of radius @xmath253 in the first case or a thin " @xmath219 in the second case .
hence @xmath254 for any @xmath73 .
using the fact that @xmath255 , it is easily seen that for any triple of distinct points @xmath256 , we have ( x_j , x_k ) + ( x_j , x_l ) + ( x_k , x_l ) . thus summing over all the triangles formed by the points @xmath246 we find that the sum of their angles should be less than or equal to @xmath249 .
therefore this sum of angles must be exactly @xmath249 .
[ l:5_circle_orbits ] if there are @xmath0 isolated circle orbits @xmath257 then , for fixed @xmath258 and points @xmath259 , @xmath260 , @xmath261 , the following hold .
* for each @xmath262 and each @xmath238 , we have @xmath263 and its slice representation is the hopf action .
* for each @xmath261 , the sets @xmath230 , consist of a single @xmath114 hopf orbit and come in mutually orthogonal pairs , that is , given @xmath262 , for each set of distinct @xmath264 , up to reordering , we can assume that @xmath265 and the remaining angles are all @xmath266 . for convenience ,
let @xmath267 . by lemma
[ l : angle_bound ] , we have _ 5jk + _ 5kl + _ 5lj= , for the 4 points @xmath268 , with @xmath269 . for @xmath270 ,
let @xmath271 observe that @xmath272 .
we have already seen in the proof of lemma [ l : angle_bound ] that the sum of the distances between any set of three distinct points in @xmath273 is equal to @xmath243 for any @xmath274 . from the geometry of the space of directions
this tells us that @xmath275 and @xmath276 are points in @xmath277 . in the case
where @xmath278 then either two of them are antipodal ( at distance @xmath279 in @xmath219 ) or all three of them lie on a great circle .
note that in this last case the three points can not lie in half of the great circle . in the second case , where @xmath277 is a thin " @xmath219 ,
two of the points must be at distance @xmath279 . since this is true for any choice of three of the four possible directions
, one may conclude that @xmath277 can not be smaller than @xmath219 .
in particular , this implies that the isotropy group of each isolated circle orbit is @xmath114 and that the action is the hopf action .
this proves part ( 1 ) of the lemma .
finally , one can conclude that any four singular points in the space of directions must lie on a great circle and consist of two pairs of antipodal points .
this implies that the sets @xmath280 lie on a great circle , consist of a single point each and must pair off as antipodal points , thus proving part ( 2 ) of the lemma .
[ l:5_circle_orbits_b ] let @xmath114 act on @xmath281 by scalar multiplication and suppose that @xmath282 .
then either @xmath283 or there exists a unique @xmath284 such that @xmath285 .
let @xmath286 be the hopf map .
the riemannian submersion metric on @xmath287 is isometric to @xmath288 .
the images of the orbits @xmath289 in @xmath287 are either separated by @xmath290 or they are joined by a unique minimizing geodesic segment @xmath291 with @xmath292 .
if @xmath293 satisfies @xmath294 and @xmath295 is a minimizing geodesic segment from @xmath296 to @xmath297 in @xmath220 , then @xmath295 projects to the minimizing curve @xmath298 .
this implies that @xmath295 is the unique horizontal lift of @xmath298 starting at @xmath296 and @xmath297 is unique .
we will assume that there are at least @xmath0 isolated circle orbits @xmath299 and will obtain a contradiction .
for each @xmath258 , let @xmath238 and observe that for each pair of points @xmath238 , @xmath259 , @xmath229 , @xmath230 is a subset of the unit normal sphere @xmath251 .
we will now show that when @xmath300 , the set @xmath301 can not be a single @xmath114 orbit in the unit normal sphere @xmath302 , in contradiction with lemma [ l:5_circle_orbits ] .
assume after relabeling points that @xmath303 .
let @xmath304 be minimal normal geodesics from @xmath305 to @xmath306 and @xmath305 to @xmath307 , respectively .
by lemma [ l:5_circle_orbits ] , @xmath308 now since @xmath309 , it follows by the discussion of the equality case in the proof of toponogov s theorem ( cf .
@xcite ) that there exists a flat , totally geodesic triangular surface @xmath310 with geodesic edges @xmath304 and @xmath311 , where @xmath312 is a minimal geodesic from @xmath306 to @xmath307 . now ,
if we replace @xmath313 with @xmath314 , where @xmath284 , we obtain another flat , totally geodesic triangular surface @xmath315 with geodesic edges @xmath316 and @xmath317 , where @xmath317 is a minimal geodesic from @xmath306 to @xmath318 . in particular , @xmath319 then , by lemma [ l:5_circle_orbits_b ] , there is a unique @xmath320 such that @xmath321 .
it then follows that , for this @xmath322 , @xmath323 and thus @xmath324 .
hence @xmath325 and we have a contradiction .
let @xmath165 be a closed , non - negatively curved manifold with an isometric @xmath326 action .
suppose that @xmath193 and that there are isolated @xmath178 orbits .
then there are at most four such isolated @xmath178 orbits .
in particular , if @xmath327 then there are none .
the first result follows directly from the proof of theorem [ t:4-circles ] .
the second result follows by proposition [ p : tn-1orbits ] .
the following lemma , easily generalized from rong @xcite , allows us to calculate the betti numbers with @xmath146 coefficients of @xmath16 .
[ l : h2a_finite_extension ] [ p : r ] suppose @xmath1 acts isometrically on @xmath16 , a closed , simply - connected @xmath0-manifold .
if there are exactly 3 isolated circle orbits , then @xmath328 has trivial free rank . if there are exactly @xmath13 isolated circle orbits , then @xmath328 has free rank equal to @xmath139 .
we will now show that a weighted graph containing a tree with a vertex of degree three , that is , a weighted graph containing a weighted claw or a weighted tree , may occur only when there are exactly @xmath14 isolated circle orbits . with this result
, we may then conclude that neither @xmath111 nor @xmath112 can never occur as the fixed point set of a finite group .
[ p : nonorientable ] let @xmath1 act isometrically on @xmath16 , a closed , simply - connected , non - negatively curved @xmath0-manifold .
if @xmath193 and there exists a non - orientable @xmath14-manifold @xmath329 fixed by a @xmath147 subgroup , then the projection of @xmath329 in @xmath45 must belong to a weighted claw and there can be no other singular points in @xmath45 corresponding to an isolated circle orbit , besides the external vertices of the claw .
let @xmath330 be the weighted graph corresponding to the set of orbits with non - trivial isotropy in @xmath45 .
there are two cases we must exclude .
the first case is where @xmath330 contains a weighted claw as a subgraph and a vertex of degree @xmath51 , @xmath139 or @xmath100 ( see , for example , figure [ f : claw_cycle ] ) .
the second case is when @xmath330 contains a weighted tree as a subgraph ( see , for example , figure [ f : tree_cycle ] )
. corresponds to an exceptional orbit with isotropy @xmath212 . ]
correspond to exceptional orbits with isotropy @xmath212 . ]
we begin with the first case .
let @xmath331 denote the center point in @xmath45 of the weighted claw , that is , whose space of directions @xmath332 is the double right - angled spherical triangle @xmath333 , and let @xmath334 , @xmath335 , denote the points in @xmath45 corresponding to the vertices of the weighted claw , each of which corresponds to an orbit with @xmath98 or @xmath336 isotropy .
we note that the space of directions for each of these external vertices is either an @xmath337 , that is a thin " @xmath100-sphere of diameter @xmath279 or a possibly thinner @xmath100-sphere of diameter @xmath279 .
if there is a fourth singular point @xmath338 corresponding to an isolated circle orbit in @xmath16 , then @xmath277 is either an @xmath219 or a thin " @xmath219 . since @xmath219 is the
largest " of these spaces of directions ( cf .
[ l : lipschitz ] ) , we will assume that @xmath278 .
it is clear that in @xmath332 the vertices of the spherical triangle correspond to the geodesic directions to the points @xmath339 and @xmath340 , and consequently @xmath341 . without loss of generality
, we will assume that @xmath342 , in which case it follows that @xmath343 .
now , by lemma [ l : angle_bound ] , the tetrahedron @xmath344 is rigid , in the sense that the angles of every face sum to @xmath243 and the angles at every vertex sum to @xmath243 . in particular , because of this rigidity and because each of the points @xmath339 and @xmath340 has space of directions a thin @xmath219 , it follows that at every one of the vertices @xmath339 and @xmath340 of @xmath344 there will be an angle of @xmath279 .
further , the maximal configuration for the spaces of directions of the points @xmath339 and @xmath340 will be where the remaining angles at each vertex in @xmath344 are all @xmath266 , that is , @xmath345 for for all @xmath346 and @xmath347 , where @xmath348 , whereas , @xmath349 will be equal to @xmath279 for one value of @xmath346 and for the remaining values it will be equal to @xmath266 . without loss of generality we may choose specific values for all angles of the form @xmath350 . once these choices are determined , the rigidity of @xmath344 will determine the remaining angles .
it now follows by toponogov s theorem that the angle sum of any triangle in any tetrahedron formed by these five singular points must be greater than or equal to @xmath243 .
when we consider the tetrahedron @xmath351 , we see that when we substitute all the known values for the angles the lower bound on the sum of the angles for any triangle forces the following two inequalities : _ 513 + _ 315 + , + _ 514 + _
+ as we saw previously , @xmath352 and one of @xmath353 or @xmath354 is equal to @xmath279 and the other is equal to @xmath266 . in particular this
tell us that one of the angles @xmath355 or @xmath356 is greater than or equal to @xmath279 and the other is greater than or equal to @xmath266 .
since @xmath357 this immediately gives us a contradiction and thus this case can not occur . for the second case , where the weighted graph contains a weighted tree ,
we observe that the addition of the singular point @xmath358 , corresponding to a @xmath203 orbit will produce an analgous contradiction and thus this case can not occur either .
we summarize the results of this subsection in the following theorem : [ t:3dimfiniteisotropy ] let @xmath17 act isometrically on @xmath16 , a closed , simply - connected , non - negatively curved @xmath0-manifold . if @xmath193 , then the fixed point set components of finite cyclic isotropy ( if they exist ) are : * @xmath20 , @xmath106 , @xmath107 , @xmath108 or @xmath110 when there are three isolated circle orbits ; + * @xmath20 , @xmath106 or @xmath107 when there are four isolated circle orbits .
we recall the following theorem of bredon @xcite : [ t : cohominequalities ] suppose that @xmath359 is a prime and that @xmath360 acts on the finite - dimensional space @xmath39 with @xmath361 closed and invariant .
suppose that @xmath28 acts trivially on @xmath362 and let @xmath363 .
then , for any @xmath364 , we have _
i0 ^k+2i(f , fb ; _
p)_i0 ^k+2i(x , b ; _ p ) .
we observe that any diffeomorphism in @xmath1 is homotopic to the identity , since it is contained in a torus .
thus we may apply this theorem to the situation at hand to obtain the following corollary : [ c : orbitconfigs ] let @xmath1 act isometrically on @xmath16 , a closed , simply - connected , non - negatively curved @xmath0-manifold . if @xmath193 and the orbit space contains a weighted claw , then @xmath16 is not @xmath18 .
this follows directly by applying the inequality in theorem [ t : cohominequalities ] , observing that if either @xmath110 or @xmath108 is contained in @xmath365 , then @xmath366 .
we will now analyze the special case where the singular set in the orbit space contains a cycle .
following work of grove and wilking @xcite , we will show that this cycle is unknotted in @xmath193 . recall that the arcs in a cycle correspond to the projection of fixed point sets of finite isotropy .
we have the following result .
[ t : unknotted_cycle ] let @xmath16 be a closed , simply - connected , non - negatively curved @xmath0-manifold with an isometric @xmath1 action and orbit space @xmath367 . if the singular set in the orbit space @xmath45 contains a cycle @xmath368 , then the following hold : * the cycle @xmath368 is the only cycle in the singular set in @xmath45 . + * if there are four isolated circle orbits , then @xmath368 comprises all of the singular set , i.e. , @xmath369 is smooth . +
* suppose there are exactly three isolated circle orbits , the cycle @xmath368 contains only two of them , and there are no exceptional orbits of isotropy @xmath212
. then the finite isotropy fixing one of the @xmath14-manifolds corresponding to one of the arcs of the cycle must be @xmath147 .
+ * the cycle @xmath368 is unknotted in @xmath45 . we will first prove parts ( 1 ) and ( 2 ) of the theorem .
our strategy is the following . for any given weighted graph containing a cycle @xmath368 in @xmath45 we will construct a branched double cover @xmath370 with branching set @xmath368 and show that @xmath371 is a simply - connected alexandrov space of non - negative curvature ( see lemmas [ l : b_nonneg_alex ] , [ l : b_space_directions ] and [ l : b_s_connected ] below ) . by lemma [ l : b_space_directions ] below and the proof of theorem [ t:4-circles ] , the cover @xmath371 can contain at most four points projecting down to isolated circle orbits in @xmath45 .
parts ( 1 ) and ( 2 ) of the theorem then follow from the fact that the weighted graph @xmath372 corresponding to the set of singular orbits can contain two cycles only if @xmath330 has four vertices . to construct the branched cover ,
first observe that a generator of @xmath373 is given by a normal circle to @xmath368 .
recall that index two subgroups of @xmath373 are in one - to - one correspondence with two - fold covers of @xmath369 .
hence there is a unique two - fold cover @xmath374 of @xmath369 .
let @xmath375 be a tubular neighborhood of @xmath368 in @xmath367 .
observe that @xmath375 is a solid torus and therefore @xmath376 .
it follows that @xmath377 also has a two - fold cover .
now let @xmath378 so that @xmath379 is a two - fold branched cover , with branching set @xmath368 .
further , @xmath371 admits a @xmath147 action , which is isometric with respect to the metric induced by the orbital distance metric from @xmath45 , with fixed point set @xmath368 .
[ l : b_nonneg_alex ] the space @xmath371 is a non - negatively curved alexandrov space .
observe that @xmath371 is locally isometric to @xmath45 outside of the branching set @xmath368 .
let @xmath380 be the union of arcs in @xmath368 with isotropy @xmath205 , @xmath195 .
we have two cases : case one , where the cycle @xmath368 contains all the singular points corresponding to isolated circle orbits , and case two , when there are exactly three isolated circle orbits and of the corresponding singular points only two belong to a cycle . for case one , proceeding as in the proof of lemma 2.3 in @xcite , one verifies that the set @xmath381 is convex in @xmath371 .
the conclusion follows after observing that any geodesic triangle in @xmath371 is the limit of geodesic triangles in @xmath381 . for case two
, there are only two graphs that correspond to this case : graph 4 of figure [ f:3_vertices_wgraphs ] and graph 2 of figure [ f : claw_wgraphs ] . for graph 4 of figure [ f:3_vertices_wgraphs ]
, we may proceed as in case one . here
we verify that the set @xmath382 is convex in @xmath371 , where @xmath305 and @xmath306 are the lifts of the point @xmath359 corresponding to the isolated circle orbit in @xmath45 that does not belong to the cycle @xmath368 . for graph 2 of figure
[ f : claw_wgraphs ] , let @xmath383 be the arc in @xmath368 with isotropy @xmath205 , @xmath195 , and let @xmath384 be the lift in @xmath371 of the arc in the claw with isotropy @xmath147 not contained in @xmath368 . observe that @xmath384 is a minimal geodesic between the lifts of the isolated circle orbit not contained in the cycle @xmath368 . as above , one verifies that the set @xmath385 is convex in @xmath371 and the conclusion follows after observing that geodesics triangles in @xmath371 are limits of geodesic triangles in @xmath385 .
[ l : b_space_directions ] let @xmath386 be a point corresponding to an isolated circle orbit in @xmath16 and consider @xmath40 as a point in @xmath371 . then @xmath387 is smaller than or equal to @xmath219 .
there is a two - fold branched cover @xmath388 .
we know that the space of directions @xmath389 is a thin " @xmath219 .
we will denote this space by @xmath390 .
observe that @xmath391 can be written as the join of a circle , @xmath392 , of diameter @xmath393 and @xmath394 , of diameter @xmath279 , where the former is the normal space of directions to @xmath368 at the point @xmath40 and the latter corresponds to the tangent space of directions at @xmath40 of @xmath368 in @xmath45 .
since @xmath371 is a two - fold branched cover of @xmath45 with branching locus @xmath368 , the corresponding space of normal directions in @xmath371 will be of twice the diameter as the space of normal directions in @xmath45 . in particular , this means that @xmath387 corresponds to the orbifold @xmath395 or @xmath396 .
since at least one of @xmath397 is greater than @xmath100 , it follows that this orbifold is smaller than or equal to @xmath219 , as we saw earlier ( cf .
lemma [ l : lipschitz ] ) .
[ l : b_s_connected ] the space @xmath371 is simply - connected .
we will prove this by contradiction , so assume that @xmath398 contains at least two elements .
observe that @xmath399 , the universal cover of @xmath371 , is a compact alexandrov space of non - negative curvature .
there are at least three singular points @xmath225 in @xmath371 , corresponding to isolated circle orbits in @xmath45 .
therefore , in @xmath399 there will be at least six points @xmath400 covering these points @xmath45 . by lemma [ l : b_space_directions ] , the spaces of directions @xmath401 are smaller than or equal to @xmath219 .
on the other hand , the extent lemma implies that there can be at most five such points in @xmath399 , yielding a contradiction .
now we prove part ( 3 ) .
let @xmath205 and @xmath402 be the finite isotropy groups fixing the two @xmath14-manifolds corresponding to the two arcs of the cycle @xmath368 . without loss of generality , we may assume that @xmath403 . in this case we may take a @xmath175-fold branched cover of @xmath45 with branching locus @xmath368 .
it follows from the proof of theorem [ t:4-circles ] that @xmath196 , since otherwise the branched cover would have more than four singular points . finally , we prove part ( 4 ) .
observe that @xmath371 is a topological @xmath14-manifold and , by lemma [ l : b_s_connected ] , @xmath371 is simply connected .
hence , by the resolution of the poincar conjecture , @xmath371 must be homeomorphic to @xmath20 . recall that we have an isometric @xmath147 action on @xmath371 fixing @xmath368 . by the equivariant version of the poincar conjecture @xcite
, it follows that this action is equivalent to a linear action on a standard @xmath220
. therefore @xmath404 .
note that @xmath147 is not equivalent to the action of @xmath405 , since there is a branching locus , that is a unique circle fixed by the @xmath147 action .
thus , without loss of generality , _ 2=\ { -1 & & & + & -1 & & + & & 1 & + & & & 1 + } .
therefore @xmath406 is not knotted .
this concludes the proof of theorem [ t : unknotted_cycle ] .
we can now classify closed , simply - connected , non - negatively curved @xmath0-manifolds with an isometric @xmath1 action , corresponding to the almost maximal symmetry rank case in dimension @xmath0 .
we summarize our results in the following theorem .
[ t : dim5 ] let @xmath16 be a closed , simply - connected @xmath0-manifold with non - negative curvature admitting an isometric @xmath1 action
. then @xmath16 is diffeomorphic to @xmath18 , @xmath407 , @xmath408 or @xmath4 . by lemma [ l : nofreeoralmostfree ] ,
the @xmath1 action is neither free nor almost free . in particular , this tells us that there is always some circle subgroup with non - empty fixed point set .
we then have two cases to consider : case a , where some circle subgroup acts fixed point homogeneously and therefore @xmath45 is @xmath161 or @xmath162 , and case b , where no circle subgroup acts fixed point homogeneously and hence @xmath193 . throughout
, our main goal will be to compute @xmath328 .
the conclusions of theorem [ t : dim5 ] will then follow by the barden - smale classification of simply - connected @xmath0-manifolds @xcite .
we remark that it is only in case b , where @xmath193 and we have finite isotropy , that the wu manifold , @xmath4 , may occur .
observe also that if one circle subgroup acts freely , then @xmath16 fibers over one of the four manifolds @xmath409 or @xmath159 ( cf .
theorem [ t : sy1 ] ) .
the corresponding total space is diffeomorphic to @xmath18 , @xmath19 or @xmath408 ( cf .
it follows that in the case where we obtain the wu manifold , no circle subgroup may act freely .
let @xmath5 be a non - negatively curved manifold with a fixed point homogeneous @xmath98 action . by definition
, this means that there is a component @xmath411 of @xmath99 of codimension @xmath100 .
the following proposition restricts the fundamental group of @xmath411 depending on the dimension of the set at maximal distance to @xmath412 in the orbit space @xmath413 of the action .
we let @xmath414 be the orbit projection map of the fixed point homogeneous circle action .
[ p:1rank ] let @xmath415 be a closed , simply - connected , non - negatively curved manifold of dimension @xmath416 with an isometric @xmath98 action and suppose that @xmath417 contains an @xmath418-dimensional component @xmath419 .
let @xmath420 be the set at maximal distance from @xmath419 in the orbit space @xmath421 .
* if @xmath420 has dimension @xmath422 , then @xmath420 is fixed by the @xmath98 action , is isometric to @xmath419 and @xmath423 is simply - connected .
+ * if @xmath420 has dimension @xmath424 , then @xmath419 is simply - connected .
+ first we prove ( 1 ) .
if we suppose that @xmath97 is not fixed by the @xmath114 action , then @xmath91 has dimension @xmath425 .
we may decompose @xmath5 as a union of neighborhoods of @xmath426 and @xmath371 , which we will denote @xmath427 and @xmath428 , respectively .
their common boundary is a circle bundle over @xmath426 which we denote by @xmath429 .
observe that both @xmath427 and @xmath429 are orientable , but that @xmath428 is not ( this follows from the mayer - vietoris sequence of the triple @xmath430 ) .
since @xmath415 is simply - connected it follows by duality that @xmath431 and that the torsion subgroup of @xmath432 is trivial .
further , since @xmath429 is a compact , orientable manifold of dimension @xmath425 , the torsion subgroup of @xmath433 is also trivial . likewise , since @xmath427 deformation retracts onto @xmath426 , @xmath434 . since @xmath371 is non - orientable , the torsion subgroup of @xmath435 is equal to @xmath147 . if we write down the first few elements of the mayer - vietoris sequence of the triple @xmath430 we have : 0h_n(m)h_n-1(v)h_n-1(u)h_n-1(v)h_n-1(m ) + h_n-2(v)h_n-2(u)h_n-2(v)h_n-2(m ) . + substituting known values
we obtain : 00 0 0 + ( h_n-2(v))((h_n-2(u))_2 ) ( h_n-2(m ) ) .
+ the sequence is not exact and thus this case can not occur .
this in turn implies that the inverse image of @xmath436 in @xmath5 must be exactly @xmath436 and thus @xmath436 is a component of @xmath437 .
it then follows from the double soul theorem [ t : sy2 ] that @xmath5 is an @xmath21 bundle over @xmath419 and hence @xmath426 must be simply - connected . to prove ( 2 ) , let @xmath295 be a loop in @xmath438 .
since @xmath415 is simply - connected , @xmath295 bounds a @xmath100-disk @xmath144 .
let @xmath439 and observe that @xmath440 . by transversality , we can perturb @xmath144 so as to lie in the complement of @xmath441 , a tubular neighborhood of @xmath442 , while keeping @xmath443 in @xmath419 .
the conclusion follows after observing that @xmath144 is now contained in @xmath444 , which deformation retracts onto @xmath419 .
the assertions in proposition [ p:1rank ] hold trivially in dimension @xmath100 , since in this case the fixed point set components of codimension @xmath100 are isolated points . in dimension @xmath14 , however , proposition [ p:1rank ] fails , since a @xmath139-dimensional fixed point set component , being compact , must be a circle and thus has infinite fundamental group
. we will now classify non - negatively curved simply connected @xmath0-manifolds @xmath16 with an isometric @xmath1 action such that some circle subgroup @xmath98 acts fixed point homogeneously .
we remark that simply connected @xmath0-manifolds with non - negative curvature and a fixed point homogeneous isometric circle action were classified in @xcite and that the resulting classification is the same in both cases .
we further remark that the methods of proof are distinct . in the following two propositions we will characterize @xmath329 in the case where @xmath91 is of dimension three .
we must consider two cases : case one , where @xmath445 is not fixed by any circle action and case two , where @xmath445 is fixed by a distinct circle subgroup of @xmath1 .
we begin with case one .
[ p : noseiferts ] let @xmath16 be a closed , simply - connected , non - negatively curved @xmath0-manifsold admitting an isometric @xmath1 action .
suppose that @xmath329 is the unique @xmath14-dimensional fixed point set component of @xmath446 for some @xmath214 and that @xmath91 is of dimension @xmath14 .
further suppose that @xmath371 is not the fixed point set of any circle subgroup of @xmath1 .
then @xmath329 is diffeomorphic to one of @xmath20 , @xmath106 , @xmath107 or @xmath447 .
observe that @xmath329 is orientable , but may not be simply - connected . since it admits a circle action it must be one of the @xmath14-manifolds listed in theorem [ t : n3.1 ] .
the circle action can be free , almost free or admit fixed points .
observe that the action can not be @xmath205-ineffective .
recall from the proof of theorem [ t : n3.1 ] that the @xmath14-manifolds with spherical geometry , @xmath448 , @xmath449 , @xmath450 ,
@xmath451 , @xmath134 , and the ones with euclidean geometry , @xmath135 , @xmath138 , @xmath136 or @xmath137 may only occur when the circle action is almost free and that @xmath109 may occur for any type of circle action .
we will first show that if the action is almost free , then the only possibilities for @xmath329 that can occur are the ones listed in the proposition , as well as @xmath109 .
finally , we will show that @xmath109 can not occur for any type of action .
thus we assume @xmath329 to be one of the @xmath14-manifolds with spherical or euclidean geometry listed above . since the circle action is almost free , the image of @xmath329 in the quotient space , @xmath452 , is a connected subspace of the boundary @xmath453 . by the soul theorem for alexandrov spaces , the distance function from @xmath453
is concave and therefore any singular points , that is , points with space of directions of diameter @xmath454 , in @xmath45 must be at maximal distance from @xmath455 .
in particular , these singular points correspond to orbits of the @xmath1 action in @xmath16 that are singular or exceptional and in the latter case they must be of the form @xmath203 ( cf .
the proof of proposition [ p:3dimfixed ] ) . since we assume that the circle action on @xmath329 is almost free , there will be fixed circles of finite isotropy . by our previous analysis of the isotropy representations
, these fixed circles must belong to a @xmath14-dimensional submanifold of finite isotropy , say @xmath205 , intersecting @xmath329 in the same circle .
we will denote this @xmath14-manifold by @xmath456 .
observe that @xmath456 admits a cohomogeneity one @xmath1 action and it must have at least one singular orbit , which is precisely the circle that belongs to @xmath457 .
now we note that if @xmath329 were to contain at least two circles fixed by the same @xmath205 subgroup , it might be possible for these two circles to correspond to the two distinct singular orbits of a @xmath456 .
if this were the case , then either there must be a unique principal orbit or there is an @xmath458 at maximal distance from @xmath329 .
if there is a unique principal orbit of @xmath456 at maximal distance from @xmath329 where @xmath456 turns around " to head back to @xmath329 , we then have a point where there are four distinct directions fixed by the corresponding isotropy subgroup and therefore all of @xmath16 would be fixed by this isotropy , making the action @xmath205-ineffective , contrary to hypothesis . likewise ,
if we have a @xmath458 at maximal distance , then we will have two points where this happens , again giving us a contradiction .
since @xmath445 is convex , it is a non - negatively curved alexandrov space admitting a @xmath1 action . by hypothesis
, this action must be of cohomogeneity one and therefore by @xcite @xmath445 is an invariant submanifold of @xmath16 ( cf . also @xcite ) .
since @xmath445 may have at most two non - principal orbits it follows easily that none of the @xmath14-manifolds with spherical geometry @xmath448 , @xmath449 , @xmath450 , @xmath451 , @xmath134 , nor the ones with euclidean geometry @xmath135 , @xmath138 , @xmath136 or @xmath137 may occur .
now we have two cases to consider : case 1 , where @xmath456 is orientable and case 2 , where @xmath456 is non - orientable .
recall that @xmath456 is only non - orientable when @xmath196 . in the first case
, it follows from the cohomogeneity one decomposition that @xmath456 has two singular orbits ( cf .
@xcite , @xcite ) and in the second case , it will have one singular and one exceptional orbit .
one of these singular orbits will belong to @xmath329 and since singular orbits not belonging to @xmath329 must be contained in the set at maximal distance from @xmath329 , the other orbit is contained in @xmath445 . for the manifolds @xmath131 , @xmath132 , @xmath133 , @xmath138 , @xmath136 and @xmath137 it follows that we must have at least two singular orbits in @xmath445 corresponding to the intersection of the corresponding @xmath456 , @xmath195 , with @xmath445 and another singular or exceptional orbit corresponding to the intersection of @xmath456 with @xmath445 for the third isotropy group @xmath205 .
this yields a contradiction , since @xmath445 can have at most two non - principal orbits .
for the manifolds @xmath130 and @xmath135 if at least two of the isotropy subgroups correspond to orientable @xmath456 , then we obtain a contradiction in exactly the same manner .
we now consider the case where at least two of the isotropy groups in @xmath130 and at least three in @xmath135 correspond to a non - orientable @xmath456 .
observe that all the @xmath147 isotropy subgroups of these manifolds belong to a unique circle subgroup of @xmath1 and are therefore all the same @xmath147 subgroup .
as we saw in our analysis of the isotropy groups in proposition [ p:3dimfixed ] , at an exceptional orbit the three @xmath14-manifolds that must share this exceptional orbit must also have distinct @xmath147 isotropy . if there is an exceptional orbit ,
then for every circle fixed by a @xmath147 in @xmath329 we would have two more @xmath14-manifolds fixed by distinct @xmath147 subgroups and whose other singular orbit must be contained in @xmath329 , since the exceptional orbit is already contained in @xmath445 .
this , however , is impossible and so this case can not occur .
thus when the action is almost free only @xmath20 , @xmath106 , @xmath107 , @xmath109 or @xmath119 may occur . finally , @xmath109 is ruled out , since @xmath109 has quotient @xmath1 modulo a circle action , and this would imply that @xmath459 , where @xmath460 .
this , however , is a contradiction , since the connected components of @xmath461 are homeomorphic to @xmath21 .
we now consider case two .
[ l : b3=f3 ] let @xmath1 act isometrically on @xmath16 , a closed , non - negatively curved , simply - connected @xmath0-manifold .
suppose that @xmath462 , and both @xmath329 and @xmath445 are fixed by different circle subgroups .
if @xmath329 is one of the spherical manifolds , @xmath448 , @xmath449 , @xmath450 , @xmath451 , @xmath134 , or one of the flat manifolds , @xmath135 , @xmath138 , @xmath136 or @xmath137 , then @xmath445 must be diffeomorphic to @xmath329 .
as we saw in proposition [ p : noseiferts ] , if @xmath329 is one of the manifolds listed in the statement of the lemma , then there will be a three - dimensional manifold @xmath456 fixed by a @xmath205 subgroup and intersecting @xmath329 in a circle .
each @xmath456 will also intersect @xmath445 in a circle and thus both @xmath329 and @xmath445 are diffeomorphic to one another .
[ p : b3fixed ] let @xmath1 act isometrically on @xmath16 , a closed , non - negatively curved , simply - connected @xmath0-manifold .
suppose that @xmath462 , and both @xmath329 and @xmath445 are fixed by different circle subgroups .
then both @xmath329 and @xmath445 are diffeomorphic to one of @xmath20 , @xmath106 , @xmath107 or @xmath447 .
observe first that ,
if @xmath329 is one of the spherical manifolds , @xmath448 , @xmath449 , @xmath450 , @xmath451 , @xmath134 , or one of the flat manifolds , @xmath135 , @xmath138 , @xmath136 or @xmath137 , by lemma [ l : b3=f3 ] , @xmath445 is diffeomorphic to @xmath329 .
now , we consider the isolated singular points in the orbit space @xmath460 corresponding to the circles fixed by a finite cyclic group in @xmath445 and @xmath329 .
for each of these points , the corresponding space of directions will be the join of an @xmath463 and a single point at distance @xmath279 from this circle .
that is , it will be a positively curved @xmath100-disc of diameter @xmath279 .
there will be six or eight such isolated singular points in @xmath45 and it is easily seen using the extent lemma [ l : xt ] that this gives us a contradiction to the non - negative sectional curvature condition .
thus none of these possibilities is allowed . as in the proof of proposition [ p : noseiferts ]
, we see that @xmath109 can not occur and thus the only manifolds left are @xmath20 , @xmath106 , @xmath107 and @xmath119 .
we are now in a position to prove the following theorem : [ t : fph ] let @xmath1 act isometrically on @xmath16 , a closed , non - negatively curved , simply - connected manifold .
if @xmath462 , then @xmath16 is diffeomorphic to @xmath18 , @xmath19 or @xmath3 . by hypothesis
, we have a circle @xmath214 acting fixed point homogeneously on @xmath16 with orbit space @xmath464 .
let @xmath329 be a component of @xmath465 .
we first observe that if there are two three - dimensional components of @xmath465 , then by the double soul theorem [ t : sy2 ] there are exactly two such components , they are isometric and @xmath16 is diffeomorphic to an @xmath21-bundle over @xmath329 .
since @xmath16 is simply - connected , @xmath329 must also be simply - connected and therefore @xmath329 must be @xmath20 and , as there is only one @xmath21 bundle over @xmath20 ( cf .
@xcite ) , @xmath16 is diffeomorphic to @xmath466 .
we now consider the case where there is a unique three - dimensional component , @xmath329 of @xmath446 , and let @xmath420 be the set at maximal distance from @xmath329 in the orbit space @xmath464 .
we have @xmath467 .
if @xmath468 , then @xmath97 is a component of @xmath469 and as we saw above , @xmath16 is diffeomorphic to @xmath19 .
we first consider the case where @xmath470 or @xmath139 .
in this case @xmath329 is simply - connected by proposition [ p:1rank ] and is therefore @xmath20
. if @xmath420 has dimension @xmath51 , then it is the soul of @xmath464 and , as in @xcite , @xmath16 is diffeomorphic to @xmath18 .
if @xmath420 is 1-dimensional , it must be an interval or a circle . since @xmath471 is homeomorphic to the boundary of a neighborhood of @xmath472 ,
we see that @xmath472 is an interval . by kleiner
s isotropy lemma [ l : kleiner ] , the only points in @xmath472 that may have finite non - trivial isotropy are the endpoints . in the presence of points with finite isotropy
, @xmath329 must be a lens space or the connected sum of two lens spaces , neither of which is possible .
hence all the points in @xmath472 are regular , @xmath464 is a manifold with boundary @xmath473 , the soul of @xmath464 is a regular point and again we conclude that @xmath16 is @xmath18 .
suppose now that @xmath420 is @xmath100-dimensional . by proposition [ p : noseiferts ] and proposition
[ p : b3fixed ] , @xmath329 can only be one of @xmath20 , @xmath106 , @xmath107 or @xmath447 .
recall that @xmath474 is the lift of @xmath475 under the orbit projection map @xmath476 . by proposition [ p : b3fixed ] , we know that if @xmath445 is fixed by some distinct circle subgroup , then it too can only be one of @xmath20 , @xmath106 , @xmath107 or @xmath447 .
thus it remains to characterize @xmath445 when it is not fixed by any circle subgroup .
to do so we use the fact that @xmath445 admits a cohomogeneity one @xmath1 action .
once we do this , we will have two cases remaining : case one , where @xmath329 admits an almost free circle action and there are then @xmath14-manifolds in @xmath16 fixed by finite isotropy and case two , where there is no finite isotropy .
we will see that in both cases we may decompose @xmath16 as the union of two disc bundles over @xmath329 and @xmath445 , respectively , and we may then directly calculate the homology groups of @xmath16 using the mayer - vietoris sequence to classify @xmath16 using the results of smale and barden ( cf .
@xcite , @xcite ) .
suppose then that @xmath1 acts on @xmath445 by cohomogeneity one . in order to understand what the possibilities for @xmath445 are we recall that @xmath477 must be an interval and therefore @xmath475
can only be one of @xmath21 , @xmath144 , @xmath171 or a cylinder @xmath145 , since it is non - negatively curved .
we first analyze the possible isotropy configurations on @xmath478=i$ ] . by our previous analysis of the isotropy representations ( cf .
proposition [ p:3dimfixed ] ) , it follows that there can be no exceptional orbits of the circle action in @xmath475 .
the action can , however , be @xmath205-ineffective , in which case @xmath445 must also be fixed by @xmath205 and the @xmath1 cohomogeneity one action is @xmath205-ineffective . in this particular case
it follows that @xmath445 can only be @xmath107 , @xmath20 or @xmath479 .
we will therefore assume that the circle action on @xmath475 is effective .
then the endpoints of @xmath480 correspond to either two singular orbits , one singular orbit and one principal orbit , or two principal orbits . in the first case
, @xmath445 corresponds to @xmath20 or @xmath479 . in the second case
, @xmath445 corresponds to @xmath107 . in the last case ,
@xmath475 must be a cylinder @xmath145 . in this case
the soul of @xmath464 is a circle and @xmath329 must be @xmath107 .
this case also corresponds to the case in which @xmath445 is also @xmath107 .
we summarize our results in the following proposition .
let @xmath1 act isometrically on @xmath16 , a closed , non - negatively curved , simply - connected @xmath0-manifold .
let @xmath462 and suppose that @xmath329 is the unique @xmath14-dimensional component of @xmath481 for some @xmath221 and that @xmath91 is of dimension three .
* if @xmath445 is fixed by a distinct circle subgroup of @xmath1 , then @xmath329 and @xmath445 are diffeomorphic to one of @xmath20 , @xmath106 , @xmath107 or @xmath447 . + * if @xmath445 is not fixed by any circle subgroup of @xmath1 , then @xmath329 is diffeomorphic to one of @xmath20 , @xmath106 , @xmath107 or @xmath447 and @xmath445 is diffeomorphic to one of of @xmath20 , @xmath106 or @xmath107 .
we may now complete the proof of the theorem .
note first that if @xmath1 acts without finite isotropy , then it is clear that we may decompose @xmath16 as a union of disc bundles over @xmath329 and @xmath445 , respectively . since @xmath16 is simply - connected , it is not possible for either @xmath329 or @xmath445 to be @xmath447 .
all other combinations of @xmath329 and @xmath445 are possible . in all cases we see from the mayer - vietoris sequence of the triple @xmath482 that @xmath483 and
thus @xmath16 is diffeomorphic to one of the @xmath20 bundles over @xmath21 by results of smale @xcite and barden @xcite . finally ,
if there is finite isotropy , in the orbit space , @xmath460 , we see that there may be at most two intervals corresponding to three - dimensional manifolds fixed by finite cyclic groups joining at most two distinct points in the image of @xmath329 in @xmath45 to at most two distinct points of the image of @xmath445 in @xmath45 .
we further observe that the angle that the interval corresponding to the image of the three - dimensional submanifold of finite isotropy forms with these points is @xmath279 .
therefore at the points of the image of @xmath329 in @xmath45 , the normal space to the corresponding points of @xmath329 will be the tangent space to the three - dimensional submanifold fixed by finite isotropy .
the same is true for the corresponding points in @xmath445 .
if we then consider the unit normal vector field to @xmath329 in @xmath16 , this then allows us to isotope the boundary of a tubular neighborhood around @xmath329 onto the boundary of a tubular neighborhood of @xmath445 and we may decompose @xmath16 as a union of two disc bundles over @xmath329 and @xmath445 , where , as above , each may be one of @xmath20 , @xmath106 or @xmath107 and neither may be @xmath119 . as in the case without isotropy , we see that @xmath16 is diffeomorphic to one of the @xmath20 bundles over @xmath21 by results of smale @xcite and barden @xcite .
this concludes the proof of theorem [ t : fph ] we consider now the case where no circle acts fixed point homogeneously , that is , @xmath193 , and there are either three or four isolated circle orbits .
we will prove the following theorem : [ t : only - fixed - circles ] let @xmath16 be a closed , simply - connected , non - negatively curved @xmath0-manifold admitting an isometric @xmath1 action .
if @xmath193 , then @xmath16 is diffeomorphic to @xmath18 , @xmath4 , @xmath19 or @xmath408 . by the proof of proposition [ p : tn-1orbits ] , in the case where there is no finite isotropy , we have the following result : [ p : m5-t2-casea ] let @xmath16 be a closed , simply - connected , non - negatively curved @xmath0-manifold admitting an isometric @xmath1 action .
suppose @xmath193 and that there is no finite isotropy .
* if there are exactly three isolated circle orbits , then @xmath16 is diffeomorphic to @xmath18 . +
* if there are four isolated circle orbits , then @xmath16 is diffeomorphic to @xmath19 or @xmath408 .
+ we now consider the case where the @xmath1 action admits finite isotropy groups .
we will devote the rest of this section to the proof of the following proposition : [ p : finiteisotropy ] let @xmath16 be a closed , simply - connected , non - negatively curved @xmath0-manifold with an isometric @xmath1 action .
suppose that @xmath193 and there is finite isotropy .
* if there are exactly three isolated circle orbits , then @xmath16 is diffeomorphic to @xmath18 or the wu manifold , @xmath4 . +
* if there are four isolated circle orbits , then @xmath16 is diffeomorphic to @xmath19 or @xmath408 .
+ we first note that combining proposition [ p : singsets ] , theorem [ t:4-circles ] and theorem [ t : unknotted_cycle ] , we may now make a complete list of all the graphs that may occur in the case where we have three isolated circle orbits and in the case where we have four isolated circle orbits .
we list these graphs in figures [ f:3_vertices_wgraphs ] , [ f : claw_wgraphs ] and [ f:4_vertices_wgraphs ] .
recall now that , by theorem [ t : unknotted_cycle ] , if the weighted graph contains a cycle , then this cycle must be unknotted in @xmath193 .
we will now show in all cases where we have a cycle that we may decompose the manifold as a union of disc bundles , where at least one of the disc bundles is over one arc of the cycle .
first show that , in all cases in which we have a cycle , the manifold decomposes as a union of disc bundles in the following fashion . in the case where there are three isolated circle orbits ,
if the cycle contains all three of them then the manifold decomposes as a union of disc bundles over one @xmath14-dimensional fixed point set of finite isotropy and an isolated circle orbit . in the case where there are three isolated circle orbits and a cycle that contains only two of them
, we will see that the manifold decomposes as a union of disc bundles over one @xmath14-dimensional fixed point set of finite isotropy and over a @xmath14-manifold that is either a fixed point set of distinct finite isotropy or corresponds to the inverse image of a minimal curve between the images of one isolated circle orbit and the image of an orbit with @xmath147 isotropy . in the case where there are four isolated circle orbits , recall that a cycle must contain all four of them by theorem [ t : unknotted_cycle ] and we can decompose the manifold as a union of disc bundles over two @xmath14-manifolds , where both correspond to a fixed point set of ( possibly distinct ) finite isotropy .
once we have established these results , we will proceed with a case - by - case analysis for the possible fixed point sets of finite isotropy corresponding to the possible weighted graphs of the @xmath1 action .
we have the following proposition : [ p : bundledecomp ] let @xmath16 be a closed , simply - connected , non - negatively curved @xmath0-manifold with an isometric @xmath1 action .
suppose that @xmath193 and there is finite isotropy .
suppose , in addition , that the weighted graph , @xmath330 , corresponding to the singular set of the action contains a cycle @xmath368 , corresponding to graphs ( 3 ) and ( 4 ) in figure [ f:3_vertices_wgraphs ] , graph ( 2 ) in figure [ f : claw_wgraphs ] , and graph ( 5 ) in figure [ f:4_vertices_wgraphs ] .
then the following are true .
* if @xmath330 is graph ( 3 ) in figure [ f:3_vertices_wgraphs ] , then @xmath16 decomposes as the union of a disc bundle over a @xmath14-dimensional submanifold @xmath485 fixed by finite cyclic isotropy , corresponding to the pre - image of an arc in @xmath368 , and a disc bundle over the remaining circle orbit not contained in @xmath486 . + * if @xmath330 is graph ( 4 ) in figure [ f:3_vertices_wgraphs ] , then @xmath16 decomposes as a union of disc bundles over two @xmath14-dimensional submanifolds .
one of these @xmath14-manifolds corresponds to the fixed point set of @xmath205 isotropy , @xmath195 , and the other corresponds to the pre - image of the arc between the remaining isolated circle orbit and an exceptional orbit @xmath209 which projects to an interior point of the arc of @xmath147 isotropy . + * if @xmath330 is graph ( 2 ) in figure [ f : claw_wgraphs ] , then @xmath16 decomposes as the union of disc bundles over two @xmath14-dimensional submanifolds .
one of these @xmath14-manifolds corresponds to the pre - image of the arc with @xmath205 isotropy , @xmath195 , and the other to the pre - image of the arc with @xmath147 isotropy containing the remaining isolated circle orbit . + * if @xmath330 is graph ( 5 ) in figure [ f:4_vertices_wgraphs ] , then @xmath16 decomposes as the union of two disc bundles over two disjoint @xmath14-dimensional submanifolds fixed by finite isotropy ( although not necessarily the same group )
. we will first prove parts ( 1 ) and ( 4 ) corresponding , respectively , to graph ( 3 ) in figure [ f:3_vertices_wgraphs ] and graph ( 5 ) in figure [ f:4_vertices_wgraphs ] .
in both cases the weighted graph is a cycle @xmath368 .
fix an arc @xmath487 in @xmath368 corresponding to a fixed point set component of finite cyclic isotropy @xmath210 . note
that whether we have three or four isolated circle orbits , the corresponding edges of the weighted cycle @xmath368 in @xmath45 form the angle @xmath279 .
thus , at isolated circle orbits corresponding to the endpoints of the arc @xmath488 , the normal space to the @xmath14-dimensional submanifold @xmath489 , fixed by @xmath210 , will be the tangent space to the @xmath14-dimensional submanifold fixed by finite cyclic isotropy , corresponding to the lift of arcs adjacent to @xmath488 . in graph
( 3 ) in figure [ f:3_vertices_wgraphs ] the cycle @xmath368 contains three edges , and there are three @xmath14-dimensional submanifolds fixed by finite isotropy , each one corresponding to the lift of an arc in @xmath368 .
consider @xmath490 , and a unit normal vector field to it . since the normal vector field is tangent to the other two @xmath14-dimensional submanifolds fixed by finite isotropy , we may isotope the boundary of a tubular neighborhood around @xmath491 along this vector field to the boundary of a tubular neighborhood around the remaining isolated circle orbit @xmath472
. therefore we may decompose @xmath16 as a union of disc bundles over @xmath472 and over @xmath491 .
the same argument in the preceding paragraph works for graph ( 5 ) in figure [ f:4_vertices_wgraphs ] , corresponding to the case of four isolated circle orbits . here
we will isotope the boundary of a tubular neighborhood around @xmath492 , corresponding to the pre - image of an arc @xmath493 in the cycle @xmath368 , to the boundary of a tubular neighborhood around @xmath494 , corresponding to the pre - image of the arc opposite to @xmath487 .
we now prove part ( 2 ) , corresponding to graph ( 4 ) in figure [ f:3_vertices_wgraphs ] .
recall that in this case one of the edges in @xmath368 corresponds to orbits with isotropy @xmath147 , while the other one corresponds to orbits with isotropy @xmath205 , @xmath495 .
we will denote the arc in @xmath368 corresponding to a fixed point set component of isotropy @xmath147 by @xmath496 , and we will let @xmath487 be the arc in @xmath368 corresponding to the fixed point set component with finite isotropy @xmath205 .
we now form an arc @xmath497 in @xmath45 by joining the vertex not contained in @xmath368 to @xmath488 via a shortest geodesic in @xmath45 .
the interior of this arc consists of principal orbits and the pre - image of this arc is a cohomogeneity one @xmath14-manifold , @xmath494 . proceeding as in cases ( 1 ) and ( 4 ) , we may decompose @xmath16 as a union of disc bundles over @xmath498 and @xmath494 . to prove part ( 3 ) , we let @xmath487 be the arc not contained in the weighted claw , that is , the arc containing two isolated circle orbits and corresponding to a fixed point set component of isotropy @xmath205 , @xmath495 . we let @xmath499 be the arc in the claw containing the isolated circle orbit not contained in @xmath488 .
proceeding as above , we may decompose @xmath16 as a union of disc bundles over @xmath498 and @xmath500 .
we are now ready to prove proposition [ p : finiteisotropy ]
. our strategy will be to analyze the weighted graphs grouped into three separate cases , where the first two cases correspond to part ( 1 ) of proposition [ p : finiteisotropy ] and the third case corresponds to part ( 2 ) of proposition [ p : finiteisotropy ] .
the first case will be all the graphs in figure [ f:3_vertices_wgraphs ] , with the exception of graph ( 4 ) .
the second case will consist of graph ( 4 ) of figure [ f:3_vertices_wgraphs ] and graphs ( 1 ) and ( 2 ) of figure [ f : claw_wgraphs ] .
the third case will be the graphs of figure [ f:4_vertices_wgraphs ] .
the weighted graph corresponding to the singular set of the action is one of those listed in figure [ f:3_vertices_wgraphs ] or figure [ f : claw_wgraphs ] .
it follows from the discussion in section [ ss : component_finite_iso ] that if the weighted graph is one of those in figure [ f:3_vertices_wgraphs ] , then a fixed point set component of finite isotropy can only be one of @xmath20 , @xmath106 or @xmath107 ; if the weighted graph is one of those in figure [ f : claw_wgraphs ] , then the fixed point set components of finite isotropy corresponding to arcs in the claw can only be @xmath110 or @xmath108 and the corresponding isotropy subgroup is a @xmath147 subgroup of @xmath1 in each case . as mentioned above , we have divided the proof into three cases : case a , where the graphs are all those from figure [ f:3_vertices_wgraphs ] with the exception of graph ( 4 ) ; case b , corresponding to graphs ( 4 ) from figure [ f:3_vertices_wgraphs ] and ( 1 ) and ( 2 ) from figure [ f : claw_wgraphs ] .
+ _ case a. _ we have the following lemma .
[ l : nolpqs2xs1 ] let @xmath1 act isometrically on @xmath16 , a closed , simply - connected @xmath0-manifold of non - negative curvature and suppose that @xmath193 .
if there are exactly @xmath14 isolated circle orbits and the weighted graph of the action is one of graphs ( 1 ) , ( 2 ) or ( 3 ) from figure [ f:3_vertices_wgraphs ] , then neither @xmath106 or @xmath107 may be a component of a fixed point set of finite isotropy .
if the weighted graph is one of graphs ( 1 ) or ( 2 ) , which do not not contain a cycle , then we may complete it to a cycle , adding edges corresponding to shortest geodesics consisting of regular points in the orbit space , so that each vertex in the graph has degree @xmath100 ( see figure [ f:3_vertices_completed_arc ] ) .
we may then decompose @xmath16 as the union of a disc bundle over a fixed point set component of finite isotropy and the remaining isolated circle orbit .
a tubular neighborhood around the isolated circle orbit will be a @xmath501-bundle over @xmath114 with boundary an @xmath20 bundle over @xmath114 .
a tubular neighborhood around the fixed point set component of finite isotropy will be a @xmath144-bundle over @xmath106 or @xmath107 and therefore the boundary of both tubular neighborhoods must be @xmath502 .
when we consider the mayer - vietoris sequence of this decomposition we immediately obtain a contradiction and therefore neither of these two manifolds may occur as a fixed point set component of finite isotropy .
_ case b. _ we have the following lemma .
[ l : caseb ] let @xmath1 act isometrically on @xmath16 , a closed , simply - connected @xmath0-manifold of non - negative curvature and suppose that @xmath193 , there are exactly three isolated circle orbits and the singular set corresponds to graph ( 4 ) of figure [ f:3_vertices_wgraphs ] or graph ( 2 ) of figure [ f : claw_wgraphs ]
. then a fixed point set component of finite isotropy @xmath205 , @xmath503 , can only be one of @xmath20 or @xmath504 .
we must rule out @xmath106 , where @xmath505 , and @xmath107 as fixed point set components of isotropy @xmath205 , @xmath195 .
recall that , for both graphs under consideration , @xmath16 decomposes as a union of disc bundles over one of @xmath106 or @xmath107 , and over one of @xmath110 or @xmath108 .
it follows from the mayer - vietoris sequence of this decomposition that only two possibilities do not give rise to a contradiction : that @xmath16 may be the union of disc bundles over @xmath504 and @xmath110 or over @xmath20 and @xmath110 . with these two lemmas
we may now complete the proof of part ( 1 ) of proposition [ p : finiteisotropy ] . from lemma
[ l : nolpqs2xs1 ] above we conclude that the only possible fixed point set components for graphs ( 1 ) , ( 2 ) , and ( 3 ) of figure [ f:3_vertices_wgraphs ] are @xmath20 . in this case
, it follows from the mayer - vietoris sequence that @xmath506 and therefore @xmath16 is diffeomorphic to @xmath18 by work of smale and barden @xcite . from lemma
[ l : caseb ] we note that for graphs ( 4 ) of figure [ f:3_vertices_wgraphs ] or graph ( 2 ) of figure [ f : claw_wgraphs ] that @xmath506 when @xmath504 is the fixed point set component of isotropy @xmath507 , and @xmath508 when @xmath20 is . both graphs ( 1 ) and ( 2 ) of figure [ f : claw_wgraphs ] contain a weighted claw and in the case of graph ( 1 ) , we may complete the graph , joining two disjoint arcs via edges corresponding to shortest geodesics consisting of regular points in the orbit space .
we may then decompose the manifold as a union of disc bundles over the pre - image of the arc joining two edges of @xmath147 isotropy and the remaining edge of @xmath147 isotropy .
we further note that in these last two cases it follows from corollary [ c : orbitconfigs ] that @xmath508 and from theorem [ t : cohominequalities ] that the arcs corresponding to @xmath14-manifolds of @xmath147 isotropy must be @xmath110 .
it now follows by @xcite that for graph ( 4 ) of figure [ f:3_vertices_wgraphs ] @xmath16 is diffeomorphic to @xmath18 or the wu manifold and for graphs ( 1 ) and ( 2 ) of figure [ f : claw_wgraphs ] @xmath16 is diffeomorphic to the wu manifold .
we have now completed the proof of part ( 1 ) of proposition [ p : finiteisotropy ] .
it remains to prove ( 2 ) of proposition [ p : finiteisotropy ] . to do this
, it suffices to show that @xmath509 for every possible fixed point set of finite isotropy and thus , by @xcite , @xmath16 is diffeomorphic to one of the two @xmath20 bundles over @xmath21 . in this case
the possible weighted graphs are shown in figure [ f:4_vertices_wgraphs ] . for graphs ( 1 ) through ( 4 ) , we may complete the weighted graph by joining disjoint isolated circle orbits or arcs via edges corresponding to shortest geodesics consisting of regular points in the orbit space . in this way we obtain a graph that is an unknotted cycle ( see figure [ f:4_vertices_completed_arc ] ) and now for all the possible graphs we may decompose @xmath16 as the union of two disc bundles over the @xmath14-dimensional manifolds that correspond to opposite arcs of the cycle . in this particular case , the @xmath14-dimensional manifold may be one of @xmath20 , @xmath106 or @xmath107 . in all cases and for all possible combinations
, we see that @xmath483 and the result follows .
given @xmath510 and @xmath511 , let ( ( _ 1 , _ 2 ) , ( z_1 , z_2 , z_3))(e^2i_1z_1 , e^2i_2z_2 , z_3 ) . here both circles @xmath512 and @xmath513 fix a @xmath14-sphere .
the corresponding singular set in the orbit space is @xmath14 isolated singular points .
given @xmath510 and @xmath514 , let ( ( _ 1 , _ 2 ) , ( z_1 , z_2 , x_1 , x_2 , x_3))(e^2i_1z_1 , e^2i_2z_2 , x_1 , x_2 , x_3 ) . here both circles @xmath512 and @xmath513 fix an @xmath107 and the action is the product of the cohomogeneity one action on @xmath20 combined with the trivial action on @xmath21 .
the corresponding singular set in the orbit space is @xmath13 isolated singular points .
we give examples of actions on @xmath18 and on @xmath19 with finite isotropy and with @xmath14 and @xmath13 isolated circle orbits , respectively .
the action on @xmath18 was given by rong @xcite and we include it here for the sake of completeness .
( ( _ 1 , _ 2 ) , ( z_1 , z_2 , z_3))(e^2i(_1 + p_2)z_1 , e^2i(_1 + q_2)z_2 , e^2i(_1 + r_2)z_3 ) .
here there are @xmath14 isolated circle orbits . if @xmath515 are pairwise relatively prime and the differences @xmath516 , @xmath517 and @xmath518 are also pairwise relatively prime , then the singular set of the action is a cycle in the orbit space and the closure of each edge corresponds to an @xmath20 fixed by finite isotropy .
[ ex : fixed_s3 ] given @xmath510 and @xmath519 , we let @xmath520 act on @xmath296 by @xmath521 where @xmath522 is the matrix @xmath523 @xmath359 , @xmath71 , @xmath524 are pairwise relatively prime integers , as are the differences @xmath516 , @xmath517 and @xmath518 , and , without loss of generality , @xmath525 . here
there are @xmath13 isolated circle orbits and the finite groups @xmath526 each fix a distinct @xmath107 that has empty intersection with the other whereas the finite group @xmath527 fixes two disjoint copies of @xmath20 , intersecting each of the fixed @xmath107 in an isolated circle orbit .
the corresponding singular set in the orbit space is a quadrangle with vertices corresponding to isolated circle orbits and edges corresponding to arcs with finite isotropy .
[ ex : wu_manifold ] let @xmath528 act canonically on @xmath4 .
there are three involutions given by the diagonal matrices with entries @xmath529 and @xmath530 .
each of these involutions will fix an @xmath531 , each of which intersects in a @xmath532 .
the corresponding singular set in the orbit space is a weighted claw .
burago , y. , gromov , m. , perelman , g. , a.d . alexandrov s spaces with curvatures bounded below " , ( russian ) uspekhi mat .
nauk , v. 47 , no . 2 ( 284 )
, 351 , 222 ( 1992 ) , translation in russian math .
surveys , v. 47 , no .
2 , 158 ( 1992 ) . devito , j. , `` the classification of simply connected biquotients of dimension at most @xmath533 and @xmath14 new examples of almost positively curved manifolds '' , ph.d .
thesis ( 2011 ) , university of pennsylvania , http://repository.upenn.edu/edissertations/311 .
grove , k. , `` developments around positive sectional curvature '' , surveys in differential geometry .
geometry , analysis , and algebraic geometry : forty years of the journal of differential geometry , 117133 , surv .
, 13 , int . press , somerville , ma , 2009 | we show that a closed , simply - connected , non - negatively curved @xmath0-manifold admitting an effective , isometric @xmath1 action is diffeomorphic to one of @xmath2 , @xmath3 or the wu manifold @xmath4 . |
entanglement is perhaps one of the most baffling features of quantum systems .
indeed , the possibility of producing entangled states was first considered as the signature of the incompleteness of quantum mechanics @xcite .
however , eventually entanglement was verified experimentally and recognized as a valid and fundamental feature of the quantum world . moreover , practical implications of quantum entanglement are foreseen nowadays , e.g. , in the fast developing fields of quantum information and computation @xcite . there , in order to achieve the highest computational power , it is desirable ( at least theoretically ) to produce states with large entanglement . for a bipartite system consisting of two subsystems of dimensions @xmath0 and @xmath1 ( with , e.g. , @xmath2 ) , several proxies were introduced to quantify the degree of entanglement , which are all functionals of @xmath0 non - negative eigenvalues @xmath16 of the reduced density matrix , satisfying the normalization constraint @xmath17 and called _ schmidt eigenvalues _ ( see section [ model ] for details ) .
these functionals are , to name a few , the entanglement entropies - von neumann entropy @xmath18 or @xmath19-parametrized rnyi entropy or @xmath20 the rnyi entropy converges respectively to the von neumann entropy @xmath21 or to @xmath22 , where @xmath23 is the largest eigenvalue .
] , @xmath24 , where @xmath25 ; the purity @xmath26 , and also its reciprocal value - the so - called schmidt number @xmath27 defined as the effective number of non - zero coefficients in the schmidt decomposition ( see below ) , i.e. , the number of effective degrees of freedom contributing to the entanglement .
the schmidt number is defined on the interval @xmath7 $ ] and @xmath28 corresponds to maximal entanglement . in some instances , the schmidt number can be directly measured experimentally @xcite . _
random pure _ states , for which the schmidt eigenvalues are strongly correlated random variables , have attracted a strong interest in recent years : they are believed to constitute a promising candidate for quantum computation since their _ average _ entanglement entropy is close to the maximal possible entropy of a completely degenerate state , when all @xmath29 and hence , @xmath30 @xcite .
furthermore , they may serve as a reference point whose entanglement content is to be compared to an arbitrary quantum state evolving in time .
they also appear in the study of quantum chaotic or non - integrable systems @xcite . finally ,
as we proceed to show , the schmidt eigenvalues for random pure states have the same joint probability density function ( jpdf ) as the so - called _ scaled eigenvalues _
( see , e.g. , @xcite ) , the eigenvalues of wishart random matrices normalized by the trace .
the latter have diverse applications both in statistics and in performance analysis of wireless communication systems , the spectrum sensing problem in cognitive radio networks being just one particular example ( for which @xmath0 and @xmath1 are the number of antennas and the number of samples per antenna , respectively ) @xcite . for other results and applications of entangled random systems ,
see @xcite . as the schmidt eigenvalues are random variables for random pure states , so are the entropies , the purity and the schmidt number .
statistical properties of the entropies and of the purity , and some other related observables ( apart from the schmidt number , which did not receive much attention ) , have been rather extensively studied in recent years @xcite , focusing either on the limiting behavior when @xmath0 and @xmath1 are small or , conversely , tend to infinity . in particular , the full distribution of the purity has been analyzed for small @xmath0 in @xcite , and for large @xmath0 its characteristic function has been determined in @xcite . for square @xmath10 systems with @xmath31 , the leading asymptotic behavior of the distributions of von neumann and rnyi entropies ( including large deviation tails ) was studied in @xcite using a coulomb gas technique .
it was there realized that , quite surprisingly , even though the average entropy of the random pure state is close to its maximal value @xmath32 , the probability of this _ closeness _ may be very small .
we will comment on this result further on . for arbitrary , not necessarily large @xmath0 and @xmath1 , which is often the most relevant case in practice , the only available results so far concern the spectral ( or one - point ) densities @xcite and moments of the purity @xcite . note ,
however , that the observations made in @xcite warn us that _ average _ values may not be representative of the actual behavior : to gain a better understanding of how meaningful they are , one has to go beyond the linear statistics and estimate the effective broadness of the corresponding distributions .
this would require , e.g. , the knowledge of the variances of these entanglement quantifiers for any @xmath0 and @xmath1 , which is lacking at present .
moreover , we stress that the von neumann entropy may not be a proper measure of the degree of entanglement as it exhibits a logarithmic growth with @xmath0 : in the limit @xmath31 , it may not be possible to distinguish whether the system attains complete or partial entanglement .
the schmidt number , in particular , seems to be a better quantifier of the degree of entanglement since it grows with @xmath0 as a power law ( see below ) . in this paper , we focus on non - linear statistics for a bipartite entanglement of random pure states .
we consider first the @xmath3-point densities_n$ ] stand for the first @xmath3 components of the vector @xmath33 . ]
@xmath34_n)$ ] of the schmidt eigenvalues of arbitrary oder @xmath3 .
we determine such densities using three complementary approaches : the first is the generalization of the method developed previously in @xcite for calculation of the one - point densities of the schmidt eigenvalues .
this allows us to show in a very compact way that the @xmath3-point densities @xmath34_n)$ ] can be expressed as a suitable laplace transform of corresponding @xmath3-point densities @xmath35_n)$ ] of the standard @xmath36-wishart - laguerre ( @xmath36-wl ) ensembles of random matrices .
our second approach hinges on the ( so far seemingly unnoticed ) fact that the fixed - trace ( ft ) schmidt eigenvalues have the same jpdf as the so - called scaled variables ( see , e.g. , @xcite ) , which also allows for a very straightforward derivation of the general relation between @xmath34_n)$ ] and @xmath35_n)$ ] .
finally , we establish a link between the cross - moments of the schmidt eigenvalues of arbitrary order and an analogous functional of the scaled variables , which allows us to relate the cross - moments of the @xmath36-ft and of the standard @xmath36-wl ensembles . employing these tools , we present explicit expressions for the two - point densities and for the variance of the von neumann entropy .
further on , we focus on the moments @xmath5 of the schmidt number @xmath6 .
we derive exact results for systems with @xmath37 and @xmath38 and arbitrary @xmath1 .
we show that in systems with a fixed @xmath0 and @xmath39 , the moment of order @xmath40 tends to @xmath41 , which implies that the schmidt number attains its maximal value @xmath0 : this is a signature that such systems become completely entangled in this limit .
next , concentrating of square systems with @xmath10 , we spot a previously unnoticed connection between @xmath5 and the probability @xmath11 that the smallest eigenvalue @xmath12 of a @xmath13 matrix belonging to the gaussian unitary ensemble ( see , e.g. @xcite for more details ) is larger than @xmath14 .
we show that @xmath42 with @xmath15 can be interpreted as the moment generating function of the purity , while @xmath42 with @xmath43 is the generating function of the moments of @xmath6 of order @xmath44 .
the moments of order @xmath40 lower than @xmath45 are determined exactly as a certain integral of @xmath42 . as a byproduct of our analysis here ,
we present an exact asymptotic expansion for @xmath42 with @xmath15 and arbitrary @xmath0 .
finally , capitalizing on the results in @xcite for the large deviation form of @xmath42 , we establish the leading asymptotic behavior of the moments of @xmath6 in the limit @xmath31 .
we show that in the square systems @xmath46 , which signifies that here @xmath6 attains only half of its maximal value and the square systems are far of being completely entangled .
this may apparently explain the paradoxical behavior observed in @xcite .
this paper is organized as follows : in sec .
[ model ] we describe the random pure state setting and introduce some basic definitions . in sec
. [ densities ] we focus on the @xmath3-point densities of the @xmath36-ft ensemble and establish a general relation with the analogous quantities of the @xmath36-wl ensembles .
eventually we also derive a series of useful relations between the cross - moments of the ft and @xmath36-wl ensembles .
next , in sec .
[ spec ] we present explicit , closed - form expressions for the two - point densities and in sec .
[ var ] we derive an exact expression for the variance of the von neumann entropy , valid for any @xmath0 and @xmath1 .
[ schmidt ] is devoted to the analysis of the schmidt number @xmath6 .
capitalizing on the known results for the distribution function of the purity @xcite , we first present the probability density function ( pdf ) of @xmath6 for @xmath8 and @xmath9 and arbitrary @xmath1 , and also discuss the forms of the right and left tails of this pdf for square @xmath10 systems in the limit @xmath31 . from these results ,
we derive exact expression for the moments @xmath5 of @xmath6 of arbitrary order for @xmath8 and @xmath38 and arbitrary @xmath1 , and analyze their asymptotic large-@xmath1 behavior .
next , taking advantage of the established relation between the cross - moments of the ft and wl ensembles , we find an exact representation of @xmath5 of arbitrary , not necessarily integer order @xmath40 in @xmath47 systems via the probability @xmath42 that the smallest eigenvalue in the gaussian unitary ensemble is larger than @xmath48 .
lastly , we discuss the asymptotic , large-@xmath0 behavior of these moments . in sec .
[ conc ] we conclude with a brief summary of our results . some technical results are then confined to the appendices .
a precise definition of our settings is as follows . consider a quantum state @xmath49 , which is bipartite into two smaller hilbert spaces @xmath50 and @xmath51 of dimensions @xmath0 and @xmath1 ( @xmath52 ) .
one example of this setting may be a spin set ( the subsystem @xmath53 ) in contact with a heat bath ( the subsystem @xmath54 ) .
@xmath55 and @xmath56 in are assumed to be two complete bases of @xmath50 and @xmath51 , respectively .
therefore , the expansion of @xmath57 on the direct product of these bases involves coefficients @xmath58 , which are the ( in general complex ) entries of a rectangular @xmath59 matrix @xmath60 . if @xmath60 is now promoted to a _ random _
matrix , the class of _ random _ states @xmath57 can be further refined by requiring that * @xmath57 must not be expressible as a direct product of two states belonging to the two subsystems @xmath53 and @xmath54 ( this condition ensures that @xmath57 is generically _ entangled _ ) , * the density matrix of the composite system is simply given by @xmath61 with the constraint @xmath62 = 1 $ ] , or equivalently , @xmath63 .
this condition ensures that @xmath57 is a _ pure _ state .
the density matrix @xmath64 for an entangled pure state @xmath57 of a bipartite system can be formally written as @xmath65 it is often convenient to consider the _ reduced _ density matrix @xmath66 $ ] as @xmath67 = \sum_{j=1}^m \br j^{b}| \rho | j^{b}\kt\ , \ ] ] whose role is to separate the contribution of the subsystem @xmath53 from the environment @xmath54 .
expectation values of quantum observables involving the subsystem @xmath53 alone can often be more easily computed invoking @xmath68 .
using now the expression in , one gets @xmath69 where @xmath70 are the entries of the @xmath47 covariance matrix @xmath71 , with @xmath72 denoting hermitian conjugation .
similarly , one might have obtained the reduced density matrix @xmath73 $ ] of the environment in terms of the @xmath74 matrix @xmath75 .
it is easy to prove that @xmath76 and @xmath77 share the same set of @xmath52 nonzero eigenvalues @xmath78 .
they are all real and positive , and are called _
schmidt eigenvalues_. the _ schmidt decomposition _ ( sd ) then takes the form @xmath79 where @xmath80 are normalized eigenvectors of @xmath81 .
this implies that the original composite state @xmath57 attains the form @xmath82 implies that @xmath83 .
each state @xmath84 is separable in the sd above .
however , their linear combination @xmath57 can not , in general , be written as a direct product @xmath85 of two states of the respective subsystems .
the state @xmath57 is therefore in general entangled , and the schmidt eigenvalues @xmath86 can be used to quantify the degree of entanglement ( see below for details ) .
a sensible way to introduce randomness in this system is to sample those entangled pure states with equal probability , ( i.e. , according to the uniform measure ) over the full hilbert space .
physically , this corresponds to assuming the minimal amount of _ a priori _ information about the quantum state under consideration .
mathematically , this implies that the coefficients @xmath87 in are uniformly distributed on the manifold @xmath88 - this condition is necessary to enforce normalization of @xmath57 .
therefore , the probability density function of the @xmath89 matrix @xmath60 with entries @xmath58 can be written as p(x)((x x^)-1 ) , which implies that @xmath60 is distributed according to a _ fixed - trace _ ensemble ( see @xcite for an excellent review ) . performing a singular value decomposition of @xmath60 and integrating out the eigenvectors ,
the jpdf of schmidt eigenvalues @xmath86 turns out to be given by @xcite p^(ft)()=z^-1_n , m ( _ i=1^n _ i -1 ) |()|^ _ i=1^n _ i^ , [ jpdf1 ] where the normalization constant @xmath90 is given explicitly by ( see , e.g. , ref .
@xcite ) [ norm ] z^-1_n , m= , with @xmath91 , @xmath92 , @xmath93 . in , @xmath94 is the dyson index and @xmath95 is the vandermonde determinant . in what follows averaging with respect to the distribution in
will be denoted by @xmath96 .
going beyond the linear statistics requires the knowledge of the @xmath3-point densities of the fixed - trace ensemble in , [ r ]
^(ft)_n([]_n)=_0^p^(ft)()_k= n+1^n d_k . in principle
, this analysis has been performed in the recent paper @xcite , which focused however on a particular scaling limit so that extracting @xmath34_n)$ ] from their general formulae is not that easy . on the other hand , we realize that the derivation of @xmath34_n)$ ] with arbitrary @xmath3 , @xmath0 and @xmath1 is rather straightforward and moreover , sheds some light on the physical meaning of the schmidt eigenvalues .
thus for the sake of completeness we present it here .
we also note that although @xmath94 in quantum context , such a derivation can be done for any value of @xmath36 in .
let @xmath97 and @xmath98_n)$ ] denote the jpdf and the normalized @xmath3-point densities ( with @xmath99 ) of the eigenvalues @xmath100 , @xmath101 , of the @xmath36-wl ensemble , respectively p^(wl)(y)= |(y)|^ _
i=1^n y_i^ e^ - y_i/2 , [ jpdf2 ] and ^(wl)_n([y]_n)=_0^p^(wl)(y ) _
k = n+1^n dy_k . averaging with respect to the distribution in
will be denoted in what follows by the symbol @xmath102 . introducing the laplace transform of a function @xmath103 @xmath104 the general relation between the @xmath3-level densities of the @xmath36-fixed - trace and the @xmath36wl ensembles
can be obtained using the following standard approach .
we define first two auxiliary functions p^(ft)(;t)=z^-1_n , m ( _ i=1^n _ i -t ) |()|^ _ i=1^n _ i^ , [ jpdf3 ] and [ rt ] ^(ft)_n([]_n;t)= _ 0^p^(ft)(;t ) _ k = n+1^n d_k , which are mere generalizations of the expressions in and for the case of a trace fixed to be equal to @xmath105 . note that @xmath106 is a normalized joint pdf only for @xmath107 .
taking now the laplace transform of @xmath108_n;t)$ ] , we have @xmath109_n;p)= \int_0^\infty \tilde{p}^{(ft)}({\bm\lambda};p ) \prod_{k = n+1}^n d\lambda_{k } \nonumber\\ & = z^{-1}_{n , m } \ , \int_0^\infty |\triangle({\bm \ld})|^{\beta } \prod_{k = n+1}^n d\lambda_{k } \ , \prod_{i=1}^{n } \ , \ld_i^\alpha\ ! \ , e^{- p \ , \lambda_i } \,.\end{aligned}\ ] ] further on , changing the integration variables @xmath110 , we formally rewrite the latter equation as @xmath111_n}{2 p};p\right)&= \frac{z^{-1}_{n , m } } { ( 2 p ) ^{\mu - n } } \int_0^\infty \prod_{k = n+1}^n d\lambda_{k } \ , |\triangle({\bm y})|^{\beta}\ , \nonumber\\ & \times \prod_{i=1}^{n } \ , y^\alpha\ ! \ , e^{- y_i/2 } \,,\end{aligned}\ ] ] from which we read off the following relation @xmath112_n;p ) \equiv \frac{2^n \ , \gamma(\mu)}{p^{\mu - n } } \ , { \rho}^{(wl)}_n(2 p [ \bm\lambda]_n ) \ .\ ] ] this relations holds for arbitrary @xmath36 , @xmath3 , @xmath0 and @xmath1 .
inversion of the laplace transform yields @xmath113_n;t ) = 2^n \ , \gamma(\mu ) \ , { \cal l}^{-1}_{t , p}\left(\frac{{\rho}^{(wl)}_n(2p[\bm\lambda]_n)}{p^{\mu - n}}\right ) \,,\ ] ] and the desired general relation between the @xmath3-point densities of the two ensembles follows setting @xmath114 @xmath115_n ) = 2^n \ , \gamma(\mu ) \ , { \cal l}^{-1}_{t=1,p}\left(\frac{{\rho}^{(wl)}_n(2p[\bm\lambda]_n)}{p^{\mu - n}}\right ) \,.\ ] ] hence , since @xmath98_n)$ ] are known from the general theory of orthogonal polynomials in terms of @xmath116 determinants of a kernel built out of laguerre polynomials , the @xmath3-level densities for fixed - trace ensembles for arbitrary @xmath3 are obtained by the inversion of the laplace transform with respect to @xmath117 , upon setting @xmath118 .
it may be instructive to re - derive the result in using a different approach , which sheds some light on the physical meaning of the schmidt eigenvalues .
we focus on the @xmath36-wl ensemble and introduce _ scaled variables _ ( see , e.g. , @xcite ) of the form @xmath119\,.\ ] ] note that such variables automatically obey the fixed - trace constraint , @xmath120 .
the physical significance of @xmath121 is evident : it measures an individual @xmath100 against the arithmetic mean eigenvalue in a given realization of the @xmath36-wl ensemble .
consequently , the one - point density of @xmath121 shows how heterogeneous the distribution of the eigenvalues @xmath100 is and how likely it is that @xmath100-s concentrate around their mean value .
note that the extreme value statistics of such random variables ( the largest , @xmath122 , and the smallest , @xmath123 , which is the reciprocal of the so called demmel condition number , see , e.g. , @xcite and more recent @xcite ) plays a key role in various scale independent hypothesis testing procedures , both in classical statistics as well as in signal processing .
classical examples ( see , e.g. , @xcite ) include testing for the presence of interactions in multi - way data and testing for equality of the population covariance to a scaled identity matrix .
modern signal processing applications include testing for the presence of signals in cognitive radio as well as non - parametric signal detection in array processing .
spectral densities of ordered @xmath124 have been determined long time ago @xcite , while analogous distributions for unordered , random scaled variables have been recently evaluated for the @xmath36-wl ensemble for small @xmath0 @xcite and the gaussian unitary ensemble for arbitrary @xmath0 @xcite .
we also note that such scaled variables have been also used to characterize the effective broadness of `` narrow '' distributions possessing moments of arbitrary order @xcite . using and taking into account the well - known fact that the distribution @xmath125 of the trace @xmath126 in the @xmath36-wl ensemble is the central @xmath127-distribution of the form @xmath128 one finds the jpdf @xmath129 of the variables @xmath130 ( see , e.g. , @xcite ) @xmath131
remarkably , ( but not counter - intuitively ) , the scaled variables @xmath130 appear to have exactly the same jpdf as the schmidt eigenvalues .
this means , in particular , that the distributions of the largest and the smallest scaled variables @xmath132 and @xmath133 coincide with the distribution of the largest and the smallest fixed - trace eigenvalue , and also implies that the @xmath3-point densities @xmath134_n)$ ] have the same functional form as the @xmath3-level densities @xmath135_n)$ ] .
next , we formally represent @xmath98_n)$ ] as @xmath136_n)&=\int f(t ) dt \int \psi_n([\bm\omega]_n ) \prod_{k = n+1}^n d\omega_{k } \ , \nonumber\\ & \times \delta\left(y_k - t \omega_k \right ) \,,\end{aligned}\ ] ] which gives , upon integration over the @xmath137-variables , @xmath138_n)=\int \frac{f(t)}{t^n } \ , \psi_n\left(\frac{[\bm y]_n}{t}\right ) dt\,.\ ] ] using the definition of @xmath125 in , changing the integration variable @xmath139 and also setting @xmath140 , we can cast the latter equation into the form @xmath141_n)= \frac{p^{\mu - n}}{2^n \gamma(\mu ) } \ , { \cal l}_{p , t}\left ( \psi_n\left(\frac{[\bm\omega]_n}{t}\right)\right ) \,,\ ] ] which yields straightforwardly @xmath142_n}{t}\right ) = 2^n \gamma(\mu ) \ , { \cal l}^{-1}_{t , p}\left ( \frac{{\rho}^{(wl)}_n(2p[\bm\omega]_n)}{p^{\mu - n}}\right ) \,.\ ] ] our previous result in follows directly from by setting @xmath114 .
further on , our aim is now to establish a relation between the cross - moments of the @xmath36-ft and @xmath36-wl ensembles . to this end
, we first note that , evidently , @xmath143 where the symbol @xmath144 on the right - hand - side denotes averaging with respect to the jpdf of the scaled variables in . further on , using the definition of the scaled variables in , we write @xmath145 next , taking advantage of the integral identity @xmath146 equation can be formally rewritten as @xmath147 changing the integration variables @xmath148 , we have , dropping the prime , @xmath149 from which we immediately obtain the following relation between the cross - moments of @xmath16-s and @xmath100-s : @xmath150 where @xmath151 and @xmath152 are arbitrary ( not necessarily integer ) numbers each greater than @xmath153 , conditioned by the constraint that their sum @xmath151 is strictly larger than zero .
this result is a direct consequence of and is valid for arbitrary @xmath36 and arbitrary @xmath154 .
while preparing this manuscript , we became aware that the same relation between the cross - moments has been recently presented in @xcite .
. entails a series of very useful identities between the powers of traces of the @xmath36-fixed - trace and the @xmath36-wl ensembles , respectively . multiplying both sides of by @xmath155 ( with all @xmath156 ) and performing summation over all positive integer @xmath157
, we get the following relation between the weighted traces of two ensembles @xmath158 where @xmath159 is a bessel function , or , equivalently , @xmath160 further on , setting @xmath161 , where @xmath162 are positive integers , multiplying both sides of by @xmath163 and performing summations over all @xmath162 , we find the following relation between the weighted squared traces @xmath164 where @xmath165 is the generalized hypergeometric series , and @xmath166 where @xmath167 is the tricomi s confluent hypergeometric function .
we note that both equations can be straightforwardly used for the derivation of the moments of purity and of its reciprocal value - the schmidt number , which we will demonstrate in sec .
[ schmidt ] . eventually , by taking advantage of our , ( or by simply using the multinomial theorem and our ) it is straightforward to show that the moments of the squared weighted traces of two ensembles obey @xmath168 where @xmath40 is an arbitrary positive integer and @xmath169 are arbitrary non - negative numbers .
more generally , for the exponential moments of the @xmath19-parametrized rnyi entropy , we have @xmath170 where @xmath19 is an arbitrary ( not necessarily integer ) positive number .
we emphasize that our equations to are valid for arbitrary @xmath36 and arbitrary @xmath154 .
in this section we present explicit results for the one and two - point densities of the ft wishart ensemble at @xmath171 .
we remark that the one - point density has been already determined for finite @xmath4 in earlier refs.@xcite also for @xmath172 . here
we present it ( for @xmath171 ) for the sake of completeness , and also to demonstrate that our general relation in yields an explicit expression for the spectral density in a most immediate manner , as compared to previous approaches , especially the one in @xcite , which relied on a quite complicated analytical approach based on the theory of the holonomic system of differential equations . before we proceed with exact calculations ,
the following simple observation is in order : note that in the limit @xmath173 , our implies that to leading order in @xmath174 and for arbitrary positive @xmath175 , one has @xmath176 this means , in turn , that the schmidt eigenvalues , as well as the scaled variables @xmath137 , have the same limiting distribution as the eigenvalues @xmath177 of the @xmath171 wl ensemble divided by @xmath178 ; that being , they obey the marenko - pastur distribution @xcite for @xmath179 ( with @xmath180 fixed ) of the form : @xmath181 where the boundaries of the support are given by @xmath182 and @xmath183 the asymptotic result in is depicted in fig.[fig1 ] together with the exact results for the spectral density .
note that the result in has been obtained earlier in several papers ( see , e.g. , @xcite ) . for the unitary fixed - trace ensemble in for @xmath184 and @xmath185 ( red ) , @xmath186 and @xmath187 ( green ) and @xmath188 and @xmath187 ( blue ) .
the dashed lines define the marenko - pastur - type distribution in corresponding to these values of @xmath1 and @xmath0.,width=302 ] now , the one - point density for the unitary ( @xmath171 ) wl ensemble at fixed @xmath4 can be written as @xmath189 where @xmath190 are numerical coefficients , which we define in explicit form in appendix [ appendix ] ( see ) .
inserting into , we get straightforwardly @xmath191 where for the @xmath171 ensemble @xmath192 .
note that this is precisely the result obtained previously in @xcite by using different approaches . in fig.[fig1 ] we plot @xmath193 in for several values of @xmath1 and @xmath0 .
the two - level density @xmath194 for the @xmath171 wishart - laguerre ensemble has the following form : @xmath195 where @xmath196 are numerical coefficients which are defined explicitly in appendix [ appendix ] ( see ) .
consequently , using , the two - level density @xmath197 of the fixed - trace ensemble can be represented in form of the inverse laplace transform of the following function : @xmath198 performing the inverse laplace transform and setting @xmath114 , we obtain the following exact result @xmath199 where @xmath200 is the heaviside step function . to the best of our knowledge , this is a new result .
the two - level density in is plotted in fig.[fig2 ] for two different values of @xmath1 and fixed @xmath187 and reveals a pronounced structuring and correlations between the eigenvalues .
finally , we present an explicit expression for the pair cross - moments of arbitrary order @xmath201 which holds for arbitrary ( not necessarily integer and positive ) @xmath202 and @xmath203 , as well as for arbitrary @xmath0 and @xmath1 .
the average von neumann entropy can be straightforwardly calculated using our to give @xmath204 which holds for any @xmath1and @xmath0 ( with @xmath205 ) .
it is straightforward to verify , e.g. , numerically , that the latter expression coincides with the one conjectured by page @xcite @xmath206 where @xmath207 is the digamma function .
this expression was subsequently proven in @xcite .
note that @xmath208 is a monotonically _ increasing _ function of both @xmath1 , ( with a fixed @xmath0 ) , and @xmath0 , ( at a fixed @xmath205 ) . in both cases , to leading order in @xmath0 , the average von neumann entropy @xmath209 , which is precisely the reason to argue that the system is close to a maximally entangled state .
we note here that there is , however , some subtlety concerning the degree of the entanglement when this issue is analyzed in terms of the entanglement entropy , which is a logarithmic function of @xmath0 .
we proceed to show in section [ schmidt ] that the situation is , as a matter of fact , more delicate . studying the behavior of the moments of the schmidt number @xmath210
, we realize that @xmath28 ( so that the complete entanglement is achieved ) only for the situations when a subsystem of some fixed size @xmath0 is coupled to a thermodynamically large bath with @xmath39 . on the contrary , when the system is partitioned in two subsystems of equal size @xmath0 , @xmath6 approaches the value @xmath211 only , when @xmath31 . consequently
, such square systems are far from being completely entangled .
next , the variance of the von neumann entropy is by definition @xmath212 to perform averaging , we may proceed in three equivalent but different ways : we may directly use the one and two - point densities in our and , apply our and use the usual replica trick , or taking advantage of our , express @xmath213 and @xmath214 via analogous moments of the unitary wl ensemble and perform averaging using the spectral and two - level densities of the unitary wl ensemble . here
we follow the first approach . using
our we readily find that the first term on the right - hand - side of is given by @xmath215 in a similar fashion , we find that the cross term in obeys @xmath216 equations , and define an exact expression for the variance of the von neumann entropy , which is valid for arbitrary @xmath0 and @xmath1 , and for any particular choice of @xmath0 and @xmath1 it can be readily evaluated using mathematica . on the other hand , it has quite a complicated structure so that its dependence on @xmath0 and @xmath1 can not be easily understood .
based on low @xmath4 evaluations , we realize eventually that the formula for the variance of the entropy can be cast into a much more compact exact form @xmath217 where @xmath218 is the trigamma function , which is defined for integer values of the argument as a truncated sum of the form @xmath219 . in fig.[fig4 ]
we plot the result in as a function of @xmath1 .
we observe that , contrary to the behavior of @xmath220 in and , the variance of the von neumann entropy is a _ decreasing _ function of @xmath1 , which signifies that in the asymptotic limit @xmath39 the distribution of the von neumann entropy tends to a delta - function so that @xmath21 becomes self - averaging . in fig .
[ fig4bis ] , we plot the full distribution of the von neumann entropy , obtained from numerical diagonalization of fixed - trace ensembles , and compare the numerical histogram with gaussian curves with mean and variance as in and , respectively . in , as a function of @xmath1 for @xmath8 ( squares ) , @xmath38 ( open circles ) , @xmath221 ( filled circles ) , @xmath222 ( triangles ) , @xmath223 ( crossed triangles ) and @xmath224 ( crosses).,width=302 ] from numerical diagonalization with @xmath225 ( red circles ) and @xmath226 ( blue triangles ) , compared with gaussian curves with mean and variance as in and , respectively.,width=302 ] finally , we turn to the asymptotic behavior of in the limits @xmath227 , @xmath39 with @xmath0 fixed , and the so - called double scaling limit when @xmath228 , @xmath229 and @xmath39 . for square systems with @xmath227 we readily find @xmath230 note that the leading asymptotic term in this expansion coincides with the result @xmath231 obtained in @xcite using a coulomb gas method .
next , in the limit @xmath39 with @xmath0 kept fixed and finite , we find the following exact asymptotic expansion @xmath232 \,.\end{aligned}\ ] ] lastly , in the double scaling limit we obtain @xmath233 note that in all three asymptotic expansions the variance decreases as the second inverse power of @xmath1 to leading order in @xmath1 .
in this section , we focus on the schmidt number @xmath6 , defined in eq . , which is a random variable supported on @xmath7 $ ] .
recall that it quantifies the number of effective degrees of freedom contributing to the entanglement , therefore @xmath234 corresponds to the maximally entangled state , while @xmath235 corresponds to the completely unentangled state .
our formalism allows us to completely characterize the statistical properties of @xmath6 . before proceeding
, it may be useful to remark that in some instances @xmath6 can be directly measured experimentally . in @xcite
, the schmidt number has been given an operational meaning using the connection between the schmidt decomposition in quantum mechanics and the coherent - mode decomposition in classical coherence theory , and measured for two - photon states entangled in a transverse mode structure .
for convenience of presentation , this rather long section is structured as follows . in the first subsection , capitalizing on known results on the behavior of the purity @xcite , we will present exact formulae for the pdf of @xmath6 with @xmath236 and arbitrary @xmath1 , and derive the corresponding exact expressions for the moments of @xmath6 of arbitrary order and determine the asymptotic , leading large-@xmath1 behavior of the variance of @xmath6 .
next , we will briefly discuss the tails of the distribution in square systems with @xmath10 in the large-@xmath0 limit , using the results of the seminal analysis of the distribution of purity in @xcite .
further on , still for square systems , we will derive exact representations for the moments of @xmath6 of arbitrary order expressing them via the probability @xmath237 that the smallest eigenvalue @xmath238 of a @xmath47 matrix belonging to the gaussian unitary ensemble is greater or equal to @xmath48 . in the final part
, we will discuss the asymptotic behavior of the moments of @xmath6 for square @xmath47 systems in the large-@xmath0 limit . for arbitrary @xmath0 and @xmath1 ,
the pdf of the purity can be formally written down as @xmath239 where @xmath90 is defined in with @xmath171 and @xmath240 .
the calculation of this pdf therefore amounts to integrating the squared vandermonde determinant , @xmath241 , over a domain formed by the part of the intersection of a hyperplane @xmath242 and a hypersphere @xmath243 , centered at the origin , and appearing within the hypercube @xmath244^n$ ] . in the general case , of course ,
a direct calculation of the integral in for finite @xmath4 is seemingly hopeless .
giraud @xcite has nonetheless managed to overcome this challenge not only for the simple case @xmath8 , analyzed earlier by scott and caves @xcite , but also for @xmath38 and @xmath187 , which required a much more involved analysis .
building on the results in @xcite , we can easily deduce that for @xmath37 the pdf @xmath245 of the schmidt number @xmath6 is given by @xmath246 for @xmath247 $ ] and is zero otherwise .
further on , for @xmath38 an analogous result is given by a more complicated formula @xmath248 where @xmath249 is the floor function , @xmath250 , @xmath251 is the kronecker delta symbol , @xmath252 for @xmath253 $ ] and @xmath254 for @xmath247 $ ] ( so that @xmath255 is piece - wise continuous ) . in fig .
[ fig5 ] we plot the distribution function in and . for @xmath37 , ( dashed lines ) , and @xmath255 for @xmath9 , ( solid lines ) .
the curves from the left to the right correspond to @xmath256 and @xmath257 . ,
width=302 ] the behavior of the pdf of the purity for square systems with @xmath10 has been analyzed in the large-@xmath0 limit in @xcite using a coulomb gas method .
it was there shown that the pdf exhibits three different regimes , associated with a different behavior of the equilibrium density of eigenvalues : a left tail , a central gaussian peak , and a right tail , similarly to what we observe in fig .
[ fig5 ] . translating these results in the language of schmidt number , the pdf @xmath258 in the vicinity of the left edge of the support , i.e.
, for @xmath6 close to @xmath259 , should tend to zero as @xmath260 in the opposite limit , i.e. when @xmath6 is close to @xmath0 so that the system tends to a maximally entangled state , @xmath261 these two asymptotic regimes are separated by a gaussian peak near @xmath211 . for @xmath37 the moments of @xmath6 of arbitrary ( not necessarily integer ) order @xmath40 can be readily found from and read @xmath263 where @xmath264 is the gauss hypergeometric function .
hence , for sufficiently large @xmath1 , @xmath265 is conveniently represented by ( see in the appendix [ appendix3 ] ) @xmath266 so that , to leading order in @xmath1 , the variance of @xmath267 is given by @xmath268 for @xmath9 , the calculations are rather involved and we relegate them to the appendix [ appendix3 ] , where we derive an explicit , albeit rather cumbersome exact result for @xmath269 ( see in the appendix [ appendix3 ] ) .
the asymptotic large-@xmath1 behavior turns out to have a rather simple form and for arbitrary , not necessarily integer @xmath40 , we have @xmath270 so that the variance of @xmath271 , to leading order in @xmath1 , obeys @xmath272 we observe that the leading terms in and are @xmath273 ( for @xmath37 ) and @xmath274 ( for @xmath9 ) .
it seems natural to conjecture that for general @xmath0 the leading term should be @xmath275 . while we are unable to prove that @xmath276 such an assumption seems to be quite plausible on physical grounds and signifies that in situations in which a small subsystem of size @xmath0 is attached to a much larger subsystem of size @xmath1 , a complete entanglement is achieved when @xmath39 .
this large-@xmath1 behavior is corroborated by numerical simulations , shown in fig .
[ figm ] . as a function of @xmath1 , for @xmath277 ( black circles ) and @xmath278 ( blue stars ) .
the convergence to the theoretical value @xmath259 is quite convincing .
, width=302 ] to close this subsection , we take advantage of the works by scott and caves @xcite and giraud @xcite , who calculated exactly the first few cumulants of the purity @xmath26 , and extract from these results the asymptotic , large-@xmath1 behavior of the moments of @xmath26 with fixed @xmath0 .
we find that , explicitly , @xmath279
@xmath280 @xmath281 and @xmath282 while the variance of @xmath26 , which is given exactly by @xcite @xmath283 obeys , to leading order in @xmath1 , @xmath284 we note that as @xmath39 and at a fixed @xmath0 , the purity tends to the value @xmath285 , i.e. , attains the lower limit of its support .
together with our , it implies that systems with fixed @xmath0 and @xmath39 become completely entangled .
having discussed the properties of @xmath286 for fixed @xmath0 and varying @xmath1 , we turn to the analysis of the moments of @xmath6 in the special quadratic case @xmath287 ( so that @xmath288 in ) . the general case
when @xmath2 can also be considered within the framework we present below but will require much lengthier calculations .
we make use of our , set @xmath94 , @xmath289 and put @xmath290 with @xmath291 . recalling next the definition of @xmath6 , we rewrite as @xmath292 using ,
the right - hand - side of can be written explicitly as @xmath293 where @xmath294 is defined in with @xmath36 set equal to @xmath295 and @xmath10 . making a linear shift of the integration variables @xmath296 , we formally rewrite the latter expression as @xmath297 one notices next that the integral in the second line in , up to the normalisation factor @xmath298 , @xmath299 is equal to the probability @xmath300 that the smallest eigenvalue @xmath238 in a @xmath47 gaussian unitary ensemble is greater or equal to @xmath301 @xcite . consequently , combining and , we establish the following intriguing representation @xmath302 where @xmath303 is the barnes g - function .
this result holds for arbitrary @xmath0 and arbitrary non - negative @xmath304 and will be used in what follows for the derivation of explicit and exact representations of @xmath286 .
moments of order @xmath306 can be obtained as follows : multiplying both sides of by @xmath307 , where @xmath308 , and integrating from zero to infinity , we get @xmath309 which defines the positive moments of the schmidt number of order @xmath306 . note that this constraint on the order of moments stems from the restriction on the convergence of the integral in
; for @xmath310 the integral in the latter equation becomes divergent on the upper integration limit so that we have to resort to a different strategy . in appendix [ appendixc ]
we present an alternative derivation of the result in using a different line of thought , closer to the derivation of the relation between the @xmath3-point densities of the ft and the wl ensembles .
the moment of order @xmath311 can be readily obtained from by noticing that @xmath312 so that , for arbitrary @xmath0 , we have @xmath313 this expression relates , at a first glance rather surprisingly , the moment of @xmath6 of order @xmath45 in the ft ensemble to the probability of the highly atypical event where all the eigenvalues of a @xmath47 matrix belonging to the gue are positive . the moments of order @xmath44 or @xmath315 , @xmath316 being an integer , can be accessed in the following way . from , recall the taylor series expansion of the tricomi s confluent hypergeometric function and express @xmath300 with non - negative @xmath304 in terms of the moments of @xmath6 , which gives @xmath317 the latter equation states that @xmath300 can be interpreted as the generating function of moments of @xmath6 of order higher than @xmath45 .
in other words , is the taylor series expansion of @xmath318 in powers of @xmath319 and the coefficients in this expansion are just the derivatives of the latter function at @xmath320 : @xmath321 and @xmath322 therefore , the moments of the schmidt number @xmath6 of order greater than @xmath45 probe the behavior of @xmath300 right in the middle of the wigner sea . since @xmath238 is typically @xmath323 , i.e. , @xmath324 , we conclude that the behavior of the moments of @xmath6 of such a high order is supported by atypical , rare events .
lastly , we consider the negative moments of @xmath6 , i.e. , positive moments of the purity @xmath26 .
we recall that formally exact expressions for @xmath326 have already been computed by giraud ( see @xcite and some corrections in @xcite ) .
our goal here is to provide an alternative derivation , relating @xmath326 to the behavior of @xmath300 in the limit @xmath15 .
this derivation will also allow us to establish an exact asymptotic expansion for @xmath300 for finite @xmath0 .
setting in all @xmath327 and @xmath3 equal to @xmath0 , we have that for any integer @xmath328 , @xmath329 further on , using the relation @xmath330 we find , after some straightforward manipulations , the following expression @xmath331 the latter expression shows that the moments of the purity , or the inverse moments of @xmath6 , are dominated by the right tail of the probability @xmath300 corresponding to the limit @xmath15 . as a certain by - product of the above considerations , which will be also useful for our further analysis
, we obtain from the following _ exact _ asymptotic representation of the probability @xmath300 in the limit @xmath15 @xmath332 which is valid for arbitrary @xmath0 .
note that , in contrast to the result in , which shows that the coefficients in the expansion of @xmath300 right in the middle of the wigner sea are related to the moments of the schmidt number of order @xmath333 , the coefficients in the taylor series for @xmath15 are proportional to the inverse moments of @xmath6 , i.e. , the positive moments of the purity @xmath26
. therefore , @xmath300 in the limit @xmath15 can be thought of as the moment - generating function of the purity .
next , capitalizing on the exact results for the moments of purity calculated by giraud @xcite , we rewrite the exact asymptotic large-@xmath304 expansion in in a closed form : @xmath334 where @xmath335 which is a finite sum over partitions of @xmath3 into @xmath0 numbers greater or equal to 0 .
these partitions can be easily generated for any @xmath3 by some suitable algorithm .
in particular , the first four terms in the expansion in ( or , equivalently , in ) are given explicitly by @xmath336 to the best of our knowledge , the result in is new . in this final section
we focus on the asymptotic large-@xmath0 behavior of the moments of @xmath6 of order @xmath306 and of the purity @xmath26 for square , @xmath47 systems , a question that seemingly has not been addressed as yet .
the large-@xmath0 behavior of the moments of @xmath26 will be simply extracted from the available exact expressions for the first few cumulants of the purity@xcite . on the other hand , to deduce the large-@xmath0 behavior of the moments of @xmath6 we will resort to the classical papers by dean and majumdar @xcite , who have shown that for sufficiently large
@xmath0 the probability @xmath300 has the form @xmath337 where the symbol @xmath338 means equivalence on a logarithmic scale , and the large deviation function @xmath339 in is given explicitly by @xcite : @xmath340 note that , as evidenced by numerical simulations performed in @xcite , the result in and is reasonably accurate already for quite modest values of @xmath0 .
consequently , in this subsection we will posit that the moments of @xmath6 of order less than @xmath45 , defined by our , can be approximated with logarithmic accuracy by @xmath341 where the constant @xmath342 will be defined below . before we proceed with the analysis of the expression in
, we find it necessary to emphasize several points on the possible limitations of the result in and , and consequently , on the errors these limitations may incur in determining the asymptotical behavior of the moments of the schmidt number for square systems .
\(i ) note that due to the omitted @xmath343 terms in , the limiting value @xmath344 in is not equal to @xmath259 , as it should .
we introduce _ ad hoc _ in a normalization constant @xmath342 , which will _ enforce _ the condition @xmath345 . clearly , this normalization constant should not dependent on @xmath0 stronger then @xmath346 .
we present the calculation of this constant in appendix [ appendixd ] and show that @xmath342 is given explicitly by @xmath347 we observe that , indeed , this @xmath342 depends very weakly on @xmath0 ( see the appendix [ appendixd ] ) .
\(ii ) let us examine next the large-@xmath304 behavior of the large deviation form @xmath237 in and and compare it with our exact asymptotic expansion in .
the large-@xmath304 asymptotic of the expression in and can be readily determined and reads @xmath348 further on , the @xmath0-dependent numerical factor in the first line in can be written down , for sufficiently large @xmath0 , as @xmath349 where @xmath350 is the glaisher constant and @xmath351 is the base of the natural logarithm .
comparing next both expansions , we first observe that the leading terms coincide , once we discard the correction terms in the second line in .
in fact , the constant @xmath342 in , chosen to enforce the condition @xmath352 , is actually equal to these discarded correction terms .
next , the subdominant terms in both expansion are series in powers of @xmath353 ( multiplied by @xmath354 ) and the coefficients in this series are polynomials of @xmath0 . inspecting the expansion coefficients in and , we note that the first two terms coincide exactly , while the third and the fourth ones are only slightly different , which means that the omitted @xmath343 terms start to contribute only at this level . concluding this discussion
, we may expect that @xmath237 in defines correctly the leading and the first sub - leading terms in the moments of purity . on the other hand ,
since the moments of @xmath6 are defined as an integral over @xmath304 and we do not know how accurately @xmath237 in approximates the true @xmath237 for small and moderate @xmath304 , we can not expect the same level of accuracy .
in appendix [ appendixd ] we construct an exact asymptotic expansion of the integral in .
we realize , indeed , that the approximation based on allows us to deduce only the exact leading-@xmath0 behavior of @xmath286 , while already the first subleading term is incorrect .
the leading behavior , for @xmath355 and @xmath31 , has the following form @xmath356 this result , on the one hand , is somewhat trivial because it corresponds to the ultimate regime where the pdf of @xmath6 attains the form of the delta function . on the other hand
, it nonetheless provides an important , physically meaningful information : namely , it states that for @xmath31 the schmidt number @xmath357 and hence , in contrast to the systems with fixed @xmath0 and @xmath39 , the entanglement in square systems is far from being complete . in fig .
[ figsquare ] , we provide results of numerical diagonalization for the ratio @xmath358 as a function of @xmath359 , which show a convincing convergence to @xmath259 as predicted theoretically . as a function of @xmath359 .
numerical simulations for @xmath360 ( purple dots ) and @xmath361 ( blue stars ) confirm that the ratio converges to @xmath259 , as predicted theoretically .
, width=302 ] finally , taking advantage once again of the general results due to scott and caves @xcite and giraud @xcite , for square @xmath10 systems we present the large-@xmath0 asymptotics of the moments of the reciprocal of the schmidt number - the purity @xmath26 .
we have that the first four moments obey @xmath362 @xmath363 @xmath364 and @xmath365 from @xcite , we have that the variance of the purity for square systems follows in the asymptotic limit @xmath31 @xmath366 this means that the variance of the purity in square systems decays , to leading order , as an inverse fourth power of @xmath0 , i.e. much faster than the variance of the von neumann entropy ( see our ) .
this implies , in turn , that the distribution of the purity is much more narrow than that of the von neumann entropy .
to summarize , we have analyzed the nonlinear statistics of schmidt eigenvalues for entangled random pure states at finite @xmath4 - the dimensions of the hilbert space partitions .
we have established a general relation between the @xmath3-point densities and the cross - moments of the @xmath36-ft ensemble and the analogous quantities of the @xmath36-wl ensembles .
building on these results , we presented explicit , closed - form expressions for the two - point densities and also an exact expression for the variance of the von neumann entropy , valid for any @xmath0 and @xmath1 .
further on , we derived a wealth of results for the schmidt number @xmath6 . capitalizing on the known results for the distribution function of the purity @xcite , we calculated the probability density function of @xmath6 for @xmath8 and @xmath9 and arbitrary @xmath1 , and also discussed the forms of the right and left tails of this distribution for square @xmath10 systems in the limit @xmath31 . from these results , we derived exact expression for the moments @xmath5 of @xmath6 of arbitrary order for @xmath8 and @xmath38 and arbitrary @xmath1 , and analyzed their asymptotic large-@xmath1 behavior .
next , taking advantage of the established relation between the cross - moments of the ft and wl ensembles , we found an exact representation of @xmath5 of arbitrary , not necessarily integer order @xmath40 in @xmath47 systems by spotting a previously unnoticed connection with the statistics of the smallest eigenvalue of gaussian unitary matrices .
lastly , we discussed the asymptotic , large-@xmath0 behavior of these moments .
our results have been corroborated via numerical simulations whenever possible , with excellent agreement . as a by - product of our analysis
, we also established an exact asymptotic expansion of the probability @xmath42 that the smallest eigenvalue in the gaussian unitary ensemble is larger than @xmath48 , in the limit @xmath15 for fixed @xmath0 , by identifying the coefficients in this expansion via the moments of the purity in the fixed - trace ensemble .
the authors acknowledge helpful discussions with d.s .
dean , o. giraud and s. n. majumdar .
pv acknowledges the stimulating research environment provided by the epsrc centre for doctoral training in cross - disciplinary approaches to non - equilibrium systems ( canes , ep / l015854/1 ) .
go acknowledges a partial support from the onrg grant n62909 - 15 - 1-c076 and wishes to thank for warm hospitality and financial support the institute for mathematical sciences of the national university of singapore , where some part of this work has been done
. 99 a. einstein , b. podolsky and n. rosen , phys . rev . * 47 * , 777 ( 1935 ) .
nielsen and i.l .
chuang , _ quantum computation and quantum information _
( cambridge univ .
press , cambridge , 2000 ) .
r. horodecki , p. horodecki , m. horodecki and k. horodecki , rev .
phys . * 81 * , 865 ( 2007 ) . h. di lorenzo pires , c. h. monken and m.p .
van exter , phys .
a * 80 * , 022307 ( 2009 ) .
e. lubkin , j. math .
* 19 * , 1028 ( 1978 ) ; s. lloyd and h. pagels , ann .
( n.y . ) * 188 * , 186 ( 1988 ) .
page , phys .
lett . * 71 * , 1291 ( 1993 ) . s.k .
foong and s. kanno , phys .
lett * 72 * , 1148 ( 1994 ) ; j. snchez - ruiz , phys .
e * 52 * , 5653 ( 1995 ) ; s. sen , phys .
lett . * 77 * , 1 ( 1996 ) .
o. bohigas , m.j .
giannoni and c. schmit , phys .
lett . * 52 * , 1 ( 1984 ) .
bandyopadhyay and a. lakshminarayan , phys .
lett . * 89 * , 060402 ( 2002 ) and references therein .
k. yczkowski , k.a .
penson , i. nechita and b. collins , j. math . phys . * 52 * , 062201 ( 2011 ) .
h. kubotani , s. adachi and m. toda , phys .
* 100 * , 240501 ( 2008 ) ; s. adachi , m. toda and h. kubotani , annals of physics * 324 * , 2278 ( 2009 ) .
krishnaiah and f.j .
schuurmann , j. multivariate anal . * 4 * , 265 ( 1974 ) .
l. wei , o. tirkkonen , p. dharmawansa and m.r .
mckay , ieee international conference on communications ( icc ) 2422 - 2426 , doi : 10.1109/icc.2012.6364410 ( 2012 ) .
b. nadler , j. multivariate anal .
* 102 * , 363 ( 2011 ) .
p. vivo , j. stat . mech .
* p01022 * ( 2011 ) .
cunden , p. facchi , g. florio and s. pascazio .
j. phys . plus
* 128 * , 48 ( 2013 ) ; o.c.o .
dahlsten , c. lupo , s. mancini and a. serafini , j. phys . a : math
. theor . * 47 * , 363001 ( 2014 ) .
p. facchi , g. florio , u. marzolino , g. parisi and s. pascazio , j. phys .
a : math . theor . * 42 * , 055304 ( 2009 ) ; _ ibid . _ * 43 * , 225303 ( 2010 ) ; new j. phys .
* 12 * , 025015 ( 2010 ) .
a. de pasquale , p. facchi , g. parisi , s. pascazio and a. scardicchio , phys .
a * 81 * , 052324 ( 2010 ) ; a. de pasquale et al .
, j. phys .
a : math . theor . * 45 * , 015308 ( 2012 ) .
cunden , p. facchi and g. florio , j. phys .
* 46 * , 315306 ( 2013 ) . k. yczkowski and h - j .
sommers , j. phys .
a : math . gen . * 34 * , 7111 ( 2001 ) .
scott and c.m .
caves , j. phys .
a : math . gen . * 36 * , 9553 ( 2003 ) .
o. giraud , j. phys .
a. : math . theor . *
40 * , 2793 ( 2007 ) .
o. giraud , j. phys .
a. : math . theor .
* 40 * , 1053 ( 2007 ) .
m. znidaric , j. phys . a : math .
theor . * 40 * , f105 ( 2007 ) . s.n .
majumdar , o. bohigas and a. lakshminarayan , j. stat . phys . *
131 * , 33 ( 2008 ) .
p. facchi , g. florio , g. parisi and s. pascazio , phys .
a * 77 * , 060304 ( r ) ( 2008 ) .
p. facchi et al . , phys .
* 101 * , 050502 ( 2008 ) . c. nadal , s.n .
majumdar and m. vergassola , phys .
lett . * 104 * , 110501 ( 2010 ) . c. nadal , s.n .
majumdar and m. vergassola , j. stat . phys . *
142 * , 403 ( 2011 ) .
sommers and k. yczkowski , j. phys .
a : math . theor . * 37 * , 8457 ( 2004 ) .
p. vivo , j. phys .
a : math . theor . * 43 * , 405206 ( 2010 ) .
s. kumar and a. pandey , j. phys . a : math
* 44 * , 445301 ( 2011 ) .
mehta , _ random matrices _ ,
( amsterdam : elsevier / academic press , 2004 ) .
tracy and h. widom , comm .
phys . * 163 * , 33 ( 1994 ) d.s . dean and s.n .
majumdar , phys .
lett . * 97 * , 160201 ( 2006 ) ; phys .
e * 77 * , 41108 ( 2008 ) .
majumdar , _ extreme eigenvalues of wishart matrices : application to entangled bipartite system _ , chapter in _ the oxford handbook of random matrix theory _ , edited by g. akemann , j. baik , and p. di francesco ( oxford university press , 2011 ) .
g. akemann and p. vivo , j. stat .
p05020 ( 2011 ) .
g. livan and p. vivo , acta phys .
b * 42 * , 1081 ( 2011 ) .
liu and d .- s .
zhou , int .
res . notices * 2011 * , 725 ( 2011 ) .
a. edelman , math .
58 * , 185 ( 1992 ) .
l. wei , m.r .
mckay and o. tirkkonen , ieee commun .
letters * 15 * , 175 ( 2011 ) .
p. dharmawansa , m.r .
mckay and y. cheng , siam j. matrix anal . appl . * 34 * , 257 ( 2013 ) . c. mejia - monasterio , g. oshanin and g. schehr , phys .
e * 84 * , 035203 ( 2011 ) .
pato and g. oshanin , j. phys .
a : math . theor . * 46 * , 115002 ( 2013 ) . c. mejia - monasterio , g. oshanin and g. schehr , j. stat .
p06022 ( 2011 ) .
mattos , c. mejia - monasterio , r. metzler and g. oshanin , phys .
e * 86 * , 031143 ( 2012 ) .
m.r . atkin and s. zohren ,
a * 92 * , 012331 ( 2015 ) .
marenko and l.a .
pastur , math .
ussr - sb . * 1 * , 457 ( 1967 ) .
the spectral density @xmath369 of the @xmath171 wl ensemble reads ( see , e.g. , @xcite ) @xmath370 \ , , \nonumber\\\end{aligned}\ ] ] where @xmath371 is the generalized laguerre polynomial .
further on , we use the following representation of the product of two generalized laguerre polynomials @xmath372 where @xmath373 is the binomial coefficient such that @xmath374 for @xmath375 and @xmath376 , and zero otherwise . inserting into , and collecting terms with the same power of @xmath177 , we find the following explicit representation of the coefficients @xmath190 in @xmath377 \,.\end{aligned}\ ] ] further on , we present the derivation of the coefficients @xmath368 . the normalized two - point density of the @xmath171 wl ensemble is given explicitly by @xmath378 the first term in square brackets , which is a product of two one - point densities , produces a trivial contribution to @xmath196 - the product of two corresponding coefficients @xmath190 .
we therefore focus on the second term .
the sum of products of laguerre polynomials entering can be written down as @xmath379 where the numerical coefficients @xmath380 are given by @xmath381 consequently , the squared sum on the left - hand - side of admits the following expansion @xmath382 where @xmath383 consequently , we arrive at the following explicit result for the coefficients @xmath196 @xmath384
for @xmath38 and arbitrary @xmath385 , it is convenient to calculate the moments of @xmath6 directly from their formal definition : @xmath386 one of the integrals , say , over @xmath387 , can be simply performed using the delta function .
changing then the integration variable @xmath388 as @xmath389 we cast the expression in into the form @xmath390 the integrals in are coupled via the expression in the denominator of the kernel and in order to factorise them we use the following expansion @xmath391 plugging the latter expansion in and performing the integration , we obtain the following result : @xmath392 where @xmath393 [ r2 ] & r^(2)_n , k = + & _ 2f_1(-k , m ; m+ ; ) , and @xmath394
lastly , we note that the hypergeometric functions entering to can be simply expressed via the jacobi polynomials @xmath395 with the argument @xmath396 :
@xmath397 where @xmath398 therefore , our with to defines an exact result for the moments @xmath399 of the schmidt number for @xmath9 and arbitrary @xmath1 and ( not necessarily integer ) @xmath40 in form of an infinite series . in principle , summation over @xmath400 and @xmath3 can be performed giving an explicit result in terms of a ( rather cumbersome ) combination of generalized hypergeometric functions . for any fixed @xmath40 and @xmath1
, this series can be straightforwardly computed using mathematica .
the expression in is , however , not very useful since it has a too complicated structure and does not permit to easily observe the @xmath1- and @xmath40-dependence of the moments of @xmath6 . to this purpose ,
we focus next on the asymptotic , large-@xmath1 behavior of the expression in . expanding
the ratios of the gamma functions entering to in taylor series in inverse powers of @xmath1 , and taking advantage of the following asymptotic expansion of the hypergeometric functions for large values of the parameters , @xmath401 \nonumber\ , \\\end{aligned}\ ] ] we have @xmath402 multiplying the last line in by @xmath403 and summing over @xmath3 , we arrive at the result in .
in this appendix , we present an alternative derivation of the expression defining moments of the schmidt number @xmath6 of order @xmath306 . our approach here is , in fact , essentially the same as the one we developed for the derivation of the @xmath3-point densities of the ft ensemble . first of all , we introduce an auxiliary function @xmath404 ( with @xmath405 defined in with @xmath10 , @xmath171 and hence , @xmath406 ) which describes for @xmath114 the pdf of the purity @xmath26 on @xmath47 ft ensembles with @xmath94 , see . taking
the double laplace transform of the expression in with respect to both @xmath407 and @xmath26 , we have @xmath408 changing the integration variables @xmath409 , we can cast the latter expression into the form @xmath410 which can be immediately written in the following more appealing form @xmath411 here , @xmath412 is the inverse of the normalization constant of the @xmath47 gaussian unitary ensemble , defined explicitly in , and @xmath413 is as before the probability that the smallest eigenvalue in this ensemble is greater or equal to @xmath414 .
consequently , the pdf of the purity can be formally written down as @xmath415 note now that the moments of the schmidt number @xmath6 can be straightforwardly expressed in terms of the function @xmath416 as @xmath417 changing the integration variable @xmath418 , we formally rewrite the latter expression as @xmath419 which gives upon the inversion of the laplace transform with respect to @xmath117 , ( for @xmath420 ) , our result in .
here we detail the evaluation of the integral in , which defines the moments of @xmath6 of arbitrary order @xmath306 under the assumption that the probability @xmath42 can be well approximated by its large deviation form calculated in @xcite .
our aim is to present the first three terms in the asymptotic large-@xmath0 expansion stemming out of this approximation .
we start with an analysis of the behavior of the integrand in .
the integrand , i.e. , @xmath421 is a bell - shaped function of @xmath304 , which vanishes for @xmath422 and @xmath15 , and has a maximum at @xmath423 the position of the maximum @xmath424 is a monotonically increasing function of @xmath0 , for a fixed @xmath40 , and is a monotonically decreasing function of @xmath40 , at a fixed @xmath0 , for @xmath425 $ ] ; @xmath426 for @xmath31 .
given that there is a large parameter @xmath427 in the exponential , it is tempting to resort to the saddle - point approximation .
this approach , however , yields a very poor result for the integral .
the reason is that , even though @xmath428 can be very well approximated by a gaussian in the vicinity of @xmath429 , this gaussian is flanked on both sides by power - law tails : @xmath430 when @xmath43 , since @xmath431 in this limit , and using the expansion in , @xmath432 as @xmath15 . as a matter of fact , these power - law tails provide the dominant contributions to the integral in ; namely , the left tail dominates the behavior of the moments of @xmath6 when @xmath40 is close to @xmath45 , and the right tail does the same for sufficiently small values of @xmath40
. hence , we have to resort here to a different approach .
lastly , we note that there is some subtlety in the behavior of the integral in ( and also in the exact expression in ) in the limit @xmath433 , since in this case @xmath434 , ( see ) .
this implies that the integrals are formally logarithmically divergent on the upper limit of integration . on the other hand , the expressions in and contain a factor @xmath435 , which vanishes as @xmath433 . as a matter of fact , these two conflicting factors compensate each other and @xmath344 exists , when one first performs the integrals at a fixed @xmath436 and only afterward takes the limit @xmath433
now , we turn to the evaluation of the integral in . changing the integration variable as @xmath437
the expression for the moments of @xmath6 in can be conveniently cast into the form @xmath438 further on , using the well - known expression for the generating function of the hermite polynomials @xmath439 , we represent the exponential in the latter equation as a sum over hermite polynomials , @xmath440 where @xmath441 . inserting this expression into and performing the integral
, we arrive at the following representation of the moments of @xmath6 in the form of an infinite series @xmath442 which is completely equivalent to the expression in .
we are now in the position to determine the @xmath0-dependent constant @xmath342 .
taking in the limit @xmath433 , and requiring that @xmath443 , we find the expression in .
the normalization constant @xmath342 as a function of @xmath0 is depicted in fig .
one observes that , indeed , @xmath342 is a very slowly varying function of @xmath0 . as a matter of fact
, the asymptotic behavior of @xmath342 is well described by a slow power law of the form @xmath444 where @xmath445 is the glaisher s constant .
one infers from fig .
[ cnn ] that this asymptotic form sets in starting from very moderate values of @xmath0 .
moreover , we note that @xmath342 defines precisely the terms in the second line in . next , we take advantage of the explicit representation of the hermite polynomials @xmath446 where @xmath447 is the floor function .
note that the series in are arranged in descending order with respect to powers of the argument , so that the term @xmath448 corresponds to the highest power of @xmath449 . inserting into ,
taking into account and performing summation over @xmath316 , we have @xmath450 \,,\end{aligned}\ ] ] where @xmath451 is the kummer s confluent hypergeometric function .
note that in a similar fashion we can obtain an explicit expression for the average logarithm of @xmath6 , which describes the `` typical '' behavior of the schmidt number .
using our , we find @xmath452 where @xmath453 is the digamma function , defined in the text after .
we turn to the analysis of the asymptotic behavior of the result in in the limit @xmath31 .
we note first that the kummer s functions entering the series have the form @xmath454 , in which @xmath455 is independent of @xmath0 , while the parameter @xmath456 and the argument @xmath449 are both proportional to @xmath427 , and hence , tend to infinity as @xmath31 . setting @xmath457 , where in the case at hand @xmath458 is a bounded function such that
@xmath459 , we have that in the limit @xmath460 the kummer s functions obey @xmath461 \,,\nonumber\\\end{aligned}\ ] ] which implies that all @xmath454 in approach constant values as @xmath31 , and the dominant @xmath0-dependence of each term in the expansion in will come from the ratio of the gamma functions .
noticing next that as @xmath31 , @xmath462 and @xmath463 we can conclude that for large @xmath0 the series in represents an expansion in the inverse powers of @xmath427 .
moreover , the dominant contribution to the large-@xmath0 behavior of @xmath286 will be provided by the zeroth term , while the terms of higher order will contribute only to the subdominant behavior .
more precisely , we have that the zeroth term ( @xmath448 ) of the series is explicitly given by @xmath464 \,,\end{aligned}\ ] ] the first one ( @xmath465 ) obeys @xmath466 \,,\end{aligned}\ ] ] while the second one ( @xmath467 ) ( contributing only to the order @xmath468 ) is explicitly given by @xmath469 \,.\end{aligned}\ ] ] the terms with higher values of @xmath400 contribute to the order @xmath470 and higher , and hence can be safely neglected , given that we are interested in the large-@xmath0 behavior of the moments of @xmath6 to the order @xmath471 at most .
summing the contributions given in , and , we arrive at the following result @xmath472 \,.\end{aligned}\ ] ] we notice immediately that already the first subdominant term in this expansion is not correct , since it has a spurious quadratic dependence on @xmath40 . due to this dependence , predicts that the variance @xmath473 as @xmath31 , which is evidently incorrect .
therefore , albeit may serve as useful approximation ( having the same level of accuracy as an approximate form of @xmath237 in @xcite ) , the only reliable term in it is the first leading term , . | we analyze the properties of entangled random pure states of a quantum system partitioned into two smaller subsystems of dimensions @xmath0 and @xmath1 . framing the problem in terms of random matrices with a fixed - trace constraint , we establish , for arbitrary @xmath2 , a general relation between the @xmath3-point densities and the cross - moments of the eigenvalues of the reduced density matrix , i.e. the so - called schmidt eigenvalues , and the analogous functionals of the eigenvalues of the wishart - laguerre ensemble of the random matrix theory .
this allows us to derive explicit expressions for two - level densities , and also an exact expression for the variance of von neumann entropy at finite @xmath4 .
then we focus on the moments @xmath5 of the schmidt number @xmath6 , the reciprocal of the purity .
this is a random variable supported on @xmath7 $ ] , which quantifies the number of degrees of freedom effectively contributing to the entanglement .
we derive a wealth of analytical results for @xmath5 for @xmath8 and @xmath9 and arbitrary @xmath1 , and also for square @xmath10 systems by spotting for the latter a connection with the probability @xmath11 that the smallest eigenvalue @xmath12 of a @xmath13 matrix belonging to the gaussian unitary ensemble is larger than @xmath14 . as a byproduct ,
we present an exact asymptotic expansion for @xmath11 for finite @xmath0 as @xmath15 .
our results are corroborated by numerical simulations whenever possible , with excellent agreement . |
this work is supported by the royal society , euryi , mext - ct-2006 - 039047 , korea research foundation grant ( krf-2005 - 215-c00040 ) , the basic research program ( 2009 - 0092809 ) through the national research foundation of korea , and the national research foundation of singapore .
n. reyren , s. thiel , a. d. caviglia , l. fitting kourkoutis , g. hammerl , c. richter , c. w. schneider , t. kopp , a .- s .
ruetschi , d. jaccard , m. gabay , d. a. muller , j .-
m . triscone and j. mannhart , _ science _ * 317 * , 1196 ( 2007 ) h. yamada , p. h. xiang and a. sawa , _ phys . rev .
b _ , * 81 * , 014410 ( 2010 ) t. koida , m. lippmaa , t. fukumura , k. itaka , y. matsumoto , m. kawasaki and h. koinuma , _ phys
b _ , * 66 * , 144418 ( 2002 ) s. dong , r. yu , s. yunoki , g. alvarez , j .-
liu and e. dagotto , _ phys .
b _ * 78 * , 201102(r ) ( 2008 ) s. j. may , a. b. shah , s. g. e. te velthuis , m. r. fitzsimmons , j. m. zuo , x. zhai , j. n. eckstein , s. d. bader and a. bhattacharya , _ phys . rev .
b _ * 77 * 174409 ( 2008 ) s. smadici , p. abbamonte , a. bhattacharya , x. zhai , b. jiang , a. rusydi , j. n. eckstein , s. d. bader and jian - min zuo , _ phys
_ * 99 * 196404 ( 2007 ) h. b. zhao , k. j. smith , y. fan , g. lupke , a. bhattacharya , s. d. bader , m. warusawithana , x. zhai and j. n. eckstein , _ phys .
lett . _ * 100 * , 117208 ( 2008 ) a. bhattacharya , s. j. may , s. g. e. te velthuis , m. warusawithana , x. zhai , bin jiang , j .-
zuo , m. r. fitzsimmons , s. d. bader and j. n. eckstein , _ phys .
* 100 * , 257203 ( 2008 ) r. v. chamberlin , _
phys . _ * 57 * 3377 ( 1985 ) d. x. li , t. yamamura , s. nimori , k. yubuta and y. shiokawa , _ appl .
lett . _ * 87 * , 142505 ( 2005 ) s. j. may , p. j. ryan , j. l. robertson , j .- w .
kim , t. s. santos , e. karapetrova , j. l. zarestky , x. zhai , s. g. e. te velthuis , j. n. eckstein , s. d. bader and a. bhattacharya , _ nature materials _ * 8 * , 892 ( 2009 ) k. ueda , h. tabata , and t. kawai , _ phys .
b _ , * 60 * r12561 ( 1999 )
l. del bianco , d. fiorani , a. m. testa , e. bonetti , l. savini and s. signoretti _ phys .
b _ , * 66 * , 174418 ( 2002 ) | we have investigated the magnetic properties of transition metal oxide superlattices with broken inversion symmetry composed of three different antiferromagnetic insulators , [ ndmno@xmath0/srmno@xmath0/lamno@xmath0 ] . in the superlattices studied here
, we identify the emergence of a relaxor , glassy - like behavior below t@xmath2 = 36k .
our results offer the possibility to study and utilize magnetically metastable devices confined at nano - scale interfaces .
ks heterostructures of materials with strong electron - electron and electron - lattice interactions , the so - called correlated electron systems , are potential candidates for emergent interfacial properties including various forms of spin , charge and orbital ordering absent in bulk materials .
promising examples include multilayers composed of insulators of laalo@xmath0 and srtio@xmath0 with interfaces displaying properties of quasi - two - dimensional electron gases @xcite , superconductivity @xcite , metallic conductivity @xcite and ferromagnetism ( fm ) @xcite .
the multilayers composed of antiferromagnetic ( af ) insulators in bulk forms , lamno@xmath0 and srmno@xmath0 are also examples of emergent electromagnetic properties at the interfaces between dissimilar manganites @xcite . it has been demonstrated that these superlattices could posses fm order at the interfaces due to a charge reconstruction , although each parent material is an af @xcite .
the competitive interaction between the reconstructed fm at interfaces and the af states present far from the interface regions has been suggested to lead to a frustrated / glass like behavior @xcite . small external perturbation in glass - like correlated electron thin film devices , at a caged " nano - structured interface , in particular ,
is expected to lead to high degree of tunability .
these include , magnetoelectronics such as spin and charge memory devices at the atomic scale .
( a ) schematic drawing of the superlattice [ ( ndmno@xmath3)@xmath4/(srmno@xmath3)@xmath4/(lamno@xmath3)@xmath4]@xmath6 . the octahedral structures and spheres represent bo@xmath7 and a - site atoms in the abo@xmath0 perovskite structure , respectively .
the arrays of the arrows represent corresponding antiferromagnetic types .
( b ) synchrotron x - ray diffraction for different superlattices .
( c ) in - situ reflection high energy electron diffraction for the superlattice with @xmath8=5 .
( d ) the topography image of atomic force microscopy for the superlattice with @xmath8=5 . ] here we report on the relaxor and spin glass - like properties arising at the interface of superlattices , composed of insulating manganites : lamno@xmath0 , srmno@xmath0 and ndmno@xmath0 , which are a- , g , and a - type af , respectively . superlatties of [ ( ndmno@xmath3)@xmath9/(srmno@xmath3)@xmath9/(lamno@xmath3)@xmath9]@xmath11 were grown epitaxially on single crystalline srtio@xmath3 substrates at an ambient oxygen / ozone mixture of 10@xmath12 torr by layer - by - layer growth technology using the laser molecular beam epitaxy technique .
the details were reported in an earlier work @xcite .
figure 1 ( a ) depicts a schematic drawing of a superlattice with @xmath8=5 studied here , with alternate a sites around the blue octahedra representing mno@xmath7 in the abo@xmath0 pervoskite .
the total thickness of the superlattices was kept approximately 500 , varying ( @xmath8 , @xmath13 ) = ( 1 unit cell , 42 ) , ( 2 , 21 ) , ( 5 , 8) , and ( 12 , 4 ) in order to investigate the effect of the period .
structural characterization using synchrotron x - ray diffraction ( fig .
1(b ) ) along with the in - situ reflection high energy electron diffraction ( fig .
1(c ) ) indicate the presence of sharp interfaces with roughness less than one unit cell .
the topography image performed using atomic force microscopy ( fig .
1(d ) ) confirms the surface roughness to be less than one unit cell .
( a ) temperature dependence of the magnetization measured in a magnetic field of 0.1 t applied parallel to the plane of the films .
( b ) hystereses loops obtained at 10k with a magnetic field applied parallel to the plane of the films . ] the bulk magnetic properties of the superlattices were investigated using a superconducting quantum interference device magnetometer .
figure 2(a ) shows magnetization curves as a function of temperature .
data were taken by warming the sample in a field ( after cooling to 10k in zero - field ) ( zfc : dashed lines ) , and by cooling in the presence of a field ( fc : solid lines ) .
the discrepancy between fc and zfc curves at low temperature broadly resembles spin - glasses ( sg ) .
we will discuss this later . for superlattices with @xmath8@xmath142 ,
the magnetization values are significantly larger ( 8 times larger in the cases of @xmath8=5 ) than those of bulk lamno@xmath3 @xcite , srmno@xmath3 @xcite and ndmno@xmath3 @xcite . on the other hand , for @xmath8=1
there is a weak magnetic moment which is comparable to that of bulk lamno@xmath3 , srmno@xmath3 , or ndmno@xmath3 .
the magnetic properties of ( ndmno@xmath3)@xmath15/(lamno@xmath3)@xmath15/(srmno@xmath3)@xmath16 may be similar to the solid solution states of ( nd , sr , la)mno@xmath0 due to charge spreading through the interfaces , resulting in a three dimensional uniform charge distribution while keeping chemically sharp interfaces @xcite .
the weaker magnetization observed for the superlattice with @xmath8=1 compared to ( lamno@xmath0)@xmath17/(srmno@xmath0)@xmath16 , is due to a decreased magnetization caused in a la@xmath18sr@xmath19mno@xmath0 solid solution by replacing the la by nd , pr or y which have smaller ionic radii @xcite .
the tendency for an increase in magnetization and curie temperature with increasing period until a critical period , @xmath8=5 , and a decrease with an increase of n above 5 ( fig .
2(a ) ) agrees with earlier suggestions for [ ( lamno@xmath3)@xmath20/(srmno@xmath3)@xmath9 ] @xcite .
hystereses loops ( fig .
2 ( b ) ) were measured at 10k after field cooling in 0.1 t applied along the plane of the film .
( the linear part of the hysteresis due to the paramagnetic substrate has been subtracted . )
the coercive fields of the samples are @xmath21=0.04 , 0.14 , 0.36 and 0.28 t for @xmath8=1 , 2 , 5 and 12 , respectively .
the coercive fields along with magnetic moments reveal a critical period of @xmath8=5 , indicating the presence of fm phases as previously reported for a similar superlattice of [ lamno@xmath22srmno@xmath3 ] @xcite . to further investigate the presence of the thermal hysteresis at low temperatures ( fig .
2(a ) ) , we examined the sample with @xmath8=2 by measuring the dc magnetic susceptibility ( @xmath24 ) as a function of temperature in different magnetic fields ( 0.05 t
1.5 t ) . in fig .
3 the dashed and solid lines depict the susceptibility obtained in zfc and fc , respectively . the shift of the peaks of the zfc curves to lower temperatures with increasing field is a characteristic of sg / relaxors .
the normalized spin glass order parameter @xmath25 is defined as @xcite @xmath26/(c / t)\ ] ] or @xmath27 where c , f@xmath28 , @xmath29 , @xmath30 , and @xmath31 are the curie constant , the scaling function , the reduced temperature @xmath29=(t - t@xmath2)/t@xmath2 ( here t@xmath2 is the sg temperature ) , and the critical exponents characterizing the sg behavior , respectively . through the scaling analysis ( fig .
3 ( inset ) ) we obtain @xmath32=36k , @xmath30=0.7 and @xmath31=1.95 .
these values are in good agreement with experimental reports for other sg such as cdin@xmath19cr@xmath33s@xmath34 ( @xmath30=0.75 and @xmath31=2.3 @xcite ) .
notably there is deviation from the scaling function at low fields ( 0.05 and 0.1 t ) .
this behavior may be due to an inhomogeneous sg order in the superlattices , such as coexistance of the former with af and fm regions whose volume ratio may be changed by an applied magnetic field . for samples with @xmath355 we do not observe the scaling law because the coexistence and modulation of the sg , af , and fm phases as a function of thickness hinders the characterization of the sg behavior from the other regions .
magnetic susceptibility ( @xmath24 ) for magnetic fields from 0.05 t to 1.5 t for the sample with @xmath8=2 . the dashed and solid lines depict the susceptibility obtained with zfc and fc conditions , respectively .
( inset ) spin glass scaling for the superlattice with @xmath8=2 . ] the time decay of the magnetization for the superlattice with @xmath8=2 ( fig .
4 ) adds credence to the glassy characteristics .
( the other films also show time relaxation but we do not present the data here . )
the relaxation of the thermoremanent magnetization ( fig .
[ aging ] ( a ) ) was measured by the following method .
the sample was cooled from room temperature to 10k in the presence of a magnetic field of 0.1 t applied parallel to the film s plane .
when the temperature was stable , the magnetic field was switched off and the magnetization decay was recorded as a function of time for 60 min .
the decay curve is fitted by a stretched - exponential function ( solid line ) @xcite @xmath36 .
we find @xmath37 = 0.7 which is typical of other sg systems such as agmn @xcite .
the slow increase of the magnetization after switching on the magnetic field is depicted in fig .
4(b ) .
here the sample with @xmath8=2 was cooled down from room temperature to 10k in the absence of a magnetic field . when at 10k , a field of 0.1 t was applied parallel to the surface of the film and
data was recorded .
the data is fitted by the logarithmic function ( solid line ) @xcite @xmath38 .
time response of the magnetization for [ ( ndmno@xmath0)@xmath17/(srmno@xmath0)@xmath17/(lamno@xmath0)@xmath17 ] .
( a ) the relaxation of the magnetization is measured at 10k after cooling in a magnetic field of 0.1 t applied parallel to the plane of the film .
the solid line is a fit to @xmath40 = @xmath41exp(-@xmath42 ) ( b ) the increase of the magnetization is measured at 10k in a field of 0.1 t applied parallel to the film s plane after cooling in zero field .
the solid line is a fit to @xmath40 = @xmath41 + @xmath43 . ]
the thermal hysteresis ( fig .
2(a ) ) , the scaling curve ( fig .
3 : inset ) and the ageing signatures ( fig .
4 ) reveal the presence of a sg like behavior .
possible origins for sg in this system include : ( 1 ) sg characteristics present in each layer , ( 2 ) miscut between substrate and the first layer from the substrate , lamno@xmath0 , and ( 3 ) magnetic frustration between fm and af regions , where fm and af regions are possibly present at interfaces and in the core of each layer , respectively . the possibility for this behavior being due to each layer can be ruled out , however , by the systematic changes in the amplitudes and the irreversible temperatures of the magnetization for samples with different periods ( fig .
2 ) . also
magnetization curves of each individual manganite show af not a sg . a possible source for
the sg characteristics may be the interface between the substrate ( srtio@xmath0 ) and the first deposited layer of lamno@xmath0 .
this too , can not be the origin for our observations , since the magnetization curves as a function of temperature for a 60 unit cells lamno@xmath0 layer grown on srtio@xmath0 indicates af not a sg @xcite .
we believe a competition between the fm and af layers may account for our observations .
in fact , such a competition has already been proposed for superlattices of [ ( lamno@xmath3)@xmath44/(srmno@xmath3)@xmath9 ] @xcite .
the modulated magnetization of af and fm layers as a function of depth was studied using polarized neutron reflectivity ( pnr ) in a superlattice with @xmath8=12 , whose period is most suitable for pnr .
figure 5(a ) shows the pnr results at 300k ( above t@xmath45 ) with non - polarized neutrons since there is no magnetic signature at this temperature .
the solid line depicts a fit of the calculated reflectivity obtained from the scattering length density ( sld ) model as a function of depth .
the sld profile ( inset ) for ndmno@xmath0 , srmno@xmath0 , and lamno@xmath0 gives 3.65 , 3.55 and 3.75@xmath4610@xmath47 @xmath48 , respectively , in good agreement with calculated and experimental values @xcite .
the weak bragg peak at @xmath25 = 0.045 @xmath49 is due to the similarity in the nuclear scattering length for la , sr and nd atoms .
the reflectivity measured in a magnetic field of 0.6 t applied parallel to the film s surface , after field cooling to 10k in 0.6 t , shows strong bragg peaks and significant difference between r@xmath50 and r@xmath51 , indicating the presence of a magnetic modulation in the superlattice .
r@xmath50 and r@xmath51 are obtained by the polarized neutrons with spin states parallel and antiparallel to the magnetic field , respectively . from our best fit to the pnr data
, we obtained the magnetic profile shown in the inset of fig .
5 . as in earlier reports on superlattices composed of srmno@xmath0 and lamno@xmath0 layers @xcite
, our data also reveal an enhancement in the magnetization at the interfaces of ndmno@xmath0/srmno@xmath0 ( 1.1@xmath52/unit cell ) and srmno@xmath0/lamno@xmath0 ( 3.3@xmath52/unit cell ) .
the obtained thickness of the interfaces is around 10 .
notably , there is no signature of an enhancement at interfaces of lamno@xmath0/ndmno@xmath0 .
this may be due to the absence of polarity - discontinuity between these layers @xcite . in the regions
far from the interfaces , ndmno@xmath0 ( 0.7@xmath52/unit cell ) , srmno@xmath0 ( @xmath530.1@xmath52/unit cell ) , and lamno@xmath0 ( 1.5@xmath52/unit cell ) layers have comparable values to single films grown on srtio@xmath0 or in bulk @xcite .
an integrated magnetization estimated from the values we obtained by the fitting in fig .
5 for the film with @xmath8=12 is within 10% of the saturated magnetization moment obtained by bulk magnetization ( fig .
2(b ) ) .
we assumed that the magnetization and the thickness of the interfaces for the film with @xmath8=5 are same to those values obtained from the films with @xmath8=12 .
an integrated magnetization for the film with @xmath8=5 which is obtained based on the above - mentioned assumptions is also within 10% of the value obtained by bulk magnetization .
therefore , we may interpret the relaxor behavior being due to the competitive interaction between fm mainly present at interfaces and af regions in a magnetically modulated system @xcite .
the large coercive fields ( fig .
2(b ) ) commonly occurring by pinning the fm spins nearby an af layer support the competition at fm / af interfaces .
( color online ) ( a ) polarized neutron reflectivity ( pnr ) measurements for the sample with @xmath8=12 taken at 300k ( above t@xmath54 ) .
the orange , green and blue regions depict the regions of ndmno@xmath0(n ) , srmno@xmath0(s ) , and lamno@xmath0(l ) layers , respectively .
the solid line is a fit of the calculated reflectivity obtained using the scattering length density ( sld ) model .
( b ) pnr taken at 10k ( below t@xmath54 ) in 0.6 t after field cooling in the same field .
the red ( upper ) and green ( lower ) circles are the r@xmath50 and r@xmath55 data , respectively .
( inset ) the magnetic structure obtained from a calculation which reproduce the pnr data as represented with the solid lines . ] in summary , we fabricated a series of superlattices stacked repeatedly by different types of af insulators namely , lamno@xmath3 , srmno@xmath3 and ndmno@xmath3 . the magnetic properties obtained by bulk magnetometry have revealed the presence of fm , af and spin glass ( sg ) phases .
the thermal hysteresis and time dependent magnetization indicate a sg like behavior below t@xmath2 ( = 36k ) .
scaling shows the critical exponents to be @xmath30=0.7 and @xmath31=1.95 .
the possible origin of the sg characteristics may be due to the competing interactions between fm and af regions .
a modulation of fm and af regions have been detected by polarized neutron reflectivity .
this study may be potentially applicable to metastable magnetic memory devices which can offer a gateway to engineer sub - nano - scale metastates confined at oxide interfaces . |
the metric fixed point theory ( @xcite ) and its variations ( @xcite ) are far - reaching developments of banach s contraction principle , where _ metric conditions _ on the underlying space and maps under consideration play a fundamental role .
this paper addresses fixed points of nonlinear maps in _ modular spaces _ introduced recently by the author ( @xcite@xcite ) as generalizations of orlicz spaces and classical modular spaces ( @xcite , @xcite@xcite ) , where _ modular structures _ ( involving nonlinearities with more rapid growth than power - like functions ) , play the crucial role . under different contractive assumptions and the supplementary @xmath1-condition on modulars fixed
point theorems in classical modular linear spaces were established in @xcite .
we begin with a certain motivation of the definition of a ( metric ) _ modular _ , introduced axiomatically in @xcite . a simple and natural way to do it is to turn to physical interpretations . informally speaking , whereas a metric on a set represents nonnegative finite distances between any two points of the set , a modular on a set attributes a nonnegative ( possibly , infinite valued ) `` field of ( generalized ) velocities '' : to each `` time '' @xmath2 ( the absolute value of ) an average velocity @xmath3 is associated in such a way that in order to cover the `` distance '' between points @xmath4 it takes time @xmath5 to move from @xmath6 to @xmath7 with velocity @xmath3 .
let us comment on this in more detail by exhibiting an appropriate example .
if @xmath8 is the distance from @xmath6 to @xmath7 and a number @xmath2 is interpreted as time , then the value @xmath9 is the average velocity , with which one should move from @xmath6 to @xmath7 during time @xmath5 , in order to cover the distance @xmath10 .
the following properties of the quantity from ( [ e : dxy ] ) are quite natural .
\(i ) two points @xmath6 and @xmath7 from @xmath11 coincide ( and @xmath12 ) if and only if any time @xmath13 will do to move from @xmath6 to @xmath7 with velocity @xmath14 ( i.e. , no movement is needed at any time ) .
formally , given @xmath4 , we have : @xmath15 where ` iff ' means as usual ` if and only if ' .
\(ii ) assuming the distance function to be symmetric , @xmath16 , we find that for any time @xmath13 the average velocity during the movement from @xmath6 to @xmath7 is the same as the average velocity in the opposite direction , i.e. , for any @xmath4 we have : @xmath17 \(iii ) the third property of ( [ e : dxy ] ) , which is , in a sense , a counterpart of the triangle inequality ( for velocities ! ) , is the most important .
suppose the movement from @xmath6 to @xmath7 happens to be made in two different ways , but the _ duration of time is the same _ in each case : ( a ) passing through a third point @xmath18 , or ( b ) straightforward from @xmath6 to @xmath7 .
if @xmath5 is the time needed to get from @xmath6 to @xmath19 and @xmath20 is the time needed to get from @xmath19 to @xmath7 , then the corresponding average velocities are @xmath21 ( during the movement from @xmath6 to @xmath19 ) and @xmath22 ( during the movement from @xmath19 to @xmath7 ) .
the total time needed for the movement in the case ( a ) is equal to @xmath23 .
thus , in order to move from @xmath6 to @xmath7 as in the case ( b ) one has to have the average velocity equal to @xmath24 . since ( as a rule ) the straightforward distance @xmath10 does not exceed the sum of the distances @xmath25 , it becomes clear from the physical intuition that the velocity @xmath24 does not exceed at least one of the velocities @xmath21 or @xmath22 .
formally , this is expressed as @xmath26 for all points @xmath27 and all times @xmath28 ( `` triangle '' inequality ) .
in fact , these inequalities can be verified rigorously : if , on the contrary , we assume that @xmath29 and @xmath30 , then multiplying the first inequality by @xmath5 , the second inequality by @xmath20 , summing the results and taking into account ( [ e : dxy ] ) , we find @xmath31 and @xmath32 and it follows that @xmath33 which contradicts the triangle inequality for @xmath34 . inequality ( [ e : axi3 ] ) can be obtained in a little bit more general situation .
let @xmath35 be a function from the set of positive reals into itself such that the function @xmath36 is nonincreasing on @xmath37 .
setting @xmath38 ( note that @xmath39 in ( [ e : dxy ] ) ) , we have @xmath40 a nonclassical example of `` generalized velocities '' satisfying ( [ e : axi])([e : axi3 ] ) is given by : @xmath41 if @xmath42 , and @xmath14 if @xmath43 .
a ( _ metric _ ) _ modular _ on a set @xmath11 is any one - parameter family @xmath44 of functions @xmath45 $ ] satisfying ( [ e : axi])([e : axi3 ] ) . in particular ,
the family given by ( [ e : dxy ] ) is the canonical ( = natural ) modular on a metric space @xmath46 , which can be interpreted as a field of average velocities . for a different interpretation of modulars related to the joint generalized variation of univariate maps and their relationships with classical modulars on linear spaces we refer to @xcite ( cf . also section [ s : exas ] ) .
the difference between a metric ( = distance function ) and a modular on a set is now clearly seen : a modular depends on a positive parameter and may assume infinite values ; the latter property means that it is impossible ( or prohibited ) to move from @xmath6 to @xmath7 in time @xmath5 , unless one moves with infinite velocity @xmath41 .
in addition ( cf .
( [ e : dxy ] ) ) , the `` velocity '' @xmath3 is _ nonincreasing _ as a function of `` time '' @xmath13 .
the knowledge of `` average velocities '' @xmath3 for all @xmath13 and @xmath4 provides more information than simply the knowledge of distances @xmath10 between @xmath6 and @xmath7 : the distance @xmath10 can be recovered as a `` limit case '' via the formula ( again cf .
( [ e : dxy ] ) ) : @xmath47 now we describe briefly the main result of this paper . given a modular @xmath48 on a set @xmath11
, we introduce the _ modular space _
@xmath49 around a point @xmath50 as the set of those @xmath51 , for which @xmath52 is finite for some @xmath53 .
a map @xmath54 is said to be _ modular contractive _ if there exists a constant @xmath55 such that for all small enough @xmath13 and all @xmath56 we have @xmath57 .
our main result ( theorem [ t : main ] ) asserts that if @xmath48 is _ convex _ and _ strict _ , @xmath58 is _ modular complete _ ( the emphasized notions will be introduced in the main text below ) and @xmath54 is modular contractive , then @xmath59 admits a ( unique ) fixed point : @xmath60 for some @xmath61 .
the successive approximations of @xmath62 constructed in the proof of this result converge to @xmath62 in the modular sense , which is weaker than the metric convergence .
in particular , banach s contraction principle follows if we take into account ( [ e : dxy ] ) .
this paper is organized as follows . in section [
s : mm ] we study modulars and convex modulars and introduce two modular spaces . in section [ s : seq ]
we introduce the notions of modular convergence , modular limit and modular completeness and show that they are `` weaker '' than the corresponding metric notions .
these notions are illustrated in section [ s : exas ] by examples .
section [ s : fp ] is devoted to a fixed point theorem for modular contractions in modular complete modular metric spaces .
this theorem is then applied in section [ s : appl ] to the existence of solutions of a carathodory - type ordinary differential equation with the right - hand side from the orlicz space @xmath63 . finally , in section [ s : cr ] some concluding remarks are presented .
in what follows @xmath11 is a nonempty set , @xmath64 is understood in the sense that @xmath65 and , in view of the disparity of the arguments , functions @xmath66 $ ] will be also written as @xmath67 for all @xmath64 and @xmath68 , so that @xmath44 with @xmath45 $ ] . [ d : md ] a function @xmath66 $ ] is said to be a ( metric ) _ modular on @xmath11 _ if it satisfies the following three conditions : \(i ) given @xmath68 , @xmath69 iff @xmath70 for all @xmath64 ; \(ii ) @xmath71 for all @xmath64 and @xmath68 ; \(iii ) @xmath72 for all @xmath73 and @xmath74 .
if , instead of ( i ) , the function @xmath48 satisfies only ( i@xmath75 ) @xmath76 for all @xmath64 and @xmath51,then @xmath48 is said to be a _
pseudomodular _ on @xmath11 , and if @xmath48 satisfies ( i@xmath75 ) and ( i@xmath77 ) given @xmath4 , if there exists a number @xmath13 , possibly depending on @xmath6 and @xmath7 , such that @xmath14 , then @xmath69,the function @xmath48 is called a _ strict modular _ on @xmath11 .
a modular ( pseudomodular , strict modular ) @xmath48 on @xmath11 is said to be _ convex _ if , instead if ( iii ) , for all @xmath28 and @xmath27 it satisfies the inequality \(iv ) @xmath78 .
a motivation of the notion of _ convexity _ for modulars , which may look unexpected at first glance , was given in ( * ? ? ?
* theorem 3.11 ) , cf .
also inequality ( [ e : fla ] ) ; a further generalization of this notion was presented in ( * ? ? ?
* section 5 ) . given a metric space @xmath46 with metric @xmath34 , two _
canonical _ strict modulars are associated with it : @xmath79 ( denoted simply by @xmath34 ) , which is independent of the first argument @xmath5 and is a ( nonconvex ) modular on @xmath11 in the sense of ( i)(iii ) , and the _ convex _ modular ( [ e : dxy ] ) , which satisfies ( i ) , ( ii ) and ( iv ) . both modulars @xmath34 and ( [ e : dxy ] ) assume only finite values on @xmath11 . clearly , if @xmath48 is a strict modular , then @xmath48 is a modular , which in turn implies @xmath48 is a pseudomodular on @xmath11 , and similar implications hold for convex @xmath48 . the essential property of a pseudomodular @xmath48 on @xmath11 ( cf .
* section 2.3 ) ) is that , for any given @xmath4 , the function @xmath80 $ ] is _ nonincreasing _ on @xmath37 , and so , the limit from the right @xmath81 and the limit from the left @xmath82 exist in @xmath83 $ ] and satisfy the inequalities : @xmath84 a _ convex _ pseudomodular @xmath48 on @xmath11 has the following additional property : given @xmath4 , we have ( cf .
* section 3.5 ) ) : @xmath85 i.e. , functions @xmath86 and @xmath87 are _ nonincresing _ on @xmath37 . throughout the paper
we fix an element @xmath50 arbitrarily .
[ d : modsp ] given a pseudomodular @xmath48 on @xmath11 , the two sets @xmath88 and @xmath89 are said to be _ modular spaces _
( around @xmath90 ) .
it is clear that @xmath91 , and it is known ( cf .
* sections 3.1,3.2 ) ) that this inclusion is proper in general .
it follows from ( * ? ? ?
* theorem 2.6 ) that if @xmath48 is a _ modular _ on @xmath11 , then the modular space @xmath92 can be equipped with a ( nontrivial ) metric @xmath93 , generated by @xmath48 and given by @xmath94 it will be shown later that @xmath93 is a well defined metric on a larger set @xmath58 . if @xmath48 is a _ convex _
modular on @xmath11 , then according to ( * ? ? ?
* section 3.5 and theorem 3.6 ) the two modular spaces coincide , @xmath95 , and this common set can be endowed with a metric @xmath96 given by @xmath97 moreover , @xmath96 is _ specifically _ equivalent to @xmath93 ( see ( * ? ? ?
* theorem 3.9 ) ) . by the convexity of @xmath48 ,
the function @xmath98 is a modular on @xmath11 in the sense of ( i)(iii ) and ( cf .
* formula ( 3.3 ) ) ) @xmath99 where the last inclusion may be proper ; moreover , @xmath100 on @xmath101 .
even if @xmath48 is a nonconvex modular on @xmath11 , the quantity ( [ e : dw * ] ) is also defined for all @xmath56 , but it has only few properties ( cf .
* theorem 3.6 ) ) : @xmath102 and @xmath103 . in this case
we have ( cf .
* theorem 3.9 and example 3.10 ) ) : if @xmath104 , then @xmath105 , and if @xmath106 , then @xmath107 .
let us illustrate the above in the case of a metric space @xmath46 with the two canonical modulars @xmath34 and @xmath48 from ( [ e : dxy ] ) on it .
we have : @xmath108 , and given @xmath4 , @xmath109 , @xmath110 , @xmath111 , @xmath112 and @xmath113 .
thus , the convex modular @xmath48 from ( [ e : dxy ] ) plays a more adequate role in restoring the metric space @xmath46 from @xmath48 ( cf .
@xmath114 on @xmath115 , whereas @xmath116 , @xmath117 and @xmath118 ) , and so , in what follows any metric space @xmath46 will be considered equipped only with the modular ( [ e : dxy ] ) .
this convention is also justified as follows .
now we exhibit the relationship between convex and nonconvex modulars and show that @xmath93 is a well defined metric on @xmath58 ( and not only on @xmath92 ) .
if @xmath48 is a ( not necessarily convex ) modular on @xmath11 , then the function ( cf .
( [ e : dxy ] ) where @xmath10 plays the role of a modular ) @xmath119 is always a _ convex _ modular on @xmath11 .
in fact , conditions ( i ) and ( ii ) are clear for @xmath120 and , as for ( iv ) , we have , by virtue of ( iii ) for @xmath48 : @xmath121 moreover , because @xmath122 , we find from ( [ e : inclu ] ) that @xmath123 . since @xmath124 for all @xmath56 , i.e. , @xmath125 on @xmath58 , and
@xmath126 is a metric on @xmath127 , then we conclude that @xmath93 is a _
well defined metric on @xmath58 _ ( the same conclusion follows immediately from ( * ? ? ?
* theorem 1 ) ) with @xmath128 ) .
this property distinguishes our theory of modulars from the classical theory : if @xmath129 is a classical modular on a linear space @xmath11 in the sense of musielak and orlicz ( @xcite ) and @xmath130 , , @xmath4 , then the expression @xmath131 is _ not allowed _ as a classical modular on @xmath11 . since @xmath120 is convex and @xmath125 on @xmath58 , given @xmath56 , by virtue of (
* theorem 3.9 ) , we have : @xmath132 @xmath133 more metrics can be defined on @xmath58 for a given modular @xmath48 on @xmath11 in the following general way ( cf .
* theorem 1 ) ) : if @xmath134 and @xmath135 is superadditive ( i.e. , @xmath136 for all @xmath137 ) and such that @xmath138 for @xmath139 and @xmath140 , then the function @xmath141 is a well defined metric on @xmath58 . given a pseudomodular ( modular , strict modular , convex or not ) @xmath48 on @xmath11 , @xmath13 and @xmath4 , we define the _ left _ and _ right _ _ regularizations _ of @xmath48 by @xmath142 since , by ( [ e : lapm ] ) , @xmath143 , and @xmath144 it is a routine matter to verify that @xmath145 and @xmath146 are pseudomodulars ( modulars , strict modulars , convex or not , respectively ) on @xmath11 , @xmath147 , @xmath148 , @xmath149 on @xmath92 and @xmath150 on @xmath58 . for instance , let us check the last two equalities for metrics .
given @xmath56 , by virtue of ( [ e : lapm ] ) , we find . in order to see that @xmath151 , we let @xmath152 be arbitrary , choose @xmath20 such that @xmath153 , which , by ( [ e : la2la1 ] ) , gives @xmath154 , and so , @xmath155 , and then let @xmath156 . in order to prove that @xmath157 , we let @xmath158 be arbitrary , choose @xmath20 such that @xmath159 , which , by ( [ e : la2la1 ] ) , implies @xmath160 , and so , @xmath161 , and then let @xmath162 . in this way we have seen that the regularizations provide no new modular spaces as compared to @xmath92 and @xmath58 and no new metrics as compared to @xmath93 and @xmath96 .
the right regularization will be needed in section [ s : fp ] for the characterization of metric lipschitz maps in terms of underlying modulars .
the notions of modular convergence , modular limit , modular completeness , etc . , which we study in this section , are known in the classical theory of modulars on linear spaces ( e.g. , @xcite ) . since the theory of ( metric )
modulars from @xcite@xcite is significantly more general than the classical theory , the notions mentioned above do not carry over to metric modulars in a straightforward way and ought to be reintroduced and justified .
[ d : wconv ] given a pseudomodular @xmath48 on @xmath11 , a sequence of elements @xmath163 from @xmath92 or @xmath58 is said to be _ modular convergent _
( more precisely , _ @xmath48-convergent _ ) to an element @xmath51 if there exists a number @xmath13 , possibly depending on @xmath164 and @xmath6 , such that @xmath165 .
this will be written briefly as @xmath166 ( as @xmath167 ) , and any such element @xmath6 will be called a _ modular limit _ of the sequence @xmath164 .
note that if @xmath165 , then by virtue of the monotonicity of the function @xmath168 , we have : @xmath169 for all @xmath170 .
it is clear for a metric space @xmath46 and the modular ( [ e : dxy ] ) on it that the metric convergence and the modular convergence in @xmath11 coincide .
we are going to show that the modular convergence is much weaker than the metric convergence ( in the sense to be made more precise below ) .
first , we study to what extent the above definition is correct , and what is the relationship between the modular and metric convergences in @xmath92 and @xmath58 .
[ t : corr ] let @xmath48 be a pseudomodular on @xmath11 .
we have : \(a ) the modular spaces @xmath92 and @xmath58 are closed with respect to the modular convergence , i.e. , if @xmath171 ( or @xmath58 ) , @xmath51 and @xmath166 , then @xmath172 ( or @xmath173 , respectively ) ; \(b ) if @xmath48 is a strict modular on @xmath11 , then the modular limit is determined uniquely ( if it exists ) .
\(a ) since @xmath166 , there exists a @xmath174 such that @xmath175 as @xmath167 .
first we treat the case when @xmath171 .
let @xmath176 be arbitrarily fixed .
then there is an @xmath177 such that @xmath178 . since @xmath179
, we have @xmath180 as @xmath181 , and so , there exists a @xmath182 such that @xmath183
. then conditions ( iii ) and ( ii ) from definition [ d : md ] imply @xmath184 the function @xmath185 is nonincreasing on @xmath37 , and so , @xmath186 implying @xmath187 as @xmath181 , i.e. , @xmath172 .
now suppose that @xmath188 .
then there exists an @xmath189 such that @xmath190 . since @xmath191
, there is a @xmath192 such that @xmath193 .
now it follows from conditions ( iii ) and ( ii ) that @xmath194 and so , @xmath173 .
\(b ) let @xmath171 or @xmath58 and @xmath4 be such that @xmath166 and @xmath195 . by the definition of the modular convergence ,
there exist @xmath196 and @xmath197 such that @xmath198 and @xmath199 as @xmath167 . by conditions ( iii ) and
( ii ) , @xmath200 it follows that @xmath201 , and so , by condition ( i@xmath77 ) from definition [ d : md ] , we get @xmath69 .
it was shown in ( * ? ? ?
* theorem 2.13 ) that if @xmath48 is a modular on @xmath11 , then for @xmath171 and @xmath172 we have : @xmath202 and so , the metric convergence ( with respect to the metric @xmath93 ) implies the modular convergence ( cf . definition [ d : wconv ] ) , but not vice versa in general . as the proof of ( *
* theorem 2.13 ) suggests , ( [ e : dwwla ] ) is also true for @xmath188 and @xmath173 .
an assertion similar to ( [ e : dwwla ] ) holds for cauchy sequences from the modular spaces @xmath92 and @xmath58 .
now we establish a result similar to ( [ e : dwwla ] ) for _ convex _ modulars .
[ t : dil = w ] let @xmath48 be a convex modular on @xmath11 . given a sequence @xmath164 from @xmath203 and an element @xmath173
, we have : @xmath204 a similar assertion holds for cauchy sequences with respect to @xmath96 .
step 1 . _ sufficiency .
_ given @xmath176 , by the assumption , there exists a number @xmath205
such that @xmath206 for all @xmath207 , and so , the definition ( [ e : dw * ] ) of @xmath96 implies @xmath208 for all @xmath207 .
first , suppose that @xmath209 .
given @xmath176 , we have : either ( a ) @xmath210 , or ( b ) @xmath211 . in case ( a ) , by the assumption , there is an @xmath205 such that @xmath212 for all @xmath207 , and so , by the definition of @xmath96 , @xmath213 for all @xmath207 . since @xmath214 and @xmath210 , inequality ( [ e : mpc ] )
yields : @xmath215 in case ( b ) we set @xmath216 , where @xmath217 is as above
. then , as we have just established , @xmath218 for all @xmath219 . now , assume that @xmath220 . again , given @xmath176 , we have : either ( a ) @xmath210 , or ( b ) @xmath211 . in case ( a )
there is an @xmath221 such that @xmath222 for all @xmath223 , and so , @xmath206 for all @xmath223 .
since @xmath210 and @xmath220 , by virtue of ( [ e : mpc ] ) , we find @xmath224 in case ( b ) we put @xmath225 , where @xmath226 is as above .
then it follows that @xmath218 for all @xmath227 .
thus , we have shown that @xmath198 as @xmath167 for all @xmath13 .
the assertion for cauchy sequences is of the form : @xmath228 its proof is similar to the one given in step 1 with suitable modifications .
theorem [ t : dil = w ] shows , in particular , that in a metric space @xmath46 with modular ( [ e : dxy ] ) on it the metric and modular convergences are equivalent . [
d : delta2 ] a pseudomodular @xmath48 on @xmath11 is said to _ satisfy the _
( sequential )
_ @xmath1-condition _
( on @xmath58 ) if the following condition holds : given a sequence @xmath188 and @xmath173 , if there exists a number @xmath13 , possibly depending on @xmath164 and @xmath6 , such that @xmath165 , then @xmath229 .
a similar definition applies with @xmath58 replaced by @xmath92 . in the case of a metric space @xmath46 the modular ( [ e : dxy ] ) clearly satisfies the @xmath1-condition on @xmath11 . the following important observation , which generalizes the corresponding result from the theory of classical modulars on linear spaces ( cf .
* i,5.2.iv ) ) , provides a criterion for the metric and modular convergences to coincide .
[ t : dew ] given a modular @xmath48 on @xmath11 , we have : the metric convergence on @xmath58 ( with respect to @xmath93 if @xmath48 is arbitrary , and with respect to @xmath96 if @xmath48 is convex ) coincides with the modular convergence iff @xmath48 satisfies the on @xmath58 .
let @xmath188 and @xmath173 be given .
we know from ( [ e : dwwla ] ) and theorem [ t : dil = w ] that the metric convergence ( with respect to @xmath93 if @xmath48 is a modular or with respect to @xmath96 if @xmath48 is a convex modular ) of @xmath230 to @xmath6 is equivalent to @xmath231 ( @xmath232 ) suppose that the metric convergence coincides with the modular convergence on @xmath58 .
if there exists a @xmath233 such that @xmath175 as @xmath167 , then @xmath230 is modular convergent to @xmath6 , and so , @xmath230 converges to @xmath6 in metric ( @xmath93 or @xmath96 ) .
it follows that ( [ e : woc ] ) holds implying , in particular , @xmath234 as @xmath167 , and so , @xmath48 satisfies the @xmath1-condition .
( @xmath235 ) by virtue of ( [ e : woc ] ) , the metric convergence on @xmath58 always implies the modular convergence , and so , it suffices to verify the converse assertion , namely : if @xmath166 , then ( [ e : woc ] ) holds . in fact , if @xmath166 , then @xmath175 as @xmath167 for some constant @xmath174 . the @xmath1-condition implies @xmath234 as @xmath167 , and so , the induction yields @xmath236 as @xmath167 for all @xmath237 .
now , given @xmath13 , there exists a @xmath238 such that @xmath239 . by the monotonicity of @xmath240
, we have : @xmath241 by the arbitrariness of @xmath13 , condition ( [ e : woc ] ) follows .
[ d : wcauchy ] given a modular @xmath48 on @xmath11 , a sequence @xmath188 is said to be _ modular cauchy _ ( or _ @xmath48-cauchy _ ) if there exists a number @xmath242 such that @xmath243 as @xmath244 , i.e. , @xmath245 it follows from theorem [ t : dil = w ] ( step 2 in its proof ) and definition [ d : wcauchy ] that a sequence from @xmath58 , which is cauchy in metric @xmath93 or @xmath96 , is modular cauchy .
note that a modular convergent sequence is modular cauchy .
in fact , if @xmath166 , then @xmath198 as @xmath167 for some @xmath13 , and so , for each @xmath176 there exists an @xmath205 such that @xmath246 for all @xmath207 .
it follows from ( iii ) that if @xmath247 , then @xmath248 , which implies that @xmath164 is modular cauchy
. the following definition will play an important role below .
[ d : wcomp ] given a modular @xmath48 on @xmath11 , the modular space @xmath58 is said to be _ modular complete _ ( or _ @xmath48-complete _ ) if each modular cauchy sequence from @xmath58 is modular convergent in the following ( more precise ) sense : if @xmath188 and there exists a @xmath242 such that @xmath249 , then there exists an @xmath173 such that @xmath165 .
the notions of modular convergence , modular limit and modular completeness , introduced above , are illustrated by examples in the next section .
it is clear from ( [ e : dxy ] ) that for a metric space @xmath46 these notions coincide with respective notions in the metric space setting .
we begin with recalling certain properties of @xmath250-functions and convex functions on the set of all nonnegative reals @xmath134 .
a function @xmath251 is said to be a _
@xmath250-function _ if it is continuous , nondecreasing , unbounded ( and so , @xmath252 ) and assumes the value zero only at zero : @xmath253 iff @xmath254 . if @xmath251 is a convex function such that @xmath253 iff @xmath254 , then it is ( automatically ) continuous , strictly increasing and unbounded , and so , it is a convex @xmath250-function . also , @xmath250 is superadditive : @xmath255 for all @xmath256 ( cf .
* section i.1 ) ) .
moreover , @xmath250 admits the inverse function @xmath257 , which is continuous , strictly increasing , @xmath258 iff @xmath254 , @xmath259 , and which is subadditive : @xmath260 for all @xmath256 .
the function @xmath250 is said to _ satisfy the @xmath1-condition at infinity _
* section i.4 ) ) if there exist constants @xmath261 and @xmath262 such that @xmath263 for all @xmath264 .
let the triple @xmath265 be a _
metric semigroup _ , i.e. , the pair @xmath266 is a metric space with metric @xmath34 , the pair @xmath267 is an abelian semigroup with respect to the operation of addition @xmath268 , and @xmath34 is translation invariant in the sense that @xmath269 for all @xmath270 .
any normed linear space @xmath271 is a metric semigroup with the induced metric @xmath272 , @xmath273 , and the addition operation @xmath268 from @xmath274 .
if @xmath275 is a convex cone ( i.e. , @xmath276 whenever @xmath277 and @xmath278 ) , then the triple @xmath279 is also a metric semigroup .
a nontrivial example of a metric semigroup is as follows ( cf .
let @xmath280 be a real normed space and @xmath274 be the family of all nonempty closed bounded convex subsets of @xmath281 equipped with the hausdorff metric @xmath34 given by @xmath282 , where and @xmath283 .
given @xmath284 , we define @xmath285 as the closure in @xmath281 of the minkowski sum @xmath286 .
then the triple @xmath287 is a metric semigroup ( actually , @xmath274 is an abstract convex cone ) . for more information on metric semigroups and their special cases , abstract convex cones , including examples we refer to @xcite and references therein . given a closed interval @xmath288\subset{\mathbb{r}}$ ] with @xmath289 , we denote by @xmath290}$ ] the set of all mappings @xmath291\to m$ ] . if @xmath250 is a _
@xmath250-function on @xmath292 , we define a function @xmath293 $ ] for all @xmath13 and @xmath294 by ( note that @xmath48 depends on @xmath250 ) @xmath295 where the supremum is taken over all partitions @xmath296 of the interval @xmath288 $ ] , i.e. , @xmath297 and @xmath298 .
it was shown in ( * ? ? ?
* sections 3 , 4 ) that @xmath48 is a _
convex pseudomodular _ on @xmath299 .
thus , given @xmath300 , the modular space @xmath301 ( here @xmath90 denotes also the constant mapping @xmath302 for all @xmath303 $ ] ) , which was denoted in ( * ? ? ? * ( 3.20 ) and section 4.1 ) by @xmath304;m)$ ] and called the _ space of mappings of bounded generalized @xmath250-variation _ , is well defined and , by the translation invariance of @xmath34 on @xmath274 , we have : @xmath305;m)$ ] iff @xmath291\to m$ ] and there exists a constant @xmath53 such that @xmath306 note that @xmath52 from ( [ e : wlax0 ] ) is independent of @xmath300 ; this value is called the _ generalized @xmath307-variation _ of @xmath6 , where @xmath308 , @xmath309 .
since @xmath48 satisfies on @xmath299 conditions ( i@xmath75 ) , ( ii ) and ( iv ) ( and not ( i ) in general ) from definition [ d : md ] , the quantity @xmath96 from ( [ e : dw * ] ) is only a pseudometric on @xmath310 and , in particular , only @xmath102 holds for @xmath311 ( note that @xmath312 was denoted by @xmath313 in ( * ? ? ?
* equality ( 4.5 ) ) ) . in
order to `` turn '' ( [ e : wlaxy ] ) into a modular , we fix an @xmath300 and set @xmath314\to m\mid x(a)=x_0\}\subset{\mathbb{x}}$ ] .
we assert that @xmath48 from ( [ e : wlaxy ] ) is a _ strict _ convex modular on @xmath11 .
in fact , given @xmath4 and @xmath315 $ ] with @xmath316 , it follows from ( [ e : wlaxy ] ) that @xmath317 and so , by the translation invariance of @xmath34 and the triangle inequality , @xmath318 now , if we suppose that @xmath14 for some @xmath13 , then for all @xmath303 $ ] , @xmath319 , we get ( note that @xmath320 ) @xmath321 thus , @xmath322 for all @xmath303 $ ] , and so , @xmath69 as elements of @xmath11 . it is clear for the modular space @xmath49 that @xmath323;m)\cap x,\ ] ] i.e. , @xmath173 iff @xmath291\to m$ ] , @xmath324 and ( [ e : wlax0 ] ) holds for some @xmath13 .
moreover , the function @xmath96 from ( [ e : dw * ] ) is a _
metric _ on @xmath58 . in this subsection
we show that if @xmath265 is a _
complete _ metric semigroup ( i.e. , @xmath266 is complete as a metric space ) , then the modular space @xmath58 from ( [ e : gvx ] ) is _ modular complete _ in the sense of definition [ d : wcomp ] .
let @xmath188 be a @xmath48-cauchy sequence , so that @xmath243 as @xmath244 for some constant @xmath242 .
given @xmath325 and @xmath303 $ ] , @xmath326 , it follows from ( [ e : fwl ] ) with @xmath327 , @xmath328 and @xmath329 that ( again note that @xmath330 for all @xmath331 ) @xmath332 this estimate , the modular cauchy property of @xmath164 , the continuity of @xmath333 and the completeness of @xmath265 imply the existence of an @xmath291\to m$ ] , @xmath324 ( and so , @xmath51 ) , such that the sequence @xmath164 converges pointwise on @xmath288 $ ] to @xmath6 , i.e. , @xmath334 for all @xmath303 $ ] .
we assert that @xmath165 . by the ( sequential ) lower semicontinuity of the functional @xmath335 from ( [ e : wlaxy ] ) ( cf .
* assertion ( 4.8 ) on p. 27 ) ) , we get @xmath336 now , given @xmath176 , by the modular cauchy condition for @xmath164 , there is an @xmath205 such that @xmath337 for all @xmath207 and @xmath338 , and so , @xmath339 since the limit inferior does not exceed the limit superior ( for any real sequences ) , it follows from the last displayed line and ( [ e : liminf ] ) that @xmath340 for all @xmath207 , i.e. , @xmath198 as @xmath167 . finally , since , by theorem [ t : corr](a ) , @xmath58 is closed with respect to the modular convergence , we infer that @xmath173 , which was to be proved . in order to be able to calculate explicitly , for the sake of simplicity we assume furthermore that @xmath341 with @xmath272 , @xmath342 , and the function @xmath250 satisfies the _ orlicz condition at infinity _ : @xmath343 as @xmath344 . in this case the value @xmath345 ( cf .
( [ e : wlax0 ] ) with @xmath346 ) is known as the _ @xmath250-variation _ of the function @xmath291\to{\mathbb{r}}$ ] ( in the sense of f.riesz , yu.t.medvedev and w.orlicz ) , the function @xmath6 with @xmath347 is said to be _ of bounded @xmath250-variation _ on @xmath288 $ ] , and we have : @xmath348}.\ ] ] denote by @xmath349 $ ] the space of all absolutely continuous real valued functions on @xmath288 $ ] and by @xmath350 $ ] the space of all ( equivalence classes of ) lebesgue summable functions on @xmath288 $ ] .
the following criterion is known for functions @xmath291\to{\mathbb{r}}$ ] to be in the space @xmath351={\mathbb{x}}_w^*$ ] ( for more details see @xcite , ( * ? ? ?
* sections 3 , 4 ) , @xcite , ( * ? ? ?
* section 2.4 ) , @xcite ) : @xmath352 $ ] iff @xmath353 for some @xmath53 ( i.e. , @xmath354 is of bounded @xmath250-variation on @xmath288 $ ] ) iff @xmath355 $ ] and its derivative @xmath356 $ ] ( defined almost everywhere on @xmath288 $ ] ) satisfies the condition : @xmath357 given @xmath358 , we set @xmath314\to{\mathbb{r}}\mid x(a)=x_0\}$ ] , and so ( cf.([e : gvx ] ) ) , @xmath359:x(a)=x_0\}.\ ] ] thus , the modular @xmath48 is strict and convex on @xmath11 and the modular space ( [ e : xax ] ) is modular complete .
note that @xmath58 is _ not _ a linear subspace of @xmath351 $ ] , which is a normed banach algebra ( cf .
* theorem 3.6 ) ) . here
we present an example when the metric and modular convergences coincide .
this example is a modification of example 3.5(c ) from @xcite .
we set @xmath288=[0,1]$ ] , @xmath341 and @xmath360 for @xmath309 .
clearly , @xmath250 satisfies the orlicz condition , but does not satisfy the @xmath1-condition at infinity . given a number @xmath361 , we define a function @xmath362\to{\mathbb{r}}$ ] by @xmath363 since @xmath364 for @xmath365 , by ( [ e : intab ] ) , for any number @xmath13 we find @xmath366 { \displaystyle}\frac{\alpha}{\lambda-\alpha } & \mbox{if } & \lambda>\alpha .
\end{array } \right.\ ] ] it follows that the modular @xmath48 can take infinite values ( although it is strict ) and that @xmath367 for all @xmath361 .
also , we have : @xmath368 thus , if we set @xmath369 and @xmath370 for @xmath331 , then we find that @xmath371 as @xmath167 and @xmath372 as @xmath167 for all @xmath13 , and , in accordance with theorem [ t : dil = w ] , these two convergences are equivalent .
here we expose an example
when the modular convergence is weaker than the metric convergence .
let @xmath288 $ ] , @xmath274 and @xmath250 be as in example [ s : exas].5 . given @xmath373 , we define a function @xmath374\to{\mathbb{r}}$ ] as follows : @xmath375 and @xmath376 since @xmath377 for @xmath378 and @xmath379 $ ] , we have : @xmath380 and so , by virtue of ( [ e : intab ] ) , given @xmath13 , we find @xmath381 where @xmath382 -\log\beta & \mbox{if } & \lambda=1 , \end{array } \right.\ ] ] and @xmath383 also , @xmath384 if @xmath209 , and @xmath385 if @xmath220 ( cf .
example [ s : exas].5 with @xmath386 ) .
thus , @xmath387 for all @xmath373 .
clearly , @xmath388 converges pointwise on @xmath389 $ ] to @xmath90 as @xmath390 ( actually , the first inequality in the proof of ( * ? ? ?
* lemma 4.1(a ) ) shows that the convergence is uniform on @xmath389 $ ] ) .
now we calculate the values @xmath391 for @xmath13 and @xmath392 and invesigate their convergence to zero as @xmath390 .
since @xmath393 we have : @xmath394 and so , by virtue of ( [ e : x - y ] ) and ( [ e : intab ] ) , @xmath395 if @xmath209 , we have @xmath396 and so , @xmath397 for all @xmath398 and @xmath209 . now suppose that @xmath220
. then @xmath399 where @xmath400 \!\!&=\!\!&2^{1/\lambda}\!\cdot\!{\displaystyle}\frac{\lambda\beta}{\lambda-1}\to0\quad \mbox{as}\quad\beta\to+0\nonumber \end{aligned}\ ] ] and @xmath401 it follows that @xmath402 as @xmath390 for all @xmath220 . on the other hand , since @xmath397 for all @xmath398 and @xmath209 ( as noticed above ) , we get @xmath403 , and so , @xmath392 can not converge to zero as @xmath390 .
thus , if we set @xmath404 and @xmath405 for @xmath331 , then we find @xmath406 as @xmath167 , whereas @xmath407 as @xmath167 only for @xmath220 .
since convex modulars play the central role in this section , we concentrate mainly on them .
we begin with a characterization of @xmath96-lipschitz maps on the modular space @xmath58 in terms of their generating convex modulars @xmath48 .
[ t : lip ] let @xmath48 be a convex modular on @xmath11 and @xmath408 be a constant . given a map @xmath54 and @xmath56 , the lipschitz condition @xmath409 is equivalent to the following : @xmath410 for all @xmath13 such that @xmath411 .
first , note that , given @xmath412 , the function , defined by @xmath413 , @xmath13 , @xmath4 , is also a convex modular on @xmath11 and @xmath414 : @xmath415 .
we may suppose that @xmath416 .
for any @xmath417 , by the assumption , we find @xmath418 , whence @xmath419 .
it follows that if @xmath13 is such that @xmath411 , then , by ( [ e : dw * ] ) , @xmath161 implying , in view of ( [ e:1c ] ) , @xmath420 and so , @xmath421 . passing to the limit as @xmath422 , we arrive at the desired inequality @xmath410 . .
by the assumption , the set @xmath423 is contained in the set @xmath424 , and so , taking the infima , by virtue of ( [ e : dw * ] ) , ( [ e:1c ] ) and the equality @xmath425 , we get @xmath426 which implies that @xmath59 satisfies the lipschitz condition with constant @xmath427 .
theorem [ t : lip ] can be reformulated as follows . since ( cf .
* theorem 3.8(a ) ) and ( [ e : dw * ] ) ) , for @xmath428 , @xmath429 we have : @xmath430 iff @xmath431 for all @xmath432 . for a metric space @xmath46 and the modular @xmath48 from ( [ e : dxy ] ) on it ,
theorem [ t : lip ] gives the usual lipschitz condition : @xmath433 for all @xmath13 such that @xmath434 , i.e. , @xmath435 for all @xmath436 , and so , @xmath437 . as a corollary of theorem [ t : lip ] , we find that @xmath438 in fact , it suffices to note only that if @xmath13 is such that @xmath411 , then , by ( [ e : lapm ] ) , @xmath439 , and apply theorem [ t : lip ] .
now we briefly comment on @xmath93-lipschitz maps on @xmath58 , where @xmath48 is a general modular on @xmath11 and @xmath93 is the metric from ( [ e : dw ] ) . note that , given @xmath412 , the function @xmath440 is also a modular on @xmath11 and @xmath441 on @xmath442 . following the lines of the proof of theorem [ t : lip ] , we get [ t : lip2 ] if @xmath48 is a modular on @xmath11 and @xmath408 , given @xmath54 and @xmath56 , we have @xmath443 iff @xmath444 for all @xmath13 such that @xmath445 . the following assertion is a corollary of theorem [ t : lip2 ] : @xmath446 [ d : modcontr ] given a ( convex ) modular @xmath48 on @xmath11 , a map @xmath54 is said to be _ modular contractive _ ( or a _ @xmath48-contraction _ ) provided there exist numbers @xmath55 and @xmath233 , possibly depending on @xmath427 , such that @xmath447 a few remarks are in order .
first , by virtue of ( [ e : dxy ] ) , for a metric space @xmath46 condition ( [ e : mc ] ) is equivalent to the usual one : @xmath437 .
second , condition ( [ e : mc ] ) is a _ local _ one with respect to @xmath5 as compared to the assumption on the left in ( [ e : corlip ] ) , and the principal inequality in it may be of the form @xmath448 .
third , if , in addition , @xmath48 is _ strict _ and if we set @xmath449 , then ( [ e : mc ] ) is a consequence of the following : there exists a number @xmath450 such that @xmath451 where the supremum is taken over all @xmath56 such that @xmath416 . in order to see this
, we first note that the left hand side in ( [ e : mc2 ] ) is well defined in the sense that , by virtue of ( i@xmath77 ) from definition [ d : md ] , @xmath452 for all @xmath13 and @xmath416 .
choose any @xmath427 such that @xmath453 .
it follows from ( [ e : mc2 ] ) that @xmath454}\,\biggl ( \sup_{x\ne y}\frac{w_{h\lambda}(tx , ty)}{w_\lambda(x , y)}\biggr)\le1<\frac kh,\ ] ] and so , there exists a @xmath455 such that @xmath456 whence @xmath457 taking into account inequalities ( [ e : mpc ] ) and @xmath458 , we get @xmath459 which together with the previous inequality gives : @xmath460 setting @xmath461 and @xmath462 and noting that @xmath463 and @xmath464 , the last inequality implies @xmath465 for all @xmath466 and @xmath56 , which is exactly ( [ e : mc ] ) . the main result of this paper is the following fixed point theorem for modular contractions in modular metric spaces @xmath58 .
[ t : main ] let @xmath48 be a strict convex modular on @xmath11 such that the modular space @xmath58 is @xmath48-complete , and @xmath54 be a @xmath48-contractive map such that @xmath467 then @xmath59 has a fixed point , i.e. , @xmath60 for some @xmath61 . if , in addition , the modular @xmath48 assumes only finite values on @xmath58 , then condition ( [ e : reach ] ) is redundant , the fixed point @xmath62 of @xmath59 is unique and for each @xmath468 the sequence of iterates @xmath469 is modular convergent to @xmath62 .
since @xmath48 is convex , the following inequality follows by induction from condition ( iv ) of definition [ d : md ] : @xmath470 where @xmath471 , @xmath472 and @xmath473 . in the proof below
we will need a variant of this inequality .
let @xmath325 , @xmath474 , @xmath475 and @xmath476 .
setting @xmath477 , @xmath478 for @xmath479 , @xmath480 for @xmath481 and applying ( [ e : cla ] ) to the primed lambda s and @xmath6 s , we get : @xmath482 by the @xmath48-contractivity of @xmath59 , there exist two numbers @xmath55 and @xmath483 such that condition ( [ e : mc ] ) holds . setting @xmath484 , the assumption ( [ e : reach ] )
implies the existence of an element @xmath485 such that @xmath486 is finite .
we set @xmath487 and @xmath488 for all integer @xmath489 , and so , @xmath188 and @xmath490 , where @xmath491 designates the @xmath492-th iterate of @xmath59 .
we are going to show that the sequence @xmath164 is @xmath48-cauchy . since @xmath493 for all @xmath494 , inequality ( [ e : mc ] )
yields : @xmath495 and it follows by induction that @xmath496 let integers @xmath492 and @xmath497 be such that @xmath474 .
we set @xmath498 by virtue of ( [ e : varcla ] ) with @xmath499 and ( [ e : wki ] ) , we find @xmath500 taking into account that @xmath501 and applying ( [ e : mpc ] ) , we get : @xmath502 thus , the sequence @xmath164 is modular cauchy , and so , by the @xmath48-completeness of @xmath58 , there exists an @xmath61 such that @xmath503 since @xmath48 is strict , by theorem [ t : corr](b ) , the modular limit @xmath62 of the sequence @xmath164 is determined uniquely .
let us show that @xmath62 is a fixed point of @xmath59 , i.e. , @xmath60 .
in fact , by property ( iii ) of definition [ d : md ] and ( [ e : mc ] ) , we have ( note that @xmath504 ) : @xmath505 \!\!&\le\!\!&w_{\lambda_0}(x_*,x_n)+w_{\lambda_0}(x_*,x_{n+1 } ) \to0\quad\mbox{as}\quad n\to\infty,\nonumber \end{aligned}\ ] ] and so , @xmath506 . by the strictness of @xmath48 , @xmath60 . finally , assuming @xmath48 to be finite valued on @xmath58 , we show that the fixed point of @xmath59 is unique .
suppose @xmath507 are such that @xmath60 and @xmath508 .
then the convexity of @xmath48 and inequalities @xmath509 and ( [ e : mc ] ) imply @xmath510 and since @xmath511 is finite , @xmath512 .
thus , @xmath513 , and by the strictness of @xmath48 , we get @xmath514 .
the last assertion is clear .
it is to be noted that assumption ( [ e : reach ] ) in theorem [ t : main ] is ( probably ) too strong , and what we actually need for the iterative procedure to work in the proof of theorem [ t : main ] is only the existence of an @xmath468 such that @xmath515 , where @xmath516 is the constant from ( [ e : mc ] ) .
a standard corollary of theorem [ t : main ] is as follows : if @xmath48 is finite valued on @xmath58 and an @xmath492-th iterate @xmath491 of @xmath54 satisfies the assumptions of theorem [ t : main ] , then @xmath59 has a unique fixed point .
in fact , by theorem [ t : main ] applied to @xmath491 , @xmath517 for some @xmath61 . since @xmath518 , the point @xmath519 is also a fixed point of @xmath491 , and so , the uniqueness of a fixed point of @xmath491 implies @xmath60 .
we infer that @xmath62 is a unique fixed point of @xmath59 : if @xmath520 and @xmath508 , then @xmath521 , i.e. , @xmath522 is yet another fixed point of @xmath491 , and again the uniqueness of a fixed point of @xmath491 yields @xmath523 .
another corollary of theorem [ t : main ] concerns general ( nonconvex ) modulars @xmath48 on @xmath11 ( cf . theorem [ t : maineconv ] ) . taking into account theorem [ t : lip2 ] and its corollary , we have [ d : mcnon ] given a modular @xmath48 on @xmath11 , a map @xmath54 is said to be _ strongly modular contractive _ ( or a _
strong @xmath48-contraction _ ) if there exist numbers @xmath55 and @xmath483 such that @xmath524 clearly , condition ( [ e : mcn ] ) implies condition ( [ e : mc ] ) .
[ t : maineconv ] let @xmath48 be a strict modular on @xmath11 such that @xmath58 is @xmath48-complete , and @xmath54 be a strongly @xmath48-contractive map such that condition ( [ e : reach ] ) holds . then @xmath59 admits a fixed point . if , in addition , @xmath48 is finite valued on @xmath58 , then ( [ e : reach ] ) is redundant , the fixed point @xmath62 of @xmath59 is unique and for each @xmath468 the sequence of iterates @xmath469 is modular convergent to @xmath62 .
we set @xmath525 for all @xmath13 and @xmath4 .
it was observed in section [ s : mm ] that @xmath120 is a convex modular on @xmath11 .
it is also clear that @xmath120 is strict and the modular space @xmath127 is @xmath120-complete .
moreover , condition ( [ e : mcn ] ) for @xmath48 implies condition ( [ e : mc ] ) for @xmath120 , and ( [ e : reach ] ) is satisfied with @xmath48 replaced by @xmath120 . by theorem [ t : main ] , applied to @xmath11 and @xmath120
, there exists an @xmath526 such that @xmath60 .
the remaining assertions are obvious .
in this section we present a rather standard application of theorem [ t : main ] to the carathodory - type ordinary differential equations .
the key interest will be in obtaining the inequality ( [ e : mc ] ) . given a convex @xmath250-function @xmath250 on @xmath292 satisfying the orlicz condition at infinity , we denote by @xmath527 $ ] the orlicz space of real valued functions on @xmath288 $ ] ( cf .
* chapter ii ) ) , i.e. , a function @xmath528\to{\mathbb{r}}$ ] ( or an almost everywhere finite valued function @xmath19 on @xmath288 $ ] ) belongs to @xmath527 $ ] provided @xmath19 is measurable and @xmath529 for some number @xmath530 , where @xmath531 is the classical orlicz modular .
suppose @xmath532\times{\mathbb{r}}\to{\mathbb{r}}$ ] is a ( carathodory - type ) function , which satisfies the following two conditions : ( c.1 ) for each @xmath533 the function @xmath534 $ ] is measurable on @xmath288 $ ] and there exists a point @xmath535 such that @xmath536 $ ] ; ( c.2 ) there exists a constant @xmath537 such that @xmath538 for almost all @xmath303 $ ] and all @xmath539 .
given @xmath358 , we let @xmath58 be the modular space ( [ e : xax ] ) generated by the modular @xmath48 from ( [ e : wlaxy ] ) under the assumptions from section [ s : exas].4 . consider the following integral operator @xmath540.\ ] ] [ t : meniq ] under the assumptions ( c.1 ) and ( c.2 ) the operator @xmath59 maps @xmath58 into itself , and the following inequality holds in @xmath83 $ ] : @xmath541 we will apply the _ jensen integral inequality _ with the convex @xmath250-function @xmath250 ( e.g. , ( * ? ? ? * x.5.6 ) ) several times : @xmath542,\ ] ] where the intergral in the right hand side is well defined in the sense that it takes values in @xmath83 $ ] .
first , we show that @xmath59 is well defined on @xmath58 .
let @xmath173 , i.e. , @xmath352 $ ] and @xmath324 .
since ( cf .
section [ s : exas].4 ) @xmath355 $ ] , by virtue of ( c.1 ) and ( c.2 ) , the composed function @xmath543 is measurable on @xmath288 $ ] .
let us prove that this function belongs to @xmath350 $ ] . by lebesgue s theorem , @xmath544 for all
@xmath303 $ ] , and so , ( c.2 ) yields @xmath545 for almost all @xmath303 $ ] . since @xmath173 , and so ,
@xmath352 $ ] , there exists a constant @xmath546 such that ( cf .
( [ e : intab ] ) ) @xmath547 and since , by ( c.1 ) , @xmath548 $ ] , there exists a constant such that @xmath549 setting @xmath550 and noting that @xmath551 by the convexity of @xmath250 , we find ( see ( [ e:6.3 ] ) ) @xmath552\biggr)\le\nonumber\\[4pt ] & \le\!\!&\frac{l\lambda_1(b\!-\!a)}{\lambda_0}\,{\varphi}\biggl ( \frac1{b\!-\!a}\int_a^b\!\frac{|x'(s)|}{\lambda_1}\,ds\biggr)\!+\ ! \frac1{\lambda_0}\,{\varphi}\bigl(l|x_0\!-\!y_0|\bigr)\!+\ !
\frac{\lambda_2}{\lambda_0}\,{\varphi}\biggl(\frac{|f(\cdot , y_0)|}{\lambda_2}\biggr ) , \nonumber \end{aligned}\ ] ] and so , ( [ e:6.3 ] ) and jensen s integral inequality yield @xmath553 now , it follows from ( [ e : jensen ] ) that @xmath554 implying @xmath555 thus , @xmath556\in\mbox{l}^1[a , b]$ ] . as a consequence ,
the operator @xmath59 is well defined on @xmath58 and , by ( [ e : intop ] ) , @xmath557 $ ] for all @xmath173 , which implies that the almost everywhere derivative @xmath558 belongs to @xmath350 $ ] and satisfies @xmath559.\ ] ] \2 .
it is clear from ( [ e : intop ] ) that , given @xmath173 , @xmath560 , and so , @xmath561\to{\mathbb{r}}\mid y(a)=x_0\}$ ] .
now we show that @xmath562 .
in fact , by virtue of ( [ e : intab ] ) , ( [ e : aetxt ] ) and ( [ e : abfa ] ) , we have @xmath563 and so , @xmath59 maps @xmath58 into itself . \3 . in order to obtain inequality ( [ e : lba ] ) , let @xmath13 and @xmath56 . taking into account ( [ e : x - y ] ) , ( [ e : intab ] ) and ( [ e : aetxt ] ) , we find @xmath564 applying ( c.2 ) and lebesgue s theorem , we get , for almost all @xmath303 $ ] ( note that @xmath320 ) , @xmath565 and so , by ( [ e : jensen ] ) , the monotonicity of @xmath250 , ( [ e : intab ] ) and ( [ e : x - y ] ) , @xmath566 now , inequality ( [ e : lba ] ) follows from ( [ e : wl ] ) . as a corollary of theorems [ t : main ] and [ t : meniq ] , we have [ t : cde ] under the conditions ( c.1 ) and ( c.2 ) , given @xmath358 , the initial value problem @xmath567 $ \,\,and \,\,$x(a)=x_0$}\ ] ] admits a solution @xmath568 $ ] with @xmath569 such that @xmath570 . we know
from section [ s : exas].4 that @xmath48 is a strict convex modular on the set @xmath571\to{\mathbb{r}}\mid x(a)=x_0\}$ ] and that the modular space @xmath572\cap x$ ] is @xmath48-complete . by theorem [ t : meniq ] , the operator @xmath59 from ( [ e : intop ] ) maps @xmath58 into itself and is @xmath48-contractive .
since the inequality @xmath57 with @xmath573 holds for all @xmath13 , in the iterative procedure in the proof of theorem [ t : main ] it suffices to choose any @xmath468 such that @xmath574 for some @xmath575 .
since @xmath576 , by virtue of ( [ e : txx ] ) and ( [ e : abfa ] ) , we find @xmath577 ( the constants @xmath578 and @xmath579 being evaluated on the interval @xmath580 $ ] ) with @xmath581 , and so , we may set @xmath582 .
now , by theorem [ t : main ] , the integral operator @xmath59 admits a fixed point : the equality @xmath583 on @xmath580 $ ] for some @xmath173 is , by virtue of ( [ e : intop ] ) and ( [ e : aetxt ] ) , equivalent to ( [ e : ivp ] ) .
it is not our intention in this paper to study the properties of solutions to ( [ e : ivp ] ) in detail : after theorem [ t : cde ] on local solutions of ( [ e : ivp ] ) has been established , the questions of uniqueness , extensions , etc .
, of solutions can be studied following the same pattern as in , e.g. , @xcite .
theorems [ t : meniq ] and [ t : cde ] are valid ( with the same proofs ) for mappings @xmath291\to m$ ] and @xmath532\times m\to m$ ] satisfying ( c.1 ) and ( c.2 ) , where @xmath271 is a reflexive banach space ; the details concerning the equality ( [ e : intab ] ) in this case can be found in @xcite@xcite . in the theory of the carathodory differential equations ( [ e : ivp ] )
@xcite ) the usual assumption on the right hand side is of the form @xmath584 for almost all @xmath303 $ ] and all @xmath533 , where @xmath585 $ ] , and the resulting solution belongs to @xmath586 $ ] for some @xmath587 .
however , it is known from ( * ? ? ? *
ii.8 ) that @xmath350=\bigcup_{{\varphi}\in\mathcal{n}}\mbox{l}^{\varphi}[a , b]$ ] , where @xmath588 is the set of all satisfying the orlicz condition at infinity . also , it follows from ( * ? ? ?
* corollary 11 ) that @xmath349=\bigcup_{{\varphi}\in\mathcal{n}}{\mathop{\rm gv}_{\!{\varphi}}\nolimits}[a , b]$ ] .
thus , theorem [ t : cde ] reflects the _ regularity _ property of solutions of ( [ e : ivp ] ) .
note that , in contrast with functions from @xmath349 $ ] , functions @xmath6 from @xmath351 $ ] have the `` qualified '' modulus of continuity ( ( * ? ? ?
* lemma 3.9(a ) ) ) : @xmath589 for all @xmath315 $ ] , where @xmath590 and @xmath591 is a subadditive function given by @xmath592 for @xmath139 and @xmath593
. theorem [ t : meniq ] does not reflect all the flavour of theorem [ t : main ] , namely , the _ locality _ of condition ( [ e : mc ] ) and the _ modular convergence _ of the successive approximations of the fixed points , and so , an appropriate example is yet to be found ; however , one may try to adjust example 2.15 from @xcite ( note that proposition 2.14 from @xcite is similar to our assertion ( [ e : corlip ] ) with @xmath594 ) .
a.ait taleb , e.hanebaly , a fixed point theorem and its application to integral equations in modular function spaces .
128(2 ) ( 2000 ) 419426 .
mr1487352 ( 99j:47082 ) v.v.chistyakov , mappings of bounded variation with values in a metric space : generalizations , pontryagin conference , 2 , nonsmooth analysis and optimization ( moscow , 1998 ) .
j. math .
( new york ) 100(6 ) ( 2000 ) 27002715 .
mr1778990 ( 2001k:26012 ) v.v.chistyakov , on multi - valued mappings of finite generalized variation , mat .
zametki 71(4 ) ( 2002 ) 611632 ( in russian ) .
english translation : math . notes 71(34 ) ( 2002 ) 556575 .
mr1913590 ( 2003m:47093 ) m.a.khamsi , w.m.kozowski , s.reich , fixed point theory in modular function spaces .
nonlinear anal .
14(11 ) ( 1990 ) 935953 .
mr1058415 ( 91d:47042 ) m.a.khamsi , w.m.kozowski , shu tao chen , some geometrical properties and fixed point theorems in orlicz spaces .
j. math .
. 155(2 ) ( 1991 ) 393412 .
mr1097290 ( 92b:47092 ) w.a.kirk , b.sims ( eds . ) , handbook of metric fixed point theory , kluwer acad .
dordrecht , 2001 .
mr1904271 ( 2003b:47002 ) m.a.krasnoselski , ja.b.ruticki , convex functions and orlicz spaces , fizmatgiz , moscow , 1958 ( in russian ) .
english translation : p. noordhoff ltd . ,
groningen , the netherlands , 1961 . mr0126722 ( 23#a4016 ) | the notion of a ( metric ) _ modular _ on an arbitrary set and the corresponding _ modular space _ , more general than a metric space , were introduced and studied recently by the author [ v.v .
chistyakov , metric modulars and their application , dokl .
math .
73(1 ) ( 2006 ) 3235 , and modular metric spaces , i : basic concepts , nonlinear anal .
72(1 ) ( 2010 ) 114 ] .
in this paper we establish a fixed point theorem for contractive maps in modular spaces .
it is related to contracting rather `` generalized average velocities '' than metric distances , and the successive approximations of fixed points converge to the fixed points in a weaker sense as compared to the metric convergence .
fixed point , metric modular , modular space , convex modular , modular convergence , modular completeness , modular contraction , mappings of bounded generalized @xmath0-variation , @xmath1-condition primary : 47h10 , 46a80 , secondary : 47h09 , 54e35 , 26a45 |
[ [ ams - subject - classification - msc2010 ] ] ams subject classification ( msc2010 ) the use of computer - based tests in which questions are randomly generated in some way provides a means whereby a large number of different tests can be generated ; many universities currently use such tests as part of the student assessment process . in this paper
we present findings that illustrate that , although the number of different possible tests is high and grows very rapidly as the number of alternatives for each question increases , the average number of tests that need to be generated before all possible questions have appeared at least once is surprisingly low .
we presented preliminary findings along these lines in @xcite .
a computer - based test consists of @xmath0 questions , each ( independently ) selected at random from a separate bank of @xmath1 alternatives .
let @xmath2 be the number of tests one needs to generate in order to see all the @xmath3 questions in the @xmath0 question banks at least once .
we are interested in how , for fixed @xmath1 , the random variable @xmath2 grows with the number of questions @xmath0 in the test .
typically , @xmath1 might be 10i.e
. each question might have a bank of 10 alternatives but we shall allow any value of @xmath1 , and give numerical results for @xmath4 and @xmath5 as well as for @xmath6 . in the case
@xmath7 , i.e. a one - question test , we re - notate @xmath2 as @xmath8 , and observe that we have an equivalent to the classic coupon - collector problem : your favourite cereal has a coupon in each packet , and there are @xmath1 alternative types of coupon .
@xmath8 is the number of packets you have to buy in order to get at least one coupon of each of the @xmath1 types .
the coupon - collector problem has been much studied ; see e.g. @xcite .
we can write @xmath8 as @xmath9 where each @xmath10 is the number of cereal packets you must buy in order to acquire a new type of coupon , when you already have @xmath11 types in your collection .
thus @xmath12 , @xmath13 is the number of further packets you find you need to gain a second type , and so on .
the random variables @xmath14 , , @xmath15 are mutually independent . for the distribution of @xmath16 , clearly @xmath17 we say that @xmath18 , or @xmath19 has a geometric distribution with parameter @xmath20 , if @xmath21 for @xmath22 , 2 , .
thus @xmath23 . as the @xmath24 distribution has expectation @xmath25 it follows that @xmath26 for different values of @xmath1 we therefore have the following .
the amount of variability can be better appreciated through the standard deviation .
the asymptotic bounds on the standard deviation of @xmath2 are @xmath27 and some values for these are in table [ ta : sd ] .
the lower bound is non - trivial , i.e. positive , in each case .
we are grateful to dave pidcock , a colleague in the mathematics education centre at loughborough university , for raising the query in the first place . as a member of staff using computer - based tests to assess students , he was concerned about this issue from a practical viewpoint . that led rc to post a query on allstat .
cmg was not the only person to respond to the query , and we also acknowledge the others who responded , particularly simon bond .
cornish , r. , goldie , c. m. , and robinson , c. l. 2006 .
computer - assisted assessment : how many questions are enough ? _ computer - aided assessment in mathematics _
, 9pp . ; http://mathstore.ac.uk/articles/maths-caa-series/feb2006 . | computer - based tests with randomly generated questions allow a large number of different tests to be generated . given a fixed number of alternatives for each question ,
the number of tests that need to be generated before all possible questions have appeared is surprisingly low . |
cellular automata are statistical - mechanical models which present a complex behavior , despite their relatively simple dynamic rules . in general
these models are defined in @xmath4-dimensional lattices with linear dimension @xmath5 , in which each site @xmath6 @xmath7 is occupied by a variable @xmath8 . the dynamic rules of a typical cellular automaton have two basic characteristics : firstly , the update of a variable @xmath9 depends only on its neighborhood ( short - range interactions ) . secondly , given a configuration of the system at instant @xmath2 , at instant @xmath10 all variables @xmath9 @xmath11
are updated at once .
the use of cellular automata in order to mimic biological immune systems has improved the understanding about the microscopic mechanisms that lead to the macroscopic behavior of these systems @xcite .
for example , cellular automata are as usefull to describe the t - helper cells response under parasitic infections brass94,brass94a as they are to model the course of the human immunodeficiency virus ( hiv ) infection @xcite .
brass _ et al .
_ divised a simple cellular automaton model of @xmath12 cell interactions to mimic the immune system of mice exposed to parasitic infections @xcite .
the model is defined in a cubic ( @xmath13 ) lattice in which each site is occupied by one of three different @xmath14 cell types : @xmath15 , @xmath16 and @xmath17 cells .
t - helper cells that have not yet been presented to the antigen are denoted by @xmath18 . in the model , two distinct routes ( equivalent to two populations of antigen presenting cells )
govern the maturing of a @xmath18 cell : antigen presented by the first ( second ) route elicit a @xmath19 ( @xmath20 ) response . in order to take into account the competition between
mature @xmath14 cells , the induction @xmath21 ( @xmath22 ) is forbidden when the neighborhood of the @xmath15 cell has an overall majority of @xmath17 ( @xmath23 ) cells .
at last , a mature cell dies ( it is substituted by a @xmath24 cell ) if such a cell is not restimulated by the appropriate antigen within the time interval @xmath25 . the model can display a spontaneous symmetry breaking as one varies the antigen density or the cutoff @xmath25 , in agreement with experimental results @xcite .
a probabilistic version of the model was proposed by tom and drugowich de felcio @xcite by allowing the death of cells @xmath19 and @xmath20 at each time step with a probability @xmath26 .
thus , in this modified version of the model , the lifetime @xmath25 was substituted by a mean lifetime related to the probability @xmath26 .
also , the development of @xmath24 cells into either @xmath27 or @xmath28 cells occurs with a probability that depends on the neghborhood of the @xmath29 cell and on a parameter @xmath30 , related to the antigen density .
such probabilistic version presents the intrinsic spontaneous symmetry breaking found in the brass _ et al .
_ automaton , besides being amenable to an analytical approach @xcite .
the critical behavior of the probabilistic model proposed in ref .
@xcite was studied in a subsequent work by ortega _
_ @xcite . by using a finite - size scaling analysis , from monte carlo simulations of square lattices , ortega _
et al . _ determined the ratios of exponents @xmath31 and @xmath32 , as well as the critical point of the model .
their results suggest that the probabilistic automaton , although not satisfying detailed balance , belongs to the same universality class of the two - dimensional kinetic ising model @xcite , thus supporting the _ up - down _ conjecture , introduced by grinstein _
_ grinstein85 . in a subsequent work , tom and
drugowich @xcite obtained the dynamic critical exponent @xmath0 of the probabilistic automaton in two dimensions , by performing short - time monte carlo simulations by studying the collapse of the fourth order binder s cumulant for different lattice sizes . in the present work
we have revisited the probabilistic automaton of ref .
we have re - obtained the dynamic critical exponent @xmath0 of the automaton by using a recent technique that mixes different initial conditions @xcite .
such technique is based upon the short - time critical dynamics introduced by janssen _
_ @xcite , who showed that even far from equilibrium the short - time relaxation of the order parameter follows a universal scale form @xmath33where @xmath34 is the order parameter at instant @xmath2 ( measured in monte carlo steps per spin - mcs ) , @xmath35 is a small initial @xmath36 value of the order parameter , and @xmath37 is the dynamic critical exponent , related to the increasing of the order parameter after the quenching of the system ( when dealing with monte carlo simulations , the time @xmath2 is discrete and measured in monte carlo steps per spin - mcs ) .
( [ m0tatheta ] ) demands working with sharply prepared initial states with a precise value of @xmath38 . after obtaining the critical exponent @xmath37 for a number of @xmath38 values , the final value for @xmath37 is obtained from the limit @xmath39 .
starting from an ordered state @xmath40 , the order parameter @xmath34 decays in time , at the critical temperature , according to the power law zheng98 @xmath41 where @xmath42 is the average of the quantity @xmath43 over different samples with initial order parameter value @xmath44 , @xmath45 and @xmath46 are the usual static critical exponents , related to the order parameter and to the correlation length , respectively , and @xmath0 is the dynamic critical exponent , defined as @xmath47 , where @xmath48 and @xmath49 are time and spatial correlation lengths , respectively . starting from a disordered configuration with @xmath50 , the second moment of the order parameter increases after the power law @xmath51 where the exponent @xmath52 is given by @xmath53 and @xmath4 is the dimension of the system . by combining eqs .
( [ bdnz ] ) and ( [ mquad ] ) , da silva _ et al . _
@xcite obtained the ratio @xmath54 which corresponds to a function with mixed initial conditions . at this point , it is important to stress here that , from eq .
( [ f2 ] ) above , two independent runs are necessary in order to calculate the ratio @xmath55 : in one of them @xmath50 ( numerator ) , while in the other one @xmath56 ( denominator ) .
the ratio @xmath55 has proven to be useful in determining the exponent @xmath0 , according to recent studies of the 2d ising model @xcite , @xmath57 and @xmath58 states potts models @xcite , ising model with next - nearest - neighbor interactions @xcite , baxter - wu model arashiro03 , at the tricritical point of the 2d blume - capel model dasilva02a , at the lifshitz point of the 3d annni model @xcite and in other studies concerning models with one absorbent state dasilva04 .
in addition , in this work we have also calculated the global persistence exponent @xmath59 @xcite , related to the probability that the order parameter of the model , @xmath34 , does not change its sign up to time @xmath2 , after a quench of the system to the critical temperature .
the layout of this paper is as follows : in section [ model ] we explain the model . in section [ opdcet ]
we define the order parameter @xmath34 and we describe the methodology used in order to obtain the dynamic critical exponents @xmath59 and @xmath0 . in section [ results ] the results obtained for the exponents @xmath0 and @xmath59 are shown
. finally , in section conclusion we present the main concluding remarks .
in this work we have studied a two - dimensional probabilistic cellular automaton in which dynamics is governed by local stochastic rules . at each site @xmath6 of the square lattice
we have attached a variable @xmath60 assuming the value @xmath61 or @xmath62 , depending on whether the site is occupied by a @xmath18 , a @xmath16 or a @xmath20 cell , respectively .
considering @xmath63 the total number of sites in the lattice , we have defined the set @xmath64 to represent the microscopic state of the system .
the probability of state @xmath65 at time @xmath2 , @xmath66 , evolves in time according to the equation @xmath67 where the transition probability @xmath68 from state @xmath69 to state @xmath70 obeys the condition @xmath71 on the other hand , for a system that evolves at discrete time steps , in which all the sites are updated at once , as is the case for the cellular automaton in this work , the transition probability @xmath72 is written in the form @xmath73 where @xmath74 is the conditional probability that site @xmath6 be in the state @xmath75 at time @xmath10 , given that the state of the system is @xmath65 at instant @xmath2 .
such conditional probability satisfies the condition @xmath76 what implies immediately that eq .
( [ soma1 ] ) is fulfilled .
the cellular automaton investigated in this work belongs to the class of totalistic cellular automata @xcite .
thus , the transition probability @xmath77 depends on @xmath60 and on the sum @xmath78 , where the sum runs over the neighborhood of site @xmath6 .
more specifically , we are considering a particular kind of totalistic automaton for which @xmath74 depends only on the sign of the sum @xmath49 . by defining
@xmath79 we may use the notation @xmath80 in order to explicit the transition probability dependence on @xmath60 and @xmath81 .
the transition probabilities ( dynamical rules ) are given by @xmath82 @xmath83 @xmath84 where , as discussed in section [ intro ] , @xmath26 is the death probability of @xmath85 and @xmath28 cells , and @xmath30 is a parameter related to the antigen density @xcite . the dynamical rules ( [ dinamica1 ] ) , ( [ dinamica2 ] ) and ( [ dinamica3 ] ) above have `` up - down '' , _
i.e. _ , @xmath86 therefore , following grinstein _
@xcite , we expect that the probabilistic cellular automaton investigated in this work be at the same universality class of kinetic ising models .
in the short - time monte carlo simulations performed in this work the order parameter @xmath34 , as well as its higher moments @xmath88 , were obtained from averages over a certain number of samples , @xmath89 . by defining the @xmath90th momentum of the order parameter in sample number @xmath91 at instant @xmath2 as @xmath92
the order parameter is written in the form @xmath93 as defined in eqs .
( [ mag_1_amostra ] ) and ( [ magmedia ] ) above , for @xmath94 the order parameter is exactly the mean magnetization , @xmath95 .
( [ magmedia ] ) defines the order parameter and its higher moments for a set of @xmath89 samples .
thus , if @xmath96 sets of samples are considered at instant @xmath2 , there are @xmath96 measurements of the magnetization ( and its higher moments ) , @xmath97 , where @xmath98 . by considering such sets of samples
, the final value of the magnetization and its higher moments are obtained from the average over the @xmath96 sets of samples , _
i.e. _ , @xmath99 where @xmath97 is given by eq .
( [ magmedia ] ) and the corresponding standard deviation is given by @xmath100 = \frac{1}{\sqrt{n_{b}(n_{b}-1)^{2 } } } \left ( \sum\limits_{l=1}^{n_{b}}\left ( m_{l}^{(k)}(t)-\overline{m^{(k ) } } ( t)\right ) ^{2}\right ) ^{1/2}. \label{desvio_padrao_media}\ ] ] in this section we define the dynamic critical exponent @xmath59 and we describe two methods used in this work to obtain estimates for @xmath59 .
in the first method , we have performed short - time monte carlo simulations in order to calculate the global persistence probability @xmath101 , _ i.e. _ , the probability that the magnetization does not change its sign up to time @xmath2 . on the other hand ,
the probability @xmath101 is numerically equal to the complement of the accumulated distribution @xmath102 , according to which the magnetization changes its sign for the first time exactly at instant @xmath2 , _
i.e. _ , @xmath103 where @xmath104 is the number of samples for which the magnetization changes its sign for the first time at instant @xmath105 and @xmath89 is the total number of samples .
the exponent @xmath59 may be obtained directly from the power law scale relation @xcite @xmath106 from which we obtain @xmath107 , where @xmath108 is constant and each run requires a sharply prepared initial state , with a precise small value of @xmath109 , as discussed in eq .
( [ m0tatheta ] ) . in a second way to obtain the exponent @xmath59 we have used the fact that the initial magnetization dependence of @xmath101 can be cast in the following finite - size scaling relation @xcite , @xmath110 which renders a different method to obtain the exponent @xmath1 from lattice sizes @xmath111 and @xmath112 @xcite . for this end
we define @xmath113 , which turns out to be a function of @xmath114 . therefore ,
if we fix the dynamic exponent @xmath0 , the exponent @xmath115 can be obtained by collapsing the time series @xmath116 onto @xmath117 as follows . under re - scaling , with @xmath118 , @xmath119 , we obtain @xmath120 and the best estimate for @xmath1 corresponds to the minimization of @xmath121 by interpolating @xmath122 to the time values @xmath123 . in order to obtain the exponent @xmath59 using the collapse method described above , it is not necessary to fix a precise value of the initial magnetization @xmath109 in the short - time simulations , once the scaling relation in eq .
( leiescalapersisti ) does not take into account the initial conditions of the system .
so , we have used initial states in which @xmath124 . on the other hand ,
the collapse method demands the dynamic exponent @xmath0 to be known beforehand .
therefore , in this work we have used the scaling relation of eq .
( [ f2 ] ) in order to obtain estimates for the exponent @xmath0 .
although both methods described by eqs .
( powerlawthetag ) and ( [ leiescalapersisti ] ) were proposed in order to calculate the exponent @xmath59 of the two - dimensional ising model majumdar96 , such methods were used recently for estimates of @xmath59 along the critical line and at the tricritical point of the 2d blume - capel model @xcite .
in this section we present details about the short - time monte carlo simulations performed for the cellular automaton considered in this work , as well as the results obtained for both dynamic critical exponents @xmath0 and @xmath125 .
initially we refine the critical parameter @xmath126 obtained in ref .
tome96 ) for @xmath127 . from eq . ( [ bdnz ] )
we expected a straight line for the log - log plot of @xmath34 against @xmath2 at the critical point @xmath128 .
however , from log - log plots of eq .
( [ bdnz ] ) obtained for different values of @xmath26 , we observed that more accurate straight lines are obtained for @xmath129 , as depicted in fig .
[ fig1 ] for @xmath130 and @xmath131 . in the short - time simulations performed in order to obtain the curves shown in fig [ fig1 ] we have used square lattices @xmath132 with linear dimensions @xmath133 , @xmath134 samples , @xmath135 sets of samples and 1000 mcs . for each value of @xmath26 used in fig .
[ fig1 ] we calculated the goodness of fit @xmath136 for the same time interval @xmath137 $ ] and obtained the values shown in table [ table1 ] , from which we obtain the critical value @xmath138 .
.values of @xmath136 for different values of @xmath26 obtained from linear fitting of log - log plots of the magnetization @xmath34 versus @xmath2 . [
cols="^,^,^",options="header " , ] from table [ table3 ] above , the best estimate for the global persistence exponent is @xmath139 , corresponding to @xmath140 .
this result is in very good agreement with the estimates @xmath141 and @xmath142 obtained from the collapse method .
in this work we have revisited the probabilistic cellular automata proposed by tom and drugowich @xcite on the basis of a previous cellular automaton devised by brass _ et al .
_ @xcite . from short - time monte carlo simulations , we have obtained the dynamic critical exponent @xmath143 by using a recent technique that mixes different initial conditions @xcite .
such result is slightly smaller than the value @xmath144 obtained from the fourth order binder s cumulant @xcite .
we have also performed short - time monte carlo simulations in order to obtain estimates for the dynamic critical exponent @xmath59 , the global persistence exponent , by using two distinct approaches : the collapse method described by eqs .
( [ leiescalapersisti ] ) , ( [ leiescalapersistii ] ) and ( [ desvquadmed ] ) , and directly from the power law scaling given by eq .
( [ powerlawthetag ] ) . from the collapse method , by using square lattices with linear dimensions @xmath145 and @xmath146 , we have obtained @xmath141 and @xmath147 , respectively . directly from eq .
( [ powerlawthetag ] ) we have obtained @xmath139 , in very good agreement with the estimates yielded from the collapse method .
n. alves jr . acknowledges financial support from brazilian agencies fapesp and capes .
r. da silva thanks gppd of the institute of informatics of federal university of rio grande do sul ( ufrgs ) , for computational resources .
k. j. else , g. m. entwistle , and r. k. grencis .
correlations between worm burden and markers of th1 and th2 cell subset induction in an inbred strain of mouse infected with trichuris - muris . , 15(10):595600 , oct 1993 .
r. da silva , n. a. alves , and j. r. drugowich de felcio .
universality and scaling study of the critical behavior of the two - dimensional blume - capel model in short - time dynamics .
, 66(2):026130 , aug 2002 . | in this work a three - states cellular automaton proposed to describe part of a biological immune system @xcite is revisited .
we obtain the dynamic critical exponent @xmath0 of the model by means of a recent technique that mixes different initial conditions . moreover , by using two distinct approaches , we have also calculated the global persistence exponent @xmath1 , related to the probability that the order parameter of the model does not change its sign up to time @xmath2 [ @xmath3 . |
in this section we give a brief overview of the interaction between the light and spin atomic ensembles .
we derive simple input - output equations describing the evolution of the light - oscillator system and relate the coupling constant to experimental parameters . in the limit of low saturation parameter ( @xmath105 , where @xmath106 is the rabi frequency , @xmath107 the probe detuning from the atomic resonance and @xmath108 is the fwhm homogeneous broadening of the electronic transition ) the coherent interaction of light with atomic spin
can be written in the form of a polarizability hamiltonian @xcite ( throughout this letter we take @xmath109 , with @xmath110 being the speed of light ) : @xmath111 \bigg\ } , \label{eq : hamiltonian0}\ ] ] where @xmath112 the wavelength of atomic transition , @xmath113 is the cross - sectional area of interaction , @xmath114 is the photon flux , @xmath115 are the components of the dimensionless atomic spin ( sum of nuclear and electron spin ) , @xmath116 are the detuning dependent scalar , vector and second - rank tensor polarizabilities respectively . for the cesium d2 transition ( @xmath117 ) from the @xmath77 hyperfine manifold of the ground state , the polarizabilities are given by ( see fig .
[ fig : cslevels ] ) : @xmath118 where @xmath119 is the frequency spacing between the @xmath120 and the @xmath121 excited states and @xmath107 is the light detuning from the @xmath122 transition ( @xmath107 is negative for blue detuning ) . in eq .
[ eq : hamiltonian0 ] the operators @xmath123 are the polarization stokes components of light and can be described in terms of creation ( annihilation ) operators of the light field in the transverse polarizations .
assuming light propagation along the @xmath124 direction : @xmath125 where the raising and lowering field operators @xmath126 , @xmath127 are defined so that in the vacuum state @xmath128=\delta ( t - t')$ ] and @xmath129 is the flux of photons at time @xmath130 . the operator @xmath34 measures the light polarization in the circular basis , whereas @xmath131 measure the polarization in complementary linear basis and are normalized so that for a coherent field linearly polarized in the @xmath132 direction : @xmath133 and @xmath134 . in describing polarization spectroscopy ,
the scalar part of the hamiltonian can be neglected , since it does not affect the evolution of atomic spin and the stokes operators . for detunings much larger than the hyperfine splitting of the excited state @xmath135 : @xmath136 . when the light interacts with an ensemble of @xmath25 atoms , all prepared in the same state and coupled with the same strength to the light mode , the coherent interaction of a light pulse of duration @xmath33 with the ensemble
can be described in the limit of large detuning with the hamiltonian : @xmath137 where @xmath138 and @xmath19 refer to the ensemble collective spin components and @xmath139 is the number of photons in the pulse . in the @xmath140 ( d2 line ) , whereas the pump and repump light are on resonance with the d1 ( @xmath141 ) and d2 line respectively .
( b ) the cs atomic ensemble is contained in a glass microchannel 300 @xmath142 m @xmath143 300 @xmath142 m in cross section and 10 mm in length . ]
room - temperature atomic ensembles may introduce two complications concerning the interaction hamiltonian : doppler broadening and thermal motion in and out of the beam .
for the experiment of this work , the detuning ( @xmath144 mhz ) was much larger than the doppler width ( @xmath145 mhz ) , so that the doppler broadening has a small effect on the observed noise ; this can also be seen from the fact that for a detuning much larger than the doppler broadening the voigt profile of the atomic line approximates a lorentzian . similarly , for the experimentally relevant time scales of the atomic state evolution the thermally moving atoms cross the beam many times and the effect of motion averages out @xcite . in this case , @xmath113 in eq .
[ eq : hamiltoniankappa]-[eq : kappa ] refers to the cross sectional area of the atomic cloud . in the following , we will assume coherent linearly polarized light field in the @xmath132 direction probing a highly polarized atomic ensemble experiencing a dc magnetic field @xmath11 along the @xmath132 direction .
for highly polarized atomic ensembles , the transverse collective spin components can be mapped to canonical position and momentum variables of an oscillator : @xmath146 the ground state of the oscillator corresponds to the coherent spin state of maximum spin in the @xmath132 direction ; in this state @xmath147 . neglecting retardation effects in the light propagation and assuming without loss of generality @xmath148 , the hamiltonian evolution of @xmath149 and @xmath150 are described by the following equations @xcite : @xmath151 \hat{s}_z ( t ' ) , \label{eq : jzevol } \\
\hat{p}(t ) = & -\sin ( \omega t ) \hat{x}(0 ) + \cos ( \omega t ) \hat{p}(0 ) + \beta \sqrt{j_x } \int_0^t \ ! \mathrm{d}t ' \ , \cos \left [ \omega ( t - t ' ) \right ] \hat{s}_z ( t ' ) , \label{eq : jyevol}\end{aligned}\ ] ] where @xmath3 is the larmor angular frequency @xmath152 , with @xmath153 being the atomic gyromagnetic ratio and @xmath154 is a parameter defined by : @xmath155 due to the hamiltonian interaction with the atomic ensemble the @xmath47 light operator evolves as : @xmath156 where the ( in ) , ( out ) superscripts denote the input and output of the ensemble .
in this section , we will estimate the squeezing that can be realized using stroboscopic probing of the oscillator .
initially , decoherence will be neglected ( we will assume only hamiltonian dynamics ) ; the effect of decoherence will be considered at the end of the section . in the hamiltonian dynamics
, we will neglect the effect of second rank polarizability .
we consider the experimentally relevant case of linearly polarized probe beam in the @xmath157 direction , with stroboscopic intensity - modulation at twice the spin oscillator frequency ; the stokes element @xmath158 is detected and the fourier component @xmath159 is measured by a lock - in amplifier . in this way , the atomic - ensemble variable that is measured is @xmath160 integrated over one oscillator cycle and weighted with a cosine wave ( from the lock - in amplifier demodulation ) and a pulse - shaped function from the stroboscopic modulation of the probe : @xmath161 where @xmath162 is an integer number ( @xmath163 ) , @xmath164 is the oscillation period and the stroboscopic function @xmath43 with duty cycle @xmath42 is given by ( see fig . 2 ( b ) in @xcite ) : @xmath165 the variable @xmath43 is defined so that the overlap with the lock - in cosine quadrature is maximized .
the position variable @xmath166 can be defined in the same way as in eq .
[ eq : defx ] by replacing @xmath73 with @xmath167 .
the squeezing @xmath168 realized after a measurement time @xmath169 , with @xmath170 being the number of complete oscillation cycles in the measurement , is given by : @xmath171}{\text{var } \big [ \hat{x}\big]_0 } , \label{eq : xi0}\ ] ] where @xmath172_0 $ ] is the noise in the ground state of the oscillator ( coherent spin state of maximum spin component along the @xmath132 direction ) and @xmath173 is the polarimetry measurement record .
for the experiment of this work ( large number of atoms and photons ) , @xmath174 and @xmath173 can be considered as gaussian variables .
therefore , the conditional variance in eq . can be written as @xcite : @xmath175 = \text{var } \big [ \hat{x } ( n_{\text{m } } t ) \big]-\frac{\text{cov}^2 \big [ \hat{x } ( n_{\text{m } } t ) , \hat{s}_{y,\tau}^{\text{out } } \big]}{\text{var } \big [ \hat{s}_{y,\tau}^{\text{out } } \big ] } \phantom{a}. \label{eq : condvargaussian}\ ] ] in the oscillator ground state : @xmath176 , with @xmath177 , so that : @xmath178_0 = \frac{\left [ 1+\text{sinc } ( \pi d ) \right]^2 } { 8 } \ , .
\label{eq : groundstatenoise}\ ] ] for simplicity , we will consider no weighting of the data with a mode function .
the measurement record : @xmath179 can be written in the form : @xmath180 , \label{eq : measrecord } \\
\hat{y}(k t ) & = \int_{k t}^{(k+1)t } \ !
\mathrm{d}t \ , \hat{s}_y^{\text{in}}(t ) \phi(t ) \cos ( \omega t ) \
, , \label{eq : ydef}\end{aligned}\ ] ] where the bar denotes average over the oscillator period .
the first sum in eq .
[ eq : measrecord ] leads to the photon shot noise , while the second sum has the information about the oscillator state . since there are no correlations between the light at the input of the ensemble ( @xmath181 ) and the oscillator : @xmath182 .
the correlation related to the photon shot noise of light is : @xmath183 \delta_{k_1 k_2 } = \langle \hat{y}^2 \rangle_0 \delta_{k_1 k_2 } \ , .
\label{eq : yycor}\ ] ] with @xmath184 being the kronecker delta and @xmath185 the average photon flux over the oscillator period .
the variable @xmath186 can be written as the sum of two uncorrelated contributions : @xmath187 , with @xmath188 .
the term @xmath189 is related to the initial ( before the interaction with the probe light ) quantum state of the oscillator , which is assumed to be the ground state : @xmath190 = \frac { \hat{x}_0 \left [ 1+\text{sinc } ( \pi d ) \right ] } { 2 } \ , .
\label{eq : noisetem1}\end{aligned}\ ] ] the term @xmath191 gives rise to the noise that describes the coupling of the quantum probe noise to the measured variable @xmath102 : @xmath192 \hat{s}_z(t ' ) .
\label{eq : ban1}\end{aligned}\ ] ] it can be shown that : @xmath193 \beta^2 j_x \bar{\phi } t } { 64 } \big [ 1-\text{sinc}(\pi d ) \big ] \big [ 1+\text{sinc}(\pi d ) \big]^2 , \label{eq : xcor2}\ ] ] where @xmath194 is a numerical factor of order unity . here , we will consider the case where we are measuring over many cycles , so that @xmath195 . in this case , the measurement variance is : @xmath196 & = n_{\text{m}}\langle \hat{y}^2 \rangle_0 + \frac{\beta^2 j_x \bar{\phi}^2 n_{\text{m}}^2 t^2}{4 } \text{var } \big [ \hat{x } \big]_0 + \frac{\beta^2 j_x \bar{\phi}^2 t^2}{4 } \displaystyle \sum_{k_1=0}^{n_{\text{m}}-1 } \sum_{k_2=0}^{n_{\text{m}}-1 } \langle \hat{x}_{\text{ba}}(k_1 t ) \hat{x}_{\text{ba}}(k_2 t ) \label{eq : varsy1a } \rangle , \\ & \approx \frac{\bar{\phi } \tau}{8 } \left [ 1+\text{sinc}(\pi d ) \right ] \left [ 1+\tilde{\kappa}^2+\frac{\tilde{\kappa}^4}{3 } \frac{1-\text{sinc}(\pi d ) } { 1+\text{sinc}(\pi d ) } \right ] , \label{eq : varsy1}\end{aligned}\ ] ] where the effective coupling constant is : @xmath197 .
\label{eq : k2sdef}\end{aligned}\ ] ] the covariance between the quantum observable at the end of the measurement ( @xmath198 ) and the measurement record ( @xmath173 ) is : @xmath199 & = \frac{\beta \sqrt{j_x } \bar{\phi } t } { 2 } \big [ n_\text{m}\text{var } \big [ \hat{x}_0 \big ] + \displaystyle \sum_{k=0}^{n_{\text{m}}-1 } \langle \hat{x}_{\text{ba } } ( n_{\text{m } } t ) \hat{x}_{\text{ba } } ( k t ) \rangle \big ] , \label{eq : covxsya } \\ & \approx \frac{\beta \sqrt{j_x } \bar{\phi } \tau}{16 } \left [ 1+\text{sinc}(\pi d ) \right]^2
\left [ 1 + \frac{\tilde{\kappa}^2}{2 } \frac { 1-\text{sinc } ( \pi d)}{1-\text{sinc } ( \pi d ) } \right ] \label{eq : covxsyb}.\end{aligned}\ ] ] the unconditional variance of @xmath198 is : @xmath200 & = \text{var } \big [ \hat{x}\big]_0 + \langle \hat{x}_{\text{ba } } ( n_{\text{m } } t ) \hat{x}_{\text{ba } } ( kn_{\text{m } } t ) \rangle , \label{eq : varxnta } \\ & \approx \text{var } \big [ \hat{x}\big]_0\left [ 1+\tilde{\kappa}^2 \frac { 1-\text{sinc } ( \pi d)}{1-\text{sinc } ( \pi d ) } \right ] .
\label{eq : varxnta}\end{aligned}\ ] ] from eqs .
[ eq : condvargaussian ] , [ eq : k2sdef ] , [ eq : covxsyb ] and [ eq : varxnta ] , we get : @xmath175 = \text{var } \big [ \hat{x}\big]_0 \bigg [ 1+\tilde{\kappa}^2 \frac { 1-\text{sinc } ( \pi d)}{1-\text{sinc } ( \pi d ) } -\frac{\tilde{\kappa}^2 \left [ 1 + \frac{\tilde{\kappa}^2}{2 } \frac { 1-\text{sinc } ( \pi d)}{1-\text{sinc } ( \pi d ) } \right]^2}{1+\tilde{\kappa}^2+\frac{\tilde{\kappa}^4}{3 } \frac{1-\text{sinc}(\pi d ) } { 1+\text{sinc}(\pi d ) } } \bigg ] \ , .
\label{eq : condvarxntsy}\ ] ] combining eqs .
[ eq : xi0 ] and [ eq : condvarxntsy ] we find that the squeezing is given by : @xmath201 ^ 2}{1+\tilde{\kappa}^2+\frac{\tilde{\kappa}^4}{3 } \frac{1-\text{sinc}(\pi d ) } { 1+\text{sinc}(\pi d ) } } .
\label{eq : squeezingformula0}\end{aligned}\ ] ] in the limit of zero duty cycle ( @xmath202 ) : @xmath203 .
so far , it has been assumed that the oscillator does not experience any decoherence .
the presence of decoherence results in reduction of the realized squeezing @xcite : @xmath204 where @xmath39 scales with the decoherence events during the measurement time @xmath33 and depends on the decay mechanism .
the decoherence can be linked to the light - probe , or it can be associated with coupling to a bath present even in the absence of the probe ( e.g. coupling to the thermal phonon bath of a mechanical oscillator , or spin decay in the dark for a spin oscillator ) . in @xcite
exact relations for the probe induced decoherence and associated noise have been derived for the case of a spin oscillator .
it is shown there that the probe induced noise can be written as : @xmath205 where @xmath206 is a numerical factor of order unity that depends on the probe polarization and the light - atom detuning , and @xmath207 is the optical depth on resonance , with @xmath208 being the atomic density , @xmath209 the absorption cross section on resonance and @xmath210 the length of the cell .
we consider a standing wave cavity with input ( output ) mirror power transmission coefficient @xmath211 ( @xmath212 ) , round - trip intracavity power loss @xmath65 and atomic single - pass absorption @xmath213 . for high finesse ( @xmath214 ) , the cavity power transmission coefficient on resonance is : @xmath215 , which in the limit of @xmath216 can be written as : @xmath217 $ ] , where @xmath218 is the cavity finesse .
the single pass atomic absorption @xmath213 is enhanced by the cavity by the factor @xmath219 . from kramers - kronig relations the atomic phase shift , and therefore the polarization rotation angle , is enhanced by the same factor .
effectively , the cavity modifies the input - output relations eq.[eq : jzevol]-[eq : syout0 ] by enhancing the light - matter coherent interaction by @xmath219 .
if @xmath220 is the coupling constant in the absence of cavity , then for the same number of detected photons in the pulse the cavity - enhanced coupling constant is : @xmath221 equation [ eq : squeezingformula0 ] for the conditional squeezing ( in the absence of decoherence ) should be modified accordingly . ,
@xmath222 , where the subscripts in ( out ) denote field in ( out in the detection side ) the cavity and @xmath223 is the power transmission coefficient of the output coupler .
the total power experienced by the atomic ensemble contributing to decoherence due to spontaneous emission is : @xmath224 . ]
the realized squeezing is limited by the noise due to decoherence events .
the probe - induced decoherence noise is proportional to the intracavity light - power experienced by the atomic ensemble @xmath225 , which is related to the power at the output of the cavity ( @xmath226 ) through : @xmath227 , where @xmath223 is the output mirror coupler power transmission ( see fig .
[ fig : cavityfields ] ) . for a standing wave cavity ,
maximum squeezing is achieved for @xmath228 , so that the quantum light field generated inside the cavity from the interaction with the atoms has very small losses to the unobserved direction . assuming only probe - induced decoherence and for @xmath229
, @xmath230 can be written in the form : @xmath231 where @xmath232 is the decoherence noise due to the light - probe in the absence of the cavity . for small but finite intracavity losses ,
optimizing with respect to the number of photons in the light pulse and the output coupler transmission , the squeezing is : @xmath233 \sqrt{\frac{\zeta}{(2 \mathscr{f } / \pi ) d_0 } } \approx \sqrt{\frac{\zeta}{(\mathscr{f } /2 \pi ) d_0 } } \phantom{a } , \label{eq : sqzoptimal}\ ] ] which is realized for @xmath234 . from the above equation
, it can be seen that the presence of the cavity effectively enhances the optical depth by @xmath235 .
equations [ eq : jzevol]-[eq : syout0 ] describe the oscillator - light system evolution in the limit of zero second - rank tensor polarizability : @xmath236 .
more complete equations , including the second - rank tensor polarizability effect , have been derived in @xcite .
it can be shown that the _ hamiltonian _ evolution of the spin - oscillator can be approximated by : @xmath237 s_y(t ' ) + \beta \sqrt{j_x } \int_0^t \ !
\mathrm{d}t ' \ , e^{-\gamma_{\text{sw}}(t - t ' ) } \sin \left [ \omega ( t - t')\right ] s_z(t')\ , , \label{eq : xevoltensor } \\
\hat{p}(t ) & = - \ , \hat{x}(0)v e^{-\gamma_{\text{sw}}t } \sin \left ( \omega t \right)+ \hat{p}_0 e^{-\gamma_{\text{sw}}t } \cos \left(\omega t \right ) \nonumber \\ & - \beta w \sqrt{j_x } \int_0^t \ ! \mathrm{d}t ' \ , e^{-\gamma_{\text{sw}}(t - t ' ) } \sin \left [ \omega ( t - t')\right ] s_y(t ' ) + \beta \sqrt{j_x } \int_0^t \ ! \mathrm{d}t ' \ , e^{-\gamma_{\text{sw}}(t - t ' ) } \cos \left [ \omega ( t - t')\right ] s_z(t ' ) \ , , \label{eq : pevoltensor}\end{aligned}\ ] ] where for the cs spin oscillator of the experiment : @xmath238 , and the dissipation rate @xmath239 , associated with the coherent interaction of the collective atomic ensemble spin with the light , is given by : @xmath240 the parameter @xmath239 can either be negative or positive depending on the polarization direction of the linearly polarized probe light . in the experiment , a linearly polarized light in the @xmath157 direction
was employed , resulting in @xmath239 being positive .
the effect of second rank tensor polarizability on the measured probe light stokes component @xmath158 can be neglected to first order in @xmath241 @xcite , and eq .
[ eq : syout0 ] does not need to be modified : @xmath242 . through the interaction parameterized by @xmath243 , noise in @xmath181 couples into the measured observable @xmath102 even in the limit of zero duty cycle . however , this back - action noise is suppressed by @xmath244 compared to the back - action noise arising from the vector polarizability hamiltonian .
the data presented in this work were acquired with a probe detuning @xmath245 ghz , leading to @xmath246 , so that the effect of tensor polarizability in the back - action noise can be ignored ( its value is on the order of the back - action noise due to the finite duty cycle of the stroboscopic measurement ) .
it can be seen from the eq .
[ eq : xevoltensor ] that the tensor polarizability dynamics can result in noise suppression for positive @xmath239 ( the case of our experiment ) .
this dissipative interaction can be used to generate unconditionally squeezed states and steady state entanglement @xcite . in the experiment ,
second - rank tensor polarizability dynamics lead to small but finite reduction in the oscillator noise ( see discussion about fig . [
fig : supsqnoise ] ) .
the cesium atomic ensemble is contained in a glass cell microchannel , shown in fig .
[ fig : cslevels](b ) filled with cs vapor and with the walls covered with an alkene coating @xcite .
the cell is enclosed in a larger glass container .
the longitudinal spin lifetime is @xmath247 ms and the transverse spin lifetime is @xmath248 ms in the absence of light fields .
the cell temperature ( and the atomic density ) can be regulated by using electrical heating .
the microcell is placed inside a standing wave optical cavity consisting of two concave mirrors .
a mirror with @xmath249 reflectivity in intensity is used as the output coupler , while the input coupler has a much higher reflectivity : @xmath250 in intensity .
the cavity is locked on resonance by a pound - drever - hall technique , applying a 10khz modulation to the piezoelectric transducer .
a digital pi controller , which incorporates feed - forward techniques , is employed to lock the cavity on resonance .
the curvature ( radius @xmath251 mm ) and the separation of the mirrors are chosen so that the resulting cavity mode has @xmath252 @xmath142 m waist diameter .
this mode has small clipping losses in the cell , and allows for efficient coupling to the atomic ensemble . due to the small clipping losses in the cell the cavity mode is gaussian to a good approximation .
the cavity finesse is @xmath74 , with a fwfm linewidth @xmath253 mhz .
optical pumping of the ensemble is realized with circularly polarized pump and repump light fields resonant with the d1 and d2 atomic transition respectively ( see fig .
[ fig : cslevels](a ) ) .
the repump light effectively transfers population from the @xmath254 hyperfine manifold to the probed @xmath77 manifold of the ground electronic state .
the pump light field creates high degree of spin orientation in the @xmath77 manifold .
the atomic polarization can be estimated by mapping the rf resonance at large magnetic fields , where the non - linear zeeman effect splits the magnetic resonances . a rf pulse , short enough so that the fourier spectrum is approximately flat over the frequency range of the zeeman resonances , is applied at the end of the pumping pulse and the spin evolution is monitored with a weak probe . in fig .
[ fig : mors ] the rf resonance ( fourier transform of the spin response to the rf pulse ) is plotted for two different degrees of spin orientation . under the assumption of spin temperature distribution @xcite , which maximizes the entropy under the constraint of a given spin orientation
, the spin ordination can be estimated by fitting the experimentally observed data to the model described in @xcite .
we note that for a weak probe the main spin - relaxation mechanism comes from the collisions with the walls of the cell and magnetic field gradients , so that a common relaxation rate for the ground state coherences can be assumed .
was close to 1 mhz .
the spin orientation is estimated by fitting the experimental data to the spin temperature distribution . ] in order to estimate the fraction of atoms in the @xmath254 manifold due to imperfect optical pumping , the spin responses to short rf field resonant to the @xmath254 manifold zeeman frequency in the cases of optical pumping phase with and without the repump light field are compared .
we note that cs has different gyromagnetic ratios in the two ground state manifolds resulting in different larmor frequencies . in the absence of repump light during the pumping interval ,
the pump light depopulates almost completely the @xmath77 manifold creating orientation in the @xmath254 manifold . for this measurement ,
the probe wavelength is tuned to be mostly sensitive to transitions from the @xmath254 manifold .
it is estimated that less than 5% of the atoms remain in the @xmath254 manifold after the optical pumping in the experiment , so that for the prediction of the measured uncertainty in the oscillator ground state approximately all atoms in the ensemble should be considered .
the number of atoms in the cell can be determined from the faraday optical rotation of a linearly polarized probe traveling through the atomic ensemble that has been polarized along the light propagation direction . for a fully polarized ensemble ( no atoms in the @xmath254 manifold and 100% orientation in the @xmath77 manifold ) ,
the number of atoms @xmath25 is related to the faraday angle @xmath255 by the equation : @xmath256 the atomic density can be also characterized by optical absorption spectroscopy .
typically , the agreement on atomic density between the faraday angle measurement and the optical absorption spectroscopy is @xmath257 . in fig.3(a ) of @xcite ,
the number of atoms was calibrated against a response to a rf field for the polarized atoms and independently by a spin noise measurement for the unpolarized atoms .
the atomic response to a rf field much shorter than the spin coherence time is @xmath258 for a fixed spin orientation , so that by monitoring the relative response to a given rf field the relative number of atoms can be found .
similarly , the spin noise of atoms in the thermal state ( where no probe back - action occurs ) scales proportionally to the number of atoms .
the photon shot noise level can be estimated from a polarimetry noise measurement when the oscillator frequency is tuned away from the lock - in detection bandwidth .
the noise associated with the spin - oscillator has a lorentzian power spectrum centered at larmor frequency with width set by the decoherence rate ; therefore if the oscillator frequency is set many line - widths away from the lock - in bandwidth , the oscillator noise is filtered out and the measurement noise only depends on the probe light noise of white spectrum . the oscillator frequency can be tuned by changing the bias magnetic field .
the probe noise is verified to scale linearly with the power as expected for the photon shot noise power .
the photodetector electronic noise is at the level of @xmath259 of the photon shot noise for the power used in the squeezing experiment .
the detection efficiency of the light field at the output of the cavity is @xmath260 , limited mainly by the detector quantum efficiency .
the ground state noise is calibrated against a measurement in the thermal state of the atomic ensemble .
including decoherence , but neglecting change in the spin variance ( e.g. due to loss of atoms from depumping into the @xmath254 manifold ) , the two - time spin correlation can be written in the form : @xmath261 $ ] , where @xmath82 is the decay rate and @xmath262 is the variance in the spin state .
it can be shown that the @xmath263 component of the @xmath158 measurement over time @xmath33 , weighted by the exponential mode function @xmath264 , is given ( in the absence of back - action ) by : @xmath265 } { \text{psn}_{\tau } } = 2 \beta^2 s_x \frac{\gamma -e^{2 \gamma_{\text{m } } \tau } \gamma+\left[1+e^{2 \gamma_{\text{m } } \tau}-2 e^{(\gamma_{\text{m}}-\gamma ) \tau}\right ] \gamma_{\text{m } } } { \left(-1+e^{2 \gamma_{\text{m } } \tau}\right ) ( \gamma_{\text{m}}-\gamma ) ( \gamma+\gamma_{\text{m}})}\langle \hat{j}_z^2\rangle \ , , \label{eq : varjzdec0}\ ] ] which for @xmath266 reduces to : @xmath265 } { \text{psn}_{\tau}}= \beta^2 s_x \tau \frac{1 + \gamma \tau - \gamma \tau \coth ( \gamma \tau ) } { \gamma \tau } \langle \hat{j}_z^2 \rangle \ , , \label{eq : varjzdec1}\ ] ] where @xmath267 is the photon shot noise ( variance ) contribution to the measurement noise , estimated by : @xmath268 with @xmath269 being the polarimetry output when the oscillator frequency is detuned away from the lock - in bandwidth . in the thermal state , where no back - action occurs , the spin variance is : @xmath270 , where @xmath77 is the total spin in the probed manifold , @xmath25 is the total number of atoms in the ensemble and the factor @xmath271 is the fraction of atoms in the @xmath77 manifold .
we note that in the experiment described here , the contribution of atoms in the @xmath254 manifold to the thermal spin noise measurement is estimated to be @xmath272 .
the ground state of the oscillator corresponds to the coherent spin state of maximum @xmath273 , leading to @xmath274 .
assuming the same decoherence rate for coherent and thermal states , we find that the contribution of a magnetic oscillator in the ground state with @xmath275 to the measured noise is given by ( in units of @xmath267 ) : @xmath276 \right|_{\text{thermal } } } { \text{psn}_{\tau}}-1 \right ] \ , .
\label{eq : groundnoisecal}\ ] ] there is a small discrepancy in the decoherence rate and the loss rate of atoms due to the probe scattering for the coherent and thermal rates ; for the timescales of the experiment , the effect of this difference is estimated to be @xmath277 .
in order to study squeezing , a scheme with two successive pulse measurements is employed ( see fig .
2(b ) in @xcite ) . the first measurement is described by : @xmath278 while the second measurement is : @xmath279 an exponentially rising ( falling ) mode is used for the first ( second ) pulse to moderate the effect of decoherence which reduces the correlations between the pulses . the measurement noise associated with the oscillator state
is extracted by subtracting the photon shot noise for the same pulse duration ( @xmath280 ) from the measured noise ; normalized to the photon shot noise variance the measured oscillator noise is : @xmath281\ , , \label{eq : xmvara } \\
\text{var } ( \hat{x}_{m , b } ) & = \left [ \frac{\text{var } ( \hat{q}_{b})}{\text{psn}_{b}}-1 \right]\ , , \label{eq : xmvarb } \\
\text{var } ( \hat{x}_{m , b|a } ) & = \left [ \frac{\text{var } ( \hat{q}_{b}|\hat{q}_{a})}{\text{psn}_{b}}-1 \right]= \left [ \frac{\text{var } ( \hat{q}_{b})}{\text{psn}_{b } } - \frac{\text{cov}^2 \left [ \hat{q}_{b},\hat{q}_{a}\right]}{\text{psn}_b \text{var}\left ( \hat{q}_{a } \right)}-1 \right]\ , .
\label{eq : xmvarcond}\end{aligned}\ ] ] the conditional variance @xmath282 effectively evaluates the oscillator position noise in the state filtered out by the first measurement .
the realized squeezing is determined by the measurement strength in the first pulse ; in eq .
[ eq : squeezingformula0 ] and eq .
[ eq : squeezingformula1 ] the relevant coupling constant @xmath36 is associated with the first measurement . ) and expected noise in the oscillator ground state ( @xmath283 ) as a function of the number of photons @xmath284 .
( b ) comparison of expected ground state noise ( @xmath285 ) with unconditional and conditional to the first pulse measured noises ( @xmath286 and @xmath283 respectively ) as a function of the number of photons in the first pulse .
see text for a discussion . ] in fig .
[ fig : supsqnoise ] the oscillator noise in the first and second pulse , conditionally and unconditionally to the first measurement , are plotted as a function of the number of photons in the first pulse @xmath284 . in the experiment ,
the probe power averaged over an oscillator period is maintained constant and the duration of the first pulse ( @xmath287 ) is varied .
the duration of the second pulse is fixed to 0.5 ms , which corresponds to @xmath288 photons . the expected noise for the ground state of the oscillator ( denoted by the subscript 0 ) is also plotted for comparison .
due to the correlation between the first and the second pulse , the conditional variance of the second pulse is significantly reduced below the position imprecision in the oscillator ground state , leading to squeezing . the data in fig .
[ fig : supsqnoise ] indicate an additional squeezing mechanism unrelated to the information gain by measurement , since the unconditional noise for both the first and second pulse drops below the ground state uncertainty .
this kind of squeezing is due to the second - rank tensor polarizability dynamics . in fig .
[ fig : supsqzplots](a ) the conditional noise is evaluated with respect to the measured unconditional imprecision . a comparison of the data in fig .
[ fig : supsqzplots](a ) with those in fig .
4 of @xcite indicates that the realized degree of squeezing is dominated by the information gain through the qnd measurement . in evaluating the squeezing degree , the measured variance is normalized to the macroscopic spin component @xmath289 at the end of the first pulse : @xmath290 , where @xmath291 accounts for the reduction of the mean spin during the first measurement . the same correction factor @xmath292
is used to estimate the squeezing in fig . 4 of @xcite .
the reduction of the mean spin @xmath19 is characterized by an independent measurement of the response to an external coherent perturbation .
the probe pulse of duration @xmath287 , applied after the optical pumping phase , is followed by a resonant rf excitation pulse , much shorter than the decoherence time , and the response , which is proportional to the mean spin @xmath19 , is recorded . in fig .
[ fig : supsqzplots](b ) the demonstrated squeezing according to the wineland criterion @xcite is plotted as a function of @xmath284 .
the wineland squeezing evaluates the effect of noise suppression in spectroscopic sensitivity : @xmath293 .
the data show that the created squeezed state offers the possibility of enhanced force measurement below the ground state uncertainty .
12 b. julsgaard , _ entanglement and quantum interactions with macroscopic gas samples _ , ph.d .
thesis , university of aarhus ( 2003 ) .
k. jensen , _ quantum information , entanglement and magnetometry with macroscopic samples and non - classical light _
thesis , university of copenhagen ( 2011 ) .
k. hammerer , a. s. srensen , and e. s. polzik , rev .
* 82 * , 1041 ( 2010 ) .
d. v. kupriyanov , o. s. mishina , i. m. sokolov , b. julsgaard , and e. s. polzik , phys .
a * 71 * , 032348 ( 2005 ) .
generation of a squeezed state of an oscillator by stroboscopic back - action - evading measurement " .
d. v. vasilyev , k. hammerer , n. korolev , and a. s. srensen , journal of physics b : atomic , molecular and optical physics * 45 * , 124007 ( 2012 ) . h. krauter , c. a. muschik , k. jensen , w. wasilewski , j. m. petersen , j. i. cirac , and e. s. polzik , phys . rev . lett . * 107 * , 080503 ( 2011 ) .
d. v. vasilyev , c. a. muschik , and k. hammerer , phys .
a * 87 * , 053820 ( 2013 ) .
m. v. balabas , t. karaulanov , m. p. ledbetter , and d. budker , phys .
lett . * 105 * , 070801 ( 2010 ) .
s. appelt , a. b .- a .
baranga , c. j. erickson , m. v. romalis , a. r. young , and w. happer , phys . rev . a * 58
* , 1412 ( 1998 ) .
b. julsgaard , j. sherson , j. l. srensen , and e. s. polzik , journal of optics b : quantum and semiclassical optics * 6 * , 5 ( 2004 ) .
d. j. wineland , j. j. bollinger , w. m. itano , f. l. moore , and d. j. heinzen , phys . rev .
a * 46 * , r6797 ( 1992 ) . | continuous observation on an oscillator is known to result in quantum back - action which limits the knowledge acquired by the measurement . a careful balance between the information obtained and the back - action disturbance leads to a limit known as the standard quantum limit .
the means to surpass this limit by modulating the measurement strength with the period proportional to half period of the oscillation has been proposed decades ago @xcite .
such modulated or stroboscopic observation leading to a squeezed state of one quadrature of the oscillator motion with the quantum noise below that of the zero - point fluctuations has been a long - standing goal . here , we report on the generation of a quadrature - squeezed state of an oscillator by stroboscopic back - action evading measurement .
the oscillator is the collective spin of an atomic ensemble precessing in magnetic field .
it is initially prepared in nearly the ground state with an average thermal occupancy number @xmath0 .
the oscillator is coupled to the optical mode of a cavity , and the cavity output field detected with polarization homodyning serves as the meter .
a back - action - evading measurement is performed by stroboscopically modulating the intensity of the light field at twice the larmor frequency , resulting in a squeezed state conditioned on the light - polarization measurement with @xmath1 db noise reduction below the zero - point fluctuations for the measured quadrature .
the demonstrated squeezing holds promise for metrological advantage in quantum sensing .
the heisenberg uncertainty sets the limit of how precisely two non - commuting variables , such as the canonical position and momentum with the commutation relation @xmath2=i$ ] , can be specified simultaneously ; however , there are no physical limitations on determining the value of an individual variable . for an oscillator with frequency
@xmath3 the position variable in the laboratory frame is @xmath4 where @xmath5 are the rotating frame variables .
quantum states for which a canonical variable has reduced uncertainty with respect to the oscillator ground state , for example , @xmath6 , are called squeezed states ( ss ) . squeezed states for matter oscillators have been first demonstrated for motion of a single ion @xcite and later for magnetic oscillators by electron - nucleous entanglement @xcite and by spin - spin interaction @xcite .
an intriguing approach towards generation of an ss in an oscillator is a stroboscopic quantum non demolition ( qnd ) or back - action evading measurement proposed in @xcite .
if a meter is coupled to the oscillator with an interaction hamiltonian @xmath7 and if the measurement strength is maximized at times @xmath8 the noise of the meter does not couple to @xmath9 and the readout of this qnd measurement yields a ss of the oscillator with a degree of squeezing defined as @xmath10 .
another version of this approach based on harmonic modulation of the measurement strength which works similarly to the stroboscopic measurement has been proposed in @xcite .
previous attempts to implement this approach with a magnetic spin oscillator @xcite and a mechanical oscillator @xcite demonstrated reduction in the back - action noise but did not result in an ss due to the insufficient strength of the qnd measurement compared to the decoherence caused by the environment . in a separate line of work back - action evasion has been demonstrated for a joint state of two oscillators @xcite .
there a continuous qnd measurement on both oscillators one of which has an effective negative mass has been shown to generate an entangled state of the two oscillators .
a key requirement for achieving the ss is that the decoherence due to interaction with the environment is reduced to the extent that the oscillator maintains its quantum state for the time longer than a qnd measurement takes .
every decoherence event is linked with loss of information about the measured variable and with the decay to the ground or thermally excited state of the oscillator depending on the decoherence mechanism . in this work
this requirement is met by firstly enhancing the rate of the optical qnd interaction by placing the spin ensemble in an optical cavity , and secondly by placing the spins in a designed spin - protecting environment .
precesses around @xmath11 with a larmor frequency @xmath3 .
normalized quantized values of the projections @xmath12 in the rotating frame are the canonical variables for this oscillator .
( b ) the energy diagram of an atomic constituent of the oscillator . in the ground state of the oscillator atoms are in the lower energy state @xmath13 .
raman scattering of photons ( dashed lines ) driven by the input light ( solid arrows ) creates collective excitations in level @xmath14 and corresponding quantum coherences @xmath15 responsible for @xmath16 and oscillating at @xmath3 .
( c ) the @xmath17-th
excited state of the oscillator corresponds to @xmath17 collective spin flips generated as in ( b ) .
a squeezed state is a coherent superposition of such number states ( comments in the text ) . ]
the oscillator used in this study is a collective spin of an atomic ensemble precessing in a bias magnetic field ( fig .
[ fig : spinoscillator]a ) .
the collective spin components in the lab frame oscillate as @xmath18 . for a macroscopic spin orientation @xmath19 along the magnetic field
, the energy spectrum of the spin ensemble can be mapped onto that of a harmonic oscillator @xcite with a resonance frequency set by the larmor frequency of the spin precession ( fig .
[ fig : spinoscillator]b , c ) .
mathematically this is expressed through the holstein - primakoff transformation @xcite .
canonical position and momentum operators can be defined through the collective symmetric spin observables in the rotating frame @xmath20=i j_x = i n_{\text{at}}f$ ] ( @xmath21 ) as : @xmath22 , and @xmath23 where it is assumed that the collective spin is well oriented , so that the population of the end state with the magnetic quantum number @xmath24 is close to the total number of spins @xmath25 and hence the macroscopic quantity @xmath19 is treated as a number rather than an operator .
the ground state of the harmonic oscillator corresponds to all atoms being in the @xmath26 state and the absence of the ensemble coherence @xmath27 ( fig .
[ fig : spinoscillator]b ) .
the mean number of excitations above the ground state can be evaluated as @xmath28 .
an excitation with the creation operator @xmath29 corresponds to a quantum of excitation , also called a polariton , distributed symmetrically among all the atoms of the oscillator .
the ground state of the oscillator belongs to the class of coherent spin states ( css ) characterized by @xmath30 @xcite .
the quantum state of the spin - oscillator can be created and probed through interaction with a light field . in the limit of large probe detuning with respect to the atomic excited state hyperfine level ( fig .
[ fig : spinoscillator]b ) , the light - spin interaction can be approximated by the qnd - type hamiltonian : @xmath31 , where @xmath32 is the number of photons in the pulse of duration @xmath33 and @xmath34 is the probe light stokes operator in the circular basis normalized so that for a coherent pulse var@xmath35 .
the coupling constant @xmath36 characterizes qnd interaction strength , which depends on the atom - light detuning , the excited - state linewidth and @xmath37 @xcite , where @xmath38 is the resonant optical depth of the atomic ensemble and @xmath39 is proportional to the fractional number of decoherence events during the measurement time @xcite . for stroboscopic probing the relevant observable for the oscillator
is the harmonic quadrature that evolves in phase with the modulation ( chosen to be the cosine quadrature ) : @xmath40 , where @xmath41 is the oscillator period , @xmath42 is the duty cycle of the probe , @xmath43 is a rectangular - pulse shaping function of unit amplitude , following the temporal evolution of the probe power , and the integration extends over one oscillator period . in the limit of zero duty cycle : @xmath44 .
the measurement record is the @xmath45 fourier component of the photocurrent integrated over the pulse length : @xmath46 , where @xmath47 is the stokes operator in the linear basis measured with polarization homodyning as shown in fig .
[ fig : experimental setup ] and @xmath48 is a mode function of unit energy over the pulse length that weights the measurement data according to the decoherence rate ( see below ) .
the evolution of variances due to the hamiltonian interaction is described by ( for @xmath49)@xcite : @xmath50 , \label{eq : polarimetryvariance } \\
\text{var } ( \hat{x}_{\text{out } } ) & = \text{var } ( \hat{x}_{\text{in } } ) + \mathcal{c } \tilde{\kappa}^2 , \label{eq : xvariance}\end{aligned}\ ] ] where the subscripts ( in ) , ( out ) indicate operators at the start and end of the interaction respectively , @xmath51 , @xmath52 is the modified coupling constant , and @xmath53 $ ] quantifies the coupling of the probe noise to the observable .
it can be shown @xcite that for a stroboscopic probe the back - action coupling constant is given by : @xmath54 for a qnd measurement @xmath55 , while for a continuous probing @xmath56 . in eq .
[ eq : polarimetryvariance ] , the first term on the right - hand side specifies the imprecision due to the quantum noise of the meter ( shot noise of light ) , whereas the second and third terms describe the oscillator input noise and back - action noise , respectively . in the following ,
the measurement noise associated with the oscillator in photon shot noise units will be denoted collectively with @xmath57 , where @xmath58 expresses the uncertainty increase due to decoherence .
conditionally on a qnd ( @xmath55 ) realization of @xmath59 the oscillator evolves to a quantum state with reduced position noise : @xmath60 , with cov denoting the covariance .
the quantum filtering then leads to conditional squeezing described by @xcite : @xmath61 where @xmath62 is the imprecision in the ground state ( zero point fluctuations ) .
the decoherence term signifies the importance of the large optical depth @xmath38 in achieving high degrees of squeezing , as @xmath63 .
we achieve an enhancement of the effective @xmath38 by enclosing the spin oscillator in an optical resonator ( fig .
[ fig : experimental setup ] ) .
optimal squeezing can be achieved with an impedance matched cavity where the output coupler transmission @xmath64 , where @xmath65 is the round - trip intensity loss @xcite . .
the projection of the spin along the probe direction is subject to stroboscopic qnd measurements every half period of oscillation as indicated with stars .
( b ) pulse sequence for generating and demonstrating conditional squeezing of an oscillator with stroboscopic qnd measurement . ] the spin oscillator is realized in room - temperature , optically - pumped cesium atoms , contained in a glass cell microchannel , @xmath66 in cross section and 1 cm in length .
an alkene coating @xcite deposited at the inner cell walls dramatically suppresses spin - relaxation due to the wall collisions .
atoms bounce off the walls and cross the optical mode cross section with the waist of @xmath67 m approximately @xmath68 times before their quantum spin state decoheres in @xmath69 msec due to wall collisions .
the atom - atom collision rate at the low cs pressure used here is negligible . as the typical light pulse duration of @xmath70
msec is much greater than the atom transient time of @xmath71sec and the oscillator period ( typical @xmath72 khz ) the thermally moving atoms cross the optical mode many times in the same state and hence the detected optical mode couples to the symmetric spin mode ( equivalently to the oscillator position @xmath73 ) .
we emphasize that the thermal motion of the atoms does not affect the oscillator temperature , which is determined by the spin distribution .
the microcell is placed inside a standing wave optical cavity with a finesse @xmath74 , the single - pass losses in the cell windows of @xmath75 and the output coupler transmission of @xmath76 , which is close to the optimal value @xmath64 .
the cavity is kept on resonance with light using the pound - drever - hall technique .
the number of atoms in the @xmath77 hyperfine ground state coupled to the light field ( fig .
[ fig : spinoscillator]b ) has been adjusted within the @xmath78 range by changing the cell temperature ( typically @xmath79c ) and optical pumping for the maximal qnd interaction strength .
the frequency of the oscillator can be tuned with @xmath11 .
duty cycle .
the arrow indicates the back - action suppression . ]
an acousto - optic modulator is used to stroboscopically modulate the intensity of the probe beam at twice the larmor frequency .
the experiment was operated with @xmath80 stroboscopic duty cycle , with probe wavelength blue detuned by 1.6 ghz with respect to the d2 transition @xcite .
the @xmath59 operator is measured by balanced polarimetry and lock - in detection .
the data are weighted with an exponential mode function : @xmath81 , where @xmath82 is the decoherence rate in the presence of the probe .
the exponential falling mode function is used to assess the measured noise , except for the squeezing investigation where the first pulse measurement is defined with a rising mode ( see fig . [
fig : experimental setup]b ) . to collect statistics for the variance estimation
, each measurement is repeated @xmath83 times .
the atoms of the magnetic oscillator are initialized by optical pumping in a state close to the ground state .
the initial state variance is calibrated against a measurement of the noise in the thermal state of the atomic ensemble obtained with unpolarized spins .
the measured spin variance of unpolarized atoms @xmath84 can serve as a robust reference for the spin noise in the coherent end state @xmath85 @xcite , since the former is insensitive to classical fields and probe - induced noise . in fig .
[ fig : spinnoisevsat]a the oscillator noise variance in the state prepared by optical pumping is plotted as a function of the atomic density . the observed linear scaling indicates a quantum - limited performance and a qnd character of the measurement . in the inset the ratio of the measured variance to the calibrated zero point imprecision is shown to be @xmath86 .
the increased measured variance in the initial oscillator state is due to the imperfect optical pumping , which leads to a finite oscillator temperature corresponding to thermal occupation @xmath87 .
the occupation probability distribution among the zeeman levels can be found from the magneto - optical resonance signal @xcite ( see fig . 3 in @xcite ) . assuming a spin - temperature distribution , it is found that after optical pumping @xmath88 of the atoms are in the end state @xmath89 , with @xmath90 occupation probability for the @xmath91 state , and negligible probabilities for the other states .
this is consistent with the 16% increase of the measured variance compared to the ground state noise .
it is instructive to compare decoherence and thermalisation properties of the magnetic oscillator and mechanical oscillators . for the former oscillator initialized in the ground state (
fully polarized spin ) , @xmath92 in the range of @xmath93 where @xmath94 is the population life time ( @xmath95 msec ) and @xmath96 in the present case .
the thermalisation time @xmath97 is at least six orders of magnitude greater than @xmath94 . for mechanical oscillators
@xmath98 at @xmath99 where the thermalisation time can reach at best a second in state - of - the - art experiments .
this comparison illustrates the difficulty of quantum state preservation without cryogenic environment for mechanical oscillators @xcite in comparison to magnetic spin oscillators .
the qnd character of the measurement is demonstrated in fig .
[ fig : spinnoisevsat]b , where the measured variance of the oscillator is plotted as a function of the number of the probe photons in the pulse . for a continuous probe ( @xmath100 duty cycle ) ,
the imprecision in units of zero point fluctuations increases with the number of photons in the pulse .
in contrast , for a stroboscopic probe with a small duty cycle ( @xmath101 ) the noise remains nearly independent of the pulse strength over the measured range .
the demonstrated reduction of probe back - action is more than 10 db compared to a continuous probing of the oscillator .
to study the generation of squeezed oscillator states , two qnd pulses are employed : the first provides information about the oscillator observable @xmath102 , and the second pulse evaluates the observable variance conditioned on the first measurement ( see fig . [
fig : experimental setup]b ) . in fig .
[ fig : spinsqueezing ] the reduction in conditional variance compared to the zero point imprecision is plotted as a function of the number of photons in the first pulse , for a fixed photon number in the second pulse @xmath103 .
the squeezing increases with the photon number in the pulse until the point where decoherence induced noise compensates for the reduction in uncertainty by measurement .
the observed conditional variance is up to @xmath104 db below the ground state noise . along with the dominant contribution of qnd measurement to the degree of squeezing ,
a smaller contribution is due to the interaction nonlinear in the spin variables ( see @xcite for the effect of the second - rank tensor polarizability ) .
the demonstrated noise reduction indicates the metrological advantage of the created state @xcite . in summary , we have demonstrated a squeezed state of an oscillator generated by a stroboscopic qnd measurement .
it has been made possible by implementing optical cavity - enhanced qnd interaction and by utilizing a long lived spin oscillator with room temperature atoms contained in a spin protecting microcell .
the techniques developed in the paper will be useful for quantum metrology and sensing , as well as for generation of entanglement between disparate oscillators @xcite .
this work was supported by the erc grant interface , darpa project quasar and eu grant siqs .
k. j. acknowledges support from the carlsberg foundation .
g.v . gratefully acknowledges help and support from p. karadaki .
21 v. braginskii , y. vorontsov , and f. khalili , jetp lett .
* 27 * , 276 ( 1978 ) .
k. s. thorne , r. w. p. drever , c. m. caves , m. zimmermann , and v. d. sandberg , phys .
rev .
lett . * 40 * , 667 ( 1978 ) .
v. b. braginsky , y. i. vorontsov , and k. s. thorne , science * 209 * , 547 ( 1980 ) .
d. m. meekhof , c. monroe , b. e. king , w. m. itano , and d. j. wineland , phys .
rev .
lett . * 76 * , 1796 ( 1996 ) .
t. fernholz , h. krauter , k. jensen , j. f. sherson , a. s. srensen , and e. s. polzik , phys .
rev .
lett .
* 101 * , 073601 ( 2008 ) . k. hammerer , a. s. srensen , and e. s. polzik , rev .
mod .
phys . * 82 * , 1041 ( 2010 ) . c. gross , t. zibold , e. nicklas , j. estve , and m. k. oberthaler , nature * 464 * , 1165 ( 2010 ) .
m. f. riedel , p. bhi , y. li , t. w. hnsch , a. sinatra , and p. treutlein , nature * 464 * , 1170 ( 2010 ) .
a. a. clerk , f. marquardt , and k. jacobs , new journal of physics * 10 * , 095010 ( 2008 ) .
g. vasilakis , v. shah , and m. v. romalis , phys .
rev .
lett . * 106 * , 143601 ( 2011 ) .
j. suh , a. j. weinstein , c. u. lei , e. e. wollman , s. k. steinke , p. meystre , a. a. clerk , and k. c. schwab , science * 344 * , 1262 ( 2014 ) .
e. s. polzik and k. hammerer , e - print ( 2014 ) , arxiv:1405.3067 [ quant - ph ] .
w. wasilewski , k. jensen , h. krauter , j. j. renema , m. v. balabas , and e. s. polzik , phys .
rev .
lett . * 104 * , 133601 ( 2010 ) .
t. holstein and h. primakoff , phys . rev . * 58 * , 1098 ( 1940 ) .
supplementary information .
m. v. balabas , t. karaulanov , m. p. ledbetter , and d. budker , phys .
rev .
lett . * 105 * , 070801 ( 2010 ) .
b. julsgaard , j. sherson , j. l. srensen , and e. s. polzik , journal of optics b : quantum and semiclassical optics * 6 * , 5 ( 2004 ) .
j. chan , t. p. m. alegre , a. h. safavi - naeini , j. t. hill , a. krause , s. groblacher , m. aspelmeyer , and o. painter , nature * 478 * , 89 ( 2011 ) .
w. happer and b. s. mathur , phys . rev . * 163 * , 12 ( 1967 ) .
d. j. wineland , j. j. bollinger , w. m. itano , f. l. moore , and d. j. heinzen , phys . rev .
a * 46 * , r6797 ( 1992 ) .
k. hammerer , m. aspelmeyer , e. s. polzik , and p. zoller , phys .
rev .
lett . * 102 * , 020501 ( 2009 ) . * supplementary information to generation of a squeezed state of an oscillator by stroboscopic back - action - evading measurement " * |
the @xmath3 methanol maser transition at 6.7 ghz has the strongest flux densities among methanol maser lines . the methanol maser emission is thought to be produced by radiative excitation in an infrared radiation field , which is formed by dust near the protostar with a temperature of @xmath2100 - 200 k ( @xcite ; @xcite ; @xcite ) .
some sources , however , are associated with shocked gas ( @xcite ; @xcite ; @xcite ) .
collisional excitation by shock wave may produce the methanol maser emission .
a number of 6.7 ghz methanol masers have been found to exhibit long - term variability .
@xcite found that the sources vary on a time - scale of several months
. showed that about 65% of methanol maser in their sample exhibit moderate or strong variability on time - scales of about four and eight years . a long - term monitoring program to investigate the variability of 54 methanol maser sources at 6.7
ghz has been conducted using the hartebeesthoek 26-m radio telescope by @xcite .
they divided the variable sources into six types according to the behavior of the variability .
for [email protected] in their sample , the periodic variations associated with massive star formation were discovered for the first time @xcite .
the very long baseline interferometer ( vlbi ) monitoring at seven epochs over three months for [email protected] was conducted within the period of the time variation @xcite .
no appearance of new spots and no change in morphology was found , suggesting that the flares were caused by a change in either the seed or pump photon levels .
they proposed a binary system as an origin of the periodic flares .
@xcite continuously observed g [email protected] , and found flares at least seven times with time delays in the range 10 - 35 days between the variability of red - shifted spectral feature and blue - shifted feature .
they discussed the delay , which can possibly be explained in terms of light - crossing time due to the spatial distribution of spots .
we have conducted daily monitoring for some maser sources to investigate a short time - scale variability with yamaguchi 32-m radio telescope .
cepheus a ( cep a ) was observed as one of our sample sources .
cep a is a co condensation at a distance of 725 pc @xcite and some radio continuum sources were detected .
methanol masers are associated with cep a - hw2 defined by @xcite , which is the brightest radio continuum source detected in the region .
the hw2 object has a radio continuum jet along a position angle ( pa ) of @xmath2 ( @xcite ; @xcite ; @xcite , ; @xcite ) . based on submillimeter observations of both dust and ch@xmath5cn line emissions ,
a flattened disklike structure was found @xcite . the structure has a size of about 1000 au and is perpendicular to the jet .
recently , it is revealed that nh@xmath5 and so@xmath6 line emissions coincided with the ch@xmath5cn disk , although the so@xmath6 structure was about 2 times smaller ( @xcite ; @xcite ) .
the 22.2 ghz water masers observed by @xcite were distributed in an elongated structure perpendicular to the radio jet , which possibly trace the circumstellar disk around the hw2 object .
the water maser has a spatial distribution similar to the so@xmath6 disk .
the methanol maser at 6.7 ghz in cep a shows variability .
it was 1420 jy in 1991 @xcite , but 815 jy in 1999 .
@xcite has found that the amplitude ratio of the lines at a radial velocity of @xmath03.8 km s@xmath1 and @xmath04.2 km s@xmath1 varies from 1.4 to 0.8 over two years . the spatial distribution of the masers at 6.7 ghz for cep a have been already reported @xcite with the japanese vlbi network ( jvn ; @xcite ) .
however , the image quality was not enough to investigate the relationship the spectral features and the spatial distribution . in this paper , we present the spectral variability with the spectral monitoring observations and a new vlbi map . in section 2 , we describe the details of these observations and data reduction , and the results
are presented in section 3 . in section 4 ,
we discuss interpretations of the rapid variability and excitation mechanism of this maser in cep a.
the daily monitoring program with yamaguchi 32-m radio telescope was made from august 4 ( corresponding to the days of year ( doy ) 216 ) to october 24 ( doy 297 ) , 2007 .
the full - width at half maximum ( fwhm ) of the beam is 5 arcmin at 6.7 ghz .
the pointing error of the antenna is smaller than 1 arcmin .
the spectrometer consists of the ip - vlbi system @xcite and a software spectrometer . both left and right circular polarizations were recorded with 2-bit sampling .
the recorded data with a bandwidth of 4 mhz , covering a velocity range of 180 km s@xmath1 , were divided into 4096 channels , yielding a velocity resolution of 0.044 km s@xmath1 .
the integration time is 14 minutes until doy 244 , and then 10 minutes from doy 245 .
the rms noise level was typically 1.2 jy and 1.4 jy with an integration time of 14 min and 10 min , respectively .
an amplitude and a gain calibration was performed by measuring the system noise temperatures by injecting the signal from noise - sources with known temperatures .
the accuracy of the calibration was estimated to be 10% .
the stability of the system was checked by daily monitoring [email protected] methanol maser emission , which showed relatively small variability in the sample of @xcite . a vlbi observation at 6.7 ghz of cep a was made on september 9 2006 from 15:00 to 22:00 ut with four telescopes ( yamaguchi 32-m , usuda 64-m , vera - mizusawa 20-m , and vera - ishigaki 20-m ) of the jvn .
the projected baselines covered from 9 m@xmath7 ( usuda mizusawa ) to 50 m@xmath7 ( mizusawa ishigaki ) , corresponding to the fringe spacing of 23 mas and 4.1 mas , respectively .
left - circular polarization was received at yamaguchi and usuda stations , while linear polarization was received at mizusawa and ishigaki stations .
a special amplitude calibration for different polarizations was made by the same procedure as described in sugiyama et al .
the data were recorded on magnetic tapes using the vsop - terminal system at a data rate of 128 mbps with 2-bit quantization and 2 channels , and correlated at the mitaka fx correlator @xcite . from the recorded 32 mhz bandwidth , 2 mhz ( 6668 - 6670 mhz )
was divided into 512 channels for maser reduction , yielding a velocity resolution of 0.176 km s@xmath1 .
this is four times broader than that of the single - dish observation .
a continuum source j2302@xmath46405 ( from cep a ) , whose coordinate is known with an accuracy of 0.62 mas in the third vlba calibrator survey ( vcs3 ) catalog @xcite , was used as a phase reference calibrator .
we alternately observed the cep a methanol maser and the continuum source with switching observational mode with a cycle of 5 min ( 2 min on cep a and 1.6 min on the continuum source ) .
cep a was sometimes continuously observed for more than 30-min to improve the uv - coverage .
the total on - source times were 2.8 hours and 0.7 hours for cep a and j2302@xmath46405 , respectively .
bright continuum sources , 3c 454.3 and 3c 84 , were also observed every one and half hours for clock and bandpass calibration .
the synthesized beam has a size of 9.0 @xmath8 3.5 mas with a pa of @xmath0 .
the data were reduced using the astronomical image processing system ( aips : @xcite ) and the difmap software @xcite .
the visibilities of all velocity channels were phase - referenced to the reference maser spot at lsr velocity of @xmath02.64 km s@xmath1 , which is the brightest spot .
the absolute coordinate of the spot of @xmath02.64 km s@xmath1 was obtained by applying the solution of phase for the velocity channel to the phase reference source j2302@xmath46405 .
we detected a rapid variability in the spectrum of 6.7 ghz methanol maser of cep a. five spectral features , at radial velocity of @xmath01.9 , @xmath02.7 , @xmath03.8 , @xmath04.2 , and @xmath04.9 km s@xmath1 , were identified during the whole observing period as shown in figure [ fig : fig1 ] .
the maser features were labeled as i , ii , iii , iv , and v , respectively . the feature ii showed a shoulder at @xmath02.5 km s@xmath1 , corresponding a second component . since we could not clearly distinguish the two components of feature ii , the second component is not discussed in this paper .
the maser features are divided into two groups ; two red - shifted features ( i , ii ) , and three blue - shifted features ( iii , iv , v ) .
these two groups are clearly separated in the spectrum .
they also exhibited different trends of variation as shown in figure [ fig : fig1]b .
for the first 30 days of the monitoring , the red - shifted group decreased to 50% of its initial value ( feature ii ) , while the blue - shifted group increased up to 300% ( feature v ) .
correlation coefficients for all combinations of spectral features for the data from doy 216 to 250 is shown in table [ tab : tab1 ] .
the absolute values of correlation coefficients are larger than 0.80 , with only exception of 0.63 for i vs iv .
the variation shows a strong correlation within each group , and an anti - correlation between two groups . a trend of the variation suddenly changed at around doy 250 .
the red - shifted group started to increase and the blue - shifted group started to decrease .
the rate of variation was smaller than that of previous period .
this change in the variation trend was clearly synchronized .
correlation coefficients for the whole period are shown in table [ tab : tab2 ] .
the coefficients show essentially the same trends with those for the first period .
the feature ii shows a strong and positive correlation ( 0.89 ) with the feature i , while anti - correlation was seen with the features iii , iv , and v. the correlation coefficients are @xmath00.91 , @xmath00.50 , and @xmath00.68 , respectively .
figure [ fig : fig2 ] shows the correlation plots for combinations with respect to the feature ii .
the positive ( i ) and anti - correlation ( iii , iv , v ) is obvious .
flux variation of feature iv and iii are quite similar each other , but the variation of feature iv is slightly advanced to that of feature iii .
cross - correlation of the time series of the features showed small time - delays .
the largest delay is 6 days for ii to iv , and the smallest is 1 day for ii to iii and v to iv .
we ignored the feature i because of its small variation .
( 80mm,80mm)figure1.eps .correlation coefficients from doy 216 to 250 [ cols="^ , > , > , > , > " , ] ( 160mm,160mm)figure2.eps with the vlbi observation , 117 spots of the methanol maser emission were detected .
peak intensities of the spots ranged from @xmath2650 mjy beam@xmath1 to 122 jy beam@xmath1 , while the rms of image noise ( 1@xmath9 ) in a line - free channel was 160 mjy beam@xmath1 .
the correlated flux accounted for over 90% of the single - dish flux .
the spatial distribution of the maser spots ( figure [ fig : fig3 ] ) showed an arclike structure .
the size from edge to edge of the arclike structure is @xmath21900 mas or @xmath21400 au at a distance of 725 pc .
the absolute coordinate of the spot at @xmath02.64 km s@xmath1 obtained in our observation is @xmath10(j2000.0 ) @xmath11 , @xmath12(j2000.0 ) @xmath11 @xmath4 , with errors less than 1 mas .
this is the origin of the image .
the peak of 43 ghz continuum emission ( star symbol ) , which may be an exciting source @xcite , located near the center of the arclike structure of the 6.7 ghz methanol maser spots .
the elongation of the arclike structure is nearly perpendicular to the radio jet .
the overall distribution of the maser spots coincide with the ch@xmath5cn and nh@xmath5 disks ( @xcite ; @xcite ) and the velocity range of the spots is similar to that of these disks , although a simple velocity gradient was not detected with the maser spots .
the water maser disk reported by @xcite locates almost the same position with the methanol arclike structure , although the size of the water maser disk is about 2 times smaller .
the ground - state hydroxyl masers around the hw2 object ( @xcite ; @xcite ) , whose internal proper motions were mainly directed away from the central source , are distributed surrounding the methanol maser distribution .
the radial velocities of both masers ( water : @xmath027.3 to @xmath48.9 km s@xmath1 ; hydroxyl : @xmath025.2 to @xmath00.6 km s@xmath1 ) cover the range of the methanol maser ( @xmath04.93 to @xmath00.36 km s@xmath1 ) .
the cluster of the spots locating near the origin of the vlbi map corresponds to the spectral features i and ii , and a cluster locating at east , south from the origin corresponds to the feature ii . the correspondences of the other clusters and the spectral features are as follows ; the western cluster ( west , north ) is iii , the eastern cluster ( east , south ) is iv , and the northern cluster ( east , north ) is v , respectively .
the clusters ii ( east , south ) and v were detected for the first time in vlbi observations . some weak spots with radial velocities from @xmath00.53 to @xmath00.36 km s@xmath1 , which are a part of the cluster ii , were detected only in the vlbi observation , and not detected in the spectral observation with yamaguchi 32-m telescope .
the red - shifted ( i , ii ) and the blue - shifted features ( iii , iv , v ) are isolated in the spatial distribution by more than 100 au , and the red - shifted features are surrounded by the blue - shifted features .
the time variation of the methanol maser features of cep a was synchronized , and ( anti-)correlated . the synchronized variation were occured at spatially isolated maser features .
the separation of clusters is larger than 100 au in the overall spatial distribution of @xmath21400 au , and it is not likely that there are some mechanical interactions between features that causes synchronized variation .
the synchronized variation may be caused by shock wave from a central star , that is arrived at the clusters simultaneously .
if we assume an outflow velocity of 4.5 km s@xmath1 which corresponds to a radial velocity range of the cep a methanol maser , it takes more than 1000 years to propagate through the spatial scale of 1000 au .
the synchronized arrival to the separated clusters with a time delay less than several days requires a very fine tuning of arrival time ( 6 days/1000 years @xmath20.002% ) , which is highly unlikely .
the synchronization , the spatial isolation and the spectral separation of the maser features suggest that it is difficult to excite the maser by collision in shock wave from a common exciting source .
these properties favor the widely accepted excitation model that the 6.7 ghz methanol masers is excited by infrared radiation from nearby warm dust ( e.g. , @xcite ) .
infrared radiation from one variable exciting source is easy to explain the synchronized variation with the spatial isolation and the spectral separation .
the exciting source might correspond to that one showed by @xcite .
the bolometric luminosity of the cep a region is about 2.5 @xmath8 10@xmath13 @xcite .
assuming that half of this luminosity is attributed to the hw2 object ( @xcite ; @xcite ) , we derive a dust temperature of 110 k at 700 au from the exciting source .
the distance 700 au is that from the supposed exciting source to the farthest maser spot .
this temperature 110 k is consistent with the suitable temperature ( @xmath2100 - 200 k ) of regions producing 6.7 ghz methanol maser ( e.g. , @xcite ) .
this time variation model of one exciting source may be applicable to the periodic variations of the 6.7 ghz methanol maser for [email protected] .
the time variation of the cep a methanol maser showed anti - correlation between the red - shifted and the blue - shifted features .
a light - crossing time of @xmath21400 au is 8 light - day .
it is consistent with the time - delays ( 1 - 6 days ) in the cross - correlation of the flux variation of spectral features .
however , the light - crossing time is too short to cause the anti - correlation with time - scale more than 20 days .
the anti - correlation could be understood in terms of excitation environment .
the 6.7 ghz methanol maser requires a suitable temperature range ( @xmath2100 - 200 k ) .
dust temperature is high at regions close to the exciting source , and is low at regions far from the source .
if the exciting source increases its luminosity , the temperature near the source would be too high to produce the maser emission , while the temperature far from the source would be suitable to produce the maser .
although the 3-dimensional distribution of maser features is uncertain , the 2-dimensional distribution of the maser features of cep a indicates that the red - shifted features are close to the supposed exciting source , and the blue - shifted features are far from it . with this spatial distribution , the anti - correlation of variation
could be explained by the dust temperature change caused by variability of the exciting source , i.e. , during the period between doy 216 and 250 , the dust temperature at feature ii , which is closer to the exciting source , increases beyond 200 k and out of the appropriate range for maser excitation , while the temperature at feature iii , iv , and v increased from 110 k , which is nearly the minimum value of maser excitation , to produce stronger maser emission .
additional observations are required to confirm this model , in particular to test if short time variation of luminosity occurs in the exciting source .
the distribution of maser features is observed as a 2-dimensional projection .
the velocity field would be a clue to understand the 3-dimensional structure of this region .
we are conducting the vlbi monitoring observations to detect the internal proper motion of methanol maser spots , and the results will be reported in future .
we have detected a rapid variability of 6.7 ghz methanol maser of cep a. five spectral features are grouped into the red - shifted ( @xmath01.9 and @xmath02.7 km s@xmath1 ) and the blue - shifted ( @xmath03.8 , @xmath04.2 , and @xmath04.9 km s@xmath1 ) .
these two groups are clearly separated in the spectrum .
they also exhibited different trends of variation .
the time variation of this maser features showed two remarkable properties ; synchronization and anti - correlation between the red - shifted and the blue - shifted features .
the spatial distribution of the maser spots obtained by the vlbi observation showed that maser spots of each feature are largely separated .
the synchronization , the spectral and the spatial isolation suggest that the collisional excitation by shock wave from a common exciting source is unlikely .
we discussed that the time variation of the cep a methanol maser could be produced by one variable exciting source and this maser could be excited by radiation from the exciting source .
it is possibly thought that the anti - correlation is caused by a variation of the dust temperature of the maser regions as a result of the distance to the exciting source , and the time variation of the exciting source s luminosity .
the authors wish to thank the jvn team for observing assistance and support .
the authors also would like to thank the anonymous referee for many useful suggestions and comments , which improved this paper .
the jvn project is led by the national astronomical observatory of japan ( naoj ) that is a branch of the national institutes of natural sciences ( nins ) , hokkaido university , tsukuba university , gifu university , yamaguchi university , and kagoshima university , in cooperation with geographical survey institute ( gsi ) , the japan aerospace exploration agency ( jaxa ) , and the national institute of information and communications technology ( nict ) . | we present the results of daily monitoring of 6.7 ghz methanol maser in cepheus a ( cep a ) using yamaguchi 32-m radio telescope as well as the results of imaging observations conducted with the jvn ( japanese vlbi network ) .
we indentified five spectral features , which are grouped into red - shifted ( @xmath01.9 and @xmath02.7 km s@xmath1 ) and blue - shifted ( @xmath03.8 , @xmath04.2 , and @xmath04.9 km s@xmath1 ) , and we detected rapid variabilities of these maser features within a monitoring period of 81 days .
the red - shifted features decreased in flux density to 50% of its initial value , while the flux density of the blue - shifted features rapidly increased within a 30 days .
the time variation of these maser features showed two remarkable properties ; synchronization and anti - correlation between the red - shifted and the blue - shifted .
the spatial distribution of the maser spots obtained by the jvn observation showed an arclike structure with a scale of @xmath21400 au , and separations of the five maser features were found to be larger than 100 au .
the absolute position of the methanol maser was also obtained based on the phase - referencing observations , and the arclike structure were found to be associated with the cep a - hw2 object , with the elongation of the arclike structure nearly perpendicularly to the radio continuum jet from the cep a - hw2 object .
these properties of the masers , namely , the synchronization of flux variation , and the spectral and spatial isolation of features , suggest that the collisional excitation by shock wave from a common exciting source is unlikely . instead , the synchronized time variation of the masers can be explained if all the maser features are excited by infrared radiation from dust which is heated by a common exciting source with a rapid variability . |
the study of the normalized ricci flow equation @xmath3 for a @xmath4-parameter family of riemannian metrics @xmath5 in a riemannian manifold @xmath6 was originally used by r. hamilton in @xcite and since then it has attracted the interest of many mathematicians ( cf .
@xcite , @xcite ) .
recently , there is an increasing interest towards the study of the ricci flow ( normalized or not ) on homogeneous spaces and under various perspectives ( @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite and references therein ) .
the aim of the present work is to study the normalized ricci flow for invariant riemannian metrics on generalized wallach spaces .
these are compact homogeneous spaces @xmath7 whose isotropy representation decomposes into a direct sum @xmath8 of three @xmath9-invariant irreducible modules satisfying @xmath10\subset \mathfrak{h}$ ] @xmath11 ( @xcite , @xcite ) . for a fixed bi - invariant inner product @xmath12 on the lie algebra @xmath13 of the lie group @xmath14 , any @xmath14-invariant riemannian metric @xmath15 on @xmath7 is determined by an @xmath9-invariant inner product @xmath16 where @xmath17 are positive real numbers . by using expressions for the ricci tensor and the scalar curvature in @xcite the normalized ricci flow equation ( [ ricciflow ] )
reduces to a system of ode s of the form @xmath18 where @xmath19 @xmath20 , are parameters of the invariant metric ( [ metric ] ) and @xmath21 here @xmath22 , @xmath23 , are the dimensions of the corresponding irreducible modules @xmath24 , @xmath25 and @xmath26 is some special nonnegative number ( see section [ wallach ] ) . if @xmath27 , then by denoting @xmath28 , @xmath23 , the functions @xmath29 can be expressed in a more convenient form ( independent of @xmath26 and @xmath22 ) as @xmath30 where @xmath31 it is easy to check that the volume @xmath32 is a first integral of the system ( [ three_equat ] ) .
therefore , on the surface @xmath33 we can reduce ( [ three_equat ] ) to the system of two differential equations of the type @xmath34 where @xmath35 it is known ( @xcite ) that every generalized wallach space admits at least one invariant einstein metric . later in @xcite ,
@xcite a detailed study of invariant einstein metrics was developed for all generalized wallach spaces .
in particular , it was shown that there are at most four invariant einstein metrics ( up to homothety ) for every such space .
it should be noted that invariant einstein metrics with @xmath36 correspond to singular points of ( [ two_equat ] ) , therefore , @xmath37 is a singular point of the system ( [ three_equat ] ) , ( [ volume ] ) if and only if @xmath38 is a singular point of ( [ two_equat ] ) .
it is our interest to determine the type of singularity of such points , and our investigation concerns this problem for some special values of the parameters @xmath39 , and @xmath40 .
the main result in this direction is theorem [ focus_center ] , which gives a qualitative answer for almost all points ( in measure theoretic sense ) @xmath41\times(0,1/2]\times ( 0,1/2]$ ] .
note that the latter inclusion is fulfilled for any triple @xmath1 corresponding to some generalized wallach spaces ( see the next section ) .
however we are interested in the behavior of the dynamical system ( [ two_equat ] ) for all values @xmath42 $ ] despite the fact that some triples may not correspond to `` real '' generalized wallach spaces .
we expect to give a more detailed study of the system ( [ two_equat ] ) for various values of the parameters @xmath43 , which could help towards a deeper understanding of the behavior of the ricci flow on more general homogeneous spaces .
also , it is quite possible that the system ( [ two_equat ] ) is interesting not only for the parameters @xmath44 $ ] , but as a more general dynamical system other than the ricci flow . _
it is clear that the system ( [ three_equat ] ) is naturally defined for all values of @xmath43 with @xmath45 , but for the system ( [ two_equat ] ) we should assume @xmath46 additionally_.
we recall the definition and important properties of generalized wallach spaces ( cf .
@xcite and @xcite ) .
consider a homogeneous almost effective compact space @xmath7 with a ( compact ) semisimple connected lie group @xmath14 and its closed subgroup @xmath47 .
denote by @xmath13 and @xmath48 the lie algebras of @xmath14 and @xmath47 respectively . in what follows ,
@xmath49 $ ] stands for the lie bracket of @xmath13 and @xmath50 stands for the killing form of @xmath13 .
note that @xmath51 is a bi - invariant inner product on @xmath13 .
consider the orthogonal complement @xmath52 of @xmath48 in @xmath13 with respect to @xmath53 .
every @xmath14-invariant riemannian metric on @xmath7 generates an @xmath9-invariant inner product on @xmath52 and vice versa ( @xcite ) .
therefore , it is possible to identify invariant riemannian metrics on @xmath7 with @xmath9-invariant inner products on @xmath52 ( if @xmath47 is connected then the property to be @xmath9-invariant is equivalent to the property to be @xmath54-invariant ) .
note that the riemannian metric generated by the inner product @xmath55 is called _ standard _ or _
killing_. let @xmath7 be a homogeneous space such that its isotropy representation @xmath52 is decomposed as a direct sum of three @xmath9-invariant irreducible modules pairwise orthogonal with respect to @xmath53 , i.e. @xmath56 @xmath57\subset \mathfrak{h } \mbox { for } i\in\{1,2,3\}.\ ] ] since this condition on each module resembles the condition of local symmetry for homogeneous spaces ( a locally symmetric homogeneous space @xmath7 is characterized by the relation @xmath58\subset \mathfrak{h}$ ] , where @xmath59 and @xmath52 is @xmath9-invariant @xcite ) , then spaces with this property were called _ three - locally - symmetric _ in @xcite .
but in this paper we prefer the term _ generalized wallach spaces _ , as in @xcite .
there are many examples of these spaces , e.g. the flag manifolds @xmath60 these spaces ( known as _ wallach spaces _ ) are interesting because they admit invariant riemannian metrics of positive sectional curvature ( see @xcite ) .
the invariant einstein metrics on @xmath61 were classified in @xcite and , on the remaining two spaces , in @xcite . in each of the cases , there exist exactly four invariant einstein metrics ( up to proportionality ) .
other classes of generalized wallach spaces are the various khler @xmath62-spaces such as @xmath63 @xmath64 the invariant einstein metrics in the above spaces were classified in @xcite .
each of these spaces admits four invariant einstein metrics ( up to scalar ) , one of which is khler for an appropriate complex structure on @xmath7 .
another approach to @xmath65 was used in @xcite .
the lie group @xmath66 ( @xmath67 ) is another example of a generalized wallach space .
being 3-dimensional , this group admits only one left - invariant einstein metric which is a metric of constant curvature ( @xcite ) . in @xcite , it was shown that every generalized wallach space admits at least one invariant einstein metric . this result could not be improve in general ( since e.g. @xmath66 admits exactly one invariant einstein metric ) . later in @xcite ,
@xcite a detailed study of invariant einstein metrics was developed for all generalized wallach spaces .
in particular , it is proved that there are at most four einstein metrics ( up to homothety ) for every such space .
denote by @xmath22 the dimension of @xmath24 .
let @xmath68 be an orthonormal basis in @xmath24 with respect to @xmath53 , where @xmath69 , @xmath70 .
consider the expression @xmath71 $ ] defined by the equality @xmath72=\sum_{\alpha,\beta,\gamma } \left\langle\big[e_i^\alpha , e_j^\beta \big],e_k^\gamma \right\rangle^2,\ ] ] where @xmath73 , @xmath74 , and @xmath75 range from 1 to @xmath22 , @xmath76 , and @xmath77 respectively .
the symbols @xmath71 $ ] are symmetric in all three indices by bi - invariance of the metric @xmath53 . moreover , for spaces under consideration , we have @xmath71=0 $ ] if two indices coincide .
therefore , the quantity @xmath78 $ ] plays an important role . by (
* lemma 1 ) we get @xmath79 for every @xmath23 with @xmath80 if and only if @xmath81=0 $ ] . note that @xmath82 if and only if the space @xmath7 is locally a direct product of three compact irreducible symmetric spaces ( see ( * ? ? ?
* theorem 2 ) ) .
suppose @xmath27 and let @xmath83 it is clear that @xmath42 $ ] .
note , that not every triple @xmath41 \times ( 0,1/2 ] \times ( 0,1/2]$ ] corresponds to some generalized wallach spaces .
for example , if @xmath84 for some @xmath85 , then there is no generalized wallach space with @xmath86 , where @xmath87 ( see lemma 4 in @xcite ) . moreover
, every @xmath88 should be a rational number for a generalized wallach space with simple group @xmath14 ( see ( [ ai ] ) , ( * ? ? ?
* lemma 1 ) and ( * ? ? ?
* table 1 ) ) .
an explicit expression for the ricci curvature of invariant metrics ( [ metric ] ) is obtained in lemma 2 of @xcite .
it immediately implies the system ( [ three_equat ] ) .
we will first give a description of the singular points of the system ( [ three_equat ] ) , as einstein metrics on generalized wallach spaces . an easy calculation
shows that for @xmath45 the singular points @xmath89 of the system ( [ three_equat ] ) can be found from the equations
@xmath90 if @xmath46 and @xmath91 , then we also get singular points of the system ( [ two_equat ] ) . recall that we are interesting only for singular points with @xmath92 , @xmath23 .
note that system ( [ two_equat_sing1 ] ) is homogeneous ( of degree 2 ) with respect to @xmath17 .
it is easy to see that @xmath93 implies @xmath94 or @xmath95 , @xmath96 .
if we have a solution with @xmath97 and @xmath98 , then we should have @xmath99 .
therefore , if @xmath100 for @xmath23 , then system ( [ two_equat_sing1 ] ) has no solution with zero component .
if @xmath101 $ ] , @xmath23 then it is proved in @xcite that this system has ( up to multiplication by a constant , for example , if we put @xmath102 ) at least one and at most four solutions with positive components . a detailed information on these solutions can be found in @xcite .
we briefly review these results below . _
the case where at least two of @xmath88 s are equal .
_ without loss of generality we may assume that @xmath103 and @xmath104
. then system ( [ two_equat_sing1 ] ) is equivalent to the following system : @xmath105 if @xmath106 then the second equation of ( [ solcase1 ] ) becomes @xmath107 thus , we have the following singular points @xmath108 where @xmath109 , @xmath110 , @xmath111 .
we observe that , for @xmath112 , we have only one family of singular points @xmath113 .
otherwise , @xmath114 , and all depends on the sign of the discriminant @xmath115 .
indeed , there exist one family of singular points for @xmath116 , two families for @xmath117 , and none for @xmath118 . if @xmath119 then @xmath120 , so the second equation of ( [ solcase1 ] ) reduces to @xmath121 if @xmath122 then equation ( [ solcase1.3 ] ) has no solution .
otherwise , @xmath123 , hence , all real roots of the equation ( [ solcase1.3 ] ) are positive .
the discriminant @xmath124 of ( [ solcase1.3 ] ) has the same sign as @xmath125 .
thus , there exist one family of singular points for @xmath126 , two families for @xmath127 , and none for @xmath128 . in particular ,
if @xmath129 , @xmath130 , then , for @xmath131 we get exactly four singular points @xmath89 up to a positive multiple , namely @xmath132 , @xmath133 , @xmath134 , or @xmath135 . for
@xmath136 we get only singular points proportional to @xmath132 .
_ the case of pairwise distinct @xmath88 s .
_ we consider two subcases here .
_ the case @xmath137 .
_ then all singular points @xmath89 have the form @xmath138 _ the case @xmath139 .
_ we look for singular points of the form @xmath140
. then system ( [ two_equat_sing1 ] ) be reduces to an equation of degree @xmath141 either in @xmath142 or in @xmath143 . by eliminating the summand containing @xmath144 , we obtain the following system equivalent to ( [ two_equat_sing1 ] ) : @xmath145 it is easy to see that @xmath146 ( see details in @xcite ) . by expressing @xmath143 from the first equation of ( [ solcase3 ] ) and inserting it into the second , we obtain the following equation of degree @xmath141 : @xmath147 denote by @xmath148 the discriminant of the polynomial in the left - hand side of ( [ solcase3.1 ] ) . it can be shown ( @xcite ) that all real solutions of ( [ solcase3.1 ] ) are positive . for @xmath149 , the equation ( [ solcase3.1 ] ) has either two or four distinct real solutions .
therefore , we get two or four families of singular points determined by ( [ solcase3.1 ] ) . note , that the condition @xmath150 holds , for example , for the homogeneous space @xmath151 ( @xmath152 , @xmath153 , @xmath154 ) .
in this special case , the equation ( [ solcase3.1 ] ) has one root of multiplicity 2 and the space under consideration admits exactly three pairwise nonhomothetic singular points ( i.e. invariant einstein metrics ) .
in this section we recall some facts about the type of singular points of the system ( [ two_equat ] ) .
our basic references are @xcite and @xcite .
the functions @xmath155 and @xmath156 ( see ( [ two_equat ] ) ) are analytic in a neighborhood of the arbitrary point @xmath38 ( where @xmath157 and @xmath158 ) and the following representations are valid : @xmath159 @xmath160 where @xmath161 , @xmath162 , @xmath163 and @xmath164 are the elements of the jacobian matrix @xmath165 the functions @xmath166 and @xmath14 are also analytic in a neighborhood of the point @xmath38 and @xmath167 the eigenvalues of @xmath168 can be found from the formula @xmath169 where @xmath170 we will use these notations in the sequel . in the non - degenerate case ( @xmath171 ) we will use the following theorem .
[ nondegenerate ] let @xmath172 be an isolated singular point of the system @xmath173 where @xmath26 and @xmath174 are analytic in a neighborhood of the origin with @xmath175 let @xmath176 and @xmath177 be the eigenvalues of the matrix @xmath178 which represents the linear part of the system at the origin .
then the following statements hold : \(i ) if @xmath176 and @xmath177 are real and @xmath179 , then @xmath172 is a saddle .
\(ii ) if @xmath176 and @xmath177 are real with @xmath180 and @xmath181 , then @xmath172 is a node . if @xmath182 ( respectively @xmath183 ) then it is unstable ( respectively stable ) .
\(iii ) if @xmath184 and @xmath185 with @xmath186 , then @xmath172 is a strong focus .
\(iv ) if @xmath187 and @xmath188 with @xmath189 , then @xmath172 is a weak focus or a center . cases ( i ) , ( ii ) and ( iii ) are known as _ hyperbolic singular points_. a direct calculation of @xmath190 and @xmath191 is often very complicated , so we will obtain more convenient formulas for @xmath191 and @xmath190 in the case of _ singular points @xmath38 of the system ( [ two_equat])_. let @xmath192 be the jacobian matrix of the map @xmath193 and let @xmath194 be the characteristic polynomial of @xmath192 . [ delta_rhon ] if @xmath38 is a singular point of the system ( [ two_equat ] ) , then @xmath195 and @xmath196 , where @xmath194 is calculated at the point @xmath197 . since the volume @xmath198 is a first integral of the system ( [ three_equat ] ) , then @xmath199 . on the other hand , @xmath200 and @xmath201 ,
hence @xmath202 .
therefore , for any @xmath23 we have @xmath203 at a singular point ( @xmath204 ) we have @xmath205 hence , @xmath206 at any singular point . by the same manner and by using ( [ simpltdif ] )
we get @xmath207 at any singular point . [ 1 ]
here we consider the case @xmath208 which has a special interest .
[ blow_up ] at @xmath208 the system of ode s ( [ two_equat ] ) has an unique ( isolated ) singular point @xmath209 which is a saddle with six hyperbolic sectors . as the calculations show the unique singular point of ( [ two_equat ] )
is indeed @xmath210 ( see figure [ degsing ] for a phase portrait in a neighborhood of this point ) .
note that in this case @xmath211 by moving @xmath212 to the origin and by the analyticity of @xmath213 and @xmath214 at @xmath212 , system ( [ two_equat ] ) can be reduced to the equivalent system @xmath215 where @xmath216 and @xmath217 ( @xmath218 ) are some homogeneous polynomials of degree @xmath85 with respect to @xmath219 and @xmath220 ( @xmath221 , @xmath222 ) .
next we use results from @xcite . using the blowing - up @xmath223 ,
@xmath224 we obtain the system @xmath225 where @xmath226 @xmath227 we have the case of subsection 6.2 of @xcite where the equation @xmath228 has three different real roots .
so it is obvious that the blowing - up system has the singular points @xmath229 , @xmath230 and @xmath231 .
we show that all of these singular points are saddles .
let @xmath232 and let @xmath233 be the matrix of the linear part of the blowing - up system at the point @xmath234 , @xmath23
. then @xmath235 with eigenvalues equal to @xmath236 and @xmath237 .
it is clear that @xmath238 for all @xmath23 .
therefore the eigenvalues of @xmath233 have different signs , and by theorem [ nondegenerate ] all the singular points @xmath234 , @xmath23 , are saddles .
the phase portrait of the blowing - up system is identical to one shown in figure 7(b ) @xcite .
the saddles @xmath234 , @xmath23 , correspond to the unique singular point @xmath172 of the initial system .
according to the qualitative classification of singular points of degree @xmath239 given by @xcite the point @xmath172 is also a saddle with six hyperbolic sectors near it ( see figure 3(12 ) in @xcite ) .
recall that the systems ( [ three_equat ] ) and ( [ two_equat ] ) are well defined for all @xmath241 with @xmath45 ( with the additional restriction @xmath46 for the system ( [ two_equat ] ) ) .
the special case considered in section [ onespc ] leads us to consider the set @xmath242 we will show a convenient way to deal with degenerate singular points of the system ( [ two_equat ] ) in lemma [ sing.one ] below .
a direct calculation using lemma [ delta_rhon ] gives the following : [ sing.zero ] let @xmath194 be the characteristic polynomial of the jacobi matrix of the system ( [ three_equat ] ) at any point @xmath89 with @xmath243
. then @xmath244 , @xmath245 and @xmath246 , where @xmath247 @xmath248 and @xmath249 .
[ sing.one ] a singular point @xmath250 of the system ( [ two_equat ] ) is degenerate if and only if the point @xmath251 satisfies the equation @xmath252 , where @xmath253 is given by ( [ singval1 ] ) . by lemma [ delta_rhon ] a singular point @xmath250 of the system ( [ two_equat ] )
is degenerate if and only if @xmath254 .
note that the right hand side of equation ( [ singval1 ] ) is homogeneous in the variables @xmath17 .
therefore , without loss of generality we may consider these variables up to a positive multiple .
obviously from lemma [ sing.zero ] it follows that @xmath255 is equivalent to @xmath252 .
now , we can represent the set @xmath240 as an algebraic surface in @xmath256 .
[ sing.two ] a point @xmath1 with @xmath45 and @xmath46 lies in the set @xmath240 if and only if @xmath257 , where @xmath258 and @xmath259 under the assumptions @xmath45 , @xmath46 , and @xmath91 , the singular points @xmath250 of the system ( [ two_equat ] ) can be found from the equations ( [ two_equat_sing1 ] ) , and according to lemma [ sing.one ] they are degenerate if and only if @xmath252 . note that the equations ( [ two_equat_sing1 ] ) and @xmath252 are homogeneous with respect to @xmath17 . setting @xmath102 and eliminating @xmath260 and @xmath261 from these three equations ( e. g. using _ maple _ or _ mathemathica _ )
we obtain the equation @xmath262 note that for @xmath263 , @xmath264 , @xmath265 , and @xmath266 we have no additional triples @xmath1 of `` degenerate '' parameters .
therefore , the first four factors in the above equation can be ignored .
another reason to ignore them is the symmetry of the problem under the permutation @xmath267 .
it is easy to see that @xmath268 is a symmetric polynomial in @xmath43 of degree 12 .
therefore , the equation @xmath257 ( without the restrictions @xmath45 and @xmath46 ) defines an algebraic surface in @xmath256 that we may identify with the set @xmath240 .
[ 1 ] _ in the rest of this section , we consider only points @xmath269\times(0,1/2]\times ( 0,1/2]$ ] . _
it is very important to describe in details the set @xmath270 ^ 3 = \{(a_1,a_2,a_3)\in ( 0,1/2]\times ( 0,1/2]\times ( 0,1/2]\ , : \,q(a_1,a_2,a_3)=0\}.\ ] ] as usual , the most complicated and interesting problem is the study of this surface in neighborhoods of singular points of @xmath240 determined by @xmath271 . for @xmath272 the equation
@xmath273 is equivalent to @xmath274 where @xmath275 and @xmath276 .
if @xmath277 $ ] then this set is a curve homeomorphic to the interval @xmath278 $ ] with endpoints @xmath279 and @xmath280 and with the singular point ( a cusp ) at the point @xmath281 ( see figure [ pictur1 ] ) .
the same is also valid under the permutation @xmath267 .
note that for @xmath282 the equation @xmath273 is equivalent to @xmath283 .
it is easy to check that for @xmath284 $ ] the equality @xmath285 holds only when @xmath286 is one of the points @xmath172 , @xmath287 , and @xmath288 . therefore , @xmath282 in the set @xmath289 ^ 3 $ ] only for points in the boundary of the triangle with vertices @xmath290 , @xmath291 , and @xmath292 . for all other points in @xmath293
^ 3 $ ] we have the inequality @xmath294 .
it is clear that @xmath295 .
note that for @xmath296 the equation @xmath273 is equivalent to @xmath297 it is not difficult to show that @xmath298 is the only point in @xmath289 ^ 3 $ ] satisfying the additional condition @xmath296 .
it turns out that the point @xmath298 is a singular point of degree @xmath299 of the algebraic surface @xmath240 ( see figure [ singsur ] ) .
the type of this point is * elliptic umbilic * in the sense of darboux ( see @xcite and @xcite ) or of type @xmath300 in other terminology ( see e. g. ( * ? ? ?
* chapter iii , sections 21.3 , 22.3 ) ) .
in this section we study the singular points of the system ( [ two_equat ] ) according to the signs of @xmath301 and @xmath191 ( see ( [ sirhodel ] ) ) .
results depend on conditions on the parameters @xmath0 .
[ sigma_minim ] the quadratic form @xmath302 where @xmath249 , is non - negative if @xmath303 ( in particular , if @xmath304 for @xmath23 ) and achieves its absolute minimum ( equal to zero ) exactly at the points @xmath305 it is easy to show that the matrix of the form @xmath14 has non negative eigenvalues @xmath306 , @xmath307 , and @xmath308 .
obviously , the last two numbers are positive for @xmath303 .
therefore @xmath14 is a non - negative form .
note that the equation @xmath309 has the solutions given in the statement of the lemma .
[ nonode ] for @xmath310 all singular points of the system ( [ two_equat ] ) are such that @xmath311 . in particular , a non - degenerate singular point ( i.e. @xmath312 ) of ( [ two_equat ] )
is either a node ( if @xmath313 ) or a saddle ( if @xmath314 ) .
let @xmath315 . by lemmas [ delta_rhon ] and [ sing.zero ]
we get @xmath316 using lemma [ sigma_minim ] for @xmath317 , @xmath318 and @xmath319 , we get that @xmath320 for all @xmath321 , @xmath322 .
so , in particular , @xmath320 holds for all singular points of the system ( [ two_equat ] ) , and by theorem [ nondegenerate ] its non degenerate singular points can be only either a node or a saddle .
[ sigmaziro ] from theorem [ nonode ] and lemma [ sigma_minim ] we get the following . if @xmath250 is a singular point of the system ( [ two_equat ] ) with @xmath323 then @xmath324 for a unique @xmath325 , @xmath326 , determined by the equality @xmath322 .
next we are interested for those values @xmath269\times(0,1/2]\times ( 0,1/2]$ ] such that the system ( [ two_equat ] ) has at least one singular point with @xmath323 .
[ a_i_ensur_sing_points_sigma=0 ] the only two families of the parameters @xmath88 , @xmath23 satisfying the conditions @xmath101 $ ] and which can give singular points of the system ( [ two_equat ] ) with the property @xmath323 , are the following : @xmath327,\ ] ] @xmath328 where @xmath329 , @xmath330 ( @xmath331 , @xmath332 ) .
consider the system of equations ( [ two_equat_sing1 ] ) and ( [ sigma=0 ] ) under the substitutions @xmath333 then @xmath334 and equations ( [ two_equat_sing1 ] ) take the form @xmath335 by eliminating @xmath336 ( and , independently , @xmath337 ) from ( [ two_equat_sing1_new ] ) we obtain the following two equations : @xmath338 @xmath339 solving this equations together ( and taking into account the equations ( [ two_equat_sing1_new ] ) ) we get the following four families of solutions of ( [ two_equat_sing1_new ] ) , that include solutions under the conditions @xmath340 ( @xmath23 ) : @xmath341 @xmath342 @xmath343 @xmath344 where @xmath345 , @xmath346 , @xmath331 , @xmath332
. now we return to the original parameters @xmath88 ( @xmath23 ) . from the first family @xmath347 we obtain ( [ first_for_sigma=0 ] ) . the second family @xmath348
, @xmath349 gives ( [ second_for_sigma=0 ] ) .
it is clear that @xmath42 $ ] for all @xmath23 if @xmath350 .
the third and fourth families of @xmath351 give @xmath352 it is easy to show that the system of inequalities @xmath304 , @xmath23 , is incompatible in this case .
[ sing_points_with_sigma=0 ] now we can find all singular points of the system ( [ two_equat ] ) corresponding to the families ( [ first_for_sigma=0 ] ) , ( [ second_for_sigma=0 ] ) and having @xmath323 . according to remark [ sigmaziro ] the family ( [ first_for_sigma=0 ] ) gives a unique singular point @xmath353 of ( [ two_equat ] ) satisfying @xmath323 for all @xmath354 $ ] .
analogously , by remark [ sigmaziro ] it follows that the family ( [ second_for_sigma=0 ] ) gives only the following singular points of ( [ two_equat ] ) satisfying @xmath323 for all @xmath355 : @xmath356 where @xmath357 is determined by the condition @xmath358 ( see ( [ volume ] ) ) .
[ roots_of_q ] using theorem [ a_i_ensur_sing_points_sigma=0 ] and lemma [ sing.two ] we can detect all values of @xmath101 $ ] , @xmath23 , such that the system ( [ two_equat ] ) has at least one singular point with @xmath359 . according to remark [ sing_points_with_sigma=0 ] for all @xmath354 $ ] the family ( [ first_for_sigma=0 ] ) gives a unique singular point @xmath353 of ( [ two_equat ] ) with @xmath323 .
in this case @xmath360 ( see ( [ singval2 ] ) ) takes the form @xmath361 and the equation @xmath273 implies that @xmath362 . then according to lemma [ sing.two ] the point @xmath212 is a degenerate singular point ( @xmath363 ) only for @xmath362 ( and its type has been determined in theorem [ blow_up ] ) . for @xmath364 $ ]
the point @xmath212 is a node .
analogously , for the family ( [ second_for_sigma=0 ] ) we have that @xmath365 and the equation @xmath273 has only three positive roots @xmath366 , @xmath367 and @xmath368 , but none of these values belongs to the interval @xmath369 .
therefore , ( [ second_for_sigma=0 ] ) can not give singular points of ( [ two_equat ] ) with @xmath359 .
next we denote by @xmath370 the set of points @xmath1 such that there is a singular point @xmath38 of the system ( [ two_equat ] ) with @xmath371 . recall that for points with @xmath372 the system ( [ two_equat ] ) is undefined .
[ sign_rho ] a point @xmath1 with @xmath45 and @xmath46 lies on the surface @xmath370 if and only if @xmath373 the equations ( [ two_equat_sing1 ] ) and ( [ singval0 ] ) are homogeneous with respect to @xmath17 . now setting @xmath102 and eliminating @xmath260 and @xmath261 from the above three equations ( using e. g. _ maple _ or _ mathemathica _ ) we get the equations @xmath374 note that for @xmath265 and @xmath266 we have no additional sets of parameters @xmath1 .
this finishes the proof .
[ singdegenp ] it is easy to show that system ( [ two_equat ] ) has a singular point with @xmath375 if and only if @xmath376 ( in this case system ( [ two_equat ] ) has exactly one degenerate singular point @xmath353 , see section [ onespc ] ) .
indeed , from the equations @xmath371 and @xmath363 we have @xmath377 .
according to remark [ roots_of_q ] , the system of equations @xmath363 , @xmath323 has a unique solution @xmath376 .
it is easy to check that this solution satisfies the equation ( [ signrho ] ) as well .
therefore , theorem [ sign_rho ] implies that @xmath371 .
we first discuss a part of the surface @xmath240 ( see section [ degenerset ] ) in the cube @xmath379 .
recall that @xmath240 is invariant under the permutation @xmath380 .
it should be noted that the set @xmath381 is connected ( it can be shown by lengthy computations using suitable geometric tools ) .
there are three curves ( `` edges '' ) of _ singular points _ on @xmath240 ( i. e. points where @xmath382 ) : one of them has parametric representation @xmath383 , and the others are defined by permutations of @xmath88 .
these curves have a common point @xmath298 which is an elliptic umbilic on the surface @xmath240 ( see figure [ singsur ] ) .
the part of @xmath240 in @xmath379 consists of three ( pairwise isometric ) `` bubbles '' spanned on every pair of `` edges '' ( cf .
pictures of elliptic umbilics at pp .
6491 of @xcite ) . the gaussian curvature at every non - singular point of the surface @xmath384 is negative , as it could be checked by direct calculations .
it should be also noted that the point @xmath376 is the unique singular point of the surface @xmath370 ( @xmath385 ) .
this is clear by reducing ( [ signrho ] ) to the simpler equation @xmath386 using the substitutions @xmath387 .
it is easy to see that @xmath370 divides the cube @xmath388 ^ 3 $ ] into three domains @xmath389 , @xmath390 , and @xmath391 containing the points @xmath392 , @xmath393 , and @xmath394 respectively . from the above discussion and some geometric considerations ( that could be rigorous but very lengthy )
, we see that the set @xmath378 has exactly three connected components .
denote by @xmath395 , @xmath396 , and @xmath397 the components containing the points @xmath398 , @xmath399 , and @xmath400 respectively .
let us fix @xmath401
. by the definition of @xmath240 , for all points @xmath402 system ( [ two_equat ] ) has only non degenerate singular points .
the number of these points and their corresponding types are the same on each component @xmath403 ( under some suitable identification for various values of parameters @xmath43 ) .
therefore , it suffices to check _ only one point in the set @xmath403_. one of the main results of this paper is the following theorem which clarifies the above observation and provides a general result about the type of the non degenerate singular points of the system ( [ two_equat ] ) . [ 1 ] [ focus_center ] for @xmath404 the following possibilities for singular points of the system ( [ two_equat ] ) can occur : \1 ) if @xmath405 or @xmath406 then there is one singular point with @xmath407 ( a node ) and three singular points with @xmath408 ( saddles ) ; \2 ) if @xmath409 then there are two singular points with @xmath408 ( saddles ) . by theorem [ nonode ] a non degenerate singular point is either a node ( if @xmath313 ) or a saddle ( if @xmath314 ) . recall that for @xmath404 all singular points of the system ( [ two_equat ] ) are not degenerate . moreover , there are no singular points @xmath89 with some zero component . therefore , * the number of singular points * and * the set of signs * of @xmath410 for these points is constant on each component @xmath403 .
it is easy to check it for a small neighbourhood of any point @xmath402 ( it follows from _ the stability of degenerate singular points _ ) , but it could be spread to all points of the connected set @xmath403 via continuous paths ( as in standard analytical monodromy theorems ) . consider the component @xmath395 containing the representative point @xmath398 .
then as the calculations show , under @xmath102 the system of equations ( [ two_equat_sing1 ] ) in the variables @xmath250 has four solutions , given by @xmath212 , @xmath411 , @xmath412 and @xmath413 . by lemma [ sing.zero ]
it follows that the point @xmath353 corresponds to the value @xmath414 ( a node with @xmath415 and @xmath416 ) .
if @xmath250 is one of the solutions @xmath411 , @xmath412 and @xmath413 , then @xmath417 equals to @xmath418 , @xmath419 and @xmath418 respectively ( so these points are saddles ) .
consider now the component @xmath396 containing the representative point @xmath399 . by the same manner using lemmas
[ delta_rhon ] and [ sing.zero ] we get the following four solutions @xmath250 of the system ( [ two_equat_sing1 ] ) : @xmath212 with @xmath420 ( a node with @xmath421 and @xmath416 ) , and @xmath422 , @xmath423 , @xmath424 with @xmath417 equal to @xmath425 , @xmath425 and @xmath426 respectively ( three saddles ) .
finally , consider the component @xmath397 containing the point @xmath427 .
in this case we get the following two solutions of the system ( [ two_equat_sing1 ] ) : @xmath428 with @xmath429 ( a saddle ) ;
@xmath430 with @xmath431 ( a saddle ) . note that all points @xmath1 with @xmath304 and @xmath137 are in @xmath395 ( see discussion after the proof of lemma [ sing.two ] ) .
therefore , by theorem [ focus_center ] we get that among the singular points determined by ( [ solcase2 ] ) there is one node and three saddles .
it is possible to get more detailed information in this case .
the first family of the solutions ( [ solcase2 ] ) corresponds to a _
unstable _ node and the other families give saddles .
analogous results can be obtained for all other singular points given in section [ sinpsec ] . namely we have also proved that in the case @xmath117 the family ( [ solcase1.2 ] ) can give only hyperbolic ( @xmath432 , @xmath433 ) or semi - hyperbolic ( @xmath434 , @xmath435 ) singular points of ( [ two_equat ] ) .
note that in the hyperbolic case , nodes are stable if @xmath436 and unstable if @xmath437 . if @xmath116 , @xmath438 then family ( [ solcase1.2 ] ) also gives hyperbolic or semi - hyperbolic singular points .
note that ( [ solcase1.2 ] ) gives no singular points of ( [ two_equat ] ) of the nilpotent type ( @xmath439 , @xmath440 , see ( [ linear ] ) ) .
we plan to discuss results of such kind in another paper .
recall also that we have considered in section [ onespc ] the special case @xmath116 , @xmath441 with a unique linearly zero ( @xmath442 ) singular point of ( [ two_equat ] ) .
theorem [ focus_center ] gives a general picture for types of singular points of the system ( [ two_equat ] ) with @xmath443 .
nevertheless , it would be interesting to study `` degenerate '' sets of parameters @xmath1 from the set @xmath2 ^ 3\cap \omega$ ] .
for the point @xmath376 we obtained suitable results in section [ onespc ] .
it should be noted that the point @xmath376 is a very special one on the algebraic surface @xmath444 ( see section [ degenerset ] and also remark [ singdegenp ] ) .
the following questions are worth for a further investigation .
einstein metrics on even - dimensional homogeneous spaces admitting a homogeneous riemannian metric of positive sectional curvature // sibirsk . mat .
( 1991 ) , v. 32 , no . 3 , p.126131 ( russian ) ; english translation in : siberian math . j. ( 1991 ) , v. 32 , no . 3 , p. 455459 | we consider the asymptotic behavior of the normalized ricci flow on generalized wallach spaces that could be considered as special planar dynamical systems .
all non symmetric generalized wallach spaces can be naturally parametrized by three positive numbers @xmath0 .
our interest is to determine the type of singularity of all singular points of the normalized ricci flow on all such spaces .
our main result gives a qualitative answer for almost all points @xmath1 in the cube @xmath2\times ( 0,1/2]\times ( 0,1/2]$ ] .
key word and phrases : riemannian metric , einstein metric , generalized wallach space , ricci flow , ricci curvature , planar dynamical system , real algebraic surface . _ 2010 mathematics subject classification : _
53c30 , 53c44 , 37c10 , 34c05 , 14p05 .
the project was supported in part by the state maintenance program for the leading scientific schools of the russian federation ( grant nsh-921.2012.1 ) and by federal target grant `` scientific and educational personnel of innovative russia '' for 2009 - 2013 ( agreement no . 8206 , application no .
2012 - 1.1 - 12 - 000 - 1003 - 014 ) . |
let @xmath0 be a polynomial ring over a field @xmath1 . then given a hilbert function @xmath2 it is easy to see that there can be more than one ideal @xmath3 such that the hilbert function of @xmath4 is @xmath2 .
one can further distinguish such ideals by passing to a finer invariant , the graded betti numbers , which gives rise to the question : given a particular hilbert function @xmath2 , what sets of graded betti numbers actually occur ?
that this problem is bounded above , and hence finite , is due to an important result by bigatti and hulett @xcite ( independently in characteristic zero ) , and pardue @xcite ( in characteristic @xmath5 ) which says that , given a hilbert function @xmath2 , the lexicographic ideal attaining @xmath2 has everywhere largest graded betti numbers . in fact , this says that the partial order on the set of sets of graded betti numbers of ideals attaining a given hilbert function has a unique maximum element .
shortly thereafter , it was shown by charalambous and evans @xcite that this order need not have a unique minimal element .
examples of infinite families of hilbert functions which did not support unique minimal elements were given by richert @xcite .
the structure of the partial order on graded betti numbers has also been considered on interesting subsets of the set of all ideals attaining a given hilbert function .
for instance , geramita , harima , and shin @xcite showed that if one restricts to the graded betti numbers arising from a certain ( dense ) set of gorenstein ideals , then the associated partial order has a unique maximal element which is constructed using a lexicographic ideal ( migliore and nagel @xcite were later able to extend this result to an even larger set of gorenstein ideals attaining a given hilbert function ) while richert @xcite showed that a unique minimal element need not exist .
for stable ideals in dimension at most three , francisco @xcite showed that , unlike the general and gorenstein cases , there is always a unique smallest element .
it is a result of aramova , herzog , and hibi @xcite that if one restricts to the graded betti numbers arising from squarefree monomial ideals attaining a given hilbert function , then the associated partial order has a unique maximal element ( which arises , not surprisingly , from the squarefree lexicographic ideal ) .
squarefree monomial ideals are particularly interesting ( to algebraists , topologists , and combinatorists alike ) because to each squarefree monomial ideal @xmath6 in @xmath7 variables can be associated a simplicial complex @xmath8 on @xmath7 vertices , while the hilbert function of @xmath4 is related to the face counts of @xmath8 and the graded betti numbers of @xmath4 can be computed by considering sums of the ranks of the reduced homologies of subsets of @xmath8 .
it was known that ( gelvin , lavictore , reed , richert @xcite ) for @xmath9 variables ( and after fixing a finite field ) , the partial orders arising from fixing a hilbert function were totally ordered , but that this failed in six variables where , in fact , there is an example of a partial order which does not have a unique smallest element . in the current paper , we continue this line of inquiry .
first , we generate an infinite family ( the size of which grows exponentially with dimension ) of hilbert functions for which the partial order on the graded betti numbers corresponding to squarefree monomial ideals fails to have a unique minimal element .
we are able to show that this same family of hilbert functions gives rise to partially ordered sets corresponding to all ideals ( not only squarefree monomial ideals ) which fail to have a unique minimal element .
we then find an infinite family ( again , growing exponentially ) of hilbert functions for which the partial order on the betti numbers ( not graded betti numbers ) of squarefree monomial ideals fails to have a unique smallest element ( and are again able to show that this family gives posets without unique minimal elements in the case of all ideals ) .
next , we find an infinite family ( growing exponentially ) of hilbert functions for which the partial order associated to the graded betti numbers of squarefree monomial ideals has a single element , and thus a unique smallest element .
we note an analogous family in the general case . finally , we find a infinite family ( growing linearly ) of hilbert functions for which the partial order associated to the graded betti numbers of squarefree monomial ideals has a unique smallest element , and a nontrivial poset tree .
let @xmath10 $ ] where @xmath1 is a field . in what follows
, all ideals will be homogeneous . in this paper
, we will be studying the graded betti numbers of ideals with a fixed hilbert function .
recall that given a homogeneous ideal @xmath11 , the _ hilbert function _ of @xmath4 in degree @xmath12 , denoted as @xmath13 is given by @xmath14 furthermore , we will write @xmath15 to be a minimal free resolution of @xmath4 , and @xmath16 to be the set of graded betti numbers of @xmath4 .
it is useful to display graded betti numbers in the following table known as a betti diagram ( using the notation of macaulay 2 @xcite ) .
@xmath17 where @xmath18 is the sum of the entries in the @xmath19 column .
then @xmath20 is the set of betti numbers .
there is an obvious partial order on the betti diagrams which arise for ideals with a given hilbert function .
if @xmath21 and @xmath22 are the graded betti numbers of the ideals @xmath6 and @xmath23 , then we say that @xmath24 if @xmath25 for all @xmath26 and @xmath27 .
furthermore , @xmath28 if @xmath24 and there is a pair @xmath29 such that @xmath30 .
we are interested in the extremal properties of this ordering .
a useful definition in this regard is that of @xmath31-linearity : a minimal free resolution of an ideal @xmath6 is called @xmath31-linear if @xmath6 is minimally generated by @xmath31-forms and @xmath32 for each @xmath33 and @xmath34 .
it turns out that the graded betti numbers are a finer invariant than hilbert functions .
they are related by the following useful equation .
@xcite given an ideal @xmath35 $ ] with graded betti numbers @xmath21 , @xmath36 it follows from this formula that the diagonal alternating sums of a betti diagram are invariant for all betti diagrams of ideals attaining a given hilbert function .
we define this alternating sum as follows : given an ideal @xmath6 , the @xmath37 diagonal alternating sum of the betti diagram of @xmath6 is : @xmath38 where the @xmath39 are the graded betti numbers of @xmath6 .
the partial ordering on betti diagrams has a unique maximal betti diagram @xcite . in order to identify an ideal attaining the maximal graded betti numbers , we need the following notation : let @xmath40 and @xmath41 be monomials such that @xmath42 ( i.e. they are of the same degree ) .
we say @xmath43 if and only if the first nonzero entry of @xmath44 is positive .
this is known as the _ lexicographic _ or _ lex _ ordering on monomials .
an ideal @xmath45 is a _ lex ideal _ if for all monomials @xmath46 and @xmath47 , then @xmath48 implies @xmath49 .
a lex ideal achieves the maximal graded betti numbers among all ideals attaining its hilbert function . in this paper
we are mainly concerned with squarefree monomial ideals .
squarefree monomial ideal _ is an ideal in @xmath50 $ ] minimally generated by elements of the form @xmath51 , where @xmath52 .
we will make use of the fact that the squarefree monomial ideals are in one - to - one correspondence with simplicial complexes .
recall the definition of a simplicial complex : a _ simplicial complex _
@xmath53 on the vertex set @xmath54 is a collection of subsets called faces or simplices , closed under taking subsets ; that is if @xmath55 is a face and @xmath56 , then @xmath57 .
a simplex @xmath55 of cardinality @xmath58 has dimension @xmath26 and is called an @xmath26-face . the ( well known , see for instance @xcite ) procedure to pass from a simplicial complex @xmath53 to a squarefree monomial ideal is to form the ideal generated by monomials corresponding to the minimal non - faces of @xmath53 , after renaming the vertex set of @xmath53 to be @xmath59 .
the simplicial complex @xmath60 corresponds to the ideal @xmath61 .
we use dashed lines in the picture to indicate minimal non - faces .
this correspondence gives a bijection between simplicial complexes and squarefree monomial ideals with no linear terms .
in general , we will conflate simplicial complexes and ideals due to this correspondence , and so we are free to talk about the betti numbers of a simplicial complex or the homology of an ideal . given a simplicial complex with @xmath62 @xmath26-faces , its corresponding _
@xmath63-vector _ is the @xmath7-tuple @xmath64 .
for example , consider the simplicial complex which has @xmath63-vector @xmath65 . here
the @xmath63-vector counts the 4 0-faces ( or points ) , 4 1-faces ( or edges ) , 1 2-face , and 0 3-faces ( or volume ) .
the @xmath63-vector is an important invariant of a simplicial complex because of the following : @xcite let @xmath6 be the stanley - reisner ideal of a simplicial complex with @xmath63-vector @xmath64 .
then @xmath66 from the theorem , it follows that @xmath63-vectors and hilbert functions are in one - to - one correspondence . given a hilbert function which arises for squarefree monomial ideals
, we will be particularly interested in the corresponding squarefree lex ideal ( because it always exists and is known to exhibit the maximal betti diagram @xcite ) .
an ideal @xmath45 is a _ squarefree lex ideal _ if for squarefree monomials @xmath67 and @xmath68 , @xmath69 implies @xmath49 .
a useful tool in studying simplicial complexes ( and therefore their associated ideals ) is simplicial homology . if @xmath53 is a simplicial complex , we let @xmath70 be the set of all @xmath71-faces in @xmath53 , and let @xmath72 denote the @xmath1 vector space whose basis consists of @xmath70 . we define the boundary map @xmath73@xmath74to be @xmath75 where the hat indicates omission .
we define the @xmath70 subspaces @xmath76@xmath77 and @xmath78 . by standard algebraic topology , we have @xmath79 , and so we can define the @xmath80 of @xmath53 , @xmath81 , to be the @xmath1 vector space @xmath82 .
in addition , the @xmath83 @xmath80 @xmath84 is given by @xmath85 , and @xmath81 when @xmath86 .
we now present the connection between the simplicial homology of certain subsets of @xmath53 and graded betti numbers , in the form of hochster s formula . to properly express this formula , we need a new notation .
if @xmath87 , then given a simplicial complex @xmath53 , let @xmath88 be the simplicial complex defined by @xmath89 where @xmath90 denotes power set .
then we have @xcite let @xmath91 be a squarefree monomial ideal , with @xmath53 the associated simplicial complex .
then we have@xmath92 for all i and j.
in this section we construct a fast - growing family of hilbert functions which fail to have unique minimal graded betti numbers among both squarefree monomial ideals , and all ideals .
we proceed using simplicial complexes , giving a method for preserving parts of the betti diagrams which guarantee incomparability .
our methods allow us to double the number of @xmath63-vectors in each dimension with this property , and so our family of @xmath63-vectors grows exponentially with the dimension of the polynomial ring .
let @xmath53 be a simplicial complex on @xmath7 vertices .
we define the @xmath27-cone of @xmath53 on the vertex @xmath93 , denoted @xmath94 , to be the simplicial complex on @xmath95 vertices such that @xmath96 if and only if either @xmath97 or @xmath98 , @xmath99 and @xmath100 .
as suggested by the definition , we define @xmath101 to be the @xmath7-cone of @xmath53 , @xmath102 .
let @xmath53 be the simplicial complex @xmath103 with @xmath63-vector @xmath65 .
then @xmath104 and , of course , @xmath105 .
here @xmath106 has @xmath63-vector @xmath107 , @xmath108 has @xmath63-vector @xmath109 , @xmath110 has @xmath63-vector @xmath111 , and @xmath112 has @xmath63-vector @xmath113 . on of the nice things about coning a simplicial complex
is that the graded betti numbers do not change .
we show below that we can similarly preserve certain of the graded betti numbers of a simplicial complex after @xmath27-coning .
[ lem : diagonalspreserved]given a simplicial complex @xmath53 , the betti diagram of the simplicial complex @xmath94 will be identical to the betti diagram of @xmath53 for the first @xmath114 diagonals .
we employ hochster s formula .
let @xmath115 .
then we have @xmath116 @xmath117 however , from the definition of @xmath27-coning and the complex @xmath88 , we have that for any @xmath118 with @xmath119 and at most @xmath27 vertices in @xmath120 , @xmath88 is a cone . by standard algebraic topology ,
a cone is contractible and so all of its reduced homology spaces are zero .
therefore , the second term in the above sum vanishes , and as the first term is @xmath121 by hochster s formula , the lemma is proved .
[ def : coningdefs]given @xmath27 , and an @xmath63-vector @xmath122 of a simplicial complex @xmath53 , define @xmath123}$ ] to be the @xmath124 entry of the @xmath63-vector @xmath125 . also , for @xmath126 , define@xmath127 if the @xmath63-vector of @xmath53 is @xmath122 , then it is easy to show that the @xmath63-vector of @xmath94 is : @xmath128 or , in the notation of definition [ def : coningdefs ] , @xmath129}_{0},f^{[1]}_{1},\dots , f^{[1]}_{j},f_{j+1},\dots , f_{n-1},0 ) & j < n\\ ( f^{[1]}_{0},f^{[1]}_{1},\dots , f^{[1]}_{n-1},f_{n-1 } ) & j\geq n\end{cases}\ ] ] we speak of this @xmath63-vector as the one generated by @xmath27-coning the initial @xmath63-vector .
consider the family of @xmath63-vectors derived by starting with an initial @xmath63-vector and at each stage , both @xmath27-coning and @xmath130-coning it .
the first iteration of such a tree is shown below , assuming @xmath131 : @xmath132_{j\text{-cone}}\ar[dr]^{\infty\text{-cone } } & \\ ( f_{0}^{[1]},\dots , f_{j}^{[1]},f_{j+1},\dots , f_{n-1})\ar@{.}[d ] & & ( f_{0}^{[1]},\dots , f_{j}^{[1]},f_{j}^{[0]}+f_{j+1}^{(1)},\dots , f_{n-1}^{(1)})\ar@{.}[d]\\ & & } } \ ] ] we will use this technique to generate families of @xmath63-vectors which grow exponentially with dimension . given a simplicial complex @xmath53 and a @xmath133-tuple @xmath134 where each @xmath135 is either a nonnegative integer or @xmath130 , define @xmath136 to be the simplicial complex generated by @xmath137-coning , then @xmath138-coning , and so on . similarly ,
if @xmath6 is the associated ideal of @xmath53 , define @xmath139 to be the ideal associated to @xmath136 .
[ lem : distincttree]given an @xmath63-vector @xmath140 of a simplicial complex @xmath53 where @xmath141 is the last nonzero element of the @xmath63-vector and @xmath142 , the @xmath63-vectors of the simplicial complexes in the tree generated by repeatedly @xmath27-coning and @xmath130-coning @xmath53 are distinct .
we can index each @xmath63-vector in the tree with the @xmath133-tuple @xmath134 where each @xmath135 is either @xmath27 or @xmath130 and @xmath136 is the corresponding simplicial complex .
let @xmath143 , @xmath144 be the @xmath145 such that @xmath146 so that @xmath147 is the number of times we @xmath130-cone
. then the @xmath63-vector of @xmath136 is : @xmath148},f_{1}^{[k]},\dots , f_{j}^{[k]},{\displaystyle \sum_{i=0}^{r-1}}f_{j}^{[t_{i}]}+f_{j+1}^{(r)},{\displaystyle \sum_{i=0}^{r-2}}f_{j}^{[t_{i}]}+f_{j+2}^{(r)},\ ] ] @xmath149}+f_{j+r}^{(r)},f_{j+r+1}^{(r)},\dots , f_{c+r}^{(r)},\dots)\text{.}\ ] ] from this @xmath63-vector we can read @xmath147 from the index of the last nonzero entry minus @xmath150 . further , as the @xmath151 are monotonically increasing and thus distinct , the elements at position @xmath114 through @xmath152 determine the @xmath143 , and as the @xmath37 element gives @xmath133 , the @xmath63-vector uniquely determines @xmath134 .
this lemma ensures that @xmath27-coning and @xmath130-coning a suitable @xmath63-vector gives a family of hilbert functions which grows as @xmath153 with dimension .
it can be shown that this growth is not necessarily achieved if we @xmath27-cone and @xmath71-cone for arbitrary @xmath27 and @xmath71 .
[ theorem : firstfamily ] given a set of squarefree monomial ideals @xmath154 corresponding to an @xmath63-vector @xmath155 , if for some @xmath27 their betti diagrams have @xmath37 diagonal such that the alternating sum @xmath156 is nonzero and , for each @xmath26 , @xmath157 , then each @xmath63-vector in the tree created by @xmath158-coning and @xmath130-coning @xmath155 has an associated hilbert function such that the poset tree of graded betti numbers of all ideals attaining this hilbert function will fail to have a unique minimum .
the poset tree of each hilbert function also fails to have a unique minimum if we restrict to the graded betti numbers of all squarefree monomial ideals .
we show that this condition on the @xmath37 diagonal guarantees that a unique minimum can not exist among the graded betti numbers . as this diagonal
is preserved under both @xmath158-coning and @xmath130-coning and the @xmath159 are squarefree monomial ideals , the theorem follows . as @xmath160 for all i ,
any diagram that is less than all @xmath161 in the partial order has all zeros on its @xmath37 diagonal .
however , as @xmath156 is nonzero , no betti diagram can have all zeros along the @xmath37 diagonal and attain @xmath156 .
thus , no ideal can be less than all @xmath161 and attain this hilbert function , and as the conditions on the @xmath162 imply that at least two of the @xmath162 are incomparable , a unique minimum can not exist .
the number of ( not necessarily distinct ) @xmath63-vectors in the tree constructed in theorem [ theorem : firstfamily ] grows as @xmath163 where @xmath150 is the number of times the initial @xmath63-vector has been coned .
if the initial @xmath63-vector satisfies the hypothesis of lemma [ lem : distincttree ] , then we are guaranteed that these @xmath63-vectors will be distinct , and so this family grows exponentially with the dimension of the polynomial ring .
we can enlarge this family by @xmath164-coning as well as @xmath130-coning where @xmath150 is the number of times we have coned already , however , our computational evidence suggests that this family still grows exponentially .
[ e : ideals ] from an exhaustive computational search , several examples were found that satisfy the hypothesis of the above theorem .
the programs written to find these examples are available from the authors on request .
the first example occurring in lex order was with the @xmath63-vector @xmath165 , and had previously been noted by gelvin , lavictore , reed , richert @xcite .
two ideals attaining this @xmath63-vector are : @xmath166 @xmath167 with graded betti numbers ( calculated over @xmath168 with macaulay 2 ) : @xmath169 [ cols="^,^,^,^,^,^,^",options="header " , ] these betti numbers satisfying the conditions laid out in lemma [ l : betti ] ( with the roles of @xmath6 and @xmath23 reversed ) .
thus , the family generated by @xmath170-coning and @xmath130-coning @xmath165 consists of @xmath63-vectors corresponding to hilbert functions whose poset trees of betti numbers fail to have unique minimums .
this family grows exponentially with dimension by lemma [ lem : distincttree ] .
the infinite family given above is also an infinite family of hilbert functions with non - unique minimal _ graded _ betti numbers , as incomparable betti numbers implies incomparable graded betti numbers .
one of the simplest ways to construct hilbert functions for squarefree monomial ideals with unique minimal graded betti numbers is to find those for which the graded betti numbers of the squarefree lex ideal are minimal .
as the squarefree lex ideal always gives maximal graded betti numbers ( among squarefree monomial ideals ) , uniqueness follows immediately . to proceed , we observe that a minimal free resolution of a squarefree lex ideal generated in a single degree @xmath12 is @xmath12-linear ( this was proved by aramova , hibi , and herzog @xcite ) .
in fact , it is true ( see herzog , reiner , and welker in @xcite ) that squarefree lex ideals are componentwise linear ( an ideal @xmath6 is componentwise linear if @xmath171 is @xmath12-linear for all @xmath12 ) . in particular , this implies that if @xmath45 is a squarefree lex ideal with no minimal generators in degree @xmath145 , then @xmath172 for @xmath173or in words , the betti diagram of a squarefree lex ideal @xmath45 may contain a nonzero entry in row @xmath26 only if @xmath174 or @xmath45 has a minimal generator in degree @xmath175 .
we now give a family , growing exponentially with dimension , of @xmath63-vectors for which the corresponding poset tree of graded betti numbers ( for squarefree monomial ideals ) has a unique ( and hence a unique minimal ) element .
suppose that @xmath45 is a squarefree lex ideal generated in a single degree .
then the poset tree of graded betti numbers of squarefree monomial ideals with the same hilbert function as @xmath176 consists of a unique element ( which is thus uniquely minimal ) .
the family of @xmath63-vectors which give rise to such @xmath45 grows exponentially with dimension .
we know that a squarefree monomial lex ideal , @xmath45 , generated in a single degree has a @xmath12-linear resolution .
thus , by lemma [ lem : diagonallemma ] , no other ideal with the same @xmath63-vector can have smaller betti numbers than @xmath45 and so it is the unique minimum . for a polynomial ring with @xmath7 variables ,
there are @xmath177 squarefree monomials of degree @xmath26 and thus @xmath178 distinct squarefree monomial lex ideals generated in a single degree .
this gives the exponential growth with dimension .
[ lem : tworow ] suppose that @xmath45 is a squarefree lex ideal with @xmath179 iff @xmath180 or @xmath181 .
then if an ideal exists that attains the same @xmath63-vector as @xmath45 , and satisfies the hypothesis of lemma [ lem : diagonallemma ] , it will have the unique minimal graded betti numbers among all squarefree monomial ideals with this @xmath63-vector .
the conditions on @xmath45 imply that any ideal with the same @xmath63-vector will have at most two nonzero betti numbers in each of its diagonal alternating sums , and these will have opposite signs in the summation .
thus , if an ideal @xmath6 satisfies the hypothesis of lemma [ lem : diagonallemma ] , it will have at most one nonzero graded betti number on each diagonal and so no other ideal can have a smaller graded betti number than @xmath6 while preserving the @xmath156 . [
theorem : hochstermin ] in the polynomial ring of dimension n , the hilbert function corresponding to the @xmath63-vector @xmath182 for @xmath183 has an ideal which attains minimal graded betti numbers among all squarefree monomial ideals with the same hilbert function . if @xmath184 we have that the given simplicial complex is the only one with the corresponding @xmath63-vector , and hence is minimal . for @xmath185
, we note that the squarefree lex ideal for this @xmath63-vector can only be generated in degree 2 and 3 , as a generator in degree 4 would correspond to a minimal non-3-face , which would certainly imply that this simplicial complex has 2-faces ; but the above @xmath63-vector has none . by lemma [ lem : tworow ] , it remains to show that there exist ideals attaining @xmath182 for each @xmath186 which satisfy the hypothesis of lemma [ lem : diagonallemma ] .
we do this in the following two results . by hochster
s formula , a resolution will be 2-linear if there is no reduced @xmath71-homology for @xmath190 , for any sub - complex @xmath88 ( because in that case @xmath191 for @xmath192 and @xmath39 , which for @xmath193 depends only on the spaces @xmath194 , are all zero ) . but note that for any @xmath87 the simplicial complex @xmath88 is homotopy equivalent to a finite set of points . since a finite set of points never has reduced @xmath71-homology for @xmath190 , the lemma is proved .
the case @xmath39 for @xmath131 is done above as if @xmath201 ( strict containment ) then @xmath88 is a simplicial complex in the form of the above lemma .
if @xmath202 , then @xmath203 depends only the space @xmath204 .
however , this simplicial complex is homotopy equivalent to a circle , so its only nonzero reduced homology space is @xmath205 , which occurs in hochster s formula when @xmath206 .
h. charalambous and e. g. evans , jr . , resolutions with a given hilbert function , in _ commutative algebra : syzygies , multiplicities , and birational algebra ( south hadley , ma , 1992 ) _ , in _ contemp .
_ * 159 * , pp .
1926 ( amer .
soc . , providence , ri , 1994 ) .
m. hochster , cohen - macaulay rings , combinatorics , and simplicial complexes , in _ ring theory , ii ( proc .
second conf .
oklahoma , norman , okla . , 1975 ) _ , 171223 , in _ lecture notes in pure and appl .
_ , * 26 * , dekker , new york , 1977 . | we give conditions for determining the extremal behavior for the ( graded ) betti numbers of squarefree monomial ideals . for the case of non - unique minima ,
we give several conditions which we use to produce infinite families , exponentially growing with dimension , of hilbert functions which have no smallest ( graded ) betti numbers among squarefree monomial ideals and all ideals . for the case of unique minima , we give two families of hilbert functions , one with exponential and one with linear growth as dimension grows , that have unique minimal betti numbers among squarefree monomial ideals . corresponding author : [email protected] ] |
it is generally recognized that there is one - to - one correspondence between the presence of nontrivial topological invariants characterizing the bulk electron states of a crystal and the appearance of specific electron modes localized at the crystal boundary @xcite .
this statement , which is known as the bulk - boundary correspondence theorem , reflects profound interrelation between the interior and exterior electron states of the truncated crystal .
while the formulated assertion is based on general arguments of the topological concept for solids , in real materials and/or heterostructures , the effect of the peculiar bulk properties on the surface may be quite intricate .
this problem has been widely discussed in the context of an existence of topological bound states on the surface of a three - dimensional topological insulator ( 3d ti ) or at the interface between 3d ti and topologically trivial material in various hybrid structures@xcite .
angle - resolved photoemission spectra have given evidence for the dirac - like dispersion and the momentum - dependent spin texture of the 3d ti surface states in a family of bi- and sb - based narrow - gap semiconductors with strong spin - orbit coupling ( soc ) @xcite , which leads to the inverted band structure characterized by nontrivial topological invariants .
although the topologically protected surface states are often considered as the most important and even decisive property of 3d tis , in practice , the specific manifestations of the boundary - related electron properties in semiconductor materials with an inverted energy gap go far beyond the bulk - boundary correspondence paradigm .
in other words , being formally correct , the topological arguments tell us too little about characteristics of the topologically protected surface / interface states ( e.g. , the details of the dispersion , actual length scale , spin texture ) or what other in - gap electron states might form at the real boundary of a 3d ti material .
while a rapid progress has been made in investigations of the vacuum - terminated ti surfaces @xcite , it is still challenging to control the properties of the dirac states under the surface modification in complex situations when a 3d ti is brought into the contact with some substances .
recent experiments on 3d tis have demonstrated that these states are notably sensitive to external perturbations , such as chemical doping of the surface via deposition of both magnetic and nonmagnetic elements @xcite , oxidation of air - exposed samples @xcite , change of the surface termination @xcite , capping layers and interfaces with other materials @xcite , applying an external gate voltage @xcite , etc .
perturbation of a bulk crystal potential exists naturally in truncated crystals and , in particular , 3d tis , and creates a surface potential ( sp ) affecting electron properties of a crystal near / at the surface . in refs .
@xcite it was argued that a bulk - truncated surface of bismuth - chalcogenides can develop complex electronic structure in which the dirac states coexist with the conventional states of the two - dimensional electron gas ( 2deg ) in the quantum well appearing near the 3d ti surface due to the band - bending effect .
the _ ab initio _
calculations @xcite have shown that an expansion of the van - der - waals ( vdw ) spacing in layered 3d tis caused by intercalation of deposited atoms leads to a simultaneous emergence of 2deg bands localized in the subsurface region .
moreover , the expansion of the vdw spacing also leads to a relocation of the dirac topological states to the lower quintuple layers @xcite .
wang _ et al _ have studied the effects of surface modification on the topological surface state in bi@xmath0se@xmath1 using first - principles calculations and shown that bi - capping and se - removing can move the dirac point upwards and slow flatten the topological surface bands @xcite .
the short - range chemical forces related to dangling - bonds on the surface of a thallium - based ternary chalcogenides tis ( tlbite@xmath0 , tlbise@xmath0 , tlsbte@xmath0 , and tlsbse@xmath0 ) produce strong surface states @xcite which can be removed by thallium adatoms @xcite . in ref .
@xcite it was suggested that the dirac point of the helical surface states can be significantly shifted by applying uniaxial strain .
many exciting physical properties of the dirac helical quasiparticles ( in particular , spin - dependent transport ) are predicted to provide good opportunities for different spintronic applications @xcite . to fully embody these promising ideas in devices ,
one requires multiple interfaces with the topologically trivial materials rather than a single pristine surface of ti . using density functional theory to design superlattice structures based on bi@xmath0se@xmath1
, it was shown that an interface state with an ideal dirac cone is caused by alternating the layers of 3d ti and 3d normal insulator ( ni ) @xcite .
the authors of ref .
@xcite have studied theoretically the sb@xmath0se@xmath1/bi@xmath0se@xmath1 heterostructures , the constituents of which possess the bloch functions of the same symmetry .
they found that the probability maximum of the dirac state largely moves from the topologically nontrivial bi@xmath0se@xmath1 into the region of the topologically trivial sb@xmath0se@xmath1 . on the other hand ,
arpes experiments @xcite provide the direct evidence that the surface state of top surface of the heterostructure containing single quintuple layer ( ql ) of bi@xmath0se@xmath1 on 19qls of bi@xmath0te@xmath1 is similar to the surface state of bi@xmath0se@xmath1 .
moreover , the transport measurements @xcite show that the studied heterostructure behaves more like bi@xmath0se@xmath1 even though there is only 1ql bi@xmath0se@xmath1 layer grown on 19qls bi@xmath0te@xmath1 . in refs .
@xcite it was established that , as a result of depositing a 3d ni overlayer ( conventional semiconductor znm , m = s , se , and te ) onto the 3d ti substrate ( bi@xmath0se@xmath1 or bi@xmath0te@xmath1 ) , the topological states can float to the top of the ni film , or stay put at the ni / ti interface , or are pushed down deeper into 3d ti .
recently berntsen and colleagues have directly observed the dirac states at the bi@xmath0se@xmath1/si(111 ) buried interface @xcite .
another photoemission study in ref .
@xcite revealed an existence of the interface topological states in the layered bulk crystal ( pbse)@xmath2(bi@xmath0se@xmath1)@xmath3 , which forms a natural multilayer heterostructure composed of ti and ni . the evidence of a large shift of the dirac point towards the conduction band edge relative to the case of the 3d ti / vacuum interface , due to the in@xmath0se@xmath1 @xcite or sb@xmath0se@xmath0te @xcite capping layer on the epitaxial bi@xmath0se@xmath1 thin film , was reported demonstrating a possibility of controlling the dirac cone in 3d ti - based systems .
thus , the experimental and theoretical data exhibit that the real 3d ti surface and 3d ti / ni interface possess very rich and diverse physics , in particular , they can hold both topological and non - topological ( ordinary ) in - gap states .
it is well known that the non - topological bound states can be created or deleted or altered , depending on the both the sign and strength of sp , when 3d ni is exposed to ambient conditions or put into the contact with other materials . on the contrary , in 3d ti ,
the topological order itself is robust against such the influences so that it can be completely destroyed only under the drastic perturbation @xcite .
nevertheless , the parameters of the topological states can undergo remarkable changes with even moderate external perturbations .
combination of the robustness of the topological states at the ti / ni interfaces with the tunability of their parameters to the external influence favours the design of the 3d ti / ni layered systems possessing suitable band structure , charge distribution and spin texture .
the efficient design of the 3d ti / ni systems can be realized by combining analytic and numerical methods . in the present work
, we consider a special type of the 3d ti / ni heterostructures of particular interest , which contain an ultrathin film of nonmagnetic ni ( overlayer ) artificially deposited on a relatively thick film of 3d ti ( substrate ) . due to a specific relation between electron affinities and band gap widths of the substrate and overlayer materials , significant modifications of the spectrum and wave function of the dirac states
are expected as compared to the pristine 3d ti surface ( i.e. the 3d ti / vacuum interface ) , thus making it possible to obtain the 3d ti - based heterostructure with tailor - made electron properties .
one assumes that the substrate film thickness is large enough to avoid sizable hybridization of the bound states appearing at the opposite boundaries of the film . at the same time , the minimal thickness of the overlayer is formally limited by the condition of an electron motion quantization in the 3d ni material . under these restrictions , to describe analytically the electron bound states near the surface of the truncated 3d ti covered by the 3d ni overlayer , the continual approach involving the method of effective surface potential ( sp ) was offered in ref .
@xcite . in
what follows , we will use the term near - surface state ( see also ref .
@xcite ) for identification of the in - gap electron state localized in the subsurface region and driven by the overlayer - induced sp . just recently , in ref .
@xcite , several preliminary results concerning the near - surface states in the 3d ti / ni heterostructures with realistic material parameters has been obtained within the numerical simulations based on density functional theory ( dft ) .
below , we employ the two complementary approaches analytical and numerical in order to elucidate thoroughly the important question how a 3d ni overlayer affects the electron properties of 3d ti / ni heterostructures . in the framework of the analytical approach
, it is instructive to re - formulate this question in terms of the boundary conditions at the ti / ni interface for the wave function of the system .
it is clear that , in systems composed of two topologically distinguishable materials , the characteristics of the near - surface state depend crucially on the choice of the boundary conditions , which still remains highly disputable subject ( for example , see refs .
the wave function at the ideal atomic interface between a pair of similar materials ( e.g. , 3d ti bi@xmath0se@xmath1 and 3d ni sb@xmath0te@xmath1 have the same crystal symmetry ) satisfies the ben - daniel&duke boundary conditions @xcite .
while matching the wave function at the contact of two dissimilar materials ( e.g. , such as si and bi@xmath0se@xmath1 ) is complicated within the @xmath4 formalism because the envelope function ( ef ) on each side of the interface are defined using distinct orbital basis ( see ref .
@xcite and reference therein ) .
however , such complication proved to be circumvented for the particular models describing different types of contacts within the effective interface potential concept @xcite .
below , in the framework of the continual approach involving the sp scheme , we formulate general boundary conditions for the long - range envelope function of the truncated 3d ti and truncated 3d ni and find the solution for the bound near - surface states .
we restrict ourselves to the situation when solely the orbital degree of freedom of electrons is manipulated by external influence at the surface .
we succeed in general qualitative understanding of the dependence of the energy spectrum and spatial profile of the near - surface states on an effective sp .
furthermore , in order to elucidate the fine details of the near - surface state transformation induced by the overlayer , we employ the material - specific dft calculations for the 3d ti / ni heterostructures . for the conceptual reasons , within an effective sp scheme
, we also discuss the near - surface states in the fictitious heterostructures of other types , in which a substrate of a 3d ni close to the quantum transition into a 3d ti phase is covered with an overlayer of either ni or ti material .
the paper is organized as follows . in sec . 2 , we discuss the effective sp concept , propose the model for a truncated ti covered with an overlayer , and introduce the main ingredients and assumptions of the problem within the continual approach . in sec . 3 , for the case of a spin - independent surface perturbation caused by an overlayer
, we thoroughly investigate how the corresponding sp modifies the electron energy spectrum and the ef spatial profile of the dirac - like near - surface state . in sec . 4
, we analyze main features of the near - surface state in a situation when a 3d ni substrate close to transition into a topological phase is covered with a ultrathin overlayer of either a ni material or a ti one . to corroborate the sp formalism results , in in sec . 5
, we apply the dft calculations and analyze the band structure and wave - function of the bound near - surface states for a set of heterostructures formed by a single tetradymite - type quintuple ( ql ) or septuple ( sl ) layer blocks and 3d ti substrates .
finally , the main conclusions are presented in sec .
apart from the aforementioned simulations of the properties of 3d tis based on the first principle calculations , various continual models have been discussed to describe relativistic fermions at the ti boundary @xcite .
there are theoretical studies of the ti properties , which are routinely based on the simple phenomenological 2d hamiltonian for helical fermions with the linear dirac - cone - like energy - momentum dispersion under an external influence @xcite : @xmath5)+\mathbb{u}$ ] , where @xmath6 is the unit vector normal to the surface , @xmath7 is the fermi velocity , @xmath8 is the vector composed of the pauli matrices .
it is generally thought that a controllable external field , @xmath9 , can be directly applied to the 3d ti surface to manage its electron states .
for instance , a spin - independent term @xmath10 could simulate the energy shift of the dirac - cone point due to an electrostatic potential caused by a nonmagnetic overlayer . in turn , a spin - dependent term @xmath11 could generate a gapped spin - polarized surface state through an exchange field proportional to the magnetization @xmath12 applied along the normal to the surface of 3d ti which is in contact with a ferromagnetic insulator @xcite . in this manner , to take into account a perturbation arising from the external influence , the additional term @xmath9 is simply included in the 2d hamiltonian @xmath13 , without a serious analysis of the microscopic origin of both @xmath13 and @xmath9 .
the vast majority of theoretical works restricts to such the description and they predict many curious effects which can be realized in the 3d ti - based structures .
however , the 2d hamiltonian @xmath13 can formally be derived from a relevant 3d hamiltonian only under the free surface stipulation in the spirit of refs . @xcite . in the framework of the consistent scheme ,
the bound near - surface states for the half - infinite 3d ti are composed of the eigen - states of the corresponding bulk 3d ti hamiltonian .
meanwhile , note that so far nobody has written down the full set of the orthogonal wave - functions ( including the bound and extended along @xmath6-direction states ) for the @xmath4 hamiltonian of 3d ti in the half - infinite geometry even under the free boundary conditions on the surface . strictly speaking
, the bound states alone do not form the full basis set suited for correct description of the @xmath9 field effect on 3d ti .
the external surface perturbation excites electron density in the bulk 3d ti region of a nanoscopic scale adjacent to the surface . both the bound and extended electron modes give rise to the response of 3d ti to this perturbation . upon placing the 3d ni on 3d ti , besides the topological bound state , the so - called ordinary bound state @xcite can arise near the interface due to the hybridization between the ni and ti atomic orbitals through the interface .
hence the near surface electron density perturbation of the 3d ti has very complicated spatial , orbital and spin configuration .
the phenomenological 2d hamiltonian @xmath13 hardly could serve as a starting point for the correct analysis of the configuration - dependent response of the 3d ti .
so , to correctly take into account the effect of an external perturbation on the surface / interface electron states in the 3d ti , one has to directly include the field @xmath9 into the `` true '' 3d hamiltonian of the system .
a basic idea to go beyond the scope of the 2d model is the use of the well - known @xmath4 method @xcite . to characterize the band electron states of a bulk semiconductor ,
@xmath14 ( @xmath15 is a wave vector , @xmath16 is a band index ) , in the region of the brillouin zone around the point of band extrema @xmath17 , the @xmath18 method is reputed to be accurate enough . under a perturbation smooth on the atomic scale
, this method makes it possible to predict evolution of the electron state wave function @xmath19 in terms of a product of a slowly varying envelope function ( ef ) @xmath20 and the bloch function of the unperturbed crystal @xmath21 at the point @xmath17 : @xmath22 , @xmath23 is the lattice periodic function .
the ef concept may also be applied to the description of localized and resonant interface states in the semiconductor junctions of different types .
however , a relevant choice of the boundary conditions for the function @xmath20 remains an unsettled question in this concept , in particular for the ti based structures .
the authors of refs .
@xcite impose the so - called open boundary conditions fixing all ef components to zero at the crystal surface .
this restriction formally simulates the effect of vanishing of the quasiparticle wave function on the infinitely high sp barrier .
nevertheless , it should be pointed out that the zero constraint is not unique , and other options for boundary conditions have been advocated in the literature .
for instance , in ref .
@xcite the problem is formulated in terms of an energy functional whose minimization yields the so - called natural boundary conditions , intermixing the magnitudes and derivatives of different ef components .
both mentioned types of the ef boundary conditions are extremely idealized and can not adequately take into consideration a sensitivity of the 3d ti electron states to the surface modifications . in this work
we propose a formalism to directly incorporate the surface perturbation effect into the 3d ti hamiltonian .
we derive the appropriate ef boundary conditions through the construction of the effective semi - phenomenological sp localized at the 3d ti surface .
the orbital and spin structure of the sp mimics induced fields resulting from a surface perturbation . as shown below ,
the structure and strength of the sp determine both the spatial and spectral features of the topological states .
the low energy and long wavelength bulk electron states of the prototypical ti , narrow - gap semiconductor of bi@xmath0se@xmath1-type , are described by the four bands @xmath24 hamiltonian with strong soc proposed in refs .
@xcite . without a loss of generality
, we make use of the simple version of this hamiltonian in the form : @xmath25 where @xmath26 , @xmath15 is the wave vector , @xmath27 , @xmath28 and @xmath29 ( @xmath30 ) denote the pauli matrices in the spin and orbital space , respectively .
the hamiltonian is written in the basis @xmath31 of the four states at the @xmath32 point of the brillouin zone with @xmath33 .
the superscripts @xmath34 denote the even and odd parity states and the arrows @xmath35 indicate the spin projection onto the @xmath36 quantization axis .
the hamiltonian ( [ h - ti ] ) captures the remarkable feature of the band structure : under the condition @xmath37 , the inverted order of the energy terms @xmath38 and @xmath39 around @xmath33 , which correctly characterizes the topological nature of the system due to strong soc .
the hamiltonian ( [ h - ti ] ) is particle - hole symmetric and isotropic , which helps us to simplify calculations .
we consider a semi - infinite 3d ti material , such as bi@xmath0se@xmath1 , occupying the region @xmath40 . the material boundary located at @xmath41 is perfectly flat and displays translational symmetry in the @xmath42 plane .
the potential at the surface of a real 3d ti material is different from the bulk crystal potential , irrespective of whether the surface is kept in ultra - high vacuum or , for example , coated with an overlayer or interfaced with another material . to demonstrate the effect of the surface modification on the topological states within a conceptually simple scheme
, we introduce the interaction of electrons with an external perturbation confined at the surface , implementing the effective sp @xmath43 into the ef calculation .
thus we write the full electron energy of the truncated 3d ti in the following form : @xmath44\theta(\mathbf{r}),\ ] ] here the operator @xmath45 determined in eq .
( [ h - ti ] ) acts in the the spinor function space @xmath46 , represented in the basis @xmath47 , the superscript @xmath48 denotes the transpose operation .
the ef components @xmath49 ( the subscript @xmath50 numbers the spinor components ) are presumed to be smooth and continuous functions in the half - space @xmath40 , while the spatial symmetry and periodicity of the system are broken due to existence of the ti surface .
it is evident that the @xmath4 approach can not provide a correct description of the wave - function behavior near the surface , where large momenta are highly important . to overcome this drawback we introduce the effective sp @xmath43 , which affects the electron states of ti at the surface
the potential @xmath43 is nonzero in a small region @xmath51 ( of the order of a lattice parameter ) around the geometrical boundary @xmath41 , where the validity of the @xmath4 scheme is questionable .
an introduction of the phenomenological sp in eq .
( [ omega ] ) enables us to correctly match the low - energy and long - range electronic states inside the truncated ti with evanescent vacuum states through the boundary conditions for ef @xmath52 .
as long as the ef spatial variation of the sought state , @xmath53 [ @xmath54 , @xmath55 , is sufficiently slow in the direction normal to the surface , one can adopt a local approximation for the sp .
namely one writes @xmath56 , where the symbol @xmath57 at the argument of the delta - function signifies that the sheet - like sp is placed inside the ti half - space but at infinitesimally small distance from the boundary @xmath41 . as a matter of course
, an electron wave function has to be continuous at a crystal boundary .
nevertheless , in the system under consideration , since the bloch factors of the wave function inside and outside ti do not coincide ( in particular , they have distinct space symmetries ) , the long - range ef @xmath58 can formally undergo a finite break ( jump ) across the boundary from @xmath59 to @xmath60 within the utilized @xmath24 method ( we refer the reader to the detailed discussion in ref .
@xcite ) . in the current work
, we do not care how the wave - function behaves in the half - space @xmath61 but next we make use of a functional @xmath62\theta(\bm{\kappa},z ) , \ ] ] where the energy @xmath63 plays a role of the lagrange multiplier , @xmath64 is an unit @xmath65 matrix , @xmath66 . the functional ( [ f ] )
is determined in the class of the smooth and continuous efs in the ti half - space @xmath40 and includes the effective surface potential @xmath67 . since , in a plane geometry , the wave - vector @xmath68 is a good quantum number , we determine the functional for each ef @xmath68-mode , @xmath58 .
varying functional @xmath69 with respect to @xmath70 yields the euler equations for the half - space @xmath40 and the boundary conditions at the surface at @xmath60 .
the corresponding equations in the compact form are : @xmath71\theta(\bm{\kappa},z)=0,\ ] ] @xmath72 in the left side of eq .
( [ current ] ) the current density operator acts on the ef spinor .
thus the right side associated with the surface perturbation plays a role of the external ( with regard to the ti bulk ) current source ( sink ) .
the equation ( [ current ] ) involves the surface potential parameters , in this sense it has something in common with the equation which was used to calculate the surface states of a crystal with a relativistic band structure in ref .
the solution of the boundary task , eqs .
( [ equation ] ) and ( [ current ] ) , answers the principal physical question how the perturbation located just at the ti boundary affects the near - surface topological states . in the half - space @xmath40 , the general solution of eq .
( [ equation ] ) for each ef spinor component obeying the condition @xmath73 can be represented as @xmath74 + \beta_{j}(\kappa , e)\exp[-q_{2}(\kappa , e)z]\},\ ] ] where @xmath75 @xmath76 here the phase factors @xmath77 forming the spinor @xmath78 depend only on the momentum polar angle , @xmath79 , @xmath80 ; the signs @xmath34 relates to lower and upper spectral branches , respectively .
the characteristic momenta @xmath81 are the solutions of the corresponding secular equation ; @xmath82 .
the boundary conditions , eq .
( [ current ] ) , determine the coefficients @xmath83 and @xmath84 as well as the dispersion relation for the near - surface states inside the bulk band gap , @xmath85 .
the parameter @xmath86 is implied to be @xmath87 .
the potential @xmath9 in eq . ( [ current ] ) is a @xmath65 matrix specifying internal properties of the ti surface and the matrix elements include different components of scattering of the ti states on sp . in principle , choosing the structure of the matrix and the strength of its components allows us to tune spatial and energy characteristics of the topological states . for example , as for the sp diagonal matrix elements , @xmath88 , the values @xmath89 and @xmath90 are proportional to the scattering intensity of particle and hole , respectively , on the spin - independent part of sp , while the quantities @xmath91 and @xmath92 are proportional to the scattering intensity of particle and hole , respectively , on the @xmath36-component of the exchange part of sp .
the off - diagonal matrix elements @xmath93 with @xmath94 result from the spin - orbit interaction at the surface , which , in general different from the soc in the ti bulk . in this work we focus on the sp that preserves time reversal symmetry , i.e. @xmath95 , where @xmath96 , @xmath97 .
such the sp structure in the basis @xmath47 results in the following relations between the ef coefficients in eq .
( [ ef ] ) : @xmath98 , @xmath99 , @xmath100 , @xmath101 .
moreover , we neglect the dependence of @xmath102 on @xmath68 in eq .
( [ current ] ) .
one can interpret the spin - independent scattering on the surface within the framework of a local band bending scheme ( which is quite reasonable for the contact of two insulators / semiconductors ) , where the band edge corresponding to the @xmath50-th spinor component is affected by the external perturbation confined at the surface : @xmath103 .
in the situation of ti covered with an overlayer the intuitive idea is that the diagonal components of sp could be heuristically adjusted to the relative offsets between the corresponding energy levels ( bands ) of the ti substrate and the overlayer . in other words , the energy @xmath104 mimics the local bending of the respective bands .
after some algebra the corresponding secular equation results in the implicit relation between the energy @xmath63 and the in - plane momentum @xmath105 for the bound state at the ti surface : @xmath106[q_{2}\pm\kappa ] -[\mathrm{b}q_{1}^{2}+\xi(\kappa)-e][\mathrm{b}q_{2}^{2}+\xi(\kappa)-e]\nonumber\\&+ & 2\biggl\{q_{1}q_{2}-\frac{\xi(\kappa)-e}{\mathrm{b}}\pm \kappa [ q_{1}+q_{2}]\biggr\ } \times \biggl\{\mathrm{b}^{2}q_{1}q_{2}-du_{1}du_{2}-\frac{\mathrm{a}^{2}}{4}\biggr\}\\&- & 2[\xi(\kappa)-e][du_{2}(q_{1}+q_{2})\pm\kappa(du_{1}+du_{2})]\nonumber\\&- & 2\mathrm{b}\{du_{1}q_{1}q_{2}(q_{1}+q_{2})\pm \kappa du_{1}(q_{1}+q_{2})^{2}\mp \kappa(du_{1}+du_{2})q_{1}q_{2}\}=0\nonumber,\end{aligned}\ ] ] where @xmath107 in accordance with eqs .
( [ q ] ) and ( [ q1,2 ] ) , @xmath108 . note that eq .
( [ evskappa ] ) is invariant under the simultaneous permutations : @xmath109 , @xmath110 and @xmath111 . at the @xmath32 point , eq .
( [ evskappa ] ) is reduced to the equation that determines the dirac ( node ) point position , @xmath112 , as a function of the sp strength @xmath113 : @xmath114 \biggl[\lambda+\sqrt{1-\frac{e^{2}}{\xi^{2}}}-\frac{du_{1}du_{2}}{\mathrm{b}\xi}\biggr]\nonumber\\&- & \sqrt{\frac{2}{\mathrm{b}\xi}}\biggl[du_{1}\sqrt{1+\frac{e}{\xi}}+du_{2}\sqrt{1-\frac{e}{\xi}}\biggr ] \times \sqrt{2\lambda-1+\sqrt{1-\frac{e^{2}}{\xi^{2}}}}=0.\end{aligned}\ ] ] the shaded areas in fig .
[ fig1]a denote the realm of the near - surface bound state with @xmath115 on the @xmath116-plane .
the dependence of the node point position on the sp strength , obtained from eq .
( [ evsu ] ) for several ratio values @xmath117 , is plotted in fig .
[ fig1]b ( for @xmath118 ) and fig .
[ fig1]c ( for @xmath119 ) .
one can see three different regions in these plots . at weak potential @xmath120 , the dirac point linearly shifts with respect to the ti bulk bands to either higher or lower binding energies depending on the sp strength sum , @xmath121 in case
the sp strength is large , @xmath122 , the node point energy approaches zero as @xmath123 on the @xmath116-plane , there are regions ( unshaded areas in fig .
[ fig1]a ) where the bound state is absent since the node point merges into the conduction or valence bulk band .
for example , if @xmath124 , the threshold values of the potential , at which the node point splits off the bulk band continuum , are @xmath125 $ ] , so that @xmath126 and @xmath127 . if the energy @xmath128 is a small deviation from the dirac linear spectrum , @xmath129 , @xmath130 , the characteristic momenta , eq .
( [ q1,2 ] ) , are found as @xmath131},\ ] ] where @xmath132 is given by eq .
( [ q0 - 12 ] ) .
the deviation @xmath128 appears to be small not only when @xmath133 but also when sp is either weak or strong . using the expression ( [ q1,2plus ] )
one can obtain the spectrum and estimate the spatial distribution of the near - surface state in these limit situations .
so , for the extremely large potential , @xmath134 , the correction approaches zero , @xmath135 , in turn , the ef coordinate dependence is described by a difference of the exponents , @xmath136-\exp[-q_{2}^{0}(\kappa)z]$ ] , so that the maximum of the electron density , @xmath137 , does not occur on the surface , where @xmath138 , but rather near the point @xmath139 ( where @xmath140 ) that is distant from the surface .
such the ef distribution , together with the linear spectrum , was found under the free boundary conditions @xcite .
our approach allows us to capture peculiarities of the surface state in 3d ti induced by the sp .
if the sp strength is much greater than the characteristic energy , @xmath141 , within the perturbation theory , one obtains the amendment to the linear dispersion law as @xmath142 the surface state spectrum acquires a curvature and a shift of the node point , @xmath143 , which are inversely proportional to the potential , however the fermion group velocity @xmath144 near the node point does not change since the amendment @xmath128 ( [ corr - u - large ] ) does not contain a contribution linear in @xmath105 . in the lowest order in @xmath145 , the relations between the coefficients in eq .
( [ ef ] ) are given by @xmath146 thus , the electron density does not vanish on the ti surface , @xmath147 .
however , under the sp influence , the ef components can vanish near the surface at @xmath148 , namely , @xmath149 at @xmath150 when @xmath151 , and @xmath152 at @xmath153 when @xmath154 .
besides , as seen from eq .
( [ q1,2plus ] ) , the sp affects the decay length of the ef nonzero harmonics .
if the sp is formally absent , @xmath155 , one arrives at the solution obtained in ref .
@xcite from using the natural boundary conditions : the surface state shows the linear spectrum @xmath156 and the ef spatial profile in @xmath36-direction is merely a sum of the two exponents , @xmath136+\exp[-q_{2}^{0}(\kappa)z]$ ] , i.e. , the probability density of the near - surface state is peaked on the boundary @xmath41 and its tail penetrates into the ti bulk with the decay length @xmath157 . in the case of weak sp , @xmath158 , the correction to the dispersion law
is given by @xmath159 the spin - independent sp is seen to entirely shift and warp the energy - momentum dependence . note , that the corrections ( [ corr - u - large ] ) and ( [ corr - u - small ] ) are opposite in the sign .
turning on the sp leads to the different contributions of the quick and slow exponents into ef : @xmath160 .
let us consider thoroughly the specific case of the staggered alignment of the matrix elements , @xmath161 , when sp does not break the particle - hole symmetry .
one can verify in eq .
( [ evskappa ] ) that in such the case the near - surface state maintains the ideal dirac spectrum @xmath156 regardless of the size and sign of @xmath162 . while the spectrum is independent of the sp , the envelope function is strongly affected by it .
the coordinate dependence of each component of the ef spinor is given by @xmath163 where @xmath164 @xmath165 the ef of eq .
( [ ef - stagg ] ) is normalized as @xmath166 .
the spatial behavior of the ef zeroth harmonic @xmath167 is illustrated in fig .
[ fig2]a and fig .
[ fig2]b for positive and negative @xmath162 , respectively . with increasing sp strength
the ef structure evolves from the sum of the exponents at @xmath168 ( black lines ) to the difference at @xmath169 ( yellow lines ) .
so , one sees a gradual change in the profile of the near - surface state such that its gravity center moves from the surface to the ti interior .
it is obvious the behavior of the ef nonzero harmonics eq .
( [ ef - stagg ] ) is insignificantly different from what is plotted in fig .
[ fig2 ] ; unless the tail at @xmath170 is slightly longer than that at @xmath171 . to study the modification of the near - surface states under the finite strength of sp we dwell at length on the situation @xmath124 .
as seen in fig .
[ fig3 ] , at a finite strength of @xmath162 , @xmath172 , apart from the aforesaid shift of the node point , the form of the spectral dependence , @xmath173 , alters ( in comparison with the limiting cases @xmath168 or @xmath174 ) under the sp influence .
the group velocity of the surface topological excitations decreases from the quantity @xmath144 to zero when the sp strength @xmath175 either increases from zero to the threshold value @xmath176 or decreases from infinity to the threshold value @xmath177 . a noticeable deviation from linearity
can be seen for the strength @xmath172 . in the limit
@xmath178 the dispersion becomes parabolic at small @xmath105 , and in the limit @xmath179 the curve @xmath173 smoothly merges into @xmath180 .
the dependence @xmath173 acquires a curvature so that the relatively strong ( @xmath181 ) and relatively weak ( @xmath182 ) potentials provide with the curvature of opposite sing . the spatial behavior of the near - surface states for @xmath124 is shown in fig .
[ fig4 ] . when the sp is weak , @xmath183 , the probability density @xmath184 is largely peaked near the surface .
the strong sp , @xmath181 , pushes the probability density towards the ti bulk .
the ef is exponentially decaying away from the surface .
we would like to emphasize that the ef decay lengths , @xmath185 ( see eq .
( [ q1,2 ] ) ) , are strongly influenced by the sp strength .
for example , when the strength @xmath162 varies either from @xmath186 to @xmath187 or from @xmath188 to @xmath189 , the momentum @xmath190 increases from @xmath191 to @xmath192 ( i.e. the long exponent of ef ( [ ef ] ) ) becomes longer ) and the momentum @xmath193 decreases from @xmath194 to @xmath186 ( i.e. the short exponent becomes shorter ) .
next , we investigate the effect of the surface modification on the near - surface bound states when the bulk is a 3d normal ( topologically trivial ) insulator . here
we address the fundamental question of whether 3d ni responds to a localized surface perturbation in a way different from 3d ti . in order to describe 3d
ni , one uses the same relativistic hamiltonian ( [ h - ti ] ) in which , however , now there is the normal ( non - inverted ) alignment of the energy terms of different parity @xmath38 and @xmath39 around @xmath33 is implied , i.e. , @xmath195 and @xmath196 . for instance , in the case of the in@xmath0se@xmath1 crystal , which shares the same crystal structure with bi@xmath0se@xmath1 , the soc is not strong enough to provide the inversion between two @xmath197 orbitals with opposite parity at the @xmath32 point @xcite . in the limit @xmath198 , when soc is negligible small , eq . ( [ h - ti ] ) defines merely the semiconductor with simple ( nonrelativistic ) two - band spectrum .
it is evident that eqs .
( [ omega])-([current ] ) and the relevant sentences are valid regardless of the @xmath199 sign .
therefore analogously to what has been done in the previous sections ( the details are omitted ) one can obtain the characteristics of the bound electron states on the ni surface subjected to the external spin - independent influence .
the existence of these states is determined by real solutions of the corresponding secular equation within the bulk gap , @xmath200 , which is given by eq .
( [ evskappa ] ) at @xmath171 and @xmath201 .
[ fig5]a shows the existence realm of the near - surface bound state , i.e. , the area on the @xmath116-plane where @xmath202 .
the energy @xmath200 as a function of the sp strength for several ratio values @xmath117 is represented in fig .
[ fig5]b ( for @xmath203 ) and fig .
[ fig5]c ( for @xmath204 ) .
as is seen , except for the quadrant @xmath205 , one may choose the ratio @xmath117 to match the sp which induces the bound electron state on the ni surface . in what follows
, we consider thoroughly only the two particular cases : @xmath206 and @xmath207 .
when the sp matrix elements are in staggered rows , @xmath161 , the bound state exists at @xmath208 , which diminishes virtually the 3d ni bulk gap on the surface . because of the presence of such the sp , the particle - hole symmetry of the system is preserved .
the relations between the energy and momentum is given by @xmath209^{2 } \biggr\}^{1/2 } , \ ] ] where @xmath210 , @xmath211 is the projection of the bulk spectrum onto the surface . in fig .
[ fig6 ] , the spectral dependence @xmath173 is illustrated for several choices of the sp strength @xmath162 .
so , the system exhibits a non - linearly dispersing surface state which is specified by the energy gap @xmath212 .
the crossing black lines in fig .
[ fig5]c show the half gap @xmath200 as a function of @xmath162 .
one can see in fig .
[ fig5]a and fig . [ fig5]c that the near - surface state stays in the bulk band gap , @xmath213 , when the sp strength is restricted by the interval @xmath214 , where @xmath215 .
while , outside this interval , it is buried in the bulk band continuum . in the close vicinity of a band - crossing point @xmath216 , where @xmath217 , the dispersion relation is given by @xmath218 , so that a degeneracy at the band - crossing point @xmath219 is lifted due to gapping @xmath220 .
it is convenient to measure the sp strength in the dimensionless units @xmath221 , then @xmath222 , where @xmath223 . in turn , in a small energy window near the bulk band edges , where @xmath224 , the dependence ( [ ni - spectrum-1 ] ) becomes @xmath225^{2 } \biggr\}^{1/2 } .\ ] ] note , in the case @xmath226 , eq .
( [ ni - spectrum-1 ] ) reduces to @xmath227 $ ] ; in other words , in the 3d ni under finite value of soc , a finite strength of sp , @xmath228 , is required to split off the in - gap near - surface state from the 3d bulk continuum .
such the behavior of the bound state location in energy axis with increasing the sp strength can intuitively be explained in the language of the local band bending scheme proposed in the previous section . when 3d ni is brought into contact with a thin dielectric overlayer , a relative weak positive sp , @xmath229 , splits off states from both the conduction bulk band and valence one due to a local narrowing of the gap , @xmath230 . at @xmath231 ,
the local band bending is so steep that the bulk band edges of 3d ni cross over the energy levels of an adjusted overlayer with opposite parity , as it would be if the 3d ni surface was in the contact with a ti overlayer .
further increase of the strength @xmath162 above the critical value @xmath177 pushes the near - surface state into the bulk continuum .
figure [ fig7 ] visualizes the effect of the external potential with @xmath161 , on the space profile of the in - gap state of the truncated 3d ni .
it is of interest to note that , while crossing the value @xmath219 , the form of the space dependence of the ef zeroth harmonic , @xmath232 , switches over from monotonically decreasing , fig .
[ fig7]b , ( in the situation of @xmath233 , which mimics the 3d ni / ti - overlayer heterostructure ) to nonmonotonically decreasing with minimum at @xmath234 , fig .
[ fig7]a , ( in the situation of @xmath235 , which mimics the 3d ni / ni - overlayer heterostructure ) .
if @xmath231 , the ef can be approximated as : @xmath236 + sgn(u_{0}-u ) ( \sqrt{1+|\lambda|}-\sqrt{|\lambda|})\exp[-q_{2}(0,0)z],\ ] ] @xmath237 when the strength @xmath238 , i.e. , @xmath239 .
this situation is depicted with the black and red curves in fig .
[ fig7 ] .
let us now draw the attention to the case when the surface perturbation has the spinor structure @xmath124 answering to the surface electrostatic potential . fig . [ fig8 ] shows the dispersion law for several values of the sp strength @xmath162 . in the vicinity of the points @xmath240 , where @xmath241 , the relations between the energy and the in - plane momentum ( in the leading order in @xmath105 ) is given by @xmath242 where @xmath243 .
thus , it is clear , given @xmath162 belonging to the interval(s ) @xmath244 ( where @xmath245 ) , the surface state consists of a single dirac cone . within the framework of a heuristical local band bending scheme , the variation of the strength @xmath162
is linked with the relative movement of the energy levels of the 3d ni substrate and the ni overlayer .
when the sp strength value exceeds the threshold quantity , @xmath246 , the band structure of this system ( which consists of the two materials with a normal gap band alignment ) is inverted , i.e. , either the substrate conduction band is lower than the overlayer valence band or the substrate valence band is higher than the overlayer conduction band . as a result ,
if the strength is in the interval @xmath244 , the near - surface state of the ni covered by the normal overlayer can display a linear dispersion dependence of the dirac - cone form , @xmath247 , where the node point location @xmath200 and propagation velocity @xmath7 are the functions of the strength @xmath162 and band structure parameter @xmath248 .
the energy @xmath200 is inside the ni bulk gap , @xmath249 .
the quasi - topological bound state is also specified by the space distribution , which is shown in fig .
the corresponding ef decays exponentially away from the surface . when the strength @xmath175 attains the quantity @xmath250 , the near - surface state merges into the bulk continuum states , in turn the decay length @xmath251 becomes large ( the black and green curves ) . at @xmath252 , the probability density
is concentrated close to the surface , in other case , it is rather smeared .
the proposed continual approach gives transparent physical explanation for evolution of the near - surface state in both momentum and real spaces with the sp superimposed on the ti or ni boundary .
this approach describes fairly well the electron density distribution of the corresponding dirac - cone - like states on the scale exceeding the lattice spacing through ef(s ) as the superposition @xmath253 ( see eq .
( [ ef ] ) , where the exponents @xmath81 and the coefficients @xmath254 , @xmath255 , are functions of the sp components and the bulk band structure parameters .
however , within the sp scheme , we are unable to elucidate the electron density features on the scale on the order of the sp spacing @xmath256 , in particular , capture the fine effect of the the topological state relocation within near - surface layers @xcite .
below , in order to provide a closer look at the wave function of the bound near - surface state and to accurately reproduce its band structure over the whole brillouin zone we present _ ab initio _ density functional theory calculation results for some systems representing the topological insulator substrate covered by the insulator ultrathin film .
.the structural and energetic characteristics of the ni / ti overlayer / substrate pairs given in the @xmath257/@xmath258 format ( except for the lattice mismatch , @xmath259 ) , with @xmath260 being experimental lattice parameter , @xmath261 , or calculated work function , @xmath262 , with the indication of the films thicknesses , or calculated band gap , @xmath263 ( @xmath264 ) . note that band gap values , @xmath263 , are given for slab and bulk in the ni overlayer ( @xmath265 ) and ti substrate ( @xmath266 ) case , respectively . [
cols="^,^,^,^,^",options="header " , ] electronic structure calculations were carried out within the density functional theory using the projector augmented - wave method @xcite as implemented in the vasp code @xcite .
the exchange - correlation energy was treated using the generalized gradient approximation @xcite .
the hamiltonian contained the scalar relativistic corrections and the spin - orbit coupling was taken into account by the second variation method @xcite . in order to take into account the effect of dispersion interactions we use the van der waals nonlocal correlation functional within dft - d2 approach @xcite .
the thin film ni / ti heterostructures were simulated within a model of repeating slabs separated by a vacuum spacing of 10 .
the overlayers were symmetrically attached to both sides of the substrate slab to preserve the inversion symmetry .
as the substrates were chosen 3d tis with tetradymite - like layered structures bi@xmath0te@xmath0s , sb@xmath0te@xmath0se @xcite composed of quintuple layer ( ql ) blocks and gebi@xmath0te@xmath267 @xcite composed of septuple layer ( sl ) blocks .
the substrates were simulated by 6 ql ( 5 sl ) slabs . as the thin insulating overlayers we used single ql(sl ) films of bi@xmath0te@xmath0s , sb@xmath0te@xmath0s , and gebi@xmath0te@xmath267 which have gapped noninverted spectrum .
the interface systems under investigation are given in table [ t1 ] .
as can be seen in table [ t1 ] , the lattice mismatch between overlayer and ti substrate in the considered heterostructures does nt exceed 0.5 % providing very good epitaxial compatibility .
the in - plane lattice parameters of the heterostructures were fixed to the experimental ones of the substrate slab @xcite .
the interlayer distances within the overlayer and the ti block , closest to the interface , were optimized .
[ fig10]a shows spectra of the free - standing gebi@xmath0te@xmath267 overlayer and bi@xmath0te@xmath0s substrate . in the spectrum of bi@xmath0te@xmath0s the topological surface state ( tss ) with the dirac point lying in the @xmath268 valley of the valence bulk states
propagates across the bulk energy gap .
the spectrum of the free - standing overlayer has the 440 mev gap .
the energies of the overlayer states are matched to substrate spectrum in accordance with work functions @xmath269 and @xmath270 given in tabl .
thus the highest occupied state of the overlayer lies @xmath271 mev above the top of the valence band of substrate .
the attaching of the topologically trivial insulating overlayer to ti substrate keeps the gapless topological surface state , however , it leads to strong modification of the spectrum ( fig .
[ fig10]b ) : the dp shifts towards the bottom of the conduction band of the substrate so that above the dp it propagates as a resonant state , mixed with bulk - like states of the bi@xmath0te@xmath0s slab . in the real space ,
the tss in the heterostructure is almost completely relocated into the overlayer and its probability maximum lies near the gebi@xmath0te@xmath267/bi@xmath0te@xmath0s interface plane ( fig .
[ fig10]c , d ) .
such a behavior of the topological state can be explained by the potential change upon the interface formation . in fig .
[ fig10]e the change in electrostatic potential of the heterostructure with respect to potentials in free - standing substrate and overlayer is shown .
one can see that due to hybridization between orbitals of the overlayer and substrate the potential within the ti substrate is smoothly bent towards the interface plane while the potential within the overlayer undergoes more noticeable changes and as a whole it shifts down by @xmath272 mev with respect to its position in the free - standing overlayer . in spite of the downward shift of the potential the resulting topological state has the dirac point position higher than in the tss of the pristine ti surface .
the change in the dp position is related to the fact that in the heterostructure the orbitals of gebi@xmath0te@xmath267 contribute more to the tss than to the orbitals of bi@xmath0te@xmath0s .
thus the modification of the topological state in the considered system is qualitatively similar to the behavior found in the continual model with @xmath273 where for negative @xmath274 beneath a critical value arise solutions with positive shift of the dirac point ( see fig .
[ fig1]b ) . in the sb@xmath0te@xmath0s / sb@xmath0te@xmath0se heterostructure ( fig . [ fig11 ] )
the tss dispersion remains almost unchanged with respect to that on the pristine sb@xmath0te@xmath0se substrate surface being shifted towards the bulk valence band by @xmath27540 mev .
along with this , the maximum of charge density of the topological state relocates into the sb@xmath0te@xmath0s ql ( fig .
[ fig11]c , d ) owing to downward shift of the overlayer potential ( fig .
[ fig11]e ) and demonstrates a resonance like behavior due to the dp proximity to the bulk continuum . in the model
this scenario is realized at @xmath276 ( fig .
[ fig1]c ) when moderate negative potential at the interface does nt leads to substantial change in the energy of the tss .
a rather distinct type of the tss modification demonstrate the bi@xmath0te@xmath0s / gebi@xmath0te@xmath267 system ( fig .
[ fig12 ] ) .
the work function of the free standing overlayer is @xmath277 ev larger than that of the gebi@xmath0te@xmath267 substrate slab ( see tabl . [ t1 ] ) .
this means that the gap - edge states of the overlayer are far below the bulk gap of ti substrate ( fig .
[ fig12]a ) .
in contrast to the previous cases , where at attaching of the overlayer the topological state remains within the bulk gap of the substrate , the tss in the bi@xmath0te@xmath0s / gebi@xmath0te@xmath267 heterostructure occurs at -0.4 ev , in the local @xmath268 gap of the bulk valence band of gebi@xmath0te@xmath267 .
this state lies in deep and abrupt potential well of @xmath278 mev depth that causes its strong localization within bi@xmath0te@xmath0s ql .
thus the modification of the tss in the bi@xmath0te@xmath0s / gebi@xmath0te@xmath267 heterostructure is qualitatively similar to the case @xmath279 ( fig .
[ fig1]c ) in the continual model where large negative value of the potential results in huge shift of the dirac point down from its position in the initial system .
in this work , we have shown that the energy spectrum and spatial profile of the near - surface electron states in the 3d ti substrate / ni overlayer type of heterostructures can be controlled by the overlayer induced sp . by choosing an appropriate overlayer it is possible to adjust the dirac point to the required position in the band gap .
the dft calculations demonstrate clearly how the overlayer - induced shift of the dirac point from its energy position on the pristine surface is associated with the parameters of the band structure of the 3d ti substrate and the quintuple / septuple overlayer .
these results are in good qualitative agreement with the tendency predicted from the analytic continual scheme , in which the overlayer - induced change in the energy spectrum is determined by the sp matrix elements , @xmath93 .
thus the sp matrix elements can be intuitively associated with the relative energy offsets between the relevant band edges of the substrate and overlayer : @xmath280 , @xmath281 , where @xmath282 and @xmath283 are bandgaps and work functions of an overlayer ( subscribe 1 ) and a ti substrate ( subscribe 2 ) . in the framework of the continual approach ,
we have succeeded in formulating general boundary conditions for the long - range ef and finding the solution for electron bound states at the ti / ni interface .
we have obtained analytical expressions for the energy spectrum and ef for different types and values of sp .
our results are strictly consistent with the limiting cases of the zero and infinite surface potentials , which were previously studied in refs .
the boundary conditions of eq .
( [ current ] ) involves the sp parameters , in this sense the represented approach has something in common with the one used to calculate the surface states of a crystal with the relativistic band structure @xcite .
note that in ref .
@xcite , the bound states at the interface between 3d ti and ni have been explored on the basis of the functional defined in the entire space . in that work ,
the explicit expressions for the matrix elements of the interface pseudo - potential , which affects electrons on the ti side of the interface , have been analytically derived . from the theoretical point of view
, one can not suggest universal recipe to impose the restrictions on the ef behavior at the 3d ti boundary for all types of the heterostructures containing 3d tis .
note , in particular , that the boundary conditions of eq .
( [ current ] ) imposed upon the near - surface states are different from those obtained in ref .
@xcite , where , for an interface between 3d ti and a topologically trivial insulator , authors succeeded in a formulation of the ef boundary task , the solution of which provides insight into the electron states at the interface . although the ef approach may be adapted for the description of the interface states in many semiconductor junctions , by taking into account the general hermiticity and symmetry requirements @xcite , this traditional description is not always able to capture the principal features of the topological states in the systems containing narrow - gap semiconductors with inverted band structure and sometimes yields rather dubious results @xcite . as for the above - stated conception ,
the appropriate ef boundary conditions are derived within the framework of the formalism of the sheet - like sp .
in such the approach , some information ( for example , about the effects of electron - electron interaction ) is evidently lost that is compensated by a relative calculation simplicity and a transparent physical interpretation .
the method , conceptually presented in our work for the study of the truncated 3d ti covered with an atomically thin non - magnetic insulating overlayer , can be straightforwardly extended to solve a wide range of problems related to a behavior of the near - surface topological states under the surface perturbations listed in introduction . indeed , as it is shown in sec . 4
, the area of applicability of the sp method significantly oversteps the formal limits of the 3d ti substrate / ni overlayer systems .
we have determined the existence or absence of the dirac - like near - surface modes in the hypothetical situations when the truncated 3d ni ( close to transition into a topological phase ) is brought into the contact with the ultrathin overlayer of the 3d ni or ti material .
we predict that the near - surface topological - like " mode can appear in the 3d ni substrate / ni overlayr , i.e. , in the system composed only of two topologically trivial materials .
this unusual item has something in common with the fact of the appearance of 2d ti state in an inas / gasb type - ii semiconductor quantum well ( qw ) @xcite .
the unique feature of inas / gasb qw is that the conduction band minimum of inas has lower energy than the valence band maximum of gasb ( due to the large band - offset ) .
consequently , when the qw thickness is large enough , the first electron subband of inas layer lies below the first hole subband of gasb layer , i.e. , an inverted band alignment , similar to that in hgte qws @xcite , happens .
the experimental study of low temperature electronic transport have shown strong evidence for the existence of helical edge modes in the hybridization gap of inverted inas / gasb qws @xcite .
in view of the aforesaid , one raises the following questions concerning different strategies to create the topological near - surface / interface states and manage their electronic properties : ( i ) how does a 3d ni overlayer influence a 3d ti ?
( ii ) when one puts an overlayer of one 3d ti on a substrate of another 3d ti , what new property will come into being ? ( iii ) how about a hypothetical 3d ti thin film on the surface states of 3d ni ?
( iiii ) finally , one could ask how to construct a quasi - topological bound state at the boundary between two topologically trivial insulators ?
our work partly answers these questions .
the approach proposed above unveils the physics of the near - surface states in the semiconductor heterostructures containing 3d tis . at the same time , it might be considered as a good guidebook for the qualitative interpretation and forecast of the topological phase behavior tendencies in 3d tis under surface perturbations .
the obtained results should open new opportunities to design various combinations of topological and conventional materials for electronic / spintronic applications .
99 and @xmath284 . in the dashed areas the dirac point is inside the bulk band gap .
( b ) position of the dirac point versus the sp strength when @xmath285 , where @xmath286 ( black line ) , @xmath287 ( red line ) , @xmath288 ( green line ) , @xmath289 ( blue line ) .
( c ) position of the dirac point versus the sp strength when @xmath285 , where @xmath286 ( black line ) , @xmath290 ( red line ) , @xmath291 ( green line ) , @xmath292 ( brown line ) , @xmath293 ( blue line coinciding with an abscissa axis ) .
the units used are , @xmath294 , @xmath295 , and @xmath296 . ] ) of the envelope function @xmath297 eq .
( [ ef - stagg ] ) in the case of @xmath161 for several values of the potential : ( a ) @xmath298 ( black line ) , @xmath299 ( blue line ) , @xmath300 ( red line ) , @xmath301 ( green line ) , @xmath302 ( brown line ) , @xmath303 ( yellow line ) ; ( b ) @xmath298 ( black line ) , @xmath304 ( blue line ) , @xmath305 ( red line ) , @xmath306 ( green line ) , @xmath307 ( brown line ) , @xmath308 ( yellow line ) ; where @xmath309 , @xmath310 , @xmath311{\frac{\mathrm{b}}{\xi}}\theta_{j}$ ] , @xmath296 .
the envelope function is normalized as @xmath312 . ] ) of the envelope function @xmath297 in the case of @xmath124 for several values of ( a ) the relative weak positive potential with @xmath316 [ @xmath298 ( black line ) , @xmath317 ( red line ) , @xmath318 ( green line ) , @xmath319 ( blue line ) ] , ( b ) the relative weak negative potential with @xmath320 [ @xmath298 ( black line ) , @xmath321 ( red line ) , @xmath322 ( green line ) , @xmath323 ( blue line ) ] , ( c ) the relative large positive potential with @xmath324 [ @xmath303 ( black line ) , @xmath302 ( red line ) , @xmath325 ( blue line ) , @xmath326 ( green line ) ] , ( d ) the relative large negative potential with @xmath327 [ @xmath308 ( black line ) , @xmath307 ( red line ) , @xmath328 ( blue line ) , @xmath329 ( green line ) ] , where @xmath294 , @xmath310 , @xmath311{\frac{\mathrm{b}}{\xi}}\theta_{j}$ ] , @xmath296 .
the envelope function is normalized as @xmath312 . ] and
@xmath284 . in
the dashed ( fill , painted ) areas ( domains ) on the @xmath116-plane , the point @xmath200 lives inside the bulk band gap .
( b ) position of the band edge energy of the near - surface state versus the sp strengths of the same sign under the stipulation @xmath330 and @xmath331 , where @xmath332 ( green line ) , @xmath333 ( red line ) , @xmath334 ( brown line ) , @xmath335 ( black line ) .
( c ) position of the band edge energy of the near - surface state versus the sp strengths of opposite signs under the stipulation @xmath330 and @xmath331 , where @xmath332 ( green line ) , @xmath336 ( red line ) , @xmath337 ( brown line ) , @xmath338 ( blue line ) , @xmath339 ( black line ) .
the dimensional units are used , @xmath221 , @xmath340 , and @xmath341 . ] at @xmath161 for several values of the sp strength : ( a ) @xmath342 [ @xmath343 ( black line ) , @xmath344 ( red line ) , @xmath345 ( brown line ) , @xmath346 ( green line ) , @xmath347 ( blue line ) ] ; ( b ) @xmath348 [ @xmath349 ( red line ) , @xmath350 ( brown line ) , @xmath351 ( green line ) , @xmath343 ( black line ) ] , where @xmath221 , @xmath340 , @xmath352 , @xmath341 . ] ) of the ni envelope function @xmath297 in the case of @xmath161 for several values of the sp strength with ( a ) @xmath353 [ @xmath354 ( black line ) , @xmath347 ( red line ) , @xmath346 ( blue line ) , @xmath345 ( green line ) , @xmath344 ( brown line ) , @xmath355 ( yellow line ) ] , ( b ) @xmath356 [ @xmath357 ( black line ) , @xmath351 ( red line ) , @xmath350 ( blue line ) , @xmath349 ( green line ) , @xmath358 ( brown line ) ] , where @xmath221 , @xmath359 , @xmath311{\frac{\mathrm{b}}{|\xi|}}\theta_{j}$ ] , @xmath341 .
the envelope function is normalized as @xmath312 . ] at @xmath124 for several values of the sp strength : @xmath350 ( red line ) , @xmath360 ( brown line ) , @xmath361 ( green line ) , @xmath347 ( blue line ) , @xmath343 ( black line ) , where @xmath221 , @xmath340 , @xmath352 , @xmath341 . ] ) of the ni envelope function @xmath297 in the case of @xmath124 for several values of the sp strength : @xmath326 ( black line ) , @xmath362 ( red line ) , @xmath363 ( blue line ) , @xmath347 ( brown line ) , @xmath364 ( green line ) , where @xmath221 , @xmath359 , @xmath311{\frac{\mathrm{b}}{|\xi|}}\theta_{j}$ ] , @xmath341 .
the envelope function is normalized as @xmath312 . ]
te@xmath0s ti substrate and free - standing septuble layer block of gebi@xmath0te@xmath267 ( red lines ) .
blue markers show the weight of the states in near - surface ql of the substrate .
shaded areas indicate projection of the bulk states of bi@xmath0te@xmath0s ti onto the surface brillouin zone ; ( b ) electronic spectrum of gebi@xmath0te@xmath267/bi@xmath0te@xmath0s heterostructure .
the size of red circles reflects the weight of the states in the gebi@xmath0te@xmath267 overlayer ; spatial distribution of the tss charge density integrated over ( @xmath365 , @xmath366 ) plane for bi@xmath0te@xmath0s ( c ) and gebi@xmath0te@xmath267/bi@xmath0te@xmath0s ( d ) ; ( e ) the change in electrostatic potential @xmath367 of gebi@xmath0te@xmath267/bi@xmath0te@xmath0s with respect to potentials in free - standing bi@xmath0te@xmath0s ti substrate and gebi@xmath0te@xmath267 overlayer . ]
te@xmath0se substrate and free - standing sb@xmath0te@xmath0s ql ( red lines ) .
blue markers show the weight of the states in near - surface ql of the ti substrate ; ( b ) electronic spectrum of sb@xmath0te@xmath0s / sb@xmath0te@xmath0se heterostructure ; spatial distribution of the tss charge density for pristine sb@xmath0te@xmath0se surface ( c ) and sb@xmath0te@xmath0s / sb@xmath0te@xmath0se heterostructure ( d ) ; ( e ) the change in electrostatic potential @xmath367 of sb@xmath0te@xmath0s / sb@xmath0te@xmath0se with respect to potentials in free - standing sb@xmath0te@xmath0se substrate and sb@xmath0te@xmath0s ql . ]
te@xmath267 slab and free - standing bi@xmath0te@xmath0s ql ( red lines ) .
blue markers show the weight of the states in near - surface sl of gebi@xmath0te@xmath267 ; ( b ) electronic spectrum of bi@xmath0te@xmath0s / gebi@xmath0te@xmath267 heterostructure ; spatial distribution of the tss charge density for pristine substrate surface ( c ) and the heterostructure ( d ) ; ( e ) the change in electrostatic potential @xmath367 of the heterostructure with respect to potentials in free - standing gebi@xmath0te@xmath267 slab and bi@xmath0te@xmath0s ql . ] | we apply both analytical and ab - initio methods to explore heterostructures composed of a three - dimensional topological insulator ( 3d ti ) and an ultrathin normal insulator ( ni ) overlayer as a proof ground for the principles of the topological phase engineering . using the continual model of a semi - infinite 3d ti we study the surface potential ( sp ) effect caused by an attached ultrathin layer of 3d ni on the formation of topological bound states at the interface .
the results reveal that spatial profile and spectrum of these near - surface states strongly depend on both the sign and strength of the sp . using ab - initio band structure calculations to take materials specificity into account
, we investigate the ni / ti heterostructures formed by a single tetradymite - type quintuple or septuple layer block and the 3d ti substrate .
the analytical continuum theory results relate the near - surface state evolution with the sp variation and are in good qualitative agreement with those obtained from density - functional theory ( dft ) calculations .
we predict also the appearance of the quasi - topological bound state on the 3d ni surface caused by a local band gap inversion induced by an overlayer . |
out of equilibrium systems may display phase transitions analogous to those found in their equilibrium counterparts .
these transitions are classified into distinct universality classes , characterized by a set of critical exponents @xcite .
particular attention has been paid to one - dimensional systems with transitions from an active state into absorbing states , i.e. , frozen configurations from which the system can not escape . for these systems , so far , only two distinct universality classes have been firmly established : the directed percolation ( dp ) and the parity conserving ( pc ) universality class . while the former is ubiquitous and found in several systems with very different dynamical rules , the latter is only known to occur when extra symmetries are present @xcite .
a model which has attracted quite some interest recently , because it may indeed belong to a novel universality class @xcite is the so - called pair contact process with diffusion ( pcpd ) . despite a rather intense activity in the past couple of years @xcite , the understanding of the pcpd and
its relation with other known models is still unsatisfactory . in the _ fermionic _ version of the model
the one studied here each lattice site is either occupied by a single particle ( @xmath0 ) or empty ( @xmath1 ) .
the reactions are @xmath2 with @xmath3 , @xmath4 .
the analysis of the critical properties of the pcpd has shown to be much more difficult than all similar models analyzed so far and several scenarios have been proposed .
first , a similarity of the exponents @xmath5 and @xmath6 ( where @xmath7 and @xmath8 are the correlation length exponents along the time and space directions and @xmath9 the order parameter exponent ) with those of the pc class @xcite was reported , although in the pcpd there is no conservation of parity .
it was later suggested that the pcpd could belong to a new universality class with exponents close to , but different from the pc values @xcite .
it has also been argued that there could be two universality classes @xcite at small and large diffusion rates ( @xmath10 ) , or continuously varying exponents @xcite , or that scaling may even be violated @xcite .
more recently the option of a slow crossover to dp was also discussed @xcite , although the general belief is that the pcpd belongs to a novel universality class @xcite .
field theoretical methods , which have been successfully applied to other reaction diffusion models @xcite failed so far to clarify the critical properties of the pcpd @xcite .
in this paper we present some insights into the pcpd .
we show that accurate numerical results from monte carlo ( mc ) simulations and density matrix renormalization group ( dmrg ) calculations convincingly demonstrate that for sufficiently long times and system lengths the exponents crossover towards the dp values .
corrections to scaling are , however , rather strong and only an accurate extrapolation of the effective critical exponents allows to identify the final asymptotic critical behavior .
we give evidence that in the near - asymptotic region the model shows effective exponents close to the pc class values , which suggests that the critical behavior of the system is described by two competing fixed points .
our mc simulations exploit a technique known as multispin coding @xcite .
the basic idea is that in a simulation of 64 systems , each with @xmath11 sites , the occupation of site @xmath12 in the @xmath13th simulation is stored in the @xmath13th bit of 64-bit word @xmath14 $ ] . to perform the reaction @xmath15 in all 64 systems at a randomly chosen site @xmath12 and its neighbors ,
one logical operation @xmath16=a[i+1]\lor ( a[i ] \land a[i-1])$ ] suffices , where @xmath17 and @xmath18 are the logical operations or and and , respectively .
other reactions require only slightly more elaborate logical operations .
a direct implementation along these lines might result in 64 simulations , each of which statistically correct , but strongly correlated to each other because the site selection is shared . to alleviate this correlation without sacrificing efficiency
, we employ random bit patterns that decide which reaction will be attempted in which system .
the strong point of multispin coding is its efficiency . as illustrated above , only a few logical operations ( each usually carried out in a single clock cycle without delay )
suffice to update a site in 64 systems simultaneously . for each combination of @xmath10 and @xmath19 we simulated 64 systems with @xmath20 sites over @xmath21 monte carlo time units , in about 15 hours on a single processor workstation
. we also simulated @xmath22 systems with @xmath20 sites over @xmath23 monte carlo time units on a parallel computer , for @xmath24 and @xmath25 , our estimate for the critical point . and
pair density @xmath26 , as a function of @xmath27 , with @xmath28 , for the pcpd with @xmath24 and at the estimated critical point @xmath25 .
this ratio approaches a constant.,width=291 ] a standard procedure to obtain critical exponents by means of monte carlo simulations is to study the decay of the particle density @xmath29 as a function of time @xmath30 , starting from a random configuration of particles . at the critical point
one has @xmath31 . to monitor the decay it is convenient to define the effective exponent @xmath32 .
typically @xmath33 is plotted as a function of @xmath34 . at the critical point , in the limit @xmath35
, it approaches a finite value ( @xmath36 ) , while it deviates upwards or downwards with respect to this value in the inactive and active phases , respectively .
this criterion allows to estimate the critical point location and the ratio @xmath37 @xcite .
some care , however , has to be taken when corrections to scaling are particularly strong , for example when @xmath38 with @xmath39 and @xmath40 a constant . in this case
@xmath33 plotted as a function of @xmath34 approaches @xmath37 with an infinite slope .
the ideal situation would be to plot @xmath33 as a function of @xmath41 as in that case the approach to the asymptotic value would be linear .
further on in this paper we will give numerical evidence that the correction - to - scaling exponent @xmath42 is close to @xmath43 ( which is much smaller than 1 ) .
a natural choice is therefore to plot @xmath33 as a function of the particle density @xmath44 instead of @xmath34 , to avoid infinite slopes .
besides the decay of the particle density @xmath44 , we also consider the decay of the pair density @xmath45 . before presenting the results for the critical exponents
we analyze the behavior of the ratio @xmath46 at the critical point .
such a quantity , which is shown in fig .
[ fig01 ] for @xmath24 , approaches a constant value ( @xmath47 ) asymptotically for long times .
the most important consequence of this fact is that in pcpd one can extract the critical exponent @xmath37 both from the decay of the particle and pair densities @xcite .
numerically the ratio is much better behaved than the individual densities @xmath44 and @xmath26 , as fluctuations in the individual densities are highly correlated and largely cancel each other in the ratio . at @xmath48 and @xmath49 ,
@xmath25 , and @xmath50 ( from bottom to top ) obtained from the decay of the particles and pairs densities .
the dashed lines are parabolic fits to the data .
the two horizontal lines show the ratio @xmath37 for the dp and pc class .
inset : @xmath51 and @xmath52 , @xmath53 , @xmath54.,width=291 ] the data of fig .
[ fig01 ] also provide an estimate of the leading correction term in the asymptotic limit @xmath35 , which appears to be of the type @xmath55 with @xmath56 , as shown by the linear approach to the asymptotic behavior of the data when plotted versus @xmath57 .
this exponent is much larger than its equivalent @xmath42 in the particle and pair densities , which implies that the leading corrections for @xmath44 and @xmath26 cancel in the ratio @xmath46 .
we tested that a similar cancellation also occurs in the process @xmath58 , @xmath59 , which belongs to the pc universality class @xcite : the leading correction in @xmath44 and @xmath26 scales as @xmath60 , whereas the leading correction in the ratio @xmath46 scales as @xmath34 .
figure [ fig02 ] shows a plot of @xmath33 at @xmath24 as a function of @xmath44 and @xmath26 , calculated in the pcpd from the decay of particles and pairs , respectively .
three different values for @xmath19 have been plotted around the critical point which we estimate as @xmath61 . for @xmath62 ( inactive phase ) and @xmath63 ( active phase ) @xmath33 rapidly veers up and down , as expected . at the critical point
, @xmath33 approaches the @xmath64 axis with a finite slope , indicating that the leading correction to scaling is most likely described by an exponent roughly equal to @xmath65 itself .
a parabolic fit through the data yields as a common estimate for the critical exponent @xmath66 .
similar calculations were repeated for other values of the diffusion coefficient @xmath67 , @xmath68 , @xmath69 , and @xmath70 ( the inset of fig .
[ fig02 ] shows the case @xmath51 ) , with the same results which we can summarize with the estimate @xmath71 .
this value is consistent with the dp class exponent @xmath72 .
( particle density ) vs @xmath30 .
( a ) @xmath24 , @xmath49 , @xmath25 and @xmath50 ( from top to bottom ) .
( b ) @xmath73 and @xmath74 , @xmath75 and @xmath76 ( from top to bottom ) .
the critical point densities are plotted as thick solid lines .
the dashed lines are linear fits to the data , shifted for clarity.,width=291 ] an alternative way to calculate critical exponents would be to use a linear fit of the densities versus time on a double - logarithmic scale .
however , such fits are hazardous when dealing with very slow convergence , as is the case here , and may lead to wrong estimates for the critical exponents .
we illustrate this in fig .
[ fig03 ] which shows the plot of the average particle density as a function of the time in a double - logarithmic scale for @xmath24 and @xmath73 . in the former case
a straight line ( dashed ) fits extremely well the critical density decay leading to the estimate @xmath77 .
the analysis of the effective exponent ( see fig . [ fig02 ] ) , however , provides a closer inspection of the local slopes of the double - logarithmic data .
this analysis reveals some remaining curvature , and the final estimate of the exponent is significantly lower , compared to that obtained from the fit in the double - logarithmic scale . in the case of higher diffusion
[ see fig . [ fig03](b ) ] the curvature is more pronounced and clearly visible also in the double - logarithmic plot , which can be fitted by two straight lines with slopes @xmath78 in the range @xmath79 , and with @xmath80 for @xmath81 .
notice that the former exponent is consistent with that expected for the pc class ( @xmath82 @xcite ) .
again an extrapolation of the effective exponent ( as done in fig .
[ fig02 ] ) shows convergence to a value consistent with dp .
note that the value @xmath83 is consistent with the most recent monte carlo estimates for the pcpd @xcite .
in particular , kockelkoren and chat , @xcite performed a series of monte carlo simulations for a bosonic version of the pcpd where the constraint of one particle per site is released .
their estimation of critical exponents is based on a straight - line fit to a double - logarithmic plot of @xmath44 versus @xmath30 , from which they find @xmath84 .
this bosonic version of the model is claimed to suffer less from corrections to scaling than the fermionic case .
notice , however , that also in the fermionic pcpd studied here the density decay at @xmath24 ( see fig .
[ fig03](a ) ) is rather straight in a double - logarithmic plot for simulation times similar to those in ref .
the advantage of the effective exponent analysis performed here is that it allows to extrapolate the numerical results to time scales _ beyond _ those actually simulated . for @xmath24 at the system critical point ( @xmath25 ) for lattices up to @xmath85 calculated from the decay of particle ( circles ) and pair ( squares )
densities as function of the lattice lengths .
the dotted and dashed lines are fits in powers of the densities .
inset : @xmath86 for @xmath87.,width=291 ] next we present some dmrg results .
dmrg @xcite allows to calculate accurate stationary state probabilities for chains of moderate lengths @xcite . as usual in dmrg , we used open boundary conditions .
in the pcpd on a lattice of finite length there are only two stationary states : a state with no particles and a state occupied by a single diffusing particle . to induce a finite density of particles we added a reaction @xmath88 at the two boundary sites .
the particle density decays from the two boundaries and forms a u - shaped profile . for chains of various lengths we calculated the density of particles @xmath89 and of pairs @xmath90 at the central site of a system of length @xmath11 . at the critical point
these quantities decay in the limit @xmath91 as @xmath92 .
figure [ fig04 ] shows the effective exponent @xmath93 versus @xmath44 for @xmath24 at the critical point . as in fig .
[ fig02 ] we include also the data for the pairs .
dotted and dashed lines are fits with polynomials in the densities . again , a test of good convergence is that both exponents extrapolate to the same asymptotic value .
this requirement seems indeed to be fulfilled and we find as extrapolation @xmath94 .
this exponent is again consistent with the dp value @xmath95 @xcite .
similar results have also been found for other values of the diffusion coefficient @xmath10 .
extrapolations for @xmath87 are shown in the inset of fig .
[ fig04 ] . at small @xmath10 the maxima in @xmath86 shift to longer @xmath11 ,
thus extrapolations are somewhat less stable . in this case
we take the estimate obtained from the particles @xmath96 .
previous dmrg results @xcite were restricted to the density of particles and to smaller systems than that studied here . the data for the effective exponent showed a monotonic behavior ( except at very strong diffusion ) and were analyzed using an extrapolation with polynomes in @xmath97 .
these extrapolations lead to a value consistent with the pc class exponent @xmath98 @xcite .
the present calculation , extended to the density of pairs and to longer systems , reveals that nonmonotonicity in the effective exponent is a common feature at all @xmath10 .
this nonmonotonicity leads to a rather strong decrease of the extrapolated exponent compared to the estimates of ref .
@xcite . as a function of @xmath97 for @xmath87 , @xmath99 , @xmath100 , and @xmath70 .
inset : blowup of the data for @xmath24 .
, width=291 ]
from the ratio of the exponents @xmath5 and @xmath37 one can estimate the dynamical exponent @xmath101 .
since both @xmath5 and @xmath37 are consistent with dp , also the dynamical exponent @xmath102 agrees with the dp value @xmath103 .
it is however instructive to show the results of an independent calculation of @xmath102 .
this quantity can be obtained from a finite size scaling analysis of @xmath104 , the gap of the master operator , which is the inverse of the relaxation time of the system ( see ref .
@xcite for details ) . as
a function of the system length @xmath11 the gap decays as @xmath105 .
figure [ fig05 ] shows a plot of the effective exponent @xmath106 versus @xmath97 .
the calculations are similar to those reported in ref .
@xcite , but now for longer systems ( up to @xmath107 compared to @xmath108 of ref .
the critical point locations were obtained from monte carlo simulations , which for this purpose are faster and more efficient than dmrg .
therefore we concentrated our computational efforts on a single value of @xmath109 and could obtain results for longer systems .
as is clear from fig .
[ fig05 ] , the exponent @xmath110 is rather sensitive to the value of the diffusion rate @xmath10 . as the estimates of @xmath5 and @xmath37 are instead rather stable as a function of @xmath10 we contribute this sensitivity to rather strong finite - size effects .
notice that the finite @xmath11 corrections change sign from the weak to the strong diffusion regime .
the border value is around @xmath24 where @xmath110 has a very weak dependence on @xmath11 .
the data ( see inset ) run extremely close to the pc value @xmath111 . at higher diffusion rates @xmath112 the effective exponent @xmath110 for the range of sizes investigated is much lower than @xmath113 . at the strongest diffusion
investigated , extrapolations with different forms for the correction to scaling terms as @xmath97 or @xmath114 yield values in the range @xmath115 , which should be compared with the dp value @xmath116 .
current monte carlo estimates from various authors @xcite place the exponent @xmath102 in the range @xmath117 and the calculations were mostly performed in the weak diffusion regime . at the estimated critical point @xmath75 for @xmath118 , @xmath119 , @xmath120 , @xmath121 , @xmath122 , and @xmath123 .
inset : plot of @xmath124 vs @xmath125 for two values of the diffusion constant .
we estimate @xmath126 for @xmath73 and @xmath127 for @xmath87 .
, width=283 ] we also performed a series of monte carlo simulations to calculate the exponent @xmath102 using finite - size scaling analysis . at the critical point and on a finite system the particle density decays as @xmath128 , with @xmath129 a scaling function . for finite @xmath11 , @xmath44 follows a power - law decay up to a characteristic time @xmath130 after which it drops exponentially .
one expects that @xmath130 scales as @xmath131 . to estimate @xmath130 we calculated @xmath44 for lattices up to @xmath132 and @xmath133 monte carlo time units .
figure [ fig06 ] shows a plot of @xmath134 versus @xmath135 for various @xmath11 and @xmath73 .
the intersection of the data with a horizontal line at @xmath136 ( with @xmath13 a constant ) provides an estimate of @xmath130 . as we work in a region where the particle density is rather low and fluctuations
are large , and as the calculation of @xmath102 requires very smooth data , we performed averages over a large number of samples ( @xmath137 ) . for the calculation we used @xmath66 , which is the value determined above , and @xmath138 ( see fig .
[ fig06 ] ) .
the inset shows a double - logarithmic plot of @xmath130 versus @xmath11 for @xmath87 and @xmath73 at their critical points . in the former case we restricted ourselves to @xmath139 as
the relevant times are typically longer at weak than at stronger diffusion , as expected .
notice that in both cases the data are well fitted by straight lines yielding the estimates @xmath127 for @xmath87 and @xmath126 for @xmath73 , where the latter value is consistent with the dynamical exponent of directed percolation @xmath116 .
the results generally confirm the dmrg findings according to which the dynamical exponent is generically smaller , for finite @xmath11 , at higher diffusivity .
we also notice that by varying the value of @xmath37 entering in the @xmath64 axis of fig .
[ fig06 ] one changes the estimate for @xmath102 .
for instance , if we take @xmath140 , as calculated in ref .
@xcite , this leads to an increase of @xmath141 in the estimated value for @xmath102 .
the estimate of @xmath102 is rather stable for changes in the constant @xmath13 . for three different values of the diffusion constant .
horizontal lines are the reference exponents for the dp and pc universality classes . , width=291 ]
boundary quantities are easily accessible in dmrg techniques @xcite , as one is basically forced to work with open boundary conditions .
surface criticality in absorbing phase transitions has been the subject of several studies in the past years both for models in the dp @xcite and in the pc @xcite universality classes . in the latter case
, it is known that there are two distinct surface exponents depending on the type of boundary conditions applied @xcite .
the results of a dmrg calculation of the surface critical exponents for a reaction - diffusion model in the pc class are presented in the appendix . here
, we report on the surface critical exponent calculations for the pcpd , using the same types of boundary conditions as in the appendix .
vs @xmath142 for different values of the diffusion constant and absorbing boundary conditions .
inset : @xmath143 vs @xmath144 calculated from the surface pair density . , width=291 ] as in the calculation of the bulk particle density of the preceding section we inject particles through the reaction @xmath88 at the boundary site labeled by the position @xmath145 in order to induce a finite density of particles in the system , and we measure the particle density @xmath146 at the opposite boundary site @xmath147 .
asymptotically for @xmath148 , we expect @xmath149 , where @xmath150 is the order parameter surface exponent .
the two different boundary conditions ( bcs ) applied at the site @xmath147 are : ( a ) no particles are allowed to leave the system from the boundary site and ( b ) particles may diffuse out of the system , i.e. , the reaction @xmath151 ( with rate @xmath10 ) is added at that site .
we refer to these as _ reflecting _ and _ absorbing _ boundary conditions , respectively . in fig .
[ fig07 ] we plot the effective exponent @xmath152 versus @xmath142 in the case of reflecting bcs .
horizontal lines show the ratio @xmath153 for dp ( @xmath154 @xcite ) and pc ( @xmath155 @xcite ) . in the dp case
the different bcs produce the same critical exponent .
effective exponents in this case grow monotonically , contrary to what is found for bulk exponents .
notice that a cubic fit yields a quite stable estimate @xmath156 in the range @xmath157 , a value actually consistent with the surface exponent for the pc class ( see appendix ) . only at higher @xmath10
we observe some deviation from pc .
the fact that the extrapolated surface exponents vary with @xmath10 , while our current estimates for the bulk exponents are independent on @xmath10 , is an indication that the former are not yet the correct asymptotic ones .
figure [ fig08 ] shows @xmath143 versus @xmath142 for the case of absorbing boundary conditions .
again for weak diffusion the exponent seems to extrapolate rather convincingly to values close to the pc class ( @xmath158 , see appendix ) , while for strong diffusion it increases to much larger values .
also in this case there is no clear signature of nonmonotonic behavior , except for the case @xmath24 where the data for the largest systems pass through a maximum .
we also analyzed the effective exponent data from the pair density @xmath144 which are shown in the inset of fig .
[ fig08 ] in the case of absorbing bcs , and plotted as functions of @xmath144 . in the range @xmath157
the data extrapolate close to the pc surface exponent @xmath158 as for the particle density . at very strong diffusion ( @xmath159 )
the surface effective exponent shows a nonmonotonic behavior with a maximum around @xmath160 .
notice that particle and pair exponents in this case are rather far apart from each other and it is quite hard to find a common extrapolation value .
we would expect for @xmath143 a similar behavior as for the bulk exponents , i.e. , an increase followed by a decrease towards the asymptotic value .
we suspect that in the present surface exponent calculation the decreasing side has barely been reached .
so we tend to distrust the extrapolation as estimates of the genuine asymptotic behavior .
they rather provide some insight on the preasymptotic region and actually point to a similarity with pc surface exponents at weak diffusion .
to conclude , by combining monte carlo and dmrg calculations we analyzed the critical properties of the pair contact process with diffusion .
this model has been the subject of increasing attention in recent years .
although the debate around it has not yet been settled , the main belief is that the pcpd belongs to a novel universality class which differs from the known dp and pc classes . in our opinion , however , the most plausible scenario for the pcpd is that it ultimately falls into the dp universality class .
the asymptotic behavior is , however , masked by rather strong finite size and time effects , characterized by small correction - to - scaling exponents , as our monte carlo simulations for the decays of the particle density @xmath44 , the pair density @xmath26 , and the ratio @xmath46 have demonstrated .
the exponents @xmath37 and @xmath5 extrapolated both from @xmath44 and @xmath26 appear to be stable as functions of the diffusion constant @xmath10 and actually consistent with the dp class values .
the data show a nonmonotonic behavior both in time and system size , which in our opinion points to a crossover phenomenon between two competing types of critical behavior .
the surface exponents , which we also investigated , turned out to be instead rather sensitive to the value of @xmath10 ; a sign , in our opinion , that the extrapolated values are probably not the true asymptotic ones .
interestingly enough , particularly at weak diffusion , the extrapolated values are rather stable and consistent with those expected for the pc class . in early numerical studies of the pcpd @xcite , restricted to shorter simulation times and system lengths compared to those considered here , several quantities as @xmath5 , @xmath37 , and @xmath161 were found to be quite consistent with the pc class values .
it is now generally agreed that the pcpd does not belong to the pc universality class , as more extensive simulations performed by several groups have shown convincingly @xcite .
still one would like to understand if the observed similarity with the pc exponent is purely fortuitous or if there is some deeper reason for it . in our opinion the evidence given above that also the surface exponents extrapolate towards pc values in an intermediate regime strongly suggests that there is a genuine nonasymptotic pc - like regime , with a crossover to dp behavior at longer time scales .
a prototype model in the pc class is the branching annihilating random walk with even offsprings ( barwe ) , defined by the reactions @xmath58 , @xmath59 plus diffusion @xcite , which differs from the pcpd only for the reaction which creates particles .
we argue that the early stages of the critical dynamics , when the system has a rather high particle density , are dominated by the annihilation process @xmath162 , so that the substitution of the barwe reaction @xmath58 with that of the pcpd @xmath163 may result in a very weak perturbation of the system . therefore a transient pc - like regime may be observed for @xmath164 , where @xmath165 is some crossover time .
this argument may help to explain features observed in the pcpd , and should be equally valid for other models where the annihilation is of the type @xmath59 and with different creation rules @xmath166 with @xmath167 , @xmath168 ; for such systems we expect a transient pc regime as well .
the study of reaction - diffusion systems where the annihilation and creation reactions involve @xmath167 particles has recently drawn some attention @xcite .
in particular , we mention here the two cases recently considered by dor @xcite ( i ) @xmath169 , @xmath59 and ( ii ) @xmath170 , @xmath59 .
in model ( i ) he estimates @xmath171 ( consistent with pc ) for small diffusion rates and @xmath172 at stronger diffusion . invoking some logarithmic corrections
he claims that all values extrapolate to @xmath173 @xcite . in case ( ii )
the estimate is @xmath171 both at high and low diffusions @xcite , again consistent with the pc class value .
the above observations suggest that these types of systems follow closely a critical behavior as described here for the pcpd , and it is thus plausible that they fall for sufficiently long times into the dp class .
however , it may turn out to be quite difficult to show this numerically , as we expect that increasing the number of particles involved in the creation and annihilation reactions will lead to models even harder to simulate and analyze than the pcpd . very recently
kockelkoren and chat analyzed a similar set of models @xcite . in their formulation
the fermionic constraint of only one particle per site is released .
all the reactions of the type @xmath174 and @xmath175 with @xmath176 were found to belong to the dp class .
surprisingly , in all those models the convergence to dp exponents seems to be quite fast ( at least for @xmath37 ) and not plagued by the strong corrections found in the fermionic models .
it would be interesting to study the same models at different values of the diffusion constant , as in the pcpd the onset of crossover behavior is quite strongly influenced by the value of @xmath10 .
we are grateful to j.d .
noh , h. hinrichsen , m. den nijs , and f. van wijland for useful discussions .
vs @xmath97 for the parity conserving process @xmath58 and @xmath59 in the case of reflecting boundary conditions .
inset : @xmath177 vs @xmath97 for absorbing boundary conditions .
dashed lines are extrapolated curves through the dmrg data .
extrapolated values are in good agreement with monte carlo simulations results of ref .
@xcite ( see text ) . ,
we present here some results on the surface critical behavior of the parity conserving process defined by the reactions @xmath59 , @xmath58 and with single particle diffusion .
we show how dmrg produces accurate surface critical exponents for this model , which are in good agreement with former monte carlo simulation results . for the single particle diffusion and pair annihilation we used the same rates as in eq .
( [ eq1 ] ) , while we assign a rate @xmath179 to the reaction @xmath180 .
we restrict ourselves to a single value of the diffusion constant @xmath24 .
we first estimated the critical point at @xmath181 by means of monte carlo simulations using a standard approach .
as mentioned above , for surface universality in pc processes there are two possible types of boundary conditions leading to two distinct surface exponents @xcite . in the first case ,
the system is truncated at one edge and no particles are allowed to cross the boundary site ; we refer to this as _ reflecting _ boundary conditions . in the second case
, particles are allowed to drop from the boundary .
we implemented this type of boundary condition adding the boundary reaction @xmath182 ( with rate @xmath10 ) which mimics the diffusion of particles out of the system .
we refer to this implementation as _ absorbing _ boundary conditions .
figure [ fig09 ] shows the effective surface exponent @xmath183 versus @xmath97 in the case of reflecting boundary conditions , calculated both from the particle ( circles ) and pair ( squares ) densities .
the same quantities are plotted in the inset in the case of absorbing boundary conditions .
notice that indeed the results confirm the existence of two distinct sets of surface exponents and that the data from pairs and particles merge for sufficiently long chains , indicating that both quantities decay with the same exponent .
our estimates @xmath184 in the former case and @xmath185 in the latter are obtained from a polynomial extrapolation in @xmath97 . as the finite - size effects are rather small ( see fig .
[ fig09 ] ) , the extrapolated values are not very sensitive to the type of correction to scaling term used in the extrapolation
. the monte carlo simulation results @xcite for the critical exponents are @xmath186 and @xmath187 , for inactive and active boundary conditions , respectively . combining these results with the pc class correlation length exponent @xmath188 @xcite , one finds @xmath189 ( reflecting bcs ) @xmath190 ( absorbing bcs ) , in very good agreement with the dmrg calculations . | the pair contact process with diffusion is studied by means of multispin monte carlo simulations and density matrix renormalization group calculations .
effective critical exponents are found to behave nonmonotonically as functions of time or of system length and extrapolate asymptotically towards values consistent with the directed percolation universality class .
we argue that an intermediate regime exists where the effective critical dynamics resembles that of a parity conserving process . |
new x - ray observatories such as the _ chandra x - ray observatory _ and _ xmm- newton _ have made high resolution spectroscopy of active galaxies possible for the first time , allowing the detection of phases of gas that were previously unknown .
the physical state of this gas is interpreted by reference to large - scale numerical simulations , hence , it is critical that physical processes that occur in an ionized plasma be fully understood ( the main topic of the conference `` spectroscopic challenges of photoionized plasmas '' , held at the university of kentucky , lexington , kentucky , usa on the 15 - 18 november , 2000 ; [ ferland & savin 2001 ] ) .
the ionization structure and temperatures in cosmic plasmas depend strongly on ionization and recombination processes .
recombination can occur via radiative recombination , charge transfer , or dielectronic recombination ( dr ) .
the dr process involves the capture of a free electron by an ion , resulting in doubly - excited recombined species with the electron in an auto - ionizing state .
this is followed either by auto - ionization ( i.e. , the reverse process ) or the emission of a photon , the latter leaving the recombined ion in a stable state .
there are two varieties of dr .
the first involves the capture of the electron to auto - ionizing states with high principal quantum number , with an accompanying core excitation with a change in principal quantum number @xmath0@xmath3 ( e.g. burgess 1964 ) .
this form of dr dominates over radiative recombination at temperatures comparable with the ionization potential , i.e collisionally ionized plasmas .
nevertheless , terms of the same parity as the ground term exist between the ground state and the first excited state of the opposite parity , and thus a series of autoionizing states converge on these terms ( see nussbaumer & storey 1983 ) .
this results in a second type of dr in which there is no change in the principal quantum number of the excited core , or @xmath0@xmath1 , which can proceed at low temperatures for which the thermal energy of the free electron is less than the excitation energy of first resonance of the recombining ion .
low - temperature dr is usually the dominant process in photoionized plasmas , with temperatures typically 5 - 10% of the ionization potential , such as those detected in x - ray spectra of seyfert galaxies ( e.g reynolds 1997 ; george et al .
1998 ; sako et al .
2000 ; kaspi et al . 2001 , 2002 ; kinkhabwala et al . 2002 ; turner et al .
2003 ) .
while low - temperature dr rates have been calculated for a number of abundant 2nd and a few 3rd row elements ( nussbaumer & storey 1983 ; nussbaumer & storey 1986 ; nussbaumer & storey 1987 ) , the full set of ionization states have only been completed for 2nd row elements .
one of the difficulties in these calculations is that the rates depend on the accurate determination of the low - level autoionizing states ( see gu 2003 ) .
ali et al . ( 1991 ) estimated low - temperature dr rate coefficients for other ions by taking the mean values from the first four ions of c , and o calculated by nussbaumer & storey ( 1983 ) , and demonstrated that the model - predicted emission - line strengths were strongly affected when these rates were introduced .
savin ( 2000 ) has shown that uncertainties in dr rates translate into similar uncertainties in predicted ionic column densities , which may hamper the determination of elemental abundances ratios in the intergalactic medium .
there has been recent work in determining both high- and low- temperature dr rates for l - shell ( the li- through ne - like iso - electronic sequences ) , both experimentally ( savin et al . 1999 , 2002 ) and computationally ( gu 2003 ; badnell et al .
2003 ) , the results of which emphasize the importance of this process in photoionized plasma .
note that , due to the difficulties in calculating the energies of the autoionizing states , experimental results are critical .
on the other hand , there has not been any recent work published on @xmath0@xmath1 rates for m - shell ions ( the na- to ar - like sequence ) .
this is unfortuitous , since m - shell fe ions , fe ix to fe xvi , contribute strongly to the unresolved transition array ( uta ) of inner shell @xmath4 lines seen in absorption in high resolution x - ray spectra of several seyfert galaxies observed with _
xmm - newton _ and _ chandra _ such as iras 13349 + 2438 ( sako et al .
2001 ) , ngc 3783 ( blustin et al .
2002 ; kaspi et al . 2002 ; netzer et al .
2003 ) , mrk 509 ( pounds et al . 2001 ; yaqoob et al .
2003 ) , and ngc 5548 ( steenbrugge et al . 2003
; steenbrugge et al . , in preparation ) . for n- and m - shell fe
, the uta features lie between @xmath5 15.5 and 17 , depending on which ions are present ( behar , sako , & kahn 2001 ) . in principal , the uta can be used to constrain the ionization state of the absorbers , which is especially valuable if the resonance lines from h- , he- , and li - like c , n , and o are saturated . however , in the case of ngc 3783 ( netzer et al .
2003 ) , the ionization state suggested by the uta wavelengths seems lower than that predicted by photoionization models generated to fit cno absorption features and the ionization state / column density of 3rd row elements such as si and s. a similar discrepancy appears to be the case for ngc 5548 ( steenbrugge et al .
, in preparation ) .
one of the possible causes netzer et al . suggested to explain this mismatch is the lack of @xmath0@xmath1 rates for the m - shell fe ions , which leads to the models predicting the fe in too high an ionization state ( compared to lower z elements ) . in order to explore the effect of @xmath0@xmath1 dr on the m - shell fe ionization balance
, we adopted the approach described in ali et al .
( 1991 ) , adding estimated rate coefficients to the development version of the photoionization code cloudy ( ferland et al .
1998 ) . in order to gauge the magnitude of the effect ,
model grids were generated with and without these rates .
we also examined the effects of uncertainties in the estimates rates .
the model details and results are described below .
since the mean @xmath0@xmath1 rate coefficients for cno show a obvious dependence on the residual charge of the ion before recombination ( ali et al . 1991 ) , we would expect a strong charge dependence among the 3rd and 4th row m - shell ions . in figure 1
, we show the rates for the l - shell fe and ar ions from gu ( 2003 ) for an electron temperature of 4 ev ( @xmath5 5 x 10@xmath6k ) , similar to what we would expect in photoionized plasmas in which the iron abundance was peaked at the m - shell .
since they span the same range in residual charge , one might expect that the m - shell fe rates would be similar to the l - shell ar rates .
hence , we estimated the rate coefficients for iron by generating 2@xmath7-order polynomial fits to the l - shells rates of fe and ar and the mean rates for the cno sequence ( ali et al .
1991 ) , with the functional form : @xmath8 where @xmath9 is the residual nuclear charge of the ion before recombination .
the individual fits are shown in figure 1 , and the values of the fit coefficients are listed in table 1 . since gu s calculations show a different dependence on @xmath9 for different elements within the l - shell , we opted not to use this fit for the other 3rd and 4th row elements .
( in some cases , the behavior within the shell can not be fit with a second order polynomial ) .
however , for the sake of completeness , we did estimate @xmath0@xmath1 rates for these ions using a linear fit to set of data including the fe and ar l - shell rates and the mean rates for the cno sequence ( ali et al .
1991 ) , as shown by the dotted line in figure 1 , with the functional form : @xmath10 since @xmath0@xmath1 dr are not fast for recombination from closed shells over the range of temperatures we consider , we did not include estimated dr for those ions .
note that we did not include temperature dependence for our estimated rate coefficients . to quantify the effects of these missing dr rates we performed a series of photoionization calculations using the best existing data , but with and without these estimates of the @xmath0@xmath1 dr . we did not introduce new high - temperature dr rates , using , instead , the fit to the rates for fe given in arnaud & raymond ( 1992 ) . for the models
, we assumed a slab geometry , solar abundances ( e.g. grevesse & anders 1989 ) , a number density n@xmath11 @xmath12 10@xmath13 @xmath14 and total hydrogen column density of n@xmath11 @xmath12 1 x 20@xmath15 @xmath16 , typical of x - ray absorbers ; note that the predicted ionization structure is insensitive to n@xmath11 , unless the densities exceed 10@xmath17 @xmath14 .
for the spectral energy distribution of the incident continuum radiation , we assumed a power law of the form f @xmath18 @xmath19@xmath20 , where @xmath21 @xmath12 1 for @xmath22 @xmath23 13.6 ev , 1.4 for 13.6 ev @xmath24
@xmath22 @xmath24 600 ev , and 0.8 for @xmath22 @xmath2 600 ev .
we generated the two sets of models over a range in ionization parameter ( number of ionizing photons per nucleon at the illuminated face of the slab ) log u = @xmath251.0 to 1.0 , in increments of 0.05 dex .
although the mean electron temperatures range from 1.8 x 10@xmath6k to 4.5 x 10@xmath26 k , the m - shell fe ions peak between log u @xmath12 @xmath250.5 and 0.5 , or 3 x 10@xmath6k @xmath27 t @xmath27 9.0 x 10@xmath6 k , justifying the energy we used in deriving the rate coefficients . over this temperature range , the dr rates may vary by factors of @xmath27 a few , which is within the uncertainty quoted above .
note that , while t increased slightly when the estimated dr rates are included , the ionic columns for 1st and 2nd row elements remained essentially unchanged .
the predicted ionic column densities for fe viii through fe xvi are shown in figures 2 through 4 . since intrinsic x - ray absorption
is typically characterized by strong oxygen bound - free edges and absorption lines , we also include the o vi , o vii , and o viii columns for comparison .
as shown in figure 2 , the ionization parameters at which the column densities of fe viii , fe ix , and fe x peak have all shifted to higher values of u. in the case of fe x , u has shifted by a factor of @xmath5 1.5 .
more importantly , the range in ionization parameter over which a significant fraction of the total fe exists as fe x ( i.e. , 10% , corresponding to a column density @xmath28 10@xmath29 @xmath16 ) has changed from @xmath30 to @xmath31 .
this suggests that , due to @xmath0n @xmath12 0 dr , relatively low ionization states of fe can contribute significantly to opacity of more highly ionized plasmas than would be predicted using the currently available atomic data .
although the peak shifts are not significant for fe viii and fe ix , the curves show substantially more of these ions at higher values of u. note that , since fe ix has a closed shell ( ar - like ) , there is no low temperature @xmath0@xmath1 process for recombination to fe viii , which is produced primarily by radiative recombination .
the increase in fe viii at higher u is due to the larger fraction of fe ix , which is produced by dr from fe x. the structure in the high ionization parameter tails of fe ix and fe x is the result of high temperature dr , although this effect is swamped for fe x in the models in which @xmath0n @xmath12 0 dr included .
fe xi through fe x iii are shown in figure 3 , and for each ion , the peaks shift to higher ionization by factors of @xmath2 2 , reflecting the large @xmath0n @xmath12 0 rates .
fe xiv through fe xvi are shown in figure 4 , and the inclusion of @xmath0n @xmath12 0 dr in this case _ decreases _ the fractional abundances of these ions over much of the adopted range in @xmath32 , due to the large recombination rates .
nevertheless , @xmath0n @xmath12 0 models predict larger columns of these ions for log u @xmath28 0.7 , as a result of the faster recombination of the l - shell ions . in order to illustrate the effects of uncertainties in the estimated dr rates
, we introduced a `` noise factor '' by using a random number generator with a gaussian distribution , with a standard deviation of 0.3 dex , suggested by the largest error in the polynomial fits ( see figure 1 ) .
furthermore , the dr rates computed by gu ( 2003 ) for l - shell ar vary by @xmath5 @xmath33 0.3 dex over the range in electron temperatures predicted by our models . each time cloudy
was run with this option , the recombination coefficient was multiplied by the random number .
the scale factor was kept constant for a particular calculation , but different random numbers were used for different calculations .
we generated a set of 20 models at each value of u , and determined the standard deviation of predicted column densities . in figure 5
we show the scatter for fe x at three values of u. clearly , uncertainties in the rate coefficients have an measurable effect on the predicted ionic columns .
nevertheless , the trend to peak abundance at higher ionization parameter is still obvious , as demonstrated by the overplot of the fe x columns from the models without the estimated dr .
the results of this simple test indicate that the model predictions of the fe ionization balance depend critically on accurate @xmath0@xmath1 dr rate coefficients .
photoionization models of x - ray absorbers that match spectral features from abundant 2nd row elements such as c , n , o , and ne will overpredict the average fe ionization state if this process is not included .
this can be easily demonstrated using figures 1 3 , by selecting an ionization parameter to give the proper ratio of the o vi , o vii , and o viii column densities and reading off the relative fe ionic columns for the models with and without the estimated dr .
the apparent discrepancy in the ionic states detected in the fe m - shell uta in ngc 3783 ( netzer et al .
2003 ) , compared to those predicted by the absorber models , is likely due to the absence of @xmath0@xmath1 rates . to illustrate this point
, we regenerated the so - called low ionization model used by netzer et al . to fit the soft x - ray absorption and inner shell si and s lines detected in a 900 ks merged _
chandra_/high energy transmission grating spectrum of ngc 3783 ( see , also , kaspi et al .
following netzer et al . , we assumed the sed described in kaspi et al .
( 2001 ) and set log u @xmath12 @xmath250.26 and n@xmath11 @xmath12 10@xmath34 @xmath16 ( note that our conversion from the x - ray ionization parameter used by netzer et al .
to the form of u used in cloudy resulted in a model prediction of slightly higher relative ionization of si and s than they predicted , which may be due to slight differences in the model sed ; in order to reproduce the same columns of si and s we adjusted the value of u downward by 10% ) .
the predicted ionic columns for fe viii through fe xii , without @xmath0n@xmath120 dr , are given in table 2 , and the combined fe xi and fe xii column is @xmath5 3.5 greater than the combination of fe viii and fe ix , while the peak ionization state is fe xi . since the strongest uta features for fe xi and fe xii lie at 16.16 and 15.98 , respectively ( behar et al . 2001 ) , the uta would peak near 16.1 , as described by netzer et al .. also shown are the predictions for the same model with @xmath0n@xmath120 dr included . in this case , the peak ionization state is fe x , and the total fe viii and fe ix column are approximately equal to those of fe xi and fe xii combined . since fe viii and feix produce their strongest features at 16.89 and 16.51 , respectively ( behar et al . ) , and the profile would be dominated by fe x ( 16.34 ) , the predicted uta will peak near 16.4 , as observed .
it should also be noted that including the new l - shell rates for 3rd row elements such as s , si , and mg will alter the predicted ionization balance for those elements as well . as noted by behar & netzer ( 2002 ) , inner shell absorption lines from these ions can be used as effective probes of high - column absorbers .
however care must be taken in using photoionization models to determine ionization states and total column densities of the absorbers using these lines as constraints , at least until a full set of accurate l - shell dr rates have become available .
seyfert galaxies are also strong soft x - ray emission - line sources ( e.g. , sako et al . 2000 ; kinkhabwala et al . 2002 ; turner et al .
. it will be important to understand the connection between this component and forbidden `` coronal '' emission lines such as [ fe x ] @xmath356374 , [ fe xi ] @xmath357892 , [ fe xiv ] @xmath355303 and [ s xii ] @xmath357611 ( e.g. kraemer & crenshaw 2000 ) , especially since the optical / uv spectral images obtained with the _ hubble space telescope_/space telescope imaging spectrograph provide tighter constraints on the spatial extent and kinematic properties of the emission - line gas .
accurate dr rates are crucial for modeling this component and exploring the link between the x - ray and optical emission lines . while the models generated for this letter clearly illustrate the magnitude of the problem , we are left with the question on how best to proceed in the absence of a full set of @xmath0@xmath1 dr rate coefficients .
our monte carlo calculations ( see figure 5 ) provide an estimate of the error in the predicted column densities due to the uncertainties in the rate coefficients .
in fact , our results suggest that comparisons between calculated and observed column densities be given error bars @xmath5 0.3 dex , assuming our rates are included in the calculations , due to the uncertainties in the rates and the unmodeled temperature dependence ( as suggested by the variations in the published l - shell rates over the range in temperature predicted by our models , e.g. , 1.8 x 10@xmath6k @xmath24 t @xmath24 4.5 x 10@xmath26 k ) .
clearly , using rates estimated from other ions is far better than assuming the process does not occur , which is what is done when it is neglected .
hence , we suggest that it is best to adopt these estimated rate coefficients with this stated error . ultimately , new atomic data , in general , and dr rates , in particular , are needed to reliably compare the ionization of third and fourth row elements . | we examine the effects of low - temperature , or @xmath0@xmath1 , dielectronic recombination ( dr ) on the ionization balance of the fe m - shell ( fe ix through fe xvi ) .
since @xmath0@xmath1 rates are not available for these ions , we have derived estimates based on the existing rates for the first four ionization states of the cno sequence and newly calculated rates for l - shell ions of 3rd row elements and fe . for a range of ionization parameter and column density applicable to the intrinsic absorbers detected in _ asca _ , _ chandra _ , and _ xmm - newton _ observations of seyfert galaxies , we generated two grids of photoionization models , with and without dr .
the results show that the ionization parameter at which the population of an fe m - shell ion peaks can increase in some cases by factor @xmath2 2 when these rates are included .
more importantly , there are dramatic changes in the range in ionization parameter over which individual m - shell ions contain significant fractions of the total fe ( e.g. @xmath2 10% ) in the plasma .
these results may explain the mismatch between the range of fe ionization states detected in the x - ray spectra of seyferts , identified by the energies of the m - shell unresolved transition array , and those predicted by photoionization models of the x - ray absorbers that reproduce lines of second and third row elements .
the results suggest that care should be taken in using 3rd and 4th row ions to constrain the physical conditions in photoionized x - ray plasmas until accurate dr rates are available .
this underscores the importance of atomic physics in interpreting astronomical spectroscopy . |
there is a growing need for a complete census of active galactic nuclei ( agns ) .
the seminal discovery that every massive galaxy in the local universe harbors a super - massive black hole ( @xmath11 @xmath12 ) implies that all massive galaxies must have hosted agn activity at some time during the past @xmath13 13 gyrs ( e.g. , rees 1984 ; kormendy & richstone 1995 ) . to trace accurately how and when these black holes grew requires a detailed census of agn activity that will provide , amongst other things , estimates of the efficiency and duty cycle of black - hole growth and any dependencies of nuclear obscuration on agn luminosity , redshift , and host - galaxy type ( e.g. , marconi 2004 ; la franca 2005 ; shankar 2007 ) . the exceptional sensitivity of the _ chandra _ deep field observations ( e.g. , brandt et al .
2001 ; giacconi et al . 2002
; alexander et al .
2003a ) has helped to unveil a @xmath14 times larger population of agns than found at most other wavelengths ( @xmath13 7200 deg@xmath15 ; e.g. , bauer et al .
optical spectroscopic follow - up observations have shown that these agns are detected out to @xmath6 5 , and detailed x - ray spectral analyses have revealed that the majority of the sources are obscured by gas and dust ( e.g. , barger et al .
2003 ; szokoly et al . 2004
; tozzi et al . 2006 ; see brandt & hasinger 2005 for a review ) .
however , although clearly effective at identifying even heavily obscured agns out to high redshift , there is compelling evidence that a large fraction of the agn population remains undetected at the @xmath16 kev observed - frame energies probed by these surveys : ( 1 ) about half of the x - ray background ( xrb ) is unresolved at @xmath17 kev ( worsley et al . 2005 ) , ( 2 ) the _ observed _ obscured : unobscured agn ratio is lower than that found for comparably luminous agns in the local universe ( e.g. , treister & urry 2005 ) , ( 3 ) the blank - field number counts are steeply rising at the faintest x - ray fluxes ( rosati et al .
2002 ) , and ( 4 ) few compton - thick agns ( @xmath18 @xmath5 ) have been identified , even though they comprise @xmath13 50% of the agn population in the local universe ( e.g. , risaliti et al .
1999 ; guainazzi et al . 2005
; tozzi et al .
2006 ) .
many of the x - ray undetected agns are expected to be intrinsically luminous sources that are heavily obscured by compton - thick material ( i.e. , @xmath19 @xmath5 ; see comastri 2004 for a review ) .
the most robust identification of a compton - thick agn is made from high signal - to - noise ratio ( s / n ) x - ray spectroscopy , where the detection of a high equivalent - width fe k emission line ( @xmath20 1 kev at rest - frame energies 6.46.9 kev ) and a steeply rising reflection component at @xmath14 kev reveals that little or no x - ray emission is being seen directly , implying that the central source is very heavily absorbed ( e.g. , george & fabian 1991 ; matt , brandt , & fabian 1996 ; maiolino et al . 1998 ; matt et al . 2000 ) . however , since the extreme obscuration toward the nucleus of a compton - thick agn renders the observed @xmath16 kev emission @xmath13
301000 times fainter than the intrinsic ( i.e. , unabsorbed ) emission , it is often challenging to identify robustly these sources on the basis of their x - ray data alone ; the range of absorption correction factors is based on the observed to intrinsic x - ray luminosity ratio for the agns in table 8.1 of comastri ( 2004 ) , with the observed x - ray fluxes from bassani et al .
for example , despite compton - thick agns comprising half of the agn population in the local universe , only @xmath13 50 sources have been reliably identified from x - ray data to date ( e.g. , comastri 2004 ) .
fortunately , although not as conclusive as high s / n x - ray spectroscopy , the presence of compton - thick absorption can also be identified in x - ray weak agns when a reliable measurement of the intrinsic power of the agn is available at other wavelengths ; we stress here the necessity for x - ray constraints in order to identify a compton - thick agn since it is only in the x - ray band that an absorbing column density can be measured or inferred .
two of the most reliable measurements of the intrinsic power of an obscured agn are the luminosity of the mid - infrared ( mid - ir ; rest - frame @xmath21 @xmath0 m ; e.g. , efstathiou & rowan - robinson 1995 ; granato et al .
1997 ; lutz et al . 2004 ) continuum and high - excitation emission lines ( i.e. , [ oiii]@xmath225007 ; e.g. , allen 1973 ; kwan & krolik 1981 ; bassani et al .
1999 ) , both of which are believed to be directly illuminated by the central source .
these emission regions provide a good proxy for the intrinsic agn luminosity , even in the presence of extreme compton - thick absorption , since they are more extended than the x - ray absorbing material ( i.e. , larger than the broad - line region ; e.g. , lamer et al .
2003 ; risaliti et al .
2007 ) . as the luminosities of both the mid - ir continuum and the high - excitation emission - line region are dependent on many factors , including the location and geometry of the region with respect to the central source , more robust constraints are placed if both measurements are available .
comprehensive evidence for a large x - ray undetected agn population has been found from a variety of x - ray based analyses using sources detected in deep _ spitzer _
surveys at mid - ir wavelengths , suggesting that compton - thick agns may be ubiquitious in the distant universe ( e.g. , donley et al .
2005 , 2007 ; alonso - herrero et al .
2006 ; polletta et al .
2006 ; daddi et al .
2007a ; steffen et al .
2007 ; martnez - sansigre et al .
2007 ; fiore et al . 2008 ) .
however , all of these studies have relied on photometric observations for the identification of candidate compton - thick agns , leading to significant uncertainties in measurements of the intrinsic agn luminosity , the absorbing column density , the degree of contamination from star - formation activity , and by implication , the space density of distant compton - thick agns . in this paper
we present _ spitzer_-irs spectroscopy and 3.670 @xmath0 m observations of an optically identified narrow emission - line agn at @xmath1 ( hdf - omd49 ; steidel et al .
2002 ) that is undetected in the published catalogs of the deepest x - ray observation currently available [ the 2 ms _ chandra _ deep field - north ( cdf - n ) ; alexander et al .
2003a ] . from a comparison of the agn energetics at x - ray wavelengths to those determined from optical and mid - ir spectroscopy
, we robustly show that hdf - omd49 hosts an intrinsically luminous ( @xmath23 erg s@xmath3 ) compton - thick agn ; i.e. , this source is a compton - thick quasar .
erg s@xmath3 .
this definition is consistent with that derived from the classical optical quasar threshold of @xmath24 ( e.g. , schmidt & green 1983 ) , assuming the @xmath25 relationship of steffen ( 2006 ) and the composite quasar spectrum of vanden berk ( 2001 ) ; this definition is also consistent with @xmath26 derived for x - ray unobscured agns ( see table 5 of hasinger 2005 ) . ]
we then use these x - ray
optical mid - ir diagnostics to identify a further six distant compton - thick quasars in the literature and place limits on the space density of compton - thick quasars at @xmath6 2 .
we adopted @xmath27 km s@xmath3 mpc@xmath3 , @xmath28 , and @xmath29 throughout .
hdf - omd49 ( optical position @xmath30
12@xmath31 37@xmath32 0434 , @xmath33 @xmath3414@xmath35 ; steidel et al .
2002 ) was originally identified in the lyman - break galaxy ( lbg ) searches of steidel ( 2002 ; 2003 ) . with a redshift of @xmath1 ,
hdf - omd49 lies at the low - redshift end of the lbg population ( median redshift @xmath6 3 ) .
hdf - omd49 is also identified as a bx / bm galaxy , based on the criteria of steidel ( 2004 ) , and it is dubbed bm1156 in the bx / bm study of reddy ( 2006 ) . the apparent contradiction in hdf - omd49
being selected as both an lbg and bx / bm is due to small differences in the photometric data used in these studies .
hdf - omd49 is detected by irac at 3.68.0 @xmath0 m and mips at 24 @xmath0 m in the _ spitzer _ observations obtained as part of the great observatories origins deep survey ( goods ) legacy project ( p.i .
m. dickinson ; r. chary et al .
, in prep . )
; the 3.6 @xmath0 m , 4.5 @xmath0 m , 5.8 @xmath0 m , 8.0 @xmath0 m , and 24 @xmath0 m fluxes are 6.1 @xmath0jy , 7.0 @xmath0jy , 14.8 @xmath0jy , 40.1 @xmath0jy , and 380 @xmath0jy , respectively .
although hdf - omd49 is bright at 24 @xmath0 m , it is undetected in the ultra - deep 70 @xmath0 m observations of frayer et al .
( 2006 ; 3 @xmath36 upper limit of @xmath37 mjy ) , indicating that it has a `` warm '' mid - ir spectral energy distribution ( sed ; e.g. , papovich et al .
hdf - omd49 is also detected at 16 @xmath0 m with @xmath38 @xmath0jy using the `` peak - up '' imaging capability of _
spitzer_-irs ( h. teplitz et al . , in prep
the _ spitzer_-irs spectroscopy of hdf - omd49 was obtained in spectral staring mode as part of program go2 - 20456 ( p.i .
r. chary ) using the long - low modules ( ll ; houck et al . 2004 ) .
hdf - omd49 was observed with ll1 ( 20.537.9 @xmath0 m ) for 3 hrs , and additional observations with ll1 ( 0.93 hrs ) and ll2 ( 1.33 hrs ; 14.321.2 @xmath0 m ) were obtained when hdf - omd49 fell serendipitously into the slit of another target from the same program .
we analyzed the 2d basic calibrated data files from the s14.0.0 _
spitzer_-irs pipeline data ; data reduction includes cleaning rogue pixels , fitting and removing the latent charge build - up , subtracting the sky , and averaging the 2d files together .
spectral extraction was performed using a 2 pixel wide window in spice and the data were calibrated using the same extraction window on a standard - star spectrum .
the average rms of the ll1 and ll2 spectra are 97 @xmath0jy and 120 @xmath0jy , respectively .
the _ spitzer_-irs fluxes measured through the 16 @xmath0 m and 24 @xmath0 m filter spectral response curves are @xmath39 @xmath0jy and @xmath40 @xmath0jy , respectively , consistent with the observed fluxes in these bands . for more details on the observations and data
reduction see pope et al .
( 2008 ) .
hdf - omd49 is undetected in the published 2 ms _ chandra _ catalogs of alexander et al .
( 2003a ) .
however , these catalogs were produced using a conservative source - detection algorithm to minimize the number of spurious detections , and it is possible to identify robustly fainter sources when searching for x - ray emission associated with a known source ( e.g. , see 3.4.2 of alexander et al .
2003a and 5 of alexander et al . 2001 ) . indeed , adopting a wavdetect ( freeman et al .
2002 ) false - positive probability threshold of @xmath41 we detected significant x - ray emission ( x - ray position @xmath30
12@xmath31 37@xmath32 0432 , @xmath33
@xmath3414@xmath42 ) within 0.25 arcsec of the optical position of hdf - omd49 in the 0.58 kev , and 48 kev bands with @xmath43 , @xmath44 , and @xmath45 counts , respectively ( see also laird et al .
hdf - omd49 remains undetected in all of the other four x - ray bands explored by alexander et al .
( 2003a ) and we calculated 3 @xmath36 upper limits following 3.4.1 of alexander et al .
( 2003a ) .
although the s / n of the x - ray data is low , the low background of _ chandra _ allows sources with very low count rates to be reliably detected .
for example , with a measured background in the detection aperture of hdf - omd49 ( @xmath13 2 pixel radius ) of only @xmath13 1 counts in the 0.52 kev band , the probability of detecting @xmath13 7 counts by chance is @xmath46 ; furthermore , hdf - omd49 is detected in two independent bands , increasing the overall probability that the detection is real .
the x - ray counts correspond to fluxes of @xmath47 erg s@xmath3 @xmath5 ( 0.58 kev ) , @xmath48 erg s@xmath3 @xmath5 ( 0.52 kev ) , and ( 48 kev ) , for @xmath49 ( i.e. , the spectral slope of the xrb ) . the detection in the 48 kev band ( rest - frame 12.825.7 kev ) is particularly notable since _ chandra _ is significantly less sensitive at than at energies , suggesting that hdf - omd49 has a flat x - ray spectral slope ; the non detection in the wider band provides support for this hypothesis .
the x - ray band ratio ( the 28 kev/0.52 kev count - rate ratio ) is @xmath50 , which corresponds to @xmath51 and is consistent with the x - ray spectral slope derived from the 48 kev/0.52 kev count - rate ratio ( @xmath52 0.4 ) ; the difference in the x - ray fluxes estimated using @xmath53 instead of @xmath49 is negligible given the low s /
n of the x - ray data ( @xmath54 15% ) .
the corresponding rest - frame 1.66.4 kev , 6.412.8 kev , and 12.825.7 kev luminosities ( or 3 @xmath36 upper limit ) of hdf - omd49 are @xmath55 erg s@xmath3 , @xmath56 erg s@xmath3 , and @xmath57 erg s@xmath3 , respectively .
lcc|ccccccc|cc|l smm j123600 + 621047@xmath58 & 2.002 & 15750 & @xmath59 & 45.3@xmath60 & & & & & & 44.7 & & 1 , 2 + hdf - omd49@xmath58 & 2.211 & 17800 & 42.1 & 45.2@xmath60 & 42.5 & 42.4 & 41.9 & 41.7 & & 44.6 & 44.4 & 3 , 4 + fsc 10214 + 4724@xmath61 & 2.285 & 18550 & 42.3 & 44.9@xmath60 & & & & & 42.6 & 44.3 & 44.4 & 5 , 6 , 7 + sw j104406 + 583954 & 2.430 & 19990 & @xmath62 & 45.8@xmath60 & 43.9 & 43.5 & 42.9 & & & 45.2 & 45.6 & 8 , 9 + bx160@xmath58 & 2.462 & 20290 & @xmath63 & 44.9@xmath64 & 42.3 & 41.8 & & & & 44.3 & 44.1 & 3 , 4 + bx1637@xmath58 & 2.487 & 20570 & @xmath65 & 44.8@xmath64 & 42.3 & & & & & 44.3 & 44.1 & 3 , 4 + sw j104409 + 585224 & 2.540 & 21100 & @xmath62 & 46.4@xmath60 & 43.8 & 43.1 & & & & 45.9 & 45.5 & 8 , 10
in fig . 1a we show the _ spitzer_-irs spectrum of hdf - omd49 .
the spectrum is noisy , particularly in the ll2 order . however , there is no significant emission from the dominant rest - frame 7.7 @xmath0 m polycyclic aromatic hydrocarbon ( pah ) feature found in star - forming galaxies , indicating that the mid - ir spectrum does not have a strong contribution from star - formation activity .
by contrast , the _ spitzer_-irs spectrum of hdf - omd49 is similar to the nearby compton - thick agn ngc 1068 and even shows weak evidence for si absorption at 9.7 @xmath0 m , a feature often seen in obscured agns ( e.g. , shi et al .
2006 ; hao et al .
2007 ; spoon et al .
the _ spitzer_-irs spectrum of hdf - omd49 is also qualitatively similar to the composite mid - ir spectrum of the x - ray obscured quasars ( @xmath66 erg s@xmath3 ; @xmath67 @xmath5 ) investigated by sturm et al .
( 2006 ) ; see fig
. 1b .
m. the error bar in the @xmath68-axis direction indicates the width that corresponds to @xmath69% of the maximum transmission in each of the photometric bands ; see footnote 3 .
the data are compared to the seds of the compton - thick agn ngc 1068 ( dashed curve ) and the starburst galaxy m82 ( long - dashed curve ) , which are shifted to @xmath1 and normalized based on the results from fitting the mid - ir spectrum of hdf - omd49 ( see 3.1 ) . in order to fit the data in the irac bands it was necessary to also add a black - body component of very hot 1000 k dust ( dotted curve ) .
the thin solid curve shows the combined model results . the data for the ngc 1068 and m82 seds were taken from rigopoulou et al .
( 1999 ) , chary & elbaz ( 2001 ) , galliano et al .
( 2003 ) , and magnelli et al .
( 2007).,width=321 ] we can quantify the contributions from agn and star - formation activity to the mid - ir luminosity of hdf - omd49 by fitting the _
spitzer_-irs spectrum with agn and star - formation templates .
adopting the approach of pope et al .
( 2008 ) , we fit the _
spitzer_-irs spectrum with an m82 starburst component and include an agn component using the nuclear spectrum of ngc 1068 , performing a @xmath70 minimization to obtain the best - fitting normalization . on the basis of this approach , we find that the mid - ir spectrum of hdf - omd49 is dominated by agn activity , with a limit of a 10% contribution from star formation in the _
spitzer_-irs band ; there is no clear signature for star formation and this limit effectively corresponds to the noise per pixel in the spectra .
we get consistent results if we fit the mid - ir spectrum with an m82 starburst component and an absorbed power - law component ( to represent the agn ) , leaving the slope of the power law and the extinction as free parameters .
the best - fitting results with this second approach give a spectral slope of @xmath71 2 ( consistent with that found for agn - dominated sources ; e.g. , alonso - herrero et al .
2003 , where @xmath72 ) with @xmath73 0.65 and a @xmath13 5% limit on the contribution from star formation . in fig .
2 we show the 3.670 @xmath0 m sed of hdf - omd49 and plot on the ngc 1068 and m82 templates , normalized to the values obtained from the mid - ir spectral fitting .
the strong observed - frame 3.68 @xmath0 m emission and the power - law - like featureless sed strengthens our conclusion from the mid - ir spectral fitting , showing that the ir sed of hdf - omd49 is dominated by agn activity .
the strongly rising continuum at @xmath74 @xmath0 m ( rest - frame @xmath75 @xmath0 m ) also indicates that the contribution from starlight ( which typically peaks at rest - frame 1.6 @xmath0 m ) is weak at these wavelengths .
the combination of the ngc 1068 and m82 templates gives a respectable fit to the overall sed of hdf - omd49 even though the wavelength coverage of the sed data is three times broader than that used when fitting the mid - ir spectrum ( rest - frame 1.121.8 @xmath0 m versus rest - frame 4.410.6 @xmath0 m ) .
however , the combination of the ngc 1068 and m82 components lies below the observed emission in the irac bands ( rest - frame @xmath76 @xmath0 m ) .
this excess emission can be fitted by adding a black - body component of very hot dust ( @xmath13 1000 k ) , as often found in obscured agns ( e.g. , alonso - herrero et al .
2001 ) . on the basis of the fitted ngc 1068 component ,
the rest - frame 6 @xmath0 m luminosity from hdf - omd49 is @xmath77@xmath78 erg s@xmath3 ( @xmath79 @xmath0jy ) , indicating that the agn is powerful ( e.g. , compare to the nearby agns explored by lutz et al .
2004 ) .
our analyses show that the ir sed of hdf - omd49 is dominated by agn activity and allow for just a small contribution from star formation . using the best fitting star - formation component normalizations and the relationship between pah luminosity and star - formation rate ( sfr ) given by pope et al .
( 2008 ; see eqn .
7 ) , we predict sfr @xmath13 60120 @xmath12 yr@xmath3 , with the range representing the two different star - formation contributions we estimated from the mid - ir spectral fitting .
the radio detection of hdf - omd49 with @xmath80 32 @xmath0jy ( g. morrison et al . , in prep . ) implies a sfr of @xmath13 500 @xmath12 yr@xmath3 ( calculated following bauer et al .
this is clearly inconsistent with the mid - ir data , suggesting that either the radio emission is dominated by the agn or the radio - sfr relationship is not applicable for high - redshift objects such as hdf - omd49 .
the contrast between the weak x - ray emission ( see 2.2 ) and the bright agn - dominated ir sed implies that the agn in hdf - omd49 is heavily absorbed at x - ray energies .
the signature of absorption should be evident in the x - ray spectrum of hdf - omd49 .
however , before examining the data it is necessary to consider the contribution to the x - ray emission from star formation . assuming a sfr of 120 @xmath12 yr@xmath3 ( see 3.1 ) and the empirically determined @xmath81sfr relationship found by bauer et al .
( 2002 ) , we predict a rest - frame 0.58 kev flux of @xmath82 erg s@xmath3 @xmath5 , which corresponds to an observed - frame 0.52 kev flux of @xmath83 erg s@xmath3 @xmath5 ( @xmath84 erg s@xmath3 ) , for a typical x - ray spectral slope for star - forming galaxies of @xmath85 .
consistent x - ray luminosities are obtained from the empricially derived ranalli ( 2003 ) relationship , although the persic ( 2004 ) relationship would predict an @xmath13 4 times lower x - ray luminosity but only takes into account the contribution from high - mass x - ray binaries .
these analyses show that the contribution from star formation could be significant to the observed - frame 0.52 kev flux ( @xmath13 40% ) .
however , the contribution will be negligible in the other x - ray band in which hdf - omd49 is detected ( observed - frame 48 kev ) . in fig .
3 we show the x - ray data of hdf - omd49 and compare them to the agn model seds adopted by gilli et al .
( 2007 ) to interpret the xrb .
the x - ray data of hdf - omd49 clearly indicate that the agn is absorbed at x - ray energies ; on the basis of the x - ray band ratio ( see 2.2 ) , the x - ray data are consistent with @xmath86 @xmath5 ( e.g.,see fig . 4 of alexander 2003b ) .
however , with our low s / n x - ray data we can not unambiguously distinguish between compton - thick and compton - thin absorption . furthermore , the x - ray spectrum of a compton - thick agn is source specific and depends upon the strength of the reflected and scattered x - ray components , in addition to the underlying power - law continuum and any additional absorption ; for example , see appendix a of guainazzi et al .
( 2005 ) for information on the difficulty of identifying compton - thick agns on the basis of x - ray colors alone .
therefore , without detailed x - ray spectral signatures that clearly indicate the presence of compton - thick absorption ( e.g.,the identification of a prominent fe k emission line or a strong reflection component ) , it is not possible to argue _ solely _ on the basis of low s / n x - ray data that an agn is absorbed by compton - thick material . in order to confirm whether such an x - ray weak source is compton thick ,
it is necessary to estimate the intrinsic agn luminosity . indicated ) and normalized to the observed - frame 48 kev flux .
the open circle shows the x - ray emission at observed - frame 0.52 kev corrected for the potential contribution from star formation ; see 3.2 .
the x - ray data of hdf - omd49 are consistent with those expected for a heavily absorbed agn , however , the low s / n x - ray spectrum alone provides limited diagnostics to distinguish between compton - thin and compton - thick absorption .
we stress that since the data in this figure is normalised to the observed 48 kev flux it does not demonstrate the @xmath87 order of magnitude difference in flux between similarly powerful compton - thin and compton - thick agns ; see fig .
4.,width=321 ] a good proxy for the intrinsic luminosity of agn activity is the rest - frame mid - ir emission , which can provide an absorption - independent measure of the intrinsic luminosity even in the presence of compton - thick absorption ( e.g. , krabbe 2001 ; lutz 2004 ; maiolino 2007 ; horst 2008 ) . when combined with the observed x - ray emission
, the rest - frame mid - ir luminosity can therefore provide a diagnostic for the presence of heavy absorption . in this study
we adopt the x - ray6 @xmath0 m luminosity relationship found by lutz et al .
( 2004 ) , which was determined by constraining the agn continuum component in large - aperture mid - ir spectra for a sample of nearby agns .
we favour the lutz ( 2004 ) x - ray mid - ir luminosity relationship over the small - aperture mid - ir imaging studies of krabbe ( 2001 ) and horst ( 2008 ) since hdf - omd49 is unresolved by _
spitzer_. furthermore , the lutz ( 2004 ) study covers a broader range in luminosity , and includes agns as luminous as those explored here .
indeed , we note that the x - ray mid - ir luminosity ratio found for x - ray obscured but compton - thin quasars ( sturm et al .
2006 ) is consistent with the lutz et al .
( 2004 ) relationship , indicating that it is applicable for the objects explored in our study ; see fig . 4 .
however , the differences in the x - ray mid - ir luminosity of these studies are small [ the lutz ( 2004 ) luminosity ratio is within a factor of @xmath13 23 of that found by krabbe ( 2001 ) and horst ( 2008 ) , assuming typical seds ( i.e. , ngc 1068 ; mrk 231 ; m82 ) when converting to a common rest - frame wavelength of 6 @xmath0 m ] . since it is now well established that the x - ray - to - optical luminosity ratio of optically selected agns is luminosity dependent ( e.g.,vignali 2003 ; steffen 2006 ) , it is also interesting to explore whether the x - ray mid - ir luminosity ratio is luminosity dependent . krabbe ( 2001 ) , lutz ( 2004 ) , and horst ( 2008 ) did not find evidence for a luminosity dependent x - ray mid - ir luminosity ratio for local agns over @xmath13 35 orders of magnitude in luminosity . by comparison , the inference from the study of distant optically selected quasars by maiolino ( 2007 ) is that the x - ray mid - ir luminosity ratio _ may _ be luminosity dependent . for example , solving eqn . 1 and eqn .
5 in maiolino ( 2007 ) gives an x - ray mid - ir luminosity relationship of @xmath81 @xmath88 @xmath89@xmath90 .
however , there is considerable uncertainty in this result , and the derived x - ray mid - ir luminosity relationship would be linear if a different x - ray optical luminosity ratio from steffen ( 2006 ) is used ( i.e. , if eqn .
1b is adopted instead of eqn .
in any case , the differences between the lutz ( 2004 ) and maiolino ( 2007 ) relationships are small ( factor @xmath91 23 ) for the luminosities of the majority of the agns explored here ( @xmath92 erg s@xmath3 ; see fig .
4 ) ; however , we indicate the cases where our results would significantly change if the luminosity dependent version of the maiolino ( 2007 ) x - ray mid - ir luminosity relationship is used . in fig .
4 we show the rest - frame 210 kev luminosity versus 6 @xmath0 m continuum luminosity of hdf - omd49 and compare it to those found for agn in the local universe ; we converted the rest - frame 6.7 @xmath0 m data used by maiolino ( 2007 ) to rest - frame 6 @xmath0 m assuming the sed of ngc 1068 , which results in small corrections ( @xmath13 10% ) . the rest - frame 210 kev luminosity of hdf - omd49 ( @xmath93 erg s@xmath3 )
is estimated from the observed - frame 0.52 kev luminosity ( rest - frame 1.66.4 kev ) , for a spectral slope of @xmath49 .
assuming both the x - ray mid - ir luminosity relationship found for local agns ( lutz et al .
2004 ) , and the relationship derived from maiolino et al .
( 2007 ) , the intrinsic rest - frame luminosity of hdf - omd49 is @xmath94 erg s@xmath3 , suggesting that the x - ray absorption toward the agn in hdf - omd49 is compton thick ; see fig . 4 and table 1 . however , to reduce further uncertainties on the x - ray ir luminosity relationship due to , for example , the geometry and covering factor of the gas and dust surrounding the x - ray emitting region , it is useful to have other measurements of the intrinsic agn luminosity .
the rest - frame uv spectrum of hdf - omd49 shows strong agn emission lines ( civ , ciii ] , heii , ly@xmath95 ; see fig . 2 of steidel et al .
2002 ) , which can be used to provide additional estimates of the intrinsic agn luminosity .
we calculated the emission - line fluxes using the steidel ( 2002 ) optical spectrum and integrating the flux over the emission - line profiles ; see reddy et al .
( 2006 ) for more details . adopting the average emission - line ratios of netzer et al .
( 2006 ) , to calculate the expected [ oiii]@xmath225007 luminosity , and the [ oiii]@xmath225007x - ray flux ratio found by mulchaey et al .
( 1994 ) , the range of predicted intrinsic rest - frame luminosity from the different emission lines is @xmath96 erg s@xmath3 .
we get similar intrinsic x - ray luminosities if we use the [ oiii]@xmath225007x - ray flux ratio found by alonso - herrero ( 1997 ) and we get higher intrinsic x - ray luminosities ( up to an order of magnitude higher for the most luminous agns ) if we use the netzer ( 2006 ) [ oiii]@xmath225007x - ray flux ratio .
we have not corrected these emission - line fluxes for extinction or contamination from star formation , which are poorly constrained at rest - frame uv wavelengths .
however , we note that the mulchaey et al .
( 1994 ) , alonso - herrero ( 1997 ) , and netzer ( 2006 ) correlations were also derived using uncorrected emission - line fluxes . the uncertainties in both the emission - line ratios and
the [ oiii]@xmath225007x - ray flux ratio are @xmath13 2 ( see table 1 of netzer 2006 and table 2 of mulchaey 1994 ) , giving a combined uncertainty of a factor of @xmath13 3 , comparable to the uncertainty in the intrinsic x - ray luminosity based on the rest - frame 6 @xmath0 m luminosity .
the intrinsic rest - fame 210 kev luminosities predicted from both the optical and the mid - ir spectroscopy are in good agreement , providing compelling evidence for compton - thick absorption of an intrinsically luminous quasar in hdf - omd49 ( @xmath2 erg s@xmath3 ; @xmath97
@xmath5 ) .
an alternative scenario for the weak x - ray emission from hdf - omd49 is that the agn has recently faded ( e.g. , guainazzi 1998 ) .
however , since the light - crossing time of the mid - ir emitting region is likely to be short ( i.e. , on the basis of the dust - sublimation radius it will be @xmath13 12 yrs ; e.g. , see table 1 of granato , danese , & franceschini 1997 and fig . 30 of suganuma 2006 ) , this scenario is only tenable for hdf - omd49 if we have caught the agn during a very brief transitionary stage where the x - ray emission has faded but the mid - ir emission is still strong .
the good agreement between the fluxes measured from the _ spitzer_-mips photometry and the _ spitzer_-irs spectroscopy , taken 2 years apart , suggests that this scenario is unlikely .
we have used _ spitzer_-irs spectroscopy , 3.670 @xmath0 m photometry , optical spectroscopy , and ultra - deep 2 ms _ chandra _ data to show that hdf - omd49 hosts an intrinsically luminous agn that is heavily obscured by compton - thick material ( @xmath2 erg s@xmath3 ; @xmath97 @xmath5 ) ; i.e. , it is a compton - thick quasar . the s / n of the x - ray data alone was too poor to show conclusively that the agn is obscured by compton - thick material . however , when combined with rest - frame 6 @xmath0 m and optical emission - line luminosity constraints , the evidence for compton - thick absorption in this x - ray faint source is compelling .
additional support for this interpretation comes from the similarity of the mid - ir spectrum of hdf - omd49 and the composite mid - ir spectrum of x - ray obscured but compton - thin quasars ; see fig .
1b . on the basis of the comparatively weak ultra - hard kev emission from hdf - omd49 , even the rest - frame kev emission appears to be heavily suppressed .
for example , assuming an x - ray spectral slope of @xmath98 , the predicted kev luminosity based on the intrinsic rest - frame 210 kev luminosity is a factor of @xmath13 20 higher than the observed 12.825.7 kev luminosity ( @xmath99 erg s@xmath3 versus @xmath100 erg s@xmath3 ; see 2.2 ) .
it is difficult to relate this factor of @xmath13 20 suppression to an accurate estimate of the absorbing column density since the observed x - ray emission in the 12.825.7 kev band could be dominated by reflected and scattered components ( e.g. , ngc 1068 ; matt 1997 ; see fig .
8.1 in comastri 2004 ) .
however , on the basis of the plcabs model ( yaqoob 1997 ) in xspec , which considers the effects of electron scattering and photoelectric absorption through an absorbing medium , this amount of suppression in the 12.825.7 kev band would correspond to an absorbing column density of @xmath101 @xmath5 .
due to the absence of a reflection component in this model , and the simplifying assumption of a spherical geometry , this absorbing column density should only be considered representative ( e.g. , see fig .
8.1 in comastri 2004 for an alternative model ) . in this final section
we adopt the approach that we used in 3 to identify other distant compton - thick agns using data obtained from the literature . utilizing these constraints
we then directly estimate the space density of compton - thick quasars at @xmath6 2 and compare our results to current observational and theoretical constraints . from a search of
the published _
spitzer_-irs spectroscopic data , there are four good candidate
@xmath6 2 compton - thick agns that have mid - ir spectroscopy and the _ essential _ sensitive x - ray constraints required to identify compton - thick agn activity : fsc 10214 + 4714 , sw j104406 + 583954 , sw j104409 + 585224 , and smm j123600 + 621047 .
fsc 10214 + 4714 is a strongly lensed seyfert 2 galaxy at @xmath102 which produces luminous 6 @xmath0 m agn continuum emission ( teplitz et al .
2006 ) but is only weakly detected at x - ray energies , despite an effective lensing - corrected _ chandra _ exposure of @xmath13 2 ms ( alexander et al .
2005a ) . in fig . 4 we plot fsc 10214 and
provide secondary support for the intrinsic luminosity of the agn using the [ oiii]@xmath225007 luminosity from alexander et al .
( 2005a ) and the [ oiii]@xmath225007x - ray flux ratio of mulchaey et al .
( 1994 ) ; see table 1 .
the combination of the x - ray data and optical mid - ir spectroscopy confirms that fsc 10214 hosts a compton - thick quasar ( @xmath103 erg s@xmath3 ) , even given the considerable scatter in the relationships .
sw j104406 + 583954 and sw j104409 + 585224 are x - ray weak candidate compton - thick agns identified in the _ chandra_/swire survey ( polletta 2006 ) that lie at at @xmath104 and @xmath105 , respectively .
the optical and mid - ir spectra of both sources show clear signatures of agn activity ( polletta 2006 , 2008 ; weedman 2006 ) , which we use to determine the intrinsic luminosities of the central sources ; see table 1 . in fig .
4 we plot their x - ray6 @xmath0 m luminosities , showing that the properties of both objects are consistent with those expected for a luminous compton - thick quasar ( @xmath106 erg s@xmath3 ) ; however , we note that the evidence is weaker for sw 104406 if the luminosity dependent x - ray - to - mid - ir luminosity relationship inferred from maiolino et al .
( 2007 ) is used .
smm j123600 + 621047 is a @xmath107 submillimeter emitting galaxy ( smg ) with a mid - ir - bright agn ( pope et al .
2008 ) that is undetected in the 2 ms cdf - n data ( alexander et al .
2003a ) ; we ran wavdetect using a false - positive probability threshold of @xmath41 but did not detect x - ray emission from this source ( see table 1 of alexander 2005b ) .
the x - ray6 @xmath0 m luminosity ratio indicates that it hosts a compton - thick agn ( see fig . 4 ) , although optical and near - ir spectroscopic observations do not reveal the signatures of agn activity ( e.g. , swinbank et al .
2004 ; chapman et al .
2005 ) , leaving some uncertainty on the intrinsic luminosity of the agn .
however , since the mid - ir luminosity of the agn in smmj 123600 is about an order of magnitude larger than that found in typical smgs ( e.g. , lutz et al .
2005 ; menndez - delmestre et al . 2007 ; valiante et al . 2007 ; pope et al . 2008 ) , the @xmath13 10 times larger intrinsic x - ray luminosity derived here seems plausible ( @xmath108 erg s@xmath3 ; i.e. , compare to alexander et al .
2005b ) ; see table 1 . in alexander ( 2005c ) , it was argued that smgs represent a rapid black - hole growth phase before the onset of optically luminous quasar activity .
the discovery of this one mid - ir luminous x - ray undetected smg in the cdf - n field increases current constraints on the _ integrated _ black - hole growth density in smgs by a factor of @xmath13 2 ( see fig . 2 of alexander et al .
2005c ) .
other studies have used optical spectroscopy to identify agns that are mid - ir bright and x - ray weak / undetected but lack _ spitzer_-irs spectroscopy .
reddy et al .
( 2006 ) found two @xmath6 2.42.5 optically identified agns ( bx160 ; bx1637 ) that are x - ray undetected in the 2 ms cdf - n survey and have power - law like ir seds [ the 24 @xmath0 m fluxes of these sources ( @xmath109 100140 @xmath0jy ) are too faint for good s / n _ spitzer_-irs spectroscopy ] ; see table 1 .
following our approach for hdf - omd49 ( see 2.2 ) , we ran wavdetect using a false - positive probability threshold of @xmath41 but did not detect x - ray emission from either source ; 3 @xmath36 x - ray upper limits are calculated following 3.4.1 in alexander et al .
( 2003a ) .
we plot the x - ray - to-6 @xmath0 m luminosity ratios of these objects in fig .
4 . the mid - ir data , when combined with the optical emission - line luminosities
, indicates that these sources are likely to be compton - thick quasars ; we determined their emission - line properties using the same procedure as for hdf - omd49 ( see 3.3 ) .
the evidence would be weaker for bx1637 if the luminosity dependent x - ray mid - ir luminosity ratio inferrred from maiolino et al .
( 2007 ) is used , although we note that the x - ray6 @xmath0 m luminosity ratio for this source is an upper limit .
we also note that other studies have selected candidate compton - thick agns using a combination of multi - wavelength data with sensitive x - ray constraints ( e.g. , donley et al .
2005 , 2007 ; alonso - herrero 2006 ; daddi et al . 2007a ; martnez - sansigre et al .
2007 ; fiore 2008 ) ; many other studies have identified obscured agns using mid - ir spectroscopy but lack the _ essential _ x - ray data to provide a case for compton - thick absorption .
however , none of these studies has optical or mid - ir spectroscopy to provide accurate measurements of the intrinsic agn luminosities .
m luminosity and the average emission - line luminosity given in table 1 .
although the four compton - thick agns that lie in the goods - n region are either x - ray weak or undetected , they are amongst the most intrinsically luminous agns in the goods - n field.,width=321 ] the intrinsic x - ray mid - ir continuum ratios , inferred using the rest - frame uv emission - line luminosities of the six compton - thick quasars with optical agn signatures , are consistent with those found for agns in the local universe ; see fig .
this is in agreement with the results of sturm ( 2006 ) , who found that the _ average _ intrinsic x - ray6 @xmath0 m luminosity of distant x - ray obscured quasars are similar to those found for local agns , suggesting that the objects explored here are the extreme compton - thick end of the obscured quasar population .
we also showed in 3.1 and fig .
1b that the _ spitzer - irs _ spectrum of hdf - omd49 is similar to the composite mid - ir spectrum produced by sturm ( 2006 ) , providing a particularly strong case for a quasar that is absorbed in the x - ray band by compton - thick material in this source .
these pieces of evidence also suggest that the accretion properties of distant optically identifiable agns are similar to those observed in local agns , in agreement with previous studies that have explored the x - ray optical and x - ray infrared properties of optically selected agns ( e.g.,steffen et al .
2006 ; jiang et al .
2006 ) . in fig .
5 we show the observed and intrinsic x - ray luminosities versus redshift of the seven compton - thick quasars identified here . four of the compton - thick quasars lie in the goods - n region ( @xmath13 160 arcmin@xmath110 ) of the cdf - n and , despite being x - ray weak / undetected , they are amongst the most intrinsically luminous agns identified in this field .
this shows that even in the deepest x - ray survey currently available , we are only able to directly probe the peak of the compton - thick agn population at @xmath6 2 .
the two luminous compton - thick quasars were identified in the wider - area swire survey ( @xmath13 0.6 deg@xmath110 ) and are amongst the most luminous obscured agns known .
four of the seven compton - thick quasars are detected in the x - ray band , most typically at @xmath111 kev .
three of these objects have hard x - ray spectra ( @xmath112 ; hdf - omd49 ; sw 104406 ; sw 104409 ; polletta 2006 ) and one object has a comparatively soft x - ray spectrum ( @xmath52 1.6 ; fsc 10214 ; alexander 2005c ) .
the comparatively soft x - ray spectrum of fsc 10214 might be dominated by scattered emission and star formation from the host galaxy , as often found in compton - thick agns ( e.g. , bassani et al .
1999 ; comastri 2004 ; guainazzi et al .
we can also place basic x - ray spectral constraints on the three x - ray undetected compton - thick quasars by stacking their x - ray data ; all of these objects lie in the goods
- n field where the x - ray data are particularly sensitive . using the x - ray stacking code adopted by worsley ( 2005 )
, we obtain @xmath13 3 @xmath36 detections in the 0.58 kev and 28 kev bands , and have a @xmath13 2 @xmath36 constraint in the 0.52 kev band ; the significance of these stacking results were calculated using 10,000 monte carlo trials .
the stacked data corresponds to fluxes ( and a 3 @xmath36 upper limit ) of @xmath113 erg s@xmath3 @xmath5 ( 0.58 kev ) , @xmath114 erg s@xmath3 @xmath5 ( 0.52 kev ) , and @xmath115 erg s@xmath3 @xmath5 ( 28 kev ) . the average rest - frame 210 kev luminosity constraint for these three objects derived from the observed - frame 0.52 kev flux is @xmath116 erg s@xmath3 .
the derived x - ray spectral slope of @xmath117 unambiguously confirms that heavily obscured agns are present in these sources ( e.g. , see fig .
2a in alexander 2005b ) .
the seven compton - thick quasars investigated here were selected on the basis of unambiguous evidence for agn activity from optical and/or mid - ir spectropscopy .
the advantage with this approach is that we have been able to estimate the intrinsic agn properties of these objects using the optical mid - ir spectroscopy .
however , the requirement for spectroscopic evidence of agn activity has led to a comparatively small sample of objects .
nevertheless , we can provide some insight into the completeness of compton - thick agn selection by comparing the properties of our reliably identified compton - thick quasars to the larger samples of candidate compton - thick agns that have been photometrically identified in other studies .
arguably , the most complete identification studies of distant compton - thick agns to date are daddi ( 2007a , b ) and fiore ( 2008 ) .
daddi ( 2007a , b ) identified a large population of x - ray undetected @xmath6 2 galaxies with an excess of mid - ir emission over that expected from star formation ( as predicted using the dust - extinction corrected uv luminosity ) .
x - ray stacking analyses revealed a hard x - ray spectral slope from these objects ( @xmath118 ) , comparable to that of the compton - thick quasars and unambiguously identifying the presence of heavily obscured agns ; see 4.2 . in fig .
6a we plot the ratio of mid - ir based to uv - based star - formation rate ( sfr ) versus 8 @xmath0 m luminosity of the compton - thick quasars and compare them to daddi ( 2007a , b ) ; we have not calculated the properties of fsc 10214 due to the significant uncertainties in the lensing magnification of the star - forming regions .
the compton - thick quasars lie significantly above the mid - ir excess threshold defined by daddi ( 2007a ) , indicating that they are extreme examples of the mid - ir excess galaxy population .
this analysis shows reliably that at least a subsample of the mid - ir excess galaxy population host compton - thick agns .
fiore ( 2008 ) took a complementary approach to that of daddi ( 2007a , b ) and selected x - ray undetected objects with extreme mid - ir - to - optical flux ratios ( @xmath119 ) and red optical colors ( @xmath120 ) .
x - ray stacking analyses of these objects revealed a hard x - ray spectrum with a slope similar to that found for the mid - ir excess galaxies ( @xmath52 1.0 , calculated from their stacked x - ray count rates ) . in fig .
6b we plot the mid - ir - to - optical flux ratio versus @xmath121 color of the compton - thick quasars and compare them to the candidate compton - thick agns studied by fiore ( 2008 ) and daddi ( 2007a ) ; as before , we did not calculate the properties of fsc 10214 .
formally , none of the compton - thick quasars match the candidate compton - thick agn selection criteria of fiore ( 2008 ) , although two of the objects lie close .
furthermore , only @xmath13 10% of the mid - ir excess galaxies of daddi ( 2007a ) match the fiore ( 2008 ) criteria .
these analyses show that the candidate compton - thick agn criteria of daddi ( 2007a ) and fiore ( 2008 ) are mutually exclusive , suggesting that the distant x - ray undetected agn population may be larger than either study has suggested .
the requirement in our study for optical spectroscopic redshifts and the optical identification of agn activity ( with the exception of smm j123600 ) probably causes a bias against selecting objects with red @xmath121 colors , which by definition will be optically faint .
our seven compton - thick quasars do not comprise a complete sample .
however , four of the objects lie in the field , providing basic constraints on the space density of compton - thick quasars at @xmath6 22.5 . on the basis of the comoving volume at @xmath122 22.5 in a region the size of the goods - n field
, we estimate a comoving space density for compton - thick quasars with @xmath123@xmath124 erg s@xmath3 of @xmath125 ( 0.72.5 ) @xmath8 mpc@xmath9 , where the range only reflects the uncertainty due to small - number statistics ( e.g.,gehrels 1986 ) . in fig .
7 we plot the compton - thick quasar space density and compare it to that found in other studies .
our results suggest that the space density of compton - thick quasars is @xmath13 15 times higher than that of comparably luminous unobscured quasars at @xmath6 22.5 ( @xmath126 mpc@xmath9 ; e.g. , hasinger 2005 ) . since the gilli ( 2007 ) model of the x - ray background constrains the number of compton - thick quasars to be the same as the number of unobscured quasars , our compton - thick quasar space density is also times higher than the compton - thick quasar predictions of gilli ( 2007 ) . as we mention in 4.1 and show in fig . 4 , the evidence for a compton - thick quasar in bx1637 is weaker if the x - ray mid - ir luminosity ratio inferred by maiolino ( 2007 ) is used . however ,
we note that a significantly more important issue is likely to be the effect of cosmic variance since we have only identified compton - thick quasars in a small region over a narrow redshift range . for example , within the goods - n field , there are also four x - ray unobscured quasars that lie at @xmath6 22.5 , indicating that the true compton - thick unobscured quasar ratio may be toward the lower end of the value given above .
nevertheless , our results clearly indicate that a large fraction of the growth of black holes was obscured by compton - thick material , in general agreement with results based on less - reliable photometrically identified objects ( e.g. , daddi 2007a ; martnez - sansigre et al . 2007 ;
fiore 2008 ) , the predictions made by theoretical models ( e.g. , marconi 2004 ; treister 2006 ) , and the compton - thick agn fraction inferred by the study of maiolino et al .
( 2007 ; fig .
@xmath124 erg s@xmath3 that lie in the goods - n region of the cdf - n field .
the open squares correspond to the results obtained by daddi ( 2007a ) and fiore ( 2008 ) using photometrically identified objects and x - ray stacking analyses ; the plotted values assume that 100% of their objects are compton - thick agns and they are therefore upper limits ( see 4.4 for discussion ) .
the dashed lines are the predicted space densities of compton - thick agns for a range of x - ray luminosity lower limits from the gilli ( 2007 ) xrb model .
the space density of the compton - thick quasars robustly identified here is a factor @xmath13 15 higher than that found for comparably luminous unobscured quasars , indicating that compton - thick accretion is ubiquitious at @xmath6 2 ; see 4.4 for a more detailed discussion.,width=340 ] we can compare our estimated space density to those determined from the photometric selection of candidate compton - thick agns by daddi ( 2007a ) and fiore ( 2008 ) . on the basis of the x - ray data and
the x - ray6 @xmath0 m luminosity , daddi ( 2007a ) argued that their candidate compton - thick agns have an average intrinsic luminosity of @xmath127 ( 14 ) @xmath128 erg s@xmath3 .
the intrinsic luminosities of these mid - ir excess galaxies are about an order of magnitude lower than the compton - thick quasars identified here , and consequently their space density is higher ( @xmath129 mpc@xmath9 ; see fig . 7 ) .
fiore ( 2008 ) estimated the intrinsic x - ray luminosities of their x - ray undetected candidate compton - thick agns using the 5.8 @xmath0 m luminosities , finding @xmath130 erg s@xmath3 . taking into account the estimated redshift range of their objects ( @xmath6 1.22.6 ) ,
we calculate a space density of @xmath131 mpc@xmath9 ; as mentioned in fiore ( 2008 ) , this space density is consistent with that predicted for compton - thick agns using the xrb model from gilli ( 2007 ) .
taken at face value , these results suggest that lower - luminosity compton - thick agns were times more common than compton - thick quasars at @xmath6 2
. however , since the primary evidence for compton - thick agn activity in these studies was from x - ray stacking analyses , there could be significant numbers of non - compton - thick agns in their samples ( i.e. , either compton - thin agns or star - forming galaxies ) .
both daddi ( 2007a ) and fiore ( 2008 ) asssessed the `` contamination '' from star - forming galaxies to their stacked x - ray results and concluded that @xmath132 50% and @xmath133% of the objects are heavily obscured agns , respectively .
however , these percentages refer to the fraction of photometrically identified objects that are heavily obscured agns and direct optical - mid - ir spectroscopic identification will be required to determine what fraction of these agns are obscured by compton - thick material .
we can also compare our results to those determined for more luminous compton - thick quasars .
polletta ( 2006 ) identified five candidate compton - thick quasars at with @xmath134 erg s@xmath3 on the basis of the _ chandra _ data in the swire survey , giving a space density of @xmath135 mpc@xmath9 .
two of these objects have been presented here as robust compton - thick quasars ( sw j104406 and sw j104409 ; see table 1 ) .
however , the other three candidate compton - thick quasars are all undetected at 24 @xmath0 m ( @xmath136 @xmath0jy ; see table 2 of polletta 2006 ) .
this suggests that either the intrinsic agn luminosities of these compton - thick quasars are more modest ( @xmath137 erg s@xmath3 ) than those estimated by polletta ( 2006 ) or that the x - ray mid - ir relationship shown in fig .
4 does not hold for these luminous objects .
martnez - sansigre et al.(2007 ) identified ten compton - thick quasars at @xmath6 1.74.2 with an average bolometric luminosity of @xmath138 erg s@xmath3 , giving a space density of @xmath139 mpc@xmath9 .
on the basis of the bolometric conversion of elvis ( 1994 ) , which martnez - sansigre et al.(2007 ) also adopted , the estimated average x - ray luminosities are @xmath140 erg s@xmath3 for these sources , suggesting that they are very luminous agns .
however , since two of their objects are compton - thin quasars that are detected at x - ray energies , we can directly estimate the intrinsic 210 kev luminosities of the candidate compton - thick quasars using the average x - ray6 @xmath0 m luminosity ratio of the compton - thin quasars . taking this approach and using the median 24 @xmath0 m flux for the sample ,
the typical intrinsic x - ray luminosities of these quasars would be more modest ( @xmath141 erg s@xmath3 ) , casting significant doubt on whether they are obscured by compton - thick material ; the rest - frame 6 @xmath0 m fluxes are derived from the observed 24 @xmath0 m fluxes using the ngc 1068 sed to apply small @xmath142-corrections .
direct spectroscopic identification of these photometrically classified compton - thick quasars is required to place more direct constraints , which should be possible with current instrumentation .
lastly , we compare our results to the compton - thick agns identified by tozzi ( 2006 ) using x - ray spectral analysis of x - ray detected agns in the _ chandra _ deep field - south survey .
fourteen of the objects investigated by tozzi ( 2006 ) had x - ray spectra that were better fitted by a pure reflection model than an absorbed power - law model , and were consequently classified as compton - thick agns ; derived space densities are @xmath143 mpc@xmath9 at @xmath6 1 and @xmath144 mpc@xmath9 at @xmath6 2 . while the tozzi ( 2006 ) study provides the most quantitative x - ray identification of distant compton - thick agns to date , it is also possible for compton - thin agns to have a pure reflection component ( i.e. , if the power - law component has temporarily decreased ) , and other pieces of evidence are required to confirm that these are compton - thick agns .
two of the seven objects with spectroscopic redshifts have strong fe k@xmath95 emission lines , confirming that they are likely to be compton thick , but supporting evidence for compton - thick absorption is lacking in the other objects .
the current study provides diagnostics to identify distant compton - thick agns but is limited in source statistics , covers a narrow redshift range , and is restricted to the brightest agns at @xmath6 2 .
future studies could focus on obtaining optical mid - ir spectroscopy of lower - luminosity @xmath6 2 candidate compton - thick agns , as well as identifying objects at lower and higher redshifts . assuming the ngc 1068 template used here and the average x - ray mid - ir relationship shown in fig . 4 , an agn with @xmath145 erg s@xmath3 will have 24 @xmath0 m fluxes of @xmath13 2 @xmath0jy , @xmath13 15 @xmath0jy , @xmath13 0.3 mjy , and @xmath13 5 mjy at @xmath146 , @xmath147 , @xmath148 , and @xmath149 , respectively ; these fluxes will typically scale linearly with luminosity ( see 3.3 ) .
mid - ir bright objects ( @xmath150 mjy ) should be identifiable with _ spitzer_-irs but mid - ir spectroscopic identification of fainter objects will need to wait until the launch of the _ james webb space telescope_. optical and near - ir spectroscopy may be easier to obtain for many of the candidate compton - thick agns , although only if the agns have identifiable emission lines , which could be weak in low - luminosity and dust - reddened agns ( e.g. , caccianiga 2007 ; see 4.3 ) .
ultimately , deep x - ray data are required to search directly for the presence of compton - thick absorption in individual objects using high s / n x - ray spectroscopy . for the majority of the objects investigated here
, this will require @xmath151 ms exposures with current x - ray telescopes ; see fig .
5 . however , due to its large lensing boost it is possible to provide reasonable x - ray spectral constraints for fsc 10214 + 4724 in a modest _ chandra _ exposure ( e.g. , a 100 ks _ chandra _ exposure would yield almost 100 x - ray counts , @xmath14 times more than found for hdf - omd49 ; see fig . 3 ) .
large improvements in the x - ray spectral constraints for distant compton - thick agns will need to wait for the next generation of x - ray observatories ( e.g. , _ constellation - x _ , _ nustar _ , _ simbol - x _ , _ xeus _ ) , where the foci are high - resolution spectroscopy and/or high - energy imaging ( @xmath14 kev ) of faint x - ray sources .
for example , a 1 ms _ xeus _ exposure will yield @xmath13 5000 counts for hdf - omd49 , allowing detailed x - ray spectral analyses and providing the potential to search for changes in the x - ray emission due to column density variations , giving direct insight into the environment around the black hole of a distant compton - thick agn ( e.g. , risaliti 2002 ) .
we have presented _ spitzer_-irs spectroscopy and photometry of hdf - omd49 , a @xmath1 optically identified agn that is formally undetected in the 2 ms observation . from a combination of optical mid - ir spectroscopy and x - ray data , we have shown that hdf - omd49 hosts an intrinsically luminous quasar that is obscured by compton - thick material ( @xmath2 erg s@xmath3 ; @xmath97 @xmath5 ) .
we selected six further @xmath6 2 agns from the literature ( four with _
spitzer_-irs spectroscopy ) and used the same x - ray
optical mid - ir diagnostics applied to hdf - omd49 to show that these objects are also likely to be compton - thick quasars with @xmath7 erg s@xmath3 .
we demonstrated that these compton - thick quasars would be classified as mid - ir excess galaxies , on the basis of the daddi ( 2007a ) definition , providing the first spectroscopic confirmation of compton - thick agn activity in a subsample of these @xmath6 2 mid - ir bright galaxies .
four of these compton - thick quasars lie in the goods - n field , and we used these objects to constrain the space density of distant compton - thick quasars , finding @xmath125 ( 0.72.5 ) @xmath8 mpc@xmath9 at @xmath6 22.5 . on the basis of our results , compton - thick quasars were as ubiquitious as unobscured quasars at @xmath6 22.5 , implying that a large fraction of the growth of supermassive black holes must have been obscured by compton - thick material .
we acknowledge support from the royal society ( dma ) , the _ spitzer space telescope _ fellowship program ( ap ) , the natural sciences and engineering research council of canada and the canadian space agency ( ap ) , the _ chandra _ fellowship program ( feb ) , and the nasa ltsa grant nag5 - 13035 ( wnb ) .
we thank r. gilli for generously providing his model tracks and compton - thick agn spectra , and for his insightful comments on an earlier draft of this paper .
we also thank f. fiore for providing the data from his sample , e. sturm for providing the composite obscured quasar spectrum , c. done , m. polletta and e. treister for scientific feedback , and the referee for useful suggestions .
this work is based in part on observations made with the _
spitzer space telescope _ , which is operated by the jet propulsion laboratory , california institute of technology under a contract with nasa .
the irs was a collaborative venture between cornell university and ball aerospace corporation funded by nasa through the jet propulsion laboratory and ames research center . | many models that seek to explain the origin of the unresolved x - ray background predict that compton - thick active galactic nuclei ( agns ) are ubiquitious at high redshift .
however , few distant compton - thick agns have been reliably identified to date . here
we present _
spitzer_-irs spectroscopy and 3.670 @xmath0 m photometry of a @xmath1 optically identified agn ( hdf - omd49 ) that is formally undetected in the 2 ms _ chandra _ deep field - north ( cdf - n ) survey .
the _
spitzer_-irs spectrum and spectral energy distribution of this object is agn dominated , and a comparison of the energetics at x - ray wavelengths to those derived from mid - infrared ( mid - ir ) and optical spectroscopy shows that the agn is intrinsically luminous ( @xmath2 erg s@xmath3 ) but heavily absorbed by compton - thick material ( @xmath4 @xmath5 ) ; i.e. , this object is a compton - thick quasar .
adopting the same approach that we applied to hdf - omd49 , we found a further six objects at @xmath6 22.5 in the literature that are also x - ray weak / undetected but have evidence for agn activity from optical and/or mid - ir spectroscopy , and show that all of these sources are likely to be compton - thick quasars with @xmath7 erg s@xmath3 .
on the basis of the definition of daddi ( 2007 ) , these compton - thick quasars would be classified as mid - ir excess galaxies , and our study provides the first spectroscopic confirmation of compton - thick agn activity in a subsample of these @xmath6 2 mid - ir bright galaxies . using the four objects that lie in the cdf - n field
, we estimate the space density of reliably identified compton - thick quasars [ @xmath8 mpc@xmath9 for @xmath10 erg s@xmath3 objects at @xmath6 22.5 ] and show that compton - thick accretion was probaby as ubiquitious as unobscured accretion in the distant universe . |
clusters of galaxies are playing an important role in the study of large scale structures of the universe .
in particular , the observed dynamical properties of clusters set stringent constrains on models of structure formation as well as on cosmological parameters .
the popular inflation plus the cold dark matter ( cdm ) cosmologies are found to be able to match the existing observations of clusters if the cosmological parameters such as the mass density @xmath0 , the cosmological constant @xmath1 and the mass fluctuation amplitude are properly selected ( e.g. bahcall & cen 1992 ; jing et al . 1993 , 1995 ; jing & fang 1994 ; eke et al . 1996 ; viana & liddle 1996 ; carlberg et al .
1997 ; bahcall , fan & cen 1997 ; xu , fang & wu 1998 ; xu , fang & deng 1999 ) . yet
, some fundamental properties of clusters are not directly measurable , among which the conventional determination of gravitating masses of clusters is related to the dynamical state of the clusters , i.e. the virialization of both baryonic and nonbaryonic matter .
for instance , the so - called dynamical mass estimator relies on the hypothesis that the optical galaxies and/or the hot intracluster gas are the tracers of the underlying gravitational potential .
nevertheless , the virialization assumption is often challenged by the existence of substructures ( e.g. richstone , loeb & turner 1992 ; jing et al .
1995 ; crone , evrard & richstone 1996 ; patrick 1999 ; mathiesen , evrard & mohr 1999 ) .
substructures in both optical and x - ray maps of clusters suggest that clusters may still be in the process of formation ( e.g. stein 1997 ; kriessler & beers 1997 ; solanes et al .
spatially - resolved measurements of gas temperature in some clusters illustrate the complex two - dimensional patterns including the asymmetric variation and the significant decline with radius ( henry & briel 1995 ; henriksen & white 1996 ; henriksen & markevitch 1996 ; markevitch 1996 ) , indicative of strong substructure merging . the unidentified or neglected substructures in clusters may lead to an overestimate of galaxy velocity dispersion , and therefore the virial mass ( smail et al .
this implies that the hydrostatic equilibrium hypothesis may be inappropriate to the dynamical state of optical galaxies and intracluster gas for some clusters . on the other hand , a statistical comparison of the cluster masses determined from optical galaxies , x - ray observations and
gravitational lensing shows that there is an excellent agreement among different mass estimates on scales much greater than the core radii .
the discrepancy of these mass estimators appears only inside the central regions of clusters where the local dynamical activities of galaxies and cooling flows may become dominant ( wu & fang 1997 ; allen 1998 ; wu et al . 1998 ) .
since gravitational lensing reveals the cluster gravitating mass regardless of the matter compositions and their dynamical status , the consistency between these three mass estimators indicates that clusters may be regarded as dynamically - relaxed systems as a whole .
further support to this argument comes from the study of the correlations between the optically determined velocity dispersion of cluster galaxies , the x - ray measured luminosity and the temperature .
namely , although many x - ray clusters show irregular configurations and are substructured , there are tight correlations between the velocity dispersion of optical galaxies , the temperature and the luminosity of the intracluster gas ( evrard , metzler , & navarro 1996 ; mohr , & evrard 1997 ; wu , fang & xu 1998 ; wu , xue & fang 1999 ; mohr , mathiesen , & evrard 1999 ) . in a word ,
the observational evidences for whether galaxy clusters are virialized seem to be contradictory . in this paper
, we intend to highlight this contradiction .
we shall show that beyond the scale of full virialization there exists a radius range in which clusters are quasi - virialized , i.e. the internal structure of _ individual _ cluster exhibits a substantial departure from dynamical relaxation , while many _ statistical _ properties of clusters are approximately the same as those of the virialized system . for instance , the virial relation @xmath2 , where @xmath3 is the velocity dispersion and @xmath4 is the mass of a system within a given radius , is a good approximation on scales larger than virial radius , even though many clusters are irregular in shape and show substructures on these scales .
actually , such a `` duality '' is an indicator of the transition from pre - virialization to virialization .
this transition can roughly be described by three stages : a ) full virialization , in which both the individual objects and their statistical properties are virialized ; b ) quasi - virialization , in which the ensemble statistical features are similar to those of the virialized systems , but the scattering from the averages is still significant ; c ) pre - virialization , in which both individual objects and their ensemble statistical features are significantly different from virialization . in hierarchical clustering scenario ,
the larger the scale , the less the virialization .
thus , the dynamical evolution from phase a ) to b ) or from phase b ) to c ) can be revealed by studying the dynamical properties of the objects on the scale of clusters and beyond .
in particular , one can address the question : on what scale will the transition of phase a ) to b ) or phase b ) to c ) occur ?
we are interested in this issue also because the virialization evolution is model - dependent .
it is hoped that the scale - dependence of virialization is useful for discrimination of various cosmological models .
the present paper is organized as follows . in 2 , we will describe our samples of cluster halos which are multi - scale identified from n - body simulations under three different cosmological models : 1 ) the standard cold dark matter ( scdm ) , 2 ) low density , flat cdm model with a non - zero @xmath5 ( lcdm ) , and 3 ) the open cdm ( ocdm ) model . the press - schechter ( ps ) formalism is also introduced so that the simulation results can be confirmed by a semi - analytical approach .
we then investigate in 3 various statistical indicators of virialization , including density contrast , configuration , substructure and velocity dispersion .
the emphasis is to find the scale - dependence of these features , and determine the scale range of virialization and quasi - virialization . in
4 we will discuss the applicable range of the virial mass estimator as a statistical measure for quasi - virialized systems . in
5 the temperature function of clusters will be constructed and used for model discrimination .
finally , we will briefly summarize our conclusions in 6 .
the samples to be used for analyzing the scale - dependence of virialization are produced by n - body simulations with the p@xmath7 m code developed by y.p.jing ( jing & fang 1994 ; jing et al .
we employed the following cosmological models : 1 ) scdm , 2 ) lcdm , and 3 ) ocdm . the cosmological parameters ( @xmath8 ) are taken to be ( 1.0,0.0,0.5,0.62 ) , ( 0.3,0.7,0.75,1.0 ) , and ( 0.3,0.0,0.75,1.0 ) for the scdm , lcdm and ocdm , respectively .
these parameters provide a consistent description for many observational properties of the universe , especially the abundance of clusters ( e.g. jing & fang 1994 ; bahcall , fan & cen 1997 ) .
other parameters in our simulations , i.e. box size @xmath9 , number of particles @xmath10 and the effective force resolution @xmath11 , are chosen to be ( @xmath9,@xmath10,@xmath12 mpc , @xmath13 mpc ) .
we have run 8 realizations for each model .
a particle has mass of @xmath14 m@xmath15 , which is small enough to resolve reliably the rich clusters of @xmath16 m@xmath15 .
we have also performed simulations with a cobe normalization @xmath17 for scdm .
three realizations were made .
there are two ways of studying the scale - dependence of dynamical status of clusters .
one is to identify clusters with a given size , like abell radius @xmath18 @xmath19 mpc , and then study the dependence of dynamical properties of these clusters on scales less or greater than @xmath20 .
another is to identify clusters by multi - scale resolution analysis , and produce samples of @xmath6-clusters , i.e. clusters identified with scales @xmath6 , and then , study the @xmath6-dependencies , and also the radius - dependence of @xmath6-clusters .
if all objects on all scales @xmath6 are virialized , @xmath6-clusters should show clouds in clouds " morphology , i.e. all smaller @xmath6 clusters locate in the centers of larger @xmath6 clusters . in this case
, the multi - scale identification will actually produce the same samples as the abell radius identification .
the @xmath20-identification and the multi - scale identification would be essentially equivalent .
however , in the evolutionary stage of pre - virialization , the two identifications are different from each other .
the multi - scale identification is more effective to study problems of scale - dependence of clusters and beyond ( xu , fang & wu 1998 ; xu , fang & deng 1999 ; fang & xu 1999 ) .
the detailed procedure of the multi - scale identifications based on the discrete wavelet transform ( dwt ) has been presented in the papers of xu , fang & wu ( 1998 ) and xu , fang & deng ( 1999 ) .
the technique of applying the dwt for large scale structure study is reviewed in fang & thews ( 1998 ) .
briefly , we first describe the distribution of the particles by a 3-d matrix , and then do the fast 3-d _ daubechies 4 _ dwt and the reversed transformations to find the wavelet function coefficients ( wfcs ) and scaling function coefficients ( sfcs ) on various scales . for each scale ,
the cells with sfcs larger than those of the random sample by a given statistical significance , say 3@xmath21 , are picked up as the halos of cluster candidates . around each of the candidates , we place a @xmath22 grid with the size of cluster diameter and
search for the accurate center .
the cluster center is taken as the position where the largest mass is surrounded .
the mass @xmath4 of a cluster is measured by counting the particles within a sphere of radius @xmath6 .
it is enough to take @xmath6 to be numbers of bi - fold , i.e. 0.75 , 1.5 , 3 , 6 , 12 and 24 @xmath19mpc .
since the @xmath6 sphere does not depend on the shape of the wavelet functions , this identification is independent of the selection of wavelet function . with the identified @xmath6-clusters , one can find the mass function ( mf ) or abundance @xmath23 , which is the number density of @xmath6-clusters with mass larger than @xmath4 . as an example the mfs of @xmath6-clusters of the scdm model
are ploted in fig .
1 . it shows scaling behavior . for other models
the mfs show similar behaviors ( fang & xu 1999 ) .
the abundance @xmath24 of clusters on the abell radius given by the dwt identification is found to be the same as those given by usual friends - of - friends ( fof ) identification ( xu , fang & wu 1998 ; xu , fang & deng 1999 ) . since each halo or cluster
is characterized by two parameters @xmath4 and @xmath6 , it is inconvenient to define richness by its mass alone . instead , the richness can be defined by the abundance , @xmath23 , or by the mean separation of neighbor clusters @xmath25 the richness of @xmath26-clusters with masses @xmath27 is equal to the richness of @xmath28- clusters with masses @xmath29 if their @xmath30 s are the same .
for instance , for the top 25 clusters on various scale @xmath6 in the simulation box 310@xmath7 @xmath31 mpc@xmath32 , their richness is @xmath33 @xmath19 mpc . for a given @xmath6
, this richness definition has its usual meaning , i.e. the larger the @xmath4 , the less the @xmath23 and the higher the richness .
the mfs given by the n - body simulation can be explained by the hierarchical clustering of cosmic structures . in the models of hierarchical clustering with gaussian initial perturbations ,
the comoving number density of collapsed halos at @xmath34 in the mass range @xmath4 to @xmath35 is given by ( press & schechter 1974 ) @xmath36 where @xmath37 almost independent of cosmologies .
@xmath38 is the linear theory @xmath39 mass density fluctuation in spheres of mass @xmath4 at redshift @xmath34 within a top - hat window of radius @xmath40 , and is determined by the initial density spectrum @xmath41 and normalization factor @xmath42mpc,0 ) . from eq.(2 ) , the cumulative number density of the collapsed halos with mass greater than @xmath4 is given by @xmath43 equations ( 2),(3 ) have been found to be good descriptions of the mass functions of collapsed clusters identified by fof method from n - body simulation ( e.g. mo , jing & white 1997 ) .
figure 1 shows that ps predications fit very well with the mass function ( mf ) of clusters on the abell radius given by the dwt identification . as pointed out by mo & white ( 1995 ) and bi & fang ( 1996 ; 1997 ) ,
the ps formalism can be employed to approximately describe not only collapsed halos , but also uncollapsed regions , i.e. both collapsed halos and uncollapsed regions can be viewed as the sum of various individual top - hat spheres .
an uncollapsed region ( or quasi - virialized halo " in terminology of this paper ) corresponds to a region in which the initial linear fluctuation is less than the threshold 1.686 at the redshift considered . as a consequence
, one can consider the initial density field as a system consisting of many spheres , each of which has initial radius @xmath40 and density contrasts @xmath44 , and mass is equal to @xmath45 .
let s consider an arbitrary spherical volume of radius @xmath6 at redshift @xmath34 .
matter in this volume can come from various initial spheres with @xmath40 and @xmath44 .
for a given @xmath6 , one can find the relationship between @xmath40 and @xmath44 by considering the nonlinear evolution of a massive shell under spherical symmetry ( padmanabhan 1993 ; mo & white 1995 ; bi & fang 1996 )
. then , we have the relation @xmath46 it means that a spherical region of @xmath47 and @xmath40 will evolve into @xmath6 at @xmath34 , which includes both collapsed and uncollapsed objects . the function @xmath48 is given in bi & fang ( 1996 ) .
if we identify clusters with radius @xmath6 in an arbitrary spatial domain @xmath49 , the fraction of the total mass @xmath50 contained in clusters of mass @xmath51 corresponds to the fraction of mass which have initial fluctuations larger than a given @xmath52 .
for a gaussian field , this fraction is given by @xmath53 where @xmath54 is the variance of density perturbation on scale @xmath40 . because a spherical region of @xmath47 and @xmath40 will evolve into @xmath6 at @xmath34 , initial spherical regions of @xmath55 and @xmath40
will evolve into radii less than @xmath6 at @xmath34 .
therefore , the comoving number density of the collapsed and uncollapsed ps spheres with mass in @xmath56 , which are spheres identified with radius @xmath6 at @xmath34 , is given by @xmath57 dm \\
\nonumber & = & -\sqrt{\frac{2}{\pi } } \frac{\rho_0}{m^2 } \nu e^{-\frac{\nu^2}{2}}(\frac{d ln\sigma}{d lnm } -\frac{d ln\delta}{d lnm } ) , % \end{array}\end{aligned}\ ] ] where @xmath58 denotes the relative height of the initial perturbation ( evaluated at redshift @xmath34 ) .
the number 2 in the equation is the notorious @xmath59 multiplication factor of ps formalism ( see discussion in bond et al .
1991 ) .
the cumulative number density , @xmath23 , of collapsed and uncollapsed @xmath6-halos with mass greater than @xmath4 should be @xmath60 these mfs are also plotted in fig .
it shows that the ps formalism basically explains the mass functions given by n - body simulation , especially the scaling behaviors .
meanwhile , fig .
1 shows that the ps calculations systematically underpredict the mfs on scale @xmath61 h@xmath62 mpc .
this is most likely caused by the hypothesis of spherical symmetry of the dynamic evolution of uncollapsed regions .
a similar discrepancy has also been found in the two - point correlation of halos in underdense ( or uncollapsed ) regions , i.e. the ps formalism significantly underpredicts the halo - halo correlation when small halos are in low - density environments ( jing 1998 ) .
nevertheless , the ps result shows the scale - dependence of statistical features given by the ps formalism is the same as that of n - body simulation .
this agreement is very useful to reveal the status of quasi - virialized ( or uncollapsed ) regions .
a popular indicator of virialization of clusters is the density contrast . for spherical collapse in a standard flat universe , the virial radius is usually measured as the radius within which the mean density contrast is @xmath63 , i.e. @xmath64 . for open universe , the density contrast within virial radius is in the range of @xmath65 to @xmath66 ( lacey & cole 1993 ) . for models of scdm , lcdm and ocdm , the density contrasts @xmath67
are equal to 178 , 335 , 402 , respectively .
observationally , the spatial range of virialization is characterized by @xmath68 , the radius where the average interior number density of galaxies is @xmath69 times higher than the average .
thus , the deviation from virialization of a @xmath6-cluster can be quantified by the mean mass contrast , @xmath70 , where @xmath71 is the mean mass within radius @xmath6 , and @xmath72 the mean mass excess of the @xmath6-clusters .
2 plots the mass contrasts of @xmath6-clusters with richness @xmath73 30 , 50 and 100 @xmath19 mpc for the scdm , lcdm and ocdm models , respectively .
the lines in lcdm almost overlap with those for ocdm . as expected , the more massive ( i.e. the larger @xmath30 ) the clusters , the closer to virialization . for the scdm clusters with richness @xmath73 30 , 50 , and 100 @xmath19 mpc , the virialized radii @xmath74 at @xmath34 are 1.09 , 1.33 , and 1.60 @xmath19 mpc , respectively .
they are 1.03 , 1.33 , 1.82 @xmath19 mpc for lcdm , and 0.96 , 1.27 , 1.70 @xmath19 mpc for ocdm .
therefore , except for the richest clusters , the full virialization generally is realized only on scales less than @xmath18 @xmath19 mpc .
this is consistent with the observed result that @xmath68 is usually smaller than the abell radius @xmath20 .
namely , except for the core regions of clusters , abell clusters as a whole are not yet fully virialized . in the case of positive bias ,
@xmath68 given by galaxies is larger than that of dark matter , and therefore , the real virialized range will be even smaller than that given by galaxy observations . as a result , in the scale range of equal to and larger than the abell radius , halos for @xmath6-clusters are mostly in the stages of quasi- or even pre - virialization .
the spatial configuration of clusters is also a useful indicator of virialization .
as discussed above , if the halos of clusters are fully virialized , their spatial configurations have to be regular clouds in clouds " , i.e. the smaller @xmath6-clusters are located in the centers of larger @xmath6-clusters .
3 plots a projected distribution of @xmath6-clusters identified from one realization of model ocdm within box 310 @xmath31 mpc@xmath7 , which illustrates the top 25 massive clusters for each @xmath6 ( the richness is @xmath75 @xmath19 mpc ) .
the scales @xmath6 are taken to be @xmath76 and 24 @xmath19mpc , respectively .
3 shows a variety of configurations of the halos .
apparently , some halos show clouds in clouds " morphology , while some do not .
according to the scale of @xmath68 in fig .
2 , it is expected that , for such a high richness , clusters of @xmath77 2 @xmath19 mpc have to be symmetrical and concentric clouds in clouds " structures .
3 , indeed , shows that 21 of the 25 @xmath78 and 1.5 @xmath19 mpc are concentric .
these @xmath79 0.75 - 1.5 clouds in clouds " objects are fully virialized
. on the other hand , there are seven @xmath6 = 3 @xmath19 mpc clusters which do not contain either @xmath78 @xmath19 mpc or 1.5 @xmath19 mpc clusters in their centers .
these are massive clusters on scale 3 @xmath19 mpc , but without fully virialized cores . on scales
@xmath80 @xmath19 mpc , more @xmath6-clusters are asymmetric , irregular , and without small - scale high density peaks within them .
these features show that identifying structures with a given scale like @xmath20 may miss massive objects which do not contain high density peaks on scale @xmath20 . only in the range @xmath81
, the multi - scale identifications find the same sample as the single - scale identification . in terms of mass and length scale ,
@xmath82 @xmath19 mpc clusters actually are superclusters .
observationally , supercluster is defined as cluster of rich clusters ( top @xmath20-clusters ) . in fig .
3 , one can see that some @xmath20-clusters are located together to form larger scale clumps or filaments .
they are obviously superclusters .
fig.3 shows , however , that some of the top 25 @xmath82 @xmath19 mpc halos contain nothing of @xmath77 24 @xmath19 mpc massive halos .
this is , although the mass and length scale of these objects are the same as superclusters , they do not contain any of the top 25 clusters on smaller scales .
this result indicates that the identification by `` cluster of rich clusters '' probably is incomplete .
recently , it has been found that the dynamical properties of superclusters are not only determined by rich clusters , but also by numerous galaxies in the intra - cluster regions ( small et al .
1998 ) . therefore , to study the dynamics on scales beyond clusters , one should directly identify structures from matter or galaxy distributions , instead of using only rich clusters .
similar to the configuration , the number of substructures in a cluster is also an effective measure of the deviation from virialization .
the asymmetry of the configuration of clusters actually is caused by their substructures .
considering that the time scale required for a sound wave in the intracluster gas to cross a cluster is shorter than the dynamical time of the cluster , it is reasonable to assume that the intracluster gas is in hydrostatic equilibrium with local gravitational potential , and therefore , x - ray images map the cluster potential locally ( jing et al .
thus , one can identify substructure as potential minimum of clusters .
we search for the potential minimums by the same algorithm of jing et al .
briefly , we first place a mesh with 0.24 @xmath19mpc resolution around each center of identified @xmath6-cluster halos . the size of this mesh
is taken to be 7.5 , 7.5 , 7.5 , 15 and 30 @xmath19mpc for clusters of @xmath79 0.75 , 1.5 , 3 , 6 and 12 @xmath19mpc , corresponding to grid points @xmath83 and @xmath84 , respectively .
we accumulate mass density values for the meshes from particles .
the mass of each particle is smoothed by a gaussian kernel @xmath85 where @xmath86 is the smoothing length .
the smoothing length @xmath87 of particle @xmath88 is equal to the local mean separation @xmath89 , calculated by counting five nearest particles around @xmath88 .
then the density value on arbitrary cell @xmath90 is given by @xmath91 where @xmath92 is the separation between cell @xmath90 and particle @xmath88 of mass @xmath93 .
the summation in eq.(9 ) is made over all particles within radius 3.75 h@xmath62 mpc . with this mass distribution , gravitational potential on the grid
can be obtained .
we then identify a cell to be a potential minimum if its potential value is smaller than those of its all 26 neighbors .
obviously , if a @xmath6-cluster contains only one potential minimum , it means no substructure .
4 shows the dependence of average number of substructures per @xmath6-cluster on the count of the particles within the cluster , which is proportional to the cluster mass .
as expected , the larger @xmath6 clusters contain more substructures on average because larger scale objects are less virialized . a `` surprising '' result shown in fig .
4 is that for @xmath94 mpc clusters , the more massive the clusters ( i.e. the more the count of particles ) , the more the substructures .
this seems to be contradicting with the general conclusion of last section : the more massive the cluster , the closer to virialization . actually , this is due to the threshold used for identifying substructures . in massive objects substructures with deep potential valleys
are easy to grow up , but not so for less massive objects . for clusters with @xmath95mpc , containing no substructures on average ( as shown in fig .
4 ) does nt mean that all @xmath96 1.5 @xmath19mpc clusters are fully virialized .
contrarily , low richness @xmath96 1.5 @xmath19mpc clusters are pre - virialized ( fig .
they simply do not have enough mass to form substructures above the defined threshold .
similar phenomena can be seen from fig .
5 which shows the probability distributions of the number of substructures per @xmath97mpc clusters with richness @xmath98 mpc , i.e. the top 59 richest clusters in the simulation box 310@xmath7 @xmath31 mpc@xmath32 . from fig .
5 it is seen that the scdm clusters generally are less substructured than those of the lcdm and ocdm . it does nt mean that the scdm clusters are more close to virialization . instead , the deep potential valleys for substructures in the scdm @xmath97mpc clusters formed later than the lcdm and ocdm . fig .
6 gives the @xmath6-dependence of the mean number of substructures per @xmath6-clusters having richness @xmath73 80 h@xmath62 mpc .
for all the three cosmological models , the number of substructures per @xmath6-clusters increases with @xmath6 . as fig . 5
, the mean number of substructures in model scdm is less than those of lcdm and ocdm .
the substructure number at redshift @xmath99 is less than that of @xmath34 , too .
this is once again to indicate that deep potential valleys have not yet developed at @xmath99 .
thus , despite substructure is an indicator of the deviation from virialization , there is no simple relation between virialization and substructures .
a system with rich substructures definitely deviates from virialization , but the existence of substructures is not a necessary condition of the deviation from virialization .
systems with more deep potential valleys are not always in a lower degree of virialization than those with fewer potential valleys .
this point is important in studying objects on scales larger than @xmath20 .
the most decisive measure of virialization certainly is the velocity dispersion of a cluster . for a virialized @xmath6-cluster we have @xmath100 , in which @xmath101 denotes the mass of a cluster within radius @xmath6 .
the velocity dispersion @xmath3 of @xmath6-cluster is calculated from the velocities of all cdm particles within radius @xmath6 with respect to the center of the cluster .
7 shows the three dimensional @xmath3 of @xmath102 mpc clusters identified from three realizations of the ocdm model at @xmath34 . in the case of full virialization , @xmath3 of the system with a given size @xmath6
should be completely determined by its masses @xmath101 .
7 shows that @xmath3 is almost completely determined by @xmath6 and @xmath101 if clusters are as massive as @xmath103 @xmath104 and @xmath105 , respectively for @xmath96 0.75 , 1.5 and 3 @xmath19 mpc .
namely , in these parameter ranges ( @xmath6 , @xmath4 ) , clusters are fully virialized . out of it ,
velocity dispersion @xmath3 is no longer completely determined by radius @xmath6 and mass @xmath101 . as shown in fig .
7 , the scattering of velocity dispersion @xmath3 around their mean @xmath106 can be as large as 15% ( see the vertical error bars of fig .
this scattering is not caused by poisson fluctuations .
for @xmath107 mpc , the scattering increases with the increase of @xmath6 and the decrease of @xmath101 .
therefore , the scattering is due to the deviation from virialization . the mean velocity dispersion @xmath106 of the panels of @xmath108 @xmath19 mpc in fig . 7 is found to be proportional to @xmath109 .
8 plots the result of @xmath110 for the data set shown in fig . 7 .
8 clearly demonstrates that for scales @xmath108 @xmath19 mpc and richness @xmath111 20 100 @xmath19 mpc , all the three models yield @xmath112 , which is the value for virialized systems .
namely , the ensembly averaged relations between mass and velocity dispersion of @xmath108 @xmath19 mpc clusters are the same as those for the virialized systems . even for ensemble of rich @xmath113 @xmath19 mpc clusters with
@xmath114 @xmath19 mpc , virialization holds approximately true .
however , this approximation is no longer valid for the @xmath82 @xmath19 and @xmath115 @xmath19 mpc clusters . strictly speaking , virial theorem is @xmath116 , where @xmath117 and @xmath118 represent , respectively , the potential and kinetic energy of the system
therefore , it includes two numbers 1 . ) the index @xmath119 from @xmath120 ; 2 .
the coefficient @xmath121 from @xmath122 . in the above used parameter space of @xmath6 and @xmath30 ,
both the index and the coefficient are found to be not different from 1/2 by 10% .
one can define quasi - virialization " to be a dynamically evolutionary stage of @xmath6-clusters , for which _ ensemble averaged _ properties approximately agree with those required by virialization , while _ individual _ clusters may significantly depart from virialization .
thus , figs . 7 and 8
show that the quasi - virialization range for the lcdm and ocdm clusters at @xmath34 is from radius @xmath74 to about 6 @xmath19 mpc with richness @xmath123 @xmath19 mpc , and to about 12 @xmath19 mpc with @xmath114 @xmath19 mpc .
for the scdm clusters of @xmath6 @xmath124 12
@xmath19 mpc , only the richest clusters ( @xmath125 @xmath19 mpc ) are quasi - virialized .
the value @xmath126 at redshift @xmath99 is also plotted in fig .
8 . for the lcdm and ocdm , the quasi - virialization
is already realized for structures of @xmath127mpc at @xmath99 . for the scdm ,
the quasi - virialization approximately holds for clusters of @xmath6 @xmath128 6 @xmath19 mpc , while @xmath113 @xmath19 mpc clusters are completely out of the quasi - virialization range . the redshift evolution of @xmath126 for models scdm
, lcdm and ocdm is shown in fig .
9 , which gives the mean @xmath126 at redshifts @xmath129 0 , 0.2 , 0.5 , 0.8 for samples of top 59 clusters , or richness @xmath73 80 @xmath19 mpc , where @xmath6 is taken to be 1.5 , 12 and 24 @xmath19 mpc . for clusters of @xmath130 12 @xmath19 mpc in the lcdm and ocdm models
, we have @xmath112 in the redshift range of @xmath131 .
but in the scdm model , the stage of quasi - virialization has not yet been reached at redshift @xmath132 for clusters of @xmath79 12 @xmath19 mpc .
with the above analysis on quasi - virialization , one can now study the applicability of the virial mass estimator .
especially , the individual - statistical ( or individual- ensemble ) duality of quasi - virialization is very useful for determining the available range of the virial mass as a statistical measure .
the virial estimator of the total mass of a cluster within radius @xmath133 is given by ( e.g. bahcall & sarazin 1977 ; mathews 1978 ) , @xmath134 where @xmath135 is the density of galaxies .
for isothermal sphere approximation of the galaxy distribution , we have @xmath136 , where @xmath137 is the core radius . basically , the applicability of eq.(10 ) relys on the following three assumptions : 1 . ) the second term in the bracket of eq.(10 ) , i.e. @xmath138 , is negligible .
namely , @xmath139 should be independent of @xmath133 ; 2 . ) for a given @xmath133 , the velocity dispersion @xmath139 is proportional to @xmath140 , i.e. @xmath141 ; 3 . )
the profile of mass distribution @xmath142 does nt significantly depend on @xmath3 and richness @xmath30 .
the assumption 2 has been studied in 3.4 .
the mean @xmath126 is indeed equal to 0.5 for quasi - virialized clusters .
we will show below that the assumptions 1 and 3 are also statistically available for quasi - virialized systems . to test the assumption 1 , we calculated the mean velocity dispersion @xmath3 within radius @xmath133 among the 50 most massive clusters ( or richness @xmath143 85 @xmath19 mpc ) at scales @xmath79 0.75 , 1.5 , 3 , 6 , 12 and 24 @xmath19 mpc .
the results averaged from 8 realizations for ocdm and scdm are plotted in fig .
the lcdm gives similar results as the ocdm .
10 shows that the profiles of velocity dispersion of @xmath144 @xmath19 mpc clusters are rather flat , and the variations of the velocity profiles over the radius range @xmath145 @xmath19 mpc are no more than @xmath146 . beyond @xmath147 @xmath19 mpc , the velocity of particles is dominated by the hubble expansion .
10 also shows that @xmath3 is virtually @xmath6-independent for clusters of @xmath148 @xmath19 mpc . when @xmath61 @xmath19 mpc , @xmath3 is lower for larger @xmath6 .
however , the velocity dispersion of @xmath149 @xmath19 mpc clusters is only @xmath124 20% lower than that of @xmath150 @xmath19 mpc clusters . even in the case of @xmath151 @xmath19 mpc ,
the @xmath3 is not different from @xmath150 @xmath19 mpc clusters by a factor of 2 . despite many top @xmath79 6 and 12 @xmath19 mpc clusters
do not contain top @xmath152 mpc clusters as clouds in clouds " , the _
mean _ profile of the velocity dispersion @xmath3 remains to be flat from 1 to @xmath124 10 @xmath19 mpc for clusters of @xmath77 6 @xmath19 mpc and @xmath153 40 @xmath19 mpc or @xmath77 12 @xmath19 mpc and @xmath153 80 @xmath19 mpc .
therefore , the assumption 1 is statistically correct for quasi - virialized clusters . to test the assumption 3 , we calculated the cross - correlation function between clusters and mass particles , @xmath154 , where @xmath155 is the number of particles within a spherical shell with radius @xmath156 , and @xmath157 is the mean number expected from random distribution . fig .
11 gives the cluster - particle cross - correlation functions of clusters with @xmath79 1.5 , 6 and 24 @xmath19 mpc and richness @xmath158 @xmath19 mpc .
it shows that the mean cross correlation functions @xmath159 are about the same for all @xmath6-clusters .
one can describe the correlation function @xmath160 as @xmath161 ( or , @xmath162 ) in the radius range from 1 to 20 @xmath19 mpc .
the index @xmath163 and their variance of each @xmath6 are listed in table 1 . for the lcdm and ocdm ,
all the indexes of @xmath130 6 @xmath19 mpc clusters are @xmath164 . for the scdm
@xmath163 is slightly lower than 0.5 at @xmath130 6 @xmath19 mpc .
12 plots @xmath159 for the @xmath79 6 @xmath19mpc clusters with richness from @xmath165 to 120 @xmath19 mpc in the ocdm .
@xmath159 shows a systematic increase with @xmath30 , indicating that the mean mass of clusters increases with @xmath30 .
however , the shapes of density profiles for all richness can approximately be written as @xmath166 , or @xmath167 , and @xmath164 in the radius range 1 - 20 @xmath19mpc .
that is , the shape of mass profile is independent of the richness .
moreover , for a given @xmath6 such as @xmath6= 6 @xmath19 mpc , the mean velocity dispersion @xmath21 increases with mass or @xmath30 ( fig .
therefore , the cluster - particle cross - correlation functions @xmath159 with different @xmath30 in fig .
12 actually is for different @xmath21 .
therefore , fig .
12 also shows that the shape of mass profile also does not depend on the mean velocity dispersion .
thus , one can conclude that the mass profile @xmath167 with @xmath164 is weakly dependent on @xmath6 , richness and velocity dispersion .
one may safely employ the assumption 3 at least on scales @xmath168 @xmath19 mpc , i.e. in the range of quasi - virialization .
actually , the independence of the cluster mass profiles on the richness and velocity dispersion has been found by jing et al .
( 1995 ) .
it should be pointed out that the radius range shown in figs .
11 and 12 are generally larger or much larger than @xmath68 .
that is , we study only on scales much larger than the cores of clusters , and therefore , the core radius is less important . accordingly , mass estimator eq.(10 ) is good as a statistical measurement for clusters on radius larger than @xmath74 or @xmath68 , and within the range of quasi - virialization .
the goodness of this statistical measurement can be seen from fig .
7 , in which the vertical error bars show the deviation of @xmath169 from their virialization average .
the error bars @xmath170 are of the order of @xmath171 .
thus , @xmath172 given by estimator ( 10 ) should have an uncertainty of about @xmath173 .
this result provides an explanation for the fact that the cluster virial masses determined from optical galaxies statistically are in a good agreement with gravitational lensing - derived masses on scales much greater than the core radii .
for instance , based on a sample of lensing and optical clusters , the correlations between the virial mass @xmath174 and weak lensing - derived mass @xmath175 of clusters in the scale range of @xmath176 @xmath19 mpc are found to be ( wu & fang 1997 ; wu et al .
1998 ) @xmath177 since the samples used for eq .
( 11 ) are not statistically as uniform as simulation sample , one can not take a detailed comparison of observed results with simulations .
nevertheless , the individual - ensemble duality of quasi - virialization gives a reasonable explanation for the eq.(11 ) which used the sample consisting of pre - dynamically - relaxed clusters .
similar to eq.(11 ) , the mass @xmath178 estimated from hydrodynamical equilibrium model of x - ray gas is found to be , _ on average _ , in agreement with the gravitational lensing - derived cluster masses @xmath179 . the correlation between @xmath180 and @xmath175 in the scale range of @xmath181 @xmath19 mpc is ( wu & fang 1997 ; wu et al .
1998 ) @xmath182 this result also shows the feature of quasi - virialization : the hydrodynamical equilibrium model - derived @xmath178 is reliable , while individual clusters may deviate from virialization .
moreover , the most recent statistics using the largest sample of 149 clusters gives ( wu , fang & xu 1998 ) @xmath183 where the parameter @xmath184 is the mean molecular weight for intracluster gas and @xmath185 is the proton mass .
eq.(13 ) means that the temperature @xmath186 of x - ray emitting intracluster gas statistically can be used as a virial indicator of the underlying gravitational potentials of clusters , and is related to the velocity dispersion by virial relation as @xmath187 the current observations have revealed a constant temperature profile within virial radius ( irwin , bergman & evrard 1999 ) . since the profile of @xmath106 is flat within radius @xmath188 6 @xmath19 mpc ( fig .
10 ) , eq.(14 ) indicates that the averaged temperature of clusters is also approximately constant within the range of quasi - virialization .
for the virialized and quasi - virialized clusters , we can define the temperature function ( tf ) of clusters , @xmath189 , which is the number density of @xmath6-clusters with temperature larger than @xmath186 . with eq.(14 )
, the tfs can be calculated from the simulated velocity dispersion function ( vdf ) @xmath190 , which is the number density of @xmath6-clusters with velocity dispersion larger than @xmath3 .
actually , eq.(14 ) has been widely employed in the calculation of tf from vdf ( e.g. klypin & rhee , g. 1994 ; jing & fang 1994 ) . yet , as we have emphasized , we are interested in the effect of dynamical state on the tfs . fig . 13 plots the differential tfs , @xmath191 , for @xmath96 0.75 to 12 @xmath19mpc clusters in the three dark matter models .
in fact , the vdf - tf transfer [ eq.(14 ) ] may not be correct for clusters of @xmath96 12 @xmath19 mpc , which is plotted only for comparison .
if the system is fully virialized on scales @xmath192 , the tfs should be basically scale - independent in this scale range , because virial temperature @xmath186 is independent on radius , and clusters identified by @xmath193 is the same as @xmath194 . fig .
13 shows the significant scale - dependence of the tfs , even at @xmath78 @xmath19 mpc .
therefore , these systems deviate from full virialization
. however , fig .
13 shows that all the tfs of @xmath108 @xmath19 mpc have similar shape , i.e. the tfs are scaling .
this feature is from quasi - virialization .
the tfs of @xmath195 mpc clusters do not join the scaling of @xmath108 @xmath19 mpc , and therefore , it is not in quasi - virialization state .
this is expected , as fig .
7 has shown that @xmath196 mpc clusters are different from that of @xmath168 @xmath19 mpc . as an application of simulated tf @xmath197 ( fig .
13 ) , we analyze the problem of the size of x - ray clusters .
observationally , x - ray clusters are described by their x - ray luminosity , flux weighted temperature and morphology .
although for some individual clusters we have known their x - ray isophotes and therefore the aperture of the x - ray emission , the x - ray luminosity is generally published as pseudo - total luminosity which does not contain information about the spatial scale of clusters .
the abell radius @xmath20 is not used as a condition in x - ray cluster identification . even for abell clusters ,
there is also the ambiguity of whether the x - ray luminosity comes from cluster halos of larger than radius @xmath198 1.5 @xmath19 mpc . if clusters are fully virialized , the size uncertainty of the tfs is not important since the tfs are size - independent .
however , in quasi - virialization state , the size - dependence of tfs is substantial .
for instance , we ploted the observed tf given by edge et al .
( 1990 ) and henry & arnaud ( 1991 ) in fig . 13 .
the lcdm and ocdm tfs are in good agreement with the observed data only if @xmath6 is in the range 0.75 - 1.5 @xmath19 mpc . in other words
, the lcdm and ocdm can pass the tf test if most x - ray clusters used for the observed tf are on the scales of 0.75 and 1.5 @xmath19 mpc . on the other hand , to fit the scdm tf with observed data , the size of x - ray clusters should be in the range 1.5 -3 @xmath19 mpc , which is larger than the results of lcdm and ocdm by a factor of 2 .
the role of the size of x - ray clusters can be more clearly illustrated by scdm model with normalization amplitude @xmath199 .
fig.14 shows the tfs of the @xmath199 scdm model . in this case
, the model - predicted tf can fit the data only if the size of x - ray clusters is in the range 6 to 12 @xmath19 mpc , which seems very unlikely .
thus , one can conclude that for either the model of scdm or lcdm and ocdm , the simulated tfs can fit with observations , if only the size of the x - ray clusters used for constructing the tfs is , on average , larger than their virial radius .
this is consistent with the fact that the configuration of x - ray clusters generally is irregular and substructured .
obviously , the scale - dependence of the tfs can be employed for discriminating among models if the information of the size of x - ray clusters is available .
currently , we are lacking of the data of x - ray cluster sizes .
instead , one can consider the test of tf plus two - point correlation function . as it has been shown in figs .
13 and 14 , the observed tfs can set a lower limit to the scale of clusters .
for instance , the data of edge et al .
( 1990 ) require that @xmath6 should be @xmath200 @xmath19 for the lcdm and ocdm , @xmath201 @xmath19 mpc for the scdm , and @xmath202 @xmath19 mpc for the @xmath17 scdm . on the other hand , for a given abundance , the larger the scale @xmath6 , the smaller the correlation length ( xu , fang & deng 1999 ) . at the abundance @xmath203 @xmath19 mpc
, the correlation length is found to be @xmath204 mpc if only @xmath108 @xmath19 mpc in the lcdm and ocdm ; and @xmath150 @xmath19 in the scdm . since the observed correlation length @xmath205 of x - ray clusters is @xmath204 mpc .
the lcdm and ocdm are able to pass the two constraints to @xmath6 simultaneously : 1 . )
tf requires @xmath206 @xmath19 mpc , and 2 . )
correlation length requires @xmath108 @xmath19 mpc . yet
, the scdm with @xmath17 is clearly in trouble as 1 . )
tf requires @xmath80 @xmath19 mpc , and 2 . )
correlation length requires at least @xmath150 @xmath19 mpc .
using n - body simulated samples in models of the scdm , lcdm and ocdm , we have analyzed the dynamical state of the dwt multi - scale identified clusters .
we showed that state of quasi - virialization exists in the dynamical evolution of these clusters , i.e. while the internal structure of individual clusters significantly differs from dynamical relaxation on scales larger than virial radius @xmath74 , some statistical features of the clusters are approximately the same as those for the virialized systems .
actually , the virial statistical features , like @xmath2 , @xmath207 , are scaling relations .
the dynamical reason of the existence of quasi - virialization is that some of the scaling properties of dynamical relaxed systems of cosmic gravitational clustering approximately hold beyond the fully virialization regime .
the scaling behavior of mass functions of the quasi - virialized clusters can also be repeated by semi - analytic calculation on collapsed and uncollapsed halos in press - schechter formalism .
consequently , in terms of statistical description , the cluster galaxies and x - ray gas can be used as virial tracers of cluster potential on scales beyond the full virialization .
this result provides a good explanation that the virial masses and x - ray masses are basically the same as the gravitational lensing determined masses on scales beyond the full virialization .
it also explained the tight correlation between the velocity dispersion of optical galaxies , and the temperature of the clusters .
the virial mass estimator based on the assumptions of isothermal and hydrostatic model is applicable on scales as large as about @xmath208mpc . in the quasi - virialization state ,
the tfs have scaling .
it is potentially important for model - discrimination .
a very preliminary result given by the test of the scale - dependence of the tfs showed that the lcdm and ocdm are favored .
cccccccc ocdm & [email protected] & [email protected] & [email protected]&[email protected] & [email protected] & [email protected] lcdm & [email protected] & [email protected] & [email protected]&[email protected] & [email protected] & [email protected] scdm & [email protected] & [email protected] & [email protected]&[email protected] & [email protected] & [email protected] | using samples of structures identified by a multi - scale decomposition from numerical simulation , we analyze the scale - dependence of the virialization of clusters . we find that beyond the scale of full virialization there exists a radius range over which clusters are quasi - virialized , i.e. while the internal structure of an _ individual _ cluster is at substantial departure from dynamical relaxation , some _ statistical _ properties of the multi - scale identified clusters are approximately the same as those for the virialized systems .
the dynamical reason of the existence of quasi - virialization is that some of the scaling properties of dynamically relaxed systems of cosmic gravitational clustering approximately hold beyond the full virialization regime .
this scaling can also be seen from a semi - analytic calculation of the mass functions of collapsed and uncollapsed halos in the press - schechter formalism .
the individual - statistical " duality of the quasi - virialization provides an explanation of the observed puzzle that the total masses of clusters derived from virial theorem are statistically the same as the gravitational lensing determined masses , in spite of the presence of irregular configuration and substructures in individual clusters .
it also explains the tight correlation between the velocity dispersion of optical galaxies and the temperature of x - ray emitting gas .
consequently , the virial mass estimators based on the assumptions of isothermal and hydrostatic model are statistically applicable to scales on which the clusters are quasi - virialized . in the quasi - virialization regime , the temperature functions of clusters also show scaling .
this feature is a useful discriminator among models . as a preliminary comparison with observation
, the discriminator yields a favor of the models of lcdm and ocdm . |
takada _ et al .
_ found superconductivity in the layered quasi - two dimensional na@xmath0coo@xmath1 when intercalating enough water ( @xmath5h@xmath1o ) to form a separate water layer between coo@xmath1 and na layers.@xcite the nonsuperconducting dehydrated na@xmath0coo@xmath1 system shows a rich phase diagram , which significantly depends on @xmath4.@xcite for @xmath6 , the system shows weakly correlated band - like behavior including pauli paramagnetism , while the phase @xmath7 reveals correlated behavior such as large enhancement in linear specific coefficient , curie - weiss susceptibility,@xcite and magnetic ordering for @xmath8 .
the most peculiar aspect of this system is an insulating phase at @xmath9@xcite with a tiny gap @xmath10 mev.@xcite as the temperature is decreased , antiferromagnetic ordering of some co spins appears at @xmath11 k , and at @xmath12 k there is a gap opening , which reflects the charge - ordering of nonmagnetic @xmath13 co1 ions and magnetic @xmath14 co2 ions.@xcite using neutron diffraction studies , williams _
_ inferred the charge difference of 0.12@xmath15 between co1 and co2.@xcite this value is much smaller than the 1@xmath15 value expected from a naive formal charge concept , but is roughly consistent with the theoretically calculated value 0.2@xmath15 using a correlated band theory lda+u method.@xcite as a result , even though there is small charge difference between the co ions , the charge - disproportionation is accompanied by local moment formation and the spins are consistent with the formal valences co@xmath16 and co@xmath17.@xcite the discovery of an unexpected insulating state in na@xmath3coo@xmath1 ( n@xmath3co ) and hydration - induced superconductivity has stimulated the study of isostructural and isovalent family a@xmath0coo@xmath1 ( a = k , rb , cs ) . in spite of a few attempts to produce superconductivity in hydrated k@xmath0coo@xmath1 ,
the amount of intercalated water is 0.8 or less , forming only a monohydrate ( k+h@xmath1o ) layer and no superconductivity has been detected yet.@xcite the k@xmath0coo@xmath1 system has been known for three decades , since hagenmuller and colleagues reported@xcite structure , transport , and magnetic data on phases with @xmath4=1.0 , 0.67 , and 0.50 .
recently , an insulating phase in k@xmath3coo@xmath1 ( k@xmath3co ) has been studied in more detail by a few groups;@xcite nakamura _
et al._@xcite in the mid-1990s had reported an almost temperature - independent resistivity well above a metallic value . in k@xmath3co , using nmr and neutron diffraction studies , watanabe _ et al .
_ observed similar temperature evolution as in n@xmath3co.@xcite at @xmath18 k , a kink in the in - plane susceptibility @xmath19 indicates onset of antiferromagnetic ordering .
the resistivity increases sharply at @xmath2 k , signaling the charge - ordering . at this temperature
, there is an additional magnetic rearrangement , indicated by kinks in both @xmath19 and @xmath20 . from @xmath21sr experiments sugiyama _ et al . _
have obtained similar transition temperatures , 60 and 16 k , in metallic k@xmath22co.@xcite the former is a magnetic ordering temperature from a paramagnetic state . based on a mean field treatment of a hubbard model
, they suggested there may be a linear spin density wave ( sdw ) state between 16 and 60 k , while a commensurate helical sdw state exists below 16 k. additionally , k or na ions order ( structurally ) , resulting in formation of a 2@xmath23@xmath24 supercell at t@xmath25 and 470 k for k and na ions , respectively.@xcite the tiny energy gap of similar magnitude with n@xmath3co has been observed by qian _
et al . _ with arpes measurements.@xcite
several characteristics of n@xmath0co , in particular the superconductivity upon hydration and effects of cation ordering , suggest that the behavior in this system is sensitive to details of the electronic structure .
the fact that k@xmath3co is similar to n@xmath3co , yet shows clear differences in behavior , indicates that a comparison of the electronic structures of these systems is warranted . in this paper , we compare and contrast the two insulating systems k@xmath3co and n@xmath3co . here
correlation effects and detailed magnetic ordering are neglected , but the observed @xmath26 supercell including na / k zigzag ordering is adopted .
( @xmath27 is the hexagonal lattice constant . )
.crystal structure comparison between k@xmath0coo@xmath1 and na@xmath0coo@xmath1 at @xmath4=0.5 .
the orthorhombic structures ( space group : @xmath28 , no .
59 ) determined from na- or k - zigzag ordering are a @xmath26 superstructure which is based on the hexagonal structure with a lattice constant @xmath27 . in this structure ,
the oxygens have three site symmetries , two @xmath29 and one @xmath30 . here , @xmath31 is an oxygen height from the co layers .
a main difference in these structures is that k@xmath3coo@xmath1 has 12 % larger @xmath32 lattice constant .
the data are from ref .
@xcite for na@xmath3coo@xmath1 and ref .
@xcite for k@xmath3coo@xmath1 . [ cols="^,^,^,^",options="header " , ] [ table2 ] these differences between two cobaltates can be clarified in two ways .
first , we can determine the effects purely due to chemical difference ( k vs. na ) as opposed to the size difference leading to structural differences .
for this , n@xmath3co is assigned the same structure as in k@xmath3co and denoted n@xmath3co@xmath33 .
the resulting band structure enlarged near @xmath34 is compared with that of k@xmath3co in fig .
[ band3 ] . even in the identical structure ,
substantial differences on an important energy scale are evident .
the top valence band in n@xmath3co@xmath33 is 20 mev higher in energy at the @xmath35 point , although the @xmath36 bandwidth is about 5% smaller ( not shown ) .
another difference is that the projected k and na dos is almost identical ( and small , of course ) through most of the @xmath36 bands , except in a @xmath37 mev region at and below the fermi level , where the na projected dos ( pdos ) is 20 - 35% larger ( more than 50% larger at e@xmath38 ) .
these distinctions indicate that the differences in electronic structure are mainly due to k substitution itself rather than indirectly through the change in structure .
second , using the mullikan charge decomposition , we obtained atom - decomposed charges , which are displayed in table [ table2 ] .
the k ion is very noticeably more ionic than the na ion , consistent with the pdos difference mentioned just above .
the compensating charge is spread over the oxygen ions ; the co charges are essentially the same for k@xmath3co and n@xmath3co@xmath33 .
this higher ionicity of k seems to be the most discernible difference between these cobaltates .
it is still unclear what water does in the system .
the only unambiguously aspect is that hydration dramatically increases the @xmath32 lattice constant , resulting in more two - dimensionality of the electronic system.@xcite however , although the isostructural system na@xmath39tas@xmath1@xmath40h@xmath1o shows very similar change in the @xmath32 lattice constant when hydrated,@xcite @xmath41 k in this system is independent of @xmath42 .
this difference in behavior established that water has effects in the cobaltates that are not present in the transition metal disulfides and diselenides . in this respect
it is interesting that ( na@xmath43k@xmath44)coo@[email protected]@xmath1o shows superconductivity with @xmath46 k and about 7 increment in @xmath32 lattice constant from k@xmath47co , which is similar in amount to that of hydrated sodium cobaltate.@xcite
using a crystal structure recently reported , we have investigated at the lda level the differences in electronic structure between k@xmath3coo@xmath1 and n@xmath3coo@xmath1 .
comparison shows a few substantial differences near @xmath34 ; smaller @xmath36 bandwidth by 60 mev in k@xmath3coo@xmath1 , and additional fermi surfaces along the @xmath48 line which are almost flat near the @xmath49 point .
these differences are due more to chemical differences ( higher ionic character of k ) rather than to structural difference between the systems . an angle - resolved photoemission comparison of the three systems a@xmath0coo@xmath1 , a = na , k , and rb , has appeared,@xcite with the differences at equal doping levels being small almost too small to quantify .
unfortunately , samples at precisely @xmath4=0.5 were not the focus of that study . since the superstructure we have studied
is confined to @xmath4=0.5 , our results can not be compared with this data .
however , the structural disorder of the alkali at @xmath500.5 , which extends to the coo@xmath1 substructure , broadens the bands and hides small distinctions.@xcite this observation suggests that carrying out spectroscopic studies of both systems in the insulating phase at @xmath4=0.5 should be an excellent way to identify and characterize more precisely the effects of the different alkali cations .
we acknowledge m. d. johannes and d. j. singh for illuminating conversations , and d. qian for clarifying the arpes data .
this work was supported by doe grant de - fg03 - 01er45876 and doe s computational materials science network . w.e.p .
acknowledges the stimulating influence of doe s stockpile stewardship academic alliance program .
n. l. wang , d. wu , g. li , x. h. chen , c. h. wang , and x. g. luo , phys .
lett . * 93 * , 147403 ( 2004 ) ; j. hwang , j. yang , t. timusk , and f. c. chou , phys .
b * 72 * , 024549 ( 2005 ) ; s. lupi , m. ortolani , l. baldassarre , p. calvani , d. prabhakaran , and a. t. boothroyd , _ ibid . _ * 72 * , 024550 ( 2005 ) .
s. nakamura , j. ohtake , n. yonezawa , and s. iida , j. phys .
jpn . * 65 * , 358 ( 1996 ) . compared with the existing data for lattice constants ,
this sample does not seem to be exact @xmath9 phase , but @xmath51 .
this assumption may be supported by the curie - weiss behavior in the sample . | k@xmath0coo@xmath1 shares many similarities with na@xmath0coo@xmath1 , as well as some important differences ( no hydration - induced superconductivity has been reported ) . at @xmath2 k , k@xmath3coo@xmath1 becomes an insulator with a tiny optical gap as happens in na@xmath3coo@xmath1 at 52 k. this similarity , with a known common structure , enables direct comparisons to be made . using
the k - zigzag structure recently reported and the local density approximation , we compare and contrast these cobaltates at @xmath4=0.5 .
although the electronic structures are quite similar as expected , substantial differences are observed near the fermi level .
these differences are found to be attributable mostly to the chemical , rather than structural difference : although na is normally considered to be fully ion , k has somewhat more highly ionic character than does na in these cobaltates . |
in first - principles calculations , we expand the wave function with localized basis sets @xcite to handle a large number of atoms in the system ( typically @xmath28 1600 atoms ) .
a single-@xmath29 for hydrogen , a singe-@xmath29 plus polarization for silicon and a mixed basis set with a single and double-@xmath29 for @xmath30- and @xmath27-orbitals of carbon atom have been used , respectively @xcite .
the kleinman - bylander s fully separable nonlocal projectors @xcite are used in the norm - conserving pseudopotentials @xcite and the local density approximation @xcite is employed in setting up the exchange - correlation potential .
we have thoroughly tested our basis set with other calculation parameters to reproduce the atomic and electronic structures of sic , graphene , other model for epitaxial graphene studied in previous literatures @xcite , respectively .
we modeled the 4@xmath0-sic(0001 ) substrate in the simulation with four alternating silicon and carbon atomic layers .
hydrogen atoms are introduced to passivate the dangling bonds in bottom of the slab . on top of the si - terminated surface of 4@xmath0-sic(0001 ) , one , two , and three graphene layers are placed for the buffer layer , monolayer graphene , and bilayer graphene , respectively .
based on the low energy electron diffraction ( leed ) experiments @xcite , the large supercell with the @xmath4 ( in short @xmath5 ) periodicity ( equivalent to @xmath31 times graphene unit cell ) is imposed to the calculations .
the atomic positions are determined by total energy minimization calculations until the forces on each atom are less than 0.06 ev / while atoms belonging to the last two silicon and carbon layers are fixed to the bulk atomic structure of 4@xmath0-sic . in the electronic structure calculations after geometric optimization process
, we use @xmath32 k - point sampling in lateral directions and set very large size of vacuum ( 50 ) in surface normal direction to prevent the spurious dipolar interactions between terminated slab geometries in supercell configuration .
-sic(0001 ) slab model , ( b ) monolayer epitaxial graphene and ( c ) bilayer epitaxial graphene on top of the buffer layer with 4@xmath0-sic(0001 ) slab , respectively .
the spectrum are drawn along the high symmetric lines of the first brillouine zone of @xmath5 supercell .
the dotted red lines are obtained by the first - principles calculations on the atomic models including one ( two ) graphene layer , buffer layer , and 4@xmath0-sic(0001 ) slab .
the solid black lines are obtained by our minimal microscopic model hamiltonian within the tight - binding approximations ( eq .
it is found that the dense flat bands near the fermi levels are from the states isolated inside the approximate @xmath6 domains of the buffer layer and sic surface underneath it , which do not contribute to arpes intensities .
, width=321 ] during the high temperature thermal decomposition process @xcite , the detailed atomic structures and shape of the domain may vary from sample to sample .
however , the domain boundary of @xmath3-orbitals of the buffer carbon atoms and the exact @xmath5 periodicity remains the same as observed in many experiments @xcite and shown in our calculations @xcite .
hence , it is sufficient to approximate the buffer layer to the coarse - grained atomic configuration of the quasi-@xmath6 periodic connections of @xmath3-electrons only .
we also found from the first - principles calculations that the interactions between @xmath3-orbital states at the domain boundary of the buffer and ones in graphene play the most significant role to determine the electronic structures of the system @xcite .
it is also noticeable that electronic states of atoms inside the domain and those under the buffer layer have negligible contributions .
hence , based on results from the first - principles calculations , our tight - binding hamiltonian for monolayer epitaxial graphene on coarse - grained buffer layer model ( fig .
1(b)-(d ) ) can be written as @xmath11 where @xmath33 and @xmath34 are the nearest neighbour hopping amplitude between carbon atoms in graphene and those in the buffer layer respectively .
the @xmath14 and @xmath15 are annihilation operators for electron in graphene and buffer layer respectively .
@xmath35 denotes the interlayer hopping amplitude between the nearest neighbour carbon atoms belong to graphene and the buffer layer with the bernal type stacking respectively . @xmath36 and @xmath37 describe the potential for graphene and the buffer layer considering charge redistributions due to the polar sic surface .
these parameters were found to be enough for fitting our first - principles energy bands of the buffer layer and monolayer epitaxial graphene , respectively ( fig .
1s ( a ) and ( b ) ) while the second and third nearest neighbour interlayer interaction terms improve the agreements a little . for bilayer epitaxial graphene
, we introduce another interlayer hopping amplitude @xmath38 between the nearest neighbours carbon atoms belong to each graphene layer ( bernal type stacking ) and potential shifts for each graphene layer , @xmath17 and @xmath39 with respect to the fermi level .
the hamiltonian for the interaction is written as @xmath40 , where @xmath41 here , @xmath42 is the annihilation operator for electron in second graphene .
we fit the energy spectrum of bilayer epitaxial graphene obtained by our model hamiltonian to the first - principles calculation results ( fig .
1s ( c ) ) and found @xmath43 being quite similar to one of graphite , but smaller than 0.48 ev @xcite and 0.46 ev @xcite obtained in recent experiments .
as shown in fig .
1s , the agreements between energy spectrums obtained by the first - principles calculations and ones by our model hamiltonians are excellent .
the dense flat bands near the fermi level in fig .
1s from the first - principles calculations are found to originate from localized states inside approximate @xmath6 domains of the buffer layer and localized states in 4@xmath0-sic(0001 ) surfaces underneath the buffer .
we found that such localized states forming flat bands are buried deep inside the surfaces and do not contribute to the simulated arpes spectrum .
[ [ simulation - of - angle - resolved - photoemission - spectroscopy - arpes - intensities ] ] simulation of angle resolved photoemission spectroscopy ( arpes ) intensities ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ we use the fermi golden rule to simulate arpes spectra @xcite .
the transition probability ( @xmath44 ) from an initial bloch state ( @xmath45 ) to an outgoing electron state ( @xmath46 ) is written as @xmath47 where @xmath48 is the photon - electron interaction hamiltonian , @xmath49 is the frequency of incident photon , and @xmath50 is an energy of @xmath51 .
tight - binding wavefunction obtained for the system is given by , @xmath52\ ] ] where @xmath53 is an amplitude for the @xmath3-orbital located at @xmath54 in the @xmath4 supercell with unit vector @xmath55 and @xmath56 is the crystal momentum of electron . by using the dipole approximation , @xmath57
, the transition probability will be expressed as @xmath58 where @xmath59 is the momentum of outgoing electron and @xmath60 is the fourier transformation of atomic orbital @xmath61 , defined by @xmath62 ( set by constant ) .
the crystal momentum @xmath56 belong to the first brillouin zone of the @xmath63 supercell and satisfys momentum conservation condition in the surface parallel direction ( i.e. , @xmath64 for the reciprocal vector @xmath65 ) of the supercell .
we include the attenuation factor for the arpes intensity considering photoelectron mean free path ( @xmath195 ) to the surface normal direction .
the conservation of energy imposed by the @xmath66-function is replaced by lorenzian function , @xmath67 , with a broadening ( @xmath22 ) of 30 mev .
we simulate the arpes spectrum ( shown in figs . 2 and 4(b ) ) with a photon energy of 50 ev and the two - dimensional constant energy maps ( shown in fig . 3 and fig .
2s ) with a photon energy of 100 ev @xcite . | we present a minimal but crucial microscopic theory for epitaxial graphene and graphene nanoribbons on the 4@xmath0-sic(0001 ) surface protopypical materials to explore physical properties of graphene in large scale .
coarse - grained model hamiltonians are constructed based on the atomic and electronic structures of the systems from first - principles calculations . from the theory
, we unambiguously uncover origins of several intriguing experimental observations such as broken - symmetry states around the dirac points and new energy bands arising throughout the brillouin zone , thereby establishing the role of substates in modifying electronic properties of graphene .
we also predict that armchair graphene nanoribbons on the surface have a single energy gap of 0.2 ev when their widths are over 15 nm , in sharp contrast to their usual family behavior .
graphene has attracted immense interests because of the unusual relativistic energy dispersions with the chiral massless dirac fermions near the fermi level ( @xmath1 ) @xcite .
the direct observation of such peculiar quasiparticle spectra in graphene is particularly important not only for understanding its novel physical properties @xcite but also for practical applications @xcite .
so , several measurements through the high resolution angle resolved photoemission spectroscopy ( arpes ) are performed on a single layered epitaxial graphene lying on the silicon carbide ( 0001 ) surfaces @xcite .
surprisingly , the reported quasiparticle spectra reveal anomalous energy dispersions around the dirac energy point ( @xmath2 ) indicating highly renormalized bands @xcite or energy gap @xcite there .
many experimental @xcite and theoretical studies @xcite address these important problems but there is still no consensus on origins of the anomalous spectrum .
moreover , there is no clear understanding on other anomalous arpes observations such as the broken six - fold symmetry near the @xmath2 and new distorted hexagonal energy bands around brillouine zone ( bz ) center @xcite .
the complex interfacial structures arise when epitaxial graphene is grown by annealing sic surfaces @xcite . during the thermal decomposition of the surfaces , a layer of carbon atoms , called the buffer layer , forms first without exhibiting the typical linear energy dispersion of @xmath3-states near the @xmath1 @xcite .
then , on top of the buffer layer , the clean honeycomb lattice of carbon atoms grows @xcite .
the lattice mismatch between the sic(0001 ) surface , buffer layer and graphene gives rise to the large scale surface reconstruction with a periodicity of @xmath4 ( in short , @xmath5 ) with respect to the sic(0001 ) surface unitcell , which is observed in the low energy electron diffraction ( leed ) measurements @xcite .
the scanning tunnelling microscopy ( stm ) image , however , indicates an approximate @xmath6 periodicity @xcite of the reconstruction .
we note that such superperiodic rearrangements shall impose constraints on possible theoretical models to explain the anomalous electronic structures mentioned above . in this paper
, we show that the interactions between epitaxial graphene and the reconstructed layer underneath it are the main driving forces to several anomalous features observed in recent experiments @xcite on electronic properties of epitaxial graphene . from the simulated arpes spectra on the systems
, it is shown that the symmetry breaking at the @xmath2 and the new hexagonal bands throughout the two - dimensional bz have the same origins .
based on the model , we predict that graphene nanoribbons or finite size fragments of graphene @xcite show an homogeneous energy gap contrary to their family behavior @xcite . moreover , by extending our microscopic model for monolayer epitaxial graphene to bilayer one , we identify the effects of the buffer layer to its energy spectrum and characteristic energy gap @xcite .
our computational results indicate that the interplays between geometries and electronic structures are pivotal in altering global energy bands of graphene , notwithstanding that the many - body interactions @xcite are expected to play some roles in modifying the quasi - particle spectrum near the @xmath2 locally . from the _ ab - initio _ pseudopotential density functional method @xcite , we find that the carbon atoms in the buffer layer with superperiodic @xmath5 unitcell are split into lattice matched regions , where carbon atoms have @xmath7-bonds to silicon atoms of the 4@xmath0-sic(0001 ) surface , and their boundaries without the @xmath7-bonds @xcite ( fig .
1(a ) ) .
the carbon atoms at the boundaries of the lattice matched regions exhibit an approximate @xmath6 domain satisfying the geometric constriction imposed simultaneously by both leed and stm measurements . to explore various initial conditions for buffer layer formation , we shift the initial atomic coordinates of the buffer layer ( shown in fig .
1(b ) ) on top of the 4@xmath0-sic(0001 ) surface by either @xmath8 ( fig .
1(c ) ) or @xmath9 ( fig .
1(d ) ) where @xmath10 is an unit vector of graphene .
we find that the final relaxed atomic geometries for all initial coordinates are essentially same to each other except for minor differences in the connectivity of @xmath3-electrons along quasi-@xmath6 domain boundaries ( figs .
1(b)-(d ) ) .
we build up a minimal ( coarse - grained ) microscopic model for interactions between epitaxial graphene and the buffer layer based on atomic and electronic structures obtained from our first - principles calculation ( see detailed method in @xcite ) . from the first - principles calculations
, it is found that the interactions between @xmath3-orbital states at the domain boundary of the buffer and ones in graphene play the most significant role to determine the electronic structures of the system while the states of atoms inside the domain and those under the buffer layer have negligible contributions to the electronic structures near the @xmath1 and @xmath2 .
hence , it is sufficient to approximate the buffer layer to the coarse - grained atomic configuration of the quasi-@xmath6 periodic connections of @xmath3-electrons only ( thick black lines in figs .
1(b)-(d ) ) .
our tight - binding hamiltonian for monolayer epitaxial graphene on coarse - grained buffer layer model ( fig .
1b ) can be written as @xmath11 where @xmath12 ( 2.70 ev ) and @xmath13 ( 1.50 ev ) are the nearest neighbour hopping amplitude between carbon atoms in graphene and those in the buffer layer respectively .
the @xmath14 and @xmath15 are annihilation operators for electron in graphene and buffer layer respectively .
@xmath16 ( 0.30 ev ) denotes the interlayer hopping amplitude between the nearest neighbour carbon atoms belong to graphene and the buffer layer with the bernal type stacking respectively .
@xmath17 ( 0.35 ev ) and @xmath18 ( 0.34 ev ) describe the potential for graphene and the buffer layer considering charge redistributions due to the polar sic surface .
since our model hamiltonians are described within the single - orbital tight - binding approximation , we can readily extend our model to explore the electronic structures of finite - sized monolayer and multilayer epitaxial graphene respectively which are beyond the reach of the first - principles calculations .
-sic(0001 ) surface with @xmath5 supercell .
epitaxial graphene ( black ) is located on top of the buffer carbon atoms ( black and grey ) .
the buffer carbon atoms consisting @xmath6 domain are denoted by grey color and its boundary by black .
the silicon and carbon atoms in the sic are denoted by red and green respectively .
( b)-(d ) coarse - grained minimal microscopic models .
the red thick arrows indicate the @xmath6 supercell unit vectors and the blue ones @xmath5 unit vectors .
black lines represent the connectivity of @xmath3-electrons of the buffer carbon atoms with approximate @xmath6 domain boundaries and grey lines denote the hexagonal network of epitaxial graphene on top of the buffer . , width=8 ] we have found that the nearest neighbour inter- and intra - layer interactions between @xmath3-electrons in the coarse - grained atomic model ( eq .
( 1 ) ) are sufficient to reproduce the electronic energy bands obtained from the first - principles calculations ( fig .
2 ) ( for detailed comparisons , see @xcite ) .
we notice that inclusion of the nearest neighbour interlayer interaction between graphene and the coarse - grained buffer layer model already breaks the symmetry between two sublattices of graphene .
the simulated arpes spectrum for monolayer epitaxial graphene agrees well with both our previous results @xcite from the first - principles calculation and ones @xcite from experiments ( fig .
2 ) ( for detailed arpes simulation method , see @xcite ) . when the initial conditions for the buffer layer formation are varied and the resulting geometries for the approximated @xmath6 domain are slightly altered as shown in figs .
1(b)-(d ) , the simulated arpes spectrum display essentially same structures with energy gaps at the @xmath2 s and the midgap states as shown in figs 2(a)-(c ) .
hence , we confirm from the simple microscopic model that the sublattice symmetry - breaking interlayer interaction indeed opens a gap of @xmath190.20 ev at the @xmath2 and the presence of midgap states give rise to high arpes intensities inside the energy gap as well as the level repulsion between upper and lower dirac cones ( fig .
2(d ) ) . from simulated arpes spectrum for bilayer epitaxial graphene
, we find that the @xmath2 approaches to the @xmath1 and the energy gap decreases from 0.20 to 0.12 ev ( fig .
2(e ) ) , agreeing well with experimental observations @xcite due to the charge transfer between the sic surfaces and graphene , perpendicular electric fields exist on the multilayer epitaxial graphene opening an energy gap @xcite .
though the fingerprints of midgap states are hardly visible for the bilayer epitaxial graphene , it is shown that the interaction between the buffer and graphene flattens the characteristic mexican - hat - shaped band @xcite at the bottom of the upper dirac cone [ fig .
2(e)and fig .
1s ( c ) in @xcite ] and shifts the second subband in the lower cone downward by 20 mev ( fig .
2(e ) ) . by simulating the two - dimensional arpes intensity maps with various fixed energies ( fig .
3 ) , we find that the interactions between the buffer layer and graphene indeed induce the symmetry breaking near the @xmath2 and produce new hexagonal energy bands throughout the first bz of graphene .
our simulated fixed - energy intensity patterns give a good agreement with experimental results @xcite ( fig .
3 ) . due to the two - source interference between photo - excited electrons from two equivalent atomic sites ( _ la _ young s
double slit ) or helical nature of the carriers in graphene , the intensity patterns at the @xmath20-points show the typical crescent shape anisotropy @xcite . with the characteristic high intensity at the @xmath20-points ,
there exist six equivalent faint replicas around each @xmath20-point forming a smaller hexagon ( inset in fig .
3(a ) ) . when approaching @xmath2 , the highest intensities at the @xmath20-points become isotropic and three of six replicas become weaker compared to other three points ( inset in fig .
3(b ) ) .
the intensities of stronger three satellites amount to 7% of the main peak at the @xmath20-points while those of the weaker to 2% agreeing with experiment results qualitatively @xcite .
the replicas are connected by the second - shortest reciprocal lattice vectors of the @xmath5 supercell ( fig .
3(b ) ) and the area of the smaller hexagon nearby is @xmath21 of the first bz of graphene @xcite . when the interlayer interaction is set to zero intentionally in our simulation , we can not find any symmetry breaking phenomena .
thus , we conclude that apparent six - fold symmetry breaking near the @xmath20-points is due to the interaction between the buffer and graphene . together with six faint replicas ,
there are global faint features throughout the first bz ( fig .
3(a)-(c ) ) .
we also find that the larger hexagonal structure around the @xmath22-point observed in the experiments @xcite ( 25% of total area ) originates from the buffer layer . by simulating constant energy map ( fig .
3(d ) ) and arpes spectrum ( fig .
2s in @xcite ) of our coarse - grained buffer layer model , we show that the underlying faint features are reminiscent of the interlayer interactions between the buffer and graphene . usually , the surface of epitaxial graphene exhibits the finite - sized terrace patterns @xcite .
thus the quantum confinement effect @xcite may play a role in determining the energy gap @xcite . to illustrate such an effect , we calculate electronic structures of epitaxial graphene nanoribbons with armchair edges on both sides ( in short , eagnr ) on top of the buffer layer . following the convention @xcite , three families of eagnrs are denoted by the number of dimer lines , @xmath23 , i.e. , @xmath23-eagnr ( @xmath24 , @xmath25 , @xmath26 families , @xmath27 is a positive integer ) . by using the same model for the single layer epitaxial graphene ( fig .
4(a ) ) , we show that the eagnrs have a similar arpes spectrum to epitaxial graphene for energy gaps , midgap states , and level repulsion between upper and lower cones , respectively ( fig .
4(b ) ) . moreover ,
when the width of eagnr is over 15 nm , the typical family behaviour of energy gaps @xcite disappears completely and converges to the energy gap ( 0.20 ev ) of two - dimensional epitaxial graphene ( fig .
4(c ) ) .
hence , we can conclude that the terrace patterns or finite - sized epitaxial graphene exhibit essentially the same electronic structures of the ideal two dimensional one . on the other hand ,
the present calculation results indicate that epitaxially grown graphene nanoribbons on the sic surface will have the homogenous energy gap if the width is over 15 nm . in summary ,
we have constructed the microscopic theory for expitaxial graphene on 4@xmath0-sic(0001 ) surface incorporating interactions between graphene and the surface .
the simulated experimental observations based on the theory have been shown to explain the several atypical aspects of epitaxial graphene from a single and unified view and , thus , shed light on understanding the quasiparticle spectrum of graphene in various circumstances .
s. k and j. i acknowledge the support of the kosef through the src program ( center for nanotubes and nanostructured composites ) .
h. j. c. was supported by the krf ( krf-2007 - 314-c00075 ) and by the kosef grant no .
r01 - 2007 - 000 - 20922 - 0 .
y .-
w .
s. was supported by quantum metamaterials research center no .
r11 - 2008 - 053 - 01002 - 0 and nano r&d program 2008 - 03670 through the kosef funded by the korean government ( mest ) .
computational resources have been provided by kisti ( ksc-2008-s02 - 0004 ) and the kias linux cluster system . |
one of the main problems in the field of primary cosmic - ray investigations is the form of the energy spectrum and the precise determination of the mass composition .
precise information about the energy spectrum and mass composition of primary cosmic - rays at vhe ( @xmath1 10@xmath3 ev ) energies is an important input for understanding the origin of cosmic - rays and their propagation through the galaxy .
cosmic - rays exhibit a nearly smooth power law energy spectrum over more than 10 decades of energy ( from @xmath4 10@xmath5 to 10@xmath6ev ) with significant kinks at @xmath4 10@xmath7 ev ( knee ) and @xmath4 10@xmath8ev ( ankle)[1,2 ] .
the knee feature at @xmath4 10@xmath7ev , which involves a change in the differential spectral exponent from @xmath4 2.7 to @xmath4 3.1 has defied a clear explanation so far . among the possibilities , one of the reasons suggested for the knee feature involves a change in the composition of the primary cosmic - rays from a mainly proton dominated(@xmath4 90% protons ) to a mixed ( protons + heavy nuclei ) composition or even heavy nuclei domination[2,3 ] .
thus , it becomes very important to measure the cosmic - ray composition in the vhe range , especially at energies around 10@xmath9 - 10@xmath7ev .
although primary cosmic - ray composition is known more or less precisely at energies up to @xmath4 10 tev from direct satellite / balloon based observations[4,5 ] , the low flux of primary cosmic - ray particles at energies @xmath1 10@xmath10 ev ( @xmath4 10@xmath11particles cm@xmath11 s@xmath13)does not allow direct measurement of primary composition with detectors carried on balloons and satellites .
attempts have , therefore , been made to infer the cosmic - ray composition indirectly from data collected with large air shower arrays and atmospheric cerenkov telescopes which detect primary cosmic - rays indirectly by recording the secondary particles and photons produced in the atmospheric cascades initiated by these primaries[6,7,8,9 ] .
these attempts have achieved limited success so far because of the difficulty in relating the observed cascade parameters to the primary mass . in the present communication
, we make an attempt to show that one of the fractal / wavelet parameters derived from the cerenkov images recorded by an imaging cerenkov telescope ( like tactic ) exhibits an unprecedented efficiency for separating primary cosmic - rays in terms of their mass .
this makes it possible to estimate the flux of cosmic - ray primaries at energies @xmath1 10@xmath10ev at large zenith angles with ground - based atmospheric cerenkov telescope systems .
primary cosmic - rays ( protons , hadrons , photons etc . ) with energy @xmath14ev , initiate extensive air showers in the atmosphere , generating in the process a large number of secondary particles ( @xmath15 , hadrons , photons ) some of which survive absorption in the atmosphere and can be detected at ground level . a large number of @xmath16 , traveling with relativistic speeds , generate cerenkov photons in the atmosphere which arrive in the form of a few ns duration optical pulse ( @xmath17@xmath18 ) spread over a radius of @xmath4 100 m at ground level[6,7,10 ] .
an imaging cerenkov telescope , comprising a large light collector with a matrix of photomultiplier tube ( pmt ) detectors in its focal plane , generates a two - dimensional picture of the resultant cerenkov photon distribution ( called the cerenkov image ) , the geometrical features of which depend upon the nature ( mass ) of the cosmic - ray primary which initiated the cascade . the technique has been successfully employed by various imaging cerenkov telescopes including tactic to detect vhe gamma - ray signals from a number of galactic and extragalactic sources[7,8 ] .
the tactic ( tev atmospheric cerenkov telescope with imaging camera ) telescope at mt.abu ( 24.6@xmath19 n ; 72.0@xmath19 e ; 1300 m asl ) comprises a 3.6 m aperture , tessellated light collector of overall parabolic shape mounted on an altitude - azimuth mount and equipped with a 349-pixel imaging camera at its focal plane [ 10,11 ] .
the imaging camera provides a field of view of 3@xmath19 radius with a uniform pixel resolution of 0.31@xmath19 . using a topological trigger based on a coincidence(@xmath20 @xmath4 20ns ) between nearest - neighbour , non - collinear triplets ( 3nct ) in the central 225 pixels ( 15@xmath2115 ) , the imaging telescope records cerenkov images of extensive air showers initiated by primary gamma - rays of @xmath1 1.5 tev and primary hadrons of energy @xmath1 3 tev from the vertical direction[10 ] .
the images are composed of cerenkov photon intensity observed by each pixel in the form of charge to digital converter counts(cdc ) . the effective collection area ( area of cerenkov light pool at ground level ) of the system has been estimated to be @xmath4 3@xmath2210 @xmath23 @xmath24[17 ] . both the threshold detection energy and the effective detection area increase with the zenith angle of observation[10,12 ] .
the results presented in this paper are based on the analysis of simulated cerenkov images produced by several primary species like gamma - rays , protons , ne and fe nuclei .
the simulations have been performed for the 349@xmath2pixel imaging camera of the tactic telescope using the corsika ( version 6.51 ) air shower simulation code[14,15,16 ] , supplemented by appropriate custom built software to take into account atmospheric absorption , reflection at the tactic light collector surface , pmt characteristics and the night sky background induced shot - noise fluctuations[7,13 ] . for each primary species 6,00,000 showers in different energy ranges ( 2tev , 4tev , 5tev ,
8tev , 10tev , 15tev , 20tev , 30tev ) with energy spectral index of -2.5 , -2.76 , -2.64 and -2.60 were simulated for gamma - rays , protons , neon and fe respectively .
all primary particles were assumed to be incident isotropically within the @xmath4 4@xmath19diameter field of view of the trigger pixels along a zenith direction of 50@xmath19 .
the large value of zenith angle was deliberately chosen to obtain a larger effective area of detection with a corresponding increase in the energy threshold as mentioned earlier .
independent simulations of the tactic telescope carried out earlier have shown that the detection threshold for gamma - rays at @xmath1 45@xmath19 zenith angle increases to @xmath4 10 tev ( 2 tev for vertical ) while the effective collection area increases to @xmath4 10@xmath5 @xmath24[12,17 ] . in this simulation study , we have also evaluated the effective collection area for 5tev gamma - rays for the tactic telescope at zenith direction of 50@xmath19 as @xmath4 7.07@xmath21 10@xmath23 @xmath24 .
all the simulated cerenkov images were subjected to similar cleaning and parameterization procedure in order to derive the hillas image parameters [ 18 ] and the recently introduced fractal and wavelet parameters as described briefly below .
the simulated cerenkov images recorded at core distances varying from 10 - 400 m have been parameterized following the well - known second order moment analysis of the derived photo - electron counts in the image pixels .
hillas shape parameters[18 ] , namely length ( l ) , width(w ) , azwidth ( a ) , distance ( d ) , miss ( m ) , orientation angle(@xmath25 ) and size ( s ) were determined for each simulated image .
the derived values were sorted as an independent set , for core distances @xmath26 400 m for a comparative analysis . the longitudinal and lateral development of air showers , represented by the height of shower maximum , the number of relativistic electrons and muons , the angular and spatial distribution of cerenkov photons and their total number varies with the primary mass .
this produces differences in the structure of cerenkov images which can be recognized through pattern recognition techniques .
fractal and wavelet parameters quantify the cerenkov image structure on different scale lengths , the former characterizing the global structure of photon distribution in the image while the latter characterize more localized structures[19,20,24,25 ] . the basic motivation and
the technique employed for fractal and wavelet analysis of cerenkov images has been described in detail in[21 - 25 ] .
we have followed an essentially similar procedure to calculate fractal moments and dimensions as well as wavelet moments and dimensions of different order .
there has been increasing interest in wavelet based approach in the recent years due to its success in several applications .
the wavelet transform is a two parameter transform[26 - 28 ] .
the two parameter domains of the wavelet tranform are dilation ( scaling or compression ) and translation ( space ) .
the advantage of using wavelets in the signal or structure analysis is the ability to perform local analysis of the signal or structure i.e to zoom on any interval of time and space . as such wavelet analysis has the ability to reveal the subtle details of the data which other analysis techniques fail to detect .
wavelet analysis is regarded as the sequence of versatile filtering approach which will allow to examine the very minute stuctures left after filtering of the images at different scale lengths i.e to zoom at any interval of frequency , time or space of the signal and enables us to detect some hidden aspect of the data or image and brings out corrections to the image analysis .
the pattern recognition technique for gamma - hadron separation employs fractal and wavelet analysis of cerenkov images to exploit differences in the structure of the recorded images due to differences in the longitudinal and lateral development of showers initiated by the two primary species .
showers initiated by primary gamma - rays and protons produce different numbers of relativistic electrons and muons and also have different heights ( and lateral extents ) of shower maximum .
these differences lead to structures in the cerenkov image on different scale lengths which can be parameterized through multifractal and wavelet dimensions and moments of different order[19 - 24 ] .
the usual procedure for calculating these parameters involves dividing the cerenkov image into m = 4 , 16 , 64 , 256 equally sized , non - overlapping cells and finding the number of cdc counts in each cell ( 6.5 cdc = 1 photo - electron ) .
the multifractal moments of order q(q = 2@xmath2 6 ) are calculated from the following relation : @xmath27^{q } \end{aligned}\ ] ] where n is the total number of cdc counts in the image , @xmath28 is the cdc count in the j@xmath29 cell and q= 2 , 3 , 4 , 5 , 6 is the order of fractal moment .
the wavelet moments are obtained as follows : @xmath30^{q } \end{aligned}\ ] ] the fractal scale length is given by @xmath31 = @xmath32 m both the fractal moment @xmath33 as well as the wavelet moment @xmath34 exhibit a proportionality with the scale length which can be expressed as , @xmath35 the fractal and wavelet dimensions of order q are then determined from the following relations : @xmath36 where @xmath37 and @xmath38 are the slopes of the line between natural logarithm with scale length @xmath39 of @xmath33(m ) and @xmath34(m ) respectively . since wavelet dimensions are more sensitive to local structures in the photon distribution in cerenkov images , they are expected to have better efficiency for gamma - hadron separation as proton - initiated showers are known to have a non - uniform structure due to contributions from pions and muons[20 - 25 ] .
fig.1 shows a composite plot of the frequency distribution of image shape parameters derived for showers initiated by 5 - 20 tev gamma - rays , protons , ne and fe primaries with shower cores located at @xmath26 400 m from the tactic imaging telescope .
we find that the distributions for the shape parameters length and width ( l , w ) show a clear distinction between gamma - ray , proton and ne / fe induced showers , especially a clear distinction between gamma - ray and hadron initiated showers . while on the other hand , the distributions for the alpha(@xmath25 ) , distance(d ) and miss(m ) parameters do not exhibit any substantial dependence on the primary mass indicating that these parameters can have a limited application in primary mass segregation .
.parameter domains and retention factors of different primary species for cerenkov events at @xmath26 400 m ; a - accepted , r - rejected [ cols="<,^,^,^,^,^ " , ]
the cerenkov imaging telescope tactic has been used to search for tev gamma - ray signals from several galactic and extragalactic sources and has successfully detected emission from the crab nebula , mkn 421 and mkn 501[29,30,31 ] .
the telescope is also used to monitor an off - source or background region where it tracks a region of sky away from the source direction in order to provide an independent measure of the cosmic - ray background rate and also to validate the gamma - ray signal retrieval methodology .
the fractal dimensions obtained from these cerenkov images showed no potential in terms of segregation of events on the basis of mass because of significant overlaps of distributions due to large background contribution .
these simulated event values also yield a cosmic - ray proton composition of 87% which is close to the experimentally measured proton abundance of @xmath4 89% at tev energies .
the gamma - ray to proton ratio determined from these values is also in close agreement with the value of 0.001 at tev energies reported from other atmospheric cerenkov telescope systems[32 ] while as the protons out number other species by few orders of magnitude .
we have also subjected cerenkov images of various cosmic - ray primaries recorded by tactic during off - source observation runs to the above discussed fractal and wavelet analysis to investigate whether the @xmath0 parameter distribution leads to the expected cosmic - ray mass composition .
fig.4 shows the frequency distribution of parameter @xmath0 for both simulated and experimentally observed cerenkov images recorded by the tactic telescope during a 37h long off - source exposure along a direction making a zenith angle of @xmath4 50@xmath19 .
the total number of events recorded during crab - off source scans was 2,76,900 corresponding to the event rate of @xmath4 2.0 hz , which translates into detection threshold energy of 4 tev for cosmic - ray proton initiated events(@xmath4 2 tev for gamma - ray initiated events ) as mentioned in the simulation process . while 2,35,635@xmath40485 events fall within the proton domain @xmath0([email protected] ) ; 1,788 @xmath40 39 events are falling in ne domain with @xmath0= ( [email protected] ) ; 1,538 @xmath40 34 events are recorded in fe domain with @xmath0 = ( [email protected] ) while as 127 @xmath40 9 events fall in the gamma domain with @xmath0=([email protected] ) . taking into account the exposure time ( 37h ) , tactic effective area 7.07@xmath21
10@xmath23 @xmath24 and the tactic field of view 3.8@xmath21 10@xmath41sr , the flux for various cosmic - primaries such as p , ne , fe and gammas above 5 tev within 40% systematic error are found to be ( [email protected])@xmath21 10@xmath41m@xmath11s@xmath13 sr@xmath13 , ( [email protected])@xmath21 10@xmath42m@xmath11 s@xmath13 sr@xmath13 , ( [email protected])@xmath21 10@xmath42m@xmath11 s@xmath13sr@xmath13 and ( [email protected])@xmath2110@xmath43m@xmath11 s@xmath13 sr@xmath13 respectively[29 ] .
the results obtained show good agreement between the simulated and experimental data .
wavelet dimension distributions of cerenkov images generated by gammas - rays and hadrons ( p , ne , fe),scaledwidth=100.0% ] wavelet dimension distribution of simulated and experimental data , scaledwidth=100.0% ]
a number of attempts have been made earlier to assess the potential of imaging atmospheric cerenkov telescopes for separating different components of primary cosmic - rays in the energy region above @xmath4 10 tev[32,33,34 ] .
it has been shown that , while hillas image parameters of atmospheric cerenkov events recorded in an imaging cerenkov telescope are able to separate protons from all other nuclei , the image parameters are rather inefficient in segregating nuclei of different masses .
our analysis confirms this fact even though we find that the image parameters like width and azwidth have the potential to separate the primaries into groups of light ( protons and ne ) and heavy ( fe ) nuclei .
several workers [ 34,35 ] have also investigated the efficiency of the temporal characteristics of atmospheric cerenkov pulses for primary mass segregation .
it has been reported that while some simple pulse shape parameters , like pulse rise time , full - width at half maximum and pulse decay time , carry information useful for cosmic - ray mass composition studies , the efficiency for primary mass segregation can not be quantified in view of the limited simulation database
. a detailed feasibility study of cosmic - ray composition measurements with cerenkov telescopes using fractal parameters of the recorded cerenkov images has been presented earlier[23 ] .
it has been shown that the mass sensitivity of all the parameters increases with the image size , possibly because of the decreasing event by event fluctuations with increasing primary energy(@xmath4 size ) . however , a separation in more than two or three mass groups seems prohibited by the relatively poor mass resolution .
our results show that for primary hadrons of different masses with individual energies pre - selected to produce different amounts of cerenkov light in the atmosphere , very efficient mass segregation is produced by the wavelet dimension @xmath0 .
as already noted , we find that the frequency distributions of @xmath0 values for the three primary nuclear species ( p , ne and fe ) , considered by us in the simulations , peaks at distinctly separate values with negligible mutual overlap . to conclude , we have shown from our preliminary simulation - based study , that cerenkov image shape parameters are able to separate primary cosmic - ray events into two or three groups of nuclei in terms of their mass but are inefficient in segregating the events in terms of primary mass on an event by event basis .
however , we find that the wavelet dimension parameter @xmath0 of the cerenkov image , has immense potential for event by event segregation(21,22,23 ) of the recorded events in terms of the primary mass .
the author is grateful to ramesh koul , head astrophysical sciences division , barc , mumbai for his encouragement and overall support during the course of this work .
thanks are also due to dr.a.k.razdan for his help in developing the software used in this work and to dr.k.k.yadav for the help rendered during the course of data analysis .
b.wiebel-sooth , p.l.biermann , 772 ( 1998 ) , max - planck - institute , radioastronomie ( germany ) j.w.fowler et al . ,
astropart.phys . , 15 ( 2000 ) , 1 , 25 s.p.wordy et al .
, astropart.phys . , 18 ( 2002 ) , 129 k.asakimori et al . ,
astrophys.j , 502 ( 1998 ) , 278 a.v.aparasenko et al . ,
proc.20@xmath29 icrc , utah . , 3 ( 1999 ) , 167 r.glasstetter et al . , ( kascade ) , proc.@xmath44 icrc , durban , 4 ( 1997 ) , 157 j.a hinton , w.hofmann , ann.rev.of astronomy and astrophysics , 47 ( 2009 ) , 523 - 565 rene a. ong , physics report , 305 ( 1998 ) , 3 , 93 t.c.weekes et al .
, asrophysical journal , 342 ( 1989 ) 379 - 395 a.k.tickoo , ph.d thesis , university of mumbai , india , ( 2002 ) r. koul et al . , nucl.instrum . and meth.a , 578 ( 2007 ) , 548 a.konopelko et al
. , j.phys.g : nucl.part.phys,25 ( 1999 ) 1989 - 2000 k.k.yadav , ph.d thesis , university of mumbai , india , ( 2010 ) d.heck , et al . , astropart.phys .
, 12 ( 1999 ) , 145 j.n.capdeville et al .
, the karlsruhe eas simulation code corsika , ( 1999 ) , report no .
kfk4998 d.heck et al . , fzka - report 6019 , forschunszentrum karlsruuhe , ( 1998 ) m.k.koul et al . , nucl.instrum . and meth.a , 30 ( 2011 ) , 646 a.m.hillas . , space science reviews , 75 , 1 - 2 ( 1996),17 - 30 b.b.mandelbrot , j.fluid mech . , 62 ( 1994 ) , 331 a.ahaony , physics a , 168 ( 1990 ) , 479 c.k.bhat , astropart.phys . , 34 ( 2010 ) , 4 , 230 c.k.bhat et al . , 16@xmath29 nsss ( india ) , 10 ( 2010 ) , 74 , 145 a.haungs , j.knapp , i.bond , r.pallassini , 27@xmath29 icrc , hamburg , 7 ( 2001 ) , 2910 , r.w.lessard , l.cayon , g.sembroski , j.a.gaidos , astropart.phys . , 17 ( 2002 ) , 4 , 427 i.daubechies , comm.pure appl.math , 41 ( 1988 ) , 909 s.s.law et al . , engineering structures , 27 , 9 , ( 2005),1339 - 1348 robi.polikar . , the wavelet tutorial part iii , ( 2004 ) , http://www.cs.ucf.edu/courses/ cap5015/wtpart3.pdf , 1 - 28 a.haungs et al . , proceedings of 27@xmath29 icrc , hamburg,(2001 ) , 315 k.k.yadav et al .
, astropart .
phys . , 27 ( 2007 ) , 447 .
et al . , j.phys.g : nucl.part.phys . 37 ( 2010 ) , 12 , 125 s.v.godambe et al . , j.phys.g : nucl.part.phys , 35 ( 2008 ) , 065202 a.k.konopelko et al . , atropart.phys . ,
10 ( 1999 ) , 275 a.k.konopelko et al . ,
25@xmath29 icrc , durban , ( 1997),og.1-,5 , 3 , 349 r.c.rannot et al . , nuclear physics b(proc.suppl ) , 52b ( 1997 ) , 269 m.d.sinnot et al . , 26@xmath29 icrc , utah , ( 1999 ) , og.1.2.36 | a preliminary flux estimate of various cosmic - ray constituents based on the atmospheric cerenkov light flux of extensive air showers using fractal and wavelet analysis approach is proposed . using a monte - carlo simulated database of cerenkov images recorded by the tactic telescope
, we show that one of the wavelet parameters ( wavelet dimension @xmath0 ) provides @xmath1 90% segregation of the simulated events in terms of the primary mass .
we use these results to get a preliminary estimate of primary flux for various cosmic - ray primaries above 5 tev energy .
the simulation based flux estimates of the primary mass as recorded by the tactic telescope are in good agreement with the experimentally determined values .
cosmic@xmath2ray composition@xmath2cosmic rays@xmath2gamma rays |
any theory based on classical concepts , such as locality and realism , predicts bounds on the correlations between measurement outcomes obtained in space - separation @xcite .
these bounds are known as bell inequalities ( see @xcite for reviews ) .
profoundly , the correlations measured on certain quantum states violate bell inequalities , implying incompatibility between the quantum and classical worldviews . which are these non - classical states of quantum mechanics ?
here , we present a tool which allows one to extend the class of non - classical states , and gives further evidence that there may exist many - particle entangled states whose correlations admit a local realistic description . despite their fundamental role , with the emergence of quantum information @xcite
, bell inequalities have found practical applications .
quantum advantages of certain protocols , like quantum cryptography @xcite or quantum communication complexity @xcite , are linked with bell inequalities .
thus , new inequalities lead to new schemes . as an example , we present communication complexity problem associated with the new multisetting inequality .
specifically , based on a geometrical argument by ukowski @xcite , a bell inequality for many observers , each choosing between arbitrary number of dichotomic observables , is derived .
many previously known inequalities are special cases of this new inequality , e.g. clauser - horne - shimony - holt inequality @xcite or tight two - setting inequalities @xcite . the new inequalities are maximally violated by the greenberger - horne - zeilinger ( ghz ) states @xcite .
many other states violate them , including the states which satisfy two - settings inequalities @xcite and bound entangled states @xcite .
this is shown using the necessary and sufficient condition for the violation of the inequalities . finally , it is proven that the bell operator has only two non - vanishing eigenvalues which correspond to the ghz states , and thus has a very simple form .
this form is utilized to show that quantum states with positive partial transposes @xcite with respect to all subsystems ( in general the necessary but not sufficient condition for entanglement @xcite ) do not violate the new inequalities .
this is further supporting evidence for a conjecture by peres that positivity of partial transposes could lead us to the existence of a local realistic model @xcite .
the paper is organized as follows . in section
ii we present the multisetting inequality . in section
iii the necessary and sufficient condition for a violation of the inequality is derived , and examples of non - classical states are given .
next , we support the conjecture by peres in section iv , and follow the link with communication complexity problems in section v. section vi summarizes this paper .
consider @xmath0 separated parties making measurements on two - level systems .
each party can choose one of @xmath1 dichotomic , of values @xmath2 , observables .
in this scenario parties can measure @xmath3 correlations @xmath4 , where the index @xmath5 denotes the setting of the @xmath6th observer .
a general bell expression , which involves these correlations with some coefficients @xmath7 , can be written as : @xmath8 in what follows we assume certain form of coefficients @xmath7 , and compute local realistic bound as a maximum of a scalar product @xmath9 .
the components of vector @xmath10 have the usual form : @xmath11 where @xmath12 denotes a set of hidden variables , @xmath13 their distribution , and @xmath14 the predetermined result of @xmath6th observer under setting @xmath15 .
the quantum prediction for the bell expression ( [ general_bell ] ) is given by a scalar product of @xmath16 .
the components of @xmath17 , according to quantum theory , are given by : @xmath18 where @xmath19 is a density operator ( general quantum state ) , @xmath20 is a vector of local pauli operators for @xmath6th observer , and @xmath21 denotes a normalized vector which parameterizes observable @xmath15 for the @xmath6th party .
assume that local settings are parameterized by a single angle : @xmath22 .
in the quantum picture we restrict observable vectors @xmath21 to lie in the equatorial plane : @xmath23 take the coefficients @xmath7 in a form : @xmath24 with the angles given by : @xmath25 the number @xmath26 is fixed for a given experimental situation , i.e. @xmath1 and @xmath0 , and equals : @xmath27_2 [ n]_2 + 1 , \label{eta}\ ] ] where @xmath28_2 $ ] stands for @xmath29 modulo @xmath30 .
the local realistic bound is given by a maximal value of the scalar product @xmath9 .
the maximum is attained for deterministic local realistic models , as they correspond to the extremal points of a correlation polytope .
thus , the following inequality appears : @xmath31 where we have shortened the notation @xmath32 . since @xmath33 and the predetermined results , @xmath34 , are real , the right - hand side of this inequality can be written as : @xmath35 moreover , since inequality ( [ ineq_deter ] ) involves the sum of all possible products of local results respectively multiplied by the cosines of all possible sums of local angles , the right - hand side can be further reduced to involve the product of sums : @xmath36 inserting the angles ( [ angles ] ) into this expression results in : @xmath37 where the factor @xmath38 comes from the term @xmath39 in ( [ angles ] ) , which is the same for all parties .
one can decompose a complex number given by the sum in ( [ re_angles ] ) into its modulus @xmath40 , and phase @xmath41 : @xmath42 we maximize the length of this vector on the complex plane .
the length of the sum of any two complex numbers @xmath43 is given by the law of cosines as @xmath44 , where @xmath45 is the angle between the corresponding vectors .
to maximize the length of the sum one should choose the summands as close as possible to each other .
since in our case all vectors being summed are rotated by multiples of @xmath46 from each other , the simplest optimal choice is to put all @xmath47 . in this case one
has : @xmath48 where the last equality follows from the finite sum of numbers in the geometric progression ( any term in the sum is given by the preceding term multiplied by @xmath49 ) .
the denominator inside the modulus can be transformed to @xmath50 $ ] , which reduces to @xmath51 .
finally , the maximal length reads : @xmath52 where the modulus is no longer needed since the argument of sine is small .
moreover , since the local results for each party can be chosen independently , the maximal length @xmath53 does not depend on particular @xmath6 , i.e. @xmath54 .
since @xmath55 is a positive real number its @xmath0th power can be put to multiply the real part in ( [ re_angles ] ) , and one finds @xmath9 to be bounded by : @xmath56^{-n } \ ! \ ! \ ! \ !
\cos \left ( \frac{\pi}{2 m } \eta + \phi_1 + ... + \phi_n \right),\ ] ] where the cosine comes from the phases of the sums in ( [ re_angles ] ) .
these phases can be found from the definition ( [ vector ] ) . as only vectors rotated by a multiple of @xmath46 are summed ( or subtracted ) in ( [ vector ] ) , each phase @xmath41 can acquire only a restricted set of values .
namely : @xmath57 with @xmath58 , i.e. for @xmath1 even , @xmath41 is an odd multiple of @xmath59 ; and for @xmath1 odd , @xmath41 is an even multiple of @xmath59 .
thus , the sum @xmath60 is an even multiple of @xmath59 , except for @xmath1 even and @xmath0 odd . keeping in mind the definition of @xmath61 , given in ( [ eta ] ) ,
one finds the argument of @xmath62 is always odd multiple of @xmath59 , which implies the maximum value of the cosine is equal to @xmath63 .
finally , the multisetting bell inequality reads : @xmath64^{-n }
\cos \left(\frac{\pi}{2 m } \right ) .
\label{inequality}\ ] ] this inequality , when reduced to two parties choosing between two settings each , recovers the famous clauser - horne - shimony - holt inequality @xcite . for higher number of parties ,
still choosing between two observables , it reduces to tight two - setting inequalities @xcite .
when @xmath0 observers choose between three observables the inequalities of ukowski and kaszlikowski are obtained @xcite , and for continuous range of settings ( @xmath65 ) it recovers the inequality of ukowski @xcite .
in this section we present a bell operator associated with the inequality ( [ inequality ] ) .
next , it is used to derive the necessary and sufficient condition for the violation of the inequality . using this condition
we recover already known results and present some new ones .
the form of the coefficients @xmath66 we have chosen is exactly the same as the quantum correlation function @xmath67 for the greenberger - horne - zeilinger state : @xmath68,\ ] ] where the vectors @xmath69 and @xmath70 are the eigenstates of local @xmath71 operator of the @xmath6th party . for this state
the two vectors @xmath72 and @xmath73 are equal ( thus parallel ) , which means that the state @xmath74 maximally violates inequality ( [ inequality ] ) .
the value of the left hand side of ( [ inequality ] ) is given by the scalar product of @xmath73 with itself : @xmath75 using the trigonometric identity @xmath76 one can rewrite this expression into the form : @xmath77.\ ] ] as before , the second term can be written as a real part of a complex number .
putting the values of angles ( [ angles ] ) one arrives at : @xmath78 note that @xmath79 is a primitive complex @xmath1th root of unity . since all complex roots of unity sum up to zero the above expression vanishes , and a maximal quantum value of the left hand side of ( [ inequality ] ) equals : @xmath80 if instead of @xmath81 one chooses the state @xmath82 $ ] , for which the correlation function is given by @xmath83 , one arrives at a minimal value of the bell expression , equal to @xmath84 , as the vectors @xmath72 and @xmath85
are exactly opposite .
since we take a modulus in the bell expression , both states lead to the same violation . the bell operator associated with the bell expression ( [ inequality ] )
is defined as : @xmath86 its average in the quantum state @xmath19 is equal to the quantum prediction of the bell expression , for this state
. we shall prove that it has only two eigenvalues @xmath87 , and thus is of the simple form : @xmath88 both operators @xmath89 and @xmath90 are defined in the hilbert - schmidt space with the trace scalar product . to prove their equivalence one should check if the conditions : @xmath91 are satisfied .
geometrically speaking , these conditions mean that the `` length '' and `` direction '' of the operators are the same .
the trace @xmath92 involves the traces @xmath93 , which are the quantum correlation functions ( averages of the product of local observables ) for the ghz states , and thus are given by @xmath94 .
their difference doubles the cosine , which is then multiplied by the same cosine coming from the coefficients @xmath7 . thus the main trace takes the form : @xmath95 where the last equality sign follows from the considerations below eq .
( [ cos_square ] ) .
the middle trace of ( [ traces ] ) is given by @xmath96 , which directly follows from the orthonormality of the states @xmath97 .
the last trace of ( [ traces ] ) is more involved . inserting decomposition ( [ b ] ) into @xmath98 gives : @xmath99 ... { \rm
tr}[(\vec m_n \cdot \vec \sigma^n)(\vec m_n ' \cdot \vec \sigma^n ) ] \end{aligned}\ ] ] the local traces are given by : @xmath100 = 2 \vec m_n \cdot \vec m_n ' = 2 \cos(\phi_{m_n}^n - \phi_{m_n'}^n).\ ] ] thus , the factor @xmath101 appears in front of the sums .
we write all the cosines ( of sums and differences ) in terms of individual angles , insert these decompositions into @xmath98 , and perform all the multiplications .
note that whenever the final product term involves at least one expression like @xmath102 ( or for the primed angles ) its contribution to the trace vanish after the summations [ for the reasons discussed in eq .
( [ complex_roots ] ) ] .
moreover , in the decomposition of @xmath103 only the products of the same trigonometric functions appear . in order to contribute to the trace
they must be multiplied by again the same functions .
since the decompositions of cosines of sums only differ in angles ( primed or unprimed ) and not in the individual trigonometric functions , the only contributing terms come from the product of exactly the same individual trigonometric functions in the decomposition of @xmath104 and @xmath105 .
there are @xmath106 such products , as many as the number of terms in the decomposition of @xmath104 .
each product involves @xmath107 squared individual trigonometric functions .
each of these functions can be written in terms of cosines of a double angle , e.g. @xmath108 , and the last cosine does not contribute to the sum [ again due to ( [ complex_roots ] ) ] .
finally the trace reads : @xmath109 thus , equations ( [ traces ] ) are all satisfied , i.e. both operators @xmath89 and @xmath90 are equal .
only the states which have contributions in the subspace spanned by @xmath97 can violate the inequality ( [ inequality ] ) . _
necessary and sufficient condition for the violation of the inequality_. the expected quantum value of the bell expression , using bell operator , reads : @xmath110 .
\label{averaged_bell_op}\ ] ] the violation condition is obtained after maximization , for a given state , over the position of the @xmath111 plane , in which the observables lie .
an arbitrary state ( density operator ) of @xmath0 qubits can be decomposed using local pauli operators as : @xmath112 where the set of averages
@xmath113 $ ] forms the so - called correlation tensor .
the correlation tensors of the projectors @xmath114 are denoted by @xmath115 . using the linearity of the trace operation and the fact that the trace of the tensor product is given by the product of local traces , one can write @xmath116 in terms of correlation tensors : @xmath117 since each of the @xmath0 local traces @xmath118
, the global trace is given by : @xmath119 the nonvanishing correlation tensor components of the ghz states @xmath97 are the same in the @xmath120 plane : @xmath121 for @xmath0 even ; and are exactly opposite in the @xmath111 plane : @xmath122 with @xmath123 indices equal to @xmath124 and all remaining equal to @xmath29 .
inserting the traces ( [ tracing ] ) into the averaged bell operator ( [ averaged_bell_op ] ) one finds that the components in the @xmath120 plane cancel out , and components in the @xmath111 plane double themselves .
finally , the necessary and sufficient condition to satisfy the inequality is given by : @xmath125 where the maximization is performed over the choice of local coordinate systems , @xmath126 includes all sets of indices @xmath127 with 2@xmath128 indices equal to @xmath124 and the rest equal to @xmath29 , and @xmath129^{-n }
\cos \left(\frac{\pi}{2 m } \right)\ ] ] denotes the local realistic bound .
we now present examples of states , which violate the new inequality . as a measure of violation , @xmath130
, we take the average ( quantum ) value of the bell operator in a given state , divided by the local realistic bound : @xmath131 _ ghz state_. first , let us simply consider @xmath97
. for the case of two settings per side one recovers previously known results @xcite : @xmath132 for three settings per side the result of ukowski and kaszlikowski is obtained @xcite : @xmath133 for the continuous range of settings one recovers @xcite : @xmath134 in the intermediate ( unexplored before ) regime one has : @xmath135 for a fixed number of parties @xmath136 the violation increases with the number of local settings .
surprisingly , the inequality implies for the cases of @xmath137 and @xmath138 that the violation decreases when the number of local settings grows .
this behaviour is shown in the fig .
[ m_plot ] . , for the @xmath0-qubit ghz state.,scaledwidth=50.0% ]
the violation of local realism always grows with increasing number of parties .
_ generalized ghz state . _
consider the ghz state with free real coefficients : @xmath139 its correlation tensor in the @xmath111 plane has the following nonvanishing components : @xmath140 , and the components with 2@xmath128 indices equal to @xmath124 and the rest equal to @xmath29 take the value of @xmath141 ( there are @xmath142 such components ) .
thus , all @xmath106 terms contribute to the violation condition ( [ ns ] ) .
the violation factor is equal to @xmath143 . for @xmath144 and @xmath145
the violation is bigger than the violation of standard two - setting inequalities @xcite .
moreover , some of the states @xmath146 , for small @xmath147 and odd @xmath0 , do not violate _ any _ two - settings correlation function bell inequality @xcite , and violate the multisetting inequality . _ bound entangled state .
_ interestingly , the inequality can reveal non - classical correlations of a bound entangled state introduced by dr @xcite : @xmath148 with @xmath149 $ ] ( @xmath150 is an arbitrary phase ) , and @xmath151 being a projector on the state @xmath152 with `` 1 '' on the @xmath153th position ( @xmath154 is obtained from @xmath151 after replacing `` 0 '' by `` 1 '' and vice versa ) . as originally shown in @xcite this state
violates mermin - klyshko inequalities for @xmath155 .
the new inequality predicts the violation factor of : @xmath156 which comes from the contribution of the ghz - like state @xmath157 to the bound entangled state .
one can follow ref . @xcite and change the bell - operator ( [ bell_operator ] ) such that the state @xmath158 becomes its eigenstate .
the new operator , @xmath159 , is obtained after applying local unitary transformations @xmath160 to the operator ( [ bell_operator ] ) , i.e. @xmath161 . the violation factor of the new inequality is higher than ( [ viol - bes ] ) , and equal to : @xmath162 if one sets @xmath163 it appears that the number of parties sufficient to see the violation ( [ tilde_vis ] ) reduces to @xmath164 @xcite . on the other hand
the result of @xcite shows that the infinite range of settings further reduces the number of parties to @xmath165 . using the new inequality , @xmath166 settings per side suffice to already violate local realism with @xmath165 parties .
in this section it is shown that all the states with positive partial transpose with respect to all subsystems satisfy the multisetting inequality ( [ inequality ] ) .
this result further supports the conjecture by peres , that all such states can admit local realistic description @xcite .
first we briefly review partial transpositions , next present an inequality that all such states must satisfy , and finally compare it with the bell inequality ( [ inequality ] ) .
the partial transpose of an operator on a hilbert space @xmath167 is defined by : @xmath168 where the superscript @xmath169 denotes transposition in the given basis .
the positivity of partial transpose is found to be a necessary condition for separability @xcite .
the operator obtained by the partial transpose of any separable state is positive ( ppt - positive partial transpose ) . in the bipartite case of two qubits or qubit - qutrit system ,
the ppt criterion is also sufficient for separability . in the multipartite case
the situation complicates as one can have many different partitions into set of particles , for example four particle system @xmath170 can be split e.g. into @xmath171 or @xmath172 .
suppose one splits @xmath0 particles into @xmath173 groups , take as an example the split into three groups @xmath172 .
the state is called @xmath173-ppt if it has positive _ all possible _ partial transposes .
fortunately , positivity of partial transpose with respect to certain set of subsystems is the same as positivity with the respect to all remaining subsystems . in the example one
should check the positivity of operator obtained after transposition of subsystem @xmath174 , next @xmath30 , and finally @xmath175 .
all the @xmath173-ppt states were recently shown to satisfy the following inequalities @xcite : @xmath176 \le 2^{1-p},\ ] ] i.e. if @xmath74 appears in the first term within the trace , @xmath177 appears in the second term , and vice versa . omitting the positive factor
@xmath178 one arrives at the bell operator form : @xmath179 \big| \le 2^{1-p}. \label{ppt}\ ] ] following the conjecture by peres let us put @xmath180 .
we shall show that if inequality ( [ ppt ] ) is satisfied , with @xmath180 , then also the bell inequality ( [ inequality ] ) is not violated . using the form of the bell operator ( [ bell_operator ] ) the upper bound of the bell inequality , for @xmath0-ppt states ,
is found to read : @xmath181 and it can never reach the local realistic bound @xmath182 .
this is shown using the violation factor : @xmath183^n}{\cos \left(\frac{\pi}{2 m } \right)}.\ ] ] since @xmath184 and @xmath185 , where we have put @xmath186 as a minimal amount of settings for which bell inequality makes sense , the violation factor is bounded by @xmath187 . the simplest system on which one can perform partial transposes
consists of @xmath137 particles , thus @xmath188 .
none of the @xmath0-ppt states violates the bell inequality ( [ inequality ] ) .
it is worth mentioning that for @xmath186 setting case the violation @xmath189 confirms the results of werner and wolf @xcite , who gave the conjecture of peres a sharp mathematical form @xcite .
bell inequalities describe a performance of quantum communication complexity protocols @xcite . in this section
we follow this general link and present communication complexity problems associated with the inequality ( [ inequality ] ) .
it is proven that the quantum protocol outperforms the best classical protocol for arbitrary number of parties and observables .
in the communication complexity problems ( ccp ) one studies the information exchange between participants _ locally _ performing computations , in order to accomplish a _ globally _ defined task @xcite .
let us focus on a variant of a ccp , in which each of @xmath0 separated partners receives arguments , @xmath190 and @xmath191 , of some globally defined function , @xmath192 .
the @xmath193 inputs are assumed to be randomly distributed , and @xmath194 inputs can in general be distributed according to a weight @xmath195 .
the goal is to maximize the probability that alice arrives at the correct value of the function , under the restriction that @xmath196 bits of overall communication are allowed .
before participants receive their inputs they are allowed to do anything from which they can derive benefit . in particular , they can share some correlated strings of numbers in the classical scenario or entangled states in the quantum case . _ the problem .
_ following @xcite one chooses for a task - function : @xmath197 = \pm 1 , \label{ff}\end{aligned}\ ] ] with the angles defined by ( [ angles ] ) . according to the angles
definition the cosine can never be zero , so the problem is well - defined for all @xmath0 and @xmath1 .
additionally , the @xmath194 inputs are distributed with the weight : @xmath198 where the normalization factor is given by @xmath199 .
after the communication takes place , if alice misses some of the random variables @xmath193 , her `` answer '' can only be random .
thus , in an optimal protocol each party must communicate one bit .
there are only two communication structures which lead to a non - random answer : ( i ) a star
each party transmits one bit directly to alice , and ( ii ) a chain sequence of a peer - to - peer exchanges with alice at the end .
the task is to maximize the probability of correct answer @xmath200 .
since both @xmath201 and @xmath202 are dichotomic variables this amounts in maximizing : @xmath203 where @xmath204 describes ( random ) distribution of @xmath193 s , and @xmath205 is a probability that @xmath206 for given inputs @xmath207 and @xmath208 .
it is useful to express the last probability in terms of an average value of a product @xmath209 , i.e. @xmath210 $ ] .
since @xmath202 is independent of @xmath201 , and for given inputs it is constant , one has @xmath211 $ ] .
finally the probability of correct answer reads @xmath212 $ ] , and it is in one - to - one correspondence with a `` weighted '' scalar product ( average success ) : @xmath213 using the definitions ( [ ww ] ) for @xmath214 and ( [ ff ] ) for @xmath202 one gets : @xmath215 with angles given by ( [ angles ] ) .
we focus our attention on maximization of this quantity .
_ classical scenario . _ in the _ best _ classical protocol each party locally computes a bit function @xmath216 , with @xmath217 , where @xmath12 denotes some previously shared classical resources .
next , the bit is sent to alice , who puts as an answer the product @xmath218 .
the same answer can be reached in the chain strategy , simply the @xmath6th party sends @xmath219 . for the given inputs the procedure is always the same , i.e. @xmath220 . to prove the optimality of this protocol ,
one follows the proof of ref .
@xcite , with the only difference that @xmath194 is a @xmath1-valued variable now .
this , however , does not invalidate any of the steps of @xcite , and we will not repeat that proof . inserting the product form of @xmath221 into the average success ( [ success ] ) , using the fact that @xmath222 , and summing over all @xmath193 s one obtains : @xmath223 which has the same structure as local realistic expression ( [ ineq_deter ] ) .
thus , the highest classically achievable average success is given by a local realistic bound : @xmath224 .
_ quantum scenario .
_ in the quantum case participants share a @xmath0-party entangled state @xmath19 . after receiving inputs each party measures @xmath194th observable on the state , where the observables are enumerated as in the bell inequality ( [ inequality ] ) .
this results in a measurement outcome , @xmath225 .
each party sends @xmath226 to alice , who then puts as an answer a product @xmath227 . for the given inputs the average answer reads @xmath228 , and the maximal average success
is given by a quantum bound of : @xmath229 the average advantage of quantum versus classical protocol can be quantified by a factor @xmath230 which is equal to a violation factor , @xmath130 , introduced before . thus , all the states which violate the bell inequality ( including bound entangled state ) are a useful resource for the communication complexity task .
optimally one should use the ghz states @xmath97 , as they maximally violate the inequality .
alternatively , one can compare the probabilities of success , @xmath231 , in quantum and classical case .
clearly , one outperforms classical protocols for every @xmath0 and every @xmath1 . as an example , in table [ table_adv ] we gather the ratios between quantum and classical success probabilities for small number of participants .
.the ration between probabilities of success in quantum and classical case @xmath232 for the communication complexity problem with @xmath0 observers and @xmath1 settings .
quantum protocol uses ghz state . [ cols="^,^,^,^,^,^",options="header " , ] one can ask about a ccp with no random inputs @xmath193 . since the numbers @xmath194 already represent @xmath233 bits of information , and only one bit can be communicated , this looks like a plausible candidate for a quantum advantage .
however , in such a case a classical answer can not be put as a product of outcomes of local computations ( compare @xcite ) , and thus there is no bell inequality which would describe the best classical protocol . since classical performance of _
all _ ccps which can lead to quantum advantage is given by some bell inequality @xcite , the task without @xmath193 s can not lead to quantum advantage .
we presented a multisetting bell inequality , which unifies and generalizes many previous results .
examples of quantum states which violate the inequality were given .
it was also proven that all the states with positive partial transposes with respect to all subsystems can not violate the inequality . finally , the states which violate it were shown to reduce the communication complexity of computation of certain globally defined function .
the bell inequality presented is the only inequality which incorporates arbitrary number of settings for arbitrary number of observers making measurements on two - level systems , to date .
we thank m. ukowski for valuable discussions . w.l .
and t.p . are supported by foundation for polish science and mnii grant no
. 1 p03b 049 27 .
the work is part of the vi - th eu framework programme qap ( qubit applications ) contract no . 015848 .
j. f. clauser and a. shimony , rep .
. phys . * 41 * , 1881 ( 1978 ) ; d. m. greenberger , m. a. horne , a. shimony , and a. zeilinger , am . j. phys . * 58 * , 1131 ( 1990 ) ; t. paterek , w. laskowski , and m. ukowski , mod .
a * 21 * , 111 ( 2006 ) . | based on a geometrical argument introduced by ukowski , a new multisetting bell inequality is derived , for the scenario in which many parties make measurements on two - level systems .
this generalizes and unifies some previous results .
moreover , a necessary and sufficient condition for the violation of this inequality is presented .
it turns out that the class of non - separable states which do not admit local realistic description is extended when compared to the two - setting inequalities .
however , supporting the conjecture of peres , quantum states with positive partial transposes with respect to all subsystems do not violate the inequality .
additionally , we follow a general link between bell inequalities and communication complexity problems , and present a quantum protocol linked with the inequality , which outperforms the best classical protocol . |
much attention in condensed - matter physics is currently directed towards understanding electronic properties of dirac fermions in three dimensional ( 3d ) topological insulators ( tis ) , which challenge fundamental concepts and hold a great potential for electronic , optic and optoelectronic applications , see e.g. ref .
@xcite .
recently nonlinear high frequency electron transport phenomena @xcite in ti systems have attracted growing interest .
there have been many theoretical and experimental works in the past few years on the helicity controlled photocurrents @xcite , linear photogalvanic effect @xcite , local photocurrents @xcite , edge photocurrents in 2d ti @xcite , coherent control of injection currents @xcite , photon drag currents @xcite , second harmonic generation @xcite , photo - induced quantum hall insulators @xcite , cyclotron resonance assisted photocurrents @xcite , quantum oscillations of photocurrents@xcite , photogalvanic currents via proximity interactions with magnetic materials @xcite and photoelectromagnetic effect @xcite .
these phenomena , scaling in the second or third order of the radiation electric fields , open up new opportunities to the study of dirac fermions , as it has been already demonstrated for graphene , for review see ref . , and
several ti materials , see e.g. ref . .
an important advantage of the nonlinear high frequency transport effects is that some of them , being forbidden by symmetry in the bulk of most 3d ti , can be applied to selectively probe the surface states even in ti materials with a finite bulk conductivity . utilizing photocurrents this advantage has been used to study sb@xmath5te@xmath6 and bi@xmath5te@xmath6 3d tis @xcite , in which conventional _ dc _ transport experiments , particularly at room temperature , are handicapped by a large residual bulk charge carrier density @xcite . it has been shown in ref . that the photocurrent excited by normal incident terahertz ( thz ) radiation is generated due to the photogalvanic effect .
the latter originated from the asymmetric scattering of dirac fermions driven back and forth by the _
ac _ electric field and is allowed in the non - centrosymmetric surface states only .
the experiments further hinted at a possible contribution of the photon drag effect - a competing photocurrent resulting from the light momentum transfer to charged carriers .
however , no experiments which provide clear evidence of the photon drag effect in ti materials have been reported so far . here
we report on the observation of the photon drag effect in ( bi@xmath0sb@xmath1)@xmath2te@xmath3 3d tis excited by thz radiation . we demonstrate that while at normal incidence the photocurrent is dominated by the photogalvanic effect , at oblique incidence it is outweighed by the photon drag effect .
the latter is shown to be caused by the _ in - plane _ component of the photon wavevector @xmath7 . strikingly
, the observed photon drag current does not change its sign upon inverting @xmath7 .
this , seemingly surprising result , is caused by the fact that in materials with trigonal symmetry the photon drag current is proportional not only to the photon wavevector but also to the product of _
in- _ and _ out - of - plane _ components of the radiation electric field . since both ,
@xmath7 and the product of the electric fields , change their sign , the total sign remains unchanged .
our experimental findings are well described by the developed theory and microscopic model , based on the boltzmann kinetic equation for the carrier distribution function .
both photon drag and photogalvanic effects are investigated in epitaxially grown ( bi@xmath0sb@xmath1)@xmath2te@xmath3 bulk materials of various composition determined by the antimony content @xmath4 .
the variation of @xmath4 enabled us to study photocurrents in different systems including binary and ternary tis with smooth changes from @xmath8- to @xmath9-type bulk conductivity , see ref . as well as in heterostructure samples , consisting of a @xmath8-type bi@xmath5te@xmath6 and a @xmath9-type sb@xmath5te@xmath6 layer ,
see ref . .
in the latter the chemical potential can be tuned by varying the thickness of the upper sb@xmath5te@xmath6 layer .
[ cols= " < , < , < , < , < , < , < " , ] experiments on photocurrents in ( bi@xmath0sb@xmath1)@xmath5te@xmath6 3d tis were performed applying radiation of a high power pulsed molecular thz laser @xcite . using nh@xmath6 , d@xmath5o and ch@xmath6f as active gases for the optically pumped laser ,
40 ns pulses with peak power of @xmath10 10 kw were obtained at different frequencies @xmath11 , see table [ table2 ] and refs . .
the radiation induces indirect ( drude - like ) optical transitions , because the photon energies are much smaller than the carrier fermi energy .
the beam had an almost gaussian form , which was measured by a pyroelectric camera @xcite .
a typical spot diameter depends on the radiation frequency and varies between 1 and 3 mm .
the electric field amplitude @xmath12 of the incoming radiation was varied from about @xmath13 to @xmath14 kv / cm ( radiation intensities @xmath15 from about @xmath13 to 1000 kw/@xmath16 ) .
the samples were illuminated at normal and oblique incidence . in experiments at normal incidence front and back
illumination was used with angle of incidence @xmath17 and @xmath18 , respectively , see figs .
[ pge ] and [ qz ] . in the measurements applying oblique incident radiation the angle @xmath19 was varied between @xmath20 and @xmath21 , see insets in figs .
[ theta_alpha ] and [ theta ] .
note , that larger angles of incidence were not used in order to avoid the illumination of contacts and edges .
the photocurrents were analysed in two directions , @xmath4 and @xmath22 , perpendicular to each other and parallel to the sample edges , see inset in fig .
most experiments at oblique incidence were carried out for ( @xmath23 ) plane of incidence . in some additional measurements
the orientation of the plane of incidence was rotated by the angle @xmath24 in respect to the ( @xmath23 ) plane , see inset in figs .
[ psi_alpha ] and [ psi ] .
the _ dc _ photocurrent @xmath25 was excited in the temperature range from @xmath26 to 296 k. it was measured as a voltage drop , @xmath27 , across a 50@xmath28 load resistor and recorded in unbiased samples with a storage oscilloscope . to control the incidence power of the laser
the signal was simultaneously measured at a reference thz detector @xcite . to examine the photocurrent behavior upon the variation of the polarization state half and quarter
wavelength plates were employed .
the initial laser radiation was linearly polarized along the @xmath22-axis . by using @xmath29 plates ,
the azimuth angle @xmath30 was varied between the linear polarization on the sample and the @xmath22 axis , see inset and top panel in fig .
[ pge ] . by applying @xmath31/4 plates
, we obtained elliptically ( and circularly ) polarized radiation . in this case , the polarization state is determined by the angle @xmath32 between the plate optical axis and the incoming laser polarization . here , the electric field vector is lying parallel to the @xmath4 axis .
the polarization states for several @xmath32 are shown on the top panel of fig .
[ phi ] . in this geometry , the radiation helicity is varied as @xmath33 @xcite .
-ray diffraction pole figure scan around the ( 1,0,5 ) reflection of the bi@xmath5te@xmath6 sample bst127 showing that one domain orientation ( highlighted by solid red line ) dominates .
it also reveals that the crystallographic axes lie parallel to the sample edges .
insets sketches domain orientations illustrated by solid red line connecting top bi atoms in the upper right panel and side view of one quintuple layer , see right bottom panel .
( b ) photocurrent @xmath34 in bi@xmath5te@xmath6 sample bst127 measured for front and back illumination at @xmath35k .
solid lines show fits after eq . .
note that the same dependencies are obtained after phenomenological , see eq . , and microscopic theory , see eq . for the photogalvanic effect
] irradiating the ( bi@xmath0sb@xmath1)@xmath5te@xmath6 3d tis with linearly polarized thz radiation we observed a _
dc _ current in both @xmath4- and @xmath22- direction .
the photocurrent was detected in the whole frequency range used from 0.6 up to 3.9 thz .
the signal followed the temporal structure of the laser pulse .
its variation upon rotation of the polarization plane is well fitted by @xmath36 e_0 ^ 2 \ , , \\
j_y = [ a(f ) \sin 2\alpha + c'(f ) ] e_0 ^ 2 \ , , \nonumber\end{aligned}\ ] ] in which @xmath37 is the squared radiation electric field and @xmath38 , @xmath39 , and @xmath40 are fitting parameters , see fig .
[ pge ] ( a ) and ( b ) .
the above polarization dependencies were observed in all samples and for all frequencies .
cooling the sample from room temperature to 4.2 k did increase the photocurrent amplitude , see fig .
[ pge ] ( c ) , whereas the overall behavior remained unchanged . measured in bi@xmath5te@xmath6 sample bst127 . ( b ) photocurrent @xmath41 measured in ( bi@xmath42sb@xmath43)@xmath5te@xmath6 sample bst641 .
plots show the dependence of the photocurrent excited by normal incident radiation with @xmath44 thz on the azimuth angle @xmath30 .
angles of incidence @xmath17 and 180@xmath45 correspond to the front and back excitation , respectively .
solid lines show fits after eq . .
note that the same dependencies are obtained after phenomenological , see eq . , and microscopic theory , see eqs . and 26 for the photogalvanic and the photon drag effect , respectively .
insets sketch the setup and the orientation of the electric field .
note , that the photocurrent is probed in the directions coinciding with the principal axes of the trigonal system .
( c ) temperature dependence of the photocurrent measured in the bi@xmath5te@xmath6 sample bst127 and sb@xmath5te@xmath6 sample bst305 .
all data are normalized to the value for @xmath26 k. ] as we have shown in ref .
two phenomena can be the cause of the thz radiation induced photocurrents , described by eqs .
( [ fit_alpha ] ) , namely the photogalvanic and the photon drag effects @xcite .
experiments applying front and back illumination with normally incident radiation allows us to distinguish them from each other . while the photogalvanic current is determined by the _ in - plane _ orientation of the radiation electric field @xcite and , consequently , remains unchanged for both geometries , the photon drag current is additionally proportional to a component of the photon momentum @xmath46 .
therefore , changing @xmath47 ( front to back illumination ) does not affect the photogalvanic but inverts the sign of the factor @xmath48 for the photon drag effect .
note that for front and back illumination at normal incidence the wavevector @xmath49 is directed parallel or anti - parallel to the @xmath50 direction . as an important result we obtained that the sign of the amplitude @xmath48 remains unchanged
( see exemplary figs .
[ pge ] and [ qz ] for samples bst127 and bst641 and table [ table ] for all investigated samples at @xmath51 thz ) .
this fact provides a clear evidence that the photocurrent at normal incidence is dominated by the photogalvanic effect in the investigated two dimensional ( 2d ) dirac fermion systems . in samples bst127 and bst641 excited with @xmath52 thz , the contribution of the photon drag is vanishingly small , see fig .
[ pge ] . at other frequencies and samples
the photon drag effect may yield a contribution up to one third compared to that of the photogalvanic effect resulting in larger signals for back illumination than front one @xcite , see fig . [ qz ] and
table [ table ] . measured in bi@xmath5te@xmath6 sample bst127 for @xmath51 thz .
( b ) @xmath41 measured in ( bi@xmath42sb@xmath43)@xmath5te@xmath6 sample bst641 for @xmath53 thz .
the data show the dependence of the photocurrent on the azimuth angle @xmath30 excited by normal incident radiation .
angles of incidence @xmath17 and 180@xmath45 correspond to the front and back excitation , respectively .
solid lines show fits after eq . .
note that the same dependencies are obtained after phenomenological , see eq . , and microscopic theory , see eqs . and for the photogalvanic and the @xmath54 - photon drag effects , respectively . ]
so far we presented data obtained for normal incidence . illuminating the samples at oblique incidence
, we found that all characteristic properties of the photocurrent including its polarization behavior remain unchanged , see fig .
[ theta_alpha ] .
in contrast to the measurements at normal incidence , the magnitude @xmath55 depends now additionally on the incident angle @xmath19 , as well as on the direction in which the current is examined : at large @xmath19 and for the plane of incidence coinciding , e.g. with ( @xmath23)-plane the photocurrent magnitudes measured in @xmath4 and @xmath22 directions become slightly different .
for some excitation frequencies the photocurrent amplitude @xmath55 did reduce upon the increase of the angle @xmath19 , exemplarily shown for the current measured in @xmath4 direction in fig .
[ theta_alpha ] ( a ) , fig .
[ theta ] ( a ) and ( b ) .
note that the change of @xmath55 is even in the angle @xmath19 .
excited in ( bi@xmath42sb@xmath43)@xmath5te@xmath6 sample bst641 by normal and oblique incident radiation .
the data demonstrate that the polarization dependence does not change upon variation of the angle of incidence : neither for the case that the current decrease at oblique incidence [ panel ( a ) , @xmath44 thz ] nor for the case it increases with @xmath56 increasing [ panel ( b ) , @xmath51 thz ] .
solid lines show fits after eq . .
note that the same dependencies are obtained after phenomenological , see eq . , and microscopic theory , see eqs . , and for the photogalvanic and both photon drag effects . ]
strikingly , at other radiation frequencies we observed that for positive as well as for negative @xmath19 the signal rises with an increase of the angle of incidence , see fig . [ theta_alpha ] ( b ) and fig .
[ theta ] ( c ) - ( f ) .
this behavior is observed for photocurrents measured in directions parallel as well as normal to the plane of incidence , see fig .
[ theta ] ( d ) .
it is also detected for any orientation of the plane of incidence .
figure [ psi_alpha ] shows the corresponding data for three positions of the incident plane determined by the angle @xmath24 .
the figure reveals that the signal varies as @xmath57 and the most pronounced change in the photocurrents polarization dependence is the appearance of a @xmath58 phase shift .
a detected small variation of @xmath59 as a function of @xmath24 can not be discussed earnestly , since precise adjustment ensuring that for different @xmath24 , technically obtained by rotating the sample , the laser spot remains on the same sample position is hard to realize . to support the conclusion that the rotation of the incident plane results in a @xmath60 phase shift , we measured photocurrents depending on the linear polarization by changing the angle @xmath24 by steps of @xmath61 in the range from 0 to @xmath62 . figure [ psi ] ( a ) demonstrates the dependencies obtained for bi@xmath5te@xmath6 sample bst127 . the corresponding dependency of the measured phase shift @xmath63 on the angle @xmath24 is shown in fig . [ psi ] ( b ) and for other samples and frequencies in fig . [ psi ] ( c ) .
the figures demonstrate that in all cases @xmath64 . finally , we discuss the results obtained applying elliptically ( circularly ) polarized radiation .
these measurements were particularly motivated by the search for the circular photogalvanic @xcite and circular photon drag effect @xcite , i.e. photocurrents changing their direction upon switching of the radiation helicity @xcite , recently observed for bi@xmath5te@xmath6 ti excited by near infrared light @xcite . applying radiation at oblique incidence and measuring the photocurrent in the direction normal to the plane of incidence ( @xmath23 ) ,
i.e. in the geometry for which circular photogalvanic @xcite and circular photon drag effects @xcite are expected , we detected a current which can be well fitted by @xmath65 , see fig .
the figure clearly shows that for circularly polarized radiation ( @xmath66 and @xmath67 ) the signal vanishes .
in fact , the term in brackets describes the degree of linear polarization in the lambda - quarter plate geometry .
therefore , the observed current is identical to the one excited by linearly polarized radiation and which was already discussed above .
examining different samples in the whole investigated thz frequency range , we observed the same result : no trace of a helicity dependent photocurrent has been detected .
excited in bi@xmath5te@xmath6 sample bst127 .
the data are obtained at oblique incidence radiation ( @xmath44 thz ) for @xmath68 and different orientations of the plane of incidence in respect to the @xmath22-direction given by the angle @xmath24 , see inset .
the data are shown for the angles @xmath69 , @xmath70 and @xmath71 .
the solid lines are calculated after eq . .
] to summarize , experiments on different types of ti samples provide a self - consistent picture demonstrating that the photocurrents ( i ) are caused by effects proportional to the second power of the radiation electric field and can be excited by both normal and oblique incident radiation , ( ii ) are excited by linearly polarized radiation , ( iii ) vary with the azimuth angle @xmath30 as @xmath72 with a possible phase shift depending on the experimental geometry and cristallographic direction in which the photocurrent is measured , see figs .
[ psi_alpha ] and [ psi ] , ( iv ) have the same sign but may have distinct magnitudes of the factor @xmath48 for front and back normal incident illumination , see fig . [ pge ] and [ qz ] , and ( v ) are described by an even function of the angle of incidence @xmath19 with the magnitude @xmath73 for different radiation frequencies , raising or decreasing with increase of @xmath19 , see figs .
[ theta ] and [ theta_alpha ] . excited in bi@xmath5te@xmath6 sample bst127 .
the data are obtained at oblique incidence radiation ( @xmath44 thz ) for @xmath68 and different orientations of the plane of incidence in respect to the @xmath22-direction given by the angle @xmath24 .
the data are shown for the angle @xmath24 changing from @xmath74 to @xmath62 with steps of @xmath75 .
solid lines show fits after eqs . .
bottom panels show the measured phase shift as a function of @xmath24 for ( b ) the data shown in panel ( a ) and ( c ) three samples bst641 , bst127 and bst305 and two radiation frequencies . here
solid and open symbols correspond to the frequencies @xmath44 and 3.3 thz , respectively .
solid line shows fit after @xmath58 . ]
now we discuss the origin of the observed photocurrents , which are induced by spatially homogeneous terahertz radiation and scale with the second power of the radiation electric field .
we begin with the standard way to treat second order effects without going into microscopic details , which makes use of the symmetry arguments .
this approach allows us to explore what kind of photocurrents are allowed in the considered system and to describe their variation upon change of macroscopic parameters , such as radiation intensity , polarization and incident angle . with this
, the response of the charge carriers ensemble to the external field can be characterized conveniently by the coordinate- and time - dependent electric current density @xmath76 .
it is expanded in a power series in the external alternating electric field @xmath77 in the form of a plane wave @xmath78 where @xmath79 is the angular frequency and @xmath49 is its wavevector . limiting the consideration to the second order effects
we obtain the photocurrent density @xmath80 in the form @xcite @xmath81 where the expansion coefficients @xmath82 and @xmath83 are third and fourth rank tensors , respectively , and @xmath84 is the complex conjugate of @xmath85 . the first term on the right - hand side of eq
. represents photogalvanic effects whereas the second term describes the photon drag effect containing additionally the wavevector of the electromagnetic field . ,
measured in bi@xmath5te@xmath6 sample bst127 at normal as well as at oblique incidence in the direction normal to the plane of incidence .
the ellipses on top illustrate the polarization states for various angles @xmath32 .
solid lines show fits after eqs . where the polarization dependent terms take for this geometry the form @xmath65 . ]
equation can be simplified considering the point group @xmath86 , which describes the symmetry of the surface states in ( bi@xmath0sb@xmath1)@xmath2te@xmath3 . to be specific we first obtain the photocurrents excited in the crystallographic directions @xmath4 and @xmath22 with the radiation plane of incidence ( @xmath23 ) . taking into account the fact that we detected only photocurrents excited by linearly
polarized radiation being even in the angle of incidence @xmath19 we can omit all contributions of photocurrents sensitive to the radiation helicity and those giving the response odd in the angle @xmath19 . under these conditions we derive for @xmath87 @xmath88 , \nonumber\\ j_y & = - 2 ( \chi + t_z
q_z ) e_x e_y - t_\parallel q_ye_xe_z \nonumber \\ & = \sin{2\alpha } \ , t_st_p\left [ ( \chi - t_z q \cos{\theta})\cos{\theta } + t_\parallel q \sin^2{\theta}/2 \right ] e_0 ^ 2 . \nonumber \end{aligned}\ ] ] here
@xmath89 and @xmath90 are the fresnel transmission coefficients for @xmath91- and @xmath9- polarized light and @xmath92 is the polarization independent offset being equal to zero for @xmath93 , see appendix 1 .
the constants @xmath94 , @xmath95 and @xmath96 are coefficients describing the photogalvanic , the photon drag effect at oblique incidence , and the photon drag effect at normal incidence , respectively . at normal incidence
we obtain the photogalvanic effect and the photon drag effect , caused by @xmath50-component of the photon wavevector .
they are given , respectively , by the terms proportional to factors @xmath94 and @xmath96 .
both effects have the same polarization dependence and vary with the azimuth angle @xmath30 according to @xmath97 and @xmath98 , being in agreement with the results of experiments , see figs .
[ pge ] , [ qz ] and eq . .
for @xmath99 reversing the radiation propagation direction ( @xmath100 ) does not affect the polarization dependence .
it may , however , change the signal magnitude because @xmath101 is different for negative and positive @xmath54 corresponding to front and back illumination .
exactly this behavior has been observed in experiments showing that at some frequencies the photon drag due to @xmath50-component of the wavevector yields a minor contribution to the total photocurrent , see values of @xmath102 at @xmath93 and @xmath18 in table [ table ] and fig . [ qz ] . at oblique incidence @xmath103 and @xmath54
components are reduced and the magnitude of both photocurrents diminish equally for positive and negative angle @xmath19 , see fig .
[ theta_alpha ] ( a ) and doted lines in fig . [ theta ] . according to eqs . at oblique incidence
the photon drag effect , coupled to the _ in - plane _
wavevector @xmath104 , can also contribute to the total photocurrent .
its dependence on the azimuth angle @xmath30 formally coincides with that of the contributions discussed so far , @xmath105 and @xmath106 .
however , it obviously vanishes at normal incidence ( @xmath107 ) and , in contrast to photogalvanic effect , increases with the rising angle of incidence , see dashed lines in fig .
[ theta_alpha ] .
moreover , the sign of the products @xmath108 and @xmath109 is staying the same for positive and negative @xmath19 .
a dominating contribution of this effect has been observed at large angles of incidence for all samples and for almost all frequencies . the most clear evidence for this conclusion is supported by measurements shown in fig .
[ theta_alpha ] ( b ) and [ theta ] ( c)-(f ) demonstrate that the photocurrent rises with the angle @xmath19 increase .
rotation of the incident plane by the angle @xmath24 changes the relative orientation of the electric field and crystallographic axes modifying eqs . .
taking into account that in all experiments described above @xmath110 and considering small angles @xmath19 , being relevant to the experimental data of figs .
[ theta_alpha ] and [ theta ] , we obtain @xmath111 here @xmath112 is the amplitude of the transmission coefficient for small @xmath19 and the offset current is omitted .
equations for arbitrary angle of incidence are given in the appendix 1 .
equations show that the rotation of the incident plane mainly results in a @xmath60 phase shift for both photocurrents @xmath113 and @xmath114 .
this phase shift has been observed for all samples and frequencies , see figs .
[ psi_alpha ] and [ psi ] .
in general , second order high frequency effects are caused by the redistribution of charge carriers in the momentum space induced by the illumination of the sample with radiation .
the resulting nonequilibrium distribution can contain components which are oscillating in time and space , as well as steady - state and spatially homogeneous ones .
hence , the irradiation may cause both _
ac _ and _ dc _ flows in a media .
their magnitudes are nonlinear functions of the field amplitude and their components are sensitive to the radiation polarization . in the following section we present models visualizing the physics of nonlinear responses . for simplicity
we assume positively charged carriers , i.e. , holes for which the directions of the carrier flow and the corresponding electric current coincide .
the model and the microscopic theory of the photogalvanic effect have been discussed in detailed in ref . , demonstrating that the photocurrent stems from the asymmetric scattering of free carriers excited by irradiation with a _ ac _ electric field . as we show below the asymmetric scattering is also responsible for the observed photon drag effects . therefore , to introduce the concepts essential for the formation of the latter effects and to provide a complete picture of the photocurrent formation in ti
, we will briefly address the model of the photogalvanic effect .
sb@xmath1)@xmath5te@xmath6 due to asymmetry of elastic scattering of holes by wedges . ]
the current generation process is illustrated in the right panel of fig .
[ modelpge ] . as addressed above , the symmetry of the surface states in ( bi@xmath0sb@xmath1)@xmath2te@xmath3 3d tis is c@xmath115 .
this point group implies that the anisotropy of carrier elastic scattering is the same as at scattering by a double triangular pyramid , whose side and top view are sketched in the left panel fig .
[ modelpge ] ( a ) .
note that for c@xmath115 symmetry scattering by top and bottom pyramids has different probabilities . in the framework of the photogalvanic effect caused by the _ in - plane _ motion of free carriers
the scatterers can be considered as randomly distributed but identically oriented wedges lying in the ql - plane .
the preferential orientation of wedges is supported by the @xmath116-rays data shown above , see fig .
[ xrd ] ( a ) and ref . .
in the absence of radiation , the flows of anisotropically scattered holes , see right panel fig . [ modelpge ] ( a ) , exactly compensate each other .
application of linearly polarized thz radiation results in an _ alignment _ of carrier momenta : the total flow of holes driven back and forth by _ ac _ electric field @xmath117 increases .
the corresponding _ stationary _ correction to the hole distribution function scales as a square of the @xmath118 electric field magnitude @xcite .
the _ stationary alignment _ of carrier momenta itself does not lead to a _
dc _ electric current but , due to asymmetric scattering by wedges , the excess of the flux of carriers moving along the field violates the balance of flows @xcite , and the linear photogalvanic current is generated @xcite .
the direction of the induced current depends on the relative orientation of the _ ac _ electric field and wedges : e.g. , a field parallel to the wedges base ( @xmath119 ) , see fig .
[ modelpge ] ( b ) , yields the current flow in @xmath4-direction while rotation of the electric field by 90@xmath45 reverses the current direction , see fig .
[ modelpge ] ( c ) .
the polarization dependence of the photogalvanic current in @xmath4- and @xmath22- directions is described by the terms with @xmath94 in eqs . and by eq . .
note that the coefficient @xmath94 has opposite signs for holes and electrons .
the trigonal photon drag effect caused by the _ in - plane _ component of the photon wavevector results in a @xmath120 current increasing with the angle of incidence increase .
it is described by an _ even _ function of the angle @xmath19 .
similar to the photogalvanic effect the photon drag current formation involves asymmetric scattering of free carriers .
the trigonal photon drag effect results from a _ dynamical _ alignment of carrier momenta .
it is generated due to the _ in - plane _ profile of the radiation electric field and implies the difference in the scattering probabilities for different half periods of the electromagnetic wave .
the process of the current generation is illustrated in fig .
[ modelpd ] . like in the model for photogalvanic effect
we consider the scatterers as randomly distributed but identically oriented pyramids in the ql - plane , see fig .
[ modelpd ] ( d ) . in the absence of radiation ,
the flows of the thermalized charge carriers which are anisotropically scattered by pyramids exactly compensate each other .
optical excitation disturbs the balance due to the action of high frequency electric field @xmath121 on charged carriers , we assume holes .
the discussed trigonal photon drag current is caused by the dynamic variation of the electric field @xmath121 in the direction of the radiation propagation , see fig .
[ modelpd ] ( a ) .
the strength of the corresponding force acting on holes is given by @xmath122 , where @xmath123 is the elementary charge @xcite .
the force is coordinate - dependent and causes the hole acceleration to be directed parallel to the @xmath4-direction for @xmath124 ( anti - parallel for @xmath125 ) and , consequently , increase of the hole flow by @xmath126 ( @xmath127 ) , see horizontal arrows in fig .
[ modelpd ] ( a ) . as a result of this _
dynamical _ momentum alignment the balance of the hole flows scattered by pyramids in the vicinity of the electric field @xmath103 nodes becomes _ locally _ violated , see fig .
[ modelpd ] ( a ) and ( d ) . the asymmetric scattering may cause equal in magnitude but oppositely directed local electric currents @xmath128 and @xmath129 in the vicinity of @xmath130 and @xmath131 ; whereas the total electric current remains zero .
however , steady - state @xmath120 electric current indeed emerges if one additionally takes into account the @xmath50-component of the radiation field and the retardation between the electric field and the instant velocity of charge carrier .
the photocurrent reaches its maximum at @xmath132 about unity .
the effect of the retardation , well known in the drude - lorentz theory of high frequency conductivity , causes a phase shift between the electric field @xmath103 and the instant change of the hole velocity @xmath133 given by @xmath134 .
consequently , the nodes of the charge carriers velocity @xmath133 are shifted in respect to that of electric field @xmath103 , these nodes are indicated in fig .
[ modelpd ] ( b ) as @xmath135 and @xmath136 .
the carriers in the vicinity of the @xmath135 and @xmath136-positions are subjected to the electric field @xmath137 and @xmath138 which have the opposite signs .
the @xmath137 [ @xmath138 ] field is pushing the carriers to the basis [ top ] of the pyramids which increase [ decrease ] the scattering probability at @xmath135 [ @xmath136 ] vicinity .
consequently , it changes the magnitudes of the local currents @xmath141 and @xmath142 caused by the asymmetric scattering .
the variation of the scattering probability upon the action of the _ out - of - plane _ electric field @xmath143 is described by eq . in sec .
[ sec : micro ] . as a result
the oppositely directed local currents do not compensate each other anymore and a @xmath120 electric current being proportional to the product @xmath144 emerges .
changing the angle of incidence form @xmath19 to @xmath145 reverses the sign of both @xmath146 and the product @xmath147 so that the direction of the _ dc _ current remains unchanged .
we emphasize that such a contribution to the photon drag effect is specific for trigonal systems and it is absent in , e.g. , hexagonal systems like graphene @xcite . .
to be specific we discuss the hole gas in the surface states excited by oblique radiation with the incidence plane ( @xmath148 ) , see panel ( c ) .
triangle in ( d ) , left picture , shows top view of the considered scattering potential .
side view of the scatters is sketched in the right picture . ]
now we turn to the microscopic theory of the photon drag effect . in the classical regime
achievable in our experiments , which is characterized by @xmath149 , the photocurrents can be well described by means of boltzmann s kinetic equation for the coordinate dependent carrier distribution function @xmath150 @xmath151 . \nonumber\end{aligned}\ ] ] where @xmath152 for holes and @xmath152 for electrons , @xmath153 is the velocity of surface charge carriers with a momentum @xmath154 , @xmath155 is the dirac fermion velocity , and @xmath156 is a probability for a charge carrier to have the momenta @xmath154 and @xmath157 before and after scattering , respectively .
lack of inversion center for the surface charge carriers makes their elastic scattering asymmetric , so that @xmath158 @xcite , and results in a _
dc _ electric current .
note that this asymmetry takes place even for isotropic scatterers like impurities or phonons .
the photocurrent can be calculated as follows @xcite @xmath159 where @xmath160 is the correction to the distribution function being quadratic in the radiation electric field amplitude and linear in the photon momentum . to calculate the asymmetric part of the scattering probability , which is responsible for the photocurrent formation , we take into account warping of the energy spectrum . without warping the energy dispersion of the surface states
is described by the hamiltonian @xcite @xmath161 where @xmath162 are pauli matrices .
the hamiltonian yields the linear energy dispersion @xmath163 and corresponding wavefunctions @xmath164/\sqrt{2 } } $ ] for electrons and holes , where @xmath165 is the angle between carrier momentum @xmath154 and the @xmath4 axis .
the warping of the energy spectrum reflects the trigonal symmetry of the surface and is described by an additional to eq .
( [ hamiltonian ] ) small term given by @xcite @xmath166 where @xmath167 is a warping constant .
this is the perturbation which leads to the hexagonal warping of the energy surfaces @xcite clearly detected by arpes , see top panels in fig .
[ arpes ] .
the perturbation caused by the terahertz radiation electric field @xmath168 changes the surface charge carrier wavefunctions due to admixture of bulk states from various bands .
the corresponding hamiltonian is linear in the coordinate @xmath50 @xmath169 taking into account both perturbations @xmath170 and @xmath171 in the first order , we obtain the corrected electron wavefunction : @xmath172 where the index @xmath91 labels the bulk orbitals from which the surface states are formed , and @xmath8 enumerates other energy bands of the bulk crystal . here
we assume that all bulk bands lie far enough away from the dirac point so the energies @xmath173 , where @xmath174 is the fermi momentum .
calculating the matrix elements of scattering by a static potential we obtain from the fermi golden rule the scattering probability in the form @xmath175 .
the field - independent part is given by the usual expression taking into account the absence of back scattering for dirac fermions @xcite .
@xmath176 where @xmath177 is the fourier image of the scattering potential , @xmath178 is the scattering angle , and the angular brackets mean averaging over positions of scatterers @xcite .
the linear in @xmath168 correction is given by @xmath179 here @xmath180 and @xmath181 are the intra- and interband matrix elements of the scattering potential , respectively .
the latter is caused by the short - range scatterers with the momentum transfer @xmath182 , where @xmath183 has an atomic scale , therefore the average product is assumed to be independent of @xmath165 and @xmath184 @xcite .
we emphasize that the obtained correction , eq .
, is responsible for the effect of @xmath168-electric field on the scattering by pyramid - like scatters discussed in the model of the photon drag effect , see sec .
[ section_modeldrag ] . using the derived scattering probability @xmath156
we solve the boltzmann equation and obtain the @xmath185-independent correction to the distribution function , @xmath160 , which allows us to calculate the photon drag current given by eq . .
with this we will search for the correction to the distribution function being responsible for the _ dynamical _ alignment momentum , @xmath186 .
first we find the linear in @xmath187 solution given by @xmath188 where @xmath189 is the fermi - dirac distribution function , and the transport relaxation time @xmath190 , determining the mobility of 2d dirac fermions is related to the symmetric part of the scattering probability as @xmath191 .
the photon wavevector is accounted by the space derivatives in the kinetic equation @xcite @xmath192 the time @xmath193 being of the order of @xmath190 describes relaxation of the above discussed alignment of charge carrier momenta .
it is defined as follows @xmath194 from eq .
we find the correction to the distribution function describing the dynamical alignment of momenta in the form @xmath195 now we take into account the anisotropic scattering which manifests itself as a correction given by eq . .
the @xmath185-independent correction to the distribution function @xmath160 is found from the following equation @xcite @xmath196 finally , from eq .
we find the trigonal photon drag current .
experiments reveal that the photocurrent in all samples is caused by the linearly polarized radiation . for excitation by oblique incidence in the @xmath197 plane it is described by @xmath198 here ,
the high - frequency conductivity is given by the drude expression for degenerate 2d carriers @xmath199 we introduce the anisotropic scattering constant which is nonzero due to c@xmath115 symmetry of the studied system @xmath200 where brackets in the denominator mean averaging over both scatterer positions and the scattering angle @xmath201 .
the derived eqs .
show that , in line with experiments , see figs .
[ theta ] , and the discussion in sec .
[ section_discussion ] the current is even in the angle of incidence and vanishes for normal incidence at which @xmath202 . for elastic scattering by coulomb impurities ,
the relaxation times are related as @xmath203 . according to ref .
, in this case the photogalvanic current is given by @xmath204 where @xmath205 is the factor describing the c@xmath115 symmetry of the studied system . in the studied samples we can estimate @xmath206 ,
see also ref . .
the frequency dependencies of trigonal photon drag and photogalvanic currents are plotted in fig .
[ freq_depend ] .
while photogalvanic current drops monotonously with frequency increase , the photon drag current has a maximum at @xmath207 . at high frequencies ,
both linear photon drag and linear photogalvanic currents decrease as @xmath208 .
the difference between the frequency dependencies may be the cause for the observed variation of the ratio between photon drag and photogalvanic currents varying in the range from about -2 to -15 for different samples and frequencies .
the ratio of the photon drag and photogalvanic currents can be estimated as @xmath209 the first factor is a dimensionless degree of warping which can be of order of unity in our samples according to arpes measurements , fig [ arpes ] . in our experiments ,
the trigonal photon drag current at oblique incidence is larger than the photogalvanic one .
this allows us to estimate the interband scattering parameter @xmath210 defined by eq . : in the studied samples @xmath211 / ev . from the difference in sign between the photogalvanic and the trigonal photon drag currents systematically observed in experiment
we conclude that the product @xmath212 , describing the trigonal photon drag and the photogalvanic constant @xmath205 , are of opposite sign .
finally we note , that the solution of the boltzmann equation also yields the circular photon drag current proportional to the radiation helicity @xmath213 it follows from these expressions that the circular photon drag current is zero at @xmath214 , a value at which the linear photon drag current is close to its maximum , fig .
[ freq_depend ] . the vanishing contribution of the circular photocurrent in the vicinity of the @xmath215 the condition corresponding to our experiments
may explain the fact that in the studied frequency range no helicity dependent current has been detected . at last but
not least we obtain the trigonal photon drag current caused by @xmath54-component of the photon wavevector .
it is given by @xmath216 where @xmath96 is a real constant .
the equations reveal that the photocurrent can be excited by linearly polarized radiation and is sensitive to the polarization plane position in respect to the crystallographic axes .
the microscopic picture of the photocurrent generation can most conveniently be described in terms of the radiation magnetic field @xmath217 rather than the transfer of the normal component of the photon wavevector @xmath54 to free carriers .
indeed , the latter is not possible in strictly 2d systems . using the relation @xmath218 , we can rewrite eq . as @xmath219 and @xmath220 in order to develop a microscopic theory for the photocurrent given by eq . , we take into account the lorentz force of the radiation magnetic field acting on the 2d carriers .
the corresponding hamiltonian has the following form : @xmath221 taking into account both @xmath222 and the warping perturbation eq . in the first order
, we derive the linear in @xmath217 correction to the elastic scattering probability @xmath223 .
it is obtained from @xmath224 , eq . , by substitution @xmath225 .\ ] ] now we solve the boltzmann kinetic eq .
accounting the correction @xmath223 to the scattering probability .
it has the form @xmath226 where @xmath227 is the correction to the distribution function linear in @xmath217 , and the linear in @xmath187 correction @xmath228 is given by eq . .
the linear in both @xmath187 and @xmath229 current density is calculated by eq . with
the above found correction @xmath227 .
the result has the form @xmath230 comparing the photocurrent amplitudes at normal and at oblique incidence , eqs . and
, we obtain @xmath231 this estimate demonstrates that , because the radiation magnetic field affects elastic scattering weaker than the electric field , @xmath96 is substantially smaller than @xmath95 . since @xmath232 ev , the factor @xmath233 is in the order of @xmath234 to @xmath235 for our samples .
this estimation explains why in all our experiments @xmath236 .
the smallest value obtained are 40 times for sample bst127 and 80 for sample bst641 . in both cases
the samples were excited by radiation of @xmath237 thz . for other conditions the ratio was even larger or
the @xmath54-related photon drag contribution was not detectable .
the only reason why we were able to detect such a small contribution at all , is that photon drag due to _ in - plane _ wavevector vanishes at normal incidence , whereas that caused by @xmath54 component achieves its maximum .
finally we note that as @xmath96 constant has the frequency dependence different from @xmath95 .
in particular , the role of @xmath54-photon drag current is enhanced at small frequencies @xmath238 .
to summarize , our experiments on a large set of _ n_- and _ p_-type ( bi@xmath0sb@xmath1)@xmath2te@xmath3 three - dimensional topological insulators demonstrated that normal incident thz radiation results in the photogalvanic current induced in the surface states . at oblique incidence ,
however , in particular at large angles of incidence , it is outweighed by the photon drag effect .
the developed microscopic model and theory show that the photon drag photocurrent is caused by the _ dynamical _ momentum alignment by time and space dependent radiation electric field and implies the difference in the scattering probabilities for different half periods of the electromagnetic wave .
both photocurrents observed even at room temperature , stem from scattering events and , therefore , can be applied to study the high frequency conductivity in ti .
phenomenological analysis for c@xmath115 symmetry at linear polarization of radiation accounting for the photogalvanic and photon drag effects yields even in @xmath19 photocurrents which can be conveniently written in the following form : @xmath239 we consider oblique incidence with an incidence plane at an angle @xmath24 with the ( @xmath240 ) plane , and @xmath30 is an angle between the radiation electric vector and the incidence plane ( @xmath241 corresponds to @xmath9 polarization ) . in these notations
we have : @xmath242 @xmath243 @xmath244 here @xmath12 is the electric field amplitude in vacuum , and @xmath90 , @xmath89 are fresnel transmission coefficients for @xmath91 and @xmath9 polarizations .
substitution of the wavevector and electric field components into eq .
yields the photocurrent in the following form : @xmath245 @xmath246 here the amplitudes are functions of the incidence angle : @xmath247,\ ] ] @xmath248 the phase shifts are given by : @xmath249 and the @xmath30-independent offset photocurrent is : @xmath250e_0 ^ 2,\ ] ] where @xmath251 .
\nonumber\end{aligned}\ ] ] we thank m.m . glazov and s.a .
tarasenko , for fruitful discussions .
the support from the dfg priority program spp1666 and virtual institute for topological insulators , the elite network of bavaria ( k - nw-2013 - 247 ) , and the russian foundation of basic research is gratefully acknowledged .
p. olbrich , l.e .
golub , t. herrmann , s.n .
danilov , h. plank , v.v .
belkov , g. mussler , ch .
weyrich , c.m .
schneider , j. kampmeier , d. grtzmacher , l. plucinski , m. eschbach , and s.d .
ganichev , phys .
lett . * 113 * , 096601 ( 2014 ) .
p. olbrich , c. zoth , p. vierling , k .-
dantscher , g.v .
budkin , s.a .
tarasenko , v.v .
belkov , d.a .
kozlov , z.d .
kvon , n.n .
mikhailov , s.a .
dvoretsky , and s.d .
ganichev , phys . rev .
b * 87 * , 235439 ( 2013 ) .
dantscher , d.a .
kozlov , p. olbrich , c. zoth , p. faltermeier , m. lindner , g. v. budkin , s. a. tarasenko , v. v. belkov , z.d .
kvon , n. n. mikhailov , s.a .
dvoretsky , d. weiss , b. jenichen , and s. d. ganichev , phys .
b * 92 * , 165314 ( 2015 ) . c. zoth , p. olbrich , p. vierling , k
dantscher , v.v .
belkov , m.a .
semina , m.m .
glazov , l.e .
golub , d.a .
kozlov , z.d .
kvon , n.n .
mikhailov , s.a .
dvoretsky , and s.d .
ganichev , phys .
b * 90 * , 205415 ( 2014 ) . c. weyrich , m. drgeler , j. kampmeier , m. eschbach , g. mussler , t. merzenich , t. stoica , i. e. batov , j. schubert , l. plucinski , b. beschoten , c. m. schneider , c. stampfer , d. grtzmacher , and th .
schpers , arxiv 1511.00965v2 ( 2015 ) .
m. eschbach , e. myczak , j. kellner , j. kampmeier , m. lanius , e. neumann , c. weyrich , m. gehlmann , p. gospodari , s. dring , g. mussler , n. demarina , m. luysberg , g. bihlmayer , th .
schpers , l. plucinski , s. blgel , m. morgenstern , c. m. schneider , and d. grtzmacher , nat . commun .
* 6 * , 8816 ( 2015 ) .
d. hsieh , y. xia , d. qian , l. wray , j. h. dil , f. meier , j. osterwalder , l. patthey , j. g. checkelsky , n. p. ong , a. v. fedorov , h. lin , a. bansil , d. grauer , y. s. hor , r. j. cava , and m. z. hasan , nature ( london ) * 460 * , 1101 ( 2009 ) .
j. kampmeier , s. borisova , l. plucinski , m. luysberg , g. mussler , and d. grtzmacher , cryst .
growth des . ,
* 15 * , 1 ( 2015 ) .
s. d. ganichev and w. prettl , _ intense terahertz excitation of semiconductors _
( oxford univ . press ,
oxford , 2006 ) .
the polarization independent offsets @xmath39 and @xmath40 , being in most samples much smaller than @xmath48 , may be caused by the non perfectly flat surface , which locally reduces the symmetry of the surface states and allows a polarization independent contribution to the both effects @xcite .
the difference in the photocurrent magnitudes @xmath48 for front and back illuminations may be additionally affected by the unequal photocurrent contributions excited in the top and interface surfaces separated by the bulk material , e.g. , due to different scattering times .
ganichev , petra schneider , v. v. belkov , e. l. ivchenko , s.a .
tarasenko , w. wegscheider , d. weiss , d. schuh , b. n. murdin , p. j. phillips , c. r. pidgeon , d. g. clarke , m. merrick , p. murzyn , e. v. beregulin , and w. prettl , phys .
b. r * 68 * , 081302 ( 2003 ) .
j. karch , p. olbrich , m. schmalzbauer , c. zoth , c. brinsteiner , m. fehrenbacher , u. wurstbauer , m. m. glazov , s. a. tarasenko , e. l. ivchenko , d. weiss , j. eroms , r. yakimova , s. lara - avila , s. kubatkin , and s. d. ganichev , phys .
lett . * 105 * , 227402 ( 2010 ) .
w. weber , l. e. golub , s. n. danilov , j. karch , c. reitmaier , b. wittmann , v. v. belkov , e. l. ivchenko , z. d. kvon , n. q. vinh , a. f. g. van der meer , b. murdin , and s. d. ganichev , phys .
b * 77 * , 245304 ( 2008 ) .
the stationary correction to the distribution function @xmath256 is obtained by writing it as an expansion in powers of the electric field @xmath257 with the oscillating in time term @xmath258 and the stationary term @xmath259 being _ second order _ in the electric field .
v. i. belinicher and b. i. sturman , sov .
* 23 * , 199 ( 1980 ) .
the oscillating in time term @xmath260 , see @xcite , does not disturb the balance of scattered carriers , and consequently , does not results in a @xmath120 current .
note that , in spite of the fact that @xmath261 , the correction to the electron energy is quadratic in @xmath167 because linear in @xmath167 terms are forbidden by time - inversion symmetry , so that @xmath262 \cos 6 \varphi_{\bm p}$ ] . note that if one first take into account @xmath224 then the correction to the distribution function contains @xmath263-like bilinear combinations of the electric field amplitudes which are @xmath185-independent , and ,
hence , its gradient is zero . | we report on the observation of a terahertz radiation induced photon drag effect in epitaxially grown _ n_- and _ p_-type ( bi@xmath0sb@xmath1)@xmath2te@xmath3 three dimensional topological insulators with different antimony concentrations @xmath4 varying from 0 to 1 .
we demonstrate that the excitation with polarized terahertz radiation results in a _
dc _ electric photocurrent . while at normal incidence a current arises due to the photogalvanic effect in the surface states , at oblique incidence it is outweighed by the trigonal photon drag effect .
the developed microscopic model and theory show that the photon drag photocurrent is due to the dynamical momentum alignment by time and space dependent radiation electric field and implies the radiation induced asymmetric scattering in the electron momentum space . |
it is well known that in quantum field theory under external conditions , quantities like vacuum energies and effective actions , which describe the influence of boundaries or external fields on the physical system , are generically divergent and require a renormalization to get a physical meaning . in this context , a powerful and elegant regularization scheme to deal with these problems is based on the use of the @xmath1-function @xcite or the heat - kernel ( for recent reviews see , for example , @xcite ) associated to the relevant differential operators appearing in the quadratic part of the actions . in this way , ground state energies , heat - kernel coefficients , functional determinants and partition functions for quantum fields can be given in terms of the corresponding @xmath1-function , where the ultraviolet divergent pieces of the one - loop contributions are encoded as poles of its holomorphic extension .
thus , it is of major interest in physics to determine the singularity structure of @xmath1-functions associated with these physical models . in particular @xcite , for an elliptic boundary value problem in a @xmath3-dimensional compact manifold with boundary , described by a differential operator @xmath4 of order @xmath5 , with smooth coefficients and a ray of minimal growth , defined on a domain of functions subject to local boundary conditions
, the @xmath1-function @xmath6 has a meromorphic extension to the complex @xmath7-plane whose singularities are isolated simple poles at @xmath8 , with @xmath9 in the case of positive definite operators , the @xmath1-function is related , via mellin transform , to the trace of the heat - kernel of the problem , and the pole structure of @xmath10 determines the small-@xmath2 asymptotic expansion of this trace @xcite : @xmath11 where the coefficients are related to the residues by @xmath12 for operators of the form @xmath13 with a singular potential @xmath14 asymptotic to @xmath15 as @xmath16 , this expansion is substantially different . if @xmath17 , the operator is essentially self - adjoint .
this case has been treated in @xcite , where log terms are found , as well as terms with coefficients which are distributions concentrated at the singular point @xmath18 . for the case
@xmath19 , the friedrichs extension has been treated in @xcite for operators in @xmath20 , and in @xcite for operators in @xmath21 , making use of the scale invariance of the operator domain and explicit representations of the resolvent .
moreover , as a particular case of a manifold with an isolated conic singularity , reference @xcite gave a description of the boundary behavior of the friedrichs heat - kernel which does not make use of the resolvent , and showed va boundary maps how it can be used to construct the heat - kernel for other self - adjoint extensions of these operators , showing explicitly the first two terms in the asymptotic expansion of the trace of their difference .
on the other hand , reference @xcite gave the pole structure of the @xmath1-function of a second order differential operator defined on the ( non compact ) half - line @xmath22 , having a singular zero - th order term @xmath23 .
it showed that , for a certain range of real values of @xmath24 , this operator admits nontrivial self - adjoint extensions in @xmath21 , for which the associated @xmath1-function ( given by an integral representation ) presents isolated simple poles which ( in general ) do not lie at @xmath25 for @xmath26 ( as would be the case for a regular @xmath14 ) , and can even take irrational values .
a similar structure has been noticed in @xcite for the singularities of the @xmath1 and @xmath27-functions of a system of first order differential operators with a singular zero - th order term @xmath28 , which also admits a family of self - adjoint extensions for real @xmath29 taking values in certain range .
it has been shown that , in the general case , the asymptotic expansion of the resolvent contains @xmath30dependent powers of @xmath0 which make the @xmath1 and @xmath27-functions to present poles lying at points which depend on the singularity , with residues depending on the self - adjoint extension .
let us mention that singular potentials @xmath31 have been considered in the description of several physical systems , like the calogero model @xcite , conformal invariant quantum mechanical models @xcite and , more recently , the dynamics of quantum particles in the asymptotic near - horizon region of black - holes @xcite .
the self - adjoint extensions of these operators have also been considered in @xcite . moreover , singular superpotentials has been considered as possible agents of supersymmetry breaking in models of supersymmetric quantum mechanics @xcite . it is the aim of the present article to analyze the behavior of the resolvent , the @xmath1-function and the trace of the heat - kernel of a second order differential operator with a regular singularity in a compact segment , @xmath32 , for those values of @xmath29 for which it admits a family of self - adjoint extensions . following the scheme developed in @xcite
, we will show that the asymptotic expansion for the resolvent in the general case presents powers of @xmath0 which depend on the singularity , and can even take irrational values .
the consequence of this behavior on the corresponding @xmath1-function is the presence of simple poles lying at points which also depend on the singularity , with residues depending on the self - adjoint extension considered .
we first construct the resolvents for two particular extensions , for which the boundary condition at the singular point @xmath18 is invariant under the scaling @xmath33 .
the resolvent expansion for these special extensions displays the usual powers , leading to the usual poles for the @xmath1-function ( and the usual structure for the asymptotic expansion of the heat - kernel trace ) .
the resolvents of the remaining extensions are convex linear combinations of these special extensions , but the coefficients in the convex combination depend on the eigenvalue parameter @xmath0 .
this dependence leads to unusual powers in the resolvent expansion , and hence to unusual poles for the zeta - function ( and unusual powers in the asymptotic expansion of the heat - kernel trace ) .
these self - adjoint extensions are not invariant under the scaling @xmath34 .
as @xmath35 they tend ( at least formally ) to one of the invariant extensions , and as @xmath36 they tend to the other .
as @xmath35 the residues at the anomalous poles tend to zero , whereas as @xmath36 these residues become infinite . the way these residues depend on the boundary condition
is explained by a scaling argument in section 7 .
the structure of the article is as follows : in section [ the - operator ] we define the operator and determine its self - adjoint extensions for @xmath37 , and in section [ the - spectrum ] we study their spectra . in section [ the - resolvent ]
we construct the resolvent for a general extension as a linear combination of the resolvent of two limiting cases , and in section [ trace - resolvent ] we consider the traces of these operators .
the asymptotic expansions of these traces , evaluated in section [ asymptotic - expansion ] , are used in section [ spectral - functions ] to construct the associated @xmath1-function and study its singularities , as well as the small - t asymptotic expansion of the heat - kernel trace . the special case @xmath38 is considered in appendix [ g=1/2 ] .
let us consider the differential operator @xmath39 with @xmath40 , defined on a domain of smooth functions with compact support in a segment , @xmath41 .
it can be easily seen that @xmath42 so defined is symmetric .
the adjoint operator @xmath43 , which is the maximal extension of @xmath42 , is defined on the domain @xmath44 of functions @xmath45 , having a locally sumable second derivative and such that @xmath46 [ lema1 - 1 ] if @xmath47 and @xmath48 , then will be considered separately , in appendix [ g=1/2 ] . ]
@xmath49\ , x^g + c_2[\phi]\ , x^{1-g}}{\sqrt{2g-1}}\right)\right| \leq \frac{\|d_x\phi(x)\|}{(3/2 -g)\sqrt{2g+1 } } \ x^{3/2}}\ ] ] and @xmath50\ , x^{g-1 } + ( 1-g)\ , c_2[\phi]\ , x^{-g}}{\sqrt{2g-1}}\right)\right| \leq \frac{3/2 \ , \|d_x\phi(x)\|}{(3/2 -g)\sqrt{2g+1 } } \ x^{1/2}}\ ] ] for some constants @xmath51 $ ] and @xmath52 $ ] , where @xmath53 is the @xmath54-norm . * proof : * let us write @xmath55 .
then , eq . ( [ dphi ] ) implies @xmath56 for some constants @xmath57 and @xmath58 .
now , taking into account that @xmath59 we immediately get eqs .
( [ lemai-1 ] ) and ( [ lemai-2 ] ) .
let @xmath60 and @xmath48
. then @xmath61^ * c_2[\phi ] - c_2[\psi]^ * c_1[\phi]\big\}+ \big\{\psi(1)^*\ , \phi'(1 ) - \psi'(1)^*\ , \phi(1 ) \big\}\ , .
\end{array}\ ] ] * proof : * from eq .
( [ dphi ] ) one easily obtains @xmath62 from which , taking into account the results in lemma [ lema1 - 1 ] , eq .
( [ ddstar ] ) follows directly .
now , if @xmath63 in eq .
( [ ddstar ] ) belongs to the domain of the closure of @xmath42 , @xmath64 , @xmath65 then the right hand side of eq .
( [ ddstar ] ) must vanish for any @xmath47 .
therefore @xmath66=c_2[\psi]=\psi(1)=\psi'(1)=0\ , .\ ] ] on the other hand , if @xmath67 belong to the domain of a symmetric extension of @xmath42 ( contained in @xmath44 ) , the right hand side of eq . ( [ ddstar ] )
must also vanish .
thus , the closed extensions of @xmath42 correspond to the subspaces of @xmath68 under the map @xmath69 , c_2[\phi ] , \phi(1 ) , \phi'(1 ) \right)$ ] , and the self - adjoint extensions correspond to those subspaces @xmath70 such that @xmath71 , with the orthogonal complement taken in the sense of the symplectic form on the right hand side of eq.([ddstar ] ) . for definiteness , in the following we will consider self - adjoint extensions satisfying the local boundary condition @xmath72 each such extension
is determined by a condition of the form @xmath73 + \beta\ , c_2[\phi ] = 0\ , , \ ] ] with @xmath74 , and @xmath75 .
we denote this extension by @xmath76 .
in order to determine the spectrum of the self - adjoint extensions of @xmath42 for @xmath77 , we need the solutions of @xmath78 satisfying the boundary conditions in eqs .
( [ bc1 ] ) and ( [ bc2 ] ) .
the general solution of the homogeneous equation for @xmath79 is @xmath80 and the boundary conditions in eqs .
( [ bc1 ] ) and ( [ bc2 ] ) imply that @xmath81 consequently , there are no zero modes except for the self - adjoint extension characterized by @xmath82 . for @xmath83 , the solutions of eq .
( [ ec - hom ] ) are of the form @xmath84 where @xmath85 , and the @xmath86-dependent coefficients have been introduced for later convenience .
taking into account that @xmath87 we get from eqs .
( [ lemai-1 ] ) and ( [ bc2 ] ) @xmath88 on the other hand , the condition in eq .
( [ bc1 ] ) implies @xmath89 for @xmath90 , eq .
( [ ec - spectrum ] ) implies @xmath91 ( dirichlet boundary conditions at the origin ) .
therefore , @xmath92 .
thus , the spectrum of this self - adjoint extension is positive and non - degenerate , with the eigenvalues of @xmath93 given by @xmath94 where @xmath95 is the @xmath96-th positive zero of the bessel function @xmath97 let us recall that large zeros of @xmath98 have the asymptotic expansion @xmath99 with @xmath100 . ] . for @xmath101 , from eqs .
( [ ec - spectrum ] ) and ( [ sol - hom-1-en1 ] ) we easily get the following transcendental equation for the eigenvalues of @xmath102 : @xmath103 where we have defined @xmath104 for the positive eigenvalues @xmath105 , both sides in eq.([spectrum ] ) have been plotted in figure [ figure ] , for particular values of @xmath106 and @xmath29 . moreover , if @xmath107 , and the extension @xmath102 has a negative eigenvalue . indeed , if @xmath108 , then @xmath109 where @xmath110 is the modified bessel function . for a plot ,
see figure [ lambda - neg ] as @xmath111 , while the corresponding eigenfunction tends to concentrate on the singularity at @xmath18 .
see also @xcite . ] .
notice that the spectrum is always non - degenerate , and there is a positive eigenvalue between each pair of consecutive squared zeroes of @xmath112 .
therefore , from eq.([large - zeroes ] ) we get @xmath113 . in particular , for the @xmath114 extension ( which we call the
n - extension " ) , @xmath115 , it can be seen fron eq.([spectrum ] ) that the eigenvalues are given by @xmath116 where @xmath117 are the positive zeroes of @xmath118 .
in this section we will construct the resolvent of @xmath42 , @xmath119 for its different self - adjoint extensions when @xmath48 .
we will first consider the two limiting cases in eq .
( [ bc2 ] ) , namely the @xmath120-extension " , for which @xmath121 = 0 $ ] , and the @xmath122-extension " , with @xmath123=0 $ ] .
the resolvent for a general self - adjoint extension will be later evaluated as a linear combination of those obtained for these two limiting cases . for the kernel of the resolvent we have @xmath124 where @xmath125 , with @xmath126 . to proceed , we need some particular solutions of the homogeneous equation ( [ ec - hom ] ) .
then , let us define @xmath127 notice that @xmath128
. we will also need the wronskians @xmath129
= \frac{2\,\cos ( g\,\pi ) } { \pi } \,j_{g-\frac{1}{2}}(\mu ) = \frac{1 } { \gamma_d(\mu)}\ , , } \\ \\
\displaystyle { w\left [ l^n(x,\mu ) , r(x,\mu ) \right ] = \frac{2\,\cos ( g\,\pi ) } { \pi } \,j_{\frac{1}{2 } - g}(\mu ) = \frac{1 } { \gamma_n(\mu)}\ , , } \end{array } \right.\ ] ] which vanish only at the zeroes of @xmath130 , for @xmath131 . in this case
, the function @xmath132 must satisfy @xmath133 and @xmath52=0 $ ] , for any function @xmath134 .
this requires that @xmath135 the fact that the boundary conditions are satisfied , as well as @xmath136 , can be straightforwardly verified from eqs .
( [ soluciones ] ) and ( [ w - d ] ) . indeed , from eqs .
( [ func - inhom ] ) , ( [ gd11 ] ) , ( [ soluciones ] ) and ( [ w - d ] ) , one gets @xmath137}{\sqrt{2\,g-1 } } \ , x^g + o({x}^{3/2})\ , , \ ] ] with @xmath138 = \frac{\pi\ , \mu^{g-\frac 1 2 } \ , \sqrt{2\,g-1 } } { 2^{\frac{1}{2}+ g } \cos(g\ , \pi ) j_{g-\frac{1}{2}}({\mu } ) \ , \gamma\left(\frac{1}{2 } + g\right ) } \
\int_0 ^ 1 r(y , \mu ) f(y ) \ , dy \ , , \
] ] for @xmath86 not a zero of @xmath139 .
notice that @xmath140\neq 0 $ ] if the integral in the right hand side of eq .
( [ c+ ] ) is non vanishing . in this case
, the function @xmath141 must satisfy @xmath133 and @xmath51=0 $ ] , for any function @xmath134 .
this requires that @xmath142 these boundary conditions , as well as the fact that @xmath143 , can be straightforwardly verified from eqs.([soluciones ] ) and ( [ w - d ] ) . in this case , from eqs .
( [ func - inhom - n ] ) , ( [ gn11 ] ) , ( [ soluciones ] ) and ( [ w - d ] ) , one gets @xmath144}{\sqrt{2\,g-1 } } \ , x^{1-g } + o({x}^{3/2})\ , , \ ] ] with @xmath145 = \frac{\pi\ , \mu^{\frac 1 2 -g } \ , { \sqrt{2\,g-1 } } } { 2^{\frac{3}{2}- g } \cos(g\ , \pi ) j_{\frac{1}{2}-g}({\mu } ) \ ,
\gamma\left(\frac{3}{2}-g\right ) } \int_0 ^ 1 r(y,\mu ) f(y ) \ , dy\ , , } \end{array}\ ] ] for @xmath86 not a zero of @xmath146 .
notice that @xmath147\neq 0 $ ] if the integral in the right hand side of eq .
( [ c- ] ) ( the same integral as the one appearing in the @xmath120-extension , eq .
( [ c+ ] ) ) is non vanishing . for the general case , we can adjust the boundary conditions @xmath148 + \beta\ , c_2[\phi ] = 0\ , , \ \alpha,\beta\neq 0\ , , \ ] ] for @xmath149 for any @xmath134 , by taking a linear combination of the resolvent for the limiting cases , @xmath150 g_d(x , y ; \lambda ) + \tau(\lambda)\ , g_n(x , y;\lambda)\ , .\ ] ] since the boundary condition at @xmath151 is automatically fulfilled
, one must just impose @xmath152 c_1^d[\phi ] + \beta\ , \tau(\lambda)\ , c_2^n[\phi ] = 0\ , .\ ] ] notice that , in view of eq .
( [ c+ ] ) , ( [ c- ] ) and ( [ spectrum ] ) , @xmath153-\beta\ , c_2^n[\phi]=0\ ] ] precisely when @xmath154 is an eigenvalue of @xmath102 . therefore , from eq .
( [ ec - tau ] ) we get the resolvent of @xmath102 by setting @xmath155 } { \alpha \ , c_1^d[\phi]-\beta\ , c_2^n[\phi ] } } = \\ \\ = \displaystyle { \frac{1}{1 - \rho ( \alpha , \beta ) \,{\mu } ^{1 - 2\,g}\ , \displaystyle { \frac { j_{g-\frac{1}{2}}(\mu ) } { j_{\frac{1 } { 2}-g}(\mu ) } } } } \ , , \end{array}\ ] ] for @xmath86 not a zero of @xmath156 .
it follows from eq .
( [ linear - comb ] ) that the resolvent of a general self - adjoint extension of @xmath42 can be expressed in terms of the resolvents of the two limiting cases , @xmath157 and @xmath158 .
moreover , since the eigenvalues of any extension grow as @xmath159 ( see section [ the - spectrum ] ) , these resolvents are trace class operators
. then , we have @xmath160\ ] ] from eqs .
( [ gd11 ] ) and ( [ gn11 ] ) we straightforwardly get ( see appendix [ integrals ] for the details ) @xmath161 and @xmath162 where we have taken into account that @xmath163 finally , we get @xmath164
using the hankel asymptotic expansion for bessel functions @xcite ( see appendix [ hankel ] ) , we get for the first term in the right hand side of eq .
( [ trg2 ] ) @xmath165 where @xmath166 for @xmath167 , and @xmath168 for @xmath169 .
the coefficients in this series can be straightforwardly evaluated from eqs .
( [ p+q ] ) and ( [ t ] ) .
notice that @xmath170 , since @xmath171 is real and @xmath172 is pure imaginary .
similarly , from ( [ jprimasobrejasymp ] ) we simply get for the second factor in the second term in the right hand side of eq.([trg2 ] ) @xmath173 finally , taking into account eq .
( [ jsobrejup ] ) , we have @xmath174 where @xmath175 ( @xmath176 ) corresponds to @xmath167 ( @xmath169 ) .
notice the appearance of @xmath29-dependent powers of @xmath86 in this asymptotic expansion .
the @xmath1-function for a general self - adjoint extension of @xmath42 is defined , for @xmath177 , as @xmath178 where the curve @xmath179 encircles counterclockwise the spectrum of the operator , keeping to the left of the origin . according to eq .
( [ trazag ] ) , we have @xmath180 where @xmath181 is the @xmath1-function for the @xmath120-extension .
since , according to the discussion in section [ the - spectrum ] , @xmath182 has a positive spectrum , and the self - adjoint extension @xmath102 has at most one negative eigenvalue , we can write @xmath183 where @xmath184 if there is a negative eigenvalue , and vanishes otherwise . we can also write @xmath185 where @xmath186 is an entire function . therefore , in order to determine the poles of @xmath187 , we can subtract and add a partial sum of the asymptotic expansion obtained in the previous section to @xmath188 in the integrands in the right hand side of eq .
( [ zeta+1 ] ) . in so doing , we get for the @xmath120-extension and for a real @xmath189 @xmath190 where @xmath191 is holomorphic in the open half plane @xmath192 .
consequently , the meromorphic extension of @xmath193 presents simple poles at @xmath194 with residues @xmath195 where the coefficients @xmath196 are given in eq.([asymp - trgd - upp ] ) .
notice that these residues vanish for even @xmath197 . in particular , for @xmath198 ( @xmath199 ) one
gets @xmath200 this is the unique pole present in @xmath193 for the @xmath201 case , where there is no singularity in the 0-th order coefficient of @xmath42 . for a general self - adjoint extension @xmath102
, we must also consider the singularities coming from the asymptotic expansion of @xmath202 in eq .
( [ trazag ] ) , given in eqs .
( [ trdifasymp ] ) and ( [ tau - asymp ] ) . from eq .
( [ zeta+ ] ) , and taking into account eq.([zeta+1 ] ) , for real @xmath189 we can write @xmath203 where @xmath204 is holomorphic for @xmath205 . therefore , @xmath206 has a meromorphic extension which presents simple poles located at negative @xmath29-dependent positions , @xmath207 with residues which depend on the self - adjoint extension given by @xmath208 } } \ , .
\end{array}\ ] ] notice that these poles are irrational for irrational values of @xmath29 .
moreover , the residues vanish for the n - extension " ( @xmath209 ) , and have a singular limit for @xmath210 .
in particular , these poles for the @xmath201 case ( for which there are no singularity in the zero - th order term of @xmath42 ) are negative half - integers , since in this case the residues vanish for even @xmath197 .
it is interesting to notice that the poles in eq.([polos - zeta ] ) are also poles of the @xmath1-function of the corresponding self - adjoint extension of the operator @xmath211 in @xmath212 considered in @xcite , with exactly the same residues , as can be easily verified .
let us remark that when @xmath213 the residue of @xmath214 at @xmath215 is a constant times @xmath216 .
this is consistent with the behavior of @xmath42 under the scaling isometry @xmath217 taking @xmath218 .
the extension @xmath102 is unitarily equivalent to the operator @xmath219 similarly defined on @xmath220 , with @xmath221 and @xmath222 : @xmath223 notice that only for the extensions with @xmath224 or @xmath114 the boundary condition at the singular point @xmath18 , eq.([bc2 ] ) , is left invariant by this scaling . therefore , we have for the @xmath1-function of the scaled problem @xmath225 and for the residues @xmath226 the factor @xmath227 exactly cancels the effect the change in the boundary condition at the singularity has on @xmath106 , @xmath228 then , the difference between the intervals @xmath229 and @xmath230 has no effect on the structure of these residues , which presumably are determined locally in a neighborhood of @xmath18 . in this way
we conclude that , for a general self - adjoint extension , the presence of poles in the @xmath1-function located at @xmath29-dependent positions is a consequence of the singular behavior ( @xmath231 ) of the zero - th order term in @xmath42 near the origin , together with a scaling non - invariant boundary condition at the singularity . finally , let us remark that the relation between the @xmath1-function and the trace of the heat - kernel of @xmath102 , @xmath232 where @xmath233 is an entire function , straightforwardly lead to the following small-@xmath2 asymptotic expansion , @xmath234k\right ) \frac{2\,g-1}{2\ , \pi}\,\rho(\alpha,\beta)^k \ { \sin\left [ \frac{\pi}{2}\left ( 2g-1\right)k \right]}\right\ } t^{\left(g-\frac{1}{2}\right)k}\ , .
\end{array}\ ] ] the first term in the right hand side , coming from eq.([trdifasymp ] ) and the first term in the asymptotic expansion of @xmath235 in eq .
( [ tau - asymp ] ) , coincides with the result reported in @xcite .
notice also the @xmath29-dependent powers of @xmath2 appearing in the asymptotic series in the right hand side of eq .
( [ heat - asymp ] ) for any general self - adjoint extension ( except for the
n - extension " , for which @xmath236 ) .
in particular , the first term in this series reduces to @xmath237 this power of @xmath2 also coincides with the result quoted in @xcite , but we find a different coefficient .
* acknowledgements : * we would like to thank prof.robert seeley for useful discussions . h.f . and p.a.g.p .
acknowledge support from universidad nacional de la plata ( grant 11/x298 ) and conicet ( grant 0459/98 ) , argentina .
acknowledge support from universidad nacional de la plata ( grant 11/x228 ) , argentina .
the case @xmath238 , for which the differential operator @xmath42 takes the form @xmath239 requires a separate consideration which we briefly present in this appendix . along the same lines as in the proof of lemma [ lema1 - 1 ] , it is straightforward to show that , if @xmath47 , then @xmath240\ , \sqrt{x } + c_2[\phi]\ , \sqrt{x}\ , \log{x}\right)\right| \leq \frac{\|d_x\phi(x)\|}{\sqrt{2 } } \ x^{3/2}}\ ] ] and @xmath241\,x^{-1/2}+ c_2[\phi]\left(x^{-1/2}+\frac{1}{2}x^{-1/2}\log{x } \right)\right]\right| \leq } \\ \\ \displaystyle{\leq \frac{3}{2\sqrt{2}}\|d_x\phi(x)\|\ , x^{1/2 } } \end{array}\ ] ] for some constants @xmath51 $ ] and @xmath52 $ ] , where @xmath53 stands for the @xmath54-norm .
therefore , it is easy to see that eq .
( [ ddstar ] ) is also valid in the present case , and the self - adjoint extensions of @xmath42 correspond again to those subspaces @xmath70 such that @xmath71 , with the orthogonal complement taken in the sense of the symplectic form on the right hand side of eq.([ddstar ] ) .
if , in addition , we select the dirichlet condition at @xmath151 , @xmath133 , the remaining self - adjoint extensions of @xmath42 correspond to a one - parameter family characterized by eq.([bc2 ] ) , @xmath102 .
there exists a particular self - adjoint extension for which @xmath52=0 $ ] , namely @xmath93 , such that the functions in its domain behave near the origin as @xmath242\,\sqrt{x}+o(x^{3/2}).\ ] ] the eigenfunction of @xmath182 corresponding to the eigenvalue @xmath0 is given by , @xmath242\sqrt{x}\ , j_0(\mu x),\ ] ] where @xmath154 and @xmath86 is a ( positive ) zero of @xmath243 . for an arbitrary self - adjoint extension @xmath102 with @xmath101 ,
the eigenfunction corresponding to the eigenvalue @xmath154 is given by @xmath244-c_2[\phi](\log{\mu}-\log{2}+\gamma ) \right\}\,\sqrt{x}\,j_0(\mu x)\ , + \\ \\
\displaystyle { + \frac{\pi}{2}\ , c_2[\phi]\,\sqrt{x}\,n_0(\mu x ) } \ , , \end{array}\ ] ] where @xmath51,c_2[\phi]$ ] are constrained by eq .
( [ bc2 ] ) .
the condition @xmath133 leads to the equation @xmath245 where @xmath246 , which determines the spectrum of @xmath102 .
notice that there are no negative eigenvalues . in order to determine the kernels of the resolvents @xmath247 and @xmath248 , we define @xmath249 to get @xmath250 } \left\{\begin{array}{lr } \mathcal{l}^{d}(x;\mu)\,\mathcal{r}(y;\mu)\ , , & x\leq y \ , , \\ \\
\mathcal{l}^{d}(y;\mu)\,\mathcal{r}(x;\mu)\ , , & x\geq y \ , , \end{array}\right . } \end{array}\ ] ] and @xmath251 } \left\{\begin{array}{lr } \mathcal{l}^{(\alpha,\beta)}(x;\mu)\,\mathcal{r}(y;\mu)\ , , & x\leq y\ , , \\ \\
\mathcal{l}^{(\alpha,\beta)}(y;\mu)\,\mathcal{r}(x;\mu)\ , , & x\geq y\ , , \end{array}\right . } \end{array}\ ] ] where the wronskians can be easily computed from ( [ lyr ] ) , @xmath252= \frac{2}{\pi}\ , j_0(\mu ) } \ , , \\ \\
\displaystyle { w[\mathcal{l}^{(\alpha,\beta)}(x;\mu ) , \mathcal{r}(x;\mu)]=\frac{2}{\pi}\ , ( \theta-\log{\mu})j_0(\mu)+ n_0(\mu)}\ , .
\end{array}\ ] ] from eq .
( [ large - zeroes ] ) , it can be seen that both @xmath253 and @xmath254 are trace class operators .
now , taking into account that @xcite @xmath255 where @xmath256 or @xmath257 , the traces of the resolvents can be readily computed to get @xmath258 from eqs .
( [ jupper ] - [ n - sigma ] ) one straightforwardly gets the same asymptotic expansion for these two traces , @xmath259 where @xmath260 ( @xmath261 ) for @xmath167 ( @xmath169 ) .
notice that the asymptotic series in eq .
( [ tr - gd-0 ] ) coincides with the right hand side of eq .
( [ asymp - trgd - upp ] ) evaluated at @xmath238 . therefore , from eq .
( [ zetad+1 ] ) one concludes that , in the present case , @xmath187 has simple poles only at @xmath262 , for @xmath263 , with residues given by @xmath264 ( vanishing for even @xmath197 ) for all the self - adjoint extensions of @xmath42 .
so , in contrast to the case of @xmath265 , the pole structure of the @xmath1-function for @xmath238 is independent of the self - adjoint extension considered and does not differ from the usual one .
in this appendix we briefly describe the evaluation of the traces appearing in section [ trace - resolvent ] . from eq .
( [ gd11 ] ) we get for the kernel of @xmath157 on the diagonal @xmath266 therefore , in order to evaluate its trace it is sufficient to know the primitives @xcite @xmath267 and @xmath268}\ , , \end{array}\ ] ] where @xmath269 these primitives , together with the relation @xmath270 necessary to simplify the intermediate results , straightforwardly lead to eq .
( [ traza - gd ] ) . similarly , for the kernel of @xmath158 on the diagonal we have @xmath271 the same argument as before leads to eq .
( [ traza - gn ] ) .
to develop an asymptotic expansion for the trace of the resolvent we employ the hankel asymptotic expansion for the bessel functions which , for completeness , we briefly describe in this appendix . for the quotient of two bessel functions we have @xmath292 where @xmath175 for @xmath287 and @xmath288 for @xmath289 .
the coefficients of this asymptotic expansion can be easily obtained , to any order , from eq .
( [ p+q ] ) , @xmath293 similarly , the derivative of the bessel function has the following asymptotic expansion @xcite for @xmath273 , @xmath297 and @xmath298 where @xmath299 and @xmath300 then , @xmath301 where the upper sign is valid for @xmath302 , and the lower one for @xmath303 .
we have also @xmath304 with @xmath305 therefore , we get @xmath306 where the upper sign is valid for @xmath302 , and the lower one for @xmath303
. the coefficients of the asymptotic expansion in the right hand side of eq .
( [ jprimasobrej ] ) can be easily obtained from eq .
( [ p+q ] ) and ( [ t ] ) , @xmath307 | we consider the resolvent of a second order differential operator with a regular singularity , admitting a family of self - adjoint extensions .
we find that the asymptotic expansion for the resolvent in the general case presents unusual powers of @xmath0 which depend on the singularity .
the consequences for the pole structure of the @xmath1-function , and for the small-@xmath2 asymptotic expansion of the heat - kernel , are also discussed . |
type - i and type - ii censoring schemes are two most popular censoring schemes which are used in practice .
they can be briefly described as follows .
suppose @xmath0 units are put on a life test .
in type - i censoring , the test is terminated when a pre - determined time , @xmath1 , on test has been reached , and failures after time @xmath1 are not observed . in type - ii
censoring , the test is terminated when a pre - chosen number , @xmath2 , out of @xmath0 items has failed .
it is also assumed that the failed items are not replaced .
so , in type - i censoring scheme , the number of failures is random and in type - ii censoring scheme , the experimental time is random .
a hybrid censoring scheme is a mixture of type - i and type - ii censoring schemes and it can be described as follows .
suppose @xmath0 identical units are put to test .
the test is finished when a pre - selected number @xmath2 out of @xmath0 items are failed , or when a pre - determined time @xmath1 on the test has been obtained . from now on
, we call this type - i hybrid censoring scheme and this scheme has been used as a reliability acceptance test in @xcite .
this censoring scheme was introduced by epstin @xcite , he also studied the life testing data under the assumption of exponential distribution with mean life @xmath3 .
epstein @xcite proposed two - sided confidence intervals for @xmath3 without any formal proof .
fairbanks et al .
@xcite moderated partly the proposition of epstein @xcite and suggested a simple set of confidence intervals .
chen and bhattacharya @xcite earned the exact distribution of the conditional maximum likelihood estimator ( mle ) of @xmath3 and implied a one - sided confidence interval .
childs et al .
@xcite proposed some simplifications of the exact distribution . from the bayesian point of view , drapper and guttmann @xcite studied the same problem , and reached a two - sided credible interval of the mean lifetime based on the gamma prior .
comparison of the different methods using monte carlo simulations , can be found in gupta and kundu @xcite .
for some related work , one may refer to ebrahimi @xcite , jeong et al .
@xcite , childs et al .
@xcite , kundu @xcite , banerjee and kundu @xcite , kundu and pradhan @xcite , dube et al . @xcite and the references cited there .
one of the disadvantages of type - i hybrid censoring scheme is that there may be very few failures occurring up to the pre - fixed time @xmath1 . because of this , childs et al .
@xcite proposed a new hybrid censoring scheme known as type - ii hybrid censoring scheme which can be described as follows .
put @xmath0 identical items on test , and then stop the experiment at the random time @xmath4 , where @xmath2 , and @xmath1 are prefixed numbers and @xmath5 indicates the time of @xmath2th failure in a sample of size @xmath0 . under the type - ii hybrid censoring scheme , we have one of the following three types of observations : + case i : @xmath6
if @xmath7 + case ii : @xmath8 if @xmath9 and @xmath10 + case iii : @xmath11 + where @xmath12 denote the observed ordered failure times of the experimental units . a schematic illustration of the hybrid censoring scheme is presented in figure [ fig1 ] .
( 10,30)(10,15 ) ( 25,25)(1,0)220 ( 35,25)(1,1)20 ( 70,25)(1,1)20 ( 200,25)(1,1)20 ( 25,15)@xmath13 ( 35,50)1-st failure ( 60,15)@xmath14 ( 90,50)2-nd failure ( 190,15)@xmath15 ( 180,50)r - th failure ( experiment stops ) ( 165,15)@xmath1 ( 122,20)@xmath16 ( 300,25)case i ( 35,25 ) ( 70,25 ) ( 170,25 ) ( 10,30)(10,15 ) ( 25,25)(1,0)220 ( 35,25)(1,1)20 ( 70,25)(1,1)20 ( 155,25)(1,1)20 ( 200,25)(1,1)20 ( 25,15)@xmath13 ( 35,50)1-st failure ( 60,15)@xmath14 ( 90,50)2-nd failure ( 110,20)@xmath16 ( 145,15)@xmath17 ( 150,50)d - th failure ( 195,15)@xmath1 ( 215,50)experiment stops ( 300,25)case ii ( 35,25 ) ( 70,25 ) ( 200,25 ) ( 10,30)(10,15 ) ( 25,25)(1,0)220 ( 35,25)(1,1)20 ( 70,25)(1,1)20 ( 155,25)(1,1)20 ( 25,15)@xmath13 ( 35,50)1-st failure ( 60,15)@xmath14 ( 90,50)2-nd failure ( 110,20)@xmath16 ( 145,15)@xmath18 ( 150,50)n - th failure ( experiment stops ) ( 195,15)@xmath1 ( 300,25)case iii ( 35,25 ) ( 70,25 ) ( 200,25 ) ( 10,-5)[fig1]figure 1 : a schematic presentation for type - ii hybrid censored scheme . in this article , we consider the analysis of type - ii hybrid censored lifetime data when the lifetime of each experimental unit follows a two - parameter weighted exponential ( we ) distribution .
this distribution was originally proposed by gupta and kundu @xcite .
the two - parameter we distribution with the shape and scale parameters @xmath19 and @xmath20 , respectively , has the probability density function ( pdf ) as : @xmath21 we denote a two - parameter we distribution with the pdf ( [ we ] ) by @xmath22 and the corresponding cumulative distribution function ( cdf ) by @xmath23 .
the aim of this article is two fold .
first , we try to earn the mle s of the unknown parameters .
it is observed that the maximum likelihood estimators can be obtained implicitly by solving two nonlinear equations , but they can not be obtained in closed form .
so mle s of parameters are derived numerically .
newton - raphson algorithm is one of the standard methods to determine the mle s of the parameters . to employ the algorithm , second derivatives of the log - likelihood
are required for all iterations .
the em algorithm is a very powerful tool in handling the incomplete data problem see dempster et al .
@xcite and mclachlan and krishnan @xcite
. then we use the em algorithm to compute the mle s .
we also evaluate the observed fisher information matrix using the missing information principle which have been used to obtained asymptotic confidence intervals of the unknown parameters .
the second aim of this article is to provide the bayes inference for the unknown parameters for type - ii hybrid censored data .
it is observed that bayes estimators can not be obtained explicitly , we provide two approximations namely lindley s approximation and gibbs sampling procedure .
so we use the gibbs sampling procedure to compute the bayes estimators , and the hpd confidence intervals .
we compare the performances of the different methods by monte carlo simulations , and for illustrative purposes we have analyzed one real data set .
the rest of the article is arranged as follows . in section 2
, we provide the mle s of the unknown parameters .
fisher information matrix is evaluated in section 3 . using lindley s approximation and gibbs sampling we obtain bayes estimators and hpd confidence intervals for the parametes in section 4 .
simulation results are presented in section 5 .
we verify our theoretical results via analyzing data set in section 6 .
in this section , we study mles of the model parameters @xmath24 and @xmath25 for @xmath26 distribution with density function : @xmath27 for simplicity , we apply a re - parametrization as @xmath24 and @xmath28 . by this
, the @xmath26 distribution can be written as : @xmath29 the likelihood function in case i is given by @xmath30 for case ii , @xmath31 and for case iii , @xmath32 where @xmath33 is presented by ( [ bl ] ) , so @xmath34 we present likelihood functions ( [ a ] ) , ( [ b ] ) and ( [ iii ] ) by : @xmath35 where @xmath36 @xmath37 taking the logarithm of equation [ c ] , we obtain @xmath38 @xmath39 then the normal equations are @xmath40 maximum likelihood estimators can be secured by solving these equations , but they can not be expressed explicitly .
so we use em algorithm to compute them .
the advantage of this method is that it is convergence for any initial value fast enough .
+ the em algorithm , originally proposed by dempster et al .
@xcite , is a very powerful tool for handling the incomplete data problem .
+ let us symbolize the observed and the censored data by @xmath41 and @xmath42 , respectively . here for a given r , @xmath43 are not observable .
the censored data vector @xmath44 can be thought of as missing data .
the combination of @xmath45 forms the whole data set .
in next we follow the method kundu and pradhan @xcite for missing data introducing .
+ if we denote the log - likelihood function of the uncensored data set by @xmath46 @xmath47 for the e - step of the em algorithm , one needs to compute the pseudo log - likelihood function as @xmath48 therefor , @xmath49 @xmath50 where @xmath51=e(z_i|z_i > c ) ~~\mbox{and}~~ b(c;\alpha,\beta)=e[\ln(1-e^{-\beta z_i})|z_i > c],\ ] ] and they are obtained in appendix a. now the m - step includes the maximization of the pseudo log - likelihood function [ 2 ] .
therefore , if at the kth stage , the estimation of @xmath52 is @xmath53 , then @xmath54 can be obtained by maximizing @xmath55 @xmath56 note that the maximization of [ 3 ] can be earned quite effectively by the similar method proposed by gupta and kundu @xcite .
first , @xmath57 can be obtain by solving a fixed - point type equation @xmath58 the function @xmath59 is defined @xmath60^{-1}\ ] ] where @xmath61 and @xmath62 one can follow iteration method .
once @xmath57 is determined , @xmath63 can be evaluated as @xmath64 . for the estimation of @xmath25 , we can use the invariance property maximum likelihood estimators and obtain @xmath65 as follow : @xmath66
one of the advantages of using em algorithm is that presents a measure of information in censored data through the missing information principle .
louis @xcite improved a procedure for extracting the observed information matrix . in this section ,
we display the observed fisher information matrix by using the missing value principles of louis @xcite . the observed fisher information matrix can be used to build the asymptotic confidence intervals .
+ using the notations : @xmath67 , x = observed data , w = complete data , @xmath68=observed information , @xmath69=complete information and @xmath70=missing information , follow the relation to @xmath71 to evaluate @xmath72 + complete information and the missing information are given respectively as : @xmath73\ ] ] @xmath74.\ ] ] as the dimension of @xmath3 is 2 , @xmath68 and @xmath70 are both of the order @xmath75 .
the elements of matrix @xmath69 for complete data set are presented in gupta and kundu @xcite .
they re - parametrized @xmath26 distribution as @xmath25 and @xmath76 .
+ we report @xmath69 which have been evaluated by them here as : @xmath77\ ] ] where @xmath78 @xmath79 @xmath80 in which @xmath81 + on the other hand , with the above re - parametrization and by using ( [ iwx ] ) , one can easily verify @xmath82,\ ] ] where @xmath83 @xmath84 @xmath85 in which @xmath86 now , @xmath68 can be computed by ( [ ix ] ) .
the asymptotic variance - covariance matrix of @xmath87 can be obtained by inverting @xmath68 .
we use this matrix to secure the asymptotic confidence intervals for @xmath25 and @xmath88 . to obtain the asymptotic confidence interval for @xmath24
, we use the non - parametric bootstrap method @xcite .
in this section , we study bayes estimators for parameters @xmath24 and @xmath25 under symmetric loss functions .
a very well known symmetric loss function is the squared error which is defined as : @xmath89 with @xmath90 being an estimate of @xmath91 . here
@xmath91 denotes some function of @xmath92 .
bayes estimators , say @xmath93 , is evaluated by the posterior mean of @xmath91 .
let @xmath94 be an observed sample from the hybrid censoring scheme , drawn from a @xmath26 distribution .
we apply re - parametrization as @xmath24 and @xmath76 .
so the likelihood function becomes @xmath95 and @xmath96-likelihood function : @xmath97 @xmath98 it is assumed that @xmath88 and @xmath24 have the following independent gamma priors : @xmath99 @xmath100 so , the joint prior distribution of @xmath24 and @xmath88 is of the form @xmath101 then the posterior distribution @xmath24 and @xmath88 can be written as @xmath102 @xmath103 where @xmath104 @xmath105 now the bayes estimators of @xmath24 and @xmath88 under the squared error loss function l are respectively obtained as : @xmath106=\frac{1}{k}\int_0^\infty\int_0^\infty\alpha^{w_4-n - r}\beta^{w_2+r-1}(\alpha+1)^re^{-\alpha w_3}e^{-\beta w_1}\]]@xmath107 and @xmath108=\frac{1}{k}\int_0^\infty\int_0^\infty\alpha^{w_4-n - r-1}\beta^{w_2+r}(\alpha+1)^re^{-\alpha w_3}e^{-\beta w_1}\]]@xmath109 since @xmath25 is a function of @xmath24 and @xmath88 , then one can obtain the posterior density function of @xmath25 and so the bayes estimator of @xmath25 under the squared error loss function @xmath110 as : @xmath111=\frac{1}{k}\int_0^\infty\int_0^\infty u^{w_4+w_2-n-1}\lambda^{w_2+r-1}(1+u)^{r}e^{-u(w_3+\lambda w_1)}\]]@xmath112 as these estimators can not be evaluated explicitly , so we adopt two different procedures to approximate them : * lindley approximation , * mcmc method . in previous section , based on type - ii hybrid censored scheme we obtained the bayes estimators of @xmath24 , @xmath88 and @xmath25 against squared error loss function @xmath110 .
it is easily observed that theses estimators have not explicit closed forms . for these evaluation ,
numerical techniques are required .
one of the most numerical techniques is lindley s method ( see @xcite ) , that for these estimators can be describe as follows . in general ,
bayes estimator of @xmath113 as a function of @xmath24 and @xmath88 is identified : @xmath114 where @xmath115 is @xmath96-likelihood function ( defined by [ log ] ) and @xmath116 .
+ by the lindley s method @xmath117 can be approximated as : @xmath118@xmath119+\frac{1}{2}[(\hat{u}_{\alpha}\hat{\sigma}_{\alpha\alpha}+ \hat{u}_{\beta}\hat{\sigma}_{\alpha\beta})(\hat{l}_{\alpha\alpha\alpha}\hat{\sigma}_{\alpha\alpha}+ \hat{l}_{\alpha\beta\alpha}\hat{\sigma}_{\alpha\beta}+\hat{l}_{\beta\alpha\alpha}\hat{\sigma}_{\beta\alpha}\]]@xmath120,\ ] ] where @xmath121 and @xmath122 are the mle s of @xmath24 and @xmath88 respectively .
also , @xmath123 is the second derivative of the function @xmath113 with the respect to @xmath24 and @xmath124 valued of @xmath123 at @xmath125 other expressions can be calculated with following definitions : @xmath126 @xmath127 @xmath128 @xmath129 where @xmath130 @xmath131 @xmath132 @xmath133 @xmath134 @xmath135 @xmath136 and we have : @xmath137 with the above defined expressions , we obtain the approximation bayes estimators .
+ also we have : @xmath138 the bayes estimator of @xmath24 under the squared error loss function @xmath110 becomes @xmath139@xmath140.\ ] ] proceeding similarly , the bayes estimator of @xmath88 under @xmath110 is given by @xmath141 @xmath142 @xmath143.\ ] ] finally the bayes estimator of @xmath25 under @xmath110 is given by @xmath144 @xmath145@xmath146+\frac{1}{2}[(\hat{u}_{\alpha}\hat{\sigma}_{\alpha\alpha}+ \hat{u}_{\beta}\hat{\sigma}_{\alpha\beta})(\hat{l}_{\alpha\alpha\alpha}\hat{\sigma}_{\alpha\alpha}+ \hat{l}_{\alpha\beta\alpha}\hat{\sigma}_{\alpha\beta}+\hat{l}_{\beta\alpha\alpha}\hat{\sigma}_{\beta\alpha}\]]@xmath147.\ ] ] the approximate bayes estimators of @xmath24 , @xmath88 and @xmath25 can be obtained using lindley approximation , but it is not possible to construct highest posterior density ( hpd ) confidence intervals using this method .
therefore , we suggest the following markov chain monte carlo ( mcmc ) method to generate samples from the posterior density function , and in turn to obtain the bayes estimators , and hpd confidence intervals . here
we study the gibbs sampling method to draw samples from the posterior density function and then compute the bayes estimators and hpd confidence intervals of @xmath24 , @xmath88 and @xmath25 under the squared errors loss function .
let @xmath148 be an observed sample from the hybrid censoring scheme , drawn from a @xmath26 distribution . from ( [ pos ] )
, we can write the joint posterior density function of @xmath24 and @xmath88 given @xmath149 as : @xmath150 @xmath151 by this , the posterior density function of @xmath88 given @xmath24 and @xmath149 is @xmath152 [ the1]the conditional distribution of @xmath88 given @xmath24 and @xmath149 is log - concave .
see appendix , part b. + by ( [ pos2 ] ) , the posterior density function of @xmath24 given @xmath88 and @xmath149 is @xmath153 [ th2]the conditional distribution of @xmath24 given @xmath88 and @xmath149 has a finite maximum point .
see appendix , part c. with the help of the acceptance rejection principle ( see devroye @xcite for details ) and the previous theorem , the generation from ( [ albe ] ) can be performed using the we generator .
+ now we use theorems [ the1 ] and [ th2 ] and pursue the idea of geman and geman @xcite , and suggest the following scheme .
+ step 1 ) take some initial value of @xmath24 and @xmath88 , such as @xmath154 and @xmath155 .
step 2 ) generate @xmath156 and @xmath157 from @xmath158 and @xmath159 .
step 3 ) repeat step 2 , @xmath160 times . step 4 )
obtain bayes estimators of @xmath24 and @xmath88 with respect to a squared error loss function : @xmath161 where @xmath162 and @xmath163 are the burn - in periods in generating of @xmath164 and @xmath165 respectively . step 5 )
obtain the hpd confidence interval of @xmath24 : order @xmath166 as @xmath167 and construct all the @xmath168 confidence intervals of @xmath24 , as : @xmath169)}),\cdots,(\alpha_{([m_1\eta])},\alpha_{(m_1)}),\ ] ] where @xmath170 $ ] symbolizes the largest integer less than or equal to @xmath171 .
the hpd confidence interval of @xmath24 is the shortest length interval .
similarly , we can construct a @xmath168 hpd confidence interval of @xmath88 . finally , using the idea of chen and shao @xcite
, we can compute the estimation and hpd confidence interval for @xmath25 .
in this section , we carry out a simulation study to compare the performance of mle s and bayes estimators . in all the cases @xmath172 and @xmath173 are taken .
we estimate the unknown parameters using the mle , bayes estimators obtained by lindley s approximations and also bayes estimators obtained by using mcmc technique .
we compare the performances of different estimators with mse .
we also obtain the average length of the asymptotic confidence intervals and the hpd confidence intervals . for computing the bayes estimators , it is assumed that @xmath88 and @xmath24 have @xmath174 and @xmath175 priors , respectively .
moreover we use the non - informative priors of both @xmath88 and @xmath24 , by considering @xmath176 .
the bayes estimators are computed under the squared error loss function and with respect to the above non - informative priors .
the simulation is performed for different choices of @xmath177 values .
we replicate the procedure for 1000 times and report the average estimators , the mse s , the average asymptotic confidence intervals length and the average hpd confidence intervals length from the mcmc technique .
the results are reported in table 1 - 4 .
the first and second rows are parameter estimators of @xmath25 and @xmath24 , respectively . from tables 1 - 4
, it is observed that for fixed n and t as r increases , the mse decrease .
the performances of the mle s and bayes estimators are very similar in all aspects .
the average hpd confidence lengths are smaller than the average asymptotic lengths in all the cases considered .
finally it should be mentioned that bayes estimators are most computationally expensive followed by mle s .
in this section , we demonstrate one data set for illustrative purposes . it has been studied by gupta and kundu @xcite that the @xmath26 distribution can be used quite to analyze them and mle s of @xmath24 and @xmath25 are 1.6232 and 0.0138 respectively .
the data set was studied by bjerkedal @xcite and is given below : + 12 15 22 24 24 32 32 33 34 38 38 43 44 48 52 53 54 54 55 56 57 58 58 59 60 60 60 60 61 62 63 65 65 67 68 70 70 72 73 75 76 76 81 83 84 85 87 91 95 96 98 99 109 110 121 127 129 131 143 146 146 175 175 211 233 258 258 263 297 341 341 376 .
+ we use them and create the following two sampling schemes : @xmath178 @xmath179 now for scheme 1 , mle of @xmath180 are @xmath181 and bayes estimators with assumed non - informative priors , i.e. , @xmath176 with lindley approximation and gibbs sampling method are @xmath182 and @xmath183 respectively .
the @xmath184 confidence intervals based on mle and bayes estimators of @xmath180 are @xmath185 and @xmath186 respectively .
+ for scheme 2 , mle of @xmath180 are @xmath187 bayes estimators with assumed non - informative priors , i.e. , @xmath176 with lindley approximation and gibbs sampling method are @xmath188 and @xmath189 respectively .
the @xmath184 confidence intervals based on mle and bayes estimators of @xmath180 are @xmath190 and @xmath191 respectively . because we see the effect of the hyper parameters on the bayes estimators and also on confidence intervals , we take the following informative priors @xmath192 + based on this , for scheme 1 , bayes estimators of @xmath180 with lindley approximation and gibbs sampling method are @xmath193 and @xmath194 respectively .
the @xmath184 confidence intervals based on bayes estimators of @xmath180 are @xmath195 for scheme 2 , bayes estimators of @xmath180 with lindley approximation and gibbs sampling method are @xmath196 and @xmath197 respectively .
the @xmath184 confidence intervals based on bayes estimators of @xmath180 are @xmath198 we plot all the different estimated density functions with non - informative priors and informative priors in figure 1 and figure 2 .
[ fig3 ] [ sch ] comparing the two schemes with informative and non - informative priors , it is observed that for scheme 1 , estimators have smaller standard errors than scheme 2 , as expected .
also it is clear that the bayes estimators depend on the hyper parameters .
because the hpd confidence intervals based on informative priors are slightly smaller than corresponding length of hpd confidence intervals based on non - informative priors , therefore the prior informative should be used if they are available .
in this article , we have studied the classical and bayes inference procedure for the type - ii hybrid censored @xmath26 distribution .
we provide the maximum likelihood estimators and it is observed that the maximum likelihood estimators of the unknown parameters can not be obtained in the closed form and we suggest the em algorithm to compute them .
we also earn the bayes estimators of the unknown parameters and show that they can not be obtained in explicit forms , and we have proposed two approximation methods to earn them .
we have compared the performance of the different methods by monte carlo simulations , and it is observed that the performance of quite satisfactory .
[ ap ] the proof can be obtained similarly as in ng et al .
+ note that using theorem [ ap ] , we can write @xmath199=\int_c^\infty \frac{\alpha+1}{\alpha}\frac{\beta}{k}xe^{-\frac{\beta}{\alpha}x}(1-e^{-\beta x})dx\;\;\;\;\left(k = e^{-\frac{\beta}{\alpha}c}(\alpha+1-e^{-\beta c})\right)\\ \\ \hspace{1in}=\frac{(\alpha+1)\alpha}{k\beta}\int_{\frac{\beta}{\alpha}c}^\infty ue^{-u}du-\frac{\alpha}{k\beta(\alpha+1)}\int_{\frac{\beta(\alpha+1)}{\alpha}c}^\infty ve^{-v}dv \;\;\;\left(put \;\;u=\frac{\beta}{\alpha}x\;\;\&\;\;v=\frac{\beta(\alpha+1)}{\alpha}x\right)\\ \\
\hspace{-0.5in}=\frac{(\alpha+1)\alpha}{k\beta}(-e^{-u})(u+1)]_{\frac{\beta}{\alpha}c}^\infty-\frac{\alpha}{k\beta(\alpha+1)}(-e^{-v})(v+1 ) ] _ { \frac{\beta(\alpha+1)}{\alpha}c}^\infty\\ \\
\hspace{-0.7in}=\frac{(\alpha+1)\alpha}{k\beta}(\frac{\beta}{\alpha}c+1)e^{-\frac{\beta}{\alpha}c}-\frac{\alpha}{k\beta(\alpha+1)}(\frac{\beta(\alpha+1)}{\alpha}c+1)e^ { -\frac{\beta(\alpha+1)}{\alpha}c}\\ \\
\hspace{-1.25in}=\frac{\alpha e^{-\frac{\beta}{\alpha}c}}{k\beta}\left[(\alpha+1)(\frac{\beta}{\alpha}c+1)-\frac{e^{-\beta c}}{\alpha+1}(\frac{\beta(\alpha+1)c}{\alpha}+1)\right]\\ \\ \hspace{-1.1in}=\frac{e^{-\frac{\beta}{\alpha}c}}{k\beta}\left[(\alpha+1)(\beta c+\alpha)-\frac{e^{-\beta c}}{\alpha+1}\left(\beta(\alpha+1)c+\alpha\right)\right ] \\ \\
c+\alpha)-\frac{e^{-\beta c}(\beta c(\alpha+1)+\alpha)}{\alpha+1}}{\beta(\alpha+1-e^{-\frac{\beta}{\alpha}c } ) } \end{array}\ ] ] and about @xmath200 , we have : @xmath201=\int_c^\infty\frac{\alpha+1}{\alpha}\frac{\beta}{k}\ln(1-e^{-\beta x})e^{-\frac{\beta}{\alpha}x}(1-e^{-\beta x})dx\]]@xmath202 @xmath203,[4,4],1-e^{-\beta c})\right.\]]@xmath204,[3],1-e^{-\beta c})\right.\]]@xmath205,[3],1-e^{-\beta c})\right)(1-e^{-\beta c})^2\]]@xmath206,[4,4],1)\]]@xmath207,[3],1)\right]\ ] ] where hypergeom ( . )
is generalized hypergeometric function .
this function is also known as barnes s extended hypergeometric function .
the definition of @xmath208 is : @xmath209 where @xmath210 $ ] , @xmath211 is the number of operands of @xmath212 , @xmath213 $ ] and @xmath214 is the number of operands of @xmath215 .
generalized hypergeometric function is quickly evaluated and readily available in standard software such as maple .
the conditional distribution of @xmath24 given @xmath88 and @xmath149 is @xmath217 in this function , we have @xmath218 and @xmath219 , now it is enough that prove @xmath220 is bounded . with simple calculation
we see that @xmath221 this function is less than the gamma function and the gamma function is a bounded function , so this function is bounded .
therefore @xmath220 has a finite maximum point .
childs.a , chandrasekhar.b , balakrishnan.n , kundu.d ( 2003 ) exact likelihood inference based on type - i and type - ii hybrid censored samples from the exponential distribution,_ann .
_ , 55 , 319 - 330 . | a hybrid censoring scheme is a mixture of type - i and type - ii censoring schemes .
we study the estimation of parameters of weighted exponential distribution based on type - ii hybrid censored data . by applying em algorithm ,
maximum likelihood estimators are evaluated . also using fisher infirmation matrix asymptotic confidence intervals
are provided . by applying markov chain monte carlo techniques
bayes estimators , and corresponding highest posterior density confidence intervals of parameters are obtained .
monte carlo simulations to compare the performances of the different methods is performed and one data set is analyzed for illustrative purposes .
+ + _ keywords _ : asymptotic distribution , em algorithm , markov chain monte carlo , hybrid censoring , bayes estimators , type - i censoring , type ii censoring , maximum likelihood estimators + + _ mathematics subject classification : _ 62f10 , 62f15 , 62n02 |
active galactic nuclei ( agn ) are believed to be powered by two fundamentally different modes of accretion . optically selected and luminous radio - loud agn ( r - agn ) are powered by the radiatively - efficient accretion of gas onto a central supermassive black hole ( smbh ) at a rate of one to ten percent of the eddington limit .
in contrast the majority of r - agn , which have a low radio luminosity , accrete gas in a radiatively inefficient manner at a rate below one percent of the eddington limit @xcite . the mode of accretion may also be related to galaxy morphology .
while agn in general are hosted by galaxies that span a range of morphological types , the hosts of r - agn are most often massive ellipticals ( e.g. , ekers & ekers 1973 , best at al .
2005b , kauffmann et al . 2008 ) .
this difference in r - agn accretion processes could be a result of differences in the fueling mechanism .
optical and luminous radio agn may be triggered by mergers and one - on - one interactions ( heckman et al .
1986 , barnes & hernquist 1996 ) which could supply the large quantities of cold gas required to maintain a high accretion rate .
low luminosity r - agn may be triggered by the accretion of hot gas from the surrounding halo ( fabian 1994 , allen et al .
2006 ) , although there could be additional triggering mechanisms @xcite .
the key to understanding these triggering mechanisms therefore lies in the kpc to mpc - scale environments of r - agn , which is populated by satellite galaxies .
many studies have examined the relationship between r - agn incidence and the environment . @xcite found that r - agn are more frequently found in central group and cluster galaxies when compared to galaxies of similar stellar mass .
@xcite studied a small sample of 91 r - agn and found that the fraction of r - agn shows little dependence on the mpc - scale local galaxy density , but is dependent on the number of galaxies in the group in which the r - agn is located , while the slightly larger sample of 212 r - agn of @xcite showed a slight increase in r - agn incidence with density at the mpc scale . @xcite and
@xcite both found a strong dependence of r - agn activity on the stellar mass of the host galaxy , but when this is accounted for , r - agn activity still has a dependence on density at the mpc scale .
@xcite found that the environments of r - agn at the @xmath3100 kpc scale are roughly twice as dense as those of radio - quiet agn or radio - quiet galaxies of similar mass .
in contrast to these studies , @xcite found that at the @xmath3100 kpc scale only powerful r - agn have higher clustering than radio - quiet galaxies of the same mass , and that for scales greater than about 160 kpc the clustering of r - agn is similar to that of radio - quiet galaxies .
the aforementioned studies find an increase in the r - agn incidence rate in denser , cluster environments .
this should not be surprising considering that r - agn likely feed on the hot gas that is plentiful in clusters .
however , it must be acknowledged that neither all ellipticals or brightest cluster galaxies ( bcgs ) are radio loud , nor are all massive field ellipticals radio quiet .
the approach taken in the current study is therefore different as it separates the general environment ( bcg , cluster member , or field galaxy ) from the small - scale environment ( @xmath4 kpc ) .
we compare the small - scale environments ( i.e. the satellite populations ) of a statistically large sample of r - agn and a matching control sample of radio - quiet galaxies .
control galaxies are chosen not only to match the mass , type , and star formation history of an r - agn but also the large scale environment it is found in , such that the controls of field r - agn are also field galaxies , the controls of cluster member r - agn are also cluster members , and the controls of bcg r - agn are also bcgs .
this careful matching allows us to compare the satellite populations of r - agn and otherwise similar radio - quiet galaxies to determine why only a subset of galaxies with similar environments become r - agn . in our study of satellite populations
we will separately focus on radio agn with high accretion rates .
the two modes of r - agn accretion can be distinguished with emission line ratios @xcite . defined an ` excitation index ' composed of four optical emission line ratios which can be used to separate r - agn with a low accretion rate , which are known as low excitation radio galaxies ( lergs ) from those with a high accretion rate , known as high excitation radio galaxies ( hergs ) . at most radio luminosities
hergs comprise only a few percent of the total r - agn population , although they become more common at the highest radio luminosities @xcite .
many studies employ the fanaroff - riley classification , which classifies r - agn based on their radio morphology @xcite .
luminous r - agn , which tend to be edge brightened , are known as fanaroff - riley type 2 ( fr2 ) sources , while core - brightened lower - luminosity sources are referred to as fanaroff - riley type 1 ( fr1 ) sources .
although lergs and hergs often exhibit fr1 and fr2 morphology , respectively , @xcite demonstrated that a number of fr2 sources have low - excitation spectra ( i.e. are lergs ) .
since the lerg & herg classifications are more closely related to accretion mode and do not require high resolution radio maps , we adopt their use in our study , as determined by @xcite .
in addition to exploring the role of satellites in r - agn triggering , the second major goal of this study is to understand if and to what extent satellites can be affected through feedback from the powerful jets that are the hallmark of r - agn .
these highly collimated jets originate in the nucleus of the agn at relativistic speeds and may extend for kiloparsecs or even megaparsecs beyond the host galaxy ( tremblay et al .
2010 , schoenmakers et al .
radio agn deposit most of their energy into the interstellar or intergalactic medium kinetically via their high - velocity jets ( see fabian 2012 for a recent review ) .
such interactions may be responsible for quenching star formation in the host galaxy , and there is some evidence that r - agn may lead to quenching of star formation in their satellite galaxies .
@xcite found that satellite galaxies in the projected jet paths of fr2 sources are redder than satellites outside the jet path , but no such trend was found for the satellites of fr1 sources .
this suggests that fr2 jets tend to quench star formation in satellite galaxies while fr1 jets do not .
we will revisit these results in our study by using a much larger sample of r - agn and by contrasting entire satellite populations of r - agn ( regardless of proximity to the jet ) to those of the radio - quiet control sample .
somewhat paradoxically , radio jets are also considered as the mechanism behind possible _ positive _ feedback , both in the host galaxy and in adjacent satellite galaxies .
the idea is that interactions with jets may induce star formation by driving shocks into dense clouds which then collapse @xcite .
such agn - driven star formation may have been important in building up the massive spheroid in the early phases of galaxy formation @xcite .
the radio jets of both hergs and lergs are believed to be capable of producing positive feedback outside of the host as well .
the powerful radio jets associated with hergs pierce external clouds in the igm , but triggered star formation may proceed in their slowly expanding radio lobes ( de young 1981 , bicknell et al .
since the jets of lergs are generally less powerful , they may be capable of inducing star formation even in head - on collisions @xcite .
a few candidates of such jet - induced star formation in the vicinity of r - agn have been observed .
the star forming region ` 09.6 ' in the eastern lobe of the nearby fr 2 source 3c 285 was first observed by @xcite .
chandra _ observations confirm that this region is indeed experiencing a starburst phase @xcite , while a satellite galaxy of the fr2 source pks225041 , which is not currently in the projected jet path , also shows evidence of jet - induced star formation @xcite . induced star formation as a result of interactions with weak r - agn has also been observed : one example is the lerg centaurus a. @xcite recently showed that the youngest stars in the inner filament of centaurus a are only a few myrs old and are probably the result of the shock induced collapse of a molecular cloud .
downstream from the inner star - forming filament , the radio jet interacts with an h i cloud and produces another filament of recently - formed stars @xcite .
minkowski s object ( mo ) is another example of lerg - induced star formation .
this star forming region is 15 kpc away from its host r - agn ngc 541 .
strong uv and h@xmath5 emission are indicative of the starburst nature of this object @xcite .
simulations by @xcite have reproduced the observational characteristics of mo , including the star formation rate ( sfr ) of 0.3 m@xmath6yr@xmath7 . while it is possible that mo is an example of
star formation being reignited in an already star - forming galaxy , this seems unlikely . although @xcite could not rule out an underlying old stellar population in mo , an hi cloud downstream from mo was observed , which suggests that unlike centaurus a the neutral hydrogen cooled out of a warm and clumpy igm and then collapsed .
while there is observational evidence that in some individual cases intense star formation may have been triggered by r - agn jets , it is not clear how common this phenomenon is and whether it can be positively stated that the star formation is indeed the result of jet interactions .
furthermore , it is not clear if this induced star formation leads to the emergence of pristine , new satellites or whether it is an enhancement in already existing , star - forming satellites . as in the case of star formation quenching , we will approach this question by studying the colors of satellites of r - agn and a matched control sample of radio quiet galaxies . in both the study of quenching and the induction of star formation in satellites
we will pay special attention to hergs in which either of these processes may be more pronounced or more common .
the layout of our paper is as follows . in section 2
, samples are presented , and in section 3 we describe our method for selecting the control sample of radio quiet galaxies as well as generating distributions of satellite properties .
our results are presented in section 4 , while in sections 5 and 6 we discuss the implications of our findings . in this work
the cosmological parameters adopted are @xmath8 , @xmath9 , and @xmath10 .
we have two primary samples : a sample of r - agn drawn from @xcite and a control sample of matching radio quiet galaxies that we have assembled .
each galaxy in the control sample was chosen to have the same type of environment as its counterpart in the r - agn sample .
control galaxies were also chosen to approximately match the galaxy mass and star formation histories of the r - agn by selecting control galaxies that provide the closest match in @xmath11 magnitude , @xmath12 color , redshift , and axis ratio .
we then compare the satellites of r - agn and radio quiet samples .
we use the term ` satellite ' to refer to nearby galaxies that are physically associated with the hosts in our samples , while ` neighbors ' refer to candidate satellites , which includes foreground or background objects not associated with the host .
this distinction is necessary because candidate satellites are faint and are generally lacking spectroscopic redshifts .
we have drawn neighbors for each of these two samples , and to correct for background contamination we have also drawn neighbor - like objects from two sets of offset positions . details of these samples are given below .
@xcite have assembled a sample of 18,286 radio galaxies by combining the 7th data release of the sdss spectroscopic sample ( dr7 ; abazajian et al .
2009 ) with the national radio astronomy observatory ( nrao ) very large array ( vla ) sky survey ( nvss ; condon et al .
1998 ) and the faint images of the radio sky at twenty centimeters ( first ; becker , white & helfand 1995 ) survey .
radio emission may arise from agn or star formation activity , so @xcite used a combination of three methods to separate r - agn from star - forming galaxies .
the first method is a comparison of the 4000 break strength to the ratio of radio luminosity per stellar mass .
since r - agn have enhanced radio luminosity relative to star - forming galaxies , they are thus separable in this plane @xcite .
the second method is the emission - line diagnostic or ` bpt ' diagram ( baldwin et al .
1981 , kauffmann et al . 2003 ) which compares the ratio of [ o iii ] 5007 and h@xmath13 line fluxes to the ratio of [ nii ] 6584 and h@xmath5 . only 30% of the galaxies in the radio source sample
could be classified using the bpt diagram : the remainder lacked detections or limits in at least one line .
the final method is a comparison of the h@xmath5 line luminosity to the radio luminosity .
this method operates under the assumption that in star - forming galaxies both the h@xmath5 line and radio luminosities are a consequence of star formation and thus correlated , while the radio luminosities of r - agn are offset because their radio luminosities are boosted by agn activity .
differing classifications are sometimes given by these three methods .
in these instances a final classification was chosen based on the properties of the galaxies in each class : see table a1 of @xcite .
= 0.1 , 0.15 , and 0.2 indicate divisions used for the redshift ranges.,width=326 ] the magnitude range of our sample is 14.5 @xmath14 _
@xmath14 17.77 , matching the limits of the sdss main galaxy spectroscopic sample @xcite .
many of the r - agn in @xcite are fainter than this limit and lie at redshifts greater than @xmath15 .
we are not interested in higher redshift r - agn since we can not draw control samples for them .
we imposed a lower redshift limit at @xmath16 to avoid excessive background contamination in the 100 kpc radius used to obtain neighbors .
thus the final redshift range of our sample is 0.04 @xmath14 _ z _ @xmath14 0.3 , and figure [ reddis ] shows this redshift distribution . at the lower limit a radius of 100 kpc corresponds to an angular radius of @xmath17 .
only 28 r - agn are removed from the sample as a result of this lower redshift limit .
we used the photometry of the eighth sdss data release ( dr8 ) @xcite rather than dr7 because photometric redshifts , used to reduce non - satellite contamination in neighbors , tend to be more accurate in dr8 than in dr7 ( see appendix [ appa ] ) .
also , in previous data releases the outer regions of large galaxies were oversubtracted , which affected the photometry of these galaxies as well as faint nearby objects @xcite .
improved sky subtraction procedures in dr8 have mostly addressed this issue .
fifty - six objects listed as r - agn in @xcite were either not present or were misidentified in dr8 , so our final r - agn sample consists of 7,220 galaxies .
figure [ cmd ] is a color - magnitude diagram that shows the relation of the r - agn sample , shown in black , to sdss galaxies in the same redshift range .
also plotted is the control sample , the selection of which is discussed below .
colors and magnitudes in this figure have been @xmath18-corrected with the @xmath18-correction provided by @xcite .
this figure shows that the r - agn hosts fall along the red sequence , with r - agn favoring luminous galaxies . and
@xmath12 have been @xmath18-corrected .
contours containing 90% , 50% , and 25% of the samples are shown .
r - agn in the sample tend to be hosted by luminous galaxies on the red sequence .
the properties of galaxies in the control sample closely match those of the r - agn.,width=326 ] one of the criteria for producing the control sample is to match the type of environment in which the galaxy is located .
three cluster membership categories were used : field galaxy , cluster member , and bcg . for each of the r - agn we performed a nearest - neighbor search to the bcgs reported in the galaxy cluster catalog of @xcite , which is based on dr7 .
the clusters in this catalog are in the redshift range @xmath19 , and the list of additional clusters at @xmath20 , available from the author s web pagejghao / gmbcg_sdss_catalog.html ] , allow us to assign cluster membership for all r - agn in the sample .
we classified as bcgs all r - agn whose dr8 coordinates were within 1.5 kpc of the bcg catalog coordinates , and found that our sample of r - agn contains 949 bcgs or 13% of the total .
those r - agn that were between 1.5 kpc and 1.5 mpc away from the nearest bcg and had a difference in spectroscopic redshift less than _ z _ @xmath21 0.01 were classified as cluster members .
we have chosen 1.5 mpc as a typical virial radius ( @xmath22 ) of clusters .
of our sample , 436 r - agn or 6% of the total were classified as cluster members , and the mean distance from cluster members to the bcg of the host cluster is 700 kpc .
the remaining 5,835 r - agn or 81% of the sample were classified as field galaxies .
while r - agn have a preference for rich environments as demonstrated in previous studies , the majority of radio agn are nevertheless not in clusters .
most ( 86% ) of the r - agn in the sample have been classified as either hergs ( 209 ) or lergs ( 6,005 ) by @xcite .
the remaining 1,006 were not classified because they lacked emission lines .
table [ tbl1 ] gives the number of hergs , lergs , and unclassified galaxies in each environment type .
interestingly , hergs are most common among the field r - agn ( 3.4% ) and rarer among cluster members ( 1.1% ) .
they are very rare among bcgs ( 0.4% ) .
this result supports the scenario in which hergs arise from cold gas accretion , which is more abundant in the field and is in agreement with the findings of @xcite , who found that lergs show an increasing rate of incidence with the local density , while hergs show a decrease , as do optically - selected agn .
we will return to the incidence rate of r - agns in different environments in section [ fraction ] .
rcccccccc field&200&4,845&790&1,115&1,488&1,711&1,521&5,835 + member&5 & 368 & 63&39&129&163&105&436 + bcg & 4&792&153&53&185&258&453&949 + total&209&6,005&1,006&1,207&1,802&2,132&2,079&7,220 for each r - agn , we have selected a control galaxy which is not an r - agn ( or is below the limits of radio surveys ) but has the same cluster membership status and similar overall physical properties as its counterpart r - agn .
we considered as control candidates those galaxies in dr8 that are within the first footprint ( @xmath23 ) and therefore subject to classification by @xcite .
the area of dr8 within the first footprint is identical to that of dr7 , so the control candidates are also subject to cluster member classification with the @xcite catalog .
this provides a pool of 842,296 control candidates .
each control galaxy was selected to closely match the @xmath11 magnitude , @xmath12 color , redshift , and axis ratio of its corresponding r - agn .
these properties were used for selection because the redshift and @xmath11 magnitude give the optical luminosity , which together with @xmath12 are a proxy for the stellar mass of the galaxy while the @xmath12 color is a proxy for the star formation history .
radio agn are predominantly found in spheroidal early - type galaxies ( ellipticals ) rather than among the flattened early types containing stellar disks ( s0s ) ( e.g. , ekers & ekers 1973 , vron - cetty & vron 2001 ) .
the two varieties of early type galaxies have similar colors and overlapping absolute magnitude ranges ( e.g. , cheng et al .
the inclusion of the axis ratio as a matching criterion helps ensure that the control sample is of the correct morphological type @xcite .
the selection was done by minimizing the metric , @xmath24 where the quantities @xmath25 , @xmath26 , @xmath27 , and @xmath28 are the differences between the r - agn and control candidate parameters , and the denominators are scaling factors that correspond to the 95 percentile ranges in redshift , @xmath11 magnitude , @xmath12 color , and axis ratio ( b / a ) . the control sample is plotted together with the r - agn sample in fig .
[ cmd ] , and it can be seen that the control sample closely matches the color - magnitude distribution of the r - agn sample .
the means and standard deviations of the differences between the control and r - agn galaxy properties also demonstrate the success of our control sample selection .
the mean difference for @xmath11 is @xmath29 ( @xmath30 0.065 ) , for redshift the difference is @xmath31 ( @xmath32 ) , for @xmath12 the difference is @xmath33 ( @xmath30 0.05 ) , and for axis ratio the difference is @xmath34 ( @xmath35 ) .
in other words , the standard deviations are roughly 2% of the 95 percentile range used for the scaling factors in the metric .
some of the control galaxies are best matches to more than one of the r - agn in the sample : this occurs for only 11% of the control galaxies . .... .... in this study , we have examined all satellites within a circular region centered on the host galaxies in our samples regardless of their position angle with respect to the jet . for the study of feedback on the satellite population it may appear better to focus only on satellites lying close to the jet
. however , jet directions are not available for the majority of the sample .
furthermore , the satellite pks2250 - 41 @xcite shows evidence of jet - induced star formation even though it is not currently in the jet path , justifying our approach .
the radius of the search circle was determined by examining the projected lengths of the 782 r - agn jets in the list provided by . of the r - agn in this list
, 90% have jets shorter than 100 kpc .
a search radius greater than 100 kpc would allow us to include the end points of a few more r - agn at the cost of a large increase in unassociated neighbors .
we restricted our analysis to satellites that lie within 100 kpc ( in projection ) of the galaxies in our samples .
an inner boundary at 5 kpc was also used to avoid spurious sources which may arise as a result of shredding of the host galaxy by the sdss pipeline .
neighbors include objects classified as galaxies by the dr8 pipeline that are brighter than @xmath36 , close to the sdss magnitude limit of @xmath37 .
we make no requirement that the magnitudes of neighbors be fainter than that of the r - agn or control galaxy , but we find that only in a few instances are neighbors more luminous than their target galaxies .
we derive properties of the satellite population by subtracting the distribution of sources selected in the same way as neighbors , but in parts of the sky that are offset ( two degrees in ra ) from the host ( r - agn or its control ) . to make this procedure more robust , we first remove from candidate neighbors those with significantly discrepant photometric redshifts .
considering the large uncertainties of available photometric redshifts ( appendix a ) we apply a very conservative cut such that the candidate neighbors are removed if their photometric redshift is more than 0.2 greater than the spectroscopic redshift of the host .
in this study we examine the co - added distribution of satellite properties for r - agn , and , separately , for their radio quiet counterparts , which span a redshift range from 0.04 to 0.3 .
as a result , at the low redshift limit we probe galaxies down to @xmath38 = @xmath39 , which shifts to @xmath38 = @xmath40 at the upper redshift limit . to account for the resulting malmquist bias we first construct various satellite distribution functions in smaller redshift ranges , in which the bias is greatly reduced . after we confirm that different redshift slices have similar distributions of satellite properties , we average them .
the four redshift ranges are 0.04 @xmath41 0.1 , 0.1 @xmath41 0.15 , 0.15 @xmath41 0.2 , and 0.2 @xmath41 0.3 . the distribution of r - agn redshifts as well as the redshift range divisions are shown in fig . [ reddis ] while table [ tbl1 ] gives the number of r - agn per environment type in each redshift range . for a given redshift range ,
the number of neighbors per bin of a distribution are given by , @xmath42 @xmath43 @xmath44 @xmath45 table [ tbl2 ] gives the number of neighbors and offset objects per redshift bin in each environment .
this table shows that roughly half of neighbors are background objects .
the number of satellites per host around r - agn is given by subtracting the offsets from the number of r - agn satellites and dividing by the number of primary galaxies in that redshift range , @xmath46 to get the error in the number of satellites per host , we use , assuming normal approximation of counting errors : @xmath47 the same method is used to calculate the number of satellites per control galaxy . however , since in 11% of cases the same control is chosen for more than one r - agn the errors calculated in the above way , which assumes independent measurements , will be underestimated for control galaxies by 5% .
this is a small effect and we do not correct for it .
this procedure gives the nearly volume - complete distributions of r - agn and control satellites separately for each of the four redshift ranges .
we then combined the four distributions with weighted averaging , and the resulting distribution for the r - agn sample is given by @xmath48 and the error is given by @xmath49 .... .... the final distribution for the control sample is computed in the same manner .
rcccccccc[h ] & @xmath50&10,023&7,899&7,337&5,815 + field & @xmath51&9,203&6,001&6,789&4,825 + & @xmath52&5,820&3,935&3,386&2,485 + & @xmath53&5,906&4,108&3,525&2,490 + & @xmath50&458&867&915&543 + cluster & @xmath51&437&784&692&454 + member & @xmath52&183&321&308&191 + & @xmath53&247&367&298&164 + & @xmath50&707&1,690&1,732&2,605 + bcg & @xmath51&658&1,503&1,577&2,330 + & @xmath52&246&535&528&712 + & @xmath53&203&549&487&709
to investigate r - agn triggering and the effects of r - agn jets on satellite galaxies , we compare the populations of r - agn and control satellites .
this is done by calculating and comparing distributions of satellites of r - agn and control galaxies as functions of projected distance from the host , satellite luminosity , and satellite color .
we have also examined the ratios of distributions of r - agn and control satellites .
figure [ distance ] shows distributions of projected radial distances of satellites from their hosts .
field galaxies are in panel a , cluster members in panel b , bcgs in panel c , and hergs , regardless of environment , in panel d. at all redshifts , our satellite samples are complete for satellites brighter than @xmath54 , so we adopt this absolute magnitude cut for the radial distance distributions .
it should be noted that 90% of the hergs are field galaxies . as expected ,
bcgs have more satellites than either field galaxies or non - bcg cluster members , regardless of whether they host an r - agn or not .
the satellite populations of field galaxies and non - bcg cluster members are similar for a given type of host .
however , what is remarkable is that the mean number of satellites differs between r - agn and radio - quiet galaxies , and it does so in all environments .
furthermore , this excess of satellites around r - agn is present over the entire 100 kpc projected radius that we explore .
the situation is ambiguous for the case of hergs , where no obvious excess is visible .
what we can confidently say is that the excess seen in panels a , b , and c reflects differences in the satellite populations of lergs and not hergs .
we construct satellite luminosity distributions to further characterize the differences in the satellite populations of r - agn versus the control sample .
absolute magnitudes were computed using the spectroscopic redshifts of the host galaxies and have been @xmath18 corrected using the corrections provided by @xcite .
the resulting @xmath38 distributions are shown in fig .
[ abmag ] with the results for the three environments in panels a , b , & c , and herg satellites in panel d. unlike the luminosity function of field galaxies , the @xmath38 distributions of satellites in fig .
[ abmag ] have a peak , which is to be expected for a flux - limited sample @xcite . the distribution of bcg satellites peaks at a brighter magnitude than field galaxies or cluster members . since bcgs are generally massive and luminous , their satellites tend to be more luminous as well .
the distributions of the satellites of r - agn and control galaxies are quite similar among themselves , for all environments , with peaks at similar positions . in terms of the average number of satellites per host that are brighter than @xmath55
, we find that field r - agn have 2.14 ( @xmath56 ) satellites per host , and field control galaxies have 1.46 ( @xmath56 ) . for cluster members ,
r - agn have 3.47 ( @xmath57 ) and control galaxies have 2.61 ( @xmath58 ) satellites per host , while r - agn that are bcgs have 4.30 ( @xmath59 ) and control bcgs have 3.72 ( @xmath60 ) satellites per host .
we find that hergs have 1.43 ( @xmath61 ) satellites per host on average , while the control galaxies of hergs have 1.48 ( @xmath62 ) .
this formally confirms that the excess is very significant in all environments .
however , no excess is formally found for hergs .
we compared the @xmath63 color distributions of r - agn and control satellites to provide context for the examination of the feedback effects of r - agn jets on nearby satellites ( section [ feedback ] ) .
although @xmath64 is more sensitive to recent star formation , we chose to use the @xmath63 colors for satellites since many of them have poor @xmath64-band photometry .
we again used an absolute magnitude cut at @xmath65 to probe the same part of the satellite population at all redshifts .
the @xmath18-correction of @xcite was also applied to the @xmath63 colors , and the analysis was restricted to satellites for which @xmath66 mag ( @xmath390% of total ) .
histograms of these distributions are shown in fig .
[ color ] .
in general the distributions of both the satellites of r - agn and control galaxies peak near @xmath67 , indicating that satellites in general are red irrespective of r - agn .
in other words , the satellite population around these massive galaxies is mostly quenched .
whether r - agn feedback plays any active role in this quenching can not be well established from these distributions and is deferred to section [ feedback ] .
the distribution for herg satellites is noisy but is consistent with no excess satellites around hergs , as seen in panel d of fig .
[ distance ] where the same @xmath38 cut was applied . from the number
histograms alone it is difficult to see if the population of excess r - agn satellites has a preferred separation from the host , a preferred luminosity , or color .
this is because the absolute excess will appear stronger where the distribution is higher .
we have therefore examined the ratio of the distributions of r - agn and control satellites .
ratio plots of the @xmath38 , color , and distance distributions for the field sample are shown in fig
. [ ratio ] , with error bars computed via simple error propagation .
individual bins with errors greater than 50% are not shown .
we have only shown plots for the field sample because it contains the most r - agn : similar results are seen for the member and bcg samples .
all three distributions in fig . [ ratio ]
are roughly flat at 1.5 , which indicates that field r - agn have @xmath350% more satellites than control galaxies . in other words ,
the satellite populations of r - agn are scaled up versions of the satellite population around radio - quiet galaxies .
we may be seeing a slight preference of the excess population for brighter and redder satellites .
the former is not surprising as any possible influence on agn activity would presumably be more likely for more massive satellites .
the latter can be explained if some of the blue satellites are preferentially further from a galaxy in real distance than their projected distance would suggest and consequently do not contribute to excess population .
previous studies ( e.g. best et al . 2005b , van velzen et al . 2012 ) have shown that the fraction of galaxies that are r - agn ( the r - agn incidence rate ) is an increasing function of the stellar mass of the host .
we revisit these results by including two new aspects in the analysis .
first , we obtain incidence rates separately for r - agn in each type of environment ( bcgs , cluster members , and field ) .
second , we consider the incidence rate to be the r - agn fraction in the _ eligible _ galaxy population , not among all galaxies in a given mass ( i.e. , luminosity ) bin .
the idea is that one is primarily interested in r - agn incidence among the galaxies that could ( and perhaps did or will ) host an r - agn , so we define the eligible population to occupy the same part of the parameter space in redshift , magnitude , color , and axis ratio as galaxies that are _ current _ r - agn hosts .
each r - agn in the sample defines a position in the parameter space around which we determine the r - agn incidence .
we take the vicinity to be @xmath68 ( eq .
1 ) and count the r - agn used to probe the parameter space as 0.5 , which leads to less bias than either not counting it or giving it a full count .
these fractions are then binned by @xmath38 , and the median of each bin is shown in fig .
[ ragnfrac ] . bins with fewer than five r - agn
were omitted .
figure [ ragnfrac ] shows that while the fraction of eligible galaxies that currently host an r - agn is generally low , with an overall average of 6% , the incidence increases from close to zero to a high near @xmath320% for the most luminous galaxies .
the incidence is generally higher in bcgs , as might be expected , while the incidence in cluster members is only slightly greater than that of field galaxies .
this suggests that fueling is primarily a _ locally _ determined process and that being in a more massive cluster halo confers modest additional benefits , unless a galaxy sits in the center of the cluster halo , as is the case for bcgs .
@xcite show a similar relation for r - agn incidence but as a function of stellar mass in panel a of their fig .
2 . the mean nvss radio luminosity of our sample is @xmath69 w hz@xmath7 at 1.4 ghz , so our results are best compared to the middle curve of panel a. the r - agn incidence of @xcite is expressed as a fraction of all galaxies from sdss dr2 without regard .... .... for the galaxies likelihood of hosting r - agn , while we have only considered galaxies from the same type of environment that are likely to host r - agn .
even so , considering that most r - agn are field galaxies and the majority of massive galaxies are eligible as r - agn , the results for the field galaxies and cluster members in our sample follow a trend similar to that of @xcite for @xmath70 w hz@xmath7 at 1.4 ghz .
@xcite found that at a stellar mass of @xmath71 the r - agn incidence in bcgs is only slightly greater than that of other galaxies , although for masses lower than @xmath72 the r - agn incidence in bcgs is over an order of magnitude greater than that of other galaxies .
this is in agreement with what we find in fig [ ragnfrac ] .
@xcite also found that except for group and cluster galaxies within @xmath73 of the center of the system , the r - agn incidence among non - bcg group and cluster galaxies is similar to that of field galaxies .
this was based on a comparison of the r - agn sample defined by @xcite to all galaxies in dr4 within the same redshift range as the @xcite r - agn sample , regardless of these galaxies likelihood of hosting an r - agn .
these results are in agreement with our finding that field and cluster member r - agn have similar incidence rates .
we have found that in all environments r - agn have more nearby satellites than radio - quiet galaxies , but our results are not a simple confirmation of what other studies have found regarding the r - agn clustering .
@xcite , @xcite , @xcite , and @xcite find that r - agn are more clustered than optical agn or radio - quiet galaxies .
the control samples of @xcite , @xcite , and @xcite were even chosen to match the stellar masses of the r - agn .
however , the results of these studies can be interpreted simply to mean that r - agn prefer denser , cluster environments or , judging by the results in section [ fraction ] that show that r - agn incidence among eligible cluster members and field galaxies is the same , that the type of galaxies that typically host r - agn ( massive ellipticals ) are more clustered , a well known result @xcite . by including cluster membership in the selection criteria of control galaxies
we have added crucial new information : the satellite population of r - agn is richer , whether the r - agn is a field galaxy , cluster member , or the very bcg .
it is now widely believed that lergs are triggered by the cooling of small amounts of gas from the hot halos in which they reside @xcite . since @xmath380% of the r - agn in our sample
are lergs , our finding that r - agn in all environments have an excess of satellites suggests that the availability of hot halo gas is not the only prerequisite for triggering activity in lergs .
for example , the dark matter halo masses of the bcgs in our r - agn and control samples are similar , which would suggest that they have similar quantities of hot halo gas available .
the difference is that bcgs that host r - agn have more nearby satellites , which therefore probably play some role in triggering .
this argument applies to field galaxies and cluster members as well , since the matched pairs of r - agn and control galaxies have similar halo masses and therefore presumably similar amounts of hot halo gas .
again , the difference between r - agn and radio - quiet galaxies in these environments is the number of nearby satellites .
results in section [ results ] suggest that the triggering is facilitated by satellites with a wide range of masses and at various projected distances from the host . while it is tempting to interpret the excess of satellites as related to the presence of r - agn , we must explore the possibility that even if the control sample is matched in stellar mass , it may be systematically different in halo mass .
if the abundance of satellites follows the halo mass rather than the stellar mass , the excess of satellites may result from a mismatch in halo masses .
we have used the dark matter halo catalog of @xcite to investigate this possibility .
this catalog presents two dark matter halo masses for each galaxy : one calculated based on the ranking of group characteristic luminosity , and the other based on the ranking of group stellar mass : we used those based on the group stellar mass .
this catalog uses the fourth sdss data release ( dr4 ) and extends out to @xmath74 , so only about 1,400 matched pairs of r - agn and control galaxies were found .
the mean of the log of r - agn masses is 0.1 dex greater than that of control galaxies . is this modest systematic offset in halo masses sufficient to explain a factor of 1.5 difference in satellite population ?
figure [ halo ] shows the number of satellites as a function of halo mass for the field control sample in the redshift range @xmath75 . there is a trend that galaxies in more massive halos indeed have more satellites .
however , this rise is not so steep . in order to bring the number of satellites from 1.5 ( the number observed on average around control sample galaxies ) to 2.1 ( the number for r - agn hosts ) the mismatch in halo masses would have to be 0.96 dex ( a factor of 9 ) while in reality it is only 0.1 dex ( 25% ) . in other words ,
the excess of r - agn satellites can not simply be a result of r - agn having dark matter halos that are systematically more massive than those of control galaxies . .
the solid line is the least - squares fit to the distribution , and the dashed horizontal lines show the average number of satellites for r - agn ( 2.1 ) and control galaxies ( 1.5 ) . for the excess r - agn satellites to be a result of mismatched halo masses ,
a difference of 0.96 dex ( dotted vertical lines ) between r - agn and controls is necessary .
since the actual difference is only 0.1 dex , r - agn must genuinely have more satellites within 100 kpc than radio - quiet galaxies .
, width=312 ] in addition to the triggering of r - agn , we also wish to examine the satellite population with the goal of understanding the effects of r - agn interactions ( presumably that of the jet ) with satellites . in the discussion that follows , we use color distributions of satellites to examine two scenarios . in the first ,
r - agn jets frequently quench star formation in their satellites , while in the second interactions with r - agn jets lead to induced star formation .
we study the effects of r - agn feedback by comparing the difference between @xmath63 color distributions of r - agn and control satellites .
let us first examine how feedback would modify the difference in color distribution between r - agn and control galaxies in the absence of an overall excess of satellites around r - agn .
if neither quenching nor induced star formation are present , the difference would be zero , as in panel a of fig .
if interactions commonly lead to quenching , there would be a deficit of blue r - agn satellites accompanied by an increase in the number of red r - agn satellites relative to the control distribution , as shown in panel b. next , we consider two scenarios of induced star formation . in the first , radio jets induce star formation in gas clouds that would not otherwise be detected as galaxies , i.e. induced star formation forms new galaxies . induced star formation would cause r - agn to have an excess of blue satellites , as shown in panel c. faint satellite galaxies that increase in luminosity as a result of induced star formation would also lead to an excess of blue satellites .
in the other scenario , interactions with r - agn jets may enhance the sfr in satellites that already form stars , making them bluer ( panel d ) . in this scenario
there is no net excess , but some moderately blue galaxies become bluer , causing a deficit at moderate value colors and an excess at very blue colors .
however , from section [ results ] we know that there is an intrinsic excess of satellites around r - agn , so the scenarios described above need to be modified to take that into account , as shown in figure [ radiomore ] .
as we have seen , this excess is primarily red simply because most satellites are red , so we obtain modified scenarios by adding a red bump to the scenarios described above . now these modified scenarios can be compared to the observations .
figure [ minus ] shows the difference plots for the three environment types as well as the herg sample .
the vertical dotted line at @xmath76 indicates the split between blue and red satellites such that these plots can be easily compared to fig .
[ radiomore ] .
panels a@xmath77c of fig . [ minus ] resemble panel a of fig .
[ radiomore ] , with no blue deficit accompanying the red bump .
this indicates that in any environment , lergs play no role in quenching star formation in satellites .
@xcite examined whether r - agn quench star formation in satellite galaxies that lie in the projected path of the radio jets .
these authors manually examined the satellites of a small sample of 21 fr1 and 72 fr2 sources and classified them , based on their spatial relation to the radio jets as lying inside or outside the jet path .
it was found that the @xmath12 color distribution of satellites in the path of fr1 radio jets is similar to that of satellites outside the path while satellites in the jet path of fr2s are redder than satellites outside the path .
@xcite therefore concluded that only the high - power fr2 galaxies quench star formation in satellites .
if we assume that most of the hergs in our sample present fr2 morphology , then we may compare our findings for herg satellites to the results of @xcite .
panel d of fig .
[ minus ] shows that , like r - agn in all environments , hergs have an excess of red satellites .
this excess is not accompanied by a deficit of blue r - agn satellites , as the sum of the bins bluer than @xmath76 is 0.04 ( @xmath78 ) .
our results indicate that high accretion r - agn , like low accretion ones , at best only rarely quench star formation in nearby satellites .
this non - detection of quenching is in apparent disagreement with the results of @xcite , but note that we have considered quenching on entire satellite populations , while @xcite only investigated satellites in the path of the jet .
if r - agn frequently induced star formation in their satellites , we would expect an excess of blue satellites between @xmath79 in fig .
[ minus ] .
such an excess is not seen for r - agn in any environment , nor for hergs .
the field sample places a limit on the incidence of star formation induction to no more than 1% .
this provides evidence that star formation is rarely induced by interactions with r - agn jets .
we have included in panel b of fig .
[ minus ] a point showing what would be 5% occurrence of satellites with the color of minkowski s object .
the actual value is consistent with zero , which indicates that r - agn rarely induce star formation in external gas clouds , nor do they lead to any significant enhancement in existing gas - rich satellites .
we have studied r - agn triggering by comparing the satellite populations of r - agn and a control sample of radio - quiet galaxies , which was matched to the r - agn in luminosity , color , redshift , axis ratio , and cluster membership .
we also compared these two satellite populations to search for evidence of quenching in satellites as a result of jet interactions as well as jet - induced star formation in satellites .
we separately analyzed hergs to determine whether either the triggering or the feedback behaves differently for powerful radio agn .
our conclusions can be summarized as follows : 1 .
the incidence rate of r - agn , which we define as the fraction of r - agn among the galaxies whose properties are such that they could host an r - agn and are found in a similar environment , is a strong function of luminosity , as found in previous studies that defined the incidence rate in a less restrictive way ( e.g. , with respect to all galaxies in a mass bin ) .
we find that the incidence rate of field r - agn and cluster member r - agn is comparable ( 4% for field , 6% for members on average ) , while that of bcgs is significantly higher at the same luminosity ( 14% on average ) .
the relative similarity of the incidence rates of field and non - bcg cluster members suggests that some previous studies that find that r - agn are more clustered on cluster scales ( 1 mpc ) either derive that signal from bcgs , which are indeed more likely to be r - agn , or from the fact that the type of galaxies that host r - agn ( massive ellipticals ) are more likely to be found in clusters .
the incidence of r - agn is consequently primarily dependent on the availability of fuel ( hot gas ) locally , in the individual halos , rather than in the cluster halo .
lerg r - agns belonging to a given environment type ( field , cluster member , bcg ) have an excess of satellites out to at least 100 kpc in projected distance compared to similar radio - quiet galaxies in respective environments .
this excess is on average 50% for field r - agn . for many galaxies this excess results in having a massive satellite vs. not having any .
this finding is similar to the @xcite result of an overdensity of r - agn with respect to optical , radio - quiet agn , with the important difference that we control for the general type of environment .
thus , we find that even r - agn bcgs have more galaxies in their vicinity than radio - quiet bcgs .
the excess is not the result of a mismatch in halo masses , but suggests a genuine relation between satellites and r - agn triggering .
because the excess of red satellites is not accompanied by a deficit of blue satellites , we conclude that neither hergs nor lergs are instrumental in quenching star formation in their satellites .
5 . similarly , since no excess of blue r - agn satellites is observed we conclude that lergs and hergs rarely ( @xmath80 of cases ) induce star formation in satellites .
our results shine new light on the scenario in which hergs are found in sparse environments and are fueled by the accretion of cold gas , while lergs prefer dense environments and accrete small quantities of hot halo gas .
we find that the relevant environment is on small scales ( @xmath81 kpc ) .
thus , the fueling of lergs either becomes more probable , or made more extensive and/or stronger by the excess of nearby satellites , which explains for example why not every bcg hosts an r - agn , despite the dense cluster environment .
alternatively , excess satellites may trace an increased infall rate onto hosts that also feeds the gas to the black hole .
we will address these possibilities in future work .
finally , our results regarding the feedback in satellite population have implications on the role of r - agn feedback in general .
we thank the referee for constructive comments and suggestions .
this study uses data from the sdss archive .
funding for sdss - iii has been provided by the alfred p. sloan foundation , the participating institutions , the national science foundation , and the u.s .
department of energy office of science .
the sdss - iii web site is http://www.sdss3.org/. sdss - iii is managed by the astrophysical research consortium for the participating institutions of the sdss - iii collaboration including the university of arizona , the brazilian participation group , brookhaven national laboratory , university of cambridge , carnegie mellon university , university of florida , the french participation group , the german participation group , harvard university , the instituto de astrofisica de canarias , the michigan state / notre dame / jina participation group , johns hopkins university , lawrence berkeley national laboratory , max planck institute for astrophysics , max planck institute for extraterrestrial physics , new mexico state university , new york university , ohio state university , pennsylvania state university , university of portsmouth , princeton university , the spanish participation group , university of tokyo , university of utah , vanderbilt university , university of virginia , university of washington , and yale university .
this research made use of the `` k - corrections calculator '' service available at http://kcor.sai.msu.ru/
in our study , photometric redshifts ( photo - z s ) are used to reduce contamination from faint background galaxies .
a number of photometric redshifts are available in the sdss database .
the primary photometric redshifts from dr7 , which are stored as ` photoz ` in the dr7 catalog archive server ( cas ) , are calculated using a technique that combines template fitting with an empirical fitting procedure to estimate the redshift @xcite . for the eighth data release
, this method was also used with the inclusion of the galaxy s inclination angle to reduce systematic bias @xcite .
these redshifts are stored in the dr8 cas as ` photoz ` .
an alternative approach using the random forest technique described in @xcite was used to estimate the photometric redshifts which are stored as ` photozrf ` . in order to determine which photometric redshift is the most consistently accurate , we compared the three photometric redshifts from dr7 and dr8 to the spectroscopic redshifts ( spec - z s ) of a sample of galaxies from the deep extragalactic evolutionary probe 2 ( deep2 ) survey @xcite .
we began with the unique redshift catalog from the deep2 data release 3 ( dr3 ) , of which only galaxies with a reliable redshift and within the extended groth strip ( egs ) were used .
these galaxies were then searched for in both dr7 and dr8 , and 3,600 were found to have photometric redshifts in both data releases .
we used the same magnitude cut of @xmath11 = 22 as was used for the galaxies from sdss , which left 1,897 galaxies .
the spectroscopic redshifts of these galaxies from deep2 together with the three photometric redshifts are plotted in fig .
[ photoz ] .
panel @xmath82 shows the photometric redshifts reported in dr7 , panel @xmath83 shows the primary photometric redshifts from dr8 , while panel @xmath84 plots the dr8 photometric redshifts derived using the random forest technique . in all panels , the spectroscopic redshift is plotted on the abscissa .
figure [ photoz ] shows that for the sample as a whole , all three methods for determining photometric redshifts tend to underestimate redshifts , and this is more pronounced for dr7 .
the scatter between photometric and spectroscopic redshifts is large at about 0.15 dex and is not symmetric as objects at redshifts greater than @xmath31 have systematically underestimated redshifts .
because of this large asymmetric scatter , we decided to use photometric redshifts to only remove objects with the most discrepant ( large ) redshifts .
we used the following process to select which photometric redshift to use . to simplify the analysis we treat our target galaxies as if they were at a redshift of @xmath85 , the mean redshift of the r - agn sample .
we then compute the fraction of galaxies in the background ( spec - z @xmath86 ) of this @xmath85 galaxy " that are identified as background galaxies by their photometric redshifts .
the large scatter in photometric redshifts means that many galaxies with a photo - z @xmath86 are in fact at a spec - z @xmath87 , so we add a 0.15 dex margin and consider as background galaxies those for which photo - z @xmath88 . of the 1,793 galaxies with spec - z @xmath89 0.15 , the dr7 ` photoz ` was greater than 0.3 for 663 galaxies , or 37% of the total . for the dr8 ` photoz ` , 1,241 galaxies had an estimated redshift greater than 0.3 or 69% of the total , while the dr8 ` photozrf ` estimated a redshift greater than 0.3 for 1,106 of the galaxies , or 62% of the total .
we therefore conclude that ` photoz ` from dr8 is the most useful for identifying background galaxies whose redshift is greater than the @xmath90 = 0.3 redshift limit of our sample of radio galaxies .
abazajian , k. n. , adelman - mccarthy , j. k. , ageros , m. a. , et al .
2009 , , 182 , 543 aihara , h. , allende prieto , c. , an , d. , et al .
2011 , , 193 , 29 allen , s. w. , dunn , r. j. h. , fabian , a. c. , taylor , g. b. , & reynolds , c. s. 2006 , , 372 , 21 antonuccio - delogu , v. , & silk , j. 2008 , , 389 , 1750 becker , r. h. , white , r. l. , & helfand , d. j. 1995 , , 450 , 559 baldwin , j. a. , phillips , m. m. , & terlevich , r. 1981 , , 93 , 5 barnes , j. e. , & hernquist , l. 1996 , , 471 , 115 best , p. n. 2004 , , 351 , 70 best , p. n. , & heckman , t. m. 2012 , , 421 , 1569 best , p. n. , kauffmann , g. , heckman , t. m. , & ivezi , . 2005a , , 362 , 9 best , p. n. , kauffmann , g. , heckman , t. m. , et al . 2005b , , 362 , 25 best , p. n. , von der linden , a. , kauffmann , g. , heckman , t. m. , & kaiser , c. r. 2007 , , 379 , 894 bicknell , g. v. , sutherland , r. s. , van breugel , w. j. m. , et al . 2000 , , 540 , 678 buttiglione , s. , capetti , a. , celotti , a. , et al .
2010 , , 509 , a6 carliles , s. , budavri , t. , heinis , s. , priebe , c. , & szalay , a. s. 2010 , , 712 , 511 chen , j. , kravtsov , a. v. , prada , f. , et al .
2006 , , 647 , 86 cheng , j. y. , faber , s. m. , simard , l. , et al .
2011 , , 412 , 727 chilingarian , i. v. , & zolotukhin , i. y. 2012 , , 419 , 1727 condon , j. j. , cotton , w. d. , greisen , e. w. , et al .
1998 , , 115 , 1693 crockett , r. m. , shabala , s. s. , kaviraj , s. , et al .
2012 , , 421 , 1603 croft , s. , van breugel , w. , de vries , w. , et al .
2006 , , 647 , 1040 davis , m. , guhathakurta , p. , konidaris , n. p. , et al .
2007 , , 660 , l1 de young , d. s. 1981 , , 293 , 43 donoso , e. , li , c. , kauffmann , g. , best , p. n. , & heckman , t. m. 2010 , , 407 , 1078 dressler , a. 1980 , , 236 , 351 ekers , r. d. , & ekers , j. a. 1973 , , 24 , 247 fabian , a. c. 1994 , , 32 , 277 heckman , t. m. , smith , e. p. , baum , s. a. , et al .
1986 , , 311 , 526 inskip , k. j. , villar - martn , m. , tadhunter , c. n. , et al .
2008 , , 386 , 1797 kauffmann , g. , heckman , t. m. , & best , p. n. 2008 , , 384 , 953 kauffmann , g. , heckman , t. m. , tremonti , c. , et al .
2003 , , 346 , 1055 laing , r. a. , jenkins , c. r. , wall , j. v. , & unger , s. w. 1994 , the physics of active galaxies , 54 , 201 liu , f. k. , & zhang , y. h. 2002 , , 381 , 757 mandelbaum , r. , li , c. , kauffmann , g. , & white , s. d. m. 2009 , , 393 , 377 sabater , j. , best , p. n. , & argudo - fernndez , m. 2013 , , 430 , 638 khochfar , s. , kaviraj , s. , et al .
2006 , , 442 , 888 schoenmakers , a. p. , de bruyn , a. g. , rttgering , h. j. a. , van der laan , h. , & kaiser , c. r. 2000 , , 315 , 371 shabala , s. s. , kaviraj , s. , & silk , j. 2011 , , 413 , 2815 silk , j. , & norman , c. 2009 , , 700 , 262 strauss , m. a. , weinberg , d. h. , lupton , r. h. , et al .
2002 , , 124 , 181 tremblay , g. r. , odea , c. p. , baum , s. a. , et al .
2010 , , 715 , 172 van breugel , w. j. m. , & dey , a. 1993 , , 414 , 563 van breugel , w. , fragile , c. , croft , s. , et al .
2004 , the interplay among black holes , stars and ism in galactic nuclei , 222 , 485 van velzen , s. , falcke , h. , schellart , p. , nierstenhfer , n. , & kampert , k .- h .
2012 , , 544 , a18 vron - cetty , m. p. , & vron , p. 2001
, , 375 , 791 wake , d. a. , croom , s. m. , sadler , e. m. , & johnston , h. m. 2008 , , 391 , 1674 worpel , h. , brown , m. j. i. , jones , d. h. , floyd , d. j. e. , & beutler , f. 2013 , arxiv:1305.2673 yang , x. , mo , h. j. , van den bosch , f. c. , et al . 2007 , , 671 , 153 yip , c .- w . , szalay , a. s. , carliles , s. , & budavri , t. 2011 , , 730 , 54 | we study the effects of radio jets on galaxies in their vicinity ( satellites ) and the role of satellites in triggering radio - loud active galactic nuclei ( agns ) .
the study compares the aggregate properties of satellites of a sample of 7,220 radio agns at @xmath0 ( identified by best & heckman 2012 from the sdss and nvss+first surveys ) to the satellites of a control sample of radio - quiet galaxies , which are matched in redshift , color , luminosity , and axis ratio , as well as by environment type : field galaxies , cluster members and brightest cluster galaxies ( bcgs ) .
remarkably , we find that radio agns exhibit on average a 50% excess ( 17@xmath1 significance ) in the number of satellites within 100 kpc even though the cluster membership was controlled for ( e.g. , radio bcgs have more satellites than radio - quiet bcgs , etc . ) .
satellite excess is not confirmed for high - excitation sources , which are only 2% of radio agn .
extra satellites may be responsible for raising the probability for hot gas agn accretion via tidal effects or may otherwise enhance the intensity or duration of the radio - emitting phase .
furthermore , we find that the incidence of radio agns among potential hosts ( massive ellipticals ) is similar for field galaxies and for non - bcg cluster members , suggesting that agn fueling depends primarily on conditions in the host halo rather than the parent , cluster halo .
regarding feedback , we find that radio agns , either high or low excitation , have no detectable effect on star formation in their satellites , as neither induced star formation nor star formation quenching is present in more than @xmath2 of radio agn . |
in this section we give an overview of the general problem , and describe what is known in the @xmath7 case .
most of this material was first discussed by low ( see @xcite , @xcite @xcite , @xcite , @xcite ) , and can be found in greater detail in @xcite .
let @xmath1 be a @xmath0-dimensional globally hyperbolic spacetime with a fixed time orientation , and consider the set @xmath2 of all its null geodesics ( we define a null vector to be a _
vector @xmath10 such that @xmath11 , and hence the constant geodesic is _ not _ a null geodesic ) .
if @xmath12 is a cauchy surface , then every null geodesic intersects @xmath13 exactly once . on the other hand , at any event @xmath14 two future - pointing null vectors are initial conditions for the same null geodesic _
iff _ they are linearly dependent , or , equivalently , _
iff _ their orthogonal projections on @xmath15 are linearly dependent .
consequently , @xmath2 can be identified with the tangent sphere bundle @xmath16 ( recall that @xmath13 endowed with the metric induced by @xmath17 is a riemannian manifold of dimension @xmath18 ) .
we use this fact to define a differentiable structure on @xmath2 , and notice that , due to smooth dependence of the solutions of the geodesic equation on its initial data , this structure is independent of the choice of cauchy surface .
thus the set of all null geodesics of a globally hyperbolic @xmath0-dimensional spacetime is a differentiable manifold of dimension @xmath19 , which we call its _ manifold of light rays_.
the _ sky _ of an event @xmath4 is the subset @xmath20 if @xmath14 then @xmath21 is a fibre of the tangent sphere bundle @xmath16 , and therefore is an embedded submanifold of @xmath2 diffeomorphic to @xmath6 . since we are free to regard @xmath2 as a fibre bundle over _ any _ cauchy surface , and every event in @xmath22 belongs to _ some _ cauchy surface , we see that any sky is an embedded @xmath6 .
let us now assume that @xmath22 is orientable .
then by choosing a global time function @xmath23 and using the globally defined nonvanishing future - pointing timelike vector field @xmath24 we can orient all cauchy surfaces @xmath25 ; hence we can define an orientation on each sky by orienting tangent spheres on each cauchy surface ( it is easy to check that this orientation does not depend on the choice of global time function ) . if @xmath8 are not in the same null geodesic then @xmath26 is a _ smooth link _
, i.e. , a disjoint union of embedded @xmath6s in a smooth manifold of dimension @xmath27 .
recall that a smooth one - parameter family of diffeomorphisms @xmath28 \times n \to n$ ] is said to be a _
smooth ( ambient ) isotopy _ if @xmath29 is the identity map .
two links @xmath30 and @xmath31 are said to be _ equivalent _ if there exists a smooth isotopy @xmath28 \times
n \to n$ ] such that @xmath32 and @xmath33 .
it can be shown that two links @xmath30 and @xmath31 are equivalent _ iff _ there exists a _
smooth motion _
carrying one into the other , i.e. , _ iff _ there exist smooth one - parameter families of embeddings @xmath34 and @xmath35 such that @xmath36 , @xmath37 , @xmath38 , @xmath39 and @xmath40 for all @xmath41 $ ] .
if @xmath42 and @xmath43 are two pairs of non - causally related events , one can easily construct smooth curves
@xmath44 \to m$ ] such that @xmath45 , @xmath46 , @xmath47 , @xmath48 and @xmath49 are not in the same null geodesic for all @xmath41 $ ] .
this induces a smooth motion of @xmath30 into @xmath31 , which are therefore equivalent .
we conclude that the skies of any two non - causally related events belong to the same equivalence class , which we define as the _ unlink _ in @xmath2 ( notice that although there exists a natural choice for the unlink in @xmath50 or @xmath51 , this is not the case in general for an arbitrary smooth @xmath3-dimensional manifold ) .
having said that , we now focus on the case when @xmath7 and @xmath13 is diffeomorphic to a subset of @xmath52 . in this case , @xmath2 is diffeomorphic to a subset of the tangent sphere bundle of @xmath52 , which in turn is diffeomorphic to the interior of a solid torus in @xmath53 .
it will prove useful to fix a particular embedding @xmath54 ( which we shall call the _ standard embedding _ ) .
thus , if @xmath55 are the usual polar coordinates in @xmath52 , and @xmath56 is the coordinate in the fibres of @xmath57 corresponding to the angle with the positive @xmath58-direction , we define @xmath59 this particular embedding has the advantage of carrying the unlink in @xmath2 ( as defined above ) to the usual unlink in @xmath53 .
notice that the skies of events on @xmath13 are mapped to circles of constant @xmath55 , which for convenience we call _
meridians_. from now on we shall identify @xmath2 with @xmath60 .
clearly a smooth motion of a smooth link in @xmath2 is a smooth motion of this link in @xmath53 , and consequently if two links are equivalent in @xmath2 they must be equivalent in @xmath53 .
recall that in @xmath53 there is a well - known link invariant , namely the _ linking number_. given a link @xmath26 , this is simply the integer given by the _
gauss integral _ , @xmath61 ( where @xmath62 are embeddings for @xmath63 ) , or , equivalently , the integer such that @xmath64 $ ] is @xmath65 times an appropriate generator of @xmath66 , depending on the orientation of @xmath67 ( see @xcite ) . in particular
if @xmath26 is the unlink then @xmath68 .
notice also that , as is easily seen from the gauss integral formula , @xmath69 .
let @xmath8 be two events not in the same null geodesic .
recall we can always assume that @xmath70 , i.e. , that @xmath67 is a meridian . in general
, however , @xmath71 .
since we can think of @xmath2 as @xmath16 , we have a natural smooth projection @xmath72 .
clearly , @xmath73 is the intersection of the light cone of @xmath74 with @xmath13 , and is a piecewise smooth closed curve not containing @xmath75 ; we call this the _ wavefront _ generated by @xmath74 at @xmath13 .
since @xmath21 is an oriented curve , @xmath73 can be given the induced orientation .
we can think of @xmath76 as quotienting @xmath2 by its meridians , and consequently we can identify @xmath13 with the quotient space .
thus any smooth surface on @xmath2 intersecting each meridian exactly once can be identified with @xmath13 . in this manner we can think of @xmath75 and @xmath73 as a point and a curve on @xmath77 .
one can always choose @xmath77 such that @xmath78 is finite .
it is then possible to construct an isotopy of @xmath2 deforming @xmath21 into a curve which approaches ( with any required accuracy ) the wavefront @xmath73 plus a finite number of meridians , one hanging from each intersection of @xmath21 and @xmath13 . indeed
, if @xmath56 is the angular coordinate along the meridians such that @xmath79 , one has but to consider isotopies of the form @xmath80 , where @xmath81 \times [ 0,2 \pi ] \to [ 0 , 2 \pi]$ ] is a nonnegative smooth function vanishing for @xmath82 and @xmath83 and approaching @xmath84 from below as @xmath85 . since the linking number of any two disjoint meridians is zero , we see that @xmath86 , where @xmath87 is the winding number of the curve @xmath88 around the point @xmath70 ( recall that @xmath13 is diffeomorphic to a subset of @xmath52 ) .
( this result plus the fact that @xmath69 allows us to make the following nontrivial observation : if @xmath89 are not in the same null geodesic and @xmath90 are arbitrary cauchy surfaces through @xmath89 then the winding number of the wavefront generated by @xmath74 at @xmath91 around @xmath75 is equal to the winding number of the wavefront generated by @xmath75 at @xmath92 around @xmath74 ) . as an example , consider minkowski s @xmath93-dimensional spacetime , i.e. , @xmath53 endowed with the line element @xmath94 , and take hypersurfaces of constant @xmath95 as cauchy surfaces .
then all wavefronts are circles , and all linking numbers are therefore either zero or one .
since an event on the cauchy surface is causally related to the event generating the wavefront _ iff _ it is either inside the wavefront ( in which case the winding number is @xmath96 ) or on it , we see that two events in minkowski @xmath93-dimensional spacetime are causally related _ iff _ their skies either intersect or are linked with linking number @xmath96 .
another example is provided by schwarzschild s @xmath93-dimensional static spacetime , i.e. , the region of @xmath53 given in cylindrical coordinates @xmath97 by @xmath98 endowed with the line element @xmath99 ( we ve taken the schwarzschild radius as our length unit ) .
if we again take hypersurfaces of constant @xmath95 as cauchy surfaces , then it is possible to show that the wavefronts are as shown in figure [ picture ] , wrapping around the event horizon any number of times .
it is then easy to see that all winding numbers are either zero or positive , and that an event in the cauchy surface not on the wavefront is causally related to the event generating the wavefront _ iff _ the winding number is positive .
consequently , two events in schwarzschild @xmath93-dimensional static spacetime are causally related _ iff _ their skies either intersect or are linked with positive linking number ( but links do occur with any positive linking number ) .
it is not true in general that all links formed by the skies of causally related events have nonvanishing linking number .
a simple counter - example is provided by @xmath53 endowed with the line element @xmath100 , where @xmath101 is an appropriate smooth function ( equal to @xmath96 except on the circles @xmath102 or @xmath103 , where it increases radially towards the centre ) .
the wavefront of the event @xmath104 on the cauchy surface @xmath105 is as depicted in figure [ picture ] , each pair of cusps corresponding to scattering by one of the circles where the metric is not flat .
the appearance of the second pair of cusps allows the existence of events @xmath75 that , although clearly causally related to @xmath74 , are such that the winding number of @xmath73 around @xmath75 is zero , and consequently @xmath68 .
however @xmath26 is _ not _ the unlink .
to see that , we notice that the riemannian metric induced on @xmath13 is conformally related to the euclidean metric , and that we can therefore use the usual angle with the @xmath74-axis as a coordinate on @xmath16 .
to decide on the value of this coordinate along the wavefront , we recall that the tangent vector @xmath106 to any null geodesic is orthogonal to the light cone of @xmath74 .
consequently , the element of @xmath16 corresponding to the null geodesic is the normal vector ( on @xmath13 ) to the wavefront , oriented so that it points outwards when the wavefront is in the boundary of the causal future . using these rules
one constructs the link shown in figure [ picture ] , which is the so - called _ whitehead link _ ( and famously _ not _ the unlink ) .
the above examples have led low to conjecture that then @xmath8 are causally related _ iff _ @xmath9 are linked .
we now proceed to prove low s conjecture for a ( large ) class of static spacetimes ( which in particular includes the above example ) .
recall that a @xmath0-dimensional spacetime @xmath1 is said to be _
static _ if it admits a nonvanishing timelike killing vector field @xmath107 whose orthogonal distribution is integrable .
the integral submanifolds of the orthogonal distribution endowed with the induced riemannian metric are isometric riemannian manifolds ( the isometry being generated by the flow of @xmath107 ) , which we call _ space manifolds_. notice that if @xmath13 is any space manifold then the integral lines of @xmath107 provide a natural projection @xmath108 .
a static @xmath93-dimensional spacetime is said to be _ regular _ if 1 .
@xmath1 is globally hyperbolic and orientable ; 2 .
the space manifolds are cauchy surfaces diffeomorphic to a subset of @xmath52 ; from this point on we shall assume that @xmath1 is a regular static @xmath93-dimensional spacetime .
notice that all three examples considered in the previous section are regular .
any chart @xmath109 defined on a space manifold @xmath13 can be extended to a chart @xmath110 by using the parameter @xmath95 along the integral curve of @xmath107 from the point in @xmath13 with coordinates @xmath111 . in this chart
we have @xmath112 .
if @xmath113 are the components of the space manifold ( riemannian ) metric on the chart @xmath109 , the line element of the full spacetime manifold on the chart @xmath110 is @xmath114 where @xmath115 one has the following _ fermat principle _ ( see @xcite ) : [ fermat ] let @xmath1 be a globally hyperbolic spacetime , @xmath23 a global time function , @xmath4 an event and @xmath116 a smooth timelike curve such that @xmath117 .
for each future - directed null curve @xmath118 through @xmath74 and @xmath119 ( in that chronological order ) let @xmath120 , where @xmath121 is the first intersection of @xmath122 and @xmath119 to the future of @xmath74 .
then the null geodesics through @xmath74 and @xmath119 are the critical points of @xmath107 .
any segment of null geodesic in the chart @xmath110 satisfies @xmath123 using this plus theorem [ fermat ] ( with the integral curves of @xmath107 in the role of @xmath119 ) it is easy to prove let @xmath1 be a regular static @xmath93-dimensional spacetime and @xmath124 a space manifold . then the projections of null geodesics on @xmath13 parametrized by the time coordinate @xmath95 are the geodesics of the riemannian metric @xmath125 on @xmath13 . for convenience , we make the following let @xmath1 be a regular static @xmath93-dimensional spacetime and @xmath124 a space manifold
. then the riemannian metric @xmath126 on @xmath13 is called the _ light metric _ , its geodesics the _ light geodesics _ , the corresponding geodesic spheres the _ light spheres _ and the corresponding geodesic balls the _ light balls_. clearly light spheres on a space manifold are just the wavefronts generated by various events . consequently , the light geodesics are perpendicular to the wavefronts ( both in the light and space metrics , as they are conformally related ) .
this can also be seen to be a simple consequence of the fact that null geodesics are orthogonal to the null cone , as the tangent vector of the light geodesic is parallel to the orthogonal projection of the tangent vector of the null geodesic on @xmath127 .
since the space manifolds are diffeomorphic to a subset of @xmath52 , we can fix global coordinates @xmath128 on any particular space manifold @xmath13 .
again we set the section defined on @xmath16 by the vector field @xmath129 as the surface @xmath130 on @xmath2 , and define the ( multivalued ) coordinate @xmath56 in @xmath2 as the angle with this vector at each point ( measured in either metric ) .
notice that for our purposes it is immaterial whether we use the space metric @xmath131 or the conformally - related light - metric @xmath132 in the construction of @xmath16 .
for obvious simplicity reasons we shall prefer the latter .
the same nondegenerate light sphere on a space manifold is the wavefront generated by exactly _ two _ events . indeed ,
if the centre of the sphere is the point in @xmath133 with coordinates @xmath134 and its radius is @xmath135 , then the sphere is the wavefront generated by the events with coordinates @xmath136 in the induced chart .
although the skies of these two events project down to the same wavefront on @xmath13 , they are disjoint curves in @xmath2 , as each point on the wavefront is the projection of two antipodal points in the corresponding fibre of @xmath16 ( one in each sky ) .
thus any sky is completely defined by its wavefront as long as we indicate a unit normal vector on the wavefront ( it suffices to do so at a particular point of the wavefront , as the rest follows from continuity and orthogonality ) . this extra bit of information
is called a _
coorientation _ of the wavefront ; it can be thought of as an indication of which way the wavefront is moving .
consider the event @xmath4 with coordinates @xmath137 .
for convenience , let us define @xmath138 to be the event with coordinates @xmath139 , so that @xmath140 and @xmath141 .
it should be clear that an event @xmath70 is causally related to @xmath74 _ iff _ there is a null geodesic through @xmath75 and @xmath142 for some @xmath143 .
consequently the set of all events in @xmath13 which are causally related to @xmath74 is the union of the light spheres with centre @xmath144 and radius @xmath143 , i.e. , is the light ball @xmath145 of radius @xmath146 and centre @xmath144 .
@xmath145 can also be thought of as the union of all light geodesic segments with endpoint @xmath144 and length @xmath146 .
if one gives the light spheres the obvious coorientation , or , equivalently , the light geodesics the corresponding orientation ( away from @xmath144 ) , one can lift @xmath145 onto a surface @xmath147 on @xmath148 by lifting each cooriented light sphere to the corresponding sky , or , equivalently , by lifting each point in the oriented light geodesic to its unit tangent vector .
the surface @xmath149 is called the _ causality surface _ of the event @xmath4 .
the reason for this name is given in the following an event @xmath70 is causally related to @xmath74 _ iff _ @xmath150 .
we ve seen that @xmath70 is causally related to @xmath74 _ iff _
@xmath151 this is a fundamental step in the proof of low s conjecture , inasmuch it gives us a precise criterion to decide which skies in @xmath2 correspond to causally related events .
notice that if one is given a pair of skies it is not obvious in general whether the corresponding events are causally related , even when one of the skies projects down to a point in @xmath13 , as the wavefront generated by the other event can easily be a very complicated curve .
we now analyse the structure of the causality surface @xmath147 in detail .
since @xmath147 can be seen as a union of skies of events in the line segment @xmath152\right\}$ ] and one can move from sky to sky along the lifts of light geodesics , it is easy to see that the causality surface is a immersion of @xmath153 $ ] in @xmath2 .
consequently the boundary of @xmath147 has two connected components : one is the sky of @xmath74 , and the other is the sky of @xmath144 , which is a meridian . for completeness , we identify how @xmath147 can fail to be an embedding : if the causality surface is not embedded then it self - intersects along segments of lifts of light geodesics . the number of such self - intersection curves is finite if @xmath144 is not conjugate to itself with respect to the light metric . if @xmath147 is not an embedded submanifold of @xmath2 , then @xmath154 for some @xmath155 , r < s$ ] ( as @xmath153 $ ] is compact ) .
thus @xmath142 and @xmath156 are in the same null geodesic @xmath122 . in other words , @xmath157 is a light geodesic containing a segment of length @xmath158 which connects @xmath144 to itself .
consequently , @xmath159 for @xmath160 $ ] , and @xmath147 self - intersects along a curve .
finally , considering the simultaneous time evolution of the light spheres of centre @xmath144 and radii @xmath161 and @xmath162 it is easy to see that this curve is simply the lift of the segment of light geodesic given by @xmath163 .
the fact that if @xmath144 is not conjugate to itself there are only finitely many self - intersection curves is a consequence of theorem [ conjugate ] .
recall that @xmath147 is ruled by the lifts of all light geodesics through the point @xmath140 .
each point in @xmath164 is the initial condition for one such geodesic , and since @xmath165 is a meridian it can be parametrized by the coordinate @xmath56 in @xmath2 .
consequently , an immersion @xmath166 \to n$ ] of @xmath147 is obtained by defining @xmath167 to be the unit tangent vector to the light geodesic with initial condition in @xmath165 corresponding to @xmath168 at the point a distance @xmath95 from @xmath144 along that geodesic ( in particular @xmath169 ) . also , notice that @xmath170 is the unit tangent vector to the light geodesic corresponding to the appropriate value of @xmath168 , whereas @xmath171 is the jacobi field corresponding to this family of light geodesics . since @xmath172 and @xmath173
, it is therefore clear that the critical points of @xmath174 are the zeroes of @xmath175 , i.e. , the conjugate points to @xmath144 in the light metric .
we now introduce some useful terminology .
a vector @xmath176 is said to be _ vertical _ if @xmath177 , i.e. , if it is a multiple of @xmath178 . a vertical vector is said to point _ up _ if it is a _ positive _ multiple of @xmath178 , and to point _ down _ if it is a _ negative _ multiple of @xmath178 .
a vector @xmath176 is said to be _ horizontal _ if it is orthogonal to @xmath178 ( in the usual euclidean metric of @xmath179 ) .
these definitions have obvious extensions to curves and surfaces : a curve or surface will be said to be _ vertical _ at a given point if its tangent space at that point contains a nonzero vertical vector , and will be said to be _ horizontal _ at a given point if all vectors in its tangent space at that point are horizontal .
thus meridians are vertical curves , whereas curves of constant @xmath56 are horizontal surfaces . to prove low s conjecture we just have to show that for every @xmath70 such that @xmath150 and @xmath180 the link @xmath26 is not the unlink .
we start by showing that we can assume @xmath181 to be finite .
[ conjugate ] if there exist infinitely many light geodesics of length @xmath182 connecting @xmath183 then these points are conjugate .
let @xmath184 be _ different _ initial conditions at @xmath165 for light geodesics connecting @xmath144 to @xmath75 , and @xmath185 their lengths .
since @xmath186 $ ] is compact , we can assume that @xmath187 as @xmath188 ( otherwise extract a convergent subsequence ) .
it is a simple consequence of the continuous dependence of geodesics on their initial conditions that the light geodesic with initial condition @xmath189 connects @xmath144 to @xmath75 , and that the light distance between @xmath144 and @xmath75 measured along this geodesic is @xmath190 .
clearly we can assume @xmath191 to be on the same chart of @xmath192 .
let us define @xmath193 and @xmath194 ( where @xmath195 is the usual euclidean norm in @xmath52 ) .
since @xmath196 , it has at least one sublimit @xmath197 .
we assume that @xmath198 was already chosen so that @xmath199 is in fact convergent .
setting @xmath200 we have from lagrange s theorem that there exists @xmath201 such that @xmath202 where the partial derivatives on the right - hand side are calculated at point @xmath203 for some @xmath201 .
thus at this point @xmath204 and taking the limit we get @xmath205 at point @xmath206 .
the same is true for @xmath207 , and we therefore conclude that @xmath197 is in the kernel of @xmath208 at @xmath206 . consequently , @xmath144 is conjugate to @xmath209 .
for every @xmath210 there exists @xmath211 such that at most finitely many light geodesics connect @xmath144 to @xmath212 .
we just have to notice that the set of points in @xmath13 which are conjugate to @xmath144 has zero lebesgue measure ( see for instance @xcite ) .
there exists @xmath213 such that @xmath214 is equivalent to @xmath26 and @xmath215 is finite .
since the light sphere @xmath73 is a closed set and @xmath216 , there exists @xmath217 such that @xmath218 .
any path in @xmath219\to b_\varepsilon(y)\subset\sigma$ ] with @xmath220 , @xmath221 induces a smooth motion of @xmath26 into @xmath214 .
thus we just have to choose @xmath222 such that at most finitely many light geodesics connect @xmath144 to @xmath212 .
therefore we shall assume @xmath181 to be finite .
it is clear from the above that we can assume that @xmath75 is not conjugate to @xmath144 along any light geodesic , and hence that @xmath67 is not tangent to @xmath147 at any of the intersections .
recall that @xmath147 is ruled by lifts of light geodesics and has @xmath165 and @xmath21 for boundaries .
consequently , each point of @xmath21 is connected to a unique point of @xmath165 through the lift of a light geodesic . since @xmath165 is a meridian , and
therefore is a much simpler curve than @xmath21 , a natural idea is to deform @xmath21 into @xmath165 using the light geodesics connecting them .
each intersection of @xmath67 and @xmath147 is then an obstruction to deforming @xmath26 into @xmath223 ( which is the unlink ) in this fashion .
let us assume then that @xmath150 , @xmath224 .
suppose that @xmath225 is one of the intersections .
then the @xmath226-_ribbon _ corresponding to this intersection is the subset @xmath227 ( for @xmath228 ) .
notice that this set is a ( closed ) neighbourhood in @xmath147 of the lift of a light geodesic segment joining @xmath144 to @xmath75 .
since the lifts of light geodesics can not intersect , we conclude that for @xmath226 sufficiently small the ( finitely many ) @xmath226-ribbons do not intersect .
see figure [ picture2 ] for a schematic illustration in the simplest case ( minkowski spacetime ) .
let @xmath229 be a disk with boundary @xmath165 not intersecting @xmath67 , so that @xmath230 is a disk with boundary @xmath21 .
let @xmath231 be the union of @xmath232 with all @xmath226-ribbons corresponding to all intersection points in @xmath181 .
notice that @xmath230 must always have self - intersections : since the derivative of the geodesic exponential map at the origin is the identity , for small values of @xmath95 the light geodesics are close to flat metric geodesics , and hence @xmath233 must approach a circle with centre @xmath144 and radius @xmath95 ; consequently , for @xmath234 small enough , @xmath165 and @xmath235 are linked with linking number @xmath96 , and hence @xmath235 must intersect any disk spanned by @xmath165 .
however , @xmath231 does nt necessarily have to self - intersect .
a regular static @xmath93-dimensional spacetime @xmath1 is said to be _ hyper - regular _ if : 1 .
no light geodesic is closed or self - intersects ( i.e. , all causality surfaces are embedded ) ; 2 .
for any two causally related points @xmath8 it is always possible to choose @xmath232 and @xmath228 such that @xmath231 does not self - intersect .
although this definition may seem somewhat artificial ( and indeed is introduced to accommodate the proof of the main result ) , the class of hyper - regular static @xmath93-dimensional spacetimes is fairly large .
for instance , all sufficiently small static perturbations of minkowski spacetime are hyper - regular , as we now show .
a curve @xmath236 is said to _ join _
@xmath237 _ to infinity _ if it is unbounded .
the set of points in all light geodesics joining two points @xmath238 is represented @xmath239 .
it is not difficult to see that the following result holds : if no light geodesic self - intersects and for all @xmath240 there exists a curve @xmath241 joining @xmath242 to infinity such that @xmath243 then @xmath1 is hyper - regular .
this proposition implies in particular that any sufficiently small perturbation of minkowski spacetime is hyper - regular .
also , it is possible to show that from the three examples in the previous section only schwarzschild s spacetime is not hyper - regular . if @xmath1 is hyper - regular then for @xmath228 small enough , @xmath26 can be deformed into @xmath244 .
since @xmath1 is hyper - regular the causality surface is embedded and hence we can deform @xmath21 into along @xmath245 along the lifts of light geodesics without ever intersecting @xmath67 .
we shall assume from this point on that @xmath1 is hyper - regular .
we see that to understand the link @xmath26 we just have to understand the behaviour of the ribbons .
since we can take @xmath217 as small as we like , we just have to consider the vector field @xmath246 along the lift of the corresponding light geodesic , which we do by studying the jacobi field @xmath175 . at @xmath82
we have @xmath247 , and hence the ribbon starts out vertical . since the derivative of the geodesic exponential map at the origin is the identity , for small @xmath95 the light geodesics are close to flat metric geodesics , and hence @xmath175 points towards the _ left _ with respect to the unit tangent vector @xmath248 .
this means that the ribbon twists _ clockwise _ around @xmath249 as one follows the light geodesic in that direction . in any time interval @xmath250
such that @xmath251 we can deform the ribbon so that it stays horizontal ( thus initially the ribbon twists by an angle of @xmath252 with respect to the vertical direction ) . however , when a conjugate point is reached this is no longer possible . at such a point one
has @xmath247 , but since any jacobi field satisfies a second order linear differential equation we must have @xmath253 ( where @xmath254 is the levi - civita connection for the light metric ) .
thus if prior to reaching the conjugate point the jacobi field points _ left _ , afterwards it must point _ right _ , and vice - versa .
consequently , either @xmath175 or @xmath255 go from left to right .
this means that the corresponding boundary line of the ribbon goes from left to right , and it is easily seen that it does so at smaller values of @xmath56 than those of the geodesic ; in other words , the ribbon twists _ clockwise _ around @xmath249 by an angle of @xmath76 at each conjugate point as one follows the light geodesic in that direction ( as once it has passed the conjugate point the ribbon can be deformed back into the horizontal position ) . since we are assuming that @xmath67 is not tangent to @xmath147 at the intersections , all ribbons are horizontal at the intersections .
consequently we have proved as one follows a light geodesic connecting @xmath144 to @xmath75 in the direction of @xmath249 , each ribbon rotates clockwise around this vector by an angle of @xmath256 with respect to the vertical direction , where @xmath257 is the number of conjugate points along the light geodesic .
one of the key ingredients in the proof of the main result will be the following trivial observation : as one follows a light geodesic connecting @xmath75 to @xmath144 in the direction of @xmath258 , each ribbon rotates _ clockwise _ around this vector by an angle of @xmath256 with respect to the vertical direction , where @xmath257 is the number of conjugate points along the light geodesic .
the fact that the rotation is _ still _ clockwise is a consequence of the fact that we have reversed _ both _ the direction in which we are following the curve and the direction of the rotation axis .
notice that @xmath259 is a disk , and hence orientable and two - sided . on the other hand
, @xmath245 has a natural orientation as a deformation of @xmath21 .
we can therefore introduce in @xmath231 the orientation induced by the orientation of @xmath245 , and use it to attribute a sign to each of the intersections in @xmath260 . as is well known
( see for instance @xcite ) , @xmath261 is exactly the sum of all signs of all these intersections .
let @xmath1 be a hyper - regular static @xmath93-dimensional spacetime , and @xmath8 two causally related events not in the same null geodesic .
then @xmath26 is not the unlink .
we argue by contradiction : assume that @xmath262 is the unlink in @xmath2 ; then it is the unlink in @xmath53 , i.e. , there exists a smooth isotopy carrying @xmath244 to the boundaries of two disjoint disks , say .
consequently there exists a disk @xmath263 such that @xmath264 and @xmath265 . since @xmath266 , however , we know that @xmath267 .
we can assume that @xmath231 and @xmath263 are transverse ( since both @xmath245 and @xmath263 are compact , we just have to deform @xmath263 in an open neighbourhood of itself not intersecting an open neighbourhood of @xmath245 to get rid of any tangency points ) .
then @xmath268 is a manifold , and hence consists either of closed curves or curves joining two points in the nonempty finite set @xmath269 ( which in particular must have an even number of points ) .
take any curve of the latter kind . as a subset of @xmath263 it joins two points in the boundary , and
hence we can turn it into a closed curve @xmath270 by adding a segment of @xmath271 between these two points .
since the intersection curves can not intersect , it is possible to choose the initial curve such that @xmath270 contains exactly two intersection points in @xmath269 .
notice that the subset of @xmath263 bound by @xmath270 is a disk @xmath272 .
consider the curve @xmath273 obtained from @xmath270 by moving it by a small amount to the interior of @xmath272 .
the subset of @xmath272 bound by @xmath273 will be another disk @xmath274 , and @xmath275 .
consequently , we must have @xmath276 . on the other hand
, @xmath270 is just a curve in @xmath231 joining two intersection points at the end of two ribbons plus a segment of @xmath67 , which is a meridian .
the curve @xmath273 , in turn , is the push - off of @xmath270 away from @xmath231 along the transverse surface @xmath263 plus a segment close to the corresponding segment of @xmath67 in @xmath270 .
this segment can not wind around @xmath67 , because @xmath277 and consequently @xmath278 .
let @xmath279 be the number of conjugate points along the two light geodesics corresponding to the two ribbons . since @xmath280 , we must have @xmath281 ; consequently , the intersection points in @xmath270 must have opposite signs , and hence @xmath282 is necessarily odd .
as @xmath273 is the push - off of @xmath270 away from @xmath231 , @xmath273 must wind around @xmath270 by an angle of @xmath283 with respect to the vertical direction along the ribbons ( see figure [ picture3 ] ) .
deforming @xmath270 and @xmath273 along the meridians we then see that we must have @xmath284 , in contradiction with our previous assertion .
we conclude that @xmath262 can not be the unlink .
i would like to thank prof .
k. p. tod for all his help and encouragement ; prof .
sir roger penrose for suggesting this problem ; and dr .
marc lackenby for many useful discussions . | given a @xmath0-dimensional spacetime @xmath1 , one can consider the set @xmath2 of all its null geodesics .
if @xmath1 is globally hyperbolic then this set is naturally a smooth @xmath3-manifold .
the _ sky _ of an event @xmath4 is the subset @xmath5 and is an embedded submanifold of @xmath2 diffeomorphic to @xmath6 .
low conjectured that if @xmath7 then @xmath8 are causally related _ iff _ @xmath9 are linked ( in an appropriate sense ) .
we prove low s conjecture for a ( large ) class of static spacetimes . |
despite several decades of intense scrutiny , our understanding of active galactic nuclei ( agn ) remains incomplete .
fundamental issues such as the number density of agn , their contributions to the far - infrared and x - ray backgrounds , the role of orientation to our line of sight , the processes that form and power the nuclei , and their evolution still need to be resolved .
recent large - scale x - ray , optical , and near - infrared surveys have begun to reveal new populations of agn with number densities comparable to , or exceeding , the number density of agn derived from previous optical and radio surveys ( e.g. , * ? ? ?
in particular , the two micron all sky survey ( 2mass ; * ? ? ?
* ) is finding large numbers of agn that have been missed by traditional uv / optical search techniques .
new survey wavebands can uncover past observational biases and reveal a wider range of agn properties .
polarization is an important property that can be used to explore the physics and structure of agn , and polarimetry yields valuable information about the various classes of active nuclei .
for instance , early surveys of the most luminous known agn quasi - stellar objects ( qsos ) and quasi - stellar radio sources ( quasars ) showed that they could be divided into two classes by their optical polarization properties @xcite .
the vast majority of objects , both radio and optically selected , show low optical polarization ( @xmath6% ; @xcite ) .
about 1% of the sample exhibited higher and variable polarization @xcite , in part earning them the designation of optically violent variable quasars ( ovvs ) .
almost all of these objects are associated with compact radio sources , and their variability behavior , both in flux and in polarization , clearly signifies that optically thin synchrotron radiation dominates the uv / optical continua .
identification of the polarization mechanism(s ) in low - polarization qsos has not been as straightforward largely because of the low levels of polarization encountered . except for a handful of broad absorption - line qsos ( balqsos ; @xcite ) and qsos discovered by the _ infrared astronomical satellite _ ( _ iras _ ) @xcite , only the ovvs exhibit polarization above the traditional high - polarization threshold of @xmath7% . a good illustration of the difficulty in identifying the polarizing mechanism in most qsos is provided by the blue - excess palomar - green ( pg ) sample @xcite .
@xcite have shown that this predominantly radio - quiet sample has a mean intrinsic optical polarization of only 0.5% and the maximum observed polarization is 2.5% .
unambiguous statistical tests designed to determine the source of the polarized flux have been extremely hard to obtain , even though high quality broad - band polarimetry exists for the entire pg qso sample . this is not the case for the the lower luminosity analogs to radio - quiet qsos , the seyfert 1 galaxies .
these objects show overall polarization properties similar to the pg qsos , but their polarization distribution extends to higher values .
the high - polarization seyfert 1 `` tail '' better facilitates comparison of the polarization distribution with dust extinction and dust emission indicators , and led @xcite to conclude that the primary polarizing mechanism in seyfert 1s is the scattering of nuclear light by dust located within the narrow emission - line region ( nlr ) .
variations of this mechanism have been successful in explaining the polarization of several seyfert 2 galaxies and showing that these objects would look like seyfert 1s if viewed from sightlines that intercept the scattering material and the nlr ( e.g. , * ? ? ? * ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
_ iras _ found a small number of infrared - luminous agn which suggest the existence of a population of red , optically - obscured radio - quiet qsos @xcite .
spectropolarimetry of several of these `` hyperluminous '' infrared galaxies ( higs ) has confirmed that their tremendous ir luminosities are powered in large part by qsos hidden from our direct view , and that polarization arises from scattering of nuclear light by material with a relatively unobstructed view of the agn @xcite .
a similar situation has been found for several radio galaxies @xcite .
for these objects , light scattered by dust located up to several kiloparsecs from the nucleus reveals the rest - frame ultraviolet spectrum of a qso hidden from our direct view .
@xcite have initiated a near - infrared , high galactic latitude survey for agn using 2mass .
the survey is designed to find new red , radio - quiet objects and has , in fact , revealed a large population of previously unidentified agn .
the new sample generally consists of low - redshift ( @xmath8 ) objects with redder near - ir colors ( a criterion of the object selection ) and much higher near - ir to optical flux ratios than optically selected agn .
many of these agn can be classified as qsos in their own right since their @xmath0-band luminosities are comparable to those of optically selected qsos . higher levels of optical polarization relative to uv - excess qsos are expected from the 2mass sample if , as with the ir - selected higs , many of these objects are obscured along our line of sight .
nuclear light scattered by material ( dust and/or electrons ) distributed around the agn will make a larger contribution to the total observed flux than if , as inferred for optically selected qsos , we have a direct , unobstructed view to the bright unpolarized nucleus .
given an asymmetric distribution and/or illumination of the scattering material , high polarization can result .
the 2mass agn survey presents a large , well - defined sample of qsos that invites comparisons with agn selected by other means . in this paper
we present observations of 2mass qsos aimed at determining their broad - band optical polarization properties .
we test the conclusion reached by @xcite , based on the near - ir colors of the sample , that some of the objects in the general low-@xmath9 qso population found by 2mass are obscured along our line of sight .
in addition , the polarization distribution of the new near - ir selected qsos is compared to three largely radio - quiet qso samples ( pg , balqso , and _ iras _ higs ) for which polarization data are available .
the objects selected for observation were 89 spectroscopically - confirmed agn from @xcite .
of these , 70 objects meet the criteria that we adopt for 2mass `` red '' qsos .
that is , @xmath10 and @xmath2 .
absolute magnitudes are determined assuming @xmath11 km s@xmath12 mpc@xmath12 , @xmath13 , and @xmath14 which are used throughout this paper , and are k - corrected using the near - ir spectral indices derived from the _ jhk@xmath15 _
2mass magnitudes .
the @xmath0 luminosity criterion adopted for inclusion into the sample was chosen to encompass the range of @xmath16 shown by the pg qsos .
only three 2mass qsos in this sample have @xmath17 , reflecting the fact that the pg qso sample includes many more objects with @xmath18 than the 2mass sample .
in addition , all 2mass agn identified by @xcite have @xmath19 @xmath20 30@xmath21 .
the sample of 2mass qsos and their optical polarization measurements are listed in table 1 .
table 2 lists 18 additional agn and one unidentified object observed in the polarization survey .
most of these objects do not meet the @xmath10 criterion , though their absolute @xmath0 magnitudes place them in the range adopted for qsos .
2massi j004125.3 + 134335 and 2massi j145744.9 + 202809 ( 2m004125 and 2m145744 ; hereafter objects will be designated as 2m_hhmmss _ in the text ) do not meet the @xmath16 criterion .
mostly because of the relatively bright limiting apparent magnitudes and color selection of the 2mass agn survey , all of the objects observed in the polarimetry program have confirmed redshifts @xmath22 except 2m092145 and 2m234905 .
we have excluded the objects listed in table 2 from the analysis of the 2mass qso sample so that we can solely concentrate on the polarization properties of a well - defined low - redshift sample and allow for a more straightforward comparison of these properties with other qsos in this redshift range .
a majority of the observations made during 1999 were obtained using the two - holer polarimeter / photometer @xcite on the steward observatory ( so ) 1.5 m telescope atop mt .
lemmon , az , or the bok 2.3 m reflector on kitt peak , az .
this instrument uses a semi - achromatic half - waveplate spinning at 20.65 hz to modulate incident polarization , and a wollaston prism to direct orthogonally polarized beams to two photomultiplier tubes .
all observations for this survey were made unfiltered so that these generally optically faint objects could be observed using the 1.5 m and 2.3 m telescopes . given the typically red optical colors of the qsos and the location of the h@xmath23 line , the effective wavelength of unfiltered observations usually falls within the kron - cousins @xmath24 bandpass ( @xmath2560007000 ) .
all observations were made with the smallest available circular aperture ( 0.51 mm ) .
this subtends 43 diameter at the f/16 cassegrain focus of the 1.5 m telescope . for the f/9 bok telescope ,
a lens is inserted before the aperture to slow the beam to f/16 and the aperture translates to 29 .
the lens has poor transmission at @xmath26 and therefore unfiltered observations made at the bok telescope will have slightly redder effective wavelengths .
sky subtraction is accomplished by moving the object out of the aperture several times during an observation .
data taking and reductions are thoroughly described in @xcite and @xcite .
uncertainties in the percent linear polarization ( @xmath27 ) and polarization position angle ( @xmath28 , where @xmath29 is the polarization position angle measured in the usual astronomical sense 0@xmath21 is north and 90@xmath21 is east ) quoted in tables 1 and 2 are based on photon counting statistics .
the linear polarization measurements listed are not corrected for statistical bias .
the polarization position angle is essentially undetermined when @xmath30 ( see * ? ? ?
* ) , and in these cases no value for @xmath29 is given .
polarization position angles are calibrated using the grid of polarization standard stars measured by @xcite .
unpolarized standard stars measured by @xcite were occasionally observed to check for instrumental polarization , which was found to be @xmath31% .
observations made with two - holer are identified in the `` comments '' column of tables 1 and 2 .
observations of 2mass agn during 2000 , along with a few 1999 observations , were obtained using the ccd imaging / spectropolarimeter @xcite in its imaging mode at the bok telescope .
the field of view for imaging polarimetry at the bok telescope is @xmath32 with an image scale of @xmath331.9 pixel / arcsecond . like two - holer
, a semi - achromatic @xmath34/2-waveplate is placed in front of a wollaston prism to analyze incident linear polarization . using a plane mirror in the grating location
, the two orthogonally polarized beams are imaged onto a thinned , anti - reflection - coated 1200@xmath35800 loral ccd .
four separate reads of the ccd chip are made with the waveplate rotated through 16 positions to determine the linear stokes parameters , @xmath36 and @xmath37 , using exposure times for each waveplate position of 30200 s. all observations were filtered using a kpno `` nearly mould '' @xmath24 filter ( @xmath2560007000 ) .
aperture photometry is performed on the co - added total flux images as well as the @xmath36 and @xmath37 flux images to extract the normalized linear stokes parameters for all usable targets in the field of view .
tables 1 and 2 list the results using a circular digital aperture with a diameter of 6 pixels ( @xmath3332 ) . for consistency with surveys of other qso samples
, no attempt has been made to correct the polarization for starlight from the qso host galaxies that are apparent in many of the images ( see 3 and 4 ) .
there is no evidence for extended polarized emission , but this does not strongly constrain the size of possible scattering regions since the seeing / image quality of the imaging polarimetry over the long exposure sequences was typically 1525 fwhm .
values for @xmath38 , @xmath29 , and their uncertainties are presented in the same manner as the two - holer data . for 16 objects ,
observations were made on two or three nights to improve on an initially low signal - to - noise ratio ( s / n ) measurement .
weighted means of the multiple observations are given in table 2 , though in most cases , the average values are dominated by the higher s / n measurements resulting from the imaging polarimetry . the multiple observations
do not reveal any polarization variations larger than twice the observational uncertainty , and we use the weighted means in the subsequent analysis . for a high galactic latitude sample of objects , such as the 2mass - selected agn , polarization due to transmission through aligned dust grains within the milky way
is minimal . @xcite and @xcite
both show that observing a level of interstellar polarization ( @xmath39 ) higher than 0.6% is rare among objects with @xmath19 @xmath20 30@xmath21 .
we observed at least one star or galaxy in the fields of 44 of the target agn .
for those cases where @xmath40 , we find evidence for @xmath41% in only 10 fields .
these are summarized in table 3 .
most of these 10 measurements come from the region 120@xmath21@xmath42@xmath21 and close to the galactic latitude limit for the survey ( @xmath4330@xmath21@xmath44@xmath21 ) .
no measurement of galactic interstellar polarization with @xmath45 yields @xmath46% .
comparison of the interstellar polarization measurements with the polarimetry of the agn suggests that the polarization of only 2m032458 and 2m095504 could be solely due to dust within our own galaxy . at the same time
, the intrinsic polarization of only 2m005010 appears to be significantly suppressed by a local interstellar component that is orthogonal to the measured polarization of the object . correcting for the interstellar polarization measured in this field ,
the @xmath24-band polarization of 2m005010 is @xmath47% ( not corrected for statistical bias ) at @xmath48@xmath21@xmath49@xmath21 . because the average uncertainty in @xmath38 for the entire sample , @xmath330.5% , is similar to the highest levels of @xmath39 that are encountered ,
the observed polarization of the 2mass objects have not been adjusted for the presence of galactic interstellar polarization .
the distribution of linear polarization for the 70 qsos is shown in figure 1 .
the unfiltered and @xmath24-band measurements have been combined to construct the histogram , and lacking multi - color polarization information , no adjustments to the observed polarizations have been made for differences between the effective bandpasses of the two - holer and imaging polarimetry .
the optical broad - band polarimetry shows that the 2mass qsos are a generally highly polarized sample of agn .
more than 10% of the sample ( nine of 70 objects ) have polarization above 3% and a maximum polarization of @xmath3311% is observed .
the results can be contrasted to the unfiltered optical polarization of the pg qso sample @xcite that is shown in the lower panel of figure 1 . for the pg sample ,
we only include the objects with @xmath50 to match the redshift range for the two qso samples .
the average redshift of the 2mass sample is @xmath51 ( median = 0.24 ) , whereas @xmath52 ( median = 0.16 ) for the 75 pg qsos with @xmath50 .
inclusion of the 21 higher - redshift pg qsos does not change the obvious difference between the general polarization levels of the two samples , since the highest polarization observed in the entire pg catalog is @xmath53% for pg 1114 + 445 ( @xmath54 ) and pg 1425 + 267 ( @xmath55 ) . for the purposes of comparisons between the pg and 2mass qsos ,
the pg sample will be limited to the 75 qsos with @xmath50 and @xmath56 .
the distributions of optical polarization show a dramatic difference between optically selected qsos represented by the pg sample , and near - ir selected agn having similar @xmath0-band luminosities ( @xmath57 ) .
the high polarization of the 2mass qsos relative to uv / optically selected qsos supports the conclusion that the red near - ir colors and high near - ir to optical flux ratios arise from some degree of obscuration by dust in our line of sight .
extinction and reddening by dust bias these objects against discovery in traditional optical ( uv - excess ) agn surveys .
high polarization is expected from a reddened , partially obscured qso population either because of the transmission of agn light through aligned dust grains ( the interstellar polarization mechanism ) within the agn host galaxy , or the scattering of nuclear light by material distributed fairly close to the agn .
optical spectropolarimetric results for several highly polarized 2mass qsos support the view that scattering of nuclear light in regions near the nucleus is indeed the dominant polarizing mechanism @xcite , and not dichroic absorption by aligned dust grains . polarized light from analogous scattering regions in uv
/ optically selected qsos appears to be completely swamped by the unpolarized , direct light from the nucleus , leading to the low polarization observed for most qsos measured in past surveys ( e.g. , * ? ? ?
* ; * ? ? ?
we further explore this conclusion in the following sections .
if dust obscuration in our line of sight plays a role in producing the high optical polarization of 2mass qsos relative to presumably unobscured qsos found optically , then a correlation might be expected between polarization and color in the sense that redder , more obscured objects would tend to be highly polarized .
figures 2 and 3 show the observed polarization as a function of @xmath58 and @xmath59 color indices , respectively . the pg qsos with @xmath50
are also plotted in both figures using data from @xcite and @xcite .
it is readily apparent , especially with respect to @xmath59 , that the likelihood of detecting @xmath3% is higher for redder objects .
a non - parametric correlation test ( kendall s tau ) applied to 2mass sample confirms the trends suggested in figures 2 and 3 , and these results are summarized in table 4 .
the probability of finding a correlation this strong between polarization and @xmath58 or @xmath59 if these parameters are uncorrelated is @xmath60% .
high polarization is observed over the range of @xmath61 spanned by the 2mass sample , though most 2mass qsos with @xmath62 have @xmath6% .
the three objects with @xmath63 are also not highly polarized .
it is interesting to note that the near - ir color criterion for selection into the 2mass sample seems to be near something of a threshold for observing high polarization .
three of the nine highly polarized 2mass qsos are at the red end of the @xmath58 range spanned by the pg sample ( @xmath64 ) , although none of the 12 pg qsos that meet the 2mass color selection criterion have @xmath65% .
the @xmath59 color index is much more sensitive to the presence of dust in our line of sight to the agn than @xmath58 , and as expected , the 2mass qsos span a range of at least 5.7 magnitudes in @xmath59 .
four objects are undetected in the blue palomar digital sky survey ( dss ) .
this is in stark contrast to the low-@xmath9 pg qsos that span only about 2 magnitudes in @xmath59 .
all qsos with @xmath3% have @xmath66 with the exception of 2m151653 ( @xmath67 ; @xmath68% ) .
if the 2mass qsos are similar to pg qsos except that they are obscured by dust , high polarization objects appear in the sample when there is @xmath691.5 mag of visual extinction of the direct nuclear light above that experienced by an average pg qso .
a strong trend is observed among the 2mass qsos between optical polarization and the near - ir luminosity inferred from the apparent @xmath0 magnitudes ( figure 4 and table 4 ) .
six of the 10 most luminous objects at 2.2@xmath70 m are highly polarized , and all objects with @xmath3% have @xmath71 .
there are two probable effects that might preferentially select intrinsically luminous qsos in this sample to be highly polarized .
first , under the assumption that 2.2@xmath70 m brightness is a more accurate measure of intrinsic luminosity than optical brightness , the most luminous objects are better able to illuminate scattering regions distributed around the nuclear source .
brighter and/or more extensive scattering regions around the nucleus will generally result in more polarized , scattered light that contributes a larger fraction to the total observed flux .
the second effect is the suppression of polarization by unpolarized sources of optical light such as starlight from the host galaxy .
the relative contribution of the host galaxy to the total flux is typically higher for lower luminosity agn , and therefore , the observed polarization is lower .
the 2mass agn sample has no morphological selection criteria and inspection of the dss and frames obtained from the imaging polarimetry easily reveal host galaxies for many of the 2mass qsos . in several cases , the extended galactic emission outshines the qso in the optical .
for instance , only @xmath3350% of the @xmath24-band flux measured in a 20-diameter aperture for 2m022150 is included in a 6 aperture . by comparison , sophisticated optical observational techniques or near - ir imaging
have typically been required to detect the host galaxies of uv / optically selected qsos ( see e.g. , * ? ? ?
* ; * ? ? ?
clearly , the effect on the measured optical polarization by host galaxy starlight will tend to be greater for the 2mass sample than it is for other qso samples .
this issue is discussed further in 3.4 and 4 .
one of the hallmarks of the 2mass color - selected sample is the wide variety of agn types revealed by optical spectroscopy @xcite . the full range of agn emission - line spectral types are represented in the 70 2mass qsos , from type 1 to type 2/liner .
a majority of the objects show some evidence for broad emission lines ( type 11.9 ) with only 9 objects not showing a broad component in their balmer line profiles . we have divided the qso sample into categories using the spectral classifications determined by @xcite : type 1 ( 22 objects ; which includes type 1.2 objects ) , type 1.5 ( 19 objects ) , and type 1.82/liner / starburst ( 15 objects ) . objects that show broad h@xmath72 emission lines , but whose available spectra either do not include h@xmath73 ( `` type 1.b '' ; 3 objects ) , or have no detected h@xmath73 emission line ( `` type 1.x '' ; 10 objects )
have been combined into a fourth category ( type 1 . ?
spectral classification has been ambiguous for 2m232745 and we have included it with the type 1.x and 1.b objects .
the properties of these categories and the @xmath50 pg qsos are summarized in table 5 .
inspection of figures 2 , 3 , or 4 reveals that all categories , with the exception of type 1 . ?
, are represented among the nine highly polarized examples . over
one quarter of the 19 qsos classified as type 1.5 are highly polarized . in the case of 2m165939
, the polarized flux spectrum shows a blue continuum , broad balmer emission lines , and that the nlr features are unpolarized @xcite . although there is a broad emission - line component in the total flux spectrum ,
these polarization properties imply that the agn is partially hidden from our direct view .
moreover , the scattering material must lie further away from the continuum source than the broad emission - line region ( blr ) , but be close to , or within the nlr .
the relative lack of high polarization among type 1 objects ( two of 22 objects ) is reminiscent of the @xmath50 pg sample that is composed of type 11.5 objects ( adopting the classifications compiled by @xcite ) , where none have @xmath3% .
although 2mass detects redder type 1 objects , many that are presumably reddened by dust , direct unpolarized light from the nucleus is apparently still strong enough to swamp any scattered , polarized light sources . alternatively , as with any of the low - polarization objects , the scattered light component could also be reddened ,
keeping its contribution to total observed flux too small to yield @xmath3% .
one of the two highly polarized type 1 qsos , 2m151653 , has been shown by @xcite to be polarized in the same manner as 2m165939 .
that is , material close to or within the nlr scatters light into our line of sight from an active nucleus partially obscured from our direct view .
all of the highly polarized qsos discovered in this survey exhibit evidence for a broad - line region in their total flux spectra , albeit weakly in the cases of 2m010835 ( type 1.9 ; @xmath74 ) and 2m171559 ( type 1.8 ; @xmath75 ) .
therefore , the high polarization of the 2mass sample is not the result of the inclusion of type 2 objects with hidden continuum sources and blrs like those that exist in several highly polarized seyfert 2 nuclei ( see e.g. , * ? ? ? * ) . on the other hand , the two highly polarized narrow emission
line - dominated objects are the most luminous objects of their type in the sample .
no other type 1.82 objects have @xmath76 and all but 2m145410 , 2m163736 , and 2m230304 have @xmath77 . within the context of the unified agn scheme , type 2 objects
are thought to be more heavily obscured by dust than type 1s .
this naturally explains why the @xmath0 luminosity of the type 2 objects is generally lower than that of type 11.5 and why @xmath59 is typically larger for type 2s ( table 5 ) .
the prevalence of intermediate - type qsos among the highly polarized 2mass qsos suggests that unpolarized light sources dilute the optical polarization of objects of more extreme spectral type .
type 2 objects may exhibit low polarization because of the relatively large flux contribution from stars in the host galaxy , assuming that there is no large systematic difference in host galaxy luminosity between types , or from other unpolarized sources such as the featureless continuum ( fc2 ) that is argued to be present in the spectra of some seyfert 2 nuclei ( e.g. , * ? ? ?
the effect of the host galaxy on the polarized agn flux can be roughly estimated for the different spectral types of 2mass qsos summarized in table 5 .
given the median luminosity and colors of the type 11.5 objects , the observed @xmath24-band polarization is @xmath330.8@xmath78 , where @xmath78 is the intrinsic @xmath24-band polarization of the agn , assuming that the host galaxy is similar to the average host found for uv - excess qsos ( @xmath79 for radio - quiet qsos ; * ? ? ?
* ) . in comparison , a similar host for a typical type 1.82 2mass qso will dilute @xmath78 by about a factor of 2 . if the 2mass type 2 qsos are high - luminosity analogs to seyfert 2 nuclei , it is not surprising that high broad - band optical ( @xmath24 ) polarization is not observed in these objects .
even the archetype of seyfert 2 nuclei that harbor type 1 polarized spectra , ngc 1068 , does not show @xmath3% in the red @xcite .
much of the strong increase in polarization in the blue in seyfert 2s is attributed to the decreased dilution by starlight ( see e.g. , * ? ? ?
* ; * ? ? ?
the 2mass qso polarization measurements have an effective wavelength @xmath206000 , and the redshifts of the objects in the sample are not high enough that the polarimetry includes much flux blueward of the ca h and k break where the starlight contribution is diminished . in addition , any reasonable aperture will include a majority of the flux from the host galaxy for a typical 2mass qso , whereas much of the host galaxy starlight is avoided using similar apertures for seyfert 2 galaxies simply because of their much lower redshifts . finally , there are no highly polarized examples among the objects placed in the type 1 . ?
this category is dominated ( 10 of 14 objects ) by qsos that show no evidence for h@xmath73 emission in their confirmation spectra , though broad h@xmath72 is observed @xcite .
the objects have a median @xmath16 similar to types 1 and 1.5 qsos in the 2mass sample , but their @xmath59 color indices are more typically like the much redder type 1.82 objects .
the red colors and large balmer decrements suggest that we are not seeing much agn light at optical wavelengths and , as with the type 1.82 objects , the host galaxy dilutes the polarized flux .
the low optical polarization of the highly reddened type 1.x qsos could also be the result of dust obscuring the scattering regions .
the optical broad - band polarization results are consistent with the interpretation that the 2mass sample includes a fraction of low-@xmath9 , radio - quiet qsos that are obscured along our line of sight @xcite . in
3 the 2mass qsos were compared with the predominantly radio - quiet pg qsos in the same redshift range .
this comparison is motivated not only by the similar numbers of objects and the high s / n polarimetry that is available , but also because the two samples have similar luminosity distributions at @xmath0 .
however , it can be seen in figure 4 that the 2mass sample lacks objects as bright at @xmath0 as the most luminous pg qsos .
@xcite suggest that this is caused by selection biases in the 2mass agn sample such as the exclusion of known agn , and that the extinction necessary to achieve @xmath80 essentially rules out finding new objects as bright as the most luminous unobscured qsos
. this should not be the case in the mid- to far - ir where emission from dust near the nucleus will dominate the observed flux .
an important test of the equivalence of the 2mass and optically - selected qsos is a comparison of their spectral energy distributions ( seds ) from mid - ir to radio wavelengths where orientation effects are mitigated .
nearly half ( 33 of 70 ) of the 2mass qsos are detected at 1.4 ghz by either the nrao vla sky survey ( nvss ; * ? ? ? * ) or the vla faint images of the radio sky at twenty - centimeters ( first ) survey @xcite , and about half ( 5 out of 9 ) of the highly polarized qsos are detected .
the detection rate is about the same for the pg qsos with @xmath50 ( 38 of 75 objects ) , even though the 2mass sample by design omits known agn in radio catalogs .
no object has a 1.4 ghz radio power ( p@xmath81 ) above @xmath82 w hz@xmath12 , while the pg sample boasts 10 objects with @xmath83 ( figure 5 ) .
in addition to having a number of powerful radio sources , the radio - detected pg sample includes 14 objects with low radio luminosity ( @xmath84 ) .
eleven of these objects have @xmath18 .
the low - luminosity objects are missing from the 2mass sample primarily because of its higher average redshift .
thus , although selection biases strongly affect the 2mass sample , the 2mass qsos are roughly consistent in radio luminosity with the radio - quiet pg qsos .
there is no evidence for any relationship between radio luminosity and high optical polarization among the 2mass objects . despite the similarity in the overall 1.4 ghz detection statistics for the 2mass and pg qso samples , the type 1 2mass objects are under - represented in the radio surveys .
only three of 22 type 1 qsos are detected by first or nvss , whereas @xmath3360% of objects in the other 2mass spectral categories are detected at 1.4 ghz ( table 5 ) .
the reason behind the lack of radio sources associated with type 1 objects in the near - ir selected sample compared to the optically selected pg qsos is not readily apparent , but this discrepancy hints at possible intrinsic differences in the central engine and/or agn environment between the two samples .
comparison between the 2mass and pg qso samples in the mid infrared is difficult because of the small number of objects detected by _
iras_. only 15 of the 70 2mass qsos are detected at 60@xmath70 m in the _ iras _ faint source catalog ( fsc ) .
nineteen pg qsos with @xmath50 were detected and the slightly higher success rate is due to the pg sample containing more objects with @xmath18 that are more likely to be bright enough to be seen by _
iras_. figure 6 shows that the
_ iras_-detected 2mass qsos are generally more luminous at 60@xmath70 m for a given near - ir luminosity than the pg qsos , but this is at least partially a selection bias caused by the higher redshift of the 2mass sample .
two highly polarized 2mass qsos were detected by _
iras _ and neither of these objects show an excess of 60@xmath70 m emission above that expected from a pg qso . as with the radio surveys , the type 1 2mass qsos are rare in the sample detected at 60@xmath70 m , with only 2m171442 being represented . the remainder of the 60@xmath70 m detections summarized in table 5 are distributed among types 1.5 ( 6 of 19 objects ) , 1.x ( 6 of 10 objects ) , 1.8 ( 2m022150 ) , and 1.b ( 2m130700 ) .
the large representation of type 1.x qsos in figure 6 is striking .
these objects tend to be the most luminous qsos at 60@xmath70 m for a given @xmath16
. it may be that these systems contain more dust than other types of qsos , a difference that could account for the observed infrared emission , redder optical and near - ir colors , and large balmer decrements .
one possibility is that the excess dust has been added via a merger with another galaxy .
this would also help to explain the low polarizations observed for the type 1.x objects since a large amount of dust in the system could obscure the scattering regions .
the presence of two stellar systems would also help to dilute any polarized flux from the agn .
future high - resolution optical and near - ir imaging of the 2mass sample can explore the possible role that galaxy mergers play in qsos . the high sensitivity of the _ space infrared telescope facility _
( _ sirtf _ ) is well suited to examine the mid - ir seds of agn samples .
sirtf _ observations may resolve the issue of whether near - ir selected , red qsos are similar to optically - selected qsos in their mid- and far - ir properties as expected if orientation to our line of sight is indeed the major discriminant between agn types at shorter wavelengths , or are systematically more luminous as suggested by the few _ iras _ detections .
broad - band optical polarization surveys exist for the balqsos and _ iras _ higs that help place the new 2mass sample in context with current ideas concerning agn . in figure 7
we plot the cumulative distributions of optical polarization for these two samples along with those of the 2mass and pg qsos .
results of kolomagorov - shmirnov ( ks ) tests on the distributions are given in table 6 , where it can be seen that the observed differences between the pg qsos and the other three samples are far larger than would be expected if the samples were drawn from the same parent population .
@xcite have shown that the polarization distributions for the pg qsos and balqsos can be made similar if @xmath38 for the pg sample is multiplied by @xmath332.4 .
that is , the higher polarization observed in balqsos can be explained if the direct , unpolarized nuclear light from these objects is attenuated in our line of sight by @xmath331 mag above that experienced by a typical pg qso .
this interpretation is in keeping with evidence that low - ionization balqsos are reddened by dust @xcite and with detailed spectropolarimetry that implies a covering factor @xmath85 for the absorbing gas and material partially obscuring the nucleus ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the implications of the broad - band polarimetry are not changed by the addition of the observations of @xcite and @xcite that increases the balqso sample to 85 objects .
the cumulative distributions of the 2mass qsos and balqsos ( figure 7 ) follow each other closely for low polarization levels ( @xmath86% ) . @xcite
have shown that the larger statistical bias associated with the higher average observational uncertainty of the fainter balqsos does not explain the generally higher levels of polarization observed for balqsos relative to the pg sample for @xmath872% . because the mean observational uncertainties of the 2mass qso and balqso polarization measurements are similar , the divergence of the 2mass and pg qso distributions at low polarization levels also can not be an artifact of statistical bias in the measurements .
therefore , the 2mass qso sample is not distinguished from the pg qso sample simply by the addition of a small fraction ( @xmath331015% ) of highly polarized objects . like the balqsos , the 2mass qso sample _ as a whole _ exhibits higher polarization than the pg qsos .
the congruence of the 2mass qso and balqso distributions at low polarization levels suggests a possible relationship between these two samples .
in fact , despite a divergence at high polarization , a ks test yields a reasonable probability that the samples could be drawn from the same parent population ( p@xmath88 ) .
of the four known high - redshift ( @xmath89 ) agn found by 2mass , one shows broad mg ii absorption @xcite .
it is possible that the 2mass qso sample is made up in part of objects that are low - redshift analogs to the high-@xmath9 balqsos , but there are some effects that must be taken into account to properly compare the polarization properties of the two samples . first , the balqsos tend to be at a much higher redshift ( @xmath90 ) than the 2mass qsos ( @xmath51 ) . as a result , optical polarimetry of most balqsos measures @xmath38 in the rest - frame ultraviolet . any general wavelength dependence in the continuum polarization between the optical and near - uv will result in a systematic difference between the samples .
possibly more important is the effect of host galaxy starlight . unlike the 2mass objects , polarization measurements of balqsos are unlikely to be affected by the host galaxy because of the overwhelming brightness of the nucleus and the fact that the observations are made blueward of the ca h and k break where the flux from galaxies is diminished . correcting the polarimetry of 2mass qsos for host galaxy starlight will result in a greater differentiation between the polarization distributions of the 2mass and balqso samples .
we have attempted to measure the effect on @xmath38 of the host galaxy in several 2mass qsos using the imaging polarimetry data .
various digital apertures were tried to explore the dependence of @xmath38 on aperture size . however , with an image scale of @xmath3305 pixel@xmath12 , even the smallest practical apertures admit a large fraction of the total stellar flux , and in no case is the expected trend of decreasing @xmath38 with increasing aperture size unambiguously observed . a final difference to be noted between the polarization properties of 2mass qsos and balqsos
is presented by spectropolarimetry .
@xcite summarize the general spectropolarimetric properties found for balqsos from various studies which imply that scattering produces the observed polarization ( see also , * ? ? ?
* ) . in most cases ,
the broad emission lines are not as highly polarized as the continuum , and in several instances there is no evidence for broad emission lines in the polarized flux spectrum at all .
spectropolarimetry of highly polarized 2mass qsos , on the other hand , shows that the broad balmer lines tend to be polarized to about the same level as the continuum @xcite .
this is consistent with the view that the continuum source is more heavily obscured than the blr in balqsos , but the two are attenuated by about the same amount in the 2mass objects .
future ultraviolet spectroscopy will provide a direct test of the possible relationships between the 2mass qsos and balqsos .
determining the fraction of near - ir selected qsos that show broad absorption lines and comparing the uv properties of 2mass qsos with other samples will likely yield important information about the structure of luminous , obscured agn and about the role that orientation plays in the selection and identification of these objects .
in a similar vein , soft x - ray observations of 2mass qsos can measure the x - ray absorption in the line of sight and these values can be compared to the absorbing column densities exhibited by other agn samples ( see e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
results from the _ chandra _ x - ray observatory suggest that 2mass qsos exhibit column densities ,
@xmath91 @xmath92 @xcite .
the higs are another important qso sample @xcite that may be closely related to objects found from their near - ir colors . like the 2mass sample , the higs are composed of agn of various optical spectral types and show larger infrared - to - optical flux ratios than are observed for uv / optically selected qsos . @xcite and @xcite have found that the _ iras _ higs as a class are optically highly polarized ( see figure 7 ) .
spectropolarimetry of the most luminous higs with type 2-like spectra has revealed that these objects are powered by qsos obscured in our line of sight @xcite .
the small _ iras _ sample occupies a larger region of redshift space ( @xmath932.3 ) than do the 2mass qsos ( @xmath50 ) , and it is the high - redshift ( most luminous ) higs that tend to be the most highly polarized .
this resembles the trend between @xmath38 and near - ir luminosity found for the 2mass qsos ( 3.3 ) .
another parallel between the two samples is that objects with the highest infrared - to - optical flux ratios ( as measured by @xmath94 for the 2mass sample and @xmath95 for the higs ) tend to be the most highly polarized .
these trends in the polarization properties suggest that the higs may represent the high - luminosity tail of the dust - obscured qso population .
the type 2 _ iras _ higs are among the most luminous objects known and are highly polarized ( see e.g. , * ? ? ?
* ; * ? ? ?
* ) , whereas the lower luminosity type 2 2mass qsos do not show high optical broad - band polarization .
we ascribe this difference to the dominance of the scattered light over host galaxy starlight in the higs , and to the fact that the polarimetry of the 2mass qsos does not measure the polarization in the rest - frame uv .
the broad - band optical polarization of qsos selected by their near - ir colors are consistent with the view that many of these objects are obscured to some degree by dust . over 10% of a sample of 70
`` red '' qsos discovered by 2mass @xcite shows high polarization ( @xmath3% ) , and values range up to @xmath3311% .
these high levels of polarization are not present among objects in standard uv - excess catalogs of qsos except for ovvs and a small number of balqsos .
the polarization of 2mass qsos is correlated with near - ir luminosity and color , and with near - ir to optical flux ratio in the sense that more luminous and redder objects tend to show high polarization .
these trends suggest that the polarization arises from the scattering of nuclear light by material located close to a partially obscured agn .
spectropolarimetry of two highly polarized 2mass qsos confirms that scattering is the dominant polarizing mechanism in these objects with the scattering material located close to , or within , the nlr @xcite . determining the nature of the scattering material ( either dust and/or electrons )
, however , likely awaits careful modeling of high - quality spectropolarimetry at uv , optical , and infrared wavelengths ( see e.g. , * ? ? ?
the near - ir color selection of the 2mass agn survey is unbiased with regard to optical spectral type , and we find that high polarization is found among nearly all types , especially those classified as type 1.51.9 by @xcite .
like optically selected qsos , few of the broad emission line - dominated ( type 1 ) qsos have @xmath3% .
this is presumably because the obscuration of direct agn light is sufficiently low that the relative contribution of any scattered component is small .
high polarization is not observed in type 2 qsos in this sample , although their colors and @xmath0 luminosities are consistent with these objects being among the most heavily obscured in the sample . dilution of scattered , polarized light by starlight from the host galaxy is suggested as the reason for low observed optical polarization in this case .
this assertion is consistent with the fact that most seyfert 2 galaxies that are highly polarized in the uv show @xmath6% for @xmath96 .
ultraviolet polarimetry and high quality imaging provide obvious tests of the effect of host galaxy starlight on the polarization of 2mass qsos .
the distribution of optical polarization for the 2mass qsos further substantiates the trend that higher polarization is seen in qso samples that have higher apparent ir to optical luminosity ratios ( @xmath97 ) .
the 2mass objects are not as extreme in @xmath97 as the _ iras _ higs , and neither is their general level of polarization . on the other hand , the 2mass sample is much more highly polarized than the pg qsos , which generally have the lowest @xmath97 .
this is presumably because our view of the nuclear region is not hindered by dust in uv - excess agn .
although the 2mass and balqso distributions in @xmath38 are fairly similar , corrections for starlight and for differing observing bandpasses are likely to better differentiate these two samples and thereby increase the significance of the correlation between @xmath38 and @xmath97 .
this correlation implies that orientation effects and/or the dust covering factor are largely responsible for differences in the optical and ir properties between various samples of high - luminosity agn .
alternatively , the near - ir selected qsos may represent a different evolutionary stage than the optically selected , uv - excess qsos .
further orientation - independent investigations of these samples , such as imaging of host galaxies and environments , extended radio emission , and ir observations using _ sirtf _ will be necessary to distinguish between evolutionary and orientation effects .
finally , the polarization properties of the 2mass qsos suggests that this new near - ir selected sample may include a large proportion of low - redshift balqsos .
ultraviolet observations can directly test this possibility , and together with x - ray measurements , investigate the nature of the obscuring material .
we thank george rieke and frank low for useful discussions and the national aeronautics and space administration ( nasa ) and the jet propulsion laboratory ( jpl ) for support through _ sirtf_/mips and science working group contracts 960785 and 959969 to the university of arizona .
rmc and bon acknowledge the support of the jpl operated by the california institute of technology under contract to nasa .
polarimetric instrumentation at steward observatory is maintained , in part , through support by national science foundation ( nsf ) grant ast 9730792 .
this publication makes use of data products from the two micron all sky survey , which is a joint project of the university of massachusetts and the infrared processing and analysis center / california institute of technology , funded by nasa and the nsf .
the dss is based on photographic data of the national geographic society palomar observatory sky survey obtained using the oschin telescope on palomar mountain and was produced at the space telescope science institute under us government grant nag w-2166 .
luke moore assisted with preliminary analysis of the initial observations .
angel , j. r. p. , stockman , h. s. , woolf , n. j. , beaver , e. a. , & martin , p. g. 1976 , , 206 , l5 antonucci , r. r. j. , & miller , j. s. 1985 , , 297 , 621 bahcall , j. n. , kirkakos , s. , saxe , d. h. , & schneider , d. p. 1997
, , 479 , 642 becker , r. h. , white , r. l. , & helfand , d. j. 1995 , , 450 , 559 beichman , c. a. , soifer , b. t. , helou , g. , chester , t. j. , neugebauer , g. , gillett , f. c. , & low , f. j. 1986 , , 308 , l1 berriman , g. 1989 , , 345 , 713 berriman , g. , schmidt , g. d. , west , s. c. , & stockman , h. s. 1990 , , 74 , 869 brandt , w. n. , laor , a. , wills , b. j. 2000 , , 528 , 637 cohen , m. h. , ogle , p. m. , tran , h. d. , goodrich , r. w. , & miller , j. s. 1999 , , 118 , 1963 cohen , m. h. , ogle , p. m. , tran , h. d. , vermeulen , r. c. , miloler , j. s. , goodrich , r. w. , & martel , a. r. 1995 , , 448 , l77 condon , j. j. , cotton , w. d. , greisen , e. w. , yin , q. f. , perley , r. a. , taylor , g. b. , & broderick , j. j. 1998 , , 115 , 1693 cutri , r. m. , nelson , b. o. , kirkpatrick , j. d. , skrutskie , m. f. , & huchra , j. p. 2001 , in preparation cutri , r. m. , huchra , j. p. , low , f. j. , brown , r. l. , & vanden bout , p.
a. 1994 , , 424 , l65 di serego alighieri , s. , cimatti , a. , & fosbury , r. a. e. 1994 , , 431 , 123 gallagher , s. c. , brandt , w. n. , laor , a. , elvis , m. , mathur , s. , wills , b. j. , & iyomoto , n. 2001 , , 546 , 795 glenn , j. , schmidt , g. d. , & foltz , c. b. 1994 , , 434 , l47 goodrich , r. w. 1997 , , 474 , 606 goodrich , r. w. , & miller , j. s. 1995 , , 448 , l73 goodrich , r. w. , miller , j. s. , martel , a. , cohen , m. h. , tran , h. d. , ogle , p. m. , & vermeulen , r. c. 1996 , , 456 , l12 green , p. j. , aldcroft , t. j. , mathur , s. , wilkes , b. j. , & elvis , m. 2001 , , 558 , 109 hines , d. c. 1994 , phd dissertation , university of texas hines , d. c. , schmidt , g. d. , gordon , k. d. , smith , p. s. , wills , b. j. , allen , r. g. , & sitko , m. l. 2001 , , in press hines , d. c. , schmidt , g. d. , smith , p. s. , cutri , r. m. , & low , f. j. 1995 , , 450 , l1 hines , d. c. , schmidt , g. d. , wills , b. j. , smith , p. s. & sowinski , l. g. 1999 , , 512 , 145 hines , d. c. , & wills , b. j. 1995 , , 448 , l69 hutsemkers , d. , lamy , h. , & remy , m. 1998 , , 340 , 371 impey , c. d. , & tapia , s. 1990 , , 354 , 124 jannuzi , b. t. , elston , r. , schmidt , g. d. , smith , p. s. , & stockman , h. s. 1995 , , 454 , l111 lamy , h. , & hutsemkers , d. 2000 , , 142 , 451 low , f. j. , cutri , r. m. , kleinmann , s. g. & huchra , j. p. 1989
, , 340 , l1 low , f. j. , huchra , j. p. , kleinmann , s. c. , & cutri , r. m. 1988 , , 327 , l41 mathur , s. , elvis , m. , & singh , k. p. 1995
, , 455 , l9 mcleod , k. k. , & rieke , g. h. 1994 , , 420 , 58 miller , j. s. , & antonucci , r. r. j. 1983 , , 271 , l7 miller , j. s. , & goodrich , r. w. 1990 , , 355 , 456 moore , r. l. , schmidt , g. d. , & west , s. c. 1987 , , 314 , 176 moore , r. l. , & stockman , h. s. 1981 , , 243 , 60 moore , r. l. , & stockman , h. s. 1984 , , 279 , 465 neugebauer , g. , green , r. f. , matthews , k. , schmidt , m. , soifer , b. t. , & bennett , j. 1987 , , 63 , 615 ogle , p. m. , cohen , m. h. , miller , j. s. , tran , h. d. , goodrich , r. w. , & martel , a. r. 1999 , , 125 , 1 sanders , d. b. , & mirabel , i. f. 1996 , , 34 , 749 schmidt , g. d. , elston , r , & lupie , o. l. 1992 , , 104 , 1563 schmidt , g. d. , & hines , d. c. 1999 , , 512 , 125 schmidt , g. d. , stockman , h. s. , & smith , p. s. 1992 , , 398 , l57 schmidt , m. , & green , r. f. 1983 , , 269 , 352 sitko , m l. , schmidt , g. d. , & stein , w. a. 1985 , , 59 , 323 skrutskie , m. f. , et al .
1997 , in the impact of large scale near - ir surveys , f. garzon et al .
smith , p. s. , hall , p. b. , allen , r. g. , & sitko , m. l. 1992 , , 400 , 115 smith , p. s. , schmidt , g. d. , hines , d. c. , cutri , r. m. , nelson , b. o. 2000a , , 545 , l19 smith , p. s. , schmidt , g. d. , hines , d. c. , cutri , r. m. , nelson , b. o. 2000b , , 32 , 1456 sprayberry , d. , & foltz , c. b. 1992 , , 390 , 39 stockman , h. s. , moore , r. l. , & angel , j. r. p. 1984
, , 279 , 485 tran , h. d. 1995a , , 440 , 565 tran , h. d. 1995b , , 440 , 578 tran , h. d. 1995c , , 440 , 597 tran , h. d. , cohen , m. h. , & villar - martin , m. 2000 , , 120 , 562 veron - cetty , m. p. , & veron , p. 2001
, , 374 , 92 wardle , j. f. c. , & kronberg , p.
, , 194 , 249 wilkes , b. j. , schmidt , g. d. , cutri , r. m. , ghosh , h. , hines , d. c. , nelson , b. , & smith , p. s. 2001 , letters , in press wilkes , b. j. , schmidt , g. d. , smith , p. s. , mathur , s. , & mcleod , k. k. 1995 , , 455 , l13 wills , b. j. , & hines , d. c. 1997 , in mass ejection from active galactic nuclei , n. arav , i. shlosman , and r. j. weymann , eds .
wills , b. j. , wills , d. , evans ii . , n. j. , thompson , k. l. , berger , m. , & sitko , m. l. 1992 , , 400 , 96 young , s. , hough , j. h. , efstathiou , a. , wills , b. j. , bailey , j. a. , ward , m. j. , & axon , d. j. 1996 , , 281 , 1206 000703.6 + 155423 & 0.114 & 1.8 & 13.10 & 2.13 & 4.30 & @xmath4325.42 & 991011 & 0.99 & 0.81 & & & 2h90 + 000810.8 + 135452 & 0.185 & 2 & 14.40 & 2.10 & 5.70 & @xmath4325.29 & 000107 & 0.79 & 0.47 & 101.9 & 22.6 & + 002614.5 + 184612 & 0.319 & 1 & 14.56 & 2.15 & 3.94 & @xmath4326.57 & 991012 & 1.01 & 1.27 & & & 2h90 + 002924.5 + 242430 & 0.370 & 1 & 14.29 & 2.16 & 3.11 & @xmath4327.25 & 991012 & 1.70 & 0.88 & 89.2 & 18.5 & 2h90 + 004118.7 + 281640 & 0.194 & 1 & 12.50 & 2.04 & 3.40 & @xmath4327.29 & 990909 & 2.16 & 0.27 & 103.5 & 3.5 & 2h60 + 005010.1 + 280619 & 0.277 & s & 15.13 & 2.64 & @xmath205.87 & @xmath4325.84 & 991016 & 2.28 & 0.75 & 49.2 & 9.5 & + 005055.7 + 293328 & 0.136 & 2 & 13.22 & 2.11 & 5.68 & @xmath4325.72 & 990911 & 2.47 & 0.49 & 98.8 & 5.7 & 2h60 + & & & & & & & & & & & & + 010230.1 + 262337 & 0.194 & 1.x & 13.83 & 2.25 & 6.87 & @xmath4326.03 & 991016 & 2.11 & 0.61 & 120.5 & 8.3 & + 010607.7 + 260334 & 0.411 & 1 & 14.61 & 2.69 & @xmath206.39 & @xmath4327.58 & 991016 & 6.49 & 0.89 & 117.0 & 3.9 & + & & & & & & & 000108 & 5.79 & 0.69 & 112.2 & 3.4 & + & & & & & & & ave .
& 6.03 & 0.54 & 114.1 & 2.6 & + 010835.1 + 214818 & 0.285 & 1.9 & 13.46 & 2.75 & 6.54 & @xmath4327.64 & 990913 & 4.45 & 1.92 & 116.9 & 14.4 & 2h60 + & & & & & & & 991011 & 5.93 & 1.15 & 119.7 & 5.6 & 2h90 + & & & & & & & ave .
& 5.45 & 0.99 & 119.1 & 5.2 & + 012031.5 + 200327 & 0.087 & s & 12.47 & 3.85 & 5.63 & @xmath4325.66 & 990912 & 3.65 & 1.54 & 125.9 & 14.0 & 2h60 + & & & & & & & 991012 & 0.74 & 1.67 & & & 2h90 + & & & & & & & 000108 & 0.65 & 0.27 & 128.9 & 13.8 & + & & & & & & & ave . & 0.74 & 0.26 & 128.5 & 11.3 & + 015721.0 + 171248 & 0.213 & 1.x & 13.16 & 2.70 & 7.34 & @xmath4327.09 & 991012 & 1.17 & 2.09 & & & 2h90 + & & & & & & & 000107 & 1.50 & 0.48 & 65.5 & 9.2 & + & & & & & & & ave . & 1.44 & 0.47 & 66.6 & 9.9 & + 022150.6 + 132741 & 0.140 & 1.8 & 13.25 & 2.38 & 5.65 & @xmath4325.82 & 991011 & 1.06 & 1.72 & & & 2h90 + & & & & & & & 000107 & 0.40 & 0.25 & 12.1 & 23.7 & + & & & & & & & ave . & 0.39 & 0.24 & 13.6 & 24.0 & + 023430.6 + 243835 & 0.310 & 1.5 & 13.75 & 2.20 & @xmath207.25 & @xmath4327.33 & 991011 & 2.57 & 0.46 & 99.1 & 5.1 & 2h90 + & & & & & & & & & & & & + 024807.3 + 145957 & 0.072 & 1 & 12.65 & 2.16 & 1.85 & @xmath4324.79 & 000108 & 0.37 & 0.15 & 75.7 & 13.6 & + 032458.2 + 174849 & 0.328 & 1 & 12.82 & 2.37 & 4.18 & @xmath4328.49 & 990912 & 1.23 & 0.38 & 119.2 & 8.8 & 2h60 + 034329.6 + 132519 & 0.314 & 1 & 14.28 & 2.12 & 3.42 & @xmath4326.79 & 991012 & 1.38 & 0.76 & 0.8 & 20.1 & 2h90 + 034857.6 + 125547 & 0.210 & 1.x & 13.60 & 3.30 & 6.10 & @xmath4326.82 & 991016 & 2.23 & 0.73 & 69.5 & 9.4 & + 081652.2 + 425829 & 0.235 & 1 & 13.73 & 2.11 & 3.57 & @xmath4326.57 & 991012 & 1.13 & 0.56 & 169.3 & 17.4 & 2h90 + 082311.3 + 435318 & 0.182 & 1.5 & 12.85 & 2.08 & 3.65 & @xmath4326.80 & 991012 & 1.31 & 0.71 & 73.6 & 19.7 & 2h90 + 085045.3 + 172003 & 0.343 & 1.5 & 14.46 & 2.22 & 3.74 & @xmath4326.91 & 000106 & 2.09 & 0.23 & 100.3 & 3.2 & + & & & & & & & & & & & & + 085116.8 + 120028 & 0.370 & 1 & 14.71 & 2.26 & 3.89 & @xmath4326.89 & 000106 & 0.61 & 0.36 & 93.3 & 22.1 & + 091602.6 + 210617 & 0.422 & 1.5 & 14.99 & 2.06 & 3.32 & @xmath4326.86 & 000107 & 0.81 & 0.31 & 1.9 & 12.2 & + 091848.6 + 211717 & 0.149 & 1.5 & 12.55 & 2.25 & 5.95 & @xmath4326.65 & 000107 & 6.30 & 0.14 & 154.1 & 0.6 & + 092049.0 + 190320 & 0.156 & 1.b & 14.92 & 2.06 & 5.78 & @xmath4324.34 & 000107 & 1.15 & 0.52 & 19.9 & 15.5 & + 092151.2 + 175855 & 0.356 & 1.5 & 14.79 & 2.06 & 2.71 & @xmath4326.59 & 000107 & 0.40 & 0.32 & & & + 094636.4 + 205610 & 0.280 & 1.5 & 13.72 & 2.08 & 3.98 & @xmath4327.03 & 000107 & 0.22 & 0.24 & & & + 094927.7 + 314110 & 0.308 & 1.x & 13.28 & 2.09 & 3.22 & @xmath4327.72 & 000107 & 0.44 & 0.16 & 148.8 & 12.0 & + & & & & & & & & & & & & + 095504.5 + 170556 & 0.139 & 1 & 13.44 & 2.03 & 3.96 & @xmath4325.54 & 000108 & 0.34 & 0.18 & 87.2 & 18.6 & + 100121.1 + 215011 & 0.248 & s & 14.68 & 2.18 & 5.52 & @xmath4325.79 & 000505 & 2.12 & 0.98 & 138.0 & 15.7 & + & & & & & & & 000509 & 0.95 & 0.62 & 124.7 & 25.1 & + & & & & & & & ave . & 1.25 & 0.52 & 130.9 & 13.8 & + 101400.4 + 194614 & 0.110 & 1.5 & 12.37 & 2.01 & 4.13 & @xmath4326.05 & 000108 & 0.67 & 0.12 & 98.7 & 5.2 & + 101649.3 + 215435 & 0.257 & 1 & 13.94 & 2.03 & 3.16 & @xmath4326.56 & 000108 & 0.24 & 0.18 & & & + 102724.9 + 121920 & 0.231 & 1.5 & 13.22 & 2.06 & 5.28 & @xmath4327.02 & 000107 & 1.83 & 0.31 & 141.2 & 4.9 & + 105144.2 + 353930 & 0.158 & l & 13.54 & 2.11 & 5.06 & @xmath4325.77 & 000108 & 0.54 & 0.29 & 47.8 & 19.0 & + 125807.4 + 232921 & 0.259 & 1 & 13.45 & 2.07 & 3.85 & @xmath4327.09 & 000108 & 1.22 & 0.13 & 105.8 & 3.0 & + & & & & & & & & & & & & + 130005.3 + 163214 & 0.080 & 1.x & 11.86 & 2.20 & 5.24 & @xmath4325.84 & 000108 & 1.68 & 0.14 & 43.9 & 2.3 & + 130700.6 + 233805 & 0.275 & 1.b & 13.45 & 3.34 & 7.58 & @xmath4327.79 & 000505 & 1.58 & 0.96 & 29.1 & 23.1 & + & & & & & & & 000508 & 3.18 & 0.83 & 40.9 & 7.5 & + & & & & & & & ave . & 2.45 & 0.63 & 37.7 & 7.3 & + 132917.5 + 121340 & 0.203 & 1 & 14.12 & 2.02 & 4.58 & @xmath4325.78 & 000505 & 2.30 & 0.37 & 38.3 & 4.6 & + 134915.2 + 220032 & 0.062 & 1.5 & 12.24 & 2.21 & 3.27 & @xmath4324.87 & 000505 & 2.88 & 0.22 & 105.4 & 2.2 & + 140251.2 + 263117 & 0.187 & 1 & 12.67 & 2.11 & 3.83 & @xmath4327.05 & 000505 & 0.21 & 0.21 & & & + 145331.5 + 135358 & 0.139 & 1.x & 13.09 & 2.23 & 4.51 & @xmath4325.93 & 000505 & 0.81 & 0.29 & 73.4 & 11.4 & + 145406.6 + 195028 & 0.260 & 1.5 & 14.71 & 2.37 & 5.69 & @xmath4325.96 & 000505 & 0.92 & 0.44 & 58.3 & 16.6 & + & & & & & & & & & & & & + 145410.1 + 195648 & 0.243 & 1.9 & 14.17 & 2.23 & 5.33 & @xmath4326.27 & 000505 & 0.42 & 0.70 & & & + 145608.6 + 275008 & 0.250 & 1 & 13.38 & 2.01 & 2.92 & @xmath4327.04 & 000505 & 0.17 & 0.31 & & & + 150113.1 + 232908 & 0.258 & 1.x & 13.46 & 2.48 & 5.84 & @xmath4327.24 & 000505 & 3.07 & 0.46 & 154.1 & 4.2 & + 151621.1 + 225944 & 0.190 & 1.b & 14.12 & 2.11 & 5.28 &
@xmath4325.64 & 000508 & 1.02 & 0.28 & 150.7 & 8.0 & + 151653.2 + 190048 & 0.190 & 1 & 11.41 & 2.12 & 4.39 & @xmath4328.35 & 000508 & 9.37 & 0.08 & 103.5 & 0.3 & spol + 151901.5 + 183804 & 0.187 & 1 & 14.25 & 2.15 & 4.65 & @xmath4325.48 & 000508 & 0.67 & 0.22 & 131.7 & 10.5 & + 152151.0 + 225120 & 0.287 & 1.x & 14.30 & 2.33 & 6.20 & @xmath4326.62 & 000509 & 1.02 & 0.48 & 117.5 & 16.3 & + & & & & & & & & & & & & + 154307.7 + 193751 & 0.228 & 1.5 & 12.74 & 2.29 & 3.86 & @xmath4327.55 & 000505 & 1.33 & 0.26 & 29.6 & 5.6 & + 163700.2 + 222114 & 0.211 & 1.x & 13.59 & 2.10 & 5.41 & @xmath4326.44 & 990913 & 4.91 & 1.89 & 111.0 & 12.4 & 2h60 + & & & & & & & 000505 & 2.27 & 0.41 & 101.2 & 5.1 & + & & & & & & & ave . & 2.34 & 0.40 & 102.1 & 4.9 & + 163736.5 + 254302 & 0.277 & 1.9 & 14.17 & 2.34 & 5.23 & @xmath4326.66 & 991012 & 5.05 & 3.25 & 109.3 & 25.4 & 2h90 + & & & & & & & 000508 & 0.60 & 0.39 & 53.3 & 25.6 & + & & & & & & & 000509 & 1.34 & 0.41 & 158.5 & 8.7 & + & & & & & & & ave .
& 0.39 & 0.28 & & & + 165939.7 + 183436 & 0.170 & 1.5 & 12.91 & 2.17 & 5.29 & @xmath4326.59 & 990913 & 6.30 & 0.73 & 162.4 & 3.3 & 2h60;spol + 170003.0 + 211823 & 0.596 & 1.5 & 14.88 & 2.46 & 7.21 & @xmath4328.30 & 000508 & 11.11 & 0.80 & 109.3 & 2.0 & + 170536.6 + 210137 & 0.271 & 1.x & 14.31 & 2.48 & 5.99 & @xmath4326.52 & 991011 & 0.90 & 1.43 & & & 2h90 + 171442.7 + 260248 & 0.163 & 1 & 13.08 & 2.25 & 3.82 & @xmath4326.34 & 990911 & 0.86 & 0.33 & 64.5 & 12.1 & 2h60 + & & & & & & & & & & & & + 171559.7 + 280717 & 0.524 & 1.8 & 14.63 & 2.46 & @xmath206.37 & @xmath4328.14 & 000509 & 6.08 & 1.28 & 6.8 & 6.0 & + 222202.2 + 195231 & 0.366 & 1.5 & 13.30 & 2.88 & 6.20 & @xmath4328.60 & 991011 & 7.19 & 1.14 & 109.1 & 4.5 & 2h90 + 222221.1 + 195947 & 0.211 & 1 & 12.92 & 2.11 & 4.58 & @xmath4327.10 & 990911 & 0.95 & 0.23 & 138.9 & 7.0 & 2h60 + 222554.2 + 195837 & 0.147 & 2 & 13.49 & 2.15 & 5.31 & @xmath4325.65 & 990911 & 1.38 & 0.56 & 58.6 & 13.4 & 2h60 + 223742.6 + 145614 & 0.277 & 1 & 14.00 & 2.05 & 3.40 & @xmath4326.71 & 990913 & 1.37 & 1.07 & & & 2h60 + & & & & & & & 990914 & 2.26 & 1.97 & & & 2h60 + & & & & & & & 991012 & 0.65 & 0.87 & & & 2h90 + & & & & & & & ave . & 0.60 & 0.64 & & & + 223946.0 + 192955 & 0.194 & s & 14.67 & 2.12 & 4.23 & @xmath4325.15 & 991011 & 0.60 & 1.33 & & & 2h90 + 230304.3 + 162440 & 0.289 & 2/s & 14.67 & 2.35 & 5.73 & @xmath4326.28 & 991011 & 1.11 & 1.43 & & & 2h90 + & & & & & & & 000108 & 0.22 & 0.55 & & & + & & & & & & & ave .
& 0.24 & 0.51 & & & + & & & & & & & & & & & & + 230307.2 + 254503 & 0.331 & 1.5 & 14.50 & 2.09 & 6.21 & @xmath4326.70 & 991011 & 4.38 & 1.16 & 140.0 & 7.6 & 2h90 + 230442.4 + 270616 & 0.237 & 1.5 & 14.77 & 2.13 & 5.03 & @xmath4325.57 & 991011 & 0.25 & 1.90 & & & 2h90 + & & & & & & & 000108 & 0.29 & 0.57 & & & + & & & & & & & ave . &
0.24 & 0.55 & & & + 232250.4 + 261845 & 0.328 & 1 & 14.36 & 2.12 & 3.25 & @xmath4326.83 & 990913 & 1.68 & 1.83 & & & 2h60 + & & & & & & & 991012 & 1.82 & 1.22 & & & 2h90 + & & & & & & & ave .
& 1.76 & 1.02 & 135.9 & 25.2 & + 232745.6 + 162434 & 0.364 & ? & 14.48 & 2.41 & 5.62 & @xmath4327.16 & 991011 & 3.43 & 4.73 & & & 2h90 + & & & & & & & 991016 & 1.05 & 0.57 & 90.6 & 19.8 & + & & & & & & & ave . & 1.08 & 0.57 & 90.0 & 18.9 & + 233200.6 + 291423 & 0.301 & 1.5 & 14.67 & 2.09 & 3.63 & @xmath4326.27 & 991012 & 2.29 & 1.46 & 77.4 & 24.7 & 2h90 + 234259.3 + 134750 & 0.299 & 1.5 & 14.19 & 2.02 & 4.02 & @xmath4326.70 & 990911 & 0.51 & 0.46 & & & 2h60 + 234449.5 + 122143 & 0.199 & 1 & 12.91 & 2.07 & 4.49 & @xmath4326.95 & 990911 & 1.01 & 0.24 & 127.2 & 6.7 & 2h60 + 004117.7 + 155129 & 0.080 & l & 13.91 & 1.52 & 2.70 & @xmath4323.70 & 000107 & 0.29 & 0.22 & & & + 004125.3 + 134335 & 0.033 & 2/l & 14.32 & 1.50 & 2.88 & @xmath4321.31 & 000108 & 0.42 & 0.35 & & & + 004930.9 + 153216 & 0.240 & 1 & 14.01 & 1.83 & 3.69 & @xmath4326.24 & 991011 & 0.93 & 0.80 & & & 2h90 + 005219.9 + 170241 & 0.250 & 1.b & 14.19 & 1.90 & 3.01 & @xmath4326.19 & 991012 & 1.28 & 1.08 & & & 2h90 + 005812.8 + 160201 & 0.211 & 1.b & 13.30 & 1.92 & 3.60 & @xmath4326.66 & 990909 & 0.67 & 0.28 & 15.2 & 13.8 & 2h60 + 024026.6 + 123715 & 0.321 & 1.5 & 13.90 & 1.91 & 3.80 & @xmath4327.23 & 990909 & 0.79 & 0.70 & & & 2h60 + 024140.3 + 163322 & 0.278 & 2 & 15.06 & 1.79 & 6.64 & @xmath4325.55 & 991016 & 0.37 & 0.81 & & & 2h60 + & & & & & & & & & & & & + 030750.1 + 160931 & 0.325 & 1.8 & 14.84 & 1.72 & 2.96 & @xmath4326.12 & 990111 & 3.04 & 2.96 & & & 2h90 + & & & & & & & 000107 & 0.64 & 0.44 & & & + & & & & & & & ave . & 0.64 & 0.44 & & & + 031302.2 + 210714 & 0.094 & l & 13.84 & 1.90 & 4.05 & @xmath4324.18 & 000107 & 1.53 & 0.43 & 147.5 & 8.0 & + 092145.7 + 191812 & 1.800 & 1.b & 14.55 & 2.20 & 6.45 & @xmath4332.27 & 000108 & 3.45 & 3.19 & & & + 095430.4 + 145250 & 0.173 & 2/l & 14.47 & 1.85 & 4.53 & @xmath4324.99 & 000108 & 0.78 & 0.42 & 123.8 & 19.6 & + 135852.5 + 295413 & 0.113 & 1.b & 12.85 & 1.80 & 4.15 & @xmath4325.58 & 000505 & 4.83 & 0.24 & 18.5 & 1.4 & + 142834.0 + 213014 & 0.285 & 1 & 14.84 & 1.69 & 2.86 & @xmath4325.78 & 000505 & 0.22 & 0.36 & & & + & & & & & & & & & & & & + 145744.9 + 202809 & 0.069 & 2/l & 14.87 & 1.40 & 3.93 & @xmath4322.39 & 000505 & 0.72 & 0.68 & & & + 154443.6 + 271850 & 0.172 & 2 & 14.67 & 1.59 & 4.63 & @xmath4324.70 & 000508 & 0.44 & 0.40 & & & + 155059.3 + 212808 & 0.373 & 1 & 14.31 & 1.80 & 2.99 & @xmath4327.04 & 000508 & 0.66 & 0.24 & 83.3 & 12.7 & + 170056.0 + 243928 & 0.509 & 1.5 & 14.26 & 1.71 & 2.54 & @xmath4327.86 & 990911 & 1.56 & 0.29 & 83.4 & 5.4 & 2h60 + 225902.5 + 124646 & 0.199 & 1.x & 14.11 & 1.93 & 4.49 & @xmath4325.71 & 990913 & 1.90 & 1.18 & 132.2 & 23.7 & 2h60 + & & & & & & & 991011 & 0.39 & 1.03 & & & 2h90 + & & & & & & & ave .
& 1.03 & 0.78 & & & + 234905.2 + 183315 & & agn ? & 14.05 & 2.37 & 8.09 & & 991016 & 4.30 & 1.24 & 157.6 & 8.2 & + 004118.7 + 281640 & gsc 1744 - 2794 & 00:41:25.6 & + 28:15:43 & 0.74 & 0.08 & 117.1 & 3.2 + 005010.1 + 280619 & field galaxy & 00:50:10.5 & + 28:06:24 & 2.40 & 0.92 & 139.8 & 12.3 + 005055.7 + 293328 & gsc 1745 - 651 & 00:50:45.3 & + 29:33:36 & 0.73 & 0.06 & 114.7 & 2.3 + 023430.6 + 243835 & gsc 1771 - 1156 & 02:34:28.3 & + 24:37:04 & 1.07 & 0.26 & 114.1 & 6.9 + 030750.1 + 160931 & & 03:07:53.7 & + 16:10:09 & 0.74 & 0.20 & 143.3 & 7.6 + 031302.2 + 210714 & & 03:13:02.6 & + 21:07:44 & 1.28 & 0.50 & 115.9 & 12.5 + 032458.2 + 174849 & & 03:25:02.8 & + 17:49:28 & 0.87 & 0.25 & 117.0 & 8.1 + 034857.6 + 125547 & & 03:48:58.6 & + 12:55:36 & 1.65 & 0.12 & 53.0 & 2.0 + 095504.5 + 170556 & & 09:55:03.5 & + 17:05:55 & 1.28 & 0.53 & 84.4 & 13.7 + 234449.5 + 122143 & gsc 1173 - 752 & 23:44:43.0 & + 12:20:41 & 0.60 & 0.06 & 111.0 & 2.7 + 2mass qsos & 70 & @xmath4326.6 & 2.1 & 4.8 & 9 & 33 & 15 + & & & & & + type 1 & 22 & @xmath4326.9 & 2.1 & 3.8 & 2 & 3 & 1 + type 1.5 & 19 & @xmath4326.7 & 2.1 & 4.1 & 5 & 12 & 6 + type 1.82 & 15 & @xmath4325.8 & 2.2 & 5.6 & 2 & 9 & 1 + type 1 . ?
& 14 & @xmath4326.6 & 2.3 & 5.8 & 0 & 9 & 7 + & & & & & + + pg qsos ( @xmath50 ) & 75 & @xmath4326.7 & 1.8 & 3.1 & 0 & 38 & 19 + & & & & & + type 1 & 63 & @xmath4326.7 & 1.8 & 3.1 & 0 & 32 & 13 + type 1.5 & 12 & @xmath4325.9 & 1.7 & 3.2 & 0 & 6 & 6 + 2mass & 70 & 1.9 & 0.5 & 0.211.1 & @xmath101 + pg & 75 & 0.6 & 0.2 & 0.12.5 & + & & & & & + 2mass & 70 & 1.9 & 0.5 & 0.211.1 & @xmath102 + balqso & 85 & 1.2 & 0.4 & 0.17.6 & + & & & & & + 2mass & 70 & 1.9 & 0.5 & 0.211.1 & @xmath103 + _ iras _ higs & 19 & 4.3 & 0.4 & 0.316.4 & + & & & & & + pg & 75 & 0.6 & 0.2 & 0.12.5 & @xmath104 + balqso & 85 & 1.2 & 0.4 & 0.17.6 & + & & & & & + pg & 75 & 0.6 & 0.2 & 0.12.5 & @xmath105 + _ iras _ higs & 19 & 4.3 & 0.4 & 0.316.4 & + & & & & & + balqso & 85 & 1.2 & 0.4 & 0.17.6 & @xmath106 + _ iras _ higs & 19 & 4.3 & 0.4 & 0.316.4 & + | optical broad - band polarimetry is presented for near - infrared color - selected active galactic nuclei ( agn ) classified as quasi - stellar objects ( qsos ) based on their @xmath0-band luminosity .
more than 10% of a sample of 70 qsos discovered in the two micron all sky survey ( 2mass ) with @xmath1 and @xmath2 show high broad - band linear polarization ( @xmath3% ) , and values range to a maximum of @xmath4% .
high polarization tends to be associated with the most luminous objects at @xmath0 , and with qsos having the highest near - ir to optical flux ratios .
the 2mass qso sample includes objects possessing a wide range of optical spectral types .
high polarization is seen in two of 22 broad emission - line ( type 1 ) objects , but @xmath5 of the qsos of intermediate spectral type ( type 1.51.9 ) are highly polarized . none of the nine qsos classified as type 2 exhibit @xmath3% .
it is likely that the unavoidable inclusion of unpolarized starlight from the host galaxy within the observation aperture results in reduced polarization for the narrow emission - line objects .
the high polarization of 2mass - discovered qsos supports the conclusion inferred from their near - ir and optical colors , that the nuclei of many of these objects are obscured to some degree by dust .
correlations between optical polarization and near - ir luminosity and color imply that the dominant polarizing mechanism in the sample is scattering of agn light into our line of sight by material located in relatively unobscured regions near the nucleus .
the broad - band polarization properties of the 2mass qso sample are compared to those of other , predominantly radio - quiet , qsos and are found to be consistent with the idea that the orientation of agn to the line of sight plays a major role in determining their observed properties . |
the large experimental statistics which are now available permit femtoscopic correlations of identical particles ( see e.g. ref.@xcite ) as a function of the full three - dimensional momentum difference @xmath0 and often also of the average pair momentum @xmath1 . increasing attention has therefore been paid to the detailed description in these higher - dimensional spaces of the second - order correlation function @xmath2 with @xmath3 the density of like - sign pairs in sibling events and @xmath4 the reference pair density usually determined by a combination of event mixing and monte carlo simulation . after removing irrelevant correlations,@xcite
the correlation function can yield information on the spacetime statistical properties of particle emission as embodied in the source function @xmath5 .
the source function is obtained from the two - particle emission function , the density of particle emission points with near - equal momenta , by projection from the two four - coordinates onto the relative three - coordinate @xmath6 ( as measured in the pair rest frame of emission point densities)@xcite .
the momentum- and coordinate - space correlations of noninteracting identical particles in their centre - of - mass system are related by a fourier transform @xmath7 where @xmath8 is the correlation strength parameter and @xmath9 for noninteracting identical spin-0 and unpolarized spin-1/2 particles emitted at independent spacetime points respectively .
the @xmath8-parameter may be modified by particle impurity or a contribution of the particles emitted in some multiparticle quantum states analogous to the coherent states in quantum optics .
note that @xmath8 is eliminated by the respective normalisation in eqs .
( [ nmq])([nmx ] ) . ] . suppressing the @xmath10-dependence in a given global system , the two - particle correlation in both spaces can in this case
be written as a normalised probability density function ( pdf ) @xmath11 in @xmath12-space and a pdf @xmath13 in @xmath14-space , related by @xmath15 a gaussian @xmath11 immediately yields a gaussian @xmath13 in any dimension .
experimental data , however , is often nongaussian , sometimes strongly so .
this raises two problems : first , to systematically describe the nongaussian shape of @xmath16 ( or @xmath17 ) in momentum space , and second , to determine parameters of @xmath18 ( or @xmath19 ) in coordinate space , given only the kernel transform and measurements in @xmath12-space .
approaches towards systematic description of nongaussian shapes in @xmath12-space can be found in e.g. refs .
@xcite , while the source function @xmath18 is reconstructed by means of higher - order coefficients in @xmath14-space using imaging techniques @xcite and cartesian harmonics @xcite . in this paper
, we wish to address the second problem of a systematic description of @xmath13 in terms of given measurements in @xmath12-space , based on the fundamental statistical properties of cumulants ; the corresponding approach treating the first problem of measurements in @xmath12-space has been treated in part in the literature @xcite @xcite and will be more fully elaborated elsewhere .
while fully three - dimensional formulations have been in part set out in e.g. ref .
@xcite , we shall here work in one dimension using so that the above expressions become @xmath20 and so on , our purpose being first to test and improve the convergence properties of series expansions in a simpler environment .
given a measured normalised correlation function @xmath21 , its @xmath12-moments @xmath22 and @xmath12-cumulants @xmath23 of lowest orders @xmath24 provide fundamental information on its properties : the ordinary mean @xmath25 is a measure of the _ location _ of the peak of @xmath21 , while the variance @xmath26 measures the dispersion and @xmath27 the _ width _ or _ scale _ of the pdf , the skewness @xmath28 measures its _ asymmetry _ and the kurtosis @xmath29 is a first description of the pdf tail s _ decay rate_. higher - order `` generalised kurtoses '' @xmath30 would provide successively more detail .
kurtoses @xmath31 can also be generally viewed as cumulants of the pdf @xmath32 of the standardised variable @xmath33 .
equivalent relations hold in coordinate space between @xmath14-moments , @xmath14-cumulants and @xmath34 , e.g. @xmath35 , @xmath36 and so on .
@xmath12-moments are derivatives of the generating function @xmath37 , @xmath38 writing @xmath39 for short , while the related derivation of @xmath12-cumulants from @xmath40 fixes relations between moments and cumulants to all orders . for identical particles , both @xmath41 and @xmath13 are symmetric , so that moments and cumulants of odd order vanish and the even - order relations in both @xmath12-space and @xmath14-space become @xmath42 cumulants form a natural basis for near - gaussian expansions since a gaussian pdf is fully determined once @xmath43 and @xmath44
are known : all its @xmath45 are identically zero .
they also have important properties such as invariance under translation and a null result for uncorrelated variables .
while for purely gaussian sources , the second - order cumulants are related by @xmath46 and all higher - order cumulants are identically zero , neither of these statements is true in general .
we will therefore consider both the modification of @xmath47 resulting from nonzero @xmath31 as well as the @xmath14-kurtosis @xmath48 .
since @xmath14-moments are found from the generating function @xmath49 through @xmath50 we can through eqs .
( [ kpt])([kps ] ) obtain @xmath14-cumulants as combinations of measured @xmath12-moments .
while experimental measurement of derivatives of @xmath52 is of course impossible , the above expressions can nevertheless be evaluated since gram - charlier and edgeworth series expansions also probe regions of nonzero @xmath12 .
both expansions start with choosing a reference pdf @xmath53 which , given the close relation between cumulants and gaussians , is almost invariably chosen by textbooks @xcite to be a gaussian @xmath54 with the free parameter @xmath55 fixed to the experimentally measured @xmath56 .
the resulting `` gauss gram - charlier '' ( ggc ) series , also known as the `` gram - charlier type a '' series , and the corresponding gauss edgeworth ( gew ) series are closely related , being mere re - orderings of one another , and are therefore commonly considered to be one and the same . as we will show , however , the gew far outperforms the ggc series at any order of the partial sums .
as shown elsewhere,@xcite the ggc series results from expanding the generating function for the nongaussian @xmath32 in powers of @xmath57 @xmath58 = e^{-x'^2/2 } \sum_{m=0}^\infty \frac{c_m(\bm{\zeta})}{m ! } \,(ix')^m\ ] ] where each @xmath59 is a polynomial in the set of @xmath12-kurtoses @xmath60 . taking the inverse fourier transform of @xmath61 term by term
, one obtains an expansion in terms of chebychev - hermite polynomials @xmath62 , @xmath63 , \\ h_r(q ' ) & = & f_0^{-1}(q')\;(-d_{q'})^r f_0(q')\,,\end{aligned}\ ] ] with lowest - order terms ( writing @xmath64 for short ) @xmath65.\end{aligned}\ ] ] using @xmath66 , the @xmath67-th derivative of the @xmath14-moment generating function is , for even @xmath67 , @xmath68,\ ] ] from which the @xmath14-cumulants follow as ratios of generating functions at @xmath69 in terms of generalised @xmath12-kurtoses @xmath70 and using @xmath71 @xmath72\end{aligned}\ ] ] while the @xmath14-kurtosis in fourth order is @xmath73\end{aligned}\ ] ] with a similar expression for @xmath74 . note
firstly that @xmath75 depends only on @xmath31 but not directly on @xmath76 ; this is true also for higher - order @xmath51 .
secondly , the above relations reduce to the gaussian relation @xmath46 and @xmath77 if and when the measured correlation function is gaussian since as mentioned all @xmath78 are then identically zero . in general , however , the `` radius '' @xmath79^{1/2}$ ] of the source distribution is a function also of higher - order @xmath12-cumulants , with both increasing orders @xmath31 and increasing powers of lower - order @xmath31 entering the expansions .
given the symmetry between @xmath14 and @xmath12 , the corresponding expansions for @xmath56 and @xmath31 in terms of @xmath47 and @xmath51 would have the same form as the above , apart from some changes in sign .
any measured @xmath56 is therefore itself the result of contributions from higher - order cumulants of @xmath34 or , in physics terms , the nongaussian shape of the emission region .
statistical errors on @xmath12-cumulants rise with increasing order so that only those lower - order ones accessible to available experimental statistics can be included .
series expansions such as ( [ xqs ] ) are known to be asymptotic , so that the question arises : how accurately can a series truncated at some maximum order @xmath80 and/or a maximum power @xmath81 estimate the @xmath51 ? to quantify this issue , we make use of the normal inverse gaussian ( nig ) probability density @xcite as a solvable toy model for @xmath32 which yields exact expressions for both coordinate- and momentum - space cumulants . while the nig has four parameters @xmath82 , @xmath83 , @xmath84 and @xmath85 , in the present symmetric case @xmath86 , so that we need only the two - parameter symmetric normal inverse gaussian ( snig ) , @xmath87 where @xmath88 is the modified bessel function .
the snig reverts to a gaussian in the limit @xmath89 and has @xmath12-moment generating function @xmath90 $ ] .
experimentally measured @xmath56 and @xmath91 would fix the parameters : writing @xmath92^{1/2}$ ] and @xmath93 for short , @xmath94^{1/2}$ ] and @xmath95^{1/2}$ ] , so that higher - order cumulants and kurtoses can be expressed in terms of measured quantities @xmath96 and @xmath19 as @xmath97^{\tfrac{r}{2}-1}\ , , \\
\label{snc } \gamma_{r,\scriptscriptstyle{\rm snig}}^\qbr & = & ( r-1)!!\,(r-3)!!\ ; [ \tfrac{1}{3}\,g]^{\tfrac{r}{2}-1}.\end{aligned}\ ] ] using the snig pdf as @xmath14-moment generating function in the form ( [ phiq ] ) @xmath98 we obtain exact expressions for @xmath14-cumulants via ( [ mpx ] ) and the moment - cumulant relations . omitting the argument of the bessel functions , which is @xmath99 in every case
, these `` exact '' @xmath14-cumulants are @xmath100 with these exact @xmath14-cumulants as reference , we test the accuracy of various truncations of eqs .
( [ ktx])([kfx ] ) as a function of the gram - charlier order @xmath101 of eq .
( [ xqs ] ) .
the results are disastrous . in fig . 1
, we show respectively the percentage deviation of ggc expansions ( [ ktx ] ) and ( [ kfx ] ) , truncated at @xmath102th order , from the exact answers ( [ snd ] ) and ( [ snf ] ) , in the form @xmath103 and @xmath104 . at @xmath105 , of course , all series reduce to a gaussian and all approximations become exact .
even small values of @xmath91 lead to large deviations , however , and the size of the deviations increases with order @xmath102 .
ggc series fail completely to approximate the exact @xmath14-cumulants .
found by eq .
( [ ktx ] ) vs. ( [ snd ] ) and of @xmath75 found by ( [ kfx ] ) vs. ( [ snf ] ) , as a function of the measured @xmath12-kurtosis @xmath91 for various gram - charlier orders @xmath102 . ,
title="fig:",width=283 ] + found by eq .
( [ ktx ] ) vs. ( [ snd ] ) and of @xmath75 found by ( [ kfx ] ) vs. ( [ snf ] ) , as a function of the measured @xmath12-kurtosis @xmath91 for various gram - charlier orders @xmath102 . ,
title="fig:",width=287 ] +
in his 1946 treatise on statistics , cramr@xcite derived the gauss - edgeworth ( gew ) series by considering the random variable ( @xmath106 in our case ) to be a convolution of @xmath107 identically - distributed independent ( iid ) random variables @xmath108 each with pdf @xmath109 , a corresponding generating function @xmath110 and second - order cumulant @xmath111 , in terms of which the generating function for @xmath57 , the dual to @xmath112 , is @xmath113 = \left[\phi_1\left(\frac{x'}{\sigma_1 \sqrt{n}}\right)\right]^n\,.\ ] ] for such convolutions , cumulants of @xmath32 are related to cumulants of @xmath114 by @xmath115 , so that @xmath116 and hence @xmath61 depends on @xmath117 through @xmath118 \ = \ e^{-x'^2/2 } \exp\biggl[\ , \sum_{j=3}^{\infty } ( ix')^j\,\zeta_j(1)\,\ell^{j-2 } \biggr].\ ] ] expanding the exponential in powers of @xmath119 rather than of @xmath57 and again inverting term by term , we obtain the gauss - edgeworth series @xmath120,\end{aligned}\ ] ] again writing @xmath121 for short and with @xmath122 here understood as @xmath123 . unlike the equivalent ggc expansion of eq .
( [ xqqd ] ) , in which the order of the expansion was determined by the order of @xmath124 , a given term of order @xmath125 in the gew series is a linear combination of hermite polynomials of different order .
the relation between gram - charlier and edgeworth ordering is summarised in table 1 , with terms listed in ascending order for @xmath126 . a given term @xmath127 is characterised by the set of partition coefficients @xmath128 which are constrained to the gram - charlier and edgeworth orders by @xmath129 the re - ordering becomes important already for the second - lowest order @xmath130 .
+ [ cols="^,>,^,^ " , ] + table 1 : re - ordering of terms between gram - charlier ( gc ) and edgeworth ( ew ) series + edgeworth re - ordering of terms in the derivatives @xmath131 leads to expressions for the @xmath14-cumulants as ratios of power series and @xmath75 contain products of generating functions , terms of order higher than @xmath125 are generated .
such terms must of course be omitted in a consistent @xmath132 calculation . ] in @xmath119 . for the snig test case ,
these series simplify to @xmath133 , \\[6pt ] \label{kefx } \gamma_4^\xbr & = & \qquad \left [ \frac { \gamma_4\ell^2 + \frac{5}{4 } \gamma_4 ^ 2\ell^4 + \frac{35}{96 } \gamma_4 ^ 3\ell^6 -\frac{35}{1152 } \gamma_4 ^ 4\ell^8 + \ldots } { 1+\frac{5}{4 } \gamma_4\ell^2 + \frac{55}{96 } \gamma_4 ^ 2\ell^4 + \frac{35}{384 } \gamma_4 ^ 3\ell^6 + \frac{35}{18432 } \gamma_4 ^ 4\ell^8 + \ldots } \right]\,,\end{aligned}\ ] ] with @xmath134 .
the algebraic simplicity of the above compared to the equivalent ggc relations ( [ ktx])([kfx ] ) and the gew relation ( [ ewz ] ) is due to the fact that snig kurtoses obey @xmath135^{\tfrac{j}{2}-1}\ ] ] with constants @xmath136 fully determined by the snig pdf . while relation ( [ snigpower ] ) is of course fulfilled by all convolutions through ( [ zetaconv ] ) ,
it is true for the snig case even without convolution . in fig . 2
, we show the percentage deviations of the edgeworth - truncated approximations ( [ ketx])([kefx ] ) from their respective exact snig values as a function of the @xmath12-kurtosis @xmath91 .
the improvement in accuracy over the ggc ordering is dramatic . unlike the ggc
, the gew series also continues to improve as higher orders of @xmath126 are included . found by eq .
( [ ketx ] ) vs. ( [ snd ] ) and of @xmath75 found by ( [ kefx ] ) vs. ( [ snf ] ) , as a function of the @xmath12-kurtosis @xmath91 for various edgeworth orders @xmath126 . the dashed line at zero marks perfect agreement between the truncated expansion and the exact snig answer.,title="fig:",width=291 ] + found by eq .
( [ ketx ] ) vs. ( [ snd ] ) and of @xmath75 found by ( [ kefx ] ) vs. ( [ snf ] ) , as a function of the @xmath12-kurtosis @xmath91 for various edgeworth orders @xmath126 . the dashed line at zero marks perfect agreement between the truncated expansion and the exact snig answer.,title="fig:",width=287 ] + what structure or principle underlies the strong superiority of gew over ggc ordering ?
clearly , the expansion parameter @xmath137 must be playing a crucial role .
( [ xndv ] ) characterises the generating function @xmath61 and therefore the pdf @xmath32 as `` @xmath107-divisible '' or more prosaically as precisely that convolution @xmath138\,\prod_i f(q_i)$ ] which led to eq .
( [ xndv ] ) .
the gew method might therefore be expected to work well if @xmath107-divisibility could be established for a given experimental data set .
however , it is usually not possible to directly establish whether the @xmath32 of an experimental data set is @xmath107-divisible , and there is no obvious physics reason to believe that a momentum difference @xmath106 between two particles is the result of an underlying summation .
the reason for the success of gew relies not on physics assumptions , but on a better description of nongaussian systems . in statistics terms ,
any deviation from the gaussian is captured by the gew series in the _ rate of approach _ of higher - order cumulants to zero with increasing @xmath107 .
indeed , many proofs of the central limit theorem rely on the fact that @xmath139 immediately yields the fixed rate of convergence for any generalised kurtosis as shown in eq .
( [ zetaconv ] ) .
the success of gew ordering is therefore based on the fact that all contributing terms in a given order of @xmath117 have the same rate of convergence to the gaussian limit .
furthermore , due to the alternating sign of @xmath71 , the sum of contributions within a given @xmath132 term tends to be substantially smaller than the individual contributions ; for example the @xmath130 term for the snig test case is made up of @xmath140 and @xmath141 , adding up to @xmath142 .
it is not necessary to know the value of @xmath107 to make use of gew ordering : using ( [ zetaconv ] ) , we can merely re - absorb the @xmath125 in ( [ ewz ] ) to write the gew expansion in @xmath107-independent notation ( now with @xmath143 , the measured kurtosis ) @xmath144,\end{aligned}\ ] ] but keeping in mind that each expression within a round bracket represents a certain rate of convergence and must be included or excluded as a whole .
it is not even necessary to require @xmath107-divisibility as such : for the gew ordering to be effective we require only that @xmath21 is reasonably close to a gaussian , where `` reasonable '' is typically quantified by the errors shown in fig . 2 .
the derivation also does not rely on a particular form of @xmath109 other than requiring existence of its cumulants .
we have calculated expressions for @xmath14-cumulants of second and fourth order for the emission region in terms of measured @xmath12-cumulants . on using a nongaussian test function to quantify accuracy of expansions , we have shown that the textbook gram - charlier series is unsuitable at any level of approximation .
by contrast , the gauss - edgeworth expansion , which orders terms based on the rate of approach to a gaussian , does give results which become increasingly accurate as more terms are added .
the gew series is robust in all respects ; for example it does not require @xmath107-divisibility as such but only that @xmath21 be close enough to a gaussian in the sense that the measured @xmath12-cumulants entering truncated gew expansion should have reasonable error bars .
the present one - dimensional calculation clearly can not be applied immediately to experimental data , but is meant to show that , even on the fundamental level of expansions , there are major questions which must be addressed first . in sorting out the fundamental issue of re - ordering , the present results represent an important step towards a consistent framework for shape description .
application to experimental data will require generalisation to three dimensions using the existing 3d machinery of refs @xcite .
furthermore , sampling fluctuations of experimental cumulants will have to be taken into account . in this connection
, we also note that the gew ordering has the additional advantage of placing terms with higher powers of @xmath91 into lower orders of @xmath126 , making it unnecessary to measure higher - order kurtoses . based on slightly different arguments ,
cramr @xcite also concluded that gew was superior to ggc ordering ; this has also been verified ggc vs gew comparisons of the nongaussian pdf itself @xcite .
the gram - charlier series can therefore be considered to be inferior to the edgeworth equivalent in all respects .
note that it is not necessary to measure the correlation function @xmath21 at @xmath145 , despite the fact that @xmath14-cumulants rely formally on the generating function ( [ mpx ] ) at zero .
the @xmath12-cumulants themselves are functions of @xmath21 over the whole range of @xmath12 , while the correlation strength parameter @xmath8 cancels in the normalisation ( [ nmq ] ) .
our final comment pertains to the usual practice of obtaining information on the correlation function through fits of nongaussian parametrisations
. fits rely on an _ a priori _ choice of parametrisation , guided only by the minimisation of @xmath146 , and suffer from increasing ambiguity in higher dimensions . by relying on direct measurement of coefficients , the present method and those of refs @xcite etc
leave less room for arbitrary choices and put the uncertainty where it belongs : in the sampling fluctuations of measured experimental quantities . | lowest - order cumulants provide important information on the shape of the emission source in femtoscopy . for the simple case of noninteracting identical particles
, we show how the fourth - order source cumulant can be determined from measured cumulants in momentum space .
the textbook gram - charlier series is found to be highly inaccurate , while the related edgeworth series provides increasingly accurate estimates .
ordering of terms compatible with the central limit theorem appears to play a crucial role even for nongaussian distributions .
keywords : femtoscopy , edgeworth series , correlations , interferometry * determining source cumulants in femtoscopy with gram - charlier and edgeworth series * + h.c .
eggers , m.b .
de kock + + * j. schmiegel * + pacs nos . : 13.85.hd , 13.87.fh , 13.85.-t , 25.75.gz |
the d detector , at the tevatron , is a general purpose particle detector , composed of tracking , calorimeter , and muon systems .
the tevatron collides opposing beams of protons and anti - protons , each with an energy of 0.98 tev , the 1.96 tev center of mass energy an energy makes the tevatron the first , and still only , top quark factory .
top quarks at the tevatron are produced both singly and in pairs .
however the single top production mode has both a smaller cross section , and more importantly a large irreducible background .
when produced in pairs , both top and anti - top decay into a @xmath1 boson and @xmath2 quark , the @xmath1 boson can then decay into a lepton and neutrino or a pair of quarks , and finally the @xmath2 quarks , and any light quarks ( @xmath3,@xmath4,@xmath5 ) produced by the @xmath1 decay undergo particle showering and hadronization . since 2001
the tevatron has produced @xmath6 @xmath7 pairs , though only a fraction of those events are useful for the mass analyses .
there are two lepton+jets channels used for the mass analysis , electron+jets ( e+jets ) and muon+jets ( @xmath8+jets ) .
events in both channels are required to have a single isolated lepton with @xmath9 and @xmath10 ( @xmath11 for muons ) , to have four jets with @xmath9 and @xmath12 , and to have large missing transverse energy @xmath13 ( @xmath14 for muons ) .
the main backgrounds in the lepton+jets channels come from multijet qcd , and @xmath1+jets production .
the qcd signal is reduced by placing quality cuts on the leptons , and also by placing a cut on the azimuthal angle between the lepton and the @xmath15 .
the @xmath1+jets background is reduced by requiring that one jet be tagged using a neural network discriminant trained to distinguish b - jets from light - jets .
at the tevatron , we can measure the top mass to a high precision , using three different channels , all - jets , lepton+jets , and dileptons . in all cases
we assume that the mass of the top and anti - top are the same , such an equality is demanded by cpt ( charge , parity , time ) invariance @xcite . however , we know that studies of c and p individually led to the discovery that neither symmetry is conserved by the weak interaction , and similarly a violation of cpt would require new physics to explain .
the top mass measurement uses the matrix element method ( me ) .
the me uses probability densities determined by integrating over the cross section for both the signal ( @xmath7 ) , and the main background ( @xmath1+jets ) .
a likelihood is constructed having the form : @xmath16 \nonumber\end{aligned}\ ] ] where @xmath17 represents the lepton and jet momenta measured in the detector and @xmath18 is the fraction of @xmath7 events in the sample . the signal probability @xmath19 has the form : @xmath20 @xmath21 where @xmath22 corrects for the acceptance of the event selection , @xmath23 is the total cross section , y represents the momenta of the @xmath7 decay products , @xmath24 is the matrix element of the process @xmath25 , @xmath26 is the 6-body phase space integral , @xmath27 and @xmath28 are the momenta of the incoming partons , and @xmath29 is the transfer function which is the probability of parton level momenta @xmath30 producing the measured momenta @xmath31 .
the background probability ( @xmath32 ) has a similar form . for convenience we transform the original variables ( @xmath33,@xmath34 )
are transformed into two new variables : @xmath35 and @xmath36 .
when quoting a final value for @xmath37 , @xmath38 is integrated out , with the final likelihood having the form : @xmath39 signal samples for this analysis have been generated using the pythia monte carlo event generator @xcite .
top and anti - top masses are given values of 165 , 170 , 175 , and 180 gev , with 14 samples in total ( see fig .
[ mc_gen_diff ] ) .
the @xmath1+jets background was generated using the leading order matrix element generator alpgen @xcite , with pythia used to shower the partons generated by alpgen .
@xmath33 and @xmath34 values for simulated signal.,width=302 ] we have constructed ensembles of pseudoexperiments to test the performance of the method , and remove any bias on the measurement .
the pseudoexperiments are formed using simulated events , and contain signal and background events consistent with that expected in data .
the measurement procedure is performed for each pseudoexperiment in the same way as for data .
the average of @xmath40 is compared to @xmath41 and the small bias on @xmath40 is calibrated out ( see fig . [ dmt_calib ] ) .
similarly , we look at the rms of the pull ( @xmath42 ) , and the bias on the measured error on @xmath40 is removed ( see fig . [ dmt_calib ] ) .
@xmath43 and pull rms vs @xmath41.,width=302 ] systematic uncertainties are determined by varying the composition of the pseudoexperiments .
the largest systematic errors come from signal modelling , qcd , and jet resolution .
the total systematic error of @xmath44 is half the size of the measured statistical error @xmath45 , which is to say that this measurement is statistics limited .
table [ tab : syst_dmt ] contains a list of systematic uncertainties .
lc source & uncertainty ( gev ) + + signal & @xmath46 + pdf uncertainty & @xmath47 + background modeling & @xmath48 + heavy flavor scale factor & @xmath49 + @xmath2-fragmentation & @xmath50 + + @xmath2/light response ratio & @xmath51 + jet identification & @xmath52 + jet resolution & @xmath53 + trigger & @xmath54 + overall jet energy scale & @xmath55 + residual jet energy scale & @xmath49 + muon resolution & @xmath54 + wrong charge leptons & @xmath49 + asymmetry in @xmath56 response & @xmath57 + + mc calibration & @xmath58 + @xmath2-tagging efficiency & @xmath58 + multijet contamination & @xmath59 + signal fraction & @xmath60 + total ( in quadrature ) & @xmath61 + the measurement in data is @xcite : @xmath62 figure [ mt_data ] shows the likelihood distribution in e+jets and @xmath8+jets data .
the measurement is consistent with the standard model expectation of zero , and is the first such measurement of the mass difference between a quark and its antiquark .
@xmath37 measured in data for a. ) e+jets b. ) mu+jets.,title="fig:",width=151 ] @xmath37 measured in data for a. ) e+jets b. ) mu+jets.,title="fig:",width=151 ]
the me mass measurement uses the same method as the mass difference measurement , but with @xmath63 , and with the reconstructed jet energies adjusted by a constant factor @xmath64 , which is treated as a variable in the likelihood , and which calibrates the jet energy scale using the hadronic w mass . as in the mass difference measurement , ensembles of pseudoexperiments are constructed to test the performance of the mass measurement technique , and biases on @xmath33 , @xmath65 , @xmath64 , and @xmath66 are removed ( see fig .
[ melj_calib ] ) .
@xmath67 and pull rms vs @xmath68.,title="fig:",width=151 ] @xmath67 and pull rms vs @xmath68.,title="fig:",width=151 ] the largest systematic uncertainty come from uncertainties in the calibration of the energy of b - jets .
table [ tab : syst_melj ] lists the systematic uncertainties .
lc source & uncertainty ( gev ) + + signal modeling & @xmath59 + pdf uncertainty & @xmath69 + background modeling & @xmath60 + @xmath2-fragmentation & @xmath48 + + @xmath2/light response ratio & @xmath70 + jet identification and resolution & @xmath47 + trigger & @xmath71 + residual jet energy scale & @xmath60 + muon resolution & @xmath60 + + mc calibration & @xmath47 + @xmath2-tagging efficiency & @xmath72 + multijet contamination & @xmath69 + signal contamination & @xmath73 + signal fraction & @xmath54 + total & @xmath74 +
the measurement in data is @xcite : @xmath75 here the error coming from the in - situ jes has been added to the systematic uncertainty .
figure [ melj_data ] shows contours of the likelihood in the @xmath76 vs. @xmath64 plane
. contour plot showing @xmath33 vs @xmath64 in data .,width=302 ]
there are five dilepton channels , three of which use only well identified electrons and muons , the @xmath77 , @xmath78 , and @xmath79 channels , and two which require one well identified electron or muon matched with an isolated charged track , the e+track and @xmath8+track channels .
all five channels require lepton @xmath80 , all channels require two jets with @xmath9 .
the main backgrounds are : * events with fake leptons ( qcd , @xmath1+jets ) * @xmath81 * @xmath82 * @xmath83 the background from fake leptons is controlled with lepton quality cuts , the @xmath84 background is controlled by requiring large @xmath15 and removing events within the z peak , the @xmath85 background is controlled by requiring a large missing et significance ( @xmath86 ) where the error is calculated per event using the known resolutions of the objects in the event . the @xmath87 background is reduced in the @xmath78 channel with a cut on the sum of the leading lepton @xmath88 plus the sum of the @xmath88 of the two leading jets .
the matrix element is employed here in the same way as for the lepton+jets measurement , the difference is that there is no jes factor applied to jets , as there is no hadronic w in dilepton events . as in the l+jets mass measurement , ensembles of pseudoexperiments
are constructed to test the performance of the mass measurement technique , and biases on @xmath33 and @xmath65 are removed ( see fig .
[ mell_calib ] ) .
@xmath67 and pull rms vs @xmath68.,title="fig:",width=151 ] @xmath67 and pull rms vs @xmath68.,title="fig:",width=151 ] the largest systematic uncertainties come from the jet energy scale calibration ( both the light jet scale , and the b - jet light - jet ratio ) ( see fig .
[ mell_syst ] ) .
systematic uncertainties for dilepton matrix element mass measurement.,width=302 ] the measurement in data is : @xmath89 figure [ mell_data ] shows the likelihood curves for the @xmath90 and @xmath91 . likelihood distribution in data for runiia , and runiib.,title="fig:",width=151 ] likelihood distribution in data for runiia , and runiib.,title="fig:",width=151 ]
the two template based methods start by using the kinematic information in the event , along with a hypothesized top mass , to reconstruct one or more kinematic @xmath7 configurations consistent with the observed event . in addition
, both methods smear the lepton and jet momenta within their known detector resolutions . the neutrino weighting ( @xmath92wt ) method samples the neutrino @xmath93 distribution from expected distributions , then uses this information along with the lepton and jet momenta to solve the event .
finally , a weight is constructed which compares the @xmath15 calculated using the neutrino momenta to the measured @xmath15 : @xmath94 \exp\left[\frac{(\mety^{calc}-\mety^{obs})^2}{2(\sigma^{\rm ue})^2}\right]\ ] ] the matrix weighting method uses the measured @xmath15 and uses an analytical technique to solve directly for the kinematics@xcite@xcite .
each solution is then weighted by the formula : @xmath95 where @xmath96 and @xmath97 are parton distribution functions for the incoming partons , and @xmath98 and @xmath99 give the probability for the top to decay into a lepton with energy @xmath100 in the rest frame of the top , given by the expression : @xmath101 both methods sum the weights produced at each hypothetical top mass for some number of resolution smearings ( 100 ) , and for the different solutions and jet - lepton assignments ( see fig .
[ nuwt_weight ] ) : @xmath102 weight distribution in @xmath92wt.,width=302 ] the @xmath92wt method takes the mean and rms of this sum - of - weights distribution , and constructs 2d distributions of signal and background known as templates ( see fig .
[ nuwt_template ] , note that @xmath92wt uses two correlated techniques referred to as @xmath92wt@xmath103and @xmath92wt@xmath104 ) , while the mwt method takes the mass at which the weight curve is maximized @xmath105 , and forms 1d templates ( see fig .
[ mwt_template ] ) .
the templates are then used in a likelihood to measure the top mass .
template distribution in @xmath92wt.,width=302 ] template distribution in mwt.,width=302 ] as before , ensembles of pseudoexperiments are constructed to test the performance of the mass measurement technique , and biases on @xmath33 and @xmath65 are removed .
table [ tab : calibration ] gives the slope and offset for these calibrations , while fig .
[ nuwt_mwt_calib ] show the uncalibrated plots for the @xmath92wt and mwt .
[ cols="^,^,^,^,^,^,^,^,^,^,^,^ , > , < , < " , ] the measurement in data is @xcite : @xmath106 figure [ nuwt_mwt_data ] shows @xmath107 for the @xmath92wt and mwt methods .
negative log likelihood distribution in data for @xmath92wt and mwt.,title="fig:",width=151 ] negative log likelihood distribution in data for @xmath92wt and mwt.,title="fig:",width=151 ] negative log likelihood distribution in data for @xmath92wt and mwt.,title="fig:",width=151 ]
the d experiment has performed the first measurement of the mass difference between the top quark and its antiquark , which is the first such measurement for any quark .
in addition the d experiment has measured the top mass to a precision of @xmath108 .
figure [ d0_mass_comb ] summarizes the current mass measurements for d .
9 j. a. r. cembranos , a. rajaraman and f. takayama , europhys .
lett . * 82 * , 21001 ( 2008 ) [ arxiv : hep - ph/0609244 ] . t. sjostrand , p. eden , c. friberg , l. lonnblad , g. miu , s. mrenna and e. norrbin , comput .
commun .
* 135 * , 238 ( 2001 ) [ arxiv : hep - ph/0010017 ] .
m. l. mangano , m. moretti , f. piccinini , r. pittau and a. d. polosa , jhep * 0307 * , 001 ( 2003 ) [ arxiv : hep - ph/0206293 ] .
v. m. abazov _ et al . _
[ d0 collaboration ] , arxiv:0906.1172 [ hep - ex ] . v.
m. abazov _ et al . _
[ d0 collaboration ] , phys .
* 101 * , 182001 ( 2008 ) [ arxiv:0807.2141 [ hep - ex ] ] .
l. sonnenschein , phys .
d * 72 * , 095020 ( 2005 ) [ arxiv : hep - ph/0510100 ] . l. sonnenschein , phys . rev .
d * 73 * , 054015 ( 2006 ) [ erratum - ibid .
d * 78 * , 079902 ( 2008 ) ] [ arxiv : hep - ph/0603011 ] . v. m. abazov _ et al . _
[ d0 collaboration ] , arxiv:0904.3195 [ hep - ex ] . | i present recent results related to the measurement of the top quark mass , using @xmath0 collisions recorded with the d detector at the fermilab tevatron collider .
the results are : a direct measurement of the mass difference between top and anti - top quarks , measurement of the top quark mass in the lepton+jets channel and in the dilepton channel . |
combinations of x - ray and radio observations have revealed the large - scale interactions between radio jets from active galactic nuclei ( agn ) and the intra - cluster medium ( icm ; see e.g. @xcite ) .
x - ray surface brightness depressions , or x - ray cavities , are seen in the icm , the position of which correlates with that of the synchrotron emitting radio galaxy lobes at ghz - frequencies , e.g. ngc 1275 in perseus @xcite , cygnus a @xcite , hydra a ( @xcite , and references therein ) and several other clusters @xcite .
agn inject large amounts of energy into the icm in the form of relativistic , magnetised and collimated plasma jets .
these seem to be active for periods of some tens of myr @xcite .
the jets drive strong shocks in the icm which is consequently heated ( @xcite ; reynolds , heinz & begelman 2001 ; @xcite ) .
jet plasma and magnetic fields inflate a cavity , the `` cocoon '' , which expands and displaces the ambient medium plasma @xcite .
the material inside cocoons is responsible the observed radio synchrotron emission .
the interaction between agn jets and the icm has been modelled as a hydrodynamical process @xcite .
it is well established however that magnetic fields thread the icm .
radio polarimetry is the major source of information about cluster magnetic fields ( cmfs ) .
the agn radio emission is typically polarized @xcite and the intervening magnetised icm gives rise to faraday rotation ( @xcite , and references therein ; for a review see @xcite ) , which is characterized by the rotation measure ( rm ) .
maps of this quantity have been produced using radio sources located within the icm or behind it , as well as using cluster radio halos ( for a review see @xcite ) .
the distribution in rm images , along with the assumption that cmfs vary over a single scale , has lead to estimates of the strength of cmfs within in the cool - core clusters cygnus a , hydra a , a1795 and 3c 295 ( see @xcite , respectively ) , within in the non - cool core clusters coma , a119 , 3c129 ( see @xcite , respectively ) and also in a2634 and a400 @xcite .
the rm also suggests that cmfs located in the cores of clusters are stronger than those in their outer regions @xcite .
a correlation between the rm values and the mass deposition rates of cluster cooling flows has been detected as well @xcite .
the structure of cmfs seems to be mainly related to turbulence .
this has been inferred using the spatial variation of observed faraday rotation images and their statistical properties @xcite , in particular , the shape of the rm structure functions @xcite .
@xcite used the rm images of a119 and modelled its magnetic fields with a three - dimensional multi - scale tangled geometry .
they used spatial variations of the field characterized by a magnetic power spectrum following a power law in fourier space .
murgia et al .
found that both the mean and the standard deviation of the rm , which are used to estimate the strength of cmfs , depend on the magnetic spectral index , @xmath0 .
they conclude that @xmath12 produces the best fit to the observations in a119 . with a similar approach
@xcite reported spectral power law slope fits at the core of a2255 , and of @xmath24 at the outer parts .
also , indexes close to 11@xmath33 have been estimated in a2382 @xcite , a400 , a2634 and hydra a @xcite .
models of the evolution of cmfs suggest that the fields structure is shaped by the dynamics of the icm via magnetic flux freezing and ideal mhd which is a good approximation in clusters ( see @xcite , and references therein ) . in this context
, cmfs should follow the icm advection closely , particularly the one driven by agn jets .
turbulence in the icm is expected to be produced during cluster formation and mergers @xcite , as well as by agn outflows ( @xcite , and references therein ) .
turbulent icm flows are thus believed to amplify cmfs from weak seed fields to the observed microgauss regime by the formation of local small - scale dynamos ( see @xcite , for a numerical review ) .
characteristic depolarization gradients are often observed between the main hotspots of several powerful radio galaxies ( see e.g. @xcite ) . @xcite and @xcite have explained such features using path length differences between each of the radio lobes and the edge of the observed x - ray emitting icm that contains them .
the rm map of the fanaroff - riley class i ( fr i , @xcite ) radio source hydra a shows opposite signs in the approaching and the receding radio lobes @xcite . @xcite have described these rm fluctuations with synthetic three - dimensional faraday screens characterized by ( i ) a random distribution of magnetic fields , ( ii ) a spherically symmetric gas distribution , ( iii ) axisymmetric prolate - spheroid cavities resembling the geometry of the faraday screen projected by hydra a. these 3d differential faraday screen models are consistent with the radio source expansion framework of @xcite and also with observations of x - ray cavities .
the rm gradients in question , therefore , imply that agn jets make their way by pushing the icm gas and its magnetic fields at once .
this process was recently studied by @xcite with high resolution multi - frequency images of faraday rotation and depolarization of the radio galaxies 0206 + 35 , m 84 , 3c 270 and 3c 353 .
the latter is an extended fanaroff riley class ii ( fr ii ) radio source such as the ones that we model in this paper .
guidetti et al
. reported , for the first time , well defined rm bands , with very little small - scale structure , perpendicular to the major axis of the radio lobes of these galaxies .
also , they use a simple time - independent model of both the radio source compression and the ambient medium magnetic fields to generate the rm bands in synthetic observations .
observations of cmfs are limited by the resolution and sensitivity of telescopes .
many questions about these fields are still open ( see @xcite , for a review ) . at least for powerful
fr ii sources there seems to be little mixing of icm plasma with the synchrotron - emitting one ; the observed rm distribution is caused by plasma external to the radio source @xcite .
the interaction of expanding radio sources with the magnetised icm should happen locally to the sources .
an important question here is how significant this process is when interpreting rm observations as indicators of the properties of both the cmfs and the icm gas themselves .
the ska @xcite , evla , e - merlin , lofar and xeus ( the new generation of telescopes ) are planned to have ultra - high sensitivity and sub - arc - second resolution which will allow a better understanding about the cosmological evolution and structure of cmfs .
e.g. we will understand important details about the icm turbulence ( see e.g. @xcite ) and its heat conduction ( see e.g. * ? ? ?
* ; * ? ? ?
this is the second of two papers in which we use numerical simulations to study the evolution of magnetic fields in fanaroff - riley class ii jets and their immediate cluster surroundings . in this work
we introduce a combination of 3dmhd and synthetic rm numerical simulations .
we follow the large - scale interaction of powerful radio jets from an fr ii radio source with the magnetic fields in the core of a non - cool core cluster .
the structure , the evolution and the polarised synchrotron emission of the radio sources themselves are discussed in our earlier paper ( huarte - espinosa , krause & alexander , 2011 ; paper 1 ) . here we concentrate on the structure of the icm and the feedback from radio sources .
we only follow the evolution of the sources while jets are active .
the behaviour of both energetics and magnetic power spectra ( in fourier space ) of cmfs is explored as a function of position in the two - dimensional parameter space defined by the velocity and density of the jets .
synthetic rm maps are produced at different viewing angles to investigate the observational signatures of expanding radio sources .
this paper is organised as follows : in section [ simul ] we briefly describe the formalism of ideal mhd , the numerical methods we use and write about our implementation of the icm , the cmfs and agn jets .
the results of our simulations are presented in section [ results ] along with their interpretation and analysis from the point of view of energetics , rm statistics and magnetic power spectra in fourier space .
both the implications and applications of our work are discussed in section [ discussion ] .
there , we also examine the well - studied case of hydra a and finish by commenting about our model assumptions .
finally , in section [ conclu ] we summarise the main results of our work and present our conclusions .
to describe the dynamics of the plasma in the icm and agn radio jets , we use the system of nonlinear time - dependent hyperbolic equations of ideal compressible mhd . in three dimensions and non - dimensional conservative form ,
these are given by : @xmath4 & = & \dot{e}_j \\
\label{eq : induction } \frac{\partial { \bf b}}{\partial t } - \nabla\times ( { \bf v } \times { \bf b } ) & = & 0 , \ ] ] where @xmath5 , @xmath6 , * v * and * b * are the plasma density , thermal pressure , flow velocity and magnetic fields , respectively . in ( [ eq : energy ] ) , and is the total energy density , while @xmath7 is the ratio of specific heats . in the right hand side of ( [ eq : mass ] ) , ( [ eq : momentum ] ) and ( [ eq : energy ] ) we have source terms to implement jets via the injection of mass , @xmath8 , momentum , @xmath9 , and kinetic energy , @xmath10 ( see section [ jets ] ) , as well as a newtonian gravitational acceleration , @xmath11 , in order to keep the initial plasma in a magneto - hydrostatic equilibrium .
we solve the equations of ideal mhd in three dimensions using the numerical code flash 3.1 @xcite .
flash s new multidimensional unsplit constrained transport ( ct ) solver is employed to maintain the divergence of magnetic fields down @xcite . a diffusive hllc solver @xcite prevents spurious low pressure and density values from appearing in the grid .
we use a courant - friedrichs - lewy parameter of 0.25 and periodic boundary conditions in all the domain s faces .
these boundary conditions prevent numerical noise from polluting the magnetic spectrum in the grid ( section [ cmfs ] ) .
our computational domain is a cube with edges @xmath12 computational units ( which represent a volume of 200kpc@xmath13 ) , cartesian coordinates and a uniform grid with 200@xmath13 cells .
we present six simulations to study the effects of active agn jets on the magnetic fields in the core of a non - cool core cluster .
five simulations , which we refer to as _ jets - simulations _ , are designed to experiment with the power of the jets .
we do this by varying their velocities and densities .
these jet - simulations are the same that we present in paper 1 .
the other simulation , the _ no - jets _ model , is carried out in order to follow the evolution of the icm plasma without any perturbations throughout the same simulation timesteps as the jet - simulations .
all of these models are labelled descriptively and summarised in table 1 .
see paper 1 for a discussion about the low densities and hypersonic velocities of the jets .
.simulations and parameters .
[ cols="<,^,^,^,^,^ " , ] [ table2 ] in order to investigate the effects that jets have on the structure of cmfs , we calculate three - dimensional magnetic power spectra in fourier space using the output from our simulations . specifically , we use the magnetic fields of the five jets - simulations which were produced at @xmath14 ( see table 1 ) .
the spectra are shown in figure [ pow - jets ] .
for this computation we exclude the volume that is occupied by the sources ( we use the spatial distribution of the jets tracer , section [ cocoon ] ) . in general we find that the spectra preserve the initial condition of a power - law with a kolmogorov index ( section [ cmfs ] ) .
yet the energy injection from the jets produces power enhancements on scales that range from the resolution limit , 1kpc , to the source size , @xmath1540kpc .
an interesting feature that we see is a flattening on the power spectra at the scale defined by the cocoon size figure [ pow - jets ] .
this effect depends on the power of the jets .
as radio sources expand , the energy of the jets is transferred to the icm gas and cmfs in the shocked ambient region . there , magnetic field lines are compressed and stretched tangentially to the surface area of the radio cocoons ( section [ visit ] ) .
hence , we find that fatter or more spherical cocoons , such as the ones produced by the `` lighter '' jets , are more efficient in increasing and redistributing the energy of the cmfs . the scales at which the jets affect the cmfs are more evident in figure [ pow - ratio ] .
there we plot the ratio of the power spectra in figure [ pow - jets ] over the corresponding spectra computed using the `` no - source '' simulation .
i.e. for each plot in figure [ pow - ratio ] , we use the corresponding profile in figure [ pow - jets ] as the numerator .
the spectrum in the denominator is calculated in the following way : ( i ) we combine the magnetic fields of the `` no - source '' run at @xmath16 with the cocoon spatial distribution of the corresponding jets - simulation at @xmath16 .
( ii ) the magnetic fields inside the cocoon are removed .
this yields a cube filled with magnetic fields , which have not been affected by the jets at all , and an empty cavity in the middle .
the geometry of such cavity exactly matches the shape and orientation of the cocoon s contact surface .
( iii ) we compute the 3d magnetic power spectrum of these fields .
we note the numerical diffusion of the simulations is cancelled in the computation of the profiles shown in figure [ pow - ratio ] .
gradients in these spectral ratio profiles persistently show maxima close to , as well as a roughly constant and progressive increase from scales close to 10kpc down see figure [ pow - ratio ] .
these features are related to the source energy and momentum injection scales and also to the thickness of the ambient medium swept - up region .
this gas component expands as the sources evolve towards a thermal pressure equilibrium state with the icm gas ( figure [ pressevo ] ) , and reaches scales of about 10kpc at the end of the simulations .
we see that the expansion and size of our model cocoons are clearly linked to the enhancement of the cmf power spectra at large - scales .
the cocoons which are formed by the lightest , fast jets are bigger and have more isotropic gas and energy distributions ( see section [ energetics ] ) .
figure [ figtopo ] illustrates the deformation of the cmfs topology that is caused by the `` lighter - fast '' source at @xmath1790@xmath18 and @xmath195.3myr .
the cocoon is marked by dots and it clearly stretches and compresses the neighbouring magnetic field lines , particularly ahead of the jets working surface .
+ .2 cm .2 cm @xcite have demonstrated that there is a correlation between the statistical properties of the rm distribution , in the real space , and the structure of magnetic fields in the three - dimensional frequency space .
based on these facts we compute the rm structure functions of those of our synthetic rm maps that correspond to @xmath16 and @xmath1745@xmath18 .
the structure function is given by @xmath20 ^ 2,\ ] ] where the sum is taken over the @xmath21 non - zero rm pixels that are inside bins ( annuli ) centred at @xmath22 and have radii of @xmath23 @xcite .
figure [ struct ] shows the rm structure functions in question . at the end of the simulations ,
the structure functions generally still show the power law that we have used as the initial conditions of the simulations ( section [ cmfs ] ) . yet these curves considerably flatten at scales of order tens of kpc .
the structure functions rise again from a certain scale on .
we see that this scale is larger for that sources with fatter cocoons .
the flat regions therefore seem to be directly related to the cocoon width .
this is consistent with our findings in the 3d magnetic power spectra at the frequency space figure [ pow - jets ] .
a combination of 3d - mhd numerical simulations and synthetic rm observations have been produced and presented .
these simulations follow the evolution and observational signatures of cmfs under the effects of powerful jets from an agn in the core of a non - cool core cluster . to understand the basic physics of this interaction , the effects from other cluster galaxies , synchrotron cooling , magnetic reconnections and observational factors ( such as beaming , depolarization and light travel effects )
have not been considered in the model .
we have focussed on the role that the jet velocity , the jet density and the observational viewing angle play on the jets - cmfs interaction .
plasma energetics , rm statistics and magnetic power spectra consistently indicate that cmfs are affected by agn jets .
the effect depends on the jets properties ; the lighter and the faster the jets , the more important the results they have on the ambient medium .
the shape of the cmf energetics profiles ( figure [ fig : energy ] ) is consistent with the following arguments .
radio sources expand due to the energy injection from jets .
the cocoons displace the icm gas and magnetic fields from the cavity , compressing them into the region located between the contact surface and the bow shock ( figures [ colormaps ] and [ figtopo ] ) .
these cmf lines are also stretched in this process .
the magnetic component tangential to the surface of the radio cocoons is therefore enhanced .
as the sources expand , the cocoon pressure decays and the compression caused by bow shocks weakens ( figure [ pressevo ] ) .
this is in good agreement with radio source evolution models ( e.g. see @xcite ) .
once the sources with mach 40 jets ( green and dark - blue curves , figure [ pressevo ] ) have expanded for about 100kpc , their cocoon thermal pressure is comparable with that in the environment .
thus only these relatively slow sources are not powerful enough to significantly affect the magnetic energy of the icm .
our rm integration neglects the contribution from icm gas and magnetic fields beyond the computational domain .
jet - produced rm enhancements are local , important at the edge of the sources and should not depend on this outer rm component .
observations show that both the cluster ion density and the strength of the cmfs decreases away from the cluster core ( @xcite ; @xcite , and references therein ) .
thus we expect to capture the major contribution to the rm by our procedure .
since the region we use for the analysis is spherical , any additional contribution to the rm is expected to be statistically spatially uniform .
[ fig : hydra ] @xcite have recently reported bands in the observed rm images of the fr ii radio sources 3c353 and cygnus a , as well as in the fri sources 0206 + 35 , m84 and 3c270 .
these bands are perpendicular to the major axis of the radio lobes and are characterized by very little small - scale rm structure . to explain the nature of such features , guidetti et al .
present synthetic rm observations generated with a simple model of gas with constant density and magnetic field distributions .
this plasma is then affected by compression induced by an elliptical cavity that is artificially placed inside the gas .
the model allows these authors to conclude that there are three magnetic field components associated with observed rm : an isotropic field which has large - scale fluctuations , plausibly associated with the undisturbed icm ; a well - ordered field draped around the leading edges of the radio lobes ; a field with small - scale fluctuations which is located in the shells of compressed gas surrounding the inner lobes .
we do not use an ordered magnetic field component in our model and do not observe rm bands in our synthetic images .
thus we confirm the findings of guidetti et al .
, namely that without any ordered field component in the icm , the radio source expansion will not lead to banded rm structures . in addition , our model shows that source - induced rm variations can be as large as 70% , are correlated with the radio jet properties , depend on time and are stronger at the edge of the cocoons .
this is a very important point because the observed rm signal is used to infer the strength of the cmfs ( e.g. see * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
.2 cm + .5 cm .2
cm [ lobes ] our synthetic rm maps persistently show pronounced gradients at regions corresponding to the receding radio lobes ( the left ones ) relative to the approaching ones ( figures [ rm - lighter - fast90][rm - light - fast45 ] ) .
such rm differences have been observed in the fr iis cygnys a @xcite and 3c353 @xcite as well as in several fr is ( e.g. 3c31 and hydraa , @xcite ; 3c449 , @xcite ; 0206 + 53 , 3c270 and m84 , @xcite ) .
although we simulate rm enhancements caused by fr ii radio sources , fr is which are known to form icm x - ray cavities as well may well produce similar effects .
hydraa is a well studied central cluster radio source .
the faraday rotation in this case is thought to be dominated by the local icm contribution .
this cluster presents an icm magnetic spectral index of which has been inferred against the emission from the embedded luminous fr i radio source hydraa .
the inclination angle of this source has been inferred to be @xcite . in terms of rm statistical fluctuations
that are asymmetrical between the source s lobes , we find that our `` lighter - faster '' simulation ( table 1 ) agrees to a certain degree with the studies of @xcite and @xcite on hydra a. this is partly unexpected , as we have chosen a particular realisation for the magnetic power spectrum of the cmfs .
however , we focus here on the edge effects , which is where we believe that radio sources have the strongest effect : two panels are shown in figure [ fig : hydra ] where we present curves of @xmath24 ( left panel ) and @xmath25 ( right panel ) as a function of position along a horizontal line between the lobes .
the mean and standard deviation of the rm are taken on vertical slices of the map corresponding to the `` lighter - faster '' simulation at @xmath1745@xmath18 and @xmath194.7myr .
vertical error bars are computed by taking the difference between the red and the black curves in figure [ rm - lighter - relativistic45 ] .
both statistical measures have large gradients near the left edge .
this effect is produced by the expanding radio source . towards the right edge
, the radio source expansion has also strongly amplified the rm values .
however , the initial magnetic field in that region happened to be small , which is why this region does not look spectacular .
this is quite comparable to the rm distribution in observed radio sources , e.g. see ( * ? ? ?
* figure 22 , panels _ a _ and _ b _ ) who present the rm observations of hydra a made by @xcite along with synthetic rm observation .
it is therefore well possible that such large rm - gradients towards the edges of radio sources are strongly influenced by the lobe expansion .
note that we consider here only one realisation for the random faraday screen .
a detailed match of the considered rm gradient with any observation is therefore not intended .
our model jets inject energy at scales that range from the resolution limit to the source size , but preferentially near the latter . as we have seen , this results in a flattening in both the magnetic power spectra and the rm structure functions at large - scales .
@xcite , @xcite and @xcite have reported such flat gradients in the observed structure functions of the fr i sources 3c 31 , 3c270 , 3c 449 and hydraa .
our simulations show that the physics of radio source expansion may explain these structure function gradients naturally .
thus conclusions about the nature of the icm turbulence may not be drawn from the gradients in question .
we have calculated the mean and the standard deviation of the rm for the approaching ( @xmath24@xmath26 and @xmath27 ) and the receding ( @xmath24@xmath28 and @xmath29 ) source lobes , as a function of the viewing angle . in the top left panel of figure [ lobes ]
we show @xmath24@xmath26 in solid lines and @xmath24@xmath28 in dashed lines . in all of these plots ,
positive ( negative ) angles correspond to integrations along ( against ) the line of sight , and again jets are on the plane of the sky at . in the bottom left panel of figure [ lobes ]
we show @xmath30@xmath27 .
in addition , to demonstrate the effect of the jets on the faraday screens , which will in turn affect the rm computations , we produce analogous ( @xmath24@xmath28 , @xmath24@xmath26 and @xmath30@xmath27 ) plots but use the data from the no - source rm maps ( section [ rmmapsb ] ) .
we show these no - source rm plots in the right column of figure [ lobes ] , side by side to their source counterparts .
together , all of these profiles show that the general dependence of the lobe rm statistics on the viewing angle is shaped by the intrinsic nature of the faraday screens , as it is generally assumed in the interpretation of the laing - garrington effect .
the result of the radio source expansion is to amplify several rm features . only for positive inclinations our realisation results in stronger average rms in the receding lobes ( figure [ lobes ] , top panel ) .
we see strong amplification at inclinations of @xmath31 .
@xmath32 is linked to the depolarisation @xcite .
already intrinsically , the depolarisation is always higher for the receding lobe ( @xmath30@xmath331 ) .
panels in the bottom row ( figure [ lobes ] ) show that this trend is mostly amplified by the radio sources , apart from the fatter ( lighter ) ones for which the @xmath34-ratio drops below unity near @xmath3590@xmath36 .
thus the laing - garrington effect is only moderately affected by the radio source expansion , in such a way that the associated trends tend to be amplified .
we have shown that powerful radio sources affect the structure of cmfs which is characterized by the magnetic power spectrum index @xmath0 .
this process is important because the icm energy transformation , either from or to magnetic form , as well as the observations @xcite , depend on @xmath0 .
hence we agree with @xcite in that the structure of cmfs inferred by using the emission of cluster radio sources should take a piecewise defined form over the range of the spatial frequencies .
our simulations show , moreover , that such functional forms change with time ( see figure [ pow]a , b ; note that the profiles in panel _ b _ would be completely flat otherwise ) .
faraday rotation measurements from clusters with broad differences in age will be key in the determination of evolving magnetic structure functions of the cmfs .
we have presented a combination of 3d - mhd numerical simulations and synthetic rm observations to model the evolution and observational signatures of cluster magnetic fields under the effects of powerful agn jets at the core of a non - cool core galaxy cluster .
we prescribe the cluster magnetic field as a gaussian random field with a power law energy spectrum tuned to resemble the expectations for turbulence .
we focus on how both the jets power , in terms of their density and velocity , and the observational angle affect the observed rm signal .
such signal is used by observational studies to infer the strength of cluster magnetic fields .
our results are very consistent and show the following .
powerful jets increase the icm energy in proportion to their velocity and in inverse proportion to their density .
light jets are more efficient in this process because they inflate fatter cocoons than relatively heaver jets .
only jets with mach numbers of 80 and 130 , which are over - pressured with respect to the ambient medium , significantly increase the icm magnetic energy .
we see that the time evolution of the cmf energetics is correlated with that of the cocoon pressure .
jets compress and displace the magnetised icm .
this results in a flattening of the 3d magnetic power spectra and the associated field rm structure functions .
this happens at scales from the resolution limit to the source size , but preferentially near the latter .
we see the effect is correlated with the jet velocity and pronounced for fat cocoons .
the rm structure functions of 3c 31 , 3c270 , 3c 449 and hydraa show flat gradients at large scales , and the cause of such shapes has been previously modeled @xcite in terms of icm turbulence
. our model suggests , however , that the physics of radio source expansion may as well explain this feature .
the general distribution of the faraday depth is determined by the intrinsic nature of the icm .
powerful jets enhance the rm locally by factors up to @xmath151.73 .
we find this process is proportional to the jet velocity and in inversely proportional to the beam density .
expanding hypersonic sources compress the icm gas and magnetic fields in the shocked ambient region . the strength of the fields that are perpendicular to the local lobe expansion direction increases , specially near the hotspots .
the alignment of compressed cmf components and the line of sight is favoured ( i ) at the cocoon edge ( ii ) by sources with a fat cocoon ( iii ) at @xmath37 within 2070 degrees , due to the cocoon geometry .
jet - produced rm enhancements are thus more important under these conditions and should be more evident in radio galaxies than in quasars .
both @xmath24 and @xmath38 continuously change and tend to increase as the radio sources develop .
`` lighter '' sources with jet velocities @xmath3940mach yield @xmath40 and @xmath38 enhancement of factors within 1.2 to 1.7 , in proportion to the jet velocity .
these rm enhancements may therefore lead to overestimations of the cmf strength by up to about 70% .
it is a pleasure to thank robert laing and malcolm longair for useful discussions about this paper , as well as the anonymous referee for comments that helped to improve this work .
the software used in these investigations was in part developed by the doe - supported asc / alliance center for astrophysical thermonuclear flashes at the university of chicago .
mhe acknowledges : financial support from conacyt ( the mexican national council of science and technology , 196898/217314 ) ; useful discussions with david titterington and jongsoo kim ; dongwook lee for the 3d - usm - mhd solver of flash3.1 ; volker gaibler for the routines to map magnetic field geometry . | a combination of three - dimensional ( 3d ) magnetohydrodynamics ( mhd ) and synthetic numerical simulations are presented to follow the evolution of a randomly magnetised plasma that models the intra - cluster medium ( icm ) , under the isolated effects of powerful , light , hypersonic and bipolar fanaroff - riley class ii ( fr ii ) jets .
we prescribe the cluster magnetic field ( cmf ) as a gaussian random field with a kolmogorov - like energy spectrum .
both the power of the jets and the viewing angle that is used for the synthetic rotation measure ( rm ) observations are investigated .
we find the model radio sources introduce and amplify fluctuations on the rm statistical properties which we analyse as a function of time as well as the viewing angle .
the average rm and the rm standard deviation are increased by the action of the jets .
energetics , rm statistics and magnetic power spectral analysis consistently show that the effects also correlate with the jets power , and that the lightest , fastest jets produce the strongest changes in their environment .
we see jets distort and amplify the cmfs especially near the edges of the lobes and the jets heads .
this process leads to a flattening of the rm structure functions at scales comparable to the source size .
the edge features we find are similar to ones observed in hydra a. the results show that jet - produced rm enhancements are more apparent in quasars than in radio galaxies .
globally , jets tend to enhance the rm standard deviation which may lead to overestimations of the cmfs strength by about 70% .
this study means to serve as a pathfinder for the ska , evla and lofar to follow the evolution of cosmic magnetic fields .
[ firstpage ] galaxies : jets galaxies : active intergalactic medium methods : numerical mhd turbulence |
in this paper we continue the investigation of the relationship between the geometry of a hyperbolic domain @xmath2 in @xmath3 and the geometry of the boundary @xmath1 of the convex core of its complement .
the surface @xmath1 is hyperbolic in its intrinsic metric ( @xcite ) and the nearest point retraction @xmath4 is a conformally natural homotopy equivalence .
marden and markovic @xcite showed that , if @xmath2 is uniformly perfect , the nearest point retraction is a quasi - isometry with respect to the poincar metric on @xmath2 and the quasi - isometry constants depend only on the lower bound for the injectivity radius of @xmath2 in the poincar metric . in this paper , we show that the nearest point retraction is a quasi - isometry with respect to the thurston metric for _ any _ hyperbolic domain @xmath2 with quasi - isometry constants which do not depend on the domain .
we recover , as corollaries , many of the previous results obtained on the relationship between a domain @xmath2 and @xmath1 and obtain explicit constants for the first time in several cases .
we also see that the nearest point retraction is lipschitz if and only if the domain is uniformly perfect , which was originally conjectured by marden and markovic @xcite .
we will recall the definition of the thurston metric in section [ metrics ] .
it is 2-bilipschitz ( see kulkarni - pinkall @xcite ) to the perhaps more familiar quasihyperbolic metric originally defined by gehring and palka @xcite . recall that if @xmath5 and @xmath6 denotes the euclidean distance from @xmath7 to @xmath8 , then the quasihyperbolic metric is simply @xmath9 .
the thurston metric has the advantage of being conformally natural , i.e. invariant under conformal automorphisms of @xmath2 .
[ projective ] let @xmath0 be a hyperbolic domain .
then the nearest point retraction is 1-lipschitz and a @xmath10-quasi - isometry with respect to the thurston metric @xmath11 on @xmath2 and the intrinsic hyperbolic metric on @xmath1 , where @xmath12 and @xmath13 . in particular ,
@xmath14 furthermore , if @xmath2 is simply connected , then @xmath15 where @xmath16 and @xmath17 .
marden and markovic s @xcite proof that the nearest point retraction is a quasi - isometry in the poincar metric if @xmath2 is uniformly perfect does not yield explicit bounds , as it makes crucial use of a compactness argument . combining theorem [ projective ] with work of beardon and pommerenke @xcite , which explains the relationship between the poincar metric and the quasihyperbolic metric
, we obtain explicit bounds on the quasi - isometry constants .
we recall that @xmath2 is _ uniformly perfect _ if and only if there is a lower bound for the injectivity radius of @xmath2 in the poincar metric .
[ poincquant ] let @xmath2 be a uniformly perfect domain in @xmath3 and let @xmath18 be a lower bound for its injectivity radius in the poincar metric , then @xmath19 is @xmath20-lipschitz and a @xmath21-quasi - isometry with respect to the poincar metric . in particular , @xmath22 for all @xmath23 where @xmath24 .
examples in section [ round ] , show that the lipschitz constant can not have asymptotic form better than @xmath25 as @xmath26 .
when @xmath2 is simply connected , we obtain : [ qiinsc ] let @xmath2 be a simply connected domain in @xmath3 .
then @xmath27 for all @xmath23 .
in particular , this gives a simple reproof of epstein , marden and markovic s @xcite result that the nearest point retraction is 2-lipschitz .
it only relies on the elementary fact that @xmath28 is 1-lipschitz ( see lemma [ finiteptoh ] ) with respect to the thurston metric and a result of herron , ma and minda @xcite concerning the relationship between the thurston metric and the poincar metric in the simply connected setting .
marden and markovic @xcite also showed that the nearest point retraction is a quasi - isometry ( with respect to the poincar metric ) if and only if @xmath2 is uniformly perfect .
as another corollary of theorem [ projective ] we obtain the following alternate characterization of uniformly perfect domains which was conjectured by marden and markovic .
corollaries [ poincquant ] and [ mmconj ] together give an alternate proof of marden and markovic s original characterization .
[ mmconj ] the nearest point retraction @xmath29 is lipschitz , with respect to the poincar metric on @xmath2 and the intrinsic hyperbolic metric on @xmath1 if and only if the set @xmath2 is uniformly perfect . moreover ,
if @xmath2 is not uniformly perfect , then @xmath28 is not a quasi - isometry with respect to the poincar metric .
sullivan @xcite showed that there exists a uniform @xmath30 such that if @xmath2 is simply connected , then there is a conformally natural @xmath30-quasiconformal map from @xmath2 to @xmath1 which admits a bounded homotopy to the nearest point retraction .
( we say that a map from @xmath2 to @xmath1 is _ conformally natural _ if it is @xmath31-equivariant whenever @xmath31 is a group of mbius transformations preserving @xmath2 . )
epstein , marden and markovic @xcite showed that the best bound on @xmath30 must lie strictly between 2.1 and 13.88 . while , combining work of bishop @xcite and epstein - markovic @xcite , the best bound of @xmath30 lies between 2.1 and 7.82 if we do not require the quasiconformal map to be conformally natural .
marden and markovic @xcite showed that a domain is uniformly perfect if and only if there is a quasiconformal map from @xmath2 to @xmath1 which is homotopic to the nearest point retraction by a bounded homotopy .
moreover , they show that there are bounds on the quasiconformal constants which depend only on a lower bound for the injectivity radius of @xmath2 .
we note that again these bounds are not concrete . in order to use the nearest point retraction to obtain a conformally natural quasiconformal map between @xmath32 and @xmath1
we show that @xmath28 lifts to a quasi - isometry between the universal cover @xmath33 of @xmath2 and the universal cover @xmath34 of @xmath1 .
although they do not make this explicit in their paper , this is essentially the same approach taken by marden and markovic @xcite .
we recall that quasi - isometric surfaces need not even be homeomorphic and that even homeomorphic quasi - isometric surfaces need not be quasiconformal to one another .
notice that the quasi - isometry constants for the lift of the nearest point retraction are , necessarily , not as good as those for the nearest point retraction .
examples in section [ round ] show that the first quasi - isometry constant of @xmath35 can not have form better than @xmath36 as @xmath26 .
[ nprlift ] suppose that @xmath2 is a uniformly perfect hyperbolic domain and @xmath18 is a lower bound for its injectivity radius in the poincar metric .
then the nearest point retraction lifts to a quasi - isometry @xmath37 between the universal cover of @xmath2 ( with the poincar metric ) and the universal cover of @xmath1 with quasi - isometry constants depending only on @xmath38 .
in particular , @xmath39 for all @xmath40 where @xmath41 as @xmath26 and @xmath42 .
we then apply work of douady and earle @xcite to obtain explicit bound on the quasiconformal map homotopic to the nearest point retraction when @xmath2 is uniformly perfect .
[ upqc ] there exists an explicit function @xmath43 such that if @xmath2 is uniformly perfect and @xmath18 is a lower bound for its injectivity radius in the poincar metric , then there is a conformally natural @xmath43-quasiconformal map @xmath44 which admits a bounded homotopy to @xmath28 .
moreover , if @xmath2 is not uniformly perfect , then there does not exist a bounded homotopy of @xmath28 to a quasiconformal map .
we will give explicit formulas for @xmath45 and @xmath30 in section [ thelift ] .
the table below summarizes the main results concerning the nearest point retraction in the cases that the domain is simply connected , uniformly perfect and hyperbolic but not uniformly perfect .
[ cols="^,^,^,^ " , ] the only constant in the table above which is known to be sharp is the lipschitz constant in the simply connected case
. it would be desirable to know sharp bounds ( or better bounds ) in the remaining cases .
we refer the reader to the excellent discussion of the relationship between the nearest point retraction and complex analytic problems , especially brennan s conjecture , in bishop @xcite .
one of the main motivations for the study of domains and the convex hulls of their complements comes from the study of hyperbolic 3-manifolds where one is interested in the relationship between the boundary of the convex core of a hyperbolic manifold and the boundary of its conformal bordification . if @xmath46 is a hyperbolic 3-manifold , one considers the domain of discontinuity @xmath47 for @xmath31 s action on @xmath3 .
the _ conformal boundary _
@xmath48 is the quotient @xmath49 and @xmath50 is the natural conformal bordification of @xmath51 .
convex core _
@xmath52 of @xmath51 is obtained as the quotient of the convex hull of @xmath53 and its boundary @xmath54 is the quotient of @xmath1 .
the nearest point retraction descends to a homotopy equivalence @xmath55 in this setting , sullivan s theorem implies that if @xmath48 is incompressible in @xmath56 , then there is a quasiconformal map , with uniformly bounded dilatation , from @xmath48 to @xmath54 which is homotopic to the nearest point retraction .
since the thurston metric is @xmath31-invariant and the nearest point retraction is @xmath31-equivariant we immediately obtain a manifold version of theorem [ projective ] .
[ projectivemanifold ] let @xmath46 be a hyperbolic 3-manifold .
then the nearest point retraction @xmath57 is a @xmath58-quasi - isometry with respect to the thurston metric @xmath11 on @xmath48 and the intrinsic hyperbolic metric on @xmath54 .
in particular , @xmath59 we also obtain a version of corollary [ poincquant ] in the manifold setting .
[ poincquantmanifold ] let @xmath46 be a hyperbolic 3-manifold such that there is a lower bound @xmath18 for the injectivity radius of @xmath47 ( in the poincar metric ) , then @xmath60 for all @xmath61 where @xmath24 .
we also obtain a version of corollary [ upqc ] which guarantees the existence of a quasiconformal map from the conformal boundary to the boundary of the convex core .
[ qcmanifold ] if @xmath46 is a hyperbolic 3-manifold and @xmath18 is a lower bound for the injectivity radius of @xmath47 ( in the poincar metric ) , then there is a @xmath43-quasiconformal map @xmath62 which admits a bounded homotopy to the nearest point retraction @xmath35 .
if @xmath51 is a hyperbolic 3-manifold with finitely generated fundamental group , then its domain of discontinuity is uniformly perfect ( see pommerenke @xcite ) , so corollaries [ poincquantmanifold ] and [ qcmanifold ] immediately apply to all hyperbolic 3-manifolds with finitely generated fundamental group . more generally , analytically finite hyperbolic 3-manifolds ( i.e. those whose conformal boundary has finite area ) are known to be uniformly perfect @xcite .
we recall that it was earlier shown ( see @xcite ) that the nearest point retraction @xmath63 is @xmath64-lipschitz whenever @xmath51 is analytically finite and has a @xmath65-lipschitz homotopy inverse where @xmath66 ( for some explicit constant @xmath67 ) as @xmath26 . moreover , if @xmath46 is analytically finite and @xmath47 is simply connected , then @xmath35 has a @xmath68-lipschitz homotopy inverse .
remark ( 2 ) in section 7 of @xcite indicates that one can not improve substantially on the asymptotic form of the lipschitz constant for @xmath35 ( see also section 6 of @xcite ) , while proposition 6.2 of @xcite shows that one can not improve substantially on the asymptotic form of the lipschitz constant of the homotopy inverse for @xmath35 .
* outline of paper : * many of the techniques of proof in this paper are similar to those in our previous papers @xcite and @xcite .
the consideration of the thurston metric allows us to obtain more general results , while the fact that the thurston metric behaves well under approximation allows us to reduce many of our proofs to the consideration of the case where @xmath2 is the complement of finitely many points in @xmath3 and the convex hull of its complement is a finite - sided ideal polyhedron .
therefore , our techniques are more elementary than much of the earlier work on the subject . in section 2
we recall basic facts and terminology concerning the nearest point retraction and the boundary of the convex core . in section 3
we discuss the poincar , quasihyperbolic and thurston metrics on a hyperbolic domain . in section 4
we establish the key approximation properties which allow us to reduce the proof of theorems [ projective ] and [ nprlift ] to the case where @xmath69 is a finite collection of points . in section 5
we study the geometry of the finitely punctured case and introduce a notion of intersection number which records the total exterior dihedral angle along a geodesic in @xmath1 .
if @xmath70 is an arc in @xmath1 , then the difference between its length and the length of @xmath71 in the thurston metric is exactly this intersection number , see lemma [ finiteptoh ] .
section 6 contains the key intersection number estimates used in the proof of theorem [ projective ] .
the first estimate , lemma [ thickintbound ] , bounds the total intersection numbers of moderate length geodesic arcs in the thick part of @xmath1 .
( lemma [ thickintbound ] is a mild generalization of lemma 4.3 in @xcite and we defer its proof to an appendix . )
we then bound the total intersection number of a short geodesic ( lemma [ shortgeointbound ] ) and the angle of intersection of an edge of @xmath1 with a short geodesic ( lemma [ anglebound ] ) .
we give a quick proof of a weaker form of lemma [ shortgeointbound ] and defer the proof of the better estimate to the appendix . in a remark at the end of section 6 , we see that one could use this weaker estimate to obtain a version of theorem [ projective ] with slightly worse constants .
section 7 contains the proof of theorem [ projective ] .
it follows immediately from lemma [ finiteptoh ] that the nearest point retraction is 1-lipschitz with respect to the thurston metric , but it remains to bound the distance between two points @xmath72 and @xmath73 in @xmath2 in terms of the distance between @xmath74 and @xmath75 in @xmath1 .
we do so by first dividing the shortest geodesic joining @xmath74 to @xmath75 into segments of length roughly @xmath76 ( for some specific constant @xmath76 ) .
the segments in the thick part have bounded intersection number by lemma [ thickintbound ] .
we use lemmas [ shortgeointbound ] and [ anglebound ] to replace geodesic segments in the thin part by bounded length arcs which consist of one segment missing every edge and another segment with bounded intersection number .
we then obtain a path joining @xmath74 to @xmath75 with length and intersection number bounded above by a multiple of .
lemma [ finiteptoh ] can then be applied to complete the proof . in section 8
we derive consequences of theorem [ projective ] , including corollaries [ poincquant ] , [ qiinsc ] and [ mmconj ] . in section 9
we prove theorem [ nprlift ] , using techniques similar to those in the proof of theorem [ projective ] .
we then use work of douady and earle @xcite to derive corollary [ upqc ] . in section 10
, we study the special case of round annuli and observe that the asymptotic form of our estimates in corollary [ poincquant ] and theorem [ nprlift ] can not be substantially improved .
the appendix , section 11 , contains the complete proofs of lemmas [ thickintbound ] and [ shortgeointbound ] .
the proof of lemma [ thickintbound ] is considerably simpler than the proof of lemma 4.3 in @xcite as we are working in the finitely punctured case .
if @xmath77 is a hyperbolic domain , we let be the hyperbolic convex hull of in @xmath78 .
we then define @xmath79 and recall that it is hyperbolic in its intrinsic metric ( @xcite ) . in the special case that lies in a round circle , is a subset of a totally geodesic plane in @xmath78 . in this case
, it is natural to consider @xmath1 to be the double of along its boundary ( as a surface in the plane ) .
we define the _ nearest point retraction _ @xmath80 by letting @xmath81 be the point of intersection of the smallest horosphere about @xmath82 which intersect @xmath1 .
we note that is a conformally natural homotopy equivalence .
the nearest point retraction extends continuously to a conformally natural map @xmath83 where @xmath74 is the ( unique ) nearest point on if @xmath84 ( see @xcite ) .
we recall that a _ support plane _ to @xmath1 is a plane @xmath85 which intersects @xmath1 and bounds an open half - space @xmath86 disjoint from @xmath1 .
the half - space @xmath86 is associated to a maximal open disc @xmath87 in @xmath2 . since @xmath87 is maximal
, @xmath88 contains at least two points in @xmath8 .
therefore , @xmath85 contains a geodesic @xmath89 .
notice that @xmath85 intersects @xmath1 in either a single edge or a convex subset of @xmath85 which we call a _ face _ of @xmath1 .
if @xmath7 , then the geodesic ray starting at @xmath81 in the direction of @xmath82 is perpendicular to a support plane @xmath90 through @xmath81 .
given a hyperbolic domain @xmath2 in @xmath3 there is a unique metric @xmath91 of constant curvature @xmath92 which is conformally equivalent to the euclidean metric on @xmath2 .
this metric @xmath91 is called the _ poincar metric _ on @xmath2 and is denoted @xmath93 alternatively , the poincar metric can be given by defining the length of a tangent vector to be the infimum of the hyperbolic length over all vectors @xmath94 such that there exists a conformal map such that and in @xcite , beardon and pommerenke show that the quasihyperbolic metric and the poincar metric are closely related .
we recall that , the _ quasihyperbolic metric _ on @xmath95 is the conformal metric @xmath96 where @xmath6 is the euclidean distance from @xmath82 to @xmath8 .
they let @xmath97 alternatively , one can define where @xmath30 is the maximal modulus of an essential round annulus ( i.e. one which separates ) in @xmath2 whose central ( from the conformal viewpoint ) circle passes through @xmath82 .
we recall that if @xmath98 is any annulus in @xmath3 , then it is conformal to an annulus of the form and we define its _ modulus _ to be @xmath99 the central circle of @xmath98 is the pre - image of the unit circle in @xmath100 .
( beardon - pommerenke @xcite ) [ bpresult ] if @xmath2 is a hyperbolic domain in @xmath101 , then @xmath102 for all @xmath103 , where @xmath104 . moreover , @xmath105 for all @xmath7 .
if @xmath2 is simply connected , then @xmath106 using the fact that the core curves of large modulus essential annuli are short in the poincar metric , one sees that a lower bound on the injectivity radius of @xmath2 gives an explicit bilipschitz equivalence between the poincar metric and the quasihyperbolic metric when @xmath2 is uniformly perfect .
[ poincandqh ] if @xmath2 is a uniformly perfect hyperbolic domain and @xmath18 is a lower bound for the injectivity radius of @xmath2 in the poincar metric , then @xmath107 for all @xmath103 .
corollary 3.3 in @xcite asserts that if , then therefore , if , then .
the result then follows from theorem [ bpresult ] .
another metric closely related to the quasihyperbolic metric is the thurston metric .
the _ thurston metric _
@xmath108 on @xmath2 is defined by letting the length of a vector be the infimum of the hyperbolic length of all vectors such that there exists a mbius transformation @xmath109 such that and . by definition ,
the thurston metric is conformally natural and conformal to the euclidean metric .
furthermore , by our alternate definition of the poincar metric , we have the inequality @xmath110 for all @xmath103 .
the thurston metric is also known as the projective metric , as it arises from regarding @xmath2 as a complex projective surface and giving it the metric thurston described on such surfaces .
kulkarni and pinkall defined and studied an analogue of this metric in all dimensions and it is also sometimes called the kulkarni - pinkall metric .
see herron - ma - minda @xcite , kulkarni - pinkall @xcite , mcmullen @xcite and tanigawa @xcite for further information on the thurston metric .
kulkarni and pinkall ( see theorem 7.2 in @xcite ) proved that the thurston metric and the quasihyperbolic metric are 2-bilipschitz .
( @xcite ) [ projtoqh ] if @xmath2 is a hyperbolic domain in @xmath101 then @xmath111 in the simply connected setting , herron , ma and minda ( see theorem 2.1 and lemma 3.1 in @xcite ) prove that the poincar metric and the thurston metric are also 2-bilipschitz .
( @xcite ) [ scthurpoin ] if @xmath2 is a simply connected hyperbolic domain in @xmath3 , then @xmath112 we will make use of the following alternate characterization of the thurston metric . we note that it is not difficult to derive theorem [ projtoqh ] from this characterization .
[ projhoro ] if @xmath2 is a hyperbolic domain in @xmath101 , then the thurston metric @xmath113 on @xmath2 is given by @xmath114 where @xmath115 is the radius of the maximal horoball @xmath116 based at @xmath82 whose interior misses @xmath1 .
recall the disk model @xmath117 for @xmath118 with the metric @xmath119 by normalizing , we may define the length of a vector in the thurston metric to be the infimum of the length of vectors such that there exists a mbius transformation @xmath120 such that with @xmath121 .
the hyperbolic length of @xmath94 is then simply thus @xmath122 where the infimum is taken over all mbius transformations @xmath120 such that @xmath123 and @xmath124 .
given such a mbius transformation @xmath120 , we let and let @xmath125 be the hyperbolic plane with the same boundary as @xmath126 .
we further let @xmath127 be the horoball based at @xmath82 which is tangent to @xmath125 . as , @xmath125 does not intersect the interior of the convex hull .
similarly , the interiors of @xmath127 and are disjoint .
let @xmath128 be the radius of the horoball @xmath127 .
notice that @xmath129 .
let @xmath130 be a parabolic mbius transformation with fixed point @xmath82 , which sends @xmath131 to the point vertically above @xmath82 .
then , @xmath130 fixes @xmath127 and sends @xmath126 to a disk centered at @xmath82 with radius @xmath132 .
therefore , since @xmath133 sends a disk of radius @xmath134 to a disk of radius @xmath135 and sends centers to centers , @xmath136 since @xmath130 is parabolic and fixes @xmath82 , , which implies that therefore @xmath137 on the other hand , let @xmath90 be the support plane to @xmath81 which is perpendicular to the geodesic ray joining @xmath81 to @xmath82 .
then @xmath90 is tangent to @xmath116 at @xmath81 .
let @xmath138 be the ( open ) disk in @xmath2 containing @xmath82 whose boundary agrees with @xmath139 . if @xmath120 is a mbius transformation such that @xmath140 and @xmath141 , then , by the above analysis , @xmath142 .
it follows that @xmath143 as desired .
given a hyperbolic domain @xmath2 , let be a dense set of points in . if , then we call @xmath144 a _ nested finite approximation _ to . notice that @xmath145 converges to @xmath69 in the hausdorff topology on closed subsets of @xmath3
we let , denote the quasihyperbolic metric on @xmath146 by @xmath147 , and let be the nearest point retraction . since , @xmath147 and @xmath148
are both defined on @xmath2 .
results of bowditch @xcite imply that @xmath149 converges to and that converges to @xmath1 in the hausdorff topology on closed subsets of @xmath78 .
the following approximation lemma will allow us to reduce the proof of many of our results to the case where @xmath2 is the complement of finitely many points .
[ quasiapprox ] let @xmath144 be a nested finite approximation to @xmath69 . if , for all @xmath130 , @xmath150 , @xmath147 is the quasihyperbolic metric on @xmath146 , and @xmath151 is the thurston metric on @xmath146 , then 1 .
the sequence @xmath152 of metrics converges uniformly on compact subsets of @xmath2 to the quasihyperbolic metric @xmath153 on @xmath2 , 2 .
the sequence @xmath154 of metrics converges uniformly on compact subsets of @xmath2 to the thurston metric @xmath11 on @xmath2 , 3 .
the associated nearest point retractions converge to the nearest point retraction @xmath155 uniformly on compact subsets of @xmath2 , 4 . if @xmath23 , then converges to , and 5 . if @xmath103 , then @xmath156 let @xmath157 denote the distance from to @xmath158 .
it is clear that @xmath159 converges to @xmath160 uniformly on compact subsets of @xmath2 .
therefore , @xmath152 converges uniformly to @xmath153 on compact subsets of @xmath2 . if @xmath161 , let @xmath162 denote the maximal horoball based at @xmath82 whose interior is disjoint from @xmath163 and whose closure intersects @xmath164 at @xmath165 .
suppose that converges to @xmath103 .
since @xmath149 converges to , @xmath166 converges to a horoball @xmath167 based at @xmath82 whose interior is disjoint from and whose closure intersects @xmath168 at . by definition ,
@xmath81 is this intersection point , so .
therefore , @xmath169 converges uniformly to @xmath28 on @xmath2 , establishing ( 3 ) .
moreover , applying lemma [ projhoro ] , we see that @xmath170 converges to @xmath113 , so @xmath154 converges uniformly to @xmath11 on @xmath2 , which proves ( 2 ) .
suppose that @xmath23 .
if @xmath171 is a path in @xmath164 of length @xmath172 then @xmath171 converges , up to subsequence , to a path in @xmath1 of length at most @xmath173 joining @xmath81 to @xmath174 .
therefore , @xmath175 on the other hand , if @xmath176 is a path in @xmath1 of length , then @xmath177 is a path of length at most @xmath178 on @xmath164 joining @xmath179 to @xmath180 .
notice that and both converge to 0 , since converges to @xmath1
. therefore , it follows that @xmath181 , establishing ( 4 ) .
the proof of ( 5 ) follows much the same logic as the proof of ( 4 ) .
in this section we consider the situation where @xmath2 is the complement of finitely many points in @xmath3 . in this case
, the convex core of the complement of @xmath2 is an ideal hyperbolic polyhedron and its boundary @xmath1 consists of a finite number of ideal hyperbolic polygons meeting along edges .
we give a precise description of the thurston metric in this case ( lemma [ finiteproj ] ) and show that @xmath28 is 1-lipschitz with respect to the thurston metric ( lemma [ finiteptoh ] ) .
we also introduce an intersection number for a curve @xmath182\to\chb$ ] which is transverse to all edges and give a formula for the length of its pre - image @xmath71 in the thurston metric on @xmath2 .
let @xmath2 be the complement of a finite set in @xmath3 .
each face @xmath183 of the polyhedron is contained in a unique plane @xmath184 .
furthermore , @xmath184 bounds a unique open half - space @xmath185 which does not intersect .
the half - space @xmath185 defines an open disk @xmath186 in @xmath2 .
let @xmath187 be the ideal polygon in @xmath186 with same vertices as @xmath183 .
abusing notation , we will call @xmath187 a _ face _ of @xmath2 . if two faces @xmath188 and @xmath189 meet in an edge @xmath190 with exterior dihedral angle @xmath191 then the spherical bigon is adjacent to the polygons @xmath192 and @xmath193 and has angle @xmath191 .
thus we have a decomposition of @xmath2 into faces @xmath187 and spherical bigons @xmath167 corresponding to the edges of .
if @xmath187 is a face of @xmath2 then it inherits a hyperbolic metric @xmath194 which is the restriction of the poincar metric on @xmath186 .
each spherical bigon @xmath195 inherits a euclidean metric in the following manner .
let @xmath196 be a mbius transformation taking @xmath195 to the euclidean rectangle .
we then give @xmath195 the euclidean metric @xmath197 obtained by pulling back the standard euclidean metric on @xmath198 .
since any mbius transformation preserving @xmath198 is a euclidean isometry , this metric is independent of our choice of @xmath196 .
lemma [ finiteproj ] shows that the thurston metric agrees with the metrics defined above .
one may view this as a special case of the discussion in section 2.1 of tanigawa @xcite , see also section 8 of kulkarni - pinkall @xcite and section 3 of mcmullen @xcite .
[ finiteproj ] let @xmath2 be the complement of finitely many points in @xmath3
. then 1 .
the thurston metric on @xmath2 agrees with @xmath199 on each face @xmath187 of @xmath2 and agrees with @xmath197 on each spherical bigon @xmath195 .
2 . with respect to the thurston metric ,
the nearest point retraction @xmath200 maps each face @xmath187 isometrically to the face @xmath183 of @xmath1 and is euclidean projection on each bigon .
that is , if @xmath195 is a bigon , there is an isometry @xmath201 such that @xmath202 @>\pi_1 > > { \bf r } \end{cd}\ ] ] * proof : * first suppose that @xmath187 is a face of @xmath2
. we may normalize by a mbius transformation so that @xmath186 is the unit disk and @xmath203 . in this case , the maximal horoball at @xmath82 has height 1 , so the thurston metric for @xmath2 at @xmath82 agrees with the poincar metric on @xmath186 at @xmath82 .
moreover , one sees that @xmath28 agrees with the hyperbolic isometry from the unit disk to @xmath184 on @xmath187 .
now let @xmath190 be an edge between two faces @xmath188 and @xmath189 .
we may normalize so that @xmath190 is the vertical line joining @xmath204 to @xmath205 and define the isometry @xmath206 by .
we may further assume that @xmath195 is the sector for a point , the maximal horoball at @xmath82 is tangent to @xmath190 at height @xmath207 .
therefore the thurston metric on @xmath167 is @xmath208 , which agrees with @xmath197 .
moreover , , so @xmath28 has the form described above .
if is a curve on @xmath1 which is transverse to the edges of @xmath1 , then we define @xmath209 to be the sum of the dihedral angles at each of its intersection points with an edge .
we allow the endpoints of @xmath70 to lie on edges .
( in a more general setting , @xmath209 is thought of as the intersection number of @xmath70 and the bending lamination on @xmath1 , see @xcite . )
we denote by @xmath210 the length of @xmath70 in the intrinsic metric on @xmath1 ( which is also its length in the usual hyperbolic metric on @xmath78 ) .
if @xmath211\to\omega$ ] is a rectifiable curve in @xmath2 , then let @xmath212 denote its length in the thurston metric .
similarly , we can define @xmath213 to be its length in the quasihyperbolic metric and @xmath214 to be its length in the poincar metric .
( since all these metrics are complete and locally bilipschitz , a curve is rectifiable in one metric if and only if it is rectifiable in another . )
suppose that @xmath2 is the complement of finitely many points in @xmath3 . 1 .
if @xmath211 \rightarrow \omega$ ] then @xmath215 2 . if @xmath182 \rightarrow \chb$ ] is a curve transverse to the edges of @xmath1 then @xmath216 [ finiteptoh ] by lemma [ finiteproj ] , @xmath28 is an isometry on the faces of @xmath1 and euclidean projection on the bigons .
it follows that @xmath28 is 1-lipschitz .
thus , for @xmath211 \rightarrow \omega$ ] , @xmath215 now let @xmath182 \rightarrow \chb$ ] be a curve transverse to the edges of @xmath1 .
let @xmath217 be the finite collection of intersection points of @xmath70 with the edges of @xmath1 .
let @xmath218 and @xmath219 .
( notice that if @xmath220 lies on an edge , then , while if @xmath221 lies on an edge . )
let @xmath222 be the restriction of @xmath70 to .
then @xmath223 as @xmath28 is an isometry on the interior of the faces , @xmath224 if @xmath225 lies on the edge @xmath226 of @xmath1 with exterior dihedral angle @xmath227 , then , by lemma [ finiteproj ] , @xmath228 has length @xmath227 in the thurston metric .
therefore , @xmath229 thus , @xmath230 as claimed .
* remark : * part ( 2 ) is closely related to theorem 3.1 in mcmullen @xcite which shows that @xmath231
we continue to assume throughout this section that @xmath2 is the complement of finitely many points in @xmath3 .
we obtain bounds on @xmath209 for short geodesic arcs in the thick part ( lemma [ thickintbound ] ) and short simple closed geodesics in @xmath1 ( lemma [ shortgeointbound ] ) .
we use lemmas [ thickintbound ] and [ shortgeointbound ] to bound the angle of intersection between an edge of @xmath1 and a short simple closed geodesic ( lemma [ anglebound ] ) .
we first state a mild generalization of lemma 4.3 from @xcite which gives the bound on @xmath209 for short geodesic arcs . in @xcite we define a function @xmath232 and we let the function @xmath183 is monotonically increasing and has domain the function @xmath233 has domain @xmath234 , has asymptotic behavior as @xmath235 tends to @xmath204 , and @xmath233 approaches @xmath236 as @xmath72 tends to @xmath205 .
[ thickintbound ] suppose that @xmath2 is the complement of finitely many points in @xmath3 .
if is a geodesic path and @xmath237 for some @xmath238 $ ] , then @xmath239 in the appendix , we give a self - contained proof of lemma [ thickintbound ] .
the proof uses the same basic techniques as in @xcite , but the arguments are much more elementary , as we need not use the general theory of bending measures . in this section
, we recall a version of the collar lemma ( see buser @xcite ) which gives a complete description of the portion of a hyperbolic surface with injectivity radius less than @xmath240 .
if @xmath176 is a simple closed geodesic in a hyperbolic surface @xmath241 , then we define the _ collar _ of @xmath176 to be the set @xmath242 a _ cusp _ is a subsurface isometric to @xmath243 with the metric @xmath244 * collar lemma : * _ let @xmath241 is a complete hyperbolic surface . _ 1 .
the collar @xmath245 about a simple closed geodesic @xmath176 is isometric to @xmath246 $ ] with the metric @xmath247 2 . a collar about a simple closed geodesic is disjoint from the collar about any disjoint simple closed geodesic and from any cusp of @xmath241 3 .
any two cusps in @xmath241 are disjoint .
4 . if @xmath248 , then @xmath72 lies in a cusp or in a collar about a simple closed geodesic of length at most @xmath249 .
if @xmath250 , then @xmath251 if @xmath252 is a cusp or a collar about a simple closed geodesic , then we will call a curve of the form @xmath253 , in the coordinates given above , a _ cross - section_. the following observations concerning the intersections of cross - sections with edges will be useful in the remainder of the section .
[ cuspcross ] suppose that @xmath2 is the complement of a finite collection @xmath254 of points in @xmath3 and @xmath252 is a cusp of @xmath1 .
then every edge which intersects @xmath252 intersects every cross - section of @xmath252 and does so orthogonally .
moreover , if @xmath176 is any cross - section of @xmath252 , then @xmath255 . identify the universal cover @xmath34 of @xmath1 with @xmath118 so that the pre - image of @xmath252 is the horodisk which is invariant under the covering transformation if @xmath256 is the pre - image of an edge , then @xmath256 either terminates at @xmath257 , or has two finite endpoints .
if @xmath256 terminates at infinity then it crosses every cross - section of @xmath258 orthogonally .
otherwise , since the edge @xmath190 is simple , the endpoints of @xmath256 must have euclidean distance less than 2 from one another and therefore @xmath256 does not intersect @xmath258 ( see also lemma 1.24 in @xcite ) .
every cusp is associated to a point @xmath259 , which we may regard as an ideal vertex of @xmath260 .
if we move @xmath72 to @xmath257 , then each cross - section is a euclidean polygon in a horosphere , so has total external angle exactly @xmath261 .
[ collarcross ] suppose that @xmath2 is the complement of finitely many points in @xmath3 and @xmath245 is collar about a simple closed geodesic @xmath176 on @xmath1 .
if @xmath190 is an edge of @xmath1 , then each component of @xmath262 intersects every cross - section of @xmath245 exactly once .
again identify the universal cover of @xmath1 with @xmath118 .
let @xmath263 be a component of the pre - image of @xmath176 and let @xmath264 be the metric neighborhood of @xmath265 of radius @xmath266 . then @xmath264 is a component of the pre - image of @xmath245 .
notice that @xmath265 is the axis of a hyperbolic isometry @xmath267 in the conjugacy class determined by the simple closed curve @xmath176 .
notice that every edge @xmath190 of @xmath1 is an infinite simple geodesic which does not accumulate on @xmath176 .
let @xmath256 be the pre - image of an edge which intersects @xmath264 , then neither of its endpoints can be an endpoint of @xmath176 .
if the endpoints of @xmath256 lie on the same side of @xmath263 , one may use the fact that they must lie in a single fundamental domain for @xmath267 and the explicit description of @xmath264 to show that @xmath256 can not intersect @xmath264 ( see also lemma 1.21 in @xcite ) . if its endpoints lie on opposite sides of @xmath265 , then it intersects every cross - section of @xmath264 exactly once , as desired .
we now observe that if @xmath176 is a homotopically non - trivial curve on @xmath1 , then [ intlowerbound ] if @xmath2 is the complement of finitely many points in @xmath3 and is homotopically non - trivial in @xmath1 , then @xmath268 if @xmath176 is a closed geodesic , then , by the gauss - bonnet theorem , its geodesic curvature is at least @xmath261 . but @xmath269 is the sum of the dihedral angles of the edges @xmath176 intersects , and is therefore at least as large as the geodesic curvature of @xmath176 , so @xmath270 in this case . if @xmath176 is any homotopically non - trivial curve on @xmath1 , then it is homotopic to either a geodesic @xmath271 or a ( multiple of a ) cross - section @xmath252 of a cusp and either or in either case , ( recall that lemma [ cuspcross ] implies that ) as an immediate corollary of lemmas [ intlowerbound ] , [ finiteproj ] and [ quasiapprox ] we see that every essential curve in a hyperbolic domain has length greater than @xmath261 in the thurston metric .
[ thurstonlengthbound ] if @xmath272 is a hyperbolic domain and @xmath273 is homotopically non - trivial in @xmath2 , then @xmath274 we may use a variation on the construction from lemma [ thickintbound ] to uniformly bound @xmath269 for all short geodesics . [ shortgeointbound ] suppose that @xmath2 is the complement of finitely many points in @xmath3 .
if @xmath176 is a simple closed geodesic of length less than @xmath249 , then @xmath275 [ short ] the lower bound @xmath276 follows immediately from lemma [ intlowerbound ] . for the moment
, we will give a much briefer proof that and defer the proof of the better upper bound to the appendix .
* proof that @xmath277 : * let @xmath72 be a point in @xmath245 with injectivity radius @xmath240 and let @xmath252 be the cross - section of @xmath245 which contains @xmath72 .
part ( 5 ) of the collar lemma implies that @xmath278 and part ( 1 ) then implies that @xmath279 we may therefore divide @xmath271 into @xmath280 segments of length at most @xmath281 , denoted @xmath282 , @xmath283 and @xmath284 .
each @xmath285 is homotopic ( relative to its endpoints ) to a geodesic arc @xmath222 and @xmath286 for all @xmath287 , since no edge can intersect either @xmath222 or @xmath285 twice .
lemma [ thickintbound ] implies that @xmath288 .
therefore , @xmath289 lemma [ collarcross ] implies that @xmath290 completing the proof .
the final tool we will need in the proof of theorem [ projective ] is a bound on the angle between an edge of @xmath1 and a short simple closed geodesic on @xmath1 .
[ anglebound ] suppose that @xmath2 is the complement of finitely many points in @xmath3 .
if @xmath176 is a simple closed geodesic of length @xmath291 , then there is an edge @xmath292 intersecting @xmath176 at an angle @xmath293 furthermore , any other edge @xmath226 intersecting @xmath176 intersects in an angle @xmath294 let @xmath176 intersect edges @xmath295 at points @xmath296 and with angles @xmath297 .
we may assume that the largest value of @xmath298 is achieved at @xmath299 .
furthermore , let @xmath300 be the exterior dihedral angle of the faces which meet at @xmath226 and let @xmath301 be the exterior angle of @xmath176 at @xmath225 . a simple ( euclidean ) calculation yields @xmath302 consider @xmath303 \rightarrow [ 0,\pi]$ ] , given by @xmath304 then @xmath305 .
one may compute that @xmath306 since @xmath307 is an increasing concave down function , @xmath308 and @xmath309 we see that @xmath310 let @xmath311 be the geodesic curvature of @xmath176
. then @xmath312 lemma [ short ] , and the fact that @xmath313 , imply that @xmath314 therefore , @xmath315 we now use the bound on @xmath299 to bound the other angles of intersection .
let @xmath226 be an edge , where @xmath316 , intersecting @xmath176 .
identify the universal cover of @xmath1 with @xmath118 and let @xmath267 be a geodesic in @xmath118 covering @xmath176 .
there is a lift @xmath317 of @xmath292 which is a geodesic intersecting @xmath267 at point @xmath318 with angle @xmath299 .
translating along @xmath267 we have another lift @xmath319 of @xmath292 intersecting @xmath267 at point @xmath320 a distance @xmath321 from @xmath318 .
let @xmath322 be a lift of @xmath226 whose intersection @xmath82 with @xmath267 , lies between @xmath318 and @xmath320 . without loss of generality , we may assume that .
as edges do not intersect , @xmath322 lies between @xmath317 and @xmath319 .
let @xmath323 and @xmath153 be the endpoints of @xmath317 and let @xmath324 ( respectively @xmath325 ) be the triangle with vertices @xmath318 , @xmath82 , and @xmath323 ( respectively @xmath318 , @xmath82 and @xmath153 ) .
let @xmath326 and @xmath327 be the internal angles of @xmath324 and @xmath325 at @xmath82 .
we may assume that the internal angle of @xmath324 at @xmath318 is and hence that the internal angle of @xmath325 at @xmath318 is @xmath299 . by construction , @xmath328 basic calculations in hyperbolic trigonometry ( see ( * ? ? ?
* theorem 7.10.1 ) ) , give that @xmath329 and @xmath330 one may then check that @xmath331 therefore , @xmath332 since , and @xmath333 * remark : * using the elementary estimate @xmath277 in lemma [ anglebound ] we obtain an angle bound of @xmath334 . using this we can prove a weaker form of theorem [ projective ] where @xmath335 is replaced by @xmath336 and @xmath337 is replaced with @xmath338 .
we are now prepared to prove theorem [ projective ] which asserts that the nearest point retraction is a uniform quasi - isometry in the thurston metric .
our approximation result , lemma [ quasiapprox ] , allows us to reduce to the finitely punctured case .
we begin by producing an upper bound on the distance between points in @xmath2 whose images on @xmath1 are close .
[ distbound ] suppose that @xmath2 is the complement of finitely many points in @xmath3 .
if @xmath339 and @xmath340 then @xmath341 let @xmath182\to\chb$ ] be a shortest path joining @xmath81 to @xmath174 and let @xmath342 by assumption , @xmath343 if @xmath70 contains a point of injectivity radius greater than @xmath240 , then lemma [ thickintbound ] implies that lemma [ finiteptoh ] then gives that @xmath344 if @xmath70 does not contain a point of injectivity radius greater than @xmath240 , then , by the collar lemma , @xmath70 is either entirely contained in a cusp or in the collar @xmath245 about a simple closed geodesic @xmath176 with let @xmath221 lie on cross - section @xmath252 of a cusp or collar .
every cross - section of a cusp has length at most 2 , and we may argue , exactly as in section [ shortgeo ] , that if @xmath252 is a cross - section of a collar , then @xmath345 in either case , if @xmath252 is a cross - section of a cusp , then by lemma [ cuspcross ] , while if @xmath252 is a cross - section of a collar , then , by lemmas [ collarcross ] and [ shortgeointbound ] , therefore , lemma [ finiteptoh ] implies that @xmath346 if @xmath70 is entirely contained in a cusp , then we may join @xmath220 to a point @xmath323 on the cross - section @xmath252 by an arc @xmath167 orthogonal to each cross - section such that also , since each edge is orthogonal to every cross - section , @xmath167 either misses every edge of @xmath1 or is a subarc of an edge .
if @xmath167 misses every edge , then @xmath347 and is an arc joining @xmath82 to a point @xmath72 on @xmath348 of length if @xmath167 is a subarc of an edge @xmath190 , then @xmath349 contains a vertical ( in the euclidean coordinates described in lemma [ finiteproj ] ) arc @xmath350 which joins @xmath82 to a point @xmath323 on @xmath348 such that in both cases , since @xmath323 can be joined to @xmath351 by an arc of length at most @xmath352 in the thurston metric .
therefore , @xmath353 finally , suppose that @xmath70 is contained in the collar @xmath245 about a simple closed geodesic @xmath176 with @xmath291 .
we may assume that @xmath220 lies on cross - section @xmath354 , @xmath221 lies on the cross - section @xmath252 given by @xmath355 , and as every edge that intersects the collar @xmath245 must intersect the core geodesic @xmath176 ( see lemma [ collarcross ] ) , we can decompose @xmath245 along edges to obtain a collection of quadrilaterals with two opposite sides corresponding to edges . as the edges intersect @xmath176 in an angle at least @xmath356 , we can foliate each quadrilateral by geodesics which also intersect @xmath176 in an angle greater than @xmath356 .
thus , we obtain a foliation of @xmath245 by such geodesics .
we let @xmath167 be the subarc of a leaf in the foliation joining @xmath220 to a point @xmath323 on the cross - section @xmath252 .
then , either @xmath167 does not intersect any edges of @xmath1 or is a subset of an edge . in either case
, @xmath167 belongs to geodesic @xmath267 which intersects @xmath176 at an angle we now use the angle bound to obtain an upper bound on the length of @xmath167 , which will allow us to complete the argument much as in the cusp case .
the hyperbolic law of sines implies that @xmath357 if we let then since @xmath109 is odd and increasing and @xmath358 since @xmath359 is decreasing on ( since for all @xmath360 ) it achieves its maximum value when @xmath361 , which implies that @xmath362 since @xmath167 is either disjoint from all edges of @xmath1 or is a subset of an edge , we argue as before to show that @xmath349 contains an arc @xmath350 joining @xmath82 to a point @xmath323 in @xmath348 such that @xmath363 therefore , as before , @xmath364 therefore , we see that in all cases @xmath365 as a nearly immediate corollary , we show that the nearest point retraction is a uniform quasi - isometry in the finitely punctured case .
[ finiteqi ] suppose that @xmath2 is the complement of finitely many points in @xmath3 .
then , @xmath366 where @xmath367 and @xmath368 the lower bound follows from the fact that @xmath28 is 1-lipschitz in the thurston metric ( see lemma [ finiteptoh ] ) . to prove the upper bound we consider a shortest path @xmath70 in @xmath1 from @xmath81 to @xmath174 .
we decompose @xmath70 into geodesic segments of length at most then , applying lemma [ distbound ] , @xmath369 where @xmath370
. we may now combine corollary [ finiteqi ] and lemma [ quasiapprox ] to prove theorem [ projective ] in the general case .
* theorem [ projective ] . *
_ let @xmath2 be a hyperbolic domain .
then the nearest point retraction @xmath200 is a @xmath10-quasi - isometry with respect to the thurston metric @xmath11 on @xmath2 and the intrinsic hyperbolic metric on @xmath1 where @xmath12 and @xmath13 . in particular , @xmath14 furthermore , if @xmath2 is simply connected , then @xmath15 where @xmath16 and @xmath17 .
_ let @xmath144 be a nested finite approximation for @xmath69 , let with projective metric @xmath371 and let @xmath372 be the associated nearest point retraction .
then , by lemma [ quasiapprox ] , converges to @xmath373 and converges to @xmath374 for all corollary [ finiteqi ] implies that @xmath375 for all @xmath130 and all @xmath23 , so the result follows .
now suppose that is @xmath2 simply connected and let @xmath376 be the minimum injectivity radius in @xmath158 along the shortest curve @xmath377 joining @xmath165 and @xmath378 .
lemma [ quasiapprox ] implies that @xmath379 as @xmath380 . as in the proof of corollary [ finiteqi ]
, we decompose @xmath377 into segments of length @xmath381 and conclude that @xmath382 as @xmath383 , we again apply lemma [ quasiapprox ] to obtain @xmath384 where @xmath385 and @xmath386 .
in this section , we derive a series of corollaries of theorem [ projective ] .
we begin by combining theorem [ projective ] with beardon and pommerenke s work @xcite and theorem [ projtoqh ] to obtain explicit quasi - isometry bounds for the nearest point retraction with respect to the poincar metric on a uniformly perfect domain .
* corollary [ poincquant ] . *
_ let @xmath2 be a uniformly perfect domain in @xmath3 and let @xmath18 be a lower bound for its injectivity radius in the poincar metric , then @xmath387 for all @xmath23 where @xmath24 . _ combining corollary [ poincandqh ] , theorem [ projtoqh ] and the fact that we see that @xmath388 for all @xmath103 .
it follows that @xmath389 for all @xmath23 .
the result then follows by applying theorem [ projective ] .
if @xmath2 is simply connected , then we may apply theorem [ scthurpoin ] to obtain better quasi - isometry constants .
bishop ( see lemma 8 in @xcite ) first showed that the nearest point retraction is a quasi - isometry in the simply connected case , but did not obtain explicit constants .
* corollary [ qiinsc ] . * _ if @xmath2 is a simply connected domain in @xmath3 , then @xmath390 for all @xmath23 .
_ we now establish marden and markovic s conjecture that the nearest point retraction is lipschitz ( with respect to the poincar metric on @xmath2 ) if and only if @xmath2 is uniformly perfect .
* corollary [ mmconj ] . * _ if @xmath2 is a hyperbolic domain , then the nearest point retraction is lipschitz ( where we give @xmath2 the poincar metric ) if and only if @xmath2 is uniformly perfect . moreover , if @xmath2 is not uniformly perfect , then @xmath28 is not a quasi - isometry with respect to the poincar metric . _
if @xmath2 is uniformly perfect , then corollary [ poincquant ] implies that @xmath28 is lipschitz . if @xmath2 is not uniformly perfect , then there exist essential round annuli in @xmath2 with arbitrarily large moduli .
for each @xmath130 , let @xmath391 be a separating round annulus in @xmath2 of modulus @xmath392 .
we may normalize so that @xmath393 and and choose and if @xmath82 lies in the subannulus then @xmath394
so @xmath395 as every path from @xmath396 to @xmath397 contains an arc joining the boundary components of @xmath398 , this implies that @xmath399 therefore , @xmath400 , so , by theorem [ projtoqh ] , theorem [ projective ] then implies that @xmath401 on the other hand , if @xmath176 is the segment of the real line joining @xmath396 to @xmath397 , then for all @xmath402 , so by theorem [ bpresult ] , @xmath403 for all @xmath130 and some constant @xmath404 .
therefore , @xmath405 for all @xmath130 .
it follows that @xmath28 is not lipschitz with respect to the poincar metric on @xmath2 .
moreover , @xmath28 is not even a quasi - isometry .
in this section we show that if @xmath2 is uniformly perfect , then the nearest point retraction lifts to a quasi - isometry , with quasi - isometry constants depending only on the geometry of the domain .
the proof requires only lemma [ thickintbound ] and does not make use of the analysis of the thin part obtained in lemmas [ shortgeointbound ] and [ anglebound ] .
we then use this result to show that @xmath28 is homotopic to a quasiconformal homeomorphism if @xmath2 is uniformly perfect . * theorem [ nprlift ] . * _ suppose that @xmath2 is a uniformly perfect hyperbolic domain and @xmath18 is a lower bound for its injectivity radius in the poincar metric
. then the nearest point retraction lifts to a quasi - isometry @xmath37 between the universal cover of @xmath2 ( with the poincar metric ) and the universal cover of @xmath1 where the quasi - isometry constants depend only on @xmath38 .
in particular , @xmath406 for all @xmath40 , where , @xmath407 @xmath408 _ corollary [ poincquant ] implies that @xmath28 is @xmath64-lipschitz , so therefore @xmath409 is also @xmath64-lipschitz
. given , let be the geodesic arc in @xmath410 joining them and let be its projection to @xmath2 ( i.e. where is the universal covering map . )
then is the length of the geodesic arc @xmath70 on @xmath1 joining @xmath411 and @xmath411 in the proper homotopy class of @xmath412 .
let @xmath145 be a nested finite approximation to @xmath69 , let , let @xmath371 be the thurston metric on @xmath146 , and let be the nearest point retraction .
let @xmath377 be the geodesic arc in @xmath164 joining @xmath413 and @xmath414 in the proper homotopy class of @xmath415 .
let @xmath376 be the minimum of the injectivity radius of @xmath164 at points in @xmath377 .
if , then we may divide @xmath377 into segments of length @xmath416 and apply lemmas [ finiteptoh ] and [ thickintbound ] , as in the proof of corollary [ finiteqi ] , to obtain the bound , @xmath417 since @xmath418 converges to @xmath1 , @xmath419 converges to @xmath28 , for all @xmath130 , and @xmath420 converges to @xmath11 ( see lemma [ quasiapprox ] ) , @xmath421 converges , up to subsequence , to a geodesic path @xmath422 on @xmath1 joining @xmath411 to @xmath423 . since @xmath424 lies arbitrarily close to @xmath377 ( for large enough @xmath130 in the subsequence ) which is properly homotopic to @xmath425
, we see that @xmath422 is homotopic to @xmath426 , and hence agrees with @xmath70 .
lemma 9.1 in @xcite gives that if @xmath18 is a lower bound for the injectivity radius of @xmath2 , in its poincar metric , then @xmath427 is a lower bound for the injectivity radius of @xmath1 . since @xmath421 converges to @xmath70 , up to subsequence , for all @xmath7 ( see lemma [ quasiapprox ] again ) , and the injectivity radius function is 1-lipschitz on any hyperbolic surface , we see that .
so , we may choose @xmath428 so that , therefore , by taking limits , we see that @xmath429 the curve @xmath71 contains an arc joining @xmath411 to @xmath423 in the homotopy class of @xmath430 , so @xmath431 therefore , @xmath432 * remark : * the proof of corollary [ mmconj ] also implies that @xmath409 is not a quasi - isometry with respect to the poincar metric if @xmath2 is not uniformly perfect . since any homotopically non - trivial curve in @xmath2 has length greater than @xmath261 in the thurston metric ( by lemma [ thurstonlengthbound ] ) , @xmath409 is not even a quasi - isometry with respect to the ( lift of the ) thurston metric on @xmath33 if @xmath2 is not uniformly perfect .
we now use the douady - earle extension theorem to obtain a quasiconformal map homotopic to the nearest point retraction whenever @xmath2 is uniformly perfect .
the constants obtained in this proof are explicit , but clearly far from optimal .
* corollary [ upqc ] . *
_ if @xmath2 is uniformly perfect and @xmath18 is a lower bound for its injectivity radius in the poincar metric , then there is a conformally natural @xmath43-quasiconformal map @xmath44 which admits a bounded homotopy to @xmath28 , where @xmath433 and @xmath434 .
moreover , if @xmath2 is not uniformly perfect , then there does not exist a bounded homotopy of @xmath28 to a quasiconformal map .
_ a close examination of the standard proof that quasi - isometries of @xmath118 extend to quasisymmetries of , see , e.g. , lemma 3.43 of @xcite and lemma 5.9.4 of @xcite , yields that a -quasi - isometry of @xmath118 extends to a -quasisymmetry of @xmath435 .
the beurling - ahlfors extension of a @xmath436-quasisymmetry of @xmath435 is a @xmath437-quasiconformal map ( see @xcite ) .
proposition 7 of douady - earle @xcite then implies that every @xmath436-quasisymmetry admits a conformally natural quasiconformal extension with dilatation at most first suppose that @xmath2 is uniformly perfect and identify both @xmath33 and @xmath34 with @xmath118 .
then @xmath409 lifts to a -quasi - isometry of @xmath410 and so extends to a -quasisymmetry of @xmath435 , which itself extends to a conformally natural @xmath43-quasiconformal map .
it follows from proposition 4.3.1 and theorem 4.3.2 in fletcher - markovic @xcite that the straight - line homotopy between @xmath409 and @xmath438 is bounded .
since @xmath409 is the lift of a conformally natural map and the douady - earle extension is conformally natural , we see that @xmath438 descends to a conformally natural @xmath43-quasiconformal map and that the straight - line homotopy between @xmath409 and @xmath438 descends to a bounded homotopy .
a quasiconformal homeomorphism between hyperbolic surfaces is a quasi - isometry ( see ( * ? ? ?
* theorem 4.3.2 ) ) .
thus , if @xmath28 admits a bounded homotopy to a quasiconformal map , it must itself be a quasi - isometry .
therefore , by corollary [ mmconj ] , if @xmath28 admits a bounded homotopy to a quasiconformal map , @xmath2 must be uniformly perfect .
in this section , we will consider the special case of round annuli .
a hyperbolic domain contains points with small injectivity radius ( in the poincar metric ) if and only if it contains round annuli with large modulus , so round annuli are natural test cases for the constants we obtain .
moreover , hyperbolic domains which are not uniformly perfect contain round annuli of arbitrarily large modulus .
let @xmath439 denote the round annulus lying between concentric circles of radius @xmath134 and @xmath440 about the origin .
one may calculate that in the poincar metric @xmath439 is a complete hyperbolic annulus with core curve of length @xmath441 and that @xmath442 is a complete hyperbolic annulus with core curve of length @xmath443 ( see example 3.a in herron - minda - ma @xcite and theorem 2.16.1 in epstein - marden @xcite ) .
more explicitly , @xmath439 is homeomorphic to @xmath444 with the metric @xmath445 and @xmath442 is homeomorphic to @xmath444 with the metric @xmath446 the nearest point retraction @xmath447 takes the core curve of @xmath439 to the core curve of @xmath442 . more generally , it takes any cross - section of @xmath439 to a cross - section of @xmath442 ( and restricts to a dilation ) and takes any vertical geodesic perpendicular to the core curve of @xmath439 to a vertical geodesic perpendicular to the core curve of @xmath442 .
let be the minimal value for the injectivity radius on @xmath439 .
one may check that there exists a constant @xmath67 such that if , then we assume from now on that .
let @xmath448 as @xmath449 .
choose points @xmath450 and @xmath451 in @xmath439 for fixed @xmath452 . notice that and @xmath453 and @xmath454 lie on opposite sides of a geodesic perpendicular to the core curve of @xmath439 .
therefore , @xmath455 and @xmath456 lie on opposite sites of a geodesic perpendicular to the core curve of @xmath442 . since
, we see that @xmath457 as @xmath458 it follows that the lipschitz constant of @xmath459 is at least @xmath460 as @xmath461 .
but , since @xmath462 , this expression is also as @xmath463 .
notice that the lipschitz constant in corollary [ poincquant ] is as @xmath26 , suggesting that our constants are close to having the correct form .
( we remark that corollary 1.2 contradicts part ( 3 ) of theorem 2.16.1 in @xcite , which is incorrect as stated . )
if the lift of the nearest point retraction to a map from @xmath464 to @xmath465 is a -quasi - isometry , then @xmath466 as @xmath449 , since it takes the lift of the core curve of @xmath439 to the lift of the core curve of @xmath442 equivariantly . notice that this expression is also as @xmath463 , and that in theorem [ nprlift ] , so again our constants are close to having the right form .
this also illustrates the fact that the quasi - isometry constants of the lift will typically be larger than the quasi - isometry constants of the original map .
we now consider the thurston metric on the round annulus @xmath439 . in this case , the nearest point retraction is a homeomorphism .
one may extend the analysis in section [ fingeom ] to show that the nearest point retraction is an isometry , with respect to the thurston metric , on lines through the origin . on a circle about the origin , with length @xmath45 in the thurston metric ,
the nearest point retraction is a dilation with dilation constant @xmath467 .
in particular , each such circle has length greater than @xmath261 and the core curve of @xmath439 has length in the thurston metric .
( one may also derive these facts more concretely using lemma [ projhoro ] to compute the thurston metric .
see example 3.a in @xcite for an explicit calculation . )
it follows that the nearest point retraction is a -quasi - isometry for all @xmath439 .
it is natural to ask what the best possible general quasi - isometry constants are for the nearest point retraction .
in this appendix we give the proofs of lemmas [ thickintbound ] and [ shortgeointbound ] .
the proof of lemma [ thickintbound ] follows the same outline as the proof of lemma 4.3 in @xcite , but is technically much simpler as we are working in the setting of finitely punctured domains .
we give the proof of lemma [ thickintbound ] both to illustrate the simplifications achieved and to motivate the proof of lemma [ shortgeointbound ] which has many of the same elements .
we begin by associating a family of support planes to a geodesic arc @xmath70 in @xmath1 .
if these support planes all intersect , we will obtain a bound on @xmath209 .
the proof of this bound is much easier in our simpler setting than in @xcite .
our lemma is essentially a special case of lemma 4.1 in @xcite .
let @xmath182\to \chb$ ] be a geodesic arc transverse to the edges of @xmath1 .
let be the finite collection of intersection points of the interior @xmath470 of @xmath70 with edges of @xmath1 .
if the initial point of @xmath70 lies on an edge , we denote the point by @xmath471 and the edge by @xmath472 , while if the endpoint of @xmath70 lies on an edge , we denote the point @xmath396 and the edge by @xmath473 .
suppose that @xmath225 lies on edge @xmath226 and that @xmath226 has exterior dihedral angle @xmath227 .
we let @xmath474 be the face containing @xmath475 .
( if @xmath220 lies on an edge @xmath472 , we let @xmath476 be the face abutting @xmath472 which does not contain @xmath477 . ) at each point @xmath225 , there is a 1-parameter family of support planes where @xmath478 contains @xmath479 and the dihedral angle between @xmath480 and @xmath481 is concatenating these one parameter families we obtain the one parameter family of support planes along @xmath70 , called the _ full family of support planes _ to @xmath70 . [ sptbnd ] suppose that @xmath2 is the complement of a finite collection @xmath254 of points in @xmath3 and @xmath182 \rightarrow \chb$ ] is a geodesic path ( which may have endpoints on edges ) . if @xmath487 is a family of support planes along @xmath70 such that the associated half - spaces satisfy @xmath488 for all @xmath489 $ ] , then @xmath490 where @xmath491 is the exterior dihedral angle between @xmath492 and @xmath493 .
moreover , @xmath70 intersects each edge at most once .
* proof : * let @xmath494 be the intersection points of @xmath70 with edges @xmath495 .
let be the support planes to @xmath226 in the family of support planes @xmath487 .
as then @xmath496 is a geodesic if @xmath497 . if @xmath498 , then @xmath499 for all @xmath500 $ ] such that @xmath501 .
if @xmath226 is not on @xmath492 then is a continuously varying family of disjoint geodesics lying between @xmath502 and @xmath503 on @xmath492 .
if every edge @xmath70 intersects is contained in @xmath492 , then the result is obvious , so we will assume that there is an edge which @xmath70 intersects which does not lie in @xmath492 .
choose @xmath504 so that @xmath505 is the first edge @xmath70 intersects that does not belong to @xmath492 . if an edge @xmath506 with @xmath507 lies in @xmath492 choose @xmath506 to be the first edge such that @xmath508 and @xmath506 is on @xmath492 .
otherwise , choose @xmath509 to be the last edge which @xmath70 intersects . by construction , since @xmath510 does not lie on @xmath492 .
notice that for and that is a continuous family of geodesics on @xmath492 .
if the family of geodesics are not all disjoint , then there must exist @xmath511 , such that if then @xmath512 lies on the same side of @xmath513 as @xmath514 .
however , this implies that there exists such that @xmath515 intersects the interior of the face of @xmath1 contained in @xmath516 .
since @xmath515 bounds a half - space @xmath517 disjoint from @xmath260 , this is a contradiction .
therefore , is a disjoint family of geodesics .
if @xmath506 lies in @xmath492 , then there is a support plane @xmath518 separating some point on @xmath519 from a point on @xmath506 .
this is a contradiction .
therefore no edge @xmath506 with lies in @xmath492 and so is a disjoint family of geodesics on @xmath492 .
it follows that @xmath70 intersects each edge exactly once , since otherwise there would be distinct values such that and hence let @xmath520 be the exterior dihedral angle between @xmath492 and @xmath516 let be the dihedral angle between @xmath521 and @xmath516 for all let @xmath120 be the maximal value of @xmath287 for an edge @xmath226 and let be the exterior dihedral angle between @xmath492 and @xmath493 and let denote the dihedral angle between @xmath522 and @xmath523 .
we let @xmath524 for @xmath525 and @xmath526 .
now for we let @xmath85 be the unique plane perpendicular to the planes @xmath492 ,
@xmath527 , and @xmath528 ( see figure [ mono ] ) . considering the triangle @xmath529 on @xmath85 given by the intersection with the three planes , we see that @xmath530 then , by induction , @xmath531 as @xmath532 , we obtain @xmath533 ( ( * ? ? ?
* lemma 3.1 ) ) [ isosceles ] suppose that @xmath529 is a hyperbolic triangle with a side of length @xmath252 and opposite angle equal to @xmath176 .
then the triangle has perimeter at most @xmath534 and this maximal perimeter is realized uniquely by an isosceles triangle .
[ arcanglebound ] ( ( * ? ? ?
* lemma 3.2 ) ) let @xmath535 , @xmath536 and @xmath537 be planes in @xmath78 such that @xmath538 and @xmath539 are tangent in @xmath3 and @xmath539 and @xmath540 are also tangent in @xmath3 . if @xmath541 and @xmath542\to { \mathbb{h}^3}$ ] is a curve of length at most @xmath45 which intersects all three planes , then @xmath535 and @xmath537 intersect with interior dihedral angle @xmath543 , where @xmath544 [ trianglefact ] let @xmath535 , @xmath536 and @xmath537 be planes in @xmath78 such that @xmath538 and @xmath539 are tangent in @xmath3 and @xmath539 and @xmath540 are also tangent in @xmath3 . if @xmath535 and @xmath537 intersect at an interior angle of @xmath543 and @xmath542\to { \mathbb{h}^3}$ ] is a closed curve which intersects all three planes , then @xmath545 , where @xmath546 let @xmath85 be the plane which intersects @xmath535 , @xmath536 and @xmath537 orthogonally and let @xmath529 be the triangle they bound in @xmath85
. if @xmath547 is the projection of @xmath430 onto @xmath85 , then and @xmath547 intersects all three sides of @xmath529 .
therefore , it suffices to bound the length of a curve in @xmath85 which intersects all three sides of @xmath529 . since @xmath529 is convex it suffices to consider curves which lie in @xmath529 .
let @xmath548 be the non - ideal vertex of @xmath529 and let @xmath549 be length of the perpendicular @xmath11 which joins a point @xmath72 in the opposite side of @xmath529 to @xmath548 .
then , by hyperbolic trigonometry , e.g. ( * ? ? ?
* theorem 7.11.2 ) , @xmath550 in @xmath118 we take two copies @xmath551 and @xmath552 of @xmath529 and glue @xmath551 to left edge of @xmath529 containing @xmath548 and @xmath552 to the right edge containing @xmath548 .
we label the points in the copies corresponding to @xmath72 by @xmath553 and @xmath554 .
we can join @xmath553 and @xmath554 by edges of length @xmath549 to @xmath548 which subtend an angle @xmath555 at @xmath548 in the union of the three triangles ( see figure [ triangle ] ) . if we join the points @xmath553 and @xmath556 by a geodesic @xmath267 in the union of the three copies and let then we have an isoceles triangle with vertices @xmath553 , @xmath556 , and @xmath548 and two sides of length @xmath549 which make an angle @xmath555 .
we may decompose this triangle into two right angled triangles with hypotenuse of length @xmath549 and again apply ( * ? ? ?
* theorem 7.11.2 ) to show that @xmath557 if @xmath558 , then we join @xmath553 to @xmath556 by a path @xmath267 of length @xmath559 made from copies of @xmath11 in @xmath551 and @xmath552 . in either case , @xmath267 projects to a closed curve in @xmath529 meeting all three sides . to complete the proof , we must show that @xmath267 projects to the minimal length such curve .
if @xmath560 is the minimal curve , then we may unfold it to a geodesic joining a point @xmath561 to the side of @xmath551 containing @xmath553 to a point @xmath562 in the side of @xmath552 containing @xmath556 .
moreover , if we join @xmath561 and @xmath562 to @xmath548 by geodesics , the two geodesics each have length at least @xmath549 and make an angle of @xmath555 .
it follows that the length of @xmath560 is at least as great as that of @xmath267 , with equality if and only if @xmath563 .
* lemma [ thickintbound ] * _ if @xmath2 is the complement of a finite collection @xmath254 of points in @xmath3 and @xmath182 \rightarrow \chb$ ] is a geodesic path with @xmath564 for some @xmath565 $ ] , then @xmath239 _ * proof : * our assumptions imply that there exists @xmath566 $ ] such that @xmath567 ( since @xmath568 ) .
we may assume that @xmath569)\le l_h(\alpha)/2 $ ] .
( if this is not the case we simply consider @xmath570\to \chb$ ] given by @xmath571 and notice that @xmath572 . )
we first handle the case where @xmath1 is contained in a plane . in this case , if @xmath573 , then there exists a subarc @xmath574 containing @xmath575 with @xmath576 . recall that @xmath1 is the double of @xmath260 in this case .
both endpoints of @xmath574 must lie in the same `` copy '' of @xmath260 .
one may then join the endpoints of @xmath574 by a geodesic arc @xmath430 in this copy , so that @xmath577 is homotopically non - trivial in @xmath1 and @xmath578 .
it follows that @xmath579 which is a contradiction . * case 3 : * if we are not in case ( 1 ) or ( 2 ) , we let @xmath590 be the first support plane such that consider the configuration of three planes @xmath535 , @xmath583 , and @xmath590 .
let be the subarc of @xmath70 of minimum length having as a family of support planes .
let @xmath72 and @xmath73 be the endpoints of @xmath574 .
lemma [ arcanglebound ] implies that @xmath535 and @xmath590 intersect and have dihedral angle @xmath543 satisfying @xmath591 we join @xmath72 to @xmath73 by a shortest path @xmath430 on which intersects at a point we take the triangle @xmath529 with vertices @xmath72 , @xmath73 , and @xmath82 .
then the interior angle @xmath592 of @xmath529 at @xmath82 satisfies since @xmath72 and @xmath73 lie in @xmath574 , then , by lemma [ isosceles ] , @xmath593 substituting the bound for @xmath543 we get @xmath594 if @xmath577 , then @xmath595 we next show that @xmath596 is homotopically non - trivial on @xmath1 .
since the restriction of @xmath28 to is a homotopy equivalence onto @xmath1 , it suffices to show that @xmath160 is homotopically non - trivial in let @xmath597 be the first point on @xmath70 which has support plane @xmath583 . then @xmath598 lies on an edge @xmath190 of @xmath1 which is contained in @xmath583 .
let @xmath599 be the plane through @xmath598 which is perpendicular to @xmath583 and let @xmath600 be the intersection of @xmath599 with the closed half - space bounded by @xmath583 which is disjoint from @xmath601 .
lemma [ sptbnd ] implies that @xmath574 intersects @xmath598 exactly once at @xmath602 and the intersection is transverse . by construction
, @xmath430 can not intersect @xmath600 and @xmath574 can not intersect @xmath603 .
therefore , @xmath160 intersects @xmath600 exactly once and does so transversely . since @xmath600 is a properly embedded half - plane in we see that @xmath160 is homotopically non - trivial in as desired .
since @xmath28 is 1-lipschitz on @xmath78 , @xmath604 in particular , since @xmath596 is homotopically non - trivial and passes through @xmath220 , we see that @xmath605 for all @xmath606 .
if , i.e. if then we have achieved a contradiction and we are done . if @xmath575 does not lie in @xmath574 , then let then is a homotopically non - trivial loop through @xmath575 such that @xmath607 since @xmath608 we have again achieved a contradiction . * lemma [ shortgeointbound ] * _ suppose that @xmath2 is the complement of finite collection @xmath254 of points in @xmath3 .
if @xmath176 is a simple closed geodesic of length less than @xmath249 , then @xmath275 _ for the remainder of the proof , we assume that @xmath260 does not lie in a plane .
we choose a parameterization of @xmath176 which is an embedding on @xmath610 and such that does not lie on an edge of @xmath1 .
recall that lemma [ intlowerbound ] implies that let be the full family of support planes to @xmath70 .
if we proceed as in the proof of lemma [ thickintbound ] , we must be in case ( 3 ) .
therefore , we obtain support planes @xmath535 , @xmath583 and @xmath590 , such that @xmath538 and @xmath611 are tangent on @xmath3 , as are @xmath583 and @xmath590 , and a subarc @xmath574 of @xmath70 which begins at ends at and intersects @xmath583 .
lemma [ arcanglebound ] implies that @xmath535 and @xmath590 intersect .
lemma [ trianglefact ] implies that if @xmath543 is the angle of intersection of @xmath535 and @xmath590 , then as @xmath612 is decreasing and we see that let @xmath616 be the shortest path on joining @xmath220 to @xmath602 and let @xmath617 be the point of intersection of @xmath616 with since @xmath618 lemma [ isosceles ] implies that @xmath619 so @xmath620 since @xmath621 and @xmath291 , we see that @xmath622 it follows that @xmath623 .
we let we may argue , exactly as in the proof of lemma [ shortgeointbound ] , that @xmath624 is homotopically non - trivial on @xmath1 . since and @xmath245 has width we see that @xmath624 is contained in @xmath245 , so is homotopic to a non - trivial power of @xmath176 .
we now describe a process to replace @xmath616 by a path @xmath625 on @xmath1 such that and @xmath625 is homotopic to @xmath626 .
once we have done so , we can form and observe that @xmath627 our result follows , since @xmath628 is homotopic to a non - trivial power of @xmath176 and a geodesic minimizes intersection number in its homotopy class .
otherwise @xmath616 intersects an edge on @xmath631 labelled @xmath472 , such that @xmath630 and @xmath472 are disjoint .
we choose @xmath472 to be the last edge on @xmath632 that @xmath70 intersects .
let @xmath633 be the first point of intersection of @xmath616 and @xmath472 .
let @xmath188 be the face of @xmath1 which abuts @xmath472 and does not contain a subarc of @xmath616 .
let @xmath634 be the support plane containing @xmath188 .
define @xmath635 .
since @xmath630 and @xmath636 are both contained in support planes to @xmath472 , they are disjoint . since @xmath637 , @xmath630 must separate @xmath636 from @xmath220 in @xmath535 . therefore , @xmath616 intersects @xmath636 at a point @xmath638 between @xmath617 and @xmath220 .
we construct @xmath639 by replacing the subarc of @xmath616 joining @xmath633 to @xmath638 by the geodesic arc in @xmath188 joining @xmath633 to @xmath638 .
notice that @xmath640 is homotopic to @xmath626 in @xmath1 . if we let @xmath641 be the exterior dihedral angle between @xmath632 and @xmath634 and let @xmath642 be the exterior dihedral angle between @xmath634 and @xmath535 , then @xmath643 if @xmath636 is not an edge , we define @xmath625 by an iterative procedure . assume @xmath645 are defined as above , for such that @xmath646 is not an edge .
define @xmath647 to be the exterior dihedral angle between @xmath648 and @xmath645 and @xmath649 to be the exterior dihedral angle between @xmath645 and @xmath535 .
we assume , by induction , that @xmath650 let @xmath317 be the last edge on @xmath651 which @xmath652 intersects , and let @xmath653 be the first point of intersection of @xmath652 and @xmath317 .
let @xmath654 be the face adjacent to @xmath292 which does not agree with @xmath655 and let @xmath656 be the support plane it is contained in . as before ,
if we let then @xmath657 separates @xmath317 from @xmath220 .
therefore , @xmath652 intersects @xmath657 in a point @xmath658 .
we construct @xmath659 by replacing the piecewise geodesic subarc of @xmath652 connecting @xmath660 to @xmath658 by the geodesic arc in @xmath654 joining @xmath660 to @xmath658 . again , notice that @xmath661 is homotopic to @xmath662 .
furthermore , @xmath663 therefore , @xmath664 notice that the support planes @xmath665 in this procedure have disjoint intersection with @xmath535 and each contain a face @xmath666 .
therefore , as there are only a finite number of faces , the process terminates and some @xmath667 is an edge of @xmath1 . then , @xmath625 lies on @xmath1 and our argument is complete .
epstein and a. marden , `` convex hulls in hyperbolic space , a theorem of sullivan , and measured pleated surfaces , '' in _ analytical and geometrical aspects of hyperbolic space _
, cambridge university press , 1987 , 113253 .
sullivan , `` travaux de thurston sur les groupes quasi - fuchsiens et les varits hyperboliques de dimension @xmath669 fibres sur @xmath435 , '' in _ bourbaki seminar , vol .
1979/1980 _ , _ lecture notes in math .
_ * 842*(1981 ) , 196214 . | we show that the nearest point retraction is a uniform quasi - isometry from the thurston metric on a hyperbolic domain @xmath0 to the boundary @xmath1 of the convex hull of its complement . as a corollary ,
one obtains explicit bounds on the quasi - isometry constant of the nearest point retraction with respect to the poincar metric when @xmath2 is uniformly perfect .
we also establish marden and markovic s conjecture that @xmath2 is uniformly perfect if and only if the nearest point retraction is lipschitz with respect to the poincar metric on @xmath2 . |
let @xmath9 be the simple random walk on @xmath0 , @xmath1 .
we write @xmath10 for the law of the random walk starting from position @xmath11 , and @xmath12 for the associated expectation .
independently of @xmath13 , we give ourselves a family @xmath3 of independent random variables , which we may call the _ potential _ , or also the _ environment_. these random variables are distributed according to a common probability measure @xmath14 on @xmath4 .
we write @xmath15 for their joint distribution , and @xmath16 for the associated expectation .
let @xmath17 be a vector of unit euclidian norm , and @xmath18 be the first time at which the random walk crosses the hyperplane orthogonal to @xmath19 lying at distance @xmath20 from the origin .
our main goal is to study the _ quenched _ and _ annealed _ point - to - hyperplane _ lyapunov norms _ ( also called _ lyapunov exponents _ ) , defined respectively by @xmath21,\ ] ] @xmath22,\ ] ] for @xmath5 tending to @xmath7 ( the first limit holds almost surely ; see @xcite for proofs that these exponents are well defined ) .
intuitively , these two quantities measure the cost , at the exponential scale , of travelling from the origin to a distant hyperplane , for the random walk penalized by the potential @xmath2 .
the quenched lyapunov norm is a measure of this cost in a typical random environment , while the annealed lyapunov norm measures this cost after averaging over the randomness of the environment .
these norms are related to the point - to - point lyapunov norms by duality , and to the large deviation rate function for the position of the random walk at a large time under the weighted measure ( see @xcite for details ) .
recently , @xcite studied this question under the additional assumption that @xmath23 $ ] is finite ( where we write @xmath23 $ ] as shorthand for @xmath24 $ ] ) .
they found that , as @xmath6 tends to @xmath7 , @xmath25}\ ] ] ( and they showed that this relation also holds for @xmath26 ) .
this means that when @xmath23 $ ] is finite , the first - order asymptotics of the lyapunov exponents are the same as if the potential were non - random and uniformly equal to @xmath23 $ ] .
our goal is to understand what happens when we drop the assumption on the integrability of the potential . from now on , @xmath27 = + \infty,\ ] ] and
write @xmath28.\ ] ] here is our main result .
[ t : estimen ] let @xmath29 .
there exists @xmath30 such that for any @xmath6 small enough and any @xmath20 , @xmath31 where @xmath32 and @xmath33 is the probability that the simple random walk never returns to its starting point , that is , @xmath34.\ ] ] this result is a first step towards a proof that , as @xmath6 tends to @xmath7 , @xmath35 one can check that @xmath36 , and theorem [ t : estimen ] provides the adequate lower bound on @xmath37 for to hold . in order to complete the proof of , there remains to provide a matching upper bound for @xmath38 .
this will be done in a companion paper .
[ r : integr ] let us write @xmath39 so that @xmath40 it is easy to see that @xmath41 is concave increasing and that @xmath42 as @xmath43 tends to @xmath7 . as a consequence , for any @xmath44 , @xmath45,\ ] ] while , since @xmath46 , @xmath47.\ ] ] when @xmath23 $ ] is finite , we thus obtain that the right - hand sides of ( [ resultatkmz ] ) and ( [ conject ] ) are equivalent as @xmath6 tends to @xmath7 , and thus holds indeed in this case .
an interesting feature of and is that their right - hand sides do not depend on @xmath19 . in other words ,
asymptotically , the balls associated to the quenched and annealed lyapunov norms look like scaled euclidian balls .
the main motivation behind @xcite was related to questions concerning the spectrum of the discrete anderson hamiltonian @xmath48 , where @xmath49 is the discrete laplacian : @xmath50 powerful techniques have been devised to transfer finite - volume estimates on the green function of @xmath51 within some energy interval into information on the spectrum in this interval ( see @xcite for the multiscale method , and @xcite for the fractional - moment approach ) .
for instance , it is known that for any @xmath5 , the spectrum of @xmath51 is pure point in a neighbourhood of @xmath7 and corresponding eigenfunctions are exponentially localized . in @xcite , extending the techniques developed in @xcite , the author gave quantitative estimates on the green function within an explicit energy interval at the edge of the spectrum , as @xmath6 tends to @xmath7 .
these were then refined in @xcite .
these results imply in particular that if @xmath23 $ ] is finite and the distribution @xmath14 is absolutely continuous with respect to the lebesgue measure , then for any @xmath52 and any @xmath6 small enough , the spectrum of @xmath51 is pure point in the interval @xmath53-\beta\eta]$ ] , with exponentially decaying eigenfunctions .
theorem [ t : estimen ] can be seen as a first step towards a study of these questions in the case when the potential is not assumed to be integrable .
we conjecture that when this integrability condition is dropped , the upper energy @xmath54 $ ] appearing in the above result should be relaced by @xmath55 the reason why this is the natural integral to consider will be explained in section [ s : extension ] .
we now give a heuristic description of the typical scenario responsible for the behaviour of @xmath56 described in theorem [ t : estimen ] .
different strategies can be used to reduce the cost of travel to the distant hyperplane .
( 1 ) one approach is to reach the hyperplane in a small number of steps .
( 2 ) a second approach is to avoid sites where @xmath57 is too large , or else , to try not to return to such sites too many times .
naturally , one should look for the optimal strategy as a combination of these two methods . in order to quantify method ( 1 )
, one can observe that , for small @xmath58 , @xmath59 \approx \frac{d v}{2 } \
n.\ ] ] the quantity @xmath58 represents the velocity of the particle . on the other hand , roughly speaking , we will show that , for small @xmath58 , @xmath60 \\
\gtrsim { \mathfrak}{i}{_\beta}\ \frac{n}{v},\ ] ] which quantifies the gains obtained by method ( 2 ) .
assuming that these observations hold , theorem [ t : estimen ] can be derived by optimizing @xmath58 so that the sum of the costs in ( [ met1 ] ) and ( [ met2 ] ) is minimized .
this is achieved choosing @xmath61 let us explain the meaning of .
recall that for any @xmath62 , holds .
relation shows that sites whose potential is bounded by @xmath62 contribute as if they were replaced by their expectation . in other words , for these sites , method ( 2 ) is simply too costly to be effective , and we may say that these sites are in a `` law of large numbers '' regime .
in fact , in this reasoning , we could allow @xmath62 to grow with @xmath6 , as long as it remains small compared to @xmath63 .
the picture changes when we consider sites whose potential is very large compared to @xmath63 .
observe that the number of distinct sites visited by the random walk at time @xmath64 grows like @xmath65 , where @xmath66 is the mean number of visits to any point visited ( conditionally on the event @xmath67 , this is true in the limit of small @xmath58 ) . under the annealed measure , the sequence of potentials attached to the distinct sites visited forms a sequence of i.i.d .
random variables with common distribution @xmath14 .
the cost of not meeting any site whose potential lies in the interval @xmath68 up to time @xmath67 is thus approximately @xmath69 when @xmath62 is large , since @xmath70 as @xmath43 tends to infinity , this is roughly @xmath71 and thus the strategy concerning sites whose potential is much larger than @xmath63 is simple to describe : simply avoid them . to sum up , formula ( [ met2 ] )
reveals the following picture .
sites whose potential is much smaller than @xmath63 stay in a law of large numbers regime .
sites whose potential is much greater than @xmath63 are simply avoided .
now , for sites whose potential is of order @xmath63 , an adequate intermediate strategy is found .
heuristically , for sites whose potential is roughly @xmath72 , the strategy consists in ( i ) lowering the proportion of such sites that are visited by a factor @xmath73 ; ( ii ) once on such a site , to go back to it with a probability @xmath74 instead of @xmath75 .
the picture described above , and in particular , must however be taken with a grain of salt .
our basic approach relies on coarse - graining arguments .
we identify _ good _ boxes , which are such that we understand well the cost and the exit distribution of a coarse - grained displacement of the walk started from a point in a good box . in our arguments
, we do not try to control what happens when a coarse - grained piece of trajectory starts within a _ bad _ box .
as was noted in @xcite , this is indeed a delicate matter , since the time spent in bad boxes does not have finite exponential moments in general .
instead , we introduce a _ surgery _ on the trajectories .
the surgery consists in removing certain _ loops _ , which are pieces of coarse - grained trajectory that start and end in the same bad box .
we show rigorous versions of and , where @xmath76 is replaced by the total time spent outside of these loops ; from these estimates , we then derive theorem [ t : estimen ] . * related works .
* we already mentioned @xcite and the connection with anderson localization . in @xcite
, it is proved that under the annealed weighted measure , the random walk conditioned to hit a distant hyperplane satisfies a law of large numbers ( see also @xcite ) .
it would be interesting to see whether their techniques can be combined with our present estimate to show that indeed , the right - hand side of gives the asymptotic behaviour of the speed as @xmath6 tends to @xmath7 ( our results do not show this directly , due to the surgery on paths discussed above ) .
another motivation relates to recent investigation on whether the disorder is _ weak _ of _ strong_. the disorder is said to be weak if the quenched and annealed lyapunov exponents coincide . to our knowledge , this question has only been investigated for potentials of the form @xmath77 with @xmath78 , see @xcite for weak disorder results when @xmath79 and @xmath6 is small , and @xcite for strong disorder results when @xmath80 .
this additional @xmath78 is very convenient since it introduces an effective drift towards the target hyperplane ( indeed , the problem can be rewritten in terms of a drifted random walk in the potential @xmath2 using a girsanov transform ) . in particular
, the asymptotic speed of travel to the hyperplane remains bounded away from @xmath7 in this case .
one of our motivations was to get a better understanding of the behaviour of the walk when we set @xmath81 .
of course , showing that the lyapunov exponents are equivalent as @xmath6 tends to @xmath7 does not touch upon the question whether they become equal for small @xmath6 or not .
recently , a continuous - space version of @xcite was obtained in @xcite . there
, the author investigates brownian motion up to reaching a unit ball at distance @xmath20 from the starting point , and evolving in a potential formed by locating a given compactly supported bounded function @xmath82 at each point of a homogeneous poisson point process of intensity @xmath83 .
it is shown that if @xmath84 for some constant @xmath85 , then the quenched and annealed lyapunov exponents are both asymptotically equivalent to @xmath86 .
* organization of the paper .
* as was apparent in the informal description above , the most interesting phenomena occur for sites whose associated potential is of the order of @xmath63 .
section [ s : lowerbd ] adresses this case , and proves theorem [ t : estimen ] with @xmath87 replaced by @xmath88 where @xmath89 is arbitrary .
this is however not sufficient to prove theorem [ t : estimen ] in full generality , since for some distributions [ and although we always assume ] , it may be that whatever @xmath89 , the integral @xmath90 is too small compared to @xmath87 as @xmath6 tends to @xmath7 . in other words , there are cases for which the integral @xmath91 can not be neglected , even if we are free to choose @xmath89 beforehand . in the very short section [ s : intermediate ] ,
we take a specific choice for @xmath92 and distinguish between three cases , depending on whether @xmath90 , @xmath93 , or both integrals have to be considered .
section [ s : intermediate2 ] tackles the most delicate case when both integrals must be accounted for .
section [ s : intermediate_only ] concludes the proof of theorem [ t : estimen ] , covering the case when @xmath90 is negligible compared to @xmath93 .
finally , section [ s : extension ] presents natural extensions of theorem [ t : estimen ] , and spells out the link with the green function of the operator @xmath8 .
* notations .
* we write @xmath94 for the euclidian norm on @xmath95 . for @xmath96 and @xmath97 ,
let @xmath98 be the ball of centre @xmath11 and radius @xmath99 .
for @xmath100 and @xmath101 , we call @xmath102 of centre @xmath11 and size @xmath99 the set defined by @xmath103 for @xmath104 , we write @xmath105 to denote the cardinality of @xmath106 .
let @xmath29 . this @xmath107 will play the role of the allowed margin of error in our subsequent reasoning .
we will need to assume that is is sufficiently small ( and we may do so without mentioning it explicitly ) , but it will be kept fixed throughout . as a start , we fix @xmath89 and focus on sites whose associated potential lies in the interval @xmath108 .
we want to approximate the integral @xmath109 by a riemann sum ( recall that the function @xmath41 was defined in ( [ deff ] ) ) .
let @xmath110 be a positive integer and @xmath111 be such that for every @xmath112 , one has @xmath113 .
this provides us with a subdivision of the interval @xmath114 , and @xmath115 from this subdivision , we extract those intervals which have a non - negligible weight : @xmath116 we let @xmath117 denote the cardinality of @xmath118 , and let @xmath119 be such that @xmath120 although @xmath121 , @xmath122 and @xmath123 may depend on @xmath6 , we keep this dependence implicit in the notation .
( since @xmath121 remains bounded by @xmath110 , this dependence will not be a problem . ) noting that @xmath124 and that @xmath41 is bounded by @xmath125 , we obtain that @xmath126 and thus , letting @xmath127 we are led to @xmath128 up to a multiplicative error controlled by @xmath107 , we can thus consider @xmath129 as a good approximation of @xmath90 .
we let @xmath130 we call elements of @xmath131 _ important sites_. a relevant length scale of our problem is @xmath132 defined by @xmath133.\ ] ] this scale is interesting since it is such that , if the random walk runs up to a distance @xmath132 from its starting point , then it meets roughly one important site .
clearly , @xmath132 tends to infinity as @xmath6 tends to @xmath7 .
let also @xmath134 = { \mathbb{p}}{\left}[0 \in t_l { \right}].\ ] ] although this is not explicit in the notation , @xmath135 , @xmath136 , and @xmath131 depend on @xmath6 .
note that @xmath137 we may write @xmath138 to denote the fact that there are constants @xmath139 such that @xmath140 . similarly ,
we have @xmath141 so @xmath142 .
although not really necessary , arguments developed in section [ s : intermediate2 ] will be clearer if instead of @xmath132 , we choose from now on @xmath143 such that @xmath144 as our length scale of reference , so that @xmath145 . note that @xmath146 .
let @xmath147 be a positive integer , which we refer to as the _ mesoscopic scale_. we define a coarse - graining of the trajectory at this scale .
that is , we let @xmath148 , and define recursively @xmath149 we will need to say that , most of the time , the values of the potential around the position of the random walk are `` typical '' . for @xmath150 ,
let @xmath151 .
the boxes @xmath152 form a partition of @xmath0 at the mesoscopic scale . roughly speaking
, we will ask a `` nice '' box to contain sufficiently important sites that are not too close from one another , and are evenly spread across the box . in order to make this informal description precise ,
we introduce two additional scales @xmath153 such that @xmath154 .
we ask that we can partition a box @xmath155 of size @xmath147 by subboxes of size @xmath156 ( that is , we ask @xmath157 to be a multiple of @xmath158 ) , and write @xmath159 for the partition of @xmath160 into subboxes of size @xmath156 .
similarly , we ask that any box @xmath161 of size @xmath156 can be partitioned into subboxes of size @xmath162 , and write @xmath163 for this partition .
let @xmath164 , @xmath165 , and @xmath166 .
if there exists @xmath167 such that @xmath168 , and if moreover @xmath169 is the only important site inside @xmath170 , then we define @xmath171 . otherwise , we set @xmath172 . in other words
, we have @xmath173 the value of @xmath174 is chosen so that @xmath175 @xmath176 observe that the event that @xmath160 is balanced depends only on the values of the potential inside the box @xmath160 .
we say that the box @xmath160 is _ good _ if for any @xmath177 such that @xmath178 , the box @xmath179 is balanced ; we say that it is a _ bad _ box otherwise . the construction is summarized in figure [ f : boxes ] .
first , we want to ensure that the walk @xmath13 does not visit any important site during a typical mesoscopic displacement . that is to say , we want @xmath180 ( by this , we mean @xmath181 as @xmath182 ) .
while of course we ask @xmath183 , we will need to have a growing number of important sites inside the intermediate boxes of size @xmath156 . in order for this to be true , we need @xmath184 . on the contrary ,
we want a typical box of size @xmath162 to contain no important site , that is , @xmath185 . to sum up , we require @xmath186 it is convenient to make a specific choice regarding these scales . for @xmath187 , we define @xmath188 and fix @xmath189 small enough so that ( [ compscales ] ) holds ( this is possible since @xmath1 ) .
an additional requirement on the smallness of @xmath189 will be met during the proof , see .
thus defined , the scales may not satisfy our requirement that @xmath157 be a multiple of @xmath158 , and that @xmath190 be a multiple of @xmath191 .
it is however easy to change the definitions slightly while preserving the asymptotics of the scales , and we will thus no longer pay attention to this problem .
we start by recalling some classical large deviations results about bernoulli random variables . for @xmath192 $ ] and @xmath193 $ ] , we define @xmath194 \\ & = & ( p+\eta ) \log{\left } ( \frac{p+\eta}{p } { \right } ) + ( 1-p-\eta ) \log{\left } ( \frac{1-p-\eta}{1-p } { \right } ) \notag.\end{aligned}\ ] ] we let @xmath195 when @xmath196 $ ]
. if @xmath197 are independent bernoulli random variables of parameter @xmath198 under the measure @xmath199 ( with the convention @xmath200 $ ] ) , then for any @xmath201 , one has @xmath202 { \leqslant}e^{-n \psi_p(\eta)},\ ] ] and @xmath203 { \leqslant}e^{-n \psi_p(-\eta)}.\ ] ] a simple calculation shows that , for @xmath193 $ ] , @xmath204 @xmath205 we now proceed to show that the probability for a box to be good is very close to @xmath206 .
[ l : subsubbox ] for @xmath207 a box of size @xmath162 , @xmath166 and @xmath6 small enough , one has @xmath208 { \geqslant}{\left}(1-\frac{3{\varepsilon}}{4}{\right } ) p_l \ |b'|,\ ] ] where @xmath135 was defined in . without loss of generality
, we may assume that the box @xmath207 is centred at the origin . by the inclusion - exclusion principle
, we have @xmath209 \\ { \geqslant}\sum_{x \in b(0,(1-{\varepsilon}_1 ) r{_\beta } ' ) } { \mathbb{p}}[x \in t_l ] - \sum_{\substack{x , y \in b(0,(1-{\varepsilon}_1 ) r{_\beta } ' ) \\ x < y } } { \mathbb{p}}[x , y \in t_l ] , \end{gathered}\ ] ] where , say , @xmath210 refers to lexicographical order .
the first sum is equal to @xmath211 while the second sum in is bounded by @xmath212 .
we have seen in ( [ comparlbp ] ) that @xmath213 , and since @xmath185 , we have that @xmath214 tends to @xmath7 as @xmath6 tends to @xmath7 , so the right - hand side of ( [ e : subsubbox1 ] ) is equal to @xmath215 to conclude , it suffices to show that the probability of having two or more important sites within @xmath170 is negligible ( see ) .
but this is true since the probability is bounded by @xmath216 .
[ p : subbox ] let @xmath161 be a box of size @xmath156 , and @xmath166 .
there exists @xmath217 ( depending only on @xmath107 ) such that for any @xmath6 small enough , @xmath218 { \leqslant}e^{-c_0 p_l |b|}.\ ] ] from lemma [ l : subsubbox ] and the fact that @xmath219 are independent random variables , we know that @xmath219 stochastically dominate i.i.d . bernoulli random variables with parameter @xmath220 we are thus in the situation of inequality , with @xmath221 the proposition follows using the observation made in . [ p : probbadbox ] there exists @xmath222 such that for @xmath6 small enough and any @xmath223 , @xmath224 { \leqslant}\frac{|b_i|}{|b| } \
e^{-c_1 l{_\beta}^{-2 } \
|b|},\ ] ] where @xmath161 is a box of size @xmath156 . considering the definition of balanced boxes given in , one obtains the result with lemma [ p : subbox ] , a union bound and the fact that @xmath213 .
we say that @xmath225 are @xmath226__-neighbours _ _ , and write @xmath227 , if @xmath228 .
we say that a subset of @xmath0 is @xmath226__-connected _ _ if it is connected for this adjacency relation .
we call _ lattice animal _ a subset of @xmath0 which is @xmath226-connected and contains the origin
. we will be interested in the set of @xmath229 s such that @xmath160 is visited by the coarse - grained trajectory , which indeed forms a lattice animal if the walk is started at the origin .
the next proposition is similar to an argument found in @xcite .
[ animals ] recall the definition of the scales in terms of @xmath189 given in , and let @xmath230 . for any @xmath6 small enough
, @xmath231 { \leqslant}e^{-n}. \ ] ] it suffices to show that , for some @xmath232 and for @xmath6 small enough , @xmath233 { \leqslant}\exp{\left}(-c l{_\beta}^{\delta/4 } n{\right}).\ ] ] first , as observed in @xcite , there are at most @xmath234 lattice animals of cardinality @xmath235 ( to see this , one can encode the lattice animal by a @xmath226-nearest - neighbour trajectory starting from the origin , of length at most @xmath236 , that is the `` depth - first search '' of a spanning tree of the lattice animal ) .
now , given @xmath237 and a lattice animal @xmath106 of cardinality @xmath235 , the probability @xmath238\ ] ] is of the form of the left - hand side of ( [ bernoullige ] ) , with , according to proposition [ p : probbadbox ] , @xmath239 { \leqslant}\frac{|b_i|}{|b| } \
e^{-c_1 l{_\beta}^{-2 } \ |b|},\ ] ] and @xmath240 with these parameters , we obtain that @xmath241 .
recalling that @xmath242 , we infer that for some @xmath232 and @xmath6 small enough , the probability in ( [ e : animals2 ] ) is smaller than @xmath243 . to conclude , note that an unbalanced box can be responsible for no more than @xmath244 bad boxes .
hence , choosing @xmath245 , we arrive at @xmath246 \\ { \leqslant}{\mathbb{p}}\big [ { \left}|
\ { i \in a : b_i \text { is not balanced}\ } { \right}| { \geqslant}c_2 \eta{_\beta}n \big ] { \leqslant}\exp{\left}(-c l{_\beta}^{\delta/2 } n { \right}).\end{gathered}\ ] ] multiplying this by our upper bound @xmath234 on the number of lattice animals , we have thus bounded the probability in the left - hand side of ( [ e : animals0 ] ) by @xmath247 this proves that holds for @xmath6 small enough , and thus finishes the proof .
in fact , the definition in of a balanced box asks that important sites are `` nowhere missing '' , but it may happen that they are in excess . since we want to bound from below the sum of the @xmath248 s seen by the random walk , this should not be a problem .
however , for the purpose of the proof , it will be convenient to extract a selection of important sites which will not be too numerous .
let @xmath249 . by definition , @xmath171 if and only if @xmath250 in this case , we let @xmath251 be the unique element of these sets .
given a balanced box @xmath160 , and @xmath252 , we know that the cardinality of the set @xmath253 is at least @xmath254 .
we choose in some arbitrary deterministic manner a subset @xmath255 of @xmath256 whose cardinality lies in the interval @xmath257 $ ] ( this is possible for @xmath6 small since @xmath258 tends to infinity as @xmath6 tends to @xmath7 ) .
we further define @xmath259 note that any two elements of @xmath260 are at a distance at least @xmath261 from one another ( see figure [ f : boxes ] ) .
we define @xmath262 and @xmath263 clearly , this last quantity is a lower bound on the `` cost '' of the first piece of the coarse - grained trajectory .
the advantage of having dropped some important sites as we just did is this .
now , we will be able to show that if the walk starts within a good box , then with high probability the quantity @xmath264 is simply @xmath7 , and the probability that two or more sites in @xmath265 actually contribute to the sum @xmath264 is negligible .
moreover , any important point contributing to @xmath264 is far enough from the boundary of @xmath266 so that if visited , the returns to this site will most likely occur before exiting @xmath266 .
see figure [ f : annulus ] for an illustration of the construction .
[ p : laplaces ] if @xmath6 is small enough , then any @xmath100 lying in a good box satisfies @xmath267 { \leqslant}(1+{\varepsilon } ) q_d \ r{_\beta}^2 \ l{_\beta}^{-2}\ ] ] and @xmath268 { \leqslant}1-(1 - 3{\varepsilon } ) r{_\beta}^2\ i{_\beta},\ ] ] where @xmath33 was defined in and @xmath129 in .
the most important step is contained in this lemma .
[ l : laplaces ] for @xmath100 and @xmath166 , define @xmath269.\ ] ] if @xmath6 is small enough , then any @xmath11 lying in a good box satisfies @xmath270 for greater clarity , we assume @xmath271 , and comment on the necessary modifications to cover arbitrary @xmath11 at the end of the proof .
consider the set @xmath272 elements of @xmath273 are of the form @xmath274 , with @xmath275 .
in particular , any point contained in @xmath276 is at distance at least @xmath156 from the origin .
a lower bound on @xmath277 is @xmath278.\ ] ] since by assumption @xmath7 lies in a good box , any @xmath276 belongs to a balanced box , and thus @xmath255 is well defined , and moreover , @xmath279 $ ] .
we learn from ( * ? ? ?
* lemma a.2 ( 147 ) ) that @xmath280 { \geqslant}{c_d } \ { q_d}{\left}(|y|^{2-d } - r{_\beta}^{2-d}{\right } ) + o{\left}(|y|^{1-d}{\right}),\ ] ] where @xmath281 is the euclidian norm of @xmath169 , @xmath282 and @xmath283 is euler s gamma function .
let @xmath284 denote the point of the box @xmath161 which is the furthest from the origin , and @xmath285 be the closest ( with respect to the euclidian norm , and with some deterministic tie - breaking rule ) . using the lower bound in , we arrive at @xmath286.\ ] ]
we first show that the error term is negligible , that is , @xmath287 let @xmath288 the left - hand side of ( [ werror ] ) is equal to @xmath289 on one hand , we have @xmath290 where @xmath291 is the surface area of the unit sphere in @xmath95 . on the other hand , @xmath292 so that holds indeed .
the same type of argument shows that @xmath293 \sim r{_\beta}^2 \int_{|z| { \leqslant}1 } { \left } [ |z|^{2-d } - 1 { \right } ] \
\d z , \ ] ] with @xmath294 \
\d z = { \left}(\frac{d}2 - 1{\right } ) \frac{\omega_d}{d}.\ ] ] recalling moreover that @xmath295 we have proved that @xmath296 which implies the lower bound in ( [ wbounds ] ) .
we now turn to the upper bound .
recall that , as given by ( * ? ? ?
* theorem 1.5.4 ) , there exists @xmath30 such that @xmath297 { \leqslant}c \ |y|^{2-d}.\ ] ] we will also use the more refined estimate that can be found in ( * ? ? ?
* lemma a.2 ( 149 ) ) stating that for @xmath298 , @xmath299 \\ { \leqslant}c_d \ q_d { \left}(|y|^{2-d } - r{_\beta}^{2-d } + o(|x|^{1-d}){\right}){\left } ( 1+o((r{_\beta}- |y|)^{2-d } ) { \right}).\end{gathered}\ ] ] we first treat the contribution of important sites lying in @xmath300 . by definition ,
any important site that contributes to the sum must be at distance at least @xmath301 .
moreover , since @xmath7 is assumed to belong to a good box , one has @xmath302 . using also
, we can bound the contribution of sites lying in @xmath303 by @xmath304 it suffices to choose @xmath189 small enough to ensure that this quantity is @xmath305 . more precisely , in order to have @xmath306 , one should impose @xmath307 which is clearly true for any small enough @xmath189 .
we now turn to the contribution of the important sites lying outside of @xmath303 .
let @xmath308 the contribution of the important sites belonging to a box of @xmath309 is bounded from above by @xmath310.\ ] ] we will now use the estimate given in . note first that since the @xmath169 s considered in are all in @xmath311 , the contribution of the error term @xmath312 appearing in is negligible . forgetting about this error term , and using also the fact that @xmath313 , we get that the sum in is smaller than @xmath314 to conclude the analysis , one can proceed in the same way as for the lower bound ( compare with ) .
to finish the proof , we discuss how to adapt the above arguments to the case when @xmath11 is not the origin . in both arguments
, we treated separately the box @xmath315 .
it has the convenient feature that any point outside of this box is at distance at least @xmath156 from the origin . for general @xmath11 , the box @xmath316 of the form
@xmath274 ( @xmath317 ) containing @xmath11 need not have this feature .
in this situation , one can consider separately the box @xmath316 together with its neighbouring boxes on one hand , and all the other boxes on the other , and the above arguments still apply . in order to prove
, it suffices to observe that @xmath318 & { \leqslant } & \sum_{\substack{y \in { \mathcal}{i } \cap d(x , r{_\beta}-r{_\beta } ) \setminus d(x,{\varepsilon}_1 r{_\beta } ' ) } } { \mathbf{p}}_x{\left}[s \text { visits } y \text { before exiting } d(x , r{_\beta}){\right } ] \notag \\ & { \leqslant } & \sum_{l=1}^\kappa w_{x , l},\end{aligned}\ ] ] and to apply lemma [ l : laplaces ] .
let @xmath319 be the event @xmath320 from the above computation , one can also infer that @xmath321 = o(r{_\beta}^2 i{_\beta}).\ ] ] indeed , the computation in shows that the probability to hit one element of @xmath322 is bounded by a constant times @xmath323 ( recall that @xmath324 ) . conditionally on hitting such a site , say @xmath43 , one can apply the same reasoning to bound the probability to hit another trap by a constant times @xmath323 .
the key point is to observe that there is no other element of @xmath260 within distance @xmath301 from @xmath43 .
the probability to hit another trap is thus bounded by @xmath325.\ ] ] there is no harm in replacing @xmath326 by @xmath327 in the above sum , which thus allows us to follow the proof of lemma [ l : laplaces ] and obtain that the sum in is @xmath328 . to sum up , we have thus shown that the probability in the left - hand side of is @xmath329 .
let us write @xmath330 for the event @xmath331 and @xmath332 . by the inclusion - exclusion principle , one has @xmath333 \\
{ \geqslant}w_{x , l } \ - \sum_{\substack{y < z \in { \mathcal}{i}_l \\ y , z \in d(x , r{_\beta}-r{_\beta } ) \setminus d(x,{\varepsilon}_1 r{_\beta } ' ) } } { \mathbf{p}}_x{\left}[s \text { visits } y \text { and } z \text { before exiting } d(x , r{_\beta}){\right}].\end{gathered}\ ] ] the sum in the right - hand side above is smaller than @xmath334 $ ] , and is thus @xmath335 .
using lemma [ l : laplaces ] , we arrive at @xmath336 { \geqslant}(1 - 2{\varepsilon } ) q_d \ p_l \ r{_\beta}^2(1+o(1)).\ ] ] similarly , @xmath337 { \geqslant}(1 - 2{\varepsilon } ) q_d \ l{_\beta}^{-2 } \ r{_\beta}^2(1+o(1)).\ ] ] we can now decompose the expectation under study the following way @xmath338 { \leqslant}{\mathbf{p}}_x[s=0 ] + \sum_{l=1}^\kappa { \mathbf{e}}_x{\left}[e^{-s } , { \mathcal}{e}_{l}{\right}].\ ] ] on one hand , we have @xmath339 = 1- { \mathbf{p}}_x[s \neq 0 ] \stackrel{\text{\eqref{boundone}}}{{\leqslant } } 1-(1 - 3{\varepsilon } ) q_d \ l{_\beta}^{-2 } \ r{_\beta}^2.\ ] ] on the other hand , conditionally on hitting a point @xmath340 , the walk does a geometrical number of returns to @xmath169 with a return probability equal to @xmath341.\ ] ] this probability tends to @xmath75 uniformly over @xmath342 . as a consequence , up to a negligible error , @xmath343 $ ]
is bounded by @xmath344 combining , and , we obtain @xmath345 { \leqslant}1-(1 - 3{\varepsilon } ) q_d \ r{_\beta}^2 \
\sum_{l=1}^\kappa p_l { \left } ( 1- \frac{q_d \ e^{-a_l}}{1 - ( 1-q_d ) e^{-a_l } } { \right}),\ ] ] which is precisely the bound .
we will need some information regarding the displacements at the coarse - grained scale .
let @xmath346 be the position of the particle when exiting the ball @xmath266 , that is , @xmath347 , where @xmath348 is the exit time from @xmath266 , defined in .
[ unitx ] let @xmath349 be such that @xmath350 . as @xmath6 tends to @xmath7 ,
one has @xmath351
{ \leqslant}1+\frac{\lambda{_\beta}^2 r{_\beta}^2}{2d } \ ( 1 + o(1)).\ ] ] observe that , since @xmath352 , one has @xmath351 = 1 + \frac{\lambda{_\beta}^2}{2 } \ { \mathbf{e}}_x{\left}[(x \cdot \ell)^2{\right } ] + o(\lambda{_\beta}^2 r{_\beta}^2).\ ] ] the functional central limit theorem ensures that the distribution of @xmath353 approaches the uniform distribution over the unit sphere as @xmath6 tends to @xmath7 . writing @xmath354 for the latter distribution , we need to show that @xmath355 in order to do so , one can complete @xmath19 into an orthonormal basis @xmath356 , and observe that , by symmetry , @xmath357 [ unitsx ] let @xmath349 be such that @xmath350 .
for @xmath6 small enough and any @xmath11 lying in a good box , @xmath358 { \leqslant}1 - ( 1 - 4{\varepsilon})r{_\beta}^2 \ i{_\beta}+\frac{\lambda{_\beta}^2 r{_\beta}^2}{2d } \ ( 1 + { \varepsilon}).\ ] ] we have the following decomposition @xmath359 = \\ { \mathbf{e}}_x{\left } [ \exp{\left}(\lambda{_\beta}\ x \cdot \ell{\right } ) \
{ \mathbf{1}}_{s = 0 } { \right } ] + { \mathbf{e}}_x{\left } [ \exp{\left}(-s + \lambda{_\beta}\ x \cdot \ell{\right } ) \ { \mathbf{1}}_{s \neq 0}{\right } ] .
\end{gathered}\ ] ] let us first evaluate the first term in the right - hand side above . @xmath360 & = & { \mathbf{e}}_x{\left } [ \exp{\left}(\lambda{_\beta}\ x \cdot \ell{\right } ) { \right } ] - { \mathbf{e}}_x{\left } [ \exp{\left}(\lambda{_\beta}\ x \cdot \ell{\right } ) \ { \mathbf{1}}_{s \neq 0 } { \right } ] \\ & { \leqslant } & 1+\frac{\lambda{_\beta}^2 r{_\beta}^2}{2d}(1+o(1 ) ) - { \mathbf{p}}_x[s \neq 0]{\left } ( 1+o(\lambda{_\beta}r{_\beta } ) { \right}),\end{aligned}\ ] ] where we used proposition [ unitx ] . for the second term in the right - hand side of , we have @xmath361 = { \left}(1 + o(\lambda{_\beta}r{_\beta}){\right } ) { \mathbf{e}}_x{\left } [ e^{-s } \
{ \mathbf{1}}_{s \neq 0}{\right}].\ ] ] moreover , @xmath362 = { \mathbf{e}}_x{\left } [ e^{-s}{\right } ] - { \mathbf{p}}_x[s = 0].\ ] ] we learn from proposition [ p : laplaces ] that @xmath363 { \leqslant}1-(1 - 3{\varepsilon})r{_\beta}^2 \ i{_\beta}.\ ] ] to sum up , we have shown that @xmath364 { \leqslant}1+\frac{\lambda{_\beta}^2 r{_\beta}^2}{2d}(1+o(1 ) ) - { \mathbf{p}}_x[s \neq 0]{\left } ( 1+o(\lambda{_\beta}r{_\beta } ) { \right } ) \\
+ { \left}(1 + o(\lambda{_\beta}r{_\beta}){\right } ) { \left}({\mathbf{p}}_x[s \neq 0 ] -(1 - 3{\varepsilon})r{_\beta}^2 \ i{_\beta}{\right}).\end{gathered}\ ] ] since proposition [ p : laplaces ] ensures that @xmath365 { \leqslant}2 r{_\beta}^2 l{_\beta}^{-2}$ ] , and since @xmath366 , the result follows .
we recall that @xmath367 denote the successive steps of the trajectory coarse - grained at scale @xmath147 .
let @xmath368 be defined by @xmath369 where @xmath370 is the half - space not containing @xmath7 delimited by the hyperplane orthogonal to @xmath19 and at distance @xmath20 from the origin . by the definition of @xmath76 , we have @xmath371 we first want to discard overly slow behaviours . out of the sequence of @xmath368 coarse - graining instants @xmath372
, we extract a maximal subsequence @xmath373 such that for any any @xmath374 , @xmath375 lies in a good box . for @xmath374
, we define @xmath376 where , for @xmath377 , @xmath378 denotes the time translation by @xmath379 on the space of trajectories , that is , @xmath380 , and we recall that @xmath264 was defined in and @xmath347 . note that the @xmath381 are stopping times ( under @xmath382 for every fixed environment , and with respect to the natural filtration associated to @xmath13 ) .
the number @xmath383 counts how many coarse - graining instants prior to @xmath384 are such that the walk at these instants lies in a good box .
[ noslowmotion ] for @xmath6 small enough , any @xmath385 and almost every environment , one has @xmath386 { \leqslant}\exp{\left}(-\frac{\sqrt{i{_\beta}}}{2 c_3 } \ n{\right}).\ ] ] for small @xmath6 , the walk makes on average @xmath387 steps in each coarse - grained unit of time . roughly speaking
, the event @xmath388 corresponds to asking @xmath76 to be in the interval @xmath389 .
we will use this proposition with @xmath390 large , so as to discard the possibility that @xmath383 be too large .
note that with probability one , the sequence @xmath391 can be defined for any @xmath392 ( that is , we may as well not stop the sequence at @xmath383 ) .
the left - hand side of is smaller than @xmath393 { \leqslant}{\mathbf{e}}_0{\left } [ \exp { \left } ( -\sum_{k = 0}^{n/(c_3 r{_\beta}^2 \sqrt{i{_\beta}})-1 } s_k { \right } ) { \right}].\ ] ] using the markov property and proposition [ p : laplaces ] , we obtain that for @xmath6 small enough , the latter is smaller than @xmath394 where in the last step , we used the fact that @xmath395 as @xmath11 tends to @xmath7 , and that @xmath396 . before discussing the costs associated to a range of speeds that should contain the optimal speed , we introduce a `` surgery '' on the trajectory of the random walk , which consists in erasing certain annoying loops .
we first introduce general notations . given @xmath397 , we write @xmath398 = \{k \text { s.t . } \exists i : g_i { \leqslant}k < h_{i+1 } \}.\ ] ] we call @xmath399 $ ] a _ surgery _ of @xmath400 with _ at most
_ @xmath177 _ cuts _ , where the cuts are the sets @xmath401 whose union is the complement of @xmath399 $ ] in @xmath400 .
note that , since we allow the possibility that @xmath402 or @xmath403 , it is not possible to recover @xmath177 if given @xmath399 $ ] , hence the phrasing `` _ _ at most _ _ @xmath177 cuts '' .
let us now discuss why some surgery is needed , and how we choose the surgery @xmath399 $ ] . for simplicity ,
we write @xmath404 for the coarse - grained random walk . by the definition of @xmath368 ( see ) , we have @xmath405 . based on our previous work
, we should be able to argue that there are only few different bad boxes visited by @xmath406 , and we would like to conclude that @xmath407 ( recall that @xmath408 is the sum of the increments of the coarse - grained walk that start within a good box ) . in other words , we would like to say that the sum of increments that start within a bad box gives a negligible contribution in the direction of @xmath19 .
this may however fail to hold , even if the number of bad boxes visited is really small , since it may happen for instance that the walk visits many times the same bad box and every time makes a jump in the direction of @xmath19 . instead of trying to control the trajectory of the walk on bad boxes ( which would be a very delicate matter ) , we introduce a surgery on the path . each time a new bad box is discovered ,
we remove the piece of trajectory between the first and last visits to the bad box .
we may call the piece of trajectory which is removed a _ loop _ since , although it possibly does not intersect itself , the starting and ending points are in the same box .
once these pieces of trajectory have been removed , the remaining increments should satisfy an inequality like .
more precisely , we define @xmath409 @xmath410 and then , recursively , as long as @xmath411 is well defined , @xmath412 @xmath413 we let @xmath414 be the largest @xmath177 such that @xmath411 is well defined , and set @xmath415 , @xmath416 ( see figure [ f : surgery ] for an illustration ) .
for any @xmath177 , one has @xmath417 since @xmath418 and @xmath419 lie in the same box ( this is the loop , that we cut out ) .
hence , @xmath420 and we may rewrite the last double sum as @xmath421 } ( y_{k+1 } - y_k ) \cdot \ell.\ ] ] clearly , @xmath414 is smaller than the number of different bad boxes visited .
moreover , by definition , if @xmath422 is such that @xmath423 for some @xmath177 ( that is , if @xmath424 $ ] ) , then @xmath425 lies in a good box . in other words , the summands in form a subsequence of the summands in the left - hand side of . considering and the fact that @xmath426
, we obtain a lower bound on the sum in : @xmath427 } ( y_{k+1 } - y_k ) \cdot \ell { \geqslant}n - r{_\beta}- 4 \sqrt{d } j r{_\beta},\ ] ] which we be useful as soon as we have a good upper bound on @xmath414 , the number of different bad boxes visited .
the set @xmath428 $ ] is a surgery on the set @xmath429 that indexes the successive jumps of the coarse - grained walk , and moreover , we recall that any @xmath424 $ ] is such that @xmath425 lies in a good box .
we now transform it into a surgery @xmath430 $ ] of the set @xmath431 that indexes the successive jumps of the coarse - grained walk that start in a good box ( call this a `` good increment '' ) .
more precisely , for @xmath432 , define @xmath433 to be the index of the last good increment occurring before @xmath434 , @xmath435 to be the index of the first good increment occurring at or after @xmath436 , together with @xmath437 and @xmath438 . with this notation
, we have @xmath439 } ( y_{k+1 } - y_k ) \cdot \ell = \sum_{k \in [ g',h ' ] } x_k \cdot \ell,\ ] ] where we recall that @xmath440 was defined in . the passage from the right - hand side to the left - hand side in
is purely a re - indexation of the terms , each of them appearing in both sides and in the same order .
note that @xmath430 $ ] has at most @xmath414 cuts , and @xmath414 is bounded by the number of distinct bad boxes visited by the walk .
a negative aspect of the surgery is that @xmath430 $ ] depend on the full trajectory of the walk up to hitting the hyperplane . to overcome this problem
, we will use a union bound on all reasonable surgeries , and then examine each deterministic surgery @xmath399 $ ] separately .
we thus need a bound on the number of these surgeries .
the bound we need can grow exponentially with @xmath20 , but the prefactor must be small compared to @xmath441 .
we start with a combinatorial lemma .
[ combinat ] let @xmath442 , and @xmath443 be the total number of possible surgeries of the set @xmath444 by at most @xmath445 cuts .
we have @xmath446n { \right}).\ ] ] the surgery in is defined by giving @xmath447 , and in our present setting , we impose @xmath448 and @xmath449 . hence , in order to define such a surgery , it is sufficient to give oneself an increasing sequence ( in the wide sense ) of @xmath450 elements in @xmath451 .
we now proceed to count these objects .
consider such a sequence @xmath452 .
we think of @xmath453 as a string of characters , and for each @xmath229 , we insert a character @xmath454 in this string just before the value taken by @xmath455 ( for instance , @xmath456 is the string obtained from the sequence @xmath457 ) .
one can see that there is a bijective correspondence between increasing sequences and the set of positions of the character @xmath454 within the string , provided we do not allow a @xmath454 as the last character .
the number of increasing sequences of length @xmath450 in @xmath451 is thus equal to @xmath458 the result is then obtained using the fact that @xmath459 ( and the latter can be checked by induction on @xmath20 ) .
recalling that the space is partitioned into the family of boxes @xmath152 , we let @xmath460 be the set of indices of the boxes visited by the coarse - grained trajectory . since the boxes are of size @xmath147 , the set @xmath461 is a lattice animal when the walk is started at the origin ( see figure [ f : surgery ] ) .
moreover , the box containing @xmath462 is within distance @xmath463 from the hyperplane , which is itself at distance @xmath20 from the origin .
it thus follows that , for any @xmath464 and @xmath6 small enough , @xmath465 let @xmath466 be the event defined by @xmath467 where we recall that @xmath468 .
we get from proposition [ animals ] and inequality that @xmath469 { \leqslant}e^{-n / c_4 r{_\beta}}.\ ] ] discarding events with asymptotic cost @xmath470 , we can focus our attention on the environments for which @xmath466 holds .
we now fix @xmath471 this way , the lower bound for the cost obtained in proposition [ noslowmotion ] is @xmath472 , which is much larger than the cost we target to obtain in the end , that is , @xmath473 ( to this end , we could as well choose a larger @xmath390 , but having a separation of the order of @xmath474 will prove useful in section [ s : intermediate2 ] ) .
in other words , with this choice of @xmath390 , we can assume from now on that the number of good steps made by the random walk satisfies @xmath475 since each good box has to be visited by the coarse - grained trajectory at least once , when condition holds , we have @xmath476 using this together with , we obtain that @xmath477 where @xmath478 [ controlcuts ] there exists @xmath479 such that for all @xmath6 small enough , when @xmath466 and are both satisfied , the number of possible values of @xmath430 $ ] ( that is , the cardinality of the set of surgeries having non - zero probability ) is bounded by @xmath480 n { \right}),\ ] ] and moreover , if @xmath481 , then @xmath482 } x_k \cdot \ell { \geqslant}(1-{\varepsilon})n.\ ] ] for the first part , considering that the number of bad boxes visited by the coarse - grained walk is an upper bound for the number @xmath414 of cuts in the surgery @xmath430 $ ] , together with , we can apply lemma [ combinat ] with @xmath483 and @xmath484 equal to the left - hand side of .
the conclusion follows since @xmath485 is then the logarithm of some power of @xmath143 , so is smaller than @xmath486 when @xmath6 is small enough .
the second part also follows from the bound on @xmath414 , together with and
. we can now give precise estimates for the cost of speeds of the order of @xmath441 or higher .
we write @xmath487|$ ] for the cardinality of the set @xmath430 $ ] , and for @xmath488 , we write @xmath489 for the conjunction of the events @xmath490| < \frac{n}{r{_\beta}^2 v{_\beta } } , } \\ \displaystyle{{\mathcal}{a}_n \text { and \eqref{boundonkn } hold } , } \end{array}\ ] ] where we recall that @xmath466 is the event defined in .
we write @xmath491 for the complement of the event @xmath492 , that is , for the event when either @xmath466 or fails to hold . recalling that on average , the walk makes @xmath387 steps during each coarse - grained displacement , and forgetting about the path surgery
, one can interpret the event @xmath493 as asking the random walk to move with a speed contained in the interval @xmath494 .
we further define @xmath495,\ ] ] @xmath496.\ ] ] [ costofspeed ] 1 . for @xmath6 small enough
, one has @xmath497 2 .
let @xmath498 .
if @xmath499 satisfy @xmath500 and either @xmath501 or @xmath502 , then for any small enough @xmath6 , one has @xmath503 \ n { \right}),\ ] ] thinking about @xmath504 , @xmath505 , this can be seen as a rigorous form of the informal statement given in the introduction .
for part ( 1 ) , we saw in that @xmath506 { \leqslant}e^{-n/(c_4 r{_\beta})}$ ] . since @xmath507 , this term is smaller than @xmath508 for small enough @xmath6 .
the claim is then a consequence of proposition [ noslowmotion ] and of our choice of @xmath390 ( see ) .
concerning part ( 2 ) , the claim holds if @xmath6 is small enough and @xmath509 . from now on , we consider only the case when @xmath481 .
an upper bound on @xmath510 is given by @xmath511 } s_k { \right } ) , { \mathcal}{e}_n(v{_\beta},v{_\beta } ' ) { \right}].\ ] ] note that it follows from the assumptions that @xmath512 for some fixed @xmath513 .
we let @xmath514 be the set of surgeries @xmath399 $ ] such that @xmath515 = [ g , h ] , { \mathcal}{e}_n(v{_\beta},v{_\beta}'){\right } ] > 0.\ ] ] in view of , proposition [ controlcuts ] ensures that for @xmath6 small enough , the cardinality of the set @xmath514 is smaller than @xmath516 \
n { \right}).\ ] ] it moreover guarantees that on the conjunction of the events @xmath430 = [ g , h]$ ] and @xmath489 , one has @xmath517 }
x_k \cdot \ell { \geqslant}(1-{\varepsilon})n.\ ] ] it thus suffices to show that , for any sequence of cuts @xmath518 \in { \mathcal}{s}_n$ ] , one has @xmath519 } s_k { \right } ) , \sum_{k \in [ g_n , h_n ] } x_k \cdot \ell { \geqslant}(1-{\varepsilon})n { \right } ] \\ { \leqslant}\exp{\left } ( -(1 - 4{\varepsilon } ) { \left}[\frac{dv{_\beta}}{2 } + \frac{i{_\beta}}{v{_\beta } ' } { \right } ] n{\right}).\end{gathered}\ ] ] for @xmath350 to be determined ,
the left - hand side above is bounded by @xmath520}- s_k+\lambda{_\beta}\ x_k \cdot \ell { \right}){\right}].\ ] ] using proposition [ unitsx ] and the markov property , for all @xmath6 small enough , we can bound the latter by @xmath521| } \\ { \leqslant}\exp{\left}(-(1-{\varepsilon})\lambda{_\beta}\
n + |[g_n , h_n]| { \left}(- ( 1 - 4{\varepsilon})r{_\beta}^2 \ i{_\beta}+ ( 1 + { \varepsilon } ) \ \frac{\lambda{_\beta}^2 r{_\beta}^2}{2d } { \right } ) { \right}),\end{gathered}\ ] ] where we used the fact that @xmath522 . for @xmath518 \in { \mathcal}{s}_n$ ] ,
inequality transfers into an inequality on the cardinality of @xmath518 $ ] , and thus the latter is bounded by @xmath523 n { \right}).\ ] ] choosing @xmath524 , we arrive at the bound @xmath525 n { \right}),\ ] ] which implies the bound , and thus finishes the proof .
[ corcase1 ] there exists @xmath30 ( depending on @xmath107 ) such that for any @xmath6 small enough , one has @xmath526 with @xmath527 .
the fact @xmath527 was seen in .
let @xmath528 be a subdivision of @xmath529 .
the events @xmath530 form a partition of the probability space . applying proposition [ costofspeed ]
, we thus get that for @xmath6 small enough , @xmath56 is smaller than @xmath531 \sqrt{i{_\beta } } \ n { \right } ) \\ + 2 \exp{\left}(-(1 - 5{\varepsilon})\frac{dx_l}{2 } \sqrt{i{_\beta } } \ n { \right } ) .\end{gathered}\ ] ] when @xmath532 is taken small enough , and @xmath533 large enough , the dominant exponential has an exponent which we can take as close as we wish ( that is , up to a multiplicative factor going to @xmath206 ) to the minimum of the function @xmath534 \sqrt{i{_\beta}}.\ ] ] this minimum is @xmath535 , and we thus obtain the proposition .
[ existm ] let @xmath536 be such that for any @xmath537 , one has @xmath538 . define @xmath539.\
] ] we have @xmath540 as @xmath6 tends to @xmath7 , and moreover , @xmath541 where we recall that @xmath87 was defined in .
the fact that @xmath542 is clear since we assume @xmath23 = + \infty$ ] . for the second part , observe that , since @xmath46 , @xmath543 while , by the definition of @xmath544 , @xmath545.\ ] ] the result follows comparing and . from now on , we fix @xmath546 .
we split sites according to the value of their attached potential into three categories , according to whether the potential belongs to @xmath547 , to @xmath548 , or to @xmath549 . in view of proposition [ existm ] ,
sites in the first category are always negligible ( under our present assumption that @xmath23 $ ] is infinite ) .
we call sites in the second category _ intermediate sites_. sites in the last category are the important sites considered in the previous section .
let us write @xmath550 and recall the definition of @xmath90 in .
three cases can occur .
note that we may switch infinitely many times from one case to another as @xmath6 tends to @xmath7 .
if @xmath6 is sufficiently small and @xmath552 , then corollary [ corcase1 ] gives us a sharp upper bound on @xmath56 ( that is , an exponential bound with exponent @xmath553 , up to a multiplicative error controlled by @xmath107 ) .
in other words , case 1 is now under control .
we treat separately cases 2 and 3 in each of the following sections .
among the three cases displayed in , the case when @xmath554 which we now investigate , is the most delicate , since both intermediate and important sites have to be taken into account .
we want to approximate the integral @xmath555 by a riemann sum .
let us write @xmath556 .
we split the interval @xmath557 along the subdivision given by the successive powers of @xmath558 : @xmath559 where @xmath560 since @xmath46 and @xmath561 , we have @xmath562 where @xmath563 in order to lighten the notation , we keep implicit the fact that @xmath564 depends on @xmath6 . to begin with ,
we want to exclude the @xmath112 s such that @xmath564 is not much larger than @xmath87 . roughly speaking
, we will then show that for @xmath112 s such that @xmath564 is indeed much larger than @xmath87 , it is too costly to have a deviation of @xmath565 from its typical value , so that potentials falling in @xmath566 are in a `` law of large numbers '' regime .
let @xmath567 note that @xmath568 and since @xmath569 ( see ) , we obtain @xmath570 and thus @xmath571 we also define @xmath572 where @xmath129 was introduced in .
in view of , and proposition [ existm ] , we have @xmath573 for @xmath574 , we let @xmath575 and we call elements of @xmath576 @xmath112-_intermediate sites_. the relevant length scale for these sites is @xmath577 defined by @xmath578 = { \tilde}{p}_l { \geqslant}\rho^{-l_0/2 }
\ { \mathfrak}{i}{_\beta}{\geqslant}{\varepsilon}^{-4 } \ { \mathfrak}{i}{_\beta}.\ ] ] we have defined in the length scale @xmath143 in such a way that @xmath579 , and recall that @xmath580 . as a consequence , @xmath581 in particular ,
all the length scales @xmath577 associated to @xmath112-intermediate sites are smaller than the length scale @xmath143 we used in section [ s : lowerbd ] . with the same @xmath189 as in section [ s : lowerbd ] and for any @xmath574
, we can define the scales @xmath582 , @xmath583 , and @xmath584 .
the space is partitioned into the boxes @xmath585 .
each box @xmath586 is itself partitioned into boxes of size @xmath587 , and we write @xmath588 to denote this partition . in turn , each box @xmath589 is partitioned into boxes of size @xmath590 , and we write @xmath591 to denote this partition .
let @xmath592 be a box of size @xmath590 .
define @xmath593 we say that a box @xmath586 is @xmath112__-balanced _ _ if for any box @xmath594 ( of size @xmath587 ) , one has @xmath595 as the reader has noticed , this definition closely parallels the definition of a balanced box given in . consider a box @xmath160 of size @xmath147 ( as introduced in subsection [ ss : coarse - grained ] ) .
we know from that for any @xmath574 , one has @xmath596 . as usual , we assume that we can partition the box @xmath160 into subboxes of size @xmath597 , for any @xmath574 , and write this partition @xmath598 .
we say that the box @xmath160 is _ very balanced _
if the following two conditions hold : 1 .
the box is balanced ( in the sense introduced in subsection [ ss : coarse - grained ] ) , 2 . for any @xmath574 , the proportion of @xmath112-balanced boxes in @xmath598 is at least @xmath599 .
we say that the box @xmath160 is _ very good _ if for any @xmath177 such that @xmath178 , the box @xmath179 is very balanced .
similarly , we say that a box of size @xmath587 is @xmath112__-good _ _ if the box and all its @xmath226-neighbours are @xmath112-balanced .
we now show that , at the exponential scale , the probability that a box is very balanced is of the same order as the probability that it is simply balanced .
[ p : verygood ] there exists @xmath600 such that for @xmath6 small enough and any @xmath223 , @xmath601 { \geqslant}{\tilde}{c}_1 l{_\beta}^{d\delta}.\ ] ] an estimate of the right order is provided by proposition [ p : probbadbox ] for the probability that @xmath160 is not balanced .
an examination of the proofs of lemmas [ l : subsubbox ] and [ p : subbox ] shows that there exists @xmath217 such that , for any @xmath574 , if @xmath6 is small enough to ensure @xmath602 then @xmath603 { \leqslant}\frac{|{\tilde}{b}_{l , i}|}{|{\tilde}{b}| } \ e^{-c_0 { \tilde}{l}{_{l,\beta}}^{d\delta}},\ ] ] where @xmath604 is a box of size @xmath587 . augmenting possibly the value of @xmath605
, we can assume that holds for every @xmath6 and @xmath112 , regardless of condition .
we now want to estimate @xmath606.\ ] ] the probability inside the logarithm is of the form of with @xmath607 and @xmath608 as a consequence , the quantity in is larger than @xmath609 for some constant @xmath232 .
observe that the derivation above , which holds for @xmath6 small enough , is valid uniformly over @xmath112 . indeed , @xmath610 so that all length scales go to infinity uniformly as @xmath6 tends to @xmath7 .
now , a union bound on @xmath611,\ ] ] together with the fact that @xmath612 gives the upper bound @xmath613 and thus proves the claim .
although being a very balanced box is more difficult than being simply a balanced box , proposition [ p : verygood ] gives an upper bound on the probability for a box not to be very balanced which is of the same order as the upper bound we obtained in proposition [ p : probbadbox ] . as a consequence ,
nothing changes if we replace `` good boxes '' by `` very good boxes '' throughout section [ s : lowerbd ] . * from now on , we perform this replacement * : that is , whenever we refer to section [ s : lowerbd ] or to objects defined therein , it is with the understanding that the `` good boxes '' appearing there are in fact `` very good boxes '' . for any @xmath574
, we can define a coarse - grained trajectory at the scale @xmath597 . for every @xmath574 , we define @xmath614 , and recursively , @xmath615 recall the definition of @xmath616 from as @xmath617 and let @xmath618 be the largest @xmath422 such that @xmath619 .
in words , the walk makes @xmath618 coarse - grained steps of size @xmath597 inside the ball @xmath620 , and then exits this ball during the next step .
we further define @xmath621 and @xmath622 .
the superscripts @xmath623 and @xmath161 stand for `` good '' and `` bad '' , respectively .
we start by giving a tight control of @xmath618 .
[ controltlk ] for @xmath6 small enough , the following properties hold for any @xmath100 and @xmath574 . 1 . @xmath624 { \leqslant}{\left}(1+\frac{{\varepsilon}}{2}{\right } ) \ { \left}(\frac{r{_\beta}}{{\tilde}{r}{_{l,\beta}}}{\right})^2.\ ] ] 2 . for any @xmath625 , @xmath626 \ { \right } ] { \leqslant}\frac{1}{2^m}.\ ] ] 3 .
there exists @xmath627 , @xmath628 independent of @xmath6 and @xmath107 such that , for any @xmath629 , @xmath630 } { \right}){\right } ] { \leqslant}c_0,\ ] ] @xmath631 } - 1 { \right } ) { \right}){\right } ] { \leqslant}\exp(c_0 \lambda^2).\ ] ] note that @xmath632 $ ] depends neither on @xmath11 , nor on the environment .
one can check that @xmath633 is a submartingale with respect to the filtration generated by the random walk at the successive times @xmath634 ( recall that we write @xmath635 for the euclidian norm ) .
hence , @xmath636 { \leqslant}{\mathbf{e}}_x{\left } [ |s_{j_{l , k\wedge { \mathcal}{t}_{l } } } - s_{0}|^2 { \right}].\ ] ] using the monotone convergence theorem for the left - hand side , and the dominated convergence theorem for the right - hand side , we obtain @xmath637 { \leqslant}{\tilde}{r}_{l,\beta}^{-2 } \ { \mathbf{e}}_x{\left } [ |s_{j_{l,{\mathcal}{t}_{l } } } - s_{0}|^2 { \right}].\ ] ] by the definition of @xmath618 , we know that
@xmath638 so @xmath637 { \leqslant}{\left } ( \frac{r{_\beta}}{{\tilde}{r}{_{l,\beta } } } + 1 { \right})^2.\ ] ] now , it follows from that @xmath639 or equivalently , @xmath640 and this leads to the upper bound in part ( 1 ) . the lower bound
can be obtained in the same way .
indeed , fails to be a proper martingale only due to lattice effects . taking these lattice effects into account
, we observe that @xmath641 is a supermartingale . keeping the subsequent arguments unchanged
, we obtain the lower bound of part ( 1 ) .
we have thus seen that starting from @xmath11 , the expectation of the number of @xmath597-steps performed before exiting @xmath326 does not exceed @xmath642 .
clearly , if instead we start from a point @xmath643 , then the number of @xmath597-steps performed before exiting @xmath326 is bounded by the number of steps required to exit @xmath644 , so in particular , its expectation is bounded by @xmath645.\ ] ] part ( 2 ) follows using the markov property and chebyshev s inequality .
part ( 3 ) is a direct consequence of part ( 2 ) .
( for the second part , one can use the uniform exponential integrability obtained in the first part to justify that @xmath631 } - 1 { \right } ) { \right}){\right } ] = 1+\frac{\lambda^2}{2 } { \mathbf{e}}_x{\left } [ { \left } ( \frac{{\mathcal}{t}_{l}}{{\mathbf{e}}_0[{\mathcal}{t}_{l } ] } - 1 { \right})^2 { \right } ] + o(\lambda^2),\ ] ] and observe that due to part ( 2 ) , the expectation on the right - hand side remains bounded uniformly over @xmath6 . ) with the next proposition , we have a first indication that @xmath646 is not very large when @xmath647 is in a very good box . [
p : tb1 ] for @xmath6 small enough and @xmath574 , if @xmath11 lies in a very good box , then @xmath648 { \leqslant}{\varepsilon}^3 \ { \mathbf{e}}_0[{\mathcal}{t}_{l}].\ ] ] for simplicity , we assume that @xmath271 . under this circumstance , the box @xmath649 is very good .
in particular , the set @xmath650 has cardinality at most @xmath651 .
we will need the observation that there exists a constant @xmath652 , independent of @xmath6 and @xmath112 , such that @xmath653 { \leqslant}\frac{c_1 } { |i|^{d-2}}.\ ] ] this fact is ( * ? ? ? * proposition 1.5.10 ) with @xmath654 . for the reader
s convenience , we now give a sketch of the proof .
let @xmath655 be the green function of @xmath13 , and @xmath656 be the hitting time of @xmath586 .
let us also write @xmath657 to denote the centre of the box @xmath586
. then @xmath658,\ ] ] and the conclusion , that is a bound on @xmath659 $ ] , is obtained through the following estimates on the green function : @xmath660 in view of , it is easy to show that the expected number of visits of the @xmath597-coarse - grained random walk inside a fixed box of size @xmath597 is bounded , uniformly over @xmath6 . indeed , from this box , one has some non - degenerate probability to move at a distance a constant multiple of @xmath597 in a bounded number of @xmath597-steps , and once there , ensures that there is a non - degenerate probability to never go back to the box . to sum up , in order to prove the proposition , it suffices to bound the expectation of the number of @xmath597-boxes of @xmath273 that are visited by @xmath13 .
using , we get that this number is bounded by @xmath661 whatever the set @xmath273 with cardinality smaller than @xmath662 is , the sum above is bounded by the sum obtained when @xmath273 is a ball centred around @xmath7 and of radius @xmath663 . comparing this sum with an integral leads to the upper bound @xmath664 this yields the desired result , thanks to part ( 1 ) of proposition [ controltlk ] .
now , for @xmath665 , the same reasoning applies , the only difference being that one has to consider not only the @xmath147-box that contains @xmath11 , but also its @xmath226-neighbours .
[ p : tb2 ] let @xmath666 , where @xmath667 appears in part ( 3 ) of proposition [ controltlk ] .
we have @xmath668 } { \right } ) { \right } ] { \leqslant}\exp{\left}({\varepsilon}^{5/4}{\right}).\ ] ] let us write @xmath106 for the event @xmath669 } > { \varepsilon}^{3/2}.\ ] ] decomposing the expectation under study along the partition @xmath670 gives the bound @xmath671 } { \right } ) , \ a { \right}].\ ] ] decomposing this new expectation according to the event @xmath672 } { \right } ) > { \varepsilon}^{-1/8 } { \right}\},\ ] ] we can bound by @xmath673 + { \mathbf{e}}_0{\left}[\exp{\left } ( \lambda_1 \frac{{\mathcal}{t}_l^{(b)}}{{\mathbf{e}}_0[{\mathcal}{t}_l ] } { \right } ) , \ a ' { \right}].\ ] ] we learn from proposition [ p : tb1 ] that @xmath674 { \leqslant}{\varepsilon}^{3/2}/\lambda_1 $ ] . finally ,
the last expectation in is bounded by @xmath675 } { \right } ) { \right}],\ ] ] and since @xmath676 , the result follows from part ( 3 ) of proposition [ controltlk ] .
let @xmath677 we would like to derive a sharp control of @xmath678 [ laplacetds ] assume that @xmath6 is small enough , and that @xmath11 lies in an @xmath112-good box . for every @xmath574 and every @xmath679 , one has @xmath680
{ \leqslant}1 - ( 1 - 3{\varepsilon } ) \ { \tilde}{r}{_{l,\beta}}^2 \ { \tilde}{p}_l \ f(\gamma \rho^l),\ ] ] where we recall that @xmath41 was defined in .
the starting point is to define a quantity smaller than @xmath681 that matches the definition of @xmath264 given in , where the length scale @xmath577 replaces @xmath143 throughout .
the analysis is then identical to the one we have performed to prove proposition [ p : laplaces ] .
the fact that the identity holds for @xmath6 small enough , uniformly over @xmath574 , follows again from .
we now construct a sequence of coarse - grained instants on the scale @xmath597 .
we do it however with a twist , since each time the walk exits a ball of radius @xmath147 , the current @xmath597-step is simply discarded , and we start the @xmath597-coarse - graining afresh .
more precisely , recall that @xmath682 denote the maximal subsequence of @xmath147-coarse - graining instants such that for any @xmath422 , @xmath375 lies in a very good box ( we may call these the instants of @xmath147-very good steps ) .
we let @xmath683 @xmath684 where @xmath685 is the time shift .
then @xmath686 is obtained as the concatenation of the sequences @xmath687 out of the sequence @xmath686 , we extract a maximal subsequence @xmath688 such that for any @xmath422 , @xmath689 lies in an @xmath112-good box ( we may call this an @xmath112-good step ) .
note that all @xmath690 are stopping times ( with respect to the natural filtration of @xmath13 , for every fixed environment ) .
recall that @xmath383 is such that @xmath691 . in words
, @xmath383 is a lower bound on the number of very good @xmath147-steps performed by the walk before the time it reaches the distant hyperplane .
we let @xmath692 this gives us a lower bound on the number of @xmath112-good @xmath597-steps performed by the walk before reaching the distant hyperplane .
let @xmath693 and define similarly @xmath694 , @xmath695 . by definition ,
the number of @xmath112-good steps performed in the @xmath422-th @xmath147-very good step is @xmath694 , hence @xmath696 we also introduce @xmath697 where @xmath698 was defined in .
let @xmath399 $ ] be a fixed surgery .
we can , out of the concatenation of @xmath699}$ ] , extract a maximal subsequence @xmath700}_{l , k } , 0 { \leqslant}k < { \tilde}{k}^{[g , h]}_{l})$ ] such that for every @xmath422 , @xmath701}_{l , k}}$ ] lies in an @xmath112-good box , and define @xmath702}_{l } = \sum_{k \in [ g , h ] } { \mathcal}{t}_{l , k}^{(g)}.\ ] ] for @xmath703}_{l}$ ] , we let @xmath704}_{l , k } = + \infty$ ] .
the important thing to notice is that for any @xmath422 , @xmath704}_{l , k}$ ] is a stopping time ( with respect to the natural filtration of @xmath13 , for every fixed environment ) . finally , we let @xmath705 } = { \tilde}{s}_l \circ \theta_{{{\tilde}{\jmath}}^{[g , h]}_{l , k}}.\ ] ] the next proposition ensures two things : first , that if @xmath706|$ ] is of order @xmath707 , then outside of a negligible event , @xmath708}_{l}$ ] is at least @xmath709 ; second , that the contribution of the @xmath112-intermediate sites , cut according to the surgery @xmath399 $ ] , is in a law of large numbers regime , outside of a negligible event ( recall that the average contribution of @xmath112-intermediate sites is of the order of @xmath710 ) . [ lotsoflgood ] there exists @xmath711 ( depending on @xmath107 ) such that the following holds .
let @xmath712 and let @xmath518 $ ] be a sequence of surgeries such that @xmath713| .\ ] ] 1 .
we have @xmath714}_{l } { \leqslant}(1 - 2 { \varepsilon } ) \
\frac{n}{{\tilde}{r}{_{l,\beta}}^2 v{_\beta } ' } { \right } ] { \leqslant}2 \exp{\left } ( - \frac{c_6}{r{_\beta}^2 v{_\beta } ' } \ n{\right}).\ ] ] 2 .
let @xmath715)$ ] be the event @xmath716}_{l}-1 } { \tilde}{s}_{l , k}^{[g_n , h_n ] } > ( 1 - 6 { \varepsilon } ) \ \rho^l \ { \tilde}{p}_l\ \frac{n}{v{_\beta}'},\ ] ] and @xmath717)$ ] be its complement .
we have @xmath718 ) { \right } ] { \leqslant}2 \exp{\left } ( - \frac{{\varepsilon}^2 \ { \tilde}{p}_l}{6 \ v{_\beta } ' } \ n{\right}).\ ] ] 3 . moreover , @xmath719){\right } ] { \leqslant}c_2 \exp{\left } ( - \frac{{\varepsilon}^{-2 } \ { \mathfrak}{i}{_\beta}}{6 \ v{_\beta } ' } \ n{\right}).\ ] ] for part ( 1 ) , it is sufficient to show the result with a shortened surgery @xmath720 $ ] , which coincides with @xmath518 $ ] for its first @xmath721 terms , and then stops . considering and the fact that @xmath722 , it suffices to show that the following two probabilities are sufficiently small : @xmath723 } { \mathcal}{t}_{l , k } { \leqslant}(1- { \varepsilon } ) \
\frac{n}{{\tilde}{r}{_{l,\beta}}^2 v{_\beta } ' } { \right}],\ ] ] @xmath724 } { \mathcal}{t}_{l , k}^{(b ) } { \geqslant}{\varepsilon}\ \frac{n}{{\tilde}{r}{_{l,\beta}}^2 v{_\beta } ' } { \right}].\ ] ] let us start by examining . with the help of part ( 1 ) of proposition [ controltlk ] , we can bound this probability by @xmath725 } \frac{{\mathcal}{t}_{l , k}}{{\mathbf{e}}_0[{\mathcal}{t}_l ] } { \leqslant}(1-\frac{{\varepsilon}}{3 } ) \
\frac{n}{r{_\beta}^2 v{_\beta } ' } { \right } ] \\ { \leqslant}{\mathbb{p}}{\mathbf{p}}_0{\left }
[ \sum_{k \in [ g_n',h_n ' ] } { \left}(\frac{{\mathcal}{t}_{l , k}}{{\mathbf{e}}_0[{\mathcal}{t}_l ] } - 1{\right } ) { \leqslant}-\frac{{\varepsilon}}{3 } \
\frac{n}{r{_\beta}^2 v{_\beta } ' } { \right}],\end{gathered}\ ] ] since @xmath726| = n/(r{_\beta}^2 v{_\beta}')$ ] .
let @xmath727 , where @xmath667 is given by part ( 3 ) of proposition [ controltlk ] .
the probability above is smaller than @xmath728 } { \left}(\frac{{\mathcal}{t}_{l , k}}{{\mathbf{e}}_0[{\mathcal}{t}_l ] } - 1{\right } ) { \right } ) { \right } ] \\
{ \leqslant}\exp{\left}(-(\lambda \frac{{\varepsilon}}{3 } - c_0 \lambda^2 ) \frac{n}{r{_\beta}^2 v{_\beta}'}{\right } ) , \end{gathered}\ ] ] where @xmath729 is given by part ( 3 ) of proposition [ controltlk ] , and in the last step , we used the markov property and the fact that @xmath730| { \leqslant}{n}/({r{_\beta}^2 v{_\beta}})$ ] .
it then suffices to take @xmath731 small enough to get an appropriate bound on .
we now turn to . using part ( 1 ) of proposition [ controltlk ] , we can bound the probability appearing there by @xmath732 } \frac{{\mathcal}{t}_{l , k}^{(b)}}{{\mathbf{e}}_0[{\mathcal}{t}_l ] } { \geqslant}\frac{{\varepsilon}}{2 } \ \frac{n}{r{_\beta}^2 v{_\beta } ' } { \right}].\ ] ] this in turn
is bounded by @xmath733 } { \mathcal}{t}_{l , k}^{(b)}{\right } ) { \right}].\ ] ] as before , using the markov property and proposition [ p : tb2 ] , we get the bound @xmath734 and this proves the claim ( provided we fixed @xmath107 small enough ) . we now prove . in view of part ( 1 ) and of the fact that @xmath735
, we can assume that @xmath736}_{l } > m_n \stackrel{\text{(def)}}{= } ( 1 - 2 { \varepsilon } ) \
\frac{n}{{\tilde}{r}{_{l,\beta}}^2 v{_\beta}'}.\ ] ] let @xmath679 to be determined . on this event ,
the probability of @xmath717)$ ] is bounded by @xmath737 } { \right } ) { \right}].\ ] ] using proposition [ laplacetds ] , the fact that @xmath738 for @xmath739 , and the markov property , we obtain the bound @xmath740 substituting by the value of @xmath741 , this transforms into @xmath742 { \right}).\ ] ] a simple computation shows that @xmath743 and since @xmath744 , for @xmath107 small enough , one has @xmath745 for @xmath746 such that @xmath747 , the quantity in square brackets appearing in is bounded from below by @xmath748 and this proves .
we now examine how to go from to .
we will actually show that @xmath749){\right } ] { \leqslant}2 { \left } ( 1 + \frac{6 \ v{_\beta}'}{n \
{ \varepsilon}^2 \ { \mathfrak}{i}{_\beta } } { \right } ) \
\exp{\left } ( -n \ \frac{{\varepsilon}^{-2 } \ { \mathfrak}{i}{_\beta}}{6 \ v{_\beta } ' } { \right}).\ ] ] since the probability is always bounded by @xmath206 , proving is sufficient . for greater clarity ,
let us write @xmath750 since for @xmath574 , one has @xmath751 , it suffices to bound @xmath752 the sum above is smaller than @xmath753 for @xmath754 , one has @xmath755 , so the latter integral is bounded by @xmath756 and since @xmath757 , this proves , and thus also .
[ closetothere ] there exists @xmath758 such that the following holds .
let @xmath499 satisfy @xmath759 for any @xmath6 small enough , if holds , then @xmath760 \ n { \right } ) , \ ] ] where @xmath510 was defined in and @xmath761 in .
an upper bound on @xmath510 is given by @xmath762}_{l}-1 } { \tilde}{s}_{l , k}^{[g',h ' ] } -\sum_{k \in [ g',h ' ] } s_k { \right } ) , { \mathcal}{e}_n(v{_\beta},v{_\beta } ' ) { \right}].\ ] ] indeed , this corresponds to our partition of sites into @xmath112-intermediate sites ( with @xmath574 ) and important sites , where we drop certain contributions according to the surgery and the coarse - graining .
let us first see that one can find @xmath484 such that when both and hold , one has @xmath502 .
this is true since @xmath763 , and @xmath764 under the assumption , and finally @xmath765 .
hence , as in the proof of proposition [ costofspeed ] , proposition [ controlcuts ] ensures that it suffices to show that , for any sequence of cuts @xmath518 $ ] satisfying @xmath766| { \leqslant}\frac{n}{r{_\beta}^2 v{_\beta}},\ ] ] one has @xmath767}_{l}-1 } { \tilde}{s}_{l , k}^{[g_n , h_n]}-\sum_{k \in [ g_n , h_n ] } s_k { \right } ) , \sum_{k \in [ g_n , h_n ] } x_k \cdot \ell { \geqslant}(1-{\varepsilon})n { \right } ] \\ { \leqslant}c_3 \exp{\left}(-(1 - 6{\varepsilon}){\left}[\frac{dv{_\beta}}{2 } + \frac{{\tilde}{i}{_\beta}}{v{_\beta}'}{\right } ] \ n { \right}).\end{gathered}\ ] ] consider the bound given by part ( 2 ) of proposition [ lotsoflgood ] on the probability of the event @xmath768).\ ] ] in order to see that this bound is smaller than the right - hand side of , it is enough to check that @xmath769 since @xmath770 .
the infimum over all possible values of the right - hand side above is @xmath771 .
it thus suffices to observe that @xmath772 this is true under condition since @xmath773 .
we can thus evaluate the expectation in the left - hand side of , restricted on the event @xmath774).\ ] ] on this event , by definition , the contributions of @xmath112-intermediate sites is bounded from below by a deterministic quantity .
more precisely , part ( 2 ) of proposition [ lotsoflgood ] ensures that the expectation in the left - hand side of , once restricted on the event , is smaller than @xmath775 } s_k { \right } ) , \sum_{k \in [ g_n , h_n ] } x_k \cdot \ell { \geqslant}(1-{\varepsilon})n { \right}].\ ] ] where we recall that @xmath776 was defined in .
the remaining expectation is the same as the one met during the proof of proposition [ costofspeed ] .
one can thus follow the same reasoning ( since we have checked that @xmath502 at the beginning of this proof , it is also true that @xmath500 ) , and arrive at the conclusion , since @xmath777 .
[ corcase2 ] there exists @xmath30 ( depending on @xmath107 ) such that for any @xmath6 small enough , if holds , then @xmath778 with @xmath779 . the fact that @xmath779 was seen in .
the proof of the estimate is similar to that of corollary [ corcase1 ] .
let @xmath528 be a subdivision of @xmath529 , with @xmath780 .
note that since @xmath773 , we have indeed @xmath781
. we can thus apply proposition [ closetothere ] with @xmath782 for any @xmath783 .
the events @xmath784 form a partition of the probability space .
we decompose the expectation defining @xmath56 according to this partition .
we use proposition [ costofspeed ] to evaluate the two last terms thus obtained , and proposition [ closetothere ] to evaluate the other terms , thus obtaining the bound @xmath785 \sqrt{{\tilde}{i}{_\beta } } \ n { \right}).\end{gathered}\ ] ] the use of proposition [ costofspeed ] to evaluate the expectation restricted on the event @xmath786 is legitimate since when holds , one has @xmath787 .
when @xmath532 is taken small enough , the dominant exponential in the sum over @xmath229 that appears in has an exponent which we can take as close as we wish to @xmath788 .
the first two exponentials appearing in are negligible , since as explained above , we have @xmath789 , and remember that we chose @xmath780 .
we now proceed to examine the case when only intermediate sites contribute to the integral , that is , when @xmath790 this case is a minor adaptation of the arguments developed in section [ s : intermediate2 ] . since in this regime , important sites are negligible , it will be harmless to redefine their associated length scale to be @xmath791 by doing so , the most important change is that the length scale @xmath143 is no longer tied with @xmath129 ( that is , we do not have @xmath366 ) .
relation is preserved .
we say that a box @xmath160 ( of size @xmath792 ) is _ locally balanced _ if for any @xmath574 , the proportion of @xmath112-balanced boxes in @xmath598 is at least @xmath599 .
clearly , the notion of a locally balanced box is weaker than the one of a very balanced box . as before
, we say that a box @xmath160 is _ locally good _ if for any @xmath177 such that @xmath793 , the box @xmath179 is locally balanced .
when we refer to previous sections , it is now with the understanding that `` good '' or `` very good '' boxes are replaced by `` locally good '' boxes .
we will now see that techniques developed previously can handle the situation under consideration without additional complication .
we start by discarding the possibility of slow motions .
recall that in our new setting , @xmath383 is a lower bound on the number of @xmath147-locally good steps performed by the random walk .
the property we want to prove is analogous to the one obtained in proposition [ noslowmotion ] .
[ p : inter : noslow ] for any @xmath794 and any @xmath6 small enough , @xmath795 { \leqslant}2 \exp{\left } ( - \frac{\sqrt{i{_\beta}'}}{2c_7 } \ n { \right}).\ ] ] let @xmath796 , and @xmath797 be the event @xmath798 it is a consequence of proposition [ lotsoflgood ] that @xmath799 & { \leqslant } & c_2 \exp{\left } ( - \frac{{\varepsilon}^{-2 } \ { \mathfrak}{i}{_\beta}}{6 \ v{_\beta } ' } \
n{\right } ) \\ & { \leqslant } & c_2 \exp{\left } ( - \frac{{\varepsilon}^{-2 } \ \sqrt{i{_\beta}'}}{6 \
c_7 } \ n{\right}).\end{aligned}\ ] ] this is negligible compared to the right - hand side of .
when one computes the expectation on the left - hand side of restricted to the conjunction of events @xmath800 one finds the upper bound @xmath801 and we thus obtain the result . from now on ,
we fix @xmath802 so that the cost associated to the event @xmath803 is much larger than @xmath804 .
[ controlcuts2 ] recall that @xmath466 is the event defined in .
there exists @xmath479 such that the following holds .
for any @xmath6 small enough and satisfying , the cardinality of the set of surgeries @xmath399 $ ] such that @xmath515 = [ g , h ] , \
\text { and \eqref{inter : noslow } are both satisfied}{\right } ] > 0\ ] ] is bounded by @xmath805 the proof is similar to that of proposition [ controlcuts ] .
the important point is to notice that when holds , one has @xmath806 ( see ) .
we change the definition of the event @xmath489 to the following : @xmath807|
< \frac{n}{r{_\beta}^2 v{_\beta } } , } \\ \displaystyle{{\mathcal}{a}_n \text { and \eqref{inter : noslow } hold } } \end{array}\ ] ] ( compare this definition with ) .
we write @xmath491 for the complement of the event @xmath492 , that is , for the event when either @xmath466 or fails to hold .
we define @xmath510 and @xmath808 as in and , respectively .
[ costofspeed2 ] 1 . for @xmath6 small enough
, one has @xmath809 2 .
let @xmath810 satisfy @xmath500 .
for any small enough @xmath6 , if holds , then @xmath811 { \leqslant}2 \exp{\left } ( -(1 - 5{\varepsilon } ) \frac{d v{_\beta}}{2 } \ n { \right}).\ ] ] 3 . let @xmath499 satisfy @xmath812 . for any small enough @xmath6 , if holds , then @xmath813 \ n { \right}).\ ] ] the proof of part ( 1 ) is identical to the proof of part ( 1 ) of proposition [ costofspeed ] , except that we use proposition [ p : inter : noslow ] instead of proposition [ noslowmotion ] . part ( 2 ) is obtained in the same way as part ( 2 ) of proposition [ noslowmotion ] , while part ( 3 ) is proved as proposition [ closetothere ] .
[ corcase3 ] there exists @xmath30 such that for any @xmath6 small enough , if holds , then @xmath814 with @xmath815 . in order to see that @xmath815 , we recall first from that @xmath816 from proposition [ existm ] , we know that @xmath817 using also , we thus obtain @xmath818 and the announced result follows , since @xmath107 can be chosen arbitrarily small . the proof of the upper bound is identical to that of corollary [ corcase2 ] , except that it is now proposition [ costofspeed2 ] that provides us with the necessary estimates .
we begin this last section by extending theorem [ t : estimen ] to cases when the potential itself may depend on @xmath6 . to this end
, we consider for every @xmath6 small enough , a family of i.i.d .
random variables @xmath819 under the measure @xmath199 , whose common distribution will be written @xmath820 .
we are now interested in @xmath821,\ ] ] and the integral @xmath822 will play the role the integral @xmath87 had previously . [
t : ext ] let @xmath29 . if @xmath823{\text{prob . } } 0,\ ] ] then for any @xmath89 , there exists @xmath30 such that for any @xmath6 small enough and any @xmath20 , @xmath824 if moreover , @xmath825 \xrightarrow[\beta \to 0 ] { } + \infty,\ ] ] where @xmath544 is as in proposition [ existm ] , and for every @xmath52 , @xmath826 { } 0,\ ] ] then there exists @xmath30 such that for any @xmath6 small enough and any @xmath20 , @xmath827 the first part is an adaptation of the results of section [ s : lowerbd ] .
the only difference is that we replace the measure @xmath14 by @xmath820 everywhere .
we need to check that the reference length - scale @xmath143 goes to infinity as @xmath6 tends to @xmath7 . for this
, it suffices to verify that for every @xmath89 , @xmath828 { } 0,\ ] ] and this is a consequence of the assumption in . the second part is obtained in the same way , but following sections [ s : intermediate ] to [ s : intermediate_only ] .
assuming enables to proceed through section [ s : intermediate ] , and the condition in ensures that the scales of reference @xmath829 go to infinity as @xmath6 tends to @xmath7 , uniformly over @xmath112 ( see for instance ) .
it suffices to apply theorem [ t : ext ] , checking that the conditions displayed in , and are satisfied .
since we assume that @xmath831 converges in law , the condition in is clear .
for the same reason , is clear if we can prove that holds . in order to check
, we may appeal to skorokhod s representation theorem , which provides us with random variables @xmath832 which are distributed as @xmath831 for each fixed @xmath6 , and which converge almost surely to @xmath833 distributed as @xmath248 . for convenience , we may assume that these random variables are defined with respect to the measure @xmath199 .
now , by fatou s lemma , @xmath834 \\ & { \geqslant } & { \mathbb{e}}{\left } [ \liminf_{\beta \to 0 } { \overline}{v}{_\beta}\ { \mathbf{1}}_{\beta { \overline}{v}{_\beta}{\leqslant}z_0}{\right } ] \\ & = & { \mathbb{e}}{\left } [ { \overline}{v } { \right } ] = { \mathbb{e}}{\left}[v{\right } ] = + \infty,\end{aligned}\ ] ] and this finishes the proof .
there are simple relations linking @xmath835 to the average of the green function ( in the probabilistic sense ) @xmath836 defined by @xmath837.\ ] ] we refer to ( * ? ? ?
* ( 10 ) ) for precise statements ( note that since we consider only @xmath1 , the function @xmath838 is uniformly bounded ) . as was noted in ( * ? ? ?
* proposition 2 ) , if we define @xmath839 then the function @xmath838 is also the green function of the operator @xmath48 ( in the usual sense of @xmath840 ) , where @xmath49 was defined in . assuming that @xmath23 = + \infty$ ] , the conditions of corollary [ cor : ext ] are satisfied
hence , for any @xmath29 , there exists @xmath30 such that @xmath841 { \leqslant}c \exp{\left } ( -(1-{\varepsilon } ) \sqrt{2d \ { \overline}{{\mathfrak}{i}}{_\beta } } \ n{\right}).\ ] ] ( in fact , a stronger statement can be derived from corollary [ cor : ext ] by replacing the green function by a `` point - to - hyperplane '' version of it . ) moreover , recall that @xmath842 where @xmath820 is the distribution of @xmath831 defined in . by a change of variables
, we can rewrite the integral @xmath843 as @xmath844 where @xmath14 is the distribution of @xmath248 . using the definition of @xmath41 in
, we obtain that this integral is equal to that displayed in .
the decay of the green function as given in is the signature of lifshitz tails and of localized eigenfunctions for energies smaller than @xmath843 ( see @xcite ) , and hence our conjecture .
d. ioffe , y. velenik .
crossing random walks and stretched polymers at weak disorder .
_ * 40 * , 714 - 742 ( 2012 ) . .
self - attractive random walks : the case of critical drifts .
_ , to appear ( 2012 ) . | we consider the simple random walk on @xmath0 , @xmath1 , evolving in a potential of the form @xmath2 , where @xmath3 are i.i.d .
random variables taking values in @xmath4 , and @xmath5 . when the potential is integrable , the asymptotic behaviours as @xmath6 tends to @xmath7 of the associated quenched and annealed lyapunov exponents are known ( and coincide ) . here
, we do not assume such integrability , and prove a sharp lower bound on the annealed lyapunov exponent for small @xmath6 .
the result can be rephrased in terms of the decay of the averaged green function of the anderson hamiltonian @xmath8 .
msc 2010 : 82b44 , 82d30 , 60k37 .
keywords : lyapunov exponents , random walk in random potential , anderson model . |
@xmath0 decay phenomenology is a rapidly growing area of particle physics , as we expect that the precise measurements of various @xmath0 meson decays provide rich information of the standard model and the physics beyond it .
following the successful cleo , lep and cdf experiments , the next generation @xmath0 factories ( babar and belle ) have just started operation , and other experimental projects will follow .
lattice qcd may contribute to this program by calculating the @xmath0 meson decay matrix elements starting from the first principles .
its application is extending from the decay constant @xmath1 and bag parameter @xmath2 to various semileptonic decay form factors .
results are being checked by several groups using different lattice formulations of heavy quark , and systematic errors are carefully estimated .
in addition , studies of new applications have been started , such as the zero recoil form factor of @xmath3 , the width difference of @xmath4 meson , and the lifetime ratio of @xmath5 hadrons . in this
talk i review recent developments in such efforts .
this talk is organized as follows . before discussing the matrix elements calculation
, i will summarize recent progress in the lattice determination of @xmath5 quark mass in the next section .
semileptonic decays are discussed in section [ sec : semileptonic_decays ] .
these are divided into two parts : heavy - to - heavy transitions ( @xmath6 , section [ sec : b->dlnu ] ) and heavy - to - light decay ( @xmath7 , section [ sec : b->pilnu ] ) .
i then describe the @xmath8 mixing , where @xmath1 and @xmath2 are relevant .
recent unquenched calculations of @xmath1 is summarized in section [ sec : f_b ] , and some updates on @xmath2 is discussed in section [ sec : b_b ] .
finally i will discuss new applications of lattice qcd , which include the calculation of @xmath4 width difference ( section [ sec : b_s_width_difference ] ) and @xmath5-hadron lifetime ratios ( section [ sec : lifeteime_ratios ] ) .
those calculations involve similar matrix elements as in @xmath2 .
a summary of the status as of 1998 was given by t. draper at the last lattice conference @xcite .
although the @xmath5 quark mass is not a matrix element related to any exclusive decay of @xmath0 meson , it is necessary in the calculation of many inclusive decay rates .
it has become clear , however , that there is an ambiguity in the definition of pole mass of heavy quark itself ( renormalon ambiguity ) , and its determination can not be better than of order @xmath9 @xcite . in order to avoid the ambiguity one has to use a short distance mass definition , such as @xmath10 quark mass @xmath11 . in the lattice qcd calculation ,
important progress was made by martinelli and sachrajda @xcite , when they obtained the two - loop coefficient in the perturbative matching between continuum qcd and lattice hqet ( static approximation ) @xmath12
\nonumber\\ & & \times \left [ 1-\frac{4}{3}\frac{\alpha_s(\overline{m}_b)}{\pi } -11.66\left(\frac{\alpha_s(\overline{m}_b)}{\pi}\right)^2 \right],\end{aligned}\ ] ] where @xmath13 is obtained with the `` binding energy '' @xmath14 measured in lattice simulations . in this expression
the perturbative expansion in the first parenthesis matches the energy shift in the lattice hqet to the heavy quark pole mass .
a power divergence exists in the loop integral that cancels with power divergent behavior of nonperturbatively calculated @xmath15 .
the last line connects the pole mass to the @xmath10 mass .
the renormalon ambiguity exists in both of these two expansions , while it cancels between them . with the perturbative expansion to two - loop
, these cancellations are expected to become more precise , making the result more stable .
numerical results with the two - loop matching are very stable among three @xmath16 values ( 6.0 , 6.2 and 6.4 ) as shown in figure [ fig : b - quark_mass ] , where corresponding one - loop results are also plotted with various definitions of the coupling constant .
this result suggests that the cancellation of the power divergence and the renormalon ambiguity is working fairly well at two - loop level . a dynamical ( @xmath17 ) result along this line has been reported at this conference by the ape collaboration ( v. gimnez @xcite ) , which is shown by a square in figure [ fig : b - quark_mass ] .
their result does not show significant shift from the quenched results .
the nrqcd collaboration presented a @xmath5 quark mass calculation using nrqcd action at lattice 98 @xcite .
they use a one - loop matching between the kinetic mass in nrqcd and the @xmath10 mass , where no power divergence appears in the perturbative calculation and the one - loop coefficient is not large . their results with ( plus ) and without ( stars )
two - flavour dynamical fermions are consistent with the hqet calculations .
the best estimate from the lattice calculations @xmath11 = [email protected] gev , which i take from the two dynamical results , is in good agreement with recent continuum calculations 4.25(8 ) @xcite , 4.20(6 ) @xcite , 4.20(10 ) @xcite on the @xmath19 resonances , and with delphi s measurement 3.91(67 ) gev at @xmath20 pole @xcite .
the lattice calculation of @xmath0 meson semileptonic decay form factors may be used to determine @xmath21 and @xmath22 with corresponding experimental results of exclusive decays @xmath3 and @xmath23 .
one of the most promising methods to determine @xmath21 is to measure @xmath24 from the differential decay rate of @xmath3 , and extrapolate it to the zero recoil limit of daughter meson @xmath25 , at which the form factor @xmath26 is normalized to unity in the heavy quark mass limit .
theoretical calculations of the power correction to the heavy quark limit have been made using qcd sum rule , but their uncertainty is still quite large ( @xmath27 9% for @xmath28 ) @xcite .
lattice calculation could be an important alternative , if it provides determination better than @xmath29(5% ) .
recently , the fermilab group has proposed a method to calculate the deviation from the heavy quark limit by actually measuring the heavy quark mass dependence of the form factors on the lattice @xcite .
their key observation is that most of statistical and systematic errors cancel in a ratio of matrix elements @xcite @xmath30 where @xmath31 denotes a zero recoil form factor of @xmath32 decay through a vector current @xmath33 .
the statistical error is less than one per cent for typical values of @xmath34 and @xmath35 , and thus it is possible to study its mass dependence .
the renormalization factor to match the heavy - heavy current on the lattice to its continuum counter part is perturbatively calculated , depending on the heavy quark mass , and found to be small for the above ratio @xcite .
they fit their data for @xmath36 as well as for @xmath37 and @xmath38 , which are form factors of @xmath39 and @xmath40 modes obtained through similar ratios , with the form of @xmath41 expansion predicted by the heavy quark symmetry @xcite @xmath42 the parameters @xmath43 , @xmath44 and @xmath15 of @xmath45 terms are determined with the fit .
their preliminary result for @xmath46 form factor obtained at @xmath16=5.7 is @xmath47 = 0.935(22)(@xmath48)(8)(20 ) , where errors represent statistical , mass determination , perturbative and unknown @xmath49 in the given order @xcite .
the systematic error associated with the fermilab formalism of heavy quark @xcite is a subtle issue , especially because the @xmath50 terms in the effective hamiltonian are not correctly tuned with the use of the clover action , while the correction to be measured is of @xmath45 .
nevertheless , using the @xmath41 expansion , it is generally shown that only @xmath51 terms in the action and currents contribute to the ratios calculated above , and thus the systematic error is well under control @xcite .
the result is consistent with the recent qcd sum rule calculation [email protected] @xcite , and the estimated error is already slightly smaller , demonstrating a possibility to improve the determination of @xmath21 using the lattice calculation in near future . in recent high statistics experiments ,
the extrapolation to the zero recoil limit without theoretical constraints seems good enough @xcite .
it is , however , informative to compare the shape of form factors with theoretical predictions in order to check the reliability of the extrapolation , and also to check the theoretical methods .
two new calculations of heavy - to - heavy form factors ( isgur - wise function ) have been presented by ukqcd @xcite and by hein _
@xcite at this conference .
ukqcd used the non - perturbatively @xmath52-improved action at @xmath16=6.0 and 6.2 , and calculated @xmath53 form factor @xmath54 @xcite .
they found no significant dependence on the heavy quark mass , and their preliminary result for the slope parameter is @xmath55 = 1.10(@xmath56)(@xmath57 ) and 1.12(@xmath58)(@xmath59 ) at @xmath16=6.0 and 6.2 , respectively .
the precision is the best among previous results @xcite , and the scaling with two lattice spacings is remarkable .
hein _ et al .
_ presented a very preliminary study with a nrqcd action at @xmath16=5.7 @xcite .
form factor @xmath54 is obtained for @xmath60 and also for @xmath61 mode , where @xmath62 denotes a radially excited state of @xmath63 meson .
perturbative matching factors recently calculated by boyle @xcite are incorporated for heavy - heavy ( axial-)vector currents .
exclusive semileptonic decays @xmath64 could be used for the determination of @xmath22 , provided that corresponding form factors are theoretically calculated .
unfortunately , in the lattice simulations , it is difficult to put large recoil momentum on the daughter meson , so that momentum transfer squared @xmath65 is restricted in the large @xmath65 region ( small recoil momentum ) .
the lattice calculations can be , however , still useful , if statistics in the experiment is precise enough to measure the partial decay rate for the large @xmath65 region , and such work has already been done by cleo for @xmath66 @xcite .
there are two form factors @xmath67 and @xmath68 involved in @xmath69 .
@xmath68 has a negligible contribution to the physical decay rate , as it is proportional to the lepton mass .
it is , however , interesting to study @xmath68 , since the soft pion theorem relates @xmath70 to @xmath1 , which may be used for a test of the lattice calculations . on the other hand ,
the calculation of @xmath67 is of practical importance , as it is directly related to the physical decay rate and may be used for a precise determination of @xmath22 of @xmath71 .
based on the lsz reduction formula and current algebra , the soft pion theorem predicts @xmath72 in the limit @xmath73 and @xmath74 , thus @xmath75 .
it was pointed out by onogi at lattice 97 that the lattice results for @xmath70 and @xmath1 do not seem to satisfy this relation @xcite .
the inconsistency has become even clearer with new data as shown in figure [ fig : soft_pion ] , in which i plot currently available results for @xmath76 ( with renormalization group correction ) , comparing them with representative data for @xmath77 .
the heavy quark scaling law predicts both quantities should be constant , up to @xmath78 corrections .
possible cures for this problem have been proposed at this conference : nonperturbative renormalization and a method of chiral and @xmath65 extrapolation , which i discuss in the following .
the jlqcd collaboration has calculated a ratio of renormalization factor @xmath79 nonperturbatively , using the chiral ward identity @xcite @xmath80 where @xmath81 and @xmath82 denote the light - light axial - current and pseudoscalar density defined on the lattice , and @xmath83 represents the renormalization factor for @xmath81 available nonperturbatively @xcite .
@xmath84 and @xmath85 are temporal component of heavy - light vector and axial - vector currents , which appear in the definition of @xmath70 and of @xmath1 respectively .
their simulation method follows that of maiani and martinelli @xcite , while they use @xmath52-improved action and current for light quark and static heavy quark action .
a preliminary result for the static - light currents at @xmath16=6.0 , @xmath79=0.72(1 ) , is significantly smaller than the one - loop value 0.86(4 ) .
this result indicates that the one - loop calculation of the renormalization factor of the static - light current may contain a large systematic error ( @xmath27 20% ) , and suggests larger @xmath70 and/or smaller @xmath1 in the static limit , which gives right direction to satisfy the soft pion relation , but magnitude is not enough .
an ongoing study to calculate @xmath86 nonperturbatively using the scrdinger functional method has also been presented at this conference by kurth and sommer @xcite .
ukqcd has pointed out that a term proportional to @xmath87 is necessary in the chiral extrapolation as well as the usual @xmath88 term @xcite .
that dependence appears solely from the @xmath87 dependence of @xmath89 : @xmath90 . as a result
the chiral extrapolation could be subtle , since @xmath87 and @xmath88 behave similarly in the region where simulations are made . in order to remove @xmath87 term
they interpolate @xmath68 to several fixed @xmath65 values for each of the active and spectator light quark masses .
then , the chiral extrapolation can be done at each fixed @xmath65 with the @xmath88 term only .
the range of available @xmath65 for all light quark mass is substantially lower than the physical @xmath89 , so they employ a pole dominance model and @xmath91 when they fit with dipole - pole ansatz @xcite or with becirevic - kaidalov model @xcite .
the functional form of @xmath68 are pole - type in both models . ] to obtain the physical @xmath70 from data points .
this extrapolation to the physical @xmath89 makes @xmath70 significantly high 1.3(@xmath92)(@xmath92 ) , which is consistent with @xmath93=1.35(@xmath94)(@xmath95 ) .
although it is a nice observation , the method is , to some extent , model dependent .
therefore , a cross check seems necessary with the direct extrapolation method including @xmath87 and @xmath88 terms , which requires high statistics data at several ( active and spectator ) light quark masses .
milc collaboration have started such a study using fatlink clover quark action for both heavy and light quarks on their dynamical quark configurations @xcite .
they presented a very preliminary result that soft pion theorem is satisfied when extrapolated including @xmath87 term albeit with a large error .
model independent calculation of @xmath67 has great phenomenological importance , since the differential decay rate @xmath96 is proportional to @xmath97 and its measurement may be used for the determination of @xmath22 . near @xmath98 , where lattice calculation is most powerful
, the pole dominance picture should give a good approximation , as the corresponding pole of @xmath99 meson is very close .
using the @xmath100 coupling @xmath101 and @xmath99 meson decay constant @xmath102 , defined by @xmath103 the form factor @xmath67 is given by @xmath104 up to the contribution of higher excited states , which is relatively small for large @xmath65 .
many previous lattice calculations supported or assumed the shape of the pole model @xcite . at this conference jlqcd
has further tested the heavy quark scaling of the pole model using the nrqcd action @xcite .
they fitted observed the measured form factor with ( [ eq : pole_dominance ] ) and extracted the numerator ( pole residue @xmath105 ) .
the heavy quark scaling law predicts @xmath106 and @xmath107 , so that the pole residue should behave as @xmath108 .
itself ( @xmath109 ) , if the daughter pion mass is kept finite .
the denominator of ( [ eq : pole_dominance ] ) behaves as @xmath110 . ]
they confirmed this behavior as shown in figure [ fig : f+_scaling ] , and obtained a very preliminary result @xmath111=0.33(4 ) for the @xmath100 coupling , where @xmath111 is defined through @xmath112 . a direct lattice calculation of the @xmath100 coupling
has recently been made by ukqcd collaboration @xcite .
they use the reduction formula to relate the matrix element @xmath113 to a ` semi - leptonic transition ' amplitude @xmath114 , where @xmath81 is a light - light axial current .
then they use the stochastic propagator to evaluate the latter matrix element .
their result @xmath111=0.42(4)(8 ) is consistent with a phenomenological determination through @xmath115 @xcite and other phenomenological model calculations @xcite . in the other limit @xmath116 , where the pion recoil momentum is large
, the light cone scaling law @xmath117 holds , which is not compatible with the pole dominance model that predicts @xmath118 .
becirevic and kaidalov proposed a model to parametrize @xmath67 , introducing a term representing the effects of higher excited state contributions @xcite . by choosing a parameter of the new term it is possible to produce an additional suppression of @xmath119 , and
the model becomes consistent with the light cone scaling law . using this and other models , such as the pole - dipole model ,
one may extrapolate the lattice data to @xmath116 .
ape @xcite and ukqcd @xcite presented @xmath120 , which are consistent with each other and with previous calculations @xcite . as i discussed before , the comparison with experiment and extraction of @xmath22 is possible with partial decay rate without introducing model dependence .
fermilab group proposed to compare the differential decay rate @xmath121 in the region 400 mev @xmath122 850 mev , where systematic error is minimized @xcite .
an update was reported at this conference @xcite , where they found good scaling between @xmath16=5.7 and 5.9 .
the mass difference of two neutral @xmath123 mesons is measured quite precisely @xmath124 = [email protected] ps@xmath125 , and a bound is known for @xmath4 meson @xmath126 14.3 ps@xmath125 @xcite . to extract the ckm matrix elements @xmath127 and @xmath128 , hadronic parameters @xmath129 and @xmath130
must be obtained theoretically , for which the lattice calculation has been proven to be the best tool . here
i describe updates on the calculation of these quantities . in the quenched approximation ,
the lattice results have stabilized very well @xcite ; many groups agree with each other using different formulation / approach for heavy quark and different lattice spacings as shown in figure [ fig : fb_quenched ] , where recent quenched results with @xmath52-improved actions are plotted as a function of @xmath131 .
the milc collaboration has also done an extensive study with unimproved action , and their result extrapolated to the continuum limit is consistent with the above improved action results @xcite .
the dominant source of systematic errors ( apart from quenching ) depends on method and parameters used , but their typical size is about 10% .
then a natural question is what happens with unquenching . before this conference , two groups ( milc @xcite , collins _
@xcite ) reported dynamical ( @xmath17 ) simulations , and both suggested increase of @xmath1 with unquenching . at this conference milc
updated their calculation with three new dynamical lattices at @xmath16=5.6 ( three different sea quark masses ) , which is consistent with their previous dynamical result @xcite .
a problem in their result is a large @xmath131 dependence seen in the quenched data due to the use of unimproved wilson quark .
a linear extrapolation to the continuum limit gives a substantially lower value compared to the data at finite @xmath131 . on the other hand
, their dynamical results do not show a similar @xmath131 dependence and the continuum limit remains high .
for this reason , although their result suggests @xmath132 , the conclusion is not solid enough
. therefore , they started a new calculation using the fatlink clover action for heavy quark , with which scaling behavior is expected to be improved .
a preliminary result favors lower value of @xmath1 , albeit with large statistical error .
the cp - pacs collaboration presented two new calculations of @xmath1 on their dynamical lattices ( @xmath133=2 ) generated with an rg improved gauge action @xcite : ali khan discussed a nrqcd calculation at @xmath16=1.95 ( @xmath134 1 gev ) with two sea quark masses @xcite . a correction of order @xmath135 of the heavy - light current induced by the operator mixing @xcite is included .
they compared the dynamical result with their quenched result obtained at a similar lattice spacing , and found clear increase with unquenching ( @xmath27 1520% ) , while no difference was found between two sea quark masses .
shanahan presented another systematic study of unquenching at three @xmath16 values ( @[email protected] gev ) with four sea quark masses ( for each @xmath16 ) @xcite .
they used the @xmath52-improved relativistic action for both heavy and light quarks and applied the fermilab reinterpretation for heavy quark @xcite .
the chiral extrapolation was performed with @xmath137 , so the real unquenching was made for the first time .
all dynamical results for @xmath1 are plotted in figure [ fig : fb_dynamical ] together with the recent quenched data .
we observe clear upward shift of @xmath1 with the inclusion of dynamical quarks .
although it is a difficult task to combine the results from different groups , we can crudely say that all available data is consistent with the following estimates : [ cols= " < , < , < " , ] where i also list the results for @xmath138 and @xmath139 .
i do not attempt to extrapolate these results to the physical @xmath140 limit . to do so
, it seems necessary to understand the systematic errors coming from the use of different actions and lattice spacings .
the sea quark mass dependence should also be clarified .
in contrast to the achievement for @xmath1 , the lattice calculation of @xmath2 is still premature . in the static approximation
, the @xmath52-improved results by gimnez and martinelli @xcite and by ukqcd @xcite have been reanalyzed in a recent paper by gimnez and reyes @xcite using corrected one - loop matching coefficient , and a disagreement , which existed between the static - clover and static - wilson results@xcite , has been greatly reduced .
the hiroshima group performed a calculation using the nrqcd action @xcite , and found significant decrease of @xmath141 as one includes @xmath78 corrections .
the matching , however , was done with the coefficient in the static limit , and thus large @xmath142 systematic error is expected for the slope in @xmath78 . in the calculation with relativistic actions
, ukqcd presented the first calculation with the @xmath52-improved action at @xmath16=6.0 and 6.2 at lattice 98 @xcite , which has recently been updated @xcite . to obtain the result at the @xmath0 meson mass , an extrapolation from charm mass regime is necessary , and they found a clear negative slope in @xmath78 . at this conference ,
ape group @xcite has presented the first result obtained using nonperturbative renormalization @xcite .
they found a similar dependence of @xmath2 on @xmath78 , but their final numerical results are not yet available at the time i wrote this contribution .
figure [ fig : b_b ] presents a compilation of lattice data for @xmath143 with @xmath144=5 gev as a function of @xmath145 .
the renormalization factor is introduced to cancel the @xmath146 dependence appearing in the matching factor @xcite .
it is encouraging that all relativistic results including the early works @xcite show a reasonable agreement with each other , and that the recent ukqcd data show a nice scaling between @xmath16=6.0 and 6.2 .
the extrapolation to the static limit ( @xmath27 0.92 ) , however , seems considerably higher than the @xmath52-improved results in that limit .
it suggests that there is unknown sources of systematic error in either or both of static ( nrqcd ) and relativistic calculations .
higher order perturbative corrections ( for both ) and @xmath147 uncertainty in the relativistic calculations are their potential candidates . for this reason , my summary of the current available data includes a large systematic uncertainty : @xmath141 = 0.80(15 ) .
the allowed region on the @xmath148 plane of the ckm matrix is shown in figure [ fig : ckm ] .
the two flavour result @xmath1 = 210 @xmath18 30 mev and the conservative estimate of @xmath2 are used to draw the constraint from @xmath124 . due to the upward shift of @xmath1 from the previous quenched results ,
the allowed region favors @xmath149 0 .
the width difference in the @xmath150 mixing is given as @xmath151 where @xmath152 represents the @xmath153=1 effective hamiltonian @xcite .
only the final states into which both of @xmath4 and @xmath154 can decay contribute .
the @xmath155 expansion induces two four - quark @xmath153=2 operators , whose matrix elements with @xmath4 and @xmath154 states are @xmath2 and @xmath156 , where @xmath156 is defined through @xmath157 and @xmath158 . at this conference ,
the hiroshima group presented a calculation of @xmath156 using the nrqcd action @xcite .
their calculation method is the same as that of @xmath2 and they obtain @xmath159=1.19(2)(20 ) . using a next - to - leading order formula of beneke _ et al . _
@xcite , the width difference is obtained as @xmath160 = 0.16(3)(4 ) , where errors are from @xmath138 and @xmath156 respectively . the two - flavour result for @xmath138 discussed in section [ sec : f_b ]
the latest experimental bound from delphi is @xmath161 0.42 @xcite .
the ratios of lifetime of @xmath5 hadrons , such as @xmath162 and @xmath163 , provide an important test of the theoretical method to calculate the inclusive hadronic decay rates @xcite . in the @xmath155 expansion , the leading contribution to the decay rate comes from a diagram in which the @xmath5-quark decay proceeds without touching the spectator quark , so that it does not contribute to the lifetime ratios .
the @xmath164 correction to the ratios is also small for the same reason , and the first correction involving the spectator quark effect is of @xmath165 , which is parametrized by the ` @xmath0 parameters ' of @xmath153=0 four - quark operators .
ukqcd computed these matrix elements for the first time and obtained @xmath162 = 1.03(2)(3 ) @xcite , which is consistent with the recent experimental result 1.07(2 ) @xcite .
it is a known problem that the lifetime of @xmath166 is surprisingly shorter than that of @xmath0 mesons @xmath163 = 0.79(5 ) @xcite . it is , therefore , interesting to see whether it is explained with the theoretical calculation , in which the similar matrix elements of four quark operators for @xmath166 are required .
the ukqcd group has studied these matrix elements at @xmath16=5.7 with 20 gauge configurations @xcite , and found that the spectator effect is large @xmath1676% .
although their result @xmath163 = 0.91(1)@xmath270.93(1 ) , depending on the light quark mass , is much higher than the experimental value , higher statistics calculations at higher @xmath16 values seem necessary to draw a definite conclusion .
lattice calculations provide model independent predictions for many important @xmath0 decay matrix elements .
progress made for the zero recoil @xmath168 from factors is essential for precise determination of @xmath21 , and the shape of the form factors is also being studied with improved techniques ( np improved action , nrqcd , etc . ) .
more study is necessary to achieve a complete understanding of the @xmath7 form factors : the violation of soft pion theorem and the shape of @xmath67 . the determination of @xmath22 with 10% precision will become possible , once we understand these questions .
the dynamical quark simulation has become practical by several groups , and its effect on @xmath1 has been identified .
further systematic study like that of milc and cp - pacs is necessary to understand systematic errors and eventually to obtain physical prediction at @xmath140 .
an unquenched study of other quantities should also be important .
several new applications have also been studied , such as the width difference of @xmath4 meson , and the lifetime ratios of @xmath5 hadrons . those will become useful theoretical calculations , provided that statistical and systematic errors are improved .
i thank a. ali khan , s. aoki , d. becirevic , c. bernard , s. collins , c. detar , g. douglas , v. gimnez , l. giusti , j. hein , k - i .
ishikawa , a. kronfeld , l. lellouch , c .- j.d .
lin , c. maynard , t. onogi , s. ryan , h. shanahan , j. simone , and n. yamada for communications and useful discussions .
i also thank m. okawa and a. ukawa for comments on the manuscript .
is supported in part by the ministry of education under grant no .
11740162 .
99 t. draper , nucl .
b ( proc . suppl . ) * 73 * ( 1999 ) 43 . for a recent review , see , m. beneke , phys .
* 317 * ( 1999 ) 1 .
g. martinelli and c.t .
sachrajda , hep - lat/9812001 .
v. gimnez _ et al .
_ ( ape collaboration ) , these proceedings , hep - lat/9909138 .
k. hornbostel _ et al . _
( nrqcd collaboration ) , nucl .
b ( proc . suppl . )
* 73 * ( 1999 ) 339 .
m. beneke and a. signer , hep - ph/9906475 .
a.h . hoang , hep - ph/9905550 .
k. melnikov and a. yelkhovsky , phys . rev . *
d59 * , 114009 ( 1999 ) delphi collaboration ( p. abreu _ et al .
_ ) , phys . lett . *
b418 * , 430 ( 1998 ) . for a recent summary ,
see i.i .
bigi , hep - ph/9907270 .
s. hashimoto , a.x .
el - khadra , a.s .
kronfeld , p.b .
mackenzie , s.m .
ryan , j.n .
simone , hep - ph/9906376 .
simone _ et al .
_ , these proceedings .
kronfeld and s. hashimoto , nucl .
b ( proc . suppl . )
* 73 * ( 1999 ) 387 .
falk , m. neubert , phys .
rev . * d47 * , 2965 ( 1993 ) .
t. mannel , phys . rev . * d50 * , 428 ( 1994 ) .
el - khadra , a.s . kronfeld and p.b .
mackenzie , phys . rev . *
d55 * , 3933 ( 1997 ) .
kronfeld , these proceedings , hep - lat/9909085 .
see , for example , delphi collaboration ( m. margoni _ et al .
_ ) , delphi 99 - 107 conf 294 , reported at eps - hep99 .
ukqcd collaboration ( g. douglas _ et al .
_ ) , these proceedings , hep - lat/9909126 .
j. hein _ et al .
_ , these proceedings , hep - lat/9908058 . for a review ,
see j.m . flynn and c.t .
sachrajda , in heavy flavours ( 2nd ed . ) , ed .
buras and m. linder ( world scientific , singapore ) , hep - lat/9710057 .
boyle , these proceedings .
cleo collaboration , ( b.h .
behrens _ et al .
_ ) , hep - ex/9905056 .
t. onogi , nucl .
b ( proc . suppl . ) * 63a - c * ( 1998 ) 59 .
a. abada _
et al . _ ,
* b416 * ( 1994 ) 675 .
ukqcd collaboration ( d.r .
_ ) , nucl . phys . *
b447 * ( 1995 ) 425 . s. hashimoto , k - i .
ishikawa , h. matsufuru , t. onogi , n. yamada , phys .
* d58 * , 014502 ( 1998 ) .
jlqcd collaboration ( s. aoki _ et al .
_ ) , nucl . phys .
b ( proc . suppl . ) * 63a - c * ( 1998 ) 380 .
s. ryan _ et al .
b ( proc . suppl . )
* 73 * ( 1999 ) 390 .
ape collaboration ( d. becirevic _ et al .
_ ) , these proceedings .
jlqcd collaboration ( t. onogi _ et al .
_ ) , these proceedings .
jlqcd collaboration ( s. aoki _ et al .
_ ) , phys . rev .
80 * , 5711 ( 1998 ) .
jlqcd collaboration ( k - i .
ishikawa _ et al .
_ ) , hep - lat/9905036 .
m. lscher , s. sint , r. sommer and h. wittig , nucl . phys . *
b491 * ( 1997 ) 344 . l. maiani and g. martinelli , phys . lett . *
178b * ( 1986 ) 265 .
m. kurth and r. sommer , these proceedings , hep - lat/9908039 .
ukqcd collaboration ( c.m .
_ ) , these proceedings , hep - lat/9909100 .
d. becirevic and a.b .
kaidalov , hep - ph/9904490 .
milc collaboration ( c. detar _
_ ) , these proceedings , hep - lat/9909076 . ape collaboration ( c.r .
_ ) , phys . lett . *
b345 * ( 1995 ) 513 .
ukqcd collaboration ( g.m .
de divitiis _ et al .
_ ) , jhep * 10 * ( 1998 ) 010 . i.w .
stewart , nucl .
phys . * b529 * ( 1998 ) 62 . for a review
see , r. casalbuoni _ et al .
_ , phys . rept .
* 281 * ( 1997 ) 145 .
s. ryan _ et al .
_ , these proceedings , hep - lat/9910010 .
lep b oscillation working group , see their web page for the latest summary , http://www.cern.ch / lepbosc/. a.x .
el - khadra _ et al .
_ , phys . rev . *
d58 * , 014506 ( 1998 ) .
d. becirevic , p. boucaud , j.p .
leroy , v. lubicz , g. martinelli , f. mescia , f. rapuano , phys . rev .
* d60 * , 074501 ( 1999 ) ; see also @xcite for an update .
ukqcd collaboration ( l. lellouch and c.j.d .
lin ) , to appear in the proceedings of heavy flavours 8 , southampton , july 1999 .
a. ali khan _ et al .
b427 * ( 1998 ) 132 .
cp - pacs collaboration ( a. ali khan _ et al .
_ ) , these proceedings . c. bernard _ et al . _ ,
rev . lett . * 81 * , 4812 ( 1998 ) .
s. collins _ et al .
_ , phys . rev . *
d60 * , 074504 1999 .
booth , phys .
d51 * , 2338 ( 1995 ) ; s.r .
sharpe , y. zhang , phys . rev . *
d53 * , 5125 ( 1996 ) .
de divitiis , r. frezzotti , m. masetti and r. petronzio , phys . lett .
* b382 * ( 1996 ) 398 .
milc collaboration ( s. gottlieb _ et al .
_ ) , these proceedings , hep - lat/9909121 .
r. burkhalter ( cp - pacs collaboration ) , nucl .
b ( proc . suppl . )
* 73 * ( 1999 ) 3 .
cp - pacs collaboration ( a. ali khan _ et al .
_ ) , these proceedings . c.j .
morningstar and j. shigemitsu , phys . rev .
* d57 * ( 1998 ) 6741 ; see also , j. shigemitsu , nucl .
suppl . ) * 60a * ( 1998 ) 134 . c. bernard and t. degrand , these proceedings , hep - lat/9909083 .
ishikawa _ et al .
_ , these proceedings , hep - lat/9909159 .
cp - pacs collaboration ( h. shanahan _ et al .
_ ) , these proceedings , hep - lat/9909052 .
v. gimnez and g. martinelli , phys . lett .
* b398 * ( 1997 ) 135 .
ukqcd collaboration , ( a.k .
_ ) , phys . rev .
* d54 * ( 1996 ) 3526 .
v. gimnez and j. reyes , nucl . phys .
* b545 * ( 1999 ) 576 .
ukqcd collaboration ( m. di pierro and c.t .
sachrajda ) , nucl . phys .
* b534 * ( 1998 ) 373 .
j. christensen , t. draper and c. mcneile , phys .
* d56 * , 6993 ( 1997 ) .
s. hashimoto , k - i .
ishikawa , h. matsufuru , t. onogi , s. tominaga , n. yamada , phys .
d60 * , 094503 ( 1999 ) .
n. yamada _
_ , these proceedings , hep - lat/9910006 .
ukqcd collaboration ( l. lellouch and c .- j.d .
lin ) , nucl .
b ( proc . suppl . )
* 73 * ( 1999 ) 357 .
a. donini , v. gimnez , l. giusti and g. martinelli , these proceedings , hep - lat/9909041 . g. martinelli , c. pittori , c.t .
sachrajda , m. testa , a. vladikas , nucl . phys .
* b445 * ( 1995 ) 81 . c. bernard , t. draper , g. hockney and a. soni , phys . rev . * d38 * , 3540 ( 1988 ) .
a. abada _
et al . _ ,
* b376 * ( 1992 ) 172 .
a. soni , nucl .
b ( proc . suppl . )
* 47 * ( 1996 ) 43 .
r. gupta , t. bhattacharya and s. sharpe , phys .
* d55 * , 4036 ( 1997 ) .
m. beneke , g. buchalla , i. dunietz , phys . rev . *
d54 * , 4419 ( 1996 ) m. beneke , g. buchalla , c. greub , a. lenz and u. nierste , phys . lett . *
b459 * ( 1999 ) 631 .
delphi collaboration ( a. borgland _ et al .
_ ) , delphi 99 - 109 conf 296 , reported at eps - hep99 .
m. neubert and c.t .
sachrajda , nucl . phys .
* b483 * ( 1997 ) 339 .
lep b lifetime working group , see their web page for the latest summary , http://home.cern.ch/ claires / lepblife.html .
ukqcd collaboration ( m. di pierro , c.t .
sachrajda and c. michael ) , hep - lat/9906031 . | i review recent developments in lattice calculations of @xmath0 decay matrix elements and other related quantities . |
given the importance of supersymmetry in both particle physics and string theory , searching for supersymmetry seems to be a crucial task in the ongoing and forthcoming colliders .
the forthcoming lhc collider would be able to explore the supersymmetric particles up to a few tev and the ilc collider would allow precision test of supersymmetry . of course , there is no guarantee to find supersymmetry at these collders since the masses of the supersymmetric particles are basically unknown . as is well known , in order to solve the fine - tuning problem in particle theory , the supersymmetric particles should be below tev scale and thus the lhc would be a factory of supersymmetric particles .
however , in the recently proposed split supersymmetry @xcite , the supersymmetric solution of fine - tuning problem in particle physics is given up ( inspired by the need of fine - tuning for the cosmological constant ) , while the virtues of supersymmetry in preserving grand unification as well as providing the cosmic dark matter candidate are still retained . as a result , the mass scale of all sfermions as well as the several heavy higgs bosons can be very high while the gaugino / higgsino mass scale may be still below the tev scale . while the split supersymmetry has the obvious virtue of naturally avoiding the notorious supersymmetric flavor problem , it predicts that no supersymmetric scalar particles except a light higgs boson are accessible at the foreseeable particle colliders .
thus , if split supersymmetry is the true story , the only way to reveal supersymmetry at the colliders is through gaugino or higgsino productions . among the gauginos and higgsinos
, the gluino is the only colored particle and thus may be most copiously produced in the gluon - rich environment of the lhc .
however , the gluino is usually speculated to be much heavier than other gauginos and higgsinos .
it was shown @xcite that the grand unification requirement can allow a gluino as heavy as 18 tev . furthermore ,
if the dark matter is assumed to be the gravitino produced from the late decay of the meta - stable gluino which freezed out at the early universe , it was found @xcite that the gluino must be heavier than about 14 tev and thus impossibly accessible at the lhc .
so , to explore split supersymmetry we should not focus only on gluino productions ; the productions of the electroweakly interacting gauginos and higgsinos ( which mix into two charged particles called charginos and four neutral particles called neutralinos ) should also be considered although their production rates at the lhc are much lower .
of course , if the lhc can discover supersymmetry and then the ilc takes the task of precision test , the productions of charginos and neutralinos at the ilc will play the dominant role . for both the lhc and the ilc , the production of charginos will give good signatures since the subsequent decays yield energetic leptons . to facilitate the collider searches for the charginos and neutralinos , the pre - estimation of their allowed mass regions are important .
while the direct experimental searches , such as lep and tevatron experiments , gave robust lower bounds for the masses of charginos and neutralinos , the cosmic dark matter can give some upper bounds and thus have important implications for the searches at future colliders .
therefore , although the consequence of split supersymmetry in the dark matter issue has been considered to some extent in the literature @xcite , we in this work scrutinize the dark matter constraints on the masses of charginos and neutralinos and evaluate the corresponding production rates at the lhc and ilc .
this work is organized in the follows . in sec .
ii we recapitulate the parameter space of the sector of charginos and neutralinos . in sec .
iii we examine the dark matter constraints .
we will show the constraints on ( a ) the original parameter space , ( b ) the masses of charginos and neutralinos , and ( c ) the production rates at the lhc and the ilc .
the conclusions are given in sec .
iv . note that for the supersymmetry parameters we adopt the notation in ref .
we work in the framework of the minimal supersymmetric model ( mssm ) and assume the lightest supersymmetric particle is the lightest neutralino , which solely makes up the cosmic dark matter .
also , we fix the parameter @xmath2 since a large value of @xmath3 is favored by current experiments and , in the region of large @xmath4 , our results are not sensitive to @xmath3 .
the chargino mass matrix is given by @xmath5 and the neutralino mass matrix is given by @xmath6 where @xmath7 and @xmath8 are respectively the @xmath9 and @xmath10 gaugino mass parameters , @xmath11 is the mass parameter in the mixing term @xmath12 in the superpotential , and @xmath13 is ratio of the vacuum expectation values of the two higgs doublets .
the diagonalization of ( [ mass1 ] ) gives two charginos @xmath14 with the convention @xmath15 ; while the diagonalization of ( [ mass2 ] ) gives four neutralinos @xmath16 with the convention @xmath17 .
so the masses and mixings of charginos and neutralinos are determined by four parameters : @xmath7 , @xmath8 , @xmath11 and @xmath3 .
the current constraints @xcite on these parameters are divided into two classes : ( 1 ) direct constraints from experimental searches of supersymmetric particles ; ( 2 ) indirect constraints from some precisely measured low - energy processes or physical quantities via supersymmetric quantum effects . for split supersymmetry , almost all indirect constraints from low - energy processes , such as various @xmath18-decays , drop out since the supersymmetric loop effects in these processes usually involve sfermions which are superheavy in split supersymmetry .
the most stringent direct bounds are from lep experiments @xcite : ( i ) the lighter chargino @xmath19 must be heavier than about 103 gev ; ( ii ) the lsp must be heavier than about 47 gev ; ( iii ) the value of @xmath3 must be larger than 2 .
note that in addition to the direct lower bound from lep ii , theoretically a large @xmath3 helps to push up the lightest higgs boson mass and thus ameliorate the stress between the experimental lower bound and the theoretical upper bound on the lightest higgs boson mass .
so in our analyses we assume a large @xmath3 and fix it to be 40 .
we checked that in the region of large @xmath4 , our results are not sensitive to @xmath3 .
we assume the cosmic dark matter is solely composed of the lsp , which is assumed to be the lightest neutralino @xmath20 .
( if the gravitino is assumed to be the lsp and make up the dark matter , it will incur some severe cosmoligical constraints @xcite . ) the thermal relic density of the lightest neutralino from the freeze - out can be calculated from the boltzmann equation which involves the thermal averaged cross section of neutralino annihilations . in our work
we use the package darksusy @xcite and we checked that the package micromegas @xcite gives the similar results in our study . in calculating the cross section of neutralino annihilations ,
many additional supersymmetric parameters are involved , among which the most important ones are sfermion masses and @xmath21 ( the mass of cp - odd higgs boson ) . since we focus only on split supersymmetry , all sfermion masses and @xmath21 are superheavy .
so any diagrams involving a sfermion or a heavy higgs boson makes negligible contributions to neutralino annihilations .
actually , we found that as long as the sfermion mass or @xmath21 gets heavier than about 10 tev , the effects of sfermions or heavy higgs bosons decouple , as shown in figs . 1 and 2 .
the peak in fig.1(b ) happens around the @xmath22-funnel resonance point @xmath23 .
the @xmath24 allowed region for the dark matter relic density is @xmath25 which can be inferred from the wilkinson microwave anisotropy probe ( wmap ) measurements @xcite . in the following
we present the @xmath24 allowed regions for four cases : ( 1 ) @xmath0 ; ( 2 ) @xmath11 is superheavy ; ( 3 ) @xmath8 is superheavy ; and ( 4 ) @xmath7 is superheavy . in our calculations
we fix a superheavy mass to be 100 tev since it is high enough for the relevant supersymmetric particles to decouple from the neutralino annihilations . in each case
we present the allowed region for ( a ) the original parameter space ; ( b ) the chargino mass @xmath26 versus the neutralino mass @xmath27 ; ( c ) the cross section of chargino pair production at lhc and ilc versus the chargino mass @xmath26 .
note that we just give the tree - level cross sections and do not include the one - loop corrections @xcite .
the center - of - mass energy is 14 tev for lhc and is assumed to be 1 tev for ilc @xcite .
the results are resented in figs .
2 - 5 , where the dot - dashed line is the lep ii lower limit on chargino mass . _
( 1 ) @xmath0 : _ this case is well motivated since the supergravity models predict the unification relation @xmath28 . in the low mass region for both @xmath8 and @xmath11 in fig .
2(a ) , the dark matter is the mixing of gauginos and higgsinos .
the strip with very large @xmath8 corresponds to higgsino dark matter , while the strip with very large @xmath11 corresponds to gaugino dark matter ( the mixing of bino and wino ) . from fig .
2(b ) we see that both the chargino mass @xmath26 and the neutralino mass @xmath27 is upper bounded by about 1 tev .
2(c ) shows that in the allowed region the cross section of chargino pair production at lhc can reach a few pb for a light chargino , but drops rapidly as the chargino gets heavy .
the cross section at ilc can reach 100 fb for a light chargino in the allowed region . _
( 2 ) superheavy @xmath11 : _ this case was proposed and favored by some authors @xcite because the @xmath11 problem @xcite is avoided and a crude gauge coupling unification is preserved . in the low mass region for both @xmath8 and @xmath7 in fig .
3(a ) , the dark matter is the mixing of bino and wino ; while the region with large @xmath7 corresponds to wino dark matter .
3(b ) shows that both the chargino mass @xmath26 and the neutralino mass @xmath27 is upper bounded by about 3 tev .
3(c ) shows that in the allowed region the cross section of chargino pair production at lhc can reach 10 pb for a light chargino , but drops rapidly as the chargino gets heavy .
the cross section at ilc can reach 200 fb for a light chargino in the allowed region . _
( 3 ) superheavy @xmath8 : _ in the low mass region for both @xmath7 and @xmath11 in fig .
4(a ) , the dark matter is the mixing of bino and higgsinos .
when @xmath7 ( @xmath11 ) gets very large , a strip is remained , which corresponds to higgsino ( bino ) dark matter .
the chargino mass @xmath26 is upper bounded by about 3 tev and the neutralino mass @xmath27 is upper bounded by about 1 tev . for a light chargino in the allowed region ,
the cross section of chargino pair production can reach the level of pb at lhc and 100 fb at ilc . _
( 4 ) superheavy @xmath7 : _ in fig .
5(a ) the strip with large @xmath8 ( @xmath11 ) corresponds to higgsino ( wino ) dark matter . as shown in fig .
5(b ) , both the chargino mass @xmath26 and the neutralino mass @xmath27 are lower bounded by about 1 tev and upper bounded by about 2.5 tev .
thus the charginos can not be pair produced at the ilc with c.m .
energy of 1 tev .
although the charginos can be pair produced at the lhc , the cross section is very small , as shown in fig .
note that while the chargino pair production with a rate of 100 fb at ilc may not be hard to observe due to the clean enviroment of the ilc ( chargino pair production is regarded as a good way to test split supersymmetry at ilc @xcite ) , searching for the chargino pair production with a cross section of pb level at the lhc may be quite challanging .
the chargino @xmath19 decays into a neutralino @xmath20 and a pair of fermions ( two jets or a charged lepton plus a neutrino ) .
so the signature of chargino pair production is ( i ) two energetic leptons plus missing energy , or ( ii ) one energetic lepton plus two jets plus missing energy .
let us take the latter signature , i.e. , @xmath29 , as an example .
the huge background comes from @xmath30 . in order to substantially reduce this background
, we may apply a cut on the transverse mass defined by @xmath31 @xmath32 is always less than @xmath33 ( and peaks just below @xmath33 ) if the only missing energy comes from a neutrino from @xmath34 decay , which is the case for the @xmath30 background events . for the signal
, @xmath32 is spread about equally above and below @xmath33 , due to the large extra missing energy from the neutralinos .
therefore , we may , for example , require @xmath35 gev .
given the importance of chargino pair production as a test of split supersymmetry at the lhc , detailed monte carlo studies with the consideration of various backgrounds are needed , which is beyond the scope of this work .
in split supersymmetry , gauginos and higgsinos are the only supersymetric particles which are possibly accessible at the lhc or the ilc collider .
the masses of these particles are subject to the stringent constraints from the cosmic dark matter . under the assumption that the lightst neutralino is the lsp and constitute the dark matter in the universe , we scrutinized the dark matter constraints on the masses of charginos and neutralinos .
we considered several cases : ( 1 ) @xmath0 ; ( 2 ) @xmath11 is superheavy ; ( 3 ) @xmath8 is superheavy ; and ( 4 ) @xmath7 is superheavy .
we found that the lightest chargino @xmath19 must be lighter than about 1 tev in the first case and about 2 or 3 tev in other cases . in the first three cases ,
the corresponding production rate of the chargino pair at the lhc ( ilc ) can reach the level of pb ( 100 fb ) in some parts of the allowed region and thus hopefully observable .
but in the last case , the chargino must be heavier than about 1 tev and thus has a too small production rate to be observale at the lhc .
so , overall , there is no guarantee to find the charginos of split supersymmetry at the lhc or ilc collider .
we thank bing - lin young and fuqiang xu for discussions . this work is supported in part by national natural science foundation of china . 11 n. arkani - hamed , s. dimopoulos , hep - th/0405159 ; g.f .
giudice , a. romanino , 699 , 65 ( 2004 ) .
n. arkani - hamed , s. dimopoulos , g. f. giudice , a. romanino , 709 , 3 ( 2005 ) . c. kokorelis , hep - th/0406258 ; 732 , 341 ( 2006 ) ; b. kors , p. nath , 711 , 112 ( 2005 ) ; k. s. babu , t. enkhbat , b. mukhopadhyaya , 720 , 47 ( 2005 ) ; e. dudas , s. k. vempati , 727 , 139 ( 2005 ) .
l. senatore , 71 , 103510 ( 2005 ) ; a. masiero , s. profumo , p. ullio , 712 , 86 ( 2005 ) .
f. wang , w. y. wang , j. m. yang , hep - ph/0507172 .
a. pierce , 70 , 075006 ( 2004 ) ; a. arvanitaki , p. w. graham , hep - ph/0411376 .
a. masiero , s. profumo , p. ullio , 712 , 86 ( 2005 ) . h. e. haber and g. l. kane , phys . rep . * 117 * , 75 ( 1985 ) ; j. f. gunion and h. e. haber , 272 , 1 ( 1986 ) .
s. eidelman , _ et al .
_ , particle data group , 592 , 1 ( 2004 ) .
lep2 susy working group , http://lepsusy.web.cern.ch / lepsusy/. j. cao , j. m. yang , 71 , 111701 ( 2005 ) . j. l. feng , a. rajaraman , f. takayama , 91 , 011302 ( 2003 ) ; 68 , 063504 ( 2003 ) ; j. l. feng , s. su , f. takayama , hep - ph/0404198 ; hep - ph/0404231 ; f. wang , j. m. yang , 38 , 129 ( 2004 ) ; k. hamaguchi , y. kuno , t. nakaya , m. m. nojiri , 70 , 115007 ( 2004 ) .
p. gondolo , j. edsjo , l. bergstrom , p. ullio , m. schelke , e. a. baltz , astro - ph/0406204 .
g. belanger , f. boudjema , a. pukhov , a. semenov , hep - ph/0112278 .
wmap collaboration , astrophys .
* 148 * , 1 ( 2003 ) ; * 148 * , 175 ( 2003 ) .
w. beenakker , r. hpker , m. spira and p. m. zerwas , 492 , 51 ( 1997 ) ; w. beenakker , m. klasen , m. krmer , t. plehn , m. spira and p. m. zerwas , 83 , 3780 ( 1999 ) ; prospino2.0 , http://pheno.physics.wisc.edu/@xmath36plehn k. abe _ et al .
_ , acfa linear collider working group , hep - ph/0109166 ; t. abe _ et al .
_ , american linear collider working group , hep - ex/0106056 ; j. a. aguilar - saavedra _ et al .
_ , ecfa / desy lc physics working group , hep - ph/0106315 .
k. cheung , c .- w .
chiang , 71 , 095003 ( 2005 ) .
j. e. kim and h. p. nilles , 138 , 150 ( 1984 ) ; y. nir , 354 , 107 ( 1995 ) ; m. cvetic and p. langacker , 54 , 3570 ( 1996 ) s. h. zhu , 604 , 207 ( 2004 ) ; w. kilian , t. plehn , p. richardson and e. schmidt , 39 , 229 ( 2005 ) . | in split supersymmetry , gauginos and higgsinos are the only supersymmetric particles which are possibly accessible at foreseeable colliders .
while the direct experimental searches , such as lep and tevatron experiments , gave robust lower bounds on the masses of these particles , the cosmic dark matter can give some upper bounds and thus have important implications for the searches at future colliders .
in this work we scrutinize such dark matter constraints and show the allowed mass range for charginos and neutralinos ( the mass eigenstates of gauginos and higgsinos ) .
we find that the lightest chargino must be lighter than about 1 tev under the popular assumption @xmath0 and about 2 or 3 tev in other cases .
the corresponding production rates of the lightest chargino at the cern large hadron collider ( lhc ) and the international linear colldier ( ilc ) are also shown .
while in some parts of the allowed region the chargino pair production rate can be larger than 1 pb at lhc and 100 fb at the ilc , other parts of the region correspond to very small production rates and thus there is no guarantee to find the charginos of split supersymmetry at future colliders .
= # 1,nucl .
phys .
b * # 1 * , # 1,phys .
lett .
b * # 1 * , # 1,eur .
phys .
j. c * # 1 * , # 1,phys . rev .
d * # 1 * , # 1,phys .
rev .
lett . * # 1 * , # 1,mod .
phys . lett .
a * # 1 * , # 1,j .
phys .
g*#1 * , # 1,commun .
theor .
phys . * # 1 * , # 1tt@xmath1#1 # 1#20.5ex |
observations of type ia supernovae ( sne ia ) at high redshift has led to the astonishing deduction that the universal expansion is accelerating @xcite .
a key factor in this discovery has been the painstaking calibration of the relation between the sne ia luminosity and the behaviour of the light curves @xcite .
however , the size of the effect ( about 0.25 mags ) which indicates acceleration is several times smaller than the intrinsic range of luminosity in local sne ia _
i.e. _ the correction that has to be made using the calibration relation is large . given that the calibration relation is derived empirically , and that the physics of sne ia is only partially understood , it is therefore important to exploit every opportunity to test for intrinsic differences in the sne or in their environments between the local and high - z universe . in this letter
we consider the rates , luminosities and extinction of sne ia at @xmath8 .
the current consensus on the diversity of sne ia is that it arises primarily from the mass of @xmath10ni produced in the explosion , with a higher relative @xmath10ni producing a more luminous and more slowly declining lightcurve .
close - binary evolutionary calculations @xcite of possible progenitors indicate that the c / o ratio is lower for higher mass co cores .
it has been suggested , although not demonstrated , that a smaller c / o ratio in the progenitor white dwarf yields less @xmath10ni and so a lower luminosity @xcite .
thus , co white dwarf progenitors arising from higher mass main - sequence stars may yield less luminous sne ia .
metallicity can have a significant effect via the white dwarf wind which affects the minimum co white dwarf mass that can accrete as far as the chandrasekhar limit @xcite .
a high metallicity implies a higher white dwarf wind .
this allows white dwarfs with a lower co mass to reach the chandrasekhar limit and hence produce brighter sne ia .
thus , there will be a tendency for a larger number of , and more luminous sne ia to occur in high metallicity populations and vice versa .
it is predicted that metallicity in spirals will decrease with increasing redshift @xcite .
such a decrease is not expected to occur in ellipticals since they have had essentially no star formation over this look - back time .
consequently we might expect that , relative to ellipticals , sne ia in high-@xmath9 spirals should exhibit reduced luminosities and lower rates .
conversely , modelling of the total sne ia rate ( per @xmath11 @xmath12 ) with redshift predicts little change in the total sne ia rate with increasing redshift up to @xmath13 @xcite . this prediction has been shown to be true up to at least @xmath14 by independent observations @xcite .
moreover , @xcite also model the variation in sne ia rate with redshift for host galaxies of different morphological types , and find that significant changes are anticipated only beyond [email protected] . at lower redshifts ,
the absolute and relative rates are predicted to remain roughly the same as the local values for spirals and ellipticals .
another important issue in current studies of high-@xmath9 sne ia is that of extinction . both supernova search teams correct for galactic extinction but find that very few of their supernovae show a residual colour excess , implying negligible extinction in the host galaxies .
no significant extinction is found in 14 out of 16 sne ia discovered by @xcite and in 40 out of 42 sne ia discovered by @xcite .
@xcite also argue against the presence of neutral extinction .
they point out that the observed dispersion in mlcs - derived distances is substantially smaller than would be expected if grey dust extinction were the main cause of the lower brightness of high-@xmath9 sne ia .
however , @xcite has highlighted inconsistencies in the treatment of host galaxy extinctions between the two sne ia search teams .
moreover , @xcite has shown that by employing a consistent treatment of extinction and excluding those sne ia observed only after maximum light , a positive @xmath16 is only required at the @xmath17 level .
the apparent lack of host galaxy extinction is , perhaps , surprising for two reasons .
firstly , local sne ia are observed with a range of host galaxy extinctions spanning @xmath18 , with most lying in the range @xmath19 @xcite .
secondly , the increasing gas content in galaxies with increasing redshift means that , contrary to the findings of the sn search teams , we might expect mean extinction levels to _ increase _ with redshift due to increased dust opacities caused by the higher gas column densities @xcite .
however , it is possible that the apparent lack of extinction at high - z is due to observational bias .
sne ia behind a dust screen will be fainter , and thus would be selected against in searches near the magnitude limit .
in addition , follow - up spectroscopy favours those sne ia that are a large distance from the host galaxy centres in order to minimise contamination from host galaxy light .
these sne ia may be expected to have less extinction than those found near the host galaxy centres .
.high redshift type ia supernovae host galaxies [ sne1ahosts ] [ cols="<,^,^,^,^,^ " , ] j2000 coordinates are quoted .
peak b band magnitudes , given in the rest - frame of the supernova , are taken from @xcite and include corrections using the stretch - factor method .
sn1997eq is present in both the @xcite sample and our sample in table [ sne1ahosts ] .
the final f814w host galaxy images are presented in figure [ sn_images ] and morphological classifications are given in table [ sne1ahosts ] .
spectra for several of the supernovae in our sample have been published @xcite , although no spectra of the host galaxies themselves have been published .
where possible we have used the contamination from host galaxy light in the published sne ia spectra to refine the morphological classifications further .
we note however that classification of host galaxy type based on the observed morphology only is still robust in distinguishing between elliptical and spiral systems . however , the images on their own are insufficiently detailed to be used to divide the spiral systems into subtypes ( e.g. sbc , sbbc etc . ) .
the sample exhibits morphologies ranging from pure ellipticals ( e.g. the hosts of sn 1997cj , sn 2000ec ) to systems with clear signs of spiral structure ( the hosts of sn 1997ce , sn 1997ek , sn 1998bi ) .
the host galaxy of sn 1997es appears to be disturbed .
overall , four hosts have a morphology consistent with an e / s0 classification , while the remaining eleven either exhibit spiral structure or are disturbed . a wide range of sn distances from the galaxy centres is also found , ranging from @xmath0kpc ( sn 1997es , sn 2000ee ) to @xmath1kpc ( sn 1998 m .
note that the object directly above the supernova is a foreground galaxy . )
the projected distances are presented in table [ sne1ahosts ] .
the colour maps for the host galaxies of sn 2000ea and sn 2000eg are presented in figure [ sn_colmaps ] .
sky background regions in both maps have been masked out . in the rest frame of the galaxies
, the colour corresponds closely to @xmath20 .
we note that , owing to the way in which the maps were constructed , they can not be used to accurately measure the supernova colours .
large - scale , irregular colour variation of order @xmath21 is seen across both galaxies , with extreme differences as high as @xmath22 .
to improve the statistics we added seven more sne ia host galaxies at a similar redshift , imaged by @xcite using stis on hst .
these galaxies are presented in table [ ellishosts ] .
two of these are classified as e / s0 and five as spirals .
an eighth galaxy from @xcite , the host of sn 1997eq , is common to both samples .
we agree with @xcite that it is a spiral type .
thus , for a combined sample of 22 sne ia host galaxies , @xmath23 are e / s0 systems and @xmath24 are spirals or disturbed systems .
these proportions are similar to that observed locally by @xcite .
these authors also find a ratio of about @xmath23 to @xmath24 when their results are expressed as sn rate per unit bolometric luminosity of the host galaxy .
adopting the same conversion between sne ia counts per galaxy type and sne ia rates per unit galaxy luminosity as @xcite , we infer that the relative sne ia rate in spirals and ellipticals has not changed between @xmath8 and the local universe .
this is consistent with the prediction of @xcite that the relative rates should be unchanged in going from z=0.6 to the local universe .
peak rest - frame b band magnitudes for the sne in our sample have so far been published for only two objects @xcite ; sn 1997ce ( @xmath5=19.26 ) and sn 1997cj ( 19.25 ) .
these are the magnitudes obtained following application of mlcs correction .
however , stretch - factor - corrected b band peak magnitudes for 7 out of the 8 host galaxies presented in @xcite have been published in @xcite .
these magnitudes are listed in table [ ellishosts ] .
the total of 9 sne ia comprises 3 in ellipticals and 6 in spirals .
although the two teams observe their supernovae using different filter sets , the light curve fitting methods employed by both teams give stretch factor corrected magnitudes in the same filter , namely b band in the rest - frame of the supernovae .
we thus intercompare published magnitudes from both teams .
the mean peak b band magnitude for the sne ia in the 6 spiral systems is @xmath25 .
for the 3 sne ia in elliptical systems the mean peak b band magnitude is @xmath26 .
thus , we find no significant difference in the average light - curve - shape - corrected @xmath5 for high - z sne ia between spirals and ellipticals .
we thus find that light - curve shape correction methods @xcite appear to be valid for sne ia at @xmath8 in all host galaxy types .
the irregular colour structure in two of the host galaxies ( figure [ sn_colmaps ] ) is most plausibly interpreted as being due to variation in a significant dust extinction ( this also supports their classification as spirals ) .
we also note that simulations of host galaxy extinctions and radial distributions of supernovae @xcite have shown that sne ia within @xmath27kpc of the centres of spiral galaxies will be observationally dimmer and have a larger magnitude dispersion than sne ia in the outer regions , although part of the observed dispersion will probably be due to projection effects .
inspection of figure [ sn_images ] shows that the sne ia are found over a range of projected distances from the host galaxy centre .
these distances are listed in table [ sne1ahosts ] .
it is possible that the real distances from the host galaxy centres are in fact all @xmath7kpc and that the observed distances are simply due to projection effects , however we argue against this for two reasons .
firstly , 10 out of the 15 sne in our sample have projected distances from the host galaxy nuclei of @xmath28kpc , which argues against a distribution in a shell of @xmath7kpc in radius .
secondly , the distances from galaxy centres for local sne ia range from @xmath29kpc to @xmath1kpc , and there is no compelling physical reason why the distribution of galactocentric distances of sne ia at z=0.6 should differ from the distribution observed locally .
overall , our results are consistent with a similar range of real separations of sne ia from the centres of their host galaxies as is observed locally .
we find no evidence to suggest that sne ia at @xmath8 are found preferentially far out in the host galaxies where extinction levels might be expected to be small .
moreover , current observational and theoretical evidence does not favour any significant changes in typical galaxy size at @xmath8 relative to that observed locally @xcite .
consequently we can also state that there is no difference in the distribution with galaxy scalelength between local sne ia and those at @xmath8 .
we are led to suspect that a significant fraction of the 22 events may be subject to some extinction . given that dustier environments at higher redshifts are also to be expected @xcite , the general lack of extinction found in high-@xmath9 sne ia discovered by @xcite and @xcite remains puzzling .
we do however note the following two caveats .
firstly , the possibility of some observational bias ( _ cf .
_ section 1 ) operating within a region closer than @xmath1510 kpc is not ruled out .
secondly , details about the sne ia in our two `` dusty '' spirals ( figure [ sn_colmaps ] ) have yet to be published , and so we are unable to comment directly on the degree of extinction that these events might have encountered . in summary , we find that , at @xmath8 , the observed relative rates and relative luminosities of type ia supernovae in elliptical and spiral galaxies are consistent with predictions based on the locally - derived understanding of sne ia physics and the influence of progenitor mass and metallicity @xcite .
we find no reason to question this understanding .
unless some level of observational bias is present , there is however still some difficulty with the apparently low observed extinction towards sne ia at high - z .
we thank bruno leibundgut for helpful discussions . we would like to thank the referee for very helpful comments .
the data were obtained from the multimission archive at the space telescope science institute ( mast ) .
stsci is operated by the association of universities for research in astronomy , inc .
, under nasa contract nas5 - 26555 .
support for mast for non - hst data is provided by the nasa office of space science via grant nag5 - 7584 and by other grants and contracts .
the nasa / ipac extragalactic database ( ned ) is operated by the jet propulsion laboratory under contract with nasa .
the digitized sky surveys were produced at the space telescope science institute under u.s .
government grant nag w-2166 .
the images of these surveys are based on photographic data obtained using the oschin schmidt telescope on palomar mountain and the uk schmidt telescope .
this work was in part supported by pparc ( grant number gr / k98728 ) .
financial support was provided to s.m . by osk . | we examine the host galaxies of high redshift type ia supernovae ( sne ia ) using archival i and r band data from the hubble space telescope .
the sne ia host galaxies show a wide variety of morphologies , including undisturbed ellipticals , spirals and disturbed systems .
sne ia are also found over a wide range of projected distances from the host galaxy centres , ranging from @xmath0kpc to @xmath1kpc . for a sample of 22 sne ia at @xmath2 , @xmath3
are found in spiral galaxies and @xmath4 are found in elliptical systems , similar to the proportions observed locally . including data from @xcite
, we find no significant difference in the average light - curve - shape - corrected @xmath5 for high - z sne ia between spirals and ellipticals .
these results are consistent with predictions based on the locally - derived understanding of sne ia physics and the influence of progenitor mass and metallicity .
we also construct colour maps for two host galaxies and find that both show a non - uniform colour structure with typical variations of rest - frame @xmath6 .
this is most plausibly attributed to the presence of , and variation in dust extinction in these galaxies .
moreover , we find no evidence that the sne ia are preferentially found in outer regions ( @xmath7 kpc ) of the host galaxies where extinction would be low .
this suggests that the range of host galaxy extinctions of sne ia at @xmath8 should be comparable to those of local sne ia . although observational bias can not be completely ruled out , this appears to be in conflict with the finding of low extinction for sne ia found in the high-@xmath9 supernova search studies .
[ firstpage ] supernovae : general galaxies : high - redshift galaxies : stellar content
cosmological parameters |
the need to communicate secretly has always been an important issue for military strategists during war time .
the one - time pad , first proposed by vernam , has been shown to be one of the most secure means of encrypting a message provided the key is truly random and the key is as long as the message @xcite . however , a major problem with the one - time pad is the establishment of a secure key between the two physically separated parties without the services of a courier .
recently , there has been a major proposal to apply the laws of quantum mechanics to establish this crucial key .
this new proposal , called quantum key distribution ( qkd ) protocols , therefore involves the use of quantum features such as uncertainty principle or quantum correlations to establish a the necessary key and hence provides unconditionally secure communication .
the first quantum key distribution was proposed by bennett and brassard ( bb84 ) in 1984 based on the fact that any measurement on an unknown state of a polarized photon by a third party will always disturb the state and hence detectable .
an extension of the scheme to three - dimensional quantum states has recently been done @xcite and it was shown to be more secure than two - dimensional case .
another well - known variation of qkd is based the idea of an entangled pair and detecting the presence of the eavesdropper using violations of the bell - clauser - horne - shimony - holt ( bell - chsh ) inequality @xcite .
this protocol ( ekert protocol ) is fundamentally interesting as it provides an example of how a fundamental problem in quantum mechanics , namely bell - chsh inequality and violation of local realism , can be applied to a physical problem . naturally , one questions if it is possible to extend this latter protocol involving bell - chsh inequality to higher dimensional system .
the extension of bell - chsh inequality to higher dimensions is a non - trivial and interesting problem . as higher dimensional quantum systems require much less entanglement to be non - separable than two - dimensional systems ( qubits ) ,
it was suspected that higher dimensional entangled systems may lead to stronger violations of local realism .
these results have been shown numerically using linear optimization method by searching for an underlying local realistic joint probability distribution that could reproduce the quantum predictions @xcite and confirmed analytically @xcite .
the quantum channel we consider consists of a source producing two qutrits @xcite , which we denote by @xmath0 and @xmath1 , in the maximally entangled state @xmath2 , where @xmath3 and @xmath4 are the @xmath5-th basis state of the qutrit @xmath0 and @xmath1 respectively ( these basis states can represent , for instance , spatial degrees of freedom of photons ) . qutrit @xmath0 flies towards alice whereas qutrit @xmath1 flies towards bob .
each observer has at his or her disposal a symmetric unbiased six - port beamsplitter .
an unbiased symmetric six - port beamsplitter performs a unitary transformation between `` mutually unbiased '' bases in the hilbert space @xcite .
such devices were tested in several quantum optical experiments @xcite , and also various aspects of such devices were analyzed theoretically @xcite .
this quantum optical device has three input and three output ports . in front of each input port
there is a phase shifter .
when all the phase shifters are set to zero an incoming photon through one of the input ports has an equal chance to leave the device through any of the output ports .
the elements of the unitary transformation , which describes its action , are given by @xmath6 where @xmath7 and the indices @xmath5 , @xmath8 ( @xmath9 ) denote the input and exit ports respectively ; @xmath10 are the phase shifters .
these phase shifters can be changed by an observer . for convenience
, we will denote the values of the three phase shifts in the form of a three dimensional vector @xmath11 . in our protocol
both observers perform three distinct unitary transformations on their qutrits .
the transformations at alice s side are defined by the following vectors of phases @xmath12 , @xmath13 , @xmath14 whereas the transformations at bob s side are defined by @xmath15 , @xmath16 , @xmath17 .
the observers choose their transformations randomly and independently for each pair of incoming qutrits . after performing the transformation defined by the vectors of phases @xmath18 the state @xmath19 reads @xmath20 .
the observers perform the measurement of the state of the qutrit in the basis in which @xmath19 is defined , that is , @xmath21 ( @xmath22 ) . we have adopted an uncommon but useful complex value assignment to the results of the measurements , first used in @xcite : namely , for the result of the measurement of the ket @xmath23 we ascribe the value @xmath24 .
this value assignment naturally leads to the following definition of the correlation function @xmath25 ( @xmath26 for short ) between the values of alice s and bob s results of measurements @xcite @xmath27 where @xmath28 denotes the probability of obtaining the result @xmath29 by alice and the result @xmath30 by bob for the respective values of the phase shifts they have used .
it can be shown that the above correlation function reads @xmath31 \ , , \label{corr}\ ] ] where , for instance , @xmath32 denotes the second component of the @xmath5-th vector of phases for alice .
note that @xmath33 .
this means that the results of the measurement obtained by alice and bob are strictly correlated . when alice obtains the results @xmath34 bob must register the results @xmath35 respectively .
thus , only the following pairs of the results are possible @xmath36 ( denoted subsequently by @xmath37 ) and each pair of correlations occurs with the same probability equal to @xmath38 .
let us also define the following quantity @xmath39 it can be shown @xcite , using the recently discovered bell inequality for two qutrits @xcite , that according to local realistic theory @xmath40 can not exceed @xmath41 .
however , when using the quantum mechanical correlation function ( [ corr ] ) , @xmath40 acquires the value @xmath42 .
therefore , to violate the above bell inequality in this case one must reduce the correlation function by the factor @xmath43 ( such reduction is possible by adding the symmetric noise to the system ) .
it has been proved @xcite that the above bell inequality gives necessary and sufficient conditions for local realism in this case .
after the transmission has taken place , alice and bob publicly announce the vectors of phase shifts that they have chosen for each particular measurement and divide the measurements into two separate groups : a first group for which they have used the vectors @xmath44 , @xmath45 and @xmath46 , @xmath47 , and a second group for which they have used @xmath48 .
subsequently , alice and bob announce in public the results of the measurements they have obtained but only within the first group . in this way
they can compute the value of @xmath40 .
if this value is not equal to @xmath42 it means that the qutrits have somehow been disturbed .
the source of this disturbance can be either an eavesdropper or noise . in case of no disturbance
the results from the second group allow them , due to the mentioned correlations , to generate a ternary cryptographic key .
for instance when alice gets the sequence of values , say @xmath49 then bob must get the following sequence of results , @xmath50 .
let us consider a symmetric incoherent attack in which the eavesdropper ( eve ) controls the source that produces pairs of qutrits used by alice and bob to generate the cryptographic key .
naturally , if eve wants to acquire any information about the key , she must introduce some disturbance to the state of the qutrits .
her only chance of being undetected is to hide herself behind what , to alice and bob , may look like an environmental noise in the channel .
we assume that the noise is symmetrical in the sense that the correlation function in the presence of it reads @xmath51 where @xmath52 .
this requirement can only be fulfilled if the reduced state for alice and bob ( after tracing out eve s degrees of freedom ) is of the form @xmath53 where the real ( not necessarily all positive ) numbers @xmath54 , and where the maximally entangled orthogonal states @xmath55 ( @xmath56 ) read @xmath57 this choice of states stems from the fact that only the above states generate correlation functions that are proportional to @xmath58 . to be more specific , the state @xmath59 gives the correlation function @xmath60 whereas the state @xmath61 gives the correlation function @xmath62 .
thus , if we compute the correlation function on the state @xmath63 , we arrive at the following formula @xmath64 from eq.([noise ] ) , we obtain the condition @xmath65 , which is only possible if @xmath66 ( @xmath67 is real ) .
eve can prepare the reduced density operator ( [ state ] ) by preparing an entangled state of the form , @xmath68 are the computational basis states of the two qutrits , and @xmath69 are states of ancilla . without loss of generality , we can assume that they are normalized ( which implies that @xmath70 ) . note that the most general state of the joint system of alice s and bob s qutrits and eve s ancilla reads @xmath71 . however , eq . ( [ state ] ) and the requirement that @xmath72 imposes the following conditions on the states of the ancilla @xmath73 denoting @xmath74 by @xmath75 we arrive at the following set of conditions @xmath76 eve s strategy is the following .
she prepares the state ( [ ancilla ] ) , sends the qutrits to alice and bob and keeps her ancilla .
she then waits for public communication between alice and bob .
when the settings of alice s and bob s apparatus ( phase shifts ) are revealed , eve adopts the following algorithm : ( i ) if the chosen settings are not the ones used for the key generation she ignores the ancilla ; ( ii ) if the settings are the ones for which the key is generated , i.e. , @xmath77 , she identifies the ancilla state .
let us first find the transformed state in case ( ii ) , i.e. , the state @xmath78 .
a straightforward computation yields @xmath79 where the un - normalized states @xmath80 read @xmath81 note that ( [ eq : transstate ] ) can also be written more conveniently as where we have grouped the terms into three orthogonal subspaces associated with alice and bob generating the correct key @xmath37 , and the two incorrect keys , @xmath83 or @xmath84 . note also that the ancilla states of one subspace are orthogonal to the ancilla states of the other subspaces .
the probability that eve projects into the subspaces spanned by the states @xmath85 and @xmath86 are @xmath87 respectively .
we have considered the fact that the states within each bracket in eq.([groupeq ] ) have the same norms with the same mutual scalar products . moreover , these scalar products are all real .
eve now has to determine the state of her ancilla , given that alice and bob have projected the whole state into one of three subspaces associated with the three cases .
these subspaces are orthogonal so that eve can , in principle , determine without error , which of these cases alice and bob have .
the three ancilla vectors in each subspace corresponding to the result obtained by alice and bob are symmetric and equiprobable .
this makes eve s task of discrimination easier as this case has an analytic optimal solution @xcite using the so - called `` square - root measurement '' .
we define the operator @xmath88 , where @xmath89 are the ancilla states spanning the subspace associated with alice and bob s measurement outcomes .
since we are discriminating 3 vectors in a 3-dimensional space , the optimum measurement directions , @xmath90 are orthogonal , hence eve simply performs a projective measurement on her ancilla ( fig .
[ fig : discrim ] ) .
thus , eve s error rate is given by @xmath91 where @xmath92 is the probability of correctly identifying the three states of the ancilla in the i - th subspace .
these probabilities are given by @xmath93 where @xmath94 due to the symmetry of the noise introduced by eve , the error rate between alice and bob determined using eq.([state ] ) and the conditions in eq.([condition ] ) is @xmath95 we also note that whenever eve eavesdrops , the correlation function obtained by alice and bob is reduced by @xmath96 .
therefore , if this factor is less than @xmath97 , the bell inequality is not violated @xcite and so alice and bob will abort the protocol .
this implies that eve must keep this factor above this value .
fig.[fig : error ] shows the three dimensional plots of the error rates of eve as a function of the parameters @xmath98 and @xmath75 ( labeled by surface i ) as well as the error rate between alice and bob ( labeled by surface ii ) .
the region in which the factor @xmath96 is greater than the threshold value ( @xmath99 ) is demarcated by the `` wall '' labeled @xmath100 . in the region bounded by @xmath101 , the error rate of eve is always greater than the error rate between alice and bob .
an alternative approach to test the security of the protocol against such incoherent symmetric attack is to consider the mutual information between alice and eve and compare it with the mutual information between alice and bob .
the mutual information between alice and eve is given by the following expression fig .
[ mutual ] shows the plan elevation of the 3-dimensional plots of the mutual information as a function of the parameters @xmath98 and @xmath75 .
the line of intersection between @xmath104 and @xmath105 clearly lies behind the wall separating the region in which the bell inequality is violated from the region ( @xmath106 ) in which local realistic description is possible ( @xmath107 ) . in the region
@xmath106 , @xmath108 . from numerical calculation ,
the maximum value of v for which eve s mutual information equals alice and bob s is 0.6629 .
thus , alice and bob have a buffer region in which to operate securely from this kind of attack by eve . to summarize ,
we have presented a cryptographic protocol using qutrits which is resistant to a form of symmetric , incoherent attacks .
the qutrit bell inequality provides a sufficient condition for secure communication .
however , this attack may not be optimal so the bell inequality may prove to be necessary .
shannon , bell syst . tech .
j. , * 28 * 656 ( 1949 ) .
d. bruand c. macchiavello , phys .
88 * , 127901 ( 2001 ) .
ekert , phys .
lett . * 67 * , 661 ( 1991 ) .
d. kaszlikowski , p. gnaciski , m. ukowski , w. miklaszewski and a. zeilinger , phys .
* 85 * , 4418 ( 2000 ) .
d. kaszlikowski , l. c. kwek , j .- l .
chen , m. ukowski and c. h. oh , quant - ph//0106010 .
d. collins , n. gisin , n. linden , s. massar , s. popescu , quant - ph//0106024 .
t. durt , d. kaszlikowski , and m. ukowski , private communication ( 2000 ) .
j. schwinger , proc .
acad . sc . *
46 * , 570 ( 1960 ) .
i. d. ivanovic , j. phys .
a * 14 * , 3241 ( 1981 ) .
w. k. wooters , found .
* 16 * , 391 ( 1986 ) . c. mattle , m. michler , h. weinfurter , a. zeilinger and m. ukowski , appl .
b * 60 * , s111 ( 1995 ) .
m. reck , phd thesis ( supervisor : a. zeilinger ) ( university of innsbruck , 1996 , unpublished ) .
m. reck , a. zeilinger , h. j. bernstein and p. bertani , phys .
lett . * 73 * , 58 ( 1994 ) .
i. jex , s. stenholm and a. zeilinger , opt . comm . * 117 * , 95 ( 1995 ) .
jing - ling chen , d. kaszlikowski , l. c. kwek and c. h. oh .
d. kaszlikowski , l. c. kwek , jing ling chen , m. ukowski , and c. h. oh , phys . rev .
a * 65 * , 032118 ( 2002 ) .
chen , d. kaszlikowski , l.c .
kwek , c.h .
oh and m. zukowski , phys .
a , * 64 * , 052109 ( 2001 ) .
a. chefles , _ contemporary physics _ * 41 * , 401 ( 2000 ) . | we present a crytographic protocol based upon entangled qutrit pairs .
we analyse the scheme under a symmetric incoherent attack and plot the region for which the protocol is secure and compare this with the region of violations of certain bell inequalities .
= msbm10 scaled # 1_#1 _ # 1*#1 * # 1#1 |
a new form of generalized nonextensive entropy recently proposed by us@xcite has been shown in to give small but interesting departures from the shannon case in terms of thermodynamic properties of a system in a manner similar to but also somewhat different from tsallis entropy @xcite .
conceptually , the new entropy appears from a novel definition of entropy in terms of the rescaled phase cells due to correlated clusters , and in a limit similar to the tsallis case approaches shannon s classical extensive entropy .
kaniadakis has also suggested @xcite the use of deformed functions leading to unusual forms of entropy , from a kinetic principle related to phase space , which gives excellent results for cosmic ray spectra . in our case
the definition of entropy is particularly simple , as it can be expressed simply as the divergence of a vector representing the modified probabilities for the different possible states taking into account a rescaling due to correlations or clustering due to interactions between the microsystems . in microscopic systems
quantum entanglement of states is also a relevant issue .
some authors @xcite have studied the problem of quantum entanglement of two states in the picture of tsallis type nonextensive entropy .
the generalization of shannon entropy to the very similar von neumann entropy using density operators in place of probability distributions @xcite reveals common features of the stochastic and the quantum forms of uncertainties and this treatment can be extended to tsallis form too .
our purpose here is to present a combined study of stochasticity and quantum entanglement , so that the former emerges from the quantum picture in a natural way , and then we intend to show that our new approach of defining entropy also allows us to obtain a measure of mutual information that involves stochasticity and entanglement together in a clear comprehensible way .
the fact that our new definition of entropy , which is conceptually very simple , also gives the probability distribution function in a closed form in terms of lambert @xmath0 functions @xcite allows one to carry out many calculations with the same ease as for tsallis entropy . in this work , however , the probability distribution will not be needed for explicit use .
entropy is intuitively associated with randomness , because it is a measure of the loss of information about a system , or the indeterminacy of its exact state , which in turn depends on the probability distribution for various states .
a uniform probability distribution function ( pdf ) among all states indicates maximal uncertainty in state space and gives the maximal entropy , whereas dirac / kronecker delta ( continuous or discrete states ) pdf with no uncertainty has zero entropy .
combinatorics gives the boltzmann form @xmath1\ ] ] because in equilibrium the @xmath2 are simply @xmath3 , where @xmath4 is the total number of subsystems , and @xmath2 is the number of subsystems in the i - th state . in terms of the @xmath5 themselves
one gets the shannon form given below .
it is well - known that maximizing the entropy with the constraints @xmath6 and ( with @xmath7 the energy of the i - th state , and u the total energy , which is fixed ) @xmath8 gives the exponential probability distribution @xmath9 where the lagrange multiplier constant @xmath10 can be identified as the inverse of the temperature .
let us now consider shannon coding theorem @xcite : when the letters @xmath11 of the alphabet used in a code have the probabilities @xmath5 , then it can be shown fairly easily that a stream of random letters coming out of the source with the given probabilities in the long run will relate the entropy to the probability of the given sequence : @xmath12\ ] ] where @xmath13 is the entropy per unit and @xmath4 is the large number of letters in the sequence .
we shall now define entropy from a somewhat different viewpoint which takes into account interaction among the units , producing clusters of size @xmath14 units for the i - th state .
this effective size may in general be a fraction , and if the interaction is weak , the average cluster size @xmath14 is just over unity . if we think of liquid clusters , the typical subsystem in state @xmath15 may bean assembly of @xmath16 molecules , but this may change due to environmental factors , such as ph value , to @xmath17 , so that we have a rescaling value of @xmath18 , which may be greater or less than 1 .
in general we allow @xmath14 to be different for each @xmath15 .
since @xmath5 is the probability of a single occurrence of the i - th state , i.e. for a cluster of size unity ( which may consist of a typical number of subunits ) , the probability for the formation of a cluster of size @xmath14 is @xmath19 . let us now consider the vector @xmath20 this is @xmath21-dimensional , where @xmath21 is the number of single unit states available to the system components .
let us now consider the phase space " defined by the @xmath14 co - ordinates . as we have said above ,
the deviations of these parameters from unity give the effective ( which may be fractional when an average is taken ) cluster sizes in each of the states . a value smaller than unity indicates a degeneration of the micro - system to a smaller one in a hierarchical fashion , partially if it is a fraction . in other words
we are considering a scenario where clusters may form superclusters or be composed of subclusters , with a corresponding change of scale in terms of the most basic unit obtainable .
we have dealt elsewhere with the interesting question an oligo - parametric hierarchical structure of complex systems @xcite , but here , we restrict ourselves to cluster hierarchy changes that do no qualitatively change the description of the system . hence , if we take the divergence of the vector @xmath22 in the @xmath14 space , it is a measure of the escape of systems from a given configuration of correlated clusterings . and , inversely , the negative of the divergence shows the net influx of systems into an infinitesimal cell with cluster sizes @xmath14 . if all the @xmath14 are unity , then we have unfragmented and also non - clustered , i.e. uncorrelated units at that hierarchy level
. we can argue first from the point of view of statistical mechanics that this negative divergence or influx of probability , may be interpreted as entropy .
@xmath23 we know that the free energy is defined by @xmath24 where @xmath25 is the free energy , @xmath26 is the internal energy , and @xmath13 is the entropy .
@xmath27 is the temperature , or a measure of the average random thermal energy per unit with boltzmann constant @xmath28 chosen to be unity , and hence @xmath29 is a measure of the random influx of energy into the system due to the breaking / making of correlated clusters due to random interactions in a large system .
this allows the subtracted quantity @xmath25 to be the free energy or the useful " energy .
there are usual thermodynamic phase space factors in dealing with a macroscopic system , which we drop as common factors in what follows . in terms of the shannon coding theorem
also we arrive at the same expression .
since , with @xmath14 average clustering the i - th state occurs with probability @xmath30 , a stream of units ( clustered ) emitted will correspond to the probability @xmath31 = \prod_i p_i^{p_i^{q_i}}\ ] ] which too gives us eqn .
[ ent ] . in @xcite
we have developed the statistical , mechanics of this entropy in detail , by first obtaining its probability distribution function in terms of the lambert w function .
however , we used , for simplicity an isotropic ( in state space ) correlation and rescaling , i.e. we had a single common @xmath32 , as in tsallis entropy . in the rest of this work we shall use the same simpler expression .
when the state vector in the combined hilbert space @xmath33 of two particles ( or subsystems in @xmath34 and @xmath35 ) can not be expressed as the factorizable product of vectors in the hilbert spaces of the two subsystems , it is by definition entangled .
hence entanglement is actually a property related to projection in the subspaces , and can not be expected to be measurable by properties in the bigger space alone . given the state @xmath36 which gives density matrix for the @xmath37 space for the trace over the @xmath38 part @xmath39 where the @xmath40 are now the coefficient matrices . in terms of density matrices
an entangled state ( for an explicit example ) of two qubits ( a qubit " , or quantum bit , being a quantum superposition of two possible states ) may be expressed by the reduced @xmath41 matrix from the @xmath42 basis sub - set : @xmath43\ ] ] with @xmath44 for the pure quantum ( entangled ) state @xmath45 , and @xmath46 .
this entanglement occurs in the subspace of the product hilbert space involving only the two basis vectors @xmath47 and @xmath48 .
other entangled combinations are equivalent to this form and may be obtained from it simply by relabeling the basis vectors , and hence we shall use this as the prototype . for @xmath49 , we have an impure state with a classical stochastic component in the probability distribution , although we still have probability conservation as @xmath50 which remains unchanged under any unitary transformation .
a possible measure of the factorizability ( purity " @xcite ) @xmath51 of a quantum state , or its quantum non - entanglement , remains invariant under changes of @xmath52 @xmath53 \nonumber \\
= c^4 + s^4 .
\end{aligned}\ ] ] so , for the maximum entanglement @xmath54 when @xmath55 , and the minimal entanglement corresponds to @xmath56 ( pure factorizable states ) when @xmath57 .
quantum impurity represented by classical stochasticity attains the maximum value when @xmath58 , and is nonexistent when @xmath44 corresponding to a pure entangled state .
we note that @xmath51 does not involve the stochasticity - related parameter @xmath52 at all , but remains the quantifier of the quantum entanglement .
another equivalent but conceptually possibly more interesting way of quantifying entanglement may be the parameter @xmath59 - tr[\rho_a]tr[\rho_b ] ) = \sin^2(2 \theta)\end{aligned}\ ] ] which is more symmetric in the two subspaces and is similar to a correlation function .
it gives 0 for no entanglement when @xmath60 , and maximal entanglement 1 for @xmath55 , as desired .
this definition of entanglement is in the spirit of mutual information , though we have not used the entropy at this stage , but only the probabilities directly .
it too does not involve the stochasticity in terms of the purity parameter @xmath52 . in the relation above we have used @xmath61\ ] ] and similarly for @xmath62 . in our specific case , for @xmath25 or for @xmath63 , @xmath64\ ] ] with @xmath65 = 1 $ ] ensured .
it is possible to formulate the stochasticity by coupling the entangled state @xmath66 to the environment state @xmath67 quantum mechanically and then taking the trace over the environment states .
@xmath68 and the trace over the environment yields @xmath69 for the entangled mixture of @xmath47 and @xmath48 in @xmath70 and the couplings @xmath71 with @xmath72 and @xmath73 , the trace over @xmath74 states gives the density matrix @xmath75\ ] ] so , classical stochasticity has been introduced by taking the trace over the environment space with @xmath76
let us consider a single system @xmath25 interacting with the environment @xmath77 .
the product space @xmath78 contains entanglement between the measured system and the environment , and hence the density operator for the combined system - environmental space is as given in eqn .
[ rhogam ] with @xmath44 to indicate a pure entangled state , and the environment - traced density is given by eqn .
[ rhoabtr ] . here
@xmath79 and @xmath80 are equal .
mutual information may be identified as the entanglement , and defined by @xmath81 - ( tr_{a}[\rho_a])^2 ) = \sin^2(2\theta')\end{aligned}\ ] ] as before , with @xmath82 the angle of entanglement .
hence , measurements on the system @xmath25 reflects the coupling of the system to the environment to the environment , and the mutual information is contained in the parameters of the system itself . in terms of von neumann entropy , which , with mixing in an orthogonal quantum basis , becomes similar to shannon entropy @xcite , @xmath83\ ] ] .
we know from the araki - lieb relation@xcite @xmath84 that , with @xmath85 in a pure quantum state , we must have @xmath86 , and hence @xmath87\end{aligned}\ ] ] which too confirms the view that the system itself contains in its parameters the mutual information in such a case , as we found above . if we use our new form of entropy@xcite with the hypothesis that the mutual information is still given by the same form with the parameter @xmath32 not equal to @xmath88 , which is the case for shannon entropy , then we get @xmath89 \nonumber \\ = - 2c'^{2q}\log(c'^{2q } ) -2 s'^{2q } \log(s'^{2q})\end{aligned}\ ] ] with the 3-system entanglement shown in eqn .
[ psiabe ] and the relatively simple choice of couplings in eqn .
[ carr ] , we have already shown @xmath90 in eqn .
[ rhoc ] .
similar construction of @xmath91 , @xmath92 and @xmath93 , and defining the 3-system mutual information as @xmath94 and with @xmath95 for any @xmath32 , for a single 3-system pure state , we may find the 3-system mutual information .
tracing over the @xmath63 space gives @xmath91 , which , using as basis @xmath47 , @xmath96 , @xmath97 and @xmath48 in the @xmath98 product space , yields @xmath99\ ] ] and an identical matrix for @xmath92 .
finally we get using the relevant eigenvalues @xmath100 where the eigenvalues @xmath101 and @xmath102 are for the @xmath90 matrix obtained after tracing over @xmath77-space .
@xmath103)\ ] ] with @xmath52 given by eqn .
[ gamma ] .
had we started with a stochastic picture of entangled impure a - b system , with @xmath104 representing the stochasticity , as we have suggested above , then the mutual information would be @xmath105 in fig .
[ fig1 ] we first show the mutual information ( mi ) calculated according to the shannon form of the entropy , which is equivalent to our form at @xmath106 , as a function of the @xmath107 entanglement angle @xmath108 and the entanglement angle of @xmath109-system with the environment @xmath82 , which is related to the stochasticity @xmath52 as explained above .
we note that mutual information is virtually independent of the angle of entanglement with the environment @xmath82 .
hence , it seems that traditional entropy in this case is insensitive to details of coupling with the environment when the mutual information between two systems is measured .
@xmath110 as a function of entanglement angle @xmath108 in a - b space and the entanglement angle @xmath82 with the environment , which is related to the stochasticity.,width=302 ] in fig .
[ fig2 ] and in fig .
[ fig3 ] we show the deviations @xmath111 of our mi from the shannon mi , as a function @xmath32 and entanglement angles @xmath82 and @xmath108 respectively , keeping the other angle at @xmath112 in each case .
there is symmetry around @xmath112 .
the variation is fairly smooth for fixed @xmath82 , i.e. fixed entanglement with the environment .
however , if the entanglement between @xmath25 and @xmath63 is kept fixed at near @xmath112 , then the mutual information using our form of entropy changes sharply with @xmath82 near the symmetry value @xmath113 .
one can see that this comes from one of the eigenvalues of the density matrix @xmath90 approaching zero for this mixing value , and with @xmath114 , the there is either a sharp peak or dip compared to the shannon entropy case , which has fixed @xmath115 .
difference of mi from our entropy with that from shannon entropy at @xmath116,width=302 ] same as fig .
[ fig2 ] with @xmath117,width=302 ] fig .
[ fig2 ] shows little variation with changing @xmath82 for almost any @xmath32 .
[ fig3 ] shows pronounced changes at small q ( @xmath118)for different @xmath108 .
mi difference between our entropy form and shannon for @xmath119,width=302 ] same as fig .
[ fig4 ] but for @xmath120,width=302 ] in fig .
[ fig4 ] and fig .
[ fig5 ] we show the difference between our mi and the shannon mi as a function of @xmath108 and @xmath82 simultaneously , keeping @xmath32 fixed at @xmath121 and at @xmath122 .
of course at @xmath106 , we get no difference , as our entropy then coincides with the shannon form . here
too we notice that the mixing angle between @xmath25 and @xmath63 shows fairly smooth variation , but @xmath82 or equivalently the stochasticity , causes pronounced peak ( for @xmath123 ) , or dip ( for @xmath124 ) . here
too we can conclude that our method of entropy calculation can indicate a greater role of the entanglement with the environment when this mixing is nearly equal for the @xmath107 entangled states .
difference between mi from our entropy and tsallis s with @xmath55,width=302 ] same as fig .
[ fig6 ] but with @xmath117.,width=302 ] in view of the prevalent familiarity with the tsallis form of nonextensive entropy , we have previously @xcite compared our form with results from tsallis entropy in formulating a general thermodynamics .
there we showed that despite the conceptual and functional differences between tsallis entropy and our new form , the results are very similar if we take the tsallis @xmath32 to be twice as far from unity ( the shannon equivalent value ) as our value of @xmath32 . in fig .
[ fig6 ] and fig .
[ fig7 ] we show the difference of our mi from that derived from tsallis s entropy .
we again note that among @xmath108 and @xmath82 differences are both relatively more significant for @xmath32 values different from @xmath88 , for both angles near @xmath125 , with peaks and dips similar to the comparison with the mutual information calculated with shannon entropy .
it is interesting to note here that recently in a study of the entropy of a chain of spins in a magnetic field @xcite it has been found that both the usual von neumann and renyi forms of entropy yield nonzero and surprisingly simple closed expressions . though this work does not mention entanglement explicitly , the correlation functions presented here , which determine the density matrix and hence its diagonalized form needed for entropy calculation , actually are manifestations of the entanglement among the spins and between the spins and the magnetic field .
the chain has been split into two parts , similar to our a and b subsystems , and the external magnetic field acts like the environment we have introduced in this work . though they carry out their extensive calculations at zero temperature , unlike our finite temperature treatment , the fact they obtain nonzero @xmath126 for the first @xmath127 spins is apparently due to the segmentation of the pure state of the fully entangled quantum system and the consideration of part @xmath25 only for the entropy calculation , which effectively is equivalent to summing the states of part @xmath63 and the entanglement with the environment , and produces entropy due to the corresponding loss of information about state of the whole system .
hence , their results for this explicit model is consistent with our general result that classical stochasticity and entropy may be a reflection of segmented consideration of bigger complete systems .
the values of the entropy of different types such as the canonical shannon form or generalized forms such as the renyi form , which goes to the shannon form in the usual limit , like that of the related parameter we have mentioned for tsallis entropy and for our new form of entropy in this work , reflect the extent of entanglement or interaction or , equivalently , correlation . in their work
a length scale comes out of this segmentation , which appears to be similar to the angle of entanglement in our case .
we do not get a phase transition as they do , because we have considered a simplified general finite system of only two or three component subsystems , not an infinite chain , and finite systems can not show any phase transitions .
we have shown how a simple definition of the entropy in terms of influx of states into cells of any given cluster sizes in various states can give us a new nonextensive form of entropy with a closed form of the probability distribution function , and which coincides with the usual form for uncorrelated microsystems .
we have then seen that classical stochasticity can be derived from quantum entanglement with the environment , and it influences the mutual information between quantum states .
the functional form of the mi differs as per the definition of the entropy , but the numerical differences in mi resulting from various forms of entropy usually differ rather subtly according the parameters of entanglement , within the system , and with the environment , as well as the scaling type parameter @xmath32 introduced in the tsallis form and in the new form introduced by us .
however , for the angle of entanglement with the environment our entropy differs from both the shannon entropy and tsallis entropy when both angles of entanglement are near the symmetry point @xmath112 .
the differences between the forms become more pronounced as @xmath32 varies from unity .
entanglement and mutual information are such fundamental concepts that experimental tests need to be designed to distinguish from minute quantitative differences the appropriateness of various theoretical forms of entropy .
theoretical works like the study of large systems such as spin chains @xcite may also help differentiate the appropriateness of various forms of entropy such as shannon , tsallis , renyi or our suggested new form , and the quantification of mutual information . the notion of clusters changing constantly into various sizes , which is the basis of our definition of the new form of generalized entropy may be the most relevant concept for the treatment of liquids and other material , where such phenomena form an integral part of the dynamics .
the author would like to thank andrew tan and ignacio sola for discussions and g. kaniadakis and j.f .
collet for useful feedback .
00 1999 quantum entanglement inferred by the principle of maximum tsallis entropy .
a _ * 60 * , 34613466 . 2002 nonadditive entropies and quantum entanglement . _ physica a _ * 306 * , 316 1970 entropy inequalities .
phys . _ * 18 * , 160170 .
2001 classical and quantum complexity and non - extensive thermodynamics .
_ chaos , fractals and solitons _ * 13 * , 367370 .
2004 quantum spin chains , toeplitz determinants and the fisher - harwig conjecture .
_ j. stat .
_ * 116 * , 7995 .
2001 nonlinear kinetics underlying generalized statistics . _
physica a _ * 296 * , 405425 .
2002 statistical mechanics in the context of special relativity .
e _ * 66 * , 056125 .
( john wiley , ny , 1998 ) p. 368 .
( cambridge u.p . ,
ny , 2000 ) 1994 the classical n - body problem within a generalized statistical mechanics._j .
* 27 * , 57075757 . 2007
the lambert function and a new nonextensive entropy _
i m a j. appl .
( in press ) 2007 oligo - parametric hierarchical structure of complex systems .
_ neuroquantology journal _ * 5 * , 8599 1988 possible generalization of boltzmann - gibbs statistics . _ j. stat .
phys . _ * 52 * , 479487 .
2002 entanglement versus bell violations and their behaviour under local filtering operations .
lett . _ * 89 * , 170401 .
1999 entanglement and nonextensive statistics .
lett a _ * 260 * , 335 - 339 . | we first show how a new definition of entropy , which is intuitively very simple , as a divergence in cluster - size space , leads to a generalized form that is nonextensive for correlated units , but coincides exactly with the conventional one for completely independent units .
we comment on the relevance of such an approach for variable - size microsystems such as in a liquid .
we then indicate how the entanglement and purity of a two - unit compound state can depend on their entanglement with the environment .
we consider entropies of tsallis , which is used in many different real - life contexts , and also our new generalization , which takes into account correlated clustering in a more transparent way , and is just as amenable mathematically as that of tsallis , and show how both purity and entanglement can appear naturally together in a measure of mutual information in such a generalized picture of the entropy , with values differing from the shannon type of entropy .
this opens up the possibility of using such an entropy in a quantum context for relevant systems , where interactions between microsystems makes clustering and correlations a non - ignorable characteristic . |
the _ kepler _ mission has discovered from its first 16 months data over @xmath1 planetary candidates @xcite .
over one third ( @xmath2800 ) of these candidates are in multiple transiting candidate planetary systems , and one remarkable feature of them , as shown by @xcite and @xcite , is that the vast majority of candidate pairs are neither in nor near low - order mean motion resonance ( mmr hereafter , see also in @xcite ) , however there are small but significant excesses / deficits of candidate pairs slightly wider / narrow of the exact resonance ( or nominal resonance center ) , particularly near the first order mmr , such as 2:1 and 3:2 mmr . such an intriguing asymmetric period ratio distribution has stimulated a number of theorists recently , who developed different models to understand and interpret it .
@xcite consider that such an asymmetric period ratio distribution around mmr could be an outcome of resonant couples having underwent eccentricity damping during some dissipative evolutions , such as tidal dissipation ( see also in @xcite ) . on the other side , @xcite attempts to interpret it as a result of the combination of stochastic and smooth planet migrations . beside and before the _ kepler _ transit survey ,
many near mmr planets had been found by various radial velocity ( rv hereafter ) surveys .
as we will show below ( section [ dis_rv ] ) , similar , if not exactly the same , features of the period ratio distributions seen in _ kepler _ planets , have been also shown in rv planets .
one question is how all these features / clues in both the _ kepler _ and rv samples could be understood systematically in a common context .
this paper is such an attempt and it is organized as the following .
we first analytically study the dynamics of planets near / in mmr in section [ sec_analytic ] , and confirm the analytical results with numerical simulations in section [ sec_numerical ] .
we find that planets orbital distribution could be asymmetric around the mmr center under certain conditions .
we then discuss its implications to _ kepler _ and rv planets in section [ dis ] .
finally , we summarize this paper in section [ sum ] . some analytical derivations are also given in the appendix a and b as supplementary .
we note that @xcite posted their paper to arxiv.org just a few days before submitting this paper , which , independently and in a different way , arrived at many of the results presented in this paper .
we study the orbital evolutions of two planets ( orbiting a central star ) near / in first order mmr . as we will show below
, the orbit distribution could be asymmetric near the mmr center under certain circumstances . for simplicity , we assume both planets orbits are coplanar .
the total energy , or hamiltonian , is @xcite @xmath3 where @xmath4 is the gravity constant , @xmath5 is the stellar mass , and following @xcite , the disturbing function due to the @xmath6 resonance is @xmath7 where @xmath8 are the two resonance angles for @xmath9 hereafter , we adopt the convention that properties with subscripts `` 1 '' and `` 2 '' belong to the inner and outer planets respectively . in the above , \{@xmath10 , @xmath11 , @xmath12 , @xmath13 , @xmath14 }
are the mass and standard orbital elements for planets . @xmath15 and @xmath16 are relevant laplace coefficients , which are on order of unity and tabulated in @xcite and @xcite . using the lagrange s planetary equation ( on the lowest order terms in @xmath12 )
, we derive the evolutions of planets semi major axes and eccentricities , @xmath17 @xmath18 where , @xmath19 and @xmath20 are the mean motion of the inner and outer planets respectively . using equation [ edot ] to eliminate @xmath21 and @xmath22 , we can rewrite equation [ adot0 ] as @xmath23 which integrate to give @xmath24 where we have defined @xmath25 we note equations [ adot1 ] or [ const1 ] are equivalent to the well known constants of motion in resonance ( see appendix a ) .
a worth noting implication of equation [ adot1 ] or [ const1 ] is that if planet pairs initially formed with circular orbit near mmr , they will shift to a little bit larger orbital period ratio as their eccentricities are excited , inducing an asymmetric orbit distribution near mmr ( see numerical confirmation in section [ sec_numerical ] ) .
using equation [ adot1 ] , this small shift extent in period ratio ( @xmath26 ) can be estimated as @xmath27 \nonumber\end{aligned}\ ] ] according to @xcite ( see their eqn .
8.209 and 8.210 ) , the maximum eccentricity increase in @xmath28 and @xmath29 ( or critical eccentricities ) are @xmath30 setting @xmath31 and @xmath32 , then equation [ dp1 ] will give an estimate of the largest asymmetric shift of period ratio .
dissipation processes ( e.g. , tidal evolution , disk migration ) may play an import role during planet formation and evolution .
generally they cause changing on planets orbital semi major axes and damping in eccentricities . to include these effects , we consider the following changing / damping terms ( i.e. , inverse of the damping timescales ) of semi major axes and eccentricities , @xmath33 where ( hereafter ) @xmath34 for the inner and outer planets respectively .
note , @xmath35 could be negative , which indicates outward migration , and @xmath36 is generally positive , i.e. , eccentricity is damped in dissipation process .
following @xcite ( see the appendix b for the derivation ) , the evolutions of the semi major axes of two planets ( after adding above damping terms ) are .
@xmath37 where @xmath38 is the proximity to the nominal resonance center , and thus its evolution follows , @xmath39 if @xmath40 , then @xmath41 will be always positive , namely the two planets will always keep divergent migration , i.e. , their period ratio will always increase .
this is the case if the planetary system undergoes tidal evolution @xcite .
if @xmath42 otherwise , then there is an stable equilibrium with @xmath43^{1/2 } , \label{deq } \end{aligned}\ ] ] wider / narrower than which , the two planets will undergo convergent / divergent migration , thus eventually they will be locked at @xmath44 .
interestingly , the above equation can be roughly written as @xmath45 where @xmath46 is the typical planet - star mass ratio of the system and @xmath47 is the well known model parameter describing the ratio between the damping rate of orbital eccentricity and that of semimajor axis . from equation ( [ deq2 ] )
, we see that the theoretical parameter @xmath0 is linked to an observable @xmath48 .
we will discuss this more in section [ dis_rv ] .
for comparison against the above analytical results , we perform some 3-body ( 1 star + 2 planets ) simulations using the well - tested n - body integrator mercury @xcite .
for all the simulations , the central star is set with a mass @xmath49 , and all angular orbital elements , except for orbital inclinations , are initially randomly set . for most simulations , the semi major axis of the inner planet
is set at 0.1 au if not specified . from equations
[ adot1]-[dp1 ] , we expect that planets orbits have an asymmetric distribution near the mmr center . here
, we numerically show such an asymmetry and its dependence on the initial period ratio , orbital eccentricities , inclinations and planetary masses .
figure [ fig_asym ] shows the orbital evolutions of two equal mass ( @xmath50 ) planets initially with circular and coplanar orbits but different orbital ratios .
planets semimajor axes and eccentricities follow periodical oscillations , and their period ratios increase with eccentricities as expected from equation [ adot1 ] . on average the planets spend more time on orbits wider than the initial ones , causing an asymmetric distribution in their period ratio .
the asymmetry become weaker as the planet pair is further away from mmr .
however , the most prominent asymmetric feature is not realized at the mmr center but at a little bit narrower than the center .
the reason is that planets eccentricities get most excited when they are at the separatrix which is at narrower than the nominal resonance center for the first order mmr @xcite .
figure [ fig_ecc ] shows how the asymmetry is affected by the initial orbital eccentricities . as expected from equation [ adot1 ] , if the eccentricity is initially larger , then it will have larger possibility ( compared to the case of zero initial eccentricity ) to decrease in the future , thus the period ratio will become more symmetric around the initial one .
the critical eccentricity , greater than which the asymmetry will be very weak , could be estimated using equation [ ecr ] , which is consistent with the numerical results and the results within the context of the restricted 3-body problem @xcite .
figure [ fig_mas ] shows the effect of planetary mass on the asymmetry . as expected from equation [ adot1]-[ecr ] , increasing mass leading to larger eccentricity excitation and thus larger period ratio shift extent .
roughly , systems with similar total masses ( regardless of mass ratio ) have similar shift extents . figure [ fig_inc ] shows the role of relative inclination @xmath51 in the asymmetry .
generally the asymmetry becomes very weak for @xmath52 .
this is not surprised , as the above analytical studies are all based on an assumption of low @xmath51 . for large @xmath51 , more terms ( e.g. , on the oder of @xmath53 )
should be considered in the disturbing function in equation [ rj ] , and in such cases , planets could be involved in second order of mmr , which is symmetric around nominal resonance center @xcite .
the case of divergent migration ( i.e. , @xmath54 ) has been studied recently in detail recently by @xcite . here we focus on the other case where @xmath55 . for simplicity , we assume @xmath56 , and damping is only added on the outer planet with @xmath57 and @xmath58 .
the two planets are started at 0.2 and 0.35 au respectively with an initial orbital period @xmath59 .
we study 7 different k values from 0 to 10000 and 3 different planetary mass sets .
the results are plotted in figure [ fig_mig ] .
the left 4 panels of figure [ fig_mig ] plot the results of one simulation with @xmath60 and @xmath61 .
the out planet moves inward and captures into 2:1 mmr with the inner planet at about @xmath62 d. after that , the two planets still moving inward together but with resonance angles , eccentricities and period ratios reaching a relatively stable state .
the orbital period ratio at the later state is asymmetric around the nominal mmr center , and the its mean value is roughly consistent with the analytical estimate from equation [ deq ] .
the right panel of figure [ fig_mig ] shows how @xmath48 depends on planetary mass and damping ratio @xmath0 .
generally we see that @xmath48 is proportional to planetary mass and increases with @xmath0 .
not surprised , the analytical predictions are consistent with the numerical simulations only for relative large @xmath48 and @xmath0 ( see appendix b ) . for low @xmath0 values ( @xmath63 )
, @xmath48 do not approach zero but become a positive constant which is proportional to planetary mass .
such a tiny constant @xmath48 may reflect the intrinsic asymmetry of the mmr .
however , we note that here the constant @xmath48 is much smaller than the maximum asymmetry estimated by equations [ dp1 ] and [ ecr ] , which is reasonable because large eccentricity leads to weak asymmetry as seen in figure [ fig_ecc ] .
the period ratio distribution of _ kepler _ multiple planet candidate systems show an intriguing asymmetric feature near mmr , especially for 2:1 and 3:2 mmr , namely there are small deficits / excesses just a little bit narrow / wide of the nominal mmr center @xcite . to interpret such an asymmetric feature ,
@xcite consider that it could be a result of planets undergoing some dissipative evolution , such as tidal dissipation .
in such a case , as discussed in section [ dissipation1 ] , @xmath64 , thus the planet period will always increase . to quantitively explain the observed asymmetric period ratio distribution
, one needs to put a right amount of dissipation on them .
in addition , as tidal effect is only efficient for short period planet , e.g. , less 10 days , one needs to resort to other dissipations at larger orbital period where the observed asymmetry is still significant . @xcite
then considers if the observed period ratio is consistent with the scenario of planets migrating in disks .
first , he considers smooth migration and finds that the excess or pileup of planet pairs is too large and too close to the mmr center .
his result is expected from our analytical results in figure [ fig_mig ] and equation [ deq ] , which shows @xmath65 ( 2 order of magnitude lower than the observed one ) if assuming a typical _ kepler _ planet mass on order of 10 @xmath66 and @xmath67 .
nevertheless , he further shows that by including certain amount stochastic forces due to disk turbulence during migration , the large pileup at mmr center could be smeared out and a period ratio distribution similar to that of _ kepler _ planets could be reproduced .
all the above attempts belong to the case with dissipation . as we have shown
( section [ nod1 ] and [ nod2 ] ) , the period ratio distribution is intrinsically asymmetric near the mmr center even if there is no dissipation . in order to see whether and how the intrinsic asymmetry can reproduce _ kepler _ planets period ratio distribution , we perform the following n - body simulations . specifically , we draw 4000 planets pairs initially with a uniform period ratio distribution near mmr , rayleigh eccentricity and inclination distributions , and uniformly random distribution for all the other angular orbital elements .
we use the mercury integrator to simulate these 4000 systems individually on a timescale of @xmath68 days and intensively output their period ratio very 200 days .
the final period ratio distribution is calculated with these output period ratios of all 4000 systems . as _ kepler _
multiple planet systems are believed to be highly coplanar within a few degree @xcite , we assume the mean inclination @xmath69 . for simplicity , we only study equal mass pairs , i.e , @xmath70 because different mass ratios lead to similar results as long as their total masses are the same ( fig.[fig_mas ] ) .
figure [ fig_obs ] compares the observed period ratio distribution to those from above simulations with different planetary masses from @xmath50 to @xmath71 and mean eccentricities from @xmath72 to @xmath73 .
the simulated period ratio distributions have an asymmetric feature resembling the observation , i.e. , a trough / pile up just a little bit narrow / wide of mmr center . as expected ( fig.[fig_ecc ] and [ fig_mas ] ) , the asymmetric feature become weaker with increasing eccentricity and more extended with increasing mass . in order to reproduce the observed period ratio distribution
, it requires a mean eccentricity less than a few percents and planetary mass about 10 - 20 @xmath66 for 3:2 mmr and @xmath74 for 2:1 mmr .
the eccentricity requirement is consistent with recent eccentricity estimate with transit timing variation @xcite . as for the typical mass of _ kepler _ planets ,
it is expected to be 4 - 9 @xmath66 given the typical radii of 2 - 3 @xmath75 and a mass radio distribution either based on fitting of the solar system , @xmath76 @xcite , or transit timing variation , @xmath77 @xcite . even considering a relatively large uncertain of mass measurements , say 100% , such an expected mass is still too low to meet the requirement for 2:1 mmr , although it is comparable to the mass requirement for 3:2 mmr . therefore , we conclude that the intrinsic mmr asymmetry ( without any damping ) could partially explain _ kepler _ planets asymmetric period ratio distribution near 3:2 mmr but not 2:1 mmr . for the latter , other mechanisms , e.g. , dissipation , should play a more important role . at the time of writing this paper ,
there are 409 exoplanets detected with radial velocity ( rv ) method ( exoplanet.org ) and about @xmath78 of them reside in multiple planet systems .
these rv planets have a wide mass range featured with a bimodal distribution @xcite as shown in the left panel of figure[fig_rvobs ] .
the boundary is at about 0.2 @xmath79 , which separate the light rv planets ( with a media mass of @xmath80 ) and the massive ones ( with a media mass of @xmath81 ) .
this bimodal distribution may indicate planets undergo different formations and evolutions for the light and massive groups @xcite .
interestingly , we find that these two groups may have different period ratio distributions . as shown in the right panels of figure [ fig_rvobs ] , there is a strong pileup of planet pairs near 2:1 mmr in the massive planet group , which is not seen in the light group .
those massive planets piled up near 2:1 mmr seems unlikely formed in situ within a small annulus , but they are more likely formed with larger distance in a disk then brought into 2:1 mmr through convergent migration .
interestingly , we note that the pile up is just a few percent ( in period ratio ) wide of the 2:1 mmr center , which is expected from our analytical and numerical prediction with planetary migration ( e.g. , fig.[fig_mig ] ) . furthermore , from the location of the pileup ( i.e. , @xmath48 )
, we can infer the damping ratio between eccentricity and semi major axis during planetary migration ( i.e. , @xmath0 ) by using equation [ deq ] .
the result of such an exercise is shown in figure [ fig_k12 ] .
here we considered two migration scenarios . in scenario 1 , only the outer planet undergoes migration , i.e. , @xmath58 and @xmath82 . in scenario 2 ,
the inner one migrates outward and the outer one migrates inward , i.e. , @xmath58 , @xmath83 and @xmath84 . as can be seen from figure [ fig_k 12 ] , the @xmath0 value is constrained in a relative wide range about 1 - 100 on order of magnitude .
we note this @xmath0 range is consistent with the hydrodynamical simulations by @xcite which predicts a @xmath0 value of order of unity , and with dynamical modeling of the well - studied system gj876 by @xcite which prefers @xmath85 .
in this paper , we analytically and numerically study the dynamics of planet pairs near first order mmr . focusing on the evolution of orbital period ratio , we find it could have an asymmetric distribution around the nominal mmr center regardless of whether dissipation is included or not . applying the asymmetric nature of mmr to the _ kepler _ planets , we find that , without dissipation , _ kepler _ planets asymmetric period ratio distribution could be _ partly _ explained for the case of 3:2 mmr but _ not _ for 2:1 mmr , suggesting that dissipation or other mechanisms may play a more important role in 2:1 than in 3:2 mmr . beside the _ kepler _ planets , similar asymmetric feature , i.e. , planets piled up wide of mmr , is also seen in rv planets .
nevertheless , planets in multiple rv systems are bimodal distribution on mass , and the pileup is currently only seen in the higher mass group .
the location of the pileup is consistent with the scenario that planetary migration toward mmr , and it infers that the ratio of damping rate between eccentricity and semimajor axis ( i.e. , k value ) during planet migration is @xmath86 on order of magnitude for massive planets .
jwx thanks the referee for helpful comments and suggestions , yanqin wu and hanno rei for valuable discussions and the _ kepler _ team for producing such an invaluable data set .
jwx was supported by the national natural science foundation of china ( nos.10833001 and 10925313 ) , phd training grant of china ( 20090091110002 ) , fundamental research funds for the central universities ( 1112020102 ) and the ontario government .
99 batalha , n. m. , rowe , j. f. , bryson , s. t. , et al .
2012 , arxiv:1202.5852 batygin , k. , & morbidelli , a. 2012 , arxiv:1204.2791 borucki , w. j. , koch , d. g. , basri , g. , et al .
2011 , , 736 , 19 brouwer , d. 1963 , aj , 68 , 152 chambers , j. e. , & migliorini , f. 1997 , bulletin of the american astronomical society , 29 , 1024 chiang , e. , & laughlin , g. 2012 , arxiv:1211.1673 delisle , j .- b . , laskar , j. , correia , a. c. m. , & bou , g. 2012 , arxiv:1207.3171 fabrycky , d. c. , lissauer , j. j. , ragozzine , d. , et al .
2012 , arxiv:1202.6328 fabrycky , d. c. , ford , e. b. , steffen , j. h. , et al .
2012 , , 750 , 114 fang , j. , & margot , j .-
2012 , arxiv:1207.5250 figueira , p. , marmier , m. , bou , g. , et al .
2012 , a&a , 541 , a139 hansen , b. m. s. , & murray , n. 2012 , , 751 , 158 kley , w. , peitz , j. , & bryden , g. 2004 , , 414 , 735 kley , w. , & nelson , r. p. 2012
, , 50 , 211 latham , d. w. , rowe , j. f. , quinn , s. n. , et al .
2011 , , 732 , l24 lee , m. h. , & peale , s. j. 2002 , , 567 , 596 lissauer , j. j. , ragozzine , d. , fabrycky , d. c. , et al .
2011 , , 197 , 8 lithwick , y. , & wu , y. 2012 , , 756 , l11 lithwick , y. , xie , j. , & wu , y. 2012 , arxiv:1207.4192 mordasini , c. , alibert , y. , benz , w. , & naef , d. 2009 , , 501 , 1161 moons , m. 1996 , celestial mechanics and dynamical astronomy , 65 , 175 murray , c. d. , & dermott , s. f. 1999 , solar system dynamics by murray , c. d. , 1999 , petrovich , c. , malhotra , r. , & tremaine , s. 2012 , arxiv:1211.5603 pepe , f. , mayor , m. , lovis , c. , et al .
2011 , iau symposium , 276 , 13 rein , h. 2012 , arxiv:1208.3583 scholl , h. 1979 , dynamics of the solar system , 81 , 217 schweizer , f. 1969 , aj , 74 , 779 terquem , c. , & papaloizou , j. c. b. 2007 , , 654 , 1110 veras , d. , & ford , e. b. 2012 , , 420 , l23 wu , y. , & lithwick , y. 2012 , arxiv:1210.7810 xie , j .- w .
2012a , arxiv:1208.3312 xie , j .- w . 2012b , submitted to iau 293 proceeding
here we show that the equations [ adot1 ] and/or [ const1 ] are equivalent to the well known constants of motion of mmr . for j :
mmr there are two constants of motion in addition to the energy ( see chapter 8.8 of @xcite ) , i.e. , @xmath87 where , @xmath88 and @xmath89 are the poincar momenta ( see chapter 2.10 of @xcite ) , and the subscript `` 1 '' and `` 2 '' denotes the inner and outer planets respectively . changing the above equation to basic orbital elements , we have @xmath90 & = & { \rm const . } \nonumber \\ m_{2}\sqrt{a_{2 } } -j\left[m_{1}\sqrt{a_{1}}\left(1-\sqrt{1-e_{1}^{2}}\right)+m_{2}\sqrt{a_{2}}\left(1-\sqrt{1-e_{2}^{2}}\right)\right ] & = & { \rm const.}\end{aligned}\ ] ] in the leading term of @xmath12 , we then have @xmath91 take the differential form of above equations and keep the leading term in @xmath12 , we then have @xmath92 using the approximation , @xmath93^{2/3}$ ] , above equations can be rewritten as equation [ adot1 ] .
compared to the original formulas of the constants , the new formulas ( eqn .
[ adot1 ] ) solve @xmath94 and @xmath95 out ( they are not coupled together as in eqn.[const2 ] ) and they are dimensionless and simpler .
following @xcite , it is convenient to introduce the compact eccentricity @xmath96 where @xmath14 is the longitude of the periastron and @xmath34 for the inner and outer planets respectively . in terms of which
, the disturbing function can be expressed as @xmath97 where the superscript `` * '' denotes the complex conjugate of the variable and `` @xmath98 '' denotes the complex conjugate of the proceeding term .
then the eccentricity equation ( after adding the damping term ) for the two planets is @xmath99 or specifically , @xmath100 adopting the following approximation @xmath101 then we can solve the eccentricities as @xmath102 where @xmath103 and @xmath104 are free solutions ( free eccentricities ) . in terms of the compact eccentricity , the evolution of semi major axes ( eqn.[adot0 ] ) can be rewritten ( after adding damping terms ) as @xmath105 -\gamma_{a1 } , \,\ , \ , \ , \ , \ , \ , \frac{\dot{a_{2}}}{a_{2 } } = \frac{jgm_{1}}{n_{2}a_{2}^{3 } } \left[(f_{1}z_{1}^{*}+f_{2}z_{2}^{*})ie^{i\lambda^{j } } + c.c .
\right ] -\gamma_{a1 } , \label{adotz}\end{aligned}\ ] ] which can be finally written as ( with the help of eqn.[ef ] ) , @xmath106 where @xmath107 and @xmath108 are the terms caused by the free eccentricities , i.e. , @xmath109 for @xmath110 defined as the free eccentricities of the system . .
in the case where it is not too close to mmr ( modest @xmath111 ) and eccentricity damping is efficient ( large @xmath112 ) , @xmath113 and thus the two oscillation terms @xmath107 and @xmath108 can be ignored ( i.e. , eqn.[da4 ] and [ deq ] )
. otherwise , if it is very close to mmr ( very small @xmath111 ) and the eccentricity damping is weak , then the system could get significant free eccentricities ( probably by approaching the separatrix ) , thus @xmath107 and @xmath108 can not be ignored and the equilibrium can not be well estimated by using equation [ deq ] ( see also in fig.[fig_mig ] ) . | many multiple planet systems have been found by the _ kepler _ transit survey and various radial velocity ( rv ) surveys .
_ kepler _ planets show an asymmetric feature , namely there are small but significant deficits / excesses of planet pairs with orbital period spacing slightly narrow / wide of the exact resonance , particularly near the first order mean motion resonance ( mmr ) , such as 2:1 and 3:2 mmr .
similarly , if not exactly the same , an asymmetric feature ( pileup wide of 2:1 mmr ) is also seen in rv planets , but only for massive ones . we analytically and numerically study planets orbital evolutions near / in mmr .
we find that their orbital period ratios could be asymmetrically distributed around the mmr center regardless of dissipation . in the case of no dissipation , _ kepler _ planets asymmetric orbital distribution could be partly reproduced for 3:2 mmr but not for 2:1 mmr , implying dissipation might be more important to the latter .
the pileup of massive rv planets just wide of 2:1 mmr is found to be consistent with the scenario that planets formed separately then migrated toward mmr .
the location of the pileup infers a @xmath0 value of 1 - 100 on order of magnitude for massive planets , where @xmath0 is the damping rate ratio between orbital eccentricity and semimajor axis during planet migration . |
max - convolution occurs frequently in signal processing and bayesian inference : it is used in image analysis @xcite , in network calculus @xcite , in economic equilibrium analysis @xcite , and in a probabilistic variant of combinatoric generating functions , wherein information on a sum of values into their most probable constituent parts ( _ e.g. _ identifying proteins from mass spectrometry @xcite ) .
max - convolution operates on the semi - ring @xmath7 , meaning that it behaves identically to a standard convolution , except it employs a @xmath8 operation in lieu of the @xmath9 operation in standard convolution ( max - convolution is also equivalent to min - convolution , also called infimal convolution , which operates on the tropical semi - ring @xmath10 ) . due to the importance and ubiquity of max - convolution ,
substantial effort has been invested into highly optimized implementations ( _ e.g. _ , implementations of the quadratic method on gpus ; @xcite ) .
max - convolution can be defined using vectors ( or discrete random variables , whose probability mass functions are analogous to nonnegative vectors ) with the relationship @xmath11 .
given the target sum @xmath12 , the max - convolution finds the largest values @xmath13 $ ] and @xmath14 $ ] for which @xmath15 .
@xmath16 & = & \max_{\ell , r : \,m = \ell+r } l[\ell ] r[r ] \\ & = & \max_\ell l[\ell ] r[{m-\ell}]\\ & = & \left ( l ~*_{\max}~ r \right)[m ] \\\end{aligned}\ ] ] where @xmath17 denotes the max - convolution operator . in probabilistic terms
, this is equivalent to finding the highest probability of the joint events @xmath18 that would produce each possible value of the sum @xmath11 ( note that in the probabilistic version , the vector @xmath19 would subsequently need to be normalized so that its sum is 1 ) .
although applications of max - convolution are numerous , only a small number of methods exist for solving it @xcite .
these methods fall into two main categories , each with their own drawbacks : the first category consists of very accurate methods that are have worst - case runtimes either quadratic @xcite or slightly more efficient than quadratic in the worst - case @xcite .
conversely , the second type of method computes a numerical approximation to the desired result , but in @xmath20 steps ; however , no bound for the numerical accuracy of this method has been derived @xcite .
while the two approaches from the first category of methods for solving max - convolution do so by either using complicated sorting routines or by creating a bijection to an optimization problem , the numerical approach solves max - convolution by showing an equivalence between @xmath17 and the process of first generating a vector @xmath21 for each index @xmath22 of the result ( where @xmath23 = l[\ell ] r[{m-\ell}]$ ] for all in - bounds indices ) and subsequently computing the maximum @xmath24 = \max_\ell u^{(m)}[\ell]$ ] .
when @xmath25 and @xmath26 are nonnegative , the maximization over the vector @xmath21 can be computed exactly via the chebyshev norm @xmath16 & = & \max_\ell u^{(m)}[\ell ] \\ & = & \lim_{p \to \infty } \| u^{(m ) } \|_p\\\end{aligned}\ ] ] but requires @xmath27 steps ( where @xmath28 is the length of vectors @xmath25 and @xmath26 ) .
however , once a fixed @xmath29-norm is chosen , the approximation corresponding to that @xmath29 can be computed by expanding the @xmath29-norm to yield @xmath30 \right)}^{p } \right)}^{\frac{1}{p } } \\ & \approx & { \left ( \sum_\ell { \left ( u^{(m)}[\ell ] \right)}^{p^ * } \right)}^{\frac{1}{p^*}}\\ & = & { \left ( \sum_\ell { l[\ell]}^{p^ * } ~ { r[{m-\ell}]}^{p^ * } \right)}^{\frac{1}{p^*}}\\ & = & { \left ( \sum_\ell { \left(l^{p^*}\right)}[\ell ] ~ { \left(r^{p^*}\right)}[{m-\ell } ] \right)}^{\frac{1}{p^*}}\\ & = & { \left ( l^{p^ * } ~*~ r^{p^ * } \right)}^{\frac{1}{p^*}}[m]\end{aligned}\ ] ] where @xmath31 \right)}^{p^ * } , { \left ( l[1 ] \right)}^{p^ * } , ~\ldots,~{\left ( l[{k-1 } ] \right)}^{p^ * } ~\rangle$ ] and @xmath32 denotes standard convolution .
the standard convolution can be done via fast fourier transform ( fft ) in @xmath33 steps , which is substantially more efficient than the @xmath27 required by the naive method ( * algorithm [ algorithm : numericalmaxconvolvegivenpstar ] * ) . to date
, the numerical method has currently demonstrated the best speed - accuracy trade - off on bayesian inference tasks , and can be generalized to multiple dimensions ( _ i.e. _ , tensors ) .
in particular , they have been used with probabilistic convolution trees @xcite to efficiently compute the most probable values of discrete random variables @xmath34 for which the sum is known @xmath35 @xcite . the one - dimensional variant of this problem ( _ i.e. _ , where each @xmath36 is a one - dimensional vector ) solves the probabilistic generalization of the subset sum problem , while the two - dimensional variant ( _ i.e. _ , where each @xmath36 is a one - dimensional matrix ) solves the generalization of the knapsack problem ( note that these problems are not np - hard in this specific case , because we assume an evenly - spaced discretization of the possible values of the random variables ) .
however , despite the practical performance that has been demonstrated by the numerical method , only cursory analysis has been performed to formalize the influence of the value of @xmath29 on the accuracy of the result and to bound the error of the @xmath29-norm approximation .
optimizing the choice of @xmath29 is non - trivial : larger values of @xmath29 more closely resemble a true maximization under the @xmath29-norm , but result in underflow ( note that in * algorithm [ algorithm : numericalmaxconvolvegivenpstar ] * , the maximum values of both @xmath25 and @xmath26 can be divided out and then multiplied back in after max - convolution so that overflow is not an issue ) .
conversely , smaller values of @xmath29 suffer less underflow , but compute a norm with less resemblance to maximization . here
we perform an in - depth analysis of the influence of @xmath29 on the accuracy of numerical max - convolution , and from that analysis we construct a modified piecewise algorithm , on which we demonstrate bounds on the worst - case absolute error .
this modified algorithm , which runs in @xmath37 steps , is demonstrated using a hidden markov model describing the relationship between u.s .
unemployment and the s&p 500 stock index .
we then extend the modified algorithm and introduce a second modified algorithm , which not only uses a single @xmath1-norm as a means of approximating the chebyshev norm , but instead uses a sequence of @xmath1-norms and assembles them using a projection as a means to approximate the chebyshev norm . using numerical simulations as evidence , we make a conjecture regarding the relative error of the null space projection method . in practice , this null space projection algorithm is shown to have similar runtime and higher accuracy when compared with the piecewise algorithm .
we begin by outlining and comparing three numerical methods for max - convolution . by analyzing the benefits and deficits of each of these methods ,
we create improved variants .
all of these methods will make use of the basic numerical max - convolution idea summarized in the introduction , and as such we first declare a method for computing the numerical max - convolution estimate for a given @xmath29 as numericalmaxconvolvegivenpstar ( * algorithm [ algorithm : numericalmaxconvolvegivenpstar ] * ) .
@xmath38 \gets { l[\ell ] } ^{p^*}$ ] @xmath39 \gets { r[r ] } ^{p^*}$ ] @xmath40 @xmath41 \gets { vm[m ] } ^{\frac{1}{p^*}}$ ] @xmath42 the effects of underflow will be minimal ( as it is not very far from standard fft convolution , an operation with high numerical stability ) , but it can still be imprecise due to numerical `` bleed - in '' ( _ i.e. _ error due to contributions from non - maximal terms for a given @xmath21 because the @xmath29-norm is not identical to the chebyshev norm ) .
overall , this will perform well on indices where the exact value of the result is small , but perform poorly when the exact value of the result is large . as noted above , will offer the converse pros and cons compared to using a low @xmath29 : numerical artifacts due to bleed - in will be smaller ( thus achieving greater performance on indices where the exact values of the result are larger ) , but underflow may be significant ( and therefore , indices where the exact results of the max - convolution are small will be inaccurate ) . the higher - order piecewise method formalizes the empirical cutoff values found in serang 2015 ; previously , numerical stability boundaries were found for each @xmath29 by computing both the exact max - convolution ( via the naive @xmath27 method ) and via the numerical method using the ascribed value of @xmath29 , and finding the value below which the numerical values experienced a high increase in relative absolute error .
those previously observed empirical numerical stability boundaries can be formalized by using the fact that the employed numpy implementation of fft convolution has high accuracy on indices where the result has a value @xmath45 relative to the maximum value ; therefore , if the arguments @xmath25 and @xmath26 are both normalized so that each has a maximum value of 1 , the fast max - convolution approximation is numerically stable for any index @xmath22 where the result of the fft convolution , _
i.e. _ @xmath46 $ ] , is @xmath45 .
the numpy documentation defines a conservative numeric tolerance for underflow @xmath47 , which is a conservative estimate of the numerical stability boundary demonstrated in * figure [ figure : baderrorsforallpstar ] * ( those boundary points occur very close to the true machine precision @xmath48 ) . because cooley - tukey implementations of fft - based convolution ( _ e.g. _ , the numpy implementation ) are widely applied to large problems with extremely small error , we will make a simplification and assume that , when constraining the fft result to reach a value higher than machine epsilon ( + tolerance threshold ) , the error from the fft is negligible in comparison to the error introduced by the @xmath29-norm approximation .
this is firstly because the only source of numerical error during fft ( assuming an fft implementation with numerically precise twiddle factors ) on vectors in @xmath49}^k$ ] will be the result of underflow from repeated addition and subtraction ( neglecting the non - influencing multiplication with twiddle factors , which each have magnitude @xmath50 ) .
the numerically imprecise routines are thus limited to @xmath51 ; when @xmath52 ( _ i.e. _ , @xmath53 , the machine precision ) , then @xmath51 will return @xmath54 instead of @xmath55 . to recover at least one bit of the significand , the intermediate results of the fft must surpass machine precision @xmath56 ( since the worst case addition initially happens with the maximum @xmath57 ) .
the maximum sum of any values from a list of @xmath28 such elements can never exceed @xmath28 ; for this reason , a conservative estimate of the numerical tolerance of an fft ( with regard to underflow ) will be the smallest value of @xmath55 for which @xmath58 ; thus , @xmath59 .
this yields a conservative estimate of the minimum value in one index at the result of an fft convolution : when the result at some index @xmath22 is @xmath60 , then the result should be numerically stable . for this reason
, we use a numerical tolerance @xmath61 , thereby ensuring that the vast majority of numerical error for the numerical max - convolution algorithm is due to the @xmath29-norm approximation ( _ i.e. _ , employing @xmath62 instead of @xmath63 ) and not due to the long - used and numerically performant fft result .
furthermore , in practice the mean squared error due to fft will be much smaller than the conservative worst - case outlined here , because it is difficult for the largest intermediate summed value ( in this case @xmath64 ) to be consistently large when many such very small values ( in this case @xmath55 ) are encountered in the same list .
although @xmath65 could be chosen specifically for a problem of size @xmath28 , note that this simple derivation is very conservative and thus it would be better to use a tighter bound for choosing @xmath65 .
regardless , for an fft implementation that is nt as performant ( _ e.g. _ , because it uses float types instead of double ) , increasing @xmath65 slightly would suffice . therefore , from this point forward we consider that the dominant cause of error to come from the max - convolution approximation .
using larger @xmath29 values will provide a closer approximation ; however , using a larger value of @xmath29 may also drive values to zero ( because the inputs @xmath25 and @xmath26 will be normalized within * algorithm [ algorithm : numericalmaxconvolvegivenpstar ] * so that the maximum of each is 1 when convolved via fft ) , limiting the applicability of large @xmath29 to indices @xmath22 for which @xmath46 \geq \tau$ ] . through this lens ,
the choice of @xmath29 can be characterized by two opposing sources of error : higher @xmath29 values better approximate @xmath66 but will be numerically unstable for many indices ; lower @xmath29 values provide worse approximations of @xmath62 but will be numerically unstable for only few indices .
these opposing sources of error pose a natural method for improving the accuracy of this max - convolution approximation . by considering a small collection of @xmath29 values
, we can compute the full numerical estimate ( at all indices ) with each @xmath29 using * algorithm [ algorithm : numericalmaxconvolvegivenpstar ] * ; computing the full result at a given @xmath29 is @xmath67 , so doing so on some small number @xmath68 of @xmath29 values considered , then the overall runtime will be @xmath69 . then
, a final estimate is computed at each index by using the largest @xmath29 that is stable ( with respect to underflow ) at that index .
choosing the largest @xmath29 ( of those that are stable with respect to underflow ) corresponds to minimizing the bleed - in error , because the larger @xmath29 becomes , the more muted the non - maximal terms in the norm become ( and thus the closer the @xmath29-norm becomes to the true maximum ) . here
we introduce this piecewise method and compare it to the simpler low - value @xmath43 and high - value @xmath44 methods and analyze the worst - case error of the piecewise method .
@xmath70 $ ] @xmath71 $ ] @xmath72}$ ] @xmath73}$ ] @xmath74{\log_2(p^*_{\max } ) } } ] $ ] @xmath75 \gets$ ] ` fftnonnegmaxconvolvegivenpstar`(@xmath76 , @xmath77 , @xmath78 $ ] ) @xmath79 \gets \max \ { i:~ { \left ( resforallpstar[i][m ] \right)}^{\text{allpstar[$i$ ] } } \geq \tau ) \}$ ] @xmath80 $ ] @xmath81 \gets resforallpstar[i][m]$ ] @xmath82 \times r[r_{\max } ] \times result$ ]
this section derives theoretical error bounds as well as a practical comparison on an example for the standard piecewise method .
furthermore the development of an improvement with affine scaling is shown .
eventually , an evaluation of the latter is performed on a larger problem .
therefore we applied our technique to compute the viterbi path for a hidden markov model ( hmm ) to assess runtime and the level of error propagation .
we first analyze the error for a particular underflow - stable @xmath29 and then use that to generalize to the piecewise method , which seeks to use the highest underflow - stable @xmath29 .
we first scale @xmath25 and @xmath26 into @xmath76 and @xmath77 respectively , where the maximum elements of both @xmath76 and @xmath77 are @xmath50 ; the absolute error can be found by unscaling the absolute error of the scaled problem : @xmath83 - numeric(l',r')[m ] |\\ = \max_\ell l[\ell ] ~ \max_r r[r ] \ ; | exact(l',r')[m ] - numeric(l',r')[m ] |.\end{gathered}\ ] ] we first derive an error bound for the scaled problem on @xmath84 ( any mention of a vector @xmath21 refers to the scaled problem ) , and then reverse the scaling to demonstrate the error bound on the original problem on @xmath85 .
for any particular `` underflow - stable '' @xmath29 ( _ i.e. _ , any value of @xmath29 for which @xmath86 ) , the absolute error for the numerical method for fast max - convolution can be bound fairly easily by factoring out the maximum element of @xmath21 ( this maximum element is equivalent to the chebyshev norm ) from the @xmath29-norm : @xmath87 - numeric(l',r')[m ] |\ ] ] @xmath88 where @xmath89 is a nonnegative vector of the same length as @xmath21 ( this length is denoted @xmath90 ) where @xmath89 contains one element equal to @xmath50 ( because the maximum element of @xmath21 must , by definition , be contained within @xmath21 ) and where no element of @xmath89 is greater than 1 ( also provided by the definition of the maximum ) .
@xmath91 thus , since @xmath92 , the error is bound : @xmath93 - numeric(l',r')[m]\\ & = & \| u^{(m ) } \|_\infty \left(\| v^{(m ) } \|_{p^ * } - 1 \right)\\ & \leq & \| v^{(m ) } \|_{p^ * } - 1\\ & \leq & k_m^\frac{1}{p^ * } - 1,\\\end{aligned}\ ] ] because @xmath94 for a scaled problem on @xmath84 . however , the bounds derived above are only applicable for @xmath29 where @xmath95 .
the piecewise method is slightly more complicated , and can be partitioned into two cases : in the first case , the top contour is used ( _ i.e. _ , when @xmath96 is underflow - stable ) .
conversely , in the second case , a middle contour is used ( _ i.e. _ , when @xmath96 is not underflow - stable ) . in this context ,
in general a contour comprises of a set of indices @xmath22 with the same maximum stable @xmath29 . in the first case ,
when we use the top contour @xmath97 , we know that @xmath96 must be underflow - stable , and thus we can reuse the bound given an underflow - stable @xmath29 . in the second case , because the @xmath29 used is @xmath98 , it follows that the next higher contour ( using @xmath99 ) must not be underflow - stable ( because the highest underflow - stable @xmath29 is used and because the @xmath29 are searched in log - space ) .
the bound derived above that demonstrated @xmath100 can be combined with the property that @xmath101 for any @xmath102 to show that @xmath103.\ ] ] thus the absolute error can be bound again using the fact that we are in a middle contour : @xmath104 the absolute error from middle contours will be quite small when @xmath105 is the maximum underflow - stable value of @xmath29 at index @xmath22 , because @xmath106 , the first factor in the error bound , will become @xmath107 , and @xmath108 ( qualitatively , this indicates that a small @xmath29 is only used when the result is very close to zero , leaving little room for absolute error ) .
likewise , when a very large @xmath29 is used , then @xmath109 becomes very small , while @xmath110 ( qualitatively , this indicates that when a large @xmath29 is used , the @xmath111 , and thus there is little absolute error ) .
thus for the extreme values of @xmath29 , middle contours will produce fairly small absolute errors .
the unique mode @xmath112 can be found by finding the value that solves @xmath113 which yields @xmath114 an appropriate choice of @xmath96 should be @xmath115 so that the error for any contour ( both middle contours and the top contour ) is smaller than the error achieved at @xmath112 , allowing us to use a single bound for both .
choosing @xmath116 would guarantee that all contours are no worse than the middle - contour error at @xmath112 ; however , using @xmath116 is still quite liberal , because it would mean that for indices in the highest contour ( there must be a nonempty set of such indices , because the scaling on @xmath76 and @xmath77 guarantees that the maximum index will have an exact value of @xmath50 , meaning that the approximation endures no underflow and is underflow - stable for every @xmath29 ) , a better error _ could _ be achieved by increasing @xmath96 .
for this reason , we choose @xmath96 so that the top - contour error produced at @xmath96 is not substantially larger than all errors produced for @xmath29 before the mode ( _ i.e. _ , for @xmath117 ) . choosing any value of @xmath118 guarantees the worst - case absolute error bound derived here ; however , increasing @xmath96 further over @xmath112 may possibly improve the mean squared error in practice ( because it is possible that many indices in the result would be numerically stable with @xmath29 values substantially larger than @xmath112 ) .
however , increasing @xmath119 will produce diminishing returns and generally benefit only a very small number of indices in the result , which have exact values very close to @xmath50 . in order to balance these two aims ( increasing @xmath96 enough over @xmath112 but not excessively so )
, we make a qualitative assumption that a non - trivial number of indices require us to use a @xmath29 below @xmath112 ; therefore , increasing @xmath96 to produce an error significantly smaller than the lowest worst - case error for contours below the mode ( _ i.e. _ @xmath117 ) will increase the runtime without significantly decreasing the mean squared error ( which will become dominated by the errors from indices that use @xmath117 ) .
the lowest worst - case error contour below the mode is @xmath120 ( because the absolute error function is unimodal , and thus must be increasing until @xmath112 and decreasing afterward ) ; therefore , we heuristically specify that @xmath96 should produce a worst - case error on a similar order of magnitude to the worst - case error produced with @xmath120 . in practice , specifying the errors at @xmath96 and @xmath120 should be equal is very conservative ( it produces very large estimates of @xmath96 , which sometimes benefit only one or two indices in the result ) ; for this reason , we heuristically choose that the worst - case error at @xmath96 should be no worse than square root of the worst case error at @xmath120 ( this makes the choice of @xmath96 less conservative because the errors at @xmath120 are very close to zero , and thus their square root is larger ) . the square root was chosen because it produced , for the applications described in this paper , the smallest value of @xmath96 for which the mean squared error was significantly lower than using @xmath116 ( the lowest value of @xmath96 guaranteed to produce the absolute error bound ) .
this heuristic does satisfy the worst - case bound outlined here ( because , again , @xmath118 ) , but it could be substantially improved if an expected distribution of magnitudes in the result vector were known ahead of time : prior knowledge regarding the number of points stable at each @xmath29 considered would enable a well - motivated choice of @xmath96 that truly optimizes the expected mean squared error . from this heuristic choice of @xmath96 , solving @xmath121 ( with the square root of the worst - case at @xmath120 on the left and the worst - case error at @xmath96 on the right ) yields @xmath122 for any non - trivial problem ( _ i.e. _ , when @xmath123 ) , and thus @xmath124 indicating that the absolute error at the top contour will be roughly equal to the fourth root of @xmath65 . by setting @xmath96 in this manner ,
we guarantee that the absolute error at any index of any unscaled problem on @xmath125 is less than @xmath126 ~ \max_r r[r ] ~ \tau^\frac{1}{2 p^*_{mode } } \left ( 1 - k_m^\frac{-1}{p^*_{mode } } \right)\ ] ] where @xmath112 is defined above . the full formula for the middle - contour error at this value of @xmath112 does not simplify and is therefore quite large ; for this reason , it is not reported here , but this gives a numeric bound of the worst case middle - contour error that is bound in terms of the variable @xmath28 ( and with no other free variables ) .
the piecewise method clearly performs @xmath127 ffts ( each requiring @xmath20 steps ) ; therefore , since @xmath96 is chosen to be @xmath128 ( to achieve the desired error bound ) , the total runtime is thus @xmath129 for any practically sized problem , the @xmath130 factor is essentially a constant ; even when @xmath28 is chosen to be the number of particles in the observable universe ( @xmath131 ; @xcite ) , the @xmath130 is @xmath132 , meaning that for any problem of practical size , the full piecewise method is no more expensive than computing between @xmath50 and @xmath133 ffts .
we first use an example max - convolution problem to compare the results from the low - value @xmath43 , the high - value @xmath44 and piecewise methods . at every index ,
these various approximation results are compared to the exact values , as computed by the naive quadratic method ( * figure [ figure : doublefigmethodsbimodalexa ] * ) .
.48 .48 * figure [ figure : doublefigmethodsbimodalexb ] * depicts a scatter plot of the exact result vs. the piecewise approximation at every index ( using the same problem from * figure [ figure : doublefigmethodsbimodalexa ] * ) .
it shows a clear banding pattern : the exact and approximate results are clearly correlated , but each contour ( _ i.e. _ , each collection of indices that use a specific @xmath29 ) has a different average slope between the exact and approximate values , with higher @xmath29 contours showing a generally larger slope and smaller @xmath29 contours showing greater spread and lower slopes .
this intuitively makes sense , because the bounds on @xmath134 $ ] derived above constrain the scatter plot points inside a quadrilateral envelope ( * figure [ figure : approxexactlinearzoom ] * ) .
( worst - case approximation ) .
the points are well described by an affine function fit using the left - most and right - most points.,width=480 ] the correlations within each contour can be exploited to correct biases that emerge for smaller @xmath29 values . in order to do this ,
@xmath135 must be computed for at least two points @xmath136 and @xmath137 within the contour , so that a mapping @xmath138 can be constructed .
fortunately , a single @xmath63 can be computed exactly in @xmath139 ( by actually computing a single @xmath21 and computing its max , which is equivalent to computing a single index result via the naive quadratic method ) . as long as the exact value @xmath135 is computed for only a small number of indices ,
the order of the runtime will not change ( each contour already costs @xmath20 , so adding a small number of @xmath139 steps for each contour will not change the asymptotic runtime ) . if the two indices chosen are @xmath140 @xmath141 then we are guaranteed that the affine function @xmath142 can be written as a convex combination of the exact values at those extreme points ( using barycentric coordinates ) : @xmath143 @xmath144\ ] ] thus , by computing @xmath145 and @xmath146 ( each in @xmath139 steps ) , we can compute an affine function @xmath142 to correct contour - specific trends ( * algorithm [ algorithm : numericalmaxconvolvepiecewiseimproved ] * ) .
.48 .48 @xmath70 $ ] @xmath71 $ ] @xmath72}$ ] @xmath73}$ ] @xmath74{\log_2(p^*_{\max } ) } } ] $ ] @xmath75 \gets$ ] ` fftnonnegmaxconvolvegivenpstar`(@xmath76 , @xmath77 , @xmath78 $ ] ) @xmath79 \gets \max \ { i:~ { \left ( resforallpstar[i][m ] \right)}^{\text{allpstar[$i$ ] } } \geq \tau ) \}$ ] @xmath147`affinecorrect`@xmath148 @xmath82 \times r[r_{\max } ] \times result$ ] @xmath149 \gets 1 $ ] @xmath150 \gets 0 $ ] @xmath151 @xmath152 = i \}$ ] @xmath153[m]$ ] @xmath154[m]$ ] @xmath155[mmin]$ ] @xmath156[mmax]$ ] @xmath157 @xmath158 @xmath159 \gets \frac{ymax - ymin}{xmax - xmin}$ ] @xmath160 \gets ymin - slope[i ] \times xmin$ ] @xmath159 \gets \frac{ymax}{xmax}$ ] @xmath80 $ ] @xmath81 \gets resforallpstar[i][m ] \times slope[i ] + bias[i]$ ] @xmath161 by exploiting the convex combination used to define @xmath142 , the absolute error of the affine piecewise method can also be bound .
qualitatively , this is because , by fitting on the extrema in the contour , we are now interpolating .
if the two points used to determine the parameters of the affine function were not chosen in this manner to fit the affine function , then it would be possible to choose two points with very close x - values ( _ i.e. _ , similar approximate values ) and disparate y - values ( _ i.e. _ , different exact values ) , and extrapolating to other points could propagate a large slope over a large distance ; using the extreme points forces the affine function to be a convex combination of the extrema , thereby avoiding this problem .
@xmath162\\ \shoveleft{\subseteq \left [ \lambda_m \frac{\| u^{(m_{\max } ) } \|_{p^*}}{k^\frac{1}{p^ * } } + \left ( 1 - \lambda_m \right ) \frac{\| u^{(m_{\min } ) } \|_{p^*}}{k^\frac{1}{p^ * } } , \right.}\\ \left .
\lambda_m \| u^{(m_{\max } ) } \|_{p^ * } + \left ( 1 - \lambda_m \right ) \| u^{(m_{\min } ) } \|_{p^ * } \right]\\ \shoveleft { = \left [ k^\frac{-1}{p^ * } \left ( \lambda_m \| u^{(m_{\max } ) } \|_{p^ * } + \left ( 1 - \lambda_m \right ) \| u^{(m_{\min } ) } \|_{p^ * } \right ) , \right.}\\ \left .
\lambda_m \| u^{(m_{\max } ) } \|_{p^ * } + \left ( 1 - \lambda_m \right ) \| u^{(m_{\min } ) } \|_{p^ * } \right]\\ \shoveleft{= \left [ k^\frac{-1}{p^ * } \| u^{(m ) } \|_{p^ * } , \| u^{(m ) } \|_{p^ * } \right]}\end{gathered}\ ] ] the worst - case absolute error of the scaled problem on @xmath84 can be defined @xmath163 because the function @xmath164 is affine , it s derivative can never be zero , and thus lagrangian theory states that the maximum must occur at a boundary point .
therefore , the worst - case absolute error is @xmath165 which is identical to the worst - case error bound before applying the affine transformation @xmath142 . thus applying
the affine transformation can dramatically improve error , but will not make it worse than the original worst - case .
one example that profits from fast max - convolution of non - negative vectors is computing the viterbi path using a hidden markov model ( hmm ) ( _ i.e. _ , the _ maximum a posteriori _ states ) with an additive transition function satisfying @xmath166 for some arbitrary function @xmath167 ( @xmath167 can be represented as a table , because we are considering all possible discrete functions ) .
this additivity constraint is equivalent to the transition matrix being a `` toeplitz matrix '' : the transition matrix @xmath168 is a toeplitz matrix when all cells diagonal from each other ( to the upper left and lower right ) have identical values ( _ i.e. _ , @xmath169 ) . because of the markov property of the chain , we only need to max - marginalize out the latent variable at time @xmath170 to compute the distribution for the next latent variable @xmath171 and all observed values of the data variables @xmath172 .
this procedure , called the viterbi algorithm , is continued inductively : @xmath173 and continuing by exploiting the self - similarity on a smaller problem to proceed inductively , revealing a max - convolution ( for this specialized hmm with additive transitions ) : @xmath174 \pr(d_{i-1 } | x_{i-1}=x_{i-1 } ) \delta[x_i - x_{i-1 } ] = } \\ \shoveright{\left(fromleft[i-1]~likelihood[d_{i-1 } ] \right ) ~*_{\max}~ \delta[x_i - x_{i-1}].}\\\end{gathered}\ ] ] after computing this left - to - right pass ( which consisted of @xmath175 max - convolutions and vector multiplications ) , we can find the _ maximum a posteriori _ configuration of the latent variables @xmath176 backtracking right - to - left , which can be done by finding the variable value @xmath177 that maximizes @xmath178[x_i ] \times \delta[x_{i+1}^ * - x_i]$ ] ( thus defining @xmath179 and enabling induction on the right - to - left pass ) .
the right - to - left pass thus requires @xmath180 steps ( * algorithm [ algorithm : viterbi ] * ) .
note that the full max - marginal distributions on each latent variable @xmath36 can be computed via a small modification , which would perform a more complex right - to - left pass that is nearly identical to the left - to - right pass , but which performs subtraction instead of addition ( _ i.e. _ , by reversing the vector representation of the pmf of the subtracted argument before it is max - convolved ; @xcite ) .
@xmath181 \gets prior$ ] @xmath178 \gets fromleft[i ] \times likelihood[data[i]]$ ] @xmath182 \gets fromleft[i ] * _ { \max } \delta$ ] @xmath183 \gets fromleft[n ] \times likelihood[data[n]]$ ] @xmath184 \gets { \operatornamewithlimits{argmax}}_{j } fromleft[n-1][j]$ ] @xmath185 @xmath186 @xmath187 \times \delta[l - path[i+1]]$ ] @xmath188 @xmath189 @xmath190 \gets argmaxprodposterior$ ] @xmath191 we apply this hmm with additive transition probabilities to a data analysis problem from economics .
it is known for example , that the current figures of unemployment in a country have ( among others ) impact on prices of commodities like oil .
if one could predict unemployment figures before the usual weekly or monthly release by the responsible government bureaus , this would lead to an information advantage and an opportunity for short - term arbitrage .
the close relation of economic indicators like market prices and stock market indices ( especially of indices combining several stocks of different industries ) to unemployment statistics can be used to tackle this problem . in the following demonstration of our method
, we create a simple hmm with additive transitions and use it to infer the _ maximum a posteriori _ unemployment statistics given past history ( _ i.e. _ how often unemployment is low and high , as well as how often unemployment goes down or up in a short amount of time ) and current stock market prices ( the observed data ) .
we discretized random variables for the observed data ( s&p 500 , adjusted closing prices ; retrieved from yahoo !
historical stock prices : http://data.bls.gov/cgi-bin/surveymost?bls series cuur0000sa0[http://data.bls.gov / cgi - bin / surveymost?bls series cuur0000sa0 ] ) , and `` latent '' variables ( unemployment insurance claims , seasonally adjusted , were retrieved from the u.s .
department of labor : https://www.oui.doleta.gov/unemploy/claims.asp ) .
stock prices were additionally inflation adjusted by ( _ i.e. _ divided by ) the consumer price index ( cpi ) ( retrieved from the u.s .
bureau of labor statistics : https://finance.yahoo.com/q?s=^gspc[https://finance.yahoo.com/q?s=^gspc ] ) . the intersection of both `` latent '' and observed data was available weekly from week 4 in 1967 to week 52 in 2014 , resulting in 2500 data points for each type of variable . to investigate the influence of overfitting
, we partition the data in two parts , before june 2005 and after june 2005 , so that we are effectively training on @xmath192 of the data points , and then demonstrate the viterbi path on the entirety of the data ( both the @xmath193 training data and the @xmath194 of the data withheld from empirical parameter estimation ) .
unemployment insurance claims were discretized into @xmath195 and stock prices were discretized into @xmath196 bins .
simple empirical models of the prior distribution for unemployment , the likelihood of unemployment given stock prices , and the transition probability of unemployment were built as follows : the initial or prior distribution for unemployment claims at @xmath197 was calculated by marginalizing the time series of training data for the claims ( _ i.e. _ counting the number of times any particular unemployment value was reached over all possible bins ) .
our transition function ( the conditional probability @xmath198 ) similarly counts the number of times each possible change @xmath199 occurred over all available time points .
interestingly , the resulting transition distribution roughly resembles a gaussian ( but is not an exact gaussian ) .
this underscores a great quality of working with discrete distributions : while continuous distributions may have closed - forms for max - convolution ( which can be computed quickly ) , discrete distributions have the distinct advantage that they can accurately approximate any smooth distribution .
lastly , the likelihoods of observing a stock price given the unemployment at the same time were trained using an empirical joint distribution ( essentially a heatmap ) , which is displayed in * figure [ figure : likelihoodheatmap]*. we compute the viterbi path two times : first we use naive , exact max - convolution , which requires a total of @xmath200 steps .
second , we use fast numerical max - convolution , which requires @xmath201 steps . despite the simplicity of the model ,
the exact viterbi path ( computed via exact max - convolution ) is highly informative for predicting the value of unemployment , even for the @xmath194 of the data that were not used to estimate the empirical prior , likelihood , and transition distributions .
also , the numerical max - convolution method is nearly identical to the exact max - convolution method at every index ( * figure [ figure : viterbi ] * ) . even with a fairly rough discretization ( _ i.e. _ , @xmath202 ) , the fast numerical method used @xmath203 seconds compared to the @xmath204 seconds required by the naive approach .
this speedup will increase dramatically as @xmath28 is increased , because the @xmath205 term in the runtime of the numerical max - convolution method is essentially bounded above @xmath206 .
although the @xmath29-norm provides a good approximation of the chebyshev norm , it discards significant information ; specifically the curve @xmath66 for various @xmath207 could be used to identify and correct the worst - case scenario where @xmath208 ; using only two points , the exact value of @xmath63 can be computed for those worst - case @xmath21 vectors by computing the norms at two different @xmath29 values and solving the following equations for @xmath209 : @xmath210 where the proportionality constant is @xmath211 and where the computed value @xmath209 yields the exact chebyshev norm @xmath135 . more generally , when there are @xmath212 unique values ( @xmath213 ) in @xmath21 , we can model the norms perfectly with @xmath214 where @xmath215 is an integer that indicates the number of times @xmath213 occurs in @xmath21 ( and where @xmath216 ) . this multi - set view of the vector @xmath21 can be used to project it down to a dimension @xmath217 : @xmath218 \cdot \left [ \begin{array}{c } n_1\\ n_2\\ n_3\\ \vdots\\ n_r\\ \end{array } \right ] = \left [ \begin{array}{c } \| u^{(m ) } \|_{p^*}^{p^ * } \\ \| u^{(m ) } \|_{2 { p^*}}^{2 { p^*}}\\ \| u^{(m ) } \|_{2 { p^*}}^{2 { p^*}}\\ \vdots\\ \| u^{(m ) } \|_{\ell { p^*}}^{\ell { p^*}}\\ \end{array } \right].\ ] ] by solving the above system of equations for all @xmath219 , the maximum @xmath220 can be used to approximate the true maximum @xmath221
. this projection can be thought of as querying distinct moments of the distribution @xmath222 that corresponds to some unknown vector @xmath21 , and then assembling the moments into a model in order to predict the unknown maximum value in @xmath21 .
of course , when @xmath217 , the number of terms in our model , is sufficiently large , then computing @xmath217 norms of @xmath21 will result in an exact result , but it could result in @xmath223 execution time , meaning that our numerical max - convolution algorithm becomes quadratic ; therefore , we must consider that a small number of distinct moments are queried in order to estimate the maximum value in @xmath21 .
regardless , the system of equations above is quite difficult to solve directly via elimination for even very small values of @xmath217 , because the symbolic expressions become quite large and because symbolic polynomial roots can not be reliably computed when the degree of the polynomial is @xmath224 . even in cases
when it can be solved directly , it will be far too inefficient .
for this reason , we solve for the @xmath219 values using an exact , alternative approach : if we define a polynomial @xmath225 , then @xmath226 .
we can expand @xmath227 , and then write @xmath228 \cdot \left [ \begin{array}{cccc } \alpha_1^{p^ * } & \alpha_2^{p^ * } & & \alpha_r^{p^ * } \\ \alpha_1^{2 { p^ * } } & \alpha_2^{2 { p^ * } } & \cdots & \alpha_r^{2 { p^ * } } \\
\alpha_1^{3 { p^ * } } & \alpha_2^{3 { p^ * } } & & \alpha_r^{3 { p^ * } } \\ \vdots & \vdots & & \vdots \\ \alpha_1^{\ell { p^ * } } & \alpha_2^{\ell { p^ * } } & & \alpha_r^{\ell { p^ * } } \\ \end{array } \right ] \cdot \left [ \begin{array}{ccccc } n_1\\ n_2\\ n_3\\ \vdots\\ n_r\\ \end{array } \right ] = \\ \left [ \begin{array}{ccccc } \alpha_1^{p^ * } \gamma(\alpha_1^{p^ * } ) & \alpha_2^{p^ * } \gamma(\alpha_2^{p^ * } ) & \alpha_3^{p^ * } \gamma(\alpha_3^{p^ * } ) & \cdots & \alpha_r^{p^ * } \gamma(\alpha_r^{p^ * } ) \end{array } \right ] \cdot \left [ \begin{array}{ccccc } n_1\\ n_2\\ n_3\\ \vdots\\ n_r\\ \end{array } \right ] = \\ \left [ \begin{array}{ccccc } 0 & 0 & 0 & \cdots & 0 \end{array } \right ] \cdot \left [ \begin{array}{ccccc } n_1\\ n_2\\ n_3\\ \vdots\\ n_r\\ \end{array } \right ] = 0,\end{gathered}\ ] ] which indicates that @xmath229 \cdot \left [ \begin{array}{c } \| u^{(m ) } \|_{p^*}^{p^ * } \\ \| u^{(m ) } \|_{2 { p^*}}^{2 { p^*}}\\ \| u^{(m ) } \|_{2 { p^*}}^{2 { p^*}}\\ \vdots\\ \| u^{(m ) } \|_{\ell { p^*}}^{\ell { p^*}}\\ \end{array } \right ] = 0.\ ] ] furthermore , @xmath230 ; therefore we can write @xmath231 \cdot \left [ \begin{array}{c } \| u^{(m ) } \|_{p^*}^{p^ * } \\ \| u^{(m ) } \|_{2 { p^*}}^{2 { p^*}}\\ \| u^{(m ) } \|_{2 { p^*}}^{2 { p^*}}\\ \vdots\\ \| u^{(m ) } \|_{\ell { p^*}}^{\ell { p^*}}\\ \end{array } \right ] = \\ \left [ \begin{array}{ccccc } \| u^{(m ) } \|_{p^*}^{p^ * } & \| u^{(m ) } \|_{2 { p^*}}^{2 { p^ * } } & \| u^{(m ) } \|_{3 { p^*}}^{3 { p^ * } } & \cdots & \| u^{(m ) } \|_{(r+1 ) { p^*}}^{(r+1 ) { p^ * } } \\ \| u^{(m ) } \|_{2 { p^*}}^{2 { p^ * } } & \| u^{(m ) } \|_{3 { p^*}}^{3 { p^ * } } & \| u^{(m ) } \|_{4 { p^*}}^{4 { p^ * } } & \cdots & \| u^{(m ) } \|_{(r+2 ) { p^*}}^{(r+2 ) { p^ * } } \\ \| u^{(m ) } \|_{3 { p^*}}^{3 { p^ * } } & \| u^{(m ) } \|_{4 { p^*}}^{4 { p^ * } } & \| u^{(m ) } \|_{5 { p^*}}^{5 { p^ * } } & \cdots & \| u^{(m ) } \|_{(r+3 ) { p^*}}^{(r+3 ) { p^ * } } \\ & & \vdots & & \\
\| u^{(m ) } \|_{(\ell - r-1 ) { p^*}}^{(\ell - r-1 ) { p^ * } } & \cdots & \| u^{(m ) } \|_{(\ell-2 ) { p^*}}^{(\ell-2 ) { p^ * } } & \| u^{(m ) } \|_{(\ell-1 ) { p^*}}^{(\ell-1 ) { p^ * } } & \| u^{(m ) } \|_{\ell { p^*}}^{\ell { p^ * } } \\
\end{array } \right ] \cdot \left [ \begin{array}{c } \gamma_0\\ \gamma_1\\ \gamma_2\\ \vdots\\ \gamma_r \end{array } \right ] = 0.\end{gathered}\ ] ] therefore , @xmath232
\in null\left ( \left [ \begin{array}{ccccc } \| u^{(m ) } \|_{p^*}^{p^ * } & \| u^{(m ) } \|_{2 { p^*}}^{2 { p^ * } } & \| u^{(m ) } \|_{3 { p^*}}^{3 { p^ * } } & \cdots & \| u^{(m ) } \|_{(r+1 ) { p^*}}^{(r+1 ) { p^ * } } \\
\| u^{(m ) } \|_{2 { p^*}}^{2 { p^ * } } & \| u^{(m ) } \|_{3 { p^*}}^{3 { p^ * } } & \| u^{(m ) } \|_{4 { p^*}}^{4 { p^ * } } & \cdots & \| u^{(m ) } \|_{(r+2 ) { p^*}}^{(r+2 ) { p^ * } } \\ \| u^{(m ) } \|_{3 { p^*}}^{3 { p^ * } } & \| u^{(m ) } \|_{4 { p^*}}^{4 { p^ * } } & \| u^{(m ) } \|_{5 { p^*}}^{5 { p^ * } } & \cdots & \| u^{(m ) } \|_{(r+3 ) { p^*}}^{(r+3 ) { p^ * } } \\ & & \vdots & & \\
\| u^{(m ) } \|_{(\ell - r-1 ) { p^*}}^{(\ell - r-1 ) { p^ * } } & \cdots & \| u^{(m ) } \|_{(\ell-2 ) { p^*}}^{(\ell-2 ) { p^ * } } & \| u^{(m ) } \|_{(\ell-1 ) { p^*}}^{(\ell-1 ) { p^ * } } & \| u^{(m ) } \|_{\ell { p^*}}^{\ell { p^ * } } \\
\end{array } \right ] \right).\ ] ] because the columns of @xmath218\ ] ] must be linearly independent when @xmath233 are distinct ( which is the case by the definition of our multiset formulation of the norm ) , then @xmath234 will determine a unique solution ; thus the null space above is computed from a matrix with @xmath235 columns and @xmath217 rows , yielding a single vector for @xmath236 .
this vector can then be used to compute the roots of the polynomial @xmath237 , which will determine the values @xmath238 , which can each be taken to the @xmath239 power to compute @xmath240 ; the largest of those @xmath219 values is used as the estimate of the maximum element in @xmath21 .
when @xmath21 contains at least @xmath217 distinct values ( _ i.e. _ , @xmath241 ) , then the problem will be well - defined ; thus , if the roots of the null space spanning vector are not well - defined , then a smaller @xmath217 can be used ( and should be able to compute an exact estimate of the maximum , since @xmath21 can be projected exactly when @xmath217 is the precise number of unique elements found in @xmath21 ) . note that this projection method is valid for any sequence of norms with even spacing : @xmath242 .
in general , the computation of both the null space spanning vector @xmath236 and of machine - precision approximations for the roots of the polynomial @xmath244 ( which can be approximated by constructing a matrix with that characteristic polynomial and performing eigendecomposition @xcite ) are both in @xmath245 for each index @xmath22 in the result ; however , by using a small @xmath243 , we can compute a closed form solution of both the null space spanning vector and of the resulting quadratic roots .
this enables faster exploitation of the curve of norms for estimating the maximum value of @xmath21 ( although it does nt achieve the high accuracy possible with a much larger @xmath246 ) .
this is equivalent to approximating @xmath247 , where @xmath248 . in this case , the single spanning vector of the null space of @xmath249\ ] ] will be @xmath250 = \left [ \begin{array}{c } \| u^{(m ) } \|_{2 p^*}^{2 p^ * } \| u^{(m ) } \|_{4 p^*}^{4 p^ * } - { \left ( \| u^{(m ) } \|_{3 p^*}^{3 p^ * } \right)}^2\\ \| u^{(m ) } \|_{p^*}^{p^ * } \| u^{(m ) } \|_{4 p^*}^{4 p^ * } - \| u^{(m ) } \|_{2 p^*}^{2 p^ * } \| u^{(m ) } \|_{3 p^*}^{3 p^*}\\ \| u^{(m ) } \|_{p^*}^{p^ * } \| u^{(m ) } \|_{3 p^*}^{3 p^ * } - { \left ( \| u^{(m ) } \|_{2 p^*}^{2 p^ * } \right)}^2\\ \end{array } \right]\ ] ] and thus @xmath251 can be computed by using the quadratic formula to solve @xmath252 for @xmath64 , and computing @xmath253 using the maximum of those zeros : @xmath254 .
when the quadratic is not well defined , then this indicates that the number of unique elements in @xmath21 is less than 2 , and thus can not be projected uniquely ( _ i.e. _ , @xmath255 ) ; in this case , the closed - form linear solution can be used rather than a closed - form quadratic solution : @xmath256 when the closed - form linear solution is not numerically stable ( due to division by a value close to zero ) , then the @xmath29-norm approximation can likewise be used . because the norms must have evenly spaced @xmath29 values in order to use the projection method described above
, the exponential sequence of @xmath29 values used in the original piecewise algorithm will not contain four evenly spaced points ( which are necessary to solve the quadratic formulation , _
i.e. _ @xmath243 ) .
one possible solution would be to take the maximal stable value of @xmath29 for any index ( which will be a power of two found using the original piecewise method ) , and then also computing norms ( via the fft , as before ) for @xmath257 ; however , this will result in a @xmath258 slowdown in the algorithm , because for every @xmath29-norm computed via fft before , now four must be computed .
an alternative approach reuses existing values in the @xmath259 sequence of @xmath29 : for @xmath29 sufficiently large , then the exponential sequence is guaranteed to include these stable @xmath29 values : @xmath260 . by considering @xmath261 in @xmath29 candidates , then we can be guaranteed to have four evenly spaced and stable @xmath29 values .
this can be achieved easily by noting that @xmath262 meaning that we can insert all possible necessary @xmath29 values for evenly spaced sequences of length four by first computing the exponential sequence of @xmath29 values and then inserting the averages between every pair of adjacent powers of two ( and inserting them in a way that maintains the sorted order ) : @xmath263 becomes @xmath264 .
thus , if ( for some index @xmath22 ) 16 is the highest stable @xmath29 that is a power of two ( _ i.e. _ , the @xmath29 value that would be used by the original piecewise algorithm ) , then we are guaranteed to use the evenly spaced sequence @xmath265 . by interleaving the powers of two with the averages from the following powers of two , we reduce the number of ffts to @xmath266 that used by the original piecewise algorithm . for small values of @xmath217 ( such as the @xmath243 used here ) , the estimation of the maximum from each sequence of four norms is in @xmath267 , meaning the total time will still be @xmath268 , which is the same as before . because the spacing in this formulation is @xmath269 , and given the maximal root of the quadratic polynomial @xmath270 , then @xmath271 ( taking the maximal root @xmath272 to the power @xmath273 instead of @xmath274 , which had been the spacing used in the description of the projection method ) .
the null space projection method is shown in * algorithm [ algorithm : piecewisewithprojection]*. @xmath70 $ ] @xmath71 $ ] @xmath72}$ ] @xmath73}$ ] @xmath275{\log_2(p^*_{\max } ) } } ] $ ] @xmath276 \gets allpstar[i ] $ ] @xmath277 \gets 0.5 \times ( allpstar[i]+allpstar[i+1 ] ) $ ] @xmath75 \gets$ ] ` fftnonnegmaxconvolvegivenpstar`(@xmath76 , @xmath77 , @xmath278 $ ] ) @xmath79 \gets \max \ { i:~ { \left ( resforallpstar[i][m ] \right)}^{\text{allpstarinterleaved[$i$ ] } } \geq \tau ) \}$ ] @xmath279 -= maxstablepstarindex[o ] \% 2 $ ] @xmath280 $ ] @xmath281 @xmath282 $ ] @xmath283 - 1]$ ] @xmath284 \gets$ ] ` maxlin`@xmath285 @xmath286 - 2]$ ] @xmath287 - 4]$ ] @xmath284 \gets$ ] ` maxquad`@xmath288 @xmath147`affinecorrect`@xmath148 @xmath82 \times r[r_{\max } ] \times result$ ] @xmath289 @xmath290 @xmath291 @xmath292 @xmath293 @xmath294 @xmath295 @xmath296 @xmath297 @xmath298 ` maxlin`@xmath285 @xmath291 the full closed - form of the quadratic roots used above ( which solve the projection when @xmath243 ) will be @xmath299 where @xmath300 in the pseudocode ( _ i.e. _ , the maximum numerically stable @xmath29 used by the piecewise algorithm at that index ) .
note that @xmath301 can be factored out because the exponents in every term in the numerator will be @xmath302 ( _ i.e. _ , @xmath303 in the square root ) .
similarly the terms in the denominator each contain @xmath304 .
factoring out the maximum value is then the same as operating on scaled vectors @xmath305 ( instead of @xmath306 ) with the maximum entry being @xmath50 , and at least one element of value @xmath50 .
furthermore , the denominator @xmath307 ; even though the terms summed to compute @xmath308 are not exclusively nonnegative , symmetry can be used to demonstrate that every negative term is outweighed by a unique corresponding term : @xmath309 thus , for well - defined problems ( _ i.e. _ , when @xmath310 ) , the denominator @xmath311 , and therefore , the maximum root of the quadratic polynomial will correspond to the term that adds ( rather than subtracts ) the square root term : @xmath312 the relative absolute error is defined as @xmath313 ; therefore , a bound on the relative error of the projection method can be established by bounding @xmath314 where the length of the @xmath21 ( respectively @xmath89 ) is @xmath90 . using this reformulation , @xmath315 indicates a zero - error approximation .
this can be rewritten to bound its value before taking to the power @xmath274 : @xmath316 where @xmath317 the extreme values of @xmath318 can be found by minimizing and maximizing over the possible values of @xmath319}^{k_m } : \exists i , v_i = 1 , \exists j , v_j \in ( 0,1 ) \}$ ] .
the final constraint on @xmath320 in ( 0,1 ) is because any @xmath305 containing only one unique value ( which must be @xmath50 in this case since dividing by the maximum element in @xmath21 to compute @xmath89 has divided the value at that index by itself ( @xmath321 ) will lead to instabilities . when values in @xmath305 are identical to one another , using @xmath322 yields an exact solution , and thus solving with @xmath243 is not well - defined because @xmath323 . because all elements @xmath324 $ ] and @xmath102 , we can perform a change of variables @xmath325 , thereby eliminating references to @xmath29 : @xmath326 @xmath327 r|ccccc @xmath90 & 3 & 4 & 5 & 6 & 7 + + minimum & 0.935537 & 0.902161 & 0.895671 & 0.880487 & 0.85343 + maximum & 1 & 1 & 1 & 1 & 1 + for small vector lengths , the exact bounds of @xmath328 are shown in * table [ table : boundsmathematica]*. notice that the upper bound is fixed , but the lower bound grows monotonically smaller as @xmath90 , the length of the vector considered , increases . for larger vectors ,
mathematica does not find optima in a matter of hours , and for arbitrary - length vectors , the karush - kuhn - tucker criteria do not easily yield minima or maxima ; however , we do observe that all maxima are achieved by vectors that are permutations ( order does not influence the result ) of @xmath329 ( again , when only two unique values are found in @xmath305 , the approximation is exact and thus @xmath330 ) . likewise ,
the minima are achieved by permutations of @xmath331 .
for this reason , we perform further empirical estimation of the bound by randomly sampling vectors of the form @xmath332 with @xmath333 degrees of freedom ( d.o.f . ) and sampling vectors of the form @xmath334 ( with @xmath335 d.o.f . ) , whose extrema are shown in * table [ table : boundspysim]*. r|ccccc @xmath90 & 4 & 64 & 1024 + + minimum ( @xmath333 d.o.f . ) & 0.90221268 & 0.74942834 & 0.81858283 + maximum ( @xmath333 d.o.f . ) & 0.99999986 & 0.92482416 & 0.86795636 + + minimum ( vectors of form @xmath336 , @xmath335 d.o.f . ) & 0.90216688 & 0.75455478 & 0.71695386 + maximum ( vectors of form @xmath336 , @xmath335 d.o.f . ) & 1.00000000 & 1.0000000 & 1.00000000 + at length @xmath337 we see that due to an extreme value scenario , an unconstrained vector scores slightly lower than a vector holding the worst - case pattern @xmath336 , because both forms of sampling approach the true lower bound , but one of the unconstrained @xmath338 d.o.f .
is slightly closer . from these results
, we conjecture that @xmath339 is bounded above @xmath340 ( this is achievable at any length @xmath90 by letting @xmath305 contain exactly two distinct values ) . in this manner
, we achieve our predicted upper bound of @xmath50 regardless of the length @xmath90 .
likewise , we conjecture that at any @xmath90 ( not simply the lengths investigated , where this principle is true ) , the lower bound is given by vectors of the form @xmath341 .
qualitatively , this conjecture stems from the fact that since the estimate is perfect when @xmath305 contains exactly two distinct elements , then the worst - case lower bound when @xmath305 contains three distinct values will concentrate the points at some value far from the other two distinct values .
when four distinct values are permitted , then we conjecture that the optimal choice ( for minimizing @xmath339 ) of the fourth value will equal the choice for the third distinct value , since that was already determined to be the best point for deceiving the quadratic approximation . from this conjecture
, we can then use the fact that the bounds should only grow more extreme as @xmath90 increases , since @xmath342 ( _ i.e. _ lower - dimensional solutions can always be reached in a higher dimension by setting some of the values to @xmath54 ) .
thus the minimum for any possible vector should be conservatively bounded below on vectors of the form @xmath341 and is achieved by letting @xmath90 approach @xmath343 : @xmath344 the minimum value of this expression over all @xmath345 , b \in [ 0,1]$ ] is @xmath346 ( computed again with mathematica ) .
overall , assuming our conjecture regarding the forms of the vectors achieving the minima and maxima , then it follows that @xmath347 $ ] , and the worst - case relative error at the @xmath96 contour will be bounded @xmath348 the steeper decrease in relative error as @xmath29 is increased means that the same procedure can be used to achieve an absolute error bound : @xmath349 which achieves a unique maximum at @xmath350 as before , the worst - case absolute error of the unscaled problem will be found by simply scaling the absolute error at @xmath112 : @xmath126~ \max_r r[r]~ \tau^{\frac{1}{2 p^*_{mode } } } \left ( 1 - { 0.7}^{\frac{4}{p^*_{mode } } } \right).\ ] ] because @xmath112 ( the value of @xmath29 producing the worst - case absolute error ) for the null space projection method it is invariant to the length of the list @xmath28 ( enabling us to compute a numeric value ) , and because its numeric value is so small , even a fairly small choice of @xmath96 will suffice ( now @xmath351 rather than in @xmath352 as it was with the original piecewise method ) . for example , the approximation of the viterbi path to infer the unemployment data is slightly superior with the null space projection method , even when @xmath353 is used ( in contrast to the @xmath354 used in the figure [ figure : viterbi ] ) . the null space projection method required @xmath355 seconds ( slightly faster than the @xmath203 seconds required by the original piecewise method ) .
the one caveat of this worst - case absolute error bound is that it presumes at least four evenly spaced , stable @xmath29 can be found ( which may not be the case by choosing @xmath29 from the sequence @xmath259 in cases when @xmath356 ) ; however , assuming standard fast convolution can be performed ( a reasonable assumption given it is one of the essential numeric algorithms ) , then four evenly spaced @xmath29 values could be chosen very close to @xmath50 ; therefore , these values of @xmath29 could be added to the sequence so that the algorithm is slightly slower , but essentially always yield this worst - case absolute error bound . in practice , we can demonstrate that the null space projection method is very accurate .
first we show the impact of using the quadratic ( _ i.e. _ , @xmath243 ) projection method on unscaled single @xmath21 vectors .
the projection method was tested on vectors of different lengths drawn from different types of beta distributions and are compared with the results of the @xmath1-norms with the highest stable @xmath1 ( figure [ figure : noextrapolvsextrapol ] ) .
the relative errors between the original piecewise method and the null space projection method are compared using a max - convolution on two randomly created input pmfs of lengths 1024 ( figure [ figure : affinevsaffineextrapol ] ) .
note that the null space projection can also be paired with affine scaling on the back end , just as the original piecewise method can be . in practice
, the null space projection increases the accuracy demonstrably on a variety of different problems , although the original piecewise method also performs well .
although the worst - case runtime of the null space projection method is roughly @xmath266 that of the original piecewise method , the error bound no longer depends on the length of the result @xmath28 .
thus , for a given relative error bound on the top contour ( _ i.e. _ , the equivalent of the derivation of @xmath96 in the original piecewise algorithm ) , the value of @xmath96 is fixed and no longer @xmath357 .
for example , achieving a @xmath358 relative error in the top contour would require @xmath359 meaning that choosing @xmath360 would achieve a very high accuracy , but while only performing @xmath361 ffts .
for very large vectors , this will not be substantially more expensive than the original piecewise algorithm , which uses a higher value of @xmath96 ( in this case , @xmath362 , which continues to grow as @xmath28 does ) to keep the error lower in practice . as a result
, the runtime of the null space projection approximation is @xmath363 rather than @xmath364 , despite the similar runtime in practice to the original piecewise method ( the null space projection method uses @xmath266 as many ffts performed per @xmath29 value , but requires slightly fewer @xmath29 values ) .
r|ccccccc @xmath28 & @xmath365 & @xmath366 & @xmath367 & @xmath368 & @xmath369 & @xmath370 & @xmath371 + + naive & * 0.0142 * & 0.0530 & 0.192 & 0.767 & 3.03 & 12.1 & 48.2 + naive ( vectorized ) & 0.0175 & 0.0381 & 0.0908 & 0.251 & 0.790 & 2.75 & 10.1 + fill1 ( @xcite ) & 0.0866 & 1.09 & 7.21 & 19.4 & 457 & & + max .
stable @xmath29 , affine corrected & 0.0277 & 0.0353 & 0.0533 & 0.0848 & 0.149 & 0.274 & 0.537 + projection , affine corrected & 0.0236 & * 0.0307 * & * 0.0467 * & * 0.0760 * & * 0.137 * & * 0.258 * & * 0.520 * + to compare the actual runtimes of the final algorithm developed in this manuscript with a naive max - convolution and a previously proposed method from @xcite , all methods were run on vectors of different random ( uniform in @xmath372 $ ] ) composition and length ( @xmath28 ) .
the first and second input vector were generated seperately but are always of same length .
* table [ table : runtimesinclbussieck ] * shows the result of this experiment .
all methods were implemented in python , using numpy where applicable ( _ e.g. _ to vectorize ) . a non - vectorized version of naive max - convolution was included to estimate the effects of vectorization .
the approach from bussieck et al . ran as a reimplementation based on the pseudocode in their manuscript . from their variants of proposed methods ,
fill1 was chosen because of its use in their corresponding benchmark and its recommendation by the authors for having a lower runtime constant in practice compared to other methods they proposed .
the method is based on sorting the input vectors and traversing the ( implicitly ) resulting partially ordered matrix of products in a way that not all entries need to be evaluated , while only keeping track of the so - called cover of maximal elements .
fill1 already includes some more sophisticated checks to keep the cover small and thereby reducing the overhead per iteration .
unfortunately , although we observed that the fill1 method requires between @xmath373 and @xmath374 iterations in practice , this per - iteration overhead results in a worst - case cost of @xmath375 per iteration , yielding an overall runtime in practice between @xmath376 and @xmath377 . as the authors state , this overhead is due to the expense of storing the cover , which can be implemented _
e.g. _ using a binary heap ( recommended by the authors and used in this reimplementation ) . additionally , due to the fairly sophisticated datastructures needed for this algorithm it had a higher runtime constant than the other methods presented here , and furthermore we saw no means to vectorize it to improve the efficiency . for this reason ,
it is not truly fair to compare the raw runtimes to the other vectorized algorithms ( and it is not likely that this python reimplementation is as efficient as the original version , which @xcite implemented in ansi - c ) ; however , comparing a non - vectorized implementation of the naive @xmath374 approach with its vectorized counterpart gives an estimated @xmath378 speedup from vectorization , suggesting that it is not substantially faster than the naive approach on these problems ( it should be noted that whereas the methods presented here have tight runtime bounds but produce approximate results , the fill1 algorithm is exact , but its runtime depends on the data processed ) . during investigation of these runtimes , we found that on the given problems , the proposed average case of @xmath373 iterations was rarely reached .
a reason might be an unrecognized violation of the assumptions of the theory behind this theoretical average runtime in how the input vectors were generated .
in contrast to the exact method from @xcite , the herein proposed approximate procedure are faster whenever the input vectors are at least @xmath196 elements long ( shorter vectors are most efficiently processed with the naive approach ) .
the null space projection method is the fastest method presented here ( because it can use a lower @xmath96 ) , although the higher density of @xmath29 values it uses ( and thus , additional ffts ) make the runtimes nearly identical for both approximation methods .
both piecewise numerical max - convolution methods are highly accurate in practice and achieve a substantial speedup over both the naive approach and the approach proposed by @xcite .
this is particularly true for large problems : for the original piecewise method presented here , the @xmath379 multiplier may never be small , but it grows so slowly with @xmath28 that it will be @xmath380 even when @xmath28 is on the same order of magnitude as the number of particles in the observable universe .
this means that , for all practical purposes , the method behaves asymptotically as a slightly slower @xmath20 method , which means the speedup relative to the naive method becomes more pronounced as @xmath28 becomes large . for the second method presented ( the null space projection ) , the runtime for a given relative error bound will be in @xmath33 . in practice ,
both methods have similar runtime on large problems . the basic motivation of the first approach described _ i.e. _ , the idea of approximating the chebyshev norm with the largest @xmath29-norm that can be computed accurately , and then convolving according to this norm using fft also suggests further possible avenues of research .
for instance , it may be possible to compute a single fft ( rather than an fft at each of several contours ) on a more precise implementation of complex numbers .
such an implementation of complex values could store not only the real and imaginary components , but also other much smaller real and imaginary components that have been accumulated through @xmath9 operations , even those which have small enough magnitudes that they are dwarfed by other summands .
with such an approach it would be possible to numerically approximate the max - convolution result in the same overall runtime as long as only a bounded `` history '' of such summands was recorded ( _ i.e. _ , if the top few magnitude summands whether that be the top 7 or the top @xmath381was stored and operated on ) . in a similar vein , it would be interesting to investigate the utility of complex values that use rational numbers ( rather than fixed - precision floating point values ) , which will be highly precise , but will increase in precision ( and therefore , computational complexity of each arithmetic operation ) as the dynamic range between the smallest and largest nonzero values in @xmath25 and @xmath26 increases ( because taking @xmath76 to a large power @xmath29 may produce a very small value ) . other simpler improvements could include optimizing the error vs. runtime trade - off between the log - base of the contour search : the method currently searches @xmath127 contours , but a smaller or larger log - base could be used in order to optimize the trade - off between error and runtime .
it is likely that the best trade - off will occur by performing the fast @xmath29-norm convolution with a number type that sums values over vast dynamic ranges by appending them in a short ( _ i.e. _ , bounded or constant size ) list or tree and sums values within the same dynamic range by querying the list or tree and then summing in at the appropriate magnitude .
this is reminiscent of the fast multipole algorithm @xcite .
this would permit the method to use a single large @xmath29 rather than a piecewise approach , by moving the complexity into operations on a single number rather than by performing multiple ffts with simple floating - point numbers .
the basic motivation of the second approach described _ i.e. _ , using the _ sequence _ of @xmath29-norms ( each computed via fft ) to estimate the maximum value generalizes the @xmath29-norm fast convolution numerical approach into an interesting theoretical problem in its own right : given an oracle that delivers a small number of norms ( the number of norms retrieved must be @xmath382 to significantly outperform the naive quadratic approach ) about each vector @xmath21 , amalgamate these norms in an efficient manner to estimate the maximum value in each @xmath21 .
this method may be applicable to other problems , such as databases where the maximum values of some combinatorial operation ( in this case the _ maximum a posteriori _ distribution of the sum of two random variables @xmath383 ) is desired but where caching all possible queries and their maxima would be time or space prohibitive . in a manner reminiscent of how we employ fft , it may be possible to retrieve moments of the result of some combinatoric combination between distributions on the fly , and then use these moments to approximate true maximum ( or , in general , other sought quantities describing the distribution of interest ) . in practice
, the worst - case relative error of our quadratic approximation is quite low .
for example , when @xmath384 is stable , then the relative error is less than @xmath385 , regardless of the lengths of the vectors being max - convolved .
in contrast , the worst - case relative error using the original piecewise method would be @xmath386 , where @xmath28 is the length of the max - convolution result ( when @xmath387 , the relative error of the original piecewise method would be @xmath388 ) . of course
, the use of the null space projection method is predicated on the existence of at least four sequential @xmath29 points , but it would be possible to use finer spacing between @xmath29 values ( _ e.g. _ , @xmath389 to guarantee that this will essentially be the case as long as fft ( _ i.e. _ @xmath120 ) is stable . but more generally , the problem of estimating extrema from @xmath29-norms ( or , equivalently , from the @xmath29-th roots of the @xmath29-th moments of a distribution with bounded support ) , will undoubtedly permit many more possible approaches that we have not yet considered .
one that would be compelling is to relate the fourier transform of the sequential moments to the maximum value in the distribution ; such an approach could permit all stable @xmath29 at any index @xmath22 to be used to efficiently approximate the maximum value ( by computing the fft of the sequence of norms ) .
such new adaptations of the method could permit low worst - case error without any noticable runtime increase .
the fast numerical piecewise method for max - convolution ( and the affine piecewise modification ) are both applicable to matrices as well as vectors ( and , most generally , to tensors of any dimension ) .
this is because the @xmath29-norm ( as well as the derived error bounds as an approximation of the chebyshev norm ) can likewise approximate the maximum element in the tensor @xmath390 generated to find the max - convolution result at index @xmath391 of a multidimensional problem , because the sum @xmath392 computed by convolution corresponds to the frobenius norm ( _ i.e. _ the `` entrywise norm '' ) of the tensor , and after taking the result of the sum to the power @xmath274 , will converge to the maximum value in the tensor ( if @xmath29 is large enough ) .
this means that the fast numerical approximation , including the affine piecewise modification , can be used without modification by invoking standard multidimensional convolution ( _ i.e. _ , @xmath32 ) .
matrix ( and , in general , tensor ) convolution is likewise possible for any dimension via the row - column algorithm , which transforms the fft of a matrix into sequential ffts on each row and column .
the accompanying python code demonstrates the fast numerical max - convolution method on matrices , and the code can be run on tensors of any dimension ( without requiring any modification ) .
the speedup of fft tensor convolution ( relative to naive convolution ) becomes considerably higher as the dimension of the tensors increases ; for this reason , the speedup of fast numerical max - convolution becomes even more pronounced as the dimension increases .
for a tensor of dimension @xmath393 and width @xmath28 ( _ i.e. _ , where the index bounds of every dimension are @xmath394 ) , the cost of naive max - convolution will be in @xmath395 , whereas the cost of numerical max - convolution is @xmath396 ( ignoring the @xmath397 multiplier ) , meaning that there is an @xmath398 speedup from the numerical approach .
examples of such tensor problems include graph theory , where adjacency matrix representation can be used to describe respective distances between nodes in a network . as a concrete example
, the demonstration python code computes the max - convolution between two @xmath399 matrices .
the naive method required @xmath400 seconds , but the numerical result with the original piecewise method was computed in @xmath401 seconds ( yielding a maximum absolute error of @xmath402 and a maximum relative error of @xmath403 ) and the numerical result with the null space projection method was computed in @xmath404 seconds ( using @xmath360 , which corresponds to a relative error of @xmath405 in the top contour , yielding a maximum absolute error of @xmath406 and a maximum relative error of @xmath407 ) and in @xmath408 seconds ( using @xmath409 , which corresponds to a relative error of @xmath410 in the top contour , yielding a maximum absolute error of @xmath411 and a maximum relative error of @xmath412 ) . not only does the speedup of the proposed methods relative to naive max - convolution increase significantly as the dimension of the tensor is increased , no other faster - than - naive algorithms exist for max - convolution of matrices or tensors .
multidimensional max - convolution can likewise be applied to hidden markov models with additive transitions over multidimensional variables ( _ e.g. _ , allowing the latent variable to be a two - dimensional joint distribution of american and german unemployment with a two - dimensional joint transition probability ) .
the same @xmath29-norm approximation can also be applied to the problem of max - deconvolution ( _ i.e. _ , solving @xmath413 for @xmath26 when given @xmath19 and @xmath25 ) .
this can be accomplished by computing the ratio of @xmath414 to @xmath415 ( assuming @xmath25 has already been properly zero - padded ) , and then computing the inverse fft of the result to approximate @xmath416 ; however , it should be noted that deconvolution methods are typically less stable than the corresponding convolution methods , computing a ratio is less stable than computing a product ( particularly when the denominator is close to zero ) .
although the largest absolute error of the affine piecewise method is the same as the largest absolute error of the original piecewise method , the mean squared error ( mse ) of the affine piecewise method will be lower than the square of the worst - case absolute error . to achieve the worst - case absolute error for a given contour the affine correction must be negligible ; therefore , there must be two nearly vertical points on the scatter plot of @xmath417 vs. @xmath418 , which are both extremes of the bounding envelope from * figure [ figure : approxexactlinearzoom]*. thus , there must exist two different indices @xmath136 and @xmath137 with vectors where @xmath419 and where @xmath420 ( creating two vertical points on the scatter plot , and forcing that both can not simultaneously be corrected by a single affine mapping ) . in order to do this , it is required to have @xmath421 filled with a single nonzero value and for the remaining elements of @xmath421 to equal zero .
conversely , @xmath422 must be filled entirely with large , nonzero values ( the largest values possible that would still use the same contour @xmath29 ) . together , these two arguments place strong constraints on the vectors @xmath76 and @xmath77 ( and transitively , also constrains the unscaled vectors @xmath25 and @xmath26 ) : on one hand , filling @xmath421 with @xmath423 zeros requires that @xmath423 elements from either @xmath25 or @xmath26 must be zero ( because at least one factor must be zero to achieve a product of zero ) . on the other hand , filling @xmath422 with all large - value nonzeros requires that @xmath424 elements of _ both _ @xmath25 and @xmath26 are nonzero .
together , these requirements stipulate that both @xmath425 , because entries of @xmath25 and @xmath26 can not simultaneously be zero and nonzero . therefore , in order to have many such vertical points , constrains the lengths of the @xmath426 vectors corresponding to those points .
while the worst - case absolute error bound presumes that an individual vector @xmath21 may have length @xmath28 , this will not be possible for many vectors corresponding to vertical points on the scatter plot .
for this reason , the mse will be significantly lower than the square of the worst - case absolute error , because making a high affine - corrected absolute error on one index necessitates that the absolute errors at another index can not be the worst - case absolute error ( if the sizes of @xmath25 and @xmath26 are fixed ) .
code for exact max - convolution and the fast numerical method ( which includes both @xmath427 and null space projection methods ) is implemented in python and available at https://bitbucket.org/orserang/fast-numerical-max-convolution .
all included code works for numpy arrays of any dimension , _
i.e. _ tensors ) .
we would like to thank mattias frnberg , knut reinert , and oliver kohlbacher for the interesting discussions and suggestions .
j.p . acknowledges funding from bmbf ( center for integrative bioinformatics , grant no . 031a367 ) .
o.s . acknowledges generous start - up funds from freie universitt berlin and the leibniz - institute for freshwater ecology and inland fisheries .
david bremner , timothy m chan , erik d demaine , jeff erickson , ferran hurtado , john iacono , stefan langerman , and perouz taslakian .
necklaces , convolutions , and x+ y. in _ algorithms esa 2006 _ , pages 160171 .
springer , 2006 . | max - convolution is an important problem closely resembling standard convolution ; as such , max - convolution occurs frequently across many fields . here
we extend the method with fastest known worst - case runtime , which can be applied to nonnegative vectors by numerically approximating the chebyshev norm @xmath0 , and use this approach to derive two numerically stable methods based on the idea of computing @xmath1-norms via fast convolution : the first method proposed , with runtime in @xmath2 ( which is less than @xmath3 for any vectors that can be practically realized ) , uses the @xmath1-norm as a direct approximation of the chebyshev norm .
the second approach proposed , with runtime in @xmath4 ( although in practice both perform similarly ) , uses a novel null space projection method , which extracts information from a sequence of @xmath1-norms to estimate the maximum value in the vector ( this is equivalent to querying a small number of moments from a distribution of bounded support in order to estimate the maximum ) .
the @xmath1-norm approaches are compared to one another and are shown to compute an approximation of the viterbi path in a hidden markov model where the transition matrix is a toeplitz matrix ; the runtime of approximating the viterbi path is thus reduced from @xmath5 steps to @xmath6 steps in practice , and is demonstrated by inferring the u.s .
unemployment rate from the s&p 500 stock index . |
billiards on surfaces of negative curvature were first investigated by j. hadamard @xcite . the case of constant negative curvature , known as the pseudosphere , has enjoyed considerable popularity since gutzwiller s realization @xcite of its potential as a paradigm of quantum chaos .
complete hyperbolicity , the availability of symbolic dynamics , the equality of the lyapunov exponents of all periodic orbits , and the validity of selberg s trace formula are among the attractive features of that system .
useful introductions can be found in refs .
@xcite . for a tesselation by
maximally desymmetrized octagons ( see below ) aurich and steiner found the spectral fluctuations of the quantum energy spectrum faithful to the gaussian orthogonal ensemble ( goe ) of random - matrix theory @xcite , as illustrated in fig . 1 for the so - called form factor , the fourier transform of the energy dependent two - point correlator of the density of levels .
one of the urgent problems in quantum chaology is to understand the rather universal validity of random - matrix type spectral fluctuations for chaotic dynamical systems , nowadays known as the bohigas - giannoni - schmit conjecture @xcite .
an important first step was made by berry @xcite on the basis of gutzwiller s trace formula @xcite which expresses the oscillatory part of the density of energy levels as a sum over periodic orbits , @xmath0 , with @xmath1 the action and @xmath2 a classical stability amplitude of the periodic orbit @xmath3 .
berry realized that the double sum over periodic orbits in the form factor , which involves the building block @xmath4 in each summand , must draw an important contribution from the diagonal terms @xmath5 since pairs of orbits with action differences larger than planck s unit @xmath6 can be expected to interfere destructively and thus to cancel in the form factor . for time reversal
invariant systems each orbit @xmath3 and its time reverse @xmath7 have equal action such that berry s `` diagonal approximation '' generalizes to include pairs of mutually time reversed orbits and then gives the time dependent form factor as @xmath8 where @xmath9 is the time in units of the so called heisenberg time , which is given in terms of the mean level density @xmath10 as @xmath11 .
recently , sieber and richter @xcite employed the pseudosphere in their pioneering move beyond the diagonal approximation ; they found a one - parameter family of orbit pairs within which the action difference can be steered to zero .
one orbit within each `` sieber - richter pair '' undergoes a small - angle self - crossing which the partner orbit narrowly avoids .
the form factor receives the contribution @xmath12 from the family .
together with berry s @xmath13 we thus have a semiclassical understanding of at least the first two terms in the taylor series of the random - matrix form factor @xcite @xmath14 .
the search is now on for further families of orbit pairs which might yield the higher - order terms of the expansion . at the same time , one would like and will eventually have to go beyond the pseudosphere , in order to find the conditions for universal behavior ; interesting first steps have been taken for quantum graphs @xcite and other billiards @xcite . in the present paper we remain with the pseudosphere for a thorough investigation of action correlations which we feel necessary as a basis for further progress towards an understanding of the random - matrix like spectral fluctuations in this prototypical dynamical system ( and beyond ) .
we shall make extensive use of symbolic dynamics in ( i ) establishing a certain local character of the relation between an orbit and its symbol sequence ( only a section of the symbol sequence is necessary to determine the associated section of the orbit ) , ( ii ) constructing very long periodic orbits ( up to hundreds of thousands of symbols ) with full control of accuracy , ( iii ) identifying the ergodic ones among them and establishing a simple and instructive rederivation of huber s exponential - proliferation law , ( iv ) expressing the angle @xmath15 of a self - crossing and the closest - approach distance @xmath16 in the partner orbit of a sieber - richter pair in terms of the mbius - transformation matrices associated with the loops joined in a crossing , ( v ) revealing a useful analytic - continuation kinship of @xmath15 and @xmath16 , ( vi ) explicitly relating the joint density @xmath17 for crossing angles and loop lengths @xmath18 in orbits of total length @xmath19 to the associated density @xmath20 for avoided crossings , and ( vii ) constructing an expression for the action difference of a sieber - richter pair valid to all orders in @xmath15 .
some of our findings are based on ( overwhelming and exceptionless ) numerical evidence based on large numbers of long periodic orbits and thus call for mathematical substantiation . even though we strictly confine ourselves to the pseudosphere we expect ( and
have indeed begun to check ) generalizability to other systems for which symbolic dynamics is available .
the hadamard - gutzwiller model @xcite is a two - dimensional billiard on the so called poincar disc , i.e. the unit disc @xmath21 endowed with the metric @xmath22 the distance @xmath23 between two points @xmath24 , measured along the unique geodesic connecting them , reads @xmath25 the geodesics are circles intersecting the boundary at right angles .
the total length of the geodesics between its two crossing points with the disc boundary is infinite signaling non - compactness .
free motion on that space is completely hyperbolic .
all trajectories ( geodesics ) have the same lyapunov exponent , @xmath26 , and that fact makes for great simplifications as compared to usual billiards in flat space .
the poincar disc can be regarded as the stereographic projection of the pseudosphere , the surface of constant negative curvature .
the symmetry of the poincar disc is described by the non - compact lorentz group @xmath27 ; we shall only be interested in its hyperbolic elements which are matrices with the structure @xmath28 inner points of the poincar disc are transformed to inner points , and boundary points @xmath29 to boundary points . two boundary points @xmath30 remain invariant , @xmath31 and similar for @xmath32 ; the indices signaling `` stable '' and `` unstable '' . using @xmath33
we find @xmath34 where @xmath35 refers to the @xmath36 sign , and @xmath37 to the @xmath38 sign in the case @xmath39 ; if @xmath40 the opposite sign assignment applies .
repeated application of the transformation @xmath41 on any point @xmath42 leads to the stable fixed point , @xmath43 for @xmath44 , while iteration of the inverse map leads to the unstable fixed point , @xmath45 .
the geodesic passing through @xmath46 and @xmath32 ( to be called the `` own '' geodesic of the matrix @xmath47 ) is invariant with respect to @xmath41 : if @xmath48 belongs to this geodesic then so does @xmath49 .
the distance @xmath50 is the same for all points of the geodesic and equals @xmath51 the inverse @xmath52 of a matrix @xmath41 is obtained by the replacement @xmath53 in ( [ su11matrix ] ) .
these two matrices have their stable and unstable points interchanged .
consequently , a matrix and its inverse shift points along their common `` own '' geodesic by the same distance but in opposite directions . to obtain a model with compact configuration space
the poincar disc is tesselated with tiles of equal area and shape .
each type of tesselation is connected with a particular discrete infinite subgroup @xmath54 of @xmath27 such that by acting on all inner points of a tile by some fixed matrix @xmath55 we get all inner points of some other tile ; the boundaries are mapped to boundaries of the same or other tiles .
using all @xmath56 we get all tiles and restore the whole disc from a single initial tile . a particular tile is distinguished by containing the origin @xmath57 and is called the fundamental domain .
in fact , different tiles are identified by identifying each point @xmath42 of the poincar disc with all its images @xmath58 , and so a compact configuration space is indeed arrived at . in the present paper we exclusively work with octagonal tiles which are all identified with the fundamental domain .
the opposite sides of the latter are glued together such that we arrive at a riemann surface of genus 2 , i.e. a two - hole doughnut .
matrices @xmath59 giving rise to octagonal tiles are arbitrary products of four elementary group elements @xmath60 and their inverses . for simplicity
, we shall mostly consider tesselation with `` regular octagons '' ( fig . [ nozwei ] ) which have the highest possible symmetry ; the pertinent elementary matrices @xmath61 are @xmath62 and their inverses .
the inverse of @xmath63 is easily checked to be @xmath64 . since the index @xmath65 describes a phase in the off - diagonal matrix elements we conclude @xmath66 where we have introduced @xmath67 .
the full set of elementary matrices is thus @xmath68 .
the sides of the boundary of the fundamental domain can now be labeled @xmath69 and constructed as follows .
tesselation identifies opposite sides of the regular octagon as 0 @xmath70 4 , 1 @xmath70 5 , 2 @xmath70 6 , and 3 @xmath70 7 .
points on side @xmath71 are thus images of their mirror symmetric counterparts on side @xmath65 ; the group element responsible for such imaging is @xmath64 . in particular , the points on side 0 solve the quadratic equation @xmath72 and form a circle of radius @xmath73 , which is in fact a geodesic .
side @xmath65 is obtained by a rotation of side 0 by @xmath74 .
= 0.35 elementary matrices other than @xmath64 lead from points on side @xmath65 of the fundamental domain to boundary sides in the next generation of tiles , which may be called `` higher brillouin zones '' , using an analogy with periodic lattices . obviously , there are eight brillouin zones neighboring to the fundamental domain . in all other brillouin zones of all generations ,
opposite sides are still identified .
all brillouin zones have the same octagonal shape and the same area as the fundamental domain ; their visual difference in size and form is caused by the metric ( [ metric ] ) . in
what follows , we shall often call the elementary matrices @xmath63 `` letters '' , and their products @xmath75 `` words '' . as a convenient shorthand for a word we shall also employ
just the string of indices of the letters , i.e. @xmath76 . = 0.3 as already mentioned , the identification of points @xmath77 implies gluing together opposite sides of the octagon ; the result is a riemann surface of genus @xmath78 . on that surface ,
the eight corner points of the octagon in the unit disc coincide .
the identification of the corner points is shown in fig .
starting from any corner , ( for instance @xmath79 in the figure ) , taking into account that elementary matrices map opposite points of the boundaries of the fundamental domain onto each other , we go through all corners and come back after eight steps , and read off @xmath80 . due to the fact that the fixed - point equation @xmath81 for @xmath82 has unimodular solutions @xmath83
, we conclude that the matrix product @xmath84 must be the identity , since the corner point @xmath79 obviously fulfills @xmath85 .
in fact , using the explicit form ( [ elem ] ) of the @xmath63 , one checks that the foregoing product of eight matrices is the @xmath86 unit matrix , @xmath87 any geodesic in the tesselated unit disc can be folded into the fundamental domain where it will look like a sequence of disjoint circular segments starting and ending on the boundaries of the octagon .
of course , when the fundamental tile is regarded as a surface of genus 2 the circular sections in question no longer appear as disjoint : a trajectory leaving the fundamental domain through one of the eight sides of the octagon and reentering on the opposite side appears as a smooth curve on the genus 2 surface .
it is only after representing the whole unit disc by a single tile with opposite sides glued together ( or , equivalently by a surface of genus 2 ) that a geodesic is capable of self - crossings .
consider now inertial motion along the `` own '' geodesic of a matrix @xmath88 .
any point @xmath42 of the geodesic and its image @xmath89 ( which we know to belong to the same geodesic ) are identified by the tesselation .
geodesically moving from @xmath42 to @xmath58 we are in fact traversing a periodic orbit associated with @xmath59 , and the length of that periodic orbit is given by ( [ 1.1 ] ) ; the time reversed orbit is similarly associated with the matrix @xmath90 .
since each @xmath88 can be written as a product of elementary matrices @xmath91 and encoded by the symbolic word @xmath92 we have in fact symbolic dynamics at our disposal as a tool for investigating of periodic orbits . for words in the sense of symbol sequences and @xmath93 for the associated mbius - transformation matrices . ]
all matrices from an equivalence class @xmath94 , where @xmath95 is any matrix from @xmath96 , have their `` own '' geodesics identified by tesselation ; there is thus one and one only periodic orbit per equivalence class in @xmath54 . an infinite number of words pertaining to the same equivalence class and thus referring to the same periodic orbit can be obtained from one representative word @xmath97 ; this is done by cyclic permutations of the letters , by similarity transforms ( replacing @xmath59 by the longer word @xmath98 where @xmath95 is another word , and its associate @xmath99 stands for the matrix inverse to the one represented by @xmath100 ; the superscript `` tr '' is read as `` time reversed '' ) , and by using the group identity ( [ ide ] ) in all its various forms .
a code of the time reversed periodic orbit is produced from the original @xmath59 by writing its symbols @xmath101 in the opposite order and replacing each @xmath101 by @xmath102 . like any geodesic
the periodic orbit can be folded into the fundamental domain where it will consist of a finite number @xmath103 of disjoined circular segments .
after that a distinguished @xmath103-letter word can be introduced for the orbit which is simply the sequence of the `` landing '' sides of the orbit segments .
( the `` launching '' side is always opposite to the `` landing '' side of the preceding segment and is not admitted to the symbolic code . )
this encoding contains the least possible number of symbols among all members of the equivalence class , and is unique up to cyclic permutations .
all @xmath103 `` own '' geodesics of a matrix @xmath59 encoded by the distinguished word and its cyclic permutations , cross the fundamental domain .
= 0.25 as an example of how to construct an orbit from its symbol sequence we consider the word @xmath104 .
its `` own '' geodesic can be found calculating the matrix product @xmath105 and geodesically joining the respective fixed points ( eq.([fp ] ) ) , the curve @xmath106 in fig .
the full orbit within the fundamental domain could be obtained by folding the geodesic @xmath106 back into that domain .
however , it is easier to find the``own '' geodesic of the cyclicly permuted matrix @xmath107 and joining its fixed points @xmath108 .
the orbit in the fundamental domain is now given by the `` inner parts '' of the two geodesics ( bold intervals of @xmath106 and @xmath109 in fig .
the `` non - primed '' segment of the orbit starts on side 2 and ends on side 3 , while the `` primed '' one starts on 7 and ends on 6 .
the time reversed orbit would have the launching and landing sides interchanged such that its word would be @xmath110 .
tesselation with less symmetric octagons , also corresponding to a genus 2 riemann surface , can be implemented with four elementary matrices @xmath111 and their inverses @xmath112 , chosen such that all eight matrices obey the group identity ( [ ide ] ) .
interestingly , completely desymmetrized genus 2 surfaces do not exist : the inversion symmetry under @xmath113 can only be destroyed for @xmath114 @xcite .
we here propose a new method for constructing very long periodic orbits .
let us consider a many - letter word @xmath115 and its matrix @xmath116 . in order to explicitly construct the associated periodic orbit we can proceed stepwise , so as to determine each of the @xmath103 circular segments separately .
each step uses only part of the word @xmath59 ( the symbol assigned to a segment and its near neighbors ) ; the necessary length of that part is determined by the required accuracy and not at all by the length of @xmath59 . in the preceding section
, we discussed the connection between the fixed points @xmath117 of the mbius transformation associated with @xmath118 and the periodic orbit .
we would also like to recall that the two fixed points determine the circular segment associated with the first letter @xmath119 of the word ; to find the next circular segment one has to determine the two fixed points for the equivalent word obtained by cyclically permuting @xmath119 to the right end .
this is how all @xmath103 circular segments of the orbit in question can be found , one after the other .
it is convenient to first determine the stable fixed point for the first circular segment . to that end
, we consider the sequence of matrices @xmath120 which are obtained by truncating @xmath59 .
for each matrix in the sequence we solve the quadratic fixed - point equation .
the sequence of unstable fixed points thus obtained behaves erratically .
the sequence of stable fixed points , however , converges rapidly , in fact with exponentially growing accuracy .
the limiting point itself is none other than the stable fixed point @xmath46 of the whole word beginning with @xmath119 .
the convergence of the series of stable fixed points of @xmath121 can be ascertained analytically . to that end
we first observe that the matrix @xmath121 has its elements grow exponentially with the length @xmath19 of its corresponding periodic orbit .
this is obvious from the relation ( [ 1.1 ] ) of the length of an orbit to the trace of its matrix , @xmath122 , and from the unimodularity of the determinant , @xmath123 .
while under other circumstances such exponential growth gives rise to inaccuracy growing out of control , we can now rejoice in the growth working in our favor when determining the stable fixed point of @xmath124 : for large matrix elements , eq .
( [ fp ] ) simplifies to @xmath125 when proceeding to @xmath126 by including the letter @xmath127 and comparing the stable fixed points @xmath128 of @xmath129 , we find for their difference , using the approximation ( [ qwq ] ) and the unimodularity of @xmath130 , @xmath131 since the elements @xmath132 are of order unity , the difference @xmath133 is of order @xmath134 , i.e. small once the word @xmath121 is long .
exponentially fast convergence of the sequence @xmath135 is thus obvious .
we can now turn to the task of determining the unstable fixed point of @xmath136 .
as explained in the last section , the time reversed orbit is obtained using the reversed mbius transformation matrix @xmath137 , which has stable and unstable fixed points interchanged relative to @xmath59 .
consequently , we get the unstable fixed point of the circular segment of @xmath59 associated with the letter @xmath119 as the limit of the sequence of stable fixed points of the matrices @xmath138 ; these matrices are truncations of @xmath139 .
the sequence of stable fixed points of these truncations converges as rapidly as the one considered before and yields the unstable fixed point @xmath32 for the first circular segment of @xmath59 . with both @xmath32 and @xmath46 determined to the desired accuracy for the segment of the orbit corresponding to @xmath119 , one repeats the procedure for the subsequent circular segment , i.e. the one corresponding to @xmath140 . as an obvious property of our method
, we find that the computational effort needed to construct a periodic orbit with @xmath141 symbols grows only linearly with @xmath103 .
more importantly , we only have to do local work for local properties : each of the @xmath103 circular segments of the orbit inside the fundamental domain is found with a number of operations independent of @xmath103 ; for each segment one has to calculate a product of a moderate number of elementary matrices ( about 18 for a 15-digit accuracy ) .
as we proceed segmentwise , no inaccuracy can accumulate .
there is thus practically no limit on the length of the accessible orbits .
for example , we had no difficulty in calculating a periodic orbit corresponding to a sequence of @xmath142 randomly selected symbols
. it may be useful to again comment on the fact that any periodic orbit is , in principle , determined by any one of its circular segments within the fundamental domain .
one just has to take the full geodesic running between the fixed points @xmath143 , to which the segment belongs , and fold that geodesi c into the fundamental domain .
however , such folding is numerically unstable and can not be implemented with finite precision for very long orbits .
inasmuch as we work with fixed points of mbius transformations , our method seems to be restricted to the hadamard - gutzwiller model .
however , underlying all technicalities is a locality in the relationship between orbits and symbolic words , and that locality does in fact make our method generalizable , as will be explained in a separate publication .
the locality of the word - orbit relationship makes our method a promising pruning tool .
pruning generally means recognizing and deleting symbolic words not corresponding to physically realizable orbits @xcite . in the hadamard - gutzwiller model every word @xmath59
does correspond to an allowed orbit . on the other hand ,
there are infinitely many equivalent words which must be counted only once whenever sums over orbits are to be taken , as for instance in selberg s trace formula .
the particular word we are interested in is simply the ordered list of `` landing sides '' of the regular octagon for the succession of circular segments of the orbit ( section [ hadamard ] ) .
the task of selecting this distinguished word can be regarded as a variant of pruning .
a first step is to discard words which can be shortened , considering that the distinguished representation contains the least possible number of symbols ; in particular , the word must not contain pairs of symbols @xmath144 with @xmath145 . the group identity ( [ ide ] )
is yet another source of `` badness '' : whenever it allows to replace a word by an equivalent shorter one , only the latter needs to be further scrutinized .
real problems are due to the fact that the group identity ( [ ide ] ) allows to write some four - letter parts of words in reversed order , e.g. @xmath146 .
the average number of such reversible 4-sequences in a word with @xmath103 random symbols is estimated in appendix a as @xmath147 .
each of the 4-sequences must have a uniquely defined direction in the distinguished word ; it is not known _ a priori _ , however , which direction is the `` correct '' one .
the suitable word has thus to be selected among @xmath148 candidates .
for , say , @xmath149 no hope can be set on any trial - and - error procedure like building the orbits corresponding to the equivalent words one by one and discarding those not lying inside the fundamental domain .
with our algorithm we had no problem finding the correct direction of the @xmath150-letter sequences just mentioned .
whenever at some stage our method produced an arc which did not cross the fundamental domain , the letter for that arc turned up within a reversible @xmath150-sequence . upon reversing that particular @xmath150-sequence ( out of the thousands present in the code ! ) the arc always returned to the fundamental domain , and construction of the periodic orbit could be continued .
pruning thus becomes a local problem , which can be attacked efficiently .
it is well known that the geodesic flow in the hadamard - gutzwiller model is ergodic @xcite such that almost all trajectories cover the phase space homogeneously . introducing limited phase - space resolution
we can extend the notion of ergodicity to periodic orbits : almost all sufficiently long periodic orbits cover the coarse - grained phase space uniformly .
the overwhelming prevalence of ergodic periodic orbits does not imply that a _ random _ sequence of _ uncorrelated _ symbols must yield an ergodic orbit .
in fact , practically every periodic orbit so constructed is extremely non - ergodic , as illustrated in fig .
[ cake](a ) by the configuration - space density for 100000 randomly chosen symbols ; the grossly non - uniform density thus incurred is anything but ergodic .
= 0.95 in configuration space , a long periodic orbit intersects itself many times , and the distribution of self - crossing angles @xmath15 provides another sensitive test for ergodicity
. we shall consider it in some detail because of its role in the sieber - richter theory .
the angle @xmath151 , defined to lie in the interval @xmath152 , is complementary to the angle between the velocities at the crossing ; see fig .
[ schnee ] .
the number of self - crossings with crossing angles in the interval @xmath153 in periodic orbits of length @xmath19 yields a density @xmath154 . since no direction is distinguished the probability that an element @xmath155 of the orbit intersects with the element @xmath156 is given by the geometric projection @xmath157 . comparing that prediction with the plot of @xmath154 , numerically obtained for our periodic orbit with 100000 randomly chosen symbols
, we encounter blatant disagreement , despite the enormous length of the orbit .
in particular , for small @xmath15 , the number @xmath154 of self - crossings decreases like @xmath158 rather than @xmath159 .
= 0.6 it is important to stress that the non - ergodic patterns in figs .
[ cake](a ) and [ schnee ] are not due to an unlucky choice of the periodic orbit by the random - number generator employed to pick symbols .
as we shall see presently , two precautions must be taken for our way towards ergodic periodic orbits .
first , we should fix the geometric length @xmath19 rather than the number @xmath103 of symbols and , second , wave good - bye to the assumption of uncorrelated symbols . within an ensemble of orbits of fixed number of symbols
@xmath103 the orbit length @xmath19 will fluctuate .
ascribing equal probability to all allowed symbol sequences and invoking , for large @xmath103 , the central limit theorem we have the distribution of @xmath19 as a gaussian , with mean and variance both proportional to @xmath103 , @xmath160 @xmath161 and @xmath162 are the mean length and variance per symbol .
the values of these quantities are system specific ; using an ensemble of 100000 periodic orbits with @xmath163 randomly selected symbols each for the regular octagon we numerically find @xmath164 the concentration of @xmath165 in the vicinity of its maximum becomes ever more pronounced as @xmath103 grows
. the mean length per symbol for a long periodic orbit obtained by throwing honest dice for its symbols will therefore almost certainly be very close to @xmath161 . of greater interest
is a different ensemble of periodic orbits , which has the length interval @xmath166 fixed rather than the number of symbols @xmath103 . in particular , it is the fixed - length ensemble that one has in mind when speaking about the overwhelming prevalence of ergodic orbits . to find the number of orbits @xmath167 in the length interval @xmath168
we need the number of different periodic orbits with @xmath103 symbols , @xmath169 where the effective number of symbols @xmath170 deviates only slightly from the naive estimate @xmath171 obtained by excluding @xmath144 patterns ; the slight deviation is due to the group identity ( [ ide ] ) ; the factor @xmath172 reflects the equivalence of cyclically permuted symbol sequences ; a detailed discussion of @xmath173 will be presented in appendix [ peffa ] .
we proceed to @xmath174 by summing over all possibilities for the length to lie in the interval @xmath168 as @xmath175 or , approximating the sum by an integral , @xmath176 in the interesting case of large @xmath19 the integral can be evaluated using the saddle - point approximation .
the saddle of the integrand lies at @xmath177 = 0.4 the orbit density is thus found to obey the familiar exponential proliferation law @xmath178 with the growth rate @xmath179 the latter rate must coincide with the topological entropy . for billiards on surfaces of constant negative curvature that entropy equals unity and ( [ huber ] )
becomes huber s law @xcite . with our approximate numerical values ( [ numval ] )
inserted , our expression ( [ etaeq1 ] ) yields @xmath180 , in nice agreement with the value @xmath181 required by huber s law .
the orbits described by the proliferation law ( [ huber ] ) are known to be ergodic in their overwhelming majority . since we have just seen this law to arise mostly from contributions of orbits with the number of symbols close to @xmath182 ,
it is clear that the ergodic orbits must have the mean length per symbol given by @xmath183 for the regular octagon , with its values for @xmath184 given in ( [ numval ] ) we have @xmath185 .
note that this latter value strongly deviates from the mean length per symbol @xmath186 obtained from randomly chosen sequences of @xmath103 symbols .
therefore , orbits generated using random sequences of @xmath103 symbols have an exceedingly small probability to be ergodic . to see this we must realize that the orbits constituting the maximum of the integrand in the distribution at fixed length ( [ fixedl ] ) ( known to be ergodic ) are not those forming the maximum of the gaussian distribution ( [ gauss ] ) at fixed number of symbols , but rather belong to its ultra - left wing ; the difference of the locations of the maxima of the two distributions in question is of order @xmath19 , while the widths are only of order @xmath187 .
the enormous difference between the two ensembles is due to the exponential growth of the total number of orbits with @xmath103 symbols , @xmath188 .
the orbits with the length per symbol close to @xmath189 are in extremely small proportion to all orbits with the same number of symbols .
however , they are dominant among the orbits with a given length since they have many more symbols .
the fact that the maximum of the integrand in ( [ fixedl ] ) is formed by ergodic orbits can be regarded as due to a theorem of bowen s @xcite .
the number of orbits within the ergodic subset ( characterized by @xmath189 ) grows in the limit @xmath190 as fast as the total number of orbits .
correlations must be accounted for in the symbol sequence if ergodic orbits are to be constructed .
we propose to establish such correlations with the help of geometric considerations .
the number @xmath191 of times an ergodic orbit intersects a line element of infinitesimal length @xmath192 , with the crossing angle in the interval @xmath193 , obeys @xmath194 indeed , @xmath195 and the distribution of self - crossing angles @xmath196 must have the same sinusoidal angle dependence ; for self - crossings the line element @xmath192 is a stretch of the periodic orbit itself rather than an element of some fixed line in the billiard .
the behavior in question is also similar to the `` cosine law '' of optics for the angular distribution of the intensity of light emitted diffusely by some surface element .
we now apply the rule ( [ dnphi ] ) to an element of one side of the regular octagon , say number @xmath197 .
the segment of an ergodic orbit launched from that element succeeds the orbit segment with the symbol @xmath150 . in search
is now the probability for the orbit segment launched from side 0 to land on side @xmath65 , i.e. the probability @xmath198 that a certain symbol @xmath199 follows the symbol @xmath150 in the symbolic code .
an orbit segment is uniquely determined by the position of its initial point on , and the crossing angle with , the `` launching side '' @xmath197 .
we thus find the transition probability by first integrating ( [ dnphi ] ) over the angle range spanned by side @xmath65 with respect to the launching element , and subsequently integrating along side @xmath197 .
[ uebergang ] explains the pertinent geometry .
= 0.25 for an irregular octagon the procedure just explained must be applied to all launching sides @xmath200 to get the probabilities for the symbol @xmath201 to be followed by @xmath202 . in the regular octagon
the transition probabilities depend only on the absolute value of @xmath203 which can be 1,2 or 3 ; they are to be compared with the value @xmath204 which would correspond to the random symbol sequence without correlation .
alternatively , symbol correlations can be read off non - periodic trajectories with random initial conditions .
both approaches lead to good agreement for the single - step transition probabilities @xmath205 , as documented in the following table , @xmath206 symbol sequences respecting the single - step correlations just explained yield periodic orbits exhibiting ergodicity to an impressive degree of approximation .
in particular , the average length per symbol of periodic orbits in the regular octagon decreases from 2.257 to 1.694 , i.e. a value rather close to the ergodic limit @xmath207 . to further faithfulness to ergodicity we must account for higher correlations in the symbol sequence .
the next step involves two - step transition probabilities @xmath208 , i.e. the conditional probabilities for a certain symbol to follow any given two - symbol sequence .
for the regular octagon two - step correlations are embodied in a set of 16 numbers which can be calculated upon suitably refining the geometric reasoning applied above to find the four transition probabilities comprising the single - step correlations . by imposing two - step correlations on the symbol sequence we brought the average length per symbol for periodic orbits down to 1.643
; the coverage of configuration space thus obtained is almost uniform , as required by ergodicity ; see fig .
[ cake](b ) .
it is worth noting that @xmath209 differ from the product of the single - step transition probabilities @xmath210 by up to 30% .
an interesting alternative way to account for all correlations of importance is to read off symbol sequences of the desired length from a non - periodic trajectory .
the symbolic word thus obtained can be used for generating an almost perfectly ergodic periodic orbit which will almost everywhere coincide , to a high precision , with the corresponding section of the non - periodic trajectory .
only a dozen or so segments of the periodic orbit , namely those corresponding to the beginning and the end of the word , will deviate from the non - periodic progenitor .
the ergodicity of the orbit thus produced will be close to perfection , as witnessed by the density distribution in fig .
( [ cake])(c ) .
moreover , the two trajectories under discussion will have very nearly the same length per symbol ; in fact , considering @xmath142 orbits with @xmath142 symbols each , generated from non - periodic trajectories , we got the average length per symbol as 1.614 .
the latter value differs slightly from the prediction @xmath211 based on ( [ dmax ] ) and the numerical data ( [ numval ] ) on the ensemble of the orbits with random symbol sequences ; we have not bothered to look for an explanation of that ( minute ) discrepancy . to finally illustrate the suitability of the mean length per symbol as an indicator of the fidelity of periodic orbits to ergodicity , fig .
[ eit ] depicts the distribution @xmath212 of the angle @xmath213 between landing side and arriving periodic orbit , for groups of orbits of different mean lengths @xmath214 per symbol .
such groups were constructed from an ensemble of periodic orbits pertaining to a large fixed number of random uncorrelated symbols .
the distribution of lengths @xmath19 within such an ensemble is the gaussian depicted in fig . [ saddle ] .
the approach to ergodicity with @xmath215 is clearly visible in fig .
[ eit ] = 0.55
we can now turn to the statistics of self - crossings and associated avoided crossings which is believed to be an important classical ingredient in the explanation of universality of fluctuations in quantum energy spectra of chaotic dynamics @xcite .
each crossing divides a periodic orbit of length @xmath19 into two loops whose lengths @xmath216 add up to @xmath19 . in what follows
, we shall focus on the shorter one and its length @xmath217 . for the statistical considerations to follow we employ the number of loops @xmath218 with lengths in the interval @xmath219 and crossing angles in the interval @xmath153 in ergodic orbits of length @xmath19 .
the density of loop lengths and crossing angles , @xmath220 , be calculated analytically for the hadamard - gutzwiller model due to the fact that each loop can be continuously deformed to a loop with crossing angle @xmath221 , that is , to a periodic orbit @xmath3 , as indicated in fig .
[ deformation ] . due to the equality of the lyapunov exponents for all trajectories in the hadamard - gutzwiller model the loop length @xmath222
is uniquely determined by the crossing angle @xmath15 and the length @xmath223 of the periodic orbit @xmath3 as @xcite @xmath224 for @xmath225 large that relation simplifies to @xmath226 we immediately infer that a loop is longer than its periodic - orbit deformation @xmath3 by an amount independent of @xmath3 and the longer the smaller the crossing angle ; even for the shortest orbits with the length @xmath227 ( the value refers to the regular octagon ) the error of the simplified relation ( [ smeps ] ) relative to ( [ 1.2 ] ) is less than 2% .
= 0.1 using the basic relation ( [ 1.2 ] ) , sieber expresses @xmath228 as a sum over periodic orbits @xcite , @xmath229 here @xmath230 is a repetition number equaling 1 if @xmath3 is a primitive periodic orbit , and @xmath231 is the area of the octagon . due to @xmath232 only orbits with @xmath233 contribute to the sum . as @xmath18 grows , more and more periodic orbits contribute and the sum will eventually be dominated by long orbits for which @xmath234 so that the denominator may be replaced by @xmath235 .
taking into account ( [ smeps ] ) and applying the sum rule of hannay and ozorio de almeida @xcite we obtain the ergodic result , @xmath236 where @xmath237 gives the length of the shortest possible loop created by deformation of the shortest periodic orbit of the system with length @xmath238 , again according to the basic relation ( [ 1.2 ] ) .
two reduced distributions can be obtained from the general expression ( [ dds ] ) .
integrating over the angle from zero to some @xmath239 , we obtain the length distribution @xmath240 for loops with crossing angles smaller than @xmath239 in the periodic orbit of length @xmath19 .
the resulting expression is astonishing simple , @xmath241 the only @xmath239-dependence lies in the summation condition @xmath242 and indicates that only those periodic orbits @xmath3 contribute to the sum which can be deformed to a loop of length @xmath243 smaller than the given length @xmath18 .
therefore , we obtain a staircase function for the quantity @xmath244 . in fig [ loop ] ( a ) , ( c ) , we display @xmath245 for @xmath246 and @xmath247 , whereas ( b ) displays the staircase obtained from ( a ) after multiplication with @xmath248 ; these plots reveal perfect agreement between the numerical results obtained from all loops in an ensemble of @xmath249 ergodic periodic orbits with the prediction of the periodic - orbit sum ( [ lolgh ] ) .
= 1.00 note that a gap arises in the length distribution latexmath:[$p(l orbit and the related shortest possible loop of angle @xmath15 and length @xmath251 .
the gap is given by @xmath252 .
in particular , in fig .
[ loop](a ) , @xmath253 is just the length of the smallest periodic orbit of the system , whereas in fig .
[ loop ] ( c ) , @xmath254 is the length of the shortest possible loop with angle @xmath255 . generally , in accord with the relation ( [ smeps ] ) between the lengths @xmath256 and @xmath225 , the peaks related to all periodic orbits experience practically the same shift to the right , @xmath257 ; therefore , when @xmath258 is decreased the length distribution appears to shift to the right rather rigidly .
when ( [ dds ] ) is integrated over all loop lengths @xmath18 , we obtain the density of crossing angles @xmath154 in ergodic periodic orbits of length @xmath19 as @xmath259 the term in the right hand side proportional to @xmath260 could be obtained immediately using the ergodic approximation ( [ ergo ] ) .
the actual number of crossings is smaller due to the already mentioned gap in the loop length distribution , such that the effective integration interval in ( [ lout ] ) is shorter than @xmath261 .
this is taken into account by the introduction of the effective minimal loop length @xmath262 ; that latter quantity is close to , but somewhat smaller than , the gap visible in fig [ loop](a ) , ( c ) , i.e. the length @xmath251 of the loops related to the shortest periodic orbit .
the discrepancy arises from the shortest periodic orbits whose contribution is greatly in excess of the ergodic prediction .
a periodic - orbit sum for @xmath262 is obtained by substituting the loop length density ( [ dds ] ) for the integrand in ( [ lout ] ) and comparing the left and right hand sides .
recalling that all loop lengths exceed the lengths of their associated periodic orbits by the same quantity ( see ( [ smeps ] ) ) and employing the sum rule , we can make the summation interval independent of @xmath15 , @xmath263 for long orbits ( large @xmath19 ) the last expression is @xmath19-independent .
indeed , the increment of @xmath264 , when @xmath19 is replaced by @xmath265 , is given by @xmath266 which follows from the proliferation law for periodic orbits .
the @xmath15 dependence of @xmath264 comes from the shortest and thus non - generic orbits ; it is so weak that for most purposes we can set @xmath267 and work with @xmath268 the constant @xmath269 was obtained by choosing @xmath270 and correspondingly allowing for periodic orbits with lengths @xmath271 in @xmath272 .
inasmuch as every periodic orbit with a small - angle self - intersections has a partner orbit which is almost identical save for avoiding the said crossing , one would expect the statistics of crossings and avoided crossings to be the same , at least in the limit of small angles . in the present subsection we shall confirm that expectation , mainly by showing that there is a formal analytic - continuation relationship between crossings and avoided crossings .
however , it is important to notice that two different types of avoided crossings arise which have different properties .
we generalize the considerations of the preceding subsection by imagining the division of a periodic orbit into two `` loops '' by an avoided crossing . as shown in fig .
[ geom ] , two geodesics , the intra - octagon segments of which belong to a periodic orbit , may either cross inside the unit disc once or not at all .
as before , we let @xmath15 be the angle complementary to the angle between the velocity vectors of the two geodesics at an intersection as defined in fig .
[ geom ] ( a ) .
when the direction of velocity in one of the two geodesics is reversed , the role of the angles @xmath273 is interchanged .
elementary euclidian geometry provides the relation between the radii @xmath274 of the two geodesic circles , the distance @xmath18 between its midpoints and the angle @xmath15 as @xmath275 where all quantities are measured in euclidian units .
the two solutions with @xmath276 correspond to the two complementary angles @xmath273 . in case
the two segments do not cross there exists a minimal distance @xmath16 between them .
that distance is measured using the metric ( [ metric ] ) , along the geodesic orthogonal to the said segments .
the two intersection points of this geodesic with the segments of the orbit divide the orbit into two loops , much the same as in the case of a crossing .
= 0.55 surprisingly , the distance @xmath16 is given by analytic continuation ( @xmath277 ) of equation ( [ huhu ] ) , @xmath278 where , in contrast to @xmath16 , the radii @xmath279 and the distance @xmath18 are still meant in the euclidian sense , as in ( [ huhu ] ) .
we do not want to pause and give the straightforward but dull proof of ( [ anal ] ) here but shall actually recover both ( [ huhu ] ) and its analytic continuation ( [ anal ] ) through symbolic dynamics in sect .
[ symbdyn ] .
the geodesic along which the width of an avoided crossing is measured may consist of several disjoint arcs within the fundamental domain ; we shall come back to this complication in sect .
[ partnership ] . in
what follows , it will be important to distinguish two different types of avoided crossings : we call `` antiparallel '' the avoided crossings with antiparallel velocities at the point of minimal distance as in fig .
[ geom ] ( b ) while avoided crossings with parallel velocities are called `` parallel ''
. we shall be interested in the respective distributions @xmath280 and @xmath281 of the loop length @xmath18 and the minimal distance @xmath16 , both meant for loops inside ergodic orbits of total length @xmath19 ; clearly , these densities correspond to the distribution of loop lengths and crossing angles considered in the preceding subsection .
= 1.0 when determining the density @xmath282 , we exploited the idea that every loop with length @xmath18 and crossing angle @xmath15 in a given orbit of length @xmath19 is related to one periodic orbit @xmath3 , of which it can be viewed as a deformation ; it is that orbit @xmath3 which fulfills @xmath283 . in this sense
, we have the association @xmath284 .
similarly , each loop defined by an antiparallel avoided crossing can be viewed as a deformation of one particular periodic orbit @xmath3 with length @xmath225 , and the same is true for parallel avoided crossings .
the analytic expressions for the relationships @xmath285 and @xmath286 appear as analytic continuations of ( [ 1.2 ] ) , in analogy to the above continuation of the simple trigonometric identity ( [ huhu ] ) for the crossing angle to ( [ anal ] ) for the closest - approach distance .
the precise meaning of this analytic continuation will be explained in the following section , see equation ( [ myfo ] ) .
here , we just mention that antiparallel avoided crossings correspond to solutions of eq .
( [ 1.2 ] ) with @xmath15 replaced by @xmath287 , and parallel avoided crossings to solutions with @xmath15 replaced by @xmath288 .
the lengths @xmath289 and @xmath290 of loops generated by avoided crossings as defined in fig .
[ ally](a ) can be related to the lengths of periodic orbits @xmath3 as @xmath291 of course , in general @xmath3 and @xmath292 are different periodic orbits . for small closest - approach distances @xmath16 , the foregoing expressions for the loop lengths are simplified to @xmath293 in part ( b ) of fig .
[ ally ] , we schematically depict loops and their lengths @xmath256 for small crossing angles ( @xmath294 ) , similarly in part ( c ) for @xmath289 with narrowly avoided antiparallel crossings ( @xmath295 ) , in part ( d ) for @xmath256 with large angle @xmath296 , and finally in part ( e ) for @xmath290 with narrowly avoided parallel crossing ( @xmath297 . in analogy to the preceding section
, we proceed to the marginal distributions @xmath298 defined as @xmath299 most interesting is the limiting case of narrowly avoided crossings . in fig .
[ joop ! ] ( a ) and ( b ) , the loop length distributions @xmath300 for crossings and @xmath301 for antiparallel avoided crossings are displayed with @xmath302 .
practically no difference can be observed between these two distributions , as predicted by @xmath303 in eq .
( [ limit ] ) for small argument ; we here encounter the expected statistical correlation between crossings and avoided crossings . on the other hand , for parallel avoided crossings , the distribution @xmath304 with @xmath305 looks dramatically different : as displayed in ( c ) , spikes exactly at the position of periodic orbits are observed , as predicted by @xmath306 in ( [ limit ] ) in the limit of narrowly avoided parallel crossings .
we here propose to reveal how self - intersections with small angles and narrowly avoided crossings are reflected in symbolic words .
that correspondence will lead us to some exact results for the hadamard - gutzwiller model , as well as a better understanding of the analytic - continuation relationship of crossings and avoided crossings .
the ideas to be presented may also become important for generalizations of the sieber - richter theory to other models such as billiards , quantum graphs and maps . in what follows , symbol sequences with brackets @xmath307 refer to periodic orbits , whereas sequences without brackets refer to parts thereof .
consider a crossing with angle @xmath308 dividing a periodic orbit into two loops .
the symbolic word of the orbit will also be divided by the crossing into two parts which refer to a loop each and will be denoted by @xmath309 and @xmath310 where @xmath311 are symbols from 0 to 7 ; see fig .
[ fuebl ] .
the word of the whole orbit will be @xmath312 , with the number of symbols @xmath313 and the length equal to the sum of the loop lengths , @xmath314 .
the word is assumed pruned ( see section iv ) .
we recall that the loops @xmath315 can be continuously deformed to periodic orbits @xmath316 , whose lengths @xmath317 related to the lengths of the loops by ( [ 1.2 ] ) , i.e. , as @xmath318 for @xmath319 ; we shall also have to relate the lengths of both orbits to the traces of the corresponding mbius - transformation matrices and the associated matrices @xmath320 according to ( [ 1.1 ] ) , i.e. @xmath321 . here @xmath322 are the @xmath323 matrices obtained by taking the product of elementary matrices ( [ elem ] ) , in the order given by the symbol sequences @xmath315 ; the foregoing expression for @xmath324 also holds for non - regular octagons , with the elementary matrices ( [ elem ] ) replaced by the pertinent generators .
= 0.45 summing up the loop lengths we get @xmath325 and , solving for @xmath326 @xmath327 except that the elementary derivation yields the first term in the numerator of the second member as @xmath328 .
we have , somewhat frivolously , dropped the modulus operation since numerical evidence suggests @xmath329 whenever the loops @xmath315 arise from a crossing ; we regret not having found a proof , all the more so since only the form conjectured in ( [ myfo ] ) allows for analytic continuation to the case of avoided crossings ( see below ) .
incidentally , the foregoing symbolic - dynamics expression for the crossing angle is the previously announced variant of the elementary - geometry result ( [ huhu ] ) . for long orbits and loops we may approximate ( [ myfo ] ) as @xmath330 inasmuch as a crossing
is connected with the loops @xmath315 , the right hand side of ( [ myfo ] ) must obey @xmath331 . as regards the inverse statement
, one must be a little careful .
there can be several divisions of the word @xmath332 into parts @xmath333 , @xmath334 leading to exactly the same value of @xmath335 .
in this situation , the orbit has one and only one crossing with the angle defined by ( [ myfo ] ) ( provided @xmath336 ) , and only one of the divisions mentioned actually describes the loops associated to the crossing . as an important application of the foregoing relation between crossing angles and symbolic words we can now find the condition on the word for the crossing angle to be small . as obvious from ( [ smaa ] )
we must have @xmath337 .
a crude estimate of these traces can be obtained by recalling that the length of an ergodic orbit with the ( pruned ! ) word @xmath332 is roughly proportional to the number @xmath338 of letters in the word , @xmath339 , where @xmath189 is the average length per symbol ( close to 1.6 in the regular octagon ; see section v ) . combining with ( [ 1.1 ] ) we find that the trace of the mbius transform @xmath340 grows exponentially with @xmath338 , @xmath341 if @xmath315 are by themselves pruned words we have for the traces in ( [ smaa ] ) @xmath342 with the trivial conclusion @xmath343 ; the crossing angle will not be small then .
suppose , however , that the matrices @xmath344 and @xmath345 possess the structure @xmath346 with @xmath347 generic mbius transforms while the `` insertions '' @xmath348 are products of @xmath349 elementary matrices .
obviously the insertions disappear from the traces of the matrices @xmath344 and @xmath350 since @xmath351 and similarly for @xmath350 ; they do not cancel in the trace of the product of @xmath352 , however . denoting @xmath353 we have from ( [ smaa ] ) @xmath354 and
can thus approximate the crossing angle as @xmath355 very long insertions @xmath348 will therefore yield very small crossing angles .
= 0.45 to understand the physics behind condition ( [ assummoeb ] ) we recall from section ii that inverting a mbius matrix means time reversal for the corresponding code ; and that the time reverse of the code @xmath356 is @xmath357 with @xmath358 . condition ( [ assummoeb ] ) for the mbius matrices therefore means @xmath359 the code of the total orbit can be written as @xmath360 , with @xmath361 .
the shorter periodic orbits associated with the loops will have the codes @xmath362 and @xmath363 . since cyclic permutation is allowed in the code of orbits ( not in loops ! ) , the insertions can be brought together to form the self - canceling sequences @xmath364 in @xmath365 and @xmath366 in @xmath367 . using the locality of the code - to - trajectory connection ( see section iii ) we may say , somewhat loosely , that the loop @xmath333 is obtained by smoothly joining the periodic orbit @xmath368 with two pieces of trajectory , one visiting the boundaries of the octagon according to @xmath369 , and the other , encoded by @xmath370 , running through the same sequence of boundaries in the opposite direction .
such is precisely the behavior of a loop with a small crossing angle : the sections in the vicinity of the crossing nearly coincide up to the sense of traversal .
the loop structure ( [ loopstruct ] ) seems in fact necessary to produce a small crossing angle .
indeed , the beginning and ending parts of a loop being `` nearly parallel '' near a small - angle crossing the corresponding symbol sequences can not help being locally identical , apart from time reversal .
impressive numerical evidence for the equivalence is provided by fig . [ noangles ] : depicted are the densities @xmath196 of crossing angles for a large generic set of symbol sequences and the subset purged of sequences of the indicated type ; in the latter case no crossings with small angles are present .
the symbolic dynamics of the avoided crossings is investigated in much the same way .
one has to replace the loop length ( [ 1.2 ] ) by the corresponding results ( [ anti ] ) and ( [ para ] ) .
that replacement suggests an interpretation as the analytical continuation @xmath371 and @xmath372 for antiparallel and parallel avoided crossings , respectively . the width of an avoided crossing dividing the orbit into the loops @xmath315 becomes @xmath373 where the approximate equalities refer to long orbits and loops , as in ( [ smaa ] ) .
it is obviously necessary that @xmath374 ( @xmath375 for the existence of an antiparallel ( resp .
parallel ) avoided crossing .
the width of the antiparallel avoided crossing will be small under the same assumptions ( [ loopstruct ] ) about the loops as in the case of small - angle crossings ; the code of the whole orbit in both cases has the structure @xmath376 such that the distinction can be made only after calculating and thus determining the sign of @xmath377 .
the sieber - richter theory is based on the fact that each orbit with a small crossing angle @xmath308 has a partner with almost the same length which avoids that crossing by the width @xmath378 ( see the next subsection for precise relations between the loop lengths , @xmath15 , and @xmath16 ) ; the partner has practically the same loops but one of them is traversed in the opposite sense relative to the original orbit .
we can now translate these assertions into the language of symbolic dynamics .
as we have just seen , a small - angle crossing divides the orbit into loops @xmath315 with the structure ( [ loopstruct ] ) ; consequently the partner must consist of the loops @xmath379 . accounting for long insertions near the crossing as after ( [ loopstruct ] )
we can state that each sieber - richter pair of orbits with small @xmath15 and @xmath16 must have codes @xmath376 and @xmath380 .
the crossing angle in one of the orbits and the closest - approach distance in its partner are defined by @xmath381 anticipating again @xmath382 we conclude @xmath383 or , with the help of @xmath384 , @xmath385 .
the approximate equality of the absolute values of the two traces is clearly necessary for the length of the two orbits in the pair to be close due to ( [ 1.1 ] ) ; the minus sign is less obvious .
even though we are mainly interested in pairs of orbits with small @xmath15 and @xmath16 it is interesting to further dwell on the crossing / avoided - crossing partnership for large angles and widths .
we shall employ symbolic dynamics to rigorously relate the crossing angle @xmath15 to the closest - approach distance @xmath16 for such pairs and give exact expressions for the action difference in a pair and the distribution @xmath386 .
such results might be of importance for future investigations of higher - order terms in the expansion of the form factor .
we again imagine the symbolic code of an orbit bisected as @xmath387 ; then the partner orbit with respect to that bisection will have the symbolic word @xmath388 ; the function @xmath389 defined in ( [ myfo ] ) will become @xmath390 for the partner . as already indicated previously , numerical checks revealed the following behavior of @xmath391 under this replacement : ( i ) if the bisection @xmath392 has @xmath393 the partner orbit @xmath394 will also have @xmath395 ; ( ii ) @xmath396 entails @xmath397 ; ( iii ) finally , if @xmath398 then , the partner orbit has @xmath399 .
if @xmath333 or @xmath334 contains less than @xmath400 symbols , and @xmath391 very close to 1 , very rare exceptions to this rule occur .
considering the previously established interpretations of the three cases ( [ aac ] ) , ( [ pac ] ) , and ( [ myfo ] ) we can say that partial time inversion of an orbit has the following respective consequences : ( i ) a parallel avoided crossing is replaced by another parallel avoided crossing ; ( ii ) an antiparallel avoided crossing becomes a crossing in the partner ; ( iii ) a crossing becomes an antiparallel avoided crossing in the partner .
for very short loops with at most @xmath400 symbols and angles close to @xmath221 very rare exceptions occur .
we can use the last of these findings to more precisely _ define _ an antiparallel avoided crossing as the object emerging after the partial time reversal of some orbit @xmath401 ( or , equivalently , @xmath402 ) provided the bisection @xmath392 has a crossing according to @xmath403 while @xmath399 .
as always , the word @xmath392 is assumed pruned in the sense of sect .
[ longorb ] , i.e. written in the standard representation such that each of its letters indicates a side of the fundamental domain visited by the orbit .
the code @xmath404 of the partner orbit with an avoided crossing will then , however , not always be pruned , and the partner orbit will sometimes not have all its circular sections within the fundamental domain .
an example of a bisection of a pruned word for which partial time reversal yields a non - pruned word is provided by @xmath405 .
then the symbolic word of the partner @xmath406 obviously contains mutually canceling parts , and @xmath407 may become the standard encoding of an orbit in the fundamental domain .
however , the avoided - crossing width in this case will be determined by the function @xmath408 in whose arguments the insertion @xmath95 may not be dropped .
this situation occurs when the geodesic determining the width of the avoided crossing consists of several segments when depicted inside the octagon ( see the preceding section ) .
there is no reason not to take into account avoided crossings of the type just characterized .
an orbit with an avoided crossing may thus contain fewer symbols and be appreciably shorter than its partner with a crossing .
the closer the crossing angle is to @xmath221 the more often one meets with that situation .
according to our definition there is a one - to - one correspondence between crossings and avoided crossings .
one might be tempted to infer that in each long ergodic orbit the number of crossings must be equal to the number of avoided crossings , but that temptation must be resisted . given a fixed length @xmath19 , there are fewer crossings than avoided crossings simply since by replacing a crossing with an avoided crossing one gets a shorter periodic orbit .
therefore , to obtain all avoided crossings with a certain width @xmath16 in the periodic orbits of length @xmath19 one has to make the crossing to avoided - crossing replacement in all orbits of length @xmath409 with suitable length increments @xmath410 , and longer orbits are more numerous due the exponential proliferation . to find the relation between the density @xmath411 of crossing angles and the density @xmath386 of the avoided - crossing widths we write the total number of crossings in all periodic orbits with lengths in the interval @xmath412 $ ] as @xmath413 with huber s proliferation law @xmath414 . on the other hand
, the number of avoided crossings with width @xmath16 in these orbits will equal the number of all crossings in the `` parent '' orbits whose lengths lie in @xmath415 $ ] which is @xmath416 .
we may thus write @xmath417 where the approximate - equality sign again signals large @xmath19 .
it remains to determine the length shift @xmath418 and the crossing angle @xmath419 , both as functions of @xmath16 .
clearly , if we only wanted to find the ergodic part of @xmath386 from @xmath420 we could , in analogy with our procedure in sect .
[ ergorb ] , drop the length shift @xmath410 everywhere in ( [ denaco ] ) except in the exponential proliferation factor ; actually , we must be more ambitious and go for the next - to - leading order in @xmath19 as well since the latter is of relevance for the form factor . to determine the functions @xmath421 and @xmath419 entering the foregoing expression we invoke two consequences of the definition ( [ su11matrix ] ) of @xmath27 matrices @xmath422 .
first , we note @xmath423 where @xmath424 is the unit matrix .
second , on multiplying with @xmath425 and taking the trace we get @xmath426 using the identity ( [ matrid ] ) we can rewrite the expressions ( [ myfo ] ) for the crossing angle and ( [ aac ] ) for the avoided - crossing width as @xmath427 and conclude @xmath428 herein employing ( [ 1.1 ] ) and denoting the lengths of the periodic orbits with crossing and avoided crossing by @xmath429 and @xmath430 we proceed to @xmath431 as by now familiar , slight simplifications arise in the physically interesting case of loops long enough for @xmath432 to be much larger than both @xmath433 and @xmath434 . the rigorous symbolic - dynamics expressions ( [ epsdel ] ) for @xmath15 and @xmath16 then simplify to @xmath435 and yield the previously announced relation between the crossing angle in the parent orbit and the avoided crossing width in the partner , @xmath436 which implies the familiar @xmath437 for small angles .
similarly , we may combine ( [ lengths ] ) and ( [ approco ] ) to @xmath438 and thus get the length ( and thus action ) difference for an orbit with a self - crossing and its partner avoiding that crossing , @xmath439 the latter relation generalizes the result of sieber and richter @xcite , @xmath440 , to all orders in @xmath15 .
an interesting geometric interpretation of our relations ( [ epsdeldel ] ) between @xmath441 , and @xmath410 in the framework of hyperbolic geometry will be presented in appendix [ pseudeuc ] .
it may be well to once more say that the approximate - equality sign in these relations points to an error which is exponentially small when the loop lengths are large .
we can now write the factors converting the density of crossing angles into the density of avoided - crossing widths in ( [ denaco ] ) , @xmath442 the latter identities yield @xmath443 such that the ergodic angle distribution @xmath444 yields the ergodic distribution of avoided - crossing widths @xmath445 . indulging in higher ambitions ,
we retain all terms of the next - to - leading ( i.e. first ) order w.r.t . to @xmath19
contributed by @xmath446 in ( [ denaco ] ) .
taking the full angle density ( [ lout],[leffprox ] ) @xmath447 from sect .
[ joint ] and using the above identities ( [ conversion ] ) to check @xmath448 we get @xmath449 as a most welcome surprise , after combining all the ostensibly unrelated factors in ( [ denaco ] ) , including those resulting from huber s exponential proliferation law , the final result is simply the analytical continuation of the crossing distribution to imaginary crossing angles . in the previous subsection
we have already commented on the symbolic codes for the periodic orbits in a sieber - richter pair as @xmath450 and @xmath451 . to further clarify the role of symbolic dynamics for crossings , avoided crossings , and sieber - richter pairs
we would like to present a concrete such pair in the regular octagon .
[ zmz2 ] ( a ) displays a numerically found periodic orbit with a small - angle crossing .
the reader is invited to read off the orbit segments , starting e.g. at the side 6 in the direction shown by arrows obtaining @xmath452 thus closing the loop .
the starting side is always opposite to the landing side of the preceding segment , with their numbers differing by @xmath453 .
hence it is sufficient to list the consecutive landing sides only : @xmath454 .
the sequence we got is the standard symbolic code of the orbit ; the underlined symbols indicate the pair of segments crossing at the point @xmath455 .
= 0.95 the crossing @xmath455 divides the orbit into two loops shown by the full and dashed lines in fig.[zmz2 ] ( a ) ; the symbol sequences of both loops @xmath456 and @xmath457 start with the symbol of the segment participating in the crossing .
the periodic orbits associated with the loops and the loops themselves are shown by thick and thin lines respectively in fig .
[ zmz2](b ) for loop @xmath333 and ( c ) for loop @xmath334 .
the periodic orbit @xmath458 associated with the loop @xmath333 has the same symbol sequence @xmath459 as the loop itself .
the code of the orbit associated with the loop @xmath334 is shorter than that of the loop , namely , @xmath460 while @xmath461 with @xmath462 ; of course , the insertions @xmath463 fall out of the code of the loop - associated orbit .
inspection of part of fig .
[ zmz2](c ) helps to appreciate insertions @xmath463 in a loop like @xmath464 as due to the deformation of a circular segment of the periodic orbit @xmath465 overstepping the boundary of the octagon .
speaking pictorially we may say that the middle of the upper segment of @xmath466 was dragged upwards such that the vertex of the loop crossed the boundary 2 of the octagon , reappeared at the opposite boundary 6 , and finally became the crossing point @xmath455 , hence the insertions @xmath462 and @xmath467 . on the other hand , when deforming segment 3 of @xmath468 to the loop @xmath333 ( fig .
[ zmz2](c ) ) the octagon boundary is never reached , and therefore no new symbols appeared in the loop @xmath333 compared with the periodic orbit @xmath469 . to obtain the sieber - richter partner of the orbit @xmath450 in fig .
[ zmz2 ] ( a ) we have to replace one of its loops , e.g. @xmath470 by its time reverse @xmath471 .
the partner periodic orbit thus formed @xmath472 is shown in fig [ cravcr ] ( a ) ( dotted line ) together with the original orbit ( full line ) .
the partner has an antiparallel avoided crossing whose width is approximately equal to the crossing angle in the original orbit ; the relation valid to all orders in @xmath15 is given by ( [ stefrel ] ) .
note that apart from the two segments participating in the crossing ( resp . avoided crossing ) the two orbits run very close together ; thus it is these two segments which contribute mainly to the difference in the orbit lengths .
= 0.44 financial support by the sonderforschungsbereich `` unordnung und groe fluktuationen '' of the deutsche forschungsgemeinschaft is gratefully acknowledged .
we have profited from discussions with uwe abresch , ralf aurich , predrag cvitanovi , gerhard knieper , christopher manderfeld , klaus richter , martin sieber , uzy smilansky , dominique spehner , frank steiner , and wen - ge wang .
consider a triangle @xmath473 formed by geodesics on the pseudosphere , with the angles @xmath474 named after the respective vertices and the lengths of the respective opposite sides @xmath475 .
the following generalizations of elementary euclidian - geometry relations hold true @xcite @xmath476 let us apply these identities to the triangle @xmath477 in fig .
[ cravcr ] ( b ) depicting the crossing and avoided crossing in a sieber - richter pair . again naming the angles of the triangle after the respective vertices we have
the point @xmath479 is supposed to be removed to infinity ( recall the assumption of long orbits ) such that the lengths @xmath480 and @xmath481 tend to infinity while the angle @xmath479 tends to zero .
however , the difference @xmath482 remains finite and gives , up to an obvious factor 4 , the length ( or action ) difference in search , @xmath483 . from the first of the relations ( [ geq2 ] ) we get the connection between the crossing angle @xmath15 and the closest - approach distance @xmath484 , @xmath485 which we recognize as ( [ stefrel ] ) ; similarly , the second group of relations in ( [ geq2 ] ) yields the length difference @xmath486 in agreement with ( [ deltact ] ) .
the association of symbolic dynamics and mbius - transformation matrices employed in sect .
[ symbdyn ] to derive the foregoing relations obviously incorporates the hyperbolic geometry of the hadamard - gutzwiller model .
we here determine the number @xmath487 of non - equivalent @xmath103-letter words ( alias periodic orbits with @xmath103 segments ) @xmath3 with an alphabet containing @xmath488 different letters .
as mentioned in section v , a rough estimate is @xmath489 , since a symbol @xmath490 may not be succeeded by its `` inverse '' @xmath491 , and since a factor @xmath172 is due to the identification of cyclic permutations .
the influence of the group identity @xmath492 requires more thought .
our first task is to find the shortest possible version of a given word . toward that end
we note that whenever the sequence ( [ all ] ) or any of its cyclic permutations is encountered within a word , we must delete that sequence and thus shorten the word by @xmath493 letters .
similarly , if some part of the identity with @xmath494 symbols and @xmath495 , is encountered , that part must be replaced by its complement with @xmath496 symbols . in any such instance
the word is shortened by 8,6,4 or 2 letters .
next , we have to consider the @xmath150-symbol sequences for which the group identity provides an equivalent @xmath150-symbol sequence like for example @xmath146 .
if a word contains @xmath65 such 4-letter patches there are @xmath497 different representations of the same orbit ; only one of these representations has the property that each symbol corresponds to one segment of the orbit inside the fundamental domain .
let us start with the reversible 4-letter sequences and denote by @xmath498 the probability that a symbol of a long word is the beginning of such a sequence .
a combinatorial estimate of @xmath498 can be made as follows .
there are @xmath499 different @xmath150-letter patterns of the type in question , due to 8 cyclic permutations of both the identity in ( [ all ] ) and its inverse , and each of these patterns occurs with probability @xmath500 .
however , the pattern can only be included when the first and last letter are not the inverse of the neighboring one in the complete word , such that a `` junction factor '' @xmath501 must be incorporated .
the estimate in search thus comes out as @xmath502 . assuming independent occurrence of the 4-letter sequences we expect to find exactly @xmath65 of them in an @xmath103-letter word with the binomial probability @xmath503 we refrain from attempts at improving that combinatorial estimate by accounting for the `` interaction '' of the 4-letter patches caused by their finite length . upon checking many long words with randomly chosen letters we found such interaction effects quite unimportant : the binomial distribution ( [ binalpha ] )
is borne out very well with @xmath504 .
similarly , let @xmath505 be the probability to find , in any letter , the beginning of any of the previously discussed @xmath494-symbol patterns with @xmath506 which has an equivalent shorter partner with @xmath496 symbols . again assuming independence of several such events we get a binomial distribution @xmath507 as before and the combinatorial estimate for @xmath505 is , again neglecting `` interactions '' , @xmath508 = 0.00074 $ ] . here again , the interactions were checked to be unimportant by going through a large sample of random words .
the numerical data for @xmath509 so obtained once more fit the binomial distribution , with @xmath510 . .
we can finally put all pieces together .
the number of @xmath103-letter words allowed after excluding @xmath511 sequences of symbols and accounting for the equivalence of cyclic permutations was seen to be @xmath512 ; only the fraction @xmath513 of these can not be shortened by using the group identity , so we are down to @xmath514 words .
roughly @xmath515 of these , however , contain precisely @xmath65 sequences of four letters each of which comes in two equivalent pairs due to the group identity ; to avoid overcounting we have to divide out the multiplicity @xmath497 .
finally summing over @xmath65 we get the desired number of different @xmath103-letter words as @xmath516^n = \frac{1}{n } { \rm e}^{n \ln p_{\rm eff } } \,,\\ p_{\rm eff } & = & ( 1 - \alpha/2)(1 - \beta)(p-1 ) = 6.98\ , .
\label{peff}\end{aligned}\ ] ] interestingly , the effective number of symbols is not drastically reduced from @xmath517 by the group identity equation ( [ ide ] ) . | employing symbolic dynamics for geodesic motion on the tesselated pseudosphere , the so - called hadamard - gutzwiller model , we construct extremely long periodic orbits without compromising accuracy .
we establish criteria for such long orbits to behave ergodically and to yield reliable statistics for self - crossings and avoided crossings .
self - encounters of periodic orbits are reflected in certain patterns within symbol sequences , and these allow for analytic treatment of the crossing statistics . in particular , the distributions of crossing angles and avoided - crossing widths thus come out as related by analytic continuation .
moreover , the action difference for sieber - richter pairs of orbits ( one orbit has a self - crossing which the other narrowly avoids and otherwise the orbits look very nearly the same ) results to all orders in the crossing angle .
these findings may be helpful for extending the work of sieber and richter towards a fuller understanding of the classical basis of quantum spectral fluctuations . |
the study on reionization history of our universe remains to be an open question in modern cosmology .
it is commonly believed that our universe should be neutral after recombination epoch , the so called dark age .
after that , the first generation of galaxies provide ultraviolet radiation so that the igm start to be ionized from neutral phase . such phase transition of the igm is the so called reionization history and it relates to many fundamental questions of astrophysics , since detailed reionization process depends on the formation and evolution of high energy astrophysical objects , such as mini - quasars , x - ray binaries , metal - free stars , etc . , which provide the sources for reionization transition .
a lot of astronomical observations provide the information of reionization , for example , the level of ionized state at different epochs , even though we could not observe the reionizing process directly .
lymann-@xmath4 forests , a series of absorption lines in the spectra of distant sources(quasars or galaxies ) from the lymann-@xmath4 electron transition of the neutral hydrogen atom(hi ) , could be an important tracer of igm ionizing .
if there exists homogenious distribution of hi gas in the line of sight from source to observer , the lymann-@xmath4 forest will turn into a gunn - peterson ( gp ) trough@xcite , so by observing gp trough , we can map the neutral hydrogen in the igm . from the detection of lymann-@xmath4 absorption lines or gp trough , we know that the universe is highly ionized at least until @xmath5 .
on the other hand , the astronomical observations provide some evidence that there exists hi in the igm at @xmath6 .
those observations show that reionization should last for a period of time .
some other experiments can put constraints on reionization history as well , for example 21 cm experiments which can measure the distribution of hi(@xmath7 ) , and traces the evolution of reionziation .
there are a number of 21 cm experiments , such as gmrt@xcite , lofar@xcite , mwa@xcite , 21cma@xcite , and paper@xcite .
however , due to their low signal - to - noise ratios , none of them can give convincing results about history of reionization at current stage .
measuring the temperature of igm can also give some constraints on the epoch of reionization@xcite . the cosmic microwave background radiation ( cmb )
provides useful information on reionization history .
once the igm are ionized , there will be lots of electrons and the interactions between cmb photons and electrons through thomson scattering will deform the black body distribution of cmb@xcite which can be imprinted in cmb maps .
@xmath1 is an important cosmological parameter in cmb for describing the post thomson scattering effects , and @xmath1 is an integration of the reionized fraction ( which labeled as @xmath0 ) , @xmath8 , where @xmath9 is the conformal time , @xmath10 is the present time , @xmath11 is the thomson cross section and @xmath12 is the number density of free electrons produced by reionization . in cmb power spectra of tt and ee , on small angular scales ,
there is a reducing factor of @xmath13 . for polarization power spectrum @xmath14
, it is enhanced on large angular scale since the scattering would generate extra polarization , and the enhancement has already detected from many cmb experiments .
these effects describing here is only about the globally averaged reionization , the perturbation of the reionization(inhomogeneous effect ) can change @xmath14 on small angular scale , but it is very fainter than weak lensing effect@xcite , and we wo nt consider it in this project . there has being lots of cmb experiments done in recent decades , and the data are becoming more and more accurate . also tight constraints are performed on the reionization optical depth , such as wmap 9 years data@xcite gives @xmath15 and @xmath16 , @xmath17 and @xmath18 from planck temperature and lensing data@xcite , where @xmath19 gives the epoch of the ionized fraction equals to one half .
when using the cmb data , people always adopt an instantaneous model to characterize the evolution of reionzation process , in which , the igm are suddenly reionized in a very short time , and the reionizing process is very short so that the function of ionized fraction @xmath0 can be described by a tanh - based function@xcite @xmath20,\ ] ] where @xmath21 , and @xmath22 for @xmath23 .
@xmath24 is a constant with value @xmath25 and @xmath26 , where @xmath27 is some constant , and there is one to one correspondence between the optical depth(@xmath1 ) and the redshift of reionization(@xmath19 ) when @xmath27 is given .
it is well known that the epoch of reionization is very complicated , there is no evidence that the history of reionization is just an instantaneous model .
also , considering that @xmath1 is the integration of the reionized fraction , the constraints on @xmath1 should be very model dependent , that is to say it will introduce bias with strong assumption on reionization model .
in fact , in this way , @xmath1 can not provide more detailed information on reionization history , since different reionization models can give same optical depth . as shown in fig .
[ figure : models ] , different reionized model could generate different @xmath14 ( in right panel ) even for the same @xmath1(in the left panel ) .
so , we see that the ionized fraction parameter @xmath0 are the more basic parameters for describing reionization history . very recently , planck have released their new results , the whole sky map of cmb anisotropies including both temperature and polarization , which gives the most accurate measurements on cmb .
the high quality data provides a wealth of new information on cosmology .
the tt spectrum is accurately measured to multipoles @xmath28 , and by cross checking with the high resolution ground based cmb observations , such as act@xcite , spt@xcite , the `` damping tail '' is measured with high accuracy in the planck tt power spectrum at @xmath29 @xmath30 @xmath31 .
the most important is that , the cross checking show that te and ee spectra are in good agreement with tt , which provides an important test of the accuracy of the data sets . with the new measurement , in this paper ,
we do a model independent study of the reionization evolution history .
we totally abandon the assumption of instantaneous model , instead , we separate the reionization history into several bins in redshift space , and take @xmath0 for each @xmath3 bin as a free parameter . by doing the global fitting
, we can get constraints on @xmath0 and reconstruct the reionization history .
further more , we adopt the principle component analysis ( pca ) method in our study to get tighter and more reasonable constraints .
pca are widely used in the literature for optimizing the signal to noise ratio and solving for high quality estimation on target parameters .
we have noted that extracting the information of reionization history from astronomic observations are widely studied in the literatures .
the structure of our paper is organized as follows . in section
ii we illustrate the motivation for the model independent study of reionization history . in section iii
we introduce the data sets that adopted in the fitting analysis , the global fitting procedure and principle component analysis method .
the final constraints are presented in section iv .
a summary and conclusion are given in section v.
in fact , even for the same @xmath1 , we can have different kinds of reionization history which will lead to different power spectra , especially for e mode at small @xmath29 , as is illustrated in fig .
[ figure : models ] . to see the detailed effect that reionization history have on power spectra , we perturb the instantaneous model , as shown in fig .
[ figure : perturbation ] , at different redshift bins .
we choose a bin width of @xmath32 to segment the redshift from z=0 to 15 , at each run we take the excursion value as @xmath33 at only one bin while remaining the other bins @xmath34 .
the differences of power spectra @xmath35 are shown in right panel of fig .
[ figure : perturbation ] . from the location of the peaks ,
we know that perturbations at high redshift influence relative high @xmath36 spectrum more , and low redshift perturbations have more effect at low @xmath36 .
we limit @xmath37 since the polarization power spectra at @xmath38 are irrelevant to our analysis of reionization . as we know
that , @xmath1 is a very important parameter when performing the data analysis with cmb , however , it will give bias when we only consider @xmath1 instead of reionization fraction @xmath0 . in order to see
the bias introduced by instantaneous assumption , we simulate future cmb experiments with a non - instantaneous model , and then perform the fitting with an instantaneous model , and compare the final constraints on @xmath1 with the fiducial model .
the fiducial model , a double ionized @xmath39 function is plotted in left of fig .
[ figure : simulation ] in solid line , with this model , by the integration , the optical depth is 0.055 .
we simulate the cmb power spectra of a bicepiii - like futrue cmb data with @xmath40 noise level using camb .
then we perform a global fitting analysis by using the standard @xmath41 model with instantaneous reionization history .
the final constraint derived by the mocked data is shown in the right panel of fig .
[ figure : simulation ] , the vertical line is @xmath42 , comparing with the mean value @xmath43 , it is disfavored at about @xmath44 , which shows the bias from an instantaneous assumption . considering the importance of the reionization history , in the following , we will do a model independent analysis@xcite for reconstructing the reionization fraction @xmath0 with the current data sets of planck , wmap , respectively , as well as bao and sn .
our numerical calculation of global fitting analysis are performed by using a modified cosmomc package@xcite by rewritten the reionization history relevant package .
we segment the epoch of reionization into several bins in redshift space , and we take the ionized fraction @xmath45 in each bins as a constant . in this way , the reionization should not be biased by the prior assumption .
the redshift bins @xmath46 , @xmath47 , are choosen as : @xmath48 $ ] , @xmath49 $ ] , @xmath50 $ ] , @xmath51 $ ] , @xmath52 $ ] , @xmath53 $ ] , @xmath54 $ ] , @xmath55 $ ] , @xmath56 $ ] , will be priorly introduced . however , due to the limited constraining power of the current data sets , we can take @xmath57 bins at most .
also , we have do many optimization design for the bins , for example , adjust the range for the redshift bins , the scale of the bins and so on according to the data sets .
the bins listed in the main text are the best choice form our testing . ] since from the observations we know that at @xmath58 our universe should be neutral and @xmath0 can be neglected@xcite . in order to guarantee @xmath59 in the @xmath60 bins to be a constant and the function of @xmath0 to be smooth
, we link @xmath59 by a @xmath61 function . in numerical calculation of the global fitting analysis , @xmath62 are taken as free parameters .
our procedure are performed with the power law @xmath63cdm@xmath64 model described by the basic parameters of @xmath65 , where @xmath66 is the cold dark matter energy density parameter , @xmath67 the baryon energy density parameter , @xmath68 is the ratio ( multiplied by 100 ) of the sound horizon at decoupling over the angular diameter distance to the last scattering surface , @xmath69 describing the reionization history and @xmath59 are the reionization fraction parameters @xmath59 for the @xmath70 redshift bin , @xmath71 and @xmath72 $ ] are the scalar spectral index and the primordial amplitude respectively . during our calculation , we find that most of the background cosmological parameters , for example @xmath66 , @xmath67 , @xmath68 and @xmath71 , do not have very strong correlation with @xmath0 , in order to get tight constraints , we fixed them to the best fit values listed in table [ table : fiducial ] derived from fitting the data with standard 6 parameters @xmath41+instantaneous reionization model .
we free the reionization history relevant parameters of @xmath59 in each redshift bins as well as @xmath73 which is strongly correlated with @xmath0 parameters .
the top - hat priors of free parameters are @xmath72 \in [ 2 , 4]$ ] and @xmath74 $ ] .
the pivot scale is set at @xmath75 @xmath76 , and in the calculation we assume an purely adiabatic initial condition .
.best fit values for some of the background cosmological parameters derived from fitting current observational data with instantaneous model . [ cols="^,^,^",options="header " , ] pca method would provide tighter constraints on the model parameters , since it performs an estimation on the noise of each mode from the data fitting analysis . by ignoring the noisy modes
, it provides a useful way for measuring cosmological parameters . in the following
, we will adopt pca to tighten the constraints . for normal pca
, we should diagonalize @xmath77 by a matrix @xmath78 , and realize @xmath79 .
the diagonal elements of @xmath80 are @xmath81 , and each of uncorrelated parameters @xmath82 has an error @xmath83 .
we order @xmath81 so that @xmath84 .
there are many orthogonal matrixes that can diagonal f , however , after ordering the diagonal element of @xmath80 , the decorrelation matrix @xmath78 should be fixed .
the rows of @xmath78 are the eigenvectors @xmath85 . from the errors of the eigenvectors
, we can judge which modes are better constraints . by keeping those good modes
, we can reconstruct @xmath62 as : @xmath86 where @xmath87 stands for the number of modes , i and j are the bin order and the mode order respectively . evolution using pca by fitting with data sets of wmap+sn+bao.,scaledwidth=80.0% ] evolution using pca by fitting with data sets of planck+sn+bao.,scaledwidth=80.0% ] we plot the uncorrelated parameters @xmath88 in fig .
[ figure : alphavalue ] , and with these components , we reconstruct the behavior of ionized fraction parameters . in order to get better constraints on @xmath62 ,
we truncate the badly constrained modes and just consider the contribution of the good ones . with eq.(6 ) , we compare several cases for adopting different number of modes respectively in fig .
[ figure : wmappca ] for the data combination of wmap and planck , respectively .
we plot the reconstructed @xmath0 by adopting different number of modes . with all the modes , the 9 bins
, we will get the same @xmath0 function as shown in fig . [
figure : comparison_wmap ] in previous section .
when we ignore the most noisy mode and adopt the first 8 modes , we can get a reconstructed @xmath0 in the second plot , in which we see that the featured structure becomes more obvious , for example in the results given by fitting with wmap+bao+sn , the deviation in 2nd and 8th bin increase to @xmath89 region . with ignoring more modes ,
the variance becomes smaller while the bias becomes greater .
in fact , reconstructing @xmath0 , two things should be balanced properly , a ) .
adopting only few modes into the reconstruction in order to avoid too much noise might lead to obvious bias , since when you abandon more noisy modes , at the same time you also lose information from the data sets which can lead to bias .
b ) . adopting more modes to avoid the risk of deviating from the original @xmath0 , at the same time it brings too much noise which will weaken the final constraints on reionization process . to quantify such balance ,
we do the estimation of the so called @xmath90 following the paper @xcite , where @xmath91 ^ 2+\sum_{i=1}^{n}{\sigma}^2(x_e^i(m ) ) \end{split}\ ] ] here , @xmath92 stands for the value of reconstructed @xmath62 by taking into account m modes at the redshift @xmath93 , and @xmath94 is its corresponding uncertainties .
the @xmath95 denotes the original value of @xmath62 .
@xmath96 denotes the total number of bins .
thus , by using the eq.(7 ) , the @xmath90 can be regarded as the function about number of modes to be kept , @xmath87 . in the fig .
[ figure : risk ] , we illustrate the @xmath90 value of considering different number of modes . obviously , @xmath97 can give the minimal risk , thus keeping the first 3 modes is the best choice to minimize the @xmath90 . with first three modes , the reconstructed @xmath0
are shown in the upper right panel of fig .
[ figure : wmappca ] . based on this result
, we find that most center values of the bins can be consistent with a instantaneous ionized model .
however , there are still three bins that show slight deviation from it , such as the 1st , 2nd and 7th bins , the deviation is at about @xmath98 .
the @xmath99 in the redshift @xmath100 behaves smaller than the value of instantaneous value , the @xmath62 in the redshift @xmath101 is little smaller than instantaneous model value .
while in redshift @xmath102 tends greater than instantaneous value , which imply maybe the resolution of reionization can not be simply characterized by a monotonic function .
we also use the recently released planck data to do the same analysis , almost get the similar result as we can see from the fig .
[ figure : planckpca ] .
one interesting thing for planck is that there is a very strange deviation at the last bin , and it exists at all cases , which means this deviation is the most useful information , which we need pay more attention to it . absorbing the diagonal elements of @xmath103 into orthogonal @xmath78 , and multiplying an orthogonal matrix , is another useful realization , where we adopt @xmath104 , as the decorrelation matrix , then the uncorrelated parameters can be written as @xmath105@xcite .
the advantage of this choice is that the modes are localized distributed in each redshift bin and the weight of each mode is almost positive this kind of choice is considered as a useful basis for achieving the uncorrelated quantities , and is widely used in the analysis of the uncorrelated galaxies power spectrum@xcite as well as the equation of state parameters of dark energy . in fig .
[ figure : qvalue ] , we show the final @xmath106 c.l .
constraints on the uncorrelated parameters(i.e . the parameters @xmath82 ) representing @xmath107 .
we also plot the weights that describe going from correlated @xmath108 to the uncorrelated @xmath82 in fig .
[ figure : weightvalue ] , here we make the weights for @xmath108 sum to unity , actually the significance is independent of which kind of normalization we have . as shown in fig .
[ figure : qvalue ] , most bins are consistent with the instantaneous model at @xmath106 c.l . ,
while there exists weak hint for the deviation from the instantaneous evolution in redshift around @xmath109 and @xmath102 for wmap+sn+bao , when planck+sn+bao are used , most bins value are consistent with instantaneous model , except the last bin(@xmath110 ) , the deviation is about 1@xmath111 c.l .. however , considering large errors , we need more accurate data to confirm it . in the fig .
[ figure : weightvalue ] , we plot the weight for each bin , and the weight function well shows its positive and localized properties , which make the uncorrelated parameters approximately one - to - one correspond to original @xmath62s and better represents the true @xmath0 .
reionization optical depth parameter @xmath1 is an important cosmological parameter for cmb , which describes the thomson scattering between the free streaming cmb photons and free electrons .
@xmath1 can be tightly constrained from cmb observations , since the thomson scattering in the epoch of reionization damps the cmb temperature and polarization spectra at angular scales smaller than the horizon at reionization , and generates e polarization at large angular scale .
the reionization history can influence the constraint on @xmath1 for @xmath1 is the integration of the ionized fraction parameter @xmath0 , thus the excessive assumption will bias the constraint on @xmath1 . usually , using the cosmological parameters for constraining @xmath1 will adopt the so called instantaneous reionization model which assume that the reionizing process is very fast . by simulating the future accurate data with a model which is very different from the usually adopted instantaneous model and constraining @xmath1 by adopting an instantaneous model , the final constraints can be highly biased . in order to study the bias on constraining @xmath1 introduced by the instantaneous assumption , we in this paper perform a model independent analysis .
we take @xmath0 to be free cosmological parameters and do a global fitting analysis for getting constraints on the evolution of @xmath0 by using the data sets of wmap , planck , respectively , combining with bao and supernovae .
the constraint on each bin is not tight enough , but still we can see that the result is consistent with instantaneous model , however there exists deviation at very few bins .
we also adopt the pca method to get better constraints .
we find that , by adopting less noisy modes better constraints can be obtained , and choosing the first 3 modes , it can provide the best reconstructed @xmath0 in which there s weak hints for deviation from an instantaneous evolution .
we hope that , combining with more astrophysical observational data so that we can get more information on the reionization history .
we acknowledge the use of the legacy archive for microwave background data analysis ( lambda ) .
we thank antony lewis , junqing xia , wei zheng and gongbo zhao for helpful discussion . h. l.
is supported in part by the nsfc under grant no .
11033005 and the youth innovation promotion association project and the outstanding young scientists project of the chinese academy of sciences .
x. z. are supported in part by the national science foundation of china under grants no .
11121092 , no.11375202 and no .
this paper is supported in part by the cas pilotb program .
d. j. eisenstein , i. zehavi , d.w .
hogg , _ et al .
_ , detection of the baryon acoustic peak in the large - scale correlation function of sdss luminous red galaxies , astrophys .
j. 633 ( 2005 ) 560 , arxiv : astro - ph/0501171 .
f. beutler , c. blake , m. colless , _ et al .
_ , the 6df galaxy survey : baryon acoustic oscillations and the local hubble constant , mon . not
416 ( 2011 ) 3017 , arxiv:1106.3366 [ astro-ph.co ] . a.j .
ross , l. samushia , c. howlett , _ et al .
_ , the clustering of the sdss dr7 main galaxy sample i : a 4 per cent distance measure at z=0.15 , mon . not .
( 2015 ) 835 - 847 , arxiv:1409.3242 [ astro-ph.co ] .
chuang , f. prada , f. beutler , _ et al .
_ , the clustering of galaxies in the sdss - iii baryon oscillation spectroscopic survey : single - probe measurements from cmass and lowz anisotropic galaxy clustering , arxiv:1312.4889 [ astro-ph.co ] .
n. suzuki , d. rubin , c. lidman , _ et al .
_ , the hubble space telescope cluster supernova survey : v. improving the dark energy constraints above z>1 and building an early - type - hosted supernova sample , apj 746 ( 2012 ) 85 , arxiv:1105.3470 [ astro-ph.co ] . | the study of reionization history plays an important role in understanding the evolution of our universe .
it is commonly believed that the intergalactic medium(igm ) in our universe are fully ionized today , however the reionizing process remains to be mysterious .
a simple instantaneous reionization process is usually adopted in modern cosmology without direct observational evidence . however ,
the history of ionization fraction , @xmath0 will influence cmb observables and constraints on optical depth @xmath1 . with the mocked future data sets based on featured reionization model
, we find the bias on @xmath1 introduced by instantaneous model can not be neglected . in this paper , we study the cosmic reionization history in a model independent way , the so called principle component analysis(pca ) method , and reconstruct @xmath2 at different redshift @xmath3 with the data sets of planck , wmap 9 years temperature and polarization power spectra , combining with the baryon acoustic oscillation(bao ) from galaxy survey and type ia supernovae(sn ) union 2.1 sample respectively .
the results show that reconstructed @xmath0 is consistent with instantaneous behavior , however , there exists slight deviation from this behavior at some epoch . with pca method , after abandoning the noisy modes , we get stronger constraints , and the hints for featured @xmath0 evolution could become a little more obvious . |
many charmonium and charmonium - like states were discovered at @xmath10-factories in the first decade of the 21st century @xcite . whereas some of these are good charmonium candidates , as predicted in different models , many states have exotic properties , which may indicate that exotic states , such as multi - quark , molecule , hybrid , or hadron - quarkonium , have been observed @xcite .
the besiii experiment @xcite at the bepcii storage ring started its first collisions in the tau - charm energy region in 2008 .
after a few years running at suitable energies for its well - defined physics programs @xcite , i.e. , at @xmath11 and @xmath12 peaks in 2009 and the @xmath13 peak in 2010 and 2011 , the besiii experiment started to collect data for the study of the @xmath0 particles , which were not described in the yellow book @xcite .
as the design center - of - mass ( c.m . ) energy of the bepcii was 2.04.2 gev , there were not many options for data samples relevant to the @xmath0-related physics .
besiii took its first data sample at the peak of @xmath14 in may 2011 , with the aim of searching for the well - known @xmath3 in the @xmath14 radiative transition and possibly the excited @xmath15-wave charmonium spin - triplet states in similar transitions .
this sample is about 0.5 fb@xmath16 , which is limited by the one - month running time left after the @xmath13 data taken in the 20102011 run .
data were not collected at the @xmath17 peak because the cleo - c experiment had already collected a sample of about 0.6 fb@xmath16 for the study of @xmath18 decays , which could be used for similar studies .
the upgrade of bepcii s linac in summer 2012 increased the highest beam energy from 2.1 to 2.3 gev , making it possible to collect data at higher c.m .
energies ( up to 4.6 gev ) .
this made data collection possible at almost all known vector states , including @xmath19 , @xmath20 , @xmath21 , and ( marginally ) @xmath22 .
the data collected at a c.m .
energy of @xmath23 gev turned out to be very fruitful .
one month s data of 525 pb@xmath16 ( from december 14 , 2012 to january 14 , 2013 ) produced observations of the charged charmonium - like state @xmath1 @xcite , resulting in changes to the data collection plan for the 20122013 run .
more data were accumulated at c.m .
energies of 4.26 gev and then 4.23 gev .
data from the @xmath20 peak were also obtained in spring 2013 , and data from even higher energies ( 4.42 and 4.6 gev ) were recorded in 2014 after a fine scan of the total hadronic cross - sections between 3.8 and 4.6 gev at more than 100 energy points , with a total integrated luminosity of about 800 pb@xmath16 .
the dedicated data samples for the @xmath0 study are presented in table [ ecm_lum_xyz ] , which lists the nominal c.m .
energy , measured c.m .
energy , and integrated luminosity at each energy point .
these data were used for all the analyses presented in this article . [ cols="^,^,>",options="header " , ]
the besiii experiment observed , for the first time , a charged charmonium - like state close to the @xmath25 threshold @xmath26/@xmath27 , and a charged charmonium - like state close to the @xmath28 threshold @xmath29/@xmath30 .
their neutral partners were also observed , confirming their isospins to be one .
the besiii experiment studied the @xmath31 process at a c.m .
energy of 4.26 gev using a 525 pb@xmath16 data sample @xcite . a structure at around 3.9 gev/@xmath32 was observed in the @xmath33 mass spectrum with a statistical significance larger than @xmath34 , which is referred to as the @xmath1 .
a fit to the @xmath35 invariant mass spectrum ( see fig . [ projfit ] ) , neglecting interference , results in a mass of @xmath36 and a width of @xmath37 mev .
the associated production ratio was measured to be @xmath38 . in the belle experiments ,
the cross - section of @xmath31 was measured from 3.85.5 gev using the initial state radiation ( isr ) method .
the intermediate states in @xmath39 decays were also investigated @xcite .
the @xmath1 state ( referred to as @xmath40 in the belle paper ) with a mass of @xmath41 and a width of @xmath42 mev was observed in the @xmath35 mass spectrum ( see fig . [ projfit ] ) with a statistical significance larger than @xmath43 .
( left panel from besiii and right panel from belle ) .
points with error bars are data , the curves are the best fit , the dashed histograms are the phase space distributions and the shaded histograms are the non-@xmath44 background estimated from the normalized @xmath11 sidebands.,title="fig:",height=200 ] ( left panel from besiii and right panel from belle ) .
points with error bars are data , the curves are the best fit , the dashed histograms are the phase space distributions and the shaded histograms are the non-@xmath44 background estimated from the normalized @xmath11 sidebands.,title="fig:",height=200 ] the @xmath1 state was confirmed shortly after with cleo - c data at a c.m .
energy of 4.17 gev @xcite , and the mass and width agreed very well with the besiii and belle measurements .
a neutral state @xmath45 with a significance of @xmath46 was observed at besiii in @xmath47 with c.m .
energy ranges from 4.194.42 gev @xcite .
the mass and width were measured to be @xmath48 mev/@xmath32 and @xmath49 mev , respectively .
this state is interpreted as the neutral partner of the @xmath50 , as it decays to @xmath51 and its mass is close to that of @xmath50 .
this is in agreement with the previously reported @xmath52 evidence for @xmath53 in the cleo - c data @xcite .
the measured born cross - sections of @xmath54 were about half of those for @xmath55 measured in the belle experiment @xcite , which is consistent with the isospin symmetry expectation .
the @xmath1 state observed in the @xmath56 final state is close to and above the @xmath25 mass threshold .
with the same data sample at @xmath23 gev , the besiii experiment studied @xmath57 . a structure ( referred to as @xmath27 )
was observed in the @xmath58 invariant mass distribution @xcite . when fitted to a mass - dependent - width breit wigner ( bw ) function , the pole mass and width were determined to be @xmath59 mev/@xmath32 and @xmath60 mev , respectively ( see fig . [ xuxp ] ) .
the angular distribution of the @xmath27 system favors a @xmath61 assignment for the structure and disfavors @xmath62 or @xmath63 .
the production rate was measured to be @xmath64 pb . ( left ) and @xmath65 ( right ) distributions for selected events at @xmath23 gev .
the curves show the best fits . , title="fig:",width=264 ] ( left ) and @xmath65 ( right ) distributions for selected events at @xmath23 gev .
the curves show the best fits . , title="fig:",width=264 ] an important question is whether @xmath27 is the same as @xmath1 @xcite .
the mass and width of @xmath27 are @xmath66 and @xmath67 , respectively , below those of @xmath1 , as observed by the besiii and belle experiments .
however , neither fit considers the possibility of interference with a coherent non - resonant background , which could shift the results . a spin - parity quantum number determination for @xmath26 would provide an additional test of this possibility . assuming the @xmath27 structure is caused by @xmath26
, we obtain @xmath68 .
this ratio is much smaller than typical values for decays of conventional charmonium states above the open charm threshold , e.g. , @xmath69
@xcite and @xmath70 @xcite .
this suggests very different dynamics in the @xmath71-@xmath1 system .
besiii measured @xmath72 cross - sections @xcite at c.m .
energies of 3.904.42 gev .
intermediate states were studied by examining the dalitz plot of the selected @xmath73 candidate events .
the @xmath74 signal was selected using @xmath75 gev/@xmath32 , and @xmath73 samples of 859 events at 4.23 gev , 586 events at 4.26 gev , and 469 events at 4.36 gev were obtained with purities of 65% .
although there are no clear structures in the @xmath76 system , there is clear evidence for an exotic charmonium - like structure in the @xmath77 system , as clearly shown in the dalitz plot .
figure [ 1dfit ] ( left ) shows the projection of the @xmath78 ( two entries per event ) distribution for the signal events , as well as the background events estimated from normalized @xmath74 mass sidebands . there is a significant peak at around 4.02 gev/@xmath32 ( @xmath2 ) , and there are also some events at around 3.9 gev/@xmath32 ( inset of fig .
[ 1dfit ] ( left ) ) , which could be @xmath1 .
the individual datasets at @xmath79 , 4.26 , and 4.36 gev show similar structures .
( left panel ) and @xmath80 ( right panel ) distributions at 4.23 , 4.26 , and 4.36 gev in the besiii data ; the inset in the left panel shows the sum of the simultaneous fit to the @xmath81 distributions at 4.23 and 4.26 gev with @xmath1 and @xmath2 .
dots with error bars are data ; shaded histograms are normalized sideband background ; the solid curves show the total fit , and the dotted curves the backgrounds from the fit.,title="fig:",scaledwidth=45.0% ] ( left panel ) and @xmath80 ( right panel ) distributions at 4.23 , 4.26 , and 4.36 gev in the besiii data ; the inset in the left panel shows the sum of the simultaneous fit to the @xmath81 distributions at 4.23 and 4.26 gev with @xmath1 and @xmath2 .
dots with error bars are data ; shaded histograms are normalized sideband background ; the solid curves show the total fit , and the dotted curves the backgrounds from the fit.,title="fig:",scaledwidth=45.0% ] an unbinned maximum likelihood fit was applied to the @xmath78 distribution summed over the 16 @xmath82 decay modes .
the data at 4.23 , 4.26 , and 4.36 gev were fitted simultaneously to the same signal function with common mass and width .
figure [ 1dfit ] ( left ) shows the fitted results .
the mass and width of @xmath2 were measured to be @xmath83 and @xmath84 mev , respectively .
the statistical significance of the @xmath2 signal was found to be greater than @xmath85 .
adding @xmath1 with mass and width fixed to the besiii measurements @xcite to the fit results in a statistical significance of 2.1@xmath86 ( see the inset of fig .
[ 1dfit ] ( left ) ) . at the 90% confidence level ( c.l . ) , the upper limits on the production cross - sections are set to @xmath87 pb at 4.23 gev and @xmath88 pb at 4.26 gev
. these are lower than those of @xmath89 @xcite .
besiii also observed @xmath90 at @xmath79 , 4.26 , and 4.36 gev for the first time @xcite .
the measured born cross - sections were about half of those for @xmath91 , which agree with expectations based on isospin symmetry within systematic uncertainties . a narrow structure with a mass of @xmath92 mev/@xmath32 ( for fitting , the width was fixed to that measured in the @xmath93 process @xcite because of low statistics )
was observed in the @xmath94 mass spectrum ( fig .
[ 1dfit ] ( right ) ) .
this structure is most likely the neutral isospin partner of the charged @xmath29 observed in the @xmath93 process @xcite .
this observation indicates that there are no anomalously large isospin violations in @xmath95 and @xmath96 systems .
the besiii experiment also studied the @xmath97 process at 4.26 gev using an 827 pb@xmath16 data sample @xcite .
based on a partial reconstruction technique , the born cross - section was measured to be @xmath98 pb . a structure near the @xmath99 threshold in the @xmath100 recoil mass spectrum was observed , and this is denoted as @xmath30 ( see fig . [
fig : fit ] ( left ) ) .
the measured mass and width of the structure were @xmath101 mev/@xmath32 and @xmath102 mev , respectively , from the fit with a constant - width bw function for the signal .
the associated production ratio @xmath103 was determined to be @xmath104 .
recoil mass spectrum ( left ) in @xmath105 at @xmath23 gev , and to the @xmath106 recoil mass spectrum ( right ) in @xmath107 at @xmath79 and 4.26 gev at besiii .
, title="fig:",height=226 ] recoil mass spectrum ( left ) in @xmath105 at @xmath23 gev , and to the @xmath106 recoil mass spectrum ( right ) in @xmath107 at @xmath79 and 4.26 gev at besiii .
, title="fig:",height=226 ] using data at @xmath79 and 4.26 gev , a structure was observed in the @xmath106 recoil mass spectrum in the @xmath108 process @xcite .
assuming that the enhancement is due to a neutral state decaying to @xmath109 , the mass and width of its pole position were determined to be @xmath110 mev/@xmath32 and @xmath111 mev , respectively ( see fig . [
fig : fit ] ( right ) ) .
the born cross - section @xmath112 was measured to be @xmath113 at 4.23 gev and @xmath114 at 4.26 gev .
the ratio @xmath115 is compatible with unity at @xmath23 gev , which is expected from isospin symmetry .
in addition , @xmath116 has a mass and width that are very close to those of @xmath117 , which couples to @xmath99 .
therefore , the observed @xmath116 state is a good candidate for the isospin partner of @xmath117 . as the @xmath30 parameters agree to within 1.5@xmath86 with those of @xmath2
, it is very probable that they are the same state .
as the results for @xmath117 are only from data at 4.26 gev , extending the analysis to 4.23 gev and 4.36 gev will probably provide a definite answer .
besiii observed @xmath119 , with @xmath11 reconstructed through its decays into lepton pairs ( @xmath120 or @xmath121 ) @xcite . the @xmath122 distribution ( summed over all energy points ) , as shown in fig .
[ fit - mx ] ( left ) , was fitted to extract the mass and signal yield of @xmath123 .
the isr @xmath12 signal was used to calibrate the absolute mass scale and to extract the resolution difference between the data and a monte carlo ( mc ) simulation .
figure [ fit - mx ] shows the fitted result : the measured mass of @xmath123 was @xmath124 mev/@xmath32 . from a fit with a floating width , we obtain a width of @xmath125 mev , or less than 2.4 mev at the 90% c.l .
the statistical significance of @xmath123 is @xmath126 .
distribution observed at besiii .
dots with error bars are data , the curves are the best fit .
right panel : fit to @xmath127\times \mathcal{b}[{x(3872)}\to { \pi^+\pi^- j/\psi}]$ ] measured by besiii with a @xmath71 resonance ( red solid curve ) , a linear continuum ( blue dashed curve ) , or an @xmath128-transition phase space term ( red dotted - dashed curve ) .
dots with error bars are data.,title="fig:",height=188 ] distribution observed at besiii . dots with error bars are data , the curves are the best fit .
right panel : fit to @xmath127\times \mathcal{b}[{x(3872)}\to { \pi^+\pi^- j/\psi}]$ ] measured by besiii with a @xmath71 resonance ( red solid curve ) , a linear continuum ( blue dashed curve ) , or an @xmath128-transition phase space term ( red dotted - dashed curve ) .
dots with error bars are data.,title="fig:",height=188 ] the born - order cross - section was measured , and the results are listed in table [ sec ] . for 4.009 and 4.36 gev data , because the @xmath123 signal is not significant , upper limits on the production rates are given at the 90% c.l .
the measured cross - sections at around 4.26 gev are an order of magnitude higher than the nrqcd calculation of continuum production @xcite , which may suggest the @xmath123 events come from resonance decays .
@cc@ @xmath129 ( gev ) & @xmath127 \cdot\mathcal{b}({x(3872)}\to { \pi^+\pi^- j/\psi})$ ] ( pb ) + 4.009 & @xmath130 or @xmath131 + 4.229 & @xmath132 + 4.260 & @xmath133 + 4.360 & @xmath134 or @xmath135 + [ sec ] the energy - dependent cross - sections were fitted with a @xmath71 resonance ( parameters fixed to pdg @xcite values ) , linear continuum , or @xmath128-transition phase space ( @xmath136 ) term .
figure [ fit - mx ] ( right ) shows all the fitted results , which imply that @xmath137 ( c.l .
= 92% ) , 5.5/2 ( c.l .
= 6% ) , and 8.7/3 ( c.l .
= 3% ) for a @xmath71 resonance , linear continuum , and phase space distribution , respectively .
thus , the @xmath71 resonance describes the data better than the other two options .
these observations strongly support the existence of the radiative transition process @xmath138 .
the @xmath138 process could be another previously unseen decay mode of the @xmath71 resonance .
together with the transitions to the charged charmonium - like state @xmath1 @xcite , this suggests that there might be some commonality in the nature of @xmath123 , @xmath71 , and @xmath1 , and so the model developed to interpret any one of them should also consider the other two .
as an example , the authors of ref . @xcite integrated these states into a molecular picture to calculate @xmath139 cross - sections . combining the above with the @xmath31 cross - section measurement at @xmath23 gev from besiii @xcite , we obtain @xmath127\cdot { { \cal b}}[{x(3872)}\to { \pi^+\pi^- j/\psi}]/\sigma^b({e^+e^-}\to { \pi^+\pi^- j/\psi } ) = ( 5.2\pm 1.9)\times 10^{-3}$ ] , under the assumption that @xmath123 and @xmath44 are only produced from @xmath71 decays . if we take @xmath140 = 5\%$ ] @xcite , then @xmath141}{{{\cal b}}({y(4260)}\to { \pi^+\pi^- j/\psi})}\sim 0.1 $ ] .
besiii observed @xmath142 in the @xmath143 process with a statistical significance of @xmath144 in data samples at c.m .
energies of @xmath1454.23 , 4.26 , 4.36 , 4.42 , and 4.60 gev @xcite .
figure [ x - fit ] shows the fitted results to @xmath76 recoil mass distributions for events in the @xmath146 and @xmath147 signal regions
. the fit yields @xmath148 @xmath142 signal events in the @xmath149 mode , with a measured mass of @xmath142 of @xmath150 , where the first error is statistical and the second systematic . for the @xmath151 mode
, no significant @xmath142 signal was observed , and an upper limit on its production rate could be determined .
the limited statistics do not allow a measurement of the intrinsic width of @xmath142 . from a fit using the bw function ( with a floating width parameter ) convolved with gaussian resolution , it can be determined that @xmath152<16 $ ] mev at the 90% c.l .
( including systematic errors ) .
this measurement agrees well with the values found by belle @xcite .
distribution of @xmath149 events ( left ) and @xmath151 events ( right ) , respectively .
dots with error bars are data , red solid curves are total fit , dashed blue curves are background , and the green shaded histograms are @xmath11 mass sideband events.,title="fig:",height=207 ] distribution of @xmath149 events ( left ) and @xmath151 events ( right ) , respectively .
dots with error bars are data , red solid curves are total fit , dashed blue curves are background , and the green shaded histograms are @xmath11 mass sideband events.,title="fig:",height=207 ] the production cross - sections of @xmath153 , @xmath154 were also measured at these c.m
. energies .
the cross - sections of @xmath155 were fitted with the @xmath20 shape or the @xmath21 shape , with their resonance parameters fixed to the pdg values @xcite .
figure [ fit - sec ] shows the fitted results , which give the kolmogorov
smirnov statistic @xmath156 for the @xmath20 hypothesis ( h1 ) and @xmath157 for the @xmath21 hypothesis ( h2 ) , based on the kolmogorov
smirnov test @xcite .
thus , both the @xmath20 and @xmath21 hypotheses ( @xmath158 ) are accepted at a 90% c.l .
\cdot \mathcal{b}(x(3823)\to\gamma\chi_{c1})$ ] to the @xmath20 and @xmath21 line shapes .
dots with error bars ( statistical only ) are data .
the red solid ( blue dashed ) curve shows a fit with the @xmath20 ( @xmath21 ) line shape.,height=240 ] the @xmath142 resonance is a good candidate for the @xmath5 charmonium state . according to potential models @xcite ,
the @xmath159-wave charmonium states are expected to be within a mass range of 3.823.85 gev . among these , the @xmath160 transition is forbidden because of c - parity conservation , and the amplitude for @xmath161 is expected to be small @xcite .
the mass of @xmath5 is in the @xmath162 gev/@xmath32 range predicted by several phenomenological calculations @xcite . in this case , the mass of @xmath5 was above the @xmath163 threshold but below the @xmath25 threshold . because @xmath164 violates parity ,
@xmath5 is expected to be narrow , in agreement with the observation , and @xmath165 is expected to be a dominant decay mode @xcite . from the cross - section measurement , we obtain the ratio @xmath166}{\mathcal{b}[x(3823)\to \gamma\chi_{c1}]}<0.42 $ ] ( where systematic uncertainty cancels ) at the 90% c.l . , which also agrees with expectations for the @xmath5 state @xcite .
using data samples collected at @xmath129 = 4.009 , 4.23 , 4.26 , and 4.36 gev , besiii searched for @xmath168 @xmath169 with the subsequent decay @xmath170 and @xmath171 .
evidence for the @xmath172 and @xmath173 processes was observed with a statistical significance of 3.0@xmath86 and 3.4@xmath86 , respectively @xcite .
no evidence for @xmath174 was observed .
the corresponding born cross - sections of @xmath175 at different c.m .
energies were calculated . under the assumption that @xmath176 signals were absent , the upper limits on the born cross - sections
were calculated at the 90% c.l .
these upper limits on the born cross - section of @xmath177 are compatible with the theoretical prediction from the nrqcd calculation @xcite .
besiii searched for @xmath178 via @xmath179 at @xmath180 , 4.26 , and 4.36 gev , but no significant @xmath178 signal was observed in any of the data samples @xcite .
the upper limits of the product of the cross - section and branching fraction @xmath181 \cdot \mathcal{b}(y(4140 ) \rightarrow \phi{j/\psi})$ ] at the 90% c.l . were estimated to be 0.35 , 0.28 , and 0.33 pb at @xmath180 , 4.26 , and 4.36 gev , respectively .
these upper limits can be compared with the @xmath3 production rates @xcite , which were measured from the same data samples ( see sec . [
sec : x3872 ] ) .
the latter were @xmath182 \cdot \mathcal{b}(x(3872 ) \rightarrow \pi^{+ } \pi^{- } { j/\psi } ) = [ 0.27 \pm 0.09({\rm stat } ) \pm 0.02({\rm syst})]$ ] pb , @xmath183 $ ] pb , and @xmath184 $ ] pb at @xmath185 , and @xmath186 gev , respectively , which are of the same order of magnitude as the upper limits of @xmath187 \cdot \mathcal{b}(y(4140 ) \rightarrow \phi{j/\psi})$ ] at the same energy .
the branching fraction @xmath188 has not previously been measured . using the partial width of @xmath189 calculated under the molecule hypothesis @xcite and the total width of the @xmath178 measured by cdf @xcite ,
the branching fraction was estimated to be approximately 30% .
a rough estimation for @xmath190 is 5% @xcite . combining these numbers , the ratio @xmath191}/{\sigma[{e^+e^-}\to\gamma x(3872)]}$ ]
can be estimated to be of the order of 0.1 or less at @xmath180 and 4.26 gev .
based on data samples collected between @xmath194 and 4.42 gev , the @xmath195 process was observed at @xmath79 and 4.26 gev for the first time @xcite , and the born cross - sections were measured to be @xmath196 and @xmath197 pb , respectively . for other energy points ,
no significant signals were found , and upper limits on the cross - section at the 90% c.l . were determined .
the data reveal a sizable @xmath198 production at around 4.23 gev/@xmath32 , as predicted in ref . @xcite . by assuming the @xmath198 signals come from a single resonance , the @xmath199 , mass , and width of the resonance were determined to be @xmath200 ev , @xmath201 mev/@xmath32 , and @xmath202 mev , respectively (
the fit is shown in fig .
[ cs - bw - float ] ) .
the parameters are consistent with those of the narrow structure in the @xmath203 process @xcite , but are not consistent with those obtained by fitting a single resonance to the @xmath204 cross - section @xcite ( with a mass of @xmath205 and a width of @xmath206 ) .
this suggests that the observed @xmath198 signals are unlikely to originate from @xmath19 .
the @xmath207 channels were also sought , but no significant signals were observed ; upper limits at the 90% c.l . on the production cross - sections
were determined .
the very small measured ratios of @xmath208 cross - sections to those for @xmath209 are inconsistent with the prediction in ref .
@xcite . with a resonance ( solid curve ) , or a phase space term ( dot - dashed curve ) .
dots with error bars are the dressed cross sections .
the uncertainties are statistical only . ,
scaledwidth=60.0% ] using data samples collected at energies of 3.814.60 gev , besiii analyzed @xmath211 @xcite .
statistically significant @xmath212 signals were observed , and the corresponding born cross - sections were measured .
in addition , a search for the @xmath213 process observed no significant signals , and upper limits at the 90% c.l . on the born cross - section
were set .
bes_belle_etajpsi.eps ( 19,57)*(a ) * bes_belle_pipijpsi.eps ( 19,57)*(b ) * a comparison of the born cross - sections @xmath214 in this measurement to previous results @xcite is shown in fig .
[ bes_belle ] ( a ) , indicating very good agreement .
the measured born cross - sections were also compared to those of @xmath215 obtained from the belle experiment @xcite , as shown in fig .
[ bes_belle ] ( b ) .
different line shapes can be observed in these two processes , indicating that the production mechanism of @xmath216 differs from that of @xmath217 in the vicinity of @xmath129 = 4.14.6 gev .
this could indicate the existence of a rich spectrum of the @xmath218 states in this energy region with different coupling strengths to various decay modes .
as shown in previous sections , besiii has achieved a lot in the study of the @xmath0 states and the conventional charmonium states .
however , there are more questions to be answered with the currently available data and , more importantly , with the data samples that besiii is able to collect in the next few years .
a few topics that need to be studied with more data are listed below .
* in the @xmath219 sector : * * where are the @xmath123 and @xmath5 coming from , resonance decays or continuum production ? * * can the other @xmath219 states , such as @xmath220 , @xmath221 , @xmath222 ( @xmath178 ) , and @xmath223 be produced in a similar way ? * * can the charmonium 2p , 3p states and the s - wave spin - singlet states ( 3s , 4s , 5s ) be observed in radiative transitions ? * in the @xmath224 sector : * * is the @xmath71 structure a single resonance , or does it have a more complicated sub - structure ? is @xmath225 a real resonance ? * * what are the other decay modes of @xmath20 ?
* * what is hidden in the @xmath72 line shape ?
* * is the @xmath22 observed in @xmath226 the same as the @xmath227 observed in @xmath228 @xcite ? * * what is the correlation between charm production ( @xmath229 open charm final states ) and charmonium production ? * * where is the vector charmonium @xmath230 state ? *
* are there charmonium states between @xmath231 and @xmath21 , and/or between @xmath21 and @xmath22 ? * * can the vector charmonium hybrid state be observed @xcite ? * in the @xmath232 sector : * * are the @xmath24 states produced from resonance decays or from continuum production ? * * is there a @xmath233 state decaying into @xmath234 or @xmath235 , @xmath236 ? * * are there more @xmath24 and @xmath233 states ? * in the @xmath237 sector : * * can @xmath238 be produced and studied ? * * can the other excited charmed mesons be produced and studied with high - energy data ?
besiii is going to collect about 3 fb@xmath16 of data in the vicinity of @xmath231 to study the @xmath18 decay properties in 20152016 run .
these data can be used to answer some of the questions listed above .
a few specific topics are listed here . 1 . in @xmath31
, there is a dip at around 4.17 gev and a sharp increase at around 4.23 gev @xcite .
the data may help to identify where the turning point is located .
2 . for the @xmath193 process
@xcite , there are few data points close to the threshold to support the claim of a narrow structure .
3 . for the @xmath239 process
, it seems there is a structure at around 4.2 gev @xcite .
4 . in the @xmath210 process
, the line shape differs from that of @xmath31 , with a peak at around 4.2 gev , but there are no data points on the left side .
5 . in @xmath240 , is there a structure close to the threshold ?
what does the @xmath241 line shape look like close to the threshold ?
are the @xmath8 signals from @xmath231 or @xmath19 , or something else @xcite ? 7 .
the search for @xmath242-wave charmonium states via @xmath231 decays .
the search for @xmath243 , @xmath244 , @xmath245 from @xmath231 decays .
more data are always better , but as we need to collect data at many points and the collection period is limited , we need to define lower limits for the number of data points and the luminosity needed at each point to ensure meaningful measurements of the physical quantities of interest . as we require high - precision cross - sections for all of the open charm modes and the cross - sections of some hadronic transition modes to fit the resonant parameters of the vector states , we require sufficient data points to give the excitation curve of any of the existing vector resonances and possible hidden structures in the full energy range .
some special issues need to be considered : ( 1 ) the thresholds of all the open charm and charmonium + hadron final states ; ( 2 ) possible energy regions where large interference effects are expected ; ( 3 ) energy regions where narrow structures are expected ; ( 4 ) energy regions that were not well explored before ; and ( 5 ) the beam energy spread . from previous calculations and measurements ,
it is known that the energy spread of bepcii at energy ranges of 3.85.0 gev is around 1.42.0 mev . in this case
, we would not use energy steps finer than three times the energy spread . that is , unless very necessary
, we would not take data at energy points less than 5 mev from an existing data point .
thus , we would take data with 5 mev steps in the energy regions where narrow structures or dramatic effects are expected , such as the thresholds for open charm or hidden charm final states , and the low mass shoulder of @xmath19 , where interference effects have been reported @xcite .
otherwise , 10 mev steps will be used .
the limit on the data size is set according to the precision of the hadronic transition modes , which typically have cross - sections of a few to a few tens of picobarns . from the analyses in previous sections , we find that an integrated luminosity of 500 pb@xmath16 is needed to reach a reasonably high precision for most modes of interest . to search for the @xmath0 and charmonium states via radiative transition
, a data sample of at least 500 pb@xmath16 is also needed to reach a @xmath246 level observation of a signal if the production rate is @xmath247 or higher .
a detailed mc study of the precision that can be reached or the exact luminosity needed for each energy point is necessary , as the background level may be very different at different resonant peaks . with the above principles in mind
, we propose to collect data samples at about 60 c.m .
energies from 4.0 gev to the maximum energy that bepcii can reach ( currently 4.6 gev ) in 10 mev steps . around 500 pb@xmath16 at each energy will be necessary for a comprehensive study of the @xmath0 and charmonium states . as besiii
has already accumulated about 5 fb@xmath16 ( see table [ ecm_lum_xyz ] ) , another 25 fb@xmath16 data should be accumulated .
this will take about 5 years at bepcii @xcite .
belle - ii @xcite will start collecting data in 2018 , and will accumulate 50 ab@xmath16 data at the @xmath248 peak by 2024 .
these data samples can be used to study the @xmath0 and charmonium states in many different ways @xcite , among which isr can produce events in the same energy range covered by besiii .
figure [ lum_belle2 ] shows the effective luminosity at bepcii energy in the belle - ii data samples .
we can see that , for 10 ab@xmath16 belle - ii data , we have about 400500 pb@xmath16 data for every 10 mev in the range 45 gev , comparable to the data sample proposed at besiii in the previous section .
of course , the isr analyses have a lower efficiency than in direct @xmath4 collisions because of the extra isr photons and the boost given to events along the beam direction . even taking these effects into account
, the full belle - ii data sample , which corresponds to about 2,0002,300 pb@xmath16 data for every 10 mev from 45 gev , will result in similar statistics for modes like @xmath31 .
belle - ii has the advantage that data at different energies will be accumulated at the same time , making the analysis much simpler than at besiii at 60 data points . in addition , belle - ii can produce events above 4.6 gev , which is currently the maximum energy of bepcii .
possible upgrades to increase the maximum c.m .
energy of bepcii will obviously expand the physical possibilities of besiii .
data samples.,scaledwidth=60.0% ] the hiepa project @xcite being discussed at this workshop will improve the aforementioned studies in many aspects , particularly with c.m .
energies of up to 7 gev and luminosity improvements of a factor of 100 . these will allow a finer scan in the full energy region with more integrated luminosity .
this will enable a better understanding of all the studies listed in this article .
with the world s largest data samples at energies of 3.84.6 gev , the besiii experiment made a significant contribution to the study of the charmonium and @xmath0 states . to further strengthen such studies
, besiii may collect more data from 4.04.6 gev ( or even higher , if possible ) .
these data will be complementary to the belle - ii study , with many other production mechanisms .
the hiepa project may enable a systematic understanding of the nature of the @xmath0 and charmonium states .
this work is supported in part by national natural science foundation of china ( nsfc ) under contract nos .
11235011 and 11475187 ; the ministry of science and technology of china under contract no .
2015cb856701 , and the cas center for excellence in particle physics ( ccepp ) .
m. ablikim _ et al . _
[ besiii collaboration ] , `` precision measurement of the integrated luminosity of the data taken by besiii at center of mass energies between 3.810 gev and 4.600 gev , '' arxiv:1503.03408 [ hep - ex ] .
s. godfrey and n. isgur , `` mesons in a relativized quark model with chromodynamics , '' phys . rev .
d * 32 * , 189 ( 1985 ) ; w. kwong , j. l. rosner and c. quigg , `` heavy quark systems , '' ann .
nucl . part .
sci . * 37 * , 325 ( 1987 ) ; d. ebert , r. n. faustov and v. o. galkin , `` properties of heavy quarkonia and @xmath271 mesons in the relativistic quark model , '' phys .
d * 67 * , 014027 ( 2003 ) ; e. j. eichten , k. lane and c. quigg , `` charmonium levels near threshold and the narrow state @xmath272 , '' phys .
d * 69 * , 094019 ( 2004 ) ; b. q. li and k. t. chao , `` higher charmonia and x , y , z states with screened potential , '' phys . rev .
d * 79 * , 094004 ( 2009 ) ; m. blank and a. krassnigg , `` bottomonium in a bethe - salpeter - equation study , '' phys .
d * 84 * , 096014 ( 2011 ) .
m. ablikim _ et al . _
[ besiii collaboration ] , `` measurement of the @xmath211 cross section and search for @xmath284 at center - of - mass energies between 3.810 and 4.600 gev , '' phys .
d * 91 * , no .
11 , 112005 ( 2015 ) .
p. guo , a. p. szczepaniak , g. galata , a. vassallo and e. santopinto , `` heavy quarkonium hybrids from coulomb gauge qcd , '' phys .
d * 78 * , 056003 ( 2008 ) ; j. j. dudek and e. rrapaj , `` charmonium in lattice qcd and the non - relativistic quark - model , '' phys . rev .
d * 78 * , 094504 ( 2008 ) . z. g. zhao , talk at the international workshop on physics at future high intensity collider 2 - 7 gev in china " , january 13 - 16 , 2015 , university of science and technology of china ( ustc ) , hefei , china . | with its unique data samples at energies of 3.84.6 gev , the besiii experiment made a significant contribution to the study of charmonium and charmonium - like states , i.e. , the @xmath0 states .
we review the results for observations of the @xmath1 and @xmath2 states , the @xmath3 in @xmath4 annihilation , and charmonium @xmath5 state , as well as measurements of the cross - sections of @xmath6 and @xmath7 , and the search for @xmath8 and @xmath9 .
we also present data from besiii that may further strengthen the study of the @xmath0 and conventional charmonium states , and discuss perspectives on future experiments . |
the axelrod model is one of the most popular agent - based models of cultural dynamics .
in addition to a spatial structure , which is modeled through a graph in which vertices represent individuals and edges potential dyadic interactions between two individuals , it includes two important social factors : social influence and homophily .
the former is the tendency of individuals to become more similar when they interact , while the latter is the tendency of individuals to interact more frequently with individuals who are more similar .
note that the voter model @xcite accounts for social influence since an interaction between two individuals results in a perfect agreement between them .
the voter model , however , excludes homophily . to also account for this factor
, one needs to be able to define a certain opinion or cultural distance between any two individuals through which the frequency of the interactions between the two individuals can be measured . in the model proposed by political scientist robert axelrod @xcite ,
each individual is characterized by her opinions on @xmath0 different cultural features , each of which assumes @xmath1 possible states .
homophily is modeled by assuming that pairs of neighbors interact at a rate equal to the fraction of cultural features for which they agree , and social influence by assuming that , as a result of the interaction , one of the cultural features for which members of the interacting pair disagree ( if any ) is chosen uniformly at random , and the state of one of both individuals is set equal to the state of the other individual for this cultural feature .
more formally , the axelrod model on the one - dimensional lattice is the continuous - time markov chain whose state space consists of all spatial configurations @xmath4 that map the vertex set viewed as the set of all individuals into the set of cultures . to describe the dynamics of the axelrod model , it is convenient to introduce @xmath5 where @xmath6 refers to the @xmath7th coordinate of the vector @xmath8 , which denotes the fraction of cultural features the two vertices @xmath9 and @xmath10 share . to describe the elementary transitions of the spatial configuration
, we also introduce the operator @xmath11 defined on the set of configurations by @xmath12 in other words , configuration @xmath13 is obtained from configuration @xmath14 by setting the @xmath7th feature of the individual at vertex @xmath9 equal to the @xmath7th feature of the individual at vertex @xmath10 and leaving the state of all the other features in the system unchanged .
the dynamics of the axelrod model is then described by the markov generator @xmath15 defined on the set of cylinder functions by @xmath16 { \mathbf{1}}\bigl\{\eta(x , i ) \neq \eta(y , i ) \bigr\ } \bigl[f ( \sigma_{x , y , i } \eta ) - f ( \eta)\bigr].\ ] ] the expression of the markov generator indicates that the conditional rate at which the @xmath7th feature of vertex @xmath9 is set equal to the @xmath7th feature of vertex @xmath10 given that these two vertices are nearest neighbors that disagree on their @xmath7th feature can be written as @xmath17 = f ( x , y ) \times \frac{1}{f ( 1 - f ( x , y ) ) } \times \frac{1}{2},\ ] ] which , as required , is equal to the fraction of features both vertices have in common , which is the rate at which the vertices interact , times the reciprocal of the number of features for which both vertices disagree , which is the probability that any of these features is the one chosen for update , times the probability one half that vertex @xmath9 rather than vertex @xmath10 is chosen to be updated .
note that , when the number of features @xmath18 , the system is static , while when the number of states per feature @xmath19 there is only one possible culture . also , to avoid trivialities
, we assume from now on that the two parameters of the system are strictly larger than one . the main question about the axelrod model is whether the system fluctuates and evolves to a global consensus or gets trapped in a highly fragmented configuration . to define this dichotomy rigorously
, we say that the system fluctuates whenever @xmath20 \\[-8pt ] \eqntext{\mbox{for all } x \in{\mathbb{z}}\mbox { and } i \in\{1 , 2 , \ldots , f \}}\end{aligned}\ ] ] and fixates if there exists a configuration @xmath21 such that @xmath22 \\[-8pt ] \eqntext{\mbox{for all } x \in{\mathbb{z}}\mbox { and } i \in\{1 , 2 , \ldots , f \}.}\end{aligned}\ ] ] in other words , fixation means that the culture of each individual is only updated a finite number of times , so fluctuation ( [ eqfluctuation ] ) and fixation ( [ eqfixation ] ) exclude each other .
we define convergence to a global consensus mathematically as a clustering of the system , that is , @xmath23 note that whether the system fluctuates or fixates depends not only on the number of cultural features and the number of states per feature , but also on the initial distribution .
indeed , regardless of the parameters , the system starting from a configuration in which all the individuals agree for a given cultural feature while the states at the other cultural features are independent and occur with the same probability always fluctuates . on the other hand ,
regardless of the parameters , the system starting from a configuration in which all the even sites share the same culture and all the odd sites share another culture which is incompatible with the one at even sites always fixates .
also , we say that fluctuation / fixation occurs for a given pair of parameters if the one - dimensional system with these parameters fluctuates / fixates when starting from the distribution @xmath24 in which the states of the cultural features within each vertex and among different vertices are independent and uniformly distributed .
we also point out that neither fluctuation implies clustering nor fixation excludes clustering in general .
indeed , the voter model in dimensions larger than or equal to three for which coexistence occurs is an example of spin system that fluctuates but does not cluster while the biased voter model @xcite is an example of spin system that fixates and clusters . in spite of these counter - examples
, we conjecture that fluctuation implies clustering and fixation excludes clustering for the one - dimensional axelrod model starting from the distribution @xmath24 .
we now give a brief review of the previous results about the one - dimensional axelrod model and state the new results proved in this article .
since two neighbors are more likely to interact as the number of cultural features increases and the number of states per feature decreases , one expects the phase transition between the fluctuation / clustering regime and the fixation / no clustering regime to be an increasing function in the @xmath0-@xmath1 plane .
the numerical simulations together with the mean - field approximation of @xcite suggest that the system starting from @xmath24 : * exhibits consensus ( clustering ) when @xmath25 and * gets trapped in a highly fragmented configuration ( no clustering ) when @xmath26 .
looking now at analytical results , the first result in @xcite states that the one - dimensional , two - feature , two - state axelrod model clusters .
the second result deals with the system on a large but finite interval , and indicates that , for a certain subset of the parameter region , the system gets trapped in a random configuration in which the expected number of cultural domains scales like the number of vertices .
this strongly suggests fixation of the infinite system in this parameter region , which we prove in this paper .
shortly after , lanchier and schweinsberg @xcite realized that the analysis of the axelrod model can be greatly simplified using a coupling to translate problems about the model into problems about a certain system of random walks . to visualize this coupling ,
think of each spatial configuration as a @xmath1-coloring of the set @xmath27 and @xmath28 for all @xmath29 and all cultural features @xmath7 .
we call @xmath30 a blockade when it contains @xmath0 particles , or equivalently when the two individuals on each side of @xmath30 completely disagree . when the number of states per feature @xmath31 , lanchier and schweinsberg @xcite proved that construction ( [ eqcoupling ] ) induces a system of annihilating symmetric random walks that has a certain site recurrence property , which is equivalent to fluctuation of the axelrod model , when starting from @xmath24 . from this property
, they also deduced extinction of the blockades and clustering , thus extending the first result of @xcite to the model with two states per feature and any number of features .
in contrast , the present paper deals with the fixation part of the conjecture and extends the second result of @xcite by again using the random walk representation induced by ( [ eqcoupling ] ) .
the first step is to prove that , for all values of the parameters , construction ( [ eqcoupling ] ) induces a system of random walks in which collisions result independently in either annihilation or coalescence with some specific probabilities .
coalescing events only occur when the number of states @xmath32 .
this is then combined with large deviation estimates for the initial distribution of particles to obtain survival of the blockades when starting from @xmath24 in the parameter region described in the second result of @xcite .
this not only implies fixation of the infinite system , but also excludes clustering so the system gets trapped in a highly fragmented configuration . [ thfixation - general ] assume that @xmath33 then , fixation ( [ eqfixation ] ) occurs and clustering ( [ eqclustering ] ) does not occur .
-@xmath1 plane .
the diagram on the left - hand side is simply an enlargement of the diagram on the right - hand side that focuses on small parameters .
the continuous straight line with equation @xmath34 is the transition curve conjectured in @xcite .
the set of crosses is the set of parameters for which the conjecture has been proved analytically : the vertical line of crosses on the left - hand side of the diagrams is the set of parameters for which fluctuation and clustering have been proved in @xcite while the triangular set of crosses is the set of parameters such that @xmath35 for which fixation is proved in theorem [ thfixation - general ] .
the dashed line is the straight line with equation @xmath36 where the slope @xmath37 is such that @xmath38 . ]
interestingly , though the second result in @xcite relies on a coupling between the axelrod model and a certain urn problem along with some combinatorial techniques that strongly differ from the techniques in our proof , both approaches lead to the same sufficient condition ( [ eqfixation - general ] ) .
the set of parameters described implicitly in condition ( [ eqfixation - general ] ) corresponds to the triangular set of crosses in the two diagrams of figure [ figdiagram ] , which we obtained using a computer program .
the picture suggests that this parameter region is ( almost ) equal to the set of parameters below a certain straight line going through the origin . to find the asymptotic slope ,
observe that if @xmath36 , then @xmath39 in other respects , if @xmath3 , then we have @xmath40 from which we deduce that @xmath41 this proves that the condition in the theorem holds for @xmath36 and so all @xmath2 since @xmath42 is decreasing with respect to its second variable .
in particular , fixation occurs whenever @xmath43 see figure [ figdiagram ] for a picture of the straight line with equation @xmath36 .
finally , though @xmath44 and therefore theorem [ thfixation - general ] does not imply fixation for the two - feature three - state axelrod model , our approach can be improved to also obtain fixation in this case .
[ thfixation-2 ] the conclusion of theorem [ thfixation - general ] holds whenever @xmath45 and @xmath46 .
note that this fixation result is sharp since the first result in @xcite gives fluctuation and clustering of the two - feature two - state axelrod model in one dimension .
in particular , the two - feature model fixates if and only if the number of states per feature @xmath32 . to conclude , we note that , in contrast with the techniques introduced in @xcite that heavily relies on the fact that the system starts from @xmath24 , our proof of theorem [ thfixation - general ] easily extends to show that , starting from more general product measures , the one - dimensional system fixates under a certain assumption stronger than ( [ eqfixation - general ] ) .
however , the estimates of lemmas [ lemoutcome ] and [ lemgeneral ] , and consequently the condition for fixation , in this more general context become very messy while the proof does not bring any new interesting argument . therefore , we focus for simplicity on the most natural initial distribution @xmath24 .
as pointed out in @xcite , one key to understanding the axelrod model is to keep track of the disagreements between neighbors rather than the actual set of opinions of each individual . when the number of states per feature @xmath31 , this results in a collection of nonindependent systems of annihilating symmetric random walks .
lanchier and schweinsberg @xcite have recently studied these systems of random walks in detail and deduced from their analysis that the two - state axelrod model clusters in one dimension .
when the number of states per feature is larger than two , these systems are more complicated because each collision between two random walks can result in either both random walks annihilating or both random walks coalescing . in this section , we recall the connection between the axelrod model and systems of symmetric random walks , and complete the construction given in @xcite to also include the case @xmath32 in which coalescing events take place . to begin with , we think of each edge of the graph as having @xmath0 levels , and place a particle on an edge at level @xmath7 if and only if the two individuals that this edge connects disagree on their @xmath7th feature .
more precisely , we define the process @xmath47 and place a particle at site @xmath48 at level @xmath7 whenever @xmath49 . to describe this system ,
it is convenient to also introduce the process that keeps track of the number of particles per site , @xmath50 and to call site @xmath30 a @xmath51-site whenever it contains a total of @xmath51 particles : @xmath52 . to understand the dynamics of these particles ,
the first key is to observe that , since each interaction between two individuals is equally likely to affect the culture of any of these two individuals , each particle moves one unit to the right or one unit to the left with equal probability one half .
because the rate at which two neighbors interact is proportional to the number of cultural features they have in common , a particle at @xmath53 jumps at a rate that depends on the total number of particles located at site @xmath30 , which induces systems of particles which are not independent .
more precisely , since two adjacent vertices that disagree on exactly @xmath51 of their features , and therefore are connected by an edge that contains a pile of @xmath51 particles , interact at rate @xmath54 , the fraction of features they share , conditional on the event that @xmath30 is a @xmath51-site , each particle at site @xmath30 jumps at rate @xmath55 which represents the rate at which both vertices interact times the probability that any of the @xmath51 particles is the one selected to jump . motivated by ( [ eqrate ] )
, the particles at site @xmath30 are said to be active if the site has less than @xmath0 particles , and frozen if the site has @xmath0 particles , in which case we call @xmath30 a blockade .
to complete the construction of these systems of random walks , the last step is to understand the outcome of a collision between two particles .
assume that @xmath53 and @xmath56 are occupied at time @xmath57 and that the particle at @xmath53 jumps one unit to the right at time @xmath58 , an event that we call a collision and that we denote by @xmath59 this happens when the individual at @xmath60 disagrees with her two nearest neighbors on her @xmath7th feature at time @xmath57 and imitates the @xmath7th feature of her left neighbor at time @xmath58 .
this collision results in two possible outcomes .
if the individuals at @xmath9 and @xmath61 agree on their @xmath7th feature just after the update , or equivalently the individuals at @xmath62 and @xmath61 agree on their @xmath7th feature just before the update , then @xmath56 becomes empty so both particles annihilate , which we write @xmath63 on the other hand , if the individuals at @xmath9 and @xmath61 still disagree on their @xmath7th feature after the update , then @xmath56 is occupied at time @xmath58 so both particles coalesce , which we write @xmath64 we refer to figure [ figparticles ] for an illustration of the coupling between the four - feature , three - state axelrod model and systems of annihilating - coalescing random walks .
each particle is represented by a cross and the three possible states by the colors black , grey and white . in our example
, there are two jumps resulting in two collisions : an annihilating event then a coalescing event .
we also refer the reader to figure [ figwalks ] for simulation pictures of the systems of random walks when @xmath65 .
lanchier and schweinsberg @xcite observed that , when @xmath31 , random walks can only annihilate , which was the key to proving clustering .
this is due to the fact that , in a simplistic world where there are only two possible alternatives for each cultural feature , two individuals who disagree with a third one must agree . in our context , the individuals at @xmath62 and @xmath61 must agree just before the update when @xmath31 , which results in an annihilating event .
in contrast , when the number of states per feature is larger , the three consecutive vertices may have three different views on their @xmath7th cultural feature , which results in a coalescing event .
we point out that , since the system of random walks collects all the times at which pairs of neighbors interact , the knowledge of the initial configuration of the axelrod model and the system of random walks up to time @xmath58 allows us to re - construct the axelrod model up to time @xmath58 regardless of the value of the parameters .
there is , however , a crucial difference depending on the number of states .
when @xmath31 , collisions always result in annihilating events , so knowing the configuration of the axelrod model is unimportant in determining the evolution of the random walks .
in contrast , when @xmath32 , whether a collision results in a coalescing or an annihilating event depends on the configuration of the axelrod model just before the time of the collision .
the key to all our results is that , in spite of this dependency , collisions result independently in either an annihilating event or a coalescing event with some fixed probabilities .
in particular , the outcome of a collision is independent of the past of the system of random walks though it is not independent of the past of the axelrod model itself . to prove this result , we need to construct the one - dimensional process graphically from a percolation structure and then define active paths which basically keep track of the descendants of the initial opinions .
first , we consider the following collections of independent poisson processes and random variables : for each pair of vertex and feature @xmath66 : * we let @xmath67 be a rate one poisson process ; * we denote by @xmath68 its @xmath69th arrival time : @xmath70 ; * we let @xmath71 be a collection of independent bernoulli variables with @xmath72 * and we let @xmath73 be a collection of independent @xmath74 .
the axelrod model is then constructed as follows . at time @xmath75 , we draw an arrow labeled @xmath7 from vertex @xmath9 to vertex @xmath76 to indicate that if @xmath77 then the individual at vertex @xmath10 imitates the @xmath7th feature of the individual at vertex @xmath9 . in particular , as indicated in ( [ eqrate ] ) , the rate at which the imitation occurs is equal to one half times the fraction of cultural features both vertices have in common divided by the number of features for which both vertices disagree , which indeed produces the local transition rates of the axelrod model .
the graphical representation defines a random graph structure , also called percolation structure , from which the process starting from any initial configuration can be constructed by induction based on an argument due to harris @xcite .
each arrow in this percolation structure is said to be active if condition ( [ eqactive ] ) is satisfied .
note that whether an arrow is active or not depends on the initial configuration , and that the fact that an @xmath7-arrow from vertex @xmath9 to vertex @xmath10 at time @xmath58 is active implies that the @xmath7th feature of @xmath10 must be equal to the @xmath7th feature of @xmath9 at time @xmath58 . we say that there is an active @xmath7-path from @xmath78 to @xmath79
whenever there are sequences of times and vertices @xmath80 such that the following two conditions hold : for @xmath81 , there is an active @xmath7-arrow from @xmath82 to @xmath83 at time @xmath84 . for @xmath85 , there is no active @xmath7-arrow that points at @xmath86 .
we say that there is a generalized active path from @xmath78 to @xmath87 whenever for @xmath81 , there is an active arrow from @xmath82 to @xmath83 at time @xmath84 . later
, we will use the notation @xmath88 and @xmath89 to indicate the existence of an active @xmath7-path and a generalized active path , respectively .
conditions 1 and 2 above imply that @xmath90 moreover , because of the definition of active arrows and simple induction , the @xmath7th cultural feature of vertex @xmath9 at time @xmath58 is equal to the initial value of the @xmath7th cultural feature of @xmath91 , so we call vertex @xmath91 the ancestor of vertex @xmath9 at time @xmath58 for the @xmath7th feature . in contrast , generalized active paths , which can be seen as concatenations of active @xmath7-paths for possibly different values of @xmath7 , do not have such an interpretation , but the concept will be useful later to prove fixation .
[ lemoutcome ] conditional on the realization of the system of random walks until time @xmath57 and the event that @xmath92 at time @xmath58 , we have @xmath93 let @xmath94 .
due to one - dimensional nearest neighbor interactions , active @xmath7-paths can not cross each other , from which we deduce that @xmath95 where @xmath96 denotes the ancestor at time @xmath97 for the @xmath7th feature , that is , @xmath98 moreover , conditional on the event of a collision @xmath99 at time @xmath58 , there is a particle at @xmath53 and a particle at @xmath56 at time @xmath57 , therefore @xmath100 from ( [ eqoutcome-1 ] ) and ( [ eqoutcome-2 ] ) , we deduce that , conditional on @xmath92 at time @xmath58 , @xmath101 in other respects , we have @xmath102 in particular , the outcome either an annihilating event or a coalescing event of a collision at time @xmath58 is independent of the realization of the system of random walks up to time @xmath57 .
moreover , since the initial states are independent and uniformly distributed , the conditional probability of an annihilating event is equal to the conditional probability @xmath103 where @xmath104 are independent uniform random variables over @xmath105 . by conditioning on the possible values of @xmath106 , we obtain that ( [ eqoutcome-3 ] ) is equal to @xmath107 finally , since each collision results in either an annihilating event or a coalescing event , the conditional probability of a coalescing event directly follows .
this completes the proof .
the main objective of this section is to extend a result of @xcite to the axelrod model , and obtain a sufficient condition for fixation which is based on certain properties of the active @xmath7-paths .
[ lemfixation ] for all @xmath108 , let @xmath109 then , the axelrod model fixates whenever @xmath110 extending an idea of bramson and griffeath @xcite and generalizing the technique in @xcite , we set @xmath111 for every cultural feature @xmath7 and define recursively the sequence of stopping times @xmath112 in other words , the stopping time @xmath113 is the @xmath51th time the individual at the origin changes the state of her @xmath7th cultural feature .
also , for each cultural feature @xmath7 , we define the random variables @xmath114 as well as the collection of events @xmath115 see the left - hand side of figure [ figactive - path ] for a schematic illustration of the stopping times @xmath113 and the corresponding vertices @xmath116 .
assumption ( [ eqfixation-1 ] ) together with reflection symmetry implies that , for each cultural feature @xmath7 , the event @xmath117 occurs almost surely for some @xmath118 .
it follows that @xmath119 since the event that the individual at the origin changes her culture infinitely often is also the event that at least one of the events @xmath120 occurs , in view of the previous inequality , in order to establish fixation , it suffices to prove that @xmath121 our proof of ( [ eqfixation-2 ] ) relies on some symmetry properties of the axelrod model that do not hold for the cyclic particle systems considered in @xcite .
first , we let @xmath122 be the set of descendants of @xmath123 at time @xmath58 , and denote by @xmath124 its cardinality . .
dashed lines represent active @xmath7-paths for some @xmath7 whereas the continuous thick line on the right - hand side is a generalized active path as defined in section [ secwalks ] . ] since each interaction between two individuals is equally likely to affect the culture of each of these two individuals , the number of descendants of any given site is a martingale whose expected value is constantly equal to one . in particular
, the martingale convergence theorem implies that @xmath125 therefore , for almost all realizations of the process , the number of descendants of @xmath123 converges to a finite value . since in addition
the number of descendants is an integer - valued process , @xmath126 using that simultaneous updates occur with probability zero , we deduce that the set of descendants inherits the properties of its cardinality in the sense that , with probability one , @xmath127 \\[-8pt ] \nonumber \rho(x , i)&:= & \inf \bigl\{t > 0 \dvtx i_t ( x , i ) = i_{\infty } ( x , i ) \bigr\ } < \infty,\end{aligned}\ ] ] where , due to one - dimensional nearest neighbor interactions , @xmath128 is a random interval which is almost surely finite . to conclude
, we simply observe that , conditional on @xmath117 , the last time the individual at the origin changes the state of her @xmath7th cultural feature is at most equal to the largest of the stopping times @xmath129 for @xmath130 from which it follows that @xmath131 according to ( [ eqfixation-3 ] ) .
this proves ( [ eqfixation-2 ] ) and therefore the lemma .
in view of lemma [ lemfixation ] , in order to prove fixation , it suffices to show that the probability of the event in equation ( [ eqfixation-1 ] ) , that we denote by @xmath132 , tends to zero as @xmath133 .
the first step is to extend the construction proposed by bramson and griffeath @xcite to the axelrod model , the main difficulty being that two active paths at different levels can cross each other .
let @xmath134 be the first time an active @xmath7-path for some @xmath135 that originates from @xmath136 hits the origin , and observe that @xmath137 from which it follows that @xmath138 denote by @xmath139 the initial position of this active path .
also , we set @xmath140 \\[-8pt ] \nonumber z_+ & : = & \max \bigl\{z \in{\mathbb{z}}\dvtx ( z , 0 ) \leadsto(0 , \sigma ) \mbox { for some } \sigma
< \tau\bigr\ } \geq 0\end{aligned}\ ] ] and define @xmath141 .
we point out that @xmath142 in general since vertex @xmath143 is defined from the set of active @xmath7-paths whereas vertex @xmath144 is defined from generalized active paths that are concatenations of active @xmath7-paths with different values of @xmath7 .
see the right - hand side of figure [ figactive - path ] for an illustration where the two vertices are different . now , note that each blockade which is initially in the interval @xmath145 must have been destroyed , that is , turned into a set of @xmath146 active particles through the annihilation of one of the particles that constitute the blockade , by time @xmath134 .
moreover , active particles initially outside the interval @xmath145 can not jump inside the space
time region delimited by the two generalized active paths implicitly defined in ( [ eqpaths ] ) . indeed , assuming that such particles exist would contradict either the minimality of @xmath144 or the maximality of @xmath147 .
in particular , on the event @xmath132 , all the blockades initially in @xmath145 must have been destroyed before time @xmath134 by either active particles initially in @xmath145 or active particles that result from these blockade destructions . to estimate the probability of this last event
, we first give a weight of @xmath148 to each particle initially active by setting @xmath149 to define @xmath150 when @xmath30 is initially occupied by a blockade , we observe that by lemma [ lemoutcome ] the number of collisions required to break a blockade is geometric with mean @xmath151 . moreover , each blockade destruction results in a total of @xmath146 active particles .
therefore , we set @xmath152 where @xmath153 are independent geometric random variables with mean @xmath151 .
the fact that @xmath132 occurs only if all the blockades initially in @xmath145 are destroyed by active particles initially in @xmath145 or active particles resulting from these blockade destructions , can then be written as @xmath154 \\[-8pt ] \nonumber & \subset & \biggl\{\sum _ { u = l}^r \phi(u ) \leq0 \mbox { for some $ l < - n$ and some $ r \geq0 $ } \biggr\}.\end{aligned}\ ] ] to understand the first inclusion , simply observe that the sum of the @xmath150 is equal to the number of collisions required to break all the blockades minus the total number of active particles initially in the interval @xmath145 or created from the destruction of blockades initially in @xmath145 . since the number of collisions is bounded by the number of such active particles , all the blockades initially in @xmath145 can only be destroyed if the number of such active particles exceeds the number of collisions required , which gives the first inclusion .
the second inclusion simply follows from the fact that @xmath155 the expression of @xmath156 can be understood heuristically as follows : since the @xmath150 are independent , one expects that fixation occurs if @xmath157 . but
@xmath158 which , since @xmath159 , is precisely equal to @xmath156 . to deduce rigorously fixation from the positiveness of the expected value , which is done in the next two lemmas , we now prove large deviation estimates for @xmath132 .
the first of these two lemmas will be used in the proof of the second one to show that the total number of collisions required to break all the blockades in a large interval does not deviate too much from its expected value .
[ lemgeometric ] let @xmath160 be an infinite sequence of independent geometric random variables with the same parameter @xmath161 .
then , for all @xmath162 , there exists @xmath163 such that @xmath164 let @xmath165 for all @xmath166 . since , in a sequence of independent bernoulli trials with success probability @xmath161 , the event that the @xmath167th success occurs at step @xmath69 is included in the event that @xmath167 successes occur in the first @xmath69 steps , we have @xmath168 letting @xmath169 denote the integer part of @xmath170
, we deduce that @xmath171 since large deviation estimates for the binomial distribution imply that @xmath172 for a suitable constant @xmath173 , the result follows .
[ lemgeneral ] let @xmath174 and assume that @xmath175 .
then @xmath176 for a suitable constant @xmath177 and all @xmath118 sufficiently large . to begin with
, we define @xmath178 since the random variables @xmath179 , @xmath48 , are independent , standard large deviation estimates for the binomial distribution imply that for all @xmath162 there exists @xmath180 such that @xmath181 \\[-8pt ] \eqntext{\mbox{for all } i = 0 , 1 , \ldots , f,}\end{aligned}\ ] ] where @xmath182 with @xmath183 .
the expression for @xmath184 follows from the fact that initially each level of each site is independently occupied with probability @xmath185 , which implies that the @xmath179 are independent binomial random variables .
let @xmath186 be the event that @xmath187 then , there exists a constant @xmath188 such that , on the event @xmath186 , @xmath189 in particular , letting @xmath167 be the integer part of @xmath190 , we have @xmath191 now , since @xmath175 , there exists @xmath162 small such that @xmath192 from which we deduce , also using ( [ eqgeneral-2 ] ) and lemma [ lemgeometric ] , that @xmath193 for all @xmath167 sufficiently large . combining ( [ eqgeneral-1 ] ) and ( [ eqgeneral-3 ] ) , we obtain @xmath194 for all @xmath118 sufficiently large . using the inclusion in ( [ eqinclusion ] ) and lemma [ lemgeneral ]
, we deduce @xmath195 this , together with lemma [ lemfixation ] , implies fixation whenever @xmath175 .
to begin with , note that , when @xmath196 and @xmath46 , we have @xmath197 for the comparison function @xmath150 defined in the previous section . in particular , to find a good enough upper bound for the probability of @xmath132 in the case @xmath196 and @xmath46 , one needs to define a new comparison function that also takes into account additional events that promote fixation , such as collisions between active particles and blockade formations . recall that in the comparison function of section [ secfixation - general ] , each particle which is initially active is assigned a weight of @xmath148 , which corresponds to the worst case scenario in which the active particle hits a blockade .
however , each active particle can also hit another active particle or form a new blockade with another active particle .
more precisely , there are four possible outcomes for each active particle : if the active particle hits a blockade , it is assigned a weight of @xmath148 .
if the active particle coalesces with another active particle , then at most one collision with a blockade can result from this pair of particles so the pair is assigned a total weight of @xmath148 ; that is , each particle of the pair is individually assigned a weight of @xmath198 .
if the active particle annihilates with another active particle , then no collision with a blockade can result from this pair so each active particle that annihilates with another active particle is assigned a weight of 0 . if the active particle forms a blockade with another active particle , then following the same approach as in the previous section the pair is assigned a total weight equal to @xmath148 plus a geometric random variable with mean @xmath151 . in view of cases 24 above , the weight of an active particle that either collides with another active particle or forms a blockade with another active particle is at least @xmath198 , and
therefore we define a new comparison function , again denoted by @xmath199 , as follows : @xmath200 where the random variables @xmath153 are again independent geometric random variables with the same expected value @xmath201 . the value of @xmath150 when @xmath202 is the same as in the previous section whereas we distinguish between active particles that satisfy case 1 or cases 24 above . the same reasoning and construction as in section [ secfixation - general ] again
imply that @xmath203 for this new comparison function . to prove that the probability of the event on the right - hand side converges to zero as @xmath133 , we follow the same strategy as for lemma [ lemgeneral ] but also find a lower bound for the probability that a particle initially active either collides with another active particle or forms a blockade with another active particle , which is done in the next lemma . [ lemdeviation ] assume that @xmath196 and @xmath204 .
then , there exists @xmath205 such that @xmath206 where @xmath174 as in lemma [ lemgeneral ] .
the first step is to find a lower bound for the initial number of active particles that will either collide or form a blockade with another active particle .
to do so , we introduce the following definition : an active particle initially at site @xmath48 is said to be a good particle if @xmath207 \\[-8pt ] \eqntext{\mbox{where } \{u , v \ } = \ { 2n - 1/2 , 2n + 1/2 \ } \mbox { for some } n \in{\mathbb{z}}.}\end{aligned}\ ] ] in other words , we partition the lattice @xmath208 into countably many pairs of adjacent sites , and call an active particle at time 0 a good particle if the other site of the pair is initially occupied by an active particle as well . an active particle which is not good
is called a bad particle . since initially each level of each site is independently occupied with probability @xmath185 , the variables @xmath179 are independent binomial random variables , so for @xmath209 as in ( [ eqdeviation-1 ] ) we have @xmath210 where @xmath211 .
similarly , we have @xmath212 since in addition the events that nonoverlapping pairs of adjacent sites are initially occupied by two good particles , or one bad particle , or one blockade , or one bad particle and one blockade , or two blockades are independent , standard large deviation estimates for the binomial distribution imply that there exists a positive constant @xmath213 such that @xmath214 where @xmath215 and @xmath216 denote respectively the initial number of good particles , the initial number of bad particles and the initial number of blockades in the interval @xmath217 . to estimate the probability that a pair of good particles collide or form a blockade
, we first observe that , when there are only two features , the graphical representation of the axelrod model simplifies as follows : for each pair of neighbors @xmath218 , draw an arrow @xmath219 at the times of a poisson process with intensity one fourth , which is equal to half of the rate at which neighbors who agree on one cultural feature interact .
if the two neighbors agree on exactly one cultural feature at the time of the interaction then the culture of the individual at vertex @xmath10 becomes the same as the culture of the individual at vertex @xmath9 . in this graphical representation , there are exactly six possible arrows that may affect the system of random walks at the pair of sites @xmath220 , namely @xmath221 the event that one of the two arrows in the first line of ( [ eqarrows ] ) appears before any of the four other ones occurs with probability two ( arrows ) over six ( arrows ) = 1/3 , and on the intersection of this event and the event that there is initially a pair of good particles at @xmath222 , the two particles either collide or form a blockade .
moreover , the event that one of the two arrows in the first line appears first only depends on the realization of the graphical representation in @xmath223 in particular , parts of the graphical representation associated with nonadjacent pairs do not intersect which , by independence of the poisson processes , implies that the events that the two arrows in the first line of ( [ eqarrows ] ) appears before any of the other ones are independent for nonadjacent pairs .
it follows that the initial number @xmath224 of good particles in @xmath217 that either collide or form a blockade is stochastically larger than a binomial random variable with @xmath225 trials and success probability one third .
large deviation estimates for the binomial distribution then imply that @xmath226 for a suitable constant @xmath227 .
now , let @xmath186 be the event that @xmath228 and observe that there exists a constant @xmath188 such that , on the event @xmath186 , @xmath229 in particular , letting @xmath167 be the integer part of @xmath230 , we have @xmath231 in other respects , recalling the definition of @xmath232 for @xmath233 , we have @xmath234 which , recalling the definition of @xmath235 , is equal to @xmath236 for all @xmath204 . in particular , there exists @xmath162 small such that @xmath237 since @xmath238 , the previous estimate , ( [ eqdeviation-4 ] ) and lemma [ lemgeometric ] imply that @xmath239 for all @xmath167 sufficiently large . combining ( [ eqdeviation-2 ] ) , ( [ eqdeviation-3 ] ) and ( [ eqdeviation-5 ] )
, we obtain @xmath240 for all @xmath118 sufficiently large , which completes the proof . as in the previous section , ( [ eqinclusion-2 ] ) and lemma [ lemdeviation ]
imply that @xmath241 which , together with lemma [ lemfixation ] , implies fixation when @xmath242 and @xmath46 .
the authors would like to thank an anonymous referee for her / his careful reading of the proofs and suggestions to improve the clarity of the paper , and for pointing out a mistake in a preliminary version of the proof of lemma [ lemdeviation ] . | the axelrod model is a spatial stochastic model for the dynamics of cultures which includes two important social factors : social influence , the tendency of individuals to become more similar when they interact , and homophily , the tendency of individuals to interact more frequently with individuals who are more similar .
each vertex of the interaction network is characterized by its culture , a vector of @xmath0 cultural features that can each assumes @xmath1 different states .
pairs of neighbors interact at a rate proportional to the number of cultural features they have in common , which results in the interacting pair having one more cultural feature in common . in this article
, we continue the analysis of the axelrod model initiated by the first author by proving that the one - dimensional system fixates when @xmath2 where the slope satisfies the equation @xmath3 .
in addition , we show that the two - feature model with at least three states fixates . this last result is sharp since it is known from previous works that the one - dimensional two - feature two - state axelrod model clusters . |
the existence of competing hamiltonian terms , e.g. between single - ion anisotropy and spin - spin interactions , or of competing spin - spin interactions , e.g. between next - nearest neighbours , often leads to a large ground - state degeneracy since external agents such as temperature may not be able to simultaneously minimize competing energy components @xcite . in this instance
, novel ground states such as spin liquid , spin ice , cooperative paramagnetism or magnetic coulomb phase based on magnetic monopole excitations may emerge in frustrated magnets , providing an excellent testing ground for approximations and theories @xcite . geometric frustration ,
e.g. in edge - sharing tetrahedra , corner - sharing spinels , or triangular kagom@xmath13 and pyrochlore lattices @xcite , often results in anomalous magnetic properties in lanthanide - based magnetic systems and prevents the relevant magnetic ions from ordering in the usual long - range fashion at low temperatures even far below the energy scale of the individual spin - spin pair interactions . consequently , structural distortion , short - ranged magnetic ordering , or noncollinear magnetic structures may result to release or to reflect the frustration .
this is normally accompanied by a reduction of the ordered moment from the corresponding theoretical saturation value .
the intensively - investigated titanium pyrochlore compounds @xmath12ti@xmath0o@xmath14 @xmath15 y and rare earth ) have two @xmath16 sites ( 16__c _ _ and 16__d _ _ ) in the @xmath17 space group and thus form two sublattices of corner - sharing tetrahedra .
the connections of magnetic @xmath16 sublattice lead to intense geometric frustration on one or both of them in case of an antiferromagnetic ( afm ) next - neighbour ( nn ) exchange interaction @xcite .
in addition , the frustration is also expected even if the nn exchange interaction is ferromagnetic ( fm ) ( e.g. , @xmath18 ho and dy ) , taking into account strong uniaxial anisotropy .
when @xmath18 yb , gd and er , magnetic order appears at transition temperatures of 0.25 , 0.97 and 1.25 k @xcite , respectively .
the compounds of ho@xmath0ti@xmath0o@xmath14 and dy@xmath0ti@xmath0o@xmath14 display a novel spin - ice ground state @xcite without magnetic phase transition to a long - range ordered spin state even down to 50 mk .
tb@xmath0ti@xmath0o@xmath14 does not order even at 70 mk , forming a spin - liquid ground state @xcite .
the new family of frustrated sr__re__@xmath0o@xmath1 compounds adopt an orthorhombic structure @xcite accommodating two inequivalent atomic sites for the @xmath16 ions ( @xmath19 and @xmath20 ) , too .
these compounds were first synthesized in 1967 @xcite .
a subsequent study of polycrystalline sr__re__@xmath0o@xmath1 samples @xcite reveals that the @xmath16 ions are frustrated magnetically .
this is based on observations that the short - ranged magnetic ordering of sr__re__@xmath0o@xmath1 ( @xmath16 = dy , ho , and er ) persists down to @xmath5 1.5 k , and no magnetic ordering was observed for sr__re__@xmath0o@xmath1 ( @xmath16 = tm and yb ) , which was attributed to the low magnetic moments of the heavy lanthanides tm and yb .
however , a recent neutron - scattering study of a sryb@xmath0o@xmath1 single crystal concluded that the compound undergoes an afm phase transition to a long - range commensurate noncollinear spin structure at @xmath21 = 0.9 k. both the yb1 and yb2 moments order , with a reduction of just the ordered yb2 moment ( @xmath5 2.17 @xmath8 at 30 mk ) from the theoretical full ionic value ( 4 @xmath8 ) @xcite .
therefore , polycrystalline @xcite and single - crystalline @xcite sryb@xmath0o@xmath1 studies give different results .
a similar discrepancy was also observed in the srho@xmath0o@xmath1 compound . on the one hand
, neutron scattering of the polycrystalline srho@xmath0o@xmath1 sample reveals a coexistence of the long - range two - dimensional ( 2d ) and the short - range magnetic orders @xcite . on the other hand ,
single crystal neutron - scattering study of srho@xmath0o@xmath1 demonstrates only one - dimensional ( 1d)-like diffuse scattering @xcite .
in addition , a srdy@xmath0o@xmath1 single crystal shows only weak diffuse magnetic scattering even down to @xmath5 20 mk @xcite . to shed light on the nature of the frustration in family of sr@xmath12o@xmath1 compounds , it is crucial to solve detailed spin structures and determine the relationship between the size of ordered moments and their corresponding magnetic @xmath16 ions . note that the theoretical saturation magnetic moment of the tm@xmath10 ion is 7 @xmath8 which is much larger than that ( 4 @xmath8 ) of the yb@xmath10 ion that indeed orders magnetically in sryb@xmath0o@xmath1 @xcite .
it is therefore necessary to explore the spin state of tm@xmath10 ions in a single - crystal srtm@xmath0o@xmath1 sample . in this paper
, we report on the first single - crystal study of the interesting magnetic behaviour of srtm@xmath0o@xmath1 .
the compound displays a negative paramagnetic ( pm ) curie temperature which suggests afm coupling .
however , polarized and unpolarized neutron - scattering investigations show no sign of any types of magnetic ordering , even down to @xmath5 65 mk , and the finding is consistent with field - dependent magnetization measurements .
this is quite different from the other members in the sr@xmath12o@xmath1 family with _ re _ = gd , tb , dy , ho , er , and yb @xcite .
lcl + 4@xmath22 ion & & tm@xmath10 + 4@xmath23 & & 12 + @xmath24 & & 1 + @xmath25 & & 5 + @xmath26 ( hunds rule for free tm@xmath10 ) & & 6 + @xmath27 & & 7/6 + @xmath28 & & @xmath9 + @xmath29 @xmath30 & & @xmath5 7.56 + @xmath31 @xmath30 & & 7.00 + @xmath32 ( 580 - 880 k , 1.5 t ) @xmath30 & & 7.69(1 ) + @xmath33 ( 2 k , 9 t ) @xmath30 & & 2.60(1 ) + @xmath34 ( 580 - 880 k , 1.5 t ) ( k ) & & -41.6(2 ) + @xmath35 ( eq .
[ br ] ) & & 0.504(1 ) + the synthesis of srtm@xmath0o@xmath1 samples is similar to that reported in ref .
the chemical stoichiometry of the studied single crystal was quantitatively estimated by inductively coupled plasma with optical emission spectroscopy ( icp - oes ) analysis .
the electrical resistivity of a bar - shaped single crystal was measured by standard dc four - probe technique .
the dc magnetization was measured as a function of temperature ( from @xmath5 580 to 880 k ) at 1.5 t , and as a function of applied magnetic field ( up to 9 t ) at 2 , 40 and 80 k , using a commercial physical property measurement system ( ppms ) .
the field - dependent magnetization was acquired after cooling in zero magnetic field ( zfc ) .
we cleaved a piece of the srtm@xmath0o@xmath1 single crystal with a mass of @xmath5 0.8 g for the neutron - scattering studies .
this sample was oriented in the @xmath36 scattering plane of the orthorhombic symmetry with the neutron laue diffractometer , orientexpress @xcite , at the institut laue - langevin ( ill ) , grenoble , france .
uniaxial longitudinal neutron polarization analysis was carried out on the d7 ( ill ) diffractometer @xcite with a dilution fridge at a wavelength @xmath37 = 4.8 .
the data were calibrated with vanadium for detector efficiency , and amorphous quartz for polarization corrections .
the corresponding background was measured at @xmath5 1.6 k using an empty sample holder .
measurements were carried out with the neutron polarization , @xmath38 , aligned normal to the scattering plane ( defined as the _
z_-direction ) .
high - resolution neutron powder diffraction ( npd ) patterns were collected on the structure powder diffractometer ( spodi ) @xcite at the frm ii research reactor in garching , germany .
the wavelength was fixed at @xmath37 = 2.54008(2 ) which is obtained from the fullprof @xcite refinement . the rest of the srtm@xmath0o@xmath1 single crystal ( @xmath5 3.91 g ) was gently ground into powder and then sealed in a cylindrical vanadium can . the can was then mounted in a @xmath39he insert with a normal cryorefrigerator .
the detector was scanned with a typical step size of 0.05@xmath40 .
the npd data were refined with the fullprof suite @xcite .
a pseudo - voigt function was used to model the peak profile shape .
the background was refined using a linear interpolation between automatically selected data points .
the scale factor , zero shift , wavelength , peak shape parameters , asymmetry , lattice parameters , atomic positions , isotropic thermal parameter _ b _ as well as the preferred orientation , etc , were all refined .
it is pointed out that the single crystals used for the above measurements are from the same ingot synthesized in one growth . throughout this paper , the wave vector
* * q**@xmath41 ( @xmath42 ) = ( @xmath43 , @xmath44 , @xmath45 ) is defined through ( _ h _ , _ k _ , _ l _ ) = ( @xmath46 , @xmath47 , @xmath48 ) quoted in units of r.l.u . ,
where _ a _ , _ b _ , and _
c _ are the relevant lattice parameters referring to the orthorhombic unit cell .
from our icp - oes measurements , we determine the mole chemical compositions as sr@xmath2tm@xmath3o@xmath4 , indicating that the studied single crystal is stoichiometric within our experimental accuracy .
we tried to measure resistivity of the srtm@xmath0o@xmath1 single crystal with a multimeter at ambient conditions .
unfortunately , potential resistivity is beyond the maximum range ( 10@xmath49 ohm ) of the ohmmeter .
attempts to check the resistivity with our ppms system from 2 to 300 k are also fruitless .
hence , we conclude that srtm@xmath0o@xmath1 is a robust insulator . in this case
, any attempt to understand the anomalous magnetic frustrations in this compound must be based on the model of purely - localized magnetism of the ionic tm@xmath10 ions
. a deeper understanding of the insulating state necessitates theoretical band structure calculations . to extract the intrinsic magnetic properties of srtm@xmath0o@xmath1 , we measured the high - temperature ( @xmath5 580 - 880 k ) magnetization of a small piece of the single crystal ( @xmath5 56.920 mg ) at 1.5 t as shown in fig .
[ m(t ) ] .
the linear increase of the inverse magnetic susceptibility @xmath50 with temperature in the pm state obeys well the molar susceptibility according to the curie - weiss ( cw ) law : @xmath51 where _ c _ is the curie constant , @xmath34 is the pm curie temperature , @xmath52 is the effective pm moment , @xmath53 = 6.022 @xmath54 10@xmath55 mol@xmath42 is avogadro s number , and @xmath56 = 1.38062 @xmath54 10@xmath57
j / k is the boltzmann constant . from the fit to the data with eq .
[ cw ] , shown as the dashed line in fig .
[ m(t ) ] , we derive a measured @xmath7 = 7.69(1 ) @xmath8 per tm@xmath10 ion , which is very close to the expected theoretical value @xmath58 @xmath8 of the ground state @xmath59 determined by the hund s rules .
this is consistent with the fact that the studied single crystal is almost stoichiometric .
we also deduce that @xmath60 k , implying a net afm coupling strength which is stronger than that ( -33.8(6 ) k ) derived from the susceptibility of a polycrystalline sample between 150 and 320 k @xcite .
similar differences in the magnetization between both kinds of samples can also be observed in srer@xmath0o@xmath1 @xcite and srho@xmath0o@xmath1 compounds @xcite .
our study further demonstrates that polycrystalline and single - crystalline srtm@xmath0o@xmath1 samples also differ in their magnetic properties due to different sample preparation procedures : a polycrystal is usually prepared through a solid - state reaction by first calcining the relevant raw materials and then sintering the subsequently pressed mixture , forming a coherent mass by heating but without melting ; on the other hand , to grow a single crystal , the related materials have to be melted by strong enough power or utilizing flux to decrease the melting point of the resultant compound @xcite .
in addition , this discrepancy in properties between samples prepared by different groups has been realized to be a significant issue in cuprate oxide @xcite and colossal magnetoresistance manganite samples @xcite during the past years . to explore the tm@xmath10 magnetic state and possible high magnetic - field effect on the high - level _ j_-multiplets , we measured the field - dependent magnetization as shown in fig .
[ m(h ) ] .
the measured moment size per tm@xmath10 ion at 9 t and 2 k is 2.60(1 ) @xmath8 , @xmath5 37.1% of the corresponding theoretical value 7 @xmath8 , implying a strong single - ion anisotropy due to crystal field effect @xcite .
it is pointed out that neutron scattering based on the bragg law and magnetization measurement detect different magnetic configurations .
it is thus not necessary to differentiate the two tm@xmath10 sites here .
the nonlinear field - dependence at low temperatures would result from pure pm , fm or a magnetic - polaron ( due to possible ionic vacancies , structural defects , etc . )
state @xcite , or a short - ranged afm state taking into account the processes of spin - flop and spin - flip transitions with increasing field @xcite . at 2 k
, there is no appreciable magnetic hysteresis effect .
this , along with the fact that the studied single crystal is nearly stoichiometric , clearly rules out the possibility for a fm or a magnetic - polaron state .
the field - dependent magnetization at low temperatures theoretically obeys a brillouin function modified specifically for a frustrated pm magnet : @xmath61{1.00,1.00,1.00}{a } with } \nonumber \\
b_j(x ) & = & \frac{2j+1}{2j } coth(\frac{2j+1}{2j}x)-\frac{1}{2j } coth(\frac{1}{2j}x ) , \label{br}\end{aligned}\ ] ] where @xmath35 denotes the degree of magnetic frustration , @xmath62 is the theoretical value of the saturation mole moment , @xmath63 is the total angular momentum , @xmath64 .
all these parameters are listed in table [ magp ] .
[ br ] was used to fit the measured data shown in fig .
[ m(h ) ] , giving @xmath65 at 2 k. to all appearances , the fits are good enough to conclude that srtm@xmath0o@xmath1 stays in a pm state at 2 k. figure [ d7 ] shows neutron polarization analysis data in the spin flip ( sf , flipper on ) and non - spin flip ( nsf , flipper off ) channels .
polarized neutron magnetic scattering depends on the direction of the neutron polarization @xmath38 with respect to the scattering vector @xmath66 , and also the direction of the ordered - moments @xmath67 . in our study
, @xmath38 is normal to the scattering plane , and is also parallel to the _ c _ axis @xcite . in this configuration , and assuming that the sample has no fm and ferrimagnetic components ( which is actually true for srtm@xmath0o@xmath1 based on the foregoing remarks ) , the spin - dependent cross - sections may be written as : @xmath68{1.00,1.00,1.00}{a } and } \\ \label{pol2_aw } & & \mathlarger{\mathlarger{\mathlarger{(}}}\frac{d\sigma}{d\omega}\mathlarger{\mathlarger{\mathlarger{)}}}^{^{\texttt{nsf}}}_{_{\texttt{z - off } } } = \frac{1}{3}\mathlarger{\mathlarger{\mathlarger{(}}}\frac{d\sigma}{d\omega}\mathlarger{\mathlarger{\mathlarger{)}}}_{_{\texttt{nsi } } } + \mathlarger{\mathlarger{\mathlarger{(}}}\frac{d\sigma}{d\omega}\mathlarger{\mathlarger{\mathlarger{)}}}_{_{\texttt{nuc } } } + \mathlarger{\mathlarger{\mathlarger{(}}}\frac{d\sigma}{d\omega}\mathlarger{\mathlarger{\mathlarger{)}}}^{^{\|}}_{_{\texttt{mag } } } \text { , } \\ & & \texttt{where } \nonumber \label{pol3_aw } \\ & & \mathlarger{\mathlarger{\mathlarger{(}}}\frac{d\sigma}{d\omega}\mathlarger{\mathlarger{\mathlarger{)}}}_{_{\texttt{mag } } } = \frac{2}{3}\mathlarger{\mathlarger{\mathlarger{(}}}\frac{\gamma_\texttt{n } r_\texttt{e}}{2}\mathlarger{\mathlarger{\mathlarger{)}}}^{^2 } f^2_\texttt{m}(|\textbf{q}| ) g^2_j j(j+1),\end{aligned}\ ] ] where @xmath69 = -1.913 is the neutron gyromagnetic ratio , @xmath70 is the classical electron radius , @xmath71 is the magnetic form factor at the magnetic reciprocal lattice ( * q * ) , @xmath27 and _ j _ are the land@xmath13 factor and total angular momentum , respectively . the cross - section subscript _
nsi _ refers to the nuclear spin incoherent contribution .
the subscript _ nuc _ refers to nuclear and isotopic incoherent contributions .
the magnetic contributions , given by the subscript _ mag _ , are subdivided into two parts : those components of @xmath67 that are parallel to @xmath38 , hence normal to the scattering plane and parallel to the _ c _ axis , give rise to nsf scattering , i.e. , @xmath72{1.00,1.00,1.00}{ab } } { \propto } \text{\textcolor[rgb]{1.00,1.00,1.00}{ab } } { \langle}{\hat{\mu } } \parallel \hat{\textbf{p}}{\rangle}^2;\end{aligned}\ ] ] those components that are perpendicular to both @xmath38 and @xmath66 give rise to sf scattering @xcite , i.e. , @xmath73{1.00,1.00,1.00}{ab } } { \propto } \text{\textcolor[rgb]{1.00,1.00,1.00}{ab } } { \langle}{\hat{\mu } } \perp \hat{\textbf{p } } \times \hat{\textbf{q}}{\rangle}^2.\end{aligned}\ ] ] figure [ d7 ] shows no obvious magnetic scattering observed at @xmath5 65 mk .
the nsf scattering in fig . [ d7](a ) shows bragg peaks that can be indexed with the corresponding orthorhombic structure .
there is some parasitic intensity visible in the sf scattering in fig .
[ d7](b ) due to small errors in correcting for imperfect polarization . however , there are indeed no new bragg peaks , which would imply long - ranged magnetic order with a nonzero propagation vector , nor is there any obvious diffuse scattering from short - range order , within the @xmath74 scattering plane
. we performed a npd analysis as shown in fig .
the large mass of the powder and the considerable counting time ( e.g. @xmath5 26 h at 0.5 k ) enable us to detect any possible magnetic neutron scattering signal , as verified in the studies of refs . @xcite where the same neutron diffractometer spodi was used .
the bragg peaks in the collected npd patterns can be well indexed by the orthorhombic structural model , which rules out possible magnetic ordering with a nonzero propagation wave vector consistent with our polarization analysis presented above .
figure [ str ] schematically depicts the resulting crystal structure ( fig . [ str](a ) ) , local connections of the tmo@xmath11 octahedra ( figs .
[ str](b ) and ( c ) ) , as well as the bent tm@xmath11 honeycomb and its projection to the _ ab _ plane ( figs . [ str](d ) and ( e ) , respectively ) .
the tmo@xmath11 octahedra build up a 3d network in the way of sharing edges between tm1o@xmath11 or tm2o@xmath11 .
the connections between both types of octahedra are their in - plane and spatial corners . by way of example ,
two nearest tm1o@xmath11 octahedra ( fig . [ str](b ) ) share a common o2-o3 bond along the _ c _ axis , and the tm1 and tm2 sites are connected by the o1 or o3 ions ( fig .
[ str](a ) ) . based on the irreducible representation analysis to the lower _
p_-1 symmetry @xcite , we tried all possible magnetic models with a propagation wave vector at * q * = ( 0 , 0 , 0 ) @xcite to analyze the npd data .
one representative fullprof @xcite refinement is displayed in fig .
we find a reasonable refinement only when the tm moments are along the _ c _ axis , although the values of the goodness - of - fit give no appreciable improvement in comparison with those of the refinement with only nuclear structural model ( table [ npdta ] ) .
the resulting tm1 and tm2 moments are + 0.30(21 ) and -0.20(21 ) @xmath8 at 0.5 k , respectively .
this suggests that there is no long - range magnetic ordering at all , consistent with the magnetization characterizations .
we therefore conclude that srtm@xmath0o@xmath1 is a totally - frustrated compound at least at @xmath75 65 mk .
theoretically , both tm@xmath10 and tb@xmath10 ( @xmath76 ) are non - kramers ions with saturation magnetic moments of 7 and 9 @xmath8 , respectively . however , at 0.5 k , srtb@xmath0o@xmath1 displays an incommensurate noncollinear afm structure @xcite with partially - ordered moments 1.92(6 ) @xmath8 ( at the maximum amplitude ) accommodated only at the tb1 site . in contrast , srtm@xmath0o@xmath1 shows no magnetic ordering .
the shortest interatomic tm - tm ( table [ npdta ] ) and tb - tb @xcite distances at 0.5 k are 3.3809 and 3.4523 , respectively , which , in conjunction with the previous remarks , implies that a direct exchange interaction between the magnetic @xmath16 ions in sr@xmath12o@xmath1 plays no role in the potential spin couplings .
the implication is consistent with the fact that unpaired 4__f _ _ electrons are well shielded by the @xmath77 shells as long as a purely ionic model , supported by the resistivity measurements , is assumed .
although some of the @xmath78tm - o - tm bond angles ( e.g. @xmath78tm1-o3-tm1 and @xmath78tm1-o1-tm2 ) that determine the degree of overlap of the tm@xmath10 and o@xmath79 orbitals display some temperature dependence between 0.5 and 4 k ( table [ npdta ] ) , the absence of magnetic ordering rules out indirect magnetic interactions through the mediator of nonmagnetic o@xmath79 ions . the strong spin - orbital interaction of the @xmath16 ions may result in an anisotropic exchange , i.e. the dzyaloshinsky - moriya ( dm ) interaction , which normally requires a breaking of the inversion symmetry of the centrosymmetric _ pnam _ space group . in a study of sryb@xmath0o@xmath1 @xcite ,
forbidden nuclear bragg peaks were indeed observed but finally were demonstrated to be entirely due to contamination from higher - order reflections .
similar weak structurally - forbidden peaks were also present in a neutron - scattering study of a srho@xmath0o@xmath1 single crystal @xcite where their crystallographic origin has not yet been confirmed .
a systematic determination of the temperature - dependent nuclear structural symmetry is a prerequisite to whether or not the dm interaction should be taken into account in the sr@xmath12o@xmath1 family . in our present study , no such kind of forbidden peaks
are observed in srtm@xmath0o@xmath1 , hence the dm interaction is unlikely to play a role in the magnetic exchange .
therefore , we conclude that the magnetic coupling in srtm@xmath0o@xmath1 arises mainly from an anisotropic dipole - dipole interaction which , while very weak , is additionally subjected to crystal field effects resulting from the octahedral environments ( as discussed below ) .
we calculate the average octahedral distortion parameter @xmath80 as defined by : @xmath81 where @xmath82 and @xmath83 are the six tm - o bond lengths along the six crossed directions and the mean tm - o bond length ( table [ npdta ] ) , respectively .
we note that the @xmath80 values of the tm1 and tm2 ions have a similar magnitude to that of the tb2 ion in srtb@xmath0o@xmath1 @xcite . since the magnetic moments associated with these three ions disorder entirely , it is thus reasonable to infer that the magnitude of the @xmath80 value acts as an indicator of whether the @xmath16 ions in sr@xmath12o@xmath1 may order magnetically . in this case , there should exist a critical value for @xmath80 , below which the potential magnetic moments disorder completely , but above which the @xmath16 ions display some form of magnetic ordering .
in other words , to realize a magnetically - ordered state , the spin interaction must exceed a critical value to overcome the single - ion anisotropic energy .
in addition , as temperature decreases from 4 to 0.5 k , the @xmath80 value of the tm2 ion decreases , whereas that of the tm1 ion increases ( table [ npdta ] ) .
this reflects the temperature - dependent behavior of the tb1 and tb2 ions in srtb@xmath0o@xmath1 @xcite .
it is necessary to explore this common tendency in other family members of sr@xmath12o@xmath1 .
llccclccc + _ t _ ( k ) & & & & + @xmath84 ( ) & & 9.9759(1 ) & 11.8150(1 ) & 3.3809(1 ) & & 9.9764(1 ) & 11.8148(1 ) & 3.3811(1 ) + @xmath85 @xmath86 & & 90 & 90 & 90 & & 90 & 90 & 90 + @xmath87 ( @xmath39 ) & & & & + atom & & _ x _ & _ y _ & _ b _
( @xmath88 ) & & _ x _ & _ y _ & _ b _ ( @xmath88 ) + sr & & 0.7536(2 ) & 0.6501(2 ) & 0.65(8 ) & & 0.7529(3 ) & 0.6489(3 ) & 0.54(8 ) + tm1 & & 0.4218(3 ) & 0.1084(3 ) & 0.52(7 ) & & 0.4210(3 ) & 0.1075(3 ) & 0.33(7 ) + tm2 & & 0.4244(3 ) & 0.6133(2 ) & 0.14(7 ) & & 0.4238(3 ) & 0.6125(3 ) & 0.64(8 ) + o1 & & 0.2104(3 ) & 0.1723(2 ) & 0.40(9 ) & & 0.2103(3 ) & 0.1739(3 ) & 0.53(9 ) + o2 & & 0.1236(2 ) & 0.4807(2 ) & 0.70(9 ) & & 0.1256(3 ) & 0.4806(3 ) & 0.40(10 ) + o3 & & 0.5160(3 ) & 0.7840(2 ) & 0.32(8 ) & & 0.5154(3 ) & 0.7832(2 ) & 0.52(9 ) + o4 & & 0.4250(3 ) & 0.4239(2 ) & 0.56(9 ) & & 0.4259(3 ) & 0.4247(2 ) & 0.26(10 ) + @xmath78tm1-o2-tm1 ( @xmath40 ) & & & & + @xmath78tm1-o3-tm1 ( @xmath40 ) & & & & + @xmath78tm1-o1-tm2 ( @xmath40 ) & & & & + @xmath78tm1-o3-tm2 ( @xmath40 ) & & & & + @xmath78tm2-o1-tm2 ( @xmath40 ) & & & & + @xmath78tm2-o4-tm2 ( @xmath40 ) & & & & + @xmath78o1-tm1-o2 ( @xmath40 ) & & & & + @xmath78o2-tm1-o3 ( @xmath40 ) & & & & + @xmath78o3-tm2-o4 ( @xmath40 ) & & & & + @xmath78o1-tm2-o4 ( @xmath40 ) & & & & + tm1-tm1 ( nn , nnn , nnnn ) ( ) & & & & + tm2-tm2 ( nn , nnn , nnnn ) ( ) & & & & + tm1-tm2 ( nn , nnn , nnnn ) ( ) & & & & + tm1-o1 ( @xmath54 1 ) ( ) & & & & + tm1-o2 ( @xmath54 3 ) ( ) & & & & + tm1-o3 ( @xmath54 2 ) ( ) & & & & + tm2-o1 ( @xmath54 2 ) ( ) & & & & + tm2-o3 ( @xmath54 1 ) ( ) & & & & + tm2-o4 ( @xmath54 3 ) ( ) & & & & + @xmath89tm1-o1,2,3@xmath90 ( ) & & & & + @xmath89tm2-o1,3,4@xmath90 ( ) & & & & + @xmath80 ( @xmath91 ) & & & & + @xmath92 & & & & + based on the refined tm - o bond lengths ( table [ npdta ] ) , we deduce two distortion modes for the tm1o@xmath11 and tm2o@xmath11 octahedra as shown in figs .
[ str](b ) and ( c ) , respectively .
it is interesting that both distortion modes are the same as the corresponding ones of the tb1o@xmath11 and tb2o@xmath11 octahedra @xcite , indicating a common feature .
both distortion modes may be ascribed to crystal field anisotropy , which is supported by the observation in srtb@xmath0o@xmath1 @xcite that the partially - ordered tb1 moments qualitatively point to the direction of the stress product imposed on the tb1 ion by the tb1o@xmath11 octahedral distortion .
a quantitative understanding of both distortion modes necessitates a theoretical model to simulate the tension state of the @xmath16-o bonds , and a quantitative determination of the detailed crystal field parameters by inelastic neutron scattering .
two explanations are possible for the refined @xmath65 at 2 k ( eq . [ br ] ) .
firstly , the tm1 and tm2 sites present measurable magnetization , but both sites are nearly half frustrated even in the pm state , and the frustrated spins are in a strong freezing state .
secondly , the measurable magnetization derives from only one of the two tm sites , probably the tm1 site , and the other is frustrated completely even in the pm state at 2 k. although both cases can lead to the slight increase in magnetization from @xmath93 3.6 to @xmath94 6.5 t as shown in fig .
[ m(h ) ] , the later case is more favourable because it resembles the magnetic behaviour of the tb1 and tb2 sites in the magnetically - ordered state of srtb@xmath0o@xmath1 @xcite , and , most importantly , there is no appreciable magnetic relaxation at 2 k after magnetic field has reached even 9 t. the obvious decrease in the measured magnetization above @xmath5 8.3 t at 2 k in contrast to the theoretical brillouin calculation ( fig .
[ m(h ) ] ) may suggest a zeeman splitting effect on the high - level _ j_-multiplets .
the pm curie temperature of single - crystal srtm@xmath0o@xmath1 is @xmath95 k , however , no evidence of magnetic ordering , even short - ranged , appears down to @xmath5 65 mk in our neutron - diffraction studies .
this implies that there exists an extremely strong magnetic frustration for the tm@xmath10 ions in srtm@xmath0o@xmath1 or the potential magnetic ordering is highly gapped by a strong crystal - field anisotropy . as shown in fig .
[ str](d ) , the nn tm1 or tm2 ions are respectively staggered along the _ c _ axis , while the connections between the nnn tm1 or tm2 ions form ladder - like chains in the same direction .
altogether , the nns and nnns of the tm1 or tm2 ions form isosceles triangles for zigzag chains .
these tm chains are arranged to form bent tm@xmath11 honeycombs ( figs .
[ str](d ) and ( e ) ) . in this crystalline environment ,
the similarity of the nn and nnn spin interactions and the low coordination numbers of the tm ions are believed to cause the geometric frustration of the spin lattice .
in addition , potential stronger and competing tm - tm spin interactions , by virtue of the shorter nn bond length compared to that of the tb - tb ions in srtb@xmath0o@xmath1 @xcite , may be insufficient to overcome crystal electric field , which therefore leads to a nonmagnetic ordering state .
it is interesting to compare our results with those from the pyrochlore compounds of @xmath12ti@xmath0o@xmath14 @xmath15 tm and tb ) . in tm@xmath0ti@xmath0o@xmath14 ,
magnetic tm@xmath10 ions occupy one of the two corner - sharing tetrahedral sublattices .
the crystal field effect radically dominates over the magnetic exchange interactions so that the ground state is a virtual crystal - field singlet rather than a short - ranged and frustrated magnetic state , and the lowest - lying crystal - field excitations thus do nt show any appreciable dispersion @xcite .
in addition , the magnetic susceptibility of tm@xmath0ti@xmath0o@xmath14 gets a clear plateau at low temperatures @xcite , whereas that of srtm@xmath0o@xmath1 continuously increases @xcite , implying different origins of the nonmagnetic ordering states of both compounds . although there exists a strong afm interaction based on the value of the curie - weiss temperature ( -19 k ) @xcite , no appreciable long - range magnetic order forms in tb@xmath0ti@xmath0o@xmath14 even down to 50 mk but short - ranged magnetic order
was indeed observed indicative of a spin - liquid ground state @xcite .
the nature of the magnetic ground state and the reason for the absence of long - ranged spin order of tb@xmath0ti@xmath0o@xmath14 still remain elusive , which represents a theoretical puzzle and also a real experimental challenge .
our further analysis of the @xmath96-averaged @xcite neutron scattering intensity in the sf ( eq .
[ pol1_aw ] ) and nsf ( eq . [ pol2_aw ] ) channels indeed indicates the existences of pm neutron scattering and afm coupling in srtm@xmath0o@xmath1 at @xmath5 65 mk .
however , an estimate of the moment size ( eq . [ pol3_aw ] ) is complicated by the presence of strong nuclear spin - incoherent scattering from the copper support on which the sample was mounted .
future experiments using the @xmath97 polarization method at d7 , which will give a clear separation between nuclear coherent , nuclear spin - incoherent , and magnetic cross - sections @xcite , may be attempted .
our present results provide a more mysterious compound for theorists and experimentalists , srtm@xmath0o@xmath1 , which is different from both srdy@xmath0o@xmath1 and tb@xmath0ti@xmath0o@xmath14 due to the absences of both short- and long- ranged magnetic orders . to summarize
, we have demonstrated that there is no long- or short - ranged magnetic ordering in single - crystal srtm@xmath0o@xmath1 even down to @xmath5 65 mk , though the high - temperature magnetization indicates a strong afm coupling .
this is consistent with the field - dependent magnetization measurements where the data can be adequately fit using a modified brillouin function ( eq . [ br ] ) for a frustrated pm state .
we argue that the nonmagnetic ordering state is ascribed to a strong anisotropic crystal field effect .
this kind of magnetic anisotropy may gap potential magnetic ordering , leading to a virtual nonmagnetic ordering state .
we also find some common features for both srtm@xmath0o@xmath1 and srtb@xmath0o@xmath1 single crystals : ( i ) the measured magnetization , even in the pm state , is from just one of the two crystallographic @xmath16 sites ; ( ii ) the same octahedral distortion modes as well as their evolutions with temperature ; ( iii ) the spin couplings are dominated by the dipole - dipole interactions .
these common features observed in both compounds may be vital to a complete understanding of the novel magnetic behaviour of sr@xmath12o@xmath1 .
further explorations with higher magnetic fields and uniaxial pressure along the axes , spin - density measurements in the pm state , and especially inelastic neutron - scattering studies , would be of great interest .
this work at rwth aachen university and j@xmath98lich centre for neutron science jcns outstation at ill was funded by the bmbf under contract no . 05k10pa3 . h.f.l
thanks the sample environment teams at ill and frm ii for expert technical assistances .
h. t. diep , _ frustrated spin systems _ ( world scientific , singapore , 2004 ) .
d. j. p. morris , _ et al_. , science * 326 * , 411 ( 2009 ) .
k. binder , phys .
* 45 * , 811 ( 1980 ) .
a. p. ramirez , g. p. espinosa , and a. s. cooper , phys .
rev . lett . *
64 * , 2070 ( 1990 ) .
h. kawamura , j. phys .
: condens .
matter * 10 * , 4707 ( 1998 ) .
n. p. raju , m. dion , m. j. p. gingras , t. e. mason , and j. e. greedan , phys .
b. * 59 * , 14489 ( 1999 ) .
j. a. hodges , _
et al_. , phys .
lett . * 88 * , 77204 ( 2002 ) .
m. j. harris , s. t. bramwell , p. c. w. holdsworth , and j. d. m. champion , phys .
lett . * 81 * , 4496 ( 1998 ) .
j. s. gardner , phys .
lett . * 82 * , 1012 ( 1999 ) .
j. g. pepin , j. appl .
* 14 * , 70 ( 1981 ) .
t. l. barry and r. roy , j. inorg .
, * 5 * , 1243 ( 1967 ) .
h. karunadasa , _ et al_. , phys .
b. * 71 * , 144414 ( 2005 ) .
d. l. quintero - castro , _
et al_. , phys .
b. * 86 * , 064203 ( 2012 ) .
s. ghosh , _ et al_. , j. phys . : condens .
matter * 23 * , 164203 ( 2011 ) .
o. young , _ et al_. , phys .
b. * 88 * , 024411 ( 2013 ) . t. h. cheffings , m. r. lees , g. balakrishnan , and o. a. petrenko , j. phys . :
matter * 25 * , 256001 ( 2013 ) .
o. a. petrenko , _ et al_. , phys .
b. * 78 * , 184410 ( 2008 ) .
t. j. hayes , o. young , g. balakrishnan , and o. a. petrenko , j. phys .
. jpn . * 81 * , 024708 ( 2012 ) .
li , _ et al_. , phys .
lett . * * , under review , ( 2014 ) .
g. balakrishnan , t. j. hayes , o. a. petrenko , and d. m@xmath99k .
paul , j. phys . : condens .
matter * 21 * , 012202 ( 2009 ) .
b. ouladdiaf , _ et al_. , physica b * 385386 * , 1052 ( 2006 ) . j. r. stewart , _ et al_. , j. appl . cryst . * 42 * , 69 ( 2009 ) .
m. hoelzel , _ et al_. , nuclear instruments & methods in physics research , section a : accelerators , spectrometers , detectors , and associated equipment , * 667 * , 32 ( 2012 ) .
j. rodrguez - carvajal , physica b * 192 * , 55 ( 1993 ) .
haifeng li , _ synthesis of cmr manganites and ordering phenomena in complex transition metal oxides _
( forschungszentrum j@xmath98lich gmbh , j@xmath98lich , 2008 ) .
j. orenstein and a. j. millis , science * 288 * , 468 ( 2000 ) .
li , _ et al_. , j. phys . : condens .
matter * 19 * , 016003 ( 2007 ) .
li , _ et al_. , j. phys . : condens .
matter * 19 * , 176226 ( 2007 ) .
et al_. , eur .
j. b * 67 * , 149 ( 2009 ) .
li , `` a second - order spin - flop transition in collinear two - sublattice antiferromagnets '' , to be submitted , ( 2014 ) .
li , _ et al_. , sci . rep . * 2 * , 750 ( 2012 ) .
m. p. zinkin , m. j. harris , z. tun , r. a. cowley , and b. m. wanklyn , j. phys . : condens .
matter * 8 * , 193 - 197 ( 1996 ) . | we have studied srtm@xmath0o@xmath1 using resistivity , magnetization , and polarized and unpolarized single - crystal and powder neutron - diffraction measurements .
resistivity measurements demonstrate that a nearly stoichiometric sr@xmath2tm@xmath3o@xmath4 single crystal is a robust insulator down to 2 k. high - temperature ( @xmath5 580 - 880 k ) magnetization reveals a net antiferromagnetic coupling with the paramagnetic ( pm ) curie temperature @xmath6 k and a consistent effective pm moment @xmath7 = 7.69(1 ) @xmath8 , to be compared with the theoretical value @xmath5 7.56 @xmath8 of the hund s rule ground state @xmath9 .
magnetic field - dependent magnetization at 2 k agrees well with a modified brillouin function for a non - interacting pm state albeit with a small deviation in the field range of @xmath5 3.6 - 6.5 t and suggests that the magnetization may originate from only one of the two inequivalent tm@xmath10 crystallographic sites .
an appreciable deviation from the brillouin function is visible above @xmath5 8.3 t , indicating that zeeman splitting of the high - level _ j_-multiplets may lead to a new ground state .
our single - crystal polarized neutron scattering at @xmath5 65 mk and powder unpolarized neutron diffraction at @xmath5 0.5 k show no evidence for a long - range magnetic order and even detect no sign of diffuse magnetic neutron scattering .
the data refinements reveal that the two tmo@xmath11 octahedral distortion modes are the same as those of the tbo@xmath11 octahedra in srtb@xmath0o@xmath1 , i.e. , one distortion is stronger than the other one especially at low temperatures , which is attributed to different crystal electric fields for the two inequivalent octahedra .
consequently , we conclude that srtm@xmath0o@xmath1 has no magnetic ordering , neither long - ranged , nor short - ranged , even down to @xmath5 65 mk .
therefore , srtm@xmath0o@xmath1 is a different compound from its brethren in the new family of frustrated sr@xmath12o@xmath1 ( _ re _ = gd , tb , dy , ho , er , and yb ) magnets .
we propose that crystal field anisotropy may dominate over weak dipolar spin interactions in srtm@xmath0o@xmath1 , leading to a virtually nonmagnetic ordering state . |
the formation and evolution of giant molecular clouds ( gmcs ) , the sites where stars form , are thought to be influenced by the local properties of the environment and global galactic structure ( @xcite ; @xcite ; @xcite ; @xcite ) .
although the internal physical properties of molecular clouds have been extensively studied ( @xcite ; henceforth s87 ) , a more detailed understanding of their origin and evolution in different types of environments is needed .
maps with resolution matched to the scale of individual clouds are the key to achieve this goal ; however , such maps currently exist only for a few galaxies outside the local group . due to the coupled evolution between gmcs and their host galaxies , achieving a more complete understanding of molecular cloud formation and evolution
processes represents a key goal for studies of galaxy evolution .
while several mechanisms have been proposed , the predominant mode for molecular cloud formation and evolution in spiral galaxies is not yet clear . in terms of the arm structure
, spiral galaxies can be classified as _ grand - design _
, most of them consisting of two symmetrical spiral arms ; _ multi - armed _ , which show several non - symmetrical arms ; or _ flocculent _ galaxies with multiple shorter arms . for grand - design galaxies ,
the spiral structure is generated by density waves induced by bars or companion galaxies .
the influence of the density waves in the ism in these galaxies has been commonly modeled assuming a long lived , rigidly rotating spiral potential ( @xcite ; @xcite ; @xcite ) . on the other hand , flocculent and multi - armed galaxies
are thought to develop from local gravitational instabilities , although flocculent galaxies can present weak density waves as well ( @xcite ) . in this case ,
numerical simulations model the galactic potential as time dependent structures , which can change on a timescale of @xmath10100 myr ( @xcite ; @xcite ) .
the formation of massive molecular clouds in grand - design galaxies is attributed to the passing of spiral density waves .
a particular feature of this type of galaxy is the formation of spurs perpendicular to the spiral arms due to the shearing of the molecular structures as they leave the arms ( @xcite ; @xcite ) . in their study of the northeastern spiral segment of ic 342 , a grand design galaxy ,
@xcite showed that actively star forming gmcs are more massive , have smaller velocity dispersions and are more gravitationally bound than quiescent gmcs .
they interpreted the tendency for the star - forming gmcs to lie downstream with respect to the quiescent molecular clouds as evidence in favor of cloud a coalescence scenario , where massive molecular clouds ( @xmath11 ) are formed from coagulation of smaller clouds entering the spiral arms , and the star formation is ignited once the excess kinetic energy is dissipated through inelastic collisions of the clouds ( hirota et al .
2011 ) . in another grand design galaxy ,
m51 , @xcite suggested that evolution of gmcs is driven by large scale galactic dynamics , where the most massive molecular complexes ( giant molecular associations , gmas ) are created by coagulation of gmcs in spiral arms .
as gmas leave the arms , they suffer fragmentation by the strong shear motions , which would explain the spur structures observed extending from the arms into the inter - arm regions .
the remaining fragmented gmcs are not disassociated into atomic gas , and they become part of the molecular cloud population observed in inter - arm regions which survive until the passage of another spiral arm . on the other hand , @xcite presented a numerical simulation of flocculent galaxies .
they find that the gaseous spiral structure essentially traces the potential minimum , in contrast to the case of grand - design spirals where an offset between the spiral shock and the potential minimum is expected .
according to the simulations provided by @xcite and @xcite the most massive structures are thought to be produced in regions where collision or merging between spiral arms can occur , yielding an enhanced star formation in those overdense regions , in contrast to the situation proposed for grand - design galaxies where the molecular clouds are formed by agglomeration within spiral arms .
conversely , the formation of inter - arm molecular gas structures are proposed to be the result of local instabilities rather than fragmentation of gma s , and the properties of the clouds are approximately similar across the disk , with no offsets between gas and star formation expected .
high resolution observations of flocculent galaxies have provided evidence in support of this picture .
for instance , @xcite mapped @xmath0 over the southern arm of the flocculent galaxy ngc 5055 .
they found no obvious offset between h@xmath12 and the molecular gas , and that on - arm and inter - arm clouds do not have significant differences in their properties .
resolved gmcs in the milky way have been shown to be in approximate virial equilibrium obeying scaling relations known as larson s laws ( @xcite ) .
these laws establish that velocity dispersion ( or line width ) , size , and luminosity are correlated in the milky way gmcs .
the size - line width relation may reflect turbulent conditions in the molecular ism .
the typical gmc temperature of 1525 k produces thermal co velocity dispersions of only @xmath100.1 @xmath13 , providing evidence of non - thermal supersonic turbulence within the clouds .
nevertheless , whether the turbulence in the molecular ism is internally or externally generated is still unclear (;
the other two larson relations are the luminosity - line width and the luminosity - size relations . in a recent work , @xcite ( henceforth b08 ) present a study of resolved properties of gmcs for a sample of dwarf galaxies along with two disk galaxies of the local group .
they find insignificant differences between milky way gmcs and gmcs in dwarf galaxies , with the latter following approximately the same larson relations as galactic clouds .
@xcite arrived at a similar conclusion by analyzing observations of co surveys for galaxies in the local group . due to sensitivity and resolution limitations , observations of resolved properties of gmcs in galaxies
other than milky way are in an early stage ( @xcite , hereafter dm12 ; @xcite ) .
only a few galaxies have been mapped in co emission across the full extent of the optical disk with sufficient resolution to probe the @xmath10100 pc size scales on which gmcs form .
high resolution co surveys ( @xcite ; @xcite ) have focused primarily on the central regions , whereas complete mapping using single - dish telescopes ( @xcite ; @xcite ) has been obtained at resolutions of 11@xmath14 to 15@xmath14 , corresponding to @xmath10500 pc for nearby spirals , inadequate to resolve gmcs . only for m33 ( bima survey ) and the magellanic clouds ( nanten surveys )
has it been possible to conduct a fairly complete census of gmcs ( @xcite ; @xcite ) . using a combination of @xmath0 and @xmath1 and pursuing a staged approach of starting with large - scale , low resolution maps and following up with high resolution imaging on smaller areas
, carma makes it possible to study the overall distribution of gmcs , as well as the properties of individual gmcs , in galaxies outside the local group .
we have successfully used this observational strategy in the nearby spiral galaxy ngc 6946 . as indicated by the single - dish co map from iram ( @xcite ) ,
this galaxy is extremely rich in molecular gas , with co emission extended to a large fraction of the optical radius .
table [ n6946-prop ] shows the basic parameters of ngc 6946 .
we have chosen ngc 6946 as our target because of its proximity , high co surface brightness , low inclination , and the availability of high - quality datasets at a variety of wavelengths , including hi imaging from things and multiband spitzer imaging from sings ( @xcite ) .
the spiral structure in ngc 6946 appears complex , with k - band images revealing four prominent , asymmetric spiral arms ( @xcite ) .
while the central region of the galaxy has been well - studied in co because of its high co surface brightness ( @xcite ; @xcite ; dm12 ) , the eastern region has not been mapped at high resolution . in this paper
we present observations of the eastern region of ngc 6946 .
firstly , we made observations of @xmath0 using carma covering a region of size 6@xmath66 kpc@xmath15 .
the angular resolution for the @xmath0 map was @xmath10110 pc , which is not sufficient to resolve structures with size scales similar to gmcs ( s87 ; b08 ) .
then , the second step was to make higher resolution @xmath1 observations ( @xmath10 50 pc ) towards the brightest co complexes to study the properties of individual gmcs .
we present our study as follows : in section [ obs ] we describe our observations of ngc 6946 using carma , and we describe the archival data at several wavelengths that we include in our analysis . in section [ cprops ] we summarize the technique used to identify gmcs and to measure their physical properties . in section [ cl - prop ] we present the scaling relations of the cloud properties , and we compare them with previous studies of galactic and extragalactic clouds . in section [ env - eff ]
we discuss whether the properties of the clouds differ between on - arm and inter - arm regions . in section
[ discuss ] we discuss the implications of our results , and in section [ summary ] we summarize the work presented in this paper .
lc morph . & sabcd + r.a .
( j2000)@xmath16 & 20:34:52.3 + decl .
( j2000@xmath16 & 60:09:14 + distance ( mpc)@xmath17 & 5.5 + incl .
( ) @xmath18 & 33 + p.a .
( ) @xmath18 & 243 + + + + [ n6946-prop ]
we performed high spatial resolution observations of @xmath0 and @xmath2 for the eastern part of ngc 6946 from july to august of 2009 .
we used the combined array for research in millimeter - wave astronomy ( carma ) in e and d array configurations , which have baselines of 8.5 - 66 meters , and 11 - 148 meters , respectively .
our goals are to resolve structures close to the typical sizes of gmcs , and to make a detailed study of their properties including sizes , luminosity , velocity dispersion and star formation activity . in order to achieve these goals ,
we have selected a region in the disk with a favorable observing geometry , active star formation , and where the interstellar medium transitions from being molecular gas dominated to atomic gas dominated .
we observed @xmath0 and @xmath2 in a 49-point mosaic area , which covered [email protected] arcmin@xmath15 , corresponding to a physical scale of 6@xmath66 kpc@xmath15 at a distance of 5.5 mpc ( @xcite ) .
the correlator was set to have the @xmath0 line in the upper side band ( usb ) , and the @xmath2 in the lower side band ( lsb ) . by placing two overlapping 31 mhz bands with an offset of 12.9 mhz from the rest frequency line
, we achieved a total velocity coverage of 150 @xmath13 and a channel width of 1.3 @xmath13 .
a 500 mhz wide - band was placed outside both spectral windows for calibration .
we observed every pointing for 30 seconds , yielding a 24.5-minute observation time per cycle . at the beginning of the track we observed 2232 + 117 as the passband calibrator , and at the end of every cycle
, we observed 2038 + 513 as the gain calibrator .
calibration , imaging and deconvolution were performed using standard procedures of the miriad software package . in order to have better angular resolution , we applied a briggs robustness parameter of -1.5 , which yields resolution for @xmath0 clean map of @xmath19 , a @xmath20 of 0.395 k , and a pixel size of 1@xmath14 . for the @xmath2 map , because the signal to noise is worse than @xmath0 , we have applied a briggs robustness parameter of + 0.5 .
the resulting @xmath2 map has a resolution of @xmath5 , @xmath20 of 0.212 k , and a pixel size of 1@xmath14 .
figure [ figure_ngc6946e ] shows the @xmath0 map of the region observed overlaid on a hi image from things ( @xcite ) .
c our observations at 3 mm using carma likely do not resolve the gmcs .
therefore , the second step was to make high resolution observations in d array towards some co complexes to study the properties of individual gmcs using @xmath1 .
the eleven brightest regions in @xmath0 integrated intensity were selected across the area covered by 3 mm observations .
figure [ figure_ngc6946e_13co ] illustrates the regions targeted with @xmath1 observations .
the observations were taken between july and august 2010 .
we set the correlator to have the @xmath1 line in the lsb . in this case
we use a 125 mhz band centered on the rest frequency of @xmath1 line , and seven 500 mhz wide bands to observe the continuum at 1 mm .
this correlator configuration yields a velocity coverage of 160 @xmath13 and a channel width of 0.5 @xmath13 .
every pointing was observed for 1.5 minutes , which yields a total cycle time of 18 minutes , including a 1.5-minute pointing on the test calibrator 1927 + 739 .
we observed the gain calibrator 1849 + 670 after every cycle , and 3c454.3 was used as passband calibrator .
we have applied a + 0.5 briggs robustness parameter , which produces a map with angular resolution of 245@xmath6203 corresponding to @xmath10 50 pc , marginally the scale for gmc sizes .
the rms noise of the clean maps is 0.527 k. the pixel size of the maps is 0.5@xmath14 .
the observed regions are illustrated in figure [ fig_co21maps ] .
we simultaneously obtained data for the same 1 mm regions using 14 continuum bands covering 3.5 ghz in each side band .
at @xmath21 resolution , the rms sensitivity is given by @xmath22 @xmath23 , which corresponds to a dust column density @xmath24pc@xmath25 , assuming a dust temperature of 20 k and a dust opacity at 1.2 mm of 1 cm@xmath15 @xmath27 . further assuming a dust - to - gas ratio of 0.01 ( ) , this corresponds to a gas surface density @xmath28pc@xmath25 .
we have no detections over the regions observed . in order to recover the extended flux ,
we have merged our @xmath0 carma data with a ngc 6946 single dish map from the nobeyama 45-meter single dish telescope ( nro 45 m ) .
nro 45 m observations of ngc 6946 are part of the carma and nobeyama nearby - galaxies ( canon ) co(1 - 0 ) survey project , which uses both carma and nro 45 m telescopes to perform resolved observations of gmcs in disks of nearby galaxies ( koda et al .
, in preparation ) .
dm12 recently reported observations towards the central part of ngc 6946 using the carma and nro 45 m telescopes , and we refer the reader to that paper for the observation strategy and details of nro 45 m map . here , we just summarize the most relevant information .
ngc 6946 was observed in the @xmath0 transition using the beam array receiver system ( bears ) instrument .
the fwhm of the 45 meter dish is 15@xmath14 at the @xmath0 rest frequency which is degraded to 19@xmath297 after regridding .
the velocity resolution of the map is 2.54 @xmath13 .
the rms noise of the @xmath0 single dish map is 0.13 k ( 0.57 jy beam@xmath30 ) , which corresponds to @xmath31 assuming a 10 @xmath13 line width window to recover the entire flux . in the same way as we did for @xmath0 maps , we have merged our @xmath1 carma data with a ngc 6946 single dish map from the heracles project ( @xcite ) .
heracles used the hera receiver at the iram 30 m telescope to map the @xmath1 transition towards 18 nearby galaxies over the full optical disk , at @xmath32 angular resolution and 2.6 @xmath13 velocity resolution . at that resolution ,
the @xmath20 in the data cube is 25 mk , which yields an integrated intensity map with a noise level of @xmath33 assuming that we need to integrate over 10 @xmath13 to recover the entire flux .
cc we have merged our interferometric maps from carma with the single dish maps from heracles and nro 45 m using the miriad task _
immerge_. this task linearly merges a low resolution image , which represents the short - spacing data , with a high resolution image which represents the large spacings in the @xmath34 plane .
the cubes were masked to exclude edge velocity channels , and the region where the sensitivity of the carma map falls below half of the maximum .
_ immerge _ allows one to solve for the factor to scale the low resolution image by to put it on the same flux scale as the high resolution image .
thus , this factor should be close to 1 if both data sets have correctly calibrated flux scales . in our combination procedure
, we obtained values close to 1.05 .
we have calculated the flux inside the regions of significant emission provided by cprops ( see section [ cprops ] for details ) for both the image from carma observations alone , and the image resulting from merging carma and single dish data . in the case of the @xmath0 map , we found that most of the regions of significant emission in carma+nro 45 m maps present @xmath35 20% more flux than carma maps .
nevertheless there are a few regions that can have up to 50% more flux in the carma+nro 45 m map . on the other hand
, we found that , for all the regions observed in @xmath1 , carma+iram 30 m data sets present typically 20% more flux inside the region of significant emission .
atomic gas surface density is estimated from 21 cm line maps from the hi nearby galaxy survey ( things ) described in @xcite .
things maps the 21 cm line in 34 nearby galaxies using the very large array ( vla ) .
the angular resolution of the `` robust '' weighted map is @xmath36 , the channel width is 2.6 @xmath13 , and the noise of the map is @xmath10 17 k. assuming a 20 @xmath13 integration velocity range to fully recover the flux , this noise corresponds to a surface density of @xmath37 .
ngc 6946 fuv data are available from the _ galex _ nearby galaxies survey ( ngs ; @xcite ) .
the fuv band covers the wavelength range 1350 - 1750 , and the maps have an angular resolution of 56 .
we generate a sky background subtracted fuv image , by subtracting the sky background image from the intensity map , both provided by the ngs project .
we examined the image for presence of foreground stars over the field we are studying in ngc 6946 . following the approach presented in @xcite ( henceforth , l08 ) , we search for foreground stars by using nuv / fuv flux ratio .
we did not detect the presence of foreground stars over the field we are analyzing , so we did not apply any masking to the fuv map .
we have corrected fuv maps for galactic extinction using the extinction map of @xcite .
the fuv extinction estimated by using @xmath38 from @xcite .
ccc + + + we use 3.6 @xmath9 m maps from the infrared array camera ( irac , @xcite ) onboard the _ spitzer space telescope _ to estimate the stellar surface density ( sings , @xcite ) .
we aim to identify the spiral arms through tracing the old stellar population . following a similar approach as taken by @xcite ,
we have applied a spatial filtering to the map to remove variations of 3.6 @xmath9 m surface density which are not related to stellar mass density enhancements .
we have implemented this procedure by using the gipsy task _ mfilter _ , which applies a median filtering to the pixels inside a given spatial box . in this study
we used a [email protected] arcsec@xmath15 box , which successfully removes the bright spots due to recent star formation in the eastern region of ngc 6946 . to estimate the amount of obscured star formation by dust
, we use the 24 @xmath9 m data obtained by sings with the mips instrument of the _ spitzer space telescope _ ( @xcite ) .
the angular resolution of mips is @xmath39 ( fwhm ) .
we perform a background estimation as follows .
we select a region away from the galaxy , and we did a median spatial filtering over this background region in order to remove bright spots and other features , as we did for the 3.6 @xmath9 m map .
then , the background value is represented by the mean flux over the filtered region .
the value obtained ( @xmath10 0.02 mjy / str ) is negligible compared to the flux in the eastern region of ngc 6946 , thus we did not apply a background subtraction to our map . in order to estimate the star formation activity of the molecular structures at the resolution provided by @xmath1 ( @xmath40 ) , we need a star formation tracer with resolution finer than the limiting resolution provided by 24 @xmath9 m ( @xmath41 ) . with an angular resolution of @xmath42 ,
h@xmath12 maps provide a useful way to estimate star formation for the smallest identified structures . an h@xmath12 image of ngc 6946 is available in the sings fourth data release as part of the sings ancillary data program .
we used the stellar continuum - subtracted h@xmath12 image , and we convert to flux units using the sings data release documentation . in order to remove the [ n ii ] contribution ,
we have assumed ratios of [ n ii]@xmath436584/ h@xmath12=0.5 and [ n ii]@xmath436548/[n ii]@xmath436584=0.335 ( @xcite ) .
we identify molecular cloud structures and measure their properties using the cloud properties algorithm ( cprops ) described in @xcite .
cprops identifies clouds by masking the spectral line cube to isolate regions of significant emission that are both spatially and kinematically connected , and assigns properties like cloud sizes , luminosities and line widths using moments of the emission inside the identified region .
those measurements can be corrected for the bias due to the finite resolution and sensitivity of the maps .
this methodology has already been applied in several extragalactic studies ( @xcite ; b08 ; @xcite , @xcite ) .
we identify regions of significant emission by selecting pixels with emission greater than a threshold of @xmath44 across two consecutive velocity channels .
then , we extend the region of significant emission to the adjacent regions with emission greater than @xmath45 in two adjacent velocity channels . we probed a range of values of @xmath46 and @xmath47 in each data set to suppress the presence of noise in our masked region .
because the sensitivity in @xmath0 map is better than @xmath1 cubes , we compare the regions identified in the latter with the emission found in @xmath0 data cube . regions of marginal significance that do not present a counterpart in @xmath0 are considered as spurious detections . in this study
we have used @xmath48 and @xmath49 for both @xmath0 and @xmath1 maps .
after regions of significant emission are identified , cprops will find individual clouds based on conditions that depend on the decomposition parameters used .
a cloud is considered as a separate structure if it is large enough to allow the algorithm to calculate properties , the contrast between the peak and the edge is larger than the minimum allowed , and the properties of the cloud change significantly when it is combined with a nearby identified cloud ( see appendix of @xcite for a complete description of the decomposition parameters ) . due to the observed substructure in the ism
, the parameters for decomposition must be based on the known properties of the objects we want to identify .
previous studies have used the `` physical priors '' set of parameters presented by @xcite which are based on the observed properties of gmcs in the milky way . in that case , the decomposition is performed at a spatial resolution of 15 pc and velocity resolution of 2 @xmath13 .
because we are limited by the resolution of our maps , in our study we use parameters based on the spatial and spectral resolutions of the data cube .
then , in finding local peaks in regions of significant emission we search in boxes of @xmath50 110 pc ( 5.1@xmath14 ) and 2.5 @xmath13 in velocity for the @xmath0 maps and @xmath50 50 pc ( 2@xmath14 ) and 2.5 @xmath13 in the case of @xmath1 . because the limiting resolution of the @xmath0 maps likely does not allow us to resolve individual clouds , particularly the smallest ones , henceforth the structures identified in @xmath0 maps will be referenced as `` molecular complexes '' . on the other hand
, structures identified in @xmath1 will be called `` clouds '' .
the sizes , line widths and luminosities of the identified structures are derived using moments of the brightness distribution .
the rms size @xmath51 is determined by the geometric mean of the second moments of emission along the major and minor axes .
the velocity dispersion @xmath52 is determined by the second moment of the emission along the velocity axis .
the flux of the cloud is calculated using the zeroth moment of the emission in position and velocity . as we noted at the beginning of this section
, the sensitivity and resolution of the data can affect the derived properties of the clouds .
for instance , the size of the clouds will be underestimated by the second moment of the flux in a map with finite signal to noise .
cprops reduces this bias by extrapolating the size , the velocity width and the flux to the case of infinite sensitivity ( i.e. brightness temperature of the edge of the cloud equal to 0 k ) .
the extrapolation for the size and the velocity is linear , while the extrapolation for the flux is quadratic .
@xcite recommend correcting for resolution bias as well , which can be significant for extragalactic observations .
we correct for the effect of finite spatial and spectral resolution by subtracting in quadrature the beam size and the spectral channel profile from the extrapolated measurements of size and velocity width respectively .
the physical quantities are derived from the moment measurements following the standard procedure described in @xcite and b08 . following the definition of s87
, the cloud size is defined as @xmath53 .
although this value is @xmath10 10% lower than the value we would obtain from the equivalent area of the cloud ( assuming @xmath54 in the density profile @xmath55 ) , we used this definition in order to compare our values to the sizes presented by s87 and b08 .
the co luminosity @xmath56 is given by @xmath57 where @xmath58 is the distance to ngc 6946 in parsecs .
once we have determined the basic properties of the clouds ( size , line width and luminosity ) , other cloud properties can be estimated by taking combinations of these basic properties .
for instance , the luminosity - based mass is obtained from @xmath59 by using @xmath60 where @xmath61 is the assumed co - to - h@xmath62 conversion factor .
this conversion factor is thought to depend on local ism properties such as metallicity or radiation field ( @xcite ; @xcite ) . in this study
we assume @xmath63 , which is broadly used in previous extragalactic studies ( e. g. , b08 ; l08 ) .
in their study of gmcs within the central 5 kpc of ngc 6946 , dm12 found @xmath64 within a factor of 2 uncertainty , which is consistent with the value assumed here .
equation [ mco - equ ] includes a factor of 1.36 to account for the presence of helium . in this study , we have assumed @xmath65 in order to be consistent with the analysis presented in b08 .
virial masses are normally calculated under the assumption that molecular clouds are spherically symmetric .
however , the observed cloud shapes present a more complex structure . generalizing the standard virial analysis for spherical clouds ,
@xcite calculated the virial mass for spheroidal clouds .
following appendix a of @xcite the gravitational energy of spheroidal clouds is given by @xmath66 where @xmath67 is the radius perpendicular to the axis of symmetry of the cloud , while @xmath68 and @xmath69 are the density distribution and shape - dependent factors respectively .
for a @xmath70 power law density distribution , @xmath71 is given by @xmath72 .
on the other hand , @xmath69 will change if the cloud presents an oblate or prolate form .
if the cloud is oblate , @xmath73 is given by @xmath74 , where @xmath75 is the eccentricity of the cloud , and @xmath76 is the ratio between the size along and perpendicular to the axis of symmetry .
if the cloud presents a prolate shape instead , @xmath77 . ignoring the magnetic energy and external pressure terms in the virial equilibrium equation , and assuming @xmath78 , the virial mass is given by @xmath79 which corresponds to the expression for the virial mass assuming a spherically symmetric geometry , with a correction factor of 1/@xmath73 .
the distribution of @xmath73 is observed to be narrow .
we found that the mean of @xmath73 for @xmath0 complexes was 0.77 , with a variance of 0.02 around the mean .
on the other hand , the mean of @xmath73 for @xmath1 clouds was 0.81 , with a variance of 0.02 .
thus , the shape factor @xmath73 can account for a correction of the spherically symmetric virial mass by @xmath10 30% for complexes , and @xmath80 for @xmath1 clouds .
a complete description of the sizes and orientations of the clouds in three dimensions is preferable ( e.g. @xcite ) , but that information is not available for observational data . in particular , a correction for inclination can be only applied statistically , and not for individual clouds .
thus , we assume here that the clouds can be fully described by the sky - projected major and minor axes .
we make the further assumption that the axis of symmetry is given by the major axis of the clouds , and the clouds are prolate .
the correction for the shape of the cloud can strongly affect the virial mass estimates for structures that present a large axis ratio . because we assume elongation in the plane of the sky ,
a structure elongated along the line of sight will appear roughly circular , and our correction will be inappropriate . nevertheless , given the sizes found in this regions ( @xmath10 50 - 150 pc ) , and the usual thickness of the molecular gas disk in galaxies ( e.g. , @xcite find molecular gas thickness @xmath10200 pc for ngc 891 ) , it is likely that the axes of symmetry of the clouds are directed along the disk , reducing the overcorrection we may be applying to our estimates . supporting this assumption , @xcite have provided observational evidence that molecular clouds are elongated along the galactic plane .
we followed the definition of the dimensionless virial parameter @xmath81 given by @xcite , @xmath82 which measures the relative importance of the cloud s kinetic energy compared to its gravitational energy .
this parameter is usually used to investigate whether the clouds are in virial equilibrium , @xmath83 ; for @xmath84 the self - gravity is not important , and the clouds are confined due to an external pressure if the clouds are assumed to be in steady state .
the uncertainties assessed to the moment measurements ( and thus to the physical quantities and their derivatives ) are estimated using bootstrapping methods , and this is the only source of uncertainties that we include in our analysis . additionally , we have used the @xmath2 line map to calculate lte masses over the regions with significant emission . if we assume that all energy levels are populated according to local thermodynamic equilibrium ( lte ) at the temperature @xmath85 , the column density of the @xmath86 molecule in terms of the opacity of the @xmath87 transition is given by ( e. g. , @xcite ) @xmath88 where @xmath8 is in @xmath13 .
in this study we have determined the excitation temperature by using the assumed optically thick line @xmath0 @xmath89 and assuming a filling factor of unity .
the mean of @xmath85 is 4.2 k considering the regions of significant emission in the @xmath0 map , while this mean increases to 5 k when considering the @xmath2 regions of detectable emission . in section [ mvir - lum ]
we will investigate the effect of using higher values of @xmath85 in the lte mass estimates .
the optical depth of the @xmath2 line is given by @xmath90.\ ] ] by assuming that @xmath2 line is optically thin , the total molecular mass can be computed using an appropriate abundance ratio , @xmath91 d^2 \int{n(^{13}\mathrm{co})\mathrm{d}\omega}\ ] ] where @xmath92 is the mean molecular mass per h@xmath62 molecule , @xmath58 is the distance , and @xmath93 $ ] is the abundance ratio . in this study
we have assumed the abundance ratio to be @xmath94 .
this value is close to the ratio given by the relation found by @xcite , which yields @xmath95\sim 60 $ ] for @xmath96 kpc , and assuming a @xmath97 abundance ratio of 1.1 @xmath98 ( @xcite ) .
another important cloud property is the mass surface density .
surface density has been estimated by dividing the luminosity - based mass by the corresponding area covered by the @xmath1 cloud ( or @xmath0 complex ) , thus @xmath99 .
in general the areas for @xmath1 clouds are smaller than @xmath0 complexes .
the fluxes and areas involved in these calculations are the non - extrapolated values .
we have made this choice , because we intend to compare molecular gas surface density with other tracers ( hi , 24@xmath9 m , h@xmath12 , fuv ) over the boundary of the clouds given by the mask provided by cprops .
because we do not have extrapolated or deconvolved measurements of the fluxes for these tracers , we have used the raw properties of the clouds . in this paper , we have estimated the star formation rate ( sfr ) using fuv , 24@xmath9 m and h@xmath12 images described in section [ obs ] .
the method to estimate the star formation activity is different for @xmath0 complexes and @xmath1 clouds , mainly due to the limiting spatial resolution of the sfr tracer images .
the details of the calculations are given below .
following the approach given by l08 , we estimate the star formation rate surface density in @xmath0 complexes using _ galex _ fuv and _ spitzer _ 24 @xmath9 m maps . by tracing photospheric emission of o and b stars , fuv emission measures unobscured star formation over time scales of 10 - 100 myr ( @xcite ) . on the other hand
, 24 @xmath9 m traces the small dust grains heated by uv photons radiated from young stars , thereby tracing obscured star formation over time scales of @xmath1010 myr ( @xcite ; @xcite ) .
the use of 24 @xmath9 m as tracer of obscured star formation was analyzed by @xcite .
they showed that the combination of h@xmath12 ( unobscured sfr ) and 24@xmath9 m ( obscured ) recovers the total sfr , which they estimated by using paschen-@xmath12 ( pa@xmath12 ) emission as a tracer of ionizing photons unaffected by extinction .
l08 used a similar prescription to derive the total sfr , but using fuv maps instead of h@xmath12 to trace unobscured sfr .
they found that the total star formation surface density @xmath100 can be estimated by : @xmath101 where @xmath102 and @xmath103 are in units of mjy / sr .
the coarse resolution of the 24@xmath9 m map ( @xmath39 ) does not permit us to use the sfr from equation ( [ sfr ] ) to perform a direct comparison between star - forming regions and the structures identified in @xmath1 .
we therefore use the h@xmath12 image to estimate the amount of star formation in @xmath1 clouds .
however , h@xmath12 emission can be strongly affected by dust extinction , and a correction should be applied to h@xmath12-based sfr estimates to account for embedded star formation .
following @xcite , the star formation surface density can be estimated by @xmath104 where @xmath105 is in units of ergs s@xmath30 , and @xmath106 is the h@xmath12 extinction . in the section [ sfr - result ] , we will show that applying a @xmath107 mag roughly recovers the total star formation as traced by fuv+24@xmath9 m .
although the cprops method corrects for sensitivity and resolution bias , those corrections can introduce some additional scatter in the cloud property scaling relations .
therefore , we will start this section by showing the scaling relations without any correction to assess overall correlation strengths .
we have identified 45 complexes in the @xmath0 channel map .
the mean uncorrected radius of the @xmath0 clouds is 128.5 @xmath108 3.1 pc , and the mean uncorrected line width is 4.2 @xmath108 0.1 @xmath13 . on the other hand ,
the mean uncorrected luminosity of these complexes is ( 8.6 @xmath108 0.1 ) @xmath109 @xmath110 , which corresponds to a mean luminosity - based mass of 3.8 @xmath108 0.05 @xmath111 . figure [ fig_co10_siwi_raw ] shows the scaling relations for the raw properties .
we selected the 11 most massive @xmath0 complexes to make follow - up observations with @xmath1 , and they are shown in figure [ figure_ngc6946e_13co ] .
we identified 64 giant molecular clouds in the regions observed , and their scaling relations are shown in figure [ fig_co10_siwi_raw ] as well .
for this set of clouds , the mean radius is 54.9 @xmath108 1.1 pc , the mean line width is 3.7 @xmath108 0.1 @xmath13 , the mean luminosity is ( 2.4 @xmath108 0.03 ) @xmath6 @xmath112 , and the mean luminosity - based mass is ( 1.04 @xmath108 0.01)@xmath6 @xmath113 .
.spearman rank correlation coefficient @xmath114 for both raw and bias corrected relations.[table - rank ] [ cols="^,^,^,^,^,^ " , ] the other region with high sfe ( region 6 in figure [ figure_ngc6946e_13co ] ) , on the other hand , does not present such a high velocity gradient as the region discussed above .
however , we observe a @xmath1015 @xmath13 difference in velocity between region 6 and the regions located to the north - west and to the south - east . while those regions surrounding the molecular gas complex in region 6 are moving more or less at @xmath115 @xmath13 with respect to the rotating frame , region 6 is moving in the other direction with a velocity of @xmath1010 @xmath13 .
this difference in velocity can be observed clearly in the position - velocity diagram shown in the right panel of figure [ figure_pv ] .
this may imply that the gas actively forming stars in region 6 could be the result of a recent convergence of flows .
although we do not observe signatures of a merging or collision process of spiral arms in the gas distribution near region 6 , it is possible that the process had already formed a new spiral arm , obscuring the original arm locations .
more detailed analysis of the kinematics of the gas is required to disentangle the actual process involved in the enhancement of star formation efficiency in region 6 .
c the study presented here can be improved in multiple ways .
our analysis has been limited to the north - eastern area of the molecular gas disk of ngc 6946 for @xmath0 observations , and to some of the brightest regions for @xmath1 follow - up observations .
a full census of the gmcs population across the disk is needed in order to perform a detailed study of molecular cloud properties and their relation with the surrounding environment .
nevertheless , high sensitivity maps with resolutions @xmath116 pc remain observationally challenging to obtain , although recent studies have achieved this goal with carma+nro 45 m ( e.g. dm12 ) .
new facilities like alma will provide maps of unprecedented quality , allowing us to perform an unbiased analysis of the molecular gas in nearby galaxies at scales close to gmc sizes . in our sfr calculations for @xmath1 clouds ,
we have applied a crude global extinction correction a@xmath117 mag to the sfr(h@xmath12 ) .
although we showed that such correction is a good approximation to the sfr values found using a combination of fuv+24@xmath9 m for @xmath0 complexes , a more sophisticated approach to calculate the obscured sfr at resolution close to the gmcs sizes is needed .
thus , tracers relatively unaffected by dust absorption represent a valuable alternative to estimate the total sfr at smaller scales . in this direction , observations of paschen @xmath12 ( p@xmath12 )
line could provide unbiased measurements of the number of ionizing photons in these star forming regions ( @xcite ) .
equations [ sfr ] and [ sfr_ha ] assume that stars have been formed continuously .
nevertheless , this assumption may not be longer valid for scales close to gmc sizes , as one approaches the case of a single stellar population with a discrete age .
some studies have investigated the uncertainty in using a linear conversion from uv or h@xmath12 luminosity to estimate the sfr , making use of models assuming an instantaneous burst of star formation ( @xcite ; @xcite ) .
@xcite showed that while h@xmath12 emission occurs mostly in the first 10 myr , fuv presents low , but significant , emission up to 65 myr after the burst .
they report a factor of 2 uncertainty inferring the sfr from h@xmath12 , and a factor 3 - 4 uncertainty inferring sfr from fuv .
resolved star cluster observations in the regions observed in this study could shed light on the ages involved , and provide an independent way to estimate the star formation rate which can be compared to the sfr traced by fuv or h@xmath12 .
we have performed one of the most detailed observations of gmcs outside the milky way . using carma ,
we have observed @xmath0 and @xmath2 over the northeastern disk of the non - grand design galaxy ngc 6946 , covering a region of 6@xmath66 kpc@xmath15 at a resolution of @xmath10110 pc .
although at this resolution it is possible to resolve some of the largest molecular clouds , we have complemented this map with higher ( @xmath1050 pc ) resolution maps of @xmath1 towards the brightest regions observed in @xmath0 .
the extended flux has been recovered by merging our interferometric maps from carma with single dish maps from nro 45 m and iram 30 m .
the results are summarized as follows : 1 .
we have estimated the properties of the molecular gas structures ( size , line width and flux ) using the cloud - finding algorithm cprops ( section [ cl - prop ] ) .
we have resolved the largest molecular complexes into several structures . at the resolution offered by @xmath0 observations ,
we have identified 45 co emitting complexes with typical sizes of @xmath10150 pc .
five of these complexes show significant @xmath2 emission , which appear to be regions of denser gas .
the higher resolution observations of @xmath1 towards the brightest regions detected in @xmath0 allowed us to find 64 structures which we have identified as gmcs .
2 . in section [ cl - prop - corr ]
we observed that the properties of the clouds and complexes follow relations similar to those found previously for extragalactic clouds ( figure [ figure_line - width ] ) .
in particular , the trends defined by the @xmath0 complexes presented in this paper are consistent with those found in the center of ngc 6946 by dm12 , despite the differences in resolution and cloud identification algorithms ; the exception is the set of gmcs located within 400 pc of the galactic center with large velocity dispersions ( > 10 @xmath13 ) .
additionally , our @xmath118-@xmath56 relation is located slightly above the relation found by dm12 for the central part of ngc 6946 and is consistent with our choice of the co to h@xmath62 conversion factor @xmath61=@xmath119 .
we have estimated lte masses for the regions with significant @xmath2 emission ( figure [ figure_13co ] ) .
although the lte masses are smaller by a factor of 2 than the luminosity - based masses in our sample of clouds , this discrepancy does not necessarily mean that @xmath0 intensities are not suitable to estimate true molecular cloud masses , but may reflect systematic errors arising from our lack of knowledge of the excitation temperature and co abundance , both needed to infer masses from @xmath2 emission .
the star formation rates of the @xmath0 complexes were estimated in section [ sfr - result ] using multiband _
imaging from sings and fuv galex maps of ngc 6946 . in figure [ figure_sfr ]
we observe a clear correlation between sfr surface density ( @xmath100 ) and molecular gas surface density ( @xmath120 ) .
the bisector linear regression for the relation @xmath121 yields a slope of @xmath122 for the @xmath0 emitting complexes , which is steeper than the value found by previous work ( @xmath123 , @xcite ) . in order to detect faint co complexes , but with active star formation , we have observed with higher sensitivity the @xmath0 line over a smaller region located beyond the spiral arms , where atomic gas is comparable ( or dominant ) to the molecular gas ( figure [ figure_mom0_e ] ) .
we interpret these co - faint star forming regions as isolated structures that could have departed from the main spiral arms , and have been turning their molecular gas into stars since then .
thus , it is not surprising to find atomic gas peaks in these regions , which could have been produced by photodissociated molecular gas .
higher sensitivity observations are necessary to detect faint molecular gas associated with such late stage star formation .
5 . for gmcs identified in the @xmath1 maps
, we estimate their sfr using h@xmath12 maps . we have found a steeper slope than the values found for @xmath0 complexes , with @xmath124 .
we observe that galactic massive dense clumps follow a rough extrapolation of the @xmath100-@xmath120 relation for these @xmath1 gmcs to higher gas surface densities ( figure [ figure_sfr_heid ] ) .
alma will observe the molecular gas distribution at even higher resolution and sensitivity , allowing a direct comparison between the galactic and extragalactic @xmath100-@xmath120 relations at similar scales .
6 . in section [ in - on - arm ] , on - arm and inter - arm regions have been defined based on stellar mass density enhancements traced by the 3.6 @xmath9 m image of ngc 6946 ( figure [ figure_four ] ) .
the complexes and gmcs were classified as on - arm and inter - arm structures accordingly .
although we observe that cloud properties have similar values independent of location in the disk , a small number of clouds are more massive , present broader line widths , and higher star formation rate than the remaining sample of clouds ( figure [ figure_arm10 ] ) .
we found that these structures appear to be located at two specific regions in the spiral arms ( region 6 - 7 and region 11 in figure [ figure_ngc6946e_13co ] ) .
7 . the star formation efficiency ( sfe , defined as
the ratio between @xmath100 and @xmath120 ) is roughly flat over the region observed , with values between @xmath125 and @xmath126 myr@xmath30 ( section [ rad - prop ] , figure [ figure_rad_sfr ] ) .
however , the sfe is enhanced in the regions where we observed more massive and more luminous gmcs ( see figure [ figure_map_sfe ] and table [ table - high - cld ] ) .
the velocity field of the molecular gas reveals irregular non - circular motion with values below 15 @xmath13 with respect to the rotating frame ( figure [ figure_mom1_e ] ) .
in addition , we found some evidence for convergence of gas flows in one of the regions of high sfe ( figure [ figure_pv ] ) .
the other region presenting sfe peaks does not show such a strong velocity gradient .
however , this region shows a 15 @xmath13 difference with respect to the surrounding molecular gas , which could be caused by a previous convergence of gas flow .
although a more detailed comparison with simulations is needed in order to have a more conclusive picture of the role of gas dynamics in the formation of gmcs and stars in these type of galaxies , these new observations suggest that , in agreement with numerical simulations of non - grand design spiral galaxies , the convergence of gas flows into the spiral arm , and collisions or merging of arms may increase the local gas density to higher values than the mean observed in the disk , creating more massive molecular clouds , and enhancing the rate of star formation .
the authors thank the anonymous referee for comments that have improved the presentation and the discussion of the paper .
we gratefully acknowledge the efforts by the sings , _ galex _ ngs , and things teams to make their data public , and we want to thank erwin de blok for providing the rotation curve of ngc 6946 in tabular form .
we also thank the carma staff the support in the observations presented in this work .
the construction of carma was supported by the gordon and betty moore foundation , the kenneth t. and eileen l. norris foundation , the james s. mcdonnell foundation , the associates of the california institute of technology , the university of chicago , the states of california , illinois , and maryland , and the national science foundation . the current carma development and operations
are supported by the national science foundation and the carma partner universities under a cooperative agreement .
this work has been partially supported by the comisin nacional de investigacin en ciencia y tecnologa of chile ( conicyt ) under the fulbright - conicyt fellowship program , and by nsf grant ast 08 - 38226 .
jin koda acknowledges support from the nsf through grant ast-1211680 .
the national radio astronomy observatory is a facility of the national science foundation operated under cooperative agreement by associated universities , inc .
akritas , m. g. & bershady , m. a. 1996 , , 470 , 706 ballesteros - paredes , j. , hartmann , l. w. , vzquez - semadeni , e. , heitsch , f. & zamora - avils , m. a. 2011 , , 411 , 65 - 70 bell , t. a. , roueff , e. , viti , s. & williams , d. a. 2006 , , 371 , 1865 - 1872 bertoldi , f. & mckee , c. f. 1992 , , 395,140 - 157 bicay , m. d. & giovanelli , r. 1986 , , 91 , 705 - 750 bigiel , f. , leroy , a. , walter , f. , et al .
2008 , , 136 , 2846 - 2871 blanc , g. a. , heiderman , a. , gebhardt , k. , evans , ii , n. j. & adams , j. 2009 , , 704 , 842 - 862 blitz , l. & rosolowsky , e. 2004 , , 612 , l29-l32 blitz , l. & rosolowsky , e. 2006 , , 650 , 933 - 944 blitz , l. , fukui , y. , kawamura , a. , et al .
2007 , protostars and planets v , 81 - 96 bolatto , a. d. , leroy , a. k. , rosolowsky , e. , walter , f. & blitz , l. 2008 , , 686 , 948 - 965 boselli , a. , gavazzi , g. , combes , f. , lequeux , j. & casoli , f. 1994 , , 285 , 79 - 68 briggs , f. h. 1990 , , 352 , 15 - 29 calzetti , d. , kennicutt , jr . ,
r. c. , bianchi , l. , et al .
2005 , , 633 , 871 - 893 calzetti , d. , kennicutt , r. c. , engelbracht , c. w. , et al .
2007 , , 666 , 870 - 895 catinella , b. , haynes , m. p. , giovanelli , r. , gardner , j. p. , & connolly , a. j. 2008 , , 685 , l13-l17 clarke , c. & gittins , d. 2006 , , 371 , 530 - 536 considre , s. & athanassoula , e. 1988 , , 76 , 365 - 404 crosthwaite , l. p. & turner , j. l. , 2007 , , 134 , 5 , 1827 de blok , w. j. g. , walter , f. , brinks , e. , et al . 2008 , , 136 , 2648 - 2719 dickey , j. m. & kazs , i. , 1992 , , 393 , 530 - 543 dobbs , c. l. & bonnell , i. a. 2006 , , 367 , 873 - 878 dobbs , c. l. , & bonnell , i. a. , 2008 , . 385 , 1893 - 1902 donovan meyer , j. , koda , j. , momose , et al .
2012 , , 744 , 42 drozdovsky , i. o. & karachentsev , i. d. 2000 , , 142 , 425 - 432 durbala , a. , buta , r. , sulentic , j. w. & verdes - montenegro , l. 2009 , , 397 , 1756 - 1775 elmegreen , d. m. , elmegreen , b. g. , yau , a. , et al .
2011 , , 737 , 32 engargiola , g. , plambeck , r. l. , rosolowsky , e. & blitz , l. 2003 , , 149 , 343 - 363 evans , ii , n. j. , dunham , m. m. , jrgensen , j. k. , et al .
2009 , , 181 , 321 - 350 fazio , g. g. , hora , j. l. , allen , l. e. , et al .
2004 , , 154 , 10 - 17 fouqu , p. , durand , n. , bottinelli , l. , gouguenheim , l. & paturel , g. 1990 , , 86 , 473 - 502 foyle , k. , rix , h .- w .
, walter , f. & leroy , a. k. 2010 , , 725 , 535 - 541 frerking , m. a. , langer , w. d. & wilson , r. w. 1982 , , 262 , 590 - 605 fukui , y. , kawamura , a. , minamidani , t. , et al .
2008 , , 178 , 56 - 70 genzel , r. , tacconi , l. j. , gracia - carpio , j. , et al .
2010 , , 407 , 2091 - 2108 gil de paz , a. , boissier , s. , madore , b. f. , et al .
2007 , , 173 , 185 - 255 goldsmith , p. f. , heyer , m. , narayanan , g. , et al .
2008 , , 680 , 428 - 445 gutermuth , r. a. , pipher , j. l. , megeath , s. t. , et al .
2011 , , 739 , 84 haynes , m. p. & giovanelli , r. 1984 , , 89 , 758 - 800 heiderman , a. , evans , ii , n. j. , allen , l. e. , huard , t. & heyer , m. 2010 , , 723 , 1019 - 1037 helfer , t. t. , thornley , m. d. , regan , m. w. , et al .
2003 , , 145 , 259 - 327 heyer , m. , krawczyk , c. , duval , j. & jackson , j. m. 2009 , , 699 , 1092 - 1103 hirota , a. , kuno , n. , sato , n. , et al .
2010 , , 62 , 1261 - 1275 ho , l. c. 2007 , , 2007 , 669 , 821 - 829 hughes , a. , wong , t. , ott , j. , et al .
2010 , , 406 , 2065 - 2086 isobe , t. , feigelson , e. d. , akritas , m. g. & babu , g. j. 1990 , , 364 , 104 - 113 jarrett , t. h. , chester , t. , cutri , r. , schneider , s. e. & huchra , j. p. , 2003
, , 125 , 525 - 554 karachentsev , i. d. , karachentseva , v. e. , huchtmeier , w. k. & makarov , d. i. 2004 , , 127 , 2031 - 2068 kennicutt , jr . , r. c. 1998 , , 498 , 541 kennicutt , jr .
, r. c. , armus , l. , bendo , g. , et al .
2003 , , 115 , 928 - 952 kennicutt , jr . ,
r. c. , calzetti , d. , walter , f. , et al .
2007 , , 671 , 333 - 348 kim , w .- t . & ostriker , e. c. 2002 , , 570 , 132 - 151 koda , j. , sawada , t. , hasegawa , t. , & scoville , n. z. 2006 , , 638 , 191 - 195 koda , j. , scoville , n. , sawada , t. , et al .
2009 , , 700 , l132-l136 koyama , h. & ostriker , e. c. 2009 , , 693 , 1316 - 1345 koyama , h. & ostriker , e. c. 2009 , , 693 , 1346 - 1359 krumholz , m. r. , mckee , c. f. & tumlinson , j. 2009 , , 699 , 850 - 856 krumholz , m. r. , cunningham , a. j. , klein , r. i. & mckee , c. f. 2010 , , 713 , 1120 - 1133 kuno , n. , sato , n. , nakanishi , h. , et al .
2007 , , 59 , 117 - 166 larson , r. b. 1981 , , 194 , 809 - 826 lavezzi , t. e. & dickey , j. m. 1997 , , 114 , 2437 lavezzi , t. e. & dickey , j. m. 1998 , , 116 , 2672 - 2681 leroy , a. k. , walter , f. , brinks , e. , et al .
2008 , , 136 , 2782 - 2845 leroy , a. k. , walter , f. , bigiel , f. , et al .
2009 , , 137 , 4670 - 4696 leroy , a. k. , bolatto , a. , gordon , k. , et al .
2011 , , 737 , 12 leroy , a. k. , bigiel , f. , de blok , w. j. g. , et al .
2012 , , 144 , 3 liu , g. , koda , j. , calzetti , d. , fukuhara , m. & momose , r. 2011 , , 735 , 63 markwardt , c. b. 2009 , 251 mckee , c. f. & ostriker , e. c. 2007 , , 45 , 565 - 687 milam , s. n. , savage , c. , brewster , m. a. , ziurys , l. m. & wyckoff , s. 2005 , , 634 , 1126 - 1132 miyawaki , r. , hayashi , m. & hasegawa , t. 2009 , , 61 , 39 murray , n. , quataert , e. & thompson , t. a. 2010 . , 709 , 191 - 209 nieten , c. , neininger , n. , gulin , m. , et al .
2006 , , 453 , 459 - 475 obreschkow , d. , croton , d. , de lucia , g. , khochfa , s. & rawlings , s. 2009 , , 698 , 1467 - 1484 obreschkow , d. , klckner , h .-
r . , heywood , i. , levrier , f. , & rawlings , s. 2009b , , 703 , 1890 - 1903 oka , t. , hasegawa , t. , sato , f. , et al .
2001 , , 2001 , 562 , 348 - 362 ossenkopf , v. & henning , t. 1994 , , 291 , 943 - 959 ostriker , e. c. , mckee , c. f. & leroy , a. k. 2010 , , 721 , 975 - 994 rahman , n. , bolatto , a. d. , wong , t. , et al .
2011 , , 730 , 72 rand , r. j. 1995 , , 1995 , 109 , 2444 rand , r. j. & kulkarni , s. r. 1990 , , 349 , l43-l46 rand , r. j. , lord , s. d. , & higdon , j. l. 1999 , , 513 , 720 - 732 regan , m. w. & vogel , s. n. 1995 , , 452 , l21 rieke , g. , young , e. t. , engelbracht , c. w. , et al .
2004 , , 154 , 25 roberts , w. w. 1969 , , 158 , 123 rohlfs , k. & wilson , t. l. 2004 , tools of radio astronomy , ( 4th rev . and enl .
ed ; berlin : springer ) .
rosolowsky , e. 2007 , , 654 , 240 - 251 rosolowsky , e. & leroy , a. 2006 , , 118 , 590 - 610 sakamoto , k. , okumura , s. k. , ishizuki , s. & scoville , n. z. 1999 , , 124 , 403 - 427 schlegel , d. j. , finkbeiner , d. p. & davis , m. 1998 , , 500 , 525 schmidt , m. 1959 , , 129 , 243 shetty , r. & ostriker , e. c. 2006 , , 647 , 997 shetty , r. & ostriker , e. c. 2008 , , 684 , 978 shetty , r. , vogel , s. n. , ostriker , e. c. & teuben , p. j. 2007 , , 665 , 1138 - 1158 shetty , r. , collins , d. c. , kauffmann , j. , et al .
2010 , , 712 , 1049 - 1056 solomon , p. m. , rivolo , a. r. , barrett , j. & yahil , a. 1987 , , 319 , 730 - 741 tamburro , d. , rix , h .- w . , walter , f. , et al .
2008 , , 136 , 2872 - 2885 tamburro , d. , rix , h .- w . ,
leroy , a. k. , et al .
2009 , , 137 , 4424 - 4435 tasker , e. j. & tan , j. c. 2009 , , 700 , 358 - 375 tasker , e. j. 2011 , , 730 , 11 thornley , m. d. & mundy , l. g. 1997 , , 484 , 202 tifft , w. g. & cocke , w. j. 1988 , , 67 , 1 - 75 tosaki , t. , shioya , y. , kuno , n. , nakanishi , k. & hasegawa , t. 2003 , , 55 , 605 - 613 tully , r. b. , 1988 , journal of the british astronomical association , 98 , 316 tully , r. b. & fisher , j. r. 1977 , , 54 , 661 - 673 tutui , y. & sofue , y. 1999 , , 351 , 467 - 471 vogel , s. n. , kulkarni , s. r. & scoville , n. z. 1988 , , 334 , 402 - 406 wada , k. & koda , j. 2004 , , 349 , 270 - 280 wada , k. , baba , j. & saitoh , t. r. 2011 , , 735 , 1 walter , f. , carilli , c. , bertoldi , f. , et al .
2004 , , 615 , l17-l20 walter , f. , brinks , e. , de blok , w. j. g. , et al .
2008 , , 136 , 2563 - 2647 walterbos , r. a. m. & braun , r. 1996 , the minnesota lectures on extragalactic neutral hydrogen , 106 , 1-+ wang , k. , wu , y. f. , ran , l. , yu , w. t. & miller , m. 2009 , , 507 , 369 - 376 wong , t. & blitz , l. 2002 , , 569 , 157 - 183 wong , t. , hughes , a. , ott , j. , et al .
2011 , , 197 , 16 .
wu , j. , evans , ii , n. j. , shirley , y. l. & knez , c. 2010 , , 188 , 313 - 357 wyder , t. , martin , d. c. , schiminovich , d. , et al .
2007 , , 173 , 293 - 314 yim , k. , wong , t. , howk , j. c. & van der hulst , j. m. 2011 , , 141 , 48-+ | we present high spatial resolution observations of giant molecular clouds ( gmcs ) in the eastern part of the nearby spiral galaxy ngc 6946 obtained with the combined array for research in millimeter - wave astronomy ( carma ) .
we have observed @xmath0 , @xmath1 and @xmath2 , achieving spatial resolutions of @xmath3 , @xmath4 and @xmath5 respectively over a region of 6 @xmath6 6 kpc .
this region extends from 1.5 kpc to 8 kpc galactocentric radius , thus avoiding the intense star formation in the central kpc .
we have recovered short - spacing @xmath7-@xmath8 components by using single dish observations from the nobeyama 45 m and iram 30 m telescopes . using the automated cprops algorithm we identified 45 co cloud complexes in the @xmath0 map and 64 gmcs in the @xmath1 maps .
the sizes , line widths , and luminosities of the gmcs are similar to values found in other extragalactic studies .
we have classified the clouds into on - arm and inter - arm clouds based on the stellar mass density traced by the 3.6 @xmath9 m map .
clouds located on - arm present in general higher star formation rates than clouds located in inter - arm regions . although the star formation efficiency shows no systematic trend with galactocentric radius , some on - arm clouds which are more luminous and more massive compared to inter - arm gmcs are also forming stars more efficiently than the rest of the identified gmcs .
we find that these structures appear to be located in two specific regions in the spiral arms .
one of them shows a strong velocity gradient , suggesting that this region of high star formation efficiency may be the result of gas flow convergence . |
the _ sp theory of intelligence _ aims to simplify and integrate ideas in artificial intelligence , mainstream computing , and human perception and cognition , with information compression as a unifying theme .
the theory is described in several peer - reviewed articles , and links from there . ] and most fully in @xcite .
the main purpose of this article is to describe how the sp theory may be applied to the understanding of natural vision and the development of computer vision , and to discuss associated issues . both of those themes natural vision and artificial vision
are discussed together throughout the article , since each one may illuminate the other . in broad terms , the potential benefits of the sp theory in those two areas are the simplification and integration of concepts , deeper insights , better performance ( of artificial systems ) , and the seamless integration of vision with other sensory modalities , and with other aspects of intelligence such as reasoning , planning , problem solving , and unsupervised learning . what is perhaps the main attraction of the theory is the potential for one relatively simple framework to accommodate several different aspects of intelligence , including vision . as a preliminary
, the next section describes the theory in outline , with associated ideas .
the sp theory combines conceptual simplicity with descriptive and explanatory power in several areas , including concepts of ` computing ' , the representation of knowledge , natural language processing , pattern recognition , several kinds of reasoning , the storage and retrieval of information , planning and problem solving , unsupervised learning , information compression , and human perception and cognition .
since the sp theory has been described quite fully in @xcite , only the essentials will be given here , with enough detail to ensure that the rest of the article makes sense .
the main elements of the sp theory are : * the theory is conceived as an abstract system that , like a brain , may receive ` new ' information via its senses and store some or all of it as ` old ' information . *
all new and old information is expressed as arrays of atomic symbols ( _ patterns _ ) in one or two dimensions . *
the system is designed for the unsupervised learning of old patterns by compression of new patterns . *
an important part of this process is , where possible , the economical encoding of new patterns in terms of old patterns .
this may be seen to achieve such things as pattern recognition , parsing or understanding of natural language , or other kinds of interpretation of incoming information in terms of stored knowledge , including several kinds of reasoning .
* compression of information is achieved via the matching and unification ( merging ) of patterns , with key roles for the frequency of occurrence of patterns , and their sizes . *
the concept of _ multiple alignment _ , outlined in section [ multiple_alignment_section ] , is a powerful central idea , similar to the concept of multiple alignment in bioinformatics but with important differences .
* owing to the intimate connection between information compression and concepts of prediction and probability ( see , for example , * ? ? ?
* ) , it is relatively straightforward for the sp system to calculate probabilities for inferences made by the system , and probabilities for parsings , recognition of patterns , and so on . * in developing the theory , i have tried to take advantage of what is known about the psychological and neurophysiological aspects of human perception and cognition , and to ensure that the theory is compatible with such knowledge . the way the sp concepts may be realised with neurons ( _ sp - neural _ ) is discussed in ( * ? ? ?
* chapter 11 ) .
the sp theory is realised in the form of computer models which may be regarded as first versions of the _ sp machine _ , an expression of the theory and a means for it to be applied .
the sp70 model is the most comprehensive version , with capabilities in the building of multiple alignments and unsupervised learning .
the sp62 model is the same but it lacks any ability to learn .
although sp62 is a subset of sp70 , it has proved convenient to maintain them as separate models .. ] at the heart of the sp models is a process for finding good full or partial matches between patterns ( * ? ? ?
* appendix a ) , with a flexibility that is somewhat like the winmerge utility for finding similarities and differences between files , or standard ` dynamic programming ' methods for the alignment of sequences . the main difference between the sp process and others , is that the former can deliver several alternative matches between patterns , while winmerge and standard methods deliver one ` best ' result .
multiple alignments are built in stages , with pairwise matching and merging of patterns , and with merged patterns from any stage being carried forward to later stages . at all stages , the aim is to encode new information economically in terms of old information and to weed out multiple alignments that score poorly in that regard . in the sp70 model
, there are additional processes for deriving old patterns from multiple alignments , evaluating sets of newly - created old patterns in terms of their effectiveness for the economical encoding of the new information , and weeding out low - scoring sets .
more detail about sp70 may be found in ( * ? ? ? * sections 3.9 and 9.2 ) .
the sp61 model , a precursor of sp62 which is very similar to it , is described in sections 3.9 and 3.10 ( _ ibid . _ ) .
the main limitations of current models are : * that they work with one - dimensional patterns and have not yet been generalised to work with 2d patterns ( although a preliminary attempt has been made to consider how the sp principles may be generalised to patterns in two dimensions ( * ? ? ?
* section 13.2.1 ) ) . * that the arithmetic meaning of numbers is not recognised
they are simply treated as patterns . *
that sp70 does not yet learn intermediate levels of abstraction in grammars , or discontinuous patterns in data .
i believe these problems are soluble .
potential solutions will be mentioned at relevant points below .
owing to the first of these limitations , most of the examples in this article , and much of the discussion , will relate to one - dimensional patterns . like most problems in artificial intelligence , the problems that are addressed in the sp models finding good full and partial matches between patterns , the formation of multiple alignments , and the learning of useful sets of patterns are not tractable if the requirement is to find ideal solutions .
but , as with most programs in artificial intelligence , things become much easier if one is content with solutions that are reasonably good and not necessarily perfect . like
most programs in artificial intellegence , the sp models apply constraints on the process of searching , to reduce the size of the search space so that useful results may be achieved with the available computational resources .
an example of multiple alignment in the sp system is shown in figure [ kittens_figure ] . here ,
row 0 contains a new pattern representing a sentence : ` t w o k i t t e n s p l a y ' , while each of rows 1 to 8 contains an old pattern representing a grammatical rule or a word with grammatical markers .
this multiple alignment , which achieves the effect of parsing the sentence in terms of grammatical structures , is the best of several built by the sp62 model when it is supplied with the new pattern and a set of old patterns that includes those shown in the figure and several others as well . in this example , and others in this article , ` best ' means that the multiple alignment in the figure is the one that enables the new pattern to be encoded most economically in terms of the old patterns . details of how the encoding is done may be found in (
* section 3.5 ) .
a point of interest about this multiple alignment is the way that , in row 8 , the symbols ` np ' and ` vp ' mark the grammatical dependency between the plural subject of the sentence ( ` k i t t e n s ' ) and the plural main verb ( ` p l a y ' ) .
this kind of dependency is often described as ` discontinous ' because there may be arbitrarily large amounts of intervening structure between one element of the dependency and another .
this method of marking discontinous dependencies is , arguably , simpler and more elegant than how they are marked in other grammatical systems .
much of the descriptive and explanatory power of the sp theory is due to the versatility of the multiple alignment concept in : * _ the representation of knowledge_. despite the simplicity of sp patterns , the way they are processed within the multiple alignment framework gives them the versatility to represent several kinds of knowledge , including grammars for natural languages , ontologies , class hierarchies , part - whole hierarchies , decision networks and trees , relational tuples , if - then rules , associations of medical signs and symptoms , causal relations , and concepts in mathematics and logic such as ` function ' , ` variable ' , ` value ' , and ` set ' . * _ the processing of knowledge_. the sp system has demonstrable capabilities in several areas , including natural language processing , pattern recognition , several kinds of reasoning , the storage and retrieval of information , planning , problem solving , unsupervised learning , and information compression .
it is pertinent to mention that part of the inspiration for the sp theory is research by fred attneave
( eg , * ? ? ?
horace barlow ( eg , * ? ? ?
* ) , and others , showing that aspects of visual perception ( and , more generally , the workings of brains and nervous systems ) may be understood in terms of information compression . other sources of inspiration for the sp theory include research on ` minimum length encoding ' ( eg , * ? ? ?
* ) , and evidence for the importance of information compression in the unsupervised learning of language ( eg , * ? ? ?
* ) , . ] and in mathematics and logic ( * ? ? ?
* chapters 2 and 10 ) .
at an abstract level , information compression brings three main benefits : * for any given body of information , @xmath0 , it reduces the amount of storage space required . *
reducing the size of @xmath0 can mean increases in efficiency .
it would , for example , mean less searching if we are trying to find something within @xmath0 .
* perhaps most importantly , information compression provides the key to inductive prediction . in the sp system
, it is the basis for all kinds of inference , and for calculations of probabilities . in animals
, we would expect these things to have been favoured by natural selection because of the competitive advantage they can bring .
and they are likely to be useful in artificial systems . in the sp framework ,
information compression is achieved via the discovery of recurrent patterns ( like those shown in rows 1 to 8 in figure [ kittens_figure ] and columns 1 to 6 in figure [ class_part_plant_figure ] ) , and also via the economical encoding of new information in terms of old patterns , as explained in ( * ? ? ?
* section 3.5 ) .
it is now widely accepted that , at ` low ' levels in vertebrate and invertebrate visual systems , there are processes that recognise perceptual features such as edges and corners . some relevant evidence is outlined in subsections below . in this section ,
the main focus is on features that may be regarded as ` explicit ' because they derive directly from visual input .
but it is well known that we may ` see ' things that have little or no counterpart in the visual input , such as the ` subjective contours ' in ( * ? ? ?
* figure 2 - 6 ) or the edge of one leaf where it overlaps another in ( * ? ? ?
* figure 4 - 1 ( a ) ) .
these kinds of ` implicit ' features will be considered in section [ seeing_things_that_are_not_there ] . in two respects ,
explicit perceptual features sit comfortably with the sp theory : * they may be seen to provide a means of encoding perceptual information in an economical manner .
for example , @xcite writes that `` common objects may be represented with great economy , and fairly striking fidelity , by copying the points at which their contours change direction maximally , and then connecting these points appropriately with a straight edge . '' ( p. 185 ) .
he illustrates this with the now - famous picture of a sleeping cat , reproduced in figure [ sleeping_cat_figure ] .
* at lowish levels , perceptual features may function as if they were the atomic symbols that provide the foundation for all higher - level structures , even though they themselves have been constructed from lower - level components .
as just indicated , vision begins with images as they are first projected , not perceptual features .
the latter must be somehow discovered or detected within the images .
the following subsections consider how the sp theory may be applied in this area , starting with a consideration of options for the encoding of light intensities . in the design of artificial systems for vision
, it seems natural and obvious that light intensities in images should be expressed as numbers .
but , in itself , the sp system recognises only atomic symbols that can be matched in an all - or - nothing manner with other atomic symbols .
it is true that , in principle , it may be supplied with patterns that express peano s axioms or similar information , and it may then interpret numbers correctly ( see * ? ? ?
* chapter 10 ) .
but this has not yet been explored in any depth and , in any case , numbers are probably a distraction in understanding how sp principles may be applied to vision . to simplify the discussion here
, we shall assume that we are processing monochrome images with just two categories of pixel : black and white . with that kind of representation , the lightness in any given small area may be encoded via the _ densities _ of black and white pixels in that area , without using explicit numbers .
it is true that such pixels may be represented with the symbols ` 1 ' and ` 0 ' but these are simply atomic symbols ( as required by the sp system ) , without numerical meanings .
it is relevant to this discussion to consider briefly how edges may be detected with neurons .
figure [ limulus_figure ] shows two sets of recordings from a single visual receptor ( ` ommatidium ' ) of the horseshoe crab , _
limulus_. in both sets of recordings , the eye of the crab was illuminated in a rectangular area bordered by a dark rectangle of the same size ( producing a step function as shown at the top right of the figure ) . in both cases ,
successive recordings were taken with the pair of rectangles in successive positions across the eye along a line which is at right angles to the boundary between light and dark areas .
this achieves the same effect as but is easier to implement than keeping the two rectangles in one position and taking recordings from a range of receptors across the light and dark areas .
in the top set of recordings ( triangles ) all the ommatidia except the one from which recordings were being taken were masked from receiving any light . in this case
, the target receptor responds with frequent impulses when the light is bright and at a sharply lower rate in the dark . in the bottom set of recordings ( circles )
the mask was removed so that all the ommatidia were exposed to the pattern of light and dark rectangles . in this case ,
positive and negative responses are exaggerated near the border between light and dark areas but the target receptor fires at or near a background rate in areas which are evenly illuminated ( either light or dark ) .
this kind of effect which is seen elsewhere in the animal kingdom appears to be due to lateral inhibition between neurons in the visual system ( * ? ? ?
* pp 172 - 174 ) .
it has been recognised for some time that the dampening of the response in regions of uniform illumination ( light or dark ) may be seen to achieve the effect of compressing visual information by extracting redundancy from it @xcite .
it is somewhat like the ` run - length coding ' technique for compression of information : a symbol or group of symbols that repeats in a contiguous sequence may be reduced to a single instance , perhaps marked for repetition .
, retrieved 2013 - 02 - 04 . ] a boundary between one uniform area and another may be represented economically by two such compressed representations , side - by - side . in the neural case ,
the upswing near the light / dark boundary may be seen as an economical representation of the idea that the whole of the preceding area is light , the downswing on the other side may be seen as a succinct marking of the fact that the following area is dark , while the two together may be seen to serve as a compressed representation of the boundary .
although it is less directly relevant to the present discussion , it is pertinent to mention that there are ` complex ' cells in mammalian visual systems that respond selectively to edges , and also to ` lines ' and ` slits ' ( see , for example , * ? ? ?
* pp 215 - 219 ) . in the sp framework
, the effect of run - length coding may be achieved via recursion , as illustrated in figure [ recursive_run - length_coding_figure ] . here
, each instance of `` a b c ` ' in the new pattern in row 0 is matched to an appearance of the self - referential old pattern `` x 1 a b c x 1 # x # x ` ' .
it is self - referential because `` x 1 # x ` ' in the body of the pattern may be matched and unified with `` x 1 ... # x ` ' at the start and end of the pattern .
the encoding of the new pattern which we may derive from this multiple alignment is the relatively short sequence `` x 1 # x ` . ' as before , two such encodings , side - by - side , would be an economical representation of the boundary between one uniform region and another . of course
, this does not look much like lateral inhibition with neurons , as outlined in section [ edge_detection_with_neurons_section ] .
but at an abstract level , the two things may be seen to produce the same result : the extraction of redundancy from uniform regions , leaving information about the boundaries between such regions as an economical representation of the raw data , like david marr s
@xcite ` primal sketch ' . with other developments such as the generalisation of the sp concepts to two dimensions
this kind of technique may be applied in computer vision . meanwhile , existing techniques , such as those described in ( * ? ? ?
* chapter 4 ) , may serve instead . so far
, we have said nothing about the orientations of edges or their lengths . in principle , those things may be encoded mathematically , and very economically , in the manner of computer graphics .
but that does not seem very likely in a biological system and it is not necessarily the best option for any artificial system that aspires to human - like capabilities in vision . as mentioned above , the visual cortex in mammals is populated by large numbers of ` complex ' neurons , each one of which responds to an ` edge ' , ` slit ' , or ` line ' , at a particular orientation .
there is a good coverage of different angles within each small area ( see , for example , * ? ?
* chapter 9 ) .
these observations suggests that , in natural vision , the orientation of any edge may be encoded quite simply and directly in terms of the corresponding type of neuron , and likewise in an artificial system .
a sequence of such codes would describe both the orientation and length of a line but it would contain the same kind of redundancy as is discussed in section [ recursive_run - length_coding_section ] .
so we may guess that , in natural vision , some kind of run - length coding may operate , reducing the redundancy within the body of the line and preserving information where the repetition stops at the points where the line begins and where it ends .
some relevant evidence comes from studies showing the existence of ` end stopped ' hypercomplex cells that respond selectively to a bar of a defined length , or a corner ( see , for example , * ? ? ?
* pp 216 - 217 ) . in keeping with attneave s @xcite remarks quoted earlier
, we may guess that , in mammalian vision , the orientation and length of an edge , slit or line , is to a large extent encoded via neurons that record the beginning and end of the line and any associated corners .
orientation - sensitive neurons would provide the input for this ` higher ' level of encoding .
in artificial systems , this kind of coding may in principle be done within the multiple alignment framework , as outlined in section [ recursive_run - length_coding_section ] .
as before , existing techniques may provide stop - gap solutions .
readers may , with some justice , object that real visual data is rarely as clean as the example in figure [ recursive_run - length_coding_figure ] may suggest .
most areas are some shade of grey , not purely black or purely white , and there are likely to be blots and smudges of various kinds .
what appears to be a promising answer to this kind of problem is that the sp system is designed to search for optimal solutions and is not unduly disturbed by errors of omission , commission and substitution
. there is more on this topic in section [ noisy_data_recognition_section ] ( see also section [ noisy_data_learning_section ] ) .
in some respects , object recognition is like parsing in natural language processing ( see , for example , * ? ? ? * ; * ? ? ? * ) .
since the sp system works well in parsing , as outlined in section [ multiple_alignment_section ] , it may also prove useful in computer vision .
naturally , it would be necessary for the sp machine to have been generalised to work with patterns in two dimensions . and in this discussion we shall assume that low - level perceptual features have been identified , and that they may be treated as atomic symbols , in accordance with the sp theory .
figure [ object_recognition_figure ] shows schematically how someone s face , with their ears , may be parsed within the multiple alignment framework .
row 0 in the figure contains a new pattern representing incoming information .
each part has been aligned with an old pattern representing stored knowledge of the structure of an ear , an eye , etc . and
these are aligned with a pattern in row 2 representing the higher - level structure of someone s head .
although this is schematic , i believe the approach has potential , as described in the following subsections .
contrary to the impression one might gain from figure [ object_recognition_figure ] , the sp system is quite robust in the face of errors .
this is illustrated in figure [ noisy_data_recognition_figure ] where the new pattern in row 0 is the same sentence as in figure [ kittens_figure ] but with the omission of the ` w ' in ` t w o ' , the substitution of ` m ' for `n ' in ` k i t t e n s ' , and the addition of ` x ' within the word ` p l a y ' . despite these errors , the best multiple alignment created by
the sp62 model is , as shown , the one that we judge intuitively to be ` correct ' .
this kind of ability to cope gracefully with noisy data is really essential in any system which aspires to explain or emulate our ability to recognise things despite fog , snow , falling leaves , or other things that may obstruct our view . in general terms
, the reason that the sp models can cope with noisy data is that they search for optimal solutions , without relying on the presence or absence of any particular feature or combination of features .
a strength of the multiple alignment concept is that it provides a simple but effective vehicle for the representation and processing of part - whole hierarchies , class hierarchies , and their integration .
recognition of an entity in terms of its parts is illustrated rather simply in figure [ object_recognition_figure ] and more realistically in figure [ kittens_figure ] . in the latter case ,
the sentence is divided into a noun phrase and a main verb , the noun phrase is divided into a determiner and a noun , and the noun contains the root or stem , ` k i t t e n ' , with the plural suffix , ` s ' . continuing with the feline theme but not illustrated here is the way that , in the multiple alignment framework , a cat may be recognised at several levels of abstraction : as an animal , as a mammal , as a cat , and as a specific individual , say ` tibs ' ( * ? ? ?
* figure 6.7 ) .
the framework also provides for the representation of heterarchies or cross classification : a given entity , such as ` jane ' ( or a class ) , may belong in two or more higher - level classes that are not themselves hierarchically related , such as ` woman ' and ` doctor . '
the way that part - whole relations and class - inclusion relations may be combined in one multiple alignment is illustrated in figure [ class_part_plant_figure ] . so that it fits more easily on the page .
] here , some features of an unknown plant are expressed as a set of new patterns , shown in column 0 : the plant has chlorophyll , the stem is hairy , it has yellow petals , and so on . from this multiple alignment , we can see that the unknown plant is most likely to be the meadow buttercup , _ ranunculus acris _ , as shown in column 1 . as such
, it belongs in the genus _ ranunculus _ ( column 6 ) , the family _ ranunculaceae _ ( column 5 ) , the order _ ranunculales _ ( column 4 ) , the class _ angiospermae _ ( column 3 ) , and the phylum _ plants _ ( column 2 ) .
each of these higher - level classifications contributes information about attributes of the plant and its division into parts and sub - parts .
for example , as a member of the the class _ angiospermae _ ( column 3 ) , the plant has a shoot and roots , with the shoot divided into stem , leaves , and flowers ; as a member of the family _ ranunculaceae _ ( column 5 ) , the plant has flowers that are ` regular ' , with all parts ` free ' ; as a member of the phylum _ plants _ ( column 2 ) , the buttercup has chlorophyll and creates its own food by photosynthesis ; and so on . of course , this example does not describe the visual appearance of an object .
but it should be apparent that this system , when it has been generalised to work with patterns in two dimensions , has potential as a means of representing and processing both the parts and sub - parts of an object s image , and how that information relates to any hierarchy of classes to which that object belongs . and
each of those two types of hierarchy is a very effective means of expressing visual information in a compressed form .
scene analysis may also be viewed as a kind of parsing ( see , for example , * ? ? ?
for the analysis of a seascape , for example , there may be a high - level structure recording the kinds of things that one sees in a typical seascape ( sea , beach , rocks , boats , and so on ) , with a more detailed description for each one of those things .
there seem to be two main complications in scene analysis : * any one thing may be partially obscured by another . in our seascape ,
a boat may be partially obscured by , for example , waves , sea birds , or members of the crew .
* the locations of things may be quite variable .
a boat may be in the sea or on the beach ; people can appear almost anywhere ; and so on .
of course , people cope easily with both those things , but there may be a problem with ` naive ' kinds of parsing system .
the sp framework may accommodate these aspects of scene analysis in three main ways : * as we saw in section [ noisy_data_recognition_section ] , parsing can be done successfully despite errors or omission , commission , or substitution .
thus there is reason to believe that , when the sp models have been generalised to work with patterns in two dimensions , an object may be recognised even if it is partially obscured . *
the variability of scenes is broadly similar to the variability of sentences in natural language .
artificial parsing systems , including the sp system , can cope with that variability by providing information about a wide variety of types of sentences and phrases , including recursive forms such as _ this is the man all tattered and torn that kissed the maiden all forlorn that milked the cow with the crumpled horn ... _ .
the same principles may be applied to vision . * where existing knowledge ca nt cope , the system may learn as discussed in section [ unsupervised_learning_section ] , next .
it is clear that learning is an integral part of vision since vision is an important means of gaining new information about the world . and
it is clear that , in general , we learn via vision in a manner that is ` unsupervised ' in the sense that it does not require the intervention of a ` teacher ' , or the provision of ` negative ' samples , or the grading of samples from simple to complex ( _ cf .
_ @xcite ) .
we take in information through our eyes ( and other senses ) and try to make sense of it as best we can . in this section ,
we consider unsupervised learning as it has been developed in the sp framework , and how it may be applied in vision .
but as background for what follows we first look at the ` donsvic ' principle in unsupervised learning . in our dealings with the world ,
certain kinds of structures appear to be more prominent and useful than others : in natural languages , there are words , phrase and sentences ; we understand the visual and tactile worlds to be composed of discrete ` objects ' ; and conceptually , we recognise classes of things like ` person ' , ` house ' , ` tree ' , and so on .
it appears that these ` natural ' kinds of structure are significant in our thinking because they provide a means of compressing sensory information , and that compression of information provides the key to their learning or discovery . at first sight
, this looks like nonsense because popular programs for compression of information , such as those based on the lzw algorithm , or programs for jpeg compression of images , seem not to recognise anything resembling words or objects . but those programs are designed to work fast on low - powered computers . with other programs that are slower but more thorough , natural structures can be revealed : * figure [ discovery_of_words ] shows part of a parsing of an unsegmented sample of natural language text created by the mk10 program @xcite using only the information in the sample itself and without any prior dictionary or other knowledge about the structure of language .
although all spaces and punctuation had been removed from the sample , the program does reasonably well in revealing the word structure of the text .
statistical tests confirm that it performs much better than chance .
* the same program does quite well significantly better than chance in revealing phrase structures in natural language texts that have been prepared , as before , without spaces or punctuation but with each word replaced by a symbol for its grammatical category @xcite .
although that replacement was done by a person trained in linguistic analysis , the discovery of phrase structure in the sample is done by the program , without assistance . * the snpr program for grammar discovery @xcite can ,
without supervision , derive a plausible grammar from an unsegmented sample of artificial language , including the discovery of words , of grammatical categories of words , and the structure of sentences . a key feature of both the mk10 program and the snpr program is compression of information by the matching and unification of patterns .
but much the same can be said of ordinary ` utility ' programs for data compression . what is distinctive about the mk10 and snpr programs is that they are designed to search through what is normally a wide variety of alternative ways in which patterns may be matched and unified , and to select those patterns or sets of patterns that yield relatively high levels of compression .
it seems likely that the principles that have been outlined in this subsection may be applied not only to the discovery of words , phrases and grammars in language - like data but also to such things as the discovery of objects in images , and classes of entity in all kinds of data .
these principles may be characterised as ` the discovery of natural structures via information compression ' , or ` donsvic ' for short .
although the sp theory has grown out of my earlier work on the unsupervised learning of language , the mk10 and snpr models are not well suited to the goal of simplifying and integrating concepts across several different aspects of intelligence .
it has been necessary to develop a radically new conceptual framework , with the sp concept of multiple alignment at centre - stage .
but information compression and the donsvic principles are as important in the new conceptual framework as they were before . as mentioned in section [ computer_models_section ] , the sp70 model works by creating multiple alignments , deriving old patterns from the multiple alignments , evaluating sets of newly - created old patterns in terms of their effectiveness for the economical encoding of the new information , and weeding out low - scoring sets . the first two of those processes is illustrated schematically in figure [ unsupervised_learning_figure ] .
as mentioned earlier , the sp system is conceived as an abstract system that , like a brain , may receive ` new ' information via its senses and store some or all of it as ` old ' information .
we may think of the ` brain ' as that of a baby listening to what people are saying .
let s imagine that he or she hears someone say `` t h a t b o y r u n s. '' if the baby has never heard anything similar , then , if it is stored at all , that new information may be stored as a relatively straightforward copy , something like the old pattern shown in row 1 of the multiple alignment in part ( a ) of the figure .
now let us imagine that the information has been stored and that , at some later stage , the baby hears someone say `` t h a t g i r l r u n s '' .
then , from that new information and the previously - stored old pattern , a multiple alignment may be created like the one shown in part ( a ) of figure [ unsupervised_learning_figure ] . and , by picking out coherent sequences that are either fully matched or not matched at all , four putative words may be extracted : ` t h a t ' , ` b o y ' , ` g i r
l ' , and ` r u
n s ' , as shown in the first four patterns in part ( b ) of the figure . in addition , a fifth pattern may be created , as shown in the figure , that records the sequence ` t h a t ... r u n s ' , with the category ` c # c ' in the middle representing a choice between ` b o y ' and ` g i r l ' .
this is the beginnings of a grammar to describe that kind of phrase .
this example shows how old patterns may be derived from a multiple alignment but it gives a highly misleading impression of how the sp70 model actually works . in practice , the program forms many multiple alignments that are much less tidy than the one shown and it creates many old patterns that are clearly ` wrong ' .
however , the program contains procedures for evaluating candidate sets of patterns and weeding out those that score badly in terms of their effectiveness for encoding the new information economically . out of all the muddle , it can normally abstract one or two ` best ' grammars and these are normally ones that appear intuitively to be ` correct ' , or nearly so .
as was mentioned in section [ computer_models_section ] , the sp70 model has two main weaknesses at it stands now : it does not learn intermediate levels in a grammar or discontinuous dependencies of the kind mentioned in section [ multiple_alignment_section ] .
but i believe some reorganisation of the model would solve both problems and greatly enhance the model s capabilities . as with the structures of natural language , it is clear that we have to learn the structures that are significant in vision , including objects . some insights into how this may be done may be gained from a consideration of random - dot stereograms like the one shown in figure [ stereogram_figure_1 ] . here ,
each of the two images is a random array of black and white pixels , with no discernable structure .
but there is a relationship between them , as shown in figure [ stereogram_figure_2 ] : both images are the same except that a square area near the middle of the left image is further to the left in the right image . .
reproduced from @xcite , with permission of lucent technologies inc./bell labs.,scaledwidth=90.0% ] when these images are viewed in a stereoscope , the central square appears as a discrete object suspended above the background .
the focus of interest here will be on how we come to see that discrete object , while possible implications for our understanding of depth perception are discussed in section [ space_depth_section ] .
a little analysis shows that seeing the central square means finding an alignment between pixels in the left image and pixels in the right image , that there are many alternative such alignments , and that some are better than others .
one solution is the algorithm developed by @xcite .
another solution , potentially , is the kind of processing that builds multiple alignments in the sp models , but generalised for two dimensions . as noted in section [ computational_complexity_section ] , the complexity of the matching problem can , in general , be reduced by applying constraints to the process of searching and thus reducing the size of the search space .
figure [ julesz_analogue_figure ] shows how the sp62 model can solve a one - dimensional analogue of the stereo matching problem . here
, the old pattern ( row 1 ) may be seen as an analogue of the left image and the new pattern ( row 0 ) may be seen to stand in for the right image .
both patterns have been prepared from a random sequence of digits , ) .
the results are , they say , better than with pseudo - random number algorithms because `` atmospheric noise '' is the source of randomness . ] with a displacement of the middle section , much as in figure [ stereogram_figure_2 ] .
this multiple alignment is the best of several different multiple alignments created by the sp62 model with those two patterns . in the figure
, one can see how the central sequence of 10 integers ( analogous to the central square in figure [ stereogram_figure_2 ] ) has been isolated from the ` background ' sequences to the left and right , and this despite repetitions of integers in both patterns and the formation of plenty of ` wrong ' alignments on the route to the ` correct ' result .
it seems likely that the processes can be generalised to work with patterns in two dimensions .
the kinds of processing just described may also be applied to objects in motion .
consider , for example , a flatfish with a sandy , speckled colouration , lying on a sandy and speckled area on the bed of the sea .
such a creature would be very well camouflaged but with one proviso : it must stay still .
as soon as it moves , it will become very much easier to see .
apart from the motion itself , an important reason seems to be that movement creates two images ( or more ) , rather like the two images in a random - dot stereogram . and
by a process of matching , much as described above , a predator or other observer will be able to see the fish standing out as a distinct entity with distinct boundaries like the square that can be seen when the two images in figure [ stereogram_figure_1 ] are viewed in a stereoscope . more generally , we see any object in motion such as a car travelling along a road as a single entity , not a multitude of images like the frames in a video or film . in all such cases ,
we merge the many instances into one .
the process of merging those many instances , which is likely to yield high levels of compression , requires a process of matching and unification , much as before . and
those processes serve to define the boundaries of the entity and to distinguish it from the background .
if we only ever see parts of an object perhaps a rare creature in its natural habitat that we have only seen in fleeting glimpses we can nevertheless develop a coherent concept of the whole object via alignments amongst the fragmentary views : `` a b ` ' may be aligned with `` b c ` ' and unified to create `` a b c ` ' ; `` c d ` ' may be aligned with `` d e ` ' to create `` c d e ` ' ; `` a b c ` ' may be aligned with `` c d e ` ' ... , and so on . this is like the ` sequence assembly ' technique in bioinformatics , , retrieved 2013 - 02 - 21 . ] or the stitching together of overlapping photos to create a panorama . and the matching may be achieved via multiple alignment , as developed in the sp theory .
similar things may be said about the learning of everyday concepts like ` person ' or ` house ' , or the more formal botanical categories shown in figure [ class_part_plant_figure ] .
if , for example , we see one thing with the characteristics `` a b c f l m n p x y z ` ' and another with the characteristics `` a b c g l m n q x y z ` ' , we may create a unified pattern like this : `` a b c 1 # 1 l m n 2 # 2 x y z ` ' , with the patterns `` 1 f # 1 ` ' , `` 1 g # 1 ` ' , `` 2 p # 2 ` ' , and `` 2 q # 2 ` ' , to fill in the slots .
the unified pattern may be seen to represent the class of things with the characteristics `` a b c ... l m n ... x y z ` ' .
this example is , of course , rather similar to the example shown in section [ unsupervised_learning_section ] .
that similarity is not accidental .
it derives from the principle , which is a key part of the sp theory , that , with compression of information via the multiple alignment framework , all kinds of knowledge may be represented economically with sp patterns . and
it is consistent with the long - established idea that there may be a syntax for images , not just natural languages ( see , for example , * ? ? ?
* ) , and with the previously - mentioned idea that object recognition and scene analysis may each be seen as a form of parsing ( section [ object_scene_section ] ) .
there is potential with this kind of learning to create structures that are quite subtle and expressive . despite its limitations
, the sp70 model can already discover grammatical structures with alternatives everywhere , and without any fixed elements as in `` a b c ... l m n ... x y z ` ' .
it is envisaged that , with the kind of reorganisation mentioned earlier , the system should be able discover structures that express part - whole hierarchies and class - inclusion hierarchies , both of them with multiple levels , and to abstract discontinuous dependencies in data of the kind mentioned in section [ multiple_alignment_section ] . as was noted in sections [ noisy_data_low - level_features_section ] and [ noisy_data_recognition_section ] ,
visual information is normally ` noisy ' in the sense that , compared with any stored information , it is likely to contain errors of omission , commission , or substitution , in any combination .
as shown in figure [ noisy_data_recognition_figure ] , the sp system has a capacity to cope with these kinds of errors , at least in tasks like parsing , recognition , or scene analysis .
what about learning ?
how can any system learn ` correct ' structures from noisy data in an ` unsupervised ' manner and without any help from a ` teacher ' , or from examples that are marked as ` wrong ' , or from anything else of that kind ?
this is not merely an issue in vision .
it also arises in connection with language learning , as illustrated in figure [ generalisation_figure ] . .
in ascending order size , they are : the finite sample of utterances from which a child learns ; the ( infinite ) set of utterances in @xmath1 ; and the ( infinite ) set of all possible utterances . adapted from figure 7.1 in @xcite , with permission.,scaledwidth=60.0% ]
when we learn our first language or languages , we learn from what we hear a finite sample of language shown as the smallest envelope in the figure .
but there are two apparent problems : * how we learn despite what is marked in the figure as ` dirty data ' : sentences that are not complete , false starts , words that are mis - pronounced , and more . * how we generalise from the finite sample represented by the smallest envelope to a knowledge of the language corresponding to the middle - sized envelope , without overgeneralising into the region between the middle envelope and the outer one .
one possible answer is that mistakes are corrected by parents , teachers , and others .
but the weight of evidence is that children can learn their first language without that kind of assistance .
an alternative answer favoured here is that information compression provides the key : * any particular error is , by its nature , rare and so in the search for useful patterns ( which , other things being equal , are the more frequently - occurring ones ) , it is discarded along with many other candidate structures . * as a general rule , the highest levels of compression can be achieved with grammars that represent moderate levels of generalisation , neither too little nor too much . in practice ,
the mk10 and snpr programs have been found to be quite insensitive to errors ( of omission , addition , or substitution ) in their data . and
the snpr program has been shown to produce plausible generalisations , without over - generalising @xcite .
since the principles are general , it seems likely that visual learning within the sp framework may be achieved in the face of noisy data .
as mentioned earlier , it is envisaged that , in the sp theory , all kinds of knowledge will be represented with patterns in one or two dimensions .
superficially , this seems to rule out anything with more dimensions , and suggests that there might be a need to introduce patterns with three dimensions and possibly more .
however , this has been rejected , at least for the time being , for these main reasons : * although the multiple alignment concept may in principle be generalised to patterns in three or more dimensions , it is difficult to see how it could be made to work in practice and it looks implausible as a model for any kind of structure or process in the brain . *
a tentative part of the sp theory is the idea that the cortex of the brains of mammals which is , topologically , a two - dimensional sheet may be , in some respects , like a sheet of paper on which ` pattern assemblies ' ( neural analogues of sp patterns ) may be written ( * ? ? ?
* chapter 11)as shown schematically in figure [ class_part_figure ] .
* if we exclude processes of interpretation in terms of harmonics , colours , or the like , raw sensory data may be seen to come in either one dimension ( eg sound ) or two ( eg visual images ) .
* three - dimensional structures may be represented with patterns in two dimensions , somewhat in the manner of architects drawings ( * ? ? ?
* section 13.2.2 ) . with the development of mathematical concepts within the sp framework ( * ? ? ?
* chapter 10 ) , four or more dimensions may be represented in much the same way as is done now with mathematical techniques .
this and the following two subsections consider some aspects of the visual perception of space and depth , and whether or how the sp theory may be applied . if an object is viewed from several different angles , with overlap between one view and the next ( as illustrated in figure [ 3d_object_views_figure ] ) , the several views may be stitched together to create what is at least a partial and approximate 3d model of the object .
this is similar to the piecing together of fragments to create a coherent concept , as outlined in section [ concepts_from_fragments_section ] .
as before , it may be achieved via multiple alignment as that concept has been developed in the sp theory .
the model will be partial if , for example , it excludes views from above or below . and
it is likely to be approximate because a given set of views may not be sufficient for an unambiguous definition of the object s geometry : there may be variations in the shape that would be compatible with the given set of views .
do these deficiencies matter ?
for many practical purposes , the answer is likely to be `` no '' .
if we want a rock to put in a rockery , or a stick to throw for a dog , the exact shape is not important . and
if we want more accurate information , we can inspect the object more closely , or supplement vision with touch .
evidence that people do something like what has been described is our ordinary experience that things can be harder to recognised from unfamiliar viewpoints than from familiar ones the basis of some trick photos .
that observation is confirmed in experimental studies showing that people are both slower at recognising things , and less accurate , when the viewpoint is unfamiliar @xcite .
although what has been described is like the stitching together of overlapping photos to create a panorama , the sp theory suggests that , with people , the visual information would be compressed via the encoding , within the sp system , of part - whole relations , class - inclusion relations , and other kinds of regularities .
, there is potential for the sp system to yield higher levels of compression and more natural structures . ]
that compression can be of benefit in both natural and artificial systems , as indicated in section [ compression_efficiency_prediction_section ] .
similar processes may be at work when we move around in our environment and learn about it .
successive views that overlap each other may be stitched together , as before , to create a model of the streets or other places where we have been .
this is essentially what has been and is being done with google s ` street view . ' . ]
the main difference between what has been achieved with street view and what is envisaged for the sp system is that , in the latter case , visual information would be compressed via the mechanisms in the sp system , as noted in section [ three_dimensional_objects_section ] . as with objects ( section [ three_dimensional_objects_section ] ) , a model of our environment that is created via overlapping views
may not be geometrically precise .
but , as before , some ambiguity may not matter very much for many practical purposes .
topological maps , such as the classic map of the london underground , can be quite good enough for finding one s way around .
however , if greater geometric accuracy is needed , it may be increased by gathering more information , especially information about areas between roads , paths or other routes . in connection with finding one s way around
, the sp system may be relevant in two ways : * if a robot has stored representations of one or more places , perhaps compressed via recurrent patterns as indicated in section [ compression_efficiency_prediction_section ] , then , via the building of multiple alignments ( as in section [ object_scene_section ] ) , it should be able to recognise when it has reached one of those places , using incoming visual information as new patterns and stored knowledge as old patterns . if it has stored information about an entire route or network of routes , then , within that environment , it should be able to identify where it is at any time .
similar things may be true of people . * with an appropriate set of old patterns , each one of which represents a direct connection between two places , the sp system , via the building of multiple alignments ,
can work out one or more routes between any two of the relevant places , including routes via two or more of the direct connections ( * ? ?
* chapter 8) .
the example in figure [ planning_figure ] shows one such flying route between beijing and new york . .... 0 1 2 3 4 beijing ------------------------- beijing 1 melbourne - melbourne 10a cape_town ------------- cape_town 14 paris ---- paris 25a new_york - new_york 0 1 2 3 4 .... these points about how we may build a model of our environment and find our way around relate to the topic of ` simultaneous localization and mapping ' ( slam ) in robotics . , retrieved 2013 - 03 - 07 . ] without attempting a comprehensive discussion of the complex subject of depth perception , this section offers some thoughts about stereoscopic vision , and the possible relevance of the sp theory . for any given object that we are looking at
, we can in principle work out its distance by a process of triangulation like that which has been widely used in cartography , at least as it used to be .
but there appear to be snags : * for this mechanism to work with reasonable accuracy , it would be necessary for one to have a rather accurate sense of the direction of gaze for each eye and the angle between that direction of gaze and the line between the two eyes .
it seems unlikely that we can sense the positions of our eyes with the necessary accuracy .
* there is evidence that , with the ames distorted room illusion , , the illusion persists when people view the room with two eyes @xcite , although , in that case , the effect may be reduced @xcite .
this suggests that any information about distance that may be gained via triangulation is not sufficiently clear or precise to overcome viewers preconceptions that the room has the conventional rectangular form .
* triangulation can not work with a stereoscope or a 3d film because what we are looking at is all at one distance , with nothing to differentiate one part of the picture from another .
the spear which makes us jump as we see it coming towards us out of a 3d film is no closer to us than anything else in the film .
we can not rule out triangulation altogether it may have a role in some situations but some other mechanism is needed to explain how we see depth with a stereoscope or a 3d film . with random - dot stereograms , it is clear that our brains are capable of forming an alignment between the left and right images that is good enough to identify the displaced area in the middle as a discrete entity ( section [ discovery_via_stereo_matching_section ] ) . by identifying the displaced area and distinguishing it from the surrounding area , we may also gain an accurate knowledge of the size of the displacement . how can the size of the displacement tell us about depth ?
there are at least three possible answers ( which are not necessarily mutually exclusive ) : * for any given displacement , our brains perform a geometrical calculation of what that displacement implies about relative distances , between the observer and the perceived object , and between the perceived object and the background .
* we are born with knowledge that is , in effect , a table of associations between displacements and distances .
* we learn those kinds of associations from experience .
that learning is important is suggested by the powerful influence of our experience ( of rectangular rooms ) in the ames room illusion .
building up a knowledge of associations is part of what the sp system is designed to achieve .
the sp theory has things to say about some other aspects of vision , as discussed in the following subsections .
as noted in section [ low_level_features_section_section ] , we often ` see ' things that are not objectively present in what we are looking at .
we may see ` subjective contours ' in certain kinds of images , or we may see the edge of a leaf where it overlaps another leaf despite there being little or nothing to mark the boundary .
the multiple alignment in figure [ kittens_figure ] provides an example of how the sp system may accommodate these kinds of things . here , the new pattern is the sentence ` t w o k i t t e n s p l a y ' with nothing to mark the boundary between one word and the next .
but those boundaries are clearly marked via the parsing of the sentence into its constituent parts .
more generally , we infer things that are not immediately visible : when we see the unbroken shell of a hazel nut , we expect to find an edible kernel inside ; when we see a horse partially obscured by a tree , we expect to see the whole animal when it moves into full view ; and so on . this kind of inference is an integral part of how the sp system works . in figure
[ noisy_data_recognition_figure ] , the word ` t w o ' appears in the new pattern as ` t o ' , but the parsing interpolates the missing ` w ' . in figure
[ class_part_plant_figure ] , the rather sketchy information in column 1 is extended via the information in columns 1 to 6 : we can infer that the plant photosynthesises ( column 2 ) , that it has five petals ( column 6 ) , that it is poisonous ( column 5 ) , and so on .
a prominent feature of natural vision is that we can recognise something despite wide variations in viewing distance and corresponding variations in the size of the retinal image .
although this phenomenon is not consistent with any simple pattern - matching model of vision , it appears that it can be accommodated within the sp theory .
let us suppose that , as described in section [ recursive_run - length_coding_section ] , the image to be processed is reduced to a ` primal sketch ' , showing boundaries between uniform areas but without the redundancy within those areas . for any given scene , the effect of that processing will be to reduce or eliminate variations in the size of the original image .
the primal sketch that is derived from a large version of the scene will be much the same as the primal sketch that is derived from a small version .
any residual variations in size , or noise in the image , may be overcome by the flexibility of the matching process in the sp system ( section [ computer_models_section ] ) and by the system s ability to tolerate noise ( sections [ noisy_data_low - level_features_section ] , [ noisy_data_recognition_section ] , and [ noisy_data_learning_section ] ) .
another prominent feature of natural vision is ` lightness constancy ' : the fact that , normally , we perceive the lightness of an object to be fixed , despite wide variations in the incident light and corresponding variations in the amount of light that is reflected from the object ( its ` luminence ' ) .
we would normally see a lump of coal as black and snow as white , even though the coal in bright sunlight may be reflecting more light per unit area than snow in shadow . in order to account for this phenomenon
, it seems necessary to suppose that , for each kind of object , we maintain some kind of table of associations between levels of illumination and corresponding values for luminance .
since we are unlikely to have an inborn knowledge of coal , snow , and the like , we must suppose that those tables are learned .
as noted in section [ alternatives_section ] , learning associations of that kind is part of what the sp system is designed to achieve .
notice that any given table can only be applied if we have some idea of what kind of object we are looking at , otherwise we might see coal as if it was snow , or _
vice versa_. there is some evidence that our perception of the lightness of an object does indeed depend on what we think the object is ( * ? ? ?
* chapter 16 ) . in a similar way
, our judgements of lightness seem to depend on our perceptions of how a given object is illuminated ( * ? ? ?
* figure 1.10 ) .
it seems likely that much of what has been said in this section about lightness constancy would also apply to colour constancy : the way we see the colour of an object to be fixed , despite wide variations in the colour of the incident light and corresponding variations in the colour of the light that is reflected from the object .
since information compression is central in the sp theory , it is pertinent to mention that lightness constancy and colour constancy may each be seen as a means of encoding information economically .
it is simpler to remember that a particular object is ` black ' or ` red ' than all the complexity of how its appearance changes in different lighting conditions .
it is often remarked that we recognise things more easily in their familiar contexts than in unfamiliar ones , and this is confirmed in formal studies ( see , for example , * ? ? ?
* ; * ? ? ?
this observation makes sense in terms of the sp framework because any part of a multiple alignment may be a context for any other , and because of the way the system searches for a global optimum which embraces any given entity and its context .
if , in our seascape example ( section [ scene_analysis_section ] ) , we see a beach and the sea then , in effect , we are primed to see boats because , in that context , boats are likely to yield multiple alignments with better scores than , say , office furniture .
a less common observation is that , with some kinds of image , there is more than one plausible interpretation .
an example is the ` young woman / old woman ' picture of psychology text books . in the sp framework
, this kind of ambiguity is accommodated in the way that , with some kinds of data , the system may create two or more multiple alignments that have good scores .
an example in the area of natural language processing is the way the sp62 model can produce two parsings corresponding to both readings of the ambiguous sentence
_ fruit flies like a banana _ , as shown in ( * ? ? ?
* figure 5.1 ) .
it is clear that in people and other animals , vision does not stand alone but works in close association with other senses .
our concept of a ship , for example , is an amalgam of images , sounds , smells , the flavour of food on board , textures of different surfaces , and so on . in a similar way
, vision works closely with other aspects of intelligence : different kinds of reasoning , learning , understanding and producing natural language , recalling information , and non - visual kinds of recognition .
achieving these kinds of integration without undue complexity has been a central aim in the development of the theory . and in that development , many candidate ideas have been rejected because they did not help to promote the simplification and integration of concepts .
to the extent that the theory achieves a combination of simplicity with versatility , it is down to three main things : representing all kinds of knowledge with ` patterns ' ; the multiple alignment concept as it has been developed in the sp theory ; and the overarching role of information compression via the matching and unification of patterns .
despite some limitations in how the sp theory is currently realised in computer models , it has what i believe are some useful things to say about several aspects of vision : * low level perceptual features such as edges or corners may be identified by the extraction of redundancy in uniform areas in a manner that is analogous to the run - length encoding technique for information compression , and comparable with the effect of lateral inhibition in the visual systems of animals . *
the concept of _ multiple alignment _ in the sp theory may be applied to the recognition of objects , and to scene analysis , with a hierarchy of parts and sub - parts , and at multiple levels of abstraction .
* the theory has potential for the unsupervised learning of visual objects and classes of objects , and suggests how coherent concepts may be derived from fragments .
it provides an account of how we may discover objects via stereo matching and via motion . * as in natural vision , both recognition and learning in the sp system is robust in the face of errors of omission , commission and substitution .
* the theory suggests how , via vision , we may piece together a knowledge of the three - dimensional structure of objects and of our environment that is good enough for many practical purposes , despite ambiguities in geometry .
* the theory provides an account of how we may see things that are not objectively present in an image , and how we may recognise something despite variations in the size of its retinal image .
* the theory has things to say about the phenomena of lightness constancy and colour constancy , about the role of context in recognition , and about ambiguities in visual perception .
a strength of the sp theory is that it is not simply a theory of vision .
it provides for the integration of vision with other sensory modalities and with other aspects of intelligence such as reasoning , planning , and problem solving .
h. b. barlow .
sensory mechanisms , the reduction of redundancy , and intelligence . in hmso ,
editor , _ the mechanisation of thought processes _ , pages 535559 .
her majesty s stationery office , london , 1959 . c. farabet , c. couprie , l. najman , and y. lecun .
scene parsing with multiscale feature learning , purity trees , and optimal covers . in _ proceedings of the 29th international conference on machine learning , edinburgh , scotland , uk , 2012 _ , 2012
. w. l. gehringer and e. engel .
effect of ecological viewing conditions on the ames distorted room illusion .
_ journal of experimental psychology : human perception and performance _ , 120 ( 2):0 181185 , 1986 .
a. glennerster , s. j. gilson , l. tcheang , and a. j. parker . perception of size in a ` dynamic ames room ' .
_ journal of experimental psychology : human perception and performance _ , 30 ( 9):0 490a , 2003 .
doi : 10.1167/3.9.490 .
bottom - up / top - down image parsing by attribute graph grammar . in _ proceedings of the tenth ieee international conference on computer vision ( iccv 2005 ) , 17 - 21 oct .
2005 _ , volume 2 , pages 17781785 , 2005 .
d. marr .
_ vision : a computational investigation into the human representation and processing of visual information_. the mit press , london , england , 2010 .
this book was originally published in 1982 by w. h. freeman and company .
m. j. tarr .
rotating objects to recognize them : a case study of the role of viewpoint dependency in the recognition of three - dimensional objects .
_ psychonomic bulletin and review _ , 20 ( 1):0 5582 , 1995 .
j. g. wolff . learning syntax and meanings through optimization and distributional analysis . in y. levy , i. m. schlesinger , and m. d. s. braine , editors , _ categories and processes in language acquisition _ , pages 179215 .
lawrence erlbaum , hillsdale , nj , 1988 .
see : http://bit.ly/zigjyc[bit.ly/zigjyc ] .
j. g. wolff .
_ unifying computing and cognition : the sp theory and its applications_. cognitionresearch.org , menai bridge , 2006 .
isbns : 0 - 9550726 - 0 - 3 ( ebook edition ) , 0 - 9550726 - 1 - 1 ( print edition ) .
distributors , including amazon.com , are detailed on http://bit.ly/wmb1rs[bit.ly/wmb1rs ] .
the publisher and its website was previously cognitionresearch.org.uk . | the _ sp theory of intelligence _ aims to simplify and integrate concepts in computing and cognition , with information compression as a unifying theme .
this article discusses how it may be applied to the understanding of natural vision and the development of computer vision .
the theory , which is described quite fully elsewhere , is described here in outline but with enough detail to ensure that the rest of the article makes sense .
low level perceptual features such as edges or corners may be identified by the extraction of redundancy in uniform areas in a manner that is comparable with the run - length encoding technique for information compression .
the concept of _ multiple alignment _ in the sp theory may be applied to the recognition of objects , and to scene analysis , with a hierarchy of parts and sub - parts , and at multiple levels of abstraction .
the theory has potential for the unsupervised learning of visual objects and classes of objects , and suggests how coherent concepts may be derived from fragments . as in natural vision
, both recognition and learning in the sp system is robust in the face of errors of omission , commission and substitution .
the theory suggests how , via vision , we may piece together a knowledge of the three - dimensional structure of objects and of our environment , it provides an account of how we may see things that are not objectively present in an image , and how we recognise something despite variations in the size of its retinal image . and it has things to say about the phenomena of lightness constancy and colour constancy , the role of context in recognition , and ambiguities in visual perception .
a strength of the sp theory is that it provides for the integration of vision with other sensory modalities and with other aspects of intelligence .
_ keywords _ : vision , information compression , artificial intelligence , perception , cognition , representation of knowledge , learning , pattern recognition , natural language processing , reasoning , planning , problem solving . |
this article concerns relationalism .
this word is used very differently by barbour ( and various collaborators ) and i @xcite on the one hand , and by crane @xcite and rovelli @xcite ( with subsequent use in loop quantum gravity ( ` lqg ' ) @xcite ) on the other hand .
this article is mostly about the former ( a successor to / considerable extension of @xcite ) , though it does contrast , and to some extent compose , this with the latter .
moreover , the position developed here most ly lies _ between _ barbour s @xcite and rovelli s @xcite extremes regarding the fine detail of the machian notion of time as being abstracted from change .
as regards other recent reviews , pooley s @xcite is a 3-way comparison of which three barbour s alone features in this review ( i go for the approaches that enter quantum gravity programs , whereas pooley goes for the past 2 decades philosophizing ) .
barbour s article @xcite and gryb s thesis @xcite largely only cover barbour s approach .
in sec 2 , i sketch the history of the absolute versus relational motion debate , which remains an active source of contention between a number of different approaches to the unsolved problem of quantum gravity ( ` qg ' ) today .
in fact , i consider qg to stand for _ quantum gestalt _ , a term that i coin in recognition of gr being not only a relativistic theory of gravitation but also , following einstein , a freeing of physics from absolute / background structure ; the expression ` quantum gravity ' fails to capture the latter .
thus i am taking ` quantum gestalt ' to be shorthand for quantum theory of the relativistic gravity background independence gestalt .
certainly most lqg practitioners agree that what they are studying is also intended as a quantum theory of background independence ; this is also to be understood of some approaches to m - theory ( but not of perturbative string theory ) .
[ as to other names , ` quantum gr ' will not do in this role since it implies the specific einstein field equations , whereas quantum gestalt remains open - minded as to which relativistic theory of gravitation is involved ; this distinction parallels that between ` quantum gr ' and ` quantum gravity ' .
quantum gestalt is also in contradistinction with ` background independent quantum gravity ' since the latter carries connotations of the background independent and gravitational parts being _ split _ , whereas in the former they are held to be _
joint_. finally , much as one can model quantum gravity without background independence ( at least for some purposes ) , one can likewise model quantum background independence without relativistic gravitation ; from the quantum gestalt perspective , these two ventures are complements , suggesting that one may learn considerably from the latter as well as from the former .
relational particle mechanics ( rpm ) @xcite , once quantized @xcite , is such a theory of quantum background independence without relativistic gravitation . ]
sec 2 also suggests some simple informal interpretations that might be taken to underlay some of the leibniz- and mach - type ideas that are themselves presented in subsequent sections . in sec 3 ,
i consider configuration space @xmath1 and barbour s ascribing primality to it in his relational program @xcite . whilst leibniz @xcite argued for time not to be primary , mach @xcite argued for time to be an emergent concept , more specifically that ` time is to be abstracted from change ' .
this program s subsequent build - up ( see also @xcite ) of actions that are temporally relational ( leibnizian ) and configurationally relational with respect to some group @xmath2 of transformations that are held to be unphysical , is reviewed and generalized in secs 4 to 8 .
this includes using jacobi
synge @xcite type actions , dispensing with velocity as a primary notion and recasting the canonical definition of momenta accordingly , as well as using barbour and bertotti s @xcite indirect implementation of configurational relationalism and the jacobi barbour
bertotti realization of a machian emergent time .
sec 8 discusses the alternative direct implementation @xcite of configurational relationalism that is occasionally available .
sec 9 provides a number of examples , including 1- and 2-@xmath3 rpm s for which the direct implementation is available @xcite .
it also covers how , whilst gr is usually formulated in a spacetime way that does not directly implement mach s ideas as understood in the relational program , can nevertheless be cast in a manifestly relational form @xcite that is quite akin to geometrodynamics , and even more akin to the baierlein sharp
wheeler @xcite and christodoulou chronos principle @xcite formalisms .
moreover , whilst this lmb ( leibniz mach
barbour ) program s literature usually uses this more traditional dynamical formulation of gr , i lay out how the ashtekar variables approach is also lmb - relational .
i also provide some comments on relationalism versus supersymmetry ( see @xcite for more ) , string theory and m - theory . as such
, i make contact between lmb relationalism and much of modern theoretical physics .
given these examples , i am in a position to comment further on compatibility restrictions between the @xmath1 s and the @xmath2 s in sec 10 .
i then consider four expansions of barbour s ` @xmath1 is primary ' postulate : to ( perhaps weakened ) phase space in sec 11 , to categorization in sec 12 , to propositioning in sec 13 and to ` perspectivalism ' in sec 14 .
categorizing means considering not only object spaces but also the corresponding morphisms and then functors between such pairs .
perspecting means considering the set of subsystem perspectives ; this is an arena in which the lmb and rc approaches make contact . by propositioning , i mean considering the set of propositions about a physical ( sub)system ; this makes contact with ideas of isham , doering and butterfield @xcite concerning applying topos theory to quantum theory ) .
perspecting makes substantial contact with the other main line of work termed ` relational ' in the quantum gravity literature , i.e. rovelli and crane s works . in sec 15
i consider a substantial generalization of barbour s ` best matching ' indirect implementation of configurational relationalism , that also happens to generalize the mathematically well - known notion of group - averaging .
i next consider the corresponding situation at the quantum level in secs 16 to 21 .
this is entwined with a number of facets of the notorious problem of time and of relevance as regards selection of strategies to attempt to resolve this .
sec 16 covers temporal and configurational relationalism within a quantum scheme along the lines of isham 1984 @xcite .
i include analysis of the problem of time facets and strategies in accord to the extent to which their attitude to time is leibniz mach .
i illustrate this with examples in sec 17 .
i then consider @xmath1 primality at the quantum level in sec 18 , provide arguments against considering quantization in general as a functor in sec 19 , and discuss qm perspectivalism in sec 20 with rovelli and crane inputs and qm propositioning in sec 21 with contact with isham and doering s work .
my conclusion ( sec 22 ) covers the presently - known principal applications of the current article .
1 ) pointers relating machian time interpretations to the formulations of timeless approaches @xcite , 2 ) other problem of time applications , 3 ) the issue of whether scale should be among the concepts discarded by relationalism ( which argument is tied to the semiclassical approach to quantum cosmology ) .
4 ) this article comments on the foundations of , and relational notions applicable in , ashtekar variables / lqg .
5 ) it also suggests contention between relationalism and supersymmetry @xcite .
6 ) finally , i outline various possibilities for unifying isham doering propositioning with each of rovelli crane perspecting and records theory .
this debate has been around since the inception of newtonian mechanics . in outline , newton s @xcite traditional formulation of mechanics defined ` true ' motion to be relative to an absolute space that is all - pervasive , infinite , invisible and
can not be acted upon .
newton also considered motion to occur _ in _ time .
his notion of time was additionally an absolute one .
i.e. , external , continuous , and uniformly flowing as ` needed to transform kinematic geometry into physical dynamics ' @xcite .
( in fact , much of this conception of time was not new to newton , c.f .
barrow @xcite , gassendi @xcite and even ptolemy @xcite . )
the virtue of newton s scheme is working well in practise , nature immediately vindicating its practical consequences in detail .
however , newton s formulation of mechanics has been argued to be bankrupt from a philosophical perspective .
absolute space does not comply with leibniz s famous * identity of indiscernibles * @xcite , by which e.g. our universe and a copy in which all material objects are collectively displaced by a fixed distance surely share all observable properties and thus are one and the same . also , leibniz already considered time not to be a separate external entity with respect to which things change .
bishop berkeley @xcite and mach @xcite ( see also e.g. @xcite for further discussion ) added to these arguments , e.g. mach holding absolute space to be a non - entity on account of its not being actable upon .
the alternative is for mechanics to be * relational * : spatial properties are to be entirely about the relations between material objects , and ` time to be abstracted from change ' ( mach s notion ) .
historically , there was a lack in alternative theories with such features , though the comparatively recent rpm s of barbour bertotti ( 1982 ) and barbour ( 2003 ) that are the subject of this article do now serve as such .
furthermore , leibniz , bishop berkeley and mach s arguments are philosophically compelling enough that they ought to apply to not just mechanics but to physics as a whole .
as argued in sec 9 , gr is also relational , and indeed relationalism is one foundation from which gr _ can _ be derived , for all that this was not the historical route to gr . for , while einstein was interested in ` machian issues ' , the way in which he viewed these does not coincide with the barbour - relational view @xcite , nor did einstein s historical route to gr @xcite constitute a _ direct _ implementation of machian ideas
( see sec 9 ) .
see e.g. @xcite as regards arguments for ( and against ) the value of barbour and bertotti s 1982 theory toward the absolute or relative motion debate .
relationalism would appear to have a good case for use in whole - universe situations , both from a philosophical perspective and because gr and at least some further gr - like theories do implement it .
( these are good theories for whole - universe situations on a number of further grounds . ) some intuitive if vague underlying relational principles are as follows .
relationalism 0 ) * relational physics is to be solely about the relations between tangible entities * this is taken to be the universalization of sec 2.1 s statement of relationalism for mechanics to all physics .
i use ` * tangible entities * ' in place of ` material objects ' to make it clear that this is open to fields and ` force mediators ' as well as ` matter building blocks ' , as befits modern physics .
the key properties of tangible entities are , rather , as follows .
relationalism 1 ) [ mach ] these * act testably and are actable upon*. relationalism 2 ) [ leibniz ] any such which are * indiscernible are held to be identical*. note 1 ) that which is not testably acting or actable upon is held to be , rather , a non - entity .
[ this is still held to be a type of entity as regards being able to philosophize about it or mathematically represent it , just not a _ physical _ entity ; absolute space is an obvious archetypal example . ]
this is less vague than saying ` things ' , which could be _ merely _ mathematical in nature .
note 2 ) as regards relationalism 2 ) , in physics , relationalism posits that physical indiscernibility _ must _ trump multiplicity of mathematical representation ; this multiplicity still exists mathematically , but the true mathematics corresponding to the physics in question is the equivalence class spanning that multiplicity .
a range of attitudes to the identity of indiscernibles and alternatives to it are considered by butterfield and caulton @xcite one would then only wish to attribute physical significance to computations of tangible entities which succeed in being independent of the choice of representative of the equivalence class ; the archetype of this is gauge theory .
chapter 2 of @xcite also argues that dimension is partly undefined for a universe consisting of a sufficiently small number of particles .
an important lesson from there is that a set of part - tangible and part - non entities is often altogether more straightforward to represent mathematically .
note 3 ) in this article s main sense of relational , one treats instantaneous configurations and time separately as befits the great conceptual heterogeneity between them . note 4 )
one largely accepts that the mathematics will be in general be most straightforward for mixtures of part - tangible- and part - non - entities .
note 5 ) i do not take this ssec s principles too seriously ; it is the standard use of configuration space , and relationalism 46 ) and 8) , perhaps supplemented by relationalism 3 ) and 7 ) , that are the actually mathematically - implemented starting point of the lmb approach ( and variants ) . as such , relationalism 02 ) for
now being vague or loose relative to relationalism 3 ) to 8) is not a problem for the work actually done in this article , but only for this present ssec , which represents a tentative under_laying _ rather than a rigorous foundational under_lying_.
as regards the instantaneous configuration entities of a given theory , which i denote by @xmath4 , a key further concept is the _ configuration space _ @xmath1 , i.e. the space of all the possible instantaneous configurations ( each of the points in configuration space represents one state of the instantaneous configuration , i.e. a set of one value per @xmath4 ) .
example 1 ) particle positions are configurations ; see @xcite for a clear exposition of the configuration space notion for these .
example 2 ) the more reduced inter - particle relative separations or relative cluster separations ( see sec 9.1 ) are configurations . example 3 ) sec 9.2 covers even more reduced relational mechanics configurations .
example 4 ) the values at each point of continuous extended objects , whether fields or geometrical objects that have values everywhere in space , which include both the 3-metrics ( or the more reduced 3-geometries ) of gr ( see sec 9.4 ) and objects that have their own separate _ notion of space ( nos ) _ of extent such as strings or membranes .
i denote the general nos by @xmath5 for @xmath6 the dimension of space , and i coordinatize it by @xmath7 .
however , in this article all specific examples of this are for @xmath8 ( field theories ) , in which case i drop the @xmath9 subscripts , or for @xmath10 , in which case @xmath11 is just a point corresponding to the particle position , which point ceases to possess any internal coordinatization .
i denote integration over whatever nos is appropriate by @xmath12 .
i take this to collapse to a multiplicative 1 for particle mechanics and other finite theories .
i denote finite theory cases instantaneous configurations by @xmath13 . for field theory cases ,
i denote the instantaneous configurations by @xmath14 .
thus the typeface object @xmath15 is a portmanteau of the upright space - dependent case for field theories and the slanty case for finite theories .
that is the default notation in this article as regards typeface , straight and slanty symbols .
such portmanteaux serve as a shorthand covering both cases at once to encompass both this article s rpm models and the geometrodynamical formulation of gr that they are closely analogous to .
the @xmath16 in use above is a suitable multi - index , over particle and/or continuous extended object species . as the portmanteau for @xmath17 for finite theories and @xmath18
$ ] for infinite theories , where round brackets indicate function dependence , square brackets indicate functional dependence and ( ; ] is mixed function - functional dependence , with the former prior to the semi - colon and the latter after it .
i use braces for actual brackets .
i then use a special font ( large mathfrak ) for such portmanteaux that come integrated over their corresponding notion of space ( the action @xmath19 is a such ) ; these are therefore pure functionals in terms of what they depend on , but differ in the portmanteau way in terms of what features on the computational side of the equation . ]
configuration space is generally not just a set , being , rather , further equipped with topological space , metric space and geometrical structure .
a broad example of the last of these is finsler metric geometry @xmath20 for @xmath21 a topological manifold and @xmath22 the metric function . to denote ` space of ' .
i use lower - case greek indices for spatial indices 1 to @xmath3 .
i use bold font as shorthand for configuration space indices and underline for spatial indices . ]
( semi-)riemannian geometry is then a common subcase of this which is usually written in terms of the metric @xmath23 itself : @xmath24 .
some of the relational program s examples are in fact infinite - dimensional generalizations of such geometries .
( but i still refer to such by the usual finite - dimensional geometries names , i.e. i elevate those names to be finite and field - theoretic portmanteaux of the usual finite notions ) .
relationalism 3 ) [ barbour ] this concerns the possibility that the * configurations are the primary entities*. i.e. the configuration space @xmath1 has a primary role .
note 1 ) relationalism 3 ) is technically a not a relationalist postulate but a * minimalist * one ; such go quite often hand in hand with relational postulates .
this does not necessarily mean that relationalists discard structures , but rather that they are prepared to consider the outcome of doing so .
some relational ideas are a subset of minimalism , e.g. doubting and discarding absolute space and absolute time .
see also secs 10 and 16 as regards whether scale is to be discarded from physics , secs 2 and 9 as regards relationalists questioning of minkowski s view on spacetime and its subsequent primacy over how gr is conceived of .
[ i.e. arguing that the dynamical tradition favours the primality of space , and that this , despite amounting to supposing less structure , nevertheless turns out to serve as an alternative foundation for gr . ]
relationalism 4 ) [ barbour ] one aims to build up * relational actions * from natural compound objects derived from the @xmath15 .
note 1 ) in making use of actions , we are working with , and drawing upon , the principles of dynamics .
many of the steps below follow from this as natural consequences of having an action , rather than of relationalism per se .
but the particular actions used embody relationalism , and the some of the subsequent principles of dynamics workings assist in enforcing that .
this valuing and making specific use of the principles of dynamics differs from mach s own thinking ; it is due to barbour and concerns positing that relational actions are the most fundamental starting - point for ( classical ) whole - universe physics .
note 2 ) this is a mathematical _ structural composition _
postulate ; it is relational only in that the primary objects themselves are to be taken to be relational . note 3 )
the compound objects readily include nos derivatives
( whenever the nos is nontrivial ) and contracted objects . from what the reader knows about actions
, one might expect velocities to be required too , and these involve an incipient notion of time , so we need to discuss that next before making progress with the construction of actions .
moreover , the manner of that progress may well be unexpected to the reader .
in particular , actions conventionally involve velocities .
relationalism 5 ) [ leibniz ] a physical theory is * temporally relational * if there is * no meaningful primary notion of time for the system as a whole * ( e.g. the universe ) @xcite .
this is implemented by \1 ) * manifest parametrization irrelevance * , alongside \2 ) * freedom from extraneous time - related variables * , such as external absolute newtonian time or gr s ` lapse ' variable , at the level of the action principle . perhaps the reader is objecting that the preceding ssec does not help with obtaining a notion of velocity since it denies rather than provides a primary notion of time for the system as a whole . a simple way of getting round this is to point out that mpi actions are trivially equivalent to actions that are @xmath25 ) * manifestly reparametrization - invariant ( mri)*. [ for mri , any choice of ( monotonic ) parameter will do , so that choice is empty , and any such choice is equivalent to making no choice by not parametrizing i.e. using the mpi action . ] next , mri _ has _ a parameter @xmath26 , so that , for all that it is a non - entity , one can interpret it as a _ label time _
non - entity , and write down : = / ( ) ^/ .
one can then straightforwardly build the kinetic term @xmath27 .
i am assuming for now that this takes the most physically standard form that is homogeneous quadratic in the velocities ; i lift this assumption later on in this ssec .
@xmath23 is the configuration space metric portmanteau of @xmath28 for finite theories and @xmath29 for infinite theories ( this is assuming _ ultralocality _ , i.e. no derivative dependence , collapsing the object from a functional to a mere function , which mathematical simplicity holds over the entirety of the standardly accepted fundamental theories of physics ) .
its determinant is @xmath30 and its inverse is @xmath31 .
@xmath32 denotes the corresponding @xmath33-`norm ' .
@xmath34 , the portmanteau of @xmath35 in the finite case and @xmath36 in the field - theoretic case . for functional derivatives ,
@xmath37 as ordinary , partial derivative portmanteau , @xmath38 as partial , functional derivative portmanteau , @xmath39 as ordinary , functional derivative portmanteau .
i hang ` cov ' , ` abs ' suffices on these for covariant and absolute derivatives , using @xmath40 for @xmath41 and @xmath42 for @xmath43 in these contexts .
finally , i use @xmath44 for laplacians ( suitably suffixed to say of which type ) , @xmath45 for measures and oversized @xmath46 for variations . ]
one can then form the lagrangian portmanteau , @xmath47 of the lagrangian @xmath48 and the _ ultralocal _ lagrangian density @xmath49 for field theories , and thus , integrating over @xmath26 and the nos , the relational action .
compliance with mri does make the lagrangian portmanteau in question _ look _ somewhat unusual , i.e. not be of difference - type form @xmath50 , but rather of _ product form _ @xmath51 .
for now we take on trust that this @xmath52 for @xmath53 , where @xmath54 is the potential energy portmanteau : potential @xmath55 for finite theories and potential density @xmath56 $ ] for field theories .
thus it is assumed to be a ) independent of the velocities [ this is another mathematical simplicity that happens to be in accord with the standardly - accepted fundamental physics ] .
b ) time - independent [ which makes good sense for fundamental whole - universe setting as opposed to the setting for dissipative subsystems or approximate modelling e.g. involving friction ] .
this time - independence of the potential , alongside time - independence of the kinetic metric is part of what is covered by the ` no extraneous time - like objects ' part of the given mathematical implementation of temporal relationalism .
@xmath57 is the energy - like portmanteau : total energy @xmath58 for finite theories and some kind of total energy density for field theories ( ` density ' here includes a jacobian factor ) .
[ @xmath59 and @xmath60 are , for now , just to be considered as the notational tidyings defined above ; we will justify this trust in sec [ sssec : b ] . ]
the mri relational action is thus ^ _ = _ _ p_p^ _ = 2__p_p , [ action ] where the j stands for ` jacobi ' since
the finite case of this is jacobi s principle .
a more manifestly relational manoeuvre , however , is to acknowledge that if there is no meaningful notion of time , then conventional physics definition of velocity , , [ flutt ] there is consequently no meaningful primary notion of velocity !
subsequently , nothing defined in terms of velocities such as kinetic terms , lagrangians or canonical momenta , is guaranteed to meaningfully exist either .
what _ does _ have tangible physical content , however , are the ^ themselves , so that velocities are replaced by differentials of the configuration variables .
i.e. , changes _ in time _ are meaningless in relational physics , the tangible content resides , rather , among the changes of one configuration variable with respect to another , _ 1/_2 .
[ erina ] as a follow - up of this dethroning of velocity from the realm of relational physics , the usual homogeneous quadratic kinetic term is supplanted by d@xmath61/2 for @xmath62 .
i note that this is a purely geometrical quantity .
the action is , in complete generality , to be homogeneous linear in the d(configuration variable ) so as to comply with mpi / mri ; in the homogeneous quadratic case this amounts to taking the square root .
then expression under the action s integral in the mpi action is of the form @xmath63 , where the tilde indicates a rescaling by some weight that is homogeneous of degree zero in the d@xmath64 .
[ this is usually taken to be independent of the weights so as to match the conventional approach to fundamental physics having velocity - independent potential . ]
i use @xmath65 and then the action is ^ _ : = _ _ p .
[ generalaction ] such assembly of an action may be subject to some limitations , whether from implementing physical or philosophical principles ( see e.g. sec [ sssec : 1.1.4 ] ) , or from purely mathematical simplicity postulates such as ` no derivatives higher than first ( or , in a few special cases , that include gr , second ) ' . the above particular action coming from the homogeneous quadratic physics only assumption is of _ jacobi type_. in the case of mechanics itself , the nos integration collapses to 1 and the above is precisely the jacobi action principle , which is well - known and established to be cleaner in a number of ways than euler
lagrange s ( alongside having a number of relational whole - universe features and applications ) . in parallel to the lagrangian portmanteau , @xmath66 and @xmath63 are in fact portmanteaux of arclength elements in the finite case and nontrivially riemannian elements ( no signature restriction implied ) in the field - theoretic case .
the mpi formulation does not have a primary concept of lagrangian - it has the arc element instead . in other words , it is a ` * geodesic principle * ' i.e. a variational principle that takes the mathematical form of finding the geodesic corresponding to a riemannian ( or sometimes more general ) geometry . by the above , and
how the integrand being a product and not a difference like the conventional lagrangian does _ not _ unacceptedly alter the physics resulting from the action , are explained below .
the mri or mpi action of the above form for homogeneous quadratic @xmath67 or @xmath68 physics is the jacobi - type action principle ; in the finite case of mechanics , it is the jacobi action principle itself .
this is more commonly encountered in the mri form , though i note that jacobi himself did use the mpi geodesic principle form . ^
_ = _ _ p_p = _ _ p_p = ^ _ complies with temporal relationalism , and has mutual compatibility , by @xmath69 .
[ this uses euler s theorem for homogeneous functions , and takes @xmath26 and @xmath70 to be any two monotonically - related parameters ; also , by this equivalence , i henceforth drop mpi and mri labels . ]
the ` square root of homogeneous quadratic ' jacobi - type case is then an obvious subcase of this .
moreover , the above generalization includes passage from riemannian geometry ( no signature connotation intended ) to finsler geometry ( no nondegeneracy connotation intended ; @xmath71 is cast in the role of metric function @xmath22 ) .
it is still a geodesic principle , just for a more complicated notion of geometry .
in fact , the jacobi - type action more commonly arises in the literature ( see e.g. @xcite ) not from the desire to be relational but rather from the desire to provide the natural mechanics for a given geometry .
synge s work starting with @xcite generalized this to the more general geometry above @xcite . the geometry in question enters by playing the role of configuration space geometry . in their honour ,
i denote the geometrically - natural construction of a mechanics given a metric geometry by @xmath72 .
the standard definition of conjugate momentum is @xmath73 then computing this from the mri form of the jacobi action gives @xmath74 however , manifest relationalism has been argued to have no place for velocities or lagrangians , so either the above definition of momentum , or the conceptual role played by momentum , need to be reformulated . in this case
, it turns out that the notion of conjugate momentum itself survives , by there being an equivalent manifestly - relational form for the defining formula , @xmath75 then computing this gives @xmath76 which is manifestly within the span of notion ( [ erina ] ) .
next , one inspects to see if there are any inter - relations among these momenta due to the form of the action : _ primary _ _ constraints _ @xcite .
then variation of the action with respect to the base objects provides some evolution equations and perhaps some _ secondary constraints _ @xcite .
one then finds the evolution equations .
finally , one checks whether even more constraints may arise by the requirement that the evolution equations propagate the constraints ; this is best done at the level of poisson brackets ( or dirac brackets if required by the presence of so - called _ second - class _ constraints ) .
the mri implementation of temporal relationalism leads to constraints via the following general argument of dirac @xcite ( as straightforwardly modified by me to concern differentials rather than velocities , and mpi instead of mri actions ) . since mpi actions are homogeneous of degree 1 in the differentials , the @xmath77 momenta arising from these are homogeneous of degree 0 in the differentials .
hence the momenta are functions of at most @xmath78 independent ratios of differentials .
thus the momenta must have at least 1 relation between them ( which is by definition a primary constraint ) .
moreover , for the above jacobi - type action , there is precisely one such constraint ( per relevant nos point ) .
this is due to \a ) the square - root form of the action , by which the momenta are much like direction cosines and ` their squaring to 1 ' property by which they are not all independent . here , instead , these square ( using the @xmath79 matrix ) to @xmath80 .
thus , the kinetic term in terms of momenta ( a legitimately relational object since the momenta are ) , which is half of the above square , is equal to @xmath60 .
\b ) in cases with nontrivial space of extent , due to the particular _ local _ ordering of having the square root ( and sum over @xmath81 , @xmath82 implicit in the kinetic arc element or kinetic term ) _ inside _ the integral - over - space - of - extent sign as opposed to outside it .
all - in - all , one has an energy - type constraint which is quadratic and not linear in the momenta , : = ^__/2 + = 0 .
[ genco ] again , the usual form of the evolution equations makes reference to lagrangians , times and velocities : \ { } = .
there is however a manifestly relational version of this available , @xmath83 . \ { ^ } + ^ _ ^ ^ = ^ , [ evol ] where @xmath84 are the configuration space christoffel symbols .
both the conjugate momenta and the evolution equations are simplified if one uses : = : = : = : = = .
[ emtime ] here , @xmath85 is to be interpreted as an emergent time , the portmanteau of position - independent time @xmath86 for finite systems and position - dependent time @xmath87 for infinite systems ; i term it the * jacobi emergent time*. it , and modifications of it , have long appeared in barbour s work ( e.g. @xcite ) . though i have also seen it in earlier russian literature @xcite for mechanics , and as christodoulou s chronos principle @xcite for gr , so i leave open the question of who its first proponent was .
inverting so as to obtain an expression for the emergent jacobi time itself , ^ ( ) - ^()(0 ) = ||||_/ .
this implements relationalism 6 ) [ mach ] * time is to be abstracted from change * : _ = f [ ] .
relationalism 7 ) [ leibniz barbour ] * the most perfect time is to be abstracted from the totality of change * ( my phrasing , and which is implemented here via _ all _ of the @xmath88 are involved via the @xmath60 and the @xmath89 ) : _ = f [ ] .
i refer to relationalism 6 , 7 ) together as the lmb view of time . in ( [ emtime ] )
, the first step defines a shorthand , the second is the mri computational formula : the third relates the lapse @xmath90 to it , the fourth recasts lapse as a truer _ velocity of the instant _
notion @xmath91 @xcite .
the fifth recasts this as the even truer _ differential of the instant _
notion @xcite @xmath92 , by which the instant @xmath93 itself is identified with the emergent time @xmath85 that labels that instant .
the above generalized lapse notion @xmath90 is a portmanteau of @xmath94 for finite theories and @xmath95 for field theories .
similarly , the above generalized instant notion @xmath93 is a portmanteau of @xmath96 for finite theories and @xmath97 for field theories .
the momenta in terms of @xmath98 are then just @xmath99 , while the evolution equations are @xmath100 .
note that this is but a * parageodesic equation * with respect to the kinetic metric ( meaning it has a forcing term arising from the @xmath60 ) , but is clearly going to be a true geodesic equation with respect to the physical ( tilded ) metric . of use in practical computations , one can supplant one of the evolution equations ( finite case ) or one per space point ( infinite case ) with the lagrangian form of the quadratic ` energy - type ' constraint , _ ^^/2 + = 0
. [ energy ] as first remarked upon in the subcase of mechanics @xcite , the above reveals @xmath85 to be a recovery on relational premises of the same quantity that is more usually assumed to be the absolute external newtonian time , @xmath101 ; there is a conceptually - similar recovery in the general case which covers a further number of well - known notions of time ( see sec 2 ) .
another form for the parageodesic equation is @xmath102 as @xmath85 has been cast as an expression solely in terms of the @xmath103 and @xmath104 , = ^/^ ( ) have tangible physical content , through being a type of ( [ flutt ] ) and not of ( [ erina ] ) . thus in the relational approach the above equations entirely make sense from a temporally relational perspective , unlike in the absolute newtonian counterparts of them that have the same mathematical form but pin an absolute time interpretation on the times present .
the emergent time is provided _ by _ the system . for
now , it furthermore gives the appearance of being provided by the whole of the subsystem .
these last two sentences fit mach s own conception of time as per sec 2.1 . also , the usually - assumed notion of time as an independent variable is un - leibnizian and un - machian .
however , though it is to be overall abstracted from motion , once this is done it _ is _ a convenient choice for ( emergent ) independent variable .
we shall from now on work with mpi and mri forms held to be interchangeable , with preference for writing mpi ones .
we will likewise present specific equations in terms of @xmath98 in the geodesic / lagrangian picture ( the geodesic picture is in terms of configuration variables and their differentials in parallel to how the lagrangian picture is in terms of configuration variables and their velocities ) .
the more well - known difference alias euler lagrange - type actions are , rather , _ = _ _
p_p = _ _ p_p\ { _ - } . [ lagaction ] here @xmath105 is the kinetic energy formulated in terms of @xmath106 derivatives for @xmath107 conventionally an external notion of time ( absolute time for mechanics , coordinate time for gr , which is a label time because gr is already - parametrized ) .
see e.g. @xcite or sec 2 of @xcite for how ( [ action ] ) @xmath108 ( [ lagaction ] ) in the case of mechanics by routhian reduction , and @xcite or sec 2 of @xcite for ( [ lagaction ] ) @xmath108 ( [ action ] ) by the emergence of some lapse - like quantity .
various advantages for product - type actions over difference - type actions stem from this as regards consideration of whole - universe fundamental physics , which is the setting for quantum cosmology .
the above justifies the identification of @xmath60 as the combination of well - known physical entities @xmath109 .
relationalism 8) [ my extension and part - reformulation of barbour ] .
the action is to be * configurationally relational * as regards a group of transformations if a @xmath2-transformed world - configuration is indiscernible from one that is not .
i.e. , one is to build into one s theory that a certain group of transformations @xmath2 acting upon the theory s configuration space @xmath1 are to be irrelevant , i.e. physically meaningless @xcite , transformations .
more specific detail of the @xmath1@xmath2 pairing is postponed until after we have seen details of specific examples , to sec [ patti ] .
one way to implement this is to use not ` bare ' configurations and their composites as above , but rather their arbitrary-@xmath2-frame - corrected counterparts .
this is the only known way that is sufficiently widespread for a relational program to underlie the whole of the classical fundamental physics status quo of gr coupled to the standard model ( see sec 9 ) .
the corrections are with respect to auxiliary variables @xmath110 [ a portmanteau of the finite case @xmath111 and the position - dependent case @xmath112 in field theory ] that are paired with the infinitesimal generators of @xmath2 .
originally the @xmath110 employed were in the form of multiplier coordinate corrections to the velocities ^ - _ _ ^ . here
@xmath113 denotes the infinitesimal group action .
however , this formulation is not consistent with manifest temporal relationalism since the multiplier corrections spoil this . nor
is it relational insofar as it is formulated in terms of meaningless label - time velocities . nor is it an arbitrary-@xmath2 frame approach .
the next ssec amends these things , and then the rest of the present ssec applies just as well within this amendment .
the above also seems to be taking a step in the wrong direction as regards @xmath2 being irrelevant in passing from the already - frame - redundant @xmath1 to the principal fibre bundle over @xmath1 , @xmath114 . for , from the perspective of just counting degrees of freedom
, this can be locally regarded as the product space @xmath115 , which has more degrees of freedom , as opposed to passing to the quotient space @xmath116 , which has less .
however , we shall also resolve this contention below .
moreover , the essential line of thought of this ssec is the _ only _ known approach to configurational relationalism that is general enough to cover the einstein standard model presentation of physics .
this ssec s notion of configurational relationalism is my @xcite portmanteau that spans both of the following .
\1 ) * spatial relationalism * , as in the traditional mechanics setting and the ( geometro)dynamical formulation of gr .
\2 ) * internal relationalism * , as in electromagnetism , yang
mills theory and the associated scalar and fermion gauge theories .
this wide range of cases is afforded by correspondingly wide ranges of @xmath1 and @xmath2 .
( although adopting a @xmath1 may carry connotations of there being some underlying nos with which that is compatible . also , given a @xmath1 , there are consistency limitations on what variety of @xmath2 can then have see sec [ ssec : qg - comp ] ) . for the scaled version of rpm , @xmath2 is the euclidean group of translations and rotations , eucl(@xmath3 ) , while for pure - shape version it is the similarity group of translations , rotations and dilations , sim(@xmath3 ) .
this extension and partial reformulation of barbour s spatial relationalism shows that it and the conventional notion of gauge theory ( usually internal but also applicable to spatial concepts ) bear a very close relation . following on from note 3 of sec 2
, gauge theory is indeed leibnizian and indeed an example of how such can usually only be implemented indirectly by mathematics of a mixed set of tangible entities and non - entities ( respectively true dynamical degrees of freedom and gauge degrees of freedom ) .
configurational relationalism is generalized in sec 15 .
finally , i here propose to extend the meaning of configurational relationalism to have a second clause : * ` no extraneous space structures'*. this matches temporal relationalism s second clause of no extraneous time variables , and also serves to exclude the nambu
goto string action ( which , without this extra criterion , can indeed be included under the previous literature s formulation of relationalism ) . for such reasons @xcite , theories in fixed- background contexts
are ` less relational ' ( or only relational in a weaker sense ) .
it is , rather , with background - independent m - theory ( or at least some limiting classical action for this ) that the most purely relational program would be expected to make contact ( see sec [ ssec : stri - rpm ] ) .
how generally this clause should be imposed requires further discussion in sec 9.11 .
typically , the potential term @xmath117 is manifestly a good @xmath2-scalar but the kinetic term @xmath118 has correction terms due to ` @xmath2-transformations and differentials not commuting ' . consider for now theories whose configuration space carry kinetic arc elements that are homogeneous quadratic and whose associated metric has at most dependence on the @xmath119 [ i.e. a ( semi)riemannian as opposed to finslerian or even more general geometry ]
. then one can render the action mri by supplanting the previous @xmath67 by : = || _ is now interpreted as the velocity corresponding to a cyclic coordinate ; i term the ensuing variational principle [ @xmath120 , @xmath121 , @xmath2 ) ] , the bb standing for barbour bertotti .
[ as a map , = @xmath122-bundle , for @xmath2-bundle : @xmath123 , @xmath121 , @xmath2 ) . ] on relational grounds , however , i prefer the equivalent mpi form , the argumentation behind which never involves the label - time non - entity or the subsequently - dethroned notion of velocity . here , instead of the above
, i supplant the previous configurational - relationally unsatisfactory arclength ( density ) @xmath124 by _
= ||_|| _ _ ^ : = ^ - _ _ _ _ ^^ .
the auxiliary variable in use here is now interpreted as the differential corresponding to a cyclic coordinate .
one can then repeat the preceding ssec s treatment of conjugate momenta , quadratic constraint , evolution equations and the beginning of the treatment of emergent time ` by placing @xmath125 suffices ' on the @xmath126 , @xmath93 , @xmath127 and @xmath98 ( and , if one ever makes use of them , the @xmath128 and @xmath90 ; i also find it helpful in considering the formulae inter - relating these to place a @xmath125 suffix on the @xmath67 as a reminder of the @xmath125-dependence residing within it ) .
@xmath129 is no longer a relationally - satisfactory notion of time due to manifest configurational non - relationalism . as such , i term this a @xmath2-dependent proto - time .
i also name it , and its eventual @xmath2-independent successor , after jbb rather than just after jacobi , to reflect the upgrade to a configurational - relationally nontrivial context .
the novel feature in the variational procedure is , rather , that variation with respect to each @xmath2-auxiliary produces one independent secondary constraint . in this article s principal examples , each of these uses up _ two _ degrees of freedom .
( that this is the case is testable for in a standard manner @xcite , and amounts to these constraints being first- and not second - class . ) thus one ends up on the quotient space @xmath116 of equivalence classes of @xmath1 under @xmath2 motions .
this resolves the abovementioned controversy , confirming the arbitrary @xmath2-frame method both to indeed implement configurational relationalism and to be an indirect implementation thereof .
i denote these linear constraints by @xmath130 : 0 = / ^ /^ ) : = l _ = _ \{^ } , [ linz ] where the last form manifestly demonstrates these being purely linear in the momenta .
the variational procedure behind obtaining these is justified in chapter 2 of @xcite .
one then applies the dirac procedure to be sure that the constraints breed no further constraints ( see sec [ patti ] for what the consequences are if they do ) .
barbour s ` best matching ' @xcite amounts to best matching 1 ) construct an arbitrary @xmath2-frame corrected action .
best matching 2 ) vary with respect to the @xmath2-auxiliary @xmath110 to obtain the linear constraint @xmath131 .
best matching 3 ) solve the lagrangian form of @xmath131 for @xmath110 .
( this is often an impasse . )
best matching 4 ) substitute this back in the action to obtain a new action .
best matching 5 ) elevate this new action to be one s primary starting point .
note 1 ) best matching 2 ) to 4 ) can be viewed as a minimization ( or , at least , as an extremization ) .
one is searching for a minimizer to establish the least incongruence between adjacent physical configurations .
see sec 9.5 for more .
note 2 ) best matching 2 ) to 5 ) can be viewed as a configuration space reduction procedure ( fig [ red ] ) . as a distinct application of best matching 3 ) ,
substitute its solution into the @xmath2-dependent proto - time @xmath129 to render it configurational - relationally acceptable .
[ albeit this is at the cost of being technically implicit and subject to the threats of nonuniqueness and of the extremization not being solvable in practise . ]
i.e. the jbb emergent time is ^ ( ) - ^()(0 ) = ( ||_||_/ ) .
[ kronos ] note 1 ) jbb time does look rather machian [ relationalism 6 ) ] , though a ` boson ' caveat will appear in sec 9.7 .
note 2 ) contrarily , the internal time approach to the problem of time ( c.f .
appendix a ) _ lacks _ in machianity .
this is by its finding specific degrees of freedom that make up such a time , rather than by everything contributing to the time .
configurational relationalism can also be implemented , at least in lower-@xmath3 rpm s , by directly constructing one s formulation to be @xmath2-invariant @xcite .
one can view this as working directly on the relational configuration space .
there is now no need for any arbitrary @xmath2-frame variables , nor then do any linear constraints arise nor are these to be used as a basis for reduction to pass to a new action . here , instead , one s action is already directly @xmath2-invariant i.e. on @xmath132 by construction ; this is reflected by the more complicated form taken by the kinetic arc element / kinetic term .
this approach @xcite builds on work of kendall @xcite on the spaces of shapes and the ` cones ' @xcite over these so as to include scale , by then using the jacobi
synge approach to build a natural mechanics from a metric geometry .
i.e. @xmath133(shape space ) and @xmath134(shape space ) ) = @xmath133(relationalspace ) where c stands for ` the cone over ' ( a notion explained in chapter 3 of @xcite ) .
more generally , see sec 9.2 for this terminology and secs 9.6 and 9.8 for comparison between it and configuration space notions in gr .
this * direct relationalspace construction * can be viewed as a gestalt of both the geometrization of mechanics that motivated jacobi and synge themselves and of the establishment of a temporally relational theory as above .
note 1 ) chapter 3 of @xcite establishes that the direct relationalspace implementation coincides with the configuration space reduction procedure ` best matching 2 - 5 ) ' in the case of 1- and 2-@xmath3 rpm s for which both are explicitly calculable . note 2 ) _ this provides a second foundation for rpm s that is independent of , but the output from which is coincident with , barbour s work _
indeed , it coincides with what arises from barbour s formulation upon performing reduction ( see also chapter 3 of @xcite ) . note
3 ) this formulation possesses emergent time as per ( [ emtime ] ) , quadratic constraint ( [ genco ] ) , no linear constraints , evolution equations in form ( [ evol ] ) and energy constraint ( [ energy ] ) [ the last two now have no @xmath125 subscripts ] .
as relative jacobi coordinates are inter - particle ( cluster ) separations , i take @xmath135 = @xmath136 , and @xmath2 = rot(@xmath3 ) the @xmath3-dimensional rotation group . then @xcite using mass - weighted jacobi coordinates @xmath137 and rotational auxiliary variable @xmath138 , ^_- = s^_- [ wast ] s^_- = surviving constraints are , as a primary constraint , @xmath139 , and , as a secondary constraint from variation with respect to @xmath138 , @xmath140 .
t^ ( ) - t^()(0 ) = ( ||_||/ ) .
this rpm model gives back the zero angular momentum portion of newtonian mechanics for the whole universe , or an arbitrary angular momentum newtonian mechanics for an island universe subsystem .
the mathematically simplest @xcite rpm s in this respect are the @xmath94-stop metrolands ( ` @xmath94 particles in 1-@xmath3 models ' ) , followed by the @xmath94-a - gonlands ( ` @xmath94 particles in 2-@xmath3 ' ) , the first two of which have are called , rather , triangleland @xcite and quadrilateralland @xcite .
the scaled @xmath94-stop metroland s configuration space is @xmath141 @xcite , for @xmath142 , with standard flat - space metric most lucidly expressed in ( ultra)spherical polar coordinates that pick out the shape scale split and constitute the cone over the shape space @xmath143-sphere .
more generally , the rpm with @xmath144 has translation - corrected configuration space @xmath145 and then scaling - corrected configuration space ( ` _ preshape space _ ' ) @xmath146 .
@xmath147 @xmath148 i use @xmath149 , @xmath150 for the corresponding inner product , overline to denote complex conjugate and @xmath151 to denote complex modulus .
the conjugate momenta are then @xmath152 and @xmath153 .
@xmath154 for @xmath155 the inverse of the fubini
study metric .
@xmath156 next , in the specific triangleland case s^ _ = s^_- s^ 2 _ ( ) = ^2 + ^2\{^2 + ^2 ^ 2}/4 which becomes , up to an overall factor of @xmath157 that can be passed on to the @xmath158 term , s^ 2 _ ( ) = i^2 + i^2\{^2 + ^2 ^ 2 } here , @xmath159 is the configuration space radius , equal to @xmath160 for @xmath96 the total moment of inertia of the system .
@xmath161 is the relative angle between the base and the median that are privileged by one s coordinate choice , and @xmath162 is a function of the ratio of the lengths of these ( see e.g. @xcite for more details ) .
the momenta are then @xmath163 , @xmath164 and @xmath165 .
@xmath166 for @xmath155 the inverse of the fs metric .
finally , the emergent jbb time is now given by t^ ( ) - t^()_0 = ^_()/ .
the usual covariant spacetime tensor presentation of the einstein field equations follows from the einstein
hilbert action s^ _ = ^4x ( ) for @xmath167 the spacetime metric with determinant @xmath168 and ricci scalar ric(@xmath167 ) .
this came about due to einstein reconceptualizing the nature of space and time , in good part due to the compellingness of mach s arguments .
however , there is considerable confusion as to how gr is and is not machian .
sr transmutes rather than eliminates absoluteness issues
one passes from newton s set of unexplainedly privileged notions to another set , the ` sr inertial frames ' .
doing so amounts to a destruction of simultaneity , which is replaced by the universal significance of null cones . in sr
it is often said that space and time can also be regarded as fused into the spacetime of minkowski .
( however , sr breaks only isolation of space and time , not their distinction @xcite . )
whitrow @xcite and barbour @xcite are also dismissive of minkowski s considering the individual notions of time and space to be _ doomed to fade away . _ " ) moreover , this is itself a fixed background structure , possessing a timelike killing vector and privileged spatial frames that precede the introduction of actual material physics into one s worldview .
however , next , in gr matter now exerts an influence on the form of space and time , by which it is in general curved rather than flat as minkowski spacetime is .
while in some senses that has a machian character , einstein s inception of the theory did not concretely build up on mach s ideas @xcite , so whether the theory actually implements these has been a source of quite some argument .
we shall see below , however , that directly building up on some of mach s ideas happens to lead one to ( a portion of ) gr , arriving to it in its ( geometro)dynamical form @xcite , so that einstein s gr does , in any case , happen to contain this philosophically desirable kernel .
while the role of spacetime in gr has often been touted , one should not forget that gr also admits a dynamical interpretation in terms of evolving spatial 3-geometries , i.e. geometrodynamics . in this approach
to gr , one has to preliminarily choose a residual nos in the sense of a spatial hypersurface of fixed topology @xmath169 .
this , i take to be a compact without boundary one for simplicity , and 3-@xmath3 as this suffices to match current observations . under the adm split of the spacetime metric g _ = ( ) with respect to a foliation by such spatial hypersurfaces , the einstein
hilbert action splits into the adm lagrangian action ^_- = _ ^3x \ { ^_-/^2 + ( h ) - 2 } , [ adm ] ^_-
= || _ - _ h_||^2_/4 .
this action encodes the gr hamiltonian constraint : = n_^^ - \{(h ) - 2 } = 0 [ hamm ] from variation with respect to the lapse @xmath170 and the momentum constraint _ : = - 2d_^ _ = 0 [ momm ] from variation with respect to the shift @xmath171 .
the latter is interpretable geometrically as gr not being just a dynamics of 3-metrics ( ` metrodynamics ' ) but furthermore that moving the spatial points around with 3-diffeomorphisms does not affect the physical content of the theory . the remaining information in the metric concerns the ` underlying geometrical shape ' .
this is how gr is , more closely , a dynamics of 3-geometries in this sense : geometrodynamics .
this description is less ( rather than completely non- ) redundant ( the partial redundancy being due to the hamiltonian constraint not yet being addressed )
. some useful notes about ( in some cases precursors to ) problem of time facets are as follows . note 1 ) the gr constraints @xmath172 and @xmath173 _ close _ in the form of the dirac algebra @xcite .
note 2 ) in the dynamical approach to gr , one often considers not one spatial hypersurface @xmath169 but a _ foliation _ of the spacetime by such spatial hypersurfaces .
moreover , classical gr has the further property of _ refoliation invariance _ :
the evolution between two given spatial hypersurfaces is independent of which foliation interpolating the two is used .
both the closure of note 1 ) and foliation - independence are held to be valuable properties of classical gr . moreover , both of these have relational character the first via sec 10 and the second due to foliation dependence in gr amounting to coordinate - dependent physics , at least from the point of view in which conventional gr s spacetime notion is taken as given . note 3 )
the * global problem of time * is due to a single time function not always being definable over the whole of spacetime / space . classically , this is already present but is a relatively straightforward issue of the finiteness of coordinate charts and the subsequent standard meshing condition being required in changes involving passing from one gr coordinate time to another . note 4 )
quantities commuting with both @xmath172 and @xmath173 are _ dirac observables _ ; quantities commuting only with the former are observables .
the * problem of observables * facet of the problem of time is that these are hard to find ( especially the dirac ones ) , though rovelli s own opinion on this matter is that this be sidestepped by considering only a weaker notion of partial observables ( see sec [ cl - pers ] ) .
lmb relationalism favours s stance @xcite , or , at least a partly - dirac stance @xcite ( i.e. commutativity with @xmath173 still being required alongside permission to view @xmath173 as distinct from a gauge constraint , that in itself constituting a deviation from dirac s writings @xcite ) .
in fact , following bell @xcite i prefer talking in terms of _ beables _ rather than observables , especially in the context of quantum cosmology .
i continue here with discussion of the _ thin sandwich approach _ , in which baierlein , sharp and wheeler @xcite regarded @xmath174 and @xmath175 as knowns .
[ i.e. the ` thin ' limit of taking the bounding ` slices of bread ' @xmath176 and @xmath177 as knowns .
] this is solved for the spacetime ` filling ' in between . in more detail ,
bsw 0 ) they take as primary a conventional difference - type action for geometrodynamics .
bsw 1 ) they vary this with respect to the adm lapse @xmath170 .
bsw 2 ) they solve resulting equation for @xmath170 . bsw 3 ) they substitute this back in the adm action to obtain a new action : the bsw action , ^ _ = _ ^3x^_- .
bsw 4 ) they then take this action as one s primary starting point .
bsw 5 ) they vary with respect to the shift @xmath171 to obtain the gr linear momentum constraint , @xmath178 .
bsw 6 ) they posit to solve the lagrangian form of @xmath178 for the shift @xmath171 . bsw 7 ) they posit to then substitute this into the computational formula for the lapse @xmath170
bsw 8) finally , they substitute everything into the formula for the extrinsic curvature to construct the region of spacetime contiguous to the original slice datum .
note 1 ) unfortunately the p.d.e . involved in this attempted elimination is the _
thin sandwich equation _ ; this is difficult and not much is known about it ( obstructions based on this are termed the * thin sandwich problem * , which is one of the facets of the problem of time ) .
counterexamples to its solubility have been found @xcite , while good behaviour _ in a restricted sense _ took many years to establish @xcite .
note 2 ) bsw 03 ) are a multiplier elimination ; they can be reformulated as a cyclic velocity elimination i.e. a passage to the routhian @xcite directly equivalent to the euler lagrange action to jacobi action move mentioned in the introduction , or a cyclic differential elimination .
note 3 ) bsw 46 ) directly parallel best matching 13 ) modulo the upgrade from multiplier shift to cyclic velocity / differential of the instant .
thus in the geometrodynamics case , best matching is hampered in practise by the thin sandwich problem .
a general phrasing for this is the obstruction , whether practical - computational or provable - analytical , to .
[ bmmmm ] moreover , ` sandwich ' is clearly a geometrodynamical , or more generally diff(3)- or foliability - specific name , whereas ` best matching ' is a more universal one , so in fact i choose the name * best matching problem * for the general theoretical problem ( [ bmmmm ] ) of which the thin sandwich problem is but the particular case corresponding to theories with some kind of nos diff - invariance .
this reflects that the sandwich problem is both generalized , and placed on relational terms , in the present article .
note 4 ) bsw did not consider the culmination move best matching 4 ) of this which ( merely formally ) produces an action on superspace .
note 5 ) bsw 7 ) is , however , a primitive version of the @xmath179 computation ( via the @xmath90 to @xmath91 to @xmath180 and hence @xmath179 progression ) .
there is a presupposed ( bsw ) versus emergent ( relational approach ) difference in the status of the computed object too .
note 6 ) as far as i am aware , christodoulou @xcite was the first person to cast bsw 7 ) as an emergent time [ amounting to the differential form of ( [ kronos ] ) , which he termed the ` _ chronos principle _ ' ] .
moreover , like wheeler , he did not base this on relational first principles .
[ though christodoulou was certainly aware in doing so of some of what barbour and i have held to be the leibniz mach position on relational time : _ they contain the statement that time is not a separate physical entity in which the changing of the physical system takes place .
it is the measure of the changing of the physical system itself that is time _ " ] .
he deduced the generalization of ( [ bmmmm ] ) to include gauge fields minimally - coupled to gr involving shift and the yang
mills generalizations of the electric potential @xcite .
i credit alevizos for pointing out this connection between barbour s and christodoulou s work .
note 7 ) bsw 8) is subsequent step concerning dynamical evolution which for geometrodynamics has nontrivial geometrical content .
note 8) i have thus clearly laid out how the technical content of barbour s best matching notion resides within a very immediate generalization of the original bsw paper .
wheeler did not , however , have the relational significance that underpins this procedure , or the cyclic auxiliary coordinate formulation of the next ssec .
the barbour - type indirect formulation of rpm s ( [ generalaction],[taction ] ) makes these parallels particularly clear . in the geometrodynamical counterpart of this ,
@xmath1 = riem(@xmath169 ) the space of riemannian 3-metrics on some spatial manifold of fixed topology @xmath169 ( taken to be compact without boundary for simpleness ) .
this choice of 3-metric objects is a simple one ( one geometrical object , rather than more than one ) , moreover one that is \a ) useful for physics given its significance in terms of lengths .
\b ) mathematically rich
one then gets a connection for free the metric connection as well as the notions of curvature and of hodge star .
nevertheless , one is entitled to view this choice as a possibly inessential simplicity postulate .
the group of irrelevant motions @xmath2 is now diff(@xmath169 ) , i.e. , the diffeomorphisms on @xmath169 .
this configurational relationalism is implemented indirectly by introducing auxiliary variables that represent arbitrary - diff(@xmath169 ) corrections .
superspace(@xmath169 ) = riem(@xmath169)/diff(@xmath169 ) .
temporal relationalism is implemented by building a mri / mpi action @xcite
[ this is similar to the above bsw action but now properly combining the temporal and configurational relationalisms ] .
the relational formulation of geometrodynamics is valuable in providing guidance in yet further investigations of alternative conceptual foundations for gr @xcite , and as regards addressing the problem of time in quantum gravity ( see sec [ ssec : 1.7 ] and part iii of @xcite ) . now , _
^ = 2 ^ 3x = ^3x _ ^ _ ^ : = ||_||_^2/4 _ ^ : = action is a jbb@xmath181 $ ] for a particular potential proportional to ric(@xmath182 ) .
note 1 ) the relational action bears many similarities to the better - known bsw one , but supersedes it via attaining manifest temporal relationalism .
the two become indistinguishable for minisuperspace : @xmath183 collapses to an ordinary @xmath184 matrix @xmath185 or further in the diagonal case ( a @xmath186 matrix @xmath23 ) the ` minisupermetric ' .
@xmath187 ) as a primary constraint @xcite ; this has the momenta contribute quadratically but not linearly . also , variation with respect to the auxiliary diff(@xmath169)-variables
@xmath188 gives the gr momentum constraint ( [ momm ] ) .
just as for the standard adm approach , the momentum constraint is interpretable geometrically as gr being a geometrodynamics . in the relational approach , gr has an analogue of emergent jbb time ( in this context crediting also christodoulou @xcite ) , ^()(x^ ) - ^()_0(x^ ) = .
||_|| _ / .
[ gremt ] this represents the same quantity as the usual spacetime - assumed formulation of gr s _ proper time _ , and , in the ( predominantly ) homogeneous cosmology setting , the cosmic time .
it also coincides with gr s own version of emergent semiclassical ( alias wkb ) time ( see appendix a and sec 16.4 for a brief explanation of what this is ) . interpretation 4 ) next , wheeler asked @xcite why @xmath173 takes the form it does and whether this could follow from first principles ( ` 7th route ' to gr ) rather than from mere rearrangement of the einstein equations .
to date , there are two different answers which tighten wide classes of anstze down to the gr form ( see @xcite for a comparison of these ) . answer a )
[ hojman kucha teitelboim ] @xcite is from _ deformation algebra _ first principles ( this assumes embeddability into spacetime ) . answer b ) [ barbour foster
omurchadha and i ] @xcite the ` _ relativity without relativity _ ' approach , which lies within the relational program , arrives at actions similar to the bsw one , and is the subject of the next two ssecs . furthermore , adopting the relational first principles , _ without _ assumption of additional features derived in adm s approach , leads to the _ recovery _ of the bfo - a action of gr as one of very few consistent choices within a large class of such actions .
e.g. one does not need to assume the gr form of the kinetic metric or of the potential .
one then has a family of trial actions @xmath189 $ ] for @xmath190 the general ultralocal supermetric @xmath191 and a potential ansatz such as @xmath192 for @xmath193 , @xmath194 constant , ^ _
then relational postulates alongside a few simplicities already give this since the dirac procedure @xcite prevents most other choices of potential term @xmath195 from working @xcite alongside fixing @xmath196 to take the dewitt value 1 of gr as one of very few possibilities .
this amounts to a classical precursor of the spacetime reconstruction problem facet of the problem of time , which , once again , the machinery of the classical gr constraints resolves .
the above relational formulation of gr is furthermore robust to the inclusion of a sufficiently full set of fundamental matter sources so as to describe nature .
thus electromagnetism is included in @xcite , yang
mills theory in @xcite , spin-1/2 fermions in @xcite and scalar and spin-1/2 fermion gauge theories @xcite .
moreover , it does _ not _ do so in a way by which relational postulates alone _ pick out these matter theories_.
( @xcite demonstrate how earlier claims to the contrary @xcite in fact contain additional tacit simplicity assumptions .
a further gap , not yet addressed , is that the possibility of second - class constraints is not properly catered for in any uniqueness by exhaustion proof so far given . )
one way some of these things are accommodated is through lifting the homogeneous quadraticity restriction @xcite ( which is not per se relational ) . in particular
this allows for mechanics with linear ` gyroscopic ' terms @xcite , as well as moving charges , spin-1/2 fermions and the usual theories coupling these to gauge fields , scalars and gr . the form of action here is of randers type @xcite ^_- = _ \{^ _ + ^ _ } , [ 62 ] with @xmath197 then not contributing to the quadratic constraint or to the emergent jbb time ( thus this is not entirely universal , though the potential associated with such fields does sit inside the square root and thus itself does contribute to these things ) .
@xmath198 does however contribute to the linear constraints .
a further implementation of mach [ e.a .
: relationalism 7b ) ] is that * time is abstracted from bosonic change .
* i further discuss these variants on relationalism involving boson fermion distinctions ( and the supersymmetric case ) in @xcite .
note : ( [ 62 ] ) also renders clear how to extend christodoulou s chronos principle working to include fermions .
other simplicities involve neglecting the possibility of metric
matter cross - terms and other terms that one knows appear in the canonical formulation of more general fields , as well as the restriction on allowed orders of derivatives that is much more ubiquitous in theoretical physics . by such means
, some more general / complicated possibilities can be included among the relationally - formulable theories : \1 ) brans dicke theory @xcite , \2 ) proca theory @xcite , and \3 ) local - sr - cone - violating and equivalence principle violating vector
tensor theories @xcite .
thus , contrary to the claims / speculations in @xcite , relationalism does _ not _ derive or imply the equivalence principle or a universal null cone .
( the former effectively requires some form of geometrodynamical equivalence principle postulate @xcite , while the latter requires a number of nonrelational simplicity assumptions . )
^ : = h_^/ = c ( ) [ cmcsl ] is @xcite more or less @xcite paralleled by an ` euler time ' variable @xmath199 this is all underlied for both gr and rpm s by scale shape splits , the role of scale being played by @xmath200 or @xmath96 for rpm s and by such as the scalefactor @xmath201 or @xmath202 in gr . in both cases it is then tempting to use the singled - out scale as a time variable but this runs into monotonicity problems .
these are avoided by using as times the quantities conjugate to ( a function of ) the scale , such as @xmath203 or @xmath204 .
moreover , gr in the conformogeometrical formulation on cmc slices @xcite can be reformulated @xcite as a relational theory in which @xmath2 consists both of diff(@xmath169 ) and a certain group of conformal transformations .
conf(@xmath169 ) are conformal transformations ; criem(@xmath169 ) is pointwise superspace = riem(@xmath169)/conf(@xmath169 ) .
cs(@xmath169 ) is conformal superspace = riem(@xmath169)/diff(@xmath169 ) conf(@xmath169 ) , for denoting semi - direct product @xcite . \{cs
+ v}(@xmath169 ) is conformal superspace to which has been adjoined a single global degree of freedom : the spatial volume of the universe .
this can be viewed as riem(@xmath169)/diff(@xmath169 ) vpconf(@xmath169 ) , for vpconf(@xmath169 ) the global volume - preserving conformal transformations .
cs(@xmath169 ) and ( cs + v)(@xmath169 ) have on a number of occasions been claimed to be the space of true dynamical degrees of freedom of gr @xcite . following from the alternative theory of @xcite and @xcite , @xcite used volume - preserving conformal transformations .
and @xcite additionally then considered a linking theory approach .
of the latter , i comment that using a hamiltonian formulation is cleaner , but also that these papers use in excess of the usual amount of canonical theory ; i do not know whether this is necessary ; the current article presents the case for questioning excess structures in relational approaches .
proponents of ashtekar variables often use the word ` relational ' of their approach .
however , this word has not hitherto been systematically used in their program in the full lmb or lmb - a sense ; instead they use this word with less detail and/or to carry ( a subset of ) the perspectival meanings of sec 14 .
nevertheless , i state here that the ashtekar variables scheme can be formulated in the full lmb or lmb - a relational scheme .
this is modulo a small glitch at the start of such an investigation , which nevertheless turns out to be entirely surmountable .
namely , that the pure gravity case of the ashtekar variables approach is not castable in temporally relational form at the level of the classical lagrangian . from the point of view in which spacetime is presupposed @xcite , in which one sets about elimination of the lapse to form the temporally relational action , this glitch is due to the following .
the lapse - uneliminated action is purely linear in the lapse .
this is because the well - known ` pure-@xmath205 ' character of ashtekar s canonical action for pure gr , in contrast to the more usual ` @xmath206 ' form of the geometrodynamical action .
thus the variation with respect to the lapse produces an equation independent of the lapse .
thus the lapse can not now be eliminated from its own variational equation .
however , it is clear that addition of matter fields breaks this pure linearity in the lapse , so this unusual accident that blocks the elimination of the lapse from its own variational equation ceases to occur and a temporally relational action can be obtained .
thus that the barbour / lmb - relational literature s looks to be fixated with gr in geometrodynamical form as regards which examples it uses is in fact down to a matter of taste / of each paper s applications , rather than being in any way being indicatory of lmb - relationalism only applying to the older geometrodynamics and not the newer ashtekar variables work that underlies lqg .
it is then surely of foundational importance for ashtekar variables programs for them to be revealed to be relational in this additional well - motivated and historically and philosophically well - founded way in addition to the various other meanings that they have hitherto been pinning on the word ` relational ' in the ashtekar variables / lqg program .
pointing this out is one of the main applications of this article .
configurational relationalism for ashtekar variables formulations has @xmath207 diff(3 ) .
then variation with respect to the additional @xmath208-auxiliaries produces the @xmath208 yang mills gauss constraint that is ubiquitous in the ashtekar variables literature , whilst variation with respect to the diff(3)-auxiliaries produces the ashtekar variables form of the gr momentum constraint .
the hamiltonian constraint arises as a primary constraint , and the usual form for a jbb time emerges .
an ashtekar variables formulation of a type of gr shape dynamics has recently appeared @xcite .
compared to theories preceding it , supersymmetry is conceptually strange in how the product of two supersymmetry operations produces a _ spatial _ transformation , and possibly also via how the supersymmetry constraint in supergravity is the square root of the hamiltonian constraint @xcite . as such , there is some chance that the supersymmetry transformation is one of not conceptually sound , or conceptually deeper than , other transformations , and the relational perspective could be one that had more to say about this matter .
thus , _ the relational program might be able to provide its own prediction _ as to whether to expect supersymmetry in nature .
conversely , insisting on supersymmetry might expose the current conceptualization of relationalism to be insufficient .
see secs [ ssec : qg - comp ] , [ sse ] and @xcite for more . aside from sec 6 s use of the preclusion of perturbative strings to sharpen up criteria for relationalism
, there are some conceptual parallels between passing from studying point particles to studying strings and passing from point particles to studying relational quantities only .
these are comparable from a conceptual perspective ; whether each of these is rich , tractable and has anything to say about quantum gravity is another matter .
the latter is moreover more conservative the relational quantities come from careful thought about the original problem rather than replacing it with a distinct problem as in string theory .
this gives less new structure leading perhaps to less mathematical richness , but it it would appear to be a safer bet as regards the _ hypotheses non fingo _ " tradition .
exporting the configurational relationalism / best matching ideas to other parts of physics could conceivably be of interest even if not accompanied by the further demands of no background spatial structures or of temporal relationalism .
e.g. one could consider what form these ideas take for target space theories of which strings on a given background .
finally , one would expect nonperturbative m - theory to have _ both _ of these features ( relationalism _ and _ extended objects ) .
provided that supersymmetry can be accommodated into ( a suitable generalization of ) relationalism , i conjecture that relational notions along the lines of those in this article are necessary for any truly gr - embracing notion of background independence for m - theory since background - independence amounts to a problem of time , it would then be exceedingly likely for m - theory to have one of those too .
thus study of the problem of time is likely to be a valuable investment from the perspective of developing and understanding m - theory too .
geometrodynamics ( very easily generalizable to arbitrary-@xmath3 ) could be seen as a first toy model of spatially 10-@xmath3 m - theory
. there may be better rpm s and geometrodynamics models for this particular setting , e.g. spatially 10-@xmath3 supergravity ( which is a low - energy/`semiclassical ' limit of m - theory ) .
the situation with extended objects that occupy / furnish a multiplicity of notions of space is , in addition to supersymmetry , an interesting frontier in which to further explore the relational postulates .
[ ssec : qg - comp ] now we have covered some examples , we can return to this ...
relationalism 9 ) [ e.a . ]
a ) * nontriviality * @xmath2 can not be too big [ if dim(@xmath2 ) @xmath209 dim(@xmath1 ) 1 then the relational procedure will yield a theory with not enough degrees of freedom to be relational , or even an inconsistent theory ] .
\b ) further * structural compatibility * is required .
example 1 ) if one uses such as sim(@xmath210 ) or eucl(@xmath210 ) with @xmath211 , it is usually for @xmath212 ( or at least @xmath213 ) .
example 2 ) if one uses diff(@xmath214 ) to match to riem(@xmath215 ) , it is usually for @xmath216 , or at least for the two to be obviously related ( e.g. taking out diff(@xmath217 ) from riem(@xmath218 ) \c ) more concretely , @xmath2 * has to have a ( preferably natural , faithful ) group action on * @xmath1 .
relationalism 10 ) [ barbour and e.a . ] in looking to do fundamental modelling , one may well have a strong taste for * lack of extraneity * , e.g. always taking out all rotations rather than leaving a preferred axis .
\i ) aut(@xmath219 ) is then a very obvious designate for this , though some subgroup of aut(@xmath219 ) might also be desirable , and there is also the issue of automorphisms of up to which level of structure .
then aut(@xmath219 ) and its subgroups definitely comply with a ) and stand a good chance of suitably satisfying criteria b ) and c ) .
\ii ) one may also have a strong taste for eliminating as many extraneous properties as possible [ barbour s view ] , though under some circumstances there may be strong practical reasons to not eliminate some [ e.a .
s counter - view ] .
example 1 ) one may wish to retain the scale , thus using eucl(@xmath3 ) rather than sim(@xmath3 ) for rpm s , or and superspace(@xmath169 ) or \{cs + v}(@xmath169 ) instead of cs(@xmath169 ) for geometrodynamics , so as to have a viable cosmological theory and possibly for reasons of time provision .
scale is possibly undesirable from the relational perspective via being a single heterogeneous addendum to the shapes @xcite .
the cone structure makes scale especially extraneous in rpm s it has the same sort of dynamics irrespective of what the shapes are . it
may then seem quite conceptually undesirable for scale to play so prominent a roles in cosmology . however , 1 ) there are currently no credible alternatives to this as regards explaining cosmology .
2 ) this heterogeneity can be useful by alignment with choices that would be harder to make without it , renders scale more presentable from a conceptual perspective .
this is the position i will take in the upcoming ` conformal nature of the universe ' conference at the perimeter institute .
see sec 16 for how various approaches fare in this regard .
example 2 ) quotienting out diff(@xmath169 ) and conf(@xmath169 ) does not interfere with the existence of spinors , but for further / other choices one has a lack of such guarantees @xcite , likely compromising one s ability to model spin-1/2 fermion matter .
moreover , it is well - known that some choices of @xmath169 by themselves preclude the existence of fermions ( see e.g. @xcite ) , and that orientability of @xmath169 is also often desired .
( but supersymmetric extensions are quite clearly also compatible in this sense @xcite . ) in more detail , one has a background nos @xmath220 , which has properties deemed redundant , all or some of which can be modelled as such .
there is then some level of structure @xmath221 within the @xmath222 .
one then builds the configuration space of actually mathematically manipulable part - tangible part - intangible entities to lean on the structure of @xmath220 , so i denote it in this sec by @xmath223 .
then finally one passes to ( ) / ( , p _ ) .
bb82 has a means of modelling continuous transformations at the level of riemannian geometry . as per chapter 2 of @xcite
, one can also conceive of @xmath220 in such a way that discrete isometries can be considered as well as continuous ones .
if one fails to be able to model @xmath224 , then the automorphisms in question will not be freeing one of all the levels of structure / aspects of @xmath219 .
this is clearly the case for superspace(@xmath169 ) : quotienting out diff(@xmath169 ) from riem(@xmath169 ) clearly does not free one from what choice of @xmath169 one made in setting up riem(@xmath169 ) .
this is because @xmath169 pertains to the level of topological structure , and rendering the diffeomorphisms irrelevant is only a freeing from background _
metric _ structure .
the relationalist who takes minimalism to an extreme would then aim to remove all sources of indiscernibility and actors that can not be acted upon .
see secs 10 and 16 as regards the status of scale , and chapter 14 of @xcite as regards whether spatial topology in gr is discernible and can be acted upon ; another example is how dimension is not primary for rpm s ( chapter 2 of @xcite explains how 3-particle rpm s can not distinguish between dimension 2 or higher ) .
the objective of relationalism is thus ( ) / ( , p _ ) , ( ) / ( , p _ ) .
relationalism 11 ) [ bfo a ] . note that @xmath1 ( the entity taken to have some tangible physical content ) has the * a posteriori right to reject * @xcite a proposed @xmath2 by triviality or inconsistency that go beyond 1 ) , arising instead via , for the moment , the dirac procedure yielding further constraints .
example 1 ) if one attempts to use @xmath225 i d with riem(@xmath169 ) and a bsw - like trial action ( like murchadha @xcite or i @xcite ) , one finds that enlarging @xmath2 to diff(@xmath169 ) is _ enforced _ by the @xmath172 arising as an integrability of @xmath173 .
this also exemplifies how not all subgroups of aut are guaranteed to serve .
note 2 ) additionally , a given @xmath1 , @xmath2 pair still constitutes a substantial ambiguity as to the form of the action principle .
this can sometimes be truncated with simplicity postulates like that the action is not to contain higher than first ( or occasionally second ) derivatives , but ambiguities beyond that still often remain .
example 2 ) the potential is completely free in scaled rpm and free up to homogeneity in pure - shape rpm .
example 3 ) one can have gr or euclidean gr or strong gravity for the riem(@xmath169 ) , diff(@xmath169 ) pair with action restricted to contain at most first label - time derivatives and quadratically and at most second spatial derivatives . note
3 ) this scheme retains considerable freedom .
relationalism does _ not _ exert a highly unique control over the form that theoretical physics is to take [ though relationalism 10 ) and 11 ) can sometimes help with this ] .
relationalism 11 ) quotienting out the @xmath2 is to * remove all the structure in the * @xmath1 * that is actually extraneous*. however , precisely what structures / features are extraneous in the case of the universe we live in is not so clear .
[ e.g. , again , whether to discard scale . ]
one passes to ( ) / ( , p ) for @xmath226 the structures up to a certain mathematical level ( e.g. metric geometry or topology ) .
one might ideally wish for relationalism to remove _ all _ levels of structure of @xmath219 .
in particular , @xmath2 = i d is too trivial for most of theoretical physics , whilst a number of further gr features come from @xmath2 = diff or some generalization ( diff @xmath227 conf , the superdiff of supergravity ) or possibly some slight weakening , such as by attitude to the large diffeomorphisms @xcite or restriction to the transverse diffeomorphisms @xcite .
i next consider four structural expansions on barbour s ` @xmath1 is primary ' postulate [ relationalism 3 ) ] .
\1 ) one usually considers * phase space * phase , or at least * adjoining momenta * @xmath228 to the configurations @xmath64 .
this generalizes @xmath1 to a more general concept of * state space * , which i denote by @xmath21 .
i will consider these expansions at the classical level first , and then at the quantum level , where such as hilbert spaces of quantum states play the role of @xmath21 .
\2 ) * categorizing*. one may well wish to consider whichever @xmath21 as objects alongside corresponding morphisms @xmath229 to form the category ( @xmath21 , @xmath229 ) .
examples of morphisms include the point transformations point for @xmath1 and the canonical transformations can for @xmath21 .
\3 ) * perspecting*. taking , say , @xmath21 to be primary in fact amounts to ( especially in the case of whole - universe studies ) considering some set of subsystems of @xmath21 .
\4 ) * propositioning*. if , say , we take @xmath21 to be primary , physics is in fact about the propositions about @xmath21 , prop(@xmath21 ) , rather than about @xmath21 itself . as we shall see ,
some additional issues concerning how to compose some of these expansions is crucial .
i consider passing from conceiving @xmath1 to be primary to suggesting that @xmath64 and @xmath228 are operationally distinct ; i refer to the ensuing state space as * rigphase*. from the more usual perspective , this is a privileged _ polarization _ @xcite of phase ( phase space ) that maintains the distinction between the physical @xmath228 s and @xmath64 s . on the other hand ,
a distinct and far more common viewpoint is that phase space = ( t^ * ( ) , \ { , } ) is central . here
, @xmath230 is the cotangent bundle ( for simple bosonic theories ) of momenta over the configurations and \ { , } is the poisson bracket , so that this is a poisson algebra . n.b .
the status of rigphase in this program is as an _ option _ that is more structurally minimalist .
rigphase motivation 1 ) as argued in sec 3 , considering more minimalist possibilities is very much part and parcel of relational thinking .
one such avenue then would be to investigate what happens if one weakens the amount of structure assumed in the study of the ` associated spaces ' of the principles of dynamics .
there are very strong reasons why @xmath1 ( or something with an equivalent amount of physical information in it ) is indispensable , but some parts of the conventional structure of phase space are more questionable .
much as one can envisage preferred - foliation counterparts of gr spacetime , one can also construe of preferred - polarization versions of phase space .
whereas relational perspectives usually view preferred foliations of spacetime as less physical , i contend that , at the level of the associated abstract spaces of the principles of dynamics , the relationalist s ontological hierarchy as regards preferred versions of these spaces is the reverse of that for spacetime itself .
my argument then is that the operational distinction between configuration and impact measurements can be encoded by using a preferred polarization centred about the _
@xmath1 ( as opposed to any other nontrivially canonically - related _ mathematical _
@xmath1 ) .
as a general point , one should not necessarily postulate that physical entities that are operationally distinct _ have _ to be mixable just because transformations that mix them can be mathematically defined .
one should not confuse ` is often useful to transform ' with giving unwarrantedly unconditional physical significance to the transformations in question .
i argue the practical implementation of rigphase to be as follows .
i hold that the physicality of @xmath1 is _ not _ the centre of this argument ; that is , rather , operational distinguishability between the @xmath228 s on the one hand and the @xmath64 s on the other .
if one includes fermions , one can operationally tell apart fermionic quantities that manage to be positions and momenta at once from quantities that solely manifest themselves as configurations and quantities that solely manifest themselves as momenta . in this way
, the scheme extends to cases where phase spaces are not just simple tangent bundles over configuration spaces , via operational distinguishibility now being tripartite @xcite .
see this reference also for how , were supersymmetry to exist in nature , it could impose limitations on the applicability of relationalism 3 ) and its development using less - than - phase - space constructs such as rigphase . as regards primality , firstly i only wish to consider _ direct _ measurements rather than inferences ( which could e.g. involve inferring distances from momentum - type measurements connected to photon momenta as in sight or spectral lines )
. one can dissociate from this problem by thinking in terms of simple ` blind feeling out ' measurements on a scale small enough that one can reach out to feel ( for all that there will often be strong practical limitations on this ) .
the virtue of this approach is that one s configuration measurements for particle mechanics are all _ geometrical _ : local angles , adjoining rulers to the separations between objects ; the blind geometer s work is free from the impact connotations that enter his sighted colleague s work .
[ the latter s methodology will however often have greater _ practicality _ , e.g. if both worked as surveyors ... ] next , in gr in geometrodynamical form , riemannian 3-geometry measurements of space are again cleanly separated out from gravitational momentum / extrinsic curvature .
note how in the mechanics
geometrodynamics sequence of physical theories , one can pass from a simple notion of relational geometry measurements being operationally primary to a more complicated one ( argued for e.g. in @xcite .
a theme that may then support part of the argument for the primality of configurational measurements is that geometrical measurements are primary . from this perspective
, it is somewhat incongruous for triad information to be considered as momentum - like within the ashtekar variables type approaches given that this is another form of space - geometric information
. it would likewise be somewhat incongruous if the ashtekar @xmath231 were to be only an indirectly inferrable quantity.[multiblock footnote omitted ] on the other hand , the foundations of geometrodynamics are _ not _ incongruous in this way , though the above - linked footnote does place limitations on , and give alternatives to , this particular sense of incongruence fostered by this sec s tentative expansion of the point of view that @xmath1 is primary .
the debate about whether lengths , impacts , times , frequencies ... are the most primary things to measure is indeed an old and unsettled one .
rods are arguably a less deep notion / apparatus than clocks e.g. via bondi s argument @xcite that they are made out of quantum matter which is ultimately underlaid by frequencies .
also , by their nature and function , rods are necessarily macroscopic @xcite and so interact with their environment in uncontrollable ways , whilst microscopic clocks are indeed possible .
i wish to make a somewhat different point too : that , regardless of primality , one can tell apart whether what one is measuring is instantaneous - configuration information or impact information .
rigphase motivation 2 ) one might doubt canonical transformations due to their clash with the common notion of ` adding in a potential ' .
if this is to be considered as a structurally minor change , as is often done , then one has a problem due to the asymmetry between the complication caused by adding a @xmath232 potential , say , as compared to the major structural upheaval of adding a @xmath233 kinetic term to one s hamiltonian .
of course , this can be dealt with by accepting that canonical transformations imply that ` adding a potential ' is as great a structural upheaval ( e.g. the above two additions to an ho are identical to each other under canonical transformation ) .
this translates to changing the habitual simplicity requirements on hamiltonians and lagrangians due to their canonical disparity , and that doubtlessly cuts down on widely applicable theorems that depended on such canonically - disparate simplicity requirements .
one might then hold all such results to have been empty anyway , but one might attempt to keep them by allotting particular significance to the physics in its presentation centred about @xmath1 . by the above , it is also clear that canonical transformations are at odds with statements of simplicity such as ` at most quadratic in the momenta ' . yet
further motivations for weakening phase space at the qm level are given in sec 18 , along with some theoretical consequences .
as continuation of mri and mpi formulations , aphase , daphase , arigphase and darigphase options are relevant alternatives at this point .
the a stands for ` almost ' , referring to it almost being phase , except that the unphysical auxiliary variables velocities ( rather than their momenta ) remain present ; thus its physical content is equivalent to that of phase .
the da stands for ` differential almost ' i.e. the parametrization - irrelevant form in which the unphysical auxiliary variables differentials d@xmath125 are present rather than their velocities @xmath234 .
these then have defined on them of almost - hamiltonians and differential almost - hamiltonians ; this amounts to a fullest distinction at the level of ` total ' @xcite objects : @xmath235 , versus @xmath236 and @xmath237 . upon performing classical reduction taking out lin to pass to a tilded @xmath238 with just a @xmath239 on it , as per fig 2 there is no longer any da - hamiltonian / a - hamiltonian / hamiltonian distinction here , though there does remain a distinction in smeared objects : @xmath240 versus @xmath241 and @xmath242 .
i also mention a set - back as regards the distinctiveness of the lmb relational approach ( hamiltonian collapse problem ) . in passing to the hamiltonian formalism prior to quantization , a lot of lmb - relational versus nonrelational formalism differences are ironed out @xcite .
now both have sum - form total ( ( d)a)hamiltonians whereas the actions were product - form versus difference - form ; moreover the objects of interest are hamiltonians rather than the smeared objects that exhibit a ` da ' trichotomy .
however , as we shall see in secs 16 - 21 , the lmb - relational _ ideas _ , if not precisely the same implementations of these that occur at the classical level , can be recycled at the quantum level .
* alternative : histories theory * holds that , instead of configurations being primary entities , it is * histories that are the primary entities*. this amounts to supplanting relationalism 3 ) with the consequences given in sec 16.4 .
one possible motivation for this extension is that quantization can be formally understood as a ( bad ) functor , linking the classical to the quantum - mechanical .
i argue against this in a number of ways in sec 19 . * categories * @xmath243 consist of objects @xmath244 and _ morphisms _ @xmath229 ( the maps between the objects , @xmath245 ) , obeying a number of axioms ( domain and codomain assignment , identity relations , associativity relations and book - keeping relations , see e.g. @xcite for details . ) * functors * are then maps @xmath246 that obey various further axioms concerning domain , codomain , identity and action on composite morphisms . by the above , while so far we have only been studying objects [ relationalism 3 ) , sec 12 ] , but we should have also been studying morphisms .
in particular the morphisms corresponding to @xmath1 are the so - called _ point transformations _ ( p. 15 of @xcite ) * point * : @xmath247 .
the morphisms of phase space are the canonical transformations can : phase @xmath248 phase .
i agree that momenta and poisson brackets are part of the conception of this world .
thus i am not contesting the passage to ( @xmath230,\ { , } ) .
however , i do at least _ consider _ resisting the suggestion that what mathematically preserves the poisson bracket should be the associated morphisms of this , due to the previous ssec s argument of configuration variables and momenta being operationally distinguishable .
thus it is questionable to take bracket preserving morphisms that do not respect this physical insight .
an alternative would be a rigged version rigphase of the phase space that preserves this distinction . here , the @xmath1-first interpretation of this is that the @xmath64 s are fundamental and each @xmath228 then follows by conjugation and only transforms on @xmath1 are primarily meaningful here , if the @xmath64 s change , then the @xmath228 s follow suit by being the new conjugates , without any extra freedom in doing so .
thus the morphisms of this are just point again , the conjugate momenta being held to follow whatever the coordinates are rather than having their own morphisms ( this approach s intent _ is _ to put configurations first , so one should not be surprised at these and not the momenta having the primary transformation properties .
see fig [ rigphase ] for the ensuing options .
as regards further involvement of categories , we shall see below that perspecting and propositioning , which are themselves at least partly inter - related , lead toward topos theory , a topos being a category armed with some extra structures that make it behave much like the category of sets , @xmath249 ( see sec 14 for more details ) .
rovelli uses relationalism in the sense of objects not being located in spacetime ( or space ) but being located with respect to each other ( though he does not develop this to the extent of the above lmb - a relational postulates ) .
rovelli and crane have , between them , put forward 4 perspectival postulates .
not all have substantial classical forms since some involve observers in ( something like ) quantum theory .
thus there is a gap in my classical line - up of these .
i use persp(@xmath1 ) to mean \{all sub@xmath250}. ` parts ' @xcite is a similar categorical concept ; i do not use this name since it sounds too much like partial ( as in partial observable ) . for @xmath249 ,
parts is embodied by the notion of power set .
moreover , in my own presentation , ` all ' has the connotation ` all physical ' .
e.g. in 2-@xmath3 the subset comprising of the x - component of one particle and the y - component of another lacks in physical significance as a combination ; there are some ways in which @xmath145 can be taken to remember its blockwise origin as @xmath251 .
one role for subconfigurations is in modelling records as localized subconfigurations of a single instant .
additionally , histories theory has an obvious localized subhistories counterpart of this .
perspectivalism 3 ) * rovelli s partial observables * @xcite implement the conceptually - desirable idea of physics being about correlations as well as being regarded as freeing oneself from complete / dirac observables ( which are much harder to find ) .
the partial observables approach is contextual through requiring pairs of partial observables in order to extract entirely physical information .
this is a possible alternative to the dirac or treatments of observables that involves considerably less structure - finding due to aiming to bypass rather than resolve the problem of observables .
it is not however clear exactly what objects to compute in this approach in order to investigate whether subsystems exhibit correlation , or how a number of other aspects of the problem of time can be addressed via these @xcite .
perspectivalism 4 ) * use anything as a time for anything else s motion*. this can be seen as a ` relationalism 7w ) ' alternative to relationalism 7 ) as regards how to abstract time from change ; the w is for ` weakened ' .
i point out its similarity to ` _ tot tempora quot motus _ '
( so many times as there are motions , which is the scholastics view of aristotle , see e.g. p 54 of @xcite . thus i also term this view of time _
amr time _ after aristotle , mach and rovelli .
note 1 ) perspectivalism 3 - 4 ) both have underlying senses of ` democracy ' and of ` not needing to detailedly construct ' , though i argue against these in secs 21 and 22 and in future works @xcite .
one should also take stock that a lot of uses of the word ` relational ' in ashtekar variables / lqg are what this article terms ` perspectival ' , which additionally explains a number of other senses in which the word ` relational ' has been used over the years and to some extent fits various of the uses together . thus e.g. gr in general and the ashtekar variables formulation do have manifest many other senses of ` relational ' outside of the more usually - ascribed perspectival ones .
i note that the present ashtekar variables / lqg literature does not systematically apply lmb(-a ) relationalism ( though there is some evidence for individual pieces of such thinking in various formalisms and techniques used )
. doing so could give the subject further conceptual awareness .
more generally , the extent and manner of the combineability of the rovelli , crane and lmb(-a ) types of relationalism is an interesting one and lqg is a reasonable arena in which to study this question .
the nature of the problem of time facets and strategies covered in this article are generally relevant here too .
i continue with this in @xcite .
physics involves asking and answering questions , i.e. propositions and therefore some kind of _ propositional logic_. as far as i am aware , the idea that physics _ is _ questions which are equivalent to logic was first put forward by mackey @xcite ( see also @xcite for suggestions of this for specific problem of time settings ) .
[ the relationalist will be interested here in those propositions concerning tangible physics .
] i view this as an interesting observer - centred idea that is worth considering whether to elevate to a general physical principle .
i term this ` * mackey s principle * ' isham and linden @xcite previously used this line of thought as a physical principle for histories theory ; my new consideration is to speculate whether this is a _
general _ principle . )
while this plays a bigger , more ambiguous , more subtle role at the quantum level ( sec 21 ) , one can already consider it at the classical level . from the @xmath1-centric perspective of relationalism 3 ) or some @xmath21-centric extension ,
this is an extension to prop(persp(@xmath21 ) ) , where prop denotes the set of all possible physical propositions about @xmath21 .
i am not claiming my treatment of this is exhaustive .
( e.g. i will take ` property ' to be an unproblematic word , for all that is not , see e.g. p. 67 of @xcite for more . )
my point is that the general theory of questions involves , a priori , both nontrivial temporal content and nontrivial atemporal content .
then in strategies in which one can in fact purge questions of temporal content , one s remaining theory should parallel the nontrivial atemporal content of questions , so in that sense the structure of the atemporal content of questions is also important for this article . implementing propositioning in at least quite general terms
involves * topos theory*. ( a _ topos _ is a category with three extra structures : 1 ) finite limits and colimits 2 ) power objects and 3 ) a subobject classifier @xcite .
this renders topos theory a more advanced topic than category theory itself , possessing also has ties to geometry and to logic @xcite . )
these allow a topos to bear further similarities to @xmath249 ( the familiar category of sets ) and boolean logic pinned on that . then _ constructing a theory of physics is equivalent to finding a representation in a topos of a certain formal language that is attached to the system " _
@xcite ( see also @xcite ) .
this approach also renders regions of configuration space [ or more generally of @xmath21 , sub@xmath252 primary , rather than the points out of which one usually considers the regions to be constituted of . at the classical level , propositioning
is implemented simply as @xmath249 , with boolean logic itself entering as subobject classifier .
one can consider this to be in terms of regions of configuration space or some extension @xmath21 or sub(@xmath21 ) , with union , intersection and complement playing the roles of and , or and not .
classically , the situation is straightforward @xcite : one is to use @xmath249 as one s topos , which comes with the obvious boolean logic entering as subobject classifier .
propositions about configuration space regions fit well within this classical scheme .
this is based on quantities having values in ( borel ) sets within some state space such as @xmath1 or phase .
open ( or clopen : simultaneously closed and open ) subsets are used .
one is considering maps from state objects in @xmath169 to quantity - value objects in @xmath253 . here
, @xmath169 is the linguistic precursor of @xmath21 .
there is a language which is independent of theory type ( classical / quantum ) though @xmath253 clearly does depend on this : the propositional language pl(@xmath21 ) .
finally , for further use at both the classical and quantum levels , i make this extension of the arbitrary @xmath2-frame method / best matching procedure .
one * builds up * compound objects as bona fide combinations of one s primary objects ( * tensor / bundle * structure ) .
this applies to the relational action ( ` relationalism 4 ' ) , but also more generally as a ` building up ' move , producing all of actions , notions of distance , notions of information and correlation , and quantum operators .
the actions in question are jacobi or jacobi synge type actions .
* @xmath2-act @xmath2-all move*. whilst @xmath2 s physical irrelevance can occasionally be directly implemented by @xmath2-independent objects , far more usually one has to implement it indirectly via incorporating @xmath2-*act * , @xmath2-*all * pair of moves into the sequence of maps in one s build - up .
this builds @xmath2-bundle objects / bibundle objects with one bundling involving @xmath2 .
schematically , for continuous or discrete isometries ( or a mixture ) , this is a metric background invariantizing ( mbi ) map = _ o for whichever object @xmath254 upon which @xmath2 is able to act and whichever maps render this into a physically - interesting combination such as an action , a qm operator or one of the notions of distance / information / correlation of sec 14 of @xcite . see @xcite for a topological - level counterpart .
this is a generalization of how one might go about implementing configurational relationalism .
that was done in sec 6 and 7 s best matching via the ` all ' being a variation and the act an infinitesimal continuous isometry .
however , the generalization also covers group - averaging ( sum and divide by order is ` all ' , with ` act ' being a suitable group action ) , or the similar group - summing / group - integrating , or inf - taking ( e.g. in gromov hausdorff type constructions @xcite ) . and
it is useable not only on actions , but also in the study of both classical and quantum notions of information , notions of distance , in halliwell - type class functions @xcite ( alongside combinations of these that form decoherence functionals @xcite ) , and for the quantum operators in sec 14.6 .
@xmath2-act , @xmath2-all also plays a role as regards classical notions of distance and of information , for quantum operators @xcite and for class functions and decoherence functionals in histories / halliwell - type approaches @xcite ( halliwell - type approaches are combined histories - semiclassical - records approaches to the problem of time ) .
the above is all at the level of the configuration space metric ; for brief discussion of topological relationalism , see @xcite .
i consider the quantization scheme along the lines of isham s @xcite , which , whilst quite general , is by no means all - embracing .
i take this to consist of the following steps as composed in fig 3 . * select * is the selection of a set of classical objects that are to be promoted to quantum operators , alongside passing from the classical poisson brackets to some commutation relations , algebra comal .
it is very obvious in the simplest example : @xmath255 , @xmath256 alongside the usual ( ` correspondence principle ' ) \{^ , _ } = ^ _ [ ^ , _ ] = i^ .
[ common ] however , in general it exhibits a number of subtleties @xcite .
more generally , comal : = ( a chosen set of f(@xmath257 alongside a bracket structure ) . here
, one has to in general make a _ choice _ of a preferred subalgebra of phase objects @xcite to promote to qm operators . * assoc * then associates a pre - hilbert space prehilb of pre - wavefunctions for these operators to act on . *
kinquant*:= select @xmath258 assoc is _
kinematical quantization_. * h - rep * is how to promote the hamiltonian @xmath259 to a function @xmath260 of kinquant s operators , yielding a wave equation .
* hsolve * is then solving this : passing from prehilb to hilb : the hilbert space within that is annihilated by @xmath260 .
note that hilb s inner product restricts what are valid operators by self - adjointness .
* dynquant * : = hsolve @xmath258 hrep , the dynamical quantization .
* quant * : = dynquant @xmath258 kinquant is , finally , this approach s basic notion of _
quantization_. n.b .
as formally defined here , this acts on a _
triple _ such as ( phase , can , @xmath259 ) or ( rigphase , point , @xmath259 ) .
tangibility , temporal relationalism and configurational relationalism ideas remain relevant in the quantum arena . however , _
implementations _ are different now , for at the classical level these were implemented at the level of actions , whereas in quantum theory there usually are no actions .
one widespread and simple implementation of tangibility at the quantum level is to ultimately consider not wavefunctions or operators @xmath261 but rather combinations @xmath262 ( which include as subcases _ expectations _ : @xmath263 , and _ overlap integrals _
@xmath261 = i d ) .
i note that ordinary qm itself is not leibniz - timeless .
it is whole - universe qm that is , at least at first glance at conventional forms for its equations . in more detail ,
temporal relationalism @xmath108 mpi actions @xmath108 homogeneous quadratic constraint @xmath173 @xmath108 at the quantum level the * frozen formalism facet * of the problem of time , i.e. that one has a timeless wave equation @xmath264 , in contradistinction from such as a time - dependent schrdinger equation @xmath265 , klein
gordon equation @xmath266 or dirac equation @xmath267 . as wavefunction versus @xmath268 as wavefunction _ of the universe_. @xmath269 is a dirac spinor and @xmath270 are dirac matrices . ]
each of the latter set of equations are common in ordinary qm but entail absolute time notions ( whether newtonian in the first example or minkowskian in the other two ) .
note also how in the relational approach the frozen formalism facet readily follows from the leibniz temporal relationalism principle , rather than coming as a surprise or being considered to arise from mri , which is , in the lmb - relational approach , not an explanatory principle but , rather a mathematical implementation of the actual leibniz time principle .
the problem of time @xcite occurs because ` time ' takes a different meaning in each of gr and ordinary quantum theory .
this incompatibility underscores a number of problems with trying to replace these two branches with a single framework in situations in which the premises of both apply , such as in black holes or the very early universe .
other problem of time facets are introduced below .
there are approaches to qg in which time is not primary [ thus implementing leibniz s relationalism 5 ) ] but which then have an emergent time or a semblance of time arising at the quantum level .
note how mach provides a suggestion for how to resolve timelessness .
variants of the mach time principle are discussed in secs [ mach-1 ] , [ sssec : rwr ] , the conclusion , and @xcite .
the last three of these references permit finer machian assessment of problem of time strategies than are given below .
i next analyze the usual set of problem of time strategies from a simple machian time perspective .
\a ) perhaps one is to find a hidden time at the classical level @xcite by performing a canonical transformation under which the quadratic constraint is sent to p_t^ + _
= 0 [ casta ] ( for @xmath271 the momentum conjugate to some new coordinate @xmath272 that is a candidate timefunction ) .
this is then promoted to a @xmath272-dependent schrdinger equation , i/ = _ .
particular cases of this are as follows .
a.1 ) scale as internal time .
a.2 ) superspace time @xcite is a further variant on this , following from the scale contribution giving the indefiniteness in gr kinetic term , a feature not found elsewhere in physics and which causes a number of difficulties .
( e.g. invalidation of the usual schrdinger interpretation of the inner product , and a whole new theory for the dynamical meaning of zeros in the potential factor . )
however , scale fails to be globally monotonic in general and the superspace time approach fails due to the gr potential term being incompatible with the gr kinetic term s conformal killing vector .
a.3 ) york time ( the conjugate of a particular scale variable ) . however , for this the practical scheme breaks down , and the split of @xmath174 into embedding and true degrees of freedom has a global impasse .
a.4 ) matter appended to the geometrodynamics might provide a timefunction ( matter time ) .
as criticisms of approach a ) in general , this is not in line with leibniz timelessness ( since the time in use is taken to a priori exist in general at the classical level : tempus ante quantum @xcite ) , nor with the lmb - type take on mach s time principle , since it involves using one particular change as the time for all the other changes .
moreover , matter time schemes themselves are typically additionally suspect along the lines of intangibility , via often involving such as ` reference fluids ' , and often endowed with unphysical properties to boot .
[ the intangibility might be taken as cover for the unphysicality , but is itself a conceptually suspect way of handling the problem of time along the lines exposited in the present article . ]
jbb time itself is tempus ante quantum , but is now classically emergent and thus in line with leibniz - timelessness and the lmb stance on mach s time principle .
one particular scheme for this has it arise to zeroth approximation from slow heavy h degrees of freedom , to which light fast l degrees of freedom then contribute to first order ; see below for physical discussion of the nature of h and l degrees of freedom . on the other hand , this approach does _ not _ gives a frozen formalism resolution .
however , we shall see it is approximately recovered by the next entry , which does provide a frozen formalism resolution .
\b ) perhaps instead there is no time in general but a notion of time does emerge under certain circumstances .
the most notable example of this is the semiclassical approach . here , one has slow heavy ` @xmath273 ' variables that provide an approximate timestandard with respect to which the other fast light ` @xmath274 ' degrees of freedom evolve @xcite . in the halliwell
hawking @xcite scheme for gr quantum cosmology , @xmath273 is scale ( and homogeneous matter modes ) and @xmath274 are small inhomogeneities .
l alignment with scale shape removes ambiguity of how to allot h and l roles .
the semiclassical approach involves making the born
oppenheimer ansatz @xmath275 . next
, one forms the @xmath273-equation ( @xmath276 for rpm s ) , which , under a number of simplifications , yields a hamilton
jacobi degree of freedom and with no linear constraints . ]
equation \{w / h}^2 = 2\{- ( h ) } [ hamjac ] for @xmath277 the @xmath273-part of the potential .
next , one way of solving this is for an approximate emergent semiclassical time @xmath278 .
finally , the @xmath274-equation @xmath279 can be recast ( modulo further approximations ) into an emergent- time - dependent schrdinger equation for the @xmath274 degrees of freedom , i|/t^ = _ l| .
[ tdse2 ] ( here the left - hand side arises from the cross - term @xmath280 and @xmath281 is the piece of @xmath282 that survives ) . note that the working leading to such a time - dependent wave equation ceases to work in the absense of making the wkb ansatz and approximation , which , additionally , in the quantum - cosmological context , is not known to be a particularly strongly supported ansatz and approximation to make .
then the scale part contributes an approximate timestandard with respect to which the shape part runs according to usual positive - definite kinetic term physics ( ` scale - indefiniteness alignment ' exploited by aligning h with both ) .
this can then be fed into more promising histories - records - semiclassical combination strategies .
note that the identification of the @xmath86 is unambiguous if one uses scale = h and shape = l split .
this puts that ` cursed ' heterogeneity to a use it is the cure to approximate time selection in cosmological models .
need to see the crudest estimate of t here is not very machian , but the corrected version does allow for l - changes to contribute too , thus rendering it very machian .
overall , emergent semiclassical time is machian once back - reaction of the l - system on the chroniferous h - system is incorporated .
this is qualitatively similar to the emergent jbb time scheme at the classical level .
\c ) a number of approaches take timelessness at face value .
the most natural follow - on for relationalism 3 ) is the ` tempus nihil est ' postulate .
one considers only questions about the universe ` being ' , rather than ` becoming ' , a certain way .
considering physical questions about a system amounts to an algebra of propositions ; this is implemented by regions or , better , by projectors at qm level .
particular versions of timeless approaches that enter this article s discussions are as follows .
c.i ) the _ _
@xcite concerns the ` being ' probabilities for universe properties , such as ` what is the probability that the universe is flat ? ' one obtains these via consideration of the probability that the universe belongs to region r of the configuration space that corresponds to a quantification of a particular such property , _
||^2 , [ nsi ] for @xmath283 the configuration space volume element .
c.ii ) the _ _
@xcite goes further by addressing conditioned questions of ` being ' such as ` what is the probability that the universe is flat given that it is isotropic ' ?
c.iii ) _ records theory _
@xcite involves localized subconfigurations of a single instant .
more concretely , it concerns whether these contain useable information , are correlated to each other , and a semblance of dynamics or history arises from this .
[ i do not use emergence and semblance as synonyms . ]
this requires notions of localization in space and in configuration space as well as notions of information ; in fact , this article adds to existing criteria as regards specifying a mathematically - precise records theory .
there is trickery to reduce temporal logic to atemporal logic @xcite , and i d itself is then a fine study of atemporal logic ( though this records - theoretic application of i d is , as far as i know , new to the present article ) .
analysis of attitudes to time permits classification of the times that arise from timeless approaches .
this involves , , page , barbour , gell - mann halliwell and hartle records , my records , the combined approaches of halliwell @xcite and of gambini porto pullin @xcite , alongside extensions of these that include notions of observables / beables @xcite , and also the rovelli approach , as well as suggesting a number of further hybrid programs , and is to be covered in more depth in a subsequent paper @xcite .
\d ) perhaps instead it is the histories that are primary ( _ histories theory _
@xcite ) ; this is an alternative to ` @xmath1 is primary ' or ` phase space is primary ' .
more generally , i note that temporal relationalism is implemented via mri into suitable formulations in terms of path integrals .
additionally , one can use emergent jbb / semiclassical time to formulate histories theory .
barbour - relationalism s indirect implementation at the classical level gives constraints @xmath284uad and @xmath285 in .
the reduced/ relationalspace r - scheme gives a constraint @xmath286 . both these schemes are relational , thus dirac and reduced schemes are both relational .
in the preceding literature , barbour , rovelli and smolin @xcite had largely not considered the lmb brand of relationalism along the above lines beyond the general idea that dirac quantization is an indirect scheme that implements configurational relationalism .
barbour had on the other hand considered a timeless records approach @xcite which has some relational features .
@xcite used the @xmath2-act , @xmath2-all method for timeless records theory , and @xcite used it along the lines of halliwell s combined approach .
methods such as the group integration approach ( essentially a variant / particular implementation of the dirac quantization approach ) are fairly widely used in lqg ( e.g. in the master constraint program @xcite ) . in this sense , lqg is configurationally relational at the quantum level . in this more general setting [ fig [ rawr].a ) ] , * quant * : = dynquant @xmath258 @xmath2-quant @xmath258 kinquant .
@xmath2-*quant * : = @xmath2-solve @xmath258 @xmath2-rep , each of which individual steps is defined in obvious parallel to @xmath287-rep and @xmath287-solve .
@xmath2-*solve * sends prehilb to midhilb , standing for ` middlingly physical hilbert space ' .
i also define @xmath288-*rep * = h - rep @xmath258 @xmath2-rep ( these maps commute ) , and @xmath288-*solve
* = h - solve @xmath258 @xmath2-solve .
these are particularly useful in approaches that do not discriminate between linear and quadratic constraints .
reduced - type formulations also embody configurational relationalism .
these are essentially tilded versions of fig 3 , available whenever one has the good fortune to , at the classical level , beable to directly formulate or reduce away all the redundancies from the indirect formulation .
this is covered in fig [ rawr]b ) and amounts to a tilde - ing of fig 3 ( itself following on from the classical reduction diagram of fig 1 ) .
qm makes use of further operators .
group averaging / refined algebraic quantization looks to be a good _ quantum _ implementation of configurational relationalism . given a non - configurationally - relational operator @xmath261 , _
( i _ _ ) ( - i _ _ ) ( for @xmath289 the action of infinitesimal generators ) is a relational counterpart , which is another example of @xmath2-act , @xmath2-all procedure based on the exponentiated adjoint group action followed by integration over the group
. one can do similarly to pass from non - configurationally relational states to configurationally - relational ones .
admittedly , there are well - definedness issues in general as regards the measure @xmath290 over the group @xmath2 . in particular
, treating the diffeomorphisms in this way is problematic .
see sec 14 of @xcite for further discussion .
configurational relationalism is linked to a number of further problem of time facets .
although specific properties of the diffeomorphisms are very important , quite a few features of the problem of time follow from relationalism alone rather than additional insistence of @xmath2 containing ( more or less ) diff(3 ) .
* best matching problem*. i discussed in sec 9.5 how this is the generalization and configurational - relational conceptualization of the usually so - called thin sandwich problem .
it is usually taken to be a classical facet of the problem of time , though additionally sec 9.5 s set - up of the thin sandwich set - up is in analogy with the qm set - up of transition amplitudes between states at two different times ; this was indeed wheeler s original motivation for it @xcite .
* functional evolution problem*. this concerns anomalies in the quantum commutator algebra , i.e. the further quantum counterpart of relationalism 10 ) .
more specifically , a subset of anomalies are time - related e.g. through the anomalous terms being foliation - dependent .
* problem of observables*. this , and its resolutions , are as stated at the classical level in sec 9.4 but now involving quantum objects and commutator algebra in place of poisson brackets algebra .
* global problem of time*. quantum mechanically , this is rather more involved than the classical meshing condition , now being some kind of largely unexplored statement about meshing together not of coordinate systems but of _ unitary quantum evolutions_. * multiple choice problem * this is partly due to coordinate choices ( which the global problem of time s need for multiple coordinate patches compounds ) , or due to choices of functions to promote to an operator algebra ( see sec [ qmqprim ] ) .
it then manifests itself at the quantum level in the form of different choices being capable of leading to unitarily - inequivalent quantum theories .
these last two facets are not particularly lmb relational ; we shall see , however that they are at least partially rc relational .
once one s @xmath2 includes diffeomorphisms , the * foliation dependence problem * arises as a further possible source of background dependence ( which would be in discord with lmb relationalism ) and there is also rather more of a * spacetime reconstruction / replacement problem * than in sec 9.7 s classical treatment .
here , the frozen wave equation is = -^^ij + v(^k)= e[zut ] summed from 1 to 3 in i , j and 1 to 2 in @xmath70 , alongside the quantum zero total angular momentum equation = _ _ ^3 ^i = 0 , [ qzam ] this working , unlike that of the next ssec , readily generalizes to all dimensions and particle numbers .
here , @xmath268 is the wavefunction of the ( model ) universe .
rpm dirac - qm study was developed in @xcite .
note 1 ) relational triangleland can not distinguish between dimensions 2 and 3 ; in @xcite i used this to argue that dimension is not an a priori given in rpm s .
note 2 ) @xmath2-quant restricts the am quantum numbers of the base and apex subsystems to counterbalance each other @xcite , thus causing a substantial collapse in the physically admissible spectrum of states for this theory .
as regards the indirect definition of rotation - averaged operators , triangleland s @xmath291 is a particularly simple case , for which _
g g = _ ^1 = _ = 0 ^ 2and @xmath292 is the infinitesimal 2-d rotation action via the matrix @xmath293 acting on the vectors of the model .
@xmath294 is here the absolute rotation .
note 1 ) clearly ( [ zut ] ) exhibits the frozen formalism facet of the problem of time .
note 2 ) rpm s exhibit no functional evolution problem nor have any place for discussions of foliations or of gr - like spacetime reconstruction , thus evading two further problem of time facets .
they clearly do exhibit the other problem of time facets .
direct r - qm study for rpm s was developed in @xcite .
looking at specific simple solvable examples such as in these last 2 ssecs , i find ( contrary to suggestions in e.g. @xcite ) a _ lack _ of non - standard qm . by this standardness of the actual mathematics involved
( just new relational meanings for well - known equations and special functions ) , plenty of machinery is available for solving , and performing further calculations with , simple examples of quantum rpm s .
note however that these models are appropriate as whole - universe models and for detailed analysis of the problem of time .
thus this coincidence is a triumph as regards finding tractable toy models for the investigation of closed - universe questions ( problem of time , foundations of quantum cosmology ... ) for scaled triangleland , the frozen wave equation is = - \ { \ { \ { ^2 } + \ { } + } } + = . [ sphetise ] note 1 ) some differences between absolutist and relational qm are as follows ( these are well understandable from within the usual kind of a mathematical framework for quantum theory ) .
there are differences of the subsystem versus the closed whole - system sort between absolutist qm and relational qm subsystem physics in a relational universe would be expected to be linked by global restrictions due to its constraint equations .
such differences were already found to be present in the gr case by dewitt @xcite .
there is no a priori time , but many parts of qm do not depend on time , and @xcite explains how the remaining parts can be accessed using emergent time dependent wave equations .
note 2 ) as regards the remaining problem of time facets for rpm s , in 1- and 2-@xmath3 , the best matching problem is resolved ( by the reduction process itself ! ) also , the observables are known @xcite and geometrically natural , and dirac observables can then be constructed along the lines of @xcite , so the problem of observables is here well under control . :
= -^2 \ { \ { ^ } - ( h;]}- ( x;h]+ 2= 0 [ wde3 ] , _ : = -2d_^ _ = 0 _ : = -2d_h^/h _ = 0 .
[ 5 ] the latter implements configurational relationalism , signifying that the wavefunction of the universe is to be a function of the 3-geometry alone .
whilst classical gr had its constraint algebra close and the refoliation invariance property by notes 1 ) and 2 ) of sec 9.4 , there are no known corresponding results at the quantum level , constituting the functional evolution problem and foliation dependence problem facets of the problem of time .
one now has a new reformulation of the quadratic constraint and of the momentum constraint , alongside a new linear @xmath208-yang mills constraint that reflects the extra degeneracy of the presentation s @xmath295
diff(3 ) @xmath296 .
the loop representation takes back this last degeneracy , whilst considering and knots formally uses up the diff(3 ) constraint too .
this amounts to performing the @xmath2-quant steps of a dirac - type approach .
i also comment that one source of good names for physical programs that purport to be relational and thus concern tangible entities , is to name the theories after the tangible entities rather than after convenient but meaningless parts of the formalism that happen to be in use in the mathematical study of that theory . by this criterion , ` geometrodynamics ' and ` shape theory ' @xcite/`dynamics of pure shape '
@xcite/`shape dynamics ' @xcite are very good relational names , whereas the name ` loop quantum gravity ' ( lqg ) leaves quite a lot to be desired from a relational perspective . for , in this program
the @xmath208 gauge group is put in by hand with the particular choice of spinor / bein variables that happen to simplify gr s canonical formalism .
it is only the subsequent removal of its significance that brings loops into the program s formalism , but the lion s share of the work in _ any _ canonical quantum gr program resides , rather , in removing diff(@xmath169 ) and in interpreting the hamiltonian constraint . even without the ` put in by hand '
issue , the parallel naming for rpm s would be ` sphere theory ' or ` sphere dynamics ' from this being the form the preshape space takes before the lion s share of the work is done in removing rot(@xmath3 ) ( and , in the dynamics case , interpreting the energy - type constraint ) .
but this is rather clearly _ not _ enlightening naming because the preshape space spheres do _ not _ play a deep conceptual role in shape theory . as such , i would suggest names along the lines of ` knot quantum gestalt ' or ` quantum nododynamics ' .
this is also particularly appropriate via the gestalt conceptualization of gr being , at least at the level of metric structure , very much emphasized in the ` lqg ' program itself , just not in the lqg program s _
name_. thiemann s alternative name ` quantum spin dynamics ( qsd ) ' is somewhat confuseable with such as the ising model and its ilk , and also the spin connotations come from the @xmath208 gauge groups . on the other hand , letting the s
stand for ` spin - net ' disambiguates the nature of the theory , though a spin - net still does have but a similar ontological status to a loop .
taking relational primality further , then , the final suggestion is that the s should stand for s - knot ( spin - knot ) .
a useful further project would be to provide a lmb(-a ) relational account of the quantum part of thiemann s book @xcite ( and of other schemes for modern lqg , there being a variety of such schemes ) .
qm @xmath1-primality motivation 1 ) on the one hand , one often hears that qm unfolds on configuration space , on the other hand most hold that it unfolds equally well on whichever polarization within phase space , of which configuration space is but one .
qm @xmath1-primality motivation 2 ) i note that isham s procedure @xcite for kinematic quantization itself favours @xmath1 : the canonical group comes from just this and then the @xmath298 is controlled by @xmath1 and its associated @xmath2 and @xmath299 so everything is controlled by configuration space too .
qm @xmath1-primality motivation 3 ) i expect then also for point to be relevant to pre - hilbert space , and for the operator - orderings of the constraints likewise .
this enables a connection with a conjecture of dewitt concerning operator - ordering in chapter 6 of @xcite .
note 1 ) this sec s somewhat outlandish suggestion is not the only possible reason behind motivations 2 and 3 ; indeed if can plays a role , can is an enlargement of point , and point is then the ` can that act on @xmath300 ' restriction , which suffices to cover dewitt s issue .
note 2 ) ( rigphase , point ) , as classically motivated in sec 11 and argued to be a sensible implementation of @xmath1-primality , has the same objects and brackets as ( phase , can ) and hence the same kinquant . but in restricting can to point as morphisms preserving poisson brackets and the rigging , then comal s morphisms would likewise be restricted .
a possible problem with this view is that qm s @xmath228 s and @xmath64 s may be more alike than their classical counterparts .
qm @xmath1-primality motivation 4 ) a further reason to question the licitness of all canonical transformations is that classical equivalence under canonical transformations is in general broken in the passage to qm as per e.g. the * groenewold van hove phenomenon * @xcite .
moreover , the groenewold van hove theorem continues to apply in the rigphase approach , since one can envisage choosing @xmath301 and @xmath302 there instead of @xmath303 and @xmath304 .
thus , whilst rigphase is an illustration of weakening the canonical transformations , it is not the _ correct _ weakening to take into account gronewold van hove ; i leave as an open question what the set of canonical transformations that do become unitary transformations at the quantum level is characterized by .
note 2 ) using just point does _ not _ resolve out all aspects of unitary inequivalence , though it may help / may be a step in the right direction .
a deeper question is which weakening of the can s or point s at the classical level form up into classes that are preserved as unitary equivalence at the quantum level .
( rigphase , point ) itself is at the bottom of the hierarchy , whereas the solution of the currently posed problem is likely to involve a rather larger proportion of the canonical transformations .
consequences of using rigphase in place of phase are as follows .
holding canonical transformations in doubt affects internal time and histories theory approaches to the problem of time , as well as ashtekar variables and the recent linking theory approach @xcite .
the second and fourth of these make _ more _ than the usual amount of use of canonical transformations ; this sec suggests , moreover , the possibility of developing physics in the _ opposite _ direction .
* histories primality alternative * further reasons for histories theory appear at the qm level @xcite .
this follows on from the path integral perspective , and additionally represents various modified interpretations of qm that are more suitable in the whole - universe context .
i analyze these by ` taking this functor apart ' into various simpler constituents as per figs 3 and 4 .
a number of these steps already have individual difficulties . two reasons for the kinquant functor to be ` bad ' ( or , at least , ambiguous ) are 1 ) preferred subalgebra choice ( and the nontriviality thereof by e.g. groenewold van hove and 2 ) global considerations .
\3 ) a further reason for badness is anomalies arising in the passage from the poisson brackets algebra to the commutator algebra .
\4 ) in the case of h - rep for geometrodynamics , there are also well - definedness issues ( functional derivatives , combinations with no functional - analytic reasons to be well - behaved ) and regularization issues .
\5 ) on top of these individual difficulties , the order in which some of these procedures are performed affects the outcome ( e.g. reduce and then conformal - order does not match up with conformal - order and then reduce ) ; this is covered in detail in chapter 6 of @xcite .
this bad functor conclusion , if not all the details of why , matches with baez s @xcite point of view , which was probably arrived at considering other / more quantization approaches than isham 1984 .
a different line of objection to quantization itself is that it is only desirable in the incipient parts of conceiving a quantum theory ( so it is tied to an understood classical theory , but qm is more fundamental so one needs to pass to first principles quantum theory which one then has the sometimes fatal hope of recovery of a recognized classical limit ) .
[ for some particular examples , the functor does work out , but the above demonstrates that for suitably general theories there is not at all a standard prescription for quantizing , and formatting this as a functor does not in any way help with that . ]
moreover , further and distinct use of categories in quantization are as follows .
option 1 ) quantizing other categories - treating their objects as one usually treats @xmath1 .
an simple example of this is in some histories theory work , in which one considers the ` history brackets ' of histories and their conjugate momenta @xcite .
more general example 1 ) .
this is how isham @xcite considered quantization on quite general categories , with the morphisms / arrows now playing the role of momenta .
localized subconfigurations furthermore play an underlying role in crane s thinking @xcite .
this involves the following postulate ( my nomenclature ) .
perspectivalism 1 ) * quantum theory only makes sense for subsystems * ( my caps ) . in the quantum gr arena , crane deals with this by considering splits of the universe with the observer residing on the surface of that split .
each of these splits then has its own hilbert space .
crane then maintains that the whole universe has no hilbert space ( so his version of what i term persp(@xmath1 ) has @xmath305 in place of my @xmath306 ) , though a hilbert space is to be recovered in a semiclassical limit .
i dissect this as follows due to not agreeing will all parts of it .
crane 1 ) qm makes sense for subsystems ( true , and obvious ) ; each has its own hilbert space , within which the standard interpretation of qm applies . crane 2 ) almost all qm concerns subsystems ( true at an elementary level , but sometimes forgotten in quantum cosmology ) .
crane 3 ) what would be qm for the whole universe does not possess a hilbert space or the standard interpretation of qm . a crane difference 1 ) here i disagree about the lack of hilbert space ( shrunken as it may be due to closed - universe effects as per secs 7 and 10 ) , but agree about the subsequent non - applicability of the standard interpretation ( though this is well known and plenty of different groups of authors have been willing to work under such conditions ) .
that is why i use @xmath306 and not @xmath305 ; this difference also accounts for my replacing ` all ' by ` almost ' .
i envisage the scalefactor of the universe as a possible whole - universe property that can plausibly enter one s physical propositions ; at the very least , scale is not a _ localized _ subsystem .
a crane difference 2 ) finally , crane did not stipulate his sub@xmath1 to carry the same local physical connotations as mine ( as per sec 14.1.1 ) .
crane 4 ) qm makes sense for the universe as a whole in some kind of semiclassical limit . here
i agree with crane , though in detail we may well not both mean exactly the same thing by a ` semiclassical limit ' .
moreover crane 4 ) is less necessary from my perspective ( it applies just to interpretation , rather than to the possession of a hilbert space as well ) .
perspectivalism 1w ) ( w for ` weakened ' ) is the name by which i shall refer to my above version [ crane 1 , 2 ) and my side of a - crane differences 1 , 2 ) ] .
note 1 ) perspectivalism 1w ) is what i posit as the perspectival expansion of relationalism 3 ) [ though one _ could _ instead elect to use perspectivalism 1 ) in such a role ] .
note 2 ) whilst i have no evidence for such a connection , i comment that on the one hand one has multiplicity by inequivalent quantum theories and on the other hand multiplicity by multiple observer perspectives .
are there ever any connections between these two multiplicities ?
perspectivalism 2 ) allows for the qm of * observers observing other observers observing subsystems*. this provides a further level of structure between the plethora of hilbert spaces associated with all the perspectives of perspectivalism 1 ) .
this approach contends that @xcite * qm propositions only makes sense in the context of particular observers * ( my rewording ) .
( this has no nontrivial classical counterpart . ) note 3 ) in rpm s , the perspectival approach can be toy - modelled using the notion of a localized fictitious test - observers who can measure nearby inter - particle distances and perhaps relative angles about their particular position , archetypes being considering the base subsystem of a triangleland or the base alongside the relative angle ( the ` fictitious test - observer more or less at the centre of mass of a localized binary ' ) .
this can accommodate perspectivalism 1 ) to some extent , but can not meaningfully incorporate perspectivalism 2 ) which explicitly requires non - negligibility of observer material properties .
note 4 ) the crane set - up then involves ` the entire set of subquant s ' , with perspectivalism 2 necessitating yet further structure beyond that , as well as nontrivial modelling of observers as quantum systems [ an idea nobody can really calculate with , which also features in the thinking of page @xcite and of hartle @xcite ; the state of the art here is hartle s information gathering and utilizing system ( igus ) model for observers @xcite .
my preceding ssec makes the possible advance that ` the entire set ' requires physical sub@xmath1 s rather than the greater mathematical generality of sub@xmath1 s , and that it is additionally contingent on a definition of localization in space to be obeyed as a further practical consideration in the selection of a set .
i add that these desirable features would not appear to be conducive to the implementing sets being mathematically simple ; the notion of localization in space in particular would introduce a source of ambiguity into the model : _ how _ local and by what criterion . the concepts embodied by perspectivalism 3 ) and 4 ) carry over straightforwardly to the quantum level .
a form of perspectivalism 4 ) purely at the qm level had previously been proposed by page and wootters @xcite , and this position ( if not accompanying technical details ) was also taken up by gambini porto
pullin @xcite .
represent propositions at the quantum level by projectors is a key move . in ordinary quantum theory , for state @xmath307 and proposition @xmath304 implemented by projector @xmath308 , prob(@xmath304 ; @xmath307 ) = tr(@xmath309 ) with gleason s theorem providing strong uniqueness criteria for this choice of object from the perspective of satisfying the basic axioms of probabilities ( see e.g. @xcite ) .
furthermore , in some approaches to the problem of time , one goes beyond @xcite the usual context and interpretation that are ascribed to projectors in ordinary quantum theory .
propositions were used from the inception in the @xcite . desiring a projector implementation of propositions led to isham and linden s reformulation of histories theory . the records theory within histories
theory is , more trivially , satisfactory from the perspective of having its propositions represented by projectors . on the other hand ,
the and halliwell approaches are still using configuration space regions rather than proposition - projector association @xcite .
prima facie , this is suspect .
moreover , one might be able to pin a topos - type interpretation on such practises ( though this has indeed not been worked out yet ) .
one possible more general mathematical implementation of propositioning is in * isham doering ( id)-type approaches * @xcite to using topos theory in physics provides a large amount of candidate structure as regards realizing prop(@xmath21 ) at the quantum level . quantum - mechanically , there are distributivity issues with the logic , insofar as quantum logic @xcite is not distributive .
however , alternatively , internal logic , which i d term ` intuitionistic logic ' @xcite may be more applicable than quantum logic , and this is now distributive again . given a state , truth values
are assigned to all propositions about the quantum system , in which sense this is a neo - realist approach .
this is independent of measurements / observers ; this is to be compared with the crane - type characterization of subsystems in terms of boundary observers ( sec 15 ) in sec 16.6 .
in fact , the i d study has 2 languages at the quantum level : as well as the propositional language pl(@xmath21 ) , it has a higher - order typed language , l(@xmath21 ) .
the propositional language is simpler and more directly related to standard quantum logic .
however , standard quantum logic comes with various problems , so that being an intuitionistic logic distinct from standard quantum logic is an advantageous feature of @xmath310 ) [ as opposed to @xmath311 . here , there is a map that maps projectors to subobjects of the spectral presheaf , which is the qm analogue of a classical state space .
moreover , the topos perspective itself favours the higher - order language l(@xmath21 ) as the natural kind of language that can be represented in a topos ; it is ` internal ' to it .
note 1 ) this approach does require radical revision of qm itself @xcite , amounting to avoiding the obstruction in standard qm due to the kochen specker theorem .
note 2 ) the logic role played in classical physics by the boolean algebra is played more generally in qm - relevant topoi by a heyting algebra @xcite ( this and the boolean algebra are both distributive lattices , the difference being that the heyting algebra has no law of the excluded middle ) .
note 3 ) implementing propositions in this fashion is more sophisticated than the or halliwell s combined approach , for which one does so by thinking about classical regions that have limitations as regards implementing the entirety of qm propositions and relations between them .
note 4 ) qm ends up involving the topos @xmath312 , i.e. the topos of contravariant ( ` opposite ' ) valued functors on the poset category @xmath313 of commutative subalgebras of the algebra of bounded operators on the hilbert space of the system , @xmath287 .
this article concerned relationalism . i contrasted and composed the very different lmb ( leibniz mach
barbour ) and rc ( rovelli crane ) uses of this word .
moreover , i argued strongly that relationalism is relevant to background independence and that background independence is an important part of many of the leading ` quantum gravity ' programs , to the extent that i took qg to stand , rather , for _ quantum gestalt_. this refers to the relativistic gravity background independence gestalt as exemplified by gr ( taken to be not only a set of relativistic field equations for gravitation , but also a _ philosophical perspective _ about freeing physics of background structures held by einstein among others ) .
finally , i argued that background independence is wo nt to lead to many of the facets of the problem of time , so that valuing background independence means having to , among other concerns , face up to this conceptually and technically difficult issue . in the lmb approach , jacobi
synge actions are more relationally primary than euler lagrange ones or equations of motion , since the first are leibniz - timeless whilst the last two are , from the relational perspective , in terms of a machian emergent time and thus not possibly relationally primary .
leibniz timelessness implies no primary notion of velocity and thus that the usual canonical definition of momentum needs reformulating ; i provide how to do this in terms of the jacobi
synge line element .
i gave further @xmath1 and @xmath2 compatibility requirements .
i considered qm counterparts of the classical relational implementation .
i then provided four expansions on barbour s ` @xmath1 is primary ' : to phase space ( or possibly some slightly weaker structure ) , categorization , propositioning and perspecting .
whilst categorization might be argued for as a prerequisite to quantization , i argued against this point of view in general ; however , a more complicated version of categorization ends up underlying propositioning and perspecting .
i argued that the ideas of isham , doering and butterfield concerning applying topos theory to quantum theory are one possible implementation of perspecting .
finally , i showed how perspecting is how the lmb relational approach makes contact with the rovelli
crane ( rc ) type of relational approach .
moreover , there is tension between some parts of lmb and rc relationalism , since lmb takes mach s ` time is to be abstracted from change ' to involve _ all _ change , whilst rovelli s approach takes it to involve _ any _ change ( which i term amr time : aristotle mach rovelli ) .
i argue then that it is better to hold to an intermediate interpretation : a _ sufficient totality of locally relevant change _ , as judged at the level of which relative forces significantly affect the equations of motion , which i term lmb - a time , is new to this article and is further elaborated upon in @xcite ( see sec 22.1 for an outline ) .
this covers further detail of this approach , and how each of these three approaches leads to distinct attitudes to the theory of time - keeping and to distinct timeless approaches to the problem of time in quantum gestalt .
[ [ application-1 - 3-interpretations - of - machs - time - is - to - be - abstracted - from - change - principle ] ] application 1 ) 3 interpretations of mach s ` time is to be abstracted from change ' principle ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ each of these three has a supporting argument .
rovelli s interpretation applies generically ( regardless of physical regime or material composition of the system in question ) . on the other hand ,
lmb s interpretation , by using the totality of the change , produces an _ incontestable _ time , in that there is no more change elsewhere that can run in suspiciously irregular ways that cause of one to doubt one s timestandard .
however , there is plenty of middle ground between these two positions .
note 1 ) this holds without having to go to the whole - universe extreme , which is impractical , and operationally undesirable we know little of the motions and constitutions of very distant bodies .
also , since leibniz s day , the notion of ` universe ' has gotten a whole lot more complicated with gr were this taken seriously , does it refer to the hubble radius or to the actual entirely of the universe including unobservable portions ? n.b . that this leibniz
gr discrepancy in the nature of the universe becomes a red herring in the below discussion .
note 2 ) this also raises a partial objection to the rovelli scheme , in that whilst rovelli does not state he is looking for ` good clocks ' , those are a fair notion to ask about , and yet clash conceptually with rovelli - type notions that any time , or any clock , will do . on the other hand ,
barbour s approach is very conducive toward asking and conceptually answering this question . in comparison to amr
, it has the extra element of acknowledging some times are locally more useful to consider than others . even in a generic situation
, one can have a local ranking procedure for candidate times , alongside a refining procedure until acceptable accuracy of predictability is obtained . in comparison with lmb time ,
the practically - used ephemeris - type method for the solar system is entirely adequate without having to consider the net physical effect ( tidal effect ) of distant massive bodies such as andromeda ( thus also rendering irrelevant our increasing lack of precise knowledge of the masses and positions as one considers more and more distant objects ) .
* lmb - a timekeeping * : is like astronomers ephemeris actually was , or atomic clock timekeeping from the perspective including recalibration by comparison with solar system dynamics .
this more faithful representation of the astronomers actual procedure ( such as ephemeris time or atomic time recalibrated in accord with solar - system physics ) is why i take inquiring in detail into what is meant by ` universe ' in the lmb approach to be a red herring .
relationalism 7 m ) is to be implemented by a * generalized local ephemeris procedure*. i.e. , do not just use a change to abstract a time , but also check whether using this time in the equations of motion for other changes suffices to predict these to one s desired accuracy
. further detail of this approach , and of how each of the above three approaches leads to distinct attitudes to the theory of time - keeping and to distinct timeless approaches to the problem of time in qg with distinct notions of emergent time is the subject of @xcite . n.b .
that the relationalism 7 versus relationalism 7w clash means that one can not just add the rc postulates to the lmb ones ; relationalism 7 m then represents a compromise between these [ and is also the most accurate encapsulation of astronomical ( based ) timestandards ] .
application 2 ) is the wider relational assessment of the problem of time s facets and strategies .
as regards frozen formalism problem resolving strategies , leibniz - time , mach - time and tangibility arguments hold against most tempus ante quantum strategies ( internal time , matter time ... ) . on the other hand , emergent jbb time and the approximately equivalent emergent semiclassical time comply with these criteria .
timeless and histories theory approaches generally do so too .
many other problem of time facets have also been exposited as relational in this article . in particular , i have argued that the thin sandwich problem facet is generalized by , and relationally interpreted as , the best matching problem , with other lmb - relational ties being established to the foliation dependence problem , functional evolution problem , problem of beables and spacetime reconstruction problem .
on the other hand , the global and multiple choice problems of time have been found to have at least some ties to rovelli s approach to relationalism ( which is also an alternative stance on the problem of beables and the most sympathetic of the relational positions considered in this article as regards internal time and matter time approaches ) .
application 1 ) s finer distinctions between the lmb , lmb - a , and amr attitudes to mach s time principle serve as further distinction between various timeless approaches and semiclassical - histories - timeless approach combinations , though this is largely left to a subsequent article @xcite .
application 6 ) below also concerns a way of enhancing timeless records theory / composites of at least histories theory and records theory .
finally , this article s more minimalist option of not considering ( all ) canonical transformations at least partly selects against internal time , histories brackets , ashtekar variables and linking theory based approaches .
application 3 ) concerns the status of scale in physical theories .
barbour argues against it , in good part due to its heterogeneity as regards all the other degrees of freedom ( the shapes ) .
however , i argue that this heterogeneity can be used to obtain an approximate timestandard based on the scale or some closely - related variable . whilst this in itself does not comply with
my or barbour s interpretation of the mach time principle , allowing the shape degrees of freedom to contribute perturbatively to the more accurate form of resulting timestandard does comply thus .
moreover , this looks to be in particular an accurate rendition of classical and quantum cosmology ( due to the good fortune of anisotropies and inhomogeneities being much smaller than scale effects in that setting ) . application 4 ) concerns further foundations for ashtekar variables / lqg .
ashtekar variables / lqg are often stated to be relational ; more presicsely , that means rovelli - relational . in the present article , however , i laid out how ashtekar variables are also lmb(-a ) relational .
this is of foundational importance in widening the use of ` relational ' to describe , motivate , structure and further develop the ashtekar variables / lqg program .
i pointed out how rpm s and geometrodynamics form a natural sequence of theories of geometrical configurations , whilst regarding the astekar bein variables as a momentum is then somewhat incongruous . on the other hand , electromagnetism , yang
mills theory and ashtekar variables form a _ rival _ natural sequence .
i also pointed out that whilst lqg is a relational approach in all of this article s senses , it does not presently bear a relational name , ( unlike geometrodynamics or shape dynamics , which are named for their configurationally relational contents ) .
i suggested knot quantum gestalt , quantum nododynamics or quantum spin - knot dynamics ( knots being configurationally relational whilst loops include a diff(3 ) s worth of physical redundancy ) .
i emphasize that by the comments i provide in theis article on ashtekar variables / lqg , supersymmetry , string theory and m - theory , i demonstrate contact between lmb relationalism and much of modern theoretical physics .
application 6 ) concerns bringing together i d and timeless approaches ( some of which also involve elements of rc relationalism ) .
all of these areas of study principally concern subsystems . in the case of i d , this is via at least the great majority of qm propositions that can be asked are about subsystems .
i d has the broadest mathematization ( topos theory ) , whilst crane uses categories .
records theory does not conventionally make use of these structures , and yet i d s development supports histories theory ( which contains , and shared many similarities with , records theory ) and sheaves @xcite as a means of locally attaching information may well provide an insightful new formulation for records theory .
there is however a chasm to bridge . in records
theory , records are local in a _ metric space _ sense , whilst topoi and sheaves have a structurally weaker _ topological space _ notion of locality .
thus one would need to extend / specialize ( provide an extra layer of structure ) to i d s work or consider the extent to which one can conceptualize with records using solely a topological notion of locality .
* acknowledgements * : i thank : my wife claire , amelia , sophie , sophie , anya , bryony , amy , hettie , hannah , duke , becky and one other for being supportive of me whilst this work was done .
professors julian barbour , enrique alvarez , louis crane , jonathan halliwell , chris isham , karel kucha , marc lachize - rey and jeremy butterfield , dr s cecilia flori , sean gryb , tim koslowsk and ioannis alevizos , and mr eduardo serna for discussions between 2008 and the present . professors
belen gavela , marc lachize - rey , malcolm maccallum , don page , reza tavakol and jeremy butterfield for support with my career .
i was funded by a grant from the foundational questions institute ( fqxi ) fund , a donor - advised fund of the silicon valley community foundation on the basis of proposal fqxi - rfp3 - 1101 to the fqxi .
i thank also theiss research and the cnrs for administering this grant .
this work was started at damtp , centre for mathematical sciences , cambridge , and continued at fsica terica , universidad autonoma de madrid .
e. mach , _ die mechanik in ihrer entwickelung , historisch - kritisch dargestellt _
barth , leipzig 1883 ) .
an english translation is _ the science of mechanics : a critical and historical account of its development _ ( open court , la salle , ill .
1960 ) . c. rovelli , p. 292
in _ conceptual problems of quantum gravity _ ed .
a. ashtekar and j. stachel ( birkhuser , boston 1991 ) ; l. smolin , in _ conceptual problems of quantum gravity _ ed .
a. ashtekar and j. stachel ( birkhuser , boston , 1991 ) .
o. pooley , http://philsci-archive.pitt.edu/archive/00000221/index.html ; o. pooley and h.r .
brown , brit .
sci . * 53 * 183 ( 2002 ) ; o. pooley , chronos ( proceedings of the philosophy of time society 2003 - 4 ) .
n. bodendorfer , a. thurn and a. stottmeister , arxiv:1203.6525 ; arxiv:1203.6526 . .
o. pooley , to appear in r. batterman ( ed . ) , _ the oxford handbook of philosophy of physics _ , http://philsci-archive.pitt.edu/9055/1 .
i. newton , _ philosophiae naturalis principia mathematica _ ( 1686 ) . for an english translation , see e.g i.b .
cohen and a. whitman ( university of california press , berkeley , 1999 ) .
in particular , see the scholium on absolute motion therein .
halliwell and j. thorwart , phys . rev .
* d65 * 104009 ( 2002 ) , gr - qc/0201070 ; j.j .
halliwell , in _ the future of theoretical physics and cosmology_. ( stephen hawking 60th birthday festschrift volume ) ed .
gibbons , e.p.s .
shellard and s.j .
rankin ( cambridge university press , cambridge 2003 ) , arxiv : gr - qc/0208018 . | this article contributes to the debate of the meaning of relationalism and background independence , which has remained of interest in theoretical physics from newton versus leibniz through to foundational issues for today s leading candidate theories of quantum gravity .
i contrast and compose the substantially different leibniz
mach barbour ( lmb ) and rovelli crane ( rc ) uses of the word ` relational ' .
leibniz advocated primary timelessness and mach that ` time is to be abstracted from change ' .
i consider 3 distinct viewpoints on machian time : barbour s , rovelli s and my own .
i provide four expansions on barbour s taking configuration space to be primary : to ( perhaps a weakened notion of ) phase space , categorizing , perspecting and propositioning .
categorizing means considering not only object spaces but also the corresponding morphisms and then functors between such pairs .
perspecting means considering the set of subsystem perspectives ; this is an arena in which the lmb and rovelli approaches make contact . by propositioning , i mean considering the set of propositions about a physical ( sub)system .
i argue against categorization being more than a formal pre - requisite for quantization in general ; however , perspecting is a categorical operation , and propositioning leads one to considering topoi , with isham and doering s work represents one possibility for a mathematically sharp implementation of propositioning .
further applications of this article are arguing for ashtekar variables as being relational in lmb as well as just the usually - ascribed rc sense , relationalism versus supersymmetry , string theory and m - theory .
the question of whether scale is relational is also considered , with quantum cosmology in mind .
* edward anderson@xmath0 * pacs : 04.20.cv and 04.60.-m @xmath0 [email protected] |
this paper is part of a long term program to cauchy - characteristic matching ( ccm ) codes as investigative tools in numerical tools in numerical relativity . for a review of ccm see the article of winicour@xcite .
the approach has two distinct features : ( i ) it dispenses with an outer boundary condition and replaces this with matching conditions at an interface between the cauchy and characteristic regions , and ( ii ) by employing a compactified coordinate , it proves possible to generate global solutions .
this paper is continuation of work investigating ccm in systems with vacuum cylindrical symmetry ( as a prototype system containing only one spatial degree of freedom ) . in @xcite
we developed the necessary machinery and in @xcite we applied the approach to the exact solution of weber and wheeler @xcite which has one degree of polarization and so only possesses one gravitational degree of freedom .
an agreement with the exact solution was found with a maximum error better then 1 part in @xmath1 . in the same paper we also investigated the propagation of gaussian wave packets possessing two gravitational degrees of freedom , but this work did not involve a check against any exact solution .
such a test is likely to be a more rigorous one because , unlike the weber - wheeler case , it would involve passing derivative information across the interface . one of the major problems in numerical relativity is that there are very few exact solutions known which can be used to test numerical codes . in this paper ccm is applied to an exact two - parameter family of cylindrically symmetric vacuum solutions possessing both gravitational degrees of freedom due to piran , safier and katz @xcite .
this family is somewhat unphysical because the rotational degree of freedom diverges at future null infinity .
nonetheless , it can be used to test the ccm cylindrical code more rigorously because it involves passing derivative information across the interface .
the previous ccm cylindrical code was constructed specifically for solutions which ( after using geroch decomposition to factor out the @xmath0-direction ) are asymptotically flat and therefore can not be used for processing the piran et al family . however , by working directly in terms of the geroch potential @xcite it proves possible to develop a modified version of the code which is able to process the family . in section [ sec : two ] we present the family of solutions and briefly discuss its properties . in section [ sec : char ] we review the derivation of the field equations in the characteristic region for solutions which , after factoring out the @xmath0-direction , are asymptotically flat . in section [ sec : asymp ] we discuss the asymptotic limit of the piran et al family and derive the modified field equations and the interface equations .
an appendix discusses the regular singular nature of two of the resulting equations . in section [ sec : results ] we briefly discuss the results of the modified code .
piran et al derive their metric by an indirect method starting from the kerr line element in boyer - lindquist coordinates @xmath2 , namely @xmath3 ^ 2 + { { \rho^2}\over{\delta}}d\tilde{r}^2 + \rho^2d\tilde{\theta}^2,\ ] ] where @xmath4 they first transform to a new isotropic radial coordinate @xmath5 where @xmath6 and then to cylindrical coordinates @xmath7 where @xmath8 employing the complex trick of sending @xmath9
they then introduce new cylindrical coordinates @xmath10 given by @xmath11,\\ r&=&{{\tilde{\rho}}\over{\alpha}}\biggl[1 -{{m^2+a^2}\over{4(\tilde{\rho}^2-\tilde{t}^2)}}\biggr],\\ z&=&z-2a^{-1}m(m+\sqrt{m^2+a^2})\tilde{\phi},\\ \phi&=&\alpha\tilde{\phi},\end{aligned}\ ] ] where @xmath12 and the resulting line element is in the jordan - ehlers - kompaneets form for a cylindrically symmetric spacetime @xcite , @xmath13 if we finally introduce coordinates @xmath14 then we can express the two - parameter family of solutions in the form @xmath15 where explicitly @xmath16 } } , \label{omegaval}\end{aligned}\ ] ] with @xmath17 the coordinates @xmath18 and @xmath19 are null , @xmath20 is the canonical azimuthal coordinate and @xmath0 lies along the axis of symmetry .
the two parameters are @xmath21 @xmath22 which is a length scale and @xmath23 @xmath24 which is a measure of the total energy of the system , with @xmath25 corresponding to flat space @xcite .
the solution is regular everywhere in the coordinate range @xmath26 there is no conical singularity on the axis of symmetry and the solution reduces to minkowski spacetime at past and future infinity .
however , the solution is conical at spatial infinity and singular at both past and future null infinity .
more precisely , at future null infinity ( cf eq .
( 10 ) in @xcite ) @xmath27 as @xmath28 , with analogous behaviour at past null infinity on interchanging @xmath18 and @xmath19 .
( note that eq .
( 10 ) in @xcite includes a typographical error and that @xmath18 and @xmath19 should be transposed ) .
since piran et al only derived this family of solutions by the indirect method described above , we tried to confirm that the family is vacuum by a direct computation of the ricci tensor using both of the computer algebra systems sheep @xcite and grtensor @xcite .
unfortunately , neither system was able to complete the calculation because the metric involves nested radicals and algebra systems are notorious for the difficulties such quantities present .
however , we were able to confirm the indirect derivation described above .
the standard treatment of the jordan - ehlers - kompaneets cylindrically symmetric line element ( [ jek ] ) in the cauchy region @xmath29 is described in @xcite . in particular
, it is shown that the independent set of dynamical equations for the variables @xmath30 , @xmath31 , @xmath32 and @xmath33 are @xmath34 , \label{ltildedot}\end{aligned}\ ] ] where @xmath32 and @xmath33 are defined in terms of the mixed components of the extrinsic curvature @xmath35 by @xmath36 these equations are augmented by the constraint equation @xmath37 , \label{haml}\ ] ] which serves to determine @xmath38 once the main variables are known . in the rest of this section
we review the treatment in the characteristic region since this will need modification to cope with the piran et al family of solutions .
we introduce the compactified coordinate @xmath39 in which case the line element becomes @xmath40 where the metric functions @xmath30 , @xmath31 and @xmath38 are now to be regarded as functions of @xmath18 and @xmath41 .
the characteristic region then consists of @xmath42 where the interface is taken to be at @xmath43 and future null infinity is given by @xmath44 . instead of working directly with the metric functions @xmath30 and @xmath31
, we use the related quantities @xmath45 and @xmath46 which are defined by @xmath47 where @xmath48 @xmath49 is the geroch potential @xcite , and the integration is along any path connecting a fixed point @xmath50 on the interface ( we choose @xmath51 ) to a general point @xmath52 in the characteristic region ( see fig . [ fig1 ] ) . with these definitions the vacuum equations can be written in the succinct form @xmath53 where @xmath54 and @xmath55 are explicit functions of their arguments @xcite .
( [ e1 ] ) and ( [ e2 ] ) serve to define @xmath56 and @xmath57 and ( [ e3 ] ) and ( [ e4 ] ) provide coupled propagation equations for @xmath56 and @xmath57 along the null rays ruling the hypersurfaces @xmath58 .
the initial data consists of specifying @xmath45 and @xmath46 on the initial hypersurface @xmath59 and the interface @xmath60 , together with @xmath38 at their intersection @xmath51 .
the essence of the iterative integration scheme is to use ( [ e3 ] ) and ( [ e4 ] ) to determine @xmath56 and @xmath57 on the interior null hypersurface @xmath61 and then ( [ e1 ] ) and ( [ e2 ] ) to determine @xmath62 and @xmath63 on @xmath61 , which in turn determines @xmath45 and @xmath46 on the next neighbouring null slice .
finally @xmath38 is determined from the constraint equation ( eq . ( 42 ) in @xcite ) which has the form @xmath64 where @xmath65 is an explicit function of its arguments .
the ccm method requires the exchange of values of the metric functions and their derivatives at the interface at each iteration , the details of which are given in @xcite .
the approach described in the last section breaks down for the piran et al solution in the characteristic region because @xmath66 as @xmath67 .
let us consider the behaviour of the various functions on the hypersurface @xmath61 in the asymptotic limit @xmath68 .
we have , @xmath69 where we set @xmath70 thus , asymptotically , @xmath71 and @xmath72 where @xmath73 it is clear that @xmath74 , @xmath75 , @xmath31 and @xmath46 are all divergent as @xmath76 . however , the ancillary quantities @xmath77 and @xmath78 as well as @xmath45 and @xmath49 , the geroch potential , are all regular .
this suggests rewriting the system ( [ e1])([e5 ] ) in terms of these last two variables .
we find explicitly that @xmath79\label{e8},\\ o_{,y}&=&{1\over y}o+{{yo_{,y}}\over{(1+my)}}m -{{y^2}\over{4(1+my)}}(yo_{,yy}+o_{,y } ) \nonumber\\ \phantom{m_y}&\phantom{=}&\phantom{aaaaaaaaaaaaaaa } + { { y^3}\over{2(1+my)^2}}(mo_{,y}+ym_{,y}o_{,y } ) , \phantom{aaaa } \label{e9}\\ \gamma_{,y}&=&-{y\over{8(1+my)^2}}\bigl((m+ym_{,y})^2+o_{,y}^2\bigr ) .
\label{e10}\end{aligned}\ ] ] although the equations ( [ e8 ] ) and ( [ e9 ] ) are coupled singular equations for @xmath56 and @xmath80 , they are regular singular equations and the solutions remain regular as @xmath76 ( see the appendix ) .
in fact , it is clear from the defining equations ( [ e6 ] ) and ( [ e7 ] ) , together with ( [ em ] ) and ( [ eo ] ) , that both @xmath56 and @xmath80 are of order unity in the limit @xmath76 .
these are the modified equations on which the new code is based for investigating the piran et al solutions .
we need to augment the equations with the interface equations which relate the two sets of variables in the two coordinate systems on the interface .
the relationships required for _ injection _ , i.e. going from the characteristic to the cauchy region are , @xmath81 ^ 2 \nonumber\\ \phantom{m_y}&\phantom{=}&\phantom{aaaaaaaaaaaaaaaaa } + { 1\over 8}{{y^5}\over{\lambda } } \left\{2(ym)_{,y}+ [ y(ym)_{,y}]_{,y}\right\ } , \\
\tilde{l}&=&{1\over 2}{m\over y } , \\
\omega_{,r}&=&{o\over{y^2\lambda } } , \\
\omega_{,rr}&=&-{{o_{,u}}\over{y^2\lambda}}+{{mo}\over{y\lambda } } + { { o}\over{\lambda } } + { 1\over 2}{{yo(ym)_{,y}}\over { \lambda^2 } } -{1\over 2}{{yo_{,y}}\over \lambda } , \\
l^{\phi}_{z}&=&{1\over 2}{{\lambda o}\over{y^2}}+{1\over 4}yo_{,y}.\end{aligned}\ ] ] similarly , the relationships required for _ extraction _ , i.e. going from the cauchy to the characteristic region are , @xmath82 where the integration in ( [ pot ] ) is along the interface from the point @xmath83 to the point @xmath84 as shown in fig .
in order to test the ccm code against the piran et al solution we will first compare the metric quantity @xmath30 in both the cauchy and characteristic regions .
( [ psi ] ) provides the exact solution in both cases when @xmath85 and @xmath74 , eqs .
( [ lamu ] ) and ( [ lamv ] ) , are written in terms of @xmath86 coordinates and @xmath87 coordinates for the cauchy and characteristic regions respectively .
it is not possible to use @xmath31 for a similar comparison since from ( [ omegdiv ] ) we have seen that it diverges as @xmath76 .
we use , instead , the geroch potential , @xmath49 , which is computed directly by the code in the characteristic region , but must be constructed in the cauchy region using its definition ( cf eq .
( 24 ) of @xcite ) , @xmath88 it is straightforward to evaluate the integrals in the cauchy region using finite - difference representations of @xmath89 and @xmath90 . for the exact solution @xmath89 and @xmath90
can be evaluated by differentiation , however the integral must be evaluated numerically to find a semi - analytic @xmath49 .
a similar approach is required for the ` exact ' value of @xmath49 in the characteristic region using eq .
( [ omchardef ] ) .
as mentioned previously , the two parameters @xmath21 and @xmath23 of the piran et al solution represent the length scale and strength of the gravitational wave respectively . for @xmath91 the wave moves in from infinity , reaches its highest concentration at @xmath92 , where it rebounds from the @xmath93 axis , and is outgoing for @xmath94 .
ccm allows us to evolve initial data containing an ingoing gravitational wave , something which is difficult to do for a cauchy only evolution .
since we have set the interface at @xmath95 the value of @xmath21 can be chosen such that a substantial fraction of the the wave will move onto the cauchy region and will be well resolved with the numbers of grid points we choose .
changing the interface position while keeping the ` interface distance to @xmath21 ' ratio the same results in an identical evolution on the cauchy portion of the grid .
the primary variable we used for comparison are @xmath30 and @xmath49 ( since @xmath38 is a derived quantity ) .
we considered grid resolution numbers of @xmath96 and @xmath97 , where the error is defined as @xmath98 where @xmath99 is the exact value , @xmath100 is the code computed value and @xmath101 denotes the @xmath102 norm .
we ran the modified code for a whole range of values in the @xmath103 parameter space and found an error in @xmath30 of no more than 0.01% and an error in @xmath49 of less than 0.2% .
however , in this version of the code , although the convergence rate of the solution starts as second - order it deteriorates to first order after long time evolution . in some recent work ,
colleagues in the southampton numerical relativity group have modified the code so that it is now fully second order and very long time stable @xcite .
the improvement was achieved by dispensing with variables which are related by exponential or logarithmic functions at the interface , using the geroch potential both in the cauchy and characteristic regimes and also using an implicit method at the interface .
this version of the code is currently being applied successfully to modelling dynamic cosmic strings .
surface plots are shown for evolutions of the metric quantity @xmath104 in fig .
[ fig2a ] , the geroch potential @xmath49 in fig .
[ fig2b ] , and the radial derivative @xmath105 ( which indicates the distribution of energy within the wave)in fig .
[ fig2c ] , where the parameters are @xmath106 and @xmath107 .
the surface plots use a radial coordinate @xmath0 defined by , @xmath108 so that there is a change in the coordinate system at @xmath109 .
note that the value of @xmath104 remains equal to 1 at null infinity , but is very close to zero at the axis @xmath93 .
the incoming wave hits the axis at @xmath92 and rebounds .
this can be seen very clearly in the surface plot of @xmath105 .
the large peak of fig .
2c grows without bound as @xmath23 is increased . for comparison purposes figures [ fig3a ] ,
[ fig3b ] and [ fig3c ] shows the same quantities , but for parameter values @xmath106 and @xmath110 , which represents an almost flat spacetime .
again the wave reaches its maximum concentration on the @xmath93 axis at @xmath92 , however the peak is much less dominant in this case .
let @xmath111 which by ( [ em ] ) and ( [ eo ] ) is regular as @xmath76 .
the homogeneous part of the modified propagation equations ( [ e8 ] ) and ( [ e9 ] ) is @xmath112 differentiating ( [ a2 ] ) with respect to @xmath41 and using ( [ a2 ] ) to eliminate @xmath56 and ( [ a1 ] ) to eliminate @xmath113 we get @xmath114 which has dominant singular part @xmath115 substituting in the trial solution @xmath116 we obtain the auxiliary equation @xmath117 which has roots @xmath118 and so gives rise to the regular independent solutions @xmath41 and @xmath119 .
99 p. musgrove , d. pollney , and k. lake , _ grtensor ii : a computer algebra system for general relativity _ , ( queen s university , kingston , ontario ) , web page : http://www.astro.queensu.ca/@xmath120grtensor/grhome.html . | this paper is part of a long term program to cauchy - characteristic matching ( ccm ) codes as investigative tools in numerical relativity .
the approach has two distinct features : ( i ) it dispenses with an outer boundary condition and replaces this with matching conditions at an interface between the cauchy and characteristic regions , and ( ii ) by employing a compactified coordinate , it proves possible to generate global solutions . in this paper ccm
is applied to an exact two - parameter family of cylindrically symmetric vacuum solutions possessing both gravitational degrees of freedom due to piran , safier and katz .
this requires a modification of the previously constructed ccm cylindrical code because , even after using geroch decomposition to factor out the @xmath0-direction , the family is not asymptotically flat .
the key equations in the characteristic regime turn out to be regular singular in nature . |
everybody knows @xcite that a black hole the final state of gravitational collapse @xcite emits a stationary flux of massless particles @xcite . from the perspective of _ asymptotic
_ inertial observers , this hawking flux is indistinguishable from the thermal radiation emitted by a black body at temperature @xmath1 ( in planck units ) , where @xmath2 is the surface gravity of the hole . for macroscopic black holes ( with @xmath3 )
, this value is exceedingly small indeed much smaller than the cmb temperature ; it is therefore unlikely that hawking radiation will ever be measured directly ( at least in the astrophysical context . ) but what if one could get very close to a hole ?
what would hawking radiation look like _ there _ ?
unruh asked this question in his celebrated `` notes on black hole evaporation '' @xcite , where he showed that the correct answer depends on the observer s state of motion : the temperature measured by a particle detector close to the horizon is @xmath4 , where @xmath5 is the detector s acceleration .
consequently , according to unruh , `` a geodesic detector near the horizon will not see the hawking flux of particles '' @xcite . because the near - horizon of a sufficiently massive black hole is almost flat , and the quantum vacuum is as regular there as anywhere else , it is indeed natural to assume that `` nothing special '' happens to geodesic observers at the horizon . yet
, this consensual view has been recently challenged on various grounds . for helfer @xcite ,
taking interactions into account `` completely alters the picture '' drawn by hawking and unruh : ultra - high - energy vacuum fluctuations must couple to quantum gravity , and thus quantum field theory must break down in the vicinity of a black hole . for almheiri _
_ @xcite and braunstein _ et al . _
@xcite , the devil is in the information conservation principle : for hawking radiation to be in a pure state , something `` dramatic '' must happen at the horizon .
pictorially , if alice is entangled with bob but being somewhat braver than him decides to cross the horizon and plunge into the hole , then quantum mechanics will `` burn her up '' .
these new arguments are both exciting and challenging , but let s face it : they are mere speculations . to this day , neither the physics of trans - planckian fluctuations nor the fate of unitarity in quantum gravity are properly understood ; drawing dramatic consequences from them is bold , but risky . what _ is _ well understood , on the other hand , is quantum field theory in curved spacetimes the very setup used by hawking to make his prediction . what does the old semi - classical framework have to say about alice falling into a black hole ?
( 0,-2.958047)(5.50291,2.958047 ) ( 0.6210156,2.5596485)(0.6210156,-2.4403515 ) ( 0.6210156,-2.4403515)(3.8210156,0.75964844 ) ( 3.8210156,0.75964844)(2.2210157,2.5596485 ) ( 0.6210156,2.5596485)(2.2210157,2.5596485 ) ( 1.8210156,1.1596484 ) ( 1.8210156,1.1596484 ) ( 1.8210156,1.1596484)(3.0210156,-0.040351562 ) ( 1.8210156,1.1596484)(0.6210156,-0.040351562 ) ( 0.6210156,-0.040351562)(1.8210156,-1.2403516 ) ( 1.9924707,1.4246484)@xmath6 ( 3.6724708,-0.115351565)@xmath7 ( 2.3524706,-1.4953516)@xmath8 ( 2.2210157,2.5596485)(0.64101565,1.0396484 ) ( 0.7824707,-2.7353516)@xmath9 ( 2.4824708,2.7646484)@xmath10 ( 4.1724706,0.76464844)@xmath11 ( 2.9824708,-0.79535156)@xmath12 ( 3.5824707,2.0046484)@xmath13 ( 3.9610157,-1.2203516)(5.0010157,-0.14035156 ) ( 5.1924706,-0.39535156)@xmath14 ( 1.2524707,2.1646485)@xmath15 consider the simplest possible model of gravitational collapse an ingoing null shell with adm mass @xmath16 coupled to a free , massless scalar field in @xmath17 dimensions and the simplest possible model of alice an unruh - dewitt @xcite detector with energy gap ( frequency ) @xmath18 moving along a radial geodesic @xmath19 with conserved energy ( per unit rest mass ) @xmath0 . in this setup , the answer to `` what does alice see at time @xmath20 ? '' is coded in the response function [ response ] ( , ) = 2_-^du_(u)_0^ds _ ( u - s)e^-isg((u),(u - s ) ) . here
@xmath21 is the two - point function of the field and @xmath22 a smooth , non - negative `` window function '' , such that @xmath23 and @xmath24 for @xmath25 . to compute for a given radial geodesic @xmath19 , only one ingredient from quantum field theory
is needed : the two - point function @xmath21 of the field in the in - vacuum .
the in - vacuum state is best described in terms of what may be called `` eikonal coordinates '' @xmath26 .
these are globally defined , null coordinates , defined for a given spacetime point @xmath6 by tracing back to past null infinity @xmath12 the two radial null rays which meet at @xmath6 ; the coordinates @xmath27 are the eddington - finkelstein coordinates @xmath28 of these rays on @xmath12 , see fig .
[ penrosediag ] . in these coordinates ,
the two - point function @xmath21 in the in - vacuum takes the simple form @xcite @xmath29 notice that , by virtue of the additivity of the @xmath30 , in @xmath17 dimensions alice s detector couples independently to @xmath31 ( the _ incoming _ vacuum modes ) and @xmath32 ( the _ outgoing _ vacuum modes ) .
this property does not hold in higher dimensions , or with massive fields , but for the argument made in this essay , it is key .
let us now ask : when will either of the two contributions @xmath33 to the response function corresponding to the incoming ( @xmath34 ) and outgoing ( @xmath35 ) modes respectively be thermal ? there is a simple criterion for this @xcite : it suffices that , along the trajectory @xmath19 , the quantity ( dot denotes @xmath36 ) [ quasi ] t_=|| is approximately constant , in which case @xmath33 is a thermal spectrum at temperature @xmath37 . here , by `` approximately constant '' i mean @xmath38 .
when this condition does not hold , a weaker form `` thermality '' of @xmath33 still holds in the ultraviolet limit , namely @xmath39 for @xmath40 . in the case of asymptotic inertial trajectories with @xmath41 , as in hawking s original work , one checks that @xmath42 and @xmath43 .
( this is consistent the lore that hawking radiation consists of those modes which have bounced off the center of the shell and escaped from it just before it forms a horizon at @xmath44 ) .
more generally , if @xmath45 , viz .
for inertial observers moving towards the hole at velocity @xmath46 , one computes @xmath47 and t^_-(e)=t_h(e+ ) .
the @xmath0-dependent correction is the doppler factor for an observer moving relative to the a static source located on the center of the black hole . ,
both in the asymptotic limit ( `` asymp '' ) and at the horizon ( `` hor'').,title="fig : " ] , both in the asymptotic limit ( `` asymp '' ) and at the horizon ( `` hor'').,title="fig : " ] but what are the values of when the radial geodesic @xmath19 crosses the horizon that is , when alice jumps into the hole ?
an explicit computation detailed in @xcite shows that , when she crosses the schwarzschild horizon , t^_-(e)=4et_ht^_+(e)=. thus , if alice enters the hole with sufficiently low radial velocity , in the sense that @xmath48 , ( with @xmath49 corresponding to the limit case where she is dropped from the horizon with zero initial velocity ) , she perceives the _ incoming _ modes at _ high temperature _ ! this divergence is confirmed by the computation of the ( outgoing ) _ flux _
@xmath50 , where @xmath51 is the @xmath52-velocity of alice and @xmath53 is a unit vector normal to @xmath51 : at horizon - crossing , this flux reads @xcite ^(e)=t_h^2(2e^2 + ) .
thus , the intensity of hawking radiation perceived by freely - falling observers near the horizon can in fact be arbitrarily high .
observer moreover that the velocity - dependence of @xmath54 and @xmath55 is _ not _ of doppler - type at least not with respect to a single source inside the black hole .
what is going on ? in short the hawking effect , in reverse .
the mechanism for the conversion of vacuum fluctuation into thermal radiation in the standard hawking effect ( observed from infinity ) is well known : it is the _ exponential redshift _ experienced by outgoing modes on their way from the to -
be horizon to infinity .
this is evident from our criterion : if @xmath56 is constant , then by integration @xmath57 is an exponential function of @xmath20 .
now , for a near - horizon observer , it is clear that outgoing modes have had no chance to experience such exponential redshift .
indeed , this is precisely what unruh understood in @xcite , and what led him to conclude that a near - horizon observer will not see the hawking flux of radiation .
but one must not forget that the field also contains incoming modes !
these modes , coming from infinity all the way down to the horizon , _ do experience a very large frequency shift_an exponential blueshift .
this is why @xmath58 is not zero on the horizon , and this is also why @xmath59 have a non - dopplerian dependence on alice s velocity with respect to the hole : in the @xmath49 limit , the `` source '' is infinity not the hole .
limit contributed _ positively _ to the outgoing flux @xmath60 . ] already at the purely classical level , it is well - known that in her free - fall approach to the black hole horizon , alice will feel the sky `` falling down '' on her head : when looking in the direction transverse to her radial motion , she will see an image of the entire sky compressed in a thin layer just above horizon .
the near - horizon hawking effect discussed in this essay is reminiscent of this ( apparent ) inversion effect : seen from a radially infalling geodesic , asymptotic infinity and the horizon itself appear _ dual _ to one another , with the outgoing field modes mapped to the incomes modes and the large @xmath0 regime to the small @xmath0 regime .
explicitly , t_+^(e/2)~t_-^(1/e)e0 .
this identity is , i believe , a rather remarkable feature of hawking radiation .
one may object that this `` horizon - infinity duality '' is imperfect .
first , @xmath61 , while @xmath42 ; second , the thermality of incoming modes is not perfect ( because @xmath62 , but rather @xmath63 , hence thermality holds for high - frequency detectors only ) ; and third , because the @xmath0-dependence of @xmath64 and of @xmath65 holds only for @xmath48 and @xmath66 .
one reason for this asymmetry is clear : the local spacetime geometry at infinity and on the horizon are not the same the former is flat , the latter is not .
suppose however that instead of the schwarzschild spacetime , we considered an artificial black hole geometry in which both infinity _ and _ the horizon are flat ( it is easy to design such a metric and repeat all computations using it instead of schwarzschild@xcite )
. then the local geometries at infinity and on the horizon would be the same , and computing @xmath37 in both limits would lift the first caveat .
i do not know whether the other two caveats are of importance .
after many years of work on the hawking phenomenon , the nature of hawking radiation at the horizon the fact that it appears to be emitted from _
outside _ the hole , and can have _ high _ temperature still surprises us today .
not only is it not always true that `` nothing special '' happens on the horizon , but one needs not go beyond the semiclassical approximation to find some drama near the horizon .
i have attempted to account for this effect by what i called the `` horizon - infinity duality '' : the idea that the hawking effect at the horizon is a reversed image of the standard hawking effect , where the role of ingoing and outgoing modes , and large - energy and small - energy geodesics , are reversed .
the emerging picture is intriguing : for alice dropped just above the horizon , it s not the black hole that looks like a blackbody
it s the sky .
i thank thanu padmanabhan for drawing my attention to the fate of infalling detectors with regards to hawking radiation , suprit singh for the collaboration leading to @xcite , and carlo rovelli and antonin coutant for several useful discussions on black hole physics .
i also note that the particular result @xmath67 was obtained previously in @xcite .
b. s. dewitt , `` general relativity : an einstein centenary survey , '' in _ general relativity : an einstein centenary survey _ , s. w. hawking and w. israel , eds . , pp .
680745 . cambridge university press , 1979 . | what happens when alice falls into a black hole ? in spite of recent challenges by almheiri _
et al._the `` ' '
firewall hypothesis , the consensus on this question tends to remain `` nothing special '' .
here i argue that something rather special can happen near the horizon , already at the semiclassical level : besides the standard hawking outgoing modes , alice can records a quasi - thermal spectrum of _ ingoing _ modes , whose temperature and intensity diverges as alice s killing energy @xmath0 goes to zero .
i suggest that this effect can be thought of in terms a _ horizon - infinity duality _ , which relates the perception of near - horizon and asymptotic geodesic observers the two faces of hawking radiation . |
in this paper we study phase transitions in heisenberg magnets on the pyrochlore lattice , which consists of corner sharing tetrahedra@xcite .
the motivation is partly theoretical , to understand the behavior of highly frustrated systems when there is a degeneracy , or near degeneracy , between different ordered states .
there is also experimental motivation since experiments on gd@xmath0sn@xmath0o@xmath1 and gd@xmath0ti@xmath0o@xmath1 have shown a rich behavior,@xcite including multiple field transitions@xcite which we would like to understand .
for the family of rare earth pyrochlore systems it is well known@xcite that dipole - dipole interactions are important , since the angular momentum is large ( @xmath2 for gd ) .
if one adds nearest neighbor exchange to dipole - dipole interactions , the fourier transform of the total interaction @xmath3 is virtually independent@xcite of @xmath4 ( and takes its minimum value ) along the ( 1,1,1 ) directions of the reciprocal space.@xcite this means that the magnetic ordering wave - vector could , potentially lie anywhere along these lines .
it turns out that the phases at the endpoints , @xmath5 ( denoted , following ref .
, by a ) and @xmath6 ( denoted by @xmath7 or b ) are particularly important .
the ordering expected at @xmath5 ( a - type ) has been discussed by several authors@xcite and is shown in fig .
[ astate ] . a possible ordering at @xmath8 ( b - type ) has been proposed in ref . [ ] .
although one would imagine that gd@xmath0sn@xmath0o@xmath1 and gd@xmath0ti@xmath0o@xmath1 should be quite similar , since the crystal structures are the same ( apart from a very small difference in the lattice constant ) , it is found that gd@xmath0sn@xmath0o@xmath1 has a _ strong _ first - order transition,@xcite while gd@xmath0ti@xmath0o@xmath1 has a second - order transition.@xcite furthermore , gd@xmath0ti@xmath0o@xmath1 orders at @xmath9@xcite while gd@xmath0sn@xmath0o@xmath1 appears more compatible with the a phase.@xcite while the small change in lattice parameter between gd@xmath0sn@xmath0o@xmath1 and gd@xmath0ti@xmath0o@xmath1 could change somewhat the exchange constants , it seems remarkable that the nature of the ordering changes so dramatically .
we would like to understand such a delicate dependence of ordering on exchange constants .
the first order nature of the transition observed in gd@xmath0sn@xmath0o@xmath1 is at variance with mean - field theory which predicts@xcite a second - order transition . in order to clarify the order of the transition , and to see whether it is affected by longer range exchange interactions which lift the degeneracy of @xmath3
, we have performed monte carlo simulations of the classical dipolar heisenberg model on the pyrochlore lattice .
these show that the apparent order of the transition is indeed very sensitive to the exchange constants .
we have supplemented the numerics by a perturbative renormalization group ( rg ) analysis .
it is known that `` fluctuation - induced first - order transitions '' occur when there is no stable fixed point in a perturbative rg calculation .
this frequently occurs when the number of components of the order parameter @xmath10 is larger than 4,@xcite . in general ,
such transitions are expected to be only _ weakly _ first - order .
another microscopic scenario for a first order transition is the proposal of brazovskii@xcite that the existence of a continuous set of degenerate soft modes could change the order of the transition .
it was latter shown that the rg analysis of models with soft modes along special directions lacks stable fixed points and the models are indeed likely to undergo first - order transitions.@xcite this scenario is , for instance , relevant to the description of the liquid crystal transition where the degeneracy naturally comes from the isotropy of the liquid.@xcite it is also particularly relevant in frustrated magnets where frustration may , precisely , provide a large number of quasi - degenerate soft - modes , though with different geometrical structures .
we shall consider a similar rg analysis , applicable for the symmetry of the dipolar pyrochlores which have lines of degeneracy , in this paper .
in this work , we study the order of the transition of the classical heisenberg model on the pyrochlore lattice with long - range dipole - dipole interactions and exchange interactions .
we show , by means of monte - carlo simulations , that the transition is _ strongly _ first - order when a continuous set of soft modes is present in @xmath11 , and becomes _ weakly _ first - order when that degeneracy is removed by including further neighbor interactions ( sec .
[ montecarlo ] ) . in section [ rg ] , we present a landau - ginzburg - wilson simplified model and a perturbative renormalization group analysis of the transition , that predicts a first - order transition , in agreement with the numerical results in sec .
[ montecarlo ] .
we summarize our conclusions in sec .
[ conclusion ] .
we present in this section some results of monte carlo simulations , using the parallel tempering approach,@xcite on the heisenberg model with dipolar and exchange interactions on the pyrochlore lattice .
the classical heisenberg hamiltonian is : @xmath12 where @xmath13 is a classical spin vector of length @xmath2 ( for gd@xmath14 ) on site @xmath15 , and @xmath16 is the heisenberg exchange between the neighbors : we will consider the first ( @xmath17 ) , second ( @xmath18 ) and third neighbor ( @xmath19 ) couplings.@xcite the number of spins is @xmath20 ( @xmath21 ) and periodic boundary conditions are applied .
the factor of 16 arises because the pyrochlore lattice consists of an fcc lattice of tetrahedra , each tetrahedron has 4 spins , and there are 4 sites of the fcc lattice in the conventional cubic cell . to incorporate the b - phase with periodic boundary conditions ,
we need @xmath22 to be even , so most of our results are for @xmath23 and 4 . often the long - range dipolar interactions are cut - off beyond a couple of neighbors@xcite to speed up the simulations .
however , here we have kept a large number of neighbors ( practically infinite ) in order to reproduce accurately the structure of the degenerate states . if the dipole - dipole interaction is cut - off , ripples appear in the degenerate lines of soft modes@xcite .
we constructed periodic repetitions of the monte carlo clusters and included the contributions of many blocks in performing the dipolar sums .
because there is no cut - off in the dipole interactions , the simulation becomes slow for large sizes , so we are limited to @xmath24 ( @xmath25 ) . we investigate ordering at @xmath26 ( called a ) see fig .
[ astate ] , and at @xmath27 ( called b ) , see ref .
[ ] . with a ordering the order parameter has @xmath28 components and with b ordering it has @xmath29 components as detailed in sec .
the @xmath10-component order parameters are given by @xmath30 where @xmath31 is the unit vector of spin @xmath15 assuming the system is fully ordered in component @xmath32 of ordering type a with the @xmath33 state ( resp .
@xmath34 , @xmath35 ) corresponding to @xmath36 ( resp .
@xmath37 , @xmath38 ) ( see e.g. the arrows in fig . [ astate ] ) or b. in the course of the simulation of a finite system , the spin configuration can fluctuate between different , equivalent ordered states .
we therefore compute the invariant quantities @xmath39 it is convenient to also compute the dimensionless binder ratio @xmath40 , \ ] ] for both a and b orderings , which has the property that it tends to 0 at high temperature and to 1 in an ordered state .
( remember that @xmath28 for a - type ordering and @xmath29 for b - type ordering . )
firstly we consider the dipolar model with only nearest neighbor interactions .
results for the order parameters are shown in fig .
[ montecarlosimulation1 ] .
the parameter @xmath17 is taken to be the same as in gd@xmath0ti@xmath0o@xmath1 , i.e. from the high temperature susceptibility @xmath41k , and the dipole - dipole interactions are fixed by the inter - ion distances.@xcite the order parameter associated with the a state jumps rapidly at @xmath42 k , indicating a strong first order transition at that temperature , whereas that of the b state goes to zero at the same temperature .
these results gives a much lower value of the transition temperature @xmath43 than the mean - field theory value of @xmath44 k , though it is a bit smaller than the experimental value of 1 k. the binder ratio shown in fig .
[ montecarlosimulation2 ] is also strongly discontinuous . in the vicinity of @xmath45
the binder ratio gets negative , as expected for a first - order phase transition.@xcite given the results of figs .
[ montecarlosimulation1 ] and [ montecarlosimulation2 ] , we conclude that the transition is _ strongly _ first - order for @xmath46 .
the very small preference for the b - phase@xcite indicated by the small minimum of @xmath3 at @xmath47 , which would be relevant at a second order transition , is unimportant here because the transition is so strongly first order .
we now include further neighbor interactions @xmath18 and @xmath19 that lift the degeneracy of the dipolar model and select other states , as studied in detail in ref .
the lowest part of the spectrum of @xmath3 is shown in fig .
[ degeneracyj2j3 ] for @xmath4 along the @xmath48 direction .
we see that spectrum is almost precisely degenerate for @xmath49 but that @xmath50 is preferred if @xmath18 and @xmath19 are negative , while @xmath9 is preferred if @xmath18 and @xmath19 are positive . for simplicity
we restrict ourselves to @xmath51 and study how the character of the transition is modified relative to the case @xmath49 . for @xmath52
, we find the same a state as for @xmath53 , but the transition temperature shifts to higher temperatures , see figs .
[ montecarlosimulationj1j21 ] and [ montecarlosimulationj1j22 ] .
this is expected from fig .
[ degeneracyj2j3 ] since , with @xmath52 , @xmath11 acquires a well - defined minimum at @xmath26 which gets deeper with increasing @xmath18 and @xmath19 . even with couplings as small as @xmath54 , the order parameter and the binder ratio , shown in fig . [ montecarlosimulationj1j22 ] , vary in a much more gradual way than for the degenerate case @xmath53 .
although only finite - size scaling on a bigger range of sizes could say whether the transition is first or second order , it is clear that removing the degeneracy makes the transition less first - order compared with the degenerate case .
these results are consistent with earlier simulations on model without dipole - dipole interactions ( where larger clusters could be considered ) which pointed out a continuous@xcite or _ weakly _ first - order@xcite transition for the pyrochlore lattice with @xmath55 .
next we consider @xmath56 which , from fig .
[ degeneracyj2j3 ] , is expected to favor the b - like states ( @xmath27 ) , and indeed this is the case as shown by figs .
[ montecarlosimulation3 ] and [ montecarlosimulation4 ] . the minimum of @xmath3 at @xmath27 shown in fig .
[ degeneracyj2j3 ] is independent of @xmath18 and @xmath19 which implies that the mean - field transition temperature is also independent of @xmath18 and @xmath19 .
although we do not have enough system sizes to attempt a serious estimate of @xmath43 using finite size scaling , it seems that the transition temperature is indeed quite similar to that for @xmath57 . as was also found for @xmath52 ,
the transition is much more gradual than for the degenerate case , showing that removing the degeneracy reduces the first - order character of the transition .
the sensitive dependence of both the order of the transition and the nature of the ordered phase on @xmath18 and @xmath19 points to the possible relevance of these terms in explaining the difference between gd@xmath0sn@xmath0o@xmath1 and gd@xmath0ti@xmath0o@xmath1 .
while the data for @xmath53 clearly indicate a strong first order transition , our results for @xmath18 and @xmath19 non - zero ( with either sign ) are not conclusive as to the order of the transition . the smooth behavior of the binder ratios shown in figs .
[ montecarlosimulationj1j22 ] and [ montecarlosimulation4 ] is typical for a second order transition .
however , we can not rule out the possibility of a _ weak _ first order transition where the correlation length at the critical point , @xmath58 , is large . in this case ,
if @xmath59 the behavior will look like that of a second order transition . only for sizes
where @xmath60 can one see a crossover to behavior expected at a first order transition .
hence for @xmath18 and @xmath19 non - zero , where the degeneracy is removed , the transition is either second order or weakly first order .
however , comparing figs .
[ montecarlosimulationj1j22 ] and [ montecarlosimulation4 ] with the corresponding figure for @xmath53 , fig .
[ montecarlosimulation2 ] , we see that behavior for @xmath18 and @xmath19 non - zero is very different from the strong first order behavior found in the degenerate case .
we now study a landau - ginzburg - wilson model by means of the renormalization group analysis . although the method is usually aimed to study second - order phase transitions , the lack of stable fixed points is often considered as an indication for a first - order kind of transition . given the degeneracies of the soft modes with @xmath61 along the 4 equivalent ( 1,1,1 ) directions when @xmath53 ( see fig .
[ degeneracyj2j3 ] ) , the fluctuations of all these modes must be taken into account simultaneously .
for this reason , we consider a model with an infinite - component order parameter ( extended to dimension @xmath62 ) and the fluctuations with wave - vectors close to these soft mode directions .
the quadratic part of the hamiltonian is written : @xmath63 where the @xmath64 are of norm 1 and represent the @xmath65 ( 1,1,1 ) directions given in fig .
[ degeneracies ] .
we have 8 fields , @xmath66 and @xmath67 ( @xmath68 ) , with @xmath69 . if we ignore the @xmath70 term , then , when @xmath71 , all the modes with @xmath72 become simultaneously unstable .
however , as in previous works,@xcite we include the small dispersion along the ( 1,1,1 ) lines , @xmath73 , to make the calculation well defined at intermediate stages . to study the degenerate case ,
it will be eliminated at the end of the calculation by taking @xmath74 .
the fourth - order invariants are similar to that introduced to describe the nematic smectic - c transitions@xcite and are given by : @xmath75 where @xmath76 ensures that the total momentum of the four @xmath77 is zero .
for instance , we can cancel the momentum by choosing pairs of momenta along the same ( 1,1,1 ) line , _ i.e. _ by combining @xmath78 with @xmath79 , and similarly with @xmath80 and @xmath81 .
that gives a fourth - order term @xmath82 term , where @xmath68 and @xmath83 is meant for @xmath84 if @xmath85 .
note that given the @xmath86 symmetry , @xmath87 does not depend upon @xmath15 , but only upon @xmath88 .
in addition , we could choose the first two @xmath61 s along 1 and 3 for instance ( see fig .
[ degeneracies ] ) and the other two along -2 and -4 , which gives @xmath89 .
another simplifying feature that we have adopted consists of neglecting the wave - vector dependence of the coefficients @xmath87 . omitting to write the @xmath90 wave - vectors ,
the only fourth - order invariants are given by : @xmath91 where @xmath92 .
we call @xmath93 the ( 1,1,1 ) model .
first , we consider the hartree correction@xcite to the gap @xmath94 ( self - energy ) : @xmath95 in @xmath96 , if we introduce momentum cut - offs , the new gap @xmath97 ( @xmath98 ) is given by : @xmath99 where @xmath32 is a proportionality coefficient and @xmath100 and @xmath101 the cut - offs . due to the strong singularity of the right - hand - side ,
the gap does not vanish anymore .
it suggests that the paramagnetic phase remains locally stable below the transition , together with other more stable phases . from the existence of other phases ( at least at the mean - field level )
, the transition is expected to be first - order.@xcite however , the cut - offs @xmath102 and @xmath103 enter explicitly the equation and a more controlled result can be obtained by the renormalization group analysis by restricting the integration to a shell of momentum @xmath104 with @xmath105.@xcite for this we introduce new real order - parameters ( @xmath65 ) : @xmath106 and the quadratic and quartic terms become : @xmath107 [ ( \phi_2 ^ 2 + \bar{\phi}_2 ^ 2)+(\phi_4 ^ 2 + \bar{\phi}_4 ^ 2 ) ] \nonumber \\ & + & u_2 [ ( \phi_1 ^ 2 + \bar{\phi}_1 ^ 2 ) ( \phi_3 ^ 2 + \bar{\phi}_3 ^
2 ) + ( \phi_2 ^ 2 + \bar{\phi}_2 ^ 2 ) ( \phi_4 ^ 2 + \bar{\phi}_4 ^ 2 ) ] \nonumber \\ & + & u_3 ( \phi_1 \phi_2 \phi_3 \phi_4 + \bar{\phi}_1 \bar{\phi}_2 \bar{\phi}_3 \bar{\phi}_4 ) \nonumber \\ & - & u_3 ( \phi_1 \bar{\phi}_2 \phi_3 \bar{\phi}_4 + \bar{\phi}_1 \phi_2 \bar{\phi}_3 \phi_4 ) \nonumber \\ & + & u_3 ( \phi_1 \bar{\phi}_3 + \bar{\phi}_1 \phi_3 ) ( \phi_2 \bar{\phi}_4 + \bar{\phi}_2 \phi_4 ) \label{h4}\end{aligned}\ ] ] the derivation of the rg equations for the coupling constants is then similar to that of ref .
[ ] except that we have to keep track of the field labels , given the anisotropy of @xmath108 in eq .
( [ h2 ] ) . by integrating over a shell of momentum @xmath109
, we find the recursion relations for the new coupling constants : @xmath110 \nonumber \\
u_1^{\prime } & = & b^{\epsilon } \left[u_1 - \right ( 8 u_1 ^ 2 i_1
+ u_3 ^ 2 i_1 + 32 u_0 u_1 i_0 + 8 u_1 u_2 i_0 \left ) \right ] \nonumber \\ u_2^{\prime } & = & b^{\epsilon } \left[u_2 - \right ( 8 u_1 ^ 2 i_0 + 8 u_2 ^ 2 i_2
+ u_3 ^ 2 i_2 + 32 u_0u_2 i_0 \left ) \right ] \nonumber \\ u_3^{\prime } & = & b^{\epsilon } \left[u_3 - \right ( 8 u_2 u_3 i_2 + 8 u_1 u_3 i_1 \left ) \right ] \nonumber \\ a^{\prime } & = & b^{-2(m-1 ) } a \label{dangerous}\end{aligned}\ ] ] where @xmath111 and the integrals are defined by @xmath112 the relation ( [ dangerous ] ) , together with the divergences of the integrals @xmath113 for @xmath98 ( at @xmath71 , finite @xmath114 ) implies that @xmath115 is a dangerous irrelevant variable for @xmath116 .
@xmath117 diverges indeed as @xmath118 , and @xmath119 as @xmath120 . to take into account these divergences
, we have to introduce rescaled constants @xmath121 ( ref .
[ ] and references therein ) . with @xmath122 ,
the recursion relations become ( we introduce @xmath123 ) : @xmath124 \nonumber \\ \tilde{u}_1^{\prime } & = & b^{m(5-d-1/m ) } \left[\tilde{u}_1 - \right ( 32 \tilde{u}_0 \tilde{u}_1 + 8 \tilde{u}_1 \tilde{u}_2 \left ) { \cal i } \right ] \nonumber \\ \tilde{u}_2^{\prime } & = & b^{m(5-d-1/m ) } \left[\tilde{u}_2 - \right ( 8 \tilde{u}_1 ^ 2 + 32 \tilde{u}_0 \tilde{u}_2
\left ) { \cal i } \right ] \nonumber \\ \tilde{u}_3^{\prime } & = & b^{m(5-d-1/m ) } \tilde{u}_3 \end{aligned}\ ] ] all the fixed points are unstable for @xmath125 ( the upper critical dimension is 5 for @xmath126 ) since @xmath127 .
although strictly speaking new fixed points could occur at order @xmath128 , the present calculation at order @xmath129 is compatible with the first - order transition observed in the monte - carlo simulations .
we now remove the degeneracy , _
e.g. _ by including further neighbor couplings in the microscopic hamiltonian . first of all we assume that the ordering is at @xmath131 .
it is simplest to go back to the order parameter shown in fig .
[ astate ] , realize that there are three components @xmath132 , @xmath133 , and that the symmetry is cubic .
the hamiltonian of this cubic model is therefore given by @xmath134 + u_1 \sum_{i , j } \psi_i^2 \psi_j^2 \ , .\ ] ] there has been a controversy regarding whether the stable fixed point of the cubic model is the heisenberg or the cubic fixed point .
a recent 6-loop expansion has shown that for @xmath135 , the stable fixed - point is the cubic one.@xcite depending on the initial values for the coupling constants , the transition could be either first - order or continuous . however , to stabilize the _ collinear _ states ( with @xmath77 either @xmath136 , @xmath137 , or @xmath138 and @xmath26 ) , the set of initial coupling constants leads to a first - order transition.@xcite we have seen that a first order transition is obtained both in the degenerate case and also when there is a well - defined minimum at @xmath139 .
is , then , the degeneracy of soft modes important or not ?
we note that , in the absence of degeneracy , the transition may be only _ weakly _ first - order .
the problem was studied some years ago in the context of the pyrochlore fef@xmath140 . for this compound ,
the @xmath139 state found by neutron scattering was characterized by a 3-component order parameter,@xcite similar to the one we have here .
monte - carlo simulations have shown that the collinear structures with @xmath141 are preferred , but the transition first appeared to be second - order , with unusual critical exponents@xcite contrary to the rg argument given above .
we can reconcile these results by suggesting that the transition may be _ weakly _ first - order , so that the correlation length would exceed the size of the monte - carlo cluster and the transition would appear second - order in the simulation .
this is also comforted by a reexamination of the monte - carlo results , which suggested that the transition is more likely to be indeed weakly first - order.@xcite we now assume that the degeneracy is lifted in such a way that one of the four @xmath143 wave - vectors is selected .
since @xmath143 and @xmath144 are related by a reciprocal lattice vector , we have to take into account the fluctuations of four fields only , with @xmath61 close to any of the @xmath143 wave - vectors , @xmath145 .
the critical model is given by @xmath146 + u_1 \sum_{i , j } \psi_i^2 \psi_j^2 \nonumber \\ & + & u_3 \psi_1 \psi_2 \psi_3 \psi_4 \ , .\end{aligned}\ ] ] this model is known to possess unstable fixed points at order @xmath128.@xcite therefore the transition to the @xmath143 phases is also expected to be first - order . on the basis of the lgw models alone , we would conclude that the phase transitions in the dipolar pyrochlore are _ all _ first - order in character .
such a simple analysis does not say whether the transition is _ strongly _ or _ weakly _ first - order that is quite a relevant question when one comes to compare with experiments . nonetheless , the results presented in this section are compatible with the monte - carlo simulations of section [ montecarlo ] .
the latter are important , precisely to say whether the transitions are weakly or strongly first - order .
we have considered the dipolar heisenberg model on a pyrochlore lattice with nearest neighbor interactions and a small amount of second and third neighbor interactions ( @xmath18 and @xmath19 ) . for @xmath53
the system is highly degenerate , see fig .
[ degeneracyj2j3 ] , and fluctuation effects pick out ordering at @xmath50 ( a - type ) .
monte carlo simulations show that the transition is very strongly first order in this case , in contrast to mean - field theory which predicts a second order transition .
a first order transition is also predicted by a renormalization group analysis . when the degeneracy is removed by including @xmath18 and @xmath19 the transition is more gradual , showing that the degeneracy is necessary to get a _
strong _ first order transition . given the limited range of sizes in the monte carlo simulations ,
we can not say from the simulations whether the transition is second order or weakly first order for @xmath147 .
however , according to a renormalization group analysis for the non - degenerate case , both a and b type orderings have no stable fixed points , indicating , presumably , a fluctuation induced first order transition .
usually this type of transition is only weakly first order , and this seems to be consistent with our numerical data . because of the degeneracy for @xmath53 , a small amount of second and third neighbor coupling can also change the nature of the ground state .
we find that for @xmath148 the a phase is retained but for @xmath56 we obtain a @xmath9 ( b - type ) ordering . in future work
we will study in more detail the nature of this b - type phase , and also consider other possible phases that occur when @xmath149 .
it is possible that anisotropic interactions , suggested on the basis of high - temperature esr@xcite and by epr on diluted samples,@xcite may be needed to explain the experimentally observed phases in detail .
our results provide a natural explanation for gd@xmath0sn@xmath0o@xmath1 having a strong first - order transition,@xcite while gd@xmath0ti@xmath0o@xmath1 has a second - order transition@xcite ( though a weak first - order transition is not ruled out experimentally ) ; namely second and third neighbor interactions are very weak in gd@xmath0sn@xmath0o@xmath1 , but they are stronger and positive for gd@xmath0ti@xmath0o@xmath1 . in this respect , _ ab - initio _ calculations could give some estimate of the strength of the couplings .
this picture is also consistent with the observations that gd@xmath0ti@xmath0o@xmath1 orders@xcite at @xmath9 while gd@xmath0sn@xmath0o@xmath1 should be a - type with equivalent sites and moments perpendicular to the local ( 1,1,1 ) directions.@xcite in the presence of a magnetic - field,@xcite gd@xmath0ti@xmath0o@xmath1 has a rich phase diagram .
for the future , it would also be interesting to perform a study of field - induced transitions in gd@xmath0sn@xmath0o@xmath1 , since this starts off with a quite different state in zero field .
in addition to the multiple phase transitions expected on the basis of mean - field theory , the field reduces the fluctuations and , hence should reduce the strong first - order character of the transition .
we would like to thank a. wills for sharing with us unpublished results concerning gd@xmath0sn@xmath0o@xmath1 .
o.c . would like to thank especially g. jackeli and t. ziman for very stimulating discussions and also s. bramwell , v. glazkov , e. kats , l. lvy , y. motome , r. stewart , and m. zhitomirsky . a.p.y .
acknowledges support from the national science foundation under grant no .
dmr 0337049 .
b.s.s . acknowledges support from the national science foundation under grant no .
dmr 0408247 .
a.p.y . would also like to thank the institut laue langevin , where this work was started , for its hospitality .
k. hukushima , and k. nemoto , j. phys .
65 * , 1604 ( 1996 ) ; e. marinari _ in _ advances in computer simulation ( editors j. kertsz and i. kondor , springer - verlag , berlin , 1998 ) p. 50
( also in cond - mat/9612010 ) . | we show that a continuous set of degenerate critical soft modes strongly enhances the first - order character of a fluctuation - induced first - order transition in the pyrochlore dipolar heisenberg antiferromagnet .
such a degeneracy seems essential to explain the strong first - order transition recently observed in gd@xmath0sn@xmath0o@xmath1 .
we present some evidence from monte - carlo simulations and a perturbative renormalization group expansion . |
with the aim of obtaining a fundamental basis behind the formation of super - chandrasekhar white dwarfs ( wds ) , mukhopadhyay and his collaborators @xcite initiated the exploration of highly magnetized wds and their new mass - limit , significantly exceeding the chandrasekhar limit of @xmath1.@xcite these wds are ideally suited to be the progenitors of peculiar , overluminous , type ia supernovae , e.g. sn 2003fg , sn 2006gz , sn 2007if , sn 2009dc,@xcite which are best explained by invoking the explosion of super - chandrasekhar wds having mass @xmath2 . along with the fact that several wds have been discovered with surface fields @xmath3 g , it has also been known that magnetized wds tend to be more massive than their non - magnetized counterparts.@xcite such observations motivate the theoretical investigation of the effect of a strong interior magnetic field on the mass of a wd . in this context
, we mention that our previous attempts at obtaining highly magnetized super - chandrasekhar wds , assumed _ a priori _ spherical symmetry .
while that may indeed be the case for certain magnetic field geometries , in general , highly magnetized wds tend to be deformed due to magnetic tension . with each new step
we have scientifically progressed towards a more rigorous model starting from a simplistic newtonian , spherically symmetric , constant field model and culminating in a model with self - consistent departure from spherical symmetry by general relativistic magnetohydrodynamic ( grmhd ) formulation , which we explain in the present work ( also , see ref .
we appropriately modify the _ xns _
code,@xcite which has so far been used only to model strongly magnetized neutron stars , to compute equilibrium configurations of static , strongly magnetized wds in the gr framework , for the first time in the literature to the best of our knowledge .
for a detailed description of the underlying grmhd equations , the magnetic field geometries , the numerical technique employed by the _
xns _ code and the values of various code parameters ,
we refer the readers to refs .
we construct axisymmetric wds in spherical polar coordinates @xmath4 , to self - consistently account for the deviation from spherical symmetry due to a strong magnetic field , which generates an anisotropy in the magnetic pressure.@xcite a uniform computational grid is used along both the radial @xmath5 and polar @xmath6 co - ordinates , the number of grid points being typically @xmath7 and @xmath8 respectively .
even higher resolution runs ( for e.g. with @xmath9 and @xmath10 ) require more computational time but do not lead to any significant change in the results . in this work ,
we focus on the equilibrium solutions of high density , magnetized , relativistic wds , which can be described by a polytropic equation of state ( eos ) @xmath11 , where @xmath12 is the pressure and @xmath13 the density , such that the adiabatic index @xmath14 and the constant @xmath15 is same as that obtained by chandrasekhar.@xcite hence , we neglect the possible effect of landau quantization on the above eos which could arise due to a strong magnetic field @xmath16 , where @xmath17 g , is the critical magnetic field.@xcite we recall that the maximum number of landau levels @xmath18 occupied by electrons in the presence of a magnetic field is given by equation ( 10 ) of ref .
1 . the range of central density and maximum magnetic field strength inside the wds considered in this work are @xmath19 gm/@xmath20 and @xmath21 g respectively . consequently , @xmath22 for this range of @xmath23 and @xmath24 , which is large enough not to significantly modify the value of @xmath25 we choose , hence justifying our assumption .
we now explore the effect of various magnetic field geometries on the structure and properties of wds . for a fiducial model , we choose a non - magnetized wd with @xmath26 gm/@xmath20 .
it has a baryonic mass @xmath27 and equatorial radius @xmath28 km , and is perfectly spherical with @xmath29 , @xmath30 being the polar radius ( note that for the definitions of all global physical quantities characterizing the solutions in this work , we refer to appendix b of ref . 13 ) .
# 1 ( # 1 ) figures [ poltor](a ) and ( c ) portray the distribution of baryonic density and magnetic field strength respectively , for the fiducial wd having a purely toroidal magnetic field configuration @xmath31 .
the maximum magnetic field strength inside this wd is @xmath32 g. interestingly , this is a highly super - chandrasekhar wd , having @xmath33 .
note that the value of the surface magnetic field does not affect this result , as long as it is @xmath34 g , which is satisfied in this case . in this context , we mention that the detection of very high surface magnetic fields @xmath35 g is very difficult due to the featureless spectrum.@xcite very importantly , the ratio of the total magnetic energy to the total gravitational binding energy , @xmath36 ( which is much @xmath37 ) . wds with even smaller @xmath38 are also found to be highly super - chandrasekhar ( see figs .
[ rhocf]a and d ) .
this argues for the wds to be stable ( see , e.g. ref .
. the radii ratio , @xmath39 ( which is slightly @xmath40 ) , indicates a net prolate deformation in the shape caused due to a toroidal field geometry .
figure [ poltor](a ) shows that although the central iso - density contours are compressed into a highly prolate structure , the outer layers expand , giving rise to an overall quasi - spherical shape .
interestingly , this justifies the earlier spherically symmetric assumption in computing models of at least certain strongly magnetized wds.@xcite figures [ poltor](b ) and ( d ) show the distribution of baryonic density and magnetic field strength superimposed by magnetic field lines respectively , for the fiducial wd having a purely poloidal magnetic field configuration @xmath41 .
the maximum magnetic field strength attained at its center is @xmath42 g , which also leads to a significantly super - chandrasekhar wd having @xmath43 .
the wd is highly deformed with an overall oblate shape and @xmath44 , which is expected to be stable because its @xmath45 , which is very much @xmath37.@xcite # 1 ( # 1 ) purely toroidal and poloidal field configurations are believed to be subjected to mhd instabilities,@xcite which possibly rearranges the field into a mixed configuration.@xcite hence , for completeness , we also construct equilibrium models of wds with a mixed magnetic field configuration . in figure [ mixed ]
, we present the results for the fiducial wd having the so - called _ twisted torus configuration _
, which compares the distribution of the toroidal and poloidal components of the magnetic field .
this again results in a significantly super - chandrasekhar wd having @xmath46 .
the poloidal component attains @xmath47 g at the center , while the ring - like toroidal component is an order of magnitude smaller .
the wd assumes a highly oblate shape with @xmath48 , resembling the purely poloidal case in all its attributes , and is again expected to be stable having @xmath49 . in a toroidal dominated mixed field configuration,@xcite more massive super - chandrasekhar wds could be possible and is worth further exploration . in this context , we mention that the results in this work have also been reproduced by bera & bhattacharya,@xcite albeit without any novel contribution to the topic .
we also construct equilibrium sequences of magnetized wds pertaining to different field geometries for the fiducial case with @xmath26 gm/@xmath20 .
figure [ rhocf ] shows the variations of different physical quantities as functions of @xmath24 .
the two most important revelations of figure [ rhocf ] are ( 1 ) the wd mass increases with an increase in magnetic field for all the three field configurations discussed above ( see figure [ rhocf]a ) , eventually leading to highly super - chandrasekhar wds and ( 2 ) the magnetic energy remains ( significantly ) sub - dominant compared to the gravitational binding energy for all the cases , since @xmath50 always ( see figure [ rhocf]d ) , which argues for the possible existence of stable , highly magnetized super - chandrasekhar wds .
gm/@xmath20 . ( a )
@xmath51 , ( b ) @xmath52 , ( c ) @xmath53 and ( d ) @xmath38 , as functions of @xmath24 .
the solid ( green ) , dotted ( blue ) and dashed ( red ) curves represent , respectively , wds having purely poloidal , purely toroidal , and twisted torus field configurations .
@xmath51 is in units of @xmath54 .
the filled boxes represent individual wds.,width=528 ]
since our foray into this topic , we have been persistent with our message that the versatile nature of magnetic field is paramount in the revelation of significantly super - chandrasekhar wds , irrespective of its nature of origin : quantum , classical and/or general relativistic which is re - emphasized in the present work . by carrying out extensive , self - consistent , grmhd numerical analysis of magnetized wds
, we have reestablished the existence of highly super - chandrasekhar , stable wds . in order to self - consistently study the anisotropic effect of a strong magnetic field , we have explored various geometrical field configurations , namely , purely toroidal , purely poloidal and twisted torus configurations .
interestingly , we have obtained significantly super - chandrasekhar magnetized wds for all the cases , having mass @xmath55 , and that too at relatively lower magnetic field strengths when the deviation from spherical symmetry is considered as already speculated in our earlier work.@xcite these wds can be ideal progenitors of the aforementioned peculiar , overluminous type ia supernovae .
our work also establishes the necessity of a general relativistic formalism over a newtonian approach while constructing models of magnetized super - chandrasekhar wds .
the authors thank n. bucciantini and a.g .
pili for their inputs with the _ xns _ code .
this work was partly supported by the isro grant isro / res/2/367/10 - 11 and was carried out at the department of physics , indian institute of science , bangalore , india . | our consistent effort to unravel the mystery of super - chandrasekhar white dwarfs ( wds ) , by exploiting the potential of magnetic fields , has brought this topic considerable attention .
this is also evident from the recent surge in the corresponding literature . in the present work , by means of full - scale
general relativistic magnetohydrodynamic ( grmhd ) numerical analysis , we confirm the existence of stable , highly magnetized , significantly super - chandrasekhar wds having mass exceeding @xmath0 solar mass .
we have explored various possible field configurations , namely , poloidal , toroidal and mixed , by self - consistently incorporating the departure from spherical symmetry induced by a strong magnetic field .
such super - chandrasekhar wds can be ideal progenitors of peculiar , over - luminous type ia supernovae . |
for a smooth , projective curve @xmath4 with a choice @xmath5 of marked points , and an assignment @xmath6 of parabolic subgroups @xmath7 to these points , a quasi - parabolic principal bundle of type @xmath8 is a principal bundle @xmath9 on @xmath10 and a choice @xmath11_{p_i}$ ] from the fiber of the @xmath12 associated bundle at the marked point @xmath13 .
the moduli stacks @xmath14 of these objects have been widely studied due to their place in algebraic geometry as a generalization to the jacobian variety , and because of their relationship to conformal field theory and representation theory , which we will describe further below . in @xcite
we began a program to study these stacks by describing the algebras formed by the global sections of their line bundles , sometimes called non - abelian theta functions .
our primary object of interest is the algebra formed by all of the non - abelian theta functions , obtained by endowing the direct sum of all global sections of line bundles on the stack with global section multiplication .
this algebra is known as the cox ring or total coordinate ring of the stack @xmath14 . in this paper
we describe this algebra in a special case @xmath15 where the curve @xmath10 is a projective line , and the parabolic structure at each marked point is given by the stabilizers @xmath16 of the highest weight vector in the representation @xmath1 or its dual @xmath2 .
we let @xmath17 be the number of marked points with parabolic data @xmath18 and @xmath19 be the number of marked points with parabolic data @xmath20 line bundles @xmath21 on this stack are indexed by a tuple @xmath22 of @xmath17 non - negative integers , a tuple @xmath23 of @xmath19 non - negative integers , and a non - negative integer @xmath24 called the level .
@xmath25 @xmath26 by definition the algebra @xmath27 contains each of the projective coordinate rings @xmath28 defined by the line bundle @xmath29 as a graded subalgebra .
we prove the following theorem .
[ main ] for generic arrangements of points @xmath30 the algebra @xmath27 is minimally generated by the non - abelian theta functions with @xmath31 the relations which hold among these generators are generated by the quadratic relations . in the case @xmath32 the parabolic subgroups @xmath33 are equal , in this case we set @xmath34 .
this case is studied by castravet and tevelev in @xcite , where the algebra @xmath35 is shown to be a cox - nagata ring .
this means that @xmath35 is the algebra of invariants by a certain non - reductive ( additive ) group action on a polynomial ring , and can be identified with the cox ring of a blow - up of a projective space .
sturmfels and xu also study this case in @xcite , where they construct a sagbi ( subalgebra analogue of grbner bases for ideals ) degeneration @xmath36 of @xmath35 for generic @xmath37 , for each trivalent tree @xmath38 with @xmath39 leaves .
the degenerations @xmath36 constructed in @xcite are notable as they are affine semigroup algebras which have also been studied by buczyska and winiewski @xcite in the context of mathematical biology .
the algebras @xmath36 are the coordinate rings of the @xmath40 group - based phylogenetic statistical models .
generators for the ideal of relations presenting these coordinate rings are important for applications of these models , to that end buczyska and winiewski prove results for @xmath36 which imply theorem [ main ] in the @xmath41 case , via the degeneration techniques in @xcite . in this paper we also employ degeneration methods , developed in @xcite from elements of conformal field theory ( see below ) .
the tree combinatorics is still present , however it stems from a relationship to the moduli of marked genus @xmath42 curves @xmath43 rather than a sagbi construction
. it would be interesting to find a sagbi construction for @xmath27 in the general case , and also an interpretation of this algebra as the invariants of an additive group action on a polynomial ring .
we choose the standard set of positive roots @xmath44 for @xmath0 , with corresponding weyl chamber @xmath45 dominant weights @xmath46 of @xmath0 and positive dominant weights of @xmath47 in this weyl chamber are weakly descreasing lists of @xmath48 and @xmath49 numbers , respectively .
@xmath50\\\ ] ] we let @xmath51 denote the irreducible representation corresponding to a dominant weight @xmath52 the fundamental weight @xmath53 is the list with @xmath54 @xmath55s followed by @xmath56 @xmath57s , and its corresponding irreducible represention is the exterior product @xmath58 note that @xmath59 is the determinant representation of @xmath47 and therefore the trivial representation of @xmath60 in order to prove theorem [ main ] , we first degenerate @xmath27 to an affine semigroup algebra , using theorem @xmath61 in @xcite . when specialized to the case we consider here , this theorem says that for any trivalent tree @xmath38 with @xmath62 labelled leaves , there is a degeneration of @xmath27 to an algebra @xmath63 . the algebra @xmath63 is built out of the total coordinate rings of moduli of @xmath0 quasi - parabolic principal bundles on a triple marked projective line , @xmath64 according to a specific recipe dictated by @xmath65 the structure of @xmath63 depends on the tree @xmath66 so it is critical to choose a tree topology which gives an algebra with advantageous properties . from now on , we consider the `` caterpillar '' tree @xmath67 , depicted in figure [ cat ] .
this tree has the property that each vertex of @xmath67 is connected to a leaf by a single edge .
each leaf - edge of this tree corresponds to a marked point , and is assigned a @xmath18 or @xmath68 , the internal edges are each assigned a copy of the borel subgroup @xmath69 .
each internal vertex of @xmath67 now has three parabolic subgroups assigned to its incident edges , so following @xcite , we think of this vertex as corresponding to the total coordinate ring of the moduli of quasi - parabolic principal bundles on @xmath64 with parabolic structure coming from the corresponding parabolic subgroups .
four types of total coordinate rings appear , @xmath70 @xmath71 @xmath72 and @xmath73 we call these the @xmath3pieri algebras for reasons that will be made clear .
the following theorem expresses the degeneration of @xmath27 in terms of these algebras , it follows from theorem @xmath61 from @xcite .
[ degen ] the algebra @xmath27 has a flat degeneration to the following algebra . @xmath74^{\otimes a-2 }
\otimes [ v_{0 , 3}(b , p^ * , b)]^{\otimes b- 2 } \otimes v_{0 , 3}(b , p^ * , p^*)]^{t}\\\ ] ] this is the algebra of invariants in a tensor product of @xmath3pieri algebras by a torus @xmath75 this torus is a product of @xmath76 copies of @xmath77 one for each internal edge of @xmath78 where @xmath79 is the torus of diagonal matrices , we describe the action of this torus below .
equation [ specialpic ] is a special case of a more general theorem due to laszlo and sorger , @xcite , which expresses the picard group of the moduli @xmath14 of quasi - parabolic principal bundles on a curve @xmath10 with parabolic structure @xmath80 at @xmath13 in terms of the character groups @xmath81 of the associated parabolic subgroups .
@xmath82 when each @xmath80 is the borel subgroup @xmath83 , we let @xmath84 denote the corresponding cox ring . this algebra contains the cox rings of all other moduli of principal bundles as a multigraded subalgebra , in particular @xmath85 contains each of the @xmath3pieri algebras as multigraded subalgebras . by laszlo and sorger s theorem , the cox rings @xmath70 @xmath71 @xmath72 @xmath86 are multigraded by triples of dominant @xmath0 weights .
the algebra @xmath87 is given as an example below .
@xmath88 here @xmath89 .
the weights @xmath90 are allowed to vary over all dominant weights in the weyl chamber @xmath91 whereas @xmath92 is always taken from the ray through @xmath93 corresponding to the parabolic subgroup @xmath94 the analogous expression holds for @xmath87 , @xmath95 , @xmath73 this implies that the tensor product @xmath96^{\otimes a-2 } \otimes [ v_{0 , 3}(b , p^ * , b)]^{\otimes b- 2 } \otimes v_{0 , 3}(b , p^ * , p^*)\\\ ] ] is a multigraded sum of spaces @xmath97 @xmath98 @xmath99 + the torus @xmath100 action on these spaces is by the characters determined by the weight information @xmath101 .
a graded component of the invariant algebra @xmath102 of the torus @xmath100 therefore has @xmath103 and @xmath104 this allows us to understand properties of the degeneration @xmath102 by studying the @xmath3pieri algebras . in order to understand the structure of these algebras , in particular their multigraded components , we exploit a connection between moduli of principal bundles on curves and mathematical physics
the moduli of quasi - parabolic principal bundles on a smooth marked curve have a close connection with conformal field theory , in particular the non - abelian theta functions are identified with the partition functions from the wess - zumino - witten model of conformal field theory , and so - called conformal blocks .
this theory associates a space of conformal blocks @xmath105 to each stable , marked curve @xmath106 , assignment of dominant weights @xmath107 and non - negative integer @xmath108 the theorem establishing this relationship in various cases is due to narasimhan , ramanathan , kumar ; beauville , laszlo , sorger ; faltings ; and pauly . for @xmath106 a smooth marked curve ,
@xmath109 @xmath110 conformal blocks are also the structure spaces of a monoidal operation called the fusion product on integrable highest - weight representations of the affine kac - moody algebra @xmath111 . as a consequence
, we may bring in the associated combinatorial representation theory to analyze the graded pieces of @xmath112 in particular , we use a result called the @xmath3pieri rule , that each graded component of @xmath70 @xmath71 @xmath113 and @xmath86 is multiplicity free . for a description of the @xmath3pieri rule in the context of fusion coefficients see section @xmath114 of @xcite .
this result implies that each @xmath3pieri algebra has a basis labelled by the weight data @xmath115 of their graded components @xmath116 and multiplication of basis members is simple addition on this weight data .
formally then we say that the @xmath3pieri algebras are affine semigroup algebras for semigroups of weight data @xmath117 , @xmath118 @xmath119 @xmath120 .
we must first determine the structure of these semigroups .
[ kpieri][the k - pieri rule ( reformulated ) ] the space @xmath121 has dimension @xmath42 or @xmath122 this space has dimension @xmath123 if and only if there is a positive @xmath47 dominant weight @xmath124 such that the following hold .
1 . @xmath125 + 2 .
@xmath126 + 3 . @xmath127 + 4 .
@xmath128 + the entries of the dominant weights @xmath129 from proposition [ kpieri ] can be placed in an interlacing diagram , shown in figure [ pieri ] for @xmath130 notice that @xmath124 can be recovered from the data @xmath131 namely @xmath132 , @xmath133 @xmath134 + [ pieri ] for such a pattern @xmath135 we define its boundary weights as follows .
@xmath136 @xmath137 by duality considerations , the space @xmath121 is non - zero with dimension @xmath123 if and only if the same is true for @xmath138 .
for this reason , interlacing patterns also give a mechanism to label the ( multiplicity free ) graded components of the algebra @xmath95 .
the space @xmath139 is labelled by @xmath140 weights @xmath141 and @xmath142 , which satisfy the following .
1 . @xmath143 + 2 .
@xmath144 + 3 . @xmath145 + 4 .
@xmath146 + we illustrate interlacing patterns @xmath147 for the spaces @xmath148 with the opposite orientation
. @xmath149 + [ pieridual ] we apply the boundary maps @xmath150 to interlacing patterns @xmath147 as follows .
@xmath151 @xmath152 each such interlacing pattern labels one of the multiplicity - free spaces in the multigraded decomposition of @xmath153 and @xmath87 and likewise @xmath95 and @xmath86 . the algebra @xmath102 is also graded by multiplicity - free components , and is therefore an affine semigroup algebra for an affine semigroup @xmath154 .
the elements of @xmath154 are lists of elements @xmath155 from the semigroups above , which can be glued along their boundary data , @xmath156 @xmath157 @xmath158 .
we will present this semigroup explicitely , by first analyzing each of the four algebras @xmath70 @xmath71 @xmath72 @xmath73 and then by controlling the `` gluing '' procedure .
the following proposition , proved in section [ lv ] , says that the four algebras we consider as our building components are not very complicated .
[ poly ] the algebras @xmath159 are polynomial rings on @xmath160 variables .
the algebras @xmath161 are polynomial rings on @xmath162 variables
. we will prove this proposition by carefully reproving the @xmath3pieri rule using elements of commutative algebra .
this continues a theme first established in @xcite and continued in @xcite , that the combinatorics of the conformal blocks are related to toric degenerations of the moduli of quasi - parabolic principal bundles .
the gluing operation described above should be a familiar operation for researchers who study affine semigroup algebras .
the following proposition is a consequence of the description of @xmath102 as an algebra of torus invariants in a tensor product of the @xmath3pieri algebras .
[ tfiber ] the affine semigroup @xmath154 is a toric fiber product of the affine semigroups associated to the @xmath3pieri algebras .
toric fiber products of affine semigroups are defined in @xcite , and further explored in @xcite .
informally we may think of the toric fiber product as a way to glue the @xmath3pieri semigroups corresponding to the above interlacing diagrams together to produce a composite semigroup @xmath154 , and indeed we describe the defining inequalities and corresponding diagram of this semigroup in section [ gt ] . in section [ gt ]
we also describe a polytope @xmath163 whose graded affine semigroup algebra is induced the degeneration of the projective coordinate ring @xmath164 we make use of the fiber product structure on @xmath154 , and general theorems on toric fiber products from @xcite and @xcite , to prove an explicit presentation of @xmath112 the set @xmath165 is defined to be the set of all tuples @xmath166 $ ] of @xmath167 elements from @xmath168 such that the first and last entries are @xmath48 or @xmath42 , and the difference
@xmath169 is in @xmath170 if @xmath171 and in @xmath172 if @xmath173 . on the set of tuples above
, a swap is a quadratic polynomial of the following form , @xmath174[\vec{i_1}^ * , i , \vec{i_2}^ * ] - [ \vec{i}_1 , i , \vec{i_2}^*][\vec{i_1}^ * , i , \vec{i_2}]\\\ ] ] [ toricmain ] the algebra @xmath102 is presented by generators @xmath175 subject to all possible swap relations .
this is a presentation of @xmath102 by generators and relations .
the fiber product structure of @xmath154 also allows us to prove that the algebra @xmath102 is gorenstein , and that it possesses a quadratic square - free grbner basis given by the swap relations above , in section [ lv ] .
these properties then lift to the algebra @xmath27 for generic @xmath37 by general properties satisfied by algebras in flat families , see the proof of theorem @xmath176 in @xcite . in order to prove these results , in particular proposition [ poly ] , we study an auxiliary semigroup @xmath177 , which comes from a classical algebra from invariant theory . one of our motivations in studying the algebra @xmath178 is that it bears a resemblance to the object of study in a classical theorem in invariant theory due to weyl .
we let @xmath179 be the algebra of @xmath0 invariants in the coordinate ring of the affine space @xmath180^a \times [ \c^m]^b$ ] .
@xmath181^a \times [ \c^m]^b]^{sl_m(\c)}\\\ ] ] we view the algebra @xmath182^a \times [ \c^m]^b]$ ] as a polynomial ring on @xmath183 variables , arranged in a matrix , with the columns labelled by elements of @xmath184 = [ a ] \coprod [ b]$ ] .
@xmath185 we let @xmath186 be the determinant form on the variables determined by a subset @xmath187 $ ] in @xmath180^a$ ] of size @xmath49 , and @xmath188 be the dual determinant in @xmath189^b,$ ] for @xmath190.$ ] we let @xmath191 be the column - wise inner product of the variables on the indices @xmath192 $ ] with those in @xmath193.$ ] each of the elements @xmath194 is an @xmath0 invariant , and therefore a member of @xmath179 .
weyl described a collection of quadratic relations on these elements , known as the plcker relations .
the algebra @xmath179 is generated by the @xmath194 , and all relations among these generators are generated by the plcker relations .
the algebra @xmath179 is multigraded by an action of @xmath195 copies of the maximal diagonal torus @xmath79 .
the components of this multigrading are the @xmath0 invariant spaces in the tensor products , @xmath196 in @xcite a relationship is established between @xmath179 and @xmath27 . for any collection of @xmath197
points @xmath198 the algebra @xmath27 is a @xmath199multigraded subalgebra of the polynomial ring over weyl s invariant ring , @xmath200.$ ] @xmath201 theorem @xmath176 in @xcite implies that the degeneration of @xmath27 corresponding to @xmath67 we use here also applies to the algebra @xmath179 .
this connection allows us to use the algebra @xmath179 to study @xmath202 we will show how a similar analysis of @xmath179 using the classical pieri rule yields the following presentation of this degeneration , @xmath203 let @xmath204 by the subset of tuples which are unbroken strings of non - zero entries .
[ tenstoric ] the algebra @xmath205 is presented by @xmath206 with respect to the swap relations on these generators .
this gives a presentation of @xmath205 , in section [ pr ] we will describe an explicit affine semigroup @xmath177 such that @xmath207 = r_{\tree_0}(a , b).$ ] the elements @xmath186 , @xmath188 , and @xmath191 from weyl s presentation correspond to minimal generators in the degeneration @xmath208 as follows .
for the set @xmath187,$ ] we let @xmath209 be the tuple with @xmath210 exactly for @xmath211 and we define @xmath212 for @xmath190 $ ] similarly .
we let @xmath213 be the tuple with @xmath210 , @xmath214,and all other differences equal to @xmath42 .
these elements constitute a proper subset of @xmath206 , so the presentation emerging from the degeneration @xmath205 is perhaps not as efficient as weyl s presentation , however it is better suited to aid in our description of @xmath102 .
the paper is organized as follows . in section [ tfp ] , we review the toric fiber product construction and some of its properties . in section [ pr ] ,
we review the classical pieri rule from the representation theory of @xmath60 in section [ lv ] , we study a valuation on the algebra of tensor product invariants which will help us to understand the semigroup algebra @xmath154 , we prove the @xmath3pieri rule from the classical pieri rule , we prove theorem [ main ] . in section [ gt ] ,
we construct @xmath154 as a set of weightings on an interlacing pattern .
in this section we review the toric fiber product operation on affine semigroup algebras .
this is the `` gluing '' operation we will need to build the affine semigroup algebras @xmath205 , @xmath102 from their component semigroup algebras . to this end , we will pay special attention to toric fiber products of simplicial cones over a simplicial cone base .
we will review results from @xcite and @xcite which control the algebraic behavior of these semigroups . for @xmath215 , @xmath216 , @xmath217 polyhedral cones , and @xmath218 linear maps with @xmath219
the toric fiber product is the set , @xmath220 the fiber product @xmath221 is also a polyhedral cone .
if @xmath222 @xmath223 and @xmath224 are lattices , with @xmath225 , then the set @xmath226 is called the fiber product semigroup of the affine semigroups @xmath227 and @xmath228 from now on we abuse notation and refer to the semigroup by its defining cone .
if the generators of the affine semigroups of these bodies behave well with respect to the maps @xmath229 then the affine semigroup of the associated toric fiber product also behaves well .
the following can be found in @xcite , and a variant appears in @xcite .
[ semigen ] let @xmath230 @xmath231 , and @xmath232 denote the generating sets of the affine semigroups associated to polyhedral cones @xmath233 with respect to lattices @xmath234 suppose that @xmath235 , and that @xmath236 is a simplicial cone , then the set @xmath237 generates the product semigroup @xmath238 for any semigroup element @xmath239 , we can factor @xmath240 , @xmath241 .
we then get factorizations , @xmath242 the affine semigroup @xmath236 has unique factorization , because it is a simplicial cone .
it follows then that @xmath243 and the sets @xmath244 , @xmath245 are equal . the elements of the sets @xmath246 and @xmath247 can be matched in some way @xmath248 , so that the resulting pairs have the same image under @xmath249 .
we therefore obtain a factorization , @xmath250 several of the elements @xmath246 may have the same image under @xmath251 , this could lead to distinct assignments @xmath248 which produce different factorizations of the element @xmath252 these factorizations are related by swap relations .
let @xmath253 @xmath254 with @xmath255 , then the following is a swap relation in @xmath238 @xmath256 [ semirel ] with the assumptions of proposition [ semigen ] , the relations of @xmath257 are generated by those which generate relations for the @xmath258 and the swap relations .
we consider a relation in @xmath221 , @xmath259 once again , the sets images of these two sets of elements in @xmath236 must agree , so @xmath260 .
the word @xmath261 can be transformed to @xmath262 by a series of generating relations in @xmath263 .
we claim that each such relation can be lifted to a relation on generators in @xmath238 if @xmath264 is such a relation , then the image sets under @xmath251 must agree , so each relation can be performed without changing the set @xmath265 .
this implies that at each stage , these sets can be assigned to elements in the set @xmath266 to form elements from the fiber product .
it follows that @xmath267 can be transformed to @xmath268 by relations lifted this way from @xmath263 and @xmath269 where the @xmath270 and @xmath271 are the @xmath272 and @xmath273 matched in a different way .
now this word can be rearranged into @xmath274 using swap relations . in @xcite and @xcite it
is shown that if the ideals defining @xmath275 , \c[p_2]$ ] under presentation by @xmath276 have quadratic , square - free grbner bases , then so does the ideal defining @xmath277 $ ] .
in particular , if @xmath233 are simplicial cones with @xmath235 , the ideal defining @xmath277 $ ] has a quadratic , square - free grbner basis made of swap relations .
from this we can deduce the following proposition .
[ kos ] let @xmath278 be a fiber product of simplicial cones over simplicial cones , by maps which take generators to generators . then the swap relations define a quadratic , square - free grbner basis on the defining ideal of @xmath279 $ ] under the presentation by the fiber product generating set .
affine semigroup algebras are appealing because their inherent combinatorial nature allows many of their commutative algebra properties to be expressed in therms of polyhedral geometry . the following is corollary @xmath280 of @xcite .
let @xmath10 be an affine semigroup algebra , with @xmath281 the set of interior lattice points .
the algebra @xmath282 $ ] is gorenstein if and only if @xmath283 for some @xmath284 in @xcite , we gave the following condition for a fiber product semigroup algebra of gorenstein semigroups to be gorenstein .
the conditions of this proposition are satisfied for the semigroups we consider in this paper , and indeed are satisfied for any fiber product of simplices over simplices , where the maps in question are surjective on generating sets .
[ gor ] let @xmath233 be semigroups with gorenstein semigroup algebras , and let @xmath285 be the corresponding generators of the interior sets
. then if @xmath286 , the semigroup algebra @xmath221 is gorenstein .
in this section we decribe the algebra @xmath179 in terms of the representation theory of @xmath287 and an associated flat degeneration to an affine semigroup algebra @xmath207 $ ] . in order to define the affine semigroup @xmath177 we will recall the pieri rule for decomposing tensor products of @xmath0 representations , and describe how this rule informs the combinatorial commutative algebra of @xmath288 first we introduce the algebra @xmath289 of invariant vectors in @xmath290fold tensor products of irreducible @xmath0 representations .
we let @xmath291 denote the cox ring of the full flag variety of @xmath0 , as a vector space , this algebra is a multiplicity - free direct sum of all irreducible representations of @xmath0 .
@xmath292 the multiplication operation in this algebra is given by projecting on the highest weight component of a tensor product .
@xmath293 this algebra is multigraded by the dominant weights @xmath294 and can also be constructed as the algebra of invariants in @xmath295 $ ] with respect to the unipotent group of upper triangular matrices .
we define @xmath289 to be the algebra of @xmath0 invariants in an @xmath290fold tensor product of copies of @xmath291 . as a vector space ,
this is a direct sum of all invariant spaces in @xmath290fold tensor products of @xmath0 representations .
@xmath296^{sl_m(\c)}\\\ ] ] the coordinate rings of @xmath2 and @xmath1 are graded subalgebras of @xmath291 , they can be identified with the sums of those representations of the form @xmath297 and @xmath298 , respectively .
it follows that the algebra @xmath179 is a multigraded subalgebra of @xmath299 . in @xcite and @xcite
we describe a flat degeneration of @xmath289 for each trivalent tree @xmath38 with @xmath39 leaves . in @xcite
, it is shown that the lift of these degenerations to @xmath300 $ ] is compatible with the corresponding degeneration on @xmath27 under the inclusion map in equation [ inc ] .
in particular , for the tree @xmath67 , @xmath289 degenerates to the algebra of invariants @xmath301^{t^{n-3}}$ ] , where @xmath302 is a product of @xmath303 copies of the maximal torus @xmath304 acting as in theorem [ degen ] . in order to pass this description to @xmath179 ,
we define four subalgebras of @xmath305 .
the algebras @xmath306 and @xmath307 are the multigraded subalgebras of @xmath305 whose componants have middle dominant weight equal to a multiple of @xmath308 and @xmath309 , respectively .
the algebras @xmath310 and @xmath311 are defined accordingly .
we call the algebras @xmath312 @xmath313 @xmath314 and @xmath311 the pieri algebras . theorem @xmath176 from @xcite implies that @xmath179 has a flat degeneration to the following invariant algebra . @xmath315^{\otimes a-2 } \otimes [ r_3(b , p^ * , b)]^{\otimes b-2 } \otimes r_3(b , p^ * , p^*)]^{t^{a+b-3}}\\\ ] ] here , once again , @xmath316 is a product of @xmath317 copies of the maximal torus of @xmath0 , and taking invariants by @xmath100 ensures that all graded components of @xmath205 are of the following form .
@xmath318 @xmath319 @xmath320 + the following is the classical pieri rule , it is a basic component of @xmath0 representation theory .
we refer the reader to any basic text covering this subject , for example the book of fulton and harris , @xcite .
[ p2][the pieri rule ] the invariant space @xmath321 is multiplicity free , and it has dimension @xmath123 precisely when there is a @xmath47 weight @xmath124 such that the following hold
. 1 . @xmath125 + 2 .
@xmath126 + 3 .
@xmath128 + weights which satisfy these conditions can be arranged in interlacing patterns , as depicted in figure [ pieri2 ] .
@xmath322 + interlacing conditions governing the spaces @xmath323 can be derived by duality . for this space to have dimension @xmath123
, there must be a @xmath47 weight @xmath142 such that the following hold , see figure [ pieri2dual ] .
1 . @xmath324 + 2 .
@xmath325 + 3 .
@xmath146 + @xmath326 + we define boundary maps @xmath150 for these patterns as in the @xmath3pieri case .
the algebra @xmath310 is composed of components of the form @xmath327 , following the recipe above , the interlacing inequalities give us patterns as in figure [ pieri3 ] .
dual patterns are depicted in figure [ pieri4 ] for spaces of the form @xmath328 .
@xmath329 + @xmath330 + notice that both of these diagrams are determined by @xmath114 parameters .
since each multigraded component of the pieri algebras is dimension @xmath42 or @xmath123 , these algebras are all affine semigroup algebras , we refer to their underlying semigroups by @xmath331 @xmath332 @xmath333 and @xmath334 .
the following proposition shows that each of these semigroups has a simple structure .
[ cpresent ] the algebras @xmath306 and @xmath307 are polynomial algebras on @xmath335 variables .
the algebras @xmath310 and @xmath311 are polynomial algebas on @xmath114 variables .
duality is a linear operation on dominant weights , so it follows that the statement for the algebras @xmath307 and @xmath311 is a consequence of the statement for @xmath306 and @xmath310 respectively , so we present proofs for these cases . from figure [ pieri3 ]
we directly compute that the algebra @xmath310 is generated by the patterns in figure [ pieripattern1 ] .
@xmath336 + any element of @xmath310 is a unique product of the elements reprsented by these patterns .
we label these elements by recording their boundary fundamental @xmath47 weights , from top down they are @xmath337,$ ] @xmath338 $ ] , and @xmath339 $ ] .
the identity @xmath340 corresponds to the trivial invariant in @xmath341 , we label this element with @xmath342.$ ] in general , we let @xmath343 $ ] and @xmath344 $ ] denote the patterns in figure [ pieripattern2 ] . @xmath345 + an interlacing pattern @xmath135 is decomposed uniquely into a sum of these patterns by finding the smallest non - zero entry , and pulling the pattern above with the first @xmath123 occuring at this entry off until this entry is @xmath42 .
@xmath346 + @xmath347 + ( b_{m-1 } - a_{m-1})[m-1 , m-1 ] + \ldots + ( a_1 - \sum_{j > 1 } a_j + b_{j-1 } ) [ 1 , 0]\\\ ] ] the indices @xmath348 in the generators @xmath349 $ ] of these semigroups range between @xmath123 and @xmath49 . as @xmath49 corresponds to the trivial dominant weight @xmath350 , we consider these elements as members of @xmath351 in particular @xmath352 = [ m , m].$ ] the multigraded components of @xmath205 all have dimension @xmath123 or @xmath42 , and have dimension @xmath123 exactly when each part of the tensor product satisfies the conditions of proposition [ p2 ] .
we use this fact to describe an affine semigroup @xmath177 such that @xmath353.$ ] we define @xmath177 to be the affine semigroup obtained by taking the following fiber product over the boundary maps @xmath150 .
the image of these boundary maps is the affine semigroup for the simplicial cone @xmath354 with @xmath48 generators .
@xmath355 @xmath356 @xmath357 the boundary maps @xmath150 were defined to capture the operation of taking invariants by the torus @xmath358 , so the semigroup algebra of @xmath177 can be identified with @xmath205 by definition .
proposition [ semigen ] then implies that elements composed of generators of pieri semigroups suffice to generate @xmath177 , and therefore @xmath203 the semigroups @xmath359 , @xmath360 , @xmath332 and @xmath334 are generated by elements labelled by pairs @xmath349 $ ] , where @xmath348 are elements of @xmath168 .
it is easily checked from the definition of the boundary maps @xmath150 that two of these generators @xmath349,[k , l]$ ] can be glued when @xmath361 .
this implies that a generating set of @xmath177 is given by @xmath362 tuples of elements from @xmath351 @xmath363\\\ ] ] where @xmath364 and @xmath365 can be @xmath49 or @xmath48 , the first @xmath366 differences @xmath367 must be @xmath42 or @xmath123 and the last @xmath368 differences must be @xmath42 or @xmath369 all of these conditions holding modulo @xmath370 by proposition [ semirel ] the relations on these generators are generated by the swap relations .
since the element @xmath371 $ ] represents @xmath372 , the swap relations tell us that an element of the form @xmath373 $ ] may be factored , @xmath374[0 , \ldots , 0 , \ldots , 0 ] = [ i_1 , \ldots , i_{k+2 } , 0 , \ldots , 0][0 , \ldots , 0 , i_{k+2 } , \ldots , i_{a + b -1}]\\\ ] ] from this it follows that @xmath177 is generated by unbroken strings of nonzero entries from @xmath375
in this section we study the four @xmath3pieri algebras , @xmath376 @xmath113 @xmath377 @xmath378 we will use the results of the previous section to prove that each of these algebras is a polynomial ring , and then use this fact to find a presentation of the algebra @xmath112 conformal blocks @xmath379 in the genus @xmath42 case can be recovered as a subspace of the space of @xmath380invariant vectors in the tensor product @xmath381 . the precise subspace depends on the data @xmath382 the moduli of @xmath383marked projective lines is a single point , so in this case there is a unique space of conformal blocks , @xmath384 , which can be explicitely defined as a subspace of @xmath385^{sl_m(\c)}$ ] using elements of the classical representation theory of @xmath60 we consider the copy of @xmath41 inside @xmath0 which corresponds to the longest root @xmath386 each representation @xmath387 can be decomposed into isotypical components of this subgroup @xmath388 in @xcite , ueno defines the following subspace @xmath389 .
@xmath390 the following proposition can also be found in @xcite .
[ ueno ] the space of conformal blocks @xmath384 can be identified with the space @xmath391^{sl_m(\c)}.$ ] proposition [ ueno ] can be used to define a function @xmath392 as follows . on pure graded components of @xmath305
we define @xmath393 on @xmath394 using the subspace @xmath395 , @xmath396 this is extended to mixed grade components using the @xmath397 function , @xmath398 by convention we set @xmath399 this function was studied in @xcite , where it was shown to be a valuation on @xmath305 , see also page @xmath400 in @xcite .
this means that @xmath393 has the following properties .
1 . @xmath401 + 2 .
@xmath402 + 3 .
@xmath403 + in terms of this function , the graded components of @xmath85 are the following subspaces of the graded components of @xmath404 @xmath405^{sl_m(\c)}\\\ ] ] there are corresponding inclusions of the @xmath3pieri algebras into polynomial rings over the pieri algebras .
@xmath406\\\ ] ] @xmath407\\\ ] ] @xmath408\\\ ] ] @xmath409\\\ ] ] now our task is to establish explicitely the @xmath393 value for a non - trivial vector in each graded component of the pieri algebras , this will allow us to compute a presentation for the @xmath3pieri algebras . in order to do this , we find @xmath410 for each generator @xmath100 of @xmath306 , @xmath307 , @xmath314 and @xmath411 and use this calculation to find the @xmath393 value for each monomial in the generators .
the generators of the pieri algebras are precisely the unique invariants in tensor products of the form @xmath412 and @xmath413 where @xmath414 and @xmath415 where @xmath416 note that the sum of these indices must be equal to @xmath160 or @xmath370 both invariant types can be viewed as a representation of the determinant , and have the following form as alternating tensors . @xmath417 @xmath418 @xmath419 here @xmath420 is a wedge product of basis vectors over the indicated index set , and @xmath421 is a sign function .
the following is well known in the type @xmath422 conformal blocks literature .
each generating invariant of the pieri algebras has @xmath393 value equal to @xmath122 to evaluate @xmath393 on these elements , we reduce to the case where the sum of the indices is @xmath423 as the @xmath160 case is related to this case by duality . in this case , each of the terms in the expansion above is of the form @xmath424 with @xmath425.$ ] when this term is considered with respect to the @xmath426 subgroup isomorphic to @xmath427 its properties are determined by whether or not the indices @xmath428 show up in the sets @xmath429 there are two possibilities .
if @xmath123 and @xmath39 are in separate index sets , the term corresponds to an invariant in the @xmath41 representation @xmath430 and therefore has @xmath393 value @xmath431 .
if both @xmath123 and @xmath39 appear in the same index set , the term is the trivial invariant in @xmath432 , and has @xmath393 value @xmath42 .
the value on the sums of these terms is therefore @xmath122 we focus briefly on a particular @xmath3pieri algebra , @xmath87 .
we associate each generator @xmath343 , [ i+1 , i]$ ] to the elements @xmath343 t , [ i+1 , i]t \in v_{0 , 3}(b , p , b ) \subset r_3(b , p , b ) \otimes \c[t]$ ] . now consider a non - trivial element @xmath433 the invariant represented by @xmath135 has a monomial decomposition into the generators above .
@xmath434^{n_i}[j , j]^{m_j}\\\ ] ] multiplicative properties of valuations then give us @xmath435 .
this implies that in the @xmath3pieri algebra we have @xmath436t)^{n_i } ( [ j , j]t)^{m_j}t^{k - \sum n_i - \sum m_j}.\\\ ] ] the same argument applies for the other @xmath3pieri algebras . as a consequence we obtain the following theorem .
the algebras @xmath87 and @xmath95 are polynomial algebras on @xmath160 generators .
the algebras @xmath153 and @xmath86 are polynomial algebras on @xmath162 generators .
the generators of these algebras are the generators of the corresponding pieri algebras , along with the identity @xmath342 $ ] , multiplied by the auxiliary variable @xmath437 .
this also gives the following characterization of the non - zero spaces @xmath438 the space of conformal blocks @xmath439 is multiplicity free , and dimension @xmath123 if and only if the conditions in proposition [ kpieri ] hold .
this follows from equation [ decomp ] and the fact that the @xmath393 of the patterns @xmath343 , [ i+1 , i]$ ] are given by the first entry on the top row .
@xmath346 + @xmath347 + ( b_{m-1 } - a_{m-1})[m-1 , m-1 ] + \ldots + ( a_1 - \sum_{j > 1 } a_j + b_{j-1 } ) [ 1 , 0]\\\ ] ] we are now in a position to construct the affine semigroup @xmath440 as polynomial algebras , the @xmath3pieri algebras @xmath376 @xmath113 @xmath377 and @xmath441 are affine semigroup algebras on simplicial cones , with @xmath442 generators respectively .
we construct @xmath154 as a toric fiber product of the corresponding semigroups following a procedure similar to the construction of @xmath443 the only extra detail is that , in addition to the boundary maps @xmath150 on the interlacing patterns for these algebras , we must also include the level data .
this implies that two interlacing patterns can be glued only if they share the same boundary weight data and level .
the semigroup @xmath154 is obtained as fiber product over the affine semigroup of the simplicial cone @xmath444 with @xmath49 generators .
@xmath445 @xmath446 @xmath447 + by proposition [ semigen ] , a generating set of @xmath154 is obtained by taking all tuples of level @xmath123 generating elements @xmath448 , \ldots , [ i_{a+b-2 } , j_{a + b -2}]$ ] , where @xmath449 .
note that , as all generators are tuples of level @xmath123 elements , and any product of level @xmath123 elements is a level @xmath450 element , this is already a minimal generating set of @xmath440 by the descriptions of the pieri algebra generators above , the generators of @xmath154 are exactly those @xmath167 tuples which satisfy the following properties , all elements are members of @xmath375 1 .
the difference @xmath451 is @xmath452 for @xmath171 .
the difference @xmath451 is @xmath453 for @xmath454 .
the first and last entries are members of @xmath455 + by proposition [ semirel ] , the relations for @xmath154 are generated by the swap relations among these generators .
this gives the presentation of @xmath456 and completes the proof of theorem [ toricmain ] .
note that propositions [ gor ] and [ kos ] also imply that both @xmath102 and @xmath205 are gorenstein algebras with quadratic , square - free grbner bases .
this in turn implies the following .
the algebras @xmath179 and @xmath27 are gorenstein , koszul algebras when @xmath37 is generic .
furthermore , since @xmath27 is graded by the level data , and the only elements of level @xmath42 are multiples of the identity @xmath457 , the level @xmath123 elements constitute a minimal generating set of this algebra .
in this section we describe elements of the fiber product semigroups @xmath154 and @xmath177 as labellings of interlacing diagrams .
we have defined the interlacing patterns for the @xmath3pieri and pieri semigroups and their boundary maps @xmath150 specifically with the gluing operation in mind .
we slightly modify these boundary maps , after taking @xmath458 , or @xmath459 , we always then pass from a positive @xmath47 weight to the underlying @xmath0 weight .
we consider a tensor product , @xmath460 . by the pieri rule ,
this space is non - zero if and only if there are two interlacing patterns @xmath461 such that @xmath462 @xmath463 @xmath464 and @xmath465 .
these conditions can be represented with a @xmath466step interlacing diagram with entries as depicted in figure [ pieri5 ] .
notice that the top two rows of this diagram have been modified from what they would have been had we considered the space @xmath468 by itself . in particular , the the following diagram has been added to the top two rows .
this does not change the boundary values of the pattern as it represents the trivial tensor product invariant in @xmath470 .
this operation is performed to make the gluing procedure easier , namely one pattern is modified until its boundary weights match , which is always possible when @xmath471 .
then it is simply stacked on top of the lower pattern .
patterns as in figure [ pieri5 ] label the graded components of the algebra @xmath472^t.$ ] to represent more complicated tensor products , we simply stack the appropriate patterns , taking care to modify pairs of rows by the null pattern to correctly identify the weights when necessary .
we define @xmath473 to be the `` wedge '' interlacing pattern in figure [ wedge ] , with rows of length @xmath49 , @xmath17 steps before the switch in direction , and @xmath19 steps after the switch in direction .
the next two propositions then follow from our combinatorial descriptions of the algebras @xmath205 and @xmath102 .
the level condition in proposition [ kpieri ] has been altered to account for the possible addition of the null diagram to a pair of rows . 1 .
the boundary value of the top row is @xmath476 .
the boundary value of the bottom row is @xmath477 .
the difference between the sums of the @xmath478 and @xmath479st rows is @xmath480 if @xmath481 and @xmath482 is @xmath483 + 1 .
the boundary value of the top row is @xmath476 .
the boundary value of the bottom row is @xmath477 .
the difference between the sums of the @xmath478 and @xmath479st rows is @xmath480 if @xmath481 and @xmath482 is @xmath483 + 4 . for rows
@xmath485 , the difference between the first entry of row @xmath478 and the last entry of row @xmath486 is @xmath487 .
f rows @xmath488 , the difference between the last entry of row @xmath478 and the first entry of row @xmath479 is @xmath489 + the labellings in proposition [ levelgt ] are the lattice points in a polytope @xmath163 . the graded affine semigroup algebra determined by this polytope ,
@xmath490 $ ] is the degeneration of the projective coordinate @xmath491 induced from the degeneration of @xmath27 to @xmath112 this algebra is considerably more complicated than @xmath456 it would be interesting to find a generating set of lattice points for this semigroup , or a classification of which @xmath492 yield a gorenstein algebra . | we study a toric degeneration of the cox ring of the moduli of principal @xmath0 bundles on the projective line , with quasi parabolic data given by the the stabilizer of the highest weight vector in @xmath1 and its dual @xmath2 .
the affine semigroup algebra resulting from this degeneration is described using the @xmath3pieri rule from kac - moody representation theory . along the way we give a proof of the @xmath3pieri rule which utilizes the classical pieri rule and elements of commutative algebra , and we describe a relationship between the cox ring and a classical invariant ring studied by weyl .
[ multiblock footnote omitted ] keywords : conformal blocks , pieri rule , toric degeneration , invariant theory |
the `` reactor anti - neutrino anomaly '' is the observation that the average of the experimentally determined reactor anti - neutrino flux at reactor - detector distances less than 100 m accounts for only 0.946 @xmath1 0.023 of the theoretical expectation @xcite .
one of the explanations for this anomaly is that standard neutrinos that carry the weak nuclear charge can oscillate into a
sterile " neutrino that does not contain a weak nuclear charge and escapes detection .
this is one of the few types of experiments that could be sensitive to the sterile neutrinos .
it is important to verify that the shape of the calculated anti - neutrino spectrum is correct . in @xcite
the effects of various corrections to the expected neutrino spectra were examined , but it was found that these corrections could not explain the anomaly . however , in @xcite only allowed decays were analyzed and the first forbidden(ff ) decay was left out . in @xcite the contribution of ff decays were examined by assuming some virtual ff branches in the decay , and it was found that the results could match the magnitude of the missing flux
. however , the actual ff decays are much more complicated than the form assumed in @xcite . in this work
, we will exam the effect of realistic ff decay from microscopic nuclear structure calculations , and make an estimation of their effect on the neutrino spectra .
for the neutron - induced fission accumulated yield distributions one finds two peaks for the fission products .
one of these is centered near @xmath2sr and another is centered near @xmath3xe . both or these regions could have comparable amount of ff branches .
for the region of @xmath3xe calculations with the quasi - particle random phase ( qrpa ) method were carried out @xcite , and reasonable agreement between experiment and theory for half - lives and @xmath4 ft values were obtained . also shell - model ( sm ) calculations for some nuclei in the xe region
can be carried out in a moderately large basis with previously derived hamiltonian .
thus , in this work we will focus on applying the qrpa and sm methods in this xe region to investigate the effect of ff beta decay on the shape of the neutrino spectra .
this article is arranged as follows .
first we give some background on @xmath0-decay , especially the ff decays , and the nuclear structure theories we used in our calculations .
then we present the calculated results with comparisons to experiments and the corresponding neutrino spectra .
conclusions are given at the end .
the decay rate for @xmath0-decay can be written generally as @xcite : @xmath5 with the conventions and numerical constants used in @xcite one obtains : @xmath6 here @xmath7 is the energy of the emitted electron in the units of electron mass , @xmath8 is the @xmath0-decay energy in the unit of electron mass , and @xmath9 is the momentum of the electron .
@xmath10 is the fermi factor which takes into account of nuclear charge on the shape of the spectra for the emitted electron .
the nuclear structure dependence on the shape of the emitted leptons is contained in @xmath11 .
it has different @xmath12-dependencies for different kinds of decays that lead to the different spectra for emitted electron and neutrino . for allowed - decay ,
@xmath11 is independent of @xmath12 . for the ff decay
, the dependence can be written in the form @xcite : @xmath13 for ff decays one has three different types of transitions associated with the change of spins , @xmath14 , they have different matrix elements and @xmath12 dependencies : @xmath15 the detailed expressions for the @xmath16 s can be obtained from @xcite . for 0@xmath17 , there are three matrix elements @xmath18 , @xmath19 and @xmath20 , for 1@xmath17 one has five matrix elements involving @xmath21 , @xmath22 , @xmath23 , @xmath24 and @xmath25 , and for 2@xmath17 just one matrix element @xmath26 is involved . the expressions for these matrix elements
are given in @xcite .
in @xcite only the @xmath18 , @xmath21 , @xmath23 and @xmath26 terms were used for ff branches .
our additional terms result in some differences between our results and those of @xcite . to get the electron or neutrino spectra , we take derivatives over the respective energies : @xmath27 the spectra for ff decays are different from that of allowed gt , and their shape depends on the decay modes ( @xmath28 ) . to obtain the spectra we need to know some detailed structure information for the @xmath0-active nuclei . for the nuclear structure calculations ,
the configuration interaction model or shell model ( sm ) provides an exact solution within a model space for a restricted set of valence orbitals .
realistic shell - model hamiltonians can be derived from renormalized interactions based on the nucleon - nucleon interaction with some empirical single - particle energies and modifications to reproduce experimental binding energies and excitation energies . however , as the number of valence nucleons increase , the dimensions of the configurations increase drastically making the calculations impossible . starting with a closed shell of @xmath29sn one
add nucleons in the @xmath30 " model space that consists of the five ( @xmath31 , @xmath32 , @xmath33 , @xmath34 , @xmath35 ) orbitals for protons and the six ( @xmath36 , @xmath37 , @xmath38 , @xmath39 , @xmath40 , @xmath41 ) orbitals for neutrons .
we can consider up to four neutrons and four protons in this @xmath30 model space .
the sm can be applied to the decays of nuclei with both even and odd numbers of protons or neutrons . to obtain results over a wider region of the nuclear chart
, one needs to use various approximations .
one of these is the quasi - particle random phase approximations ( qrpa ) which assumes the excited states of the nuclei are small harmonic oscillations beyond the hartree - fock - boglyubov ( hfb ) or bcs ground states .
only two quasi - particle excitations are considered in this approximation . by changing one neutron to one proton or _ vice versa _ , we obtain the spectra for odd - odd nuclei , this is the so - called pn - qrpa method @xcite which is usually used for charge exchange reactions as well as @xmath0-decay . the qrpa method can only be applied to even - even nuclei .
[ cols="^,^,^,^,^,^,^,^,^,^,^,^ " , ] [ singsum ] as we have stated above , different decay channels may have different shapes due to different dependencies over energy @xmath12 , so we need to investigate the effects of these decay channels on the neutrino spectra shape . for the odd - odd or odd -
a nuclei there is usually mixing between different decay channels as @xmath42 , but for even - even nuclei , because the ground states of the parent nuclei has always @xmath43 , @xmath44 is unique for specific final state of daughter nuclei , there will be no mixing among different channels and it is easy to isolate different shape changes in different decay channels . in fig.[sgl ] , we compare the neutrino spectra shape changes relative to the allowed shape for different channels with different methods for two even - even nuclei ( @xmath45te and @xmath3xe ) . for each nucleus
we show the 0@xmath17 , 1@xmath17 and 2@xmath17 decay branches .
the sm and qrpa methods agree well with each other . for @xmath46 decays ,
the change of the spectra is small and it is a good approximation to treat the @xmath46 decay as allowed decay . for @xmath47 decay
the change is large with the peak of the neutrino spectra shifted downwards .
this means that more neutrinos have less energy than expected from the previous simulation@xcite using the allowed type of phase space . for @xmath48 decay
the behavior of the change to the shape is a bit different from that of @xmath47 as seen from fig.[sgl ] where the shape of the neutrino spectra for this decay branch is broadened .
we also make a comparison of the full microscopic calculations to the approximations made in @xcite where 4 out of 9 matrix elements are used ( affecting the 0@xmath17 and 1@xmath17 decays ) .
for the 0@xmath17 decay , the approximation used in @xcite gives a result that is opposite to the full microscopic calculations , slightly shifting the neutrino spectra to lower energy . for 1@xmath17 decay
, the approximation completely changes the behavior of the neutrino spectra . due to
the over simplified forms in @xcite , the behavior of an overall shift of spectra to low energies disappears now .
this comes from the fact that for simplified 1@xmath17 decay in table i of @xcite one of its matrix elements ( @xmath49^{1-}$ ] or @xmath21 in this work ) has the same form as that for 2@xmath17 decay ( @xmath49^{2-}$ ] or @xmath26 in this work ) .
there is similar behavior between @xmath45te and @xmath3xe .
we would also expect the same behaviors of these ff decay channels in odd - mass or odd - odd nuclei since they have the same transition operators as the even - even nuclei . from the above results ,
we conclude that the inclusion of ff decays could eliminate the reactor anti - neutrino anomaly " if there are enough beta branches containing 1,2@xmath17 transitions with suitable end - point energies , especially 1@xmath17 .
however , if we examine the nuclear chart for the decay branching ratios , we find that 1,2@xmath17 are usually accompanied with @xmath46 decays which usually have a much smaller @xmath4 ft values ( a stronger transition probability ) .
this would reduce the overall changes to the spectra . to quantify the change in the neutrino spectrum due to the change of phase space , we integrate over the spectra with the two phase spaces as follows , @xmath50 with @xmath51 .
@xmath52 is the energy needed to trigger the interaction @xmath53 , and @xmath54 is the maximum energy of emitted neutrinos .
the reduction in the number of low - energy neutrinos is given by @xmath55 .
the change depends on the end point energy @xmath54 , which can be expressed as @xmath56 .
so we need precise excitation energies for the determination of neutrino spectra .
this result can then be compared with the value of the reactor neutrino anomaly to see if the lack of ff phase space factor in the simulation can explain the missing neutrinos .
the results for single decay branches are listed in table[singsum ] .
a comparison between qrpa and shell model shows similarities for the ratio @xmath57 , this agrees with fig.[sgl ] . for the detailed values ; the change @xmath58 of the @xmath46 decay is negligible , for 1@xmath17
, @xmath58 goes up to ten percent , and for 2@xmath17 , @xmath58 is only 2 - 3 percent .
( defined in text ) on the end - point energies for several ff decay branches of @xmath45te(bold lines for qrpa calculations and dashed lines for sm calculations ) and @xmath3xe(dashed - dot lines for qrpa calculation ) . here
we varying the q values of the two nuclei above to see how the changes are related to the end - point energies of the decay branches . ] to obtain quantitative results on the dependence of the detailed changes on the end point energies of the decay branches , we vary the q values in the calculations for the two nuclei @xmath45te and @xmath3xe . the results are plotted in fig.[dn ] where one observes that to a large extent this relation is nucleus independent .
@xmath58 for the @xmath46 branches are near zero except below end - point energies of 3 mev . for small end - point energies , @xmath58 is large due to the shape changes at the spectra tail ; but these are not important since contributions of these branches to the total spectra are small , see fig.3 of @xcite . for @xmath48 decay
the dependence of @xmath57 on the end - point energies are independent of @xmath4 ft values since it has only one component . for end - point energies from 4 - 6 mev , @xmath58 is around @xmath59 . however , for @xmath47 decays @xmath57 depends on both @xmath54 and @xmath4 ft . to see this we also plot the @xmath60 decay branches for the two nuclei fig.[dn ] .
compared to @xmath61 the ff decays to the @xmath60 states have smaller @xmath4 ft values ( table.[ffrt ] ) ( i.e. they are stonger ) and have smaller @xmath58 values ( fig.[dn ] ) .
the reason of this comes from the fact that the transition rates of @xmath47 are determined by five different components .
they are combined to give the final decay rates , and their different combinations have different energy dependencies . at @xmath62 mev
, @xmath58 is @xmath63 .
it was estimated in @xcite that @xmath64 of the decay branches of the fission products are ff .
thus , in the most extreme case where the ff is dominated by @xmath65 the change of the neutrino spectrum could be as large as @xmath66 .
in this work , explicit analysis of @xmath0-decay neutrino spectra with inclusion of the first forbidden part has been performed .
one finds that use of the allowed decay phase space factor results in a correction of up to about @xmath67 due to @xmath68 ff transitions .
an average over all types of ff transitions , end - point energies and @xmath4 ft values would result in a smaller value of @xmath69 .
the finite size effects and the weak magnetism corrections obtained in @xcite for the allowed ( gt ) decays are estimated to be @xmath70 .
if the average branching ratios for all types of ff is estimated , they can be combined with our results to obtain an improved correction for the shape of the neutrino spectra .
99 y. abe _ et al . _
[ double - chooz collaboration ] , phys .
* 108 * , 131801 ( 2012 ) g. mention , m. fechner , t. lasserre , t. a. mueller , d. lhuillier , m. cribier and a. letourneau , phys .
d * 83 * , 073006 ( 2011 ) .
t. .a .
mueller , d. lhuillier , m. fallot , a. letourneau , s. cormon , m. fechner , l. giot and t. lasserre _ et al .
_ , phys .
c * 83 * , 054615 ( 2011 ) p. huber , phys .
c * 84 * , 024617 ( 2011 ) [ erratum - ibid .
c * 85 * , 029901 ( 2012 ) ] a. c. hayes , j. l. friar , g. t. garvey , g. jungman and g. jonkmans , phys .
lett . * 112 * , 202501 ( 2014 ) .
d. -l .
fang , b. a. brown and t. suzuki , phys .
c * 88 * , 034304 ( 2013 ) t. suzuki , t. yoshida , t. kajino and t. otsuka , phys .
c * 85 * , 015802 ( 2012 ) q. zhi , e. caurier , j. j. cuenca - garcia , k. langanke , g. martinez - pinedo and k. sieja , phys .
c * 87 * , 025803 ( 2013 ) j. a. halbleib and r. a. sorensen , nucl .
a*98 * , 542(1967 ) b. a. brown and w. d. m. rae , nuclear data sheets * 120 * , 115 ( 2014 ) .
b. a. brown , n. j. stone , j. r. stone , i. s. towner and m. hjorth - jensen , phys . rev .
c * 71 * , 044317 ( 2005 ) ; erratum , phys .
c * 72 * , 029901(e ) ( 2005 ) .
national nuclear data center , information extracted from the chart of nuclides database , http://www.nndc.bnl.gov/chart/ | we examine the effect of first forbidden ( ff ) decays on @xmath0-decay neutrino spectra by performing microscopic nuclear structure calculations . by analyzing the ff decay branches of even - even nuclei
we conclude that ff decays may be responsible for part of the missing neutrinos in the so called `` reactor neutrino anomaly '' . further calculations and more experimental data are needed for a firm conclusion . |
the one impurity spin degenerated anderson model @xcite has been studied by different many bodies techniques , such as green function , renormalization group , bethe ansatz , quantum monte carlo and combination of quantum monte carlo with the of maximum entropy method .
the static and dynamic properties was obtained , such as specific heat , magnetic susceptibility , photoemission and nuclear magnetic resonance .
the electron spin resonance ( esr ) of a probe with a well deffined magnetic moment added to that model remains to be better studied , which is the main point of the present work .
this model has been useful to analyze the behavior of heavy fermions and intermediate valence compounds so that the calculation of the esr can contribute to a better understanding of the experimental results of the esr in those compounds .
the esr of a magnetic impurity embedded in a heavy fermion compounds results in an enhancement in the korringa rate krug .
however , the results for the intermediate valence compounds are controverted .
gambke _ at al _
@xcite associated the reduction in the relaxation rate of gd in cepd@xmath12 to the intermediate valence state .
on the other hand , heirinch and meyer @xcite have observed the opposite phenomena in cebe@xmath13:gd , as well barberis _ at al _
@xcite in ceir@xmath14:nd and rettori _ at al _
@xcite in ybincu@xmath15:gd . in the anderson model
the configuration of the anderson ion can be changed continuously from doubly occupied to empty orbital , which allows us to analyze the effect of the anderson impurity orbital occupation number on the esr relaxation rate , from the doubly to the singly occupied orbital regimes , passing by the valence fluctuation regime . in the present work we hve applied the numerical renormalization group ( nrg ) method of wilson @xcite to research the relaxation rate t@xmath0 of a magnetic probe embedded in a host represented by the single impurity spin degenerate anderson model .
originally wilson discretized the conduction band defining the sequence @xmath16 ( where @xmath17 is the half - width of the conduction band , @xmath18 and @xmath19 ) as a discrete set of conduction states .
the anderson hamiltonian is projected onto the basis formed by this set and the anderson ion states and is diagonalized iteratively and the thermodynamic average over those discrete eigenvalues are smooth functions of temperature .
the magnetic probe relaxation rate t@xmath11 , however , in a consequence of the fermi golden rule energy conservation , shows only discrete line transitions . to overcome this difficulty we discretize the conduction band by the sequence @xmath20 , @xmath21 , where @xmath22 is a continuous parameter fixed arbitrarily inside intervals ( @xmath23 , @xmath24 ) , with @xmath25 , as was done in references @xcite and makoto
the anderson hamiltonian is projected onto the basis formed by the new set and the impurity states , and is diagonalized iteratively .
continuum spectra are obtained by averaging numerically the t@xmath0 over @xmath22 , at each temperature , in the range @xmath26 the present paper is organized as follows : section ii details the model and describes the relaxation rate ; section iii develops the formalism to obtain the relaxation rate from the fermi golden rule ; section iv shows the results and discussion of the relaxation rate , and section v contains the conclusions .
the model is represented by the single impurity anderson model @xmath27 and a magnetic probe sitting at @xmath28 , interacting with the anderson ion , sitting at @xmath29 , via the rkky interaction @xmath30 @xcite .
the single impurity anderson model consists of the kinetic energy of the free electrons ( @xmath31 ) , the bound energy of the anderson ion orbital ( @xmath32 ) and the hybridization of this orbital with the conduction band ( @xmath33 ) , @xmath34 the hamiltonian of the conduction electrons is written as @xmath35 where @xmath36 creates ( annihilates ) an electron in the conduction band with energy @xmath37 , spin @xmath38 and momentum @xmath39 , obeying the usual anticommutation relation @xmath40 .
the second term in eq . ( [ model ] ) , corresponding to the anderson ion orbital , is given by @xmath41 where the operator @xmath42 , orthonormal to the operator @xmath43 of the conduction band , creates ( annihilates ) an electron in the anderson ion orbital ( @xmath44 ) with energy @xmath45 and spin @xmath46 , and @xmath47 is the coulomb interaction between the electrons in that orbital .
the hamiltonian @xmath33 in eq .
( [ model ] ) is written as @xmath48 where @xmath49 is the hybridization interaction between the electrons of the @xmath44 orbital and the conduction band electrons . for a moment , if we consider that there is no coupling between the anderson ion orbital and the conduction electrons ( @xmath50 ) , this orbital may be empty ( @xmath51 @xmath52 @xmath53 and energy @xmath54 ) , singly ( @xmath55 and energy @xmath56 ) or doubly ( @xmath57 and energy @xmath58 ) occupied orbital .
the introduction of the interaction @xmath59 allows transitions between those configurations at a rate @xmath60 , via the conduction band , whose density of state is @xmath61 . in this case
the model can represent the intermediate valence ( @xmath62 ) and the kondo ( @xmath63 @xmath64 ) regimes , where @xmath65 is the interconfigurational energy .
the interaction between the magnetic probe and the anderson ion orbital ( @xmath30 ) , that will be treated as a perturbation , is given by @xmath66 \text { , } \label{hx}\ ] ] where @xmath67 is the coupling constant between the magnetic probe and the conduction electrons , and the field operator @xmath68 is written as @xmath69 which annihilates an electron in the wannier state at the site of the magnetic probe ( @xmath28 ) , and @xmath70 is the lowering ( raising ) spin operator of the magnetic probe ion . before the application of the numerical renormalization group ( nrg ) to the hamiltonian given by eq .
( [ model ] ) , we will introduce a basis given by two sets of s - wave operators , one sited at the magnetic probe and the other at the anderson ion , which represent the new states of the conduction band on substitution of the operators @xmath71 . in the present work the conduction band obeys the linear dispersion relation @xmath72 , where the energies and moments are measured in relation to the fermi level and the units
are taken so that @xmath73 . considering the energy as an isotropic function of @xmath74 , _
i.e. _ , @xmath75 , the new operators are defined as @xmath76 where @xmath77 ( @xmath78 ) is the operator which annihilates an electron at the s - wave state of the anderson ion ( magnetic probe ) , with energy @xmath79 and spin @xmath46 , sited at @xmath80 ( @xmath28 ) , obeying the anticommutation relations @xmath81 these operators do not form an orthogonal basis , since @xmath82 where @xmath83 is the fermi momentum and @xmath84 @xmath1is the distance between the anderson ion and the magnetic probe . using the gram - smidth orthogonalization process we define a new operator @xmath85 , orthogonal to @xmath86 , obeying the standard anticommutation relation , @xmath87 \text{,}\ ] ] with @xmath88 . to be used later
, the operators @xmath89 and @xmath90 are defined as @xmath91 where @xmath61 is the state density of the conduction band . in terms of the operators @xmath77 , @xmath92 , @xmath89 and @xmath90 the hamiltonians @xmath33 , @xmath93 and @xmath94 are written as @xmath95 @xmath96 @xmath97\mathbf{i% } _ { -}+h.c,$% \end{tabular } \label{hxnovo}\ ] ] where @xmath98
the conduction eigenstates of the hamiltonian @xmath31 is represented by s - waves with two centers of symmetry , one sited at the anderson ion ( @xmath93 ) , @xmath99 and the other at the magnetic probe ( @xmath100 ) , @xmath101 with @xmath102 following wilson @xcite , both the conduction bands , @xmath93 and @xmath100 , are discretized as a function of the discretezation parameter @xmath18 , and the continue parameter @xmath103@xcite .
the electronic states are written as a complete set of orthonormal functions in the domain @xmath104 .
for the extreme intervals , these function are given by @xmath105 ^{1/2}}\limfunc{e}% \nolimits^{\pm\;i\omega\ell\dfrac{\varepsilon}{d}}$ & for $ \lambda^{-z}<\pm% \dfrac{\varepsilon}{d}<1 $ \\
$ 0 $ & out side,% \end{tabular } \right.\ ] ] where @xmath106 represent the positive or negative values of @xmath107 , @xmath108 is the fourier harmonic index , which takes all integer values in the interval @xmath109 , and @xmath110 is the fundamental frequency . in the internal intervals
@xmath111 ^{1/2}}\limfunc{e}\nolimits^{\pm\;i\omega_{m}\ell \dfrac{\varepsilon}{d}}$ & for $ \lambda^{-\left ( m+1+z\right ) } < \pm \dfrac{\varepsilon}{d}% < \lambda^{-\left ( m+z\right ) } $ \\ $ 0 $ & out side,% \end{tabular } \right.\ ] ] where @xmath112 label the descretization band intervals , and the parameter @xmath113 is the fundamental frequency of the _ m - sime _ interval , given by @xmath114 . the operators @xmath77 and @xmath85 , are expanded in these basis as @xmath115 + \vspace{0.08in}$ \\ & & $ \dsum \limits_{m=0}^{+\infty}\dsum \limits_{\ell=-\infty}^{+\infty}% \left [ a(\bar{a})_{m\ell\mu}\psi_{m\ell}^{\left ( + \right ) } \left ( \varepsilon\right ) + b(\bar{b})_{m\ell\mu}\psi_{m\ell}^{\left ( -\right ) } \left ( \varepsilon\right ) \right ] $ , % \end{tabular } \notag\ ] ] with @xmath116 ^{\ast}c(% \bar{c})_{\varepsilon\mu}$ ; & $ b(\bar{b})_{\ell\mu } = \dint \limits_{-d}^{+d}d\varepsilon\left [ \psi_{\ell}^{\left ( -\right ) } \left ( \varepsilon \right ) \right ] ^{\ast}c(\bar{c})_{\varepsilon\mu}\vspace{0.08 in } $ \\ $ a(\bar{a})_{m\ell\mu}=\dint \limits_{-d}^{+d}d\varepsilon\left [ \psi_{m\ell}^{\left ( + \right ) } \left ( \varepsilon \right ) \right ] ^{\ast}c(% \bar{c})_{\varepsilon\mu}$ ; & $ b(\bar{b})_{m\ell \mu}=\dint \limits_{-d}^{+d}d\varepsilon\left [ \psi_{m\ell}^{\left ( -\right ) } \left ( \varepsilon \right ) \right ] ^{\ast}c(\bar{c})_{\varepsilon\mu}$,% \end{tabular}%\ ] ] where @xmath117 , @xmath118 , @xmath119 , @xmath120 and @xmath121 , @xmath122 , @xmath123 , @xmath124 form , respectively , complete sets of orthonormal operators obeying the standard anticommutation relations @xmath125 replacing the expression for the oparator @xmath126 into the the expression for @xmath93 [ hcnovo ] and eliminating the terms with @xmath127 , approach that was successfully used to calculate the magnetic susceptibility and the specific heat of the kondo model @xcite , the magnetic susceptibility @xcite and the photoemission spectroscopy @xcite of the anderson model , @xmath93 and @xmath100are written as @xmath128 \\ h_{\bar{c } } & = & d\dfrac{1+\lambda^{-z}}{2}[(\bar{a}_{\mu}^{\dagger}\bar { a}% _ { \mu}-\bar{b}_{\mu}^{\dagger}\bar{b}_{\mu } ) \\ & & + \dsum \limits_{m=0}^{+\infty}\lambda^{-\left ( m+z\right ) } \left ( \bar{a% } _ { m\mu}^{\dagger}\bar{a}_{m\mu }
-\bar{b}_{m\mu}^{\dagger}\bar{b}% _ { m\mu}\right ) ] % \end{array}\ ] ] where @xmath129 , @xmath130 , @xmath131 and @xmath132 , for the conduction band electrons sited at the anderson ion orbital ( magnetic probe ) .
the operators @xmath133 and @xmath134 of eq.([f0g0 ] ) are now written as @xmath135 obeying the orthonormalization condition @xmath136 the logarithmic discretization of the conduction band results in the definition of a basis of operators @xmath137 , where @xmath138 , which is unsuitable for the numerical approach of the conduction electrons - anderson ion interaction for all order of the hybridization @xmath139 , since the anderson ion orbital is coupled to all conduction states @xmath140 and @xmath141 via the operator @xmath133 .
then a new basis @xmath142 will be defined , where each operator @xmath143 is coupled only to the operators @xmath144 and solely the operator @xmath133 is coupled to the anderson ion orbital @xcite .
taking an unitary transformation of the set of operators @xmath145 to the new set of orthonormal operators @xmath146 , the conduction band hamiltonians are given by @xmath147 where the parameter @xmath148 is numerically calculated .
in the limit of large @xmath149 , @xmath148 is approached by the asymptotic expression @xcite @xmath150 which for @xmath151 recovers the wilson expression @xcite .
the series in the hamltonian @xmath93 can be truncated without directly affect the interaction between the electrons of the anderson ion orbital and the conduction band electrons . as the energy associated with the opertors @xmath152 and @xmath153 , for large @xmath154is of order of @xmath155 , one can choose a @xmath156 sufficient large so that this quantity becomes very small as compared with the energy scale relevant to the problem which is defined by the temperature , @xmath157 the truncated hamiltonian @xmath93 and @xmath158 are written as @xmath159 it is interesting to be observed that the truncation of the series in the hamiltonians @xmath93 does nt affect the energy of the anderson ion orbital , since the coupling of the conduction electrons with the electrons of this orbital is only via the operator @xmath133 of the basis @xmath160 the hamiltonian @xmath27 ( eq.([model ] ) ) , in terms of the operators of the basis @xmath161 , can be written as @xmath162 where the state density of the conduction band @xmath163 . for a better interpretation of the results ,
we defne @xmath164 and @xmath165as the hamiltonian @xmath166 and @xmath100 divided by @xmath167 , respectively , so that the smallest energy scale associated to the operator @xmath168 and @xmath169 be on the order of unity : @xmath170 where @xmath171 as @xmath172 , the hamiltonians @xmath164 and @xmath173 are diagonalized separately and the eigenstates of @xmath174 are direct products of the eigenstates of @xmath164 and @xmath165 .
the diagonalization of the hamiltonians @xmath164 and @xmath165 are performed iteractively in subspaces of the same charge and spin ( @xmath175 ) .
the diagonalization process of @xmath164 begins with the diagonalization of the anderson ion hamiltonian @xmath176 in the second step we add the operator @xmath133 and diagonalize the hamiltonian @xmath177 in relation to a basis formed by the eigenstates of the hamiltonian of the first step ( @xmath178 ) and the states that are constructed by the operators @xmath179 , @xmath180 and @xmath181 applied to the eigenstates of @xmath178 . following this procedure ,
the hamiltonian @xmath182 is diagonalized in relation to a basis formed by the eigenstates of @xmath164 and the states given by the operators @xmath183 , @xmath184 and @xmath185 applied to those eigenstates .
the same procedure of iteractive diagonalization is used to diagonalize the hamiltonian @xmath165 .
the ralaxation time t@xmath186 is written as the fermi golden rule @xcite : @xmath187 where @xmath188 is the statistical boltzmann weight , @xmath189 , @xmath190 is the boltzmann constant , @xmath2 is the temperature and @xmath191 and @xmath192 are the initial and final many - particles states of the hamiltonian @xmath27 , with energy @xmath193 and @xmath194 , respectively . in terms of the eigenstates of the hamiltonian given by eq .
( [ hn ] ) , the ralaxation time t@xmath186 is written as @xmath195 where @xmath196 in the iteractive process we take @xmath197 as a constant , @xmath198 , with @xmath199 @xmath200 , and @xmath2 as a function of @xmath201 ( @xmath202 ) , according to wilson method to calcuate the magnetic susceptibility of the kondo model @xcite , @xmath203 as a consequence of the logarithmic discretization of the conduction band , eq.([rate ] ) gives discrete lines centered at the energy @xmath204 . to transform these discrete lines in a continuum spectrum the dirac delta function propriety @xmath205
was used where @xmath206 is the roots of the function @xmath207 , and average @xmath208 over @xmath22 inside the interval ( @xmath23,@xmath209 ) , with @xmath210(@xmath211 ) , @xmath212 with @xmath213
in this section we have shown the results of the relaxation time @xmath214t@xmath215 of the magnetic probe that interacts with the electrons of the anderson ion via a rkky interaction given by eq .
( [ hxnovo ] ) . in this discussion
the coupling constant @xmath67 between the magnetic probe and the conduction electrons will be taken as @xmath216 .
we begin by verifying the accuracy of our numerical results , comparing the numerical renormalization group results for @xmath217 with the analytical solution @xcite . in fig.(1 ) we present @xmath218t@xmath219 as a function of the temperature @xmath2 . for @xmath220 ,
@xmath218t@xmath221 @xmath1tends for one , as is expected for pure metal .
the transition from the valence fluctuation to the doubly occupied orbital regime occurs near the temperature in the order of @xmath222 , where @xmath218t@xmath221 presents a minimum due to the reduction of the density of state of the conduction band close to the anderson ion . for @xmath223
, @xmath218t@xmath219 tends for a constant value of @xmath224 ( traced line ) , in accordance wih the analytical result of @xmath225 ( full line ) obtained from the equation @xcite @xmath226 with an error lower than @xmath227 .
the analysis of the results will be carried out for @xmath228 and two regimes of the parameter space of the anderson model : the valence fluctuation and the singly occupied orbital .
we separate the results for @xmath229 @xmath230 ( @xmath231 ) , with the magnetic probe close to the anderson ion site , and @xmath232with the magnetic probe sited at a distance @xmath233 from the anderson ion .
the valence fluctuation regime occurs for @xmath234 and @xmath235 . in this case , the configurations of doubly and singly occupied orbital are mixed and the number of particle in the fundamental state is in the interval @xmath236 . in fig . ( 2 )
we show @xmath218t@xmath221 for @xmath237 , @xmath238 and two values of @xmath239 ( @xmath240 and @xmath241 ) , for @xmath242 , _
i.e. _ , with the magnetic probe close to the anderson ion site . for @xmath243,@xmath244 all the configurations ( @xmath245 )
are thermically populated and @xmath218t@xmath219 tends to one , as in pure metal .
lowering @xmath2 , for @xmath246 , the anderson ion orbital is in the valence fluctuation regime , where the coupling between the two configurations ( @xmath55 and @xmath247 ) reduces the number of conduction states around the magnetic probe site , with the corresponding reduction in @xmath218t@xmath219 . for @xmath248 , the anderson ion orbital
is strongly coupled to the conduction electrons and the system behaves as a fermi sea , with @xmath218t@xmath249lower than one . for
low @xmath250 @xmath218t@xmath219 increases with the increasing of @xmath239 . for high @xmath251
, the contribution of the configuration @xmath55 for the fundamental state increases and the anderson ion orbital behaves as a kondo impurity , which will be discussed as follows . the anderson ion orbital is singly occupied when @xmath251 , @xmath252 .
the interaction @xmath47 between the electrons inside this orbital increases the energy of the configuration @xmath57 , so that in the fundamental state the anderson ion orbital is in the configuration @xmath55 . at low @xmath2 ( @xmath253 ) the electrons of the anderson ion orbital
is strongly coupled to the electrons of the conduction band , with virtual transition between the configuration @xmath55 and @xmath57 , with the additional electron coming from the conduction band .
the spin states ( @xmath254 and @xmath255 ) are doubly degenerated , so that there is an effective antiferromagnetic exchange interaction between the electrons of the anderson ion orbital and the conduction electrons , breaking the degeneracy of the configurations @xmath256 and @xmath257 , arising the well known kondo effect . from the coupling between these configurations
emerge a singlet and a triplet state , with an energy separation @xmath258 , where @xmath3 is known as the kondo temperature .
a measure with characteristic time @xmath259 can detect the spin inversion of a electron of the anderson ion orbital , which occurs in this period of time . in order for the total magnetic moment of the fundamental state to remain zero ,
the electrons of the conduction band go along with the spin inversion of the electron of the anderson ion orbital , so that the electrons with energy the order of @xmath260 stay a long time around these orbital , increasing the scattering of the conduction electrons , with an increasing in the spin relaxation rate . in fig.(3 ) we present @xmath214t@xmath215 as a function of the temperature for the parameters of the model representing the kondo regime , @xmath261 , @xmath262 and three values for @xmath251 ( @xmath263 ( a ) , @xmath264 ( b ) , @xmath265 ( c ) ) , considering @xmath242 , the magnetic probe close to the anderson ion site .
the peaks occur around @xmath266 ( the kondo width ) , which is given by @xmath267 and takes the values @xmath268 ( a ) , @xmath269 ( b ) and @xmath270 ( c ) .
the kondo temperature @xmath3 is calculated from the magnetic susceptibility results @xcite , using the k. g. wilson criterion @xmath271 where @xmath272 is the magnetic susceptibility , * g * is the land factor and @xmath273 bohr magneton . for low temperature , the heavy fermi liquid behavior is better observed in fig .
( 4 ) , where we show @xmath218t@xmath219 as a function of @xmath2 , using the same parameters of fig.(3 ) . for @xmath274 ,
@xmath218t@xmath219 is drastically enhanced as compared with the normal metal results , and increases as the kondo temperature @xmath3 decreases .
the thermodynamics proprieties of the anderson model in the kondo regime , like the magnetic susceptibility , are characterized by a universal behavior @xcite when the temperature is scaled by the kondo temperature @xmath275 .
this universal behavior is also observed in the electron spin resonance when the spin relaxation rate is multiplied by the temperature.in fig .
( 5 ) we show the universality of @xmath276t@xmath215 in the kondo regime for @xmath277 , @xmath238 and three values for @xmath251 ( @xmath263 ( a ) , @xmath264 ( b ) , @xmath265 ( c ) ) , considering @xmath242 .
the universality remains for temperatures the order of 100@xmath8 . in fig .
( 6 ) we compare the spin relaxation time @xmath214t@xmath215 with the magnetic susceptibility @xmath278 of the anderson ion orbital . for temperatures
lower then @xmath3 the spin relaxation rate @xmath214t@xmath215 is proportional to @xmath279 , in qualitative agreement with the experimental results of coldea _
et al_. @xcite for kondo lattice . increasing the distance between the anderson ion orbital and the magnetic probe , @xmath280 , decreases the contributions of the term @xmath281 and enlarges the contributions of the terms @xmath282 , where @xmath283 , @xmath284 and @xmath285 . on the other hand , the contribution of the term @xmath286 , where @xmath287 , increases with the reduction of @xmath288 until reaching its maximum value @xmath289 , after which it decreases until zero at the limit @xmath290 . in this limit only the term @xmath291 survives , so that the spin relaxation rate is equal to a pure metal . in fig.(7 ) we present the spin relaxation rate @xmath214t@xmath215 for @xmath292 , and @xmath293 , with the parameter of the anderson model given by @xmath294 , @xmath295 and @xmath296 .
the kondo peak , which is sited at the temperature the order of the kondo width @xmath297 , decreases with the increasing of the distance @xmath233 between the anderson ion and the magnetic probe .
the inset of the fig.(7 ) shows that the kondo effect remains even for large distances , being eliminated only in the limit @xmath298 ( @xmath299 ) . the fermi liquid behavior is analyzed in fig .
( 8) , where we present @xmath300t@xmath219 as a function of @xmath301 at very low temperature , taking @xmath262 , @xmath302 and four values @xmath229 ( @xmath303 , @xmath304 , @xmath305 and @xmath293 ) . for a fixed value of @xmath306 ,
the ratio @xmath301 varies continuously from @xmath307 to @xmath308 , crossing three regimes : the regime of anderson ion orbital doubly occupied ( @xmath309 ) , the intermediate valence regime ( @xmath310 ) , and the kondo regime ( @xmath308 ) . for @xmath309 the anderson ion orbital
is doubly occupied and @xmath311 is up bounded by one . by increasing @xmath251 the energy of the configuration with doubly occupied orbital approaches the energy of the configuration with a singly occupied orbital , and @xmath311 decreases monotonically to zero , where these two configurations are degenerated . for @xmath308 , in the kondo regime
, the single orbital configuration becomes dominant and @xmath311 suffers a huge increase . for @xmath312 , in the intermediate valence regime , depending on the ratio @xmath301 , the constant @xmath311 can be smaller or greater than one , so that t@xmath0 can be reduced or enhanced in relation to a pure metal relaxation rate .
this is in contrast with the particular case of coulomb interaction @xmath217 used by gambke _ at al _ @xcite to analyze intermediate valence compound , in which @xmath311 is always reduced in relation to the korringa relaxation rate of pure metal .
we have calculated the relaxation rate @xmath214t@xmath215 of a magnetic probe sited at a distance @xmath233 from the impurity of the spin degenereted anderson model , using numerical renormalization group formalism .
the numerical result is in very good agreement with the analytical result for @xmath217 . in the valence fluctuation regime
the curve @xmath313t@xmath314 presents a depression at temperature the order of @xmath251 , which is the difference between the energy of the configurations @xmath55 and @xmath57 , and behaves as a fermi liquid for very low temperature , augmenting @xmath311 as @xmath251 increases . in the kondo regime
@xmath214t@xmath215 presents a peak at the kondo width @xmath315 , which height increases as @xmath8 decreases . for temperature
much lower then @xmath8 the system behaves as a fermi liquid , presenting a dramatic enhancement in the relaxation rate .
the product of the temperature by the relaxation rate ( @xmath276t@xmath215 ) as function of the temperature scaled by @xmath8 obeys an universal function , even for temperatures of the order of @xmath7 . for temperature
lower then @xmath266 , the relaxation rate @xmath214t@xmath215 is proportional to the product of the temperature by the magnetic susceptibility ( @xmath316 ) . increasing the distance @xmath233 between the magnetic probe and the anderson impurity , the peak at @xmath8 decreases until it disappears at @xmath290 .
for very low temperatures , @xmath313t@xmath314 as a function of @xmath301 approaches to one in the region where the anderson impurity is doubly occupied ( @xmath317 ) , presents a minimum in the region of valence fluctuation regime ( @xmath318 ) , and is drastically enhanced in the region of the kondo regime ( @xmath319 ) .
1 . numerical renormalization group result for @xmath214(t@xmath320 ) as a function of the temperature ( full line ) for @xmath217 and @xmath231 , as compared with the analytical result ( traced line ) [ 12 ] , with a lower than @xmath227 error .
2 . the rate @xmath321(@xmath2t@xmath215 ) as a function of the temperature for the valence fluctuation regime and @xmath231 . for @xmath220 , all the anderson ion states are thermically accessible and the system behaves as a pure metal
; for @xmath322 , @xmath214(@xmath2t@xmath215 ) has a minimum due to the reduction in the density of states around the magnetic probe ; for @xmath253 , @xmath214(@xmath2t@xmath215 ) is constant and increases with @xmath251 .
3 . the spin relaxation rate t@xmath0 as a function of the temperature for three sets of parameters of the anderson model , taking fixed @xmath45 and @xmath323 , and varying @xmath324 the curves present a peak at the kondo width @xmath268 ( curve a ) , @xmath325 ( curve b ) and @xmath326 ( curve c ) , which are marked in the temperature axis .
@xmath8 was obtained from the magnetic susceptibility , as given by eq .
( 39 ) , using the wilson relation @xmath327 .
the height of the peak grows with the reduction of the kondo temperature .
4 . the rate @xmath321(@xmath2t@xmath215 ) as a function of the temperature for the same parameters of fig.(3 ) . for high @xmath2
, the system behaves as a pure metal ; lowering the temperature , the curves present a minimum , associated with the valence fluctuation regime ; for @xmath328 , the @xmath329 is enhanced in the kondo regime , and the system behaves as a heavy fermi liquid .
5 . in the kondo regime
, @xmath276t@xmath215 behaves as a universal function of the @xmath6 .
the figure presents three sets of anderson model parameter in the kondo regime , and the universality behavior of @xmath276t@xmath215 remains until temperatures of the order of 100@xmath330 fig .
6 . in the kondo regime
, the spin relaxation rate t@xmath0 is proportional to @xmath10 .
the figure presents t@xmath0 and @xmath10 for the same set of parameters of the anderson model , in the kondo regime , where t@xmath0 is proportional to @xmath10 for temperature lower than the kondo temperature ( marked in the temperature axis ) .
7 . spin relaxation rate t@xmath0 as a function of the temperature for four distances @xmath233 between the anderson ion and the magnetic probe , corresponding to @xmath331 ( the magnetic probe very close to the anderson ion ) , and @xmath332 , @xmath333 and @xmath293 .
the kondo peak decreases with the distance @xmath334 even for long distance ( curve d ) , the peak survives , as is shown in the inset .
the rate @xmath321(@xmath2t@xmath215 ) as a function of @xmath301 , for very low temperature and deferent distances between the anderson ion and the magnetic probe . for @xmath335 ,
the anderson ion is doubly occupied , for @xmath318 is in the intermediate valence regime , and for @xmath336 , is in the kondo regime .
the curves are lower than one for the anderson ion in the doubly occupied regime , present a minimum in the valence fluctuation regime , and are enhanced in the kondo regime . | the renormalization group formalism was applied to calculate the spin - lattice relaxation rate t@xmath0 of a well - defined magnetic moment in the neighborhood of a spin degenerate anderson impurity . in the kondo regime ,
t@xmath0 in function of @xmath1the temperature @xmath2 presents a peak at the kondo temperature @xmath3 ; for @xmath4 , the system behaves as a heavy fermy liquid , with an enhanced density of states , which increases with the decreasing of the kondo temperature ; t@xmath5 remains an universal function of @xmath6 up to temperatures the order of @xmath7 , where @xmath8 is the kondo width ; for temperatures lower then @xmath3 , the spin relaxation rate , t@xmath9 is proportional to the magnetic susceptibility multiplied by the temperature , @xmath10 ; the peak of t@xmath11 at the kondo temperature decreases with the increasing of the distance between the anderson impurity and the magnetic probe . |
nonlocality is a basic property of quantum states not available in classical world , and it has been the research interest for many years @xcite . the nonlocality can be studied in the context of bell - type inequalities @xcite , it can also be investigated from other viewpoints , for example , from an information - theoretic perspective @xcite . on the other hand , the nonlocality is closely related with entanglement which plays a central role in quantum information processing @xcite , but they are not completely the same .
there are a series of work concentrated on revealing the relationship between nonlocality and entanglement , for example , it is shown that while there are entangled states without bell nonlocality @xcite , there also exists nonlocality without entanglement @xcite .
it is generally accepted that entanglement is a resource in quantum protocols and algorithms @xcite .
interestingly , recent progress shows that the separable but nonclassical states may be responsible for the speedup of a quantum algorithm @xcite .
it ignites interests of people on studying , from various points of view , the quantumness of correlations @xcite and their relations with entanglement and nonlocality @xcite .
particularly , luo and fu @xcite introduced a geometric measure of nonlocality which they termed measurement - induced nonlocality ( min ) . in comparison with bell nonlocality
which is featured by violation of bell - type inequalities satisfied by any local hidden variable theory @xcite , the min is in some sense a more general kind of correlation which was introduced from the consideration that measurements in quantum mechanics usually causes disturbance .
the extent of the maximum disturbance of locally non - disturbing measurements on the total state can reveal global effect of that state which can not be accounted for locally and therefore can be exploited to quantify nonlocality .
the min provides an interesting classification scheme for the bipartite quantum states . in the original geometric quantification scheme @xcite ,
it is obtained by maximizing the hilbert - schmidt distance between the pre- and post - measurement states . for completeness of theoretical consideration , it is also desirable to quantify it from other perspectives so as to reveal more about its meaning and properties .
this is one of our motivations in this paper .
moreover , we know that entanglement is fragile due to decoherence induced by inevitable interaction between the system and its environment . while the min can also be a powerful resource enabling fascinating quantum tasks as shown in @xcite , it is of practical significance to know the property of min , for example , to find ways to stabilize the min of a state or to minimize the devastating effects of the decohering environments .
this is also a motivation of this work . in this paper , we introduce the entropic measure of min , which is different from the corresponding geometric measure @xcite , and give an information - theoretic interpretation of it .
we will obtain the conditions for nonvanishing min , and show that there is nonlocality without quantumness in the sense that the min may persist even for certain zero - discord states .
we will also offer a comparative study of the relationships between min and different quantum correlation measures @xcite for the two - qubit case .
we will show that when interacting with the structured reservoirs , the min can be frozen by introducing detuning or by performing local unitary operations on the initial state , particularly , the strength of min can even be enhanced by interacting the system with a common reservoir .
the paper is organized as follows . in section
ii we first recall the concept of min and its quantification based on the hilbert - schmidt norm @xcite .
then we reformulate the min from an entropic way and list some general features of it . in section iii
, we evaluate dynamics of min for a certain family of two - qubit states both in the independent and common reservoirs , and present schemes for preserving and enhancing the min .
finally , section iv is devoted to a summary .
the preform of min was first introduced by luo and fu @xcite and further formulated by the same authors @xcite .
consider a bipartite state @xmath0 shared by two parties @xmath1 and @xmath2 , with marginal states @xmath3 and @xmath4 , then the min is defined as follows @xmath5 where the maximum is taken over the local von neumann measurements @xmath6 satisfying @xmath7 , and @xmath8 denotes the hilbert - schmidt norm with @xmath9 .
clearly , @xmath10 contains in fact only the measurements that do not disturb @xmath3 locally , thus as pointed out in @xcite , the min is _ in some sense dual to _ the geometric measure of quantum discord ( gmqd ) @xcite @xmath11 , with @xmath10 runs over all the local von neumann measurements . by noting that any @xmath12 bipartite state @xmath0 can always be represented as @xmath13 luo and fu derived an analytical formula of min for the @xmath14 dimensional states , which is given by @xcite @xmath15 where @xmath16 and @xmath17 are traceless hermitian operators satisfying the conditions @xmath18 and @xmath19 .
moreover , @xmath20 is the smallest eigenvalue of @xmath21 with @xmath22 being a real @xmath23 matrix with elements @xmath24 , and @xmath25 with @xmath26 . here
the superscript @xmath27 denotes transpose of vectors or matrices .
although the definitions of min and gmqd are different , they may assume equal values under certain circumstances . to see this ,
consider the two - qubit x state @xmath28 ( i.e. , the @xmath29 density matrix with nonzero elements only along the main diagonal and anti - diagonal ) , for which one can obtain @xmath30 and the eigenvalues of @xmath21 as @xmath31 moreover , it has been shown by daki _
_ @xcite that for the two - qubit quantum states the gmqd can be presented explicitly as @xmath32 where @xmath33 is the largest eigenvalue of the matrix @xmath34 . for @xmath28 ,
the eigenvalues of @xmath35 can be obtained analytically as @xmath36 eqs .
( 4 ) and ( 7 ) imply that @xmath37 , so for nondegenerate @xmath38 ( i.e. , @xmath39 ) we have @xmath40 when @xmath41 .
in fact , the eigenvector of @xmath35 corresponding to the eigenvalue @xmath42 is @xmath43 , hence the optimal measurement for obtaining the gmqd is @xmath44 @xcite .
moreover , if @xmath38 is nondegenerate , then the unique von neumann measurement leaving @xmath38 invariant is @xmath45 @xcite . the optimal measurements for obtaining the min and gmqd are completely the same ,
thus it is understandable that they assume equal values under the condition @xmath41 .
moreover , it is direct to check that @xmath46 , thus if @xmath38 is degenerate with the addition of @xmath41 and @xmath47 , we still have @xmath40 . besides the hilbert - schmidt norm which measures the min from a geometric perspective @xcite , it is also desirable to quantify it from other perspectives . in particular ,
if we accept that the quantum mutual information ( qmi ) is a good measure of total correlations in a bipartite state @xmath0 , then it is natural to quantify min by the maximal discrepancy between the qmi of the pre- and post - measurement states as @xmath48,\ ] ] where @xmath49 represents the qmi , and the minimum is taken only over the local von neumann measurements leaving @xmath3 invariant .
this measure of min quantifies in fact , the maximal loss of total correlations under locally invariant ( non - disturbing ) measurements .
since @xmath50=s(\rho^b)-s(\rho|\{\pi_k^a\})$ ] @xcite represents the average information gained about the system @xmath2 conditioned on the measurements @xmath51 , @xmath52 is in some sense dual to the quantum discord ( qd ) @xmath53 $ ] @xcite .
but the maximum in @xmath54 runs over _ all _ the local von neumann measurements .
moreover , the measure of min introduced here is also quite different from the quantity @xmath55 $ ] , which was introduced by luo @xcite and has been termed measurement - induced disturbance ( mid ) , because when defining @xmath56 the locally invariant measurements @xmath57 are performed on both parties of @xmath1 and @xmath2 .
since the projective measurement @xmath10 leaves @xmath3 invariant ( @xmath4 is obviously also invariant ) , @xmath52 defined above is equivalent to @xmath58-s(\rho),\ ] ] from which one can obtain @xmath59 because _ projective measurements always increase entropy _ ( theorem 11.9 of @xcite ) .
( 9 ) also indicates that @xmath52 quantifies in fact the maximal increment of von neumann entropy induced by the locally invariant measurements .
this result thus establishes a connection between increment of entropy and nonlocality in the quantum domain .
moreover , from an information - theoretic perspective , the entropy measures how much uncertainty there is in the state of a physical system @xcite , @xmath60 can therefore also be interpreted as the maximal increment of our uncertainty about that system induced by the locally invariant measurements .
the post - measurement state can always be written as @xmath61 @xcite , combination of this with the _ joint entropy theorem _
@xcite gives rise to @xmath62=h(p_i)+\sum_i p_i s(\rho^b_i)$ ] , where @xmath63 represents the shannon entropy .
furthermore , from theorem 11.10 of @xcite one can obtain @xmath64 , and from @xcite we have @xmath65 , thus @xmath52 is lower bounded by @xmath66 and upper bounded by @xmath67 , with @xmath68 being the conditional von neumann entropy for @xmath0 , and @xmath69 gives the lower bound of the one - way distillable entanglement . but both the lower and the upper bounds are non - trivial only when @xmath70 because we always have @xmath59 and @xmath71 .
for the two - qubit x states @xmath28 with nondegenerate @xmath38 , the only von neumann measurement leaving @xmath38 invariant is @xmath72 with @xmath73 , which gives rise to the post - measurement state as @xmath74 .
still for @xmath28 with nondegenerate @xmath38 one can check directly that @xmath75 .
thus for the special circumstance considered here we always have @xmath76 .
but if @xmath38 is degenerate , then @xmath77 in general . now we discuss conditions under which the min disappears .
first , it is clear that the min is strictly positive for any state @xmath0 with non - vanishing quantum correlations .
this can be shown directly from the definitions of min and qd based on the hilbert - schmidt distance or the discrepancy between qmi of the pre- and post - measurement states , for which we always have @xmath78 and @xmath79 .
thus one only need to consider the classical - quantum state @xmath80 @xcite .
if the reduced state @xmath81 is nondegenerate ( i.e. , @xmath82 ) , then the min disappears because for this case the only von neumann measurement leaving @xmath83 invariant is @xmath84 , which yields @xmath85 and thus there is no min @xcite .
if @xmath83 is degenerate ( i.e. , @xmath86 ) , then we need to consider two different cases .
the first one is that @xmath87 for all @xmath88 and @xmath89 . for this case
, we always have @xmath90 and there is still no min .
second , if @xmath91 , then one can always find a complete projective measurement @xmath92 so that @xmath93 , thus from theorem 11.9 of @xcite we obtain @xmath94>s(\rho_{\rm cq})$ ] , which gives rise to non - vanishing min . since there is no quantum correlations in @xmath95
, the result obtained here reveals that there exists not only nonlocality without entanglement but also nonlocality without quantumness . as an example
, we consider the two - qubit @xmath95 with degenerate @xmath83 ( i.e. , @xmath96 ) and @xmath97 . in fig .
1 we plot @xmath98 and @xmath99 versus @xmath100 and @xmath101 with @xmath102 .
it is evident that the min vanishes only when @xmath103 , that is @xmath104 , which corroborates our above statements .
also one can note that when @xmath105 , @xmath106 or @xmath107 , @xmath108 , @xmath98 assumes the value 0.25 , half of the maximum 0.5 obtained for the maximally entangled bell states @xmath109 .
but @xmath99 assumes its maximum 1 under the same conditions , equals completely to those obtained for @xmath109 .
this is because @xmath110 , and for this two special cases one can always choose corresponding complete projective measurements @xmath111 so that @xmath112 and @xmath94=2 $ ] .
in fact , the classical correlation in @xmath95 is completely destroyed by the projective measurement @xmath10 because @xmath113=0 $ ] .
for @xmath109 , however , the classical correlation is undisturbed because the optimal measurement @xmath10 ( may be different from that for @xmath95 ) for obtaining the min gives @xmath114=1 $ ] , equals to the classical correlation present in @xmath109 @xcite .
moreover , we would like to point out that for @xmath95 with non - vanishing min , the classical correlation is partially destroyed by the optimal measurement @xmath10 .
the above discussions also indicate that the different measures of min may impose different orderings of quantum states , and in particular , the maximum @xmath115 does not always indicates that the state @xmath0 is maximally entangled , which is in sharp contrast to that of min measured by @xmath116 or quantum correlations measured by qd , gmqd and mid @xcite .
* a. the model * we consider a system consists of two identical atoms with lower and upper levels denoted by @xmath117 and @xmath118 , and is coupled to two typical structured reservoirs , i.e. , the independent and common zero - temperature bosonic reservoir @xcite . for the former case ,
the two atoms interact with their own reservoir with @xmath119 , and the single `` qubit+reservoir '' hamiltonian reads @xmath120 where @xmath121 denotes the transition frequency of the atoms , and @xmath122 ( @xmath123 ) are the pauli rasing and lowering operators .
the index @xmath88 labels the reservoir field mode with frequency @xmath124 , with @xmath125 and @xmath126 being the bosonic creation and annihilation operators and @xmath127 the coupling strength .
we will assume that the two reservoirs are completely the same and denote @xmath128 and @xmath129 . for the latter case ,
the two atoms are coupled to the same reservoir and the hamiltonian reads @xmath130 in the following , we consider the structured reservoir with the spectral density having the lorentzian form @xcite @xmath131 where @xmath132 denotes the spectral width of the reservoir and is related to the reservoir correlation time @xmath133 by the relation @xmath134 , and @xmath135 is related to the decay time scale @xmath136 of the excited state of the atom in the markovian limit of flat spectrum via @xmath137 .
@xmath138 corresponds to the markovian regime , while @xmath139 corresponds to the non - markovian regime .
moreover , @xmath140 is the central frequency of the reservoir detuned from the transition frequency @xmath121 by an amount @xmath141 . for the independent reservoir , we work in the interaction picture and thus the hamiltonian of eq .
( 10 ) can be rewritten as @xmath142 .
we suppose the initial single `` qubit+reservoir '' state to be @xmath143 ( @xmath144 represents vacuum state of the reservoir ) , then by using the schrdinger equation @xmath145 ( @xmath146 represents the reservoir with one excitation in mode @xmath88 ) , one can obtain @xmath147 from the second expression of the above equation one can derive @xmath148 , substituting of which into the first expression we obtain @xmath149 where the function @xmath150 in the limit of large number of modes @xmath88 . then by performing the laplace transformation of eq .
( 14 ) one can obtain @xmath151 with @xmath152,\ ] ] where @xmath153 . if we further define @xmath154 , then @xmath155 , according to which the single - qubit reduced density matrix can be determined as @xmath156 in the standard basis @xmath157 .
for the two - qubit system considered in this work its reduced density matrix can be obtained by the procedure presented in ref .
@xcite .
if the two atoms are coupled to the common reservoir , then their dynamics can be described by the following pseudomode master equation @xcite @xmath158 + \lambda(2b\tilde{\rho } b^{\dag}-b^{\dag}b\tilde{\rho}-\tilde{\rho } b^{\dag}b),\ ] ] where @xmath159 is the density matrix for the atoms and the pseudomode , and the interaction hamiltonian @xmath160 of the atomic system and the pseudomode can be written as @xmath161 we assume throughout this work that the total system contains at most two excitations , then eq . (
17 ) turns out to be set of 64 differential equations , which can be solved by using different methods @xcite .
particularly , for the initial extended werner - like ( ewl ) states , analytical solutions of eq .
( 17 ) were obtained by using the laplace transform which turns the differential equations to the polynomial equations @xcite ( note that there may be some misprints in @xcite .
first , the @xmath162 in eq . ( a.21 ) should be @xmath163 , and @xmath164 should be multiplied by @xmath165 .
second , the solution of @xmath166 in eq .
( a.22 ) should be @xmath167 ) .
+ * b. min dynamics * now we begin our discussion about min dynamics in structured reservoirs .
we suppose the two atoms are prepared initially in the ewl states @xmath168 with @xmath169 or @xmath170 , and @xmath171 , @xmath172 . for the case of independent reservoir ,
@xmath173 and @xmath174 yield qualitatively similar min dynamics , which are independent of the phase factor @xmath175 . in fig .
2 we plotted @xmath52 versus @xmath176 and @xmath177 with the two atoms prepared in different initial states and coupled resonantly ( i.e. , @xmath178 ) to the reservoirs .
it is clear that they attain certain maxima at @xmath179 and vanish at @xmath180 or @xmath181 . for fixed @xmath176
, min decreases exponentially with increasing @xmath177 for the markovian reservoir [ see figs .
2(a ) and 2(b ) ] , while there are revivals of min for the non - markovian reservoir [ see figs .
2(c ) and 2(d ) ] .
since there are neither direct nor reservoir - mediated interactions between the two atoms , the revivals of min after a certain minimum may be induced by the non - markovian effects . if the two atoms interact with the common reservoir , however , @xmath173 and @xmath174 will exhibit completely different min dynamics .
particularly , for the initial sub - radiant state @xmath182 , the density matrix maintains its initial value ( i.e. , latexmath:[$\rho(t)\equiv decay . to see more about min dynamics for this case , we show in fig .
3 the @xmath52 versus @xmath176 and @xmath177 with different initial states . one can note that the min displays complicated behaviors .
first , it is not a symmetric quantity with respect to @xmath179 for the initial state @xmath184 , and @xmath52 can even does not behave as a monotonic function of @xmath177 for the markovian reservoir . as one can see from figs .
3(a ) and 3(b ) , there are creations of min for the initial state @xmath184 with @xmath180 and for @xmath170 with @xmath180 or @xmath181 .
the underlying reason for this is the reservoir - mediated interaction because here the two atoms interact with a common reservoir .
second , there are still revivals of min for the non - markovian case , but the combined effects of the non - markovian effects and the reservoir - mediated interaction makes it more evident than those for the independent reservoirs .
when the two atoms interact with independent reservoirs , their entanglement may be trapped by detuning the central frequency of the reservoir from @xmath121 @xcite . in fig .
4(a ) we display @xmath52 versus @xmath177 for the initial state @xmath185 with various values of detuning @xmath141 .
the result shows that @xmath52 may be enhanced by introducing detuning @xmath141 and when @xmath186 it begins to oscillating weakly around its initial value , from which it is reasonable to conjecture that up to the large detuning limit case @xmath52 will maintains its initial value during the time evolution process and thus min trapping occurs .
experimentally , for two rb rydberg atoms with lifetime @xmath187 and inside the fabry - prot superconducting resonant cavities with quality factor @xmath188 , we have the atomic decay rate @xmath189 , and the spectral width of the reservoir @xmath190 @xcite .
these values correspond to a good non - markovian regime .
moreover , the detuning needed to freeze the min can be implemented by stark - shifting the transition frequencies with a static electric field .
the typical stark shifts are @xmath191 @xcite and thus a dimensionless shift with respect to @xmath135 of the order @xmath192 can be realized .
since the reservoir always acts in some basis and therefore changing the basis may change the strength of reservoir and thus affect the durability of correlations .
particularly , when the two identical atoms equally coupled to the common reservoir , there exists a sub - radiant state @xmath193 which is decoupled from the field modes and thus does not decay with time @xcite .
this motivates us to enhance robustness of min by performing appropriate local unitary operations on the initial state , though the min itself is invariant under local unitary operations . as an explicit example
, we show in fig .
4(b ) the numerical results obtained from the initial local unitary equivalent states @xmath194 with @xmath195 , @xmath196 and @xmath197 .
it is clear that the strength of min at finite time @xmath198 can be tuned among regions bounded approximately by those obtained for the initial states @xmath174 with @xmath199 and @xmath200 , respectively .
particularly , one can note that for the initial state @xmath174 with @xmath200 ( the top red line ) , the amount of min can even be enhanced by subjecting the two atoms to a common reservoir .
in fact , for the initial state @xmath174 there exists a threshold phase angle @xmath201 ( depends on the system parameters ) beyond which the min as well as the quantum correlations measured by entanglement of formation ( denoted by @xmath202 ) @xcite , qd and gmqd can be enhanced to a larger value and further be frozen in the long time limit . in fig .
5 we display numerical results of @xmath203 versus @xmath204 for @xmath205 with @xmath195 , @xmath196 , @xmath197 and @xmath206 . clearly , the threshold @xmath201 are different for different correlation measures , but @xmath207 always increases with increasing value of @xmath208 and attains the maxima at @xmath200 .
this provides an efficient method for enhancing and storing min for future usage in external environments .
one can also note that the curves for @xmath116 and @xmath209 are overlapped when @xmath210 , this is because during this parameter region @xmath3 is nondegenerate and @xmath42 is the largest eigenvalue of @xmath35 ( see section ii ) .
finally , we compare dynamical behaviors of min quantified by different measures .
two exemplified plots with min quantified by @xmath116 and @xmath52 are presented in fig .
6 , from which one can note that while for the initial state @xmath174 the two different measures give qualitatively similar characterizations of the min , for @xmath173 there exists states @xmath211 and @xmath212 which fulfill @xmath213<n^s[\rho(t_2)]$ ] and @xmath214>n^v[\rho(t_2)]$ ] .
thus we see that the different measures of min are not only quantitatively but also qualitatively different in the sense that they may impose inequivalent orderings of quantum states .
therefore , the lack of the same states ordering under different entanglement measures @xcite and other quantum correlation measures @xcite , goes beyond entanglement and quantum correlations .
in this paper , we introduced the entropic measure of min based on the maximal difference between qmi of the pre- and post - measurement states . while the geometric measure of min @xcite quantifies the maximal hilbert - schmidt distance between @xmath0 and @xmath215 , the entropic measure of min introduced here quantifies the maximal loss of total correlations or equivalently , the maximal increment of von neumann entropy of @xmath0 by @xmath10 .
it seems naturally a definition from information - theoretical point of view .
based on this measure , we discussed conditions for nonvanishing min and consolidated the finding that there exists nonlocality without quantumness @xcite .
for the two - qubit case , we further compared min with different quantum correlation measures @xcite , showing that some of them may assume equal values under certain special conditions , and a quantum state with the maximum @xmath115 does not always indicate that the state possess the maximum quantum correlation .
since real physical system is inevitably coupled to its surroundings , in comparing with fragile entanglement , it is significant to study stability of min against decoherence . here as a concrete example , we evaluated min dynamics for a family of two - qubit states in the lorentzian structured reservoirs , and displayed their rich dynamical behaviors both for the markovian and non - markovian situations .
we showed that the min can be frozen by increasing the detuning parameter , and for some initial states the strength of min as well as the quantum correlations can even be enhanced in the short time region and further be frozen in the long time limit by interacting the system with a common reservoir , the underlying reason for which may be the existent component of the decoherence - free state @xmath193 .
finally , we studied the issue related to the relativity of different min measures , just as that of different entanglement @xcite and quantum correlation measures @xcite . by evaluating the dynamics of @xmath116 and @xmath52 in an non - markovian common reservoir
, we showed that the two min measures do not necessarily imply the same ordering of quantum states .
this is reminiscent of the cases for entanglement and quantum correlation measures and may indicate that the difference in nonlocality is rather measure dependent than state dependent . experimentally , besides atomic systems @xcite , two - qubit quantum state with different reservoirs , such as markovian and non - markovian , can be realized @xcite , and different correlations such as quantum discord have been shown recently @xcite .
it will also be interesting to explore the entropic min of this paper in experiments .
+ this work was supported by nsfc ( 10974247 , 11175248 ) , `` 973 '' program ( 2010cb922904 ) , nsf of shaanxi province ( 2010jm1011 , 2009jq8006 ) , and the scientific research program of education department of shaanxi provincial government ( 2010jk843 ) .
50 p. g. bergmann , v. de .
sabbata , and j. n. goldberg , _ classical and quantum nonlocality _
( world scientific , singapore , 2000 ) .
j. s. bell , physics * 1 * , 195 ( 1964 ) ; j. f. clauser , m. a. horne , a. shimony , and r. a. holt , phys .
* 23 * * , 880 ( 1969 ) ; r. augusiak , d. cavalcanti , g. prettico , and a. acn , _ ibid . _
* 104 * , 230401 ( 2010 ) .
n. brunner , d. cavalcanti , a. salles , and p. skrzypczyk , phys .
lett . * * 106 * * , 020402 ( 2011 ) ; c. branciard and n. gisin , _ ibid . _
* 107 * , 020401 ( 2011 ) .
r. horodecki , p. horodecki , m. horodecki and k. horodecki , rev .
phys . * 81 * , 865 ( 2009 ) .
l. masanes , y .- c .
liang , and a. c. doherty , phys .
lett . * * 100 * * , 090403 ( 2008 ) ; r. f. werner , phys .
rev . a * * 40 * * , 4277 ( 1989 ) ; a. g. kofman and a. n. korotkov , _ ibid . _
* 77 * , 052329 ( 2008 ) ; b. bellomo , r. l. franco , and g. compagno , _ ibid . _
* 78 * , 062309 ( 2008 ) .
c. h. bennett , _ et al .
_ , phys . rev . a * * 59 * * , 1070 ( 1999 ) ; r. horodecki , m. horodecki , and p. horodecki , _ ibid . _
* 60 * , 144 ( 1999 ) ; j. niset and n. j. cerf , _ ibid . _ * 74 * , 052103 ( 2006 ) ; j. walgate and l. hardy , phys .
* 89 * * , 147901 ( 2002 ) .
e. knill and r. laflamme , phys .
rev . lett . * * 81 * * , 5672 ( 1998 ) ; a. datta , a. shaji , and c. m. caves , _ ibid .
_ * 100 * , 050502 ( 2008 ) ; b. p. lanyon , m. barbieri , m. p. almeida , and a. g. white , _ ibid .
_ * 101 * , 200501 ( 2008 ) .
s. luo , phys .
a * * 77 * * , 022301 ( 2008 ) . b. daki , v. vedral , and .
brukner , phys .
* 105 * * , 190502 ( 2010 ) ; s. luo and s. fu , phys .
a * 82 * , 034302 ( 2010 ) .
k. modi , t. paterek , w. son , v. vedral , and m. williamson , phys .
* 104 * * , 080501 ( 2010 ) .
l. mazzola , j. piilo , and s. maniscalco , phys .
* 104 * * , 200401 ( 2010 ) ; t. werlang , c. trippe , g. a. p. ribeiro , and g. rigolin , _ ibid .
_ * 105 * , 095702 ( 2010 ) .
j. cui and h. fan , j. phys .
theor . * 43 * , 045305 ( 2010 ) .
m. d. lang and c. m. caves , phys .
* 105 * * , 150501 ( 2010 ) .
a. streltsov , h. kampermann , and d. bru , phys .
* 106 * * , 160401 ( 2011 ) ; m. f. cornelio , m. c. de oliveira , and f. f. fanchini , _ ibid . _ * 107 * , 020502 ( 2011 ) . s. luo and s. fu , epl * 92 * , 20004 ( 2010 ) . s. luo and s. fu , phys .
* 106 * * , 120401 ( 2011 ) .
m. a. nielsen and i. l. chuang , _ quantum computation and quantum information _
( cambridge university press , cambridge , uk , 2000 ) .
n. li and s. luo , phys .
a * 84 * , 042124 ( 2011 ) ; z. j. xi , x. m. lu , x. g. wang , and y. m. li , j. phys .
a * 44 * , 375301 ( 2011 ) .
breuer and f. petruccione , _ the theory of open quantum systems _
( oxford university press , oxford , 2001 ) .
b. bellomo , r. l. franco , and g. compagno , phys .
* 99 * * , 160502 ( 2007 ) ; b. m. garraway , phys .
rev . a * * 55 * * , 2290 ( 1997 ) ; b. j. dalton , s. m. barnett , and b. m. garraway , _ ibid . _
* 64 * , 053813 ( 2001 ) ; b. j. dalton and b. m. garraway , _ ibid .
_ * 68 * , 033809 ( 2003 ) .
l. mazzola , s. maniscalco , j. piilo , k .- a .
suominen , and b. m. garraway , phys .
rev . a * * 79 * * , 042302 ( 2009 ) . l. mazzola , s. maniscalco , j. piilo , and k. a. suominen , jpb * 43 * , 085505 ( 2010 ) .
d. z. rossato , t. werlang , l. k. castelano , c. j. villas - boas , and f. f. fanchini , phys .
rev . a * * 84 * * , 042113 ( 2011 ) .
s. maniscalco , f. francica , r. l. zaffino , n. l. gullo , and f. plastina , phys .
* 100 * * , 090503 ( 2008 ) .
m. l. hu , eur .
j. d * 59 * , 497 ( 2010 ) .
b. bellomo , r. l. franco , s. maniscalco , and g. compagno , phys .
rev . a * * 78 * * , 060302 ( 2008 )
. s. kuhr _
et al . _ ,
. lett . * 90 * , 164101 ( 2007 ) .
e. hagley _
* 79 * * , 1 ( 1997 ) . c. h. bennett , d. p.
divincenzo , j. a. smolin , and w. k. wootters , phys .
a * * 54 * * , 3824 ( 1996 ) ; w. k. wootters , phys .
* 80 * * , 2245 ( 1998 ) .
y. yeo , j .- h .
an , and c. h. oh , phys . rev .
a * 82 * , 032340 ( 2010 ) ; d. girolami and g. adesso , _ ibid . _ * 83 * , 052108 ( 2011 ) ; m. l. hu and h. fan , ann . phys . * 327 * , 851 ( 2012 ) ; m. okrasa and z. walczak , e - print arxiv : 1109.4132 [ quant - ph ]
. b. h. liu , y. f. huang , c. f. li , g. c. guo , e. m. laine , h. p. breuer , j. piilo , nature phys .
* 7 * , 931 ( 2011 ) .
j. s. xu , x. y. xu , c. f. li , c. j. zhang , x. b. zou , and g. c. guo , nature commun . * 1 * , 1005 ( 2010 ) . | we propose the entropic measurement - induced nonlocality ( min ) as the maximal increment of von neumann entropy induced by the locally non - disturbing measurement , and study behaviors of it both in the independent and common structured reservoirs .
we present schemes for preserving the min , and show that for certain initial states the min , including the quantum correlations , can even be enhanced by the common reservoir .
additionally , we also show that the different measures of min may give different qualitative characterizations of nonlocal properties , i.e. , it is rather measure dependent than state dependent . |
the experimental verification of the formulae for the perihelion precession of mercury and for the bending angle of light passing near massive bodies such as our sun , are the two most spectacular confirmations of einstein s theory of general relativity @xcite .
furthermore , the efforts of einstein in order to explain these two phenomena played an important role in the development of general relativity ( gr ) @xcite .
it is generally accepted that newton s universal gravitational law , which can not account for the observed perihelion advancement and which gives only half of the experimentally observed light bending angle , is not accurate enough under relativistic conditions @xcite .
however , when employing newton s universal gravitational law under relativistic conditions , little attention has been paid to the fact that there exist differences between the rest , relativistic , inertial and gravitational masses , of the bodies involved . the inertial mass , @xmath11 , and the gravitational mass , @xmath12 , are equal according to the equivalence principle @xcite , but the ratio of the inertial mass to the rest mass and also the ratio of the relativistic mass , @xmath13 , to the rest mass are unbounded as the particle velocity approaches the speed of light @xcite . consequently , the following important question arises when using newton s universal gravitational law under conditions where the lorentz factor @xmath14 is significantly larger than unity : should one use the particle rest mass , @xmath15 , the relativistic mass , @xmath13 , or the gravitational mass , @xmath12 , which equals the inertial mass , @xmath11 ?
this question has been addressed recently in @xcite , where it is argued that one should use the gravitational mass , @xmath12 , rather than the rest mass , @xmath15 , or the relativistic mass , @xmath13 .
perhaps one of the reasons that this question was not discussed earlier is that the velocities of the planets of our planetary system are of the order @xmath16 , thus the corresponding lorentz factors are of the order of @xmath17 and therefore the differences between rest , relativistic , and gravitational masses are very small , which is also the case for the magnitude of the perihelion precession value per orbit and of the light bending angle caused by our sun . however , this is not the case for many other astronomical systems involving heavier stars @xcite .
the inertial mass , @xmath11 , and hence according to the equivalence principle , the gravitational mass @xmath12 , is related with the rest mass , @xmath15 , via the equation @xmath18 this result for linear motion was first obtained in einstein s pioneering special relativity paper @xcite ; it has been argued recently @xcite that this result remains valid for an arbitrary particle motion , including circular motion .
consequently , newton s universal gravitational law for the gravitational force between two particles , at a distance @xmath19 , of rest masses @xmath20 and moving with a velocity @xmath21 relative to each other , is given by @xmath22 if @xmath23 , the above equation reduces to @xmath24 by employing equation ( [ eq3 ] ) it is possible to predict the existence of gravitationally confined circular rotational states with many of the properties of hadrons @xcite , or bosons @xcite , and to show that the rest masses of quarks are very small , in the mass range of neutrinos @xcite .
furthremore , using equation ( [ eq3 ] ) it can be shown that the ratio @xmath25 of the force , @xmath26 , keeping two or three - particle systems in orbit , divided by the same force computed for @xmath27 , is @xmath28 , which is exactly the value predicted by gr @xcite . and
@xmath29.,scaledwidth=25.0% ] for the motion of a particle with rest mass @xmath15 rotating around a massive particle of rest mass @xmath30 , equation ( [ eq2 ] ) remains valid .
indeed , consider for example the sun ( mass @xmath30 ) and mercury ( mass @xmath15 ) at the locations of two instantaneous inertial systems , @xmath31 and @xmath32 , ( fig .
1 ) moving with an instantaneous velocity @xmath21 relative to each other . in this case , the inertial and thus the gravitational inertial mass of mercury is given by @xmath33 according to the observer on the sun at @xmath31 , but also the inertial mass of sun is given by @xmath34 according to the observer on mercury at @xmath32 .
if one uses rest masses instead of inertial masses then , according to newtonian mechanics the particle of mass @xmath15 moves on an ellipse . consider such an ellipse with an elliptical parameter @xmath29 ( also known as the semi - latus rectum ) , with eccentricity @xmath35 , and with a semi - major axis @xmath36 ; these parameters are related by the equation @xmath37 gr implies the following expression for the precession angle , @xmath1 , @xcite @xmath38 where @xmath39 denotes angular momentum per unit mass . using the momentum - velocity equation of special relativity ( sr ) ,
namely momentum equals @xmath40 , lemmon and mondragon examined in @xcite the resulting corrections to keplerian orbits and to the precession angle @xmath1 .
their work , which does not take into consideration equation ( [ eq1 ] ) , yields one sixth of the value computed via gr ( equation ( [ eq5 ] ) ) .
it is interesting to note that silberstein already back in 1917 @xcite attempted to account for the velocity dependence of the rotating mass @xmath41 by combining @xmath42 with the empirical force expression @xmath43 he reported that he could get agreement with the gr result of equation ( [ eq5 ] ) provided that @xmath44 and @xmath45 .
he noted that it was entirely unclear to him why this particular value of @xmath46 leads to exact agreement with gr @xcite .
furthermore , silberstein only analysed the relevant equations in the limit of @xmath47 , where @xmath21 is the velocity of the rotating particle . here ,
for the sake of completness we first present a simple rederivation of the well known @xcite gr expression @xmath48 where @xmath49 in equations ( [ eq8 ] ) , @xmath50 is the angular momentum of the particle of mass @xmath41 , @xmath51 is the mass of the heavier body , @xmath52 is the gravitational constant , and @xmath39 is the angular momentum per unit mass .
we then present the first main result of our paper : if a particle of mass @xmath41 moves around a heavier mass @xmath51 under the influence of the relativistic gravitational law ( [ eq2 ] ) , then @xmath53 satisfies the ordinary differential equation @xmath54\right\}^{-3},\ ] ] where @xmath55 and @xmath56 denotes quantities computed via equations ( [ eq1 ] ) to ( [ eq3 ] ) .
it turns out that for small @xmath57 the dominant parts of the solutions of equations ( [ eq7 ] ) and ( [ eq9 ] ) are both given by the expression @xmath58 and thus , the precession angle per revolution , @xmath3 , is given in both cases by @xmath59 which using the definition of @xmath57 in the third of equations ( [ eq8 ] ) becomes the well known formulae ( [ eq5 ] ) .
it is important to note that one may consider a third instantaneous inertial system , @xmath60 , with an observer , located on earth and with a velocity @xmath61 with respect to one of the two moving bodies .
this second velocity @xmath61 , and the corresponding lorentz factor , @xmath62 , defines the ratio @xmath63 of the relativistic and rest masses of mercury as observed from @xmath60 , i.e. from the earth .
since @xmath15 is not amenable to direct observation , and also since @xmath1 should be independent of @xmath60 , we may set @xmath64 .
let @xmath65 be cylindrical coordinates in a plane containing the center of the sun , and let @xmath66 , @xmath67 , and @xmath50 denote the proper time , energy , and angular momentum respectively .
the basic equations of gr are conservation of energy and of angular momentum , as well as the schwarzschild metric equation : @xmath68 @xmath69 @xmath70 @xmath71 where @xmath72 is the schwarzschild radius of the sun and @xmath51 is the mass of the sun . solving equations ( [ eq13 ] ) and ( [ eq14 ] ) for @xmath73 and @xmath74 in terms of @xmath75 , substituting the resulting expressions into ( [ eq15 ] ) , and then solving the resulting equation for @xmath76 we find @xmath77 making the substitution @xmath78 equation ( [ eq17 ] ) becomes @xmath79 differentiating this equation with respect to @xmath80 and introducing the dimensionless quantities defined in equations ( [ eq8 ] ) we find equation ( [ eq7 ] ) .
consider a single particle of relativistic mass @xmath81 and of gravitational mass @xmath12 moving under the influence of a central force of magnitude @xmath82 .
then , @xmath83 incidentaly , in the particular case that @xmath84 , using the definition of @xmath14 , i.e. the second of equations ( [ eq1 ] ) , we find @xmath85 hence , @xmath86 it is shown in the appendix that @xmath87 and then equation ( [ eq21 ] ) takes the familiar form found in most books , namely @xmath88 starting with ( [ eq19 ] ) , it is straightforward to derive the following results : 1 .
the angular momentum @xmath89 is conserved , where @xmath89 is defined by @xmath90 furthermore , the magnitude of the angular momentun denoted by @xmath50 is given by @xmath91 where @xmath92 are cylindrical coordinates in the plane of motion which clearly is orthogonal to @xmath50 .
2 . @xmath93 satisfies the equation @xmath94 3 .
the speed of the particle @xmath21 satisfies the equation @xmath95.\ ] ] in order to derive the above results we begin by differentiating equation ( [ eq23 ] ) with respect to @xmath96 : @xmath97 thus , @xmath89 is constant . equation ( [ eq23 ] ) implies that @xmath98 is perpendicular to @xmath89 .
introducting cylindrical coordinates in the plane perpendicular to @xmath89 , we find @xmath99 where @xmath100 and @xmath101 are unit vectors along the @xmath102 and @xmath103 axis . differentiating ( [ eq28 ] ) with respect to @xmath96 we find @xmath104 thus , the definition of @xmath89 implies @xmath105 where @xmath106 is the unit vector perpendicular to @xmath89 .
let @xmath50 denote the magnitude of @xmath89 .
equation ( [ eq30 ] ) implies @xmath107 decomposing equation ( [ eq19 ] ) along the @xmath102 and @xmath103 axis we find the following equations : @xmath108 using the indentity @xmath109 we find @xmath110 replacing in this equation @xmath111 via equation ( [ eq31 ] ) and simplifying we find @xmath112 thus , the first of equations ( [ eq32 ] ) yields @xmath113 replacing in this equation @xmath111 by ( [ eq31 ] ) , and simplifying we find equation ( [ eq25 ] ) .
equation ( [ eq29 ] ) implies that @xmath114 satisfies @xmath115 noting that @xmath116 and using for @xmath117 the expression obtained from ( [ eq31 ] ) , namely @xmath118 it follows that @xmath119 using in ( [ eq35 ] ) equations ( [ eq36 ] ) and ( [ eq37 ] ) we find eq .
( [ eq26 ] ) .
in this case @xmath120 thus , equations ( [ eq25 ] ) and ( [ eq26 ] ) become @xmath121 and @xmath122,\ ] ] where @xmath123 denotes angular momentum per unit mass , i.e. @xmath124 introducing the dimensionless parameter @xmath53 via @xmath125 equations ( [ eq39 ] ) and ( [ eq40 ] ) become @xmath126 and @xmath127,\ ] ] where @xmath128 is defined by @xmath129 equations ( [ eq43 ] ) and ( [ eq44 ] ) together with the definition of @xmath14 imply @xmath130\right\}^{-3}.\ ] ] in order to solve ( [ eq46 ] ) for small @xmath128 we let @xmath131 then , equation ( [ eq46 ] ) yields @xmath132 and @xmath133 the general solution of equation ( [ eq48 ] ) is @xmath134 where @xmath35 and @xmath135 are constants .
letting for simplicity @xmath136 , equation ( [ eq49 ] ) becomes @xmath137 thus , @xmath138 hence , using ( [ eq50 ] ) with @xmath136 and ( [ eq52 ] ) into ( [ eq47 ] ) we find @xmath139.\ ] ] the dominant part of @xmath53 denoted by @xmath140 is given by @xmath141 we next solve equation ( [ eq7 ] ) for small @xmath57 .
letting @xmath142 we find that @xmath143 satisfies the ordinary differential equation @xmath144 thus , @xmath145.\ ] ] the dominant part of @xmath146 denoted by @xmath147 is given by @xmath148 in order to relate @xmath57 and @xmath128 we need to relate @xmath39 and @xmath123 . in this respect
we note that equations ( [ eq14 ] ) and ( [ eq24 ] ) yield @xmath149 thus , @xmath150 hence , using equation ( [ eq13 ] ) to compute @xmath151 , as well as using the first of equations ( [ eq8 ] ) and equation ( [ eq42 ] ) to relate @xmath152 and @xmath153 with @xmath146 and @xmath53 we find the following : @xmath154 or @xmath155 equation ( [ eq18 ] ) can be rewritten in the form @xmath156 using in this equation the expansion @xmath157 it follows that @xmath158 letting in ( [ eq60 ] ) @xmath159 and @xmath160 we find @xmath161 hence , @xmath162 thus @xmath163 coincides with @xmath140 .
the computation of the precession of the perihelion depends only on @xmath163 . indeed ,
equation ( [ eq57 ] ) implies that @xmath164 vanishes at @xmath165 , where @xmath166 satisfies the equation @xmath167 ) ( general relativity , gr ) and ( [ eq9 ] ) ( relativistic gravitational law , rg ) and comparison with the keplerian orbit ( [ eq48 ] ) for reference .
distances in @xmath168 . for illustrative purposes
the value used for gm is larger by a factor of @xmath169 than the real value of @xmath170.,scaledwidth=40.0% ] [ fig:2 ] one solution of this equation is @xmath171 . in order to obtain the second solution
we let @xmath172 and we use the identities @xmath173 then , equation ( [ eq65 ] ) implies @xmath174 equation ( [ eq53 ] ) with @xmath175 also yields for @xmath166 an expression identical with ( [ eq66 ] ) .
results of numerical integration of equations ( [ eq7 ] ) and ( [ eq9 ] ) as well as of the kepler equation ( [ eq48 ] ) are presented in figure 2 , which shows that indeed the results of equations ( [ eq7 ] ) and ( [ eq9 ] ) coincide for small @xmath176 , i.e. the @xmath177 gravitational law yields to order @xmath176 the same results with general relativity .
the present results show that the relativistic newton s gravitational law , which uses gravitational rather than rest masses and thus contains the term @xmath178 multiplying each moving mass in the familiar @xmath179 or @xmath180 term , leads to a particle trajectory @xmath0 whose dominant term is the same with the dominant term of the familiar @xmath0 expression obtained via the theory of general relativity . furthermore , the perihelion precession angle , @xmath3 , computed via the former approach is equal to the value computed via general relativity .
@xmath181 we thank mr .
d. grigoriou for helpful discussions and for the numerical integration of equations ( 7 ) and ( 9 ) shown in figure 2 .
22 einstein a ( 1916 ) die grundlagen der allgemeinen relativittstheorie .
ann der physik bd .
xlix , s. 769 - 822 linton cm ( 2004 ) from eudoxus to einstein : a history of mathematical astronomy .
cambridge university press , isbn 0521827507 .
misner cw , thorne ks , zurek wh ( 2009 ) john wheeler , relativity , and quantum information .
phys today 62(4):40 - 46 das a ( 2011 ) lectures on gravitation .
world scientific publishing co. pte .
ltd , singapore roll pg , krotkov r , dicke rg ( 1964 ) the equivalence of inertial and passive gravitational mass .
annals of physics 26(3):442 - 517 einstein a ( 1905 ) zr elektrodynamik bewegter krper .
der physik . ,
xvii , s. 17:891 - 921 ; english translation on the electrodynamics of moving bodies ( @xmath182 ) by g.b .
jeffery and w. perrett ( 1923 ) vayenas cg & souentie s ( 2012 ) , gravity , special relativity and the strong force : a bohr - einstein - de - broglie model for the formation of hadrons .
springer , isbn 978 - 1 - 4614 - 3935 - 6 .
vayenas cg , souentie s , fokas a ( 2014 ) a bohr - type model of a composite particle using gravity as the attractive force . arxiv:1306.5979v4
[ physics.gen-ph ] ; physica a 405:360 - 379 .
vayenas cg , grigoriou d ( 2015 ) microscopic black hole stabilization via the uncertainty principle .
journal of physics 574:012059 vayenas cg , fokas as , grigoriou d ( 2016 ) on the structure , masses and thermodynamics of the w@xmath183 bosons .
physica a 450:37 - 48 .
bertone g ( 2010 ) particle dark matter : observations , models and searches .
cambridge university press , cambridge .
isbn 9780521763684 das s & vagenas ec ( 2009 ) phenomenological implications of the generalized uncertainty principle .
87:233 - 240 french ap ( 1968 ) special relativity .
norton and co. , new york freund j ( 2008 ) special relativity for beginners .
world scientific publishing , singapore einstein a. erklarung der perihelbewegung des merkur aus der allgemeinen relativitatstheorie .
sitzungsberichte der kniglich preussischen akademie der wissenschaften ( berlin ) , seite 831 - 839 translated by vankov aa ( 2011 ) einstein s paper : `` explanation of the perihelion motion of mercury from general relativity theory '' http://www.wbabin.net/eeuro/vankov.pdf .
misner cw , thorne ks and wheeler ja ( 1973 ) gravitation ( san francisco : freeman ) weinberg s ( 1972 ) gravitation and cosmology : principles and applications of the general theory of relativity ( john wiley & sons ) pp .
194 - 8 wald rm ( 1984 ) general relativity ( university of chicago press ) pp .
142 - 3 taylor ef and wheeler ja ( 2000 ) exploring black holes : introduction to general relativity ( addison wesley longman ) magnan c ( 2007 ) complete calculations of the perihelion precession of mercury and the deflection of light by the sun in general relativity .
arxiv:0712.3709v1 [ gr - qc ] lemmon tj & montragon ar ( 2010 ) first - order special relativistic corrections to kepler s orbits .
arxiv:1012.5438v1 [ astro-ph.ep ] silberstein l ( 1917 ) the motion of the perihelion of mercury deduced from the classical theory of relativity
. monthly notices of the royal astronomical society , 77:503 - 510 .
here we establish the validity of the equation @xmath184 differentiating equation ( [ eq29 ] ) with respect to @xmath96 we find @xmath185 where the unit vector @xmath186 define by @xmath187 is orthogonal to the unit vector @xmath188
. equations ( [ eq28 ] ) , ( [ eq29 ] ) , and ( [ a3 ] ) imply that @xmath189 can be rewritten in the form @xmath190 hence using @xmath191 , it follows that @xmath192 thus , @xmath193 differentiating ( [ eq35 ] ) with respect to @xmath96 we find @xmath194 equations ( [ a6 ] ) and ( [ a7 ] ) imply ( [ a1 ] ) .
we will establish the following fact : suppose that @xmath19 satisfies the equation @xmath195 | let @xmath0 denote the orbit of mercury .
we compare the formulae obtained via general relativity for @xmath0 and for the corresponding perihelion precession angle @xmath1 , with the formulae obtained via the relativistic gravitational law , @xmath2 .
the latter law can be derived from newton s gravitational law by employing the gravitational rather than the rest masses of the sun and mercury .
remarkably , it is found that the two expressions for the dominant part of @xmath0 are identical and that the two expressions for @xmath3 are also identical .
a.s . fokas@xmath4 and c.g .
vayenas@xmath5 _ @xmath6department of applied mathematics and theoretical physics , university of cambridge , cambridge , cb3 0wa , uk + @xmath7department of electrical engineering , university of southern california , los angeles , california , 90089 - 2560,usa + @xmath8lcep , 1 caratheodory st . , university of patras , patras gr 26500 , greece + @xmath9division of natural sciences , academy of athens , 28 panepistimiou ave . , + gr-10679 athens ,
greece _ * pacs numbers : * 03.30.+p , 04.20.@xmath10q , 12.60.@xmath10i , 14.20.dh , 14.65.@xmath10q |
the drell - yan process has been widely studied in many past @xcite and present @xcite hadron collider experiments and played a significant role in the development of our understanding of qcd and electroweak ( ew ) interactions , both from the experimental and theoretical point of view .
certainly this will also be the case for the lhc experiments , especially because it will soon become the unique @xmath1 and @xmath0-boson production factory which is expected to collect 300 million @xmath1 and 20 million @xmath0 events per year of its operation at energies @xmath2 tev and the luminosity of @xmath3 . among various distributions of @xmath1 and @xmath0 observables ,
the transverse momentum spectrum of vector bosons in a drell - yan process is a very useful and important quantity for the experimental program at the lhc . in the case of @xmath1 production ,
the uncertainty in the shape of the spectrum directly affects the measurement of the @xmath1 mass @xcite and its mass charge asymmetry @xmath4 @xcite .
it also helps to understand the signature for higgs boson production at either tevatron or lhc @xcite .
although the experiments measure the @xmath0 transverse momentum distribution and use this to infer that of the @xmath1 boson , the extent to which the effects are non - universal limits the ultimate accuracy of the measurement , unless elaborate tricks , as proposed in ref .
@xcite are used . for these reasons
, it is of utmost importance to predict the @xmath1 and @xmath0 observables with as high as possible theoretical precision .
the sources of uncertainty in the theoretical predictions of observables , such as the transverse momentum of electroweak bosons discussed here , are of perturbative and non - perturbative origin . in this short note
we will concentate on the modelling of the latter in the framework of a backward evolution parton shower approach @xcite which is widely used in general purpose monte carlo generators such as herwig @xcite , pythia @xcite or sherpa @xcite . during the parton shower evolution , which terminates at some scale of typical hadron mass , the recoil from the emitted gluons builds up a transverse momentum for the @xmath5 . in order to fit existing data
, the conventional backward evolution parton shower approach needs to be supplemented by the so - called ` intrinsic ' ( or ` primordial ' ) transverse momentum @xmath6 distribution of partons initiating the shower .
the physical motivation behind this additional non - perturbative ingredient is the fermi motion of partons within a hadron .
therefore , its average value per parton can be estimated based solely on the proton size and uncertainty principle to be of the order of @xmath7 @xmath8 gev .
but the values extracted from data first of all are too large and secondly grow with collision energy which can not be explained by fermi motion .
for example , in herwig++ its value grows from @xmath9 gev , which is needed to describe the data taken at the energy @xmath10 gev ( experiment r209 ) , to @xmath11 gev which , is needed at the tevatron energies ( @xmath12 gev ) .
this motivated us to propose a model for backward evolution in which an additional non - perturbative component at low transverse momentum provides additional smearing at each step of the evolution . by construction
we expect more non - perturbative smearing for longer parton shower evolution ladder which might cure the problem of dependence on centre of mass energy as well as on the size of needed intrinsic smearing , which in our studies is kept , according to the fermi motion argumentation fixed at @xmath13 gev
. in the following sections we will first briefly describe the model , then present how it fits the existing data sets , and at the end of this note we will demonstrate the model s predictions for the lhc energies which we use for a comparison with other approaches .
the implementation of transverse momentum production in which non - perturbative smearing takes place throughout the perturbative evolution , was achieved by a simple modification to an initial - state parton shower algorithm .
the model was implemented in the framework of herwig++ @xcite in which the sudakov form factor for backward evolution from some scale @xmath14 down to @xmath15 takes the form @xmath16 with @xmath17 , for further details cf.ref .
@xcite .
the argument of the strong coupling @xmath18 in eq .
( [ eq : sudakov ] ) is the transverse momentum @xmath19 of an emission is a slightly simplified expression , equal to the transverse momentum to the required accuracy , but not exactly .
we have tested the implementation of our model with this simplified expression and the exact expression for transverse momentum , and find very similar results .
we therefore use the default expression . ] . the cut - off scale represented by @xmath20
is needed to avoid divergence of the strong coupling .
below the cut - off scale @xmath18 is equal to zero and consequently the derivative of the sudakov form factor is equal to zero which translates to zero probability of the gluon emission below @xmath20 .
therefore , the two arguments of the sudakov form factor , @xmath21 and @xmath20 , are not the evolution variables but only explicitly specify the available phase - space of an emission . in order to populate the phase - space below @xmath20 by additional non perturbative emissions we introduce the additional sudakov form factor @xmath22 such that @xmath23 we achieve this by extending @xmath24 into the non - perturbative region using the following model @xmath25 in order to explore the possibility of a reasonable description of experimental data , we have studied in a greater detail two simple choices of the non perturbative function @xmath26 : flat continuation of @xmath27 , with a constant value @xmath28 , @xmath29 and a quadratic interpolation between the two values @xmath30 and @xmath31 : @xmath32 in both cases our model is determined by two free parameters @xmath20 and @xmath28
in this section we present some new results of the model which were obtained after important improvements of herwig++ s parton shower , released with version 2.3.1 of the program .
the main change in the program was a fix for a wrongly applied pdf veto in the parton shower @xmath33 splittings which , by construction of our model could have influence on previously presented results @xcite .
therefore , we have repeated the procedure described in detail in @xcite and have fitted the two parameters of our model to the drell - yan data from three experiments : the fixed target @xmath34cu fermilab e605 @xcite @xmath35gev , cern isr @xmath34@xmath34 collisions experiment r209 @xcite at @xmath36gev and cdf tevatron run i experiment with energies at @xmath37 gev @xcite .
these experiments cover the whole spectrum of centre mass energy for the drell - yan process data sets which are interesting for our studies .
( left ) , r209 @xmath38 ( middle ) and cdf @xmath39 ( right ) .
the monte carlo results come from our parameter set with @xmath40gev .
, title="fig : " ] ( left ) , r209 @xmath38 ( middle ) and cdf @xmath39 ( right ) .
the monte carlo results come from our parameter set with @xmath40gev .
, title="fig : " ] ( left ) , r209 @xmath38 ( middle ) and cdf @xmath39 ( right ) .
the monte carlo results come from our parameter set with @xmath40gev .
, title="fig : " ] in the case of purely parton - level shower , with all the light - quark and gluon effective masses and cutoffs set to zero , with our model for the low - scale @xmath18 as the only non - perturbative input the fitting procedure gave the optimal value for the quadratic extrapolation with @xmath41 and @xmath42 .
the resulting low-@xmath43 distributions for the new values are presented on top of the data sets in fig .
[ massless ] .
the @xmath44 values are a little higher than before the parton shower improvements , nevertheless the agreement with data remains at a high - level ; @xmath45 for all the experimental data sets are below or equal one .
if we are only interested in the @xmath5 transverse momentum distribution , it is enough to use a parton - level study , however , if one needs to simulate fully exclusive events then a hadronization model has to be used .
the hadronization model used in herwig++ requires termination of the shower using non - perturbative effective parton masses tuned to @xmath46 data .
( left ) , r209 @xmath47 ( middle ) and cdf @xmath48 ( right ) .
the monte carlo results are from our parameter set with @xmath49gev .
, title="fig : " ] ( left ) , r209 @xmath47 ( middle ) and cdf @xmath48 ( right ) .
the monte carlo results are from our parameter set with @xmath49gev .
, title="fig : " ] ( left ) , r209 @xmath47 ( middle ) and cdf @xmath48 ( right ) .
the monte carlo results are from our parameter set with @xmath49gev .
, title="fig : " ] therefore , we performed the same analysis as above but this time with restored tuned effective parton masses . in this case
the best and most stable situation was found for @xmath50 and @xmath51gev , giving the @xmath45 per degree of freedom of @xmath52 for cdf , @xmath53 for r209 and the worst @xmath54 for e605 .
we should stress that the used parameter set may not be the optimal choice for each experiment or cm energy but rather the best compromise between the three experiments . as the fixed target data in our analysis do not even include the systematic errors quoted to be around 510% , we have deliberately put a bit more emphasis on the tevatron results .
the first remark is that the new parameter choices for both the parton and hadron - level models , are not very different from the ones obtained using the old version of herwig++ . before
parton shower improvements our best choices were , for the parton - level mode : @xmath41 , @xmath55gev , and for the hadron - level case : @xmath56 and @xmath57gev @xcite .
we have also checked how the results depend on the intrinsic momentum @xmath58 by varying its value with @xmath59 gev around our fixed value @xmath60 , which is in the range permitted by the fermi motion .
we have repeated the fitting procedure and observed that for both intrinsic momenta , @xmath61 , we are able to find a pair of parameters for which our model gives equally good description of data sets as for the central value of @xmath60 .
moreover , we have observed that the value of @xmath62 parameter for all studied intrinsic momenta remains the same but the @xmath20 value is shifted for a bigger intrinsic momentum to a higher scale and for a smaller one to a lower scale .
therefore , by changing the intrinsic momentum from @xmath13 to @xmath8 gev we can obtain exactly the same best model s parameters set as in @xcite and the same shape of @xmath18 as presented in figure 4 from @xcite .
in that case the comparison of the shape of @xmath18 in the non perturbative region of the parton - level study are in good agreement with other approaches to modelling non - perturbative corrections to inclusive observables with a modified coupling in the soft region @xcite .
the last remark is that using our model as the only non - perturbative ingredient in the simulation , i.e. removing the non - perturbative constituent parton masses that usually cut off the parton shower in herwig++ , gives a somewhat better description of the data .
this lays open the speculation that perhaps , in some way , the two approaches could be combined .
one could for example use our model for the initial - state radiation , and the usual model , tuned to describe the final states of @xmath46 annihilation , for final - state radiation .
at the end of this note we would like to compare the results for a transverse momentum distribution of the @xmath0 boson at the lhc energies using the nonperturbative gluon emission model and two other approaches : resbos @xcite and the gaussian intrinsic @xmath63 extrapolation . but first let us compare our prediction of the parton level , marked as the filled histogram in fig .
[ fig : lhc - pt ] , and of the hadron level , dot dashed blue line .
both histograms , as expected , give a consistent extrapolation .
distribution at the lhc .
our model is compared to the extrapolation of gaussian intrinsic @xmath63 to lhc energies and the result from resbos .
[ fig : lhc - pt ] ] the result from resbos in fig .
[ fig : lhc - pt ] ( solid , black ) shows a slightly different behaviour from our predictions .
we predict a slightly more prominent peak and a stronger suppression towards larger transverse momenta .
both computations match the data well at large transverse momenta as they rely on the same hard matrix element contribution for a single hard gluon emission .
let us stress the remarkable feature that we predict the same peak position with these models which is very important from the experimental point of view .
this feature is quite understandable as both models are built on the same footing : extra emissions of soft gluons .
a comparison of resbos to data from experiments at various energies including the experiments e605 and r209 was done in @xcite .
furthermore , we see the herwig++ result from only using intrinsic @xmath64gev ( dashed , red ) as recommended in @xcitegev is adjusted to be in agreement with the prediction of the non - perturbative model presented in fig . [
fig : lhc - pt ] ] .
this large value stems from an extrapolation from lower energy data with the assumption that the average @xmath63 will depend linearly on @xmath65 .
the peak is seen to lie at a considerably higher value of the transverse momentum .
it would clearly be of interest to have experimental data to distinguish these two models of non perturbative transverse momentum .
we consider the model based on soft - gluon radiation , much like the resummation program resbos , to have a more meaningful physics input than simply extrapolating the gaussian smearing of a primordial transverse momentum .
the model implemented in the improved parton shower of herwig++ ( release 2.3.1 ) gives a good ( and very similar to the older version of herwig++ ) description of data . on the other hand
, the fitting procedure shows that the best values for the model s parameters are slightly different than the previous ones .
we would like to thank w. paczek , s. sapeta and all our colleagues from the herwig++ collaboration for useful discussions .
we gratefully acknowledge support from the eu marie curie research training network mcnet ( mrtn - ct-2006 - 035606 ) and the helmholtz alliance `` physics at the terascale '' .
one of us ( as ) also would like to acknowledge men grant n202 175435 ( 2008 - 10 ) for additional support .
99 w. j. stirling and m. r. whalley , j. phys .
g * 19 * ( 1993 ) d1 .
a. a. affolder _ et al .
_ [ cdf collaboration ] , phys .
* 84 * ( 2000 ) 845 .
b. abbott _ et al .
_ [ d0 collaboration ] , phys .
d * 61 * ( 2000 ) 032004 .
n. besson , m. boonekamp , e. klinkby , t. petersen and s. mehlhase [ atlas collaboration ] , eur .
j. c * 57 * ( 2008 ) 627 [ arxiv:0805.2093 [ hep - ex ] ] .
f. fayette , m. w. krasny , w. placzek and a. siodmok , arxiv:0812.2571 [ hep - ph ] . c. balzs , j. huston and i. puljak , phys .
d * 63 * ( 2001 ) 014021 .
m. w. krasny , f. fayette , w. paczek and a. sidmok , eur .
j. c * 51 * ( 2007 ) 607 .
t. sjstrand , phys .
b * 157 * ( 1985 ) 321 .
g. corcella _ et al .
_ , jhep * 0101 * ( 2001 ) 010 [ arxiv : hep - ph/0011363 ] . t. sjostrand , s. mrenna and p. skands , jhep * 0605 * ( 2006 ) 026 [ arxiv : hep - ph/0603175 ] .
t. gleisberg , s. hoche , f. krauss , m. schonherr , s. schumann , f. siegert and j. winter , jhep * 0902 * ( 2009 ) 007 [ arxiv:0811.4622 [ hep - ph ] ] . m. bahr _ et al .
_ , eur .
j. c * 58 * ( 2008 ) 639 [ arxiv:0803.0883 [ hep - ph ] ] .
s. gieseke , p. stephens and b. webber , jhep * 0312 * ( 2003 ) 045 .
s. gieseke , m. h. seymour and a. siodmok , jhep * 0806 * ( 2008 ) 001 [ arxiv:0712.1199 [ hep - ph ] ]
. g. moreno _ et al .
_ , phys .
d * 43 * ( 1991 ) 2815 .
d. antreasyan _ et al .
_ , phys .
* 48 * ( 1982 ) 302 .
. l. dokshitzer and b. r. webber , phys .
b * 352 * ( 1995 ) 451 .
l. dokshitzer , g. marchesini and b. r. webber , nucl .
b * 469 * ( 1996 ) 93 .
s. berge , p. nadolsky , f. olness and c. p. yuan , phys .
d * 72 * ( 2005 ) 033015 .
f. landry , r. brock , p. m. nadolsky and c. p. yuan , phys .
d * 67 * ( 2003 ) 073016 . | the transverse momentum of electroweak bosons in a drell - yan process is an important quantity for the experimental program at the lhc .
the new model of non - perturbative gluon emission in an initial state parton shower presented in this note gives a good description of this quantity for the data taken in previous experiments over a wide range of cm energy .
the model s prediction for the transverse momentum distribution of @xmath0 bosons for the lhc is presented and used for a comparison with other approaches . |
knowledge of the initial mass function ( imf ) of the first stars is critical to our understanding of the earliest stages of galaxy formation , one of the key frontiers of modern cosmology ( see e.g. barkana & loeb 2001 ; ciardi & ferrara 2005 ; bromm et al . 2009 ; benson 2010 ) . while the population ( pop ) iii imf is widely believed to have been top - heavy due to the limited cooling properties of primordial gas ( see e.g. bromm & larson 2004 ) , various scenarios for the formation of low - mass primordial stars have been discussed for some time ( e.g. palla et al . 1983 ; nakamura & umemura 2001 ; omukai & yoshii 2003 ; salvaterra et al .
2004 ; machida et al . 2005
; silk & langer 2006 ) .
most recently , high - resolution simulations of pop iii star formation have challenged the view that the first stars were very massive single objects formed in isolation ( e.g. abel et al .
2002 ; bromm et al .
2002 ) , instead suggesting that many formed as binaries ( turk et al . 2009 ; stacy et al . 2010 ) or in clusters of lower mass stars ( clark et al . 2010 ; greif et al .
these results also raise a challenge to the postulated existence of a critical metallicity , below which stars can only form with a top - heavy imf ( e.g. bromm & loeb 2003 ; santoro & shull 2006 ; schneider et al .
2006 ; frebel et al . 2007
; but see also jappsen et al .
2009 ) .
further motivation to consider the formation and fate of low - mass pop iii stars comes from the fact that no low - metallicity star has yet been found to exhibit the chemical signature of pair - instability supernovae ( sne ) , one of the primary predicted end states of very massive pop iii stars ( see e.g. fryer et al .
2001 ; heger et al .
2003 ) . while this observation in itself does not imply that low - mass pop iii stars formed , it is consistent with a pop iii imf that is not dominated by very massive stars ( see karlsson et al . 2008 ) ; instead , the observed chemical abundances of low - metallicity stars ( see beers & christlieb 2005 ; frebel 2010 ) suggest that they formed from material enriched by the explosions of 15 - 40 m@xmath5 pop iii stars ( e.g. umeda & nomoto 2003 ; iwamoto et al .
2005 ; joggerst et al .
altogether , there remains the distinct possibility that some fraction of pop iii stars had masses low enough ( @xmath0 0.8 m@xmath5 ) that they may have survived to the present day . in turn
, this raises once again the long - standing question ( e.g. bond 1981 ) : where are such low - mass pop iii stars today ?
one explanation for why pop iii stars may not be identified in the galaxy today is that they do not appear as such , their surfaces having been enriched due to the accretion of heavy elements from the interstellar medium ( ism ) during the course of their long lives .
enrichment by accretion has been studied by numerous authors , in the context of the present - day milky way ( e.g. mestel 1954 ; talbot & newman 1977 ; alcock & illarionov 1980 ; yoshii 1981 ; iben 1983 ; frebel et al .
2009 ) . using
recently obtained kinematic information on a large number of metal - poor stars , frebel et al .
( 2009 ) modelled the enrichment of these stars individually over the past @xmath1 10 gyr , finding that in almost all cases the amount of metals accreted onto the stars was far below what would be needed to explain their observed surface metallicities .
based on this result , it was concluded that the accretion of enriched material during passage through the disk of the galaxy has not significantly altered the surface composition of observed low - metallicity stars .
however , as the first stars formed during the earliest stages of the assembly of the milky way , it is important to also consider the degree to which the accretion of enriched material may have altered their surface compositions during this entire assembly process .
recently , this has been studied by komiya et al .
( 2009 , 2010 ) using the extended press - schechter formalism to model the hierarchical growth of the milky way .
their findings suggest that significant accretion of enriched material occurs soon after the formation of a pop iii star , before its host minihalo undergoes a merger . here
we present a complementary study in which we evaluate the likelihood that low - mass pop iii stars may avoid enrichment via accretion from the ism , using a simple model for their dynamical evolution during the hierarchical assembly of the galaxy .
in particular , we investigate the impact that a weak , solar - like stellar wind may have in limiting accretion , as it is perhaps only by such a wind that accretion onto a low - mass star can be completely prevented . in the next two sections ,
we review the criteria for the suppression of accretion by a stellar wind and we describe our model for the dynamical evolution of low - mass pop iii stars . in section 4 , we present our findings on the conditions for , and the likelihood of , enrichment by accretion from the ism in the assembly of the galaxy .
finally , we briefly discuss the implications of our findings and give our conclusions in section 5 .
just as in the case of the sun , the outer envelope of a low - mass pop iii star is expected to be convective ( e.g. yoshii 1981 ; fujimoto et al .
1995 ) . therefore , as in the sun
, this convection may drive a dynamo which produces a magnetic field emanating from the surface of the star . in turn , as the corona of the sun is likely generated by processes associated with its magnetic field ( e.g. hassler et al .
1999 ) , such a star may also have a hot corona from which a weak , solar - like wind is launched . to evaluate the effect that such a stellar wind would have in suppressing the accretion of material from the ism
, we follow talbot & newman ( 1977 ) and define a minimum density @xmath6 of the ism required for accretion to take place when a stellar wind is launched with the same density and speed as the solar wind : @xmath7 where @xmath8 is the velocity of the star with respect to the ism and @xmath9 is the sound speed of the gas , both in units of km s@xmath10 . where in its orbit the velocity of a star is large enough that @xmath6 exceeds the density @xmath11 of the ism through which the star passes , then the accretion of gas from the ism is prevented . for simplicity , in the following we shall adopt the formula for @xmath6 given by talbot & newman ( 1977 ) directly , although it is derived for the specific case of the sun
however , the radius of a pop iii star with mass @xmath1 0.8 m@xmath5 , which may still be on the main - sequence today , is expected to be comparable to the solar radius ( e.g. weiss et al . 2000 ; suda et al .
thus , we expect that the velocity of a wind from such stars , roughly the escape velocity , will also be similar to the solar value . in turn , if the density of the wind is also comparable to that of the solar wind , then the above formula is appropriate . lacking detailed knowledge of such winds from pop iii stars , we limit ourselves to considering the single case of a solar - like wind .
here we consider the nature of stellar orbits within the progenitors of the dark matter ( dm ) halo of the galaxy , in order to estimate the typical velocities of low - mass pop iii stars with respect to the ism . as the detailed properties of the high - redshift progenitor haloes of the milky way
are unknown , we begin by adopting the following simple form for the gravitational potential , consistent with a @xmath12 @xmath13 @xmath14 matter density profile ( white 1985 ) , as well as with the observed velocity curve of the galaxy ( xue et al . 2008 ) and with cosmological simulations of the formation of the first protogalaxies ( e.g. johnson et al . 2007 ; wise & abel 2008 ) : @xmath15 here @xmath16 is arbitrary and @xmath17 is the maximum circular velocity of a halo of mass @xmath18 , given by ( e.g. barkana & loeb 2001 ) @xmath19 a stellar orbit can then be defined by an apogalactic distance @xmath20 and an eccentricity @xmath21 , given by @xmath22 where @xmath23 is the perigalactic distance of the orbit ( e.g. chiba & beers 2000 ) . with these parameters , the specific orbital angular momentum @xmath24 of the star is given by @xmath25 and the velocity of the star at radius @xmath26 is @xmath27 we note that even for a very eccentric orbit with @xmath21 = 0.99 , the velocity @xmath28 of the star changes by only a factor ln(@xmath20/@xmath26 ) @xmath29 ln((1+@xmath21)/(1-@xmath21 ) ) @xmath1 5 throughout its orbit ; hence , to within a factor of order unity , @xmath8 @xmath1 @xmath17 . strictly speaking , the velocity @xmath8 we derive here is the velocity of the star relative to the center of its host dark matter halo .
however , for simplicity , we shall treat this as the velocity of the star relative to the ism . while this assumption is clearly not valid for the case of stars formed , for example , in the disk of the present - day milky way ( see e.g. talbot & newman 1977 ; yoshii 1981 ) , it is more tenable for pop iii stars formed at high redshift , as the orbits of such stars
would be altered numerous times in the frequent mergers that their host haloes undergo in the assembly of the galaxy .
thus , we expect that any extant low - mass pop iii stars are likely to have orbits that are in general not oriented in a particular direction with respect to the motion of the ism ( see e.g. carollo et al .
2007 ) , in which case a _
typical _ such star is indeed likely to have a velocity relative to the ism that is comparable to @xmath8 @xmath1 @xmath17 .
furthermore , in the calculations that follow , we shall neglect the possibility that intracloud motions are of the order of @xmath17 and thereby impact the value of @xmath6 within a cloud , as cosmological simulations of the formation of high redshift dwarf galaxies ( wise & abel 2007a ; greif et al .
2008 ) show that the velocity dispersion within clouds with @xmath11 @xmath30 @xmath6 are well below @xmath17 .
we note , however , that cloud collisions may lead to temporarily higher intracloud velocity dispersions .
here we consider the two primary situations in which accretion onto a single low - mass star with a solar - like wind may take place : from the dense gas at the center of the dm halo in which it forms and during passage through a dense gas cloud at a later time in the hierarchical assembly of the milky way .
for the latter case , we consider separately the most massive progenitor halo of the milky way , in which the earliest pop iii stars likely formed ( e.g. brook et al .
2007 ) , and the full population of progenitor haloes which may or may not have hosted pop iii stars , due to chemical and radiative feedback ( e.g. gao et al . 2010 ) . for the density @xmath11 of the ism in a dm halo hosting the formation of a pop iii star at high redshift , here we consider the case of an isothermal density profile with the following scaling , roughly consistent with the properties of the gas in pop iii star - forming haloes , as found in numerical simulations ( e.g. oshea & norman 2007 ; wise & abel 2007b ) :
@xmath31 where the number density of baryons at the virial radius @xmath32 is given by @xmath33 here @xmath34 is the ratio of the cosmic baryon density to the critical density @xmath35 of the universe at redshift @xmath3 = 0 , and @xmath36 is the mass of the hydrogen atom . to determine where in a halo of mass
@xmath18 a star with a solar - like wind may accrete gas , we compare the gas density , given by equation ( 7 ) , to the minimum density required for accretion , given by equation ( 1 ) . taking , as discussed in section 3 , @xmath8 = @xmath17 , and assuming a sound speed @xmath9 @xmath0 @xmath17 , we have that the condition for accretion ( i.e. @xmath11 @xmath37 @xmath6 ) is only satisfied if @xmath38 simplifying this , we find that the dependences on the halo mass and redshift cancel out , leaving us with the following condition for accretion : @xmath39 that is , for the case of an isothermal gas at the virial temperature of its host halo , a star will only accrete if it is on an orbit that brings it within a physical distance of @xmath1 10 pc of the center of the halo .
given this condition , if dynamical interactions result in the ejection of a low - mass pop iii star from the central @xmath1 10 pc of the halo in which it forms , then the star will be prevented from accreting material .
if the pop iii imf is top - heavy , albeit with some low - mass stars forming , then dynamical interactions during the formation of stellar clusters are likely to eject the low - mass stars at higher velocities than in the case of present - day star formation ( e.g. greif et al .
2010a ) , which can be up to a few km s@xmath10 ( e.g. bate et al . 2003 ) .
as such ejection velocities can be comparable to the virial velocity of the minihalo hosting pop iii star formation , this could result in such low - mass stars being kicked out of the centers of their host haloes at early times .
indeed , it may be the case that low - mass stars form _ because _ they are kicked out of the center of the halo , the accretion of dense gas being halted before the star can grow to be massive ( e.g. bate et al . 2003 ; clark et al . 2010 ) .
note that even for an ejection velocity of just 3 - 5 km s@xmath10 , a low - mass pop iii star would travel 10 pc from the center of its host minihalo within the 2 - 3 myr lifetime of the most massive stars that explode as sne and enrich the ism .
therefore , low - mass pop iii stars may avoid accretion of metals from within their host haloes as a natural consequence of the process of their formation .
we also note that mergers of star - forming haloes , which for the minihalo progenitors of the first galaxies occur on timescales of 10@xmath40 - 10@xmath41 myr ( e.g. greif et al .
2008 ; gao et al .
2010 ) , are likely to result in the violent relaxation of the stars in the halo , and this may also serve to kick low - mass stars to larger radii in the merged host halo .
a further impediment to the accretion of metals is that this will not occur until the sne ejecta from massive pop iii stars re - collapses to high densities within the host halo , which generally occurs on a timescale of @xmath1 10@xmath41 yr ( see wise & abel 2008 ; greif et al .
2010b ; but see also whalen et al .
prior to this re - collapse , the ejecta will be at low density and may also be moving with high velocities , making accretion very unlikely .
if dynamical interactions within the initial stellar cluster or violent relaxation as a result of a merger , take place within this time , then a low - mass star is likely to avoid accretion in its host minihalo altogether . while the condition we have just found dramatically limits the range of circumstances in which a low - mass pop iii star can accrete metal - enriched gas within the halo in which it forms , it is still possible for accretion to take place if the star passes through a sufficiently dense molecular cloud in a more massive halo at a later time
taking it that the star has a velocity relative to a dense cloud of @xmath8 @xmath1 @xmath17 , and using equation ( 7 ) with the redshift dependence of @xmath32 , we find the minimum density @xmath6 of a cloud required for accretion to be @xmath42 for a molecular cloud with with this density and with a characteristic size @xmath43 @xmath1 10 pc ( e.g. spitzer 1978 ) , assuming the standard bondi ( 1952 ) accretion rate , we find that this would result in the accretion of a total mass of @xmath1 10@xmath44 m@xmath5 , leading to a stellar surface metallicity of @xmath45 where @xmath46 is the metallicity of the gas in the cloud , @xmath47 @xmath0 0.8 m@xmath5 is the mass of the star , and @xmath48 is the fraction of the stellar mass contained in the outer convective layer ( e.g. yoshii 1981 ) .
we have normalized this formula to the results of recent cosmological simulations of the earliest episodes of metal enrichment via powerful pop iii sne ( wise & abel 2008 ; greif et al .
2010b ) , from which it has been found that the metallicity of the gas that re - collapses into haloes of mass @xmath18 @xmath1 10@xmath41 m@xmath5 at @xmath3 @xmath30 10 is of the order of @xmath46 @xmath1 10@xmath49 @xmath50 .
this metallicity is likely to be roughly an order of magnitude lower if the pop iii imf is not top - heavy ( see e.g. tumlinson 2010 ; komiya et al .
2010 ) . for the case of the milky way at @xmath3
= 0 , with a typical cloud size @xmath43 @xmath1 10 pc , @xmath18 = 10@xmath51 m@xmath5 , and @xmath46 = z@xmath5 , we find @xmath52 @xmath1 10@xmath53 z@xmath5 , in basic agreement with the estimate of the surface enrichment for the similar ( but with @xmath43 = 100 pc ) case presented by frebel et al .
( 2009 ) . in the following ,
we consider the probability of a star encountering a cloud of sufficient density that such enrichment via accretion may take place . an important specific case to consider is the probability of accretion onto low - mass pop iii stars formed in the main progenitor halo of the galaxy , as this is the most likely formation site of the earliest pop iii stars , which formed before radiative or chemical feedback could begin to inhibit primordial star formation ( e.g. brook et al .
2007 ; gao et al . 2010 ) . for this
, we make use of the results from recent , high - resolution cosmological simulations of the formation of milky way - like haloes ( boylan - kolchin et al .
these authors provide a best - fit formula for the mass assembly histories of their simulated haloes , as a function of redshift : @xmath54 \mbox{\ , } \ ] ] where we take the present - day mass of the halo to be @xmath55 = 10@xmath51 m@xmath5 . comparing this formula with the results of zhao et al .
( 2009 ) on the early growth of milky way - like haloes , we find good agreement up to at least @xmath3 @xmath1 20 ( see also gao et al .
2010 ) , at which point pop iii star formation was likely ongoing ( e.g. bromm & larson 2004 ) .
therefore , we shall use this formula for the growth history of the halo from @xmath3 = 20 to the present - day in our calculations .
next , we shall assume the same density dependence of the volume - filling fraction @xmath56(@xmath11 ) of gas clouds with gas density @xmath37 @xmath11 as found by talbot & newman ( 1977 ) .
in particular , we shall take it that @xmath57 where @xmath58 = 0.9 for @xmath11 @xmath29 10@xmath59 cm@xmath49 and @xmath58 = 3 for @xmath11 @xmath4 10@xmath59 cm@xmath49 .
we note that this value of @xmath58 for @xmath11 @xmath4 10@xmath61 cm@xmath49 is also broadly consistent with the results of the recent _ bolocam galactic plane survey _ of dense molecular clouds ( rosolowsky et al .
as the gas content of the galaxy today is lower than in the past due to star formation and feedback , in equation ( 14 ) we have chosen a normalization for @xmath56 that is a factor of 10 larger than that derived by talbot & newman for the case of the present - day milky way , in order to conservatively account for the higher gas fraction in the high redshift progenitors of the galaxy .
assuming that the star formation rate ( sfr ) within a halo is proportional to the mass of dense gas ( e.g. @xmath11 @xmath30 30 cm@xmath49 ; wolfire et al .
2003 ; krumholz et al .
2009 ) in the halo , that the mass of dense gas is proportional to the total halo mass , and that the sfr in the galaxy today is of the order of 1 m@xmath5 yr@xmath10 , this normalization yields a sfr of order 10@xmath2 m@xmath5 yr@xmath10 for dwarf galaxies in @xmath1 10@xmath62 m@xmath5 haloes at @xmath3 @xmath0 10 , in broad agreement with the sfrs found in recently high - resolution cosmological simulations ( wise & cen 2009 ; razoumov & sommer - larsen 2010 ) . as we will show , the probability for accretion in the halo is much higher at high redshift than in the present - day milky way , and thus our choice of an elevated volume - filling fraction of dense gas compared to the present - day galaxy does not appreciably affect our results for the overall probability of accretion , even though it is clearly an overestimate at low redshift .
using the mass accretion history of the main progenitor halo given by equation ( 13 ) and the volume - filling fraction @xmath56 given by equation ( 14 ) , we estimate the number of cold ( @xmath9 @xmath63 @xmath17 ) , dense clouds @xmath64(@xmath3 ) with @xmath11 @xmath30 @xmath6 ( given by equation 11 ) through which a star has passed between redshift @xmath3 and the present - day as @xmath65 where @xmath66 @xmath1 @xmath20 / @xmath17 is the orbital period of the star , and @xmath67(@xmath3 ) is age of the universe at redshift @xmath3 . while it is largely uncertain what we should choose for the apogalactic distance of the stellar orbit , for our current purposes we choose @xmath20 = 0.1 @xmath32 , to be consistent with simulations which suggest that pop iii stars formed in the high - redshift progenitors of the galaxy end up in the central @xmath1 0.1 @xmath32 of the halo at @xmath3 = 0 ( e.g. diemand et al .
2005 ; scannapieco et al .
2006 ; brook et al . 2007 ; de lucia & helmi 2008 ; madau et al . 2008 ; salvadori et al . 2010 ; gao et al .
we shall relax this assumption in the next section , in which we consider the broader question of accretion onto pop iii stars formed within the full population of milky way progenitor haloes .
we emphasize that the formula for @xmath56 that we use here ( equation 14 ) was derived by talbot & newman ( 1977 ) as the volume - filling fraction of the disk of the milky way .
therefore , our estimate of the number @xmath64 of dense clouds encountered implicitly assumes that at every orbit a star passes through the central , star - forming regions of its host halo corresponding to the disk of the galaxy today .
as stars may instead orbit completely outside the central regions of the halo ( i.e. at radii @xmath30 0.1 @xmath32 ) , in this sense equation ( 15 ) is an upper limit for @xmath64 .
evaluating the integral in equation ( 15 ) yields a probability of accretion after redshift @xmath3 of @xmath68(@xmath69 ) @xmath1 1 - @xmath70 , which is shown in figure 1 .
one trend that is evident is that the probability for accretion is much higher at high redshift .
there are two reasons for this .
firstly , the maximum circular velocity @xmath17 of the halo is lower at higher redshift , and hence the volume - filling factor of gas at densities above @xmath6 is higher .
secondly , the orbital period @xmath66 of stars that we have chosen is also lower at high redshift , leading to more frequent passages through the ism during which accretion may take place .
we also note that there is a probability of only @xmath68 @xmath0 0.2 of accretion onto stars formed at @xmath3 @xmath0 20 , and a probability of only @xmath68 @xmath1 10@xmath71 of accretion onto such stars during passages through the disk of the galaxy over the past @xmath1 10 gyr , as treated in frebel et al .
( 2009 ) . from this
we conclude that , if they have solar - like winds , the majority of any extant low - mass pop iii stars formed in the main progenitor halo of the milky way are likely to exhibit pristine surface compositions today , having been unaffected by the accretion of metal - enriched gas . while the earliest pop iii stars are most likely to form in the main progenitor halo of the galaxy , the majority may form in other , smaller progenitor haloes .
here we evaluate the probability of accretion onto low - mass pop iii stars in these haloes .
for this purpose , we employ the extended press - schechter ( e.g. lacey & cole 1993 ) formalism to construct a merger tree of the progenitor haloes of a milky way - like halo , with a mass of 10@xmath51 m@xmath5 at @xmath3 = 0 .
in particular , we employ the monte - carlo method suggested by somerville & kolatt ( 1999 ) .
as we have shown , in haloes of lower mass , accretion is more likely to occur , due principally to the fact that a star moving at lower velocity @xmath8 @xmath1 @xmath17 can accrete gas at a lower density , which in turn has a larger volume - filling factor @xmath56 .
therefore , the minimum halo mass for which accretion of metals can take place is a critical parameter in our calculation . as discussed in section 4.1 ,
a minimum halo mass of 10@xmath72 m@xmath5 is likely a good estimate ( e.g. greif et al .
2010b ) , as this is the typical mass of a halo that is able to re - capture the metal - enriched ejecta of the first pop iii sne , which typically occur in the 10@xmath73 - 10@xmath74 m@xmath5 progenitors of these haloes ( e.g. bromm & larson 2004 ; trenti & stiavelli 2009 ) .
similar to the case of the main progenitor halo described in section 4.2.1 , we estimate the probability of accretion onto a low - mass pop iii star by integrating equation ( 15 ) from @xmath3 = 0 to the redshift at which a halo with the minimum mass forms .
however , to account for the numerous halo mergers that alter the orbital parameters of a star , and in particular the apogalactic distance of its orbit , instead of assuming @xmath20 = 0.1 @xmath32 as before , we now use @xmath20 = @xmath75 @xmath32 , where @xmath75 is a random number between zero and unity that is updated after each merger .
this yields the orbital period of a star in a given halo @xmath66 = @xmath75 @xmath32 / @xmath17 . also , as in komiya et al .
( 2010 ) , we account for the delay in a star being incorporated into its new host halo after a merger , which is taken to be the timescale for dynamical friction given by springel et al .
similar to the case of accretion onto pop iii stars with no winds considered by komiya et al . , we find that the accounting for this delay results in somewhat higher accretion probabilities , especially for stars formed in haloes at relatively low @xmath3 that undergo mergers with much more massive haloes ( see discussion below ) .
figure 2 shows the probability that low - mass pop iii stars within haloes with mass @xmath37 10@xmath41 m@xmath5 accrete metal - enriched gas during the assembly of the galaxy .
the left panel shows the probability of accretion onto stars hosted by all of the 10@xmath41 m@xmath5 progenitor haloes , as a function of the redshift at which they form .
the right panel shows the distribution of these probabilities ; assuming that each 10@xmath41 m@xmath5 halo hosts the same number of low - mass pop iii stars , this can be translated as the distribution of probabilities of accretion onto all pop iii stars in the halo of the galaxy today .
as shown by the black histogram in the right panel , the probability of accretion is generally low , with the stars in most haloes having a probability of @xmath0 0.1 to accrete metal - enriched gas .
we note that , although we have simply assumed that the stars have random orbits in their host halos , this conclusion holds for different random realizations of these orbits ( i.e. different @xmath75 ) .
interestingly , as shown in the left panel of fig .
2 , the haloes for which accretion is most likely are those which form at relatively low redshift , @xmath3 @xmath0 10 .
this is a result of the fact that such late - forming haloes also grow slowly , and so there is more time available for accretion of relatively low - density gas in these haloes compared to those that grow more rapidly at higher redshift .
another important effect is that a portion of the late - forming 10@xmath41 m@xmath5 haloes also undergo mergers with much more massive haloes than at higher @xmath3 . as the timescale for dynamical friction in these mergers can be of the order of gyr
, the stars can orbit within the haloes of lower mass ( and hence lower @xmath17 ) for much longer times than they would at higher @xmath3 . in turn , this results in considerably higher accretion probabilities for some stars formed at relatively low @xmath3 ; however , the effect of dynamical friction on the overall probability distribution , shown in the right panel of fig .
2 , is not large , as also found by komiya et al .
( 2010 ) in a similar calculation . however , it is important to note that the fraction of such haloes at low redshift that actually host pop iii stars may be relatively low .
this is due to feedback effects from stars formed in neighboring haloes , such as metal enrichment ( e.g. maio et al .
2010 ; greif et al .
2010b ) and photoevaporation ( e.g. shapiro et al .
2004 ; whalen et al .
2008b ) , which will act to diminish the formation rate of pop iii stars ( see also tsujimoto et al . 1999
; hernandez & ferrara 2001 ) . even neglecting the suppression of pop iii star formation due to metal enrichment and photoionization by massive stars , the suppression due to the photodissociation of molecules ( e.g. haiman et al .
1997 ) is likely to be significant on its own . assuming a physical number density of atomic - cooling haloes of @xmath1 0.03 kpc@xmath49 at @xmath3 @xmath0 20 , as found from simulations of the formation of milky way - like haloes ( gao et al .
2010 ) , either massive stars ( e.g. yoshida et al . 2003 ) or rapidly accreting black holes ( e.g. greif et al .
2008 ; johnson et al .
2010 ) within these galaxies would produce a background molecule - dissociating flux of the order of @xmath76 @xmath1 10@xmath2 ( i.e. 10@xmath2 @xmath77 10@xmath78 erg s@xmath10 cm@xmath2 hz@xmath10 sr@xmath10 ) , high enough to suppress the rate of formation of pop iii stars in minihaloes ( e.g. oshea & norman 2008 ) .
furthermore , the probabilities of accretion onto pop iii stars forming during and after reionization ( e.g. at @xmath3 @xmath0 10 ) ( johnson 2010 ) are upper limits , since the 10@xmath41 m@xmath5 haloes in which they form will have a substantial amount of the gas photoevaporated out of the halo and hence not available for accretion ( see e.g. thoul & weinberg 1996 ; gnedin 2000 ; dijkstra et al .
2004 ) . to evaluate the impact that external chemical and radiative feedback may have on the probability of accretion onto extant low - mass pop iii stars , the red histogram in the right panel of fig .
2 shows the distribution of accretion probabilities in the subset of haloes formed at @xmath3 @xmath4 10 . the distribution of probabilities for this subset of haloes peaks at somewhat lower values than that of the overall population , with a typical probability for accretion onto a star within these haloes of the order of 10@xmath2 .
therefore , we conclude that if chemical enrichment is rapid or radiative feedback is strong during the assembly of the milky way , then the accretion of metal - enriched gas onto a given low - mass pop iii star is quite unlikely . to account for the possibility that the metal - enriched sne ejecta of the first stars can be contained in somewhat less massive haloes ( see e.g. kitayama & yoshida 2005 ; whalen et al .
2008a ) , we have also carried out the same exercise with a minimum halo mass for accretion of 10@xmath74 m@xmath5 . in this case
, we find that the probability distribution for all such haloes peaks at @xmath1 0.2 , and that that for the subset forming at @xmath3 @xmath4 10 peaks at @xmath1 0.1 . while the probability for accretion in this case is considerably higher , we can conclude that even in this case the majority of low - mass pop iii stars are not likely to have accreted metal - enriched gas .
we have carried out an analysis of the likelihood that low - mass pop iii stars with primordial surface compositions could be found in the present - day milky way , under the assumption that such stars have solar - like stellar winds that prevent the accretion of metal - enriched material from the ism .
we have found that if low - mass pop iii stars are kicked out of the central @xmath1 10 pc of the haloes in which they form , either via dynamical interactions with more massive stars or from violent relaxation during an early merger , then the accretion of gas onto the stars can be prevented , at least until the star is incorporated into a larger halo in which dense molecular clouds are formed outside of the central @xmath1 10 pc ( e.g. pawlik et al .
2010 ) . to evaluate the likelihood that a star accretes gas during the passage through dense molecular clouds at later stages in the hierarchical assembly of the galaxy , we have modeled the dynamics of pop iii stars in their host dark matter haloes along with the density distribution of gas within these haloes .
overall , we find that the probability of enrichment via accretion is @xmath0 0.1 , and is even lower for pop iii stars formed at @xmath3 @xmath30 10 , at which times the majority of primordial star formation may have occurred ( e.g. maio et al .
2010 ) .
the primary uncertainty affecting our calculations is in the volume - filling fraction @xmath56 of dense clouds in the progenitor haloes of the milky way .
as discussed in section 4.2.1 , we have made the simple , conservative choice of a constant volume - filling fraction . while this is likely an overestimate at low redshift
, it does yield broad agreement with cosmological simulations of star formation in dwarf galaxies in haloes of mass @xmath30 10@xmath41 m@xmath5 at high redshift ( i.e. @xmath3 @xmath4 2 ) , which we have shown are the most likely sites for accretion onto low - mass pop iii stars .
we note , however , that existing observations of relatively massive galaxies ( i.e. within haloes of mass @xmath18 @xmath30 10@xmath80 m@xmath5 ) at @xmath3 @xmath0 2 suggest a steep increase of the star formation rate , which is likely to be proportional to the amount of dense gas , with redshift ( e.g. bouch ' e et al .
it is possible that such a redshift dependence , which we have not sought to model , extends to the less massive progenitors of the milky way at higher redshift .
more precise calculations of the probability of accretion onto low - mass stars will have to await observations , for instance by the _
atacama large millimeter array _
( e.g. combes 2010 ) , which provide stronger constaints on the properties of the dense gas in high redshift galaxies .
we emphasize that the low probability we find for accretion onto low - mass pop iii stars is limited to the case that such stars have weak , solar - like stellar winds . as previous authors have shown
, the amount of metal enrichment via accretion can be much higher for stars with no winds ( frebel et al . 2009 ; komiya et al . 2010 ) . importantly , unless low - mass pop iii stars do have winds , such enrichment via accretion is likely to be an inevitability and the likelihood of finding truly pristine pop iii stars in the galaxy today is practically zero .
our results thus suggest that it may be possible to test , unequivocally , theoretical predictions of whether or not low - mass pop iii stars formed in the early universe , as we have found that any such stars are likely to still exhibit primordial surface compositions in the galaxy today .
the assessment of these theories may ultimately require that such pristine stars be found ( see e.g. weiss et al .
2000 ) in large surveys of metal - poor stars , such as the _ sloan extension for galactic understanding and exploration _
( e.g. yanny et al . 2009 ) , the _ radial velocity experiment _
( e.g. fulbright et al .
2010 ) , the _ apache point observatory galactic evolution experiment _
( e.g. allende prieto et al . 2008 ) , the _ large sky area multi - object fiber spectroscopic telescope _ project ( e.g. zhao et al .
2006 ) , and the hermes project .
if , however , they are not discovered , then it may be for one of two reasons : either low - mass pop iii stars do not have sufficiently strong winds to prevent enrichment via accretion , or long - lived pop iii stars were not formed in sufficiently high numbers in the progenitor haloes of the milky way ( see e.g. madau et al .
2008 ) for them to be found in observational campaigns .
jlj is grateful to thomas greif for sharing his results on the kinematics of low - mass pop iii stars in minihaloes ahead of their publication , as well as for comments on an early draft of this work .
the authors would like to thank fabrice durier , volker gaibler , and claudio dalla vecchia for valuable feedback and discussion , as well as to the anonymous reviewer for a helpful report .
99 abel t. , bryan g. l. , norman m. l. 2002 , sci , 295 , 93 alcock c. , illarionov a. 1980 , apj , 235 , 541 allende - prieto c. , et al .
2008 , an , 329 , 1018 barkana r. , loeb a. 2001 , phr , 349 , 125 bate m. r. , bonnell i. a. , bromm v. 2003 , mnras , 339 , 577 beers t. c. , christlieb n. 2005 , ara&a , 43 , 531 benson a. j. 2010 , phr , 495 , 33 bond h. e. 1981 , apj , 248 , 606 bondi h. 1952 , mnras , 112 , 195 bouch ' e n. , et al . 2010 , apj , 718 , 1001 boylan - kolchin m. , springel v. , white s. d. m. , jenkins a. 2010 , mnras , 406 , 896 bromm v. , coppi p. s. , larson r. b. 2002 , apj , 564 , 23 bromm v. , larson r. b. 2004 , ara&a , 42 , 79 bromm v. , yoshida n. , hernquist l. , mckee c. f. 2009 , nat , 459 , 49 v. , loeb a. 2003 , nat , 425 , 812 brook c. b. , kawata d. , scannapieco e. , martel h. , gibson b. k. 2007 , apj , 661 , 10 carollo d. , et al .
2007 , nat , 450 , 1020 combes f. 2010 , aip conference series , in press ( arxiv:1004.4364 ) chiba m. , beers t. c. 2000 , apj , 119 , 2843 ciardi b. , ferrara a. 2005 , ssrv , 116 , 625 clark p. c. , glover s. c. o. , klessen r. s. , bromm v. 2010 , apj , submitted ( arxiv:1006.1508 ) de lucia g. , helmi a. 2008 , mnras , 391 , 14 diemand j. , madau p. , moore b. 2005 , mnras , 364 , 367 dijkstra m. , haiman z. , rees m. j. , weinberg d. h. 2004 , apj , 601 , 666 frebel a. 2010 , an , 331 , 474 frebel a. , johnson j. l. , bromm v. 2007 , mnras , 380 , l40 frebel a. , johnson j. l. , bromm v. 2009 , mnras , 392 , l50 fryer c. l. , woosley s. e. , heger a. 2001 , apj , 550 , 372 fujimoto m. y. , sugiyama k. , iben i. , hollowell d. 1995 , apj , 444 , 175 fulbright j. p. , et al .
2010 , apjl , accepted ( arxiv:1010.4491 ) gao l. , theuns t. , frenk c. s. , jenkins a. , helly j. c. , navarro j. , springel v. , white s. d. m. 2010 , mnras , 403 , 1283 gnedin n. y. 2000 , apj , 542 , 535 greif t. h. , springel v. , white s. d. m. , glover s. c. o. , clark p. c. , smith r. j. , klessen r. s. , bromm v. 2010a , submitted greif t. h. , glover s. c. o. , bromm v. , klessen r. s. 2010b , apj , 716 , 510 greif t. h. , johnson j. l. , klessen r. s. , bromm v. 2008 , mnras , 387 , 1021 haiman z. , rees m. j. , loeb a. 1997 , apj , 476 , 458 hassler d. m. , dammasch i. e. , lemaire p. , brekke p. , curdt w. , mason h. e. , vial j .-
c . , wilhelm k. 1999 , sci , 283 , 810 heger a. , fryer c. l. , woosley s. e. , langer n. , hatmann d. h. 2003 , apj , 591 , 288 hernandez x. , ferrara a. 2001 , mnras , 324 , 484 iben i. 1983 , mmsai , 54 , 321 iwamoto n. , umeda h. , tominaga n. , nomoto k. , maeda k. 2005 , sci , 309 , 451 jappsen a .- k .
, mac low , m .- m .
, glover s. c. o. , klessen r. s. , kitsionas s. 2009 , apj , 694 , 1161 joggerst c. c. , almgren a. , bell j. , heger a. , whalen d. , woosley s. e. 2010 , apj , 709 , 11 johnson j. l. 2010 , mnras , 404 , 1425 johnson j. l. , greif t. h. , bromm v. 2007 , apj , 665 , 85 johnson j. l. , khochfar s. , greif t. h. , durier f. 2010 , mnras , in press ( arxiv:1007.3849 ) karlsson t. , johnson j. l. , bromm v. 2008 , apj , 679 , 6 kitayama t. , yoshida n. 2005 , apj , 630 , 675 komiya y. , habe a. , suda t. , fujimoto m. y. 2009 , apj , 696 , l79 komiya y. , habe a. , suda t. , fujimoto m. 2010 , apj , 717 , 542 krumholz m. r. , mckee c. f. , tumlinson j. 2009 , apj , 693 , 216 lacey c. , cole s. 1993 , mnras , 262 , 627 machida m. n. , tomisaka k. , nakamura f. , fujimoto m. y. 2005 , apj , 622 , 39 madau p. , kuhlen m. , diemand j. , moore b. , zemp m. , potter d. , stadel j. 2008 , apj , 689 , l41 maio u. , ciardi b. , dolag k. , tornatore l. , khochfar s. 2010 , mnras , 407 , 1003 mestel l. 1954 , mnras , 114 , 437 nakamura f. , umemura m. 2001 , apj , 548 , 19 omukai k. , yoshii y. 2003 , apj , 599 , 746 oshea b. w. , norman m. l. 2007 , 654 , 66 oshea b. w. , norman m. l. 2008 , apj , 673 , 14 palla f. , salpeter e. e. , stahler s. w. 1983 , apj , 271 , 632 pawlik a. h. , milosavljevi ' c m. , bromm v. 2010 , apj , submitted ( arxiv:1011.0438 ) razoumov a. o. , sommer - larsen j. 2010 , apj , 710 , 1239 rosolowsky e. , et al .
2010 , apjs , 188 , 123 salvadori s. , ferrara a. , scheider r. , scannapieco e. , kawata d. 2010 , mnras , 401 , l5 salvaterra r. , ferrara a. , schneider r. 2004 , newa , 10 , 113 santoro f. , shull j. m. 2006 , apj , 643 , 26 scannapieco e. , kawata d. , brook c. b. , schneider r. , ferrara a. , gibson b. k. 2006 , apj , 653 , 285 schneider r. , omukai k. , inoue a. k. , ferrara a. 2006 , mnras , 369 , 1437 shapiro p. r. , iliev i. t. , raga a. c. 2004 , mnras , 348 , 753 silk j. , langer m. 2006 , mnras , 371 , 444 somerville r. s. , kolatt t. s. 1999 , mnras , 305 , 1 spitzer l. 1978 , physical processes in the interstellar medium , wiley , new york springel v. , white s. d. m. , tormen g. , kauffmann g. 2001 , mnras , 328 , 726 stacy a. , greif t. h. , bromm v. 2010 , mnras , 403 , 45 suda t. , fujimoto m. y. , itoh n. 2007 , apj , 667 , 1206 talbot r. j. , newman m. j. 1977 , apjs , 34 , 295 thoul a. a. , weinberg d. h. 1996 , apj , 465 , 608 tsujimoto t. , shigeyama t. , yoshii y. 1999 , apj , 519 , l63 trenti m. , stiavelli m. 2009 , apj , 694 , 879 tumlinson j. 2010 , apj , 708 , 1398 turk m. j. , abel t. , oshea b. 2009 , sci , 325 , 601 umeda h. , nomoto k. 2003 , nat , 422 , 871 weiss a. , cassisi s. , schlattl h. , salaris m. 2000 , apj , 533 , 413 whalen d. , van veelen b. , oshea b. w. , norman m. l. 2008a , apj , 682 , 49 whalen d. , oshea b. w. , smidt j. , norman m. l. 2008b , apj , 679 , 925 white , s. d. m. 1985 , mnras , 294 , l99 wise j. h. , abel t. 2007a , apj , 665 , 899 wise j. h. , abel t. 2007b , apj , 671 , 1559 wise j. h. , abel t. 2008 , apj , 685 , 40 wise j. h. , cen r. 2009 , apj , 693 , 984 wolfire m. g. , mckee c. f. , hollenbach d. , tielens a. g. g. m. 2003 , apj , 587 , 278 xue x. x. , et al .
2008 , apj , 684 , 1143 yanny b. , et al .
2009 , aj , 137 , 4377 yoshida n. , abel t. , hernquist l. , sugiyama n. 2003 , apj , 592 , 645 yoshii y. , 1981 , a&a , 97 , 280 zhao d. h. , jing y. p. , mo h. j. , b " orner g. 2009 , apj , 707 , 354 zhao g. , chen y .- q . , shi j .-
, liang y .- c .
, chen l. , zhang h .- w . ,
2006 , chjaa , 6 , 265 | motivated by recent theoretical work suggesting that a substantial fraction of population ( pop ) iii stars may have had masses low enough for them to survive to the present day , we consider the role that the accretion of metal - enriched gas may have had in altering their surface composition , thereby disguising them as pop ii stars .
we demonstrate that if weak , solar - like winds are launched from low - mass pop iii stars formed in the progenitors of the dark matter halo of the galaxy , then such stars are likely to avoid significant enrichment via accretion of material from the interstellar medium .
we find that at early times accretion is easily prevented if the stars are ejected from the central regions of the haloes in which they form , either by dynamical interactions with more massive pop iii stars , or by violent relaxation during halo mergers .
while accretion may still take place during passage through sufficiently dense molecular clouds at later times , we find that the probability of such a passage is generally low ( @xmath0 0.1 ) , assuming that stars have velocities of order the maximum circular velocity of their host haloes and accounting for the orbital decay of merging haloes . in turn , due to the higher gas density required for accretion onto stars with higher velocities , we find an even lower probability of accretion ( @xmath1 10@xmath2 ) for the subset of pop iii stars formed at @xmath3 @xmath4 10 , which are more quickly incorporated into massive haloes than stars formed at lower redshift .
while there is no a priori reason to assume that low - mass pop iii stars do not have solar - like winds , without them surface enrichment via accretion is likely to be inevitable .
we briefly discuss the implications that our results hold for stellar archaeology .
-1 cm accretion galaxy : halo formation stars : low - mass population iii kinematics and dynamics abundances |
Subsets and Splits