Datasets:

ccdv
/

arxiv-summarization

Dataset card Data Studio Files Files and versions Community

Subset (2)

section · 216k rows

Split (3)

train · 203k rows

article stringlengths 0 745k	abstract stringlengths 0 92.7k
we study the cauchy problem @xmath3 where @xmath1 is a network and the operator @xmath4 is continuous , convex , nonnegative and positive homogeneous in the last variable . in @xmath5 , the problem arises in flame propagation models and evolution of curves whose speed of propagation only depends on the normal direction . existence , uniqueness and evolution of level sets of the solution of have been extensively studied in the framework of viscosity solution theory ( see @xcite ) . the unique viscosity solution of is given by the hopf - lax formula @xmath6 where @xmath7 is a distance function characterized by solving the associated stationary equation @xmath8 in the recent time , there is an increasing interest in the study of nonlinear differential equations on networks since they describe various phenomena as traffic flow , blood circulation , data transmission , electric networks , etc ( see @xcite ) . concerning hamilton - jacobi equations on networks , we mention the recent papers @xcite . in this paper , following the approach in @xcite , we cope existence , uniqueness and regularity for evolutive hamilton - jacobi equations on networks . in particular we prove that the hopf - lax formula can be extended to this framework . the main issue of our investigation is to tackle transition vertices ( namely , points of the network where several edges meet each other ) . actually , in our framework a suitable definition of viscosity solution at transition vertices ( together with the standard one at points inside edges ) will ensure the well posedness of the problem . let us recall that this feature also happens for stationary first order equations ( see @xcite ) whereas , for second order equations , some transition conditions ( the so - called kirchhoff condition ) need to be imposed ( see @xcite and references therein ) . in the second part of the paper we illustrate our results with a concrete application : the _ blocking _ problem . suppose that a fire breaks up in some region of an oil pipeline . a central controller can stop the propagation of the fire by closing the junctions of the pipes , represented by the vertices of the network . the controller spends some time to reach the junctions which become unavailable when they are reached by the fire front . therefore only a subset of the vertices can be closed on time to get fire under control . the aim is to find a strategy which maximizes the part of the network preserved by the fire . we give a characterization of the optimal strategy and we study the corresponding flame propagation in the network . moreover we describe a numerical scheme for the solution of the problem and we present some numerical examples . * notations : * a network @xmath1 is a finite collection of points @xmath9 and edges @xmath10 in @xmath5 . the vertices of @xmath11 are connected by the continuous , non self - intersecting arcs of @xmath12 . each arc @xmath13 is parametrized by a smooth function @xmath14\to{{\mathbb r}}^n,\ , l_j>0 $ ] . for @xmath15 we set @xmath16 . we set @xmath17 , @xmath18 , and we denote by @xmath19 , the set of boundary vertices of @xmath1 , and by @xmath20 , the set of transition vertices . for simplicity , we assume @xmath21 ( otherwise , one can introduce appropriate boundary condition on @xmath22 ) . the network is not oriented , but the parametrization of the arcs induces an orientation which can be expressed by the _ signed incidence matrix _ @xmath23 with @xmath24 in the following we always identify @xmath25 with @xmath26 $ ] . for any function @xmath27 and each @xmath28 we denote by @xmath29\to { { \mathbb r}}$ ] the restriction of @xmath30 to @xmath13 , i.e. @xmath31 for @xmath32 $ ] . the derivative are always considered with respect to the parametrization of the arc , i.e. if @xmath33 , @xmath34 then @xmath35 . at @xmath36 , we write by @xmath37 to intend the derivative relative to the arc @xmath13 , @xmath38 . we denote by @xmath39 the path distance on @xmath1 , i.e. @xmath40 in this section we consider the hamilton - jacobi equation @xmath41 with the initial condition @xmath42 the hamiltonian @xmath43 satisfies the following assumptions @xmath44 are standard assumptions in the framework of viscosity solution theory . assumption is the independence of @xmath2 of the parametrization of the arc : if we change the parametrization of @xmath13 incident to @xmath45 and we invert its orientation then also the derivative @xmath46 for @xmath47 changes sign . by , the equation is called _ geometric _ and it is connected with front propagation ( see @xcite ) , while implies in particular the coercivity of the hamiltonian , i.e. @xmath48 for any @xmath49 . [ ex ] an hamiltonian satisfying the previous assumptions is given by @xmath50 where @xmath51 is a compact metric space , @xmath52 is a continuous function such that , for some @xmath53 , there holds @xmath54 and @xmath55 . in particular , if @xmath56 and @xmath57 with @xmath58 bounded and strictly positive , then the hamiltonian is @xmath59 ( called eikonal hamiltonian ) and it satisfies assumptions - . in the next definitions we introduce the class of test functions and the notion of viscosity solution for . [ 1:def2 ] let @xmath60 . * let @xmath61 , @xmath28 . we say that @xmath62 is test function at @xmath63 , if @xmath64 is differentiable at @xmath65 . * let @xmath66 , @xmath67 , @xmath68 , @xmath69 , @xmath70 . we say that @xmath62 is @xmath71-test function at @xmath63 , if @xmath64 and @xmath72 are differentiable respectively at @xmath65 and @xmath73 , with @xmath74 where @xmath75 is the signed incidence matrix . [ rmk_def ] by the continuity of @xmath62 and point @xmath76 we infer @xmath77 [ scsol ] * if @xmath61 , @xmath28 , then a function @xmath78 ( resp . , @xmath79 ) is called a subsolution ( resp . supersolution ) of at @xmath63 if for any test function @xmath62 for which @xmath80 attains a local maximum ( resp . , a local minimum ) at @xmath63 , we have @xmath81 * if @xmath66 , @xmath67 , @xmath70 then + @xmath82 : : a function @xmath83 is called a subsolution at @xmath63 if for any @xmath68 and any @xmath71-test function @xmath62 for which @xmath80 attains a local maximum at @xmath63 relatively to @xmath84 , then @xmath85 @xmath82 : : a function @xmath86 is called a supersolution at @xmath63 if for any @xmath87 , there exists @xmath88 , @xmath89 ( said @xmath90-feasible for @xmath91 at @xmath63 ) such that for any @xmath71-test function @xmath62 for which @xmath80 attains a local minimum at @xmath63 relatively to @xmath92 , we have @xmath93 by remark [ rmk_def ] and assumption , at a vertex @xmath94 , it holds @xmath95 in the next proposition we exploit the geometric character of to give equivalent definitions of subsolution and supersolution . [ geometric ] * a function @xmath96 is a subsolution ( respectively , @xmath79 is a supersolution ) of at @xmath61 if and only if for any @xmath97 and for any admissible test function @xmath62 which has a local minimum on @xmath98 ( resp . , a local maximum on @xmath99 ) at @xmath63 , then ( resp . , ) holds . * a function @xmath83 is a subsolution of at @xmath66 , @xmath67 , @xmath70 if and only if for any @xmath97 , for any @xmath68 with @xmath100 , and any @xmath71-test function @xmath62 which has a local minimum on @xmath101 at @xmath63 , then holds . * a function @xmath86is a supersolution of at @xmath66 , @xmath67 , @xmath70 if and only if for any @xmath97 , for any @xmath87 , there exists @xmath88 , @xmath89 such that for any @xmath71-test function @xmath62 for which @xmath102 has a local maximum on @xmath103 at @xmath63 then holds . the proof is postponed to the appendix . we prove a comparison principle for the equation . we shall use some techniques introduced for evolutive equations in the euclidean setting ( see @xcite and references therein ) and for stationary equations on networks ( see ( * ? ? ? * theorem 5.1 ) ) . note that for the next result we only exploit assumptions . let @xmath104)$ ] , @xmath105)$ ] be a subsolution and , respectively , a supersolution of such that @xmath106 for @xmath49 . then @xmath107 in @xmath108 $ ] . we proceed by contradiction , assuming @xmath109}(u - v)>0.\ ] ] for @xmath110 sufficiently small , the function @xmath111 attains a positive maximum at some point @xmath112 and we set @xmath113 for each @xmath114 , @xmath115 , we set @xmath116 where @xmath117 is the distance defined in . by the choice of @xmath118 and @xmath110 , we have @xmath119 whence , the function @xmath120 attains a positive maximum with respect to @xmath121 $ ] at some point @xmath122 . moreover , by the last relation , for every @xmath123 , we have @xmath124 for later use , we denote by @xmath125 a maximum point of @xmath126 in @xmath127 , namely @xmath128 by the inequality @xmath129 and the boundedness of @xmath1 , we get @xmath130 for some constant @xmath58 independent of @xmath110 , @xmath118 and @xmath123 ; in particular , we infer the estimates @xmath131 it follows that , since @xmath132 , @xmath133 for @xmath123 sufficiently small . we deduce that , as @xmath134 ( possibly passing to a subsequence ) , there holds @xmath135 for some @xmath136 . we claim that , in fact , there holds @xmath137 indeed , if the claim is false , by the semicontinuity of @xmath30 and @xmath138 , ( recall that @xmath139 is a maximum point for @xmath126 in @xmath127 ) passing to the limit in the inequality @xmath140 we get a contradiction to the definition of @xmath139 in . whence is proved . we now distinguish two cases : @xmath141 and @xmath142 . assume first that @xmath141 ; in order to contradict relation , observe that @xmath143 passing to the limit as @xmath144 and taking into account relation and @xmath106 on @xmath1 , we infer @xmath145 which is the desired contradiction to . assume now that @xmath146 . for @xmath123 sufficiently small , both @xmath147 and @xmath148 are strictly positive . for @xmath149 ( i.e. , @xmath150 for some @xmath28 ) , one can accomplish the proof by standard arguments so we consider only the case @xmath151 . we assume without any loss of generality that the unique path of length @xmath152 connecting @xmath153 with @xmath154 runs at most through one vertex and , if this happens , the vertex is @xmath45 . we observe that the functions @xmath155 = : u(x , t)-\psi_1(x , t)\\ ( \xi , s ) & \mapsto & v(\xi , s)-\left[-\frac{d({{\widetilde x}},\xi)^2}{2\a}-\frac{\|{{\widetilde t}}-s\|^2}{2\a}\right]=:v(\xi , s)-\psi_2(\xi , s)\end{aligned}\ ] ] attain respectively a maximum at @xmath156 and a minimum at @xmath157 . we now distinguish several cases according to the position of @xmath153 and @xmath154 . _ case 1 : @xmath158 , @xmath159 for some @xmath68 @xmath69 . _ we note that the functions @xmath160 and @xmath161 are admissible test function respectively at @xmath156 and at @xmath162 ; therefore , the definition of sub- and of supersolution entails @xmath163 subtracting the latter inequality from the former , we deduce @xmath164 by , we obtain @xmath165 by the last two inequalities and relation , passing to the limit as @xmath166 , we conclude @xmath167 which is the desired contradiction . _ case 2 : @xmath158 , @xmath168 for some @xmath87 . _ as before , the function @xmath160 is an admissible test function at @xmath156 . we claim that the function @xmath161 is an admissible @xmath71-test function for every @xmath169 ( recall that the point @xmath153 belongs to the edge @xmath13 ) . actually , for @xmath170 , we have @xmath171 hence , condition is satisfied . in particular , @xmath161 is an admissible @xmath71-test function where @xmath172 is the @xmath90-feasible edge for @xmath91 at @xmath157 in the definition of supersolution . the definition of sub- and of super solution yields again inequalities and . since the rest of the proof can be achieved following the arguments of the previous case , we shall omit it . _ case 3 : @xmath173 , @xmath174 for some @xmath87 . _ this case is similar ( and even simpler , because the definition of subsolution is less restrictive than the one of supersolution ) to case @xmath175 ; hence , we shall omit it . _ case 4 : @xmath176 . _ let us just observe that @xmath177 namely , @xmath160 ( respectively , @xmath161 ) is an admissible @xmath71-test function at @xmath156 ( resp . , at @xmath157 ) for every @xmath68 with @xmath178 . by the same arguments as those in case 2 and 3 , we infer inequalities and and we conclude as in case 1 . this section is devoted to establish a regularity result for solution of the equation . [ prp : reg ] let @xmath30 be a solution to - where @xmath179 is lipschitz continuous . then , @xmath30 is lipschitz continuous in @xmath180 $ ] . we shall follow some ideas of ( * ? ? ? * proposition 2.1 ) . let @xmath181 and @xmath182 be the lipschitz constant of @xmath179 and respectively the maximum of @xmath183 in @xmath184 $ ] . we note that the functions @xmath185 are respectively a super- and a subsolution to problem - . the comparison principle ensures @xmath186 hence @xmath30 is bounded . moreover , for every @xmath187 , the functions @xmath188 are respectively a sub- and a supersolution to problem - . invoking again the comparison principle , we infer @xmath189 for every @xmath187 sufficiently small . for each @xmath190 and for @xmath187 small , we deduce @xmath191 which amounts to the lipschitz continuity in the variable @xmath192 . hence , the function @xmath30 verifies ( in viscosity sense ) @xmath193 by the coercivity of @xmath2 and the previous estimate , we get the lipschitz continuity of @xmath30 in the variable @xmath194 ( see ( * ? ? ? ii.4.1 ) ) . in this section we exploit the geometric character of the equation to give a representation formula for the solution of - . given the hamiltonian @xmath2 , we define the support function of the sub - level set @xmath195 by @xmath196 the function @xmath197 is continuous , convex , positive homogeneous and non negative ( see @xcite ) . for example , for the eikonal hamiltonian @xmath59 ( introduced in example [ ex ] ) , there holds @xmath198 . a path @xmath199\to\g$ ] is said _ admissible _ if there are @xmath200 such that , for any @xmath201 , @xmath202)\subset e_{j_m}$ ] for some @xmath203 and @xmath204 . we denote by @xmath205 the set of the admissible path such that @xmath206 , @xmath207 . we introduce a distance function related to the hamiltonian @xmath2 on the network . for @xmath208 define @xmath209 note that the distance defined by coincides with the one defined by for @xmath210 since in this case @xmath211 . the next proposition summarizes some properties of @xmath7 ( for the definition of viscosity solution on a network in the stationary case , we refer the reader to the paper @xcite ) . [ propdist ] @xmath7 is a lipschitz continuous function on @xmath212 and it is equivalent to the path distance @xmath117 , i.e. there exists @xmath213 such that @xmath214 moreover , for any @xmath215 , closed , @xmath216 is a subsolution in @xmath1 and a supersolution in @xmath217 of the hamilton - jacobi equation @xmath218 for the proof of the previous proposition we refer to ( * ? ? ? [ existencec ] let @xmath219 be a continuous function . then the solution of - is given by @xmath220 in order to prove this result , let us first establish a preliminary lemma . [ caratstaz ] if @xmath221 is a subsolution ( resp . , supersolution ) of in @xmath1 , then @xmath222 is a subsolution ( resp . , supersolution ) of in @xmath223 . we only show that @xmath30 is a subsolution at @xmath224 in the case @xmath151 , being the case @xmath225 similar . assume by contradiction that there exists @xmath226 , @xmath68 and a @xmath71-admissible test function @xmath62 at @xmath224 such that @xmath80 has a local maximum at @xmath224 relatively to @xmath227 and @xmath228 by @xmath229 for @xmath230 , taking into account the definition of @xmath30 , we get @xmath231 @xmath232 ; by the arbitrariness of @xmath192 , we infer @xmath233 moreover , by with @xmath234 , we get that @xmath235 has a local maximum at @xmath236 relatively to @xmath237 , hence @xmath238 by and we get a contradiction to . the proof that @xmath30 is a supersolution in @xmath239 is similar . we claim that @xmath30 verifies the supersolution condition choosing as @xmath90-feasible edge for @xmath87 the same edge @xmath240 which is @xmath90-feasible for @xmath91 for the function @xmath221 at @xmath45 . the calculations follows the same arguments as before so we shall omit them . proof of theorem [ existencec ] + _ i ) @xmath30 is continuous : _ given @xmath241 $ ] , let @xmath242 $ ] be such that @xmath243 and set @xmath244 where @xmath245 is as in . we claim that @xmath246 indeed , if @xmath247 , then @xmath248 and therefore . moreover we claim @xmath249 indeed , if @xmath250 , then @xmath251 where the latter inequality is due to the subadditivity of @xmath7 ; hence , relation follows . therefore , by and , we deduce @xmath252 where @xmath253 is the modulus of continuity for @xmath179 in a neighborhood of @xmath236 . this gives @xmath254 by @xmath255 we get in a similar way @xmath256 _ ii ) @xmath30 is a supersolution : _ we only prove that @xmath30 is a supersolution at @xmath224 with @xmath257 . assume by contradiction that , for some @xmath87 and for any @xmath169 , there is a @xmath71-admissible test function @xmath72 such that @xmath258 has a local minimum at @xmath224 relatively to @xmath237 with @xmath259 and such that @xmath260 observe that @xmath261 assume first that @xmath262 and define @xmath263 . we claim that @xmath72 has a local maximum on the set @xmath264 at @xmath224 . in fact if @xmath265 then by , @xmath266 and since @xmath267 we get @xmath268 and the claim is proved . by lemma [ caratstaz ] @xmath221 is supersolution to and therefore prop.[geometric ] gives a contradiction to . if @xmath269 , we claim that @xmath224 is a local maximum point for @xmath30 . in fact , @xmath270 for some @xmath226 . if @xmath63 is such that @xmath271 with @xmath272 , then @xmath273 hence @xmath274 and the claim is proved . by @xmath224 is also a local maximum point for @xmath72 on @xmath227 for any @xmath89 . by and , we get @xmath275 and therefore a contradiction to @xmath224 being a local maximum point for @xmath72 . + _ iii ) @xmath30 is a subsolution : _ we only consider the case @xmath224 with @xmath257 . we first observe that @xmath276 given @xmath91 , @xmath88 let @xmath62 be an @xmath71-test function such that @xmath80 has a local maximum at @xmath224 relatively to @xmath277 . arguing as in the supersolution case we define @xmath263 and we show that @xmath62 has a local minimum on the set @xmath278 . then we conclude by applying prop.[geometric ] and lemma [ caratstaz ] . in this section we provide a concrete application of our results : now , the network @xmath1 represents an oil pipeline ( a network of computer , the circulatory system , etc . ) and at initial time a fire breaks up in the region @xmath279 ( a virus is detected in a subnet , an embolus occurs in some vessel ) . the speed of propagation of the fire is known but it may depend on the state variable . our aim is to determine an optimal strategy to stop the fire and to minimize the burnt region . as in the flame propagation model described in @xcite , let @xmath280 be the initial burnt region and @xmath281 the region burnt at time @xmath192 . assume that the front @xmath282 propagates in the outward normal direction to the front itself . then @xmath281 is given by the @xmath283-sublevel set of a viscosity solution of - where the initial datum @xmath179 satisfies @xmath284 . recalling the representation formula we observe that the 0-sublevel set of the solution of - is given by @xmath285 where @xmath7 is defined as . note that , since @xmath1 is composed by a finite number of bounded edges and therefore its total length is finite , then the burnt region @xmath286 coincides with @xmath1 . in other words , without any external intervention , the pipeline will be completely burnt in a finite time . we assume that an operator , located at @xmath287 ( the `` operation center '' ) , can block the fire by closing the junctions of the pipeline ( i.e. , vertices of the network ) and that this operation is effective only after a delay which depends on the distance of the junction from @xmath236 . our problem is reminiscent of other models described in literature ( for instance , see @xcite and references therein ) which concern the control of some diffusion in a network ( e.g. minimizing the spread of a virus or maximizing the spread of an information ) . in this framework , let us stress the main novelties of our setting : in our model , the diffusion has positive finite speed and it affects both vertices and edges , the spread is not reversible ( namely , `` infected '' points can not become again `` healthy '' ) and the effect of the operator s action has finite speed ( in other words , it is effective after a delay depending on the distance from the operation center ) . an admissible strategy @xmath288 is a subset of @xmath11 such that @xmath289 where @xmath118 is a given nonnegative constant . we denote by @xmath290 the vertices of the network which satisfy the admissibility condition and by @xmath291 the set of the admissible strategies . condition means that the time to reach the vertex @xmath292 from @xmath236 at the velocity @xmath293 is less than or equal to the time the fire front reaches @xmath45 . therefore the junction @xmath45 can be blocked before the front goes through it . given a strategy @xmath294 , we denote @xmath295 $ ] the distance restricted to the trajectories not going through a vertex in @xmath288 , i.e. @xmath296$}\right\}\ ] ] with @xmath297 if there is no admissible curve joining @xmath298 to @xmath194 . we also set @xmath299 which are respectively the region burnt at time @xmath192 and the total burnt region using the strategy @xmath288 . observe that * if @xmath118 is very small , then the optimal strategy is given by the extremes of the edges containing @xmath280 ; * if @xmath118 is very large and @xmath300 , then every strategy is useless since the whole pipeline will burn whatever the operator does . aside the previous simple cases an optimal strategy for the blocking problem may be not obvious and we aim to find an efficient way to compute it . to find a strategy which minimizes the burnt region , we first give a characterization of @xmath301 in terms of a problem satisfied by the distance @xmath302 . given @xmath294 , set @xmath303 and @xmath304 . then * @xmath305 and @xmath306 in @xmath307 . moreover if @xmath308 and @xmath87 is such that @xmath309 , then @xmath310 ; * @xmath30 is a viscosity solution of the problem @xmath311 * let @xmath312 be such that , defined @xmath313 , @xmath314 then @xmath315 and @xmath316 in @xmath1 . note that , for @xmath28 , either @xmath309 or @xmath317 , i.e. an edge is either completely burnt or it can not be reached by the fire . the function @xmath30 can be discontinuous at @xmath308 and * if @xmath318 , then either @xmath319 for all @xmath87 if @xmath320 or @xmath321 for all @xmath87 if @xmath322 ; * if @xmath323 , then either @xmath319 for all @xmath87 if @xmath320 or there exists @xmath87 such that @xmath321 if @xmath322 and in this case @xmath324 . actually , if @xmath308 and @xmath321 , an admissible trajectory for @xmath325 connecting @xmath45 to @xmath280 and containing the edge @xmath13 , @xmath87 , necessarily enters from @xmath45 into @xmath13 . hence @xmath326 is increasing for @xmath33 , @xmath327 and @xmath328 . in @xmath329 , @xmath325 locally behaves as the distance @xmath7 defined in . therefore the continuity of @xmath30 in @xmath330 and the sub and supersolution properties in the open set @xmath331 are obtained with the same arguments ( * ? ? ? * prop.4.1 ) . to prove _ iii ) _ , assume by contradiction that there exists @xmath332 such that @xmath333 . for any @xmath334 such that @xmath335 , a minimizing trajectory always exists since ( up to reparametrization ) there is only a finite number of trajectories connecting the two points . hence let @xmath336 be an admissible curve for @xmath325 such that @xmath337 , @xmath338 and @xmath339 . let @xmath340 such that , for any @xmath201 , @xmath202)\subset e_{j_m}$ ] for some @xmath203 , @xmath341 and @xmath204 . clearly @xmath342 for @xmath343 $ ] . if @xmath221 is a subsolution to , then by the coercivity of @xmath2 , @xmath221 is lipschitz continuous in @xmath344 and therefore @xmath345 a.e . on @xmath344 . moreover , by the definition of the support function @xmath346 , we have @xmath347 if and only if @xmath348 . hence @xmath349 $ , $ m=0,\dots , m$.}\ ] ] ( the term @xmath350 takes into account the orientation of the arc @xmath351 ) . integrating the previous relation in @xmath352 $ ] and recalling that @xmath353 , we get @xmath354 iterating the same argument in @xmath355 $ ] we finally get @xmath356 and therefore a contradiction since @xmath338 . we conclude that @xmath357 and @xmath358 . we now show that the strategy composed by all the admissible nodes which are adjacent to a non admissible node is optimal , in the sense that it maximizes the preserved region . [ optbl ] the admissible strategy @xmath359 satisfies : for any @xmath360 , @xmath361 . assume by contradiction that there exist @xmath362 and @xmath363 . hence there exists an admissible trajectory @xmath336 for @xmath364 connecting @xmath236 to @xmath280 , i.e. there exists @xmath365 , @xmath366 and @xmath367 such that @xmath368 for @xmath369 $ ] . note that @xmath364 disconnect the subgraph containing the admissible vertices @xmath290 by the one containing the non admissible vertices @xmath370 and therefore @xmath371)$ ] is contained in the subgraph with vertices @xmath372 . since @xmath373 , then @xmath336 is also admissible for @xmath374 and therefore a contradiction to @xmath375 . one could be interested in a cost functional on @xmath376 which takes into account other terms , as the cost of blocking a junction , beside the length of the burnt region . for instance , consider the cost @xmath377 where @xmath378 represents the cost of blocking the vertex @xmath45 while @xmath379 represents the damage of the destruction of edge @xmath13 by fire . clearly , the minimum of @xmath380 exists since @xmath291 is finite , but it seems more difficult to characterize the optimal strategy because it will strongly depend on the geometry of the network . in this section we propose a numerical method to compute the optimal strategy for the blocking problem . the scheme is based on a finite difference approximation of the stationary problem ; for simplicity , we only consider the case of the eikonal hamiltonian @xmath381 . on each interval @xmath382 $ ] parametrizing the arc @xmath13 , we consider an uniform partition @xmath383 with @xmath384 and @xmath385 . in this way we obtain a grid @xmath386 on the network @xmath387 . we define @xmath388 , the set of the nodes in the initial front . for @xmath389 , we say that @xmath390 and @xmath391 are adjacent and we write @xmath392 if and only if they are the image of two adjacent grid points , i.e. @xmath393 , for @xmath394 $ ] , @xmath395 , @xmath28 and @xmath396 . note that if @xmath397 is a vertex of @xmath1 , then the nodes of the grid @xmath398 adjacent to @xmath45 may belong to different arcs . + we compute the optimal strategy by means of the following algorithm , based on the results of prop . [ optbl ] . * blocking strategy [ b ] * 1 . in the first step we solve the front propagation problem on the network computing the approximated time @xmath399 at which a node @xmath400 gets burnt @xmath401 u^h(x_{j , m})=0 & x_{j , m}\in \mathcal{r}_0^h\\[6pt ] \end{array } \right.\ ] ] note that if @xmath402 coincides with a vertex @xmath397 , the approximating equation reads as @xmath403 the discrete function @xmath404 is such that @xmath405 , where @xmath406 . 2 . in the second step we determine the vertices which satisfy the admissibility condition . we define @xmath407 , where @xmath408 represents the approximated time to reach a node @xmath409 , starting from the operation center @xmath236 and moving with a constant speed @xmath410 . the function @xmath411 is computed by means of the finite difference scheme @xmath412 w^h(x_0)=0 . & \\[6pt ] \end{array } \right.\ ] ] 3 . we define the approximated optimal strategy by setting @xmath413 and , for @xmath414 , we compute the corresponding approximate distance by solving the following finite difference scheme @xmath415 u^h_\s(x_{j , m})=0 & x_{j , m}\in \mathcal{r}_0^h\\[6pt ] -\frac{1}{h_j } ( u^h_{j,\s}(x)-u^h_{j,\s } ( x_{j , m}))-c(x_{j , m})=0 & \text{$x_{j , m } \in \sigma$ , $ x\in \mathcal{g}^h$ , $ x\sim x_{j , m}$ } \end{array } \right.\ ] ] the discrete function @xmath404 is such that @xmath416 , where @xmath303 solves @xmath417 . note that as in the continuous case , the value of @xmath418 at @xmath292 can depend on the edge @xmath13 and in general the function is discontinuous at these points . in this paper we do not analyze the properties of the previous finite difference schemes . in any case , at least for and , the well - posedness and the convergence of the schemes can be studied by adapting the techniques in @xcite . we consider a network with a simple structure where the fire starts in one vertex , @xmath419 and propagates with speed @xmath420 . we first perform step @xmath421 of algorithm * [ b ] * and we compute the approximated time @xmath399 at which a node @xmath422 get burnt . the results are shown in fig.[test1grafo ] together with the graph structure . + next , we perform step @xmath423 of algorithm * [ b ] * . we suppose the operation center @xmath236 is located on the vertex @xmath424 and the velocity to reach a node @xmath45 from @xmath236 is @xmath425 . using , we compute the set of nodes @xmath426 . the result is shown in in fig.[test1vad ] , the set of nodes in @xmath426 are represented by the square markers . once computed the set of admissible nodes @xmath426 , we can compute the optimal strategy , following step @xmath427 . the result is shown in fig.[test1blocking ] . it is clear , from the simple structure of thenetwork , that any other choice of @xmath428 would lead to a greater burnt region and consequently to a smaller preserved network region . we consider a more complex network , with 20 vertices and 32 arcs . we suppose the fire starts in two vertices and propagates with a non constant normal speed @xmath429 . + we proceed as in the first example and we compute the approximated time @xmath399 at which a node @xmath422 get burnt . the results are shown in fig.[test2grafo ] together with the graph structure . + we suppose the operation center @xmath236 is located on the vertex @xmath430 and the velocity to reach a node @xmath45 from @xmath236 is @xmath431 . using , we compute the set of nodes @xmath426 . the result is shown in in fig . [ test2vad ] , the set of nodes in @xmath426 are represented by the square markers . once computed the set of admissible nodes @xmath426 , we can compute the optimal strategy , following step @xmath427 . the result is shown in fig.[test2blocking ] . in this case we get the optimal solution blocking only three vertices . by changing the set @xmath280 as in fig.[test3grafo ] , the region of the admissible node @xmath432 , shown in fig . [ test3vad ] , turns out to be much smaller . in this case the optimal strategy is formed by all the vertices in @xmath432 , and the preserved region becomes smaller than the previous case , see fig . [ test3blocking ] . this section is devoted to the proof of proposition [ geometric ] . to this end , it is expedient to establish the following result . [ cgg5.2 ] let @xmath96 ( respectively , @xmath433 ) be a subsolution ( resp . , supersolution ) to . for any function @xmath434 continuous and nondecreasing , the function @xmath435 is still a subsolution ( resp . , supersolution ) to . + in particular , if @xmath30 solves , then also @xmath435 solves the same equation . for a point inside an edge , the proof follows by the same arguments as in the proof of ( * ? ? ? * theorem 5.2 ) . for the subsolution case in a vertex @xmath45 , we observe that an usc function is a subsolution if , and only if , it is a standard viscosity subsolution on the arc @xmath237 for any @xmath68 . taking into account this observation and using again the arguments of ( * ? ? ? * theorem 5.2 ) we obtain the statement . finaly , the supersolution case at a vertex @xmath45 follows by similar arguments . @xmath437 consider a function @xmath83 . for @xmath36 , assume that , for each @xmath97 , @xmath68 with @xmath178 , @xmath71-test function @xmath62 which has a local minimum on @xmath101 at @xmath63 , inequality holds . our aim is to prove that @xmath30 satisfies the subsolution condition at @xmath63 . to this end , we assume by contradiction that there exists an admissible test function @xmath62 which verifies @xmath438 for @xmath439 , we set @xmath440 . inequality and the definition of @xmath441 yield @xmath442 namely , the function @xmath62 attains a local minimum in @xmath63 with respect to @xmath441 . invoking our assumption , we infer inequality which amounts to the desired contradiction . hence , the first implication of the statement is achieved . we now prove the reverse implication . let @xmath30 be a subsolution to . fix @xmath97 , @xmath443 and @xmath68 with @xmath178 . let @xmath62 be an admissible @xmath71-test function which attains a local minimum on @xmath444 at @xmath63 . let @xmath447 be a compact neighborhood of @xmath63 in the topology of @xmath277 such that @xmath448 in @xmath449 . we set @xmath450 ( observe that the definition of @xmath451 ensures : @xmath452 and @xmath453 ) . we define @xmath463\\ ( \a-\b)^{-1}\left[(\a - s)\inf_w \phi + ( s-\b)\phi(x , t)\right]&\qquad \textrm{for } s\in(\b,\a ) . \end{array}\right.\ ] ] now , we want to prove @xmath464 for every @xmath465 in some neighborhood of @xmath63 . to this end , we shall consider separately the cases when @xmath466 belongs to @xmath467 , @xmath468 , @xmath449 . for @xmath469 , by the monotonicity of @xmath445 and the definition of @xmath441 and @xmath468 , we have @xmath470 which amounts to inequality . for @xmath471 , taking into account the definition of @xmath472 , we have @xmath473 for @xmath474 , by the definition of @xmath441 ( recall that @xmath62 attains a local minimum on @xmath475 ) , there holds @xmath476 hence our claim is completely proved . + case-@xmath459 . for any @xmath477 , we introduce @xmath478 we observe that the sets @xmath479 are not empty for @xmath480 sufficiently large with @xmath481 and @xmath482 . moreover , the sequence @xmath483 is nondecreasing and , by the compactness of @xmath479 , it fulfills : @xmath484 and @xmath485 as @xmath486 . hence , there exists an increasing sequence @xmath487 such that @xmath488 is increasing and @xmath489 . we set @xmath490\\ \textrm{linear function}&\qquad\textrm{for } s\in(\b_{n_{m-1}}-\b_{n_m } ) . \end{array}\right.\ ] ] we want to prove studying separately the cases when @xmath466 belongs to @xmath467 , @xmath491 , @xmath492 ( @xmath493 ) and @xmath449 . in the first and last cases , inequality follows by the same arguments of point @xmath456 . for @xmath494 there holds @xmath495 and , by the monotonicity of @xmath445 , @xmath496 for @xmath497 , there holds @xmath498 and @xmath499 ; therefore , we infer @xmath500 whence , inequality is established and statement ( @xmath501 ) is completely proved . @xmath502 the proof of this part of the statement follows by the same arguments as those in the previous one reversing each inequality and choosing the test functions only along the @xmath90-feasible edge .	we consider the cauchy problem @xmath0 where @xmath1 is a network and @xmath2 is a convex and positive homogeneous hamiltonian . we introduce a definition of viscosity solution and we prove that the unique viscosity solution of the problem is given by a hopf - lax type formula . in the second part of the paper we study flame propagation in a network and we seek an optimal strategy to block a fire breaking up in some part of a pipeline . * msc 2000 * : : : 35r02 , 35f20 , 49l25 , 35f21 , 65m06 . * keywords * : : : evolutive hamilton - jacobi equation , viscosity solution , network , hopf - lax formula , finite difference approximation .
many continuous phase transitions observed in nature belong to the o(@xmath0 ) vector universality classes , which are characterized by an @xmath0-component order parameter with o(@xmath0 ) symmetry and the symmetry breaking o(@xmath0)@xmath2o(@xmath3 ) . the superfluid transition in @xmath4he , the formation of bose - einstein condensates , density wave systems , transitions in magnets with easy - plane anisotropy , and in superconductors belong to the xy or o(2 ) universality class ; the curie transition in isotropic magnets , zero - temperature quantum transitions in two - dimensional antiferromagnets , are examples for the heisenberg or o(3 ) universality class ; the o(4 ) universality class is relevant for the finite - temperature transition in two - flavor quantum chromodynamics , the theory of strong interactions . see , e.g. , refs . @xcite for reviews . in the absence of external fields , the phase transition of o(@xmath0)-symmetric vector models is driven by only one relevant parameter , which is usually associated with the temperature . the corresponding rg dimension is @xmath5 where @xmath6 is the correlation - length exponent . the leading odd perturbation , which breaks the o(@xmath0 ) symmetry , is associated with the external field @xmath7 coupled to the order parameter ; it has rg dimension @xmath8 , where @xmath9 is the exponent controlling the power - law space - dependence of the two - point correlation function of the order parameter at criticality . the asymptotic critical power - law behaviors of o(@xmath0)-symmetric vector models have been determined with high accuracy . in table [ tabcexp ] we report some of the most accurate estimates of the critical exponents @xmath6 and @xmath9 , and of the leading and next - to - leading scaling - correction exponents @xmath10 and @xmath11 , which characterize the dominant corrections to the universal scaling . 3 & lt & 0.7112(5 ) @xcite & 0.0375(5 ) @xcite & 0.773 @xcite & + & & 0.7117(5 ) @xcite & 0.0378(5 ) @xcite & & + & & 0.7116(10 ) & 0.0378(3 ) & & + & ft & 0.7073(35 ) @xcite & 0.0355(25 ) @xcite & 0.782(13 ) @xcite & 1.78(11 ) @xcite + 4 & lt & 0.749(2 ) @xcite & 0.0365(10 ) @xcite & 0.765 @xcite & + & & 0.7477(8 ) @xcite & 0.0360(4 ) @xcite & & + & & 0.750(2 ) & 0.0360(3 ) & & + & ft & 0.741(6 ) @xcite & 0.0350(45 ) @xcite & 0.774(20 ) @xcite & + in this paper we study the effects of anisotropic perturbations breaking the o(@xmath0 ) symmetry , which can not be related to an external vector field coupled to the order parameter , but which are represented by composite operators with more complex transformation properties under the o(@xmath0 ) group . an interesting question is whether they change the critical behavior , or whether they do not affect it so that the symmetry shown by the critical correlations is larger than that of the microscopic model . this issue arises in several physical contexts . anisotropy in magnetic systems may naturally arise due to the cubic structure of the underlying lattice , giving rise to anisotropic interactions terms , see , e.g. , ref . the relevance of the anisotropic perturbations determines also the nature of the multicritical behavior at the meeting point of two transition lines with different o(@xmath12 ) and o(@xmath13 ) symmetries , in particular , whether the symmetry gets effectively enlarged to @xmath14 , see , e.g. , refs . another interesting issue is the critical behavior of secondary order parameters , which are generally represented by powers of the order parameter transforming as higher representations of the o(@xmath0 ) group ; their critical behaviors can be measured in density wave systems , such as liquid crystals @xcite , see also refs . @xcite . let us consider the general problem of the o(@xmath0)-symmetric theory in the presence of an external field @xmath15 coupled to a perturbation @xmath16 . assuming @xmath16 to be an eigenoperator of the rg transformations , the singular part of the free energy for the reduced temperature @xmath17 and @xmath18 can be written as @xmath19 where @xmath20 is the rg dimension of @xmath15 , and @xmath21 is a scaling function . therefore , the rg dimensions of the anisotropic external fields quantitatively control their capability to influence or change the asymptotic critical behavior when @xmath22 . in the field - theoretical ( ft ) framework the o(@xmath0)-symmetric vector model is represented by the o(@xmath0)-symmetric landau - ginzburg - wilson theory @xmath23 , \label{phi4h}\ ] ] where @xmath24 is an @xmath0-component real field and @xmath7 an external field . the anisotropic perturbations are conveniently classified @xcite using irreducible representations of the o(@xmath0 ) internal group , characterized by the spin value @xmath25 . let us consider the perturbation @xmath26 defined by the power @xmath27 of the order parameter and the spin representation @xmath25 of the o(@xmath0 ) group @xmath28 where @xmath29 is a homogeneous polynomial of degree @xmath25 that is symmetric and traceless in the @xmath25 indices : @xmath30 etc ... the classification in terms of spin values is particularly convenient : ( i ) under the rg flow the operators with different spin never mix ; ( ii ) all parameters @xmath31 associated with the components of @xmath32 have the same rg dimension @xmath33 . on the other hand , operators with different @xmath27 but with the same @xmath25 mix under renormalization . the spin-0 operators are already present in the @xmath34 hamiltonian ( [ phi4h ] ) : the rg dimension of @xmath35 is related to the correlation length exponent , @xmath36 , while the rg dimension of @xmath37 ( after an appropriate subtraction to cancel the mixing with @xmath35 ) gives the leading scaling correction exponent , indeed @xmath38 . the spin-1 perturbation is related to the external field coupled to the order parameter , thus @xmath39 . is redundant @xcite , because a hamiltonian term containing @xmath40 can be always eliminated by a redefinition of the field @xmath41 . anyway , using the equation of motion , one obtains @xmath42 . ] close to four dimensions , thus for small @xmath43 , @xmath44 for @xmath45 , which implies that the only relevant operators have @xmath46 . it is reasonable to assume that this property holds up to @xmath47 . moreover , near four dimensions we can use standard power counting to verify that the perturbation with indices @xmath48 mixes with @xmath49 , @xmath50 , but their rg dimensions are significantly smaller . in principle , one should also consider terms with derivatives of the field , but again one can show that they are all irrelevant or redundant . the above arguments show that the most interesting anisotropic perturbations are represented by the spin-2 , spin-3 and spin-4 operators @xmath51 because they provide the leading effects of anisotropy for each spin sector . as we shall see , the leading rg dimensions within each spin sector , @xmath52 characterize interesting critical behaviors in various physical contexts . some @xmath53 have been already estimated by using ft approaches based on high - order perturbative calculations , and lattice techniques , such as high - temperature ( ht ) expansions and monte carlo ( mc ) simulations . in table [ tabrgdimp ] we report some results for @xmath54 . in most cases these results provide already a clear indication of the relevance of the perturbation , with the only exception of the spin-4 perturbation in the o(3 ) universality class , where the value of @xmath55 is close to zero . while high - order ft results indicate the relevance of the spin-4 perturbation , the mc estimate of @xmath55 appears compatible with zero . since the issue concerning its relevance is of experimental interest , an accurate determination of @xmath56 is called for to conclusively settle it . .estimates of the rg dimensions @xmath57 of the couplings @xmath58 associated with the leading anisotropic perturbations @xmath59 for the three - dimensional o(@xmath0 ) vector universality classes with @xmath54 . we report results obtained by various methods , such as ft perturbative expansions within @xmath47 and @xmath60-expansion schemes , and lattice techniques , such as high - temperature expansions ( ht ) and finite - size scaling analyses of monte carlo simulations ( fss mc ) . notice that in the mc estimates of @xmath55 reported in ref . @xcite only statistical errors are explicitly given ; the authors write that systematic errors are likely of a similar size . [ cols="^ , < , < , < , < " , ] fixing @xmath63 we obtain @xmath64 ( with @xmath65dof=9.77/12 ) as amplitude for the differences @xmath66 and @xmath67 with @xmath68 taking data only for @xmath69 we get @xmath70 @xmath71 , @xmath72 @xmath73 . it is quite clear from these numbers that a linearization of the correction amplitude as a function of @xmath74 is not sufficient to compute the estimate of @xmath75 . for the same reason , we give an asymmetric estimate of the error : @xmath76 this value is larger than @xmath77 that we quote in ref . however we are quite confident that indeed a @xmath75 exists for the o(4 ) case . note that in the limit @xmath78 for the simple cubic lattice and the given lattice action , no @xmath75 exists and that leading corrections are minimal in the limit @xmath79 @xcite . in order to obtain the critical exponent @xmath9 , we analyze the behavior of @xmath80 where , in the o(3 ) case @xmath81 is defined by @xmath82 . in the first step of the analysis we eliminate leading corrections to scaling . to this end we analyze the ratios @xmath83 we obtain a good fit starting from @xmath84 . for @xmath84 we obtain @xmath85 , @xmath86 and @xmath65dof@xmath87 . therefore in order to eliminate corrections at @xmath88 we follow the strategy discussed in section [ rescorr ] . using @xmath89 and @xmath90 ) eq . ( [ tildecorrect ] ) reads @xmath91 we fit @xmath92 with the ansaetze @xmath93 @xmath94 @xmath95 with @xmath96 or @xmath97 . in the case of the ansatz ( [ chi0 ] ) we obtain very large @xmath65dof up to @xmath98 . for @xmath99 we get @xmath100 and @xmath65dof@xmath101 . using the ansatz ( [ chi1 ] ) we get @xmath65dof@xmath102 already for @xmath103 ; for example , for @xmath104 we obtain @xmath105 and @xmath65dof@xmath106 . using the ansatz ( [ chi2 ] ) with @xmath96 we get for @xmath107 the results @xmath108 and @xmath65dof@xmath109 . and for @xmath97 and @xmath110 we get @xmath111 and @xmath65dof@xmath112 . we redo these fits for @xmath113 without correction to check the effect of the uncertainty of @xmath75 . we find that the estimates of @xmath9 change by about @xmath114 . taking into account only fits with ansaetze that include the analytic background , we arrive at @xmath115 we estimate the exponent @xmath6 from the behavior of the slope of @xmath117 and @xmath118 at @xmath119 : @xmath120 since we did not plan to compute the exponent @xmath6 from the beginning , we did not compute the second derivatives of @xmath117 and @xmath118 with respect to @xmath121 . hence we can not compute the slope at fixed values of @xmath117 or @xmath118 . at @xmath88 we performed mc simulation very close to our final value of @xmath119 . therefore it is sufficient to have a rather rough estimate of the second derivatives of @xmath117 and @xmath118 in order to compute the first derivatives of @xmath117 and @xmath118 at @xmath119 starting from the first derivatives of @xmath117 and @xmath118 at @xmath122 that we have computed in our simulations . to this end , we simulated for @xmath123 at a number of different @xmath121 values . using these data we compute the second derivatives of @xmath117 and @xmath118 with respect to @xmath121 by finite differences . the second derivatives are then estimated by @xmath124 . notice that our estimate of @xmath125 is very close to the simulation point @xmath126 . we analyze the resulting data by fitting with various ansaetze that are derived from eq . ( [ sbehavior ] ) . we arrive at @xmath127 from the analysis of the slope of of @xmath118 and @xmath128 from that of @xmath117 . the error bars take also into account the uncertainty of @xmath75 . as our final estimate we quote @xmath129 m. saito and m. matsumoto , `` simd - oriented fast mersenne twister : a 128-bit pseudorandom number generator '' , in _ monte carlo and quasi - monte carlo methods 2006 _ , edited by a. keller , s. heinrich , h. niederreiter , ( springer , 2008 ) ; m. saito , masters thesis , math . , graduate school of schience , hiroshima university , 2007 . the source code of the program is provided at `` http://www.math.sci.hiroshima-u.ac.jp/@xmath132m-mat/mt/sfmt/index.html ''	we investigate the effects of anisotropic perturbations in three - dimensional o(@xmath0)-symmetric vector models . in order to assess their relevance for the critical behavior , we determine the renormalization - group dimensions of the anisotropic perturbations associated with the first few spin values of the representations of the o(@xmath0 ) group , because the lowest spin values give rise to the most important effects . in particular , we determine them up to spin 4 for @xmath1 , by finite - size analyses of monte carlo simulations of lattice o(@xmath0 ) models , achieving a significant improvement of their accuracy . these results are relevant for several physical systems , such as density - wave systems , magnets with cubic symmetry , and multicritical phenomena arising from the competition of different order parameters .
apart from confirming , yet again , the long - standing solar neutrino problem @xcite , the recent results from the sudbury neutrino observatory ( sno ) on neutral current ( nc ) events @xcite have given strong evidence that solar neutrinos convert mainly to an active neutrino flavor . in addition , valuable spectral and day / night information has been provided @xcite . this adds to the already robust evidence that an extension of the standard model of particle physics is necessary in the lepton sector . although certainly not yet unique , at least for the case of solar neutrinos , which can be accounted well by spin - flavor precession @xcite or non - standard neutrino matter interactions @xcite , the most popular joint explanation of solar and atmospheric experiments is provided by the neutrino oscillations hypothesis , with neutrino mass - squared differences of the order of @xmath9 and @xmath10 , respectively . in the wake of the recent sno nc results we have re - analyzed the global status of current neutrino oscillation data including these and the remaining solar data @xcite as well as the current atmospheric @xcite samples , including the 1489 days super - kamiokande data @xcite and the most recent macro data @xcite . motivated by the stringent limits from reactor experiments @xcite we adopt an effective two - neutrino approach in which solar and atmospheric analyses decouple . however our effective two - neutrino approach is generalized in the sense that it takes into account that a light sterile neutrino @xcite , advocated to account for the lsnd anomaly @xcite , may take part in both solar and atmospheric conversions . the natural setting for such a light sterile neutrino is provided by four - neutrino models . in this paper we will determine the constraints on oscillation parameters in this generalized scenario following from solar and atmospheric data separately . such separate analyses are necessary ingredients towards a combined analysis of all current oscillation data , including solar , atmospheric , negative short - baseline data and the lsnd experiment @xcite . as shown in ref . @xcite such separate analyses can be performed independently of the details of the four - neutrino mass scheme . since the release of the latest sno data in april 2002 a number of global solar neutrino analyses in terms of active oscillations appeared @xcite . moreover , it has been shown by model - independent comparisons of the sno cc rate with the sno nc and super - k rates that transitions of solar neutrinos into sterile neutrinos are strongly constrained by the recent data ( see , _ e.g. _ , refs . however , so - far no dedicated global analyses exist , where a participation of a sterile neutrino in the oscillations is fully taken into account . here we present a complete solar neutrino analysis including sterile neutrinos , determining the allowed ranges for the oscillation parameters @xmath11 and @xmath12 , as well as for the parameter @xmath13 describing the active - sterile admixture . furthermore , we investigate in detail the impact of the sno neutral current , spectral and day / night data and compare with an analysis where we use only the charged current rate from sno . concerning the atmospheric data , we perform an update of previous analyses @xcite , adopting again the most general parameterization of atmospheric neutrino oscillations in the presence of sterile neutrino mixing , characterized by four parameters . we find that the recent 1489-day super - kamiokande data combined with the latest macro data lead to considerably stronger rejection against a sterile neutrino contribution to the oscillations than the previous 1289-day data sample . the plan of the paper is as follows . in sec . [ sec : sol - oscill ] we set the general parametrization for solar oscillations in the presence of active - sterile mixing . in sec . [ sec : sol - analysis ] we briefly describe the solar neutrino data and their analysis . in sec . [ sec : sol - results ] we present the results of our analysis , aimed at studying the impact of recent solar data in the determination of the solar neutrino oscillation parameters , assuming , as mentioned , that the @xmath0 may convert to a mixture of active and sterile neutrinos . we give the regions of oscillation parameters for different allowed @xmath14 values , display the global behavior of @xmath15 and @xmath16 , calculated with respect to the favored active lma solution , and evaluate the impact of the sno nc , spectral and day / night data . present solar data exhibit a higher degree of rejection against non - lma and/or non - active oscillation solutions , which we quantify , giving also the absolute goodness of fit ( gof ) of various oscillation solutions . our solar neutrino results are briefly compared with those obtained in other recent analyses in sec . [ sec : sol - compare ] . in sec . [ sec : atm - oscill ] we set our notations for atmospheric oscillations in the presence of active - sterile admixture . in sec . [ sec : atm - analysis ] we briefly describe the atmospheric neutrino data and their analysis . in sec . [ sec : atm - results ] we describe our results for atmospheric oscillation parameters in an improved global fit of current atmospheric neutrino data . we quantify the impact both of our improved analysis as well as that of the recent data in rejecting against the sterile oscillation hypothesis . we update the corresponding regions of oscillation parameters and display the global behavior of @xmath17 and @xmath18 . we compare the situation before - and - after the recent 1489-day atmospheric super - k data samples and give the present gof of the oscillation hypothesis . in sec . [ sec : atm - compare ] , we briefly compare our atmospheric neutrino results with those of other analyses . finally , in sec . [ sec : conclusions ] we present our conclusions . in the following we will analyze solar neutrino data in the general framework of mixed active - sterile neutrino oscillations . in this case the electron neutrino produced in the sun converts into a combination of an active non - electron neutrino @xmath19 ( which again is a combination of @xmath1 and @xmath20 ) and a sterile neutrino @xmath21 : @xmath22 the parameter @xmath14 with @xmath13 describes the fraction of the sterile neutrino participating in the solar oscillations . therefore , the oscillation probabilities depend on the three parameters @xmath12 , @xmath11 and @xmath14 . the natural framework of light sterile neutrinos participating in oscillations are four - neutrino mass schemes , proposed to account for the lsnd result @xcite in addition to solar and atmospheric neutrino oscillations . for previous studies of solar neutrino oscillation in a four - neutrino framework see refs . @xcite and for an exact definition of the solar parameters and adopted approximations see ref . @xcite . as experimental data , we use the solar neutrino rates of the chlorine experiment homestake @xcite ( @xmath23 snu ) , the most recent result of the gallium experiments sage @xcite ( @xmath24 snu ) and gallex / gno @xcite ( @xmath25 snu ) , as well as the 1496-days super - kamiokande data sample @xcite in the form of 44 bins ( 8 energy bins , 6 of which are further divided into 7 zenith angle bins ) . in addition to this , we include the latest results from sno presented in refs . @xcite , in the form of 34 data bins ( 17 energy bins for each day and night period ) . therefore , in our statistical analysis we use @xmath26 observables , which we fit in terms of the three parameters @xmath12 , @xmath11 and @xmath14 , with a @xmath27 of the form @xmath28 in order to fully isolate the impact of the recent neutral current , spectral and day / night information of the sno result , we also present an analysis which does not include such information . to this aim we use only the sno events with energy higher than 6.75 mev , for which the nc component is negligible @xcite . we sum these events to a single rate , combining with cl , ga rates and full super - k data , as described above . this procedure is analogous to the pre - sno - nc situation , except that we take advantage of the enhanced statistics on the cc rate provided by the new data . we will refer to this analysis as sno@xmath29 analysis and it contains 48 data points . the comparison with the analysis including the complete sno data published this year ( sno@xmath30 ) allows us to highlight the impact of the sno nc , spectral and day / night information . for the solar neutrino fluxes we use the standard solar model ( ssm ) flux @xcite , including its standard @xmath31 flux prediction measurement of ref . @xcite . ] . motivated by the excellent agreement of the recent sno nc result with the predictions of the standard solar model , we prefer to adopt a boron - fixed analysis . however , for case of the lma solution we explicitly illustrate the effect of this assumption by performing also a boron - free analysis , where we treat the solar @xmath31 flux as free parameter in the fit . for simplicity we neglect the @xmath32 and @xmath33 neutrino fluxes , whose contribution to the present solar neutrino experiments is marginal , while for the @xmath34 , @xmath35 , @xmath36 , @xmath37 , @xmath38 and @xmath39 fluxes we use the ssm value given in ref . @xcite , taking properly into account their theoretical uncertainties and cross - correlations in the calculation of the @xmath40 function . for the neutrino cross sections of chlorine , sage , gallex / gno and super - k we assume the same as used in previous papers @xcite , while for the cc and nc neutrino deuteron differential cross sections relevant for sno we use the tables given in @xcite . the contribution of the cross - section uncertainties to the covariance matrix for the chlorine and gallium experiments is calculated as suggested in ref . @xcite . for a given experiment ( chlorine or gallium ) we use full correlation of the error on the cross section for low - energy neutrino fluxes ( @xmath34 , @xmath37 , @xmath35 , @xmath38 and @xmath39 ) , but no correlation of the cross section error between the low - energy fluxes and the higher - energy @xmath31 flux . the neutrino survival probability @xmath41 is extracted from the neutrino evolution operator @xmath42 , which we factorize as a product of three factors @xmath43 , @xmath44 and @xmath45 corresponding to propagation in the sun , vacuum , and earth , respectively . the first and last factors include matter effects with the corresponding density profiles given in refs . @xcite and @xcite . as a simplifying approximation , we assume that @xmath43 depends only on the neutrino production point @xmath46 , @xmath44 only on the sun - earth distance @xmath47 and @xmath45 depends only on the zenith - angle @xmath48 of the incoming neutrinos . therefore in our calculations we neglect the small correlation between seasonal effects and day - night asymmetry @xcite . for each value of the neutrino oscillation parameters @xmath49 , @xmath11 and @xmath14 we calculate the neutrino survival probability @xmath41 by averaging over @xmath46 , @xmath47 and @xmath48 , properly accounting for all the interference terms between @xmath43 , @xmath44 and @xmath45 . special care is taken in including all the theoretical and experimental errors and their cross - correlations in the calculation of the covariance matrix , for which we follow the description of ref . @xcite ( covariance approach ) . in particular , the errors associated to the boron - flux shape , the energy - scale and the energy - resolution uncertainties of the super - kamiokande and sno experiments are recalculated for each point in parameter space . in order to determine the expected event numbers for the various solar neutrino experiments we calculate the @xmath0 survival probability for each point in parameter space of ( @xmath50 ) and convolute it with the standard solar model neutrino fluxes @xcite and the relevant neutrino cross sections . we have compared such expected event numbers with the data described above , taking into account the detector characteristics and appropriate response functions . using the above - mentioned @xmath27 we have performed a global fit of solar neutrino data , whose results we now summarize . allowed regions of @xmath51 and @xmath12 for @xmath52 ( active oscillations ) , @xmath53 and @xmath54 . the lines indicate the regions determined by the sno@xmath29 analysis ( see definition in text ) , the shaded regions correspond to sno@xmath30 ( see text ) . the confidence levels are 90% , 95% , 99% and 3@xmath55 for 3 d.o.f .. ] our global best - fit point occurs for the values @xmath56 and correspond to @xmath57 . we obtain a @xmath58 for @xmath59 d.o.f . , leading to the excellent goodness of fit of 84% . in fig . [ fig : sol - region ] we display the regions of solar neutrino oscillation parameters for 3 d.o.f . with respect to this global minimum , for the standard case of active oscillations , @xmath60 , as well as for @xmath53 and @xmath54 . the first thing to notice is the impact of the sno nc , spectral , and day / night data in improving the determination of the oscillation parameters : the shaded regions after their inclusion are much smaller than the hollow regions delimited by the corresponding sno@xmath29confidence contours . especially important is the full sno@xmath30information for excluding _ maximal _ solar mixing in the lma region and in closing the lma region from above in @xmath12 . values of @xmath61 appear only at @xmath62 . previously solar data on its own could not close the lma region , only the inclusion of data from reactor experiments @xcite ruled out the upper part of the lma region @xcite . we obtain the following @xmath62 ranges ( 1 d.o.f . ) : @xmath63 it is interesting to note that these @xmath62 intervals are essentially unchanged if we minimize with respect to @xmath14 or if we apply the constraint @xmath57 ( pure active oscillations ) . in order to compare our allowed regions given in fig . [ fig : sol - region ] with those of other groups , one has to take into account that we calculate the c.l . regions for the 3 d.o.f.@xmath51 , @xmath12 and @xmath14 . therefore at a given c.l.our regions are larger than the usual regions for 2 d.o.f . , because we also constrain the parameter @xmath14 . next we notice the enhanced discrimination against non - lma solutions implied by the new data , apparent in figs . [ fig : sol - region ] , [ fig : sol - chisq ] and [ fig : sol - etas ] . this shows that the first hints @xcite in favor of a globally preferred lma oscillation solution which followed mainly from the flatness of the super - k spectra , have now become a robust result , thanks to the additional data , to which sno has contributed significantly . one sees that , in contrast with the sno@xmath29 situation , non - lma solutions do not appear at 95% c.l .. however , the low and vac solutions still appear at 99% c.l . for 3 d.o.f .. @xmath64 as a function of @xmath12 and @xmath51 , for pure active ( @xmath60 ) , pure sterile ( @xmath65 ) and mixed neutrino oscillations ( @xmath66 ) . upper and lower panels correspond to the sno@xmath29 and sno@xmath30 samples defined in text . ] in order to concisely illustrate the above results we display in fig . [ fig : sol - chisq ] the profiles of @xmath64 as a function of @xmath12 ( left ) as well as @xmath51 ( right ) , by minimizing with respect to the undisplayed oscillation parameters , for the fixed values of @xmath57 , @xmath67 , @xmath68 . by comparing top and bottom panels in fig . [ fig : sol - chisq ] one can clearly see the impact of the full sno@xmath30 sample in leading to the relative worsening of all non - lma solutions with respect to the preferred active lma solution . ? = ? ' '' '' l\|cccc\|cccc\| & & + region & @xmath51 & @xmath69 & @xmath70 & gof & @xmath51 & @xmath69 & @xmath70 & gof + + lma & 0.46 & @xmath71 & 40.9 & 69% & 0.46 & @xmath72 & ? 65.8 & 86% + low & 0.83 & @xmath73 & 46.2 & 46% & 0.66 & @xmath74 & ? 74.4 & 62% + vac & 1.7 ? & @xmath75 & 45.0 & 51% & 1.7 ? & @xmath76 & ? 74.4 & 63% + sma & @xmath77 & @xmath78 & 57.8 & 11% & @xmath79 & @xmath78 & ? 89.3 & 20% + just - so@xmath2 & 1.0 ? & @xmath80 & 59.6 & ? 9% & 1.0 ? & @xmath80 & ? 96.8 & ? 8% + + lma & 0.46 & @xmath81 & 45.4 & 50% & 0.42 & @xmath72 & ? 74.4 & 62% + low & 0.91 & @xmath82 & 51.1 & 28% & 0.83 & @xmath73 & ? 86.3 & 27% + vac & 1.6 ? & @xmath83 & 49.4 & 34% & 0.35 & @xmath84 & ? 81.3 & 41% + sma & @xmath85 & @xmath86 & 59.7 & ? 8% & @xmath87 & @xmath86 & ? 96.3 & ? 9% + just - so@xmath2 & 1.0 ? & @xmath80 & 59.8 & ? 8% & 1.0 ? & @xmath80 & ? 97.0 & ? 8% + + lma & 0.44 & @xmath88 & 56.0 & 15% & 0.38 & @xmath88 & ? 99.0 & ? 6% + low & 1.6 ? & @xmath89 & 58.5 & 10% & 1.6 ? & @xmath90 & 101.6 & ? 4% + vac & 1.7 ? & @xmath83 & 56.1 & 15% & 0.33 & @xmath84 & ? 89.1 & 21% + sma & @xmath91 & @xmath92 & 61.2 & ? 7% & @xmath85 & @xmath92 & ? 99.4 & ? 6% + just - so@xmath2 & 1.1 ? & @xmath80 & 59.9 & ? 8% & 1.0 ? & @xmath80 & ? 97.2 & ? 8% + the corresponding best - fit values for the various solutions of @xmath12 and @xmath11 and the values of @xmath27 evaluated at the best - fit points are compiled in tab . [ tab : sol - chisq ] . this table gives results for the three cases considered above : pure active , pure sterile and mixed neutrino oscillations , both for the sno@xmath29 and the full sno@xmath30 analysis . to calculate the goodness of fit of the various solutions we evaluate in this table the @xmath40 for @xmath93 @xmath94 d.o.f . for the sno@xmath29 ( sno@xmath30 ) analysis defined previously . note that we fix @xmath14 at the three values @xmath95 , @xmath67 and @xmath68 . in the pure active case we find for low , vac , sma and just - so@xmath2 the following differences in @xmath40 relative to the global best - fit point in lma @xmath96 note that especially sma and just - so@xmath2 are highly disfavored with respect to lma . in addition to the scrutiny of the different neutrino oscillation solutions in the solar neutrino oscillation parameters @xmath12 and @xmath11 , the present solar data can test the sterile neutrino oscillation hypothesis , characterized by the parameter @xmath14 introduced above . the results can be presented in several equivalent ways . for example , rejection of sterile solar neutrino oscillations is already hinted by comparing the middle and right panels of fig . [ fig : sol - region ] with the left one , corresponding to the pure active oscillation case : clearly the solutions deteriorate as @xmath14 increases . furthermore , the lines for @xmath97 and @xmath98 shown in fig . [ fig : sol - chisq ] clearly show that sterile solutions are strongly disfavored with respect to pure active solutions . @xmath64 displayed as a function of @xmath14 with respect to favored active lma solution , for the sno@xmath29 ( left panel ) and the sno@xmath30 ( right panel ) analysis , as defined in text . ] in order to summarize the above results we display in fig . [ fig : sol - etas ] the profile of @xmath64 as a function of @xmath99 , irrespective of the detailed values of the solar neutrino oscillation parameters @xmath12 and @xmath11 . this figure clearly illustrates the degree with which the solar neutrino data sample rejects the presence of a sterile component for each one of the possible solar neutrino oscillation solutions . the figure shows how the preferred lma status survives in the presence of a small sterile component characterized by @xmath14 ( also seen in figs . [ fig : sol - region ] and [ fig : sol - chisq ] ) . further , one sees that the value @xmath57 is always preferred , so that increasing @xmath14 leads to a deterioration of all oscillation solutions . notice that there is a crossing between the lma and vac solutions , as a result of which the best pure sterile description lies in the vacuum regime . however , in the global analysis pure sterile oscillations with @xmath98 are highly disfavored . we find a @xmath40-difference between pure active and sterile of @xmath100 if we restrict to the lma solution , or @xmath101 if we allow also for vac . for 3 d.o.f . the @xmath101 implies that pure sterile oscillations are ruled out at 99.997% c.l . compared to the active case . for the lma solution we have also performed an analysis without fixing the boron flux to its ssm prediction . in this case we treat the @xmath31 flux as a free parameter in the fit , and remove the error on this flux from the covariance matrix . from fig . [ fig : sol - etas ] one can see that the constraint on @xmath14 is weaker in the boron - free case than in the boron - fixed one , since a _ small _ sterile component can now be partially compensated by increasing the total @xmath31 flux coming from the sun . from the figure we obtain the bounds @xmath102 at 99% c.l . for 1 d.o.f .. in summary , we have found that , as long as the admixture of sterile neutrinos is acceptably small , the lma is always the best of the oscillation solutions , establishing its robustness also in our generalized oscillation scheme . to round off our discussion of the solar neutrino fit update we present in fig . [ fig : sol - probab ] the @xmath0 survival probability versus energy @xmath103 for the various solutions lma , low and vac , calculated as described above at the local @xmath40-minima given in tab . [ tab : sol - chisq ] . best - fit active solar neutrino survival probabilities.,scaledwidth=90.0% ] similar plots can be made for the case of sterile oscillations . ? = ? ' '' '' l\|c\|c\|c\|c\|c\|c\|c\|c\|c\|c\|c\| & & & & & & & & & & & + d.o.f . & @xmath104 & @xmath105 & @xmath104 & @xmath106 & @xmath107 & @xmath108 & @xmath109 & @xmath59 & @xmath59 & @xmath110 & @xmath110 + best fit & + @xmath51 & 0.34 & 0.38 & 0.39 & 0.41 & 0.45 & 0.45 & 0.40 & 0.41 & 0.42 & 0.47 & 0.46 + @xmath69 [ @xmath111 & 5.0 & 6.9 & 5.6 & 6.1 & 5.8 & 7.9 & 5.4 & 6.1 & 5.8 & 5.6 & 6.6 + @xmath112 & 57.0 & 43.5 & 50.7 & 40.6 & 75.4 & 33.0 & 30.8 & 65.2 & 73.4 & 68.0 & 65.8 + gof & 90% & 58% & 97% & 66% & 53% & 94% & 80% & 85% & 63% & 81% & 86% + @xmath113 & 10.7 & ? 9.0 & ? 9.2 & 10.0 & ? 9.6 & ? 8.1 & & 12.4 & 10.0 & & ? 8.7 + @xmath114 & & 10.0 & 25.6 & 15.5 & 10.1 & 14 . ? 9.7 & ? 7.8 & & ? 8.6 + @xmath115 & & 15.4 & 57.3 & 30.4 & 25.6 & 23 . ? & & 34.5 & 23.5 & & 23.5 + @xmath116 & & & & & & 29 . ? & & & & & 33.2 + @xmath117 & & & & & & & & & & & 35.9 + @xmath118 & & & & & 26.0 & & & & & & 23.3 + @xmath119 & & & & & 39.7 & & & & & & 33.6 + before turning to the atmospheric neutrino fits let us compare our solar neutrino results with those of other groups . since the release of the latest sno data in april 2002 several analyses have appeared . taking into account the large amount of experimental input data , variations in the analysis ( such as the construction of the @xmath40 function or the treatment of theoretical errors ) and the complexity of the codes involved it seems interesting to compare quantitatively the outcomes of different analyses . in tab . [ tab : sol - compare ] we have compiled some illustrative results of the solar neutrino analyses performed by the sno and super - k collaborations @xcite , as well as theoretical ones @xcite . generally speaking , on statistical grounds , one expects the differences in the statistical treatment of the data to have little impact on the global best - fit parameter values , which lie in the lma region for all analyses and are in good agreement . these differences typically become more visible as one compares absolute values of the @xmath40 , and/or as one departs from the best - fit region towards more disfavored solutions . aware of this , we took special care to details such as the dependence of the theoretical errors on the oscillation parameters , which enter in the covariance matrix characterizing the super - k and sno electron recoil spectra . this way we obtain results which we consider reliable in the full oscillation parameter space . in the row labeled `` d.o.f . '' we show the number of analyzed data points minus the fitted parameters in each analysis as a free fit parameter , since we consider only the limiting cases @xmath57 and @xmath68 ; this is the reason for the number @xmath110 in the present analysis . ] . one can see from these numbers that various groups use different experimental input data , in particular the spectral and zenith angle information of super - k and/or sno is treated in different ways . despite obvious differences in the analyses there is relatively good agreement on the best - fit lma active oscillation parameters : the obtained best - fit values for @xmath51 are in the range @xmath120 and for @xmath12 they lie in the interval @xmath121 . there is also good agreement on the allowed ranges of the oscillation parameters ( not shown in the table ) . for example , the @xmath62 intervals given in ref . @xcite ( @xmath122 and @xmath123 ) and in ref . @xcite ( @xmath124 and @xmath125 ) agree very well with the ranges given in eq . . however , even for the favored lma solution , there are some differences in the gof of the best - fit lma solution , ranging from 53% @xcite to 97% @xcite , due to differences in the construction of the @xmath40 function by different groups . there is remarkable agreement on the rejection of the low solution with respect to lma with a @xmath126 . our result for the vacuum solution @xmath127 is in good agreement with the values obtained in refs . @xcite , whereas refs . @xcite obtain higher values . our result for the sma solution of @xmath128 is in good agreement with the values obtained in refs . @xcite ; while refs . @xcite and especially ref . @xcite obtain higher values . on the other hand in ref . @xcite sma is somewhat less disfavored goes beyond the scope of this work . ] . there had been so - far no dedicated global analysis of solar neutrino oscillations including the most recent sno data for the case where sterile neutrinos take part in solar oscillations ( @xmath129 ) . model - independent considerations of transitions into sterile neutrinos can be found in refs . solar neutrino oscillations in the presence of active - sterile admixtures are also studied in ref . @xcite , although in a different context . in the lower part of tab . [ tab : sol - compare ] we compare the partial results given in refs . @xcite and @xcite for the pure sterile case ( @xmath98 ) with the corresponding values found in the present analysis . although there are noticeable differences of the shown @xmath130-values , there is agreement on the qualitative behavior . we have also studied intermediate levels of sterile neutrino admixture , giving the corresponding regions of oscillation parameters and the full @xmath40 profiles relative to the favored active lma solution ( not shown in tab . [ tab : sol - compare ] , see figs . [ fig : sol - region ] , [ fig : sol - chisq ] and [ fig : sol - etas ] ) . we now turn to the analysis of the latest atmospheric data . as already mentioned in the introduction , separate analyses of solar and atmospheric data samples constitute the necessary ingredients towards a full combined study of all current oscillation data , including also the short - baseline data , as shown in @xcite . in our analysis of atmospheric data we will make use of the hierarchy @xmath131 and neglect the solar mass splitting . further , in order to comply with the strong constraints from reactor experiments @xcite we completely decouple the electron neutrino from atmospheric oscillations . in the following we will consider atmospheric neutrino data in a generalized oscillation scheme in which a light sterile neutrino takes part in the oscillations . the setting for such scenarios are four - neutrino mass schemes @xcite . in such schemes , besides the solar and atmospheric mass - splittings , there is also a large @xmath132 motivated by the lsnd experiment @xcite . in contrast with the case of solar @xmath0 oscillations , the constraints on the @xmath1content in atmospheric oscillations are not so stringent : in fact such constrains are provided by atmospheric data themselves @xcite . as a result to describe atmospheric neutrino oscillations in this general framework @xcite we need two more parameters besides the standard 2-neutrino oscillation parameters @xmath133 and @xmath134 . we will use the parameters @xmath135 and @xmath136 already introduced in ref . @xcite , and defined in such a way that @xmath137 and @xmath138 correspond to the fractions of @xmath1 and @xmath21 participating in oscillations with @xmath134 , respectively . hence , pure active atmospheric oscillations with @xmath134 are recovered in the limit @xmath139 and @xmath140 . in four - neutrino models there is a mass scheme - dependent relationship between @xmath136 and the solar parameter @xmath14 . for details see ref . @xcite . we will also perform an analysis by imposing the constraint @xmath139 . in such `` restricted '' analysis the @xmath1 is completely constrained to the atmospheric mass states . only in this limit the parameter @xmath136 has a similar interpretation as @xmath14 introduced in the solar case . for @xmath139 we obtain that @xmath1 oscillates into a linear combination of @xmath20 and @xmath21 with @xmath134 : @xmath141 for the atmospheric data analysis we use all the charged - current data from the super - kamiokande @xcite and macro @xcite experiments . the super - kamiokande data include the @xmath142-like and @xmath143-like data samples of sub- and multi - gev contained events ( 10 bins in zenith angle ) , as well as the stopping ( 5 angular bins ) and through - going ( 10 angular bins ) up - going muon data events . we do not use the information on @xmath20 appearance , multi - ring @xmath143 and neutral - current events since an efficient monte - carlo simulation of these data sample would require a more detailed knowledge of the super kamiokande experiment , and in particular of the way the neutral - current signal is extracted from the data . such an information is presently not available to us . from macro we use the through - going muon sample divided in 10 angular bins @xcite . we did not include in our fit the results of other atmospheric neutrino experiments , such as the recent 5.9 kton - yr data from soudan-2 @xcite , since at the moment the statistics is completely dominated by super - kamiokande @xcite . furthermore , some of the older experiments have no angular sensitivity , and thus can not be used to discriminate between active and sterile neutrino conversion , our main goal . our statistical analysis of the atmospheric data is similar to that used in ref . @xcite , except that we now take advantage of the new super - kamiokande data and of the full ten - bin zenith - angle distribution for the contained events , rather than the five - bin distribution employed previously . we have now @xmath144 observables , which we fit in terms of the four relevant parameters @xmath134 , @xmath133 , @xmath135 and @xmath136 : @xmath145 concerning the theoretical monte - carlo , we improve the method presented in ref . @xcite by properly taking into account the scattering angle between the incoming neutrino and the scattered lepton directions . this was already the case for sub - gev contained events , however previously @xcite we made the simplifying assumption of full neutrino - lepton collinearity in the calculation of the expected event numbers for the multi - gev contained and up - going-@xmath143 data samples . while this approximation is still justified for the stopping and thru - going muon samples , in the multi - gev sample the theoretically predicted value for down - coming @xmath1 is systematically higher if full collinearity is assumed . the reason for this is that the strong suppression observed in these bins can not be completely ascribed to the oscillation of the down - coming neutrinos ( which is small due to small travel distance ) . because of the non - negligible neutrino - lepton scattering angle at these multi - gev energies there is a sizable contribution from up - going neutrinos ( with a higher conversion probability due to the longer travel distance ) to the down - coming leptons . however , this problem is less visible when the angular information of multi - gev events is included in a five angular bins presentation of the data , as previously assumed @xcite . folding together the atmospheric neutrino fluxes @xcite , our calculated neutrino survival probabilities including earth matter effects with the profile of ref . @xcite , and the relevant neutrino cross sections , we determine the expected event numbers for the various atmospheric neutrino observables , taking into account the appropriate detector response characteristics . comparing with the data described in sec . [ sec : atm - analysis ] , we have performed a global fit of the atmospheric neutrino data using the above - discussed @xmath146 , following the same method used in ref . we now summarize the main features of this fit . our global best - fit point occurs at the parameter values @xmath147 and @xmath148 . we see that atmospheric data prefers a small sterile neutrino admixture . however , this effect is not statistically significant , also the pure active case ( @xmath149 ) gives an excellent fit : the difference in @xmath40 with respect to the best - fit point is only @xmath150 . for the pure active best - fit point we obtain @xmath151 with the 3@xmath55 ranges ( 1 d.o.f . ) @xmath152 allowed regions of the parameters @xmath153 and @xmath134 at 90% , 95% , 99% and 3@xmath55 for 4 d.o.f . and different assumptions on the parameters @xmath136 and @xmath135 ( see text ) . the lines ( shaded regions ) correspond to 1289 ( 1489 ) days of super - k data.,scaledwidth=95.0% ] the determination of the parameters @xmath133 and @xmath134 is summarized in figs . [ fig : atm - region ] and [ fig : atm - chisq ] . at a given c.l . we cut the @xmath146 at a @xmath8 determined by 4 d.o.f.to obtain 4-dimensional volumes in the parameter space of ( @xmath154 ) . in the upper panels we show sections of these volumes at values of @xmath140 and @xmath139 corresponding to the pure active case ( left ) and at the best - fit point ( right ) . again we observe that moving from pure active to the best - fit does not change the fit significantly . in the lower right panel we project away both @xmath135 and @xmath136 , whereas in the lower left panel we fix @xmath155 and project away only @xmath135 . comparing the regions resulting from 1489 days super - k data ( shaded regions ) with those from the 1289 days super - k sample ( hollow regions ) we note that the new data leads to a slightly better determination of @xmath133 and @xmath134 however , more importantly , from the lower left panel we see , that the new data shows a stronger rejection against a sterile admixture : for @xmath155 no allowed region appears at 3@xmath55 for 4 d.o.f .. @xmath156 as a function of @xmath153 ( left ) and @xmath134 ( right ) , using 1289 ( upper ) and 1489 ( lower ) days of super - k data , for the case of neutrino oscillations with arbitrary @xmath136 and @xmath135 , best - fit @xmath136 and @xmath135 , pure active and mixed active - sterile neutrino oscillations . ] in fig . [ fig : atm - chisq ] we display the @xmath8 with respect to the global best - fit point as a function of @xmath153 ( for both signs of @xmath133 ) and @xmath134 , minimizing with respect to the other parameter , for different assumptions on the parameters @xmath136 and @xmath135 . in contrast to the solar case shown in fig . [ fig : sol - chisq ] the atmospheric @xmath40 exhibits a beautiful quadratic behavior , reflecting the fact that the oscillation solution to the atmospheric neutrino problem is robust and unique . notice again the significant worsening of the fit for the case of a sizable sterile neutrino admixture ( see , _ e.g. _ , the line corresponding to @xmath155 ) . ( a ) allowed regions of the parameters @xmath136 and @xmath135 at 90% , 95% , 99% and 3@xmath55 for 2 d.o.f .. the lines ( shaded regions ) correspond to 1289 ( 1489 ) days super - k data . further we show @xmath156 as a function of @xmath136 ( b ) and @xmath135 ( c ) , minimized with respect to all other parameters . the subscript `` r '' refers to the restricted analysis with @xmath139.,scaledwidth=95.0% ] in fig . [ fig : atm - etas ] we summarize the behavior of atmospheric @xmath40 with respect to the parameters @xmath136 and @xmath135 . indeed , the most striking result of the present improved analysis is the stronger rejection we now obtain on the fraction of the sterile neutrino @xmath157 in atmospheric oscillations . [ fig : atm - etas ] ( b ) clearly illustrates the degree with which the atmospheric neutrino data sample rejects the presence of a sterile component . on this basis one can place a model - independent atmospheric limit on the parameter @xmath136 , @xmath158 at 99% c.l . ( 1 d.o.f . ) . for the case of the restricted analysis , in which @xmath159 , we obtain , and not with respect to the global one . ] @xmath160 by comparing eqs . and we note the importance of taking into account the finite @xmath135 value in the analysis . although there is no substantial change in the 99% c.l . bounds on @xmath157 due to the new super - k data there is a huge effect for the case of sizable sterile neutrino admixtures , @xmath161 . in tab . [ tab : atm - chisq ] we have compiled the best - fit values of @xmath153 , @xmath134 , the @xmath40 values and the gof for the various atmospheric data samples for pure active and pure sterile oscillations . in the last column we give the difference in @xmath40 between active and sterile oscillation cases . comparing these numbers for the 1289 and 1489 days super - k samples we observe that all the new data except the sub - gev sample lead to a significant higher rejection against sterile oscillations . in combination with macro data the 1289 days super - k gave a difference between pure sterile and active oscillations of @xmath162 , whereas with the recent data we obtain @xmath163 showing that pure sterile oscillations are highly disfavored with respect to the active ones . let us note that macro data give an important contribution to this effect : macro alone disfavors the sterile oscillations already with @xmath164 . these limits on the sterile admixture are significantly stronger than obtained previously @xcite and play an important role in ruling out four - neutrino oscillation solutions in a combined global analysis of the lsnd anomaly @xcite . note , however , that in contrast with the case of @xmath136 , there is no substantial improvement in constraining the parameter @xmath135 due to the new data , as seen in fig . [ fig : atm - etas ] ( c ) . ? = ? ' '' '' l\|c\|cccc\|cccc\|c\| & & & & + data sample & d.o.f . & @xmath165 & @xmath166 $ ] & @xmath167 & gof & @xmath165 & @xmath166 $ ] & @xmath168 & gof & @xmath169 + + sk sub - gev & @xmath170 & @xmath171 & @xmath172 & @xmath173 & @xmath174 & @xmath171 & @xmath175 & @xmath176 & @xmath177 & @xmath178 + sk multi - gev & @xmath170 & @xmath171 & @xmath179 & @xmath180 & @xmath181 & @xmath182 & @xmath183 & @xmath184 & @xmath185 & @xmath186 + sk stop-@xmath143 & @xmath187 & @xmath171 & @xmath188 & @xmath189 & @xmath190 & @xmath191 & @xmath192 & @xmath193 & @xmath194 & @xmath195 + sk thru-@xmath143 & @xmath196 & @xmath197 & @xmath198 & @xmath199 & @xmath200 & @xmath201 & @xmath202 & @xmath203 & @xmath204 & @xmath205 + macro & @xmath196 & @xmath171 & @xmath206 & @xmath207 & @xmath208 & @xmath209 & @xmath210 & @xmath211 & @xmath212 & @xmath213 + sk contained & @xmath214 & @xmath171 & @xmath215 & @xmath216 & @xmath181 & @xmath217 & @xmath218 & @xmath219 & @xmath220 & @xmath221 + upgoing-@xmath143 & @xmath222 & @xmath171 & @xmath223 & @xmath224 & @xmath225 & @xmath226 & @xmath188 & @xmath227 & @xmath228 & @xmath229 + sk+macro & @xmath230 & @xmath171 & @xmath231 & @xmath232 & @xmath233 & @xmath234 & @xmath235 & @xmath236 & @xmath237 & @xmath238 + + sk sub - gev & @xmath170 & @xmath171 & @xmath239 & @xmath213 & @xmath240 & @xmath241 & @xmath215 & @xmath213 & @xmath240 & @xmath242 + sk multi - gev & @xmath170 & @xmath171 & @xmath243 & @xmath244 & @xmath245 & @xmath182 & @xmath183 & @xmath246 & @xmath247 & @xmath248 + sk stop-@xmath143 & @xmath187 & @xmath171 & @xmath235 & @xmath249 & @xmath225 & @xmath250 & @xmath235 & @xmath251 & @xmath228 & @xmath252 + sk thru-@xmath143 & @xmath196 & @xmath171 & @xmath183 & @xmath253 & @xmath254 & @xmath201 & @xmath255 & @xmath256 & @xmath257 & @xmath258 + macro & @xmath196 & @xmath171 & @xmath206 & @xmath207 & @xmath208 & @xmath209 & @xmath210 & @xmath211 & @xmath212 & @xmath213 + sk contained & @xmath214 & @xmath171 & @xmath259 & @xmath260 & @xmath181 & @xmath217 & @xmath218 & @xmath261 & @xmath185 & @xmath262 + upgoing-@xmath143 & @xmath222 & @xmath171 & @xmath218 & @xmath263 & @xmath264 & @xmath250 & @xmath265 & @xmath266 & @xmath212 & @xmath267 + sk+macro & @xmath230 & @xmath171 & @xmath268 & @xmath269 & @xmath181 & @xmath191 & @xmath231 & @xmath270 & @xmath271 & @xmath272 + zenith angle dependence of the @xmath143-like data used in our fit . further we show the predicted number of atmospheric neutrino events for best - fit , pure - active and pure - sterile oscillations and no oscillations . ] in order to better appreciate the excellent quality of the neutrino oscillation description of the present atmospheric neutrino data sample we display in fig . [ fig : atm - zenith ] the zenith angle distribution of atmospheric neutrino events . clearly , active neutrino oscillations describe the data very well indeed . in contrast , no oscillations can be visually spotted as being inconsistent . on the other hand conversions to sterile neutrinos lead to an excess of events for neutrinos crossing the core of the earth , in all the data samples except sub - gev . let us briefly compare our atmospheric neutrino oscillation results with those of other groups . apart from the analyses presented in refs . @xcite there had been no other complete atmospheric neutrino analysis taking into account the most general structure of neutrino mixing in the presence of sterile neutrinos , characterized by four mixing parameters . in the analyses of refs . @xcite the @xmath1 is restricted to the atmospheric mass states , which corresponds to the constraint @xmath139 in our parameterization . however , at the corresponding limiting cases our generalized analysis can be compared with the results of other works . let us further note that the analysis of ref . @xcite is based on the 1289-days sk data sample ( 79.5 kton - yr ) and in contrast to refs . @xcite we use also data from the macro experiment . first , we find very good agreement in the case of pure active oscillations : the agreement of our best - fit values given in eq . with those obtained by the super - k collaboration ( @xmath273 @xcite ) is excellent , with good agreement also with the results of ref . @xcite ( @xmath274 ) . similarly , also the allowed ranges shown in the upper left panel of fig . [ fig : atm - region ] compare very well with the ranges obtained in refs . this shows that the determination of the active atmospheric oscillation parameters is already rather stable with respect to variations in the analysis and inclusion of additional of data . concerning admixtures of sterile neutrinos , we note that it is presently not possible to use information on @xmath20 appearance , multi - ring @xmath143 and neutral - current events outside the super - k collaboration , because to simulate these data a detailed knowledge of the detector and the applied cuts is necessary . these classes of events should provide additional sensitivity towards rejecting a possible contribution of sterile neutrinos . therefore , the fact that the value of @xmath275 @xcite between pure active and sterile oscillations obtained by the super - k collaboration is higher than our value 34.6 given in eq . is understandable , since with the super - k data accessible to us we have a reduced discrimination between active and sterile oscillations , based solely on the matter effects . prompted by the recent data on solar and atmospheric neutrinos we have reanalyzed the global status of oscillation solutions , taking into account the that both the solar @xmath0 and the atmospheric @xmath1 may convert to a mixture of active and sterile neutrinos . in addition to the sno neutral current , spectral and day / night ( sno@xmath30 ) results we add the latest 1496-day solar and 1489-day atmospheric super - k neutrino data samples . we have studied the impact of the recent solar data in the determination of the regions of oscillation parameters for different allowed @xmath14 values , displaying the global behavior of @xmath15 and @xmath16 , calculated with respect to the favored active lma solution . we have investigated in detail the impact of the full cl + ga rates + super - k spectra + the complete sno@xmath30 data set , comparing with the situation when the this year s sno data is left out . we confirm the clear preference for the lma solution of the solar neutrino problem and obtain that the low , vac , sma and just - so@xmath2 solutions are disfavored with a @xmath276 , @xmath4 , @xmath5 and @xmath6 , respectively , for the pure active case . in addition , we find that the global solar data sample constrains admixtures of a sterile neutrino to be smaller than 44% at 99% c.l .. this bound is relaxed to 61% when the solar @xmath31 flux is treated as a free parameter . a pure sterile solution is ruled out with respect to the active one at 99.997% c.l .. for allowed sterile neutrino admixtures lma is always the best of all the oscillation solutions . we remark , however , the existence of non - oscillation solutions @xcite . these will be crucially tested @xcite at the up - coming kamland reactor experiment @xcite . by performing an improved fit of the atmospheric data , we have also updated the corresponding regions of oscillation parameters for the case where the atmospheric @xmath1 convert to a mixture of active and sterile neutrinos . we have displayed the global behavior of @xmath17 and @xmath18 for different allowed values of the sterile neutrino admixture in the atmospheric channel . we have compared the situation before - and - after the recent 1489-day atmospheric super - k data samples and shown that the gof of the oscillation hypothesis is excellent . we have found that the recent 1489-day atmospheric super - k data strongly constrain a sterile component in atmospheric oscillations : if the @xmath1 is restricted to the atmospheric mass states only a sterile admixture of 16% is allowed at 99% c.l . , while a bound of 35% is obtained in the unconstrained case . pure sterile oscillations are disfavored with a @xmath7 compared to the active case . we thank mark chen , andre hamer , art mcdonald and scott oser , for help in accessing the details of the sno experiment needed for our analysis . this work was supported by spanish grant bfm2002 - 00345 , by the european commission rtn network hprn - ct-2000 - 00148 and by the european science foundation network grant n. 86 . . has been supported by the doc fellowship of the austrian academy of science and , in the early stages of this work , by a fellowship of the european commission research training site contract hpmt-2000 - 00124 of the host group . m.m . was supported by contract hpmf - ct-2000 - 01008 and m.a.t . was supported by the m.e.c.d . fellowship ap2000 - 1953 . @xmath8 as a function of @xmath12 and @xmath51 , for solar+chooz , kamland alone , and solar+reactor ( both chooz and kamland ) data . ] in a recent paper the first results of the kamland collaboration became public @xcite . the kamland experiment is a reactor neutrino experiment whose detector is located at the kamiokande site . most of the @xmath277 flux incident at kamland comes from nuclear plants at distances of @xmath278 km from the detector , making the average baseline of about 180 kilometers , long enough to provide a sensitive probe of the lma - msw region . the target for the @xmath277 flux consists of a spherical transparent balloon filled with 1000 tons of non - doped liquid scintillator , and the anti - neutrinos are detected via the inverse neutron @xmath279-decay process @xmath280 . the kamland collaboration has for the first time measured the disappearance of neutrinos produced in a power reactor . they observe a strong evidence for the disappearance of neutrinos during their flight over such distances , giving the first terrestrial confirmation of the solar neutrino anomaly and also establishing the oscillation hypothesis with man - produced neutrinos . in this appendix ( which does not appear in the published version of this paper ) we analyze the implications of the first 145.1 days of kamland data on the determination of the solar neutrino parameters . the details of our theoretical monte - carlo and statistical analysis are given in ref . @xcite ; in particular , the kamland @xmath40-function is calculated assuming a poisson distribution for the experimental data , as described in sec . iv of that paper . allowed regions of @xmath51 and @xmath12 for @xmath52 ( active oscillations ) , @xmath53 and @xmath54 . the lines correspond to the analysis of solar+chooz data , while the shaded regions correspond to the combination of solar+reactor ( both chooz and kamland ) data . both lines and shaded regions refer to 90% , 95% , 99% and 99.73% confidence intervals for 3 d.o.f .. ] the impact of the kamland result on @xmath11 and @xmath12 for the case of pure active oscillations has already been discussed in detail in ref . @xcite , and the results are summarized here in fig . [ fig : kam - chisq ] . first of all , we note that non - lma solutions , characterized by a very small value of either @xmath12 ( low , vac , just - so@xmath2 ) or @xmath11 ( sma ) are in disagreement with the evidence for @xmath277 disappearance observed in kamland . as a consequence , the relative quality of these solutions with respect to lma is worsened by an extra @xmath281 when the kamland data are also included in the analysis , so that eq . is now replaced by : @xmath282 this led to the conclusion that lma is presently the only allowed solution to the solar neutrino problem . the global best - fit point occurs for pure - active oscillations , and is practically unaffected by the inclusion of kamland ( cfr . ): @xmath283 however , as can be seen from fig . [ fig : kam - region ] the lma region is now split into two sub - regions , and from the left panel of fig . [ fig : kam - chisq ] we see that a secondary minimum characterized by @xmath284 and @xmath285 appears . the relative quality of this point with respect to the global best - fit point given in eq . is @xmath286 . the allowed @xmath62 ranges for @xmath12 and @xmath11 are ( cfr.eq . ): @xmath287 let us now consider the impact of kamland on the determination of @xmath14 . from fig . [ fig : kam - region ] we have a first indication that the bound on the fraction of sterile neutrino participating in solar neutrino oscillations is essentially unaffected by this experiment . this can be easily understood since kamland is only sensitive to the anti - neutrino survival probability @xmath41 , and is therefore unable to discriminate between different oscillation channels . taking into account that matter effects induced by the earth mantle are practically negligible given the short distance traveled by the neutrinos in their flight between the source and the detector , it is straightforward to conclude that kamland is completely insensitive to @xmath14 . left panel : @xmath64 and @xmath288 displayed as functions of @xmath14 , for both fixed and free @xmath31 flux . right panel : @xmath289 allowed regions for @xmath12 from the analysis of solar+chooz data , for both fixed ( green ) and free ( red ) @xmath31 flux . the dark ( light ) gray horizontal belts correspond to the @xmath289 ( @xmath290 allowed regions for @xmath12 from the analysis of kamland data . ] in the left panel of fig . [ fig : kam - etas ] we display the profile of @xmath64 and @xmath288 as functions of @xmath14 , irrespective of the detailed values of the solar neutrino oscillation parameters @xmath12 and @xmath11 . the dashed lines correspond to free @xmath31 flux , while for the solid lines the boron flux is fixed to its ssm prediction . as expected , for the boron - fixed case the inclusion of kamland is almost completely irrelevant for the determination of the sterile neutrino fraction , and for @xmath291 no visible difference arises between the pre - kamland and post - kamland cases . however , looking at fig . [ fig : kam - etas ] we see that for the boron - free case the inclusion of the kamland data is relevant . at 99% c.l . we obtain the bounds ( cfr.eq . ): @xmath292 to understand why fixing or not - fixing the @xmath31 flux leads to such a different behavior , we illustrate in the right panel of fig . [ fig : kam - etas ] the dependence on @xmath14 of the @xmath289 allowed range for @xmath12 from the analysis of solar data alone , both for the boron - fixed ( green ) and the boron - free ( red ) cases . these two bands should be confronted with the horizontal gray belts . ] , which correspond to the kamland @xmath289 ( dark gray ) and @xmath293 ( light gray ) allowed intervals for @xmath12 . concerning the boron - fixed case , we see that for @xmath294 the @xmath12 value preferred by solar data is almost insensitive to @xmath14 , and in very good agreement with the kamland prediction . in this regime a non - zero value of @xmath14 leads to a mild deficit in the expected number of nc and es events , thus reducing the quality of the fit . when @xmath14 exceeds @xmath295 , this deficit become relevant , and in order to compensate it the best - fit point moves towards regions of the neutrino parameter space where the electron neutrino survival probability @xmath41 is larger . from fig . [ fig : kam - probab ] it is easy to understand that this correspond to larger values of @xmath12 . in any case , as long as @xmath296 the @xmath12 regions allowed at @xmath289 by solar and kamland data still overlap , and this explains why in this regime there is no visible difference between @xmath64 and @xmath288 . the situation is different for the boron - free case . as for the previous case , an increase of @xmath14 lead to a deficit in the expected number of nc and es events , which can now be compensated by assuming a larger value of the @xmath31 flux . however , a larger boron flux results in an excess of @xmath0 arriving at the detectors , so that now the experimental data favor regions of the parameter space where the electron neutrino survival probability @xmath41 is smaller . this explains why the preferred value for @xmath12 decreases as @xmath14 increases . when @xmath297 the favored value for the @xmath31 flux rapidly decreases , and @xmath12 increases again . a consequence of this is that already for @xmath298 the @xmath289 regions for solar and for kamland data no longer overlap . this leads to a tension between the two data sets , which results in an excess of @xmath288 over @xmath64 . it is only for very large values of @xmath14 that solar data return in agreement with kamland , and the two lines @xmath64 and @xmath288 merges again . in summary , in this appendix we have investigated the impact of the first 145.1 days of kamland data on the determination of the solar neutrino parameters . we have found that all non - lma solution are now ruled out , and that the original lma region is split into two relatively narrow islands around the values of @xmath300 ( best fit point ) and @xmath285 ( local minimum ) . the bound on the sterile neutrino fraction @xmath14 is practically unaffected in the boron - fixed case , but improves from @xmath191 to @xmath241 ( at 99% c.l . ) in the boron - free case . m. smy , talk at neutrino 2002 , ` http://neutrino2002.ph.tum.de/ ` and hep - ex/0202020 ; super - kamiokande collaboration , s. fukuda _ et al . _ , phys . b * 539 * ( 2002 ) 179 [ hep - ex/0205075 ] ; y. fukuda _ et al . lett . * 81 * , 1158 ( 1998 ) ; erratum _ ibid . _ * 81 * , 4279 ( 1998 ) and _ ibid_. * 82 * , 1810 and 2430 ( 1999 ) ; y. suzuki , nucl . b ( proc . suppl . ) * 77 * , 35 ( 1999 ) ; s. fukuda _ et al . _ , hep - ex/0103032 . gallex coll . , w. hampel _ et al . _ , phys . lett . * b447 * , 127 ( 1999 ) ; m. altmann _ _ , gno coll . b 490 ( 2000 ) 16 , hep - ex/0006034 ; c. cattadori , gno coll . , nucl . * 110 * ( 2002 ) 311 ; t. kirsten , talk at neutrino 2002 , ` http://neutrino2002.ph.tum.de/ ` . q. r. ahmad _ et al . _ [ sno coll . ] , phys . 87 * , 071301 ( 2001 ) [ nucl - ex/0106015 ] . q. r. ahmad _ et al . _ [ sno coll . ] , phys . rev . lett . * 89 * , 011301 ( 2002 ) [ nucl - ex/0204008 ] . q. r. ahmad _ et al . _ [ sno coll . ] , phys . lett . * 89 * , 011302 ( 2002 ) [ nucl - ex/0204009 ] . o. g. miranda _ et al . _ , nucl . b * 595 * ( 2001 ) 360 [ hep - ph/0005259 ] ; phys . b * 521 * ( 2001 ) 299 [ hep - ph/0108145 ] , and references therein . j. barranco , o. g. miranda , t. i. rashba , v. b. semikoz and j. w. valle , hep - ph/0207326 . m. guzzo _ et al . _ , nucl . b * 629 * ( 2002 ) 479 [ hep - ph/0112310 ] , and references therein . y. fukuda _ et al . _ , kamiokande coll . b * 335 * ( 1994 ) 237 ; r. becker - szendy _ et al . _ , imb coll . b ( proc . suppl . ) * 38 * ( 1995 ) 331 ; w.w.m . et al . _ , soudan coll . b * 449 * ( 1999 ) 137 . macro coll . , m. ambrosio _ et al . _ , b * 434 * ( 1998 ) 451 ; m. spurio _ _ , hep - ex/0101019 ; b. barish , talk given at _ neutrino 2000 _ , 1521 june 2000 , sudbury , canada ` http://nu2000.sno.laurentian.ca ` . m. apollonio _ et al . _ , chooz coll . b * 466 * ( 1999 ) 415 ; f. boehm _ et al . _ , palo verde coll . d * 64 * ( 2001 ) 112001 [ hep - ex/0107009 ] . j. t. peltoniemi , d. tommasini and j. w. f. valle , phys . b * 298 * ( 1993 ) 383 . j. t. peltoniemi and j. w. f. valle , nucl . b * 406 * , 409 ( 1993 ) [ hep - ph/9302316 ] . caldwell and r.n . mohapatra , phys . d * 48 * ( 1993 ) 3259 . lsnd collaboration , c. athanassopoulos _ et al . * 77 * ( 1996 ) 3082 ; _ ibid _ * 81 * ( 1998 ) 1774 ; a. aguilar _ et al . _ , phys . d * 64 * ( 2001 ) 112007 [ hep - ex/0104049 ] . m. maltoni , t. schwetz and j.w.f . valle , phys . d * 65 * ( 2002 ) 093004 [ hep - ph/0112103 ] . m. maltoni , t. schwetz , m. a. tortola and j. w. valle , nucl . b * 643 * , 321 ( 2002 ) [ arxiv : hep - ph/0207157 ] . j. n. bahcall , m. c. gonzalez - garcia and c. pena - garay , jhep * 0207 * , 054 ( 2002 ) [ arxiv : hep - ph/0204314 ] . a. bandyopadhyay , s. choubey , s. goswami and d. p. roy , phys . b * 540 * , 14 ( 2002 ) [ hep - ph/0204286 ] . v. barger , d. marfatia , k. whisnant and b. p. wood , phys . b * 537 * , 179 ( 2002 ) [ hep - ph/0204253 ] . aliani _ et al . _ , hep - ph/0205053 . p.c . de holanda and a.yu . smirnov hep - ph/0205241 . p. creminelli _ _ , hep - ph/0102234 ( updated on 22 april 2002 including the sno nc and day / night results ) . g. l. fogli , e. lisi , a. marrone , d. montanino and a. palazzo , hep - ph/0206162 . j. n. bahcall , m. c. gonzalez - garcia and c. pena - garay , phys . c * 66 * , 035802 ( 2002 ) [ arxiv : hep - ph/0204194 ] . m. c. gonzalez - garcia , m. maltoni and c. pena - garay , phys . d * 64 * , 093001 ( 2001 ) . a. r. junghans _ et al . _ , phys . * 88 * , 041101 ( 2002 ) [ nucl - ex/0111014 ] . m. c. gonzalez - garcia , p. c. de holanda , c. pena - garay and j. w. valle , nucl . b * 573 * ( 2000 ) 3 [ hep - ph/9906469 ] . m. c. gonzalez - garcia , m. maltoni , c. pena - garay and j. w. f. valle , phys . d * 63 * ( 2001 ) 033005 [ hep - ph/0009350 ] ; n. fornengo , m.c . gonzalez - garcia and j. w. f. valle , nucl . b * 580 * ( 2000 ) 58 ; m. c. gonzalez - garcia , h. nunokawa , o. l. g. peres and j. w. f. valle , nucl . phys . b * 543 * ( 1999 ) 3 [ hep - ph/9807305 ] . j. n. bahcall , m. c. gonzalez - garcia and c. pena - garay , jhep * 0204 * , 007 ( 2002 ) [ hep - ph/0111150 ] . k. kubodera s web - page ` http://nuc003.psc.sc.edu/~kubodera/ ` . de holanda , c. pena - garay , m.c . gonzalez - garcia and j.w.f . valle , phys . d * 60 * , 093010 ( 1999 ) [ hep - ph/9903473 ] ; g.l . fogli _ et al . _ , phys . d * 62 * , 113004 ( 2000 ) [ hep - ph/0005261 ] ; e. lisi _ et al . _ , phys . d * 63 * , 093002 ( 2001 ) [ hep - ph/0011306 ] ; e. lisi and d. montanino , phys . d * 56 * , 1792 ( 1997 ) [ hep - ph/9702343 ] . j. n. bahcall , p. i. krastev and a. y. smirnov , phys . d * 60 * , 093001 ( 1999 ) [ arxiv : hep - ph/9905220 ] . a. strumia _ et al . _ , hep - ph/0205261 . m. c. gonzalez - garcia , h. nunokawa , o. l. peres , t. stanev and j. w. valle , phys . d * 58 * ( 1998 ) 033004 [ hep - ph/9801368 ] . g. barr , t. k. gaisser and t. stanev , phys . d * 39 * ( 1989 ) 3532 ; phys . d * 38 * ( 1988 ) 85 ; t. k. gaisser and t. stanev , phys . rev . d * 57 * , 1977 ( 1998 ) . fogli , e. lisi and a. marrone , phys . d * 64 * ( 2001 ) 093005 [ hep - ph/0105139 ] . e. k. akhmedov and j. pulido , hep - ph/0209192 . a. piepke , kamland coll . b ( proc . suppl . ) 91 , 99 ( 2001 ) ; ` http://kamland.lbl.gov/kamland ` .	we analyze the impact of recent solar and atmospheric data in the determination of the neutrino oscillation parameters , taking into account that both the solar @xmath0 and the atmospheric @xmath1 may convert to a mixture of active and sterile neutrinos . we use the most recent global solar neutrino data , including the 1496-day super - k neutrino data sample , and we investigate in detail the impact of the recent sno neutral current , spectral and day / night data by performing also an analysis using only the charged current rate from sno . we confirm the clear preference of the pure active lma solution of the solar neutrino problem and obtain that the low , vac , sma and just - so@xmath2 solutions are disfavored with a @xmath3 , @xmath4 , @xmath5 , @xmath6 , respectively . furthermore , we find that the global solar data constrains the admixture of a sterile neutrino to be less than 44% at 99% c.l .. a pure sterile solution is ruled out with respect to the active one at 99.997% c.l .. by performing an improved fit of the atmospheric data , we also update the corresponding regions of oscillation parameters . we find that the recent atmospheric super - k ( 1489-day ) and macro data have a strong impact on constraining a sterile component in atmospheric oscillations : if the @xmath1 is restricted to the atmospheric mass states only a sterile admixture of 16% is allowed at 99% c.l . , while a bound of 35% is obtained in the unconstrained case . pure sterile oscillations are disfavored with a @xmath7 compared to the pure active case . in the appendix we discuss the implications of the first 145.1 days of kamland data on the determination of the solar neutrino parameters . the inclusion of kamland enhances the rejection of non - lma - msw solutions by 13 units in @xmath8 . the bound on the sterile neutrino fraction is practically unaffected in the boron - fixed case , while it improves from 61% to 51% in the boron - free case .
in typical phase i studies in the development of relatively benign drugs , the drug is initiated at low doses and subsequently escalated to show safety at a level where some positive response occurs , and healthy volunteers are used as study subjects . this paradigm does not work for diseases like cancer , for which a non - negligible probability of severe toxic reaction has to be accepted to give the patient some chance of a favorable response to the treatment . moreover , in many such situations , the benefits of a new therapy may not be known for a long time after enrollment , but toxicities manifest themselves in a relatively short time period . therefore , patients ( rather than healthy volunteers ) are used as study subjects , and given the hoped - for ( rather than observed ) benefit for them , one aims at an acceptable level of toxic response in determining the dose . current designs for phase i cancer trials , which are sequential in nature , are an ad hoc attempt to reconcile the objective of finding a _ maximum tolerated dose _ ( mtd ) with stringent ethical demands for protecting the study subjects from toxicities in excess of what they can tolerate . it treats groups of three patients sequentially , starting with the smallest of an ordered set of doses . escalation occurs if no toxicity is observed in all three patients ; otherwise an additional three patients are treated at the same dose level . if only one of the six patients has toxicity , escalation again continues ; otherwise the trial stops , with the lower dose declared as mtd . as pointed out by storer ( @xcite ) , these designs , commonly referred to as 3-plus-3 designs , are difficult to analyze , since even a strict quantitative definition of mtd is lacking , `` although it should be taken to mean some percentile of a tolerance distribution with respect to some objective definition of clinical toxicity , '' and the `` implicitly intended '' percentile seems to be the 33rd percentile ( related to 2@xmath16 ) . storer ( @xcite ) also considered three other `` up - and - down '' sequential designs for quantile estimation in the bioassay literature and performed simulationstudies of their performance in estimating the 33rd percentile . subsequent simulation studies byoquigley et al . ( @xcite ) showed the performance of these designs to be `` dismal , '' for which they provided the following explanation : `` not only do ( these designs ) not make efficient use of accumulated data , they make use of no such data at all , beyond say the previous three , or sometimes six , responses . '' they proposed an alternative design , called the _ continual reassessment method _ ( crm ) , which uses parametric modeling of the dose response relationship and a bayesian approach to estimate the mtd or , more generally , the dose level @xmath2 such that the probability @xmath3 of a toxic event is @xmath4 ( @xmath5 in the case of mtd ) . letting @xmath6 and assuming the usual logistic model @xmath7 for the probability of a toxic response at dose level @xmath2 , the problem of optimal choice of @xmath8 dose levels to estimate the mtd seems to be covered by the theory of nonlinear designs . a well - known difficulty in nonlinear design theory is that the optimal design for parameter estimation involves the unknown parameter vector . to circumvent the difficulty , it has been proposed that the design be constructed sequentially , using observations made to date to estimate @xmath9 by maximum likelihood and choosing the next design point by using the mle to replace the unknown parameter value in the optimal design ; see fedorov ( @xcite ) . if @xmath9 is known , then a target probability @xmath4 of response is attained at the level @xmath10 that solves @xmath11 , that is , @xmath12/\beta$ ] . wu ( @xcite ) proposed to use at stage @xmath13 the _ certainty equivalence _ ( or plug - in ) level @xmath14 , where @xmath15 is the mle of @xmath9 based on @xmath16 , @xmath17 , and @xmath18 is the binary response at dose level @xmath19 . using some approximations , he also derived a recursive representation of @xmath14 and showed that it is asymptotically equivalent ( as @xmath20 ) to the adaptive stochastic approximation rule of lai and robbins ( @xcite ) . the likelihood version of crm proposed by oquigley and shen ( @xcite ) in response to the comments of korn et al . ( @xcite ) on bayesian designs is in fact a variant of wu s ( @xcite ) design . babb et al . ( @xcite ) pointed out that the symmetric nature of squared error loss used by crm may not be appropriate for modeling the toxic response to a cancer treatment . they proposed the _ escalation with overdose control _ ( ewoc ) method , which uses an asymmetric linear loss function that penalizes dose level @xmath21 for @xmath22 , corresponding to an overdose , more than an under - dose @xmath23 . whereas crm is equivalent to estimating the mtd at each stage by the mean of the posterior distribution of @xmath10 , ewoc is equivalent to estimating the mtd at each stage by the @xmath24th quantile of the posterior distribution of @xmath10 , where @xmath25 is the so - called _ feasibility bound _ , usually chosen to be slightly less than @xmath4 . there has also been much work on designs intended to give an accurate post - experiment estimate of the @xmath26 or other functions of the unknown parameter vector @xmath9 . for example , locally optimal designs such as @xmath0- and @xmath27-optimal designs have been investigated extensively for binary responses , and because a nonlinear model s information matrix for binary data is a function of the unknown parameters , locally optimal designs are usually applied using initial estimates , multistage methods or bayesian priors . haines , perevozskaya and rosenberger ( @xcite ) proposed a two - stage design whose first stage is a locally optimal design based on a chosen prior , which is then updated sequentially during its second stage . a comparative study of these methods is given in section [ sec:2stage ] . since the parameter @xmath9 is unknown , _ active _ statistical learning involves setting the doses at levels that give maximum information about the function of the unknown parameters of interest , the mtd , and how to do this is a problem in nonlinear experimental design theory ( abdelbasit and plackett , @xcite ; dette et al . , @xcite ) . on the other hand , there is also an ethical issue of treating patients in the phase i trial at dose levels below the unknown mtd for safety , and hopefully close to the mtd for efficacy . this dilemma between treatment of current patients and efficient experimentation to gather information for future patients was articulated by lai and robbins ( @xcite ) in a simple linear regression model @xmath28 , where , instead of the mtd , the desired level is @xmath29 , for some given value @xmath30 . whereas an asymptotic theory of how this dilemma can be resolved optimally as @xmath31 was developed by lai and robbins ( @xcite ) , it was quite recent that a tractable scheme was developed by han , lai and spivakovsky ( @xcite ) to compute an approximately optimal solution for finite sample size @xmath8 ( number of patients enrolled in the trial ) . in section [ sec2 ] we introduce a basic stochastic optimization problem that incorporates the treatment versus experimentation dilemma in the design of phase i cancer trials . this problem adopts a bayesian formulation as in crm and ewoc , for which the computation of the posterior distributions of the parameters and of the mtd is described in section [ sec2 ] . because the regression function @xmath32 for the binary response @xmath33 given by ( [ 1 ] ) is nonlinear in the parameters , the stochastic optimization problem is considerably more difficult than the linear regression model @xmath34 considered by lai and robbins ( @xcite ) . we review in section [ sec2 ] recent advances in the field of _ approximate dynamic programming _ , which we use in section [ sec : hyb_des ] to develop a new tool for tackling the stochastic optimization problem . using this tool , we derive nearly optimal _ hybrid designs _ in section [ sec : hyb_des ] . these hybrid designs are convex combinations ( and therefore hybrids ) of designs that are targeted toward treating the current patient at the best guess of the mtd ( e.g. , ewoc and crm ) and the haines perevozskaya rosenberger designs that are @xmath27- or @xmath0-optimal in estimating the model parameters for future patients . the weights in these convex combinations are determined by approximate dynamic programming and can be conveniently stored to provide simple table look - up schemes for the clinical user , as noted in section [ sec : conc ] which gives some concluding remarks . section [ sec:2stage ] provides a comparative study of the hybrid design and previous designs . it also introduces a modified hybrid design that incorporates the traditional nonparametric step - up / down scheme as a cautious first stage , followed by the model - based design in the second stage of a phase i cancer trial . to begin with , we specify a prior distribution on @xmath9 by following babb et al . ( @xcite ) who first specify a range @xmath35 $ ] of possible dose values believed to contain the mtd , with @xmath36 believed to be a conservative starting value . rather than directly specifying the prior distribution @xmath37 for the unknown parameter @xmath9 of the working model to be used in the second stage , which may be hard for investigators to do in practice , an upper bound @xmath38 on the probability @xmath39 of toxicity at @xmath36 can be elicited from investigators ; uniform distributions over @xmath40 $ ] and @xmath41 $ ] are then taken as the prior distributions for the mtd and @xmath42 , respectively . let @xmath43 denote the information set generated by the first @xmath44 doses and responses , that is , by @xmath45 . letting @xmath46 denote the mtd , it is convenient to transform from the unknown parameters @xmath47 in the two - parameter logistic model ( [ 1 ] ) to @xmath48 via the formulas @xmath49 \label{3}\beta&=&\frac{\log(1/\rho-1)-\log(1/p-1)}{\eta - x_{\min}},\vspace{-2pt}\end{aligned}\ ] ] giving @xmath50 & & { } -(x - x_{\min } ) \log(1/p-1)\bigr)\nonumber \\[-8pt]\\[-8pt ] & & /(\eta - x_{\min})\nonumber \\[-2pt ] & = & \psi(x,\rho,\eta).\nonumber\vspace{-2pt}\end{aligned}\ ] ] assuming that the joint prior distribution of @xmath48 has density @xmath51 with support on @xmath41\times [ x_{\min},x_{\max}]$ ] , the @xmath43-posterior distribution of @xmath48 has density @xmath52 & & \quad = c^{-1}\prod_{i=1}^{k}\biggl[\frac { 1}{1+e^{-\psi ( x_i,\rho,\eta)}}\biggr]^{y_i } \\[-2pt ] & & { } \qquad \hspace{34pt}\cdot \biggl[\frac{1}{1+e^{\psi(x_i,\rho,\eta)}}\biggr]^{1-y_i } \pi(\rho,\eta),\nonumber\vspace{-2pt}\end{aligned}\ ] ] where @xmath53^{y_i } \\[-2pt ] & & { } \hspace{57pt}\cdot \biggl[\frac{1}{1+e^{\psi(x_i,\rho,\eta)}}\biggr]^{1-y_i } \\[-2pt ] & & { } \hspace{42pt}\times \pi ( \rho,\eta ) \,d\rho\ , d\eta\vspace{-2pt}\end{aligned}\ ] ] is the normalizing constant . the marginal @xmath43-posterior distribution of @xmath46 is then @xmath54 the aforementioned crm and ewoc doses based on @xmath43 are the mean and the @xmath24-quantile of ( [ 6 ] ) . note that using ewoc or crm amounts to the `` myopic '' policy of dosing the ( @xmath55)th patient at the dose @xmath56 that minimizes @xmath57 $ ] , in which @xmath58 where @xmath59 and @xmath60=\int_{x_{\min}}^{x_{\max } } h(x,\eta ) f(\eta\|\mathcal{f}_k)\,d\eta.\ ] ] since the information about the dose toxicity relationship gained from @xmath61 and the response @xmath62 affects the ability to safely and effectively dose the other patients @xmath63 , one potential weakness of these myopic policies is that they may be inadequate in generating information on @xmath9 for treating the rest of the patients , as well as the post - experimental estimate of the mtd for subsequent phases . to incorporate these considerations in aphase i trial , @xmath64 should be chosen sequentially in such a way as to minimize the _ global risk _ @xmath65,\ ] ] in which the expectation is taken over the joint distribution of @xmath66 . note that ( [ 8 ] ) measures the effect of the dose @xmath67 on the @xmath44th patient through @xmath68 , its effect on future patients in the trial through @xmath69 , and its effect on the post - trial estimate @xmath70 through @xmath71 . it can therefore be used to address the dilemma between safe treatment of current patients in the study and efficient experimentation to gather information about @xmath46 for future patients . as noted in section [ sec1 ] , lai and robbins ( @xcite ) have introduced a similar global risk function to address the dilemma between information and control in the choice of @xmath67 in the linear regression model @xmath28 so that the outputs @xmath72 , @xmath73 , are as close as possible to some target value @xmath30 . specifically , they consider ( [ 8 ] ) with @xmath74 and @xmath75 . dynamic programming is a standard approach to a stochastic optimization problem of the form ( [ 8 ] ) . define @xmath76 , \qquad 0\leq k < n-1 , \cr e[h(x,\eta ) \cr { } \quad + g(\hat{\eta}(x_1,\ldots , x_{n-1},x),\eta ) \|\mathcal{f}_{n-1 } ] , \cr \qquad k = n-1 . } \hspace{-20pt } % \ ] ] to minimize ( [ 8 ] ) , dynamic programming solves for the optimal design @xmath77 by backward induction that determines @xmath78 by minimizing @xmath79\ ] ] after determining the future dose levels @xmath80 . note that ( [ 10 ] ) involves computing the conditional expectation of @xmath81 given the dose @xmath2 at stage @xmath44 and the information set @xmath82 , and that @xmath78 is determined by minimizing such conditional expectation over all @xmath2 . for @xmath83 , since @xmath84 is a complicated nonlinear function of the past observations and of @xmath85 that are not yet observed , evaluation of the aforementioned conditional expectation is a formidable task . to overcome this difficulty , we use recent advances in approximate dynamic programming , which we first review and then extend and modify for the problem of minimizing the global risk ( [ 8 ] ) . to begin with , consider the problem of minimizing ( [ 8 ] ) with @xmath74 and @xmath75 in the linear regression model @xmath28 with i.i.d . normal errors @xmath86 having mean @xmath87 . assuming a normal prior distribution of @xmath47 , the posterior distribution of @xmath47 given @xmath88 is also bivariate normal with parameters @xmath89 , in which @xmath90 denotes conditional expectation given @xmath88 . these conditional moments have explicit recursive formulas ; see section 4 of han , lai and spivakovsky ( @xcite ) . the myopic policy that chooses @xmath2 at stage @xmath91 to minimize @xmath92 $ ] is given explicitly by @xmath93\\[-8pt ] & = & \{y^e_{i-1}(\beta)-e_{i-1}(\alpha\beta)\ } /e_{i-1}(\beta^2).\nonumber\end{aligned}\ ] ] although the myopic policy is suboptimal for the global risk function ( [ 8 ] ) , han , lai and spivakovsky ( @xcite ) use it as a substitute for the intractable @xmath84 for @xmath94 in ( [ 10 ] ) , in which the conditional expectation can then be evaluated by monte carlo simulation . this method is called _ rollout _ in approximate dynamic programming . the idea is to approximate the optimal policy @xmath78 by minimizing ( [ 10 ] ) with @xmath80 replaced by some known _ base policy _ @xmath95 , which ideally is some easily computed policy that is not far from the optimum . specifically , given a base policy @xmath96 , let @xmath97 be the @xmath2 that minimizes @xmath98,\ ] ] and the expectation in the second term in ( [ 12 ] ) is typically evaluated by monte carlo simulation . the policy @xmath99 is called the _ rollout _ of @xmath100 and has been used for stochastic control problems arising in a variety of applications ; see section 2.1 of han , lai and spivakovsky ( @xcite ) . the rollout @xmath101 may itself be used as a base policy , yielding @xmath102 , and , in theory , this process may be repeated an arbitrary number of times , yielding @xmath103 letting @xmath104 $ ] , bayard ( @xcite ) showed that , regardless of the base policy , rolling out @xmath8 times yields the optimal design and that rolling out always improves the base design , that is , that @xmath105\\[-8pt ] & \ge & r\bigl(\hat{\mathbf{x}}^{(n)}\bigr)=r(\mathbf{x}^)\nonumber\end{aligned}\ ] ] for any policy @xmath100 , where @xmath106 denotes the optimal policy . for the global risk function ( [ 8 ] ) associated with phase i designs , with @xmath107 given by ( [ 7 ] ) , one can use the myopic design ewoc or crm as the base design in the rollout procedure . in contrast with the explicit formula ( [ 11 ] ) for the case of a linear regression model with normal errors @xmath108 , the posterior distribution with density function ( [ 5 ] ) does not have finite - dimensional sufficient statistics and the myopic design involves ( a ) bivariate numerical integration to evaluate @xmath109\ ] ] for @xmath110 , and ( b ) minimization of the conditional expectation over @xmath2 . the simulation studies in [ sec:2stage ] , in which the rollout is implemented with ewoc as the base design , show substantial improvements of the rollout over ewoc and crm . although ( [ 13 ] ) says that rolling out a base design can improve it and rolling out @xmath8 times yields the dynamic programming solution , in practice , it is difficult to use a rollout ( which is defined by a backward induction algorithm that involves monte carlo simulations followed by numerical optimization at every stage ) as the base policy for another rollout . to overcome this difficulty , we need a tractable representation of successive rollouts , which we develop by using other ideas from approximate dynamic programming ( adp ) . the conditional expectation in ( [ 10 ] ) , as a function of @xmath2 , is called the _ cost - to - go function _ in dynamic programming . an adp method , which grew out of the machine learning ( or , more specifically , reinforcement learning ) literature , is based on two statistical concepts concerning the conditional expectation . first , for given @xmath2 and the past information @xmath82 , the conditional expectation is an expectation and therefore can be evaluated by monte carlo simulations , if one knows how @xmath111 are generated . the second concept is that , by ( [ 9 ] ) , @xmath112 is a conditional expectation given @xmath113 , which is a regression function ( or minimum - variance prediction ) of @xmath112 , with regressors ( or predictors ) generated from @xmath114 . based on a large sample ( generated by monte carlo ) , the regression function can be estimated by least squares using basis function approximations , as is typically done in nonparametric regression . combining least squares ( ls ) regression with monte carlo ( mc ) simulations yields the following ls - mc method for markov decision problems in reinforcement learning . let @xmath115 be a markov chain whose transition probabilities from state @xmath116 to @xmath117 depend on the action @xmath118 at time @xmath119 , and let @xmath120 denote the cost function at time @xmath119 , incurred when the state is @xmath121 and the is taken . consider the statistical decision problem of choosing @xmath2 at each stage @xmath44 to minimize the cost - to - go function @xmath122 assuming that @xmath123 have been determined . let @xmath124 these functions can be evaluated by the backward induction algorithm of dynamic programming:@xmath125 , and for @xmath126 , @xmath127\\[-8pt ] & & { } \hspace{19pt}+e[v_{k+1}(s_{k+1})\|s_k = s , x_k = x]\},\nonumber\end{aligned}\ ] ] in which the minimizer yields @xmath78 . the ls - mc method uses basis functions @xmath128 , @xmath129 , to approximate @xmath130 by @xmath131 , and uses this approximation together with @xmath132 monte carlo simulations to approximate @xmath133\ ] ] for every @xmath2 in a grid of representative values . this yields an approximation @xmath134 to @xmath135 and also @xmath136 to @xmath78 . moreover , using the sample @xmath137 generated by the control action @xmath136 , we can perform least squares regression of @xmath138 on @xmath139 to approximate @xmath134 by @xmath140 . further details of this approach can be found in chapter 6 of bertsekas ( @xcite ) . although the problem ( [ 10 ] ) can be viewed as a markov decision problem with the @xmath141-posterior distribution being the state @xmath116 , the state space of the markov chain at hand is infinite - dimensional , consisting of all bivariate posterior distributions of the unknown parameter vector @xmath47 . if the state space were finite - dimensional , for example , @xmath142 , then one could approximate the value functions ( [ eq : argmin ] ) by commonly used basis functions in nonparametric regression , such as regression splines and their tensor products ; see hastie , tibshirani and friedman ( @xcite ) . however , in the infinite - dimensional case , there is no such simple choice of basis functions of posterior distributions , which are the states . as pointed out in section 6.7 of bertsekas ( @xcite ) , an alternative to approximating the value functions @xmath135 , called _ approximation in value space _ , is to approximate the optimal policy by a parametric family of policies so that the total cost can be optimized over the parameter vector . this approach is called _ approximation in policy space _ and most of its literature has focused on finite - state markov decision problems and gradient - type optimization methods that approximate the derivatives of the costs , as functions of the parameter vector , by simulation . we now describe a new method for approximation in policy space , which uses iterated rollouts to optimize the parameters in a suitably chosen parametric family of policies . the choice of the family of policies should involve domain knowledge and reflect the kind of policies that one would like to use for the actual application . one would therefore start with a set of real - valued basis functions of the state @xmath116 of the markov chain with general , possibly infinitely - dimensional , state space , on which the family of chosen policies will be based . the control policies in this family can be represented by @xmath143 , which is the action taken at time @xmath119 [ after @xmath116 has been observed and the basis functions @xmath144 have been evaluated ] and in which @xmath145 is a parameter to be chosen iteratively by using successive rollouts , with @xmath146 being the base policy for the rollout @xmath147 . using the simulated sample @xmath148 in which @xmath149 denotes the @xmath150th simulated replicate of @xmath151 , least squares regression of @xmath152 on @xmath153 is performed to estimate @xmath145 by @xmath154 ; nonlinear least squares is used if @xmath155 is nonlinear in @xmath145 . in view of ( [ 13 ] ) , each iteration is expected to provide improvements over the preceding one . a concrete example of this method in a prototypical phase i setting is given in the next section , where linear regression splines are used in iterated rollouts . in this setting the state variable @xmath116 represents the complete treatment history up to time @xmath119 in the trial all prior distributions , doses and responses up to that time and the cost function @xmath156 will be replaced by @xmath157 given by ( [ 9 ] ) . in their use of rollouts to approximate the optimum for ( [ 8 ] ) for the normal model , han , lai and spivakovsky ( @xcite ) , section 3 , used the structure of their problem to come up with an ingenious `` perturbation of the myopic rule '' as a base policy to improve the performance of the rollout , without performing second- or higher - order rollouts . in this section we explore this technique in the context of phase i designs , using such perturbations called here _ hybrid designs_both as base policies and as a way to represent highly complicated but efficient policies in a simple , clinically useful way . as pointed out in section [ sec : globrisk ] , the objective function of the dynamic programming problem ( [ 8 ] ) involves both experimentation ( for estimating the mtd ) and treatment ( for the patients in the study ) . consider the @xmath44th patient in a trial of length @xmath8 ( @xmath158 ) . if the @xmath44th patient were the last patient to be treated in the trial ( @xmath159 ) , the best dose to give him / her would be the myopic dose @xmath160 that minimizes @xmath161 , given by ( [ 9 ] ) . on the other hand , early on in the trial , especially if @xmath162 is relatively large , one expects the optimal dose to be perturbed from @xmath160 in the direction of a dose that provides more information about the dose response model , for the relatively large number of doses that will have to be set for the future patients . since the optimal design theory for learning the mtd under overdose constraints , developed by haines , perevozskaya and rosenberger ( @xcite ) , yields a @xmath0- or @xmath27-optimal design @xmath163 , we propose to use the following _ hybrid design _ representation of the optimal dose sequence : @xmath164 where @xmath163 is the chosen `` learning design . '' of course , any dosing policy admits the representation ( [ 14 ] ) with @xmath165 however , we will show that it is possible to use rollouts to choose @xmath166 of a simple form , not depending on @xmath78 , such that the resulting hybrid design given by the right - hand side of ( [ 14 ] ) is highly efficient . similar ideas have been used in `` @xmath167-greedy policies '' in reinforcement learning ( sutton and barto , @xcite , page 122 ) . from our simulation studies that include the example in section [ sec : sim ] , we have found that the sequential @xmath0-optimal design ( haines et al . , @xcite , section 5 ) with @xmath0 being the vector @xmath168 works well for learning design @xmath163 in ( [ 14 ] ) , which we now briefly explain . in general , optimal designs such as @xmath0- and @xmath27-optimal can be characterized as optimizing some convex loss function @xmath169 of the information matrix @xmath170 associated with the parameter value @xmath9 and a measure @xmath171 on the space of design points ( see fedorov , @xcite ) . here @xmath172 is interpreted as the asymptotic variance of the mle @xmath173 of @xmath9 . the optimization problem can be generalized to the sequential bayes setting , with prior distribution @xmath37 on @xmath9 , by finding the @xmath171 that minimizes @xmath174\pi(\theta\|\mathcal { f}_{k-1})\,d\theta\ ] ] at the @xmath44th stage , where @xmath175 is the empirical measure of the previous design points . in the case @xmath176 , ( [ eq : infint ] ) is replaced by @xmath177\pi(\theta)\,d\theta$ ] . for a given vector @xmath0 , the @xmath0-optimal design measure @xmath171 minimizes the asymptotic variance of the linear estimator @xmath178 of @xmath179 or , equivalently , @xmath180=c'(i(\theta,\xi ) ) ^{-1}c$ ] . taking the bayesian @xmath0-optimal design with @xmath181 as the learning design @xmath163 in ( [ 14 ] ) gives @xmath182 , hence , this design is optimal , in some sense , for learning about @xmath183 or , equivalently , about the slope @xmath184 of the dose response curve ( [ 1 ] ) at the mtd , for which @xmath4 is @xmath5 or some other prespecified value . this has the following connections to the stochastic optimization problem of lai and robbins ( @xcite ) discussed in section [ sec1 ] and to the rollout procedure of han , lai and spivakovsky ( @xcite ) . for the normal model discussed in section [ sec1 ] and as an asymptotic limiting case of other models , sacks ( @xcite ) showed that the optimal value of the step size ( a user - supplied parameter in the lai robbins procedure affecting its convergence rate ) is proportional to @xmath185 . moreover , han , lai and spivakovsky ( @xcite ) , section 3 , found that in the normal model , perturbations of the myopic policy in the direction of this @xmath0-optimal design provide a base design for a rollout that has comparable performance to that of an `` oracle policy . '' since the treatment versus experimentationdilemma discussed in section [ sec2 ] stems from the uncertainty in the current estimate of the mtd @xmath46 , it is natural to expect that the amount of perturbation from the myopic dose @xmath160 depends on the degree of such uncertainty , using little perturbation when the posterior distribution of @xmath46 is peaked , and much more perturbation when it is spread out . this suggests choosing @xmath166 as a function of the posterior variance @xmath187 , whose reciprocal is called the `` precision '' of @xmath186 in bayesian parlance . following the approach described in section [ sec : lsmc ] , we use functions of @xmath188 as basic features of the posterior distribution of @xmath46 to approximate the @xmath166 in ( [ 14 ] ) . to begin , monte carlo simulations are performed to obtain the rollout @xmath189 of ewoc , yielding a simulated sample @xmath190 , @xmath191 } , where @xmath192 is the @xmath150th simulated replicate of @xmath193 which is essentially the same as ( [ 14 ] ) with @xmath194 replaced by @xmath195 . the basic idea in section [ sec : lsmc ] can be implemented via nonparametric regression of @xmath192 on @xmath149 , yielding the estimated regression function @xmath196 . letting @xmath197 , the hybrid design @xmath198 can then be used as the base policy to form the rollout @xmath199 , and this procedure can be repeated to obtain the iterated rollouts @xmath200 linear regression splines , and their tensor products for multivariate regressors , provide a convenient choice of basis functions ; see section 9.4 of hastie , tibshirani and friedman ( @xcite ) . for the present problem , it suffices to use a truncated linear function @xmath201\\[-8pt ] \eqntext{\mbox{for } s_\le s\le s^,}\end{aligned}\ ] ] where @xmath202 and @xmath203 are the minimum and maximum of the sample values @xmath149 , @xmath191 , and to extend beyond the range @xmath204 $ ] by @xmath205 which agrees with the constraint @xmath206 and ensures that the weight assigned to experimentation does not exceed @xmath207 . a further simplification is to group the data into @xmath208 blocks so that @xmath209 does not vary with @xmath44 within each block , since it is expected that the amount of experimentation for the initial stages depends mostly on the uncertainty about @xmath46 , while for the final stages experimentation would only benefit the post - trial estimate of @xmath46 . 348pt@lccccc@ design&risk&bias&rmse&dlt&od * + ewoc & 0.84 ( 0.01 ) & @xmath2100.20 ( 0.010 ) & 0.31 ( 0.04)&29.8% ( 0.7% ) & 21.9% ( 0.6% ) + + roll & 0.75 ( 0.01 ) & @xmath2100.04 ( 0.009 ) & 0.22 ( 0.03 ) & 33.0% ( 0.7% ) & 31.2% ( 0.7% ) + + hybrid 1 & 0.75 ( 0.02 ) & @xmath2100.14 ( 0.012 ) & 0.29 ( 0.06 ) & 33.5% ( 1.5% ) & 37.5% ( 1.5% ) + + hybrid 2&0.71 ( 0.01 ) & @xmath2100.04 ( 0.005 ) & 0.22 ( 0.04 ) & 31.24% ( 0.9% ) & 27.8% ( 0.9% ) + we illustrate the method in section [ sec:4.1 ] by applying it to the following example , in which @xmath211 and @xmath212 $ ] is transformed to @xmath213 $ ] by location and scale changes . independent uniform priors on @xmath41 $ ] and @xmath213 $ ] are used for @xmath214 and the mtd @xmath46 , respectively ; see ( [ 2 ] ) and ( [ 3 ] ) and the sentence following it . we use @xmath215 and the ewoc loss with @xmath216 in ( [ 7 ] ) , and the squared error loss @xmath217 in ( [ 8 ] ) . since @xmath8 is relatively small , we can assume for simplicity that @xmath218 in ( [ eq : f1 ] ) does not vary with @xmath44 and estimate the common @xmath219 by applying least squares regression to the sample @xmath220 we also simply use ( [ eq : f1 ] ) for all @xmath121 without performing the extrapolation beyond @xmath204 $ ] . rolling out ewoc as the base design and using @xmath221 simulations , the preceding procedure gave @xmath222 . putting @xmath223 in the hybrid design @xmath224 we used @xmath189 as the base policy of a second rollout , for which the preceding procedure yielded @xmath225 . here we used the sequential @xmath0-optimal design with @xmath226'$ ] as the learning design @xmath163 ( see section [ sec : hyb_des ] ) . table [ table : op - char ] contains the operating characteristics , explained below , of ewoc and its rollout , the first hybrid design @xmath189 with @xmath166 given by ( [ eq : epsk ] ) and the second hybrid design @xmath199 in which @xmath227 in ( [ eq : epsk ] ) is replaced by @xmath228 . each result is based on 2000 simulation runs . the values of @xmath48 were generated from the prior distribution given by the joint uniform distribution on @xmath41\times[x_{\min},x_{\max}]$ ] . figure [ fig : rk ] plots the cumulative risk @xmath229 $ ] of the ewoc , rollout and hybrid designs for @xmath230 . the operating characteristics in table [ table : op - char ] are the monte carlo estimates of overall risk @xmath231 , the bias and root mean squared error ( rmse ) of the terminal mtd estimate @xmath232 , the dlt rate @xmath233 and the overdose rate od , which is the expected proportion of patients treated at doses higher than @xmath46 . standard errors are given in parentheses . the first hybrid design , which is an approximation to the rollout design , provides more than 10% improvement in terminal risk @xmath231 over the myopic policy . the second hybrid design provides an additional 5% improvement in the terminal risk @xmath231 , and also smaller values of the dlt and od rates than the rollout design . the monte carlo simulations used to evaluate the operating characteristics and to fit the hybrid designs were performed by using rejection sampling to simulate from the posterior distribution . at each stage , the posterior distribution of @xmath48 is continuous and supported on the compact set @xmath41\times[x_{\min},x_{\max}]$ ] , hence , the joint uniform distribution on @xmath41\times[x_{\min},x_{\max}]$ ] is a natural candidate for the instrumental distribution in rejection sampling ; see also the last paragraph of section [ sec : conc ] . babb et al . ( @xcite ) used ewoc to design a phase i trial to determine the mtd , with @xmath234 , of the antimetabolite 5-fluorouracil ( 5-fu ) for the treatment of solid tumors in the colon , when taken in conjunction with fixed levels of the agents leucovorin ( 20 mg@xmath1m@xmath235 ) and topotecan ( 0.5 mg@xmath1m@xmath235 ) . in this setting , a toxicity is considered a grade 4 hematologic or grade 3 or 4 nonhematologic toxicity within 2 weeks . as mentioned above , ewoc involves specifying pre - trial a set @xmath236 of possible dose values believed to contain the mtd , where @xmath36 is taken as the starting value . based on preliminary studies of 5-fu given in conjunction with topotecan , a dose of @xmath237 mg@xmath1m@xmath235 of 5-fu was believed to be safe when given with 0.5 mg@xmath1m@xmath235 of topotecan . also , a previous trial concluded that the mtd of 5-fu was 425 mg@xmath1m@xmath235 when administered without topotecan , so @xmath238 was taken to be 425 mg@xmath1m@xmath235 since 5-fu has been observed to be more toxic when given with topotecan than alone . the two - parameter logistic model ( [ 1 ] ) was chosen based on previous experience with the agents , and uniform prior distributions over @xmath40 $ ] and@xmath239 $ ] were chosen for the mtd and the probability @xmath240 , respectively . a feasibility bound of @xmath241 was chosen for ewoc and @xmath234 . we compare ewoc and other previous designs with the rollout ( abbreviated by roll ) of ewoc and the hybrid 1 design described in section [ sec : sim ] in this setting with @xmath242 . to give a feel for the computational time required for the ewoc , roll and hybrid 1 designs , on a desktop personal computer with a 2.66 ghz processor , the simulation of a single @xmath242 run of the roll design with the ewoc base design took 49 minutes , whereas the hybrid 1 design took 0.4 seconds and ewoc took 0.12 seconds . the hybrid 1 design is computationally much simpler than roll since it does not perform rollouts of a base design , but rather calculates its dose via ( [ eq : hyb_des ] ) , where @xmath160 is the ewoc dose and @xmath163 is the sequential @xmath0-optimal learning design . so although the interpolation function ( [ eq : epsk ] ) is derived from data gathered by roll during its rollouts as described in [ sec : hyb_des ] , the hybrid 1 design has computational time on the order of ewoc and the learning design @xmath163 , even though the computational time required for roll is large . @lccccc@ & & & & & + roll & 0.81 ( 0.01 ) & @xmath2100.069 ( 0.002 ) & 0.126 ( 0.022 ) & 27.68% ( 1.70% ) & 29.17% ( 1.87% ) + + hybrid 1&0.92 ( 0.03)&@xmath2100.075 ( 0.003)&0.128 ( 0.028)&24.68% ( 0.86%)&23.48% ( 0.68% ) + + ewoc & 1.13 ( 0.01 ) & @xmath2100.076 ( 0.003 ) & 0.138 ( 0.024 ) & 26.17% ( 0.98% ) & 19.69% ( 0.89% ) + + crm & 1.65 ( 0.01 ) & 0.037 ( 0.003 ) & 0.118 ( 0.021)&36.37% ( 1.10% ) & 62.69% ( 1.10% ) + + @xmath0-opt & 1.71 ( 0.01 ) & 0.060 ( 0.003 ) & 0.126 ( 0.022 ) & 23.44% ( 0.95% ) & 12.42% ( 0.74% ) + + @xmath27-opt & 1.96 ( 0.02 ) & @xmath2100.084 ( 0.006 ) & 0.143 ( 0.023 ) & 13.55% ( 0.31% ) & 3.78% ( 0.17% ) + + wu & 1.77 ( 0.04 ) & 0.038 ( 0.009 ) & 0.122 ( 0.045)&23.40% ( 0.54% ) & 40.25% ( 0.77% ) + + sa & 1.52 ( 0.02 ) & 0.063 ( 0.003 ) & 0.131 ( 0.022)&22.39% ( 0.93% ) & 35.56% ( 0.40% ) + + @xmath243 & 1.87 ( 0.01 ) & 0.060 ( 0.003 ) & 0.138 ( 0.024)&17.06% ( 0.84% ) & 0.85% ( 0.21% ) + + @xmath244 & 2.19 ( 0.02 ) & 0.070 ( 0.002 ) & 0.161 ( 0.025)&14.11% ( 0.81% ) & 0.75% ( 0.29% ) + table [ table:2stage ] first lists bayesian designs , followed by non - bayesian designs that include wu s ( @xcite ) design , stochastic approximation ( lai and robbins , @xcite ) and two dose escalation designs . the first , denoted by @xmath243 , uses 10 uniformly - spaced dose levels in @xmath212=[140,425]$ ] . the second uses 20 uniformly - spaced dose levels and is denoted by @xmath244 . besides ewoc and its rollout roll , the bayesian designs include crm , the constrained @xmath27-optimal design ( abbreviated by @xmath27-opt ) of haines et al . ( @xcite ) with constraint @xmath245 and the unconstrained sequential bayesian @xmath0-optimal design ( abbreviated by @xmath0-opt ) with @xmath0 being the vector @xmath246 . the prior density is assumed to be uniform : @xmath247^{-1}\nonumber \\[-8pt]\\[-8pt ] & & { } \cdot1\{(\rho,\eta ) \in[0,q]\times[x_{\min},x_{\max}]\}\nonumber\end{aligned}\ ] ] with @xmath248 , where @xmath249 denotes the indicator of a set @xmath250 . the values of @xmath48 were generated from the prior distribution ( [ eq : unifprior ] ) . the performance of these designs is first evaluated in terms of the global risk ( [ 8 ] ) , in which we use the squared error @xmath251 for the mtd estimate @xmath252 . we then evaluate performance exclusively in terms of the bias and root mean squared error ( rmse ) of @xmath253 without taking into consideration the risk to current patients , noting that the @xmath0- and @xmath27-optimal designs focus on errors of post - trial parameter estimates . finally , since safety of the patients in the trial is the primary concern of traditional 3-plus-3 designs , performance is also evaluated in terms of the dlt rate and the probability of overdose ( i.e. , dose level exceeding the mtd ) . each result in table [ table:2stage ] is based on 2000 simulations . the results in table [ table:2stage ] show that the effects of considering the `` future '' patients is large , with roll and hybrid 1 substantially reducing the global risk from the myopic designs : in the case of roll , about 30% from ewoc , 35% from crm , and more from the 3-plus-3 , @xmath0- and @xmath27-opt , sa and wu designs . although roll has somewhat smaller global risk than hybrid 1 , it is computationally much more expensive , as noted above . the results for the 3-plus-3 designs show that they are highly sensitive to the choice of @xmath254 in ( [ eq : levels ] ) . the @xmath243 design , using 10 uniformly - spaced levels in @xmath212 $ ] , performs perhaps surprisingly well , as it even has smaller risk than @xmath27-opt , which suffers from substantial under - dosing due to its overdose constraint of @xmath245 . this seems to be because the number 10 of dose levels was a fortuitous choice given the parameter values and sample size of this study , allowing the @xmath243 design to escalate to near the mtd in most cases . however , there is often little information about the appropriate number of doses and scale before aphase i cancer trial begins , and when dose levels are chosen over a less fortuitous range or on too fine a scale , as with the @xmath244 design , the majority of doses can end up being administered at levels far below the therapeutic range near the mtd . we emphasize that all of these designs are being evaluated using the ewoc loss function in ( [ 7 ] ) which , in particular for crm , differs from its associated loss function ; using the crm loss function in ( [ 7 ] ) results in crm having smaller global risk than ewoc , but again roll ( with crm as its base design ) yields smaller global risk than both and the relationship between the other designs remains roughly unchanged . in terms of mtd estimation accuracy , crm and wu have the smallest rmse , closely followed by @xmath0-opt and roll ; crm and wu also have the smallest absolute bias . it is interesting to note that the designs which explicitly account for the asymmetric underdose / overdose relationship , that is , roll , ewoc and @xmath27-opt ( through its overdose constraint @xmath255 ) , are negatively biased , while the others all have positive bias . in terms of safety , @xmath27-opt and the @xmath256 designs have the smallest dlt and od rates , but in view of their large risk values , this safety comes at the cost of low doses that are nontherapeutic . crm has high estimation accuracy and moderate risk , but also the largest dlt and od rates because of its symmetric loss function . the remaining designs , roll , ewoc , wu and sa , all have comparable dlt and od rates , but their risk values suggest that the magnitude of the overdoses in wu and sa are larger than ewoc , which , in turn , has larger overdoses than roll . of particular concern in phase i trials is coherence of the design ( cheung , @xcite ) , that is , whether the next patient will be given a higher dose if the current patient experiences a toxicity , and a lower dose if the current patient does not . while a theoretical investigation of the coherence of the roll and hybrid designs is beyond our scope here , as an illustrative example figure [ fig : tree ] lists the first five doses given by ewoc , roll and hybrid 1 in the 5-fu trial setting , assuming a nontoxic response to the first dose of @xmath237 . note that coherence is exhibited by all three designs in this example . @lcccccc@ * design&&risk&bias&rmse&dlt&od * + roll&(a)&1.64 ( 0.02 ) & @xmath2100.031 ( 0.003 ) & 0.142 ( 0.025 ) & 30.32% ( 1.03% ) & 39.38% ( 1.09% ) + & ( b)&1.39 ( 0.02 ) & @xmath2100.025 ( 0.003 ) & 0.145 ( 0.026 ) & 27.41% ( 1.00% ) & 33.39% ( 1.05% ) + + hybrid 1&(a)&1.82 ( 0.05)&@xmath2100.032 ( 0.002)&0.151 ( 0.027)&36.90% ( 1.52%)&42.31% ( 1.56% ) + & ( b)&1.69 ( 0.04)&@xmath2100.027 ( 0.003)&0.131 ( 0.036)&35.70% ( 1.51%)&41.11% ( 1.56% ) + + ewoc & & 2.29 ( 0.02 ) & @xmath2100.034 ( 0.003 ) & 0.155 ( 0.028 ) & 35.33% ( 1.07% ) & 45.98% ( 1.11% ) + + crm & & 3.83 ( 0.02 ) & 0.037 ( 0.004 ) & 0.179 ( 0.032 ) & 44.18% ( 1.11% ) & 65.12% ( 1.07% ) + when one may have concerns about the validity of the bayesian parametric model in this model - based approach , one can readily incorporate the hybrid designs as the second stage of a two - stage design . the first stage of such escalates the doses cautiously by using a modified 3-plus-3 design . for the batches of 3 in the 3-plus-3 design , we propose to combine the nonparametric step - up / down approach with a parametric model - based dose determining scheme , thereby checking the parametric model to be used for model - based escalation in the second stage . this modification of the traditional 3-plus-3 design uses a specified set of dose levels ( [ eq : levels ] ) . set @xmath257 . in the @xmath44th group of 3 patients , 2 patients are treated at the same dose @xmath258 and 1 patient at the ewoc dose @xmath160 , computed given the doses and responses of the previous @xmath259 patients . if no dlt occurs in the group of 3 patients , @xmath260 is increased to @xmath261 . if 1 dlt occurs , @xmath260 stays the same at @xmath258 . otherwise , 2 or 3 dlts have occurred , so the trial is stopped if @xmath262 , and continues with @xmath260 lowered to @xmath263 . ( alternatively , it may be desired to stop when 3 toxicities occur , regardless of what @xmath264 was . ) the ewoc dose @xmath265 is updated when the process is repeated with the next group of 3 patients . this process repeats until a certain fraction of the total number @xmath8 of patients has been treated , provided the trial has not been stopped at the first stage due to excess toxicities . we have found from our simulation studies that switch - over points around @xmath266 or @xmath267 seem to strike a balance between enough time for conservative dose escalation and model checking during the first stage , while leaving enough time for efficient dose escalation in the second stage . the benefit of a first stage of conservative dose escalation occurs when , unlike in table [ table:2stage ] , the prior distribution of the mtd is misspecified . for example , if the true mtd falls in the left tail of the prior distribution of @xmath46 , then the prior information about the mtd is biased upward , which can cause overdoses . in this situation , including an initial stage of modified dose escalation , like the modified 3-plus-3 scheme , provides additional safety by refining the prior to be more accurate when it begins to be used in the second stage . focusing on the crm , ewoc , roll and hybrid 1 designs , table [ table : misspec ] contains the results of a simulation study that considers a situation such as this , where the true mtd is the lower 15th percentile of the mtd s nominal uniform prior distribution on @xmath212 $ ] . that is , the data are generated with @xmath46 fixed at the 15th percentile of @xmath212 $ ] and @xmath268 uniformly distributed over @xmath41 $ ] , with @xmath248 as in table [ table:2stage ] . the nominal prior for @xmath48 used by the bayesian procedures in table [ table : misspec ] is ( [ eq : unifprior ] ) , the same as in table [ table:2stage ] , as are the values of the other parameters . to see the effects of the first stage of more conservative dose escalation , the operating characteristics of roll are recomputed using a first stage of length @xmath269 ; the dose levels ( [ eq : levels ] ) used by the modified 3-plus-3 design are 10 uniformly - spaced levels in @xmath212=[140,425]$ ] . adding this first stage to roll or hybrid 1 substantially reduces the risk , dlt and overdose rates , as shown in table [ table : misspec ] , in which ( a ) refers to the case of @xmath242 dose levels without the modified 3-plus-3 first stage , and ( b ) refers to the two - stage design using a first stage of length @xmath269 consisting of the modified 3-plus-3 design . despite their shortcomings and the development of alternative bayesian approaches since 1990 , conventional dose - escalation designs are still widely used in phase i cancer trials because of the ethical issue of safe treatment of patients currently in the trial . however , a phase i design also has the goal of determining the mtd for a future phase ii cancer trial , and needs an informative experimental design to meet this goal . von hoff and turner ( @xcite ) have documented that the overall response rates in phase i trials are low and that substantial numbers of patients are treated at doses that are retrospectively found to be nontherapeutic . eisenhauer et al . ( @xcite ) , page 685 , have pointed out that `` with a plethora of molecularly defined antitumor targets and an increasingly clear description of tumor biology , there are now more antitumor candidate therapies requiring phase i study than ever , '' and that `` unless more efficient approaches are undertaken , phase i trials may be a rate - limiting step in the process of evaluation of novel anticancer agents . '' the hybrid designs in the previous section were motivated by developing one such `` more efficient '' approach . hybrid designs with simple interpolation functions , refined through iterated rollouts and regression , can be implemented by using simple look - up tables for the parameters in ( [ eq : f1 ] ) , and thus can be relatively simple to use for clinicians . given computer packages to compute the standard myopic and learning designs , practitioners can use a look - up table for the values @xmath166 in ( [ 14 ] ) as a function of the relative posterior standard deviation @xmath270 . for given values of the prior parameters @xmath271 and @xmath24 , a computer package can generate this look - up table , which can be used at every stage of the trial . we are in the process of developing open source software for this purpose . tighiouart , rogatko and babb ( @xcite ) have shown how markov chain monte carlo ( mcmc ) can be used to compute the posterior distribution of @xmath48 when the prior distribution is supported on @xmath41\times[x_{\min},\infty)$ ] , extending the model considered abovewhere the support of @xmath46 is bounded above by @xmath238 . they note that priors in this class with a negative correlation structure between @xmath268 and @xmath46 result in an ewoc design with comparable accuracy for estimating the mtd but lower dlt and od rates , relative to its performance for priors supported on @xmath41\times[x_{\min},x_{\max}]$ ] . as noted in section [ sec:2stage ] , a two - stage design can easily address the higher dlt and od rates caused by misspecifications of such priors . on the other hand , even without a cautious first stage , the above and other generalizations of the prior of @xmath48 can be seamlessly incorporated into our hybrid design . in fact , the model @xmath272 of tighiouart et al . ( @xcite ) , which has been shown to perform well in their simulation studies , has a left - truncated , hierarchical normal prior distribution on @xmath46 , so the rejection sampling approach in the last paragraph of section [ sec : sim ] can be applied here by using , say , the exponential distribution as the instrumental distribution , since its tails are upper bounds of those of the normal distribution . we can therefore still use the monte carlo approach laid out at the end of section [ sec : sim ] . bartroff s work was supported by nsf grant dms-0907241 and lai s work was supported by nsf grant dms-0805879 .	optimal design of a phase i cancer trial can be formulated as a stochastic optimization problem . by making use of recent advances in approximate dynamic programming to tackle the problem , we develop an approximation of the bayesian optimal design . the resulting design is a convex combination of a `` treatment '' design , such as babb et al.s ( @xcite ) escalation with overdose control , and a `` learning '' design , such as haines et al.s ( @xcite ) @xmath0-optimal design , thus directly addressing the treatment versus experimentation dilemma inherent in phase i trials and providing a simple and intuitive design for clinical use . computational details are given and the proposed design is compared to existing designs in a simulation study . the design can also be readily modified to include a first stage that cautiously escalates doses similarly to traditional nonparametric step - up / down schemes , while validating the bayesian parametric model for the efficient model - based design in the second stage .
let us consider a planar polynomial vector field : @xmath2 recall that vector fields @xmath3 and @xmath4 are called _ orbitally topologically equivalent _ if there is a homomorphism @xmath5 of the phase space such that @xmath5 takes any trajectory of @xmath3 to a trajectory of @xmath4 and @xmath5 preserves the natural orientation of these trajectories ( see @xcite and @xcite ) . the phase space of a vector field below is the plane , or a closed subspace of the plane such that its boundary consists of limit cycles of the field ( see definition [ pfield ] ) . it follows from bezout s theorem that a singular point set of the polynomial vector field @xmath6 is either finite or contains an algebraic curve . in the first case this set consists of at most @xmath7 points . in the second case the equation of the algebraic curve is given by @xmath8 where @xmath9 is the greatest common divisor of @xmath10 and @xmath11 . we are mainly interested in the first case . notice that we can , in some sense , reduce the second case to the first one , dividing @xmath10 and @xmath11 by @xmath9 . [ asterisk]a planar polynomial vector field is called _ @xmath12-field _ if it has only a finite number of singular points . one of the questions in the second part of the 16th hilbert problem is the following : is it true that for any @xmath0 there is a constant @xmath13 such that every polynomial vector field of degree at most @xmath0 has at most @xmath13 limit cycles ? the answer is not known even for @xmath14 . our goal is to find some upper and lower bounds for the number of orbital topological equivalence classes of degree @xmath0 planar polynomial vector fields . the hilbert problem is an obstacle to obtaining such an estimate because orbitally topologically equivalent vector fields have the same number of limit cycles ( see corollary [ hilbert ] ) . to avoid this difficulty we shall introduce the concept of _ equivalence modulo limit cycles _ ( see definition [ modulo ] below ) . it is possible to obtain both upper and lower bounds for this type of equivalence . the aim of this paper is to obtain the upper bound . the lower bound has been found in @xcite . [ nest]a _ nest _ of a vector field @xmath6 ( see figure [ fnest ] ) is an open subset @xmath15 of its phase space such that 1 . @xmath15 is homeomorphic to an annulus ; 2 . the boundary curves of @xmath15 are limit cycles of the field ; 3 . @xmath15 contains no singular points of the field . [ modulo]let us consider vector fields @xmath3 and @xmath4 . let @xmath16 ( @xmath17 ) be the union of all the nests of @xmath3 ( @xmath4 ) . the fields @xmath3 and @xmath4 are called _ equivalent modulo limit cycles _ if the restriction of @xmath3 to @xmath18 is orbitally topologically equivalent to the restriction of @xmath4 to @xmath19 . [ main1 ] denote by @xmath20 the number of orbital topological equivalence types of planar polynomial vector fields @xmath6 such that 1 ) @xmath6 has finite number of singular points ; 2 ) @xmath6 has at most @xmath21 limit cycles and 3 ) @xmath22 . then @xmath23 where @xmath24 . [ hilbert ] for every @xmath25 the following statements are equivalent 1 . the number of orbital topological equivalence classes of planar polynomial vector fields with finite number of singular points and degree less than or equal to @xmath0 is finite . 2 . there is @xmath13 such that every planar polynomial vector field of degree less than or equal to @xmath0 has at most @xmath13 limit cycles . this is obvious from theorem [ main1 ] . we can restrict ourselves to @xmath12-fields . indeed , if we reduce components of a vector field by the common divisor , the number of limit cycles can only increase ( see the remark before definition [ asterisk ] ) . suppose that there are exactly @xmath28 orbital topological classes of @xmath12-fields of degree less than or equal to @xmath0 . take one representative @xmath29 for each class ( @xmath30 ) . let @xmath31 be the number of limit cycles of @xmath29 . by the finiteness theorem ( see @xcite , @xcite and also @xcite ) @xmath32 . thus @xmath33 [ main2 ] consider planar polynomial vector fields with finite number of singular points and degree at most @xmath0 . denote by @xmath34 the number of equivalence classes modulo limit cycles of such fields . then @xmath35 where @xmath36 , @xmath37 . both theorems above follow from a general theorem below . [ pfield]a closed set @xmath38 is called _ admissible _ for a @xmath12-field @xmath6 if its boundary is a union of some limit cycles of @xmath6 . the restriction @xmath39 of the @xmath12-field to an admissible set @xmath40 is called a _ @xmath10-field_. all the limit cycles of @xmath6 that are components of @xmath41 are considered limit cycles of @xmath39 . the _ degree _ of @xmath39 is the degree of @xmath6 . [ high ] consider @xmath10-fields of degree at most @xmath0 with at most @xmath21 limit cycles . there are at most @xmath42 orbital topological equivalence classes of such fields , where @xmath24 . notice that @xmath43 is an admissible set for any @xmath12-field . the lower bound in the theorem is proved in @xcite . the explicit constant @xmath37 has not been written out but the calculation is straightforward . take pairwise non - equivalent modulo limit cycles @xmath12-fields @xmath44 with @xmath45 for @xmath46 . let @xmath47 be the union of all the nests of @xmath29 . the set @xmath48 is admissible for @xmath29 , since @xmath47 is the disjoint union of all the maximal nests of @xmath29 . let @xmath49 be the restriction of @xmath29 to @xmath48 . by definition [ modulo ] the fields @xmath49 are pairwise orbitally topologically non - equivalent . [ cyklicity ] sum of the number of maximal nests of a @xmath12-field and the number of its limit cycles that do not belong to any nest is less than or equal to the number of singular points of this @xmath12-field . it follows from this lemma and from bezout s theorem that the number of limit cycles of @xmath49 is at most @xmath50 ( note that every maximal nest gives two boundary limit cycles ) . now we can apply theorem [ high ] : @xmath51 where @xmath24 . hence , @xmath52 . choose a limit cycle in every maximal nest . call these limit cycles and the limit cycles that do not belong to the nests _ labelled_. our goal is to assign a singular point to every labelled limit cycle . take any labelled cycle @xmath53 . let @xmath54 be all the labelled cycles , satisfying two conditions : 1 . @xmath55 is inside @xmath53 . 2 . there is no labelled cycle @xmath56 such that @xmath56 is inside @xmath53 and @xmath55 is inside @xmath56 . let @xmath57 be the domain bounded by @xmath53 , let @xmath47 be the domain bounded by @xmath55 . consider @xmath58 we claim that @xmath59 contains at least one singular point . indeed , if @xmath60 , then it follows from the fact that euler characteristic of @xmath59 is not equal to 0 ( see figure [ lcycles ] ) . suppose @xmath61 and there are no singular points in @xmath59 . then @xmath59 is a nest , so @xmath53 and @xmath55 are in the same maximal nest . we come to contradiction . assign to @xmath53 any singular point inside @xmath59 . clearly , we have assigned different singular points to different labelled cycles . the rest of the paper is devoted to the proof of theorem [ high ] . the paper is organized as follows : in section [ singularpoints ] we recall the topological classification of singular points , define _ complexity _ of a singular point , and estimate the sum of complexities of all the singular points of a @xmath10-field , using the method , which we have learnt from @xcite . in section [ structure ] we assign a graph on the sphere to a @xmath10-field . it is proved in appendix , that this graph is a complete invariant of orbital topological classification of @xmath10-fields . the proof is essentially using theorems 75 and 76 of @xcite ( see also @xcite and @xcite ) . in section [ combinatorics ] we estimate the number of graphs on the sphere , using the main result of @xcite . the problem has been posed by a. g. khovanskii and yu . s. ilyashenko , i want to express my gratitude to them . without the help of yu . s. ilyashenko this paper would be unlikely completed . i would like to thank g. iyer for correcting my english . i am grateful to the department of mechanics and mathematics of the moscow state university , the mathematical college of the independent university of moscow , and the mathematical department of the university of chicago for their hospitality . a @xmath10-field can not be flat at any point . recall ( see @xcite , chapter 5 , 3 ) that every non - flat singular point @xmath62 of a smooth planar vector field is either monodromic or has a characteristic trajectory ( i.e. a trajectory that tends to @xmath62 as @xmath63 or @xmath64 , being tangent to some line at @xmath62 ) . topological type of monodromic singular point is determined by its poincare map . this map can not have infinite number of isolated fixed points due to the finiteness theorem , since these fixed points correspond to limit cycles ( see @xcite , @xcite and also @xcite ) . thus every monodromic singular point of a @xmath10-field is either a focus , or a center . a small neighbourhood of a characteristic singular point @xmath62 splits into the union of standard sectors . there are three types of standard sectors : hyperbolic , elliptic , and parabolic ( see @xcite , chapter 5 , 3 and figure [ sectors ] ) . this splitting is not canonical : if we shrink the neighbourhood , then some parts of the elliptic sector will become parabolic sectors . nevertheless definition [ complexity ] and definition [ separatrix ] are `` invariant '' . we want to compactify the phase space by adding an infinite point to the plane . to get a chart in the neighbourhood of this point we identify the plane with the 2-sphere , punctured at the north pole , by stereographic projection . the projection from the south pole gives a chart in the neighbourhood of the infinite point . the transition between these charts is given by @xmath65 thus in the second chart our vector field has the form @xmath66 where @xmath3 is a polynomial vector field . set @xmath67 the fields of lines , corresponding to @xmath6 and @xmath68 , agree on @xmath69 . thus they can be patched together to the field @xmath70 on the sphere . the infinite point is a singular point of @xmath70 , thus the trajectories of @xmath70 are those of @xmath6 plus the infinite point . this is why we need an extra @xmath71 factor in ( [ infinity ] ) ( otherwise we could have a trajectory of @xmath70 corresponding to two trajectories of @xmath6 ) . the following convention will be taken : the infinite point is considered a singular point of @xmath6 . [ complexity]_complexity _ of a characteristic singular point is the number of its hyperbolic and elliptic sectors . complexity of a monodromic singular point is zero . we are going to estimate the total complexity ( i.e. the sum of complexities ) of singular points . first we estimate the total complexity of all the singular points except the point at infinity . [ compl ] the total complexity of finite singular points of a degree @xmath0 @xmath10-field is at most @xmath72 . consider a @xmath10-field of degree @xmath73 @xmath2 set @xmath74 . then @xmath75 is the equation for the set of singular points of @xmath6 . [ curve ] let @xmath59 be any open bounded set of the plane , containing all the singular points of @xmath6 . let @xmath76 be the curve given by the equation @xmath77 . then 1 . for sufficiently small @xmath78 the curve @xmath76 is nonsingular in @xmath59 . 2 . the ovals of this curve tend to singular points of @xmath6 as @xmath79 . 3 . for each singular point there is an oval , enveloping this point . consider the set @xmath80 of points on @xmath76 where @xmath6 is tangent to this curve . this set is given by the system of equations @xmath81 where @xmath82 is the directional derivative in the direction of @xmath6 . these equations are polynomial of degree @xmath83 and @xmath84 respectively . thus @xmath80 is algebraic . let @xmath85 be so small that @xmath76 is nonsingular . clearly , if an oval of @xmath76 contains infinitely many points of @xmath80 , then this oval is contained in @xmath80 . in this case the oval is a cycle of @xmath6 . thus @xmath80 consists of isolated points and cycles of @xmath6 . now bezout s theorem tells us that @xmath80 has at most @xmath72 isolated points . it is an easy consequence of lemma [ curve ] that in every hyperbolic or elliptic sector of a singular point of @xmath6 there is a point of @xmath76 where @xmath6 is tangent to this curve ; for @xmath85 small enough these points are distinct and isolated ( the latter follows from the fact that a characteristic singular point has a neighbourhood without cycles ) . thus the overall number of elliptic and hyperbolic sectors is bounded by @xmath72 . the singular set of the curve @xmath76 is its intersection with the set , given by the equation @xmath86 . we claim that every singular point @xmath62 has a punctured neighbourhood without points where @xmath86 . if not , then there is a smooth curve in this set such that @xmath62 is in its closure , since every algebraic set is a union of finite number of smooth strata . since @xmath87 does not change along this curve , the curve consists of singular points of @xmath6 but this is impossible . thus there is a punctured neighbourhood of @xmath62 without points of the set @xmath86 , and 1 ) is proved . let @xmath88 be all the singular points of @xmath6 . choose pairwise disjoint open neighbourhoods @xmath89 such that @xmath90 . set @xmath91 clearly @xmath92 . take any @xmath93 . we have @xmath94 this proves 2 ) . it remains to show that in @xmath89 there is an oval , enveloping @xmath95 . but if there is no such an oval , then we can join @xmath95 with some point of @xmath96 by a curve that does not intersect @xmath76 , this contradicts continuity of @xmath87 . [ compl2 ] the complexity of the infinite point is at most @xmath97 . it is similar to the proof of proposition [ compl ] . the only difference is that we should use the curve @xmath98 with @xmath99 large enough instead of the curve @xmath100 . we shall assign an equipped oriented graph on the sphere ( possibly disconnected ) to every @xmath10-field . this graph can have loops and multiple edges . this graph will be a complete invariant of the orbital topological type of a @xmath10-field . two graphs are considered equivalent if they are isomorphic as equipped graphs , embedded into the sphere . [ separatrix]a _ separatrix _ of a @xmath10-field is a boundary trajectory of a hyperbolic sector of a characteristic singular point . we want to assign an edge of the graph to every separatrix of the @xmath10-field . the problem is that the @xmath101-limit set or the @xmath102-limit set of the separatrix can consist of more than one point . also , we want to have the graph @xmath103-smooth . thus we first introduce the notion of _ small graph_. let @xmath104 or @xmath105 . an _ @xmath101-germ _ at @xmath106 is an equivalence class of maps @xmath107 , where two maps @xmath108 and @xmath109 are equivalent if their restrictions onto @xmath110 coincide for some @xmath111 . @xmath102-germ _ is defined similarly . an @xmath101-germ at @xmath112 is called _ @xmath103-smooth _ if some ( and then any ) of its representatives can be extended to a @xmath103-smooth map @xmath113 for some @xmath78 . for an @xmath101-germ at @xmath114 we consider a representative @xmath115 . the @xmath101-germ is @xmath103-smooth if the map @xmath116 can be extended to a @xmath103-smooth map @xmath117 for some @xmath78 . the similar definitions apply to @xmath102-germs . let @xmath118 be a trajectory of a @xmath10-field . assume that @xmath118 is neither a singular point , nor a cycle , then it has a natural parametrization @xmath119 ( possibly @xmath105 , @xmath120 ) . the _ @xmath101-germ _ of the trajectory is the @xmath101-germ of @xmath118 at @xmath106 , the _ @xmath102-germ _ is the @xmath102-germ of @xmath118 at @xmath121 . a separatrix of a @xmath10-field is called _ nice _ if both its @xmath101-germ and @xmath102-germ are @xmath103-smooth . other separatrices are called _ nasty_. [ nice ] the @xmath101-germ of a trajectory @xmath122 is @xmath103-smooth provided this trajectory tends to a characteristic singular point @xmath62 as @xmath64 . the similar statement is valid for @xmath102-germs . by theorem 64 of 20 of @xcite , @xmath122 tends to @xmath62 in a definite direction . choose an affine coordinate system in the neighbourhood of @xmath62 so that the direction is parallel to the @xmath123-axis . suppose the @xmath101-germ of @xmath122 is not @xmath103-smooth . _ we claim that there is a direction , not parallel to @xmath123-axis , such that the points on @xmath122 where the tangent line is parallel to this direction accumulate to @xmath62 . indeed , if in no neighbourhood of @xmath62 the projection of @xmath122 to @xmath123-axis is one - to - one , then we can take the direction of @xmath124-axis . otherwise in some neighbourhood of @xmath62 the curve @xmath122 is given by an equation @xmath125 , @xmath126 . we have @xmath127 since @xmath122 is tangent to @xmath123-axis . since the @xmath101-germ of @xmath122 is not @xmath103-smooth , @xmath128 . indeed , otherwise we can extend @xmath129 to a @xmath103-smooth function , declaring @xmath130 for @xmath131 . now it follows from ( [ deriv ] ) and the mean value theorem that @xmath132 does not exist . let @xmath133 be any nonzero number between the lower and upper limits of @xmath134 . we can take direction of the line @xmath135 . _ the set of points in the phase space where the vector field is parallel to this direction is an algebraic set . thus its intersection with small enough neighbourhood of @xmath62 is the union of smooth curves @xmath136, ,@xmath137 , where @xmath138 connects @xmath62 with some point @xmath95 . it is enough to show that for all @xmath139 the intersection points of @xmath138 with @xmath122 can not accumulate to @xmath62 . this is clear if @xmath138 is not tangent to @xmath122 at @xmath62 . otherwise we shall use a version of rolle khovanskii method . assume that the points of intersection of @xmath122 with @xmath138 accumulate to @xmath62 . since @xmath122 and @xmath138 are analytic these points can accumulate only to @xmath62 . thus we can enumerate them in the order they appear on @xmath122 : @xmath140 where @xmath141 as @xmath142 . if the field is tangent to @xmath138 everywhere , then @xmath138 is part of a trajectory , and the claim is easy . otherwise , shrinking the neighbourhood of @xmath62 we can assume that the field is transversal to @xmath138 ( except at @xmath62 ) . consider any curve @xmath143 , satisfying the following properties : ( 1 ) it connects @xmath62 with @xmath95 ; ( 2 ) it does not intersect @xmath138 ; ( 3 ) it is tangent to @xmath124-axis at @xmath62 . then @xmath138 and @xmath144 bound a domain , denote it by @xmath59 . the part of @xmath122 between @xmath145 and @xmath146 can not be entirely in @xmath59 , since the field is transversal to @xmath138 . thus there is a point of @xmath122 between @xmath145 and @xmath146 , where @xmath122 intersects @xmath144 . we see that if the intersection points of @xmath138 with @xmath122 accumulate to @xmath62 , then so do the intersection points of @xmath144 with @xmath122 . this contradicts property ( 3 ) of @xmath144 . the converse is also true but we do not need this . choose a point @xmath147 on every limit cycle ( including boundary limit cycles , see definition [ pfield ] ) . choose one trajectory in every elliptic sector of every singular point . the _ small graph _ consists of the following elements : vertices : : : the points @xmath147 and the singular points of the @xmath10-field . edges : : : the closures of the chosen trajectories in elliptic sectors , the limit cycles , and the closures of the nice separatrices . this graph is naturally embedded into the sphere . all the edges of the small graph are @xmath103-smooth in this embedding ( for trajectories in elliptic sectors it follows from lemma [ nice ] ) . consider a @xmath103-smooth curve on the 2-sphere . by _ side _ of the curve we mean the choice of co - orientation of this curve . thus every @xmath103-smooth curve has 2 sides . let @xmath136 and @xmath148 be co - oriented separatrices of the vector field ( i.e. one of the two possible co - orientations on each separatrix is chosen ) . we say that @xmath148 is the _ continuation _ of @xmath136 if the following holds : 1 . @xmath136 tends to a singular point @xmath62 as @xmath63 , @xmath148 tends to @xmath62 as @xmath64 ; 2 . @xmath136 and @xmath148 bound a hyperbolic sector of @xmath62 ; this sector is on the positive side of each of the separatrices . a _ limit polycycle _ of a vector field is a cyclically ordered finite set of singular points ( with possible repetitions ) , and a cyclically ordered set of disjoint co - oriented separatrices such that 1 . the time oriented @xmath149-th separatrix connects the @xmath149-th and @xmath150-st singular points ; 2 . the @xmath150-st separatrix is the continuation of the @xmath149-th separatrix . ( compare with @xcite , 3.1 . ) consider a half - interval with the vertex on the limit polycycle ( not at a singular point ) . assume that this half - interval is transversal to the field everywhere and it is on the positive side of the polycycle . the monodromy map @xmath151 is defined in some neighbourhood of the vertex . let @xmath152 be the coordinate on the half - interval ( @xmath153 at the vertex , @xmath154 outside of the vertex ) . the fixed points of @xmath151 correspond to limit cycles , thus there are only finitely many of them by the finiteness theorem ( see @xcite , @xcite and also @xcite ) . thus for all @xmath152 small enough we have @xmath155 , @xmath156 or @xmath157 . in these cases we call the limit polycycle _ @xmath101-limit , @xmath158-limit and @xmath102-limit _ respectively . clearly , an @xmath101-limit ( @xmath102-limit ) polycycle is the @xmath101-limit ( @xmath102-limit ) set for all nearby trajectories that are on its positive side . the converse is also true . [ limitset ] the @xmath101-limit ( @xmath102-limit ) set of any trajectory is either a point , or a limit cycle , or a limit polycycle . consider any infinite limit set @xmath159 such that @xmath159 is not a limit cycle . by theorem 68 of 23 of @xcite the separatrices , constituting @xmath159 , can be cyclically ordered so that for any two consecutive separatrices the next is the continuation of the previous . the continuation is defined differently in @xcite ( see definitions 19 and 20 of 15 ) . we leave to the reader to check that this definition matches our definition . it follows that @xmath159 is a limit polycycle . a _ cycle without contact _ of a vector field is a smooth closed curve transversal to the vector field at every point . it is easy to see ( see @xcite , lemma 2 of 24 and lemma 1 of 27 ) that we can assign a cycle without contact to every @xmath101-limit polycycle , every @xmath102-limit polycycle , every focus , and every side of a limit cycle so that 1 . consider the region bounded by a limit polycycle , a focus , or a limit cycle and the corresponding cycle without contact . this region contains neither singular points , nor limit cycles , nor other chosen cycles without contact . every trajectory , tending to a polycycle , to a focus , or to a limit cycle , intersects the corresponding cycle without contact at exactly one point . every trajectory that intersects a cycle without contact tends to corresponding focus , polycycle , or a limit cycle . 3 . these cycles without contact intersect neither each other nor the elements of the small graph . fix such a system of cycles without contact . [ nasty ] every nasty separatrix intersects exactly one cycle without contact at exactly one point . we can assume that @xmath102-limit set of a nasty separatrix @xmath118 is a characteristic limit point . it follows from lemmas [ nice ] and [ limitset ] that its @xmath101-limit set is a focus , a limit polycycle , or a limit cycle . denote the corresponding cycle without contact by @xmath160 . then @xmath118 intersects @xmath160 at exactly one point . it remains to prove that @xmath118 does not intersect other cycles without contact . indeed , if @xmath118 intersects another cycle @xmath161 , then the corresponding focus , limit cycle , or limit polycycle should coincide with @xmath101-limit set of @xmath118 . thus @xmath160 and @xmath161 correspond to the same @xmath101-limit set . there are only two situations when it happens : ( 1 ) @xmath160 and @xmath161 are the cycles without contact corresponding to the same limit cycle and ( 2 ) @xmath160 and @xmath161 correspond to limit polycycles which coincide as sets but have different co - orientation of trajectories . the first case is impossible , since @xmath118 would have to intersect this limit cycle . in the second case the polycycle is necessarily homeomorphic to the circle ; @xmath118 would have to intersect the polycycle , which is impossible . thus a nasty separatrix is cut by the point of its intersection with the corresponding cycle without contact into two parts . the part that is the boundary trajectory of a hyperbolic sector is called _ truncated separatrix_. the _ large graph _ has all the vertices and edges of the small graph and the following vertices : : : the intersection points of the nasty separatrices with the chosen cycles without contact . edges : : : the truncated nasty separatrices and the segments of the cycles without contact in which they are split by the vertices , oriented counterclockwise ( if the cycle intersects no separatrix , then this cycle is not added to the large graph ) . now we want to equip vertices and edges of the graph with some data . we indicate on a vertex whether it is a repeller , an attractor , a clockwise rotation center , a counterclockwise rotation center , or none of the above ( e.g. a saddle ) . thus there are * 5 * possible equippings of a vertex . on every edge we indicate whether it is a part of a cycle without contact , a trajectory of an elliptic sector , a separatrix , or an edge , corresponding to a limit cycle . consider a limit polycycle @xmath159 . its separatrix @xmath162 has a distinguished side with respect to this polycycle . we mark this side of @xmath162 by the symbol @xmath101 , @xmath102 , or 0 , depending on the type of polycycle . it the side of a separatrix is not marked with @xmath101 , @xmath102 , or @xmath158 ( e.g. the separatrix does not belong to any limit polycycle ) we mark this side with @xmath163 . in the same way we equip the edge , corresponding to a limit cycle . however , only @xmath101 , @xmath102 , and @xmath163 are possible for a side of a limit cycle . indeed , since the poincare map of a limit cycle is analytic there are only two possibilities : this map is either identity , or the set of its fixed points is discrete . it can not be identity , since the cycle is limit , thus 0 is impossible on a limit cycle . the combination @xmath164 is also impossible . the other combinations are possible : for example , @xmath165 means that the limit cycle is a stable boundary limit cycle ( see definition [ pfield ] ) . thus there are * 8 * possible equippings of a limit cycle . for a separatrix all the combinations are possible . if a separatrix marked with anything except @xmath164 , then it is nice and its @xmath101-germ and @xmath102-germ are both boundary trajectories of some hyperbolic sectors by definition of limit polycycle . if a separatrix is marked with @xmath164 , then we shall indicate if its @xmath101-germ and @xmath102-germ are boundary trajectories of hyperbolic sectors . there are * 3 * cases because at least one germ should be a boundary trajectory of a hyperbolic sector by definition of separatrix . notice that for a nasty separatrix the equipping is unnecessary but we shall keep it for simplicity . thus there are @xmath166 possible equippings of a separatrix . it adds up to @xmath167 possible equippings of an edge . [ technical ] suppose that two @xmath10-fields have their equipped graphs isomorphic ( as planar oriented equipped graphs ) . then these @xmath10-fields are orbitally topologically equivalent . the proof of this proposition is a reformulation of the main result of @xcite and is given in the appendix . let @xmath168 be the number of connected graphs on the sphere ( possibly with loops and multiple edges ) with at most @xmath169 edges . then @xmath170 ( two graphs @xmath171 and @xmath172 are considered the same if they are isomorphic as embedded graphs . ) the graph is said to be _ rooted _ if one of its edges is chosen together with its orientation and co - orientation ( the other edges are neither oriented , nor co - oriented ) . let @xmath173 be the number of rooted graphs with @xmath169 edges . by ( 5.1 ) of @xcite @xmath174 a root is an additional structure . thus if we forget the root , then we decrease the number of graphs : @xmath175 there is another approach due to physicists . since every graph can be completed to a triangulation , and the number of subgraphs of a triangulation grows exponentially , it is enough to estimate the number of triangulations . the following integral is the generating function for these numbers ( for graphs of different genera ) : @xmath176 where @xmath177 is the set of @xmath178 hermitian matrices , @xmath179 is the gaussian measure ( see @xcite , 7.1 ) . however , the author has never seen the proof of the required estimate , based on these methods . the _ size _ of a graph is the sum of the number of its edges and the number of its vertices . [ graphnum ] let @xmath180 be the number of oriented possibly disconnected graphs on the sphere ( possibly with loops and multiple edges ) with size at most @xmath169 . then @xmath181 consider a graph @xmath182 with @xmath121 edges and @xmath183 vertices ( @xmath184 ) . suppose that this graph consists of @xmath185 connected components , then @xmath186 . clearly , we can make the graph connected by adding @xmath187 edges . denote the new graph by @xmath188 . then the number of edges of @xmath188 is less than @xmath169 . thus every planar graph of size at most @xmath169 can be obtained from a _ connected _ graph with at most @xmath169 edges by deleting some edges . since there are at most @xmath189 subsets of edges of @xmath188 , there are at most @xmath190 planar graphs with size @xmath169 or less . putting an orientation on every edge , gives @xmath191 factor . thus @xmath192 in section [ structure ] we have assigned an equipped graph to every @xmath10-field . [ size ] the size of the graph , corresponding to a @xmath10-field of degree @xmath0 with @xmath21 limit cycles , is at most @xmath193 . denote by @xmath194 and @xmath195 the total numbers of elliptic and hyperbolic sectors respectively . denote by @xmath196 and @xmath197 the number of nice and nasty separatrices respectively . clearly , @xmath198 . the number of vertices of the small graph is at most @xmath199 ( remember the infinite singular point ! ) , the number of its edges is @xmath200 . the number of additional vertices of the large graph is @xmath197 , the number of additional edges is @xmath201 . thus the size of the graph is at most @xmath202 by propositions [ compl ] and [ compl2 ] @xmath203 ( this is the total complexity of singular points ) is at most @xmath204 ( @xmath97 for the infinite point , @xmath72 for the finite points ) . it remains to substitute the last inequality in ( [ inequality ] ) . now we can finish the proof of theorem [ high ] . the theorem trivially holds for @xmath205 , since all the constant @xmath10-fields are topologically equivalent . thus we can assume @xmath25 . set @xmath206 . using propositions [ graphnum ] and [ size ] , we see that there are at most @xmath207 possible graphs . every vertex of the graph can have one of 5 possible equippings , every edge of the graph can have one of 28 possible equippings , thus there are at most @xmath208 possible equippings . therefore there are at most @xmath209 possible equipped graphs . now we apply proposition [ technical ] , it gives the estimate @xmath209 for the number of @xmath10-fields . further , @xmath210 theorem [ high ] is proved . in @xcite to every vector field on the sphere with finite number of `` singular elements '' a _ scheme _ is assigned . the main theorems of @xcite ( theorems 75 and 76 of 29 ) assert that if two vector fields have the same schemes , then they are orbitally topologically equivalent . unfortunately , the definition of scheme ( definition 33 , 29 ) is distributed all over the book . so our goal is to show that the scheme of a @xmath10-field can be recovered from the large graph of the field . we do this examining subsequently all the elements of the scheme . according to definition 33 of 29 , we need to list all the _ singular elements _ , _ limit continua _ , their _ global schemes _ , and all the pairs of _ conjugate free continua_. we shall recall all the relevant definitions from @xcite . the other issue is that @xcite deals with bounded phase spaces . however , they mention that all the results are valid for a system on the sphere , see 29.5 . we shall have to adjust some definitions of @xcite to this case . equilibrium states * are singular points in our terminology . they are vertices of our graph . notice that we can distinguish between singular points and other vertices of the graph , using equippings of edges adjacent to this vertices ( if there are no such edges , then the vertex is necessarily a singular point ) . thus we can recover the list of singular points from the graph . orbitally unstable paths * a trajectory @xmath118 ( it is called _ path _ in @xcite ) is called _ @xmath102-orbitally stable _ at a point @xmath211 provided that @xmath212 such that every trajectory @xmath213 , passing through @xmath214-neighbourhood of @xmath215 at @xmath216 , remains in @xmath85-neighbourhood of @xmath118 for @xmath217 , _ @xmath101-orbitally stable _ trajectories are defined similarly . a trajectory is called _ orbitally unstable _ if it is not @xmath101-orbitally stable or not @xmath102-orbitally stable at least at one point ( see definitions 1417 , 15 ) . note that in @xcite this definition is applied to bounded semitrajectories only . in order to make it work on the sphere we have to use the spherical metric . we claim that orbitally unstable trajectories are exactly limit cycles and separatrices . indeed , limit cycles are orbitally unstable by theorem 37 of 15 . it is quite clear that the separatrices are orbitally unstable . conversely , assume that @xmath122 is an @xmath102-unstable trajectory that is not a limit cycle . theorem 40 of 15 shows that @xmath102-limit set of @xmath122 consists of a single point . theorem 38 of 15 tells that @xmath122 is a boundary curve of a hyperbolic sector . thus @xmath122 is a separatrix . therefore we can recover the list of all orbitally unstable trajectories , using the equippings of edges . * the remaining 6 types of singular elements deal with the boundary of a region . only so - called _ normal boundaries _ are considered in @xcite ( see 16.2 ) . the normal boundary is one that consists of finite number of arcs without contact and segments of trajectories ( these segments are called _ corner arcs _ ) . these trajectories are not allowed to be separatrices or limit cycles of the field . thus the boundary of a @xmath10-field is not normal . therefore we have to do the following : for a @xmath10-field @xmath6 remove from the phase space the areas , bounded by boundary limit cycles and their corresponding cycles without contact . then we get a new vector field @xmath218 with normal boundary . it is easy to see that @xmath10-fields @xmath3 and @xmath4 are orbitally topologically equivalent if and only if @xmath219 and @xmath220 are equivalent and the directions of rotation on the corresponding boundary cycles of @xmath3 and @xmath4 are the same . thus we shall recover from graph the scheme of @xmath218 instead of that of @xmath6 . the local scheme of a monodromic singular point can be trivially recovered from the graph . the global scheme of a monodromic singular point is read from the corresponding cycle without contact ( see proposition [ corresp ] of this paper ) . according to definition 23 of 19 , a scheme of a characteristic singular point @xmath62 is the list of * 1 ) * all the separatrices of @xmath62 ; * 2 ) * all the separatrices of other singular points that tend to @xmath62 ; * 3 ) * elliptic sectors ; * 4 ) * the cyclic order of the above ( recall that the boundary does not have corner arcs ) . elliptic sectors correspond to the loops of the small graph . they can be distinguished from other loops by their equipping . the separatrices are the other edges , adjacent to @xmath62 . we can distinguish between separatrices of @xmath62 and `` foreign '' separatrices , since we know for each germ of a separatrix whether it is a boundary trajectory of a hyperbolic sector ( see [ equip ] of this paper ) . the cyclic order is specified , since the graph is embedded into the sphere . to comply with the terminology introduced in @xcite , we use the expressions _ @xmath101-limit continuum _ and _ @xmath102-limit continuum _ as the synonyms for @xmath101-limit and @xmath102-limit sets of trajectories . a _ cell _ of a vector field is a connected component of the phase space after removal of all the singular elements . consider the cell filled by closed trajectories . this cell is doubly connected ( see theorem 50 of 16 ) . a connected component of its boundary is called _ @xmath158-limit continuum . _ [ continua ] the infinite @xmath101-limit , @xmath102-limit , and 0-limit continua that are not limit cycles are limit polycycles and vice versa , limit polycycles are limit continua . all the limit polycycles can be recovered from equipped graph . it follows from lemma [ limitset ] of this paper that any infinite @xmath102-limit continuum or @xmath101-limit continuum is a limit cycle or a limit polycycle . it can be proved similarly ( using theorem 70 of 23 ) that a 0-limit continuum is a 0-limit polycycle . conversely , it is clear that @xmath101-limit and @xmath102-limit polycycles are limit continua . consider a 0-limit polycycle @xmath159 . we need to show that it is a 0-limit continuum . consider a half - interval where the monodromy map is defined and the family of cycles , intersecting this half - interval . all these cycles belong to the same cell ; @xmath159 belongs to the boundary of this cell . thus @xmath159 is a part of 0-limit continuum . it follows from theorem 70 of 23 and the uniqueness of continuation of a separatrix that @xmath159 coincide with this 0-limit continuum . it remains to show that the limit polycycles can be recovered from the graph . a trajectory @xmath162 belongs to some limit polycycle if and only if this trajectory is equipped with @xmath158 , @xmath101 , or @xmath102 on at least one of its sides . it remains to show that we can ascertain from the graph whether two separatrices belong to the same limit polycycle . to this end we just need to check whether one separatrix is the continuation of the other ( because a co - oriented separatrix has at most one continuation ) . this is the information we can get from the graph . thus the whole limit polycycle can be recovered from the graph . we know whether a continuum is @xmath101-limit , @xmath102-limit , or @xmath158-limit continuum from the equipping of any of its separatrices . the _ global scheme _ ( see definition 28 , 25 ) of the continuum is the list of all the separatrices , tending to this continuum , with their cyclic order . this is read from the graph by looking at the corresponding cycle without contact , this is possible due to the following proposition . the large graph splits the sphere into the parts , we shall call them _ faces_. there is a natural embedding of the large graph into the phase space . the part of the phase space , corresponding to the face under this embedding , is called _ realization of a face_. consider the case of polycycle ( the other cases are similar ) . the proposition follows from the following claim : suppose @xmath159 and @xmath160 bound a face and this face is on the positive side of @xmath159 . suppose , on the contrary , that @xmath160 corresponds to another limit polycycle , limit cycle , or focus @xmath221 . since @xmath160 has been added to the graph , there is a separatrix @xmath122 such that @xmath122 intersects @xmath160 . then the @xmath101-limit set or the @xmath102-limit set of @xmath122 coincides with @xmath221 . consider the first case ( the second case is similar ) . since @xmath122 is a separatrix , its @xmath102-limit set must be a single point @xmath62 . but the @xmath102-limit set of @xmath122 is contained in the closure of the realization of a face , bounded by @xmath159 and @xmath160 , since @xmath160 is a cycle without contact . thus @xmath62 belongs to @xmath159 ( hence , @xmath159 can not be a limit cycle ) . but this contradicts the assumption that @xmath159 is a limit polycycle and the face is on the positive side of @xmath159 . the boundary ( after removing neighbourhoods of boundary limit cycles ) consists of cycles without contact , labelled by boundary limit cycles ( notice that these cycles without contact may or may not belong to the large graph ) . let @xmath160 be a cycle without contact , corresponding to a boundary limit cycle @xmath159 . since the boundary contains no corner arcs , the global scheme of @xmath160 is the list specifying ( see definition 30 of 26 ) : * 1 ) * whether @xmath160 is an outer or inner boundary curve ; * 2 ) * whether @xmath160 is positive or negative cycle ( i.e. whether @xmath159 is an @xmath101-limit cycle or an @xmath102-limit cycle ) ; * 3 ) * all singular paths , intersecting @xmath160 , enumerated in cyclic order . now * 1 ) * and * 2 ) * are read from the equipping of @xmath159 . the list of all singular paths is recovered from vertices of the large graph ( recall that if @xmath160 does not belong to the large graph , then there are no singular paths , intersecting @xmath160 ) . two 0-limit continua are called _ conjugate _ if they bound a cell , filled with closed trajectories . two cycles without contact are called conjugate if every trajectory that intersects the first one , intersects the second one as well . an @xmath101-limit continuum and an @xmath102-limit continuum are called conjugate if the corresponding cycles without contact are conjugate ( see 28 and 27.4 ) . it remains to show that we can recover all the pairs of conjugate free continua from the graph . ( the continuum is called _ free _ if no separatrix intersects its cycle without contact . the conjugate continua are obviously free . ) again , consider the faces of the large graph . it is enough to prove the following claim : _ limit continua @xmath159 and @xmath221 are conjugate if and only if they bound a face that is on the positive side of each continuum_. the `` only if '' part follows from lemma 3a of 28.2 . conversely , suppose @xmath159 and @xmath221 bound a face , and this face is on the positive side of each of them . if the realization of this face is filled with closed phase curves , then @xmath159 and @xmath221 are conjugate 0-limit continua . otherwise let @xmath160 and @xmath161 be cycles without contact , corresponding to @xmath159 and @xmath221 respectively . we need to show that @xmath160 and @xmath161 are conjugate cycles . on the contrary , assume that @xmath122 is a trajectory , @xmath122 intersects @xmath160 but does not intersect @xmath161 . since @xmath160 is a cycle without contact , @xmath122 intersects it just ones . thus either positive , or negative semitrajectory of @xmath122 belongs entirely to the region bounded by @xmath160 and @xmath161 . but this region contains neither singular points nor limit cycles , this contradicts the poincare bendixon theorem , which tells that @xmath101-limit ( @xmath102-limit ) set of any trajectory contains either a singular point or a limit cycle . andronov a. a. , leontovich e. a. , gordon i. i. , maier a. g. , _ qualitative theory of second - order dynamic systems _ , translated from the russian by d. louvish , halsted press ( a division of john wiley & sons ) , new york toronto , ont . ; israel program for scientific translations , jerusalem london , 1973 , 524 pp . anosov d. v. , aranson s. kh . , arnold v. i. , bronshtein i. u. , grines v. z. , ilyashenko yu . s. , _ ordinary differential equations and smooth dynamical systems _ , translated from the 1985 russian original by e. r. dawson and d. oshea , third printing of the 1988 translation [ _ dynamical systems . i _ ] , encyclopaedia math . , * 1 * , springer , berlin , 1988 ; springer verlag , berlin , 1997 . arnold v. i. , _ dopolnitelnye glavy teorii obyknovennykh differentsialnykh uravnenii _ ( russian ) [ _ supplementary chapters to the theory of ordinary differential equations _ ] , `` nauka '' , moscow , 1978 , 304 pp . calle j. , _ introuduction aux fonctions analysables et preuve constructive de la conjecture de dulac [ introduction to analyzable functions and constructive proof of the dulac conjecture ] _ , actualites mathematiques . [ current mathematical topics ] ( french ) hermann , paris , 1992 . ii+340 pp .	the paper deals with planar polynomial vector fields . we aim to estimate the number of orbital topological equivalence classes for the fields of degree @xmath0 . an evident obstacle for this is the second part of hilbert s 16th problem . to circumvent this obstacle we introduce the notion of _ equivalence modulo limit cycles_. this paper is the continuation of the author s paper in [ mosc . math . j. * 1 * ( 2001 ) , no . 4 ] where the lower bound of the form @xmath1 has been obtained . here we obtain the upper bound of the same form . we also associate an equipped planar graph to every planar polynomial vector field , this graph is a complete invariant for orbital topological classification of such fields .
in the first paper of this series ( i ) , the hypersonic attenuation of highly densified silica glass , @xmath0-sio@xmath1 , was investigated . it was found that in that material the damping of hypersound is completely dominated by its _ anharmonic interaction _ with the thermally dominant modes . in the present paper we consider the corresponding situation in usual vitreous silica , @xmath2-sio@xmath1 , for which a large quantity of high quality data is already available in the literature . in the early 1950 s , ultrasonic absorption peaks in function of the temperature @xmath3 were first observed in @xmath2-sio@xmath1 . these were described by anderson and bmmel in terms of a phenomenological maxwell model @xcite . following the discovery of the key role played by two - level systems ( tls ) in producing the anomalous thermal properties of glasses @xcite , it became clear that these should also be invoked in the description of acoustic relaxation . a theory including both resonant and relaxational sound damping produced by tunneling was developed by jckle @xcite . it was then extended to higher temperatures by including in the description the classical jumps over the energy barrier separating the wells of tls @xcite . the presently accepted model for thermally activated relaxation ( tar ) includes a distribution of asymmetric double - well potentials , as discussed by gilroy and phillips @xcite . the simple extrapolation of this behavior to ghz frequencies , _ i.e. _ to the hypersonic regime , has sometimes been presented as an appropriate description of the observed phenomena , _ e.g. _ in @xcite or @xcite . however , it has also been recognized that tar might not always be sufficient to account for sound dispersion and damping . a specific example has been discussed in the case of vitreous germania , geo@xmath1 @xcite . the results presented in ( i ) strongly suggest that @xmath2-sio@xmath1 should be an excellent candidate to search for a possible anharmonic contribution to the hypersonic attenuation . the paper is organized as follows . in section ii the phenomenology of tar is reviewed to formulate a description that can reasonably be extrapolated to sufficiently high-@xmath3 and to brillouin - scattering frequencies . in section iii this formalism is applied to available high quality sonic and ultrasonic data on @xmath2-sio@xmath1 , covering more than four orders of magnitude in the frequency @xmath4 . the model parameters that are thus extracted are then used in section iv to estimate the tar contributions to both velocity dispersion and damping at brillouin - scattering frequencies . the anharmonic contribution can then be extracted from the total signal . it is analyzed in terms of a mean thermal mode relaxation time @xmath5 in section v. interestingly , this time is found to be about an order of magnitude longer than that of densified silica glass , @xmath0-sio@xmath1 . section vi presents a synthesis of the above analysis for the entire range of @xmath4 , from ultrasonic to brillouin - scattering frequencies . a part of that analysis concerns the velocity changes with @xmath3 . we find an anomalous increase of the bare velocity @xmath6 with @xmath3 , nearly the same in @xmath2-sio@xmath1 and @xmath0-sio@xmath1 . this unrelaxed velocity , @xmath6 , is obtained after subtraction of the velocity changes produced by both thermally activated relaxation and anharmonicity . the @xmath3-dependence of @xmath6 suggests that silica experiences a progressive structural change with increasing @xmath3 . section vii is a discussion , which mentions the crossovers between the various frequency regimes , considers the suitability of power laws for the description of the dependence of the damping on @xmath4 , and evoques possible extension to other glasses . we consider an assembly of defects represented by double - well potentials that are separated by barriers of height @xmath7 and whose depths differ by the asymmetry @xmath8 . the energies @xmath7 and @xmath8 are randomly distributed according to a distribution @xmath9 to be discussed below . the system is thought to hop continuously between the wells . the energy difference between the wells is coupled to the strain @xmath10 of a sound wave of angular frequency @xmath11 by a deformation potential @xmath12 . owing to the delayed energy exchanged in hopping , this produces the dissipation of the sound wave . this situation has been described in great details elsewhere @xcite . to an excellent approximation , it leads to a relaxational contribution to the internal friction given by @xmath13 in this expression , @xmath14 is the material density , @xmath2 is the velocity of sound , and @xmath3 is in energy units . we remark that both @xmath15 and @xmath2 depend on the polarization of the acoustic wave , longitudinal ( la ) or transverse ( ta ) . in this writing , @xmath7 is restricted to positive values while the distribution is symmetric in @xmath8 . we also note that @xmath9 is a density , _ i.e. _ a number per unit volume . the associated velocity change , which follows from the kramers - kronig relation , is given by @xmath16 in these equations , @xmath17 is the relaxation time for hopping within the double well . it is given by @xmath18 where @xmath19 is the inverse of an attempt frequency , as shown in detail _ e.g. _ in @xcite . the key in applying these expressions is to use a reasonable distribution @xmath20 . for small @xmath8 and @xmath7 the distribution is often replaced by a constant @xmath21 . this is suggested by the @xmath3-dependence of the specific heat at low @xmath3 which only probes low values of @xmath8 and @xmath7 @xcite . of course , a constant @xmath21 can not be extended to high values of @xmath8 and @xmath7 as this leads to a diverging integral density of defects which is unphysical . a reasonable guess for @xmath20 can be obtained with the help of the soft - potential model ( spm ) @xcite . that model is characterized by a distribution of random dimensionless cubic and quadratic coefficients , @xmath22 and @xmath23 respectively , by an energy scale of the potential @xmath24 , and by a characteristic cross - over energy @xmath25 @xcite . we are only interested here in the region @xmath26 with @xmath27 which gives double wells with barriers centered at the origin of the soft mode coordinate @xmath28 . as shown in @xcite , owing to the latter choice , the variables @xmath22 and @xmath23 are not statistically independent . this leads to a sea - gull singularity in their distribution , @xmath29 , where @xmath30 is finite near the origin . for the range of values of interest here , one has @xmath31 and @xmath32 . the deformation potential of the spm is also function of @xmath33 with @xmath34 as defined in @xcite and further explained in @xcite . the terms @xmath35 of ( 1@xmath36 ) are transformed into @xmath37 in the spm . using the jacobian @xmath38 , one finds that @xmath39 @xcite . this was already used by keil _ @xcite who selected a distribution @xmath40 proportional to @xmath41 times a gaussian . these authors experimentally found that indeed @xmath42 is very near 1/4 in the case of silica . for convenience , and for lack of different compelling indications , we assume that the distribution can be factored into @xmath43 for @xmath44 , inspired by @xcite and @xcite , we use the normalized form @xmath45 since @xmath46 , this expression is integrable . the norm @xmath47 is selected so that @xmath44 integrates to 1 . this gives @xmath48 the function @xmath49 is the parabolic cylinder function @xcite . as explained in the following section , our independent analysis of a collection of data larger than that used in @xcite also leads to @xmath42 very close to 1/4 . this provides a solid support for this particular choice of @xmath44 . for @xmath50 we use a simple gaussian , rather similar to the gaussian cut - off of the linear asymmetry coefficient @xmath51 used by gil _ indeed , @xmath52 , so that a gaussian in @xmath51 is very close to one in @xmath8 , the power of @xmath7 connecting the two variables being quite small . this functional form for @xmath50 was already employed succesfully by bonnet @xcite . it is written @xmath53 where @xmath54 is defined by the normalization condition that the integral of @xmath55 equals @xmath56 . the norm @xmath57 is then dimensionless and given by @xmath58 where we defined @xmath59 @xmath60 being the cut - off value of the asymmetry . we finally remark that with the above definitions , the integral density of defects is @xmath61 to complete the calculation , we now introduce ( 4 ) and ( 5 ) into ( 1 ) , and use ( 2 ) . in performing the integrals in ( 1 ) , we make the same approximation as in @xcite that sech(@xmath62 ) is replaced by 1 for @xmath63 and by zero otherwise . this eliminates the sech factors in ( 1 ) and ( 2 ) and simply replaces the limits of integration on @xmath8 by @xmath64 . it is convenient to define a dimensionless constant @xmath65 one obtains @xmath66 @xmath67 where @xmath68 is the error function , @xmath69 for the purpose of comparison with literature results , it will be useful to relate the value of @xmath70 with the `` tunneling strength '' defined by @xmath71 in the usual tunneling model @xcite . given the distribution ( 4 ) , one can not strictly define @xmath21 , but this does not prevent defining @xmath72 in a consistent manner @xcite . we calculate this for @xmath73 . we use @xmath74 and @xmath75 from @xcite , where @xmath76 . introducing these definitions into @xmath77 , and using ( 8) , we find in the limit of small @xmath8 and @xmath7 , @xmath78 information on the numerical handling of eqs . ( 9 ) and our choice of suitable fitting parameters are found in appendix a. it is generally agreed that sonic or ultrasonic damping at temperatures above @xmath79 k is dominated by thermally activated relaxation . to obtain the model parameters @xmath80 , @xmath60 , and @xmath81 entering eqs.@xmath82 , it is necessary to analyze acoustic results over a sufficiently large range of frequencies @xmath4 , this up to high ultrasonic frequencies . it implies comparing data from various sources and generally acquired with different measurement techniques . tar leads to a peak in @xmath83 as observed by anderson and bmmel in ultrasonic pulse - echo measurements @xcite . these authors report precise data obtained on the ta - mode of vitreous silica at @xmath84 mhz . this is one of the curves that will be used for our analysis , as displayed in fig.@xmath85 . other high - frequency data were measured using the bragg diffraction of light , @xcite in a set - up similar to the original debye - sears experiment @xcite . damping results at @xmath86 mhz on both la- and ta - modes have been obtained in that manner @xcite . from these , and from results presented in @xcite , it is clear that la- and ta - modes lead to identical @xmath87 peak shapes . this implies that similar distributions of defects are active in the damping of all acoustic waves , independently from their polarizations . we use the data on the la - mode at 207 mhz from @xcite , also shown in fig.@xmath88 . the independence from polarization allows including in the evaluation sonic frequency results acquired on macroscopic vibrational modes , such as in vibrating reed measurements . we use the data at @xmath89 khz of classen , as reported in @xcite . this curve is also displayed in fig.@xmath88 . results at intermediate frequencies were collected using composite oscillators , at 660 khz on the la - mode in @xcite , and at 180 khz on a torsional mode in @xcite . as remarked in @xcite , this particular method can easily lead to instrumental background - loss contributions . this might have been the case in @xcite , as suggested from data on other glasses presented by the same authors in @xcite which show long absorption tails at high @xmath3 . this notion is also supported by fits that are explained below . hence , it is the 180 khz data from @xcite which is included in fig.@xmath85 . with these four curves , the analysed data covers more than four decades in @xmath4 with a nearly linear progression in @xmath90 . -sio@xmath1 taken from the literature , @xcite , @xcite , @xcite , and @xcite in order of decreasing frequency . the solid lines show our adjustment to eq.@xmath91 as explained below.,width=321 ] the curves shown in fig.@xmath85 have been simultaneously adjusted to eq.@xmath91 , allowing for each measurement an independent coupling parameter @xmath92 . in the marquardt - levenberg routine used to this effect , the weights have been adjusted so that each curve had approximately the same weight , independently from the number of data points at each particular frequency . the excellent results of this fit , with @xmath42 fixed at 1/4 , are illustrated by the solid lines in fig.@xmath85 . the distribution parameters corresponding to these lines are @xmath93 k , @xmath94 with @xmath81 in seconds , and @xmath95 . if the same fit is repeated with @xmath42 free , one finds @xmath96 , and similar values for the other parameters : @xmath97 k , @xmath98 , and @xmath99 . this shows that the distribution @xmath44 suggested by the spm is very adequate indeed . it also shows that @xmath42 can be fixed at 1/4 , which will be done for the rest of this paper . the rather high value obtained for @xmath100 emphasizes that one should not neglect the cut - off in the asymmetry distribution . we return to this point below . our analysis reveals a certain difficulty in comparing the absolute size of the various curves . this can be related to some extent to the different polarizations of the waves in these measurements , or also to a small part to the different qualities of the silica samples employed . however , it seems more likely that it arises to a large part from calibration inaccuracies in some of these measurements . indeed , discrepancies in the size of @xmath87 are directly observed in comparing literature reports , for example for the ta - waves at 20 mhz in @xcite and @xcite . the difficulty is also seen in comparing the la to ta values of @xmath87 in @xcite and in @xcite . while in the former case it is the attenuations that have the same amplitudes for both la and ta , in the latter it is the internal frictions which are practically the same for both polarizations . this can not be reconciled as attenuation and friction differ by a factor proportional to the sound velocity , and the latter is @xmath101 m / s for la and @xmath102 m / s for ta - waves at low @xmath3 . we also observed that the results at 660 khz in @xcite are fitted extremely well with the same model parameters , provided one allows for a relatively small constant background contribution to @xmath87 , equal to @xmath103 . the values of @xmath92 obtained for these various measurements range then from @xmath104 in the case of @xcite to @xmath105 for @xcite . these variations in @xmath92 do not seem correlated to either the measuring frequency or the wave polarization , which is another reason to suspect calibration difficulties . -sio@xmath1 taken from the literature , @xcite for 660 khz and @xcite for 6.8 mhz . the solid lines show our calculation of the tar contribution according to eq.@xmath106 with @xmath107 . the data points of the upper two curves show the experiment minus the tar contribution.,width=321 ] it is necessary to obtain a more reliable value of @xmath92 to analyze the brillouin scattering data in the following section . to this effect we remark that velocity measurements are generally both accurate and precise , allowing to follow small changes in @xmath108 with excellent accuracy . we use two measurements of the la - wave velocity , one at 660 khz from @xcite and the other at 6.8 mhz from @xcite . the experimental values @xmath109 are shown in fig.@xmath110 , together with continuous lines marked @xmath111 . these are calculated from the relaxation equation @xmath112 using the above values of @xmath80 , @xmath60 , and @xmath81 , together with @xmath107 . with the latter value , the dip around 50 k is completely accounted for . this is emphasized by the difference @xmath113 also shown in fig.@xmath114 . hence , we adopt for the following the coupling constant @xmath107 which happens to fall within the range of values obtained by fitting the various @xmath87 curves discussed above . we now return to the small cut - off value @xmath115 k which we obtained . this aspect has mostly been ignored by other workers . it was shown in @xcite that it is necessary to include a cut - off to obtain a good fit to the high-@xmath3 tails in fig . 1 . for example , forcing @xmath116 , the quality of the fits degrades appreciably above @xmath117 k , especially for the highest @xmath4 curve . more importantly , the calculated values of @xmath111 are then much too large to account properly for the dip around 60 k observed in fig . 2 . obviously , by the virtue of the kramers - kronig transform , @xmath111 integrates over a large part of the distribution , which is the reason for this problem . one might wish to gain an intuitive picture of why @xmath118 can be so much smaller than @xmath80 . to this effect , one can consider the models of tar drawn in fig . 2 of @xcite . we take as simplest examples model a , in which the two si atoms of a si@xmath119o@xmath119si bond are too close , and model b , in which they are too far apart . in either case , a double well potential for the connecting oxygen results with a barrier height @xmath7 that comes mainly from the separation of the two si . this barrier can thus be quite high . on the other hand , the asymmetry is produced by the difference in the wider environment of the two wells . in a hard glass , one might expect that these environments , which are dictated by the minimization of the energy , can be mostly quite similar . from these considerations , one intuitively anticipates that the ratio @xmath120 might depend significantly on the particular glass . finally , we remark that the temperature @xmath121 of the peak positions in fig.@xmath85 depends linearly on ln@xmath122 . a similar observation was already made in @xcite . this typical arrhenius behavior supports tar as the principal relaxation mechanism to describe fig.@xmath85 , as opposed for example to incoherent tunneling that would lead to @xmath123 as seen from eq . ( 2.95 ) of @xcite . high - resolution brillouin scattering measurements of the temperature dependence of the la - linewidth of vitreous silica have been reported in @xcite and @xcite . both experiments were performed near backscattering and below room @xmath3 , using as exciting radiation the blue argon - laser line at @xmath124 nm . measurements above room @xmath3 are reported in @xcite . these were performed at @xmath125 nm . the brillouin frequency shifts @xmath126 from @xcite and @xcite , the latter rescaled to @xmath124 nm , are displayed in fig.@xmath127 over the entire range of @xmath3 . -sio@xmath1 near backscattering , from @xcite and @xcite . the values of @xcite have been rescaled to the blue argon - laser line , @xmath124 nm . the solid line is a guide to the eye . * b * : the internal friction derived from the brillouin linewidths . the crosses are from @xcite and @xcite , while the dots are from @xcite . the solid curve shows the tar contribution calculated from eq.@xmath91 using the value of @xmath92 determined from the velocities in fig . 2.,width=321 ] the frequency widths extracted from the brillouin peaks in @xcite and @xcite , after correction for the instrumental broadening , have been converted to internal friction as shown in fig.@xmath128 . the internal friction reported by tielbrger _ et al_. in @xcite is also shown on the same figure . one notices the remarkable agreement between these two independent determinations . it emphasizes that brillouin scattering gives an absolute measurement of @xmath87 , independent from calibrations , since it suffices to take the ratio of the brillouin full - width to the brillouin shift to extract @xmath87 . this statement of course assumes that the spectroscopy can be performed with the required resolution , and in particular that the broadening produced by the finite collection aperture can be made sufficiently small . also shown in fig.@xmath128 is the curve @xmath83 calculated with eq.@xmath91 for @xmath129 ghz and with the parameters determined in section iii . it is evident that at brillouin frequencies tar can not account for the entire internal friction . the value @xmath130 is confirmed by the brillouin velocity results explained below . however , even if one would allow oneself to arbitrarily double the coupling constant @xmath92 , the line in fig.@xmath131 would still not superpose the data points . in particular the peak in @xmath83 occurs at a lower @xmath3-value than the peak in the observed @xmath87 . the difference between the two curves is attributed to its largest part to the anharmonic coupling of the la - waves with the thermally excited vibrational modes , as reported in ( i ) for densified silica glass . the velocities are extracted from the brillouin frequencies @xmath126 of fig.@xmath127 using @xmath132 , where @xmath133 is the refractive index and @xmath134 is the internal scattering angle , here close to @xmath135 . to derive precise values on @xmath136 , it is necessary to know @xmath137 with an equal precision . this information has been derived from @xcite . the results below 300 k are shown in fig . 4 , together with the velocities at 6.8 mhz from @xcite . . the upper data points show the same velocitites after subtraction of the tar contribution . we note that after this subtraction there remains a small difference between the slopes observed at high @xmath3 for the two different frequencies . these slopes are shown by the dashed lines and they are discussed in appendix b.,width=321 ] the relative changes reported in @xcite have also been converted back to velocities using @xmath138 m / s . subtracting from all three curves the @xmath139 calculated from eq.@xmath106 , one observes that below 150 k all the data collapse quite well . this confirms that the value of @xmath92 is also correct at brillouin frequencies . it also emphasizes that the @xmath4-dependence in the depth of the dip around 70 k is well predicted by tar . as explained in section vib , the velocity changes that are produced by the anharmonicity , @xmath140 , show little dispersion compared to the large dispersion in @xmath139 . hence , they do not modify the above conclusion . figure 5 shows the difference between the experimental internal friction and the calculated tar contribution . similarly to the results on @xmath0-sio@xmath1 in ( i ) , there is a region where the signal falls rapidly with decreasing @xmath3 . however , it does not seem to fall to zero sufficiently fast . in spite of the large scatter in the experimental values , there is a hint for an additional contribution to @xmath87 at low @xmath3 , in the region from @xmath141 to @xmath142 k. . the crosses are from @xcite while the dots are from @xcite . the solid line is a rough estimate for the small tunneling contribution extrapolated from data presented in @xcite.,width=321 ] at these hypersonic frequencies and temperatures , one can anticipate a contribution @xmath143 arising from the relaxation of two - level systems ( tls ) by incoherent tunneling . generally , @xmath144 exhibits a plateau separating a regime @xmath145 at low @xmath146 from @xmath147 at higher @xmath3 , as shown in @xcite where @xmath148 and @xmath149 are defined . the position , extent , and height of this plateau are @xmath150-dependent , as emphasized in fig . 6 of @xcite . according to @xcite , the extent of the plateau should shrink to zero at sufficiently high frequencies such that @xmath151 , where @xmath152 is a frequency independent characteristic temperature marking the end of the plateau region . in @xmath2-sio@xmath1 , it has been observed that @xmath153 k @xcite . the plateau becomes then a hump slightly above @xmath154 . as shown in @xcite , for @xmath155 mhz one should have @xmath156 k. this value increases with @xmath157 , as confirmed from sound - damping data between 330 and 930 mhz @xcite , showing a shoulder on the low @xmath3 side of the tar peak . hence , at 35 ghz one expects @xmath158 k , giving a broad tls hump centered around 40 k. @xmath159 below this hump , and @xmath160 above it @xcite . extrapolating the observations reported in @xcite we posit @xmath161 as a rough estimate for this relatively small contribution to @xmath87 . this curve , which peaks around 40 k , is shown in fig . the difference @xmath162 will be used below to analyse the anharmonic damping . since there seems to be a small tunneling contribution @xmath143 , then there ought to be by the kramers - kronig relation a corresponding contribution @xmath163 . as shown in eq . ( 2.96 ) of @xcite , this should produce at high-@xmath3 a frequency dependent term in @xmath164 . as shown in appendix b , there are reasons to believe that it is this term that produces the difference in slopes indicated by short dashed lines in fig . 4 . to conclude , we find a relatively large @xmath165 at hypersonic frequencies . it is of the same order of magnitude as @xmath83 . the small additional term @xmath143 that has been discussed above is a correction of minor importance compared to @xmath165 , as obvious from fig . in the spirit of ( i ) we now analyze the anharmonic damping with the expression @xmath166 where @xmath5 is the mean lifetime of the thermal modes and @xmath167 is the prefactor given in eq.@xmath168 of ( i ) , with @xmath169 the specific heat per unit volume and @xmath170 the debye velocity . we remark that all quantities entering @xmath171 are known , except for the mean - square average grneisen parameter @xmath172 . thus we can directly plot the quantity @xmath173 as it is independent from @xmath172 . the result obtained using the values @xmath165 taken from fig . 5 is drawn in fig.@xmath174 . . the solid line is a guide to the eye calculated from the corresponding line in b. the peak is located at @xmath175 ( see appendix c ) and its height gives @xmath176 . * b * : the values of @xmath5 for @xmath2-sio@xmath1 obtained from the data in a after solving eq . data for x - cut crystal quartz and @xmath0-sio@xmath1 from ( i ) are shown for comparison . the lines for @xmath2-sio@xmath1 and @xmath0-sio@xmath1 are guides to the eye which are sums of negative powers of @xmath3 , while the line for @xmath177-sio@xmath1 is calculated from eq . ( 15).,width=321 ] points below 60 k are not shown as the data is too noisy at low @xmath3 to be significant after division by @xmath178 . this plot exhibits a peak for @xmath179 . from the height of this peak one reads @xmath180 . with this determination , eq . ( 14 ) can then be solved for @xmath181 , and thus for @xmath5 . the procedure is explained in appendix c. the result is shown in fig.@xmath182 . one observes that @xmath183 is proportional to @xmath184 over the main part of the data , from @xmath185 k to @xmath186 k. at higher @xmath3 , @xmath5 tapers off , while at lower @xmath3 there is a hint that @xmath183 might diverge faster than in @xmath184 . the solid line is a guide to the eye which is a sum of powers of @xmath3 . the same law is then used to trace the guide to the eye in fig.@xmath174 . in fig.@xmath182 we have traced for comparison the values of the relaxation time of the thermal modes for two other cases . firstly , we observe that @xmath183 obtained in ( i ) for densified silica glass , @xmath0-sio@xmath1 , is substantially shorter than in @xmath2-sio@xmath1 . however , it shows a very similar functional dependence on @xmath3 . secondly , in order to deepen the understanding , we applied the same method of analysis to data on the la - mode of x - cut crystal quartz , @xmath177-sio@xmath1 . we used to this effect the attenuation coefficient measured by bmmel and dransfeld at 1 ghz @xcite . at temperatures above 150 k , we completed these data using the width of the brillouin line reported in @xcite . these data were analyzed similarly to fig.@xmath174 leading to @xmath187 . the corresponding @xmath183 is shown in fig.@xmath182 . in this case , the line through these points is _ not _ a guide to the eye as for the two glasses , but it is an independent determination of @xmath183 using the well - known kinetic expression for the thermal conductivity , @xmath188 we observe that in this case the value of @xmath5 derived from @xmath189 ( line ) is in remarkable agreement with that obtained from the measurement of anharmonic damping of the acoustic modes . this emphasizes the physical significance of @xmath5 : it really is the mean lifetime of the excitations in the thermal bath which at sufficiently high @xmath3 is entirelly controlled by umklapp ( u ) processes . as pointed out in @xcite , this produces a relaxation proportional to the phonon population , and thus at high @xmath3 one has @xmath190 . at intermediate @xmath3 , where @xmath191 and @xmath192 , eq . ( 15 ) indicates that @xmath193 . this is indeed observed in @xmath177-sio@xmath1 over a large range of @xmath3 , as seen in fig.@xmath182 . equation ( 15 ) assumes propagating thermal phonons . for this reason , it becomes invalid in glasses at soon as @xmath3 increases beyond the thermal conductivity plateau located around 10 k in @xmath2-sio@xmath1 . however , we remark that the functional dependence @xmath194 observed in the two glasses is very similar in shape to that in @xmath177-sio@xmath1 . we have no simple explanation for this . in glasses , there are two competing effects that modify the picture presented above . on the one hand , the strict quasi - momentum conservation which is invoqued in u - processes is strongly relaxed , and this must greatly enhance the interactions in the thermal bath , decreasing @xmath5 . this could account for the observed difference between crystal and glasses . on the other hand , the thermal modes are not expected to be propagating plane - waves but they are at best diffusive . this restricts the spatial extent of the modes , greatly decreasing their overlap and thereby their interactions , which increases @xmath5 . we believe it is the latter effect which produces the much longer @xmath5 in @xmath2-sio@xmath1 compared to @xmath0-sio@xmath1 . indeed , in @xmath0-sio@xmath1 the boson peak is strongly reduced @xcite . this increases the crossover frequency @xmath195 beyond which the acoustic excitations become diffusive , as recently confirmed by inelastic x - ray scattering @xcite . this in turn increases the mean spatial extent of the modes and thereby their interactions leading to a faster thermalization . finally , we remark that the above analysis gives values for the mean grneisen parameter @xmath172 . we find 0.95 for the la - mode of x - cut quartz , 3.6 for @xmath2-sio@xmath1 , and 8 for @xmath0-sio@xmath1 . in the case of crystal quartz , the agreement between the line calculated from the thermal conductivity @xmath189 and that obtained from @xmath87 gives a solid support for the value of @xmath172 . unfortunately , one can not make a similar comparison for the glasses . figure 7 summarizes the present analysis of the internal friction observed in @xmath2-sio@xmath1 with brillouin scattering around 35 ghz . one identifies two main contributions : @xmath83 arising from thermally activated relaxation and @xmath196 which results from network viscosity . around room @xmath3 , @xmath165 is nearly twice as large as @xmath83 . the incoherent tunneling contribution @xmath143 , if it really exists at these high frequency and temperatures , mainly produces the small hump around 40 k. although there is a hint for such a feature in the data shown in fig . 7 , it would need to be confirmed by more precise measurements . also , we neglected in this analysis a possible contribution arising from the quasi - harmonic oscillators of the soft - potential model @xcite . this is explained in appendix d. ) and @xcite ( + ) . the solid curve is the entire adjustment made here which is the sum of three contributions @xmath197 , @xmath198 , and @xmath199.,width=321 ] as @xmath4 is increased beyond 35 ghz , we find that the value @xmath83 calculated with eq . ( 9a ) saturates . this saturation essentially results from the cut - off in @xmath50 . for example at room @xmath3 , @xmath83 does not increase beyond @xmath2001.5@xmath20110@xmath202 . on the other hand , at sufficiently high @xmath3 , @xmath165 continues to grow with @xmath150 according to @xmath203 which applies as long as @xmath204 . we also find that @xmath205 is practically constant from 100 to 300 k. for these reasons , the total @xmath87 becomes dominated by network viscosity , it continues to increase proportionally to @xmath150 up to several hundred ghz , and it is nearly @xmath3-independent from 100 k to 300 k. this is in excellent agreement with the results on acoustic propagation reported for @xmath2-sio@xmath1 thin films by zhu , maris , and tauc @xcite . these authors find that the mean free path , @xmath206 , grows in @xmath207 from @xmath20030 ghz to @xmath200300 ghz , and that it is nearly independent of @xmath3 from 80 to 300 k. since @xmath208 , the dominance of @xmath165 , its growth @xmath209 , and its near constancy in @xmath3 , fully account for the data reported in @xcite . our predictions are also fully consistent with three data points measured with uv brillouin scattering from @xmath20050 to @xmath200100 ghz @xcite . on the other hand , at ultrasonic frequencies , _ i.e. _ much below 35 ghz , the contribution of @xmath165 which decreases proportionally to @xmath150 becomes completely negligible compared to @xmath83 . this justifies the analysis based only on tar that was performed in relation with fig . 1 . after correction for tar and network viscosity ( upper data points ) . the line is a fit of @xmath6 to wachtman s equation as a guide to the eye . the @xmath6 data for @xmath0-sio@xmath1 from ( i ) are also shown for comparison.,width=321 ] figure@xmath210 shows the velocity changes calculated from eq.@xmath106 for tar , and using @xmath211 for the network viscosity , this at two typical frequencies . one notices that the increase in @xmath4 from 6.8 mhz to 35 ghz leads to a large reduction of @xmath212 . on the other hand , @xmath213 , although quite large above 100 k , changes relatively little with @xmath4 . in particular , the dashed curve is entirely in the regime @xmath204 , so that it is practically independent of @xmath150 at lower @xmath150-values . this justifies the approach used to determine @xmath214 in fig . 2 where the anharmonicity contribution was simply neglected . at sufficiently high-@xmath3 , anharmonicity always leads to a quasi - linear decrease of @xmath2 with @xmath3 , very similar to what is observed in crystals @xcite . this has been seen in ultrasonic measurements in a large number of glasses , _ e.g. _ in @xcite where this effect is interpreted as such . at these frequencies , one is in a regime where the velocity changes are dominated by network viscosity while the damping mainly originates from tar . it is generally necessary to go up to brillouin frequencies to observe in @xmath87 both contributions simultaneously . the velocity changes in tetrahedrally coordinated glasses , in particular in @xmath2-sio@xmath1 and in @xmath0-sio@xmath1 , are more complex than described above . fig.@xmath215 shows the observed velocities , @xmath2 , and the unrelaxed value that is corrected for both tar and network - viscosity dispersions , @xmath216 $ ] . as observed , the experimental points collapse to a @xmath217 which on this scale is essentially @xmath150-independent . however , rather than being constant , @xmath217 increases considerably with @xmath3 . in the same figure , the results obtained in ( i ) for @xmath0-sio@xmath1 are shown for comparison . both solid lines are guides to the eye obtained by an _ ad hoc _ adjustment to wachtman s equation @xcite . the behaviour of @xmath217 is remarkably similar in both glasses , pointing to an anomalous hardening of silica with increasing @xmath3 . as discussed in ( i ) , the earlier proposal that this is a manifestation of large structural inhomogeneities @xcite is not supported by observations . the anomalous hardening has recently been attributed to a progressive local polyamorphic transition associated with abrupt rotations of randomly distributed si - o - si bonds , found in simulations @xcite . thermal agitation would redistribute the si - o - si bond angles in configurations where they are more resilient . however , contrary to @xcite , our results indicate that a very similar mechanism would then be active in permanently densified silica . repeating the warning made in ( i ) , although @xmath217 has a clear physical significance , it would be nearly impossible to perform at the moment an experiment to directly measure it . this is even more so in @xmath2-sio@xmath1 than in @xmath0-sio@xmath1 , owing to the lower @xmath218 of the former glass . it remains to compare the value @xmath107 found in section iii to the tunneling - strength parameter @xmath219 from the literature . to do this , both values are inserted in eq . ( 11 ) to calculate the cut - off energy @xmath220 that applies to the present case . we find @xmath221 k , as opposed to the tls value @xmath222 k derived from the low-@xmath3 specific heat data @xcite . such a larger value of @xmath220 was already reported , resulting from a similar analysis in @xcite . its origin was convincingly explained there as arising from a different weighing of the distribution of @xmath220-values , in @xmath223 for the specific heat , and in @xmath224 for the ultrasonic absorption . the value of @xmath92 is thus fully in line with expectations . we have shown that ultrasonic and hypersonic damping and dispersion in vitreous silica at temperatures above the quantum regime originate in at least two main processes : thermally activated relaxation and network viscosity . these have different @xmath150 and @xmath3 dependences which leads to a non - trivial variation of the scattering linewidth @xmath225 with @xmath150 . at a given @xmath3 , tar dominates the damping at low @xmath150 , while anharmonicity dominates at sufficiently high @xmath150 . between these two regimes there occurs a `` dynamical crossover '' which for @xmath2-sio@xmath1 at room @xmath3 falls near our brillouin frequency , as shown in fig . 10 of @xcite and confirmed in fig . 7 above . above this crossover , the damping is dominated by @xmath203 , which implies that @xmath226 . as @xmath150 is increased far beyond optical brillouin frequencies , another broadening mechanism eventually comes into play . around 300 ghz , one expects the onset of a new contribution to @xmath87 , growing with a higher power of @xmath150 , possibly in @xmath227 , as suggested in @xcite . this additional contribution is currently thought to result from the hybridization of the acoustic modes with boson - peak excitations @xcite , as inelastic x - ray scattering observations on @xmath0-sio@xmath1 strongly suggest @xcite . unfortunately , inelastic x - ray scattering at sufficiently small scattering vectors and with sufficiently high energy resolution could not yet be performed on @xmath2-sio@xmath1 to settle the issue of this onset . this hybridization eventually leads to the ioffe - regel crossover at @xmath195 , above which the excitations have lost their plane - wave character . then , the wave - vector @xmath228 does not remain a good label for the modes which are at best diffusive @xcite . the excitations of the thermal bath are expected to be of that nature in glasses , as discussed in section v. power laws @xmath229 , or @xmath230 which is equivalent as long as the frequency is well below the ioffe - regel limit so that @xmath231 , are often employed to represent the damping data @xmath232 . as just explained , @xmath226 is a reasonable approximation in the region between the dynamical crossover and the ioffe - regel one . on the opposite , such laws should be viewed as quite rough approximations in the tar - dominated regime . the reason is that the dependence of @xmath225 on @xmath150 predicted by eq . 9a is more complicated . it can not be written simply in terms of a @xmath3-dependent exponent @xmath137 . for example , in @xmath2-sio@xmath1 at room @xmath3 , a value @xmath233 is given in @xcite , while @xmath234 is derived from @xcite . we conclude that power laws in the tar - dominated regime are _ ad hoc _ devices which only represent approximately the data , especially if the latter are sufficiently noisy to mask the disagreements . at this point , it is legitimate to ask to what extent anharmonicity might be important to sound damping in other glasses . the answer to this question depends on two factors : i ) the strength , density , and distribution of the relaxing defects ; ii ) the strength of the anharmonicity . it is conceivable that in glasses that contain a large quantity of defects up to high values of @xmath80 and @xmath118 , tar would mask the anharmonicity . in that case the dynamical crossover might move up to nearly @xmath218 , so that a regime @xmath226 might not be observable at all . we have found such a case in lithium diborate , @xmath2-li@xmath1b@xmath235o@xmath236 @xcite . this type of situation is expected for most polymers in which the tunneling strength @xmath72 is quite large @xcite . on the other hand , there exist many glasses or amorphous materials for which @xmath72 is quite small , well below @xmath237 @xcite . such are the amorphous group iv semiconductors , c , si , and ge . in that case , the anharmonicity is expected to dominate over a considerable range of @xmath150 and a calculation to that effect has already been performed @xcite . of the tetrahedrally coordinated glasses , it will be particularly interesting to reinvestigate @xmath2-geo@xmath1 and @xmath2-bef@xmath1 . difficulties met in obtaining a quantitative description of damping and dispersion in germania on the basis of tar alone , described in @xcite , might find their resolution by including in the analysis the cut - off @xmath118 and the network - viscosity contribution . to summarize , tar essentially explains acoustic damping results in silica glass up to a few ghz . at higher frequencies , the anharmonic coupling of hypersound to the modes that form the entire thermal bath becomes progressively dominant . this conclusion , in line with previous observations @xcite and simulations @xcite , corrects statements to the contrary often found in the literature , _ e.g. _ in @xcite . the anharmonicity which in our view produces @xmath199 is a coupling of sound to the whole bath , in addition to the relatively small fraction of modes forming the boson peak . this process , which we call network viscosity , is in a way quite similar to the usual lattice viscosity of crystals . for the marquardt - levenberg adjustment of eqs . ( 9 ) , it is convenient to use the parameters @xmath92 , @xmath80 , log@xmath238 , @xmath42 , and @xmath239 . taking as integration variable @xmath240 , and defining the auxiliary variables @xmath241 and @xmath242 , eq.@xmath91 becomes @xmath243 good results are obtained with a simple trapezoidal integration , taking a geometric progression for the vector of @xmath28-values . we used @xmath28 starting at @xmath244 and ending at @xmath245 , in steps of 1% . eq.@xmath106 is handled similarly . one observes on the upper curves of fig . 4 that there is a small difference in the high temperature slopes for the two different frequencies . the origin of this effect might be in the velocity dispersion associated , by the kramers - kronig transform , with the relatively small damping @xmath198 produced by incoherent tunneling . at high @xmath3 , the damping is given by the third equation ( 2.94 ) of @xcite , @xmath246 the corresponding @xmath247 is derived from the third equation ( 2.96 ) of @xcite , which gives for the @xmath150-dependent part @xmath248 the temperature @xmath152 marks the end of the low-@xmath3 plateau region in @xmath87 and it is given by eqs . ( 2.78 ) and ( 2.79 ) of @xcite . owing to the logarithmic dependence in ( b2 ) , it is the difference of slopes at two frequencies which is significant , @xmath249 with the values in fig . 4 , and calculating @xmath152 following @xcite , one finds @xmath250 . as expected , this is smaller than the tunneling strength found from specific heat measurements at low temperatures @xcite , the reduction being of the order of @xmath251 . hence , it is not unreasonable to assign the difference in slopes observed at high-@xmath3 in fig . 4 to the kramers - kronig transform of @xmath198 . it is in principle trivial to solve the quadratic equation @xmath252 for @xmath253 in terms of the measured @xmath254 . however , imaginary roots do occur in the region around @xmath255 owing to fluctuations in the data leading to some points for which @xmath256 . these are the points whose ordinate lies above the maximum of the solid curve drawn in fig.@xmath174 . for these data points , we simply replaced @xmath28 by 1 , as the nearest real solution . this produces a line of data points with @xmath257 in the presentation of fig.@xmath182 . for the data points with @xmath258 , one must select between the upper and the lower root of the quadratic equation . this is set by the location of the maximum in the solid curve , @xmath259 . for @xmath260 , the upper root applies since @xmath261 , while the lower root applies at higher temperatures . we have not considered the damping arising from the weak anharmonicity of the nearly harmonic oscillators ( ho ) of the soft potential model in the domain @xmath262 @xcite . this contribution should be most active at `` intermediate '' frequencies . in @xmath2-sio@xmath1 , its strength has actually been predicted to be largest near our brillouin frequency @xcite . it must be noted that the entire @xmath263-domain covered by the propagating pulse measurements of @xcite falls within this `` intermediate '' frequency range where the additional damping obeys @xmath264 @xcite . if the strength of @xmath265 would be sufficient , one should definitely notice its effect on the @xmath263-dependence of the mean free path reported in @xcite from 80 to 300 k and from 30 to 300 ghz . on the contrary , the mean free path is found practically @xmath3-independent and it increases @xmath266 , in agreement with the network - viscosity contribution . it seems thus justified to neglect @xmath265 in our analysis of the brillouin results up to 300 k. the difficulty in estimating the size of @xmath265 apparently lies in finding the proper value for the tunneling - strength parameter @xmath72 that applies to it . the value @xmath267 3.0@xmath20110@xmath268 , appropriate for tls at very low temperatures @xcite , is certainly too large here . indeed , using this value in eq . ( 3.18 ) of @xcite , we calculate @xmath269 5@xmath20110@xmath270(k ) , which at 300 k is larger than the entire @xmath87 observed . > from an analysis of heat - release measurements there are grounds to adopt here a value of @xmath72 which is at least five times smaller @xcite . even so , the resulting @xmath265 is still too large to agree with the results of @xcite . our brillouin measurements up to 700 k shown in fig.@xmath128 suggest that there could be a small contribution @xmath271 that starts being felt above @xmath272 k. this contribution would then be of the order of @xmath273(k ) , which implies a very small effective @xmath72 for this particular damping mechanism .	the temperature dependence of the frequency dispersion in the sound velocity and damping of vitreous silica is reanalyzed . thermally activated relaxation accounts for the sound attenuation observed above 10 k at _ sonic _ and _ ultrasonic _ frequencies . its extrapolation to the _ hypersonic _ regime reveals that the anharmonic coupling to the thermal bath becomes important in brillouin - scattering measurements . at 35 ghz and room temperature , the damping due to this anharmonicity is found to be nearly twice that produced by thermally activated relaxation . the analysis also reveals a sizeable velocity increase with temperature which is not related with sound dispersion . a possible explanation is that silica experiences a gradual structural change that already starts well below room temperature .
we wish to thank dr . andrea geralico and dr . damiano tommasini for the support with the numerical analysis and prof . salvatore capozziello for useful discussions . this work was partially supported by dgapa - unam , grant no . 113514 , and conacyt , grant no . 166391 . o. luongo and h. quevedo , _ toward an invariant definition of repulsive gravity in proceedings of the twelfth marcel grossmann meeting on general relativity _ , edited by t. damour , r.t . jantzen , r. ruffini ( world scientific , singapore , 2012 ) , part b , pp . 1029 - 1031 , arxiv[gr - qc]:1005.4532	the existence of current - time universe s acceleration is usually modeled by means of two main strategies . the first makes use of a dark energy barotropic fluid entering _ by hand _ the energy - momentum tensor of einstein s theory . the second lies on extending the hilbert - einstein action giving rise to the class of extended theories of gravity . in this work , we propose a third approach , derived as an intrinsic geometrical effect of space - time , which provides repulsive regions under certain circumstances . we demonstrate that the effects of repulsive gravity naturally emerge in the field of a homogeneous and isotropic universe . to this end , we use an invariant definition of repulsive gravity based upon the behavior of the curvature eigenvalues . moreover , we show that repulsive gravity counterbalances the standard gravitational attraction influencing both late and early times of the universe evolution . this phenomenon leads to the present speed up and to the fast expansion due to the inflationary epoch . in so doing , we are able to unify both dark energy and inflation in a single scheme , showing that the universe changes its dynamics when @xmath0 , at the repulsion onset time where this condition is satisfied further , we argue that the spatial scalar curvature can be taken as vanishing because it does not affect at all the emergence of repulsive gravity . we check the goodness of our approach through two cosmological fits involving the most recent union 2.1 supernova compilation . [ [ introduction ] ] * _ introduction _ * + + + + + + + + + + + + + + + + general relativity predicts a universe characterized by a decelerated expansion , if only dark matter and baryons are involved within the energy - momentum tensor . observations , however , predict a late - time accelerated universe , pushed up by some sort of exotic constituent whose nature is today unknown @xcite . to account for this experimental fact , cosmologists employ in general two different strategies . the first one consists in extending the hilbert - einstein action , modifying general relativity by means of additional degrees of freedom @xcite . the second one takes into account an additional fluid , dubbed _ dark energy _ , which enters einstein s equations and represents a source for the cosmic speed up @xcite . both those treatments are based on the hypothesis that space - time geometry is _ fueled _ by components entering the energy - momentum tensor @xmath1 . in such a way , the energy - momentum tensor becomes the source for gravitation and modifies the geometry itself . however , if we rewrite einstein s equations in an equivalent way as @xmath2 , it is possible to reverse the interpretation of @xmath1 . indeed , one can address a dynamical problem by considering an energy - momentum tensor induced by geometry , and not _ vice - versa_. the philosophy is completely different from the one used in higher - order theories of gravity , since no extensions are required , but only a physical mechanism inside the einstein equations themselves , providing a _ geometrical source term_. it is therefore possible to imagine that under certain circumstances the geometry could correspond to a repulsive field that counterbalances the attraction of gravity and induces an accelerated universe . this _ third way _ of handling the universe dynamics becomes a natural consequence of general relativity itself , instead of a new additional ingredient that must be added into the energy - momentum tensor . indications of repulsive effects in gravity are not new , specially in the case of naked singularities @xcite . several intuitive approaches have been proposed to understand repulsive gravity , but only recently an invariant definition in terms of geometric objects was formulated @xcite . indeed , it was shown that the behavior of the curvature tensor eigenvalues can be used to find the places where repulsive gravity appears . this was explicitly shown in the case of naked singularities with black hole counterparts where it was found that there exists a repulsion region which is always situated very close to the source of gravity . here , we extend this approach to the case of a homogeneous and isotropic universe , modeled by the friedmann - robertson - walker ( frw ) metric . we consider the effects of repulsive gravity as a source for the cosmic speed up , and find the stage at which the universe starts accelerating , corresponding to a phase in which repulsive gravity dominates over attraction . we therefore interpret the acceleration of the universe in terms of a natural outcome due to the non - linearity of large - scale cosmology which is reflected in the eigenvalues of the curvature tensor . we characterize gravitational repulsion by evaluating precise values for the observable acceleration parameter @xmath3 and its variation @xmath4 . we also find that the universe dynamics does not depend upon the scalar curvature ; interestingly , this could be interpreted as an explanation of the experimental evidence for that the universe appears spatially flat at all stages of its evolution , and enables us to unify the inflationary phase with the accelerated phase in a single description . finally , we find an approximate effective density induced by repulsive gravity , assuming that the conditions valid at the moment of the repulsion onset continue holding at all times . we check the goodness of our approach through a numerical test based on supernova union 2.1 data , performed on a grid in which all cosmic parameters evolve . particularly , we employ two hierarchical experimental tests . in the first one , we take all coefficients free to span in allowed prior domains . in the second one , we fix the acceleration parameter by means of the most recent planck results . we definitively show compatible limits between our theoretical predictions and numerical outcomes . [ [ repulsive - effects - in - a - homogeneous - and - isotropic - universe ] ] * _ repulsive effects in a homogeneous and isotropic universe _ * + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + although there are several intuitive ways to define repulsive gravity , only recently an invariant definition in terms of the curvature tensor eigenvalues was formulated which gives physically reasonable results in the case of all naked singularities investigated so far @xcite . the idea is that the curvature tensor eigenvalues contain in an invariant manner all the information about the behavior of gravity . to this end , it is convenient to introduce an orthonormal frame @xmath5 which represents the simplest choice for an observer to perform local time , space and gravity measurements so that all the quantities related to this frame are coordinate independent . the orthonormal tetrad is determined by the relationships @xmath6 , with @xmath7 , and @xmath8 . furthermore , the first and second cartan equations enables one to compute explicitly the riemann curvature components in this frame . for the analysis of the eigenvalues it is convenient to consider the bivector representation of the riemann tensor , which is also useful for finding its irreducible representation with respect to the lorentz group . we use the notations and conventions introduced in @xcite , according to which each bivector index @xmath9 corresponds to two tetrad indices , i.e. , @xmath10 , with the convention @xmath11 . then , the curvature tensor can be expressed as a @xmath12matrix which , in the case of the frw orthonormal frame @xmath13 , reduces to the diagonal matrix @xmath14 @xmath15 where we have used the friedmann equations and @xmath16 . it then follows that the only independent eigenvalues of the curvature tensor are @xmath17 and @xmath18 . according to our invariant definition of repulsive gravity @xcite , the curvature eigenvalues reflect the behavior of the gravitational interaction and if gravity becomes repulsive in some regions , the eigenvalues must change accordingly ; for instance , if repulsive gravity becomes dominant at a particular point , one would expect at that point a change in the sign of at least one eigenvalue . moreover , if the gravitational field does not diverge at infinity , the eigenvalue must have an extremal at some point before it changes its sign . this means that the extremal of the eigenvalue can be interpreted as the onset of repulsion . let us now consider the eigenvalues of the frw curvature tensor . first , we note that @xmath18 is positive definite by virtue of the energy conditions and so it does not contain any information about repulsive gravity . it therefore represents a constraint eigenvalue which fixes the total energy budget pushing up the universe , but with no information on its evolution . instead , @xmath17 can change its sign in the interval where the pressure @xmath19 is negative and , so , the universe can accelerate . thus , @xmath17 represents the dynamical eigenvalue which dominates the spacetime dynamics , predicting the acceleration . moreover , @xmath17 does not depend explicitly on the value of the constant @xmath20 which is the first interesting consequence of our approach . indeed , this means that the universe does not need scalar curvature to accelerate so that the curvature density @xmath21 can be taken arbitrarily small , without affecting the universe dynamics . this theoretical result is in agreement with observations @xcite . on the other hand , this could be considered as a theoretical explanation of the spatial flatness of the universe . to further analyze the repulsive effects in the dynamics of our the universe , we introduce the acceleration parameter @xcite @xmath22 to obtain @xmath23 so that @xmath24 when the universe is accelerating , whereas for @xmath25 the universe decelerates . the condition @xmath26 determines the special point @xmath27 at which the acceleration parameter vanishes and the universe undergoes a transition from a decelerating phase into an accelerating phase . the particular time @xmath28 at which this phase transition occurs follows from the integration of eq . ( [ accond ] ) , i.e. , @xmath29 where @xmath30 is an integration constant . the behavior of the eigenvalue @xmath17 is illustrated in fig . [ figura2 ] for @xmath3 and @xmath31 of the @xmath32cdm model . in the redshift domain @xmath33 $ ] with @xmath31 and @xmath3 from the @xmath32cdm paradigm with @xmath34 and @xmath35 . _ ] as mentioned above , the onset of repulsion is determined by an extremal of the eigenvalue , i.e. @xmath36 . then , using eq . , we obtain the onset condition @xmath37 using the identity @xcite @xmath38 where the jerk parameter @xmath39 determines the change of the acceleration @xcite , the repulsion onset condition ( [ versotra ] ) reduces to @xmath40 implying that at that moment of time @xmath41 the jerk must be positive because the onset happens when the universe is still decelerating @xmath42 . by using the definitions of the cosmographic parameters , we can rewrite ( [ repcond ] ) as @xmath43 a relationship that can be integrated and yields @xmath44\ , , \label{hrep}\ ] ] where @xmath45 and @xmath46 are integration constants . this is the behavior of the hubble parameter at the onset of repulsion . from that point on , the value of the acceleration parameter @xmath3 diminishes until it vanishes at @xmath28 ( cf . fig . [ figura2 ] ) . it seems reasonable to construct a cosmological model based upon the behavior of the hubble parameter near the repulsion onset point ( [ hrep ] ) , where the acceleration begins to play an important role in the universe dynamics . in fact , one expects acceleration to be important in late and early phases of the universe evolution . hence , we can assume that eq.([hrep ] ) is valid from the transition time up to @xmath47 and also during the inflationary phase . as the redshift decreases , corresponding to the limit of @xmath48 , with @xmath49 our present time , we get [ approximatiotolambda ] @xmath50&\approx \rho_\lambda\,,\\ \alpha^2\cosh^{-2}[\alpha\left(t_0+\beta\right)]&\approx-3h_0\rho_m-\dot \rho_m\,,\end{aligned}\ ] ] where we considered @xmath51 the cosmological constant density in units in which @xmath16 , @xmath52 is the today matter density and , furthermore , we used the continuity equation @xmath53 for the case of a late - time universe , in which @xmath54 and @xmath55 . the above equations fix limits over @xmath45 and @xmath46 at late times . on the other hand , at the inflationary phase we have [ approximatiotoinfla ] @xmath56&\approx v(\phi)\,,\\ \alpha^2\cosh^{-2}[\alpha\left(t_i+\beta\right)]&\approx-\sqrt{3\,v(\phi)}\dot\phi\,,\end{aligned}\ ] ] where we considered the relation @xmath57 and @xmath58 , with @xmath59 and @xmath60 the field and potential determining the inflation , respectively . we also took into account the slow - roll approximation @xmath61 and @xmath62 , the time at which inflation occurred . those relations impose bounds on @xmath45 and @xmath46 at the times in which the universe accelerated . in other words , the mechanism of acceleration is induced for different epochs in the same way , but with different coupling constants , whose values are associated to the typology of the acceleration involved , i.e. , with dark energy or inflation . [ [ numerical - outcomes - from - experimental - fittings ] ] * _ numerical outcomes from experimental fittings _ * + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we carry out two analyses based on supernova standard candles , employing the most recent union 2.1 compilation . this survey has been built up by means of 580 measurements of supernova magnitudes . to each magnitude @xmath63 , one associates the corresponding redshift @xmath64 and magnitude error @xmath65 , assuming the redshift error to be negligible . under the hypothesis of a perfect gaussian distribution , it is possible to relate the luminosity distance @xmath66 to the magnitude @xmath63 . afterwards , since @xmath67 , one gets : @xmath68 with the coefficients @xmath69 and @xmath70 @xcite . notice that we _ a priori _ assumed @xmath71 , without losing generality , since @xmath17 does not depend upon it . we therefore consider a parameter - fitting procedure which makes use of a grid in which all parameters evolve , in their corresponding priors , i.e. @xmath72 $ ] , @xmath73 $ ] , @xmath74 $ ] . we minimize the chi - square function : @xmath75 where @xmath76 , @xmath77 , @xmath78 $ ] and @xmath79 and we get the results provided in tab . [ tabella ] . _ contours for the experimental results . we plot the graphics of @xmath80 versus @xmath81 ( top left figure ) , @xmath82 versus @xmath81 ( top right ) , @xmath82 versus @xmath80 ( button left ) and @xmath80 versus @xmath81 of the second fit ( button right ) , in which @xmath82 is fixed in the framework of the @xmath32cdm model , with @xmath34 , compatible with modern planck results . in the contour plots , we indicate @xmath83 ( inside ellipse ) and @xmath84 ( outer ellipse ) with different colors , corresponding to @xmath85 and @xmath86 confidence levels . _ , width=302 ] the advantage of using the union 2.1 compilation on a grid is that previous systematics of other catalogs have been reduced . thus , it is easy to show that the _ @xmath87-squared _ function permits to maximize the likelihood function : @xmath88 with a good approximation . in tab . [ tabella ] , we present the results derived from our experimental analysis . fig . [ contorni ] contains the corresponding contour plots with the @xmath83 and @xmath84 of our coefficients . the values of @xmath82 and @xmath80 are basically close to each other , as theoretically predicted . unfortunately , the higher errors on @xmath80 do not allow to characterize this property at a @xmath89 confidence level which would definitively suggest the validity of our model . however , incoming refined data are expected to certify this experimental fact , to see how the coefficients @xmath82 and @xmath80 differ from each other . c\|c\|c @xmath90 parameter @xmath90 & @xmath91 fit1 @xmath91 & @xmath91 fit2 @xmath91 + @xmath92 & @xmath93 & @xmath94 + @xmath81 & @xmath95@xmath96 & @xmath97@xmath98 + @xmath82 & @xmath99@xmath100 & @xmath101@xmath102 + @xmath80 & @xmath103@xmath104 & @xmath105@xmath106 + @xmath17 & @xmath107@xmath108 & @xmath109@xmath110 + [ [ conclusions - and - perspectives ] ] * _ conclusions and perspectives _ * + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we presented a new approach to describe the universe dynamics at current and early phases of its evolution . we showed that it is possible to infer a process enabling the universe to accelerate by means of repulsive effects . to investigate such a phenomenon , we use an invariant definition of repulsive gravity in terms of the curvature tensor eigenvalues and we got two independent eigenvalues : @xmath17 and @xmath18 . it turns out that the dynamics of the repulsive component is governed by a particular eigenvalue @xmath17 whose behavior at the moment of the repulsion onset , @xmath111 , allows us to construct a cosmological model based upon the behavior of the hubble parameter at that moment . moreover , the repulsion onset condition implies that the cosmographic parameters must satisfy the algebraic relationship @xmath112 , which is then shown to be in good agreement with observations . as a result of this analysis , we were able to feature an _ effective emergent dark energy fluid _ , which depends on two free integration constants whose values are fixed for dark energy and inflationary phases . in so doing , we demonstrated that this process fuels the universe acceleration without the need of scalar curvature @xmath113 . we therefore found a theoretical explanation for assuming a spatially flat universe at both late and early phases of its evolution . for these reasons , we believe our treatment to represent a _ third way _ to account the problem of cosmic speeding up , instead of considering barotropic dark energy fluids or modifications of general relativity . finally , we checked the goodness of our cosmological approach by fitting the union 2.1 compilation , inferring cosmic bounds on the observable quantities and showing that our picture is compatible with present time observations . future developments will better characterize our model at different stages of the universe expansion history , to better clarify how eigenvalues change sign . additional efforts will be also spent to better constrain the cosmological parameters of interest with different data sets and combined data surveys .
elliptic flow in relativistic nuclear interactions has its origin in the spatial anisotropy of the initial reaction volume in non - central collisions and in particle rescatterings in the evolving system which convert the spatial anisotropy into a momentum anisotropy @xcite . the spatial anisotropy decreases rapidly because of the fast expansion of the system @xcite making the momentum anisotropy measured at the end of this evolution strongly dependent on the matter properties and the equation of state ( eos ) at the early stage @xcite . comparison of measured anisotropies with hydrodynamic model calculations provide an important test of the degree of thermalisation in the produced particle system at the early stage . flow of heavy particles is affected more strongly by changes in the eos than flow of pions @xcite . the anisotropic flow parameters measured to date at sps and lower energies are mainly those of pions and protons @xcite . rhic experiments have measured elliptic flow for many particle species @xcite , including hyperons @xcite . in these experiments , the rapid rise of elliptic flow with transverse momentum ( @xmath3 ) up to 1.5 and its particle - mass dependence are well reproduced by hydrodynamic models @xcite . at higher values of @xmath3 quark number scaling of elliptic flow has been observed at rhic @xcite which indicates that the quark coalescence mechanism dominates hadron production in the intermediate @xmath3 region @xcite . in order to test the validity of this scenario at sps energies we have extended elliptic flow measurements in 158 pb+pb ( @xmath4 gev ) collisions to @xmath0 hyperons . in this paper , results on @xmath0 elliptic flow as a function of the center of mass rapidity ( @xmath5 ) and transverse momentum will be presented and compared to model calculations . the main components of the na49 detector @xcite are four large - volume time projection chambers ( tpcs ) for tracking and particle identification by energy loss ( @xmath6/@xmath7 ) measurement with a resolution of 3@xmath86% . the tpc system consists of two vertex chambers inside the spectrometer magnets and two main chambers placed behind the magnets at both sides of the beam . downstream of the tpcs a veto calorimeter detects projectile spectators and is used for triggering and centrality selection . the data sample consists of @xmath9 semi - central pb+pb events after online trigger selection of the 23.5% most central collisions . the events were divided into three different centrality bins , which correspond to the first three bins used in a previous analysis ( see table 1 in @xcite ) and are defined by centrality ranges 0 - 5% ( bin 1 ) , 5 - 12.5% ( bin 2 ) , and 12.5 - 23.5% ( bin 3 ) . many model predictions are published for impact parameter ranges similar to those of our centrality classes , which are : 0 - 3.4 fm ( bin 1 ) , 3.4 - 5.3 fm ( bin 2 ) , and 5.3 - 7.4 fm ( bin 3 ) for our centrality classes . the measurement in the centrality range @xmath10 = [email protected]% ( called mid - central ) is obtained by averaging the results of bins 2 plus 3 with weights corresponding to the fractions of the total cross section in these bins . the @xmath0 hyperon candidates were selected from the sample of v@xmath11-track configurations consisting of oppositely charged particles , which include the @xmath0 decays into proton and @xmath12 ( branching ratio 63.9% ) . the identification method @xcite relies on the evaluation of the invariant mass distribution and is enhanced by daughter particle identification applying a cut in @xmath6/@xmath7 around the expectation value derived from a bethe - bloch parametrization . the extracted @xmath0 candidates have a background contamination of 5 - 9% in the p-@xmath13 invariant mass window [email protected] , depending on centrality . the yields of @xmath0 hyperons are obtained by counting the number of entries in the invariant mass peak above the estimated background as a function of the azimuthal angle @xmath14 of @xmath0 candidates with respect to the event plane angle @xmath15 . the background is estimated from a fit of invariant mass spectra to the sum of a lorentz distribution and a polynomial . it is subtracted in 16 bins of @xmath16 . the acceptance of @xmath0 hyperons covers the range @xmath17 and @xmath18 but the detection efficiency strongly depends on azimuthal angle , @xmath3 and @xmath5 . thus we have to introduce differential corrections to avoid biases when averaging @xmath1 over rapidity and transverse momentum . multiplicative factors were introduced for every @xmath0 particle to correct the @xmath0 yields for detector and reconstruction efficiency . they were determined as ratios of published @xmath0 yields @xcite , parametrized by the blast wave model , to the measured raw @xmath0 yields . the elliptic flow analysis is based on the standard procedure outlined in @xcite to reconstruct the event plane for each event and the corrections for the event plane resolution . the event plane is an experimental estimator of the true reaction plane and is calculated from the azimuthal distribution of primary charged @xmath13 mesons . identification of pions is based on @xmath6/@xmath7 measurements in the tpcs . to avoid possible auto - correlations , tracks associated with @xmath0 candidates are excluded from the event plane calculation . the method to determine the event plane azimuthal angle @xmath15 uses the elliptic flow of pions , according to the formula : @xmath19 , \nonumber \\ & y_{2}&= \sum_{i=1}^{n}p^{i}_{{\mbox{\textrm{\scriptsize t } } } } [ \sin(2\phi^{i})-\langle \sin(2\phi)\rangle ] , \\ & { \ensuremath{\phi_{{\mbox{\textrm{\scriptsize 2ep}}}}}}&= \frac{1}{2 } \tan^{-1}\left(\frac{y_{{\mbox{\textrm{\scriptsize 2}}}}}{x_{{\mbox{\textrm{\scriptsize 2}}}}}\right ) , \nonumber\end{aligned}\ ] ] where @xmath20 are the components of the event plane flow vector @xmath21 and the sums run over accepted charged pion tracks in an event . the acceptance correction is based on the recentering method @xcite which consists of subtracting in eq . 1 the mean values @xmath22 and @xmath23 . these mean values are calculated in bins of @xmath3 and rapidity for all pions in those events which contain at least one @xmath0 hyperon candidate . the values were stored in a 3-dimensional matrix of 20 @xmath3 intervals , 50 rapidity intervals , and eight centrality bins . a second level acceptance correction is done by using mixed events . hyperons with respect to the event plane for real events ( solid symbols ) and mixed - events ( open symbols ) in two centrality bins . the curves are fourier expansion fits ( see text eq . ] we used 10 mixed events for each real event . particles for mixed events are randomly selected from different events in the same centrality bin with at least one @xmath0 hyperon . the final angular distributions are obtained by dividing the real @xmath0 angular distribution by the mixed event distribution to remove the acceptance correlations remaining after recentering . the corrected @xmath0 azimuthal distributions are then fitted with a truncated fourier series : @xmath24~~ \\ & + & v^{obs}_{4}\cos[4(\phi-{\ensuremath{\phi_{{\mbox{\textrm{\scriptsize 2ep}}}}}})]).~~ \nonumber\end{aligned}\ ] ] the elliptic flow @xmath1 is evaluated by dividing the observed anisotropy @xmath25 by the event plane resolution @xmath26 : @xmath27 the resolution , @xmath28\rangle } , \nonumber\ ] ] is calculated from the correlation of two planes ( @xmath29 ) for random sub - events with equal multiplicity . the results are @xmath26 = 0.27 , 0.34 and 0.40 for centrality bins 1 , 2 and 3 , respectively . the total errors in figs . 2 - 4 are given by the quadratic sum of contributions from the statistical error of the signal , the uncertainty due to the background subtraction , the mixed event correction and event plane resolution . the observed hexadecupole anisotropy @xmath30 is consistent with zero within statistical errors for all centrality bins . the final statistics in our sample consists of about @xmath31 @xmath0 candidates . this allows flow analysis for several rapidity and @xmath3 bins . hyperons as a function of rapidity ( top ) and @xmath3 ( bottom ) . the open points in the top graph have been reflected about midrapidity . ] two sample azimuthal distributions of @xmath0 hyperons with respect to the estimated reaction plane for real and mixed events are shown in fig . 1 . the curves represent results of fits with the truncated fourier series eq . the distributions exhibit a strong correlation for real events ( full symbols and curves ) . as expected , no correlation is observed for mixed - events ( open symbols , dashed curves ) . the @xmath3 averaged elliptic flow is obtained from all identified @xmath0 hyperons without @xmath3 cuts . it exhibits no significant dependence on rapidity as shown in fig . the absence hyperons as a function of @xmath3 from mid - central ( top ) and central ( bottom ) events measured by the star ( open symbols ) and na49 ( solid symbols ) experiments . curves are hydrodynamical model predictions at rhic energy for the two different centrality bins . ] of a rapidity dependence of @xmath32 was also observed for protons ( see fig . 6 in ref . @xcite ) in mid - central events . we use the @xmath0 sample from the full rapidity range of the data in fig . 2(top ) for the study of @xmath1 as a function of @xmath3 . the @xmath3 dependence of rapidity - averaged @xmath0 elliptic flow is shown in the bottom plot of fig . 2 for two centrality ranges . the @xmath1 parameter significantly increases with transverse momentum , the rise being stronger for more peripheral events . 3 shows a comparison of @xmath33 of @xmath0 hyperons for mid - central and central ( 0 - 5% ) events measured by the na49 and star experiments @xcite . the na45 collaboration recently also presented preliminary results in pb+au collisions at the top sps energy @xcite which agree well with the na49 results ( not shown ) . for mid - central collisions at sps energy the elliptic flow grows linearly with @xmath3 up to @xmath342 gev / c , but the increase is steeper at rhic than at sps energy . it should be noted that rhic mid - central data have been measured in the centrality range @xmath10 = 5@xmath830% while sps events are somewhat more central . the effect from mid - central events measured by the star @xcite ( open symbols ) and na49 @xcite ( solid symbols ) experiments . ] hyperons ( squares ) as a function of @xmath3 from 158 pb+pb mid - central events measured by the na49 experiment . curves are blast wave fits and hydrodynamic model predictions for @xmath4 gev . ] of different centrality ranges has been estimated by hydrodynamic calculations @xcite at rhic energy for the slightly different centrality bins of na49 and star . as shown by the corresponding curves in fig . 3 this explains only partially the difference between both measurements . for central collisions na49 and star results do not differ significantly ( fig . 3 bottom ) . a comparison of @xmath1 of pions and protons obtained by na49 @xcite and star @xcite for mid - central collisions is shown in fig . the na49 values were obtained as the cross section weighted averages of the measurements published in @xcite for the appropriate centrality range . as observed for @xmath0 hyperons , @xmath1 of pions also rises faster with @xmath3 at rhic than at sps . protons seem to exhibit a similar trend , but the limited @xmath3 range and the larger measurement errors do not allow a firm conclusion . a comparison of @xmath33 for pions , protons and @xmath0 hyperons as measured by the na49 experiment in mid - central events is displayed in fig . the elliptic flow grows linearly with @xmath3 for all particle species but the rise for pions starts from @xmath3 close to zero while for protons and @xmath0 particles it starts from @xmath35 . the elliptic flow for pions is significantly larger than that for heavier particles although at @xmath36 the flow becomes similar for all particle species . data are reproduced by blast wave fits @xcite ( dashed curves in fig . 5 ) with the following parameters : freeze - out temperature @xmath37 = 92 mev , mean transverse expansion rapidity of the shell @xmath38 = 0.82 , its second harmonic azimuthal modulation amplitude @xmath39 = 0.021 , and a spatial eccentricity parameter @xmath40 = 0.033 . the values of freeze - out temperature and expansion rapidity are consistent with those obtained from @xmath41 spectra and bose - einstein correlations @xcite . in fig . 5 the measured values of @xmath1 are also compared to hydrodynamical model calculations @xcite assuming a first - order phase transition to a qgp at the critical temperature @xmath42 mev . the initial conditions employed for the hydrodynamical calculations at sps are : initial energy density @xmath43 = 9.0 @xmath44 , baryon density @xmath45 = 1.1 @xmath46 , thermalization time @xmath47 = 0.8 @xmath48 . the initial conditions were fixed as in @xcite by requiring a good fit to the @xmath3 spectra of protons and negatively charged particles in central pb+pb collisions at the sps . with the freeze - out temperature @xmath49 mev tuned to reproduce particle spectra , the model calculations significantly overestimate the sps results for semi - central collisions ( full curves in fig . 5 ) in contrast to predictions at rhic energy which agree quite well with data for @xmath50 @xcite ( see fig.3 ) . the discrepancy at sps may indicate a lack of complete thermalisation or a viscosity effect . however , the model reproduces qualitatively the characteristic hadron - mass ordering of elliptic flow . thus the data support the hypothesis of early development of collectivity . on the other hand , the magnitude and trends of elliptic flow , measured by the second fourier coefficient @xmath1 is found to be underpredicted by a hadronic cascade model @xcite . a more comprehensive description of experimental results was found by coupling a hadronic rescattering phase to the hydrodynamical evolution and hadronisation @xcite . this model can reproduce @xmath41 spectra and elliptic flow both at top sps energy and rhic consistently , although predictions of @xmath1 for exactly the centrality range of the present analysis are not published in the literature . one should note , however , that none of the hydrodynamical models have yet been able to describe bose - einstein correlations successfully . quark coalescence models @xcite have been used to explain the quark number scaling observed at rhic for @xmath1 in the intermediate @xmath3 region . hyperons ( squares ) scaled by the number of constituent quarks @xmath51 as a function of @xmath52 ( top ) and @xmath53 ( bottom ) from 158 pb+pb mid - central events measured by the na49 experiment . ] fig . 6 shows the scaling behavior of the @xmath1 measurements of na49 at the sps . the values of @xmath1 shown in fig . 5 were divided by the number of constituent quarks @xmath54 for baryons and @xmath55 for mesons . when plotting @xmath51 versus @xmath52 approximate scaling was observed at rhic @xcite except for pions . the na49 measurements ( fig . 6 , top ) are consistent with this result in the @xmath3 range covered by the data . the deviation of the pions from the universal curve has been attributed to the goldstone nature of the pion ( its mass is smaller than the sum of masses of its constituent quarks ) or to the effect of resonance decays @xcite . the variable @xmath56 , where @xmath41 is the transverse mass and @xmath57 the rest mass of the particle , was proposed in @xcite as an alternative to @xmath3 since pressure gradients which give rise to azimuthal asymmetry may naturally lead to collective transverse energy of produced particles . good scaling for all particle species is seen when @xmath51 is plotted versus @xmath53 in the range @xmath58 gev covered by the sps data ( fig . 6 , bottom ) . this behaviour was first observed at rhic and interpreted as a result of hydrodynamic evolution @xcite . the data of na49 do not reach higher transverse momenta at which a decisive test of coalescence models would be possible . in summary , we report the first measurement of the anisotropic flow parameter @xmath1 for @xmath0 particles from pb+pb collisions at @xmath4 gev . elliptic flow of @xmath0 hyperons exhibits no significant dependence on rapidity for @xmath59 . it rises linearly with @xmath3 and is smaller than @xmath1 for pions . both features are quantitatively reproduced by the blast wave parametrization but only qualitatively by the hydrodynamic model . the increase of @xmath1 with @xmath3 is weaker at sps than at rhic energy . the observation of significant elliptic flow and its mass dependence suggest strong collective behaviour of the matter produced in collisions of heavy nuclei already at the sps . hydrodynamic models with a deconfinement phase transition and a microscopic freezeout treatment appear to provide a consistent description of @xmath1 and @xmath41 spectra at both top sps and rhic energies @xcite . quark number scaling of elliptic flow ( @xmath51 versus @xmath53 ) was shown to hold also at the sps . however , the limited @xmath3 reach of the data does not allow a decisive test of the quark coalescence hypothesis . this work was supported by the us department of energy grant de - fg03 - 97er41020/a000 , the bundesministerium fr bildung und forschung grant 06f-137 , germany , the virtual institute vi-146 of helmholtz gemeinschaft , germany , the polish state committee for scientific research ( 1 p03b 006 30 , 1 p03b 097 29 , 1 p03b 121 29 , 1 p03b 127 30 ) , the hungarian scientific research foundation ( t032648 , t032293 , t043514 ) , the hungarian national science foundation , otka , ( f034707 ) , the polish - german foundation , the korea science & engineering foundation ( r01 - 2005 - 000 - 10334 - 0 ) and the bulgarian national science fund ( ph-09/05 ) . j. y. ollitrault , nucl . a * 638 * , 195 ( 1998 ) . p. f. kolb , j. sollfrank and u. w. heinz , phys . c * 62 * , 054909 ( 2000 ) . j. y. ollitrault , phys . d * 46 * , 229 ( 1992 ) ; h. sorge , phys . lett . * 82 * , 2048 ( 1999 ) . d. teaney , j. lauret and e. v. shuryak , phys . lett . * 86 * , 4783 ( 2001 ) and nucl - th/0110037 . p. huovinen , p. f. kolb , u. w. heinz , p. v. ruuskanen and s. a. voloshin , phys . b * 503 * , 58 ( 2001 ) . r. snellings [ star collaboration ] , heavy ion phys . * 21 * , 237 ( 2004 ) n. bastid _ et al . _ [ fopi collaboration ] , nucl . a * 622 * , 573 ( 1997 ) ; a. andronic _ et al . _ [ fopi collaboration ] , nucl . a * 679 * , 765 ( 2001 ) ; j. barrette _ et al . _ [ e877 collaboration ] , phys . c * 55 * , 1420 ( 1997 ) ; h. appelshauser _ et al . _ [ na49 collaboration ] , phys . lett . * 80 * , 4136 ( 1998 ) . c. alt _ et al . _ [ na49 collaboration ] , phys . c * 68 * , 034903 ( 2003 ) . j. adams _ et al . _ [ star collaboration ] , phys . rev . c * 72 * , 014904 ( 2005 ) j. adams _ et al . _ [ star collaboration ] , phys . lett . * 95 * , 122301 ( 2005 ) . s. s. adler _ et al . _ [ phenix collaboration ] , phys . lett . * 91 * , 182301 ( 2003 ) v. greco , c.m . ko and p. levai , phys . c * 68 * , 034904 ( 2003 ) ; v. greco , c.m . ko and p. levai , phys . lett . * 90 * , 202302 ( 2003 ) ; d. molnar and s.a . voloshin , phys . rev.lett . * 91 * , 092301 ( 2003 ) . t. anticic _ et al . _ [ na49 collaboration ] , phys . * 93 * , 022302 ( 2004 ) s. a. voloshin and a. m. poskanzer , phys . b * 474 * , 27 ( 2000 ) . j. adams _ et al . _ [ star collaboration ] , phys . rev . lett . * 92 * , 052302 ( 2004 ) . j. milos@xmath60vi _ et al . _ [ na45 collaboration ] , nucl.phys . a * 774 * , 503 ( 2006 ) . p. huovinen , nucl . a * 761 * , 296 ( 2005 ) , and private communication . the initial conditions employed for the hydrodynamical calculations at rhic are : @xmath43 = 23 @xmath44 , @xmath45 = 0.12 @xmath46 , @xmath47 = 0.6 @xmath48 . f. kolb , j. sollfrank and u. w. heinz , phys . b * 459 * , 667 ( 1999 ) ; p. f. kolb , j. sollfrank , p. v. ruuskanen and u. w. heinz , nucl . a * 661 * , 349 ( 1999 ) ; p. f. kolb , p. huovinen , u. w. heinz and h. heiselberg , phys . b * 500 * , 232 ( 2001 ) . m. bleicher and h. stoecker , phys . b * 526 * , 309 ( 2002 ) .	the elliptic flow of @xmath0 hyperons has been measured by the na49 collaboration at the cern - sps in semi - central pb+pb collisions at 158 . the standard method of correlating particles with the event plane was used . measurements of @xmath1 near mid - rapidity are reported as a function of rapidity , centrality and transverse momentum . elliptic flow of @xmath0 particles increases both with the impact parameter and with the transverse momentum . it is compared with @xmath1 for pions and protons as well as with model calculations . the observation of significant elliptic flow and its mass dependence suggest strong collective behavior of the matter produced in collisions of heavy nuclei already at the sps . scaling properties of elliptic flow of different particle species have been tested at 158@xmath2 . the limited @xmath3 range of the data does not allow for a decisive test of the coalescence model .
the coronal heating problem is one of the most fundamental open questions in solar physics . the formation of the high temperature solar atmosphere is clearly related to the magnetic fields generated in the solar interior , but how this magnetic energy is converted to the thermal energy of the corona remains unknown . one important constraint on the coronal heating mechanism is the time scale for energy release . if energy is deposited into magnetic flux tubes on timescales that are very short compared to a characteristic cooling time , then the corona will be filled with loops that are close to equilibrium and appear to be steady . active region observations have generally suggested that high temperature loops are consistent with steady heating ( e.g. , * ? ? ? * ; * ? ? ? these studies have found that the observed evolution of the emission is much slower than the radiative and conductive cooling timescales . there has also been some success in modeling entire active regions with steady heating models @xcite . finally , @xcite have argued that observations of the `` moss , '' the footpoints of high temperature loops , are also consistent with steady heating . they found that the average moss intensities are typically constant over many hours and loops cooling through 1mk were not observed in the moss region they observed . there is considerable evidence that coronal loops observed at lower temperatures ( @xmath2mk ) are evolving and not in equilibrium ( e.g. , * ? ? ? ? * ; * ? ? ? * ; * ? ? ? these `` warm '' loops appear to have apex densities that are much higher than can be accounted for by steady heating models ( e.g. , * ? ? ? * ; * ? ? ? the properties of the warm loops are more consistent with impulsive heating models ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? if it is true that hot loops are close to equilibrium while the warm loops are generally heated impulsively then the coronal heating mechanism becomes even more difficult to understand . the cooling time for short , hot loops is relatively rapid ( typically a few hundred seconds ) , so heating events on these loops would need to occur very frequently . the cooling time for the warm loops is much longer ( typically a few thousand seconds ) , so impulsive heating events on these loops would be infrequent . it is tempting to conjecture that the emission at high temperatures is also consistent with impulsive heating models and that the apparent steadiness of this emission is the result of the superposition of many evolving stands along the line of sight ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? previous observations have not been able to exclude this possibility . for example , it could be that the cooling loops are not easy to detect in an active region core because they are faint relative to both the bright moss and the extended corona in which the active region is embedded . the euv imaging spectrometer ( eis ) and the x - ray telescope ( xrt ) on the _ hinode _ mission provide a new opportunity to observe active region emission in unprecedented detail . eis observes emission covering a very broad range of coronal temperatures : between and , only 4 ionization stages of fe are not present in the eis data ( xxi ) . eis also observes , xv , xvi , and xvii , providing excellent coverage of the critical temperature range around 3mk @xcite . xrt is a broadband imaging telescope that observes high temperature plasma very efficiently and at high spatial resolution . the high cadence xrt data complement eis , which often observes at a much lower cadence , and allows us to track the evolution of coronal plasma over a large field of view . in this paper we use eis and xrt observations to examine the properties of coronal plasma in the core of an active region . the active region that we have selected ( noaa active region 10960 ) is unusual in that most of the overlying warm loops are located to the north and south of the region , providing a largely unobstructed view of the active region core over a wide range of temperatures ( see figure [ fig : context ] ) . high cadence eis observations of this region , which was observed by _ hinode _ during the period 413 june 2007 , have shown that the 195.119 intensities , doppler shifts , and non - thermal widths in the moss are constant over long periods of time , suggesting steady heating @xcite . in this study we find three additional lines of evidence that also indicate that the heating of high temperature loops is steady . we examine the evolution of the emission in individual pixels in xrt and find that the vast majority of the emission is constant to within approximately 15% during this period . the inspection of emission at different temperatures in the core of the active region shows that there is no evidence for loops cooling from high temperatures . we find no relationship between the warm emission ( ) and the steady hot emission ( ) . finally , we find that we can bring all of the observed moss intensities into agreement with steady heating models , if we allow for loop constriction at the base of the loop . previous modeling work had found significant discrepancies at the lowest temperatures observed with eis @xcite . none of these observations is conclusive ; they do not necessarily exclude alternative models that involve non - equilibrium processes . however , these observations do provide very strong constraints on the mechanism responsible for producing the high temperature emission observed in solar active regions . the eis instrument on _ hinode _ is a high spatial and spectral resolution imaging spectrograph . eis observes two wavelength ranges , 171212 and 245291 , with a spectral resolution of about 22m and a spatial resolution of about 1 per pixel . there are 1 and 2 slits as well as 40 and 266 slots available . the slit - slot mechanism and the ccd both image an area on the sun 1024 in height , but a maximum of 512 pixels on the ccd can be read out at one time . solar images can be made using one of the slots or by stepping one of the slits over a region of the sun . telemetry constraints generally limit the spatial and spectral coverage of an observation . see @xcite and @xcite for more details on the eis instrument . for these observations the 1 slit was stepped over the central part of the active region and 25s exposures were taken at each position . an area of @xmath3 was imaged in about 57 minutes ( see figure [ fig : context ] ) . the observing sequence for this observation returned the complete wavelength range from each detector so all of the lines observable with eis are potentially available in these data . the raw data were processed using ` eis_prep ` to remove dark current , warm pixels , and other instrumental effects using standard software . during the processing the observed count rates are converted to physical units . intensities from the processed data are computed by fitting the observed line profiles with gaussians . one line of particular interest that requires special attention is 192.858 , which is strongly blended with 192.813 and an multiplet near 192.90 . an algorithm for deconvolving this blend using 188.216 and multi - component fitting has been developed and is discussed in detail by @xcite . we use this approach here to determine the intensities . the uncertainties for this line are significantly greater than for the other lines that we consider . however , as illustrated by the eis rasters shown in figure [ fig : eis ] , there is a strong similarity between the deconvolved 192.858 image and other rasters from ions formed at similar temperatures , such as 200.972 and 254.87 . the xrt on _ hinode _ is a high cadence , high spatial resolution ( approximately 1 pixels ) grazing incidence telescope that images the sun in the soft x - ray and extreme ultraviolet wavelength ranges . temperature discrimination is achieved through the use of focal plane filters . because xrt can observe the sun at short wavelengths , xrt images can observe high temperature solar plasma very efficiently . the thinner xrt filters allow longer wavelength euv emission to be imaged and extend the xrt response to lower temperatures . further details on xrt are given in @xcite . the xrt data taken around the time of the eis raster consisted of relatively high cadence open / ti - poly images with open / al - thick , open / be - thick , and g band images interleaved at a cadence of about 600s . for this analysis we have processed all of the open / ti - poly images with ` xrt_prep ` to remove the dark current and to do exposure time normalization . all of the images have been co - aligned to the initial frame using cross correlation . to investigate the longer term evolution of this region we have considered all of the xrt data within @xmath4 hours of the start of the eis raster . an example xrt image is given in figure [ fig : context ] . an animation of these data is available in the electronic version of the manuscript . in this analysis we also make reference to observations from _ trace _ , the _ transition region and coronal explorer _ _ trace _ is a normal incidence , multi - layer telescope . there are 3 channels for imaging the corona : /x 171 , 195 , and 284 . there is also a long wavelength channel for imaging the photosphere , chromosphere , and transition region . all of the images are projected onto a single ccd and images in different wavelengths must be taken sequentially . for these observations the data are mostly from the 171 channel , with a few white light , 1600 , and 284 images taken for context . the 171 data were processed with ` trace_prep ` , despiked , and co - aligned using cross - correlation . an example trace 171 image is given in figure [ fig : context ] . an animation of the data is available in the electronic version of the manuscript . in this section we presented a detailed analysis of the eis , xrt , and _ trace _ data available for this period . the eis data represent a snapshot of the active region accumulated over about an hour and yield little insight into the temporal evolution of the emission . the xrt and trace observations , in contrast , have excellent temporal coverage but somewhat limited diagnostic capabilities . in combination they allow us to establish a detailed understanding of the plasma properties in the core of an active region and how they evolve in time . as is suggested by the four - hour xrt movie associated with figure [ fig : context ] , the large scale pattern of the soft x - ray emission from this region is remarkably steady . to illustrate this we have selected several points from the core of the active region and plotted light curves for the emission in these individual pixels . the light curves , shown in figure [ fig : xrt ] , generally show fluctuation levels of about 1015% around the median intensity for this time interval . in figure [ fig : xrt ] we also show two points that illustrate the evolution of transient loops . the lifetimes of these brightenings are generally on the order of 1000s or less and show that the intensities in the core of the active region are steady on timescales that are long compared to a cooling time . a more systematic view of the variability is obtained by computing the median intensity ( @xmath5 ) and standard deviation ( @xmath6 ) for each point in the co - aligned xrt data cube . a spatially resolved plot of @xmath7 is shown in figure [ fig : xrt_fluct ] . note we have considered only those pixels with median intensities above 50dn s@xmath8 . low intensities well outside the core of the active region are highly variable . this calculation shows that there a few structures in the core of the active region that are highly variable , such as those illustrated in figure [ fig : xrt ] , but that the emission in most of the active region is constant to within approximately 15% . @xcite performed a comprehensive study of active region transient brightenings with the soft x - ray telescope on _ yohkoh_. he found that the distribution of event energies followed a power - law distribution and that the total energy in these events was about a factor of 5 smaller than the energy required to heat the active corona . our result is qualitatively consistent with this systematic study . the xrt light curves suggest that the emission in the core of an active region is largely steady , with relatively few significant brightenings and a very high median intensity level ( @xmath9dn s@xmath8 ) . this result is consistent with previous results from sxt ( e.g. , * ? ? ? * ; * ? ? ? * ) , which showed that the observed emission decayed on timescales that were long relative to a cooling time . to put these observations in perspective , the cooling time for a 50 mm loop cooling from equilibrium at 5mk is of the order of 800s , while the xrt emission is relatively steady for several hours . note that a 50 mm loop would connect the middle of the moss regions in this active region . the core of the active region is clearly made up of the emission from many different loops and it is unclear if the low fluctuation levels are the result of steady heating on individual loops or from the superposition of many evolving loops @xcite . models based on many evolving loops , however , predict that there should be warm emission ( @xmath10mk ) that is spatially correlated with the cooling of hot loops in the core of the active region ( e.g. , * ? ? ? furthermore , the magnitude of the warm emission should be related to the observed intensity of the hot emission . rising levels of hot emission should be accompanied by a similar increase in the warm emission . these predictions can be tested easily with the observations of this region with eis . the interpretation of many active region observations is complicated by the presence of overlying warm loops . for this region such loops are largely , but not completely , absent , giving a relatively unobstructed view of the core of the active region at warm temperatures . this is illustrated by the trace 171 movies associated with figure [ fig : context ] . in these movies we see large scale 1mk loops to the north and south of the active region core but not over the core itself . we also do not see many loops cooling through the trace 171 bandpass that are connected to the moss , similar to the observations considered by @xcite . the limited amount of overlying emission allows us to look for any faint warm emission that is related to the formation of hot coronal loops . another useful aspect of these observations is that there is a strong gradient in the intensities of the hot lines ( xvii , , and xrt ) in the core of the active region in between the moss . to illustrate the behavior of the emission at different temperatures in the core of the active region we have selected a line segment parallel to the moss and extracted the intensities from each of the eis rasters . these intensities are displayed in figure [ fig : warm_hot ] where we compare the intensities in each line with those from a hot line , 200.972 . to facilitate the comparisons we have normalized the intensities to their values at the end of the line segment where the intensities of the hot lines are generally the smallest . for the hot lines the intensity along the segment increases by a factor of 3 to 4 in this region . at the cooler temperatures ( xiv ) , however , the intensities are unchanged and show no spatial correlation with the increase of the hot emission . the behavior of the emission from and appears to fall in between the hot and warm lines , with following fairly closely and behaving more like a warm line . since the eis raster lacks temporal information it is possible to argue that the more intense emission in the hot lines is evolving and will show up at a later time as enhanced emission at warmer temperatures . we have , however , already used the xrt observations to demonstrate that hot emission in the core is constant over many hours ( figure [ fig : xrt ] ) . inspection of later xrt data for this active region shows that the strong gradients in the intensity and the dark inter - moss region actually persist for several days ( see @xcite figures 1 and 3 ) . the hot emission shown in the eis rasters is clearly not dominated by transient events . we note that the intensities of the warm lines are non - zero in the core of the active region . inspection of the trace image shown in figure [ fig : context ] suggests that while the inter - moss region is the dimmest part of the active region core , it is not the dimmest part of the active region at these temperatures . as we will show in the next section , almost all of the warm lines have inter - moss intensities that are less than 20% of the moss intensities . the 202.044 line , however , is unusually bright in the inter - moss region . the lower level of the transition responsible for this line is very sensitive to collisonal de - excitation into other levels and the emissivity for this line rises sharply with decreasing density . this sensitivity to low density background plasma may account for the enhanced intensity of the 202.044 line in the inter - moss region . the origin of the warm emission in the core of the active region is unclear . it is most likely related to unresolved , high lying loops that form around the active region ( e.g. * ? ? ? scattered light from the bright moss may also influence the observed intensities in this region . see , for example , @xcite for a discussion of scattered light in the _ trace _ instrument . scattered light levels have not yet been measured in eis , but because eis and trace have several similar design elements , such as a front entrance filter supported by a mesh and multi - layer coatings , observations of dim regions with eis are also likely to suffer from some level of stray light . the moss represents the footpoints of high temperature loops and offers important boundary conditions for physical models . observations of the moss are particularly useful for constraining physical models since the observed intensity is proportional to the loop pressure and independent of the loop length @xcite . that is , @xmath11 where @xmath12 is the observed intensity , @xmath13 is the base pressure , and @xmath14 is the filling factor . this allows for the calculation of physical models of the moss without knowing the loop length . this is important since the magnetic topology of an active region is difficult to infer , even with the use of vector magnetograms ( e.g. , * ? ? ? some previous work comparing eis moss intensities with steady heating models has been presented in an earlier paper @xcite . this work introduced the use of density sensitive lines to determine both the base pressure and the filling factor . for lines formed close to there was good agreement between steady heating models and the observations . at the lowest temperatures , however , there was a dramatic difference , with the modeled intensities being about 400% higher than what was observed . a similar temperature dependence in the contrast between moss and network intensities was noted by @xcite . their work suggested that it was difficult to identify moss regions in relatively low temperature emission . here we investigate the possibility that variations in the loop cross section are responsible for the relatively low moss intensities at low temperatures . for some time it has been recognized that magnetic flux tubes must expand rapidly in the region between the high beta photosphere and the low beta corona . that is , coronal loops must look like `` funnels '' in the transition region . this loop expansion plays an important role in determining the energy balance on coronal loops ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) and needs to be accounted for in modeling the moss intensities . the network model of @xcite shows a significant constriction at temperatures below about 1mk , qualitatively similar to what has been suggested by the eis observations of the moss . the first step in performing detailed comparisons between theory and observation with these data is to accumulate the observed moss intensities from this active region . as shown in figure [ fig : moss ] , we follow @xcite and use a density sensitive line ratio to identify the dense moss regions . for this work we focus on the 203.826/202.044 ratio . the threshold has been set at @xmath15 so that density contours encompass the bright region in the core of the active region . for each line we have extracted the intensities in this region and computed the distribution of intensities . the resulting distribution is approximately gaussian and for each line we have determined the median intensity as well as the standard deviation in the distribution of intensities . a summary of all of the observed moss intensities is given in table [ table : ints ] . to account for any contamination from overlying loops or scattered light , we have subtracted an approximate background derived from the low intensity pixels in the inter - moss region . for consistency , the same background subtraction procedure is applied to all of the emission lines . the background intensities are given in table [ table : ints ] . it is unclear what uncertainties should be associated with these median intensities . the moss intensities are relatively high and the statistical errors in the fits to the gaussian line profiles are generally small . furthermore , for simplicity we will work with the median moss intensities and the uncertainties in these values are also small . finally , the errors in our analysis are undoubtedly dominated by the systematic errors in the atomic data and by the assumptions embedded in the hydro modeling , such the assumption of ionization equilibrium . absent any compelling alternative we assume that the uncertainty in each measured intensity is 15% , which is much higher than the statistical uncertainty . note that in a previous paper we used the 186.880/195.119 line pair to identify the moss and determine the pressure and filling factor @xcite . recent work , however , has shown that there are discrepancies between the various and density sensitive line ratios @xcite , leading one to wonder which densities are the most accurate . as we will discuss in more detail in the next section , the 258.375/261.058 ratio is generally consistent with the results from and we use the ratio here . if we assume a constant loop cross section and steady , uniform heating , the only remaining parameters needed to solve the hydrodynamic loop equations are the volumetric heating rate and the loop length . as indicated by equation [ eq : ints ] , any family of solutions with the same base pressure will yield the same observed moss intensity . to illustrate this we calculate a grid of solutions that covers a range of loop lengths and heating rates . to solve the hydrostatic loop equations we use a numerical code written by aad van ballegooijen ( e.g. , @xcite ) . we consider total loop lengths in the range @xmath16100 mm and heating rates that yield maximum temperatures in the range @xmath177.5mk . in the numerical model the lower boundary condition is set at 0.02mk and the loops are assumed to be oriented perpendicular to the solar surface . for each numerical solution we calculate the total intensity in the line using the usual expression @xmath18 where @xmath19 is the emissivity computed from the chianti atomic physics database version 5.2.1 @xcite . figure [ fig : moss_pressure ] shows the calculated 203.826/202.044 ratio as a function of the pressure at 1.5mk , the temperature of formation for . the ratio is clearly a function of the pressure alone and allows us to easily convert the observed line ratio into a value for the base pressure . an example calculation for the 188.216 intensity is also shown in figure [ fig : moss_pressure ] . with the pressure determined we can now read off the calculated intensities for each line . for 188.216 , for example , the calculated intensity is 8625erg @xmath20 s@xmath8 sr@xmath8 , which is much larger than what is observed and indicates that the emission is not resolved at the spatial resolution of eis . calculated intensities for all of the emission lines are given in table [ table : ints ] . we have estimated a filling factor of 20% using the median ratio of the observed to modeled intensities for the , xi , xii , and xiii lines . this corresponds to the 188.216 line . for the emission lines from these ions there is generally good agreement between the observed and modeled intensities . at lower temperatures , however , the observed intensities are much lower than what is calculated from the simulation . ratios of the observed to modeled intensities as a function of temperature of formation are displayed in figure [ fig : table_ints ] . this plot shows that the constant area model breaks down at temperatures below about 1mk . this discrepancy is consistent with the comparisons presented previously @xcite . an alternative approach to modeling the observed emission is to assume a loop length and determine the heating rate and filling factor that best fits the observed intensities . this method is based on the assumption that the observed moss intensities are independent of the loop length , an assumption supported by theory @xcite and by the numerical solutions presented here . this method has the advantage of being extensible . our goal is to consider loops with complex geometries and the intensity - pressure approach outlined in the previous paragraphs becomes very cumbersome when applied to models with several free parameters . we have implemented such an approach using the levenberg - marquardt least - squares minimization routine ` mpfit ` for the interactive data language ( idl ) . the resulting modeled intensities are presented in table [ table : ints ] and are very similar to those derived from the intensity - pressure relationships . we have assumed a loop length of 50 mm and obtain best - fit parameters of @xmath21erg @xmath22 s@xmath8 and @xmath23 . only the emission lines formed at and above are considered in the minimization . in this approach all of the lines of interest are considered simultaneously and the pressure indicated by the lines is not necessarily reproduced by the model . to emphasize these lines in the final solution we reduce the uncertainties assumed for these lines to 5% . to explore the role of geometry in determining the observed intensities we parametrize the cross - sectional area as a function of height using the following expression @xmath24\right\},\ ] ] for @xmath25 . the area is extended to the other footpoint by symmetry . the area expansion is defined so that the minimum is @xmath26 and the maximum is @xmath27 . the gradient in area expansion peaks at @xmath28 and the magnitude of the gradient is inversely proportional to @xmath29 . smaller values of @xmath29 indicate a sharper transition . this form has been assumed because it allows for the loop expansion to be localized . previous work has considered functions of the form @xmath30 , where the expansion continues throughout the corona ( e.g. , * ? ? ? * ; * ? ? ? there is considerable empirical evidence that coronal loops have constant cross sections ( e.g. , * ? ? ? * ) and our area function is consistent with this . note that in our prescription the loop geometry is fixed and does not respond to heating within the loop . the expectation is that the constriction will reduce the calculated intensities at the low temperatures . the discrepancy between the observed intensities appears to occur around the temperature of formation of , which is approximately 1mk . in the constant cross - section case this temperatures occurs at about 0.5 mm up the loop length . to solve for the best - fit parameters we assume a loop length of 50 mm and use the levenberg - marquardt approach . the free parameters are the volumetric heating rate ( @xmath31 ) , the filling factor ( @xmath14 ) , and the loop expansion parameters @xmath32 , @xmath28 , and @xmath29 . an example solution is shown in figure [ fig : expand ] and the corresponding calculated intensities are given in table [ table : ints ] . the solutions to the constant cross section and funnel models are generally similar . the constant cross section model has somewhat higher temperatures and lower densities . the most significant impact on the calculated intensities comes from the area term in the emission measure @xmath33 the funnel model clearly reproduces the intensities at the lower temperatures much better than the constant cross - section models do . in this fit all of the emission lines are used . to emphasize the lines over the other lines the error assumed for the intensities is again reduced to 5% . previous studies on the transition region geometry ( e.g. , * ? ? ? * ; * ? ? ? * ) have considered the area as a function of temperature and for comparisons with these works we show @xmath34 derived from the model parameters and the solution to the loop equations in figure [ fig : expand2 ] . our result is qualitatively similar to the most promising funnel model derived by ( * ? ? ? * see model b4hl in their figure 12 ) . in this model the area is relatively constant until @xmath35 and then expands rapidly , similar to what we show in figure [ fig : expand ] . our result , however , is not consistent with the model derived by @xcite from sumer doppler shift measurements ( see their figure 19 ) . their model shows significant expansion at temperatures above @xmath36 . in both cases , as the authors were well aware , comparisons were made with spatially averaged quiet sun observations and the applicability of these results to our modeling of the active region moss is unclear . @xcite used spatially resolved measurements of the quiet network to examine loop expansion from the transition region to the corona . the amount of loop expansion that they measured is generally similar to what we have determined here , although it appears to begin at a somewhat lower temperature than in our model . one factor that we have not considered is the impact of chromospheric absorption on the moss intensities . @xcite have shown that the evolving , reticulated pattern of the moss is often related to obscuration by chromospheric jets on neighboring field lines . absorption leads to reduced observed intensities and impacts the filling factors and heating rates determined from the observations . this chromospheric absorption could also be stronger at the lowest heights and could explain the lower intensities of the lower temperature lines . this could also potentially complicate the inference of the transition region geometry from the observed intensities . recent analysis of coordinated eis and sumer observations suggests that emission below the lyman continuum is reduced by about a factor of 2 @xcite . it should be noted that this result is predicated on comparisons between 2 line ratios ( 186.880/195.119 and 1241.990/195.119 ) . as we have discussed , the atomic data for is problematic and there is some uncertainty in this correction . the results from also do not yield insights into the possibility of additional absorption at lower heights . clearly more analysis is needed on this important issue . in this paper we have used the 203.826/202.044 line ratio to determine the electron density while previously the 186.880/195.119 ratio was used . the ratio is sensitive over a very wide range of densities ( @xmath37 7 12 ) , making it a very useful plasma diagnostic . it has been found , however , that the and densities generally do nt agree @xcite , leading to the potential for systematic errors in the modeling . in many previous studies the atomic data for si has proven to be very robust ( e.g , * ? ? ? * ; * ? ? ? * ) and so we have compared the densities derived from the 258.375/261.058 ratio with those from the other line pairs . the ratio is sensitive over a much smaller range of densities than the and pairs so we have looked for observations of all three line pairs in small active regions , bright points , and the quiet sun , i.e. , regions where the densities are generally lower than in the moss region that we have studied here . the result of one such comparison is shown in figure [ fig : dens ] . this plot shows the densities computed in a small active region from all three line pairs . these calculations clearly show that at high densities the results are systematically higher than those from and and that the and densities are in good agreement . we have found similar results from the analysis of 4 other quiet sun and active region observations . the problems with become more pronounced at the highest densities , which is particularly problematic for the analysis of the moss . in these observations the peak density is about a factor of 3 higher in . since the observed intensity scales with the square of the density this discrepancy is amplified . we have derived filling factors of 1020% for this moss region . emphasizing the ratio would lead to filling factors almost an order of magnitude smaller . we have presented a comprehensive analysis of observations in the core of an active region using data from the eis and xrt instruments on _ hinode _ and _ trace_. the apparent steadiness of the xrt emission , the lack of spatial correlation between the hot and warm emission , and the consistency of the funnel models with the observed emission all point to frequent heating events that keep the hot loops close to equilibrium . furthermore , these results are consistent with high cadence eis measurements of moss intensities , doppler shifts , and nonthermal widths that show little evidence of dynamical events over many hours @xcite . in combination , these results provide strong evidence that the heating in the core of an active region is effectively steady , that is , the time between heating events is short relative to the relevant radiative and conductive cooling times . hinode is a japanese mission developed and launched by isas / jaxa , with naoj as domestic partner and nasa and stfc ( uk ) as international partners . it is operated by these agencies in co - operation with esa and nsc ( norway ) . _ trace _ is supported by a nasa contract to lmatc . the authors would like to thank jim klimchuk for helpful discussions on the coronal heating problem , george doschek for an explanation of the 202.044 emissivity , and the referee for a number of very helpful comments on the original manuscript . and @xmath7 as a function of position for the xrt open / ti - poly data . for the @xmath7 plot only the points with a median intensity above 50dn s@xmath8 are shown . for low intensity pixels the fluctuation level is generally very high ( @xmath38% ) . ] and @xmath7 as a function of position for the xrt open / ti - poly data . for the @xmath7 plot only the points with a median intensity above 50dn s@xmath8 are shown . for low intensity pixels the fluctuation level is generally very high ( @xmath38% ) . ] . the box indicates the region used to estimate the background contribution to the line intensities . for each emission line considered here we have computed the median intensity in the moss and background regions . ] , @xmath39 , @xmath40 , @xmath41 mm , and @xmath42 mm . bottom panels : the temperature and density along the loop . the peak temperature of formation for each ion is indicated on the plots . for comparison , the corresponding best - fit solution for the constant cross section case is also plotted . the total loop length is 50 mm and only part of the solution is shown . ] lrrrrrrrrrr 276.579 & 36.4 & 12.4 & 3.1 & 33.3 & 178.3 & 0.19 & 231.3 & 0.14 & 38.0 & 0.88 + 270.394 & 97.8 & 25.9 & 8.5 & 89.3 & 420.1 & 0.21 & 541.3 & 0.16 & 95.4 & 0.94 + 280.737 & 96.7 & 27.5 & 5.1 & 91.6 & 247.1 & 0.37 & 321.4 & 0.28 & 67.4 & 1.36 + 275.352 & 136.6 & 42.0 & 9.6 & 126.9 & 529.6 & 0.24 & 668.2 & 0.19 & 137.7 & 0.92 + 188.497 & 248.0 & 64.6 & 13.8 & 234.2 & 710.1 & 0.33 & 892.6 & 0.26 & 229.9 & 1.02 + 197.862 & 106.1 & 26.1 & 8.0 & 98.0 & 310.3 & 0.32 & 384.8 & 0.25 & 101.8 & 0.96 + 184.536 & 936.9 & 225.3 & 62.3 & 874.6 & 1107.0 & 0.79 & 1352.1 & 0.65 & 553.3 & 1.58 + 182.167 & 1114.6 & 266.3 & 59.0 & 1055.6 & 1114.1 & 0.95 & 1326.6 & 0.80 & 935.0 & 1.13 + 188.216 & 1827.7 & 398.5 & 121.4 & 1706.4 & 1706.4 & 1.00 & 1984.4 & 0.86 & 1347.6 & 1.27 + 258.375 & 1390.2 & 301.4 & 65.4 & 1324.8 & 1171.6 & 1.13 & 1354.4 & 0.98 & 1355.8 & 0.98 + 261.058 & 455.0 & 97.3 & 24.7 & 430.3 & 355.0 & 1.21 & 406.9 & 1.06 & 402.0 & 1.07 + 186.880 & 3026.4 & 676.5 & 130.8 & 2895.6 & 2608.4 & 1.11 & 3072.6 & 0.94 & 3263.0 & 0.89 + 192.394 & 1309.5 & 266.7 & 81.9 & 1227.5 & 1443.1 & 0.85 & 1622.4 & 0.76 & 1571.6 & 0.78 + 195.119 & 4102.9 & 827.4 & 271.6 & 3831.3 & 4883.6 & 0.78 & 5516.1 & 0.69 & 5395.9 & 0.71 + 202.044 & 2196.5 & 349.7 & 214.8 & 1981.7 & 1457.3 & 1.36 & 1626.5 & 1.22 & 1759.2 & 1.13 + 203.826 & 6609.6 & 1284.1 & 304.9 & 6304.7 & 4633.7 & 1.36 & 5379.5 & 1.17 & 6446.1 & 0.98 [ table : ints ]	the timescale for energy release is an important parameter for constraining the coronal heating mechanism . observations of `` warm '' coronal loops ( @xmath0mk ) have indicated that the heating is impulsive and that coronal plasma is far from equilibrium . in contrast , observations at higher temperatures ( @xmath1mk ) have generally been consistent with steady heating models . previous observations , however , have not been able to exclude the possibility that the high temperature loops are actually composed of many small scale threads that are in various stages of heating and cooling and only appear to be in equilibrium . with new observations from the euv imaging spectrometer ( eis ) and x - ray telescope ( xrt ) on _ hinode _ we have the ability to investigate the properties of high temperature coronal plasma in extraordinary detail . we examine the emission in the core of an active region and find three independent lines of evidence for steady heating . we find that the emission observed in xrt is generally steady for hours , with a fluctuation level of approximately 15% in an individual pixel . short - lived impulsive heating events are observed , but they appear to be unrelated to the steady emission that dominates the active region . furthermore , we find no evidence for warm emission that is spatially correlated with the hot emission , as would be expected if the high temperature loops are the result of impulsive heating . finally , we also find that intensities in the `` moss , '' the footpoints of high temperature loops , are consistent with steady heating models provided that we account for the local expansion of the loop from the base of the transition region to the corona . in combination , these results provide strong evidence that the heating in the core of an active region is effectively steady , that is , the time between heating events is short relative to the relevant radiative and conductive cooling times .
detailed investigation of populations in local group ( lg ) galaxies using stellar photometry has recently become very popular ( aparicio 1998 ) . this is a direct way to study the populations and evolution of galaxies . however , even for lg objects it is difficult to reach the required photometry depth . to reconstruct the star formation history of more distant galaxies a different techniques should be applied . historically , the most popular method is an evolutionary population synthesis ( tinsley 1980 ; arimoto & yoshii 1986 ) , which treats the entire galaxy as a single zone unit . such an approach is probably satisfactorily applicable only for ellipticals . however , it seems too simplified in the case of spiral disks . to improve the tools available for the evolutionary analysis of galaxies , a new method for multicolor mapping is introduced and applied for the interacting galaxy pair ngc5194/95 ( hereafter , m51 system ) . m51 is a well studied nearby system of two overlapping galaxies seen nearly face - on , and therefore suitable for testing new methods and @xmath1 mapping techniques . since the interpretation of the results of optical photometry always suffer from the influence of dust residing in galaxies , it is important to know the extinction properties of the galaxies under consideration . for this purpose , overlapping galaxies are particularly suitable , allowing precise derivation of an attenuation law and the amount of the extinction in the foreground galaxy ( white & keel 1992 ) . moreover , results of an extensive dynamical modeling of m51 are available ( barnes 1998 ) , and hence can be used for a comparison of the photometrically derived parameters with the ones determined by other independent methods . observations were carried out at kiso observatory ( japan ) using the @xmath2 schmidt telescope on february 17 & 18 and march 7 , 1994 . the ccd camera tek 1024 ( read - out noise @xmath3 , scale @xmath4 , and field of view @xmath5 ) was used . the open cluster m67 was observed every night as a standard field . more than 150 exposures of m51 of @xmath6 seconds duration and image quality of @xmath7 were taken in total . the standard @xmath8 and narrow - band @xmath9 ( @xmath10 ) filters , together with @xmath11 ( @xmath12 ) filter of vilnius system . standard reductions were made with iraf . the @xmath1 aperture photometry ( scanning the entire field with a constant step of @xmath13 and aperture of @xmath14 ) was performed using an afo package ( bridius & vanseviius 1998 ) . assuming the distance to the m51 system to be equal to @xmath15 ( feldmeier , ciardullo , & jacoby 1997 ) , each circular @xmath14 aperture corresponds to the region of @xmath16 in galaxies ( hereafter , zone ) . in the color - magnitude diagram ( cmd ) ( fig . 1 ) zones of both galaxies are shown separately . for further discussion we use only precisely measured ( @xmath17 ) zones with @xmath18 . it is worth noting the different morphologies and fine structures of both cmds ( fig . 1 ) . most of the `` streaming features '' are parallel to the reddening arrow , and can be attributed to the dust influence . the problem of dustiness of galaxies at different evolutionary stages is widely discussed recently . in particular , the dilemma of whether the spiral disks are optically thick or thin is debated ( davies & burstein 1995 ) . to clarify the problem of spiral disk dustiness , we determine the amount of extinction and the attenuation curve in ngc5194 using ngc5195 as a tracer . the facts supporting the validity of such a method are : a ) ngc5195 is behind ngc5194 ; b ) ngc5195 is a source bright enough to have a high @xmath0 ratio even in the most obscured parts of the galaxy ; c ) ngc5195 is found to be a regular s0 galaxy in the near - infrared ( thronson , rubin , & ksir 1991 ) , therefore it is possible to compare symmetrically located zones . we measured two regions ( symmetric with respect to the ngc5195 center and bar ) , each consisting of 50 zones ( @xmath19 ) at an average distance of @xmath20 from the center . visual inspection of the @xmath21 image helped to locate the region relatively free of dust signatures , and then the corresponding dusty region was chosen on the opposite side of the center and bar . the @xmath22 and @xmath23 histograms of the investigated zones are shown in figure 2 . the parameters of the derived attenuation curve are given in table i ( definition of the extinction : @xmath24 , is used ) . the derived median value of the extinction for the interarm region at @xmath25 distance from the center of ngc5194 is @xmath26 . .properties of the extinction in ngc5194 . [ cols="^,<,^,^,^",options="header " , ] numerous determinations of the amount of extinction in ngc5194 have recently been obtained . smith et al . ( 1990 ) derived @xmath27 using ngc5195 as a tracing source . rix & rieke ( 1993 ) put a limit on the global optical depth of ngc5194 disk of @xmath28 . nakai & kuno ( 1995 ) determined @xmath29 , based on a study of 37 @xmath30 regions in the disk of ngc5194 . beckman et al . ( 1996 ) concluded that optical depth in the @xmath22 band , averaged out to a radius of three scale lengths from the center of ngc5194 , is 0.7 . these results mainly ascribe properties of the arm regions , however they do not contradict our conclusion that the disk of ngc5194 is opaque in the interarm regions even at a large distance from the center . however , such a conclusion is in contradiction with the one drawn by white , keel , & conselice ( 1996 ) from a study of nine overlapping galaxy systems . they derived the extinction @xmath31 in the arm and @xmath32 in the interarm regions . ngc3314 is the only system in their sample which possesses significant extinction in the arm ( @xmath33 ) and in the interarm ( @xmath34 ) regions , which supports our conclusion that some of the spiral disks are opaque even in the interarm regions . the color excess ratios derived for the outer part of the disk are very close to those of the milky way ( mw ) extinction law . taking into account the same conclusion reported by panagia et al . ( 1996 ) for the central part of ngc5194 , we assume that color excess ratios are constant throughout the disk of this galaxy . to avoid the age - abundance degeneracy in determining the effective age of zones , global distribution of the abundance should be derived . abundance analysis of @xmath30 regions ( daz et al . 1991 ; zaritsky , kennicutt , & huchra 1994 ) indicates the mean abundance to be above solar : @xmath35 , with a gradient equal to @xmath36 . alternatively , hill et al . ( 1997 ) explained the observed uv color gradient in terms of radial extinction gradient , rather than of abundance . we define an abundance parameter ( independent of the extinction ) as follows : @xmath37 . @xmath38 is a measure of @xmath39 spectral feature strength , which is a good abundance indicator ( barbuy 1994 ; casuso et al . plotting @xmath38 vs. radial distances from the galaxy centers , we do not see the abundance gradient except for the very central region inside the radius of @xmath40 . however , a significant difference of abundances between ngc5194 and ngc5195 is found . therefore , deriving the ages of zones we assume no abundance gradient across the disk of both galaxies . the abundances z=0.05 for ngc5194 and z=0.02 for ngc5195 are adopted , respectively . such an assumption does not significantly affect differential analysis of the age distribution of populations in the disk outside the inner lindblad resonance ( radius @xmath41 ) of ngc5194 ( elmegreen , elmegreen , & seiden 1989 ) , and outside the comparable radius in ngc5195 . let us introduce a parameter @xmath42 as a measure of the effective age of each zone , which is again extinction - independent . actually , both @xmath43 formulae are only correct in the case of a star residing behind the dust cloud ( screen ) . the validity of dusty screen approximation in the case of face - on galaxies was shown by gonzlez & graham ( 1996 ) . diagram @xmath22 vs. @xmath44 constructed from our data sample , in contrast to figure 1 , shows no prominent `` streaming features '' , which can be attributed to the influence of extinction . all this indicates that @xmath44 is a reddening - free parameter in the case of face - on galaxies . figure 3a shows calibration of @xmath44 vs. age for single ( age ) stellar populations ( ssps ) calculated from the model spectra library gissel96 ( bruzual & charlot 1993 ) . figure 3b provides histograms of the parameter @xmath44 for both galaxies of the m51 system , which are simply separated by declination . a significant difference in age between the two galaxies is obvious even if the interfering influence of the northern arm of ngc5194 projected onto ngc5195 is neglected . the derived ages of the bulk of population in ngc5194 & ngc5195 are @xmath45 and @xmath46 , respectively . the gray - coded population age distribution in galaxies is shown in figure 4 . recent dynamic modeling suggests two starbursts in m51 , the first less than @xmath47 ago and the second one @xmath48 ago ( see barnes ( 1998 ) for discussion ) . we derive the age for a bulk of zones in ngc5194 to be @xmath49 , in good agreement with the dynamical prediction . zones in ngc5195 are @xmath50 older . if one assumes this age difference to be a characteristic time between two close - up passages , the predicted age @xmath51 for the last passage again is in good agreement with the dynamical modeling . therefore , our study indicates that the evolutionary scenario of m51 has had at least three periods of activity . the first starburst occurred @xmath52 ago in ngc5195 during a close - up passage with ngc5194 , when , probably , ngc5195 was transformed to an s0 galaxy . the next starburst took place after @xmath53 , when the bulk of the population in ngc5194 was formed , and finally the last one occurred again after @xmath53 when the present grand - design appearance of ngc5194 was `` touched up '' . main conclusions are the following : a ) the attenuation law determined in ngc5194 is close to mw standard extinction law ; b ) the opaque interarm region at distance @xmath25 from the center of ngc5194 supports optically thick models of spiral disks ; c ) the derived effective age @xmath48 for the bulk of ngc5194 zones agrees well with the predictions of recent dynamic modeling of m51 ; d ) the effective age @xmath52 derived for the bulk of zones in ngc5195 implies a scenario with three periods of starburst activity . however , more sophisticated radiative transfer models and realistic star - dust distributions as well as @xmath1 models of the evolution of spiral disks , instead of simple ssp approach , are crucial for the interpretation of the mapping data . i am thankful to k. kodaira and s. yoshida for their help with an application for observing time . kiso observatory s staff is also generously acknowledged . i am indebted to k. zdanaviius for providing photoelectric measurements of m51 in vilnius system . finally , i would like to thank a. kuinskas and y. mclean for careful reading of the manuscript . aparicio , a. 1998 , in : t. richtler & j.m . braun ( eds . ) , the magellanic clouds and other dwarf galaxies , shaker verlag , p. 107 arimoto , n. , & yoshii , y. 1986 , aa , 164 , 260 barbuy , b. 1994 , apj , 430 , 218 barnes , j.e . 1998 , in : r.c . kennicutt , jr . et al . , galaxies : interactions and induced star formation , springer - verlag , p. 323 beckman j.e . , et al . 1996 , apj , 467 , 175 bridius , a. , & vanseviius , v. 1998 , baltic astronomy , 6 , 661 bruzual , g.a . , & charlot , s. 1993 , apj , 405 , 538 casuso , e. , vazdekis , a. , peletier , r.f . , & beckman , j.e . 1996 , apj , 458 , 533 davies , j.i . , & burstein , d. ( eds . ) 1995 , the opacity of spiral disks , kluwer daz , a.i . , 1991 , mnras , 253 , 245 elmegreen , b.g . , elmegreen , d.m . , & seiden , p.e . 1989 , apj , 343 , 602 feldmeier , j.j . , ciardullo , r. , & jacoby , g.h . 1997 , apj , 479 , 231 gonzlez , r. , & graham , j.r . 1996 , apj , 460 , 651 hill , j.k . , 1997 , apj , 477 , 673 nakai , n. , & kuno , n. 1995 , publ . japan , 47 , 761 panagia , n. , et al . 1996 , in : d.l . block & j.m . greenberg ( eds . ) , new extragalactic perspectives in the new south africa , p. 552 rix , h .- w . , & rieke , m.j . 1993 , apj , 418 , 123 smith , j. , et al . 1990 , apj , 362 , 455 tinsley , b.m . 1980 , apj , 241 , 41 thronson , h.a . , jr . , rubin , h. , & ksir , a. 1991 , mnras , 252 , 550 white , r.e . , , & keel , w.c . 1992 , nature , 359 , 129 white , r.e . keel , w.c . , & conselice , c.j . 1996 , astro - ph/9608113 zaritsky , d. , kennicutt , r.c . , jr . , & huchra , j.p . 1994 , apj , 420 , 87	a new technique is used to derive the extinction and age of populations in the interacting galaxies ngc5194 & ngc5195 from high @xmath0 multicolor photometric data . a new evolutionary scenario of the interaction is proposed .
in the quantities related to kaon physics such as @xmath1 , @xmath2 , @xmath0 , @xmath3 and @xmath4 , the first three are constructed straightforwardly on the lattice . however , in actual numerical simulations , there are several potential sources of systematic error , namely explicit breaking of chiral symmetry , scaling violation , finite volume effect and quenching effect . although many efforts have been made to decrease these errors , yet more numerical simulations should be done to attain a conclusion . we report our results of @xmath1 , @xmath2 and @xmath0 as a conclusion of the first stage of our longstanding quenched numerical simulations using domain - wall fermion and dbw2 gauge action . by our choice of the lattice action , it is expected that the contamination from the explicit chiral symmetry breaking is vanishingly small . in particular , we estimate the effect of operator mixing in the calculation of @xmath0 through the non - perturbative calculation of the renormalization factors which weight contributions of the mixing operators . we also examine the scale dependence of our results by carrying out two kinds of numerical simulations with a similar physical lattice volume @xmath5 fm but with different scales @xmath6 2 gev and 3 gev . after comparing our results with those from previous works , we present our continuum results of @xmath0 . for other kind of rbc s numerical simulations with the dynamical quark , see ref . @xcite . .simulation parameters , statistics and results of @xmath7 , @xmath8 , @xmath9 and @xmath10 . [ cols=">,^,^ " , ] in table [ params ] , we enumerate two sets of our simulation parameters , statistics and results of the basic quantities such as lattice scale from the input @xmath11 mev , the residual quark mass @xmath12 and the renormalization factor of axial vector @xmath9 . results of the basic quantities for @xmath13 are quoted from ref . @xcite , in particular . for the finer lattice with @xmath14 , it is known that the topological charge @xmath15 changes very slowly in the ordinary markov chain @xcite . to avoid incorrect distribution of @xmath15 , we generated configurations as described in ref . @xcite . as a result , we obtained a reasonable distribution of the topological charge : @xmath16 . we employ the averaged quark propagator over those with periodic and anti - periodic boundary conditions in the temporal direction for the calculation of @xmath1 , @xmath2 and @xmath0 . therefore , temporal lattice size may be treated as 96 and 48 for @xmath14 and @xmath17 . ( filled symbols ) and @xmath2 [ gev ] ( open symbols ) versus @xmath18 $ ] . , width=207 ] as a combination of fit parameters of two point correlation functions of mesons , we calculate decay constants of the pseudo - scalar from the combination of the fit parameters : @xmath19 where @xmath20 and @xmath21 are the amplitudes of the correlation functions @xmath22 and @xmath23 . @xmath24 is determined from the simultaneous fit of these two correlation functions thus common to both . using the linear fit with @xmath25 , we extract bare values of @xmath1 at @xmath26 and @xmath2 at @xmath27 . multiplying these values by @xmath9 in table [ params ] , we show our physical results in figure [ fpi_comp ] . as well as for @xmath28 ( circles ) and @xmath17 ( squares ) , results for @xmath29 ( diamonds ) from ref . @xcite are plotted in this figure . by the linear fit using these three data , our results are @xmath30 140.1(3.3 ) mev and @xmath31153.9(2.6 ) mev in the continuum limit . they are both inconsistent with the experimental values . on the other hand , our result @xmath32 in the continuum limit is roughly consistent with the estimation from the quenched chiral perturbation theory @xcite . and rgi value @xmath33 versus @xmath34 for dbw2 @xmath14 ( left ) and @xmath13 ( right ) . in each panel , @xmath35 ( circles ) , rgi value ( squares ) and its linear extrapolation are shown.,title="fig:",width=157 ] and rgi value @xmath33 versus @xmath34 for dbw2 @xmath14 ( left ) and @xmath13 ( right ) . in each panel , @xmath35 ( circles ) , rgi value ( squares ) and its linear extrapolation are shown.,title="fig:",width=143 ] we compute @xmath36 _ i.e. _ @xmath0 on the lattice , by extracting a plateau of the ratio of the correlation functions : @xmath37 where the locations of two wall sources @xmath38 are @xmath39 and @xmath40 for @xmath41 and @xmath17 . we chose @xmath42 and @xmath43 for the fit ranges in each case . to determine the renromalization factor @xmath35 , we employ the non - perturbative calculation which was pioneered by ref . including the possibility of the contribution from the mixing operators , the renormalization condition is written by @xmath44 where @xmath45 and @xmath46 are the amputated four point vertices on the lattice and in the tree level for the four - quark operator @xmath47 with certain chiralities @xmath48 and @xmath49 . by solving ( [ renorm2 ] ) , we observe that the renormalization factors @xmath50 for the mixing operators are at most @xmath51 of the factor for @xmath52 , which corresponds to @xmath53 . these small @xmath50 s suppress the effect of the operator mixing to be less than the statistical error of @xmath36 in spite of the fact the b - parameters of the mixing operators are a few dozens times larger than @xmath36 @xcite . in figure [ zbkinv ] , we show the results @xmath54 in ri / mom scheme and the renormalization group independent ( rgi ) value @xmath55 , where the factor @xmath56 was calculated in ref . @xcite . extrapolating the data for @xmath57 linearly , we quote @xmath33 at @xmath58 . renormalization factor in the @xmath59 , ndr scheme are calculated as @xmath60 . we plot renormalized value @xmath61 as a function of @xmath62 $ ] in figure [ bkren_sum ] , where circles are from @xmath41 and squares from @xmath17 . using the chiral expansion @xcite @xmath63 + \xi_1 m_{ps}^2 , \label{bkratio_clog}\end{aligned}\ ] ] and our results of @xmath64 for the coefficient of the chiral log , we obtain the solid and dashed curves for @xmath14 and @xmath17 with @xmath65dof = 0.80 and 0.08 , respectively . result of @xmath66 for each @xmath67 from the interpolation to @xmath68 are shown by the filled symbols in figure [ bk ] . in the same figure , we plot results of cp - pacs @xcite using iwasaki gauge action with two set of similar parameters as ours , _ i.e. _ @xmath69 and @xmath70 with the open squares . on the other hand , the open circle corresponds to the previous result of rbc @xcite using wilson gauge action with @xmath71 . for @xmath72 2 gev , results from these three kinds of calculations distribute over the width of more than 10% . on the other hand , for @xmath73 gev , results from both collaboration agree with each other . in the continuum limit taken by the naive extrapolation of our two results , we extract @xmath74 and @xmath75 . 15 see contributions of c. dawson , t. izubuchi to these proceedings . y. aoki _ et al . _ ( rbc collaboration ) , phys . rev . d69 ( 2004 ) 074504 . j. noaki for rbc collaboration , nucl . * b * ( proc . suppl . ) 119 ( 2003 ) 362 . c. bernard and m. f. l. golterman , phys . rev . d46 ( 1992 ) 853 . g. martinelli _ et al . _ , nucl . phys . * b445 * ( 1995 ) 81 ; t. blum _ et al . _ ( rbc collaboration ) , phys . rev . d66 ( 2002 ) 014504 . rbc collaboration , in preparation . m. ciuchini _ et al . _ , nucl . phys . * b523 * ( 1998 ) 501 . sharpe , s. , phys . rev . d46 ( 1992 ) 3146 . a. ali khan _ et al . _ ( cp - pacs collaboration ) , phys . rev . d64 ( 2001 ) 114506 . t. blum _ et al . _ ( rbc collaboration ) , phys . rev . d68 ( 2003 ) 114506 .	we present calculations of the decay constants and kaon b - parameter @xmath0 as the first stage of rbc collaboration s quenched numerical simulations using dbw2 gauge action and domain - wall fermions . some of potential systematic errors and consistency to previous works are discussed .
the production of charmonium states at high - energy colliders has been the subject of considerable interest during the past few years . new results from @xmath2 , @xmath3 and @xmath4 experiments have become available , some of which revealed dramatic shortcomings of earlier quarkonium production models . in theory , progress has been made on the factorization between the short distance physics of heavy - quark creation and the long - distance physics of bound state formation . the colour - singlet model @xcite has been superseded by a consistent and rigorous framework , based on the use of non - relativistic qcd ( nrqcd ) @xcite , an effective field theory that includes the so - called colour - octet mechanisms . on the other hand , the colour evaporation model @xcite of the early days of quarkonium physics has been revived @xcite . however , despite the recent theoretical and experimental developments the range of applicability of the different approaches is still subject to debate , as is the quantitative verification of factorization . because the charmonium mass is still not very large with respect to the qcd scale , non - factorizable corrections @xcite may not be suppressed enough and the expansions in nrqcd may not converge very well . in this situation cross checks between various processes , and predictions of observables such as quarkonium polarization and differential cross sections , are crucial in order to assess the importance of different quarkonium production mechanisms , as well as the limitations of a particular theoretical approach . among the specific predictions of nrqcd , transverse polarization of direct ( i.e. not from @xmath5 and @xmath6 decays ) @xmath0 and @xmath1 hadroproduction at large transverse momentum has emerged as the most distinct and most accessible signature @xcite . transverse polarization also discriminates nrqcd from other approaches , like the colour evaporation model which predicts the quarkonium to be produced unpolarized . in the following , i will briefly review the general mechanism of quarkonium production at hadron colliders in the nrqcd factorization approach . the prediction of @xmath0 and @xmath1 transverse polarization at large transverse momentum is confronted with recent experimental data from the tevatron , and potential shortcomings of the theoretical analysis are discussed . in the nrqcd approach , the cross section for producing a quarkonium state @xmath7 at a hadron collider is written as a sum of factorizable terms , @xmath8 + x)\,\langle { \cal{o}}^{h}\,[n]\rangle , \end{aligned}\ ] ] where @xmath9 denotes the colour , spin and angular momentum state of an intermediate @xmath10 pair . the short - distance cross section @xmath11 can be calculated perturbatively in the strong coupling @xmath12 . the nrqcd matrix elements @xmath13 \rangle$ ] ( see @xcite for their definition ) are related to the non - perturbative transition probabilities from the @xmath10 state @xmath9 into the quarkonium @xmath7 . they scale with a definite power of the intrinsic heavy - quark velocity @xmath14 @xcite . ( @xmath15 for charmonium and @xmath16 for bottomonium . ) the general expression ( [ eq_fac ] ) is thus a double expansion in powers of @xmath12 and @xmath14 . the nrqcd formalism implies that so - called colour - octet processes associated with higher fock state components of the quarkonium wave function must contribute to the cross section . heavy quark pairs that are produced at short distances in a colour - octet state can evolve into a physical quarkonium through radiation of soft gluons at late times in the production process , when the quark pair has already expanded to the quarkonium size . the production of @xmath17-wave charmonium in @xmath2 collisions at the tevatron has attracted considerable attention and has stimulated much of the recent theoretical development in quarkonium physics . the cdf collaboration has measured cross sections for the production of @xmath0 and @xmath1 states not coming from @xmath5 or radiative @xmath6 decays , for a wide range of transverse momenta @xmath18 @xcite . surprisingly , the experimental cross sections were found to be orders of magnitudes larger than the theoretical expectation based on the leading - order colour - singlet model @xcite . this result is particularly striking because the data extends out to large transverse momenta where the theoretical analysis is rather clean and non - factorizable corrections should be suppressed . the shortcoming of the colour - singlet model can be understood by examining a typical feynman diagram contributing to the leading - order parton cross section , fig.1(a ) . at large transverse momentum , the two internal figure 1 . generic diagrams for @xmath0 production in hadron - hadron collisions via colour - singlet and colour - octet channels . + quark propagators are off - shell by @xmath19 so that the parton differential cross section scales like @xmath20 , as indicated in the figure . on the other hand , when @xmath21 the quarkonium mass can be considered small and the inclusive charmonium cross section is expected to scale like any other single - particle inclusive cross section @xmath22 . the dominant production mechanism for charmonium at sufficiently large @xmath23 must thus be via fragmentation @xcite , the production of a parton with large @xmath23 which subsequently decays into charmonium and other partons . a typical fragmentation contribution to colour - singlet @xmath0 production is shown in fig.1(b ) . while the fragmentation contributions are of higher order in @xmath12 compared to the fusion process fig.1(a ) , they are enhanced by a power @xmath24 at large @xmath23 and can thus overtake the fusion contribution at @xmath25 . when colour - singlet fragmentation is included , the @xmath23 dependence of the theoretical prediction is in agreement with the tevatron data but the normalization is still underestimated by about an order of magnitude @xcite , indicating that an additional fragmentation contribution is still missing . it is now generally believed that gluon fragmentation into colour - octet @xmath26 charm quark pairs @xcite , as shown in fig.1(c ) , is the dominant source of @xmath0 and @xmath1 at large @xmath23 at the tevatron . the probability of forming a @xmath27 particle from a pointlike @xmath28 pair in a colour - octet @xmath26 state is given by the nrqcd matrix element @xmath29 \rangle$ ] which is suppressed by @xmath30 relative to the non - perturbative factor of the leading colour - singlet term . however , this suppression is more than compensated by the gain in two powers of @xmath31 in the short - distance cross section for producing colour - octet @xmath26 charm quark pairs as compared to colour - singlet fragmentation . at @xmath32 in the velocity expansion , two additional colour - octet channels have to be included , fig.1(d ) , which do not have a fragmentation interpretation at order @xmath33 but which become significant at moderate @xmath34 @xcite . the importance of the @xmath35 and @xmath36 contributions can not be estimated from naive power counting in @xmath12 and @xmath14 alone , but rather follows from the dominance of @xmath37-channel gluon exchange , forbidden in the leading - order colour - singlet cross section . the different contributions to the @xmath0 transverse momentum distribution are compared to the cdf data @xcite in fig.2 . as mentioned above , the colour - singlet model at lowest order in @xmath12 fails dramatically when confronted with the experimental results . when colour - singlet fragmentation is included , the prediction increases by more than an order of magnitude at large @xmath23 , but it still falls below the data by a factor of @xmath38 . the cdf results on charmonium production can be explained by including the leading colour - octet contributions and adjusting the unknown non - perturbative parameters to fit the data . numerically one finds the non - perturbative matrix elements to be of @xmath39 , see e.g.@xcite , perfectly consistent with the @xmath30 suppression expected from the velocity scaling rules . similar conclusions can be drawn for @xmath1 production at the tevatron . colour - singlet and colour - octet contributions to direct @xmath0 production in @xmath40 at the tevatron together with experimental data from cdf @xcite . parameters and nrqcd matrix elements as specified in @xcite . the analysis of the cdf data @xcite alone , although very encouraging , does not strictly prove the phenomenological relevance of colour - octet contributions because free parameters have to be introduced to fit the data . however , if factorization holds the non - perturbative matrix elements extracted from the @xmath0 and @xmath1 cross sections are universal and can be used to make predictions for various processes and observables . the single most crucial test of the nrqcd approach to charmonium production at hadron colliders is the analysis of @xmath1 polarization at large transverse momentum . recall that at large @xmath23 , @xmath1 production should be dominated by gluon fragmentation into a colour - octet @xmath26 charm quark pair , fig.1(c ) . when @xmath25 the fragmenting gluon is effectively on - shell and transverse . the intermediate @xmath28 pair in the colour - octet @xmath26 state inherits the gluon s transverse polarization and so does the quarkonium , because the emission of soft gluons during hadronization does not flip the heavy quark spin at leading order in the velocity expansion . consequently , at large transverse momentum one should observe transversely polarized @xmath1 @xcite . the polarization can be measured through the angular distribution in the decay @xmath41 , given by @xmath42 , where @xmath43 denotes the angle between the lepton three - momentum in the @xmath44 rest frame and the @xmath44 three - momentum in the lab frame . pure transverse polarization implies @xmath45 . corrections to this asymptotic limit due to higher order fragmentation contributions have been estimated to be small @xcite . the dominant source of depolarization comes from the colour - octet fusion diagrams , fig.1(d ) , which are important at moderate @xmath23 . still , at @xmath32 in the velocity expansion , the polar angle asymmetry @xmath46 can be unambiguously calculated within nrqcd @xcite in terms of the non - perturbative matrix elements that have been determined from the unpolarized cross section . in fig.3 we display the theoretical prediction @xcite ( shaded band ) for @xmath46 in @xmath1 production at the tevatron as function of the @xmath1 transverse momentum , taking into account the @xmath35 and @xmath36 fusion channels and higher - order corrections to the fragmentation contributions . no transverse polarization is expected at @xmath47 gev , but the angular distribution is predicted to change drastically as @xmath23 increases . a first measurement from cdf @xcite does not support this prediction , but the experimental errors are too large to draw definite conclusions . the analysis of @xmath0 polarization is complicated by the fact that the experimental data sample @xcite includes @xmath0 that have not been produced directly but come from decays of @xmath6 and @xmath1 mesons . the contribution from the radiative decays of the higher excited states decreases , but does not eliminate , the transverse @xmath0 polarization at large @xmath23 @xcite . again , the nrqcd factorization prediction is not supported by the experimental data , see fig.4 . polar angle asymmetry @xmath46 for direct @xmath1 production in @xmath48 at the tevatron as a function of @xmath23 compared to data from cdf @xcite . see @xcite for details and the choice of parameters . also shown is the prediction for two extreme choices of nrqcd matrix elements : @xmath49 \rangle = 0 $ ] ( upper dashed line ) and and @xmath50 \rangle = 0 $ ] , @xmath51 \rangle = 5\cdot 10^{-2}~\mbox{gev}^3 $ ] , @xmath52 \rangle = 5\cdot 10^{-4}~\mbox{gev}^3 $ ] ( lower dotted line ) , as discussed in the text . + figure 4 . polar angle asymmetry @xmath46 for prompt @xmath0 production ( i.e. not from b - decay ) in @xmath53 at the tevatron as a function of @xmath23 compared to data from cdf @xcite . from @xcite . the absence of transversely polarized @xmath0 and @xmath1 at large transverse momentum , if confirmed at the tevatron run ii , represents a challenge for the application of nrqcd factorization to charmonium production . in the following , i will discuss some possible uncertainties and potential shortcomings of the present theoretical analysis : + @xmath54 _ higher - order qcd effects _ in the short - distance cross section should be included to improve the accuracy of the theoretical prediction , in particular in the intermediate @xmath23 region . while no attempt has been made to estimate the impact of higher - order corrections on the spin - dependence of the cross section , preliminary studies indicate large nlo effects for the unpolarized @xmath23-distributions @xcite . these corrections could strongly affect the determination of the nrqcd matrix elements from the tevatron data . a full nlo analysis is however needed before quantitative conclusions can be drawn . the impact of non - vanishing transverse momentum of the incoming partons on the determination of the nrqcd matrix elements has been analysed using a gaussian smearing of the @xmath23 distribution @xcite , monte carlo event generators @xcite and , most recently , the @xmath55-factorization formalism @xcite . the actual size of the effect , however , turns out to be very different for the different approaches studied in the literature . + @xmath54 the _ uncertainty in the determination of nrqcd matrix elements _ translates into an uncertainty in the predicted yield of transversely polarized @xmath0 and @xmath1 . the extraction of nrqcd matrix elements from the unpolarized cross section at the tevatron is affected by large theoretical uncertainties which have not yet been fully quantified , see e.g. @xcite . in fig.3 we also show the polarization prediction for the extreme choices of vanishing @xmath56 \rangle$ ] matrix elements ( upper dashed line ) and a very small @xmath52 \rangle$ ] matrix element ( lower dotted line , see figure caption ) . the latter choice is motivated by the recent analyses of @xmath0 and @xmath1 production at the tevatron in the @xmath55-factorization approach @xcite which predict a strongly enhanced colour - singlet contribution and a substantially decreased value of @xmath52 \rangle$ ] @xcite . a decreased value of the @xmath57 \rangle$ ] matrix element , relative to @xmath56 \rangle$ ] , would delay the onset of transverse polarization and reduce the discrepancy between nrqcd predictions and data . however , such extreme values for the nrqcd matrix elements would indicate a flaw in our understanding of the velocity scaling rules @xcite which imply that all three matrix elements are of approximately equal size @xmath58 . + @xmath54 _ heavy quark spin - symmetry _ is violated by higher - order terms in the nqrcd lagrangian and longitudinal polarization can arise if the binding of the charm quark pair into @xmath0 and @xmath1 proceeds through two chromo - magnetic dipole transitions . according to the velocity scaling rules these transitions are suppressed by @xmath30 and the corresponding corrections to transverse polarization should not exceed @xmath59 . predictions of heavy quark spin - symmetry for radiative charmonium decays agree quite well with experimental data @xcite and there is no obvious reason why spin - symmetry should fail when applied to charmonium production . still , the estimate of the spin - symmetry breaking corrections to transverse polarization relies on the conventional velocity scaling rules @xcite which have not been firmly established for the case of charmonium @xcite . alternative velocity scaling rules @xcite imply a hierarchy of nrqcd matrix elements different from the standard counting and might indicate larger spin - symmetry violating corrections . + @xmath54 finally , charmonium production mechanisms _ beyond nrqcd _ could be responsible for @xmath0 and @xmath1 depolarization . for example , the leading - twist formalism of nrqcd factorization does not include possible rescattering interactions between the intermediate heavy quark pair and a comoving colour field . it has been shown that such rescattering corrections could yield unpolarized quarkonium @xcite . the analysis @xcite , however , relies on several simplifying assumptions , and further work is needed to establish the importance of comover interactions for charmonium production at large @xmath23 . the absence of a transverse polarization in @xmath0 and @xmath1 hadroproduction at large @xmath23 represents a serious challenge for the application of the nrqcd factorization approach to charmonium production . several potential problems with the current theoretical analysis have been discussed but no conclusive picture has emerged yet . if no transverse polarization is observed with higher statistics at larger values of @xmath23 in run ii of the tevatron , one might have to conclude that the charm quark mass is not large enough for a non - relativistic approach to work in all circumstances . _ note added _ + after submission of this proceedings contribution , charmonium polarization at the tevatron has been addressed in the @xmath55-factorization approach @xcite . the authors argue that the @xmath35 production channel could be the dominant source of @xmath0 and @xmath1 in the experimentally accessible region of transverse momentum , which implies that quarkonium be produced mainly unpolarized . _ acknowledgements _ + i would like to thank martin beneke for his collaboration and the conference organizers for creating a pleasant and stimulating atmosphere . a conference grant by the royal society is gratefully acknowledged . e. l. berger and d. jones , phys . * d23 * ( 1981 ) 1521 ; r. baier and r. rckl , phys . lett . * b102 * ( 1981 ) 364 . g. t. bodwin , e. braaten and g. p. lepage , phys . * d51 * ( 1995 ) 1125 ; erratum _ ibid . _ * d55 * ( 1997 ) 5853 . h. fritzsch , phys . lett . * 67b * ( 1977 ) 217 ; f. halzen , phys . lett . * 69b * ( 1977 ) 105 ; m. glck , j. f. owens and e. reya , phys . rev . * d17 * ( 1978 ) 2324 . g. a. schuler and r. vogt , phys . * b387 * ( 1996 ) 181 ; j. f. amundson , o. j. eboli , e. m. gregores and f. halzen , phys . * b390 * ( 1997 ) 323 ; a. edin , g. ingelman and j. rathsman , phys . * d56 * ( 1997 ) 7317 . s. j. brodsky , int . j. mod . * a12 * ( 1997 ) 4087 ; p. hoyer , nucl . suppl . * 75b * ( 1999 ) 153 . p. cho and m. b. wise , phys . * b346 * ( 1995 ) 129 . g. p. lepage , l. magnea , c. nakhleh , u. magnea and k. hornbostel , phys . * d46 * ( 1992 ) 4052 . f. abe _ et al . _ [ cdf collaboration ] , phys . lett . * 69 * ( 1992 ) 3704 . f. abe _ et al . _ [ cdf collaboration ] , phys . * 79 * ( 1997 ) 572 . r. baier and r. rckl , z. phys . * c19 * ( 1983 ) 251 ; r. gastmans , w. troost and t. t. wu , nucl . phys . * b291 * ( 1987 ) 731 . e. braaten and t. c. yuan , phys . rev . * 71 * ( 1993 ) 1673 . m. cacciari and m. greco , phys . * 73 * ( 1994 ) 1586 ; e. braaten , m. a. doncheski , s. fleming and m. l. mangano , phys . * b333 * ( 1994 ) 548 ; d. p. roy and k. sridhar , phys . * b339 * ( 1994 ) 141 . e. braaten and s. fleming , phys . * 74 * ( 1995 ) 3327 . p. cho and a. k. leibovich , phys . * d53 * ( 1996 ) 150 ; phys . rev . * d53 * ( 1996 ) 6203 . m. krmer , f. maltoni and m. a. sanchis lozano , in bottom production , p. nason , g. ridolfi , o. schneider , g.f . tartarelli , p. vikas et al . , hep - ph/0003142 , published in cern - yr-2000/01 , g. altarelli and m.l . mangano editors . a. k. leibovich , these proceedings [ hep - ph/0008236 ] . m. beneke and i. z. rothstein , phys . * b372 * ( 1996 ) 157 . ; erratum _ ibid . _ * d54 * ( 1997 ) 7082 . m. beneke and m. krmer , phys . rev . * d55 * ( 1997 ) 5269 . a. k. leibovich , phys . * d56 * ( 1997 ) 4412 . t. affolder _ et al . _ [ cdf collaboration ] , phys . * 85 * ( 2000 ) 2886 . e. braaten , b. a. kniehl and j. lee , phys . * d62 * ( 2000 ) 094005 ; b. a. kniehl and j. lee , hep - ph/0007292 . a. petrelli , nucl . * 86 * ( 2000 ) 533 . b. cano - coloma and m. a. sanchis - lozano , nucl . * b508 * ( 1997 ) 753 . f. yuan and k. chao , hep - ph/0008302 . p. hgler et al . , hep - ph/0008316 . a. pineda , these proceedings [ hep - ph/0008327 ] . m. beneke , hep - ph/9703429 . g. a. schuler , int . j. mod . phys . * a12 * ( 1997 ) 3951 . n. marchal , s. peigne and p. hoyer , hep - ph/0004234 . f. yuan and k. chao , hep - ph/0009224 .	the nrqcd factorization approach for quarkonium production at hadron colliders is reviewed . the prediction of @xmath0 and @xmath1 transverse polarization at large transverse momentum is confronted with recent experimental data , and potential shortcomings of the theoretical analysis are discussed .
let @xmath0 be a compact complex curve and @xmath1 be a meromorphic @xmath2-form on @xmath3 in an open neighborhood of each point @xmath4 we can write @xmath5 where @xmath6 is a local normalizing parameter at @xmath7 the coefficient @xmath8 of the above series does not depend on the choice of parameter @xmath6 and is called the residue of @xmath1 at @xmath7 the residue is non - zero only at finitely many points @xmath9 where @xmath1 has a pole . the well - known residue formula says that the sum of residues of @xmath1 over all points of @xmath10 is zero : @xmath11 indeed , the residue at @xmath12 is equal to the integral of @xmath1 over any sufficiently small cycle enclosing @xmath13 divided by @xmath14 in the complement @xmath15 the form @xmath1 is closed , and the sum of cycles is homologous to zero . thus , the residue formula follows from the stokes theorem . although this proof is topological , the residue itself can be defined purely algebraically and one can give an algebraic proof of the residue formula which works in a much more general situation , not only in the case of complex curves ( see , for example , @xcite , @xcite ) . in the late @xmath16 s a. parshin introduced his notion of multidimensional residue for a rational @xmath17-form @xmath1 on an @xmath17-dimensional algebraic variety @xmath18 ( in @xcite parshin mostly deals with the two - dimensional case , then a. beilinson and v. lomadze in @xcite and @xcite generalized parshin s ideas to the multidimensional case ) . the main difference between the parshin residue and the classical one - dimensional residue is that in higher dimensions one computes the residue not at a point , but at a complete flag of subvarieties @xmath19 @xmath20 parshin , beilinson , and lomadze proved the reciprocity law for multidimensional residues , which generalizes the classical residue formula : fix a partial flag of irreducible subvarieties @xmath21 , where @xmath22 is omitted ( @xmath23 ) . then @xmath24 where the sum is taken over all irreducible @xmath25-dimensional subvarieties @xmath26 such that @xmath27 more precisely , the statement of the theorem is that there are only finitely many summands which are not zeros , and the sum of non - zero summands is zero . in addition , if @xmath28 is proper ( compact in the complex case ) , then one has the same relation for @xmath29 @xmath30 again , there are finitely many non - zero summands and their sum is zero . all these papers are purely algebraic . the methods used by parshin , beilinson and lomadze are applicable in very general settings , not only over complex numbers . however , in the complex case one would expect a more geometric variant of the theory . brylinski and d. a. mclaughlin give a more topological treatment of the complex case in @xcite . given a flag @xmath31 they introduce _ flag - localized _ homology groups @xmath32 and a homology class @xmath33 such that @xmath34 for any meromorphic @xmath17-form @xmath35 the class @xmath36 is obtained from of the fundamental class @xmath37 by applying the boundary homomorphisms in the appropriate flag - localized homology groups @xmath17 times . brylinski and d. a. mclaughlin mention , that the class @xmath36 could be constructed in a more geometric way , so that it is naturally represented by a union of certain real @xmath17-tori . however , they only give such a construction in the case when all elements of the flag @xmath38 are smooth . in this paper we develop a different approach to the construction of the class @xmath39 we use the geometry of the whitney stratified spaces to introduce the leray coboundary operators @xmath40 for any two consecutive strata @xmath41 ( see definition [ order on strata ] ) of a stratified space . given a flag @xmath31 and a meromorphic top - form @xmath1 on @xmath42 one can choose a stratification of @xmath42 such that the flag @xmath38 consists of closures of strata , and @xmath1 is regular on the top - dimensional stratum . then one can construct the homology class @xmath43)\in h_n(\breve v_n)$ ] ( here @xmath44 is the unique @xmath25-dimensional stratum in @xmath22 ) . in the section [ section reciprocity law ] we prove that @xmath45 the construction of the leray coboundary operators is very geometric . in particular , the class @xmath46 is naturally represented by a smooth submanifold @xmath47 which is a union of smooth @xmath17-dimensional tori @xmath48 . in fact , in the original parshin s construction , the residue at the flag @xmath38 is defined as a sum of certain more delicate residues ( we briefly review parshin s definitions in the section [ section parshin s theory reviewed ] ) . we show , that the tori @xmath49 naturally correspond to the summands in the parshin s definition . let @xmath50 be the algebraic surface , given by the equation @xmath51 consider the flag @xmath52 where @xmath53 is the surface @xmath54 @xmath28 is the @xmath55-axis ( which is the singular locus of @xmath56 ) , and @xmath57 is the origin . the intersection of @xmath56 with the real plane is the cone over a figure eight ( see figure [ figure eight ] ) . the real picture helps a lot to visualize this example . with the real space.,width=377 ] there is a natural stratification of @xmath54 consisting of @xmath58 strata : the origin , the @xmath55-axis without the origin , and the regular part of @xmath59 following the above notations , we denote the strata @xmath60 @xmath61 and @xmath62 correspondingly . @xmath57 is a point on the complex line @xmath63 one can consider a small circle @xmath64 going counterclockwise around @xmath57 on @xmath63 @xmath64 naturally represents the class @xmath65)\in h_1(\breve v_1).$ ] on the next step , we have two branches of @xmath53 at each point of @xmath66 take a point @xmath67 consider a transversal section to @xmath28 through @xmath7 its intersection with @xmath53 is a curve with two local branches at @xmath7 consider two small circles @xmath68 and @xmath69 around @xmath13 one on each branch . one can choose transversal sections to @xmath28 at each point of @xmath70 so that they depend nicely on the point @xmath67 furthermore , one can choose the circles in such a way that they form a fiber bundle over @xmath66 easy to see , that the local branches of @xmath53 do not interchange , as one goes around the origin along @xmath64 . therefore , one gets two tori @xmath71 and @xmath72 in @xmath73 the parshin residue of a meromorphic @xmath74-form @xmath1 on @xmath56 can be computed via integral over @xmath75 @xmath76 consider the whitney umbrella the surface @xmath77 given by the equation @xmath78 consider the flag @xmath52 where , again , @xmath53 is the surface , @xmath28 is the @xmath55-axis , and @xmath57 is the origin ( figure [ umbrella ] ) . notice , that @xmath28 again coincide with the singular locus of @xmath59 with the real space.,width=377 ] in the same way as in the previous example , one can consider a small loop around the origin on the @xmath63 and , again , @xmath53 has two branches at each point of the loop . however , as one goes around the origin on @xmath79 the branches interchange . therefore , the class @xmath80 is represented by just one torus in this case , and there is only one summand in the parshin s definition of the residue . coboundary operators satisfy an interesting relation : let @xmath81 be two strata such that there are @xmath25 intermediate strata @xmath82 and these intermediate strata are incomparable ( equivalently , for any @xmath83 from @xmath2 to @xmath84 @xmath85 are consecutive strata ) . then @xmath86 ( see theorem [ relation ] . ) one can illustrate this relation on an example : [ example relation ] on the figure [ picture relation ] , @xmath0 is the origin , @xmath87 is a half - line , @xmath88 is a surface with an isolated singularity at the origin , and @xmath89 we take a small sphere @xmath90 with center at the origin . then @xmath91)\in h_{\dim z_i -1}(z_i)$ ] is represented by the intersection @xmath92 take a small neighborhood of @xmath93 in @xmath94 its boundary @xmath95 represents the class @xmath96)\in h_1(y).$ ] then the sphere @xmath90 with the neighborhoods of @xmath93 s deleted gives a two - dimensional chain in @xmath97 which boundary is the union @xmath98 .,height=264 ] our approach also allows us to prove an interesting result about parshin residues : let @xmath1 be a meromorphic top - form on @xmath18 consider any whitney stratification of @xmath99 such that @xmath1 is regular on the top - dimensional stratum . then the residue @xmath100 could be non - trivial , only if all elements of the flag @xmath38 are closures of strata of the stratification ( theorem [ criteria for non - trivial residues ] ) . in particular , there are only finitely many non - trivial residues for a given form . * structure of the paper . * in the first part of the paper we introduce the leray coboundary operators for stratified spaces and prove the relation ( theorem [ relation ] ) . in the second part we use the results of the first part to express the parshin residue as an integral over a real smooth cycle and to prove the reciprocity law . in the section [ section strat review ] we give a short introduction to the theory of stratified spaces . in the section [ section parshin s theory reviewed ] we review the original parshin s definitions and the formulation of the reciprocity law . * acknowledgments . * author would like to thank his phd advisor , professor askold khovanskii , for raising the question and helpful discussions . let @xmath101 be a smooth manifold . let @xmath102 be a locally closed subset of @xmath103 by a _ whitney stratification _ @xmath104 of @xmath105 we mean a subdivision of @xmath102 into smooth strata , such that : 1 . it is _ locally finite _ - each point of @xmath102 has an open neighborhood which intersects only finitely many strata . 2 . _ condition of the frontier _ - for each stratum @xmath106 its boundary @xmath107 is a union of strata . 3 . each pair @xmath108 of strata satisfies _ whitney conditions _ * a * and * b * : 1 . for any @xmath109 and any sequence @xmath110 such that @xmath111 if the sequence of tangent planes @xmath112 converges to some plane @xmath113 ( in the appropriate grassmanian bundle over @xmath101 ) , then @xmath114 2 . for any @xmath115 any sequence @xmath110 and any sequence @xmath116 such that @xmath117 and @xmath118 if the sequence of tangent planes @xmath112 converges to some plane @xmath119 and the sequence of secants @xmath120 converges to some line @xmath121 ( in some smooth coordinate system in @xmath101 ) , then @xmath122 * remark . * actually , condition * b * implies condition * a * , so it is enough to require condition * b. one can prove , that if a pair of strata @xmath108 satisfies condition b * and @xmath123 , then @xmath124 * notation . * we say , that @xmath81 if @xmath125 one can see that this defines a partial order on the set of strata @xmath126 consider the surface in @xmath127 given by the equation @xmath128 ( see figure [ whitney example ] ) . the singular locus of the surface coincide with the @xmath55-axis . thus , the @xmath55-axis and its complement gives a subdivision of the surface in two smooth pieces . easy to prove that this pair satisfies the condition * a , * but does nt satisfy condition * b * at the origin . note , that the small neighborhood of the origin looks very different from the neighborhood of any other point of the @xmath55-axis . it is easy to improve the subdivision in such a way that it satisfies condition * b * : one only needs to consider the origin as a separate stratum . whitney showed that if conditions * a * and * b * are satisfied for the pair @xmath108 , then @xmath129 `` behaves regularly '' along @xmath3 let @xmath102 be a closed subvariety in a smooth algebraic variety @xmath103 let @xmath10 be a locally finite family of subvarieties in @xmath130 then there exists a whitney stratification of the set @xmath102 such that each element of @xmath10 is a union of strata and all strata are algebraic . detailed review of the theory of whitney stratifications can be found in @xcite . the notion of an abstract stratified space , introduced by john mather in @xcite , provides a convenient setup for working with `` nice '' stratifications : subdivisions into smooth pieces with regular behaviour along strata . j. mather proved , that any whitney stratification can be endowed with a structure of an abstract stratified space . below we introduce the notion of an abstract stratified space . let @xmath102 be a hausdorff , locally compact topological space , satisfying the second countability axiom ( i.e. there exist a countable basis in the topology of @xmath102 ) . let @xmath131 be a locally finite subdivision of @xmath102 into topological manifolds , endowed with smoothness structures . the elements of @xmath56 are called strata . let @xmath131 satisfy the condition of the frontier ( i.e. the boundary of any stratum is a union of strata ) . similarly as for whitney stratifications , the set of strata @xmath131 inherits the natural partial order ( @xmath81 if @xmath132 ) . for every @xmath133 let @xmath134 be a neighborhood of @xmath0 in @xmath105 @xmath135 be a continuous function , and @xmath136 be a retraction . one should think of @xmath137 as of the distance to @xmath3 therefore , we require that @xmath138 it is also convenient to say that @xmath139 if @xmath140 we call @xmath137 the tubular function , and @xmath134 the tubular neighborhood . let @xmath141 @xmath142 we use the following notations : @xmath143 @xmath144 @xmath145 we assume that @xmath146 is empty unless @xmath147 we also assume that if @xmath0 and @xmath129 are incomparable , then @xmath148 is empty . we have the following compatibility conditions : @xmath149 @xmath150 whenever both sides of these equations are defined . the following conditions ensure that the space @xmath102 behaves regularly along strata : for any @xmath151 the map @xmath152 is a locally trivial fibration with a compact fiber . moreover , for any @xmath153 the restriction @xmath154 is a smooth fibration . finally , we want the fiber of @xmath155 over a point @xmath4 to be a cone with the vertex at @xmath7 it does not follow from the above conditions . so , we need to add one more : let @xmath156 let @xmath157 then @xmath158 is the mapping cone over @xmath159 the triple @xmath160 is called _ control data_. the triple @xmath161 with the above conditions is called an _ abstract stratified space_. it follows that @xmath162 has a natural structure of an abstract stratified space , obtained by intersecting the strata of @xmath102 with @xmath162 and restricting tubular functions and retractions . abstract stratified spaces were introduced by john mather in @xcite . in fact , the original definition is slightly different and less restrictive . however , it follows easily that by shrinking the tubular neighborhoods and rescaling the tubular functions , one can change the control data so that they satisfy the above conditions . let @xmath163 be a smooth fibration with compact oriented @xmath25-dimensional fiber @xmath164 then one can define the gysin homomorphism on homology @xmath165 basically , one just set @xmath166,$ ] where @xmath167 is a representative of the homology class @xmath168 * remark . * we use the following convention about the orientations . let @xmath169 and @xmath170 let @xmath171 be a smooth representative of a homology class @xmath172 and @xmath173 let the differential form @xmath174 on @xmath175 be such that its restriction to @xmath167 defines the orientation of @xmath167 at @xmath176 and the differential form @xmath177 on @xmath101 be such that its restriction to the fiber @xmath178 defines the orientation of @xmath178 at @xmath179 then the orientation of the preimage @xmath180 at the point @xmath181 is given by the restriction of the form @xmath182 let now @xmath101 be an oriented manifold with boundary , and @xmath163 be a proper map to an oriented manifold @xmath175 , such that its restriction both to the boundary @xmath183 and the interior @xmath184 are submersions . then , by the ehresmann lemma for manifolds with boundary , @xmath185 is a locally trivial fibration and its restrictions to @xmath186 and @xmath187 are smooth fibrations . let @xmath188 be the gysin homomorphism . [ relation for gysin maps ] @xmath189 where @xmath190 is the embedding . one can generalize the gysin homomorphism to the described above case , when the fiber of @xmath185 is a manifold with boundary . the only difference is that now the homomorphism lands in the relative homology group : @xmath191 where @xmath192 then one immediately sees that @xmath193 where @xmath194 is the boundary homomorphism from the long exact sequence of the pair @xmath195 however , by the long exact sequence , @xmath196 we apply the above constructions to the stratified spaces . let all the strata of a stratified space @xmath102 be oriented . let @xmath106 be a stratum . restriction of the retraction @xmath136 to @xmath197 is a locally trivial fibration . moreover , for any stratum @xmath129 such that @xmath81 the restriction to @xmath198 is a smooth fibration . [ order on strata ] let @xmath81 be two strata . we say that @xmath81 are _ consecutive strata _ if there is no such @xmath199 that @xmath200 let @xmath81 be consecutive strata . then the fiber of @xmath201 is compact . since @xmath81 are consecutive strata , it follows that @xmath202 is a closed stratum of @xmath162 ( indeed , otherwise the closure of @xmath203 in @xmath162 would contain a smaller stratum ) . the fiber of the restriction of @xmath155 to @xmath203 is the intersection of the fiber of the restriction of @xmath155 to @xmath162 and @xmath204 therefore , it is compact as a closed subset of a compact set . note , that @xmath203 is orientable . indeed , it is the level set of a smooth function @xmath205 in @xmath206 let us fix the orientation of @xmath203 given as follows : we say that the restriction a differential @xmath207-form @xmath208 on @xmath129 defines the positive orientation of @xmath203 if the form @xmath209 defines the positive orientation of @xmath210 let @xmath211 and @xmath212 the _ leray coboundary operator _ @xmath40 is given by the composition @xmath213 where @xmath214 is the embedding and @xmath215 is the gysin homomorphism . [ relation ] let @xmath81 be two strata . let @xmath216 be all strata such that @xmath217 suppose that @xmath216 are incomparable . then @xmath218 we want to apply the lemma [ relation for gysin maps ] . consider @xmath219 note , that @xmath220 therefore , @xmath221 is a smooth fibration over @xmath222 denote @xmath223 according to the construction of the leray coboundary operators , we have @xmath224 where @xmath225 is the embedding and @xmath226 is the gysin homomorphism of @xmath227 here we fix the orientation of @xmath95 given in the following way : we say that the restriction of a differential @xmath228-form @xmath229 on @xmath129 defines the positive orientation of @xmath95 if the form @xmath230 defines the positive orientation of @xmath210 consider now @xmath231 the restriction @xmath232 is a smooth fibration . however , the fibers of this fibration are not compact . on the other side , if we consider the restriction of @xmath155 to the union @xmath233 then the fibers are compact . @xmath234 can be thought of as the boundary of the neighborhood @xmath235 of @xmath236 in @xmath237 denote @xmath238 by ehresmann lemma for manifolds with boundary , the restriction @xmath239 is a locally trivial fibration . indeed , @xmath240 is proper , because @xmath101 is a closed subset of @xmath241 and @xmath242 is a fibration with compact fibers ; the restrictions of @xmath155 to the interior of @xmath101 and the boundary @xmath243 are submersions . to conclude the proof by lemma [ relation for gysin maps ] , one needs to check , that the orientation of @xmath95 as a piece of the boundary of @xmath101 always coincide ( or always is opposite ) with the orientation of @xmath95 used in the first part of the proof . indeed , we fixed the orientation of @xmath95 in such a way , that if @xmath244 gives the orientation of @xmath95 then @xmath245 gives the orientation of @xmath210 let @xmath246 according to our convention about the orientation of @xmath247 @xmath208 gives the positive orientation of @xmath204 therefore , the orientation of @xmath95 as a piece of the boundary of @xmath101 is given by @xmath248 in this chapter the coefficient ring is always @xmath249 . for simplicity , we skip it in the notations . there is a natural question : _ which operator is poincare dual to the coboundary operator @xmath250 _ the manifolds @xmath0 and @xmath129 are not compact . therefore , one has to use the borel - moore homology to do the poincare duality . a nice review of the theory of borel - moore homology ( in much more details than needed here ) is given in @xcite . consider @xmath251 ( here @xmath252 @xmath253 ) . the dual operator is @xmath254 there is a natural candidate for the dual : indeed , one can show , that @xmath255 therefore , there exits the boundary operator @xmath256 * remark . * it is crucial that @xmath81 are consecutive strata . otherwise , the union @xmath257 would not be locally compact , and the boundary operator would not be defined . leray coboundary operator @xmath258 ( @xmath252 @xmath253 ) is poincare dual to the boundary homomorphism @xmath259 by poincare duality , the intersection form @xmath260 is well defined and non - degenerate ( here @xmath101 is a smooth oriented manifold and @xmath261 ) . therefore , the only thing we need to check is that for any classes @xmath262 and @xmath263 @xmath264 let @xmath265 be the embedding . according to the definition of the leray coboundary operator , @xmath266 can be factored : @xmath267 where @xmath268 is the leray coboundary operator for the stratified space with two strata : @xmath0 and @xmath269 on the other side , the boundary homomorphism @xmath270 also can be factored : @xmath271 ( here @xmath272 is the restriction homomorphism induced by the inclusion @xmath273 ) . therefore , it is enough to assume that @xmath274 we know that @xmath146 is diffeomorphic to @xmath275 therefore , there is an isomorphism @xmath276 given by taking a representative , transversal to @xmath203 and intersecting it with @xmath204 the inverse isomorphism @xmath277 is given by multiplying a representative by @xmath278 * remark . * one should be careful with the orientations . we want the following condition to be satisfied : if @xmath279 is a cycle , transversal to @xmath247 @xmath280 and @xmath281 gives the orientation of @xmath282 at some point , then @xmath283 should give the positive orientation of @xmath167 at this point . with the above orientation conventions one gets that @xmath284 where @xmath215 is the gysin homomorphism and the intersection on the right is taken inside @xmath203 . moreover , @xmath285 therefore , we only need to check , that the gysin homomorphism is dual to the @xmath286 which is obvious . * corollary . * leray coboundary operator @xmath266 does not depend on the choice of the control data at least modulo torsion . one can also investigate the relation , dual to the relation on the coboundary operators , proved in the theorem [ relation ] . consider the strata @xmath287 satisfying the conditions of the theorem [ relation ] . let @xmath288 the boundary operator @xmath289 is dual to the direct sum of the coboundary operators @xmath290 the boundary operator @xmath291 is , in turn , dual to the direct sum @xmath292 therefore , the dual relation is @xmath293 however , it is not hard to prove independently . indeed , one can use that @xmath294 and @xmath295 then the relation basically says , that the boundary of the boundary of a chain is zero , which is trivial . this provides another proof of the theorem [ relation ] modulo torsion . * remark . * it is crucial for this argument , that @xmath296 is locally compact . therefore , if one would accidentally forget one of the intermediate strata @xmath297 the relation would not hold . in this section we review the definition of the parshin residue and the reciprocity law . as it was discussed in the introduction , parshin residue at a flag @xmath38 is defined as a sum of certain more delicate residues . in fact , every flag `` contains '' finitely many _ parshin points _ , and the more delicate residues are computed at these points . 1 . @xmath301 is the normalization ; 2 . @xmath302 is the union of @xmath303-dimensional irreducible components of the preimage of @xmath304 3 . for every @xmath305 1 . @xmath306 is the normalization ; 2 . @xmath307 is the union of @xmath308-dimensional irreducible components of the preimage of @xmath309 choosing a point @xmath311 is equivalent to choosing irreducible components in every @xmath312 @xmath313 indeed , @xmath314 is normal and , therefore , locally irreducible at every point . in particular , it is locally irreducible at the image of @xmath315 let @xmath316 be the irreducible component of @xmath317 containing the image of @xmath315 let @xmath318 note , that @xmath319 is an irreducible component of @xmath320 the normalization of the whitney umbrella ( example [ umbrella ] in the introduction ) is isomorphic to @xmath321 the preimage of the @xmath55-axis is a line , which covers the @xmath55-axis twice with a branching at the origin . therefore , @xmath322 is just one point . this corresponds to the fact that there is only one torus in the example [ umbrella ] . in order to define the parshin residue , one needs to define the _ local parameters _ at a parshin point , which play the role of the normalizing parameter in one - dimensional case . after that , one uses these parameters to define a sequence of residual meromorphic forms @xmath323 on @xmath324 . @xmath325 is a hypersurface in a normal variety . it follows that there exists a ( meromorphic ) function @xmath326 on @xmath314 which has zero of order @xmath2 at a generic point of @xmath327 since meromorphic functions are the same on @xmath328 and @xmath317 one can consider @xmath326 as a function on @xmath320 then one can extend ( in an arbitrary way ) @xmath326 to @xmath329 and so on . for simplicity , we denote all these functions by @xmath330 now @xmath326 is defined on @xmath42 and can be consecutively restricted to @xmath331 for @xmath332 * remark . * one can choose local parameters in such a way , that @xmath326 has zero of order @xmath2 at generic point not only of @xmath335 but of the whole @xmath336 then these local parameters work for all parshin points with the flag @xmath337 we only use local parameters with this property . take a generic point @xmath341 both @xmath342 and @xmath343 are smooth at @xmath344 moreover , parameters @xmath345 provide an isomorphism of a neighborhood of @xmath346 to an open subset in @xmath347 and @xmath348 is given by the equation @xmath349 in this neighborhood . restrict the function @xmath185 to the transversal section to @xmath343 at @xmath350 given by fixing the parameters @xmath351 the restriction can be expanded into a laurent series in @xmath352 it is easy to see that the coefficients of this expansion depend analytically on @xmath344 moreover , one can see that the coefficients are meromorphic functions on @xmath353 let @xmath354 be the coefficient at @xmath355 in this expansion . then @xmath356 is a meromorphic @xmath303-form on @xmath353 ( @xcite,@xcite,@xcite)[parshin s reciprocity law ] let @xmath1 be a meromorphic @xmath17-form on @xmath18 fix a partial flag of irreducible subvarieties @xmath21 , where @xmath22 is omitted ( @xmath23 ) . then @xmath24 where the sum is taken over all irreducible @xmath25-dimensional subvarieties @xmath26 such that @xmath365 ( in this formula only finitely many summands are non zero . ) we want to apply the stratification theory to study the parshin points and residues . therefore , we need to stratify all the spaces in the normalization diagram in such a way that the stratifications respect the normalization maps @xmath367 the following lemma easily follows from the well known results on existence of whitney stratifications ( see section 1.7 in @xcite , for example ) : * notation . * let @xmath0 be an irreducible ( complex analytic ) variety considered with a fixed whitney stratification . then by @xmath368 we denote the stratum of maximal dimension . if @xmath0 is reducible , then by @xmath368 we denote the union of strata of maximal dimension . 1 . @xmath373 are unions of strata of @xmath374 2 . @xmath375 are unions of strata of @xmath376 correspondingly ; 3 . for all @xmath377 the local parameter @xmath326 is regular and non - vanishing on @xmath378 4 . for all @xmath377 the restriction of the normalization map @xmath379 to any stratum in the source is a covering over a stratum in the image . let @xmath384 be a point , such that @xmath314 is singular at @xmath7 note , that by dimension reasons and condition of the frontier , the only strata intersecting a small neighborhood of @xmath176 are @xmath385 and @xmath386 note , that @xmath326 is regular in @xmath387 and at a generic point of @xmath388 therefore , by extension theorem for normal varieties , @xmath326 is regular at @xmath7 note also , that @xmath326 is non - vanishing in @xmath389 and has zero of order @xmath2 at a generic point of @xmath336 therefore , @xmath390 coincide with @xmath381 near @xmath7 moreover , easy to see , that the germ of @xmath326 at @xmath176 generates the ideal of the germ of @xmath381 at @xmath7 indeed , if @xmath391 is a function , regular at @xmath176 and vanishing on @xmath392 then @xmath393 is regular at @xmath176 by the extension theorem for normal varieties . now , let @xmath394 be any coordinate system on @xmath381 at @xmath7 easy to see , that the functions @xmath395 generate the maximal ideal in the local ring of @xmath396 therefore , @xmath176 is a smooth point of @xmath383 moreover , we will show that @xmath46 naturally splits into the sum @xmath400 such that @xmath401 note , that according to the construction of the leray coboundary operator , @xmath46 is represented by a smooth compact real @xmath17-dimensional submanifold @xmath402 moreover , @xmath403 is obtained from a point by the following procedure : there are @xmath17 steps , and on each step we take the total space of an oriented fibration with @xmath2-dimensional compact fiber over the result of the previous step . therefore , @xmath403 is a union of @xmath17-dimensional tori . we ll show , that the connected components of @xmath403 are in natural one - to - one correspondence with the points of @xmath322 and the connected component @xmath49 corresponding to @xmath404 represents @xmath405 fix control data on the stratification @xmath131 of @xmath18 let us use these control data to construct the representative @xmath406 of @xmath407 let us also denote by @xmath408 the representative of @xmath409)\in h_k(\breve v_k),$ ] constructed in the same way . note , that for @xmath413 @xmath414 is the preimage of @xmath415 under the map @xmath416 since @xmath417 and @xmath418 are consecutive strata , it follows that the restriction @xmath419 is a proper submersion to @xmath420 [ lifting control data ] let @xmath102 and @xmath438 be two stratified spaces , consisting of two strata each : @xmath439 @xmath440 and @xmath441 @xmath442 let @xmath443 be a map , such that @xmath444 is a covering over @xmath0 and @xmath445 is a covering over @xmath210 let @xmath446 and @xmath447 be the control data on @xmath130 then there exist control data @xmath448 on @xmath449 such that indeed , the map from the connected components of @xmath459 to the connected components of @xmath461 is given by the retraction @xmath462 existence of the inverse to this map follows from the fact that @xmath463 is a complex hypersurface in the manifold @xmath464 indeed , if @xmath465 is a hypersurface in a complex manifold @xmath466 then there is only one connected component of @xmath101 in a neighborhood of a connected component of @xmath467 since we have chosen the control data on @xmath435 to be coherent with the control data on @xmath430 ( given by restricting the control data from the ambient space ) , we have the equality on the level of representatives . now we use the local parameters @xmath333 to construct cycles @xmath478 such that , on the one side , it is obvious that @xmath479 and , on the other , @xmath480 is homologically equivalent to @xmath481 in @xmath475 function @xmath482 is regular and non - vanishing in @xmath483 and has zero of order one at a generic point of @xmath484 it follows immediately , that @xmath482 is regular on @xmath435 and the equation @xmath485 defines @xmath461 in @xmath464 in order to get the parshin s reciprocity law from the theorem [ relation ] and the above consideration , one needs a fixed stratification of @xmath102 , such that all non - zero residues of a given form @xmath1 are in the flags consisting of closures of strata of the stratification . it turns out , that any whitney stratification , such that @xmath1 is regular on the top dimensional stratum is good enough . more precisely , we have the following theorem : [ criteria for non - trivial residues ] let @xmath102 be an @xmath17-dimensional variety and @xmath1 be a meromorphic @xmath17-form on @xmath130 let @xmath500 be a whitney stratification of @xmath105 such that @xmath1 is regular on @xmath501 . consider the normalization diagram for the flag @xmath164 let @xmath506 and let @xmath333 be local parameters . let @xmath131 be a stratification of @xmath102 satisfying conditions of the lemma [ stratification of the normalization diagram ] , and such that all strata of the stratification @xmath507 are unions of strata of @xmath508 as usual , we denote by @xmath418 the stratum of @xmath131 which is open and dense in @xmath509 1 . let @xmath510 be consecutive strata of @xmath511 and let @xmath81 be the consecutive strata of @xmath512 such that @xmath0 is an open dense subset in @xmath513 and @xmath129 is an open dense subset in @xmath514 let @xmath515 and @xmath516 be the embeddings . then it easily follows from the construction of coboundary operators @xmath266 and @xmath517 and the independence of these operators from the choice of the control data , that @xmath518 2 . let @xmath519 be a stratum of @xmath507 and let @xmath81 be consecutive strata of @xmath512 such that @xmath520 and @xmath129 is open and dense in @xmath514 then @xmath521 where @xmath522 is the embedding . moreover , if @xmath167 is a representative of a homology class in @xmath523 and @xmath282 is the representative of the @xmath524)\in h_{*+\dim y-\dim x-1}(y),$ ] constructed in the standard way , then every connected component of @xmath282 is homologically equivalent to @xmath525 in @xmath210 indeed , one can use the control data on @xmath131 to embed the mapping cone of @xmath526 into @xmath514 let @xmath25 be the largest number , such that @xmath418 is a subset of a stratum of @xmath500 of dimension bigger then @xmath527 for @xmath528 , let @xmath529 be the stratum of @xmath511 such that @xmath530 note , that @xmath531 and @xmath532 is open and dense in @xmath533 moreover , by dimension reasons and the condition of the frontier , @xmath534 and , since @xmath1 is regular in @xmath539 @xmath540 moreover , easy to see , that every connected component of the standard representative of the @xmath541)$ ] is homologically equivalent to @xmath542 therefore , @xmath543 for any @xmath544	the article consists of two main parts : an analog of the theory of leray coboundary operators for stratified spaces , and its application to the theory of parshin residues . the first part is independent from the second . it uses the theory of whitney stratifications . the second part is an application of the first . in particular , a geometric proof of the parshin s reciprocity law for residues in complex case is given . [ section ] [ section ] [ section ] [ section ] * parshin residues via coboundary operators . * by mikhail mazin , stony brook university .
at the end of their lives , stars of approximately solar - mass ( 0.88 m@xmath9 ) ascend the asymptotic giant branch ( agb ) , the final phase of nuclear burning in the lives of these stars . the agb is one of the brightest populations in the infrared ( ir ) sky , contributing up to @xmath02030% of the ir light for the small magellanic cloud ( smc ) @xcite . essentially all agb stars are variables @xcite , with periods on the order of hundreds of days . this variability is caused by hydrodynamic pulsations traveling through the extended atmosphere of the star . the dramatic changes in stellar radius these shocks produce cause brightness fluctuations on the scale of @xmath02 mag ( e.g. * ? ? ? * ) , on time scales of hundreds of days @xcite . on much longer time scales ( @xmath010@xmath10 yr ) , runaway thermonuclear reactions in the helium burning shell of an agb star , called ` thermal pulses , ' can cause much more dramatic brightness variations @xcite . thermal pulses cause readjustments in the global structure of the agb star , carrying elements synthesized in the nuclear burning regions to the surface . this process , known as the ` third dredge up , ' is responsible for the formation of carbon - rich agb stars @xcite , as well as bringing more exotic elements such as tc to the surface @xcite , and its precise details remain a topic of current research ( e.g. * ? ? ? towards the end of their evolution , agb stars exhibit extensive rates of mass - loss @xcite , driven by the aforementioned hydrodynamic pulsations and radiation pressure on the resultant dust grains @xcite . this mass - loss , enhanced by elements produced in the nuclear - burning regions of the star and dredged to the surface , makes agb stars one of the primary sources for @xmath11-elements in the universe , and an important contributor to dust in galaxies . the precise degree to which agb stars contribute dust to the interstellar medium ( ism ) is a matter of some dispute however . specifically , to what extent do agb stars or supernovae ( sne ) dominate this process ? the agb contribution to a galaxy s total dust budget can be computed if the entire population of mass - losing agb stars in the galaxy is identified , and the rate of dust production by each star in this sample is known . a comprehensive study of galactic agb stars is hampered by extinction in the plane of the milky way . the large magellanic cloud ( lmc ) offers the ideal combination of relative proximity and low line - of - sight extinction , allowing for detailed studies of galaxy - wide evolved star populations . the lmc s high galactic latitude minimizes both foreground contamination by milky way stars and reddening due to intervening dust . in addition , the distance to the lmc ( @xmath050 kpc ) is well determined ( e.g. * ? ? ? * ) . see @xcite for discussion and meta - analysis of this measurement . this is close enough that individual stars can be resolved , yet far enough away that the 3-d structure of the lmc can be neglected , accurate distances assumed for all stars , and therefore intrinsic brightnesses determined . recent large - scale photometric surveys of the lmc such as the _ magellanic clouds photometric survey _ ( mcps ; * ? ? ? * ) , the _ two micron all - sky survey _ ( 2mass ; * ? ? ? * ) , and the _ surveying the agents of a galaxy s evolution _ survey ( sage ; * ? ? ? * ) , have allowed the construction of catalogs of the entire agb population of that galaxy , with multi - band photometry for tens of thousands of sources @xcite . the sage - spec follow - on to the sage survey @xcite has produced spectral classifications for 100 sources in the lmc @xcite . the irsf survey @xcite also examined the lmc in near - ir ( jh@xmath5 ) bands . the akari mission @xcite has also been used to survey a portion of the lmc @xcite . while this survey covers a smaller area , the unique [ 11 ] and [ 15 ] @xmath12 m bands of the akari satellite are a valuable contribution to the multi - wavelength coverage of this important galaxy . the ongoing vista magellanic clouds survey ( vmc ) @xcite will push three magnitudes deeper than the akari survey , and has already been used to develop a spatially resolved star formation history of part of the lmc @xcite . as a first step towards the direct measurement of mass - loss rates for the entire lmc agb sample , ( * ? ? ? * hereafter paper i ) computed infrared excesses from sage data and used these to estimate the total dust injection rate . boyer et al . ( 2012 , in press ) have performed a similar analysis for the small magellanic cloud ( smc ) . @xcite determined a simple ir color proxy for mass - loss using mass - loss rates derived by @xcite and @xcite through detailed modeling of individual sources , and extrapolated a global gas and dust budget for the lmc . an alternative to detailed modeling is to compare the observed seds of sources to pre - computed models . with this aim , ( * ? ? ? * hereafter paper iv ) and ( * ? ? ? * hereafter paper v ) presented the grid of rsg and agb models ( grams ) for oxygen - rich and carbonaceous dust respectively . in this paper we apply the grams to the whole population of evolved stars in the lmc in order to derive more precise measurements of individual sources and the population wide return of mass to the galaxy . the remainder of this paper is organized as follows : section [ sec : data ] of this paper discusses the sources of our observational data ( [ sec : dat_source ] ) and our fitting procedure ( [ sec : var ] & [ sec : fitting ] ) . section [ sec : tour ] displays some specific fits representative of our sample . section [ sec : discuss ] describes the major results of our fitting , including the o - rich / c - rich determination ( [ sec : oc ] ) , the revealed luminosity function of the evolved stellar population of the lmc ( [ sec : lum_func ] ) , the integrated dust mass return to the ism ( [ sec : mlr ] & [ sec : int_mlr ] ) , and the development of simple observational proxies for dust mass - loss rate ( [ sec : proxy ] ) . we review our conclusions in section [ sec : conclusions ] . the original sage survey was conducted in two epochs , spaced @xmath03 months apart @xcite . the observations from these epochs have been combined into a single mosaic photometry archive and catalog , which is deeper and has smaller photometric errors than the individual epochs . our dataset consists of 33718 sources extracted from the sage mosaic photometry archive . these sources have been matched to optical data from the magellanic clouds photometric survey ( mcps ) @xcite , near - infrared photometry from the 2mass survey @xcite , and the variability information of the macho survey @xcite , allowing us to construct 12 band spectral energy distributions ( seds ) from the u band to 24 @xmath12 m for most of our sources . a 2 matching radius was used for all catalog joins . details of the matching to the mcps catalog can be found in the sage data delivery document . the join to the macho catalog is discussed more thoroughly in @xcite . all non - sage photometry has been de - reddened , with the de - reddened flux , @xmath13 , related to the observed photometry , @xmath14 by @xmath15 . our de - reddening coefficients , @xmath16 , are listed in table [ tab : dered ] . ll u & 0.5900 + b & 0.5315 + v & 0.4590 + i & 0.2708 + j & 0.1125 + h & 0.0652 + @xmath5 & 0.0372 [ tab : dered ] when extracting sources from the sage database , our initial agb classifications follow from @xcite and @xcite . a star is classified as an oxygen - rich ( o - rich ) or carbon - rich ( c - rich ) agb candidate based on its location on the @xmath5 vs. _ j _ @xmath17 @xmath5 color magnitude diagram ( cmd ) . the _ j _ @xmath17 [ 3.6 ] color ( or , in the absence of a j - band detection , the [ 3.6 ] @xmath17 [ 8.0 ] color ) is used to select extreme agb candidates . see 2.2 in @xcite for explicit definitions of these color cuts . rsg candidates are selected based on the color - magnitude criteria presented in @xcite . specifically , rsgs candidates are defined as being brighter than the tip of the red giant branch ( trgb ) , @xmath18 , and between the lines @xmath19 and @xmath20 . these cmd - based definitions are illustrated in figure [ fig : cmd_class ] . looking for young stellar objects ( ysos ) , @xcite published a list of 13 ` extremely red objects ' that they spectroscopically identified as carbon stars , based on the presence of sic absorption . twelve of these sources lie outside our defined cmd cuts and are therefore not included in our initial dataset , but we have manually added them to our list , using the photometry published in that work . all of these sources are classified as extreme agb candidates . [ cols="^,^ " , ] we present the largest sample of evolved stars subjected to radiative transfer modeling to date . by fitting @xmath030000 stars with @xmath012 bands of photometry each to pre - computed radiative transfer models from the grams model grid of o- and c - rich agb stars and rsgs , we compute individual bolometric luminosities and dust mass - loss rates for the entire evolved star population of the lmc . our work forms a useful prediction to be tested against models of stellar population synthesis and stellar evolution . we establish that the grams model grid is an important new tool for the study of stellar populations . capable of generating statistically accurate predictions for difficult - to - observe stellar parameters such as mass - loss rate and luminosity for thousands of stars in a reasonable period of time , grams will be valuable for interpreting the infrared observations of future missions ( [ sec : fitting ] ) . through comparison to previously published studies of agb stars in the lmc , we show that the grams o- and c - rich identifications are over 80% accurate ( [ sec : oc ] ) . we present a quadratic formula for the @xmath5 band bolometric correction of agb stars as a function of _ j _ @xmath17 @xmath5 color ( [ sec : lum_func ] ) . our fit is detailed in table [ tab : bc_fit ] . we characterize the luminosity functions of both o - rich and c - rich agb stars in the lmc . the o - rich stars follow a distribution between m@xmath21 = -3.5 - -5 centered at @xmath0 -4.3 ( @xmath03500 l@xmath9 ) . individual c - rich agb stars tend to be brighter , distributed between m@xmath21 = -4 -6 m@xmath21 , centered at -4.7 m@xmath21 ( @xmath05500 l@xmath9 ) . despite tending to be less luminous on an individual basis , o - rich sources dominate the integrated ir light from the lmc evolved stellar population , by virtue of their greater numbers ( [ sec : lum_func ] ) . we find that the agb stars exhibiting the ` long secondary period ' phenomenon , and lying on sequence d are significantly more o - rich than the stellar population on the other , radially pulsating sequences . we see a qualitative relationship between m@xmath7 and pulsation period for radially pulsating agb stars , in that sources with no detectable mass - loss are concentrated on sequences 3 & 4 , and do not appear at all on sequences 1 & 2 ( [ sec : mlr ] ) . we also derive hyperbolic fits to the m@xmath7 from c - rich agb stars as a function of various near - ir colors ( [ sec : proxy ] ) . presented in table [ tab : mlr_fit ] , our fits are derived from a sample much larger than those used by previous studies . we derive a total dust mass - injection rate for the entire rsg+agb stellar population of ( @xmath22 m@xmath2 yr@xmath3of the lmc through consistent , direct summation . assuming a gas - to - dust ratio @xmath23 for o - rich agb stars and rsgs , and a ratio @xmath24 for c - rich agb stars , this translates to a total mass injection rate into the ism from rsgs and agb stars of @xmath25 m@xmath2 yr@xmath3 , with half the mass coming from carbon - rich agb stars . dominated by the uncertainty in the gas - to - dust ratio , this figure is accurate to a factor of @xmath02 . that is , carbon stars inject the same amount of _ mass _ into the ism as o - rich agbs , but two and a half times as much _ 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 the national aeronautics and space administration and the national science foundation . this research was supported by nasa nag5 - 12595 , and nasa adp nnx11ab06 g . conversations with m. boyer were helpful . dr acknowledges support from nasa / jpl / spitzer contract # 1415784 .	we present results from the first application of the _ _ g__rid of _ _ r__ed supergiant and _ _ a__symptotic giant branch _ _ m__odel_s _ ( grams ) model grid to the entire evolved stellar population of the large magellanic cloud ( lmc ) . grams is a pre - computed grid of 80843 radiative transfer ( rt ) models of evolved stars and circumstellar dust shells composed of either silicate or carbonaceous dust . we fit grams models to @xmath030000 asymptotic giant branch ( agb ) and red supergiant ( rsg ) stars in the lmc , using 12 bands of photometry from the optical to the mid - infrared . our published dataset consists of thousands of evolved stars with individually determined evolutionary parameters such as luminosity and mass - loss rate . the grams grid has a greater than 80% accuracy rate discriminating between oxygen- and carbon - rich chemistry . the global dust injection rate to the interstellar medium ( ism ) of the lmc from rsgs and agb stars is on the order of @xmath1 m@xmath2 yr@xmath3 , equivalent to a total mass injection rate ( including the gas ) into the ism of @xmath4 m@xmath2 yr@xmath3 . carbon stars inject two and a half times as much dust into the ism as do o - rich agb stars , but the same amount of mass . we determine a bolometric correction factor for c - rich agb stars in the @xmath5 band as a function of _ j _ @xmath5 color , bc@xmath6 . we determine several ir color proxies for the dust mass - loss rate ( m@xmath7 ) from c - rich agb stars , such as @xmath8)+3.37}-5.93 $ ] . we find that a larger fraction of agb stars exhibiting the ` long - secondary period ' phenomenon are o - rich than stars dominated by radial pulsations , and agb stars without detectable mass - loss do not appear on either the first - overtone or fundamental - mode pulsation sequences .
many inflationary models make a key prediction : the production of a significant background of gravitational waves @xcite . a detection of this stochastic gravitational wave background would therefore provide particularly strong evidence for the inflationary paradigm , alongside the observed super - horizon correlation , near scale - invariance and gaussianity of the scalar density fluctuations ( e.g. , @xcite and references therein ) . such a detection would also give significant insights into the physics of inflation . in particular , the ratio of gravitational wave or tensor power to scalar density power @xmath0 gives a direct probe of the energy scale at which inflation occurred . the currently most promising method to measure this inflationary gravitational wave background is through its effects on the cosmic microwave background ( cmb ) . however , tensor modes can be difficult to constrain accurately via the cmb temperature alone , because scalar modes also contribute to the temperature , which leads to cosmic variance and degeneracies . unlike in temperature , the cmb sky contains two degrees of freedom in polarization ( a magnitude of polarization and a direction ) ; because of this , one can construct a null " linear combination from the two degrees of freedom which has no response to scalar perturbations at linear order , but still is sourced by inflationary gravitational waves . this combination , known as the cmb b - mode polarization , therefore allows a constraint on primordial gravitational waves that is robust and free of leading - order scalar cosmic variance @xcite . over the past year , the observational status of primordial b - mode polarization has been a subject of active debate . while the bicep2 collaboration reported a measurement of b - mode polarization on large angular scales @xcite , subsequent analyses based on planck polarization measurements suggested that this signal could be sourced by foreground emission from polarized galactic dust @xcite . a recently - released cross - correlation analysis involving data from the bicep2 , keck array and planck experiments @xcite has confirmed that a significant fraction of the signal seen in bicep2 was of galactic origin , and has thus replaced any evidence for inflationary gravitational waves with a new upper limit of @xmath4 at 95% confidence . for simplicity , and motivated by this new result , we work under the default assumption that @xmath5 ( though cosmic variance from tensor modes could straightforwardly be incorporated into our analysis if future measurements required this ) . while foreground polarization appears to be a significant obstacle when analyzing only one frequency channel , component separation with multifrequency data is a promising path towards removing much of the foreground contamination . even in the absence of foregrounds , primordial gravitational waves are not the only source of b - mode polarization . a well - understood , recently observed @xcite source of b - mode polarization is the gravitational lensing of the cmb , the deflection of cmb photons by the distribution of large - scale structure along the photons path . by remapping the cmb sky , gravitational lensing converts some of the primordial e - mode polarization ( the polarization component that can be sourced by scalars ) into b - mode polarization @xcite . the lensing b - mode power spectrum is constant at low multipoles , so that the lensing b - modes induce variance and act as a source of noise similar in effect to instrumental white noise at a level of @xmath6k - arcmin . the keck array / bicep2 collaboration recently reported the measurement of cmb maps that , for the first time , have noise levels below this value ( @xmath7k - arcmin , @xcite ) . lensing b - mode noise is thus beginning to limit even current b - mode surveys . hence , for an improved constraint on @xmath0 ( or a better characterization of the signal if it is non - zero ) , delensing methods are rapidly becoming essential @xcite . these methods involve subtracting off an estimate of the lensing b - mode map in order to reduce the lensing noise . to construct an estimate of the lensing b - mode and delens , one requires both a measurement of the e - mode polarization and an estimate of the lensing which remapped the primordial cmb sky . internal methods to delens involve reconstructing the lensing field ( using statistical methods such as a quadratic estimator , @xcite , or a full maximum likelihood approach , @xcite ) with the same cmb dataset that is used to constrain the primordial b - modes . however , very low noise levels are required for the best internal delensing performance . furthermore , much experimental progress continues to be made with cmb experiments with large beams ( @xmath8 ) , which can not reconstruct cmb lensing maps that are sufficiently signal - dominated for efficient delensing . with such experiments reaching low noise levels , delensing which makes use of external datasets to estimate the lensing field may be promising . the cosmic infrared background ( cib ) is diffuse far - infrared radiation originating in mostly unresolved dusty star - forming galaxies , where it is emitted by the uv - heated dust enshrouding young stars @xcite . in a number of studies , the cib was found to have a remarkably high correlation ( @xmath9 ) with cmb lensing ( e.g. , @xcite ) , confirming earlier predictions based on the similar high - redshift origin of both signals @xcite . this high correlation makes the cib an ideal proxy for cmb lensing . in this paper , we will investigate the use of maps of the cosmic infrared background to delens the b - mode sky . the use of external large - scale structure tracers for delensing was investigated in @xcite , focusing on tracers from surveys with well - known redshift origin . for such tracers , the authors concluded that external delensing was not a very promising technique , as futuristic surveys that mapped out all the structure to redshifts beyond @xmath10 were required to achieve good delensing performance . unlike the surveys considered , cib maps are well - suited for delensing , because even currently available cib maps trace the underlying mass distribution over a broad range of high redshifts and thus are excellent lensing tracers @xcite . in a paper discussing how well delensing methods can improve constraints on the tensor tilt @xmath11 ( which appeared as this paper was being completed ) , @xcite , the authors included forecasts for a constant - correlation cib channel . our work differs because it focuses on constraints on the tensor amplitude @xmath0 rather than tensor tilt and assumes @xmath0 is significantly less than 0.2 . we also discuss additional aspects of cib delensing , such as combinations of datasets , and , in particular , the effects of uncertainties in the cib cross- and auto - spectra . after briefly reviewing delensing in general , we discuss methods for cib delensing and for calculating the resulting reduction in lensing noise in section 2 . we then apply this formalism to models of the planck cib data in section 3 in order to forecast how much detection errors on @xmath0 can be improved by cib delensing . in section 3 , we also discuss combinations of external delensing datasets : first , the optimal combination of cib frequencies , and second , the combination of the cib with external estimates of lensing from high - resolution cmb experiments . in section 4 , we investigate the limitations of real data , focusing in particular on the problem of uncertainty in the cib cross- and auto - spectra , and investigating to what extent cib delensing can be self - calibrated without assuming models for the cib spectra . we conclude in section 5 . in this paper we assume a flat @xmath12cdm cosmology with @xmath13 , which is very close to the tt , te , ee+lowp+lensing best - fit base @xmath12cdm model from @xcite . as cmb photons travel through the universe , they are gravitationally lensed by large - scale structure . this lensing leads to a remapping of the cmb polarization anisotropies from their unlensed position @xmath14 on the sky : q ( ) = q_(+ ) ; u ( ) = u_(+ ) with an analogous remapping in temperature . here @xmath15 and @xmath16 are the cmb polarization stokes parameters , and @xmath17 is the lensing deflection field . as the curl - like part of the deflection is many orders of magnitude smaller than the gradient - like part @xcite , the lensing can be equivalently described using the convergence @xmath18 , which can be shown to equal a projection of the matter density out to high redshift @xcite . the convergence is also simply related to the often - used lensing potential @xmath19 by @xmath20 . writing the polarization not in terms of the experimentally measured @xmath15 and @xmath16 parameters but in terms of even parity e and odd parity b - modes , we find that lensing induces a conversion of a pure e mode pattern into b - modes , with the lensing b - mode given in harmonic space by ( to leading order in @xmath21 ) : b^ ( ) = w(, ) e ( ) ( - ) for w(, ) = ( 2_, ) , where @xmath22 is the unlensed e - mode , @xmath23 are multipoles written in the flat sky approximation , and @xmath24 is the angle between @xmath25 and @xmath26 . the flat - sky approximation used here gives sufficient accuracy for our simple forecast calculations and @xmath27-range , though we also verify our numerical results using the corresponding full - sky formulae @xcite . keeping only the lowest order term in @xmath21 has also been found to be an excellent approximation on the scales relevant for delensing @xcite . from this expression , we can calculate the power spectrum b^ ( ) b^^ * ( ) ( 2)^2 ^d(- ) c^bb,_l to give c_l^bb , = w^2 ( , )c^ee_l c^_\|-\| this lensing b - mode power is effectively constant on large scales @xmath28 @xcite . the full ( beam - deconvolved ) b - mode power spectrum is given by c_l^bb , = c_l^bb , r + c_l^bb , + n_l^bb where @xmath29 is the contribution of tensor modes to the b - mode power spectrum ( assumed zero by default ) and @xmath30 is the instrumental noise power n_l^bb = ( _ p / t_)^2 e^ l^2 _ ^2/(8 2 ) with @xmath31 the noise level , here written in @xmath32k - radians ( though typically specified in @xmath32k - arcmin ) , and @xmath33 the full width at half maximum of the beam in radians . assuming the realization of the lensing b - modes on the sky is unknown , the presence of lensing b - mode power will contribute to the error on a measurement of the b - mode power spectrum : ( c_l^bb , ) = ( c_l^bb,+n_l^bb ) where @xmath34 gives the sky fraction observed by the survey . while non - gaussian covariance between different b - multipoles is present because fluctuations in lensing and e - modes simultaneously affect different b - mode scales @xcite , this covariance is subdominant on the large scales that constrain @xmath0 . despite this , we include this covariance in our analysis in later sections of this paper . the error on the b - mode power determines the error on a constraint of the tensor - scalar ratio @xmath0 ( assuming the `` detection '' null hypothesis of no tensor modes in the error calculation ) : & ( r)&= ^- + & & ^-c_l^bb , + n^bb_l _ l<100 in the second line we have assumed that both the lensing b - mode and noise power spectra are constant , i.e. , white - noise - like , at @xmath28 @xcite , where nearly all the constraining power on @xmath0 arises ( note that the derivative with respect to @xmath0 is simple since @xmath0 merely normalizes a fixed tensor spectrum ) . the lensing - induced error affects any measurement of the b - mode power spectrum and therefore any constraint on @xmath0 , and persists even as instrumental noise is reduced to zero . delensing algorithms are therefore needed to remove the lensing b - mode and thus reduce the errors when the instrumental noise approaches or falls below the level of @xmath35 , which corresponds to @xmath36k - arcmin in polarization . we now consider delensing with a cosmic infrared background map @xmath37 , which traces the underlying matter distribution and which we will attempt to use as a proxy for @xmath21 . ( our analysis is general and thus @xmath37 could be any tracer of large - scale structure . ) we construct an estimate of the lensing b - mode using this field , using a weighting @xmath38 ( which we will determine ) : ^ ( ) = w(, ) f(, ) e^n ( ) i(- ) where @xmath39 is the noisy , observed e - mode ; to lowest order , the fact that it is lensed can be neglected . though we distinguish @xmath22 and @xmath39 , we will always take @xmath37 to contain all the components of the observed cib map ( e.g. , shot noise , foregrounds , instrumental noise if relevant ) . we then subtract this estimate off from the measured b - mode in order to delens . the residual lensing b - mode is given by b^ ( ) & = & b^ ( ) - ^ ( ) = w(, ) + & & ( e ( ) ( - ) - f(, ) e^n ( ) i(- ) ) we now wish to determine @xmath40 such that the residual lensing b - mode power , and hence the lensing b - mode noise , is minimized . evaluating the lensing b - mode power spectrum after the delensing term has been subtracted gives : c_l^bb , & = & w^2 ( , ) [ c^ee_l c^_\|-\| + & & -(f(,)+f^(, ) ) c^ee_l c^i_\|-\| + & & + f^(, ) f(, ) ( c^ee_l+ n^ee_l ) c^ii_\|-\| ] taking the derivative with respect to @xmath40 , we obtain the form of @xmath40 which minimizes the residual lensing b - mode : f(, ) = ( ) we note that for signal - dominated @xmath22-modes , this weighting corresponds simply to a normalization of the @xmath37 field by @xmath41 . additionally , noisy @xmath22-modes are down - weighted . we also note that the same weight @xmath40 could be obtained by treating the part of the cib that is uncorrelated with lensing as effective noise in the standard calculation of internal delensing methods as in @xcite and @xcite . however , our explicit approach has advantages when we consider uncertainty in the cib spectra in later sections of this paper . using this ( perturbatively ) optimal weight @xmath40 , we obtain the following residual lensing b - mode power after delensing : c_l^bb , & = & w^2 ( , ) c^ee_l c^_\|-\| + & & where @xmath42 is the correlation coefficient of cmb lensing and the cib : _ l= the full delensed b - mode power spectrum then becomes ( from eq . 6 ) : c_l^bb , = c_l^bb , r + c_l^bb , + n_l^bb . when the e - mode is signal dominated for all relevant @xmath43 used in the sum in eq . 14 , the delensing can be simply described by the following reduction of the lensing signal in the map : c^_l ( 1-_l^2)c_l^. if the correlation coefficient is constant with @xmath27 , the lensing b - mode power and the resulting lensing noise are then similarly reduced by a factor @xmath44 . to estimate the delensing performance and the reduction in noise for b - mode searches , we must know the residual b - mode power ; to determine the residual b - mode power with eq . 14 , we must in turn have an estimate for the correlation coefficient of the cib with cmb lensing , @xmath45 . to calculate @xmath45 in this work , we will use the cib halo model of @xcite to estimate cib lensing - cross- and auto - power spectra ; this model is a good fit to planck measurements of these spectra @xcite . as the cib power spectra in the default model do not include the shot noise contributions that remain in the measured spectra , we add shot noise terms for both bright cib and radio sources ; we obtain the relevant amplitudes ( which match the measured spectra ) from the models used in @xcite . however , in this section we do not include instrumental noise ( or any additional noise due to component separation ) . on the scales relevant for delensing , this is an excellent approximation for the higher frequency planck channels ( 545 , 857 ghz ) , but not necessarily the lower frequency channels ( 353 ghz , 217 ghz ) , which are only included for comparison . we will discuss the impact of uncertainties of cib spectra in section 4 . the correlation coefficients for the different frequencies are shown in fig . 1 . as expected from prior work @xcite , we find a high degree of correlation at low multipoles , peaking above 0.8 for all the frequencies . this again illustrates that the cib is an excellent tracer of lensing . the correlation coefficients do not appear constant with scale , unlike shown in figure 13 of @xcite ; in part , this is because we include the shot noise contribution to the cib power spectra , though even without the inclusion of shot noise the correlation in the model we use falls somewhat with @xmath27 . we will assume that only cib scales @xmath46 can be used to delens ; however , the results are not very sensitive to the exact value of the low-@xmath27 cutoff , as most of the lensing b - modes arise from higher-@xmath27 lens scales ( see fig . 1 ) . ( in arbitrary units ) , i.e. how much each scale @xmath27 in the lensing contributes to the mean b - mode power on scales @xmath47 , and thus which scales are most important to delens . ] we now consider to what extent foregrounds , in particular galactic dust emission , will reduce the correlation from the models described above and degrade the delensing performance . to determine this , we consider the effect of foregrounds on the cib maps cross - correlations with cmb lensing as well as on their power spectra . galactic foregrounds are not expected to be correlated with the true lensing maps ( so that the cross - power measured in @xcite should be accurate everywhere ) , but they will contribute to the cib auto - power spectra on large scales . however , on @xmath48 of sky , the planck collaboration @xcite have demonstrated that galactic dust can be removed by subtracting dust maps constructed from observations of galactic hi emission ( 21 cm emission from neutral hydrogen , @xcite ) combined with planck hfi and iras observations ; after subtraction , the dust contributions to the auto - spectra are small on angular scales above @xmath49 . while the planck cib analyses that demonstrate galactic dust subtraction rely on some of the cleanest areas of sky , most upcoming b - mode searches will either focus on these same low - dust regions at high galactic latitude where foreground contamination is smallest , or will have enhanced multifrequency coverage that will allow improved dust modeling and subtraction . even if large scales @xmath50 are assumed completely unusable , the reduction in delensing performance when changing from @xmath51 to @xmath52 is only @xmath53 , because smaller scales @xmath54 , which are less affected by dust , are much more important for delensing ( as can be seen in fig . 1 ) . though for the largest surveys our forecasts may be somewhat optimistic , our correlation model should thus be sufficiently accurate to make realistic forecasts of delensing performance for most surveys . in our analysis we have assumed that the e and b cmb polarization maps have been cleaned with multifrequency data so the levels of residual galactic dust and cib contamination in the cmb maps are negligible , though we revisit this in section 4 , along with other aspects of potential foreground contamination . with this model for the correlation coefficients , we can characterize the delensing performance for cib maps at different frequencies . we calculate the delensed residual b - mode power for the different planck cib channels from eq . 14 , at first setting the instrumental noise in the cmb maps to zero . the residual lensing b - mode spectra , after delensing in this idealized case , are shown in fig . 2 ; we find that for all planck frequencies cib delensing can remove a substantial part of the lensing - b mode power . we quantify this by measuring the change in the mean @xmath35 after delensing , averaged over large scales @xmath28 , where most of the constraining power for primordial b - mode polarization lies . we obtain the following results : after delensing noise - free cmb maps using the 857 ghz , 545 ghz , 353 ghz , and 217 ghz cib frequencies , the mean low - ell @xmath35 is reduced by a factor 0.46 , 0.45 , 0.47 , 0.49 respectively ( though the 217 and 353 ghz delensing performance will be degraded when noise is included . ) with our knowledge of the residual b - mode power , we can now define an improvement factor in the measurement error of @xmath0 , @xmath55 , as in @xcite : _ 0(r)/_(r)where @xmath55 depends on the b - mode power as in eq . 9 , @xmath56 indicates the error before delensing , and @xmath57 is the error after delensing ( i.e. , determined by @xmath58 rather than @xmath59 ) . when the instrumental noise in eq . 9 is neglected , this reduces to the simple expression @xmath60 because both the lensed and residual b - mode power are flat at @xmath28 ( see fig . 2 ) ; we show the results for this expression ( which is just the inverse of the reduction factors quoted earlier ) in table 1 . .ideal improvement factors on errors on a detection of @xmath0 achievable with cib delensing using four planck frequency channels , assuming zero instrumental noise in both the cmb and the cib maps ( a good approximation for 545 and 857 ghz , but not for the lower frequencies ) . [ cols="^,^",options="header " , ] going beyond the ideal noise - free limit , we now examine the improvement in measurement errors on @xmath0 as a function of instrumental noise level of a cmb polarization telescope . with non - zero experimental noise , the improvement factor becomes : = where the residual b - mode power is calculated from eq . 14 as before , now including the effects of instrumental noise in the calculation . since the noise power spectrum after beam deconvolution enters into this calculation , the impact of noise depends on the instrumental beam size ( as seen in eq . we will thus consider two reference experiments in our analysis . the first reference experiment , which we will designate the _ highres _ experiment , has high resolution a 1.4 arcmin fwhm gaussian beam and is thus capable of efficient internal delensing when noise levels are low . this experiment should be taken as a proxy for ground - based telescopes such as actpol , polarbear - i / ii , sptpol , and their successors @xcite , as well as high - resolution satellites such as core or prism @xcite . the second reference experiment , which we term the _ lowres _ experiment , is a lower angular resolution cmb telescope targeted towards measuring large - scale cmb polarization . we take it to have a 30 arcmin fwhm gaussian beam ; this resolution is too low for efficient internal delensing . the _ lowres _ experiment is intended to represent cmb telescopes such as keck array , bicep3 or litebird @xcite . for , we assume that the measured e - mode can be co - added with planck maps of the e - mode polarization on smaller scales ( for which we assume 60@xmath32k - arcmin in polarization and a 7-arcmin beam , as in @xcite ) . for both experiments , we show the improvement factor @xmath61 for delensing with the 545 ghz cib frequency in fig . 3 , as a function of polarization noise level . other cib frequencies give nearly the same results , as their correlation coefficients with lensing are very similar . for , we see that @xmath61 becomes greater than 1.5 as noise levels fall below 4@xmath32k - arcmin and reaches @xmath62 as the experimental noise approaches zero . as a cib - delensed survey gives the same constraints as an @xmath63 times larger survey that has not been delensed ( neglecting windowing effects ) , these improvement factors can correspond to significant enhancements in experimental power . for , the delensing performance , though similar , is slightly reduced for intermediate noise levels . this is due to differences in the e - mode map : while for , the e - mode map is signal - dominated on all relevant scales ( i.e. , the results do not change visibly if we set the noise in the e - mode map to zero ) , the _ lowres _ e - mode is not signal - dominated everywhere , which results in a small reduction in delensing performance . for , we also show the improvement factor @xmath61 for internal delensing , i.e. for delensing with an estimate of the lensing convergence reconstructed from cmb data . we calculate the expected noise level of cmb lensing reconstruction for the minimum variance combination of optimal quadratic estimators as in @xcite ( assuming scales of @xmath64 and @xmath65 are used in temperature and polarization reconstruction respectively ) ; we hence obtain a correlation coefficient for the reconstructed maps , which allows us to calculate the improvement factor @xmath61 as before and compare with cib delensing . in fig . 3 , we see that for noise levels above @xmath66k - arcmin , cib delensing performs slightly better ( though @xmath61 is not large ) ; for much lower noise levels , with a small beam such that lensing reconstruction is possible , internal lensing reconstruction performs better than the cib at delensing . the use of an iterative procedure for internal reconstruction will not greatly modify the reconstruction curves at the noise levels shown in fig . 3 @xcite , though it will cause large increases to @xmath61 when the noise is below @xmath67k - arcmin . it is encouraging that for moderate noise levels @xmath68k - arcmin , internal delensing and cib delensing ( with current cib data ) are fairly similar in performance . furthermore , for future _ lowres_-like experiments without the necessary angular resolution to perform lensing reconstruction , cib delensing will allow an improvement at low noise levels . we now investigate how well we can delens using combinations of large - scale - structure datasets , turning first to combinations of different cib channels . writing our combined delensing tracer @xmath37 as a linear combination @xmath69 , where the fields @xmath70 are different cib frequencies ( or other large - scale structure tracers ) , we must determine the coefficients @xmath71 that allow us to delens optimally . we note that in eq . 14 , these coefficients only enter into the correlation coefficient of @xmath37 with lensing , @xmath45 ; hence , if we seek to maximize the delensing performance , we must maximize @xmath45 with respect to the @xmath71 coefficients . in the appendix , we solve for the coefficients @xmath71 that maximize the correlation . we find that for the optimal linear combination @xmath69 , c_i = _ j ( ^-1)_ij _ j where @xmath72 is the cross - correlation coefficient of large - scale structure tracer @xmath70 with the true lensing convergence , @xmath73 is the power spectrum of @xmath70 , and @xmath74 are the cross correlation coefficients of tracers @xmath70 and @xmath75 . for the linear combination @xmath37 , the correlation coefficient with cmb lensing is ^2 = _ ij _ i ( ^-1)_ij _ j . at each @xmath27 . for the linear combination of different cib frequencies , we have all the information needed to calculate this correlation coefficient from the model for the cib we use in this paper . we consider the combination of 545 , 857 , 217 and 353 ghz cib maps , as well as the combination of only 545 and 857 ghz maps ; the cross - correlation coefficient for the combination of four frequencies is shown in fig . 1 . for both linear combinations , we note that the improvement in correlation is quite small compared to individual cib frequencies . this is expected : given how correlated the different cib frequency channels we consider are ( typically @xmath76 , @xcite ) , it is clear they must contain much of the same information . nevertheless , we can calculate the delensing improvement factor @xmath61 for these linear combinations . as for individual frequencies , we obtain the improvement factors @xmath61 by calculating the delensed residual b - mode power from eq . 14 . as can be seen in table 1 and as expected from the correlation coefficient , the value of @xmath61 for a noise - free experiment is 2.3 for both 545 ghz + 857 ghz and for all cib frequencies combined ( neglecting any cib noise ) not very different from the results of delensing with the @xmath77 ghz channel alone . larger improvements may be possible if a wider range of less correlated frequencies is used , but we defer an investigation to future work . we now consider combining maps of the cib with maps of lensing that have been reconstructed from cmb data . such a combination could provide a better proxy for the true cmb lensing field and thus increase the delensing performance . to derive the delensing performance of such a combination , let us assume that our lensing reconstruction , for a certain instrumental noise level and beam , results in a lensing map with reconstruction noise @xmath78 , so that its power spectrum is ^ _ _ = ^ + n^_l we combine this reconstructed lensing map with a cib map at one frequency ( e.g. , at 545 ghz ) , which we take to have a correlation coefficient @xmath79 with the _ true _ lensing field . how efficient is this combination at delensing ? to calculate this , we treat our lensing reconstruction as just another large - scale structure tracer @xmath70 and repeat our calculation of how well a combination of tracers can delens . the correlation matrices of eq . 21 can be simply calculated to give the total correlation coefficient ( at each @xmath27 ) : ^2_l = 1 - given our knowledge of the cib - lensing correlation coefficient and our knowledge of the lensing reconstruction noise , we can calculate the improvement factor @xmath61 . we will first consider the case where the cmb lensing is estimated internally from the same survey that is being delensed . this requires an experiment with a small beam , such as our _ highres _ reference experiment . in fig . 3 , we include a curve showing the improvement factor @xmath61 for delensing _ highres _ with the combination of 545 ghz cib maps and quadratic estimator lensing reconstruction maps ; moderate increases in @xmath61 can be seen when delensing with this combination . we now consider the delensing synergy of two different overlapping surveys an experiment similar to ( with low noise and low resolution that constrains large - scale b - modes and @xmath0 , but can not reconstruct lensing well ) and an experiment similar to ( with high resolution that reconstructs @xmath21 as a delensing tracer ) . this is a useful scenario because the high resolution experiment can be used to delens the low - resolution experiment if the surveys overlap @xcite . the polarization noise level of the high resolution experiment @xmath80 will of course generally differ from that of the low resolution experiment @xmath81 . we show the delensing performance for such a scenario as a function of @xmath81 in fig . 4 ( for _ highres _ noise values of @xmath82k - arcmin and @xmath83k - arcmin ) . the lensing reconstruction from _ highres _ is either used alone or combined with the 545 ghz cib maps to delens _ lowres_. we find that the combination of cib and reconstructed lensing from high - resolution cmb maps , even with intermediate noise levels of @xmath84k - arcmin , is quite effective as a delensing tracer , and is able to improve the detection errors by close to a factor @xmath85 for a lensing - limited survey . the addition of the cib results in a significant improvement in delensing performance compared to reconstruction with alone ; noisy @xmath86k - arcmin reconstruction coadded with cib data delenses as well as @xmath87k - arcmin reconstruction without the cib . this increased performance is due to the fact that lensing reconstruction at intermediate noise levels does not image small lenses well ; adding the cib provides information on these small lenses and increases the small - scale correlation of the coadded delensing tracer with the true lensing convergence . the combination of a low noise b - mode search ( similar to ) that can not reconstruct lensing well with a higher resolution experiment ( such as ) , even at a higher noise level , is thus quite promising for delensing , especially when this external lensing reconstruction is combined with cib data . ( we note that our calculations are only relevant when the constraint on @xmath0 from is more powerful than the constraint from alone , e.g. , because the latter has higher noise or has not been optimized to measure large - scale polarization ) . additional improvements might be obtained by combining cib data with other tracers of large - scale structure , such as galaxy surveys ( or other cib maps over a wider frequency range ) , in order to improve the overlap of the combination s redshift origin with the lensing kernel . however , we leave such an investigation to future work . in our calculations of section 3 , we assumed a model ( albeit one based on measurements ) for the lensing cross- and auto - spectra @xmath88 and @xmath89 of the cib . however , though good measurements are available , the true cross - and auto - spectra of the cib fields are of course not exactly known . this is potentially important for estimating delensing performance , not just because forecasts may be slightly inaccurate , but because the true cib lensing - cross- and auto - spectra determine the residual b - mode power spectrum ( the power spectrum of the lensing b - mode that remains after delensing ) as in eq . because the cib spectra affect the residual b - mode power spectrum , insufficiently constrained cib spectra can degrade constraints on @xmath0 . for example , if the cib cross - correlation is in fact smaller than in the fiducial model , the reduced correlation leads to an `` excess '' of residual lensing b - mode power compared to the fiducial model , which could be confused with the presence of primordial tensor b - modes . we illustrate this by constructing 10 different cib cross - correlation curves , all of which are consistent with current planck measurements , and calculating the residual b - mode power spectrum from eq . 12 for each such model . ( to construct each curve , we perturb every cib - lensing - cross model bandpower by a random amount drawn from a gaussian of width given roughly by the size of the measurement errors from @xcite ; see later for more discussion of our error assumptions ) . in fig . 5 we show the difference between the residual b - mode spectrum for each of these 10 alternative `` models '' and the residual b - mode spectrum for the fiducial model . we also show a primordial tensor b - mode power spectrum corresponding to @xmath90 for comparison . though the differences in the residual b - mode power are small ( of order 5% of the total power ) , it can be seen that they are comparable in size to primordial b - modes when @xmath91 ; uncertainty in the cib spectra could thus limit the power of cib delensing for b - mode surveys that are capable of such high precision constraints on @xmath0 ( e.g. , are sufficiently large ) . it can be seen that variation in the uncertain cib spectra can be degenerate with the presence of primordial b - modes , especially if only the largest scales of the b - mode polarization are measured . ] we now seek to quantify the effects of uncertainty in the cib spectra on the delensing procedure more rigorously . we begin by writing a theoretical model for the full delensed b - mode power spectrum , @xmath92 , as : ^bb,(r , a_i , b_i ) = ^bb , r(r ) + c_l^bb,[^i(a_i ) , ^ii(b_i)]where we have denoted the residual lensing b - mode power as a function of the cib spectra , and we have assumed a measurement without noise bias . we have also assumed that , in turn , the parameters @xmath93 ( which we will later define as effectively parametrizing the amplitude of cib bandpowers ) completely describe the true cib cross- and auto - spectra . as illustrated previously , uncertain cib spectral parameters @xmath93 could potentially be degenerate with @xmath0 and thus reduce the performance of cib delensing . to quantify this limitation , we will again forecast constraints on @xmath0 , but now marginalize them over the uncertain true cib spectra ( i.e. , the parameters @xmath93 ) . first , we discuss how the cib spectra can be parametrized for delensing analyses . while considerable progress has been made in modeling these spectra , for cosmological constraints on @xmath0 we believe it is more robust if cross- and auto - spectra can be completely constrained from measurements without assuming a certain model for the cib properties and spectra . this is especially important as foreground residuals may also contribute to the spectra and may not be simple to model . we thus propose to constrain @xmath0 and the cib spectral parameters @xmath93 simultaneously from measurements of @xmath94 as well as measurements of the cib spectra @xmath95 and @xmath96 , and then marginalize constraints on @xmath0 over @xmath93 . @xmath96 should ideally be measured with the exact cib map used for delensing , including any foreground ( e.g. , galactic dust ) residuals present in the map . as long as fiducial model spectra are close to the data , they should be used for the filtering function @xmath40 of eq . this preserves linearity and is only slightly suboptimal , because any correction to the delensing performance is second order in an error in @xmath40 . for such an empirical approach , we must have a very general parametrization of the cib spectra . as indicated previously , we will here assume that the cib lensing - cross- and auto powers can be completely described by rescaling different @xmath27-ranges of the fiducial cib spectra @xmath97 , similar to rescaling effective bandpowers : & & c^i_l ( a_i ) ^i_l(1 + a _ i u_i(l ) ) , + & & c^i i_l ( b_i ) ^i i_l(1 + b _ i u_i(l ) ) where a sum over @xmath98 is implicit , and @xmath99 are top - hat functions that are unity within an @xmath27-range corresponding to the @xmath98th bin ( we describe our binning in later sections ) , and zero otherwise , i.e. : = 1200 u_i(l ) = \ { ll 1 & l _ < l < l _ , + 0 & . . ( though there are other possible , equally complete , parametrizations ) . in writing this model , we assume that the true cib spectra in our data are sufficiently smooth that they can be described without significant losses by this discrete parametrization when integrated over the kernel of eq . 5 ( and fig . 1 ) , though this assumption appears reasonable unless the binning is very coarse . assuming that measurements of both cib spectra and the b - mode power spectrum are available , and that we are fitting both @xmath0 and the cib spectral parameters simultaneously , we now investigate to what extent uncertain @xmath95 can degrade the constraints on @xmath0 . we will neglect uncertainty in the cib auto - power , because we assume that the raw auto - spectrum of the cib map ( including any residual foregrounds ) can be measured with planck data much more accurately than the lensing cross - spectrum . this is currently the case for the relevant high frequency cib maps at 545 and 857 ghz ( despite the smaller area on which the auto - spectrum is measured ) , and is typically a good approximation because the lensing reconstruction maps are expected to have significantly higher noise than the cib maps . ( furthermore , the residual b - mode is twice as sensitive to changes in the cross - power than to the auto - power , as can be inferred from eq . 12 ) . to investigate how much constraints on @xmath0 are degraded by uncertain @xmath95 , we perform a fisher forecast for the parameters @xmath100 , obtaining a constraint on @xmath0 marginalized over the cib - lensing cross - power parameters @xmath101 . we construct the fisher matrix as f_pq & = & _ l_a = l^bb_^l^bb _ _ l_b = l^bb_^l^bb _ _ l_a , l_b ^-1 + & + & _ j where the cross - power has been binned into bandpowers @xmath102 which are identical to the effective bins of the @xmath101 parametrization . to construct this fisher matrix , we must evaluate derivatives of @xmath94 and @xmath103 with respect to all parameters @xmath104 , and estimate the measurement errors ( covariance matrices ) of the measurements of @xmath105 and @xmath106 . we begin by calculating how changing the assumed cib cross - spectrum via @xmath101 affects the delensed residual b - mode power . recalling our expression for the residual b - mode power ( eq . 12 ) , we can insert our parametrization of the cib spectra ( eq . 25 ) in terms of @xmath93 into this expression ( keeping the weight @xmath40 fixed to the fiducial model ) to obtain : & & c_l^bb,= w^2 ( , ) c^ee_l c^_\|-\| + & & where @xmath107 is the correlation coefficient for the fiducial model . we can now calculate the derivatives of the delensed b - mode power spectrum ( eq . 28 ) with respect to the @xmath101 parameters . using the chain rule it can be seen that = -2 _ l l_i c_l^ ^2_l where we have rewritten the expression in terms of the full - sky derivative @xmath108 with discrete multipoles @xmath27 . we evaluate the lensing derivative in this equation by differentiating the analytical full - sky expression of @xcite for @xmath109 , which is correct to leading order in @xmath110 . the @xmath111 superscript on the derivative indicates that when the beam is large and the e - mode is not signal dominated , the derivative must be evaluated after applying a filter @xmath112 to the e - mode power . the derivatives of the cib - lensing cross - power with respect to @xmath101 and the b - mode power with respect to @xmath0 are much simpler = _ ij c_j^i ; = c_l^bb , r(r=1 ) where we have noted that @xmath113 for all @xmath0 . we now turn to determining errors on the measurements . for a measurement of the b - mode power spectrum , the errors are determined by the covariance matrix : ^bb , bb_l_a , l_b & = & ( c_l_a^bb,+n_l_a^bb,)^2 + & + & _ l + & + & _ l where the second and third lines give the dominant part of the non - gaussian contribution to the covariance @xcite . when calculating delensed constraints , we have assumed that the off - diagonal parts of the covariance matrix are reduced by the constant delensing improvement factor @xmath114 , neglecting any scale - dependence . we also neglect the weak dependence of @xmath115 and @xmath116 on @xmath101 given current constraints . we assume that the lensing covariance itself is gaussian ^,_l_a , l_b = ( c_l_a^)^2 _ l_a , l_b and that an analogous expression holds for @xmath117 , the unlensed e - mode . we again use first order analytical derivatives of the b - mode power spectrum in constructing the off - diagonal part of the b - mode covariance matrix ( eq . 31 ) , unlike in @xcite , where full numerical derivatives are evaluated . higher order corrections can cause changes to the off - diagonal covariance , but as we find that the omission of the non - gaussian covariance contributions only causes negligible ( much less than @xmath118 ) changes to our final results , our simple approximation of the off - diagonal covariance appears sufficiently accurate for our estimates . we must also make an estimate of the measurement uncertainty on the cib - lensing cross - spectrum over the range of scales that contributes most to the signal - to - noise . we consider two cases for the errors on the cib - lensing cross - power , corresponding to current or near - future measurements : first , a constraint similar to the currently available planck cib - lensing cross - correlation data , second , a substantial improvement on this with the errors halved . we will make a simple model of the planck measurement errors as follows . we assume 23 bandpowers , of width @xmath119 ( half the width of planck bandpowers ) , with the first bandpower centered at @xmath120 , i.e. roughly spanning @xmath121 ( we maintain our cutoff below @xmath122 ) . we take the @xmath101 parameters to have the same effective bins , but will revisit the choice of effective bandpower width at the end of this section . the magnitude of the errors on each of these bandpowers is assumed to be a constant @xmath123uk.sr , which is a good approximation to the the errors on the planck cross - correlation ( @xcite , e.g. , table 2 and figure 5 ; optimistically , we only use the statistical part of the errors ) . this corresponds to a @xmath124 measurement of the cross - power , with fractional errors on the bins ranging from @xmath125 . the other , more futuristic , scenario we consider corresponds to the same assumed measurements but with the errors halved . we assume that any signal covariance between the delensed b - mode measurement and the measurement of the lensing - cib cross - correlation is negligible , because the planck lensing measurements are still reconstruction - noise - dominated on the scales most important for delensing ( and in addition , the areas of sky used for the two measurements will often differ ) . with our knowledge of the errors on both observables ( b - mode power and the cib - lensing cross power ) and the dependence of the observables on the parameters @xmath100 , we construct the fisher matrix of eq . 27 and obtain the marginalized error on @xmath0 as follows : _ ( r)= we can now compare two cases . first , a case in which we do not delens ; in this case , there is no reduction of the lensing noise in a measurement of the b - mode power , but there is also no delensing - induced degeneracy between @xmath0 and @xmath101 . second , a case in which we have delensed ; this reduces the b - mode noise by up to @xmath62 , but also introduces a degeneracy of @xmath0 with the set of uncertain cib parameters @xmath101 . to quantify this tradeoff , we define the marginalized improvement factor as the ratio of the errors on @xmath0 for these two cases , now including the effects of marginalizing over the uncertain correlation parameters @xmath101 in the delensed error : _ = _ 0(r ) / _ ( r ) we can plot this ratio as a function of @xmath34 , initially assuming that due to foregrounds and systematics , the largest scales below @xmath126 ( and hence any tensor b - modes from reionization ) can not be measured . to gain intuition , we begin by neglecting instrumental noise in the b - mode measurement , and simply impose a sharp cutoff @xmath127 beyond which b - modes are not measured ( though we later investigate more realistic scenarios with different beam and noise specifications ) . the results are shown in fig . 6 , for different ranges of @xmath27 which contribute to the b - mode power spectrum measurement ( @xmath128 , @xmath129 , @xmath130 ) as well as two different assumptions for the errors on the measurement of the cib - lensing correlation ( planck , planck/2 ) . on a constraint on @xmath0 , as a function of survey area now marginalized over the uncertain cib - lensing cross - spectrum . if the cib - lensing correlation were known perfectly , @xmath61 would be constant with area , as both delensed and non - delensed errors would change by the same factor as the area is varied . however , marginalization changes this for large surveys , as can be seen from the falling blue solid lines ( corresponding to planck - like errors on the cib - lensing correlation ) and green dashed lines ( corresponding to planck errors halved ) . the degradation in delensing performance seen is due to a delensing - induced degeneracy of @xmath0 with the cib - lensing cross - spectrum ; for large , powerful surveys , this degeneracy significantly increases the delensed errors on @xmath0 upon marginalization . to gain insight , we have here neglected instrumental noise , and have simply added a sharp cutoff @xmath127 to which the b - mode power is measured . for each set of lines , the lower line corresponds to @xmath131 , the middle line to @xmath129 and the upper line to @xmath130 . the constant value of @xmath61 when there is no uncertainty in the cib spectra is also shown as a black semicircular point at the top left of the figure . as the resolution of the b - mode power spectrum measurement is increased from @xmath131 to @xmath130 , it can be seen that the degeneracy of @xmath0 with the cross - correlation parameters is broken as the cib - lensing correlation is increasingly self - calibrated by the b - mode power spectrum at small scales . ] on a constraint on @xmath0 as a function of survey area , marginalized over the cib - lensing cross - correlation , as in fig . 6 . instead of the sharp @xmath127 cutoffs used to gain understanding in fig . 6 , we show more realistic results for different noise and beam specifications ( _ highres _ corresponds to a @xmath132 beam , _ lowres _ a @xmath133 beam ) . planck - like errors on the cib - lensing correlation are assumed . for reference , the ideal improvement factors when there is no uncertainty on the cib spectra are indicated with solid circles at the left of the plot . our results again confirm that marginalizing over the uncertain cib spectra does not greatly degrade constraints for small surveys or for surveys which can resolve small b - mode scales ( and can hence self - calibrate ) . for many surveys , cib delensing thus does not require the assumption of a model for the cib spectra . ] we first discuss our calculation when we restrict ourselves to @xmath134 . it can be seen that as the area of the survey becomes large , @xmath135 falls rapidly . as the area is increased , the marginalized @xmath136 must therefore become much greater than its value for fixed cib parameters . ( note that @xmath56 , the error when not delensing , is unaffected by marginalization ) . we can understand this as follows . with larger survey area , all constraints on @xmath0 become tighter ; with stronger constraints , the sensitivity of a constraint on @xmath0 to cross - power uncertainty increases , because smaller @xmath0 can more easily be mimicked by uncertain cross - powers as in fig . 5 . as large , powerful surveys are more sensitive to cross - power uncertainty , marginalization over the uncertain cross - powers leads to a significant increase of errors on @xmath0 when surveys are large . for a survey larger than @xmath137 square degrees , this sensitivity is such that it is no longer worth delensing if this comes at the cost of introducing a degeneracy of @xmath0 with uncertain cib cross - powers . ( however , note that our current delensing weighting is not optimized to account for this marginalization . ) the conclusion that it is not worth cib delensing the largest surveys changes either if the cross - correlation measurements are greatly increased in accuracy , or if we use the constraining power of smaller scales in @xmath94 . increasing @xmath138 to 450 , we can see that now the improvement upon delensing remains close to 2.2 , the value we obtained when neglecting uncertainty in the cib - lensing cross - power , even for large surveys . an explanation for this observation is that the degeneracy of the cross - correlation parameters and @xmath0 is being broken by measurements of the small - scale b - modes ; effectively , the higher-@xmath27 b - mode power spectrum measurements are determining the cib - lensing correlation and are fixing the amplitude of the low-@xmath27 lensing b - mode residual , uncertainty in which would otherwise be degenerate with @xmath0 . we confirm this by noting that the constraints on any @xmath101 , marginalized over @xmath0 and holding the other @xmath139 parameters at their fiducial values , are tightened significantly as @xmath138 is increased beyond @xmath140 ( whereas for @xmath140 , including the b - mode measurements does not tighten the constraints on @xmath101 ) . it is possible that this self - calibration could also have been equivalently achieved via noisy lensing reconstruction using only @xmath141 information ( though a reconstruction may have insufficient signal - to - noise per mode to delens , it may still allow a reasonable measurement of the cross - correlation given enough sky area ) . as expected , the constraints on @xmath0 are much less affected by marginalizing over uncertain cib - lensing cross - powers if the cross - powers are measured twice as accurately by a futuristic lensing survey . to make our analysis more realistic , we now repeat our calculations for the _ highres _ and _ lowres _ reference experiments introduced earlier , now including instrumental noise in our calculation . we show results for both experiments , at two different noise levels , in fig . 7 . we again see that for small surveys , marginalizing over the uncertain cib spectra barely degrades constraints from their ideal values . it can also be seen that the degradation with sky area is less severe for high - resolution experiments ( blue curves ) than for low - resolution ones ( green curves ) . this is consistent with the @xmath127-dependence discussed in the context of fig . 6 , and is due to the fact that high - resolution experiments can use the high-@xmath27 b - modes for better self - calibration . for , the improvement factor falls , though generally does not degrade as much with sky area as when a sharp cutoff is applied ( as in fig . 6 ) , because even high-@xmath27 b - modes that are individually noisy contribute when averaging over many modes . our results confirm that , for a range of realistic surveys ( in particular , small surveys and high resolution surveys ) , a specific cib spectrum need not be assumed ; the spectra can instead be marginalized over without a significant loss of delensing performance . we will briefly discuss how our results change if we now assume that the reionization b - modes at @xmath142 can be measured . in this case , we do not find a substantial reduction of delensing performance when marginalizing over the cib spectra , even for very large surveys . we explain this as follows : when reionization b - modes can be measured , most of the constraining power on @xmath0 arises from the largest scales @xmath142 ; on these scales , the effects of uncertain cib spectra are much less important ( as they are subdominant to the large lensing b - mode variance ) . finally , we briefly comment on our use of effective bandpowers @xmath101 as a very simple non - parametric description of the cib cross - power for forecasting and marginalization purposes ( see @xcite for a similar analysis in a different context ) . we focus first on choice of their width . it should be noted that to obtain the residual b - mode power of eq . 14 , the correlation is integrated over the kernel plotted in fig . 1 . our bandpower widths are chosen to be somewhat smaller than the scale on which this kernel varies strongly . still narrower bins are thus unnecessary to characterize effects on the residual b - mode power , as any sub - bandpower features in the cib cross - correlation will be averaged out by the integration . of course , since most physical processes should lead to smooth features in the cross - power , the use of wider bins , different general parametrizations that encode smoothness ( such as a limited set of fourier or polynomial coefficients ) or gaussian processes may be well - motivated . however , investigating the impact of our choice of bandpower width by doubling ( and halving ) their width , we find that our results are effectively unchanged . we thus believe that marginalizing over our current set of cib parameters should give a conservative , though reasonable , estimate of the degradation of constraints on @xmath0 . we have argued in section 3.1 that our forecasts are realistic for many surveys , despite the presence of galactic dust foregrounds in the cib maps . we have also argued that the raw cib auto - spectra ( including residual foreground power ) can be measured accurately , so that additional foreground power should not bias or degrade results if @xmath0 and the true cib power spectra are simultaneously constrained . however , there are two potential limitations of our analysis , which will need to be revisited in future work . first , we have assumed the cib maps to be gaussian , whereas some non - gaussianity is in fact expected , both due to intrinsic non - gaussianity in the cib field and due to residual foregrounds or secondary anisotropies . if we assume perfect cmb cleaning , any non - gaussianity in @xmath37 is not expected to lead to biases , as the mean residual b - mode power spectrum is obtained from two powers of the cib field @xmath37 . though non - gaussianity in @xmath37 could potentially affect the size of the error bars , integration of many independent modes is expected to reduce the already - moderate non - gaussianity present in the two - halo - term dominated @xmath54 cib scales most important for delensing . however , this will need to be quantified in future work . second , we have assumed that the e and b - mode maps have been cleaned successfully using multifrequency data , so that any foreground residuals can be neglected . if the cmb component separation and cleaning is imperfect , additional complications can arise . though cib anisotropies are expected to be negligible in the large - scale cmb polarization ( as galaxy polarization directions add incoherently ) , any remaining galactic dust intensity features in the cib may be correlated with polarized galactic dust residuals in the cmb e and b maps . in principle , this could lead to additional gaussian and non - gaussian contributions to the delensed b - mode power spectrum and its error . while dust residuals are expected to be small in the cleaned cmb and cib maps , and the galaxy - dominated low-@xmath27 modes ( which overlap with the scales used to constrain @xmath0 ) can be excluded from the delensing cib map without significant losses , the magnitude of any biases and corrections are currently uncertain . therefore , though simple arguments suggest that the impact of foreground residuals and non - gaussianity on the delensing procedure is likely to be moderate , a more detailed investigation of these effects , with simulations and data , will be required in future work . we note that further investigation of some of these potential issues is also well - motivated for internal delensing methods . we have examined in detail the methods and performance of cib delensing , including combinations of different frequencies as well as combinations of cib maps with lensing reconstruction . our results show that cib delensing can remove more than @xmath143 of the lensing power and can hence lower the noise of a b - mode measurement by a factor of @xmath62 , and significantly more when cib maps are combined with lensing reconstruction . marginalizing over the true cib spectra , we find that uncertainty in the cib spectra does not degrade delensing performance significantly for small surveys , and only moderately reduces delensing performance for large surveys with high angular resolution . this confirms that , in most cases , cib delensing can be performed without assuming a model for the cib spectra . however , further investigation into systematic errors from foregrounds and non - gaussianity will be required in future work . at noise levels of @xmath144k - arcmin ( as targeted by cmb stage - iv ) or below , internal delensing by reconstructing the lensing field from the cmb itself is significantly more powerful than cib delensing . however , for many experiments that are currently planned , cib delensing appears competitive with internal delensing . even experiments such as core , litebird or prism @xcite are expected to have polarization noise levels of @xmath145k - arcmin or higher ( especially after foreground cleaning ) . at these noise levels , internal delensing does not perform much better than cib delensing ( e.g. , in the case of iterative delensing with @xmath145k - arcmin noise , the improvement factor is @xmath146 from @xcite , whereas cib delensing gives @xmath147 ) . perhaps most significantly , many current and upcoming surveys in the next decade do not have the high angular resolution required to delens well internally ( e.g. , keck array , bicep3 , litebird ; @xcite ) , so that external delensing is the only feasible method ; for these experiments , cib delensing is a promising option . cib delensing can significantly strengthen constraints on tensor b - mode polarization while requiring only minimal assumptions about cib properties . it is thus a robust and promising method for enhancing the upcoming generation of cmb polarization surveys . we now derive the optimal combination of multiple fields which trace large - scale structure , such as cib maps measured in different frequency channels ( without assuming redshift distributions for these fields ) , in order to maximize delensing performance . this optimal combination can be described by writing the tracer @xmath37 as a linear combination of different fields i ( ) = c_i i_i where @xmath70 are the different fields to be combined , @xmath71 are the linear combination coefficients and repeated indices are summed over . to construct a field which is optimal for delensing and cross - correlating , we now solve for the coefficients @xmath71 which maximize the correlation coefficient : ^2_l = = = we maximize by taking the derivative of @xmath148 with respect to @xmath149 0= ( ^2 ) & = & 2 - + & = & - we note that this equation can be rewritten as a matrix equation involving the optimal vector of coefficients @xmath149 or @xmath150 : 0= 2 - 2 , where we have defined the vector @xmath151 and the matrix @xmath152 as follows : k_i=^i_i , q_ij = ^i_i i_j it can thus be seen that the correlation coefficient is maximized when = n where @xmath153 is an arbitrary normalization factor , set by eq . 13 . for clarity , we will assume @xmath154 for now , and will verify that our normalization is correct by checking at a later stage that the normalization eq . 13 is satisfied by this choice . the optimal coefficients are thus given by = ^-1 rewriting the original expression for the correlation coefficient in our matrix formalism and inserting the solution , the maximal correlation coefficient is given by inspired by this result we rewrite the weights in terms of the correlation coefficients of two @xmath70 fields , @xmath155 , and the correlation coefficients of the @xmath70 fields with the true lensing convergence , @xmath156 . we can rewrite the matrix @xmath15 as ( no sum ) q_ij = _ ij ; ( q^-1)_ij = ( ^-1)_ij/ and k_i = _ i the power spectrum of this linear combination is ^ii & = & ( _ iajb _ ia ) ^ + & = & ^ and the cross - spectrum with @xmath21 takes the exact same value . hence ^ii = ^i = ^2 ^ we note that because @xmath157 , the linear combination is already correctly normalized ( for our choice of @xmath154 ) . grishuk , l. p. , 1975 j. exp . 40 , 409 starobinsky , a. a. , 1976 , zh . pisma red . 30 , 719 rubakov , v. a. , sazhin , m. v. , & veryaskin , a. v. 1982 , physics letters b , 115 , 189 fabbri , r. , & pollock , m. d. 1983 , physics letters b , 125 , 445 abbott , l. f. , & wise , m. b. 1984 , nuclear physics b , 244 , 541 lyth , d. h. 1997 , physical review letters , 78 , 1861 planck collaboration , ade , p. a. r. , aghanim , n. , et al . 2014c , astron . astrophys . , 571 , aa22 seljak , u. 1997 , , 482 , 6 kamionkowski , m. , kosowsky , a. , & stebbins , a. 1997 , physical review letters , 78 , 2058 seljak , u. , & zaldarriaga , m. 1997 , physical review letters , 78 , 2054 ade , p. a. r. , aikin , r. w. , barkats , d. , et al . 2014 , physical review letters , 112 , 241101 review letters , 78 , 1861 mortonson , m. j. , & seljak , u. 2014 , j. cosm . phys . , 10 , 035 flauger , r. , hill , j. c. , & spergel , d. n. 2014 , j. cosm . 8 , 039 planck collaboration , adam , r. , ade , p. a. r. , et al . 2014 , arxiv:1409.5738 bicep / keck and planck collaborations , ade , p. a. r. , et al . 2015 , http://new.bicepkeck.org/ hanson , d. , hoover , s. , crites , a. , et al . 2013 , physical review letters , 111 , 141301 the polarbear collaboration : ade , p. a. r. , akiba , y. , anthony , a. e. , et al . 2014 , physical review letters , 112 , 131302 the polarbear collaboration : ade , p. a. r. , akiba , y. , et al . 2013 , arxiv:1312.6646 the polarbear collaboration : p. a. r. ade , akiba , y. , anthony , a. e. , et al . 2014 , , 794 , 171 van engelen , a. , sherwin , b. d. , sehgal , n. , et al . 2014 , arxiv:1412.0626 story , k. t. , hanson , d. , ade , p. a. r. , et al . 2014 , arxiv:1412.4760 zaldarriaga , m. , & seljak , u. 1998 , , 58 , 023003 ade , p. a. r. , ahmed , z. , aikin , r. w. , et al . 2015 , arxiv:1502.00643 knox , l. , & song , y .- s . 2002 , physical review letters , 89 , 011303 kesden , m. , cooray , a. , & kamionkowski , m. 2002 , physical review letters , 89 , 011304 seljak , u. , & hirata , c. m. 2004 , , 69 , 043005 hu , w. , & okamoto , t. 2002 , , 574 , 566 hirata , c. m. , & seljak , u. 2003 , , 68 , 083002 hauser , m. g. , & dwek , e. 2001 , annu . astrophys . , 39 , 249 planck collaboration , ade , p. a. r. , aghanim , n. , et al . 2014a , astron . astrophys . , 571 , aa18 holder , g. p. , viero , m. p. , zahn , o. , et al . 2013 , astrophys . j. lett . , 771 , ll16 song , y .- s . , cooray , a. , knox , l. , & zaldarriaga , m. 2003 , , 590 , 664 smith , k. m. , hanson , d. , loverde , m. , hirata , c. m. , & zahn , o. 2012 , j. cosm . phys . , 6 , 014 viero , m. p. , wang , l. , zemcov , m. , et al . 2013 , , 772 , 77 viero , m. p. , moncelsi , l. , quadri , r. f. , et al . 2013 , , 779 , 32 bthermin , m. , le floch , e. , ilbert , o. , et al . 2012 , astron . , 542 , aa58 simard , g. , hanson , d. , & holder , g. 2014 , arxiv:1410.0691 planck collaboration , ade , p. a. r. , aghanim , n. , et al . 2015 , arxiv:1502.01589 lewis , a. , & challinor , a. 2006 , , 429 , 1 hu , w. 2000 , , 62 , 043007 challinor , a. , & lewis , a. 2005 , , 71 , 103010 smith , k. m. , hu , w. , & kaplinghat , m. 2004 , , 70 , 043002 challinor , a. , & chon , g. 2005 , mon . not . r. astron . soc . , 360 , 509 smith , k. m. , hu , w. , & kaplinghat , m. 2006 , , 74 , 123002 li , c. , smith , t. l. , & cooray , a. 2007 , , 75 , 083501 benoit - lvy , a. , smith , k. m. , & hu , w. 2012 , , 86 , 123008 planck collaboration , ade , p. a. r. , aghanim , n. , et al . 2014b , astron . astrophys . , 571 , aa30 kalberla , p. m. w. , mcclure - griffiths , n. m. , pisano , d. j. , et al . 2010 , astrophys . , 521 , aa17 niemack , m. d. , ade , p. a. r. , aguirre , j. , et al . 2010 , spie , 7741 , 77411s suzuki , a. , ade , p. , akiba , y. , et al . 2014 , journal of low temperature physics , 176 , 719 austermann , j. e. , aird , k. a. , beall , j. a. , et al . 2012 , proc . spie , 8452 , 84521e the core collaboration , armitage - caplan , c. , avillez , m. , et al . 2011 , arxiv:1102.2181 andr , p. , baccigalupi , c. , banday , a. , et al . 2014 , j. cosm . phys . , 2 , 006 kernasovskiy , s. , ade , p. a. r. , aikin , r. w. , et al . 2012 , spie , 8452 , 84521b ahmed , z. , amiri , m. , benton , s. j. , et al . 2014 , proc . spie , 9153 , 91531n matsumura , t. , akiba , y. , borrill , j. , et al . 2014 , journal of low temperature physics , 176 , 733 pearson , r. , & zahn , o. 2014 , , 89 , 043516 namikawa , t. , & nagata , r. 2014 , j. cosm . , 9 , 009 joachimi , b. , & bridle , s. l. 2010 , astron . astrophys . , 523 , aa1	as confusion with lensing b - modes begins to limit experiments that search for primordial b - mode polarization , robust methods for delensing the cmb polarization sky are becoming increasingly important . we investigate in detail the possibility of delensing the cmb with the cosmic infrared background ( cib ) , emission from dusty star - forming galaxies that is an excellent tracer of the cmb lensing signal , in order to improve constraints on the tensor - to - scalar ratio @xmath0 . we find that the maps of the cib , such as current planck satellite maps at 545 ghz , can be used to remove more than half of the lensing b - mode power . calculating optimal combinations of different large - scale - structure tracers for delensing , we find that co - adding cib data and external arcminute - resolution cmb lensing reconstruction can lead to significant additional improvements in delensing performance . we investigate whether measurement uncertainty in the cib spectra will degrade the delensing performance if no model of the cib spectra is assumed , and instead the cib spectra are marginalized over , when constraining @xmath0 . we find that such uncertainty does not significantly affect b - mode surveys smaller than a few thousand degrees . even for larger surveys it causes only a moderate reduction in cib delensing performance , especially if the surveys have high ( arcminute ) resolution , which allows self - calibration of the delensing procedure . though further work on the impact of foreground residuals is required , our overall conclusions for delensing with current cib data are optimistic : this delensing method can tighten constraints on @xmath0 by a factor up to @xmath1 , and by a factor up to @xmath2 when combined with external @xmath3k - arcmin lensing reconstruction , without requiring the modeling of cib properties . cib delensing is thus a promising method for the upcoming generation of cmb polarization surveys .
the main focus of klaus work during the last years of his life was the event generator vni which describes relativistic heavy - ion collisions and high energy particle collisions in terms of a perturbative parton shower monte carlo in phase - space combined with a simple space - time hadronization prescription . a partonic starting point is certainly required at ultrarelativistic collision energies , @xmath0 gev , where perturbative contributions start to account for a significant part of the measured particle production . with his parton cascade model @xcite , later amended by a space - time hadronization algorithm @xcite and made publicly available in form of the event generator vni @xcite , klaus geiger led the way . from summer 1997 until klaus death we collaborated with him and john ellis on implementing an afterburner for bose - einstein correlations into this code and testing it in simulations for @xmath1 collisions at lepi and lepii collisions . our goal was to prepare for an experimental test of the space - time dynamics predicted by vni and to provide theoretical guidance for multi - dimensional bose - einstein analyses of various event classes with hadronic final states generated at lep . this work @xcite was interrupted prematurely ; our contribution gives an account of the present status . vni distinguishes itself from many other high energy event monte carlos by following the event history in phase - space , not only in momentum space . _ space - time aspects _ enter in the numerical simulation in various ways : on the _ microscopic _ level , the rescattering between produced partons or hadrons is controlled by geometric cross sections , and hadronization is modelled geometrically by requiring partons to get ( in their pair rest frame ) closer than 0.8 fm in coordinate space in order to form a hadronic cluster . on a _ macroscopic _ level this leads to a strong density dependence of particle production and absorption rates , further affected by the collective expansion of the system which is generated by the rescattering . finally , quantum statistical effects among the produced final state particles , in particular bose - einstein correlations in momentum space between pairs of identical pions or kaons , depend on the _ phase - space density _ of the system at the point of decoupling and thus on both the momentum - space and space - time structure of the event at `` freeze - out '' ( i.e. at the point of the last strong interaction between the particles ) . in this sense all measured quantities in a heavy - ion collision depend to some extent on the space - time structure of the reaction zone . the crucial question is , however , whether they are sufficiently sensitive to such aspects to allow for a reconstruction of the space - time geometry and dynamics . the successful reproduction of single - particle yields and spectra by event generators with different space - time features , or none at all like the popular jetset @xcite and pythia @xcite generators , seems to argue against such a possibility . two - particle correlations in momentum space , however , are sensitive to both the geometric extension of the `` fireball '' at freeze - out and to its collective expansion dynamics @xcite . the latter affects the two - particle spectra via so - called `` @xmath2-@xmath3-correlations '' in the emission function @xmath4 . this function is the quantum mechanical analogue of the single - particle phase - space density of the source at freeze - out : collective expansion correlates the average direction and magnitude of the momenta of the emitted particles with their emission points . a detailed analysis of 2-pion and 2-kaon correlations in relativistic heavy - ion collisions at the brookhaven ags and the cern sps has recently led to an unambiguous demonstration of strong collective dynamics of the fireballs created in these experiments @xcite . on a finer level , however , there remain a number of open physical questions whose resolution requires two - particle correlation data of similar quality and detail from elementary particle collisions . such data do not exist , and another motivation for our work with klaus geiger was therefore to provide stimulation for similar experimental analyses of high - statistics high energy data samples like the @xmath5 million hadronic @xmath6 decays collected at lepi . for sufficiently high secondary particle multiplicities ( i.e. at very high energies or for large collision systems ) it is reasonable to assume that the two particles of a selected pair were emitted independently . this allows to express the two - particle correlation function in terms of the single - particle wigner density @xmath4 of the source @xcite . neglecting final state interactions ( assuming that they can be corrected for experimentally @xcite or theoretically @xcite at a later stage ) , it is given by @xcite @xmath7 here @xmath8 are the on - shell momenta of the two particles in the pair while @xmath9 and @xmath10 are their average and relative 4-momenta . the numerator in the second term stands for the product of the measured single - particle spectra @xmath11 where @xmath12 . the normalization @xmath13 will be discussed below . if , for a given momentum @xmath14 , the space - time dependence of @xmath15 can be characterized with reasonable accuracy by a single set of rms widths ( i.e. the particle emission is not characterized by several widely differing length scales ) , one can approximate @xmath15 by a gaussian in @xmath2 . the correlation function then takes the simple form @xcite @xmath16 where @xmath17 describes the second space - time moments ( rms widths or `` homogeneity lengths '' ) of the _ effective source _ of particles with momentum @xmath14 . since the two measured particles are on - shell , @xmath18 , @xmath19 and only 3 of the 4 components of @xmath20 are independent . different choices for the independent components lead to different gaussian parametrizations of the correlation function @xcite . we will here use the cartesian parametrization which eliminates @xmath21 where @xmath22 is ( approximately ) the velocity of the particle pair . @xmath23 is decomposed into its cartesian components @xmath24 where @xmath25 denotes the `` longitudinal '' direction ( in heavy - ion collisions this is the direction of the beam axis , in @xmath1 collisions the direction of the thrust axis ) , @xmath26 denotes the outward direction , fixed by the azimuthal orientation of the transverse pair momentum @xmath27 around the @xmath25-axis , and @xmath28 denotes the third cartesian ( sideward ) direction ( defined by @xmath29 , @xmath30 ) . eliminating @xmath31 from the exponent in ( [ eq3b ] ) leaves a sum over 6 terms ; using furthermore the azimuthal symmetry of the event sample around the @xmath25-axis allows to further reduce this to 4 terms involving certain algebraic sums @xcite of the rms widths @xmath17 . so far , however , bose - einstein correlations in elementary particle collisions have been parametrized by much simpler gaussian forms , involving only one or at most two ( @xmath14-independent ! ) size parameters . such incomplete parametrizations have the fundamental disadvantage that the fit parameter(s ) mix(es ) the interesting space - time information contained in the rms widths @xmath17 of the effective source of particles with momentum @xmath14 in such a way that they can no longer be recovered @xcite . an important goal of our work with klaus was to make predictions for the multidimensional shape of the correlation function @xmath32 , by calculating the complete set of size parameters and their @xmath14-dependence . the latter is particularly interesting since it signals @xmath2-@xmath14-correlations in the emission function @xmath15 . while in heavy - ion collisions the dominant mechanism for such correlations appears to be collective expansion of the fireball @xcite , contaminations from resonance decays after freeze - out @xcite and from @xmath2-@xmath3-correlations in the primary hadron formation process @xcite ( e.g. from string fragmentation ) are known to exist . the latter are expected to play a much bigger role in elementary particle collisions where a multidimensional analysis of bose - einstein correlations may help to isolate them . this would provide crucial input into a quantitative discussion of bose - einstein correlations from heavy - ion collisions where , contrary to elementary particle collisions where the multiplicities are much lower , most resonance decays can _ not _ be reconstructed experimentally and their effects must therefore be simulated . event generators like vni evolve classical _ probabilities _ , not quantum mechanical _ amplitudes _ calculated from properly symmetrized many - particle wave functions . the bose - einstein correlations among pairs of identical pions or kaons must thus be included _ a posteriori_. for notational convenience , we restrict the following discussion to a single particle species , say like - sign pions . let us assume that we have generated @xmath33 collision events , and let the @xmath34th event ( @xmath35 ) consist of @xmath36 such pions in the final state , emitted as free particles from the phase - space points @xmath37}$ ] . for @xmath33 events , the event generator thus simulates a classical phase - space distribution @xmath38 what is needed to calculate two - particle bose - einstein correlations according to ( [ eq3 ] ) is a prescription which relates this classical phase - space density with the quantum mechanical single - particle wigner density @xmath4 of the pion source which is supposed to be monte carlo simulated by the event generator . in sections [ sec2a ] and [ sec2b ] , we focus on two different interpretations of the event generator output which we call `` classical '' and `` quantum '' , respectively , although these names should not distract from the fact that conceptually both are on an equal footing . each one of them leads to a different algorithm for calculating the two - particle correlation function from the event generator output : @xmath39 } \bigr\rbrace_{m\in [ 1,n_{\rm ev } ] } \longrightarrow \ c(\bq,\bk)\ , , \label{eq1}\ ] ] before discussing them we list three requirements which any such algorithm should fulfill : 1 . since the event generator evolves classical probabilities , not symmetrized production amplitudes , the generated momenta @xmath40 , @xmath41 do not show the enhanced probability at low relative momenta @xmath42 characteristic for bose - einstein final state symmetrization . the prescriptions ( [ eq1 ] ) should calculate this effect a posteriori . 2 . the strength of bose - einstein correlations depends on the distance of the identical particles in phase - space , not in momentum space . we thus require the prescriptions ( [ eq1 ] ) to use the entire phase - space information , and not only the generated momentum information . ` weighting ' or ` shifting ' prescriptions which are based only on the latter @xcite may successfully match the measured momentum correlations but obviously do not allow to test the simulated space - time structure . 3 . in general , bose - einstein statistics can affect particle multiplicity distributions during the particle production process but classical event generators do not include such effects explicitly . nevertheless they are tuned to reproduce the measured average particle multiplicities @xmath43 . in order not to destroy this tuning we require that the prescription ( [ eq1 ] ) conserves the single - particle multiplicities . if the event generator were also tuned to reproduce the average multiplicity of identical particle _ pairs _ , @xmath44 , i.e. to reproduce not only the mean , but also the width of the multiplicity distribution , then the prescription ( [ eq1 ] ) should not change that either . interpreting the correlator as a factor which relates the measured two - particle differential cross section to the one simulated by the event generator , @xmath45 , this then implies @xcite that @xmath46 in ( [ eq3 ] ) . the algorithms discussed below do not satisfy this last requirement , i.e. in general they change the width of the multiplicity distribution . however , since the space - time analysis of correlation data can be based entirely on the momentum dependence of @xmath32 , irrespective of its absolute normalization , this does not matter for our purposes . in the `` classical '' interpretation @xcite , the phase - space points @xmath47 are seen as a discrete approximation of the on - shell wigner phase - space density @xmath4 : @xmath48 this defines both the two - particle correlator via ( [ eq3 ] ) and the one - particle spectrum via ( [ eq3a ] ) . for a numerical implementation one must replace the momentum - space @xmath49-functions in ( [ eq2 ] ) by normalized `` bin functions '' with finite width @xmath50 of rectangular @xcite or gaussian @xcite shape , e.g. @xmath51 which reduce to the original @xmath49-function in the limit @xmath52 . the one - particle spectrum and two - particle correlator then read @xcite @xmath53 @xmath54 \over \sum_{m=1}^{n_{\rm ev } } \left [ \left(\sum_{i=1}^{n_m } \delta^{(\epsilon)}_{\check{\bp}_i,\bp_1}\right ) \left(\sum_{j=1}^{n_m } \delta^{(\epsilon)}_{\check{\bp}_j,\bp_2}\right ) - \sum_{i=1}^{n_m } \delta^{(\epsilon)}_{\check{\bp}_i,\bp_1 } \delta^{(\epsilon)}_{\check{\bp}_i,\bp_2 } \right]}\ , . \label{eq13}\ ] ] the correlator ( [ eq13 ] ) is the discretized version of the fourier integral in ( [ eq3 ] ) . the subtracted terms in the numerator and denominator remove discretization errors which would amount to pairs constructed from the same particles . this `` classical '' algorithm , as well as the `` quantum '' version discussed below , is numerically efficient since it involves only @xmath55 manipulations per event . the calculated observables , while being discrete functions of the input , are continuous functions of the measured momenta , i.e. no binning in @xmath56 is necessary . the `` quantum '' interpretation @xcite of the event generator output starts from the observation that , for a given event , i.e. a single term in the sum of ( [ eq2 ] ) , the simultaneous sharp definition of the particle momenta and positions at emission violates the uncertainty relation . in the limit @xmath57 the sum is still expected to be a smooth phase - space function and , to the extent that the event generator provides an accurate simulation of the underlying qcd quantum dynamics , it is also expected to respect the uncertainty constraints on any allowed wigner density . since in practice , however , one has to work with finite numbers of events , one may wish to ensure consistency with the uncertainty principle on the single - event level . this is achieved @xcite by associating the centers of single - particle wave packets with the set of generated phase - space points @xmath58 } \rbrace_{m\in [ 1,n_{\rm ev}]}$ ] : @xmath59 the @xmath60 describe quantum mechanically best - localized states , i.e. they saturate the heisenberg uncertainty relation with @xmath61 and @xmath62 . taking only two - particle symmetrized contributions into account ( `` pair approximation '' @xcite ) , all spectra can be written @xcite in terms of the single - particle spectrum @xmath63 corresponding to an individual wave packet at phase - space point @xmath64 : @xmath65 \over \sum_{m=1}^{n_{\rm ev } } \left [ \nu_m(\bp_1)\ , \nu_m(\bp_2 ) - \sum_{i=1}^{n_m } s_i(\bp_1)\,s_i(\bp_2 ) \right]}\ , , \label{eq6 } \\ s_i(\bp ) & = & \left(\sigma^2 \over\pi\right)^{3/2}\ , e^ { - \sigma^2 ( \bp-\check{\bp}_i)^2}\ , . \label{eq7 } \end{aligned}\ ] ] again , the subtracted terms in the numerator and denominator of @xmath32 are finite multiplicity corrections which become negligible for large particle multiplicities @xcite . this algorithm is consistent with an emission function @xmath66 which is obtained by folding the classical distribution @xmath67 of wave packet centers with the wigner density @xmath68 of a single wave packet : @xmath69 the latter saturates the uncertainty relation with a spatial uncertainty @xmath70 and a momentum uncertainty @xmath71 , and the folding ensures that now @xmath66 is always quantum mechanically consistent . however , in this algorithm both the one- and two - particle spectra depend on the wave packet width @xmath70 . the role of this parameter will be discussed in the context of our toy model study in section [ sec3 ] . the `` classical '' and `` quantum '' algorithms differ only with respect to two points : 1 . the `` classical '' algorithm has no analogue for the gaussian prefactor @xmath72 in ( [ eq6 ] ) which is a genuine quantum effect stemming from the quantum mechanical localization properties of ( [ eq4 ] ) . the gaussian single - particle distributions @xmath63 in the `` quantum '' algorithm are the quantum analogues of the `` bin functions '' in the `` classical '' agorithm . with the gaussian bin functions ( [ eq11 ] ) the two agree for the choice @xmath73 . finite event statistics puts a lower practical limit on @xmath50 in the `` classical '' algorithm , but to get accurate spectra one should try to choose @xmath50 as small as possible , by simulating sufficiently many events . in contrast , @xmath70 in the `` quantum '' algorithm is the finite spatial width of the single - particle wave packets , and the limit @xmath74 ( which corresponds to @xmath52 ) is not physically relevant : according to ( [ eq8 ] ) it amounts to an emission function with infinite spatial extension and thus to a correlator @xcite @xmath75 . before describing realistic event generator simulations we discuss some analytical results for a classical toy model distribution @xmath67 . these are then used to test the algorithms of sections [ sec2a ] and [ sec2b ] , by applying those to sets of phase - space points @xmath76}$ ] generated from the model distribution with a monte carlo procedure . in this way we can make quantitative statements about i ) the @xmath70-dependence of the `` quantum '' algorithm , ii ) the @xmath50-dependence of the `` classical '' algorithm ( especially : how small @xmath50 has to be chosen to extract the limit @xmath52 numerically ) and iii ) the statistical requirements for the algorithms to work . the toy model , first introduced by zajc @xcite , reads : @xmath77 the distribution is normalized to a total event multiplicity @xmath78 . the parameter @xmath28 smoothly interpolates between completely position - momentum correlated and uncorrelated sources . for @xmath79 , the position - momentum correlation vanishes and we are left with the product of two gaussians in @xmath80 and @xmath81 . in the limit @xmath82 @xmath83 the position - momentum correlation is perfect . the total phase - space volume of the distribution @xmath84 vanishes for @xmath85 . this strong @xmath28-dependence allows to study the performance of our numerical algorithms for different phase space volumes . in the following subsections , we discuss the @xmath28-dependence of the one - particle spectrum and two - particle correlator , focussing in sections [ sec3a ] and [ sec3b ] on analytical results , and comparing these in section [ sec3c ] to numerical calculations . inserting ( [ eq14 ] ) into ( [ eq10])-([eq13 ] ) one finds @xcite @xmath86 we recover the physical hbt radius @xmath87 from ( [ eq20 ] ) in the limit @xmath52 or by inserting ( [ eq10 ] ) directly into ( [ eq3 ] ) . the remarkable fact is that for sufficiently large @xmath28 @xcite , @xmath88 the hbt radius @xmath89 becomes negative and the two - particle correlator shows an unphysical rise of the correlation function with increasing @xmath90 . the change of sign in ( [ eq19 ] ) seems to be related to the violation of the uncertainty relation by the emission function for large @xmath28 when @xmath80 and @xmath81 become strongly correlated . at the critical value @xmath91 the phase - space volume @xmath92 of the source drops below 1 . only for @xmath93 the distribution @xmath94 is a quantum mechanically allowed emission function , i.e. a wigner density . the practical importance of the zajc model in the unphysical limit @xmath95 is that it provides an extreme testing ground for our numerical algorithms . analytically , we conclude already from ( [ eq20 ] ) that in order to be close to the physical limit the bin width @xmath50 has to be small on the scale of the width of the generated momentum distribution , @xmath96 in the above toy model this requirement is independent of the strength of position - momentum correlations in the source . inserting the model distribution @xmath94 into ( [ eq8 ] ) ( instead of ( [ eq10 ] ) ) , we find @xmath97 in this case , @xmath98 and @xmath99 satisfy @xmath100 independent of the value of @xmath70 and , in contrast to the `` classical '' algorithm , the radius parameter @xmath101 is now always positive , irrespective of the value of @xmath28 . even if the classical distribution @xmath102 is sharply localized in phase - space , its folding with minimum uncertainty wave packets leads to a quantum mechanically allowed emission function @xmath103 and to a correlator with falls off with increasing @xmath90 as expected . however , the spread of the one - particle momentum spectrum ( [ eq23 ] ) receives an additional contribution @xmath104 . choosing @xmath70 too small increases this term beyond the phenomenologically reasonable values , choosing it too large widens the corresponding hbt radius parameters significantly . it was argued @xcite that @xmath70 can be interpreted as quantum mechanical `` size '' of the particle , @xmath105 fm . given the heuristic nature of these arguments and the significant modifications this implies for the spectra ( [ eq23 ] ) and ( [ eq24 ] ) , it is however fair to say that presently @xmath70 mainly plays the role of a regulator of unwanted violations of the quantum mechanical uncertainty principle while a deeper understanding of its origin in the particle production dynamics is still missing . we have studied the performance of our bose - einstein algorithms by generating with a random number generator sets @xmath106}$ ] of phase - space points according to the distribution @xmath94 and comparing the numerical results of our algorithms to the analytical expressions of section [ sec3a ] and [ sec3b ] . = 9.5 cm fig . [ fig1](a ) shows the @xmath50-dependence of the hbt radius parameter ( [ eq20 ] ) . for fixed bin width @xmath50 , the approximation of the true hbt radius parameter by @xmath107 is seen to become better with increasing @xmath108 , in agreement with ( [ eq22 ] ) . for the hbt radius obtained from the `` quantum '' version of the zajc model and depicted in fig . [ fig1](b ) , the situation is both qualitatively and quantitatively different . now , the hbt radius is always positive , since the gaussian wave packets take quantum mechanical localization properties automatically into account . also , the @xmath28-dependence of the radius is seen to be much weaker since the wave packets smear out the unphysically strong position - momentum correlations present in the classical distribution @xmath94 . the hbt radius depends not only on the geometrical size @xmath109 , and on the momentum width @xmath108 of the source , but also on the wave packet width @xmath70 . as seen in fig . [ fig1](b ) , this wave packet width affects the hbt radius and its @xmath108-dependence significantly for @xmath110 . in figs . [ fig1](c , d ) we present the two - particle correlation functions corresponding to these hbt radius parameters for characteristic values of the model parameters . the analytical results , obtained by plotting ( [ eq19 ] ) and ( [ eq24 ] ) , are compared to the results from the event generator algorithms ( [ eq6 ] ) and ( [ eq13 ] ) . the plot was obtained using @xmath111 events of multiplicity @xmath112 . the differences between analytical and numerical results originate from statistical fluctuations and become smaller with increasing number of events @xmath33 or event multiplicity @xmath36 . the good agreement between analytical and numerical results in figs . [ fig1](c , d ) indicates the relatively low statistical requirements of our algorithms . the reason is that both algorithms associate with the _ discrete _ set of generated momenta @xmath40 _ continuous _ functions of the measured momenta @xmath113 , @xmath114 . this smoothens any statistical fluctuation significantly . for the `` classical '' algorithm , small deviations between numerical and analytical results are still visible in fig . [ fig1](c ) , while the results of the `` quantum '' algorithm coincide within line width . this can be traced back to the gaussian prefactor @xmath115 in ( [ eq6 ] ) which provides an additional smoothening of statistical fluctuations not present in the `` classical '' algorithm . we have applied the `` quantum '' algorithm discussed above to simulated @xmath1 collisions at lepi @xcite and lepii @xcite energies from vni . we have focussed on the channels ( see fig . [ fig2 ] ) @xmath116 = 8truecm = 8truecm which provide the `` cleanest '' environment for the study of bose - einstein correlations in high - energy particle collisions . especially for the first channel there exists an impressively extensive and accurate data sample of several million events . for high - energy @xmath1 collisions , theoretical studies of bose - einstein enhancements have mainly been performed within the context of the string models @xcite , which have been quite successful in reproducing the distributions of identical particle pairs on the level of 1-parameter gaussian fits of the correlator @xcite . our study @xcite is based on the event generator vni and aimed at a full inclusion of the space - time structure of the events and a multidimensional correlation analysis . the `` classical '' and `` quantum '' interpretation of the event generator output discussed in section [ sec2b ] and [ sec2a ] provide _ two different prescriptions _ for the calculation of two - particle correlations . this ambiguity in the calculational scheme can be traced back to the fact that both algorithms are _ a posteriori _ remedies for an incomplete quantum mechanical time evolution which does not account properly for the quantum mechanical symmetrization of identical @xmath78-particle states . the ambiguity has to be removed by a physical consistency requirement . from our discussion of the zajc model in section [ sec3 ] , first crude statements about such consistency requirements can be made : 1 . the `` quantum '' interpretation introduces a wave packet width @xmath70 which must be adjusted to data . the measured hbt radius parameters provide an upper bound on @xmath70 , @xmath117 . hence , the `` quantum '' interpretation can only be consistent with experimental data if the wave packet width is not too large . 2 . in order for the `` classical '' interpretation of the event generator output to be consistent with experimental data , the generated output must not be peaked too strongly in phase - space . otherwise , the hbt radius may be unphysically small or even show the wrong sign , cf . section [ sec3a ] . the first step in a realistic study of two - particle correlations is necessarily to refine these crude statements . to what extent do physical observables depend on the choice of algorithm ? which choice of @xmath70 is consistent with experiment ? are there phenomenological reasons to prefer one of the algorithms ? these workshop proceedings reflect the state of our work when klaus left us . at the time of the airline accident we had only completed the calculation with the `` quantum '' algorithm , and only for a single value of the wave packet width @xmath70 . a comparison of the two algorithms remains to be done . [ fig3 ] shows the correlator @xmath118 for different @xmath23-values and vanishing pair momentum @xmath119 in the c.m . frame of the collision , @xmath120 . the widths of the correlator in three cartesian directions are roughly the same . quantitatively , they are roughly given by the inverse of the wave packet width @xmath70 which suggests that they are dominated = 7.2 cm = 7.2 cm by the `` quantum mechanical smearing '' features of the `` quantum '' bose - einstein algorithm used here . a simulation with the `` classical '' algorithm remains to be done . one also sees that near @xmath121 the correlation functions show characteristic deviations from a gaussian shape . however , at non - zero values of the orthogonal @xmath23-components , the correlators become nicely gaussian . the non - gaussian features at small @xmath23 can be traced to decay contributions from long - lived resonances : fig . [ fig4 ] shows the same correlator = 13.5 cm with and without such contributions . neglecting the pions from long - lived resonance decays , the correlation function becomes gaussian and wider , reflecting a smaller source size . with long - lived resonances included , the effective pion source @xmath66 is larger , resulting in a narrower and non - gaussian correlation function . this clearly demonstrates the sensitivity of the correlation function on the space - time _ geometry _ of the source function @xmath66 . as discussed before , the dependence of the correlation function on the pair momentum @xmath119 indicates @xmath2-@xmath14-correlations in the source and can thus be sensitive to the space - time _ dynamics _ of hadron production . fig . [ fig5 ] shows the correlation function @xmath122 of same - sign pions for various values of the pair momentum @xmath123 . clearly , as @xmath119 increases , the correlation function becomes wider , indicating a smaller effective source , qualitatively ( although not quantitatively ) very similar to the corresponding observations in heavy - ion collisions . this is an interesting prediction which to our knowledge has not yet been tested in the lep experiments . according to the more detailed analysis presented in @xcite the observed @xmath119-dependence can be fully attributed to the @xmath119-dependence of resonance decay contributions ; however , this may be to some extent an artifact of the employed `` quantum '' bose - einstein algorithm whose wave packet width @xmath70 apparently dominates the widths of the correlation functions shown above . the rather abrupt ending of the above section on physics results gives a sad feeling for the gap left by klaus . in our collaboration , he was the only one who actually knew how to run vni . our original motivation came from relativistic heavy - ion collisions where the space - time geometry and dynamics of the event plays a crucial role in understanding basic measurable quantities . mainly for this reason we developed algorithms which allow to calculate identical two - particle correlation functions from an arbitrary event generator output . we tested the accuracy and statistical requirements of these algorithms in simple toy model studies and applied them in a first realistic study to the hadronic channels in @xmath124 @xmath125 annihilation at the @xmath6-peak and near the @xmath126-@xmath127 threshold . however , a comparative study of both algorithms is still missing and our main goal , the application of these algorithms to the study of event simulations of heavy ion collisions , is not achieved yet . we plan to do so in the near future . also , we plan to make the algorithms described in section [ sec2 ] publicly available , using the open standard for codes and routines ( oscar ) format , advocated in @xcite . geiger , k. , and mller , b. , ; geiger , k. , _ phys . _ d * 47 * , 133 ( 1993 ) ; _ phys . rep . _ * 258 * , 237 ( 1995 ) . ellis , j. , and geiger , k. , ; and . geiger , k. , _ comput . _ * 104 * , 70 ( 1997 ) . wiedemann , u.a . , ellis , j. , heinz , u. , and geiger , k. , in _ cris98 : measuring the size of things in the universe : hbt interferometry and heavy ion physics _ , ( s. costa et al . , eds . ) , world scientific , singapore , 1999 . geiger , k. , ellis , j. , heinz , u. , and wiedemann , u.a . , hep - ph/9811270 , submitted to _ phys . _ d. lnnblad , l. , and sjstrand , t. , _ phys . _ b * 351 * , 293 ( 1995 ) ; _ eur . c * 2 * , 165 ( 1998 ) . sjstrand , t. , _ comput . commun . _ * 39 * , 347 ( 1986 ) ; sjstrand , t. , and bengtsson , m. , _ comput . commun . _ * 43 * , 367 ( 1987 ) . wiedemann , u.a . , and heinz , u. , nucl - th/9901094 , _ phys . _ , in press . heinz , u. , jacak , b.v . , nucl - th/9902020 , _ ann . rev . nucl . part . * 49 * ( 1999 ) , in press . anchishkin , d.v . , heinz , u. , and renk , p. , _ phys . _ c * 57 * , 1428 ( 1998 ) ; sinyukov , yu . , et al . , _ phys . _ b * 432 * , 248 ( 1998 ) . heinz , u. , in _ correlations and clustering phenomena in subatomic physics _ harakeh , o. scholten , and j.h . koch , eds . ) , nato asi series b * 359 * , 137 ( 1996 ) ( plenum , new york ) . schlei , b.r . , ornik , u. , plmer , m. , and weiner , r.m . , _ phys . _ b * 293 * , 275 ( 1992 ) ; bolz , j. , et al . , _ phys . _ b * 300 * , 404 ( 1993 ) ; and _ phys . _ d * 47 * , 3860 ( 1993 ) ; wiedemann , u.a . , and heinz , u. , _ phys . rev . _ c * 56 * , r610 and 3265 ( 1997 ) . bialas , a. , zalewski , k. , _ acta phys . pol . _ b * 30 * , 359 ( 1999 ) . wiedemann , u.a . _ c * 57 * , 3324 ( 1998 ) . wiedemann , u.a . et al . , _ phys . _ c * 56 * , r614 ( 1997 ) . zhang , q.h . , wiedemann , u.a . , slotta , c. , and heinz , u. , _ phys . _ b * 407 * , 33 ( 1997 ) . merlitz , h. , and pelte , d. , _ z. phys . _ a * 357 * , 175 ( 1997 ) . zajc , w.a . , in _ particle production in highly excited matter _ gutbrod and j. rafelski , eds . ) , nato asi series b * 303 * , 435 ( 1993 ) ( plenum , new york ) . hebbeker , t. , _ phys . rep . _ * 217 * , 69 ( 1992 ) ; bethke , s. , and pilcher , j.e . , _ ann . rev . nucl . part . sci . _ * 42 * , 251 ( 1992 ) . ballestrero , a. , et al . , _ j. phys . _ g * 24 * , 365 ( 1998 ) . andersson , b. , gustafson , g. , ingelman , g. , and sjstrand , t. , _ phys . rep . _ * 97 * , 33 ( 1983 ) ; andersson , b. , gustafson , g. , and sderberg , b. , _ nucl . _ b * 264 * , 29 ( 1986 ) . bowler , m.g . , _ z. phys . _ c * 29 * , 617 ( 1985 ) ; andersson , b. , and hoffmann , w. , _ phys . _ b * 169 * , 364 ( 1986 ) ; andersson , b. , and ringner , m. , _ phys . _ b * 421 * , 283 ( 1998 ) ; hkkinen , j. , and ringner , m. , _ eur . c * 5 * , 275 ( 1998 ) . pang , y. , _ nucl . _ a * 638 * , 219c ( 1998 ) .	we report on work done in collaboration with klaus kinder - geiger and john ellis which aims at connecting the space - time structure of event generator simulations with observable output . # 1#2#3#4#1 * # 2 * , # 3 ( # 4 ) _ combined presentation of the talks given by the two authors at the workshop + `` rhic physics and beyond kay kay gee day '' + brookhaven national laboratory , 23 october 1998 _ klaus geiger was an important driving force in our search for understanding the dynamics of ultrarelativistic heavy - ion collisions and quark - gluon plasma . an unconventional and unique character this is how we will remember him . physics was his passion , but he also loved his porsche and his mg ( = `` ma geiger '' ) . it is hard for us to accept that by his untimely death klaus , whose research was dedicated so much to the future , in particular to the rhic program at brookhaven , should be imprisoned forever in the past . as his friends and collaborators , we will try to carry on his legacy .
hot molecular cores ( hereafter hmcs ) are small , dense , hot , and dark molecular clumps , usually found in the proximity of ultracompact hii regions ( e.g. , kurtz et al . 2000 ) . unlike hii regions , these objects present weak or undetectable free - free emission . the lack of free - free emission has been interpreted as due to an intense mass accretion phase that quenches the development of an hii region ( e.g. , walmsley 1995 ) . therefore , these hmcs may be the precursors of hii regions , tracing the earliest observable stage in the life of massive stars . a detailed modeling of the emission of the hmcs is a necessary requirement to assess their physical nature and their relationship with the birth of massive protostars . osorio , lizano , & dalessio ( 1999 ) showed that the spectral energy distribution ( sed ) of the dust continuum emission of several hmcs without detectable free - free emission is consistent with that of a massive envelope collapsing onto a b type star , with an age less than @xmath9 yr , and accreting mass at a rate of @xmath10@xmath11 . these intense mass accretion rates are high enough to prevent the development of an ionized region around the central star . the physical properties of the hmcs were inferred using the collapse of the singular logatropic sphere ( sls ; mclaughlin & pudritz 1997 ) as a dynamical model for the envelope . the sls collapse models were chosen because these envelopes are massive enough to reproduce the strong millimeter emission observed , without requiring excessively high mass accretion rates , thus preventing excessive emission in the near and mid infrared ( see more details in osorio et al . 1999 ) . one can further test osorio s models , including the velocity field of the accreting envelopes , by comparing their expected molecular emission with observations of appropriate molecular tracers . high - excitation ammonia transitions are well suited for this kind of comparison , since they trace the hot and warm dense gas in the hmcs , and can be observed with subarcsecond angular resolution using the very large array ( vla ) . such observations are among the highest angular resolution molecular line data of hmcs that can be obtained at present , allowing to determine the variations of the line intensity across the sources . furthermore , the ammonia lines have hyperfine structure that is sensitive to the optical depth and permit an accurate measurement of the column density . in order to do this test , we selected the hmc located @xmath12 to the southwest of the peak of the g31.41 + 0.31 ultracompact hii region . hereafter , we will call this source g31 hmc . this hmc is located at a distance of 7.9 kpc ( cesaroni et al . 1998 ) . it is an interesting hmc because it exhibits strong millimeter continuum emission , as well as molecular emission of high excitation lines ; it is associated with a group of water masers , and with a bipolar outflow ( cesaroni et al . 1994 , 1998 ; gibb , wyrowski & mundy 2004 ) . the presence of high excitation molecular emission in this luminous source indicates that g31 hmc ( together with a few other objects ; beuther & walsh 2008 , longmore et al . 2007 ) is one of the hottest molecular cores discovered so far . the lack of strong free - free centimeter emission ( araya et al . 2003 ) indicates that g31 hmc could be in a phase of intense mass accretion , being unable to develop a detectable hii region . therefore , we consider g31 hmc as a good candidate to test the basic hypothesis of osorio s model , namely , that young massive stars are formed inside hmcs with very high accretion rates . in this model the onset of a detectable ultracompact hii region is quenched by the very high mass accretion rate . g31 hmc has a database of continuum flux density measurements covering a wide range of wavelengths although , unfortunately , most of the data points were obtained with relatively poor angular resolution and the sed is poorly constrained for @xmath13 m . g31 hmc is associated with line transitions from a variety of molecular species , such as hco@xmath14 , sio ( maxia et al . 2001 ) , @xmath15co ( olmi et al . 1996 ) , cs ( anglada et al . 1996 ) , h@xmath16s , c@xmath17o ( gibb , mundy & wyrowski 2004 ) , and ch@xmath18cn ( beltrn et al . 2005 ) . particularly interesting is the association of g31 hmc with strong ammonia emission ( churchwell et al . 1990 , cesaroni et al . 1992 ) whose ( 4,4 ) inversion transition has been observed with subarcsecond angular resolution using the vla ( cesaroni et al . 1998 ; see their figs . 2c , 4c , and 9c ) . this makes g31 hmc one of the few sources where a high signal - to - noise ratio analysis of the spatial variation of the ammonia emission along the core can be carried out . the high intensity of the observed ammonia ( 4,4 ) emission , the ratio of the main to satellite lines close to unity , as well as the unusually large line widths indicate extreme physical conditions that suggest that this core is harboring an o type star . because of this , we chose g31 hmc as a candidate to test the molecular emission predictions of the osorio s model for the early stages of the formation of massive stars . we note that vla observations at 7 mm ( araya et al . 2003 ) reveal the presence of a binary system at the center of g31 hmc . however , the model of osorio et al . ( 1999 ) is spherically symmetric , with a single central source of luminosity . this assumption is valid as long as the binary separation ( @xmath19 ) is much smaller than the observed size of the hmc ( @xmath20 ) . likewise , deviations from spherical symmetry either by rotation ( disks ) or by an intrinsic elongation of the cloud are not considered by osorio s model . to test non - spherical models , one must include high angular resolution mid - ir data that would allow to properly constrain parameters such as the degree of flattening of the envelope , the centrifugal radius , or the effect of a disk and/or cavities caused by outflows . this kind of data has been obtained for other hmcs and is highly sensitive to the geometry of the cloud ( e.g. , de buizer , osorio & calvet 2005 ) . unfortunately , such observations are not yet available for g31 hmc . observations of several molecular tracers reveal , however , the existence of small velocity gradients in g31 hmc that have been interpreted as suggestive of either outflow ( gibb , wyrowski , & mundy 2004 , araya et al . 2008 ) or rotation motions ( beltrn et al . 2005 ) . sensitive high angular resolution observations are needed to clarify this issue . despite the possible presence of small velocity gradients , we expect the kinematics of the core to be dominated by the overall infall motions of the envelope . in this work we assume that the spherically symmetric models of osorio et al . ( 1999 ) are adequate to study the dust continuum and molecular line emission of g31 hmc . the paper is organized as follows . in @xmath21 , we describe the modeling procedure , in @xmath22 we present the observational data available from the literature , in @xmath23 we model the dust and ammonia data and we compare our results with the observations , as well as with other modeling attempts of this source reported in the literature . finally , in @xmath24 we summarize our conclusions . following osorio et al . ( 1999 ) , we model a hmc as a spherically symmetric envelope of dust and gas that is collapsing onto a recently formed central massive star . an accretion shock is formed at the stellar surface . thus , the stellar radiation and the accretion luminosity provide the source of heating of the envelope . the temperature of the envelope is a function of the distance to the center . the density , infall velocity , and turbulent velocity dispersion for each point of the envelope are obtained from the solution of the dynamical collapse of the sls ( mclaughlin & pudritz 1997 ) . the parameters of the sls model are chosen by calculating the emerging dust emission , assuming a constant dust to gas mass ratio along the envelope , and fitting the observed sed . once obtained the physical conditions in the envelope from the fit to the observed sed , the excitation of the ammonia molecule is calculated and the radiative transfer is performed in order to obtain the emerging ammonia spectra for different lines of sight towards the envelope . in these calculations , the gas - phase ammonia abundance , that is a function of radius , is the only free parameter . we consider two simple cases : uniform gas - phase abundance and uniform total ammonia abundance . in the latter case , the ratio of gas to solid phase ammonia as a function of radius is determined by calculating the balance between condensation and sublimation of molecules on dust grains . the gas - phase ammonia abundance is constrained by comparing the ammonia model spectra with the observed spectra towards different positions . for the collapsing envelope , we adopt the density , @xmath25 , infall velocity , @xmath26 , and turbulent velocity dispersion , @xmath27 , distributions resulting from the solution of the self - similar collapse of the sls ( adams , lizano & shu 1987 , unpublished notes ) . the sls has a logatropic pressure , @xmath28 , where @xmath29 is a constant that sets the pressure scale , and @xmath30 is an arbitrary reference density , introduced by lizano & shu ( 1989 ) to empirically take into account the observed turbulent motions in molecular clouds . the collapse of the sls has been studied by mclaughlin & pudritz ( 1996 , 1997 ; hereafter mp97 ) , and it has been generalized by reid , pudritz & wadsley ( 2002 ) , and sigalotti , de felice & sira ( 2002 ) to include three - dimensional hydrodynamic calculations of the collapse of both singular and nonsingular logatropic spheres . in the sls collapse solution an expansion wave moves outward into a static cloud and sets the gas into motion towards the central star . both the speed of the expansion wave and the mass accretion rate increase with time . the radius of the expansion wave is given by @xmath31 where @xmath32 is the time elapsed since the onset of collapse . outside the radius of the expansion wave ( @xmath33 ) , the sls envelope is static , @xmath34 and the density is given by @xmath35 inside the expansion wave ( @xmath36 ) , the infall velocity , whose direction is radial , is given by @xmath37 having a free - fall behavior for @xmath38 . the density is @xmath39 where @xmath40 is the similarity variable , and @xmath41 and @xmath42 are non - dimensional functions . with this normalization the expansion wave is located at @xmath43 . the self - similar variable , and the density and velocity functions are related to those tabulated by mclaughlin & pudritz ( 1997 ) through @xmath44 , @xmath45 , and @xmath46 , where the subindex `` mp '' labels the mclaughlin & pudritz ( 1997 ) solution . as noted by osorio et al . ( 1999 ) the sls collapse tends to produce massive envelopes , since inside the radius of the expansion wave only 3% of the mass is in the central star , while 97% is in the collapsing envelope . the velocity dispersion inside the envelope due to turbulent motions is obtained assuming that alfvn waves in the cloud induce fluid motions with velocity amplitudes , @xmath47 , of the order of the wave speed , @xmath48 , where the magnitude of the magnetic field is @xmath49 . for random polarizations and random orientations of the magnetic field in the cloud , the one - dimensional velocity dispersion in the line - of - sight direction is given by @xmath50 where the factor 1/3 comes from averaging over the solid angle the square of the magnetic field projections along the line of sight . the physical structure of the sls collapse is characterized by the constant @xmath29 and the time elapsed since the onset of collapse , @xmath32 . in terms of familiar variables , this elapsed time is given by @xmath51 where @xmath52 is the mass accretion rate , and @xmath53 is the mass of the central star . @xmath29 can be written as @xmath54 where @xmath55 is the reduced ( dimensionless ) mass . from now on , we will use @xmath53 and @xmath52 to characterize the dynamical model . the total source of heating of the envelope , @xmath56 , is assumed to be the sum of the stellar luminosity , @xmath57 , and the accretion luminosity , @xmath58 , where @xmath59 is the stellar radius . a value of @xmath60 cm has been adopted ( see osorio et al . 1999 , hosokawa & omukai 2008 ) . the stellar luminosity is related to the stellar mass using the schaller et al . ( 1992 ) evolutionary tracks . the temperature of the dust grains inside the envelope , @xmath61 , is self - consistently calculated from the total luminosity using the condition of radiative equilibrium for outer optically thin regions of the envelope , whereas for the inner optically thick regions the temperature is calculated from the standard diffusion approximation ( see details in osorio et al . 1999 ) . for temperatures @xmath62 k and densities @xmath63 @xmath64 , the gas and dust are well coupled and are described by the same temperature ( e.g. , sweitzer 1978 ; doty et al . our models fulfill this condition and , thus , the kinetic temperature of the gas , @xmath65 , is taken to coincide with the temperature obtained from the dust calculations , @xmath66 . assuming that the dust grains shield the molecules from the intense central radiation field , the dust destruction front , at the dust sublimation temperature ( @xmath67 k ) , determines the inner radius , @xmath68 , of the dust and molecular envelope . the external radius , @xmath69 , of the envelope is inferred from the observations . to solve the radiative transfer equation for the dust emission we have the geometry depicted in figure [ fig : cartoon ] . given the spherical symmetry of the envelope , the emergent specific intensity , @xmath70 , for a given line of sight is only a function of the impact parameter , @xmath71 , the distance between the line of sight and the center of the envelope . assuming that the source function of the thermal dust emission is the planck function evaluated at the dust temperature , @xmath72 , we have the following set of equations that allow us to obtain the emerging intensity as a function of frequency and impact parameter , @xmath73 @xmath74 and @xmath75 where @xmath76 is the background intensity , @xmath77 is the monochromatic dust absorption coefficient per unit mass , @xmath78 is the coordinate along the line of sight ( @xmath79 ) , positive in the direction away from the observer , and whose origin is in the plane perpendicular to the line of sight that contains the center of the envelope ( see fig . [ fig : cartoon ] ) . the integration limits @xmath80 and @xmath81 correspond to the edges of the envelope nearest and farthest to the observer , respectively , being @xmath82 . in the absence of a strong background source , @xmath83 will be taken as the cosmological background , @xmath84 , where @xmath85 k. following osorio et al . ( 1999 ) , the dust absorption coefficient , @xmath77 , at high frequencies ( @xmath86 ghz ) is obtained from dalessio ( 1996 ) , who considered spherical ( mie ) particles of graphites , silicates , iron , and water ice compounds , with the standard grain - size distribution of the interstellar medium , @xmath87 ( mathis , rumpl , & nordsieck 1977 ) , with a minimum radius of 0.005 @xmath88 m and a maximum radius of 0.3 @xmath88 m . the abundance and optical constants of the compounds were taken from draine & lee ( 1984 ) , draine ( 1987 ) , and warren ( 1984 ) . the value of the absorption coefficient at 1500 ghz obtained for this dust composition and size distribution can be extrapolated to lower frequencies in the standard power - law form , resulting @xmath89 @xmath90 g@xmath3 for @xmath91 ghz . the value of the index @xmath92 is adopted so that it is consistent with the slope of the optically thin millimeter emission of the observed sed , where @xmath93 . we assumed a constant dust - to - gas mass ratio of 0.01 . depending of the wavelength , the dalessio ( 1996 ) opacities can be a factor of up to 2 - 3 times lower than those of ossenkopf & henning ( 1994 ) . both opacity models have been successfully used to reproduce the sed of other high mass star forming cores ( osorio et al 1999 , van der tak et al 2000 ) . once the emergent intensity as a function of the impact parameter is obtained , the total flux density of the source at a given frequency is calculated as @xmath94 where @xmath95 is the distance of the observer to the source . finally , the resulting model sed is calculated and compared with the observed sed . a grid of models is run with different values of the free parameters @xmath96 and @xmath53 ( or , equivalently , @xmath57 ) . in this way , the model ( or models ) that are consistent with the observed sed are determined . note that in case the sed is not well constrained , more than one model can be consistent with the data . further details on the method are described by osorio et al . ( 1999 ) . in this section , we describe the calculation of the emerging intensity of the ammonia inversion transitions arising from a spherically symmetric envelope with the physical properties obtained from the dust modeling using the sls solution . the physics of the ammonia molecule is studied in detail in townes & slawlow ( 1975 ) , ho ( 1975 ) , and ho & townes ( 1983 ) . here we only summarize some basic information that may be relevant to our discussion . because ammonia is a symmetric top molecule , its rotational energy levels are described by two quantum numbers , namely , the total angular momentum , @xmath97 , and its projection on the symmetry axis , @xmath98 ( @xmath99 ) . rotational levels with @xmath100 are more populated and are called metastable , while rotational levels with @xmath101 are short - lived and are called non - metastable . all rotational levels ( except those with @xmath102 ) are split into inversion doublets . the transitions between these inversion doublets are called inversion transitions and occur at wavelengths of @xmath81.3 cm . as shown in appendix [ apentex ] , the excitation temperature of the ammonia inversion transitions can be obtained from a two - level model . for a given point of coordinate @xmath78 along a line of sight of impact parameter @xmath71 , we can estimate the excitation temperature , @xmath103 , from equation ( [ eq : dosni2 ] ) , taking the total number density of gas molecules as @xmath104 ) , where @xmath105 is the mean molecular weight in the envelope and @xmath106 is the hydrogen mass . in the calculation of @xmath107 we will approximate the intensity of the local radiation field by the background intensity ( @xmath108 ) . with this approximation , equation ( [ eq : dosni2 ] ) gives a lower limit for the excitation temperature . however , as shown in appendix [ apentex ] , for the physical conditions of hmcs the values obtained in this way are very close to the value of the kinetic temperature , which , for thermal lines , is an upper limit for the excitation temperature . therefore , this indicates that taking @xmath109 is a very good approximation for the calculation of the excitation temperature in hmcs . note that , with this approximation , the resulting excitation temperature will become a spherically symmetric function , @xmath110 . the spectra of the inversion transitions present a hyperfine structure due to the interaction of the electric quadrupole momentum of the nitrogen nucleus with the electric field of the electrons . this effect splits the line into five components , the central main line , and two pairs of satellite lines , separated @xmath81 - 2 mhz , symmetrically placed about the main line . these hyperfine components are further split by weaker magnetic spin interactions . since the separation between the magnetic hyperfine components is only @xmath810 - 40 khz ( 0.1 - 0.5 km s@xmath3 ) , they are distinguishable only in very high spectral resolution observations of regions with very low velocity dispersion . for this reason , we neglect the magnetic hyperfine structure , and we restrict our analysis to only the five electric quadrupole hyperfine components . as shown in appendix [ apenkappa ] ( eq . [ [ eq : opafinal ] ] ) , the absorption coefficient of the ammonia @xmath111 inversion transition , at a given point of coordinate @xmath78 along a line of sight of impact parameter @xmath71 ( see fig . [ fig : cartoon ] ) , as a function of the observed lsr velocity , @xmath112 , is given by @xmath113 where the frequency @xmath114 corresponds to that of the main line of the inversion transition ( given in table [ tbl : inver ] ) , @xmath115 is the local gas - phase ammonia abundance relative to h@xmath16 , @xmath116 is the local dispersion of the distribution of line - of - sight velocities due to turbulent and thermal motions , @xmath117 is the local rotational temperature , @xmath118 is the statistical weight of the @xmath111 rotational level ( given by equation [ [ eq : pesos ] ] ) , @xmath119 its rotational energy ( given by equation [ [ eq : energy ] ] ) , @xmath120 is the partition function ( given by equation [ [ eq : particion ] ] ) , @xmath121 is the relative strength of the @xmath122th hyperfine component ( @xmath123 ) and @xmath124 its doppler velocity shift with respect to the main line ( both given in table [ tab : propnh3 ] ) , @xmath125 is the lsr velocity of the ambient cloud , and the term @xmath126 is the line - of - sight component of the infall velocity of the envelope . in writing equation ( [ eq : opafinal2 ] ) it has been assumed that @xmath127 , corresponding to assume a 10% he abundance . note that the excitation temperature , @xmath107 , and the einstein spontaneous emission coefficient , @xmath128 , are those of the entire inversion transition , unsplit by hyperfine interactions ( eq . [ [ eq : dosni2 ] ] and table [ tbl : inver ] , respectively ) . due to the dipolar selection rules on the rotational transitions , the populations of the metastable levels are mainly determined by collisions ; therefore , @xmath129 , that is defined by the relative populations of the rotational levels is expected to be similar to the kinetic temperature . in fact , @xmath129 is a lower limit of @xmath65 , and its value is different for each pair of rotational levels considered ( see appendix [ apenkappa ] ) . however , the agreement between @xmath129 and @xmath65 is expected to improve for high metastable levels , due to the increased excitation involved , and for high densities , where all temperatures tend to thermalize to @xmath65 ( e.g. , walmsley & ungerechts 1983 ; danby et al . 1988 ; wilson et al . 2006 ) . this is the case for g31 hmc ( osorio 2000 ) , and hereafter we will take @xmath130=@xmath65 . in the above equation , the dispersion of the line - of - sight velocity distribution results from the contribution of the macroscopic ( turbulent ) and microscopic ( thermal ) components , @xmath131 where @xmath27 is given by equation ( [ eq : dispersion ] ) and @xmath132 is given by @xmath133 where @xmath134 is the mass of the ammonia molecule . the only free parameter in equation ( [ eq : opafinal2 ] ) is the gas - phase ammonia abundance , @xmath115 . the smallest values of the ammonia abundance reported in the literature are found in the coolest cores of low - mass star forming regions , with typical abundances of @xmath135-@xmath136 ( e.g. , herbst & klemperer 1973 , ungerechts et al . 1980 , estalella et al . 1993 ) , and some chemical models predict abundances as low as @xmath137-@xmath138 ( le bourlot et al . 1993 ) . the highest values , obtained in very hot regions of massive star formation ( e.g. , millar 1997 , walmsley 1997 , ohishi 1997 ) , are @xmath139-@xmath140 . neglecting chemical effects , we consider the two simplest possibilities for the function , @xmath115 , that describes the gas - phase ammonia abundance inside the envelope . first , we consider the case in which @xmath115 has a constant value across the envelope . as a second possibility , we consider the case in which the total ( solid+gas ) ammonia abundance remains constant across the envelope but the gas - phase abundance increases towards the center because of the temperature gradient . variations in the gas - phase molecular abundances have already been explored by , e.g. , van der tak et al . ( 2000 ) , and boonman et al . ( 2003 ) to explain the ch@xmath18oh , hcn , and h@xmath16o emission in some regions of massive star formation . we assume a minimum gas - phase ammonia abundance , @xmath141 , in the outer , cooler parts of the core , where the bulk of the ammonia molecules are frozen in grain mantles . the ammonia molecules are released from the grain mantles to the gas phase , reaching a maximum value of the gas - phase ammonia abundance , @xmath142 , in the inner , hotter regions of the core , where it is assumed that all the ammonia molecules are in the gas phase . under this simple approximation , the gas - phase ammonia abundance as a function of the distance to the center of the core will be given by @xmath143 where @xmath144 is the ratio between the solid and gas phases , obtained from the balance between the condensation and sublimation of molecules on dust grains following the thermal equilibrium equation of sandford & allamandola ( 1993 ) ( see appendix [ apeneta ] ) . two possible cases are considered . first , it is assumed that the ammonia molecules are directly frozen in grain mantles , with @xmath144 described by equation ( [ eq : etanh3 ] ) ; in this case , the ammonia molecules are released to the gas phase at the ammonia sublimation temperature , which is of @xmath860 k for the range of densities typical in hmcs ( see fig . [ fig : sketch ] ) . as a second possibility , it is assumed that the ammonia molecules are trapped in water ice , being released to the gas phase only after sublimation of water molecules , that occurs at a temperature of @xmath8100 k as described by equation ( [ eq : etah2o ] ) ( see fig . [ fig : sketch ] ) . as in the case of the continuum emission ( @xmath145 [ modeldust ] ) , we adopt the geometry of figure [ fig : cartoon ] , and the emergent specific intensity , @xmath146 , will be given by @xmath147 where @xmath107 is given by equation ( [ eq : dosni2 ] ) , and @xmath148 is given by equation ( [ eq : opafinal2 ] ) . as for the dust calculations , in the absence of a strong background source , the background intensity will be taken as the cosmological background , @xmath149 , where @xmath150 k. the geometrical parameters are the same as in equation ( [ eq : tralin1p ] ) . finally , in order to make an accurate comparison with the observations , the emergent intensity is convolved ( in 2-d ) with a gaussian with a fwhm equal to that of the observing beam , and a velocity smoothing is applied to reproduce the spectral resolution of the spectrometer . note that @xmath151 and @xmath142 are the only parameters to be fitted by the molecular line modeling . numerous observations of the g31.41 + 0.31 region are reported in the literature . the main results of the continuum flux density measurements in the wavelength range @xmath152 m to 1.3 cm are summarized in table [ datos ] . most of these observations did not have the angular resolution required to separate the emission of the hmc from that of the g31.41 + 0.31 uchii region , whose emission peak is only @xmath12 apart . thus , it is likely that in the data obtained with angular resolution @xmath153 the dust continuum emission of the hmc is contaminated by emission originated in the nearby uchii . therefore , we have considered these data as upper limits . in addition , we note that the flux density at 1.3 cm is likely dominated by free - free emission . then , we expect the dust thermal emission at this wavelength to be well below the upper limit set by the observations , and therefore , in our modeling , we did not attempt to reproduce the emission at 1.3 cm and longer wavelengths . finally , we considered the 7 mm data point as a lower limit since it is likely that a fraction of the flux density is missed in the very high angular resolution interferometric vla observations at this wavelength . observations of several line transitions from a variety of molecular species have been carried out towards g31 hmc . among them , the vla b - array observations of the nh@xmath18 ( 4,4 ) inversion transition ( vla project code ah0483 ) reported by cesaroni et al . ( 1998 ) ( combined with c - array data from cesaroni et al . 1994 ) constitute one of the best sets of high angular resolution data towards g31 hmc and objects of the same kind published so far . these combined data have an angular resolution of @xmath154 ( corresponding to a spatial scale of 5000 au at the distance of 7.9 kpc of the source ) , high enough to spatially resolve the structure of the core , revealing its extreme physical conditions and their variation inside the core . thus , we used this dataset to carry out the main part of our spectral line data analysis . figure 9c of cesaroni et al . ( 1998 ) shows the ammonia ( 4,4 ) spectra as a function of the impact parameter with respect to the center of g31 hmc . as can be seen in the figure , the brightness temperature is very high , reaching a value of approximately 100 k for the line - of - sight towards the center of the core , indicating high values of the kinetic temperature . the optical depth is also high , as indicated by the fact that the satellite lines , whose opacity is @xmath81/60 that of the main line , reach an intensity similar to that of the main line towards the center of the core . also , lines are unusually broad , resulting in the blending of contiguous satellite lines , indicative of large turbulent motions . this figure also indicates that the physical conditions in the core change as a function of the distance to the center , as shown by the variation of the line intensity and ratio between main line and satellites for different impact parameters . in addition to the subarcsecond resolution ammonia data of cesaroni et al . ( 1998 ) , churchwell et al . ( 1990 ) and cesaroni et al . ( 1992 ) report a multitransitional ammonia study of g31 hmc carried out with the effelsberg 100 m telescope . these observations have an angular resolution of @xmath155 . although the study lacks the angular resolution to be sensitive to variations of the line parameters across the core , providing only the integrated ammonia emission of the overall core , these data are useful to further test the model predictions derived from the analysis of the dust and high angular resolution ammonia ( 4,4 ) line data . in the next section we present a simultaneous analysis of the dust continuum and ammonia line data that will provide a powerful diagnostic to constrain the physical conditions inside g31 hmc . the observed sed of g31 hmc is characterized by strong millimeter emission suggesting the presence of either a hot or a dense envelope ( or a combination of both ) . therefore , `` a priori '' we expect the models that are able to fit the observed sed to be characterized by a high mass accretion rate . bearing this in mind , we ran a grid of models covering a relatively wide range of values of the free parameters @xmath53 and @xmath52 . the mass and luminosity of the central star have been related using the evolutionary tracks of schaller et al . ( 1992 ) for a solar metallicity . we tested the same values of the mass given in the tables of schaller et al . in the range @xmath156 ( corresponding to @xmath157 ) , and for the mass accretion rate we explored the range @xmath158 . we set the value of @xmath92 so that the slope of the model sed was consistent with that of the observed sed in the millimeter wavelength range , resulting in values of @xmath92 in the range 1 - 1.2 . we set the external radius to @xmath69 = 30,000 au , a value inferred from the observed images of the source ( e.g. , cesaroni et al . 1994a , maxia et al . 2001 ) and required to reproduce the large flux densities observed in the millimeter range . because of the incompleteness of the observed sed of g31 hmc , there is more than one model that can fit the available observational data . in table [ tbl5 ] we give the parameters of a set of models ( corresponding to values of the stellar mass listed in the tables of schaller et al . 1992 ) that are consistent with the observed sed . the higher luminosity ( and hotter envelope ) model consistent with the observed sed has a central star with a mass of @xmath159 and an accretion rate of @xmath160 . models with @xmath161 produce far- and mid - ir flux densities that exceed the observational constraints . lower luminosity models ( @xmath162 ) that are consistent with the sed data have colder envelopes , but they have a higher density ( see the values of the temperature and density at @xmath163 au listed in the table ) in order to account for the strong millimeter emission observed in g31 hmc , resulting in higher values for the mass of the envelope . note that a low value of @xmath53 does not imply that the source is a low - mass protostar since in table [ tbl5 ] the models with the smaller central mass are also the younger sources , meaning simply that the time elapsed since the onset of collapse , @xmath32 , is still too short to accumulate a massive central object . in fact , as can be seen in the table , these models are associated with the more massive envelopes and would eventually form the higher mass stars . the value of the inner radius of the envelope , @xmath68 , is @xmath164 au for model i ( @xmath159 ) and gets smaller as the mass of the central star decreases . we note that in all these models the radius of the expansion wave , @xmath165 , is smaller than the radius of the core , @xmath69 ; therefore , the outermost regions of these envelopes are static . also , the infall accretion rates in these models are so high that the accretion luminosity is the most important source of energy . for a given value of @xmath53 , the observational uncertainties in the sed data points result in a variation of @xmath166 in the possible values of @xmath52 that are consistent with the data . this results also in a formal uncertainty of the order of 10% in the derived parameters listed in table [ tbl5 ] for a given value of @xmath53 . figure [ fields ] shows the physical structure of the sls envelopes corresponding to the models listed in table [ tbl5 ] . as the figure illustrates , model i has the higher values of the velocity and temperature , and the lower values of the density and velocity dispersion . turbulent motions are the most important contributors to the local velocity dispersion at large radii of the envelope , reaching values of @xmath167 km s@xmath3 in the outermost regions . here we obtain the nh@xmath18 emission for the models that are consistent with the observed sed of g31 hmc to further test their physical properties by comparing their emission with the vla observations of the nh@xmath18(4,4 ) inversion transition reported by cesaroni et al . ( 1998 ) . in order to reproduce the setup of the cesaroni et al . observations , the emerging ammonia intensity as a function of the lsr velocity , @xmath112 , is calculated for each line of sight through the core and convolved with a gaussian beam of hpbw = @xmath154 . the spectra have been converted into a brightness temperature scale using the relation @xmath168 . a velocity range of @xmath8 150 km s@xmath3 is covered , and hanning smoothing is applied to obtain an effective spectral resolution of 4.86 km s@xmath3 . a value of @xmath169 km s@xmath3 has been adopted for the lsr velocity of the ambient cloud . figure 9c of cesaroni et al . ( 1998 ) shows the spectra obtained by averaging the observed ammonia emission over circular annuli around the position of the core center . therefore , in order to make more accurate the comparison of the model results with the observations , the spectra have been calculated for the impact parameters @xmath71 = 0 , 5000 , 10000 , and 15000 au , corresponding to the values sampled in this figure . we first ran a grid of cases with different values of the gas - phase ammonia abundance , assumed to be constant along the envelope . we tested a wide range of values , from @xmath170 to @xmath171 , for all the models resulting from the fit to the observed sed ( discussed in @xmath145 [ compardust ] ) . after running all these cases , we conclude that none of these models is able to reproduce the observed spectra . figure [ fig : grid1 ] shows the synthetic spectra obtained for model i at different impact parameters and for different ammonia abundances . the spectra observed by cesaroni et al . ( 1998 ) are also shown for comparison . this figure shows that model i can not reproduce the observed behavior of the ammonia emission as a function of impact parameter . for a low ammonia abundance ( left column ) , the main line reaches the observed brightness temperature of @xmath172 k in the @xmath173 panel ; however , the satellite lines are too weak in all the panels . increasing the ammonia abundance ( middle and right columns ) produces an increase of the intensity of the satellite lines , that reach a value close to the observed one ( @xmath172 k ) in the @xmath71 = 0 au panel of the right column . nevertheless , in this panel the main line reaches a brightness temperature of only @xmath5 k , which is much smaller than the observed value . this occurs because the main line is very optically thick ( reaching an opacity of @xmath174 ) and , thus , the brightness temperature must be roughly equal to the kinetic temperature of the outer parts of the envelope model , which are cold . note that the brightness temperature of the satellite lines is higher since their opacity is 60 times lower , and they can reach the inner parts of the envelope where the temperatures are higher . models ii to xi are colder than model i ( see fig . [ fields ] ) and therefore , they are worse in reproducing the observed behavior of the ammonia emission . as for model i , for a low abundance the satellite lines are too weak in all the panels . the intensity of the satellite lines increases with the ammonia abundance but , as in model i , for high ammonia abundances the main line becomes very optically thick in the @xmath71 = 0 au panels ( @xmath175 ) and only traces cold material from the outer envelope . since models ii to xi are colder than model i , the line intensities are lower and remain well below the observed values . clearly , there is no hope to reproduce the observations with constant values of the ammonia abundance . a high optical depth is required to reproduce the intense satellite lines , but this high opacity prevents that the main line is formed in the inner , hottest regions resulting in a main line brightness temperature too low . nevertheless , the analysis of constant abundance suggests the alternative of having a low abundance in the outer envelope and a high abundance in the inner parts . in this way , the colder , outer parts of the envelope would have a smaller contribution to the observed emission , that would instead trace inner , higher temperatures . one expects that a compromise could be reached , in which the ammonia abundance in the outer parts of the envelope is low enough for the main line not to be saturated until reaching the inner , hotter regions , where the abundance may be high enough for even the satellite lines to become optically thick . in addition , having a low abundance in the outer parts of the envelope would prevent the main line to become saturated in the outer panels , helping to explain the observed decrease of the intensity of the main line as the impact parameter increases . the rationale for a variable ammonia abundance is provided by the calculation of the sublimation of molecules from dust grain mantles ( see appendix [ apeneta ] ) . given the temperature gradient in the envelope , the outer parts have temperatures below that of sublimation of grain mantles and it is expected that most of the ammonia molecules are trapped there . the ammonia gas - phase abundance should increase drastically in the inner regions of the envelope , where the kinetic temperature reaches the sublimation temperature of the mantles and the ammonia molecules are released to the gas phase . therefore , inclusion of the process of sublimation of ammonia molecules from grain mantles should provide a more realistic description of the abundance distribution required in the calculation of the ammonia emission . in order to try to fit the observed ammonia emission using a variable gas - phase ammonia abundance , we first ran a grid of cases with the ammonia abundance calculated using equation ( [ eq : escalon ] ) , and with the ratio , @xmath176 , between the solid and gas phases given by equation ( [ eq : etanh3 ] ) of appendix [ apeneta ] ( that corresponds to direct sublimation of nh@xmath18 ices ) . this procedure assumes that most of the ammonia molecules are frozen into grain mantles in the outer parts of the envelope , being released to the gas phase at the ammonia sublimation temperature ( about 60 k ; see appendix [ apeneta ] and fig . [ fig : sketch ] ) . in this way , the gas - phase ammonia abundance behaves like a step function , with a minimum value , @xmath151 , in the external parts of the envelope ( where the temperature remains below the ammonia sublimation temperature ) , and increasing steeply to its maximum value , @xmath142 , for points inside the radius where the ammonia sublimation temperature is reached . we have explored values for the outer gas - phase nh@xmath18 abundance in the range @xmath177 , and in the range @xmath178 for the inner gas - phase nh@xmath18 abundance . however , none of the models resulting from the fit to the observed sed ( @xmath145 [ compardust ] ) is able to reproduce reasonably well the nh@xmath18(4,4 ) observations . the main problem is that the maximum brightness temperatures of the main line obtained with this kind of ammonia abundance distribution are too low ( @xmath179 k ) , to account for the peak brightness temperatures of @xmath172 k observed in g31 hmc . apparently , this occurs because the main line becomes optically thick in regions with temperatures of the order of 60 k , the temperature at which the bulk of ammonia molecules that were frozen in grain mantles are released to the gas phase . thus , the difficulties to fit simultaneously the main and satellite lines are similar to those of the case of a constant gas - phase ammonia abundance . as already mentioned , an alternative scenario is that ammonia molecules are mixed with water ice in grain mantles ( as suggested , e.g. , by ahearn et al . 1987 for comets , and by brown et al . 1988 and osorio 2000 for hmcs ) . in this case , ammonia molecules will be trapped in the grain mantles until temperatures high enough for sublimation of water molecules are reached . therefore , the gas - phase ammonia abundance will be described by equation ( [ eq : escalon ] ) , where @xmath176 corresponds now to the water sublimation and is given by equation ( [ eq : etah2o ] ) of appendix [ apeneta ] . as illustrated in figure [ fig : sketch ] , the transition from low to high gas - phase ammonia abundance occurs at a higher temperature ( @xmath172 k ) than in the case of pure ammonia sublimation . as a result , optically thick lines will trace hotter regions of the envelope , resulting in higher brightness temperatures . as in the case of pure ammonia sublimation we ran a grid of cases with values for the outer gas - phase nh@xmath18 abundance in the range @xmath180 , and in the range @xmath181 for the inner gas - phase nh@xmath18 abundance . we found that models ii to xi can not fit the nh@xmath18 observations . the envelopes of the lower luminosity models ( @xmath182 ) are apparently too cold to fit the observed spectra . for these models the sublimation temperature of 100 k is reached at a radius @xmath183 au ( see fig . [ fields ] ) , and therefore the spectra calculated for the outer impact parameters ( @xmath184 au ) behave very much like a case of constant gas - phase ammonia abundance equal to @xmath151 ( since the lines of sight for @xmath185 au only cross the outer regions of the envelope , whose temperatures remain below the sublimation temperature ) . since models ii to xi could not reproduce the observed spectra with a constant abundance , it is not expected that these models can fit the spectra in the outer panels , even with a variable gas - phase ammonia abundance . in summary , models with @xmath182 are too cold to reproduce the observed properties of the nh@xmath18(4,4 ) spectra , even with a variable gas - phase nh@xmath18 abundance . in contrast , we have been able to obtain a fairly good fit to the observed spectra using model i ( @xmath186 ) . in figure [ fig : gridvariable ] ( central column ) we show our best fit spectra , obtained using this model with a value of @xmath187 and a value of @xmath188 . this set of parameters reproduces quite well the observed spectra . it reproduces the observed intensity ( @xmath189 k ) of the main line towards the center of the core ( @xmath173 ) , as well as the variation of its intensity as @xmath71 increases . it also reproduces the observed intensity of the satellite lines in the @xmath173 and @xmath190 au spectra . in the @xmath191 and @xmath192 au spectra the observed intensity of the satellite lines is somewhat stronger than predicted , but its signal - to - noise ratio is poor and therefore we consider this discrepancy less significant . lines are broad ( note that the satellite lines on each side of the main line appear blended ) , and the model also reproduces quite well the observed line widths ( @xmath193 km s@xmath3 ) . interestingly , the best fit is obtained with values of the abundance for the cold and hot parts of the envelope that coincide respectively with the typical values inferred ( usually assuming a constant abundance ) from the observations of low mass ( cold ) and high mass ( hot ) star - forming cores ( see @xmath145 [ am1 ] ) , giving additional support to the goodness of the fit . we also show in figure [ fig : gridvariable ] two additional cases , corresponding to a lower value of the abundance in the hot region of the envelope , @xmath194 ( left column ) , and to a higher value of the abundance in the cold region , @xmath195 ( right column ) , to illustrate the behavior of the resulting spectra when varying these parameters . as figure [ fig : gridvariable ] illustrates , the satellite lines are mainly sensitive to variations in @xmath142 ( left and middle columns ) , whereas the main line is sensitive to variations in @xmath151 ( middle and right columns ) . at large values of @xmath71 ( @xmath71 = 10000 and 15000 au ) , the spectra do not change by varying @xmath142 because the temperatures along these lines of sight are lower than 100 k and , therefore , sublimation will not occur . figure [ fig : gridvariable ] ( left column ) also shows that an abundance of @xmath196 ( ten times lower than in the best fit ) in the inner regions of the envelope is high enough to reproduce the observed brightness temperature of the main line , but it is too low to reproduce the satellite lines . on the other hand , as the right column shows , an abundance of @xmath197 in the outer parts of the envelope ( ten times higher than in the best fit model ) is adequate to reproduce the brightness temperature of the satellite lines , but not for the main line , whose peak brightness temperature decreases to a value below the observed one . this is because for such a high value of @xmath141 the main line becomes very optically thick and the observed emission originates in the outer ( and , therefore , colder ) regions of the core . since the spacing between values of @xmath53 listed in table [ tbl5 ] , that correspond to those given by schaller et al . ( 1992 ) , is too coarse , especially in the high mass range , we explored additional cases obtaining the relationship between @xmath53 and @xmath57 by interpolation of the tables of schaller et al . in this way , we performed a fine tuning of the mass of the central star and determined the range of values of the parameters that is consistent with the observational uncertainties of the data . the results are summarized in table [ tbl5b ] . in summary , the dust continuum sed and the spatial variation of the ammonia spectra can be explained by a dense envelope with a central massive star of mass @xmath53 = 20 - 25 @xmath198 , undergoing an intense accretion at a rate @xmath52 = 2 - 3 @xmath199 , and with a total luminosity of @xmath200 . the mass of the envelope is @xmath201 = 1400 - 1800 @xmath198 and it has a considerable static region that surrounds the collapsing region . the value of the magnetic field is @xmath202 = 5 - 6 mg . the outermost region has a velocity dispersion of @xmath203 km s@xmath3 , dominated by turbulent motions . the temperature of the envelope ranges from @xmath5 k ( at @xmath204 au ) to @xmath67 k ( at @xmath205 = 130 - 160 au ) . the gas - phase ammonia abundance inside the envelope is described by a step - like function with a minimum value of 1 - 4 @xmath206 in the outer , colder ( @xmath207 k ) region , and 2 - 4 @xmath208 in the inner , hotter ( @xmath209 k ) region . this distribution of the gas - phase ammonia abundance results from the sublimation of ammonia molecules trapped in water ice grain mantles . the minimum and maximum values of the ammonia abundance are similar to the typical values reported in low and high mass protostellar cores , respectively . our results for g31 hmc can be compared with the calculations for this source made by other authors . cesaroni et al . ( 1998 ) carried out an analysis of the nh@xmath18(4,4 ) main line assuming local thermodynamic equilibrium and that the line is optically thick . they claimed that the observed gradient in the peak brightness temperature is a good estimate of the radial distribution of the kinetic temperature of the gas inside the core . since the observed peak brightness temperature profile of g31 hmc does not coincide with the expected kinetic temperature profile of a sphere , these authors concluded that neither a collapsing nor an expanding spherical envelope can explain the properties of the observed emission . this conclusion was based on a large velocity gradient ( lvg ) hypothesis , that implies that for any line of sight with @xmath210 the only contribution to the peak brightness temperature ( at zero velocity relative to the ambient cloud ) arises from the plane perpendicular to the line of sight that contains the center of the core ( since the line - of - sight component of the infall velocity of this material is strictly zero ) . however , if lines are broad ( as is the case in g31 hmc ) , the lvg hypothesis is not valid since material from outside this plane can also contribute to the emission at the center of the line . furthermore , if the envelope has a static region ( as suggested by our analysis ) the peak brightness temperature would have a significant contribution from the material in this static region . therefore , this analysis does not seem adequate to infer the radial temperature distribution in the envelope and , consequently , the possibility of a collapsing spherical envelope can not be discarded . cesaroni et al . ( 1998 ) favored a geometrically thin disk seen almost face on ( therefore , with a single temperature for each line of sight ) to account for the g31 hmc nh@xmath18(4,4 ) emission . however , in order to reach a very high optical depth with such a geometrically thin disk a very high surface density would be required . this hypothesis was not tested by cesaroni et al . ( 1998 ) , and it is unclear if could correspond to a realistic scenario . also , the satellite lines were not considered in the analysis , and it is also unclear how the observed variation of the main line to satellites ratio as a function of impact parameter can be reproduced . therefore , we conclude that although we followed a simplified approach ( a spherically symmetric model with a single central source of luminosity ) , ignoring deviations from spherical symmetry produced either by rotation ( disks ) or due to the intrinsic elongation of the cloud , our model accounts for some of the basic observed properties of g31 hmc , especially the highest angular resolution line observations , and constitutes a step ahead in modeling these structures . to further test our modeling , we calculated the emerging emission for ammonia transitions other than the ( 4,4 ) using the same sls model and ammonia abundances derived by fitting both the sed and the subarcsecond resolution ammonia ( 4,4 ) transition data ( table [ tbl5b ] ) . unfortunately , for ammonia transitions other than the ( 4,4 ) only single - dish data ( churchwell et al . 1990 , cesaroni et al . 1992 ) are available in the literature . these observations were obtained with the 100 m telescope , with an angular resolution of @xmath155 . therefore , a convolution of the model results with a gaussian with a fwhm=@xmath155 was performed . since this angular resolution is insufficient to resolve the emission of the core , only the spectra towards the @xmath173 position have been calculated . a value of @xmath169 km s@xmath3 has been adopted for the lsr velocity of the ambient cloud . in figure [ figbaja ] we show the model spectra for the ( 1,1 ) , ( 2,2 ) , ( 4,4 ) , and ( 5,5 ) transitions overlaid on the spectra observed by churchwell et al . ( 1990 ) and cesaroni et al . ( 1992 ) . the uncertainty in the absolute calibration of the observed spectra is estimated to be about 30% ( churchwell et al . 1990 , cesaroni et al . as can be seen in the figure , the model spectrum of the ( 4,4 ) line coincides pretty well ( within @xmath810% ) with the observed one . for the ( 2,2 ) and ( 5,5 ) lines , the brightness temperatures of the model spectra are @xmath211% lower than in the observed spectra , which is still within the calibration uncertainties . the ratio of intensities of the main to satellite lines , which is independent of the calibration , is @xmath212 for the observed ( 2,2 ) spectrum ( the uncertainty in the ratio is estimated from the rms of the spectrum ) , in agreement with the value of 3.9 obtained for the model spectrum . the ratio is @xmath213 for the observed ( 5,5 ) spectrum , which is also in agreement with the value of 3.5 obtained for the model spectrum . the observed line width ( fwhm ) of the ( 2,2 ) main line is @xmath214 km s@xmath3 , somewhat smaller than the value of @xmath215 km s@xmath3 obtained for the model spectrum . finally , the observed line width of the ( 5,5 ) main line is @xmath216 km s@xmath3 , which is very similar to the value of @xmath215 km s@xmath3 predicted by the model . given the coincidence of line ratios and line widths , and taking into account the uncertainty in the calibration , we consider that the model predictions are roughly consistent with the observations of the ( 2,2 ) , ( 4,4 ) , and ( 5,5 ) transitions . however , uncertainties in the calibration are large ( 30% ) and a definitive conclusion is difficult to attain with the present data . for the ammonia ( 1,1 ) line , the brightness temperature of the model spectrum is significantly lower than that of the observed spectrum , and the shape of the spectrum is also different . we believe that for the ammonia ( 1,1 ) transition , that traces the colder gas , there is likely a significant contribution of cold molecular gas from outside the core ( see hatchell et al . 2000 ) , which is not considered in our modeling . also , the assumption @xmath129=@xmath65 used in deriving our ammonia spectra could affect these results . we conclude that our model parameters , obtained by fitting the dust and high angular resolution nh@xmath18(4,4 ) transition data , can also reproduce , within the observational uncertainties , the intensities , line profiles , and main to satellite line ratios of the multitransitional ammonia data . our modeling is carried out by fitting essentially four parameters , namely , the mass accretion rate and the mass of the central star ( or , alternatively , the stellar luminosity ) in a physically self - consistent collapse model of a sls envelope , and the minimum and maximum gas - phase ammonia abundances inside this envelope . we conclude therefore that in massive protostars , where an important temperature gradient is expected to be present , an analysis in terms of a realistic , self - consistent model that takes into account the gradients in the physical properties inside the core is necessary to properly interpret the observational data . the modeling we carried out for g31 hmc , including the continuum sed as well as the high angular resolution properties of the molecular emission , is , to our knowledge , one of the more complete ever made for a hot molecular core . this source appears to be a good laboratory to test the very early stages of massive star formation . our main results can be summarized as follows : 1 . we have succeeded in developing a self - consistent model of the physical structure of a hot core , described as a spherically symmetric sls envelope infalling onto a massive star . the model predicts the continuum emission sed as well as the shape and intensity of ammonia lines . 2 . this model is able to reproduce the general observed properties of g31 hmc . in addition to the sed , it can reproduce the strong intensity of the main and satellite ammonia ( 4,4 ) lines , as well as their spatial variations as observed with subarcsecond angular resolution . the same model can roughly reproduce , within the observational uncertainties , the single - dish ammonia ( 2,2 ) , ( 4,4 ) , and ( 5,5 ) data . the ammonia ( 1,1 ) data likely have a significant contribution of cold gas outside the core that has not been included in the model . although the sed fitting alone can not constrain adequately the parameters of g31 hmc , the simultaneous fitting of the sed and the ammonia emission allows us to obtain a preliminary estimate of the physical parameters of this source . our best fit model for g31 hmc has a central protostar of 20 - 25 @xmath217 , with an age of 3 - 4 @xmath218 yr , a high mass accretion rate of 2 - 3 @xmath219 yr@xmath3 , and a total luminosity of @xmath220 . the magnetic field is 5 - 6 mg . the mass of the envelope is large , 1400 - 1800 @xmath217 , and its temperature is high , ranging from @xmath5 k at the outer radius of the envelope ( 30,0000 au ) to @xmath67 k at the inner radius ( 130 - 160 au ) . + a steep increase of the gas - phase ammonia abundance , from @xmath221 in the outer parts of the envelope to @xmath222 in the inner parts , is required to reproduce the observed spatial variations of the high angular resolution ammonia spectra . this is naturally explained as a result of the release ( at a characteristic temperature of @xmath223 k ) of ammonia molecules trapped in water ice grain mantles . therefore , taking into account variations in the gas - phase molecular abundances because of sublimation and freezing - out of molecules onto grain surfaces appears to be necessary to properly reproduce the highest angular resolution observations , and will probably be a key ingredient in the modeling of future line observations with alma . although the actual scenario for g31 hmc is certainly complex ( binaries , rotation , outflows , ... ) , our simple model can provide a rough description of the main features of the star formation process in this source , revealing that massive protostars appear to be able to excite high excitation ammonia transitions , not only at circumstellar scales , but also in the larger scale ( @xmath8 30,000 au ) infalling envelopes . + however , the sed of g31 hmc is still incomplete , lacking high angular resolution data at wavelengths shorter than 880 @xmath88 m . also , high angular resolution line data are only available for the ammonia ( 4,4 ) transition , and the calibration uncertainties in the lower angular resolution observations of other transitions are still too large to properly constrain the models . higher quality data as well as an improved modeling would be necessary to make firm statements on the physical properties of this source . g.a . and m.o . acknowledge support from mec aya2005 - 08523-c03 - 03 and micinn aya2008 - 06189-c03 - 01 grants ( co - funded with feder funds ) , and from junta de andaluca . s.l . and p.d . acknowledge support from dgapa - unam and conacyt ( mexico ) . we thank j.m . girart , f. wyrowski , and s. kurtz for communicating data in advance of publication . we thank the referee for a thoughtful review that help us to improve the paper . this research made use of data products from the midcourse space experiment ( msx ) . since the inversion transitions inside a given rotational level of the ammonia molecule are more frequent than the transitions to other rotational levels , we approximate any metastable inversion doublet by a two - level model and we estimate the excitation temperature , @xmath107 , of its inversion transition by considering a detailed balance between excitation and deexcitation , resulting @xmath224\right\}^{-1 } , \label{eq : dosni2}\ ] ] where @xmath114 is the rest frequency of the inversion transition , @xmath225 is the total number density of gas molecules ( mainly h@xmath16 molecules , which dominate the collisions ) , @xmath226 is the intensity of the local radiation field , @xmath65 is the kinetic temperature , @xmath227 is the collisional deexcitation rate coefficient , and @xmath128 is the einstein spontaneous emission coefficient from upper to lower level of the inversion doublet . in table [ tbl : inver ] we list the frequencies adopted for the inversion transitions of the metastable levels ( 1,1 ) to ( 6,6 ) of the ammonia molecule ( from pickett et al . 1998 ) . the collisional deexcitation rate coefficient ( from ho 1977 ) is taken as @xmath228 for all the inversion transitions , since sweitzer ( 1978 ) and danby et al . ( 1988 ) show that it does not change significantly from one transition to another . the einstein spontaneous emission coefficient from upper to lower level of the inversion doublet is given by @xmath229 where @xmath230 is the electric dipole moment of the ammonia molecule whose value is @xmath231 d ( cohen & poynter 1974 ) . the values of @xmath128 for metastable states with @xmath232 are listed in table [ tbl : inver ] . as an example , figure [ texs ] shows the excitation temperature for the ammonia ( 4,4 ) inversion transition as a function of the density for a range of kinetic temperatures from 10 to 1000 k. as can be seen from the figure , the transition is well thermalized for the density and temperature ranges of hmcs . a similar result is obtained for transitions ( 1,1 ) to ( 5,5 ) . the absorption coefficient corresponding to a transition between two levels is given by ( estalella & anglada 1999 ) : @xmath233 where @xmath234 is the number density of particles in the upper level , and @xmath235 is the profile function that contains the dependence on the line - of sight velocity . the profile function is actually the line - of - sight velocity distribution function , and is normalized so that @xmath236 . for the inversion transition of the @xmath111 rotational level of the ammonia molecule , the absorption coefficient is given by @xmath237 where @xmath114 is the rest frequency of the main line ( @xmath238 ) of the inversion transition ( the values adopted are listed in table [ tbl : inver ] for metastable levels up to @xmath239 ) , @xmath128 is obtained from equation ( [ eq : cee ] ) and listed in table [ tbl : inver ] , @xmath234 is the number density of ammonia molecules in the upper level of the @xmath111 inversion doublet , and @xmath107 is the excitation temperature that describes the relative populations of the inversion doublet and is calculated using equation ( [ eq : dosni2 ] ) . due to the quadrupole hyperfine structure of the ammonia inversion doublet , the profile function is given by @xmath240 , where @xmath121 is the lte relative intensity of the _ i_th hyperfine component given in table [ tab : propnh3 ] , and @xmath241 its profile function that is assumed to be a gaussian . since the profile function of the overall inversion transition is normalized , @xmath242 , and because @xmath243 , the profile function of each individual hyperfine component should be also normalized , so that , @xmath244 . therefore , @xmath245 where @xmath124 is the doppler velocity corresponding to the frequency shift of each hyperfine component with respect to the main line ( @xmath246 ) , @xmath125 is the line - of - sight lsr velocity of the ambient cloud , @xmath247 is the line - of - sight component of the systematic velocity field , and @xmath248 is the local dispersion of the distribution of line - of - sight velocities due to turbulent and thermal motions . the density of molecules in the upper level of the inversion doublet , that is required in equation ( [ eq : opa120 ] ) , can be written in terms of the density of molecules in the @xmath111 rotational state , @xmath249 , using the boltzmann equation and the excitation temperature , @xmath250 on the other hand , the total density of ammonia molecules is given by the sum of the densities of all the rotational states , @xmath251 solving for @xmath262 in equation ( [ eq : bolz2 ] ) and substituting in equation ( [ eq : sumatot ] ) one obtains @xmath263 solving the above equation for @xmath264 and substituting the result in equation ( [ eq : nu ] ) , the population of molecules in the upper level of the inversion doublet @xmath234 can be obtained in terms of @xmath265 , @xmath266 this equation can be finally written in terms of the number density of h@xmath16 molecules , @xmath267 , using the fractional ammonia abundance , @xmath268 . furthermore , if we assume that the relative populations of all the rotational states are characterized by the same rotational temperature , that we call @xmath129 ( this is usually the case in high density regions where collisions dominate and all temperatures thermalize to @xmath65 ) , the above equation can be written as @xmath269 where @xmath270 is the partition function defined as : @xmath271 in the calculation of @xmath270 usually only metastable rotational states ( @xmath100 ) need to be included , since non - metastable states are short lived and in general are not significantly populated . the sum is extended up to a value of @xmath272 high enough so that the contribution of higher levels is negligible . substituting equations ( [ eq : nutot3 ] ) and ( [ eq : perfinalc ] ) into ( [ eq : opa120 ] ) , the expression for the absorption coefficient for the inversion transition of the @xmath111 rotational state of the ammonia molecule becomes @xmath273 let s assume a molecular species whose total amount remains constant , while the solid versus gas phase ratio , @xmath176 , changes as a result of condensation and sublimation processes due to variations in temperature and density . let s further assume that there is a small residual fraction of molecules that remain in the gas phase even at temperatures well below the characteristic sublimation temperature , as observations suggest that the abundance of some molecular species does not drop to zero in the very low temperature regions . let s call @xmath151 the minimum gas - phase molecular abundance , reached at low temperatures ( where @xmath274 ) , and @xmath142 the maximum gas - phase abundance , reached in the hot regions ( @xmath275 ) , where all the molecules are in the gas phase . therefore , @xmath276 , the gas - phase molecular abundance at a given point will fulfill the following equation : @xmath277 where the first term corresponds to the fraction of molecules in the gas - phase and the second term corresponds to the fraction in the solid phase , after correcting for the residual fraction of molecules ( corresponding to @xmath151 ) that do not follow the condensation / sublimation process . following sandford & allamandola ( 1993 ) , the ratio , @xmath176 , of molecules in the solid to gas phase , assuming that only thermal processes are involved , can be obtained as @xmath279 where @xmath280 is the number density of dust grains , @xmath281 is the grain radius , @xmath282 is the lattice vibrational frequency of the molecule , @xmath283 is the mass of the molecule , and @xmath284 is the binding energy of the molecule on the ice surface . for the physical conditions of hmcs the exponential term , defined by the binding energy @xmath284 , dominates the behavior of this equation . assuming a given dust - to - gas mass ratio , @xmath285 , the density of dust grains can be written in terms of the density , @xmath225 , of gas molecules as @xmath286 , where @xmath287 is the mean molecular weight , @xmath106 is the mass of the h atom , and @xmath288 is the density of the material constituting the dust grains . assuming an average radius of the grains @xmath289 @xmath88 m ( see @xmath145 [ modeldust ] ) , a ratio @xmath290 , and a density @xmath291 3 g @xmath64 ( draine & lee 1984 ) , we obtain @xmath292 and equation ( [ eq : laeta ] ) can be written in terms of the gas density . in the case that the ammonia molecules are trapped into water ice , we substitute the corresponding numerical constants for water , namely @xmath296 hz and @xmath297 k ( sandford & allamandola 1993 ) into equation ( [ eq : laeta ] ) , and the equation becomes @xmath298 cccccc 12 & msx & 18 & @xmath299 & @xmath8 18 & 1 + 12 & iras & 30 & @xmath300 & @xmath8 30 & 2 + 21 & msx & 18 & @xmath301 & @xmath8 18 & 1 + 25 & iras & 30 & @xmath302 & @xmath8 30 & 2 + 60 & iras & 60 & @xmath303 & @xmath8 60 & 2 + 100 & iras & 120 & @xmath304 & @xmath8 120 & 2 + 450 & scuba & 9 & @xmath305 & @xmath8 150 & 3 + 850 & scuba & 15 & @xmath306 & @xmath8 150 & 3 + 880 & sma & 0.8 & @xmath307 & @xmath8 8 & 4 + 1300 & ovro & 3.6 & @xmath308 & @xmath8 7 & 5 + 1300 & 3 m irtf & 90 & @xmath309 & @xmath8 90 & 6 + 1350 & scuba & 22 & @xmath310 & @xmath8 22 & 3 + 1400 & pdbi & 0.7 & @xmath311 & @xmath8 10 & 7 + 1400 & bima & 0.5 & @xmath312 & @xmath8 3 & 8 + 2000 & scuba & 34 & @xmath313 & @xmath8 34 & 3 + 2700 & pdbi & 2.1 & @xmath314 & @xmath8 5 & 9 + 3300 & pdbi & 1.8 & @xmath315 & @xmath8 10 & 7 + 3400 & ovro & 4.7 & @xmath316 & @xmath8 10 & 5 + 7000 & vla & 0.05 & @xmath317 & & 10 + 13000 & vla & 2.2 & @xmath318 & @xmath8 5 & 11 + i & 25 & @xmath319 & 1.0 & 156 & @xmath320 & @xmath321 & 342 & @xmath322 & 4.9 & 0.78 & @xmath323 & @xmath324 & @xmath325 & @xmath326 & 5.2 + ii & 15 & @xmath327 & 1.0 & 115 & @xmath328 & @xmath329 & 296 & @xmath330 & 3.2 & 0.91 & @xmath331 & @xmath332 & @xmath333 & @xmath334 & 7.1 + iii & 12 & @xmath335 & 1.0 & 97 & @xmath336 & @xmath337 & 263 & @xmath338 & 2.7 & 0.95 & @xmath339 & @xmath340 & @xmath341 & @xmath342 & 7.6 + iv & 9 & @xmath343 & 1.0 & 82 & @xmath344 & 3@xmath345 & 234 & @xmath346 & 2.1 & 1.03 & @xmath347 & @xmath348 & @xmath349 & @xmath350 & 8.7 + v & 7 & @xmath351 & 1.0 & 70 & @xmath352 & @xmath353 & 212 & @xmath354 & 1.7 & 1.05 & @xmath355 & @xmath356 & @xmath357 & @xmath358 & 9.7 + vi & 5 & @xmath359 & 1.0 & 62 & 550 & @xmath356 & 194 & @xmath360 & 1.1 & 1.20 & @xmath361 & @xmath341 & @xmath362 & @xmath363 & 12.4 + vii & 4 & @xmath364 & 1.0 & 54 & 240 & @xmath365 & 180 & @xmath366 & 0.9 & 1.22 & @xmath367 & @xmath365 & @xmath368 & @xmath369 & 13.4 + viii & 3 & @xmath370 & 1.0 & 46 & 80 & @xmath371 & 163 & @xmath372 & 0.7 & 1.28 & @xmath373 & @xmath374 & @xmath375 & @xmath376 & 15.2 + ix & 2 & @xmath370 & 1.1 & 43 & 16 & @xmath377 & 160 & @xmath378 & 0.3 & 1.51 & @xmath362 & @xmath379 & @xmath380 & @xmath381 & 22.4 + x & 1.5 & @xmath382 & 1.1 & 37 & 5 & @xmath383 & 145 & @xmath384 & 0.2 & 1.60 & @xmath385 & @xmath386 & @xmath387 & @xmath388 & 26.1 + xi & 1 & @xmath389 & 1.2 & 33 & 0.7 & @xmath339 & 140 & @xmath390 & 0.1 & 1.84 & @xmath349 & @xmath391 & @xmath392 & @xmath393 & 35.8 ll @xmath69 & 30000 au + @xmath59 & @xmath394 cm + @xmath92 & 1.0 + @xmath53 & 20 - 25 @xmath217 + @xmath395 & 2 - 3 @xmath396 @xmath217 yr@xmath3 + @xmath68 & 130 - 160 au + @xmath57 & 5 - 8 @xmath397 @xmath398 + @xmath399 & 1.0 - 1.5 @xmath400 @xmath398 + @xmath401 & 40 k + @xmath402 & 1200 k + @xmath403 & 6000 - 6500 au + @xmath201 & 1400 - 1800 @xmath217 + @xmath32 & 3 - 4 @xmath397 yr + @xmath165 & 1.9 - 2.3 @xmath397 au + @xmath29 & 2 - 3 @xmath404 dyn @xmath405 + @xmath202 & 5 - 6 mg + @xmath151 & 1 - 4@xmath406 + @xmath142 & 2 - 4@xmath404 + rlrrrl & & @xmath413 & @xmath124 & + @xmath122 & @xmath414 & ( mhz ) & ( km s@xmath3 ) & @xmath121 & notes + + 1&@xmath415 & 1.531 & @xmath416 & 0.11111 & external satellite + 2&@xmath417 & 0.613 & @xmath418 & 0.13889 & internal satellite + 3&1@xmath4191 + 2@xmath4192 & 0 & 0 & 0.08333 + 0.41667 = 0.50000 & main line + 4&@xmath420 & @xmath4210.613 & 7.75 & 0.13889 & internal satellite + 5&@xmath422 & @xmath4211.531 & 19.37 & 0.11111 & external satellite + + 1&@xmath420 & 2.04 & @xmath42125.78 & 0.05000 & external satellite + 2&@xmath423 & 1.31 & @xmath42116.55 & 0.05185 & internal satellite + 3&1@xmath4191 + 2@xmath4192 + 3@xmath4193 & 0 & 0 & 0.15000 + 0.23148 + 0.41481 = 0.79629 & main line + 4&@xmath424 & @xmath4211.31 & 16.55 & 0.05185 & internal satellite + 5&@xmath417 & @xmath4212.04 & 25.78 & 0.05000 & external satellite + + 1&@xmath424 & 2.30 & @xmath42128.88 & 0.02645 & external satellite + 2&@xmath425 & 1.71 & @xmath42121.47 & 0.02678 & internal satellite + 3&2@xmath4192 + 3@xmath4193 + 4@xmath4194 & 0 & 0 & 0.21164 + 0.28009 + 0.40178 = 0.89352 & main line + 4&@xmath426 & @xmath4211.71 & 21.47 & 0.02678 & internal satellite + 5&@xmath423 & @xmath4212.30 & 28.88 & 0.02645 & external satellite + + 1&@xmath426 & 2.45 & @xmath42130.43 & 0.01620 & external satellite + 2&@xmath427 & 1.95 & @xmath42124.21 & 0.01629 & internal satellite + 3&3@xmath4193 + 4@xmath4194 + 5@xmath4195 & 0 & 0 & 0.24305 + 0.30083 + 0.39111 = 0.93500 & main line + 4&@xmath428 & @xmath4211.95 & 24.21 & 0.01629 & internal satellite + 5&@xmath425 & @xmath4212.45 & 30.43 & 0.01620 & external satellite + + 1&@xmath428 & 2.57 & @xmath42131.40 & 0.01090 & external satellite + 2&@xmath429 & 2.12 & @xmath42125.91 & 0.01094 & internal satellite + 3&4@xmath4194 + 5@xmath4195 + 6@xmath4196 & 0 & 0 & 0.26181 + 0.31148 + 0.38299 = 0.95629 & main line + 4&@xmath430 & @xmath4212.12 & 25.91 & 0.01094 & internal satellite + 5&@xmath427 & @xmath4212.57 & 31.40 & 0.01090 & external satellite + + 1&@xmath429 & 2.63 & @xmath42131.46 & 0.00783 & external satellite + 2&@xmath431 & 2.25 & @xmath42126.92 & 0.00785 & internal satellite + 3&5@xmath4195 + 6@xmath4196 + 7@xmath4197 & 0 & 0 & 0.27422 + 0.31765 + 0.37677 = 0.96863 & main line + 4&@xmath432 & @xmath4212.25 & 26.92 & 0.00785 & internal satellite + 5&@xmath430 & @xmath4212.63 & 31.46 & 0.00783 & external satellite + +	we present a model aimed to reproduce the observed spectral energy distribution ( sed ) as well as the ammonia line emission of the g31.41 + 0.31 hot core . the hot core is modeled as an infalling envelope onto a massive star that is undergoing an intense accretion phase . we assume an envelope with a density and velocity structure resulting from the dynamical collapse of a singular logatropic sphere . the stellar and envelope physical properties are determined by fitting the observed sed . from these physical conditions , the emerging ammonia line emission is calculated and compared with subarcsecond resolution vla data of the ( 4,4 ) transition taken from the literature . the only free parameter in this line fitting is the ammonia abundance . the observed intensities of the main and satellite ammonia ( 4,4 ) lines and their spatial distribution can be well reproduced provided it is taken into account the steep increase of the gas - phase ammonia abundance in the hotter ( @xmath0 k ) , inner regions of the core produced by the sublimation of icy mantles where ammonia molecules are trapped . the model predictions for the ( 2,2 ) , ( 4,4 ) , and ( 5,5 ) transitions , obtained with the same set of parameters , are also in reasonably agreement , given the observational uncertainties , with the single - dish spectra of the region available in the literature . the best fit is obtained for a model with a central star of @xmath1 , a mass accretion rate of @xmath2 yr@xmath3 , and a total luminosity of @xmath4 . the outer radius of the envelope is 30,000 au , where kinetic temperatures as high as @xmath5 k are reached . the gas - phase ammonia abundance ranges from @xmath6 in the outer region to @xmath7 in the inner region . to our knowledge , this is the first time that the dust and molecular line data of a hot molecular core , including subarcsecond resolution data that spatially resolve the structure of the core , have been simultaneously explained by a detailed , physically self - consistent model . this modeling shows that hot , massive protostars are able to excite high excitation ammonia transitions up to the outer edge ( @xmath8 30,000 au ) of the large scale infalling envelopes .
fast magnetic monopoles with dirac charge @xmath0 are attractive objects to search for with deep underwater neutrino telescopes . the intensity of monopole cherenkov radiation is @xmath1 8300 times higher than that of muons . an optical module ( om ) of the baikal experiment could detect such an object from a distance up to hundred meters . propagating through the universe , a monopole could be accelerated by magnetic fields up to energies @xmath2 ev @xcite , @xcite . hence , monopoles with mass below @xmath3 gev are expected to be relativistic . the monopole energy loss is about 10 @xmath4 for a lorentz factor @xmath5 , but rises significantly with energy ( by a factor 1000 for @xmath6 ) . monopoles with such energy losses could not cross the earth . still , for a wide mass range of @xmath7 gev one may search for relativistic monopoles from the lower hemisphere , with significant suppression of the background caused by very energetic atmospheric muons . the monopole flux is limited by the chudakov - parker bound @xmath810@xmath9 @xmath10s@xmath11sr@xmath11 which results from the requirement that galactic magnetic fields must be conserved . the strongest experimental bounds on relativistic monopoles ( @xmath12 ) have been obtained by large cherenkov detectors : baikal @xcite and amanda @xcite . in this paper we present the result of a search for relativistic monopoles with the baikal neutrino telescope nt200 , based on data taken in the years 1998 - 2003 ( 994 days ) . the present stage of the telescope nt200 @xcite takes data since april , 1998 and consists of 192 optical modules ( oms ) at 8 strings . the oms are grouped in pairs and each pair defines a channel . a trigger is formed by the requirement of @xmath13 fired channels within 500 ns . for such events amplitude and time of all fired channels within time window 2000 ns are digitized and sent to shore . the space - time pattern of light recorded from a monopole depends on the water optical characteristics . the absorption length of baikal deep water is @xmath14(480 nm ) @xmath15 m and slightly varies during years . scattering in baikal water is characterized by a strongly anisotropic function @xmath16 with a mean cosine of the scattering angle @xmath17 and a scattering length @xmath18 m. the om response to a fast monopole was calculated for a scattering length 30 m. oms facing the monopole path record about one photoelectron at 85 m distance , om turned away still see one photoelectron at 50 m. the uncertainty of the scattering length which varies between 15 m and 30 m leads to an uncertainty of nearly 20@xmath19 for the effective area for fast monopole registration . the processing chain for fast monopole starts with the selection of events with a high multiplicity of hit channels . in order to reduce the background from downward atmospheric muons we restrict ourself to monopoles coming from the lower hemisphere . for an upward going particle the times of hit channels increase with rising z - coordinates from bottom to top of the detector . to suppress downward moving particles , a cut on the value of the z time correlation has been applied : @xmath20 where @xmath21 and @xmath22 are time and z - coordinate of a fired channel , @xmath23 and @xmath24 are mean values for times and z - coordinates of the event , @xmath25 and @xmath26 the root mean square errors for time and z - coordinates . the requiments @xmath27 and @xmath28 ( cut 0 ) define the first selection stage and leave 0.015% of events that satisfy trigger @xmath29 on @xmath30 strings . this cut reduces the effective area for monopoles with @xmath31 by almost two times . the main background for monopole signatures are muon bundles and single high energy muons . the simulation chain of such muons starts with cosmic ray air showers generated with corsika @xcite . we use the qgsjet model @xcite and a primary composition according to @xcite . the mum program @xcite is used for muon propagation through water . in fig.1 we compare experimenal data ( solid curve ) and simulated atmospheric muon events ( dashed curve ) with respect to all parameters which have been used for background rejection . one sees that the simulation describes experimental data quite well , even on the level of those rare events . within 994 days live time , about @xmath32 events with @xmath33 have been recorded , with 20943 of them satisfying cut 0 ( @xmath27 and @xmath28 ) . all events have been reconstructed according to the standard algorithm @xcite . for further background supression we use additional cuts , which essentially reject muon events and at the same time only slightly reduce the area for registration of relativistic monopoles . 1 . @xmath34 and @xmath35 2 . the @xmath36 determined from reconstruction has to be smaller than 3 ( @xmath37 ) 3 . reconstructed zenith angle @xmath38 4 . reconstructed track distance from the center of nt200 @xmath39 m. nt200 took data in various configurations , due to the different numbers of temporarily operating channels . the different cuts on @xmath40 and @xmath41 correspond to different configurations of nt200 . in table 1 the rejection coefficient , i.e. the ratio of events passing cuts 1 - 4 to those passing cut 0 , is presented ( for a subset of configurations representing 50% of full statistics ) . also shown is the reduction factor for the monopole effective area @xmath42 . no events from the experimental sample pass cuts 1 - 4 . .[crflux ] rejection coefficient for experimental events and simulated downgoing atmospheric muons , as well as the reduction coefficient @xmath42 for the effective area of monopole registration , after application of cuts 1 - 4 [ cols="^,^,^,^",options="header " , ] in fig.2 , the 90% c.l . upper limit obtained with the baikal neutrino telescope for an isotropic flux of fast monopoles is compared to the final limits from the undergrund experiments ohya @xcite and macro@xcite and to the published limit of the underice detector amanda@xcite . this work was supported by the russian ministry of education and science , the german ministry of education and research and the russian fund of basic research ( grant 05 - 02 - 17476 , 04 - 02 - 17289 and 02 - 07 - 90293 ) and by the the grant of the president of russia nsh-1828.2003.2	we present a limit on the flux of relativistic monopoles obtained during 994 days of operation of the baikal neutrino telescope nt200 . the search for relativistic monopoles is based on the enormous amount of cherenkov radiation emitted by these particles .
we would like to thank profs . w. c. chang , c. w. kao , atsushi hosaka , t .- s . h. lee , yongseok oh , sho ozaki , and a. i. titov , for useful discussions and/or correspondences . this work is supported in part by the national science council of taiwan under grant nsc98 - 2112-m002 - 006 . we would also like to acknowledge the help from national taiwan university high - performance computing center in providing us with a fast and dependable computational environment which is essential in completing this work . t. h. bauer , r. d. spital , d. r. yennie and f. m. pipkin , rev . phys . * 50 * , 261 ( 1978 ) . a. donnachie and p. v. landshoff , phys . b * 185 * , 403 ( 1987 ) ; nucl . b * 244 * , 322 ( 1984 ) ; nucl . b * 267 * , 690 ( 1986 ) ; nucl . b * 311 * , 509 ( 1989 ) . a. i. titov , y. s. oh and s. n. yang , phys . rev . lett . * 79 * , 1634 ( 1997 ) ; a. i. titov , y. oh , s. n. yang , and t. morii , nucl . a * 684 * , 354 ( 2001 ) . r. a. williams , phys . c * 57 * , 223 ( 1998 ) . a. i. titov , t .- s . h. lee , h. toki , and o. streltsova , phys . c * 60 * , 035205 ( 1999 ) . y. s. oh and h. c. bhang , phys . c * 64 * , 055207 ( 2001 ) . a. i. titov , t .- s . h. lee , phys . c * 67 * , 065205 ( 2003 ) . a. i. titov and b. kmpfer , phys . rev . c * 76 * , 035202 ( 2007 ) . ( leps collaboration ) , phys . lett . * 95 * , 182201 ( 2005 ) and references therein . q. zhao , j .- didelez , m. guidal and b. saghai , nucl . a * 660 * , 323 ( 1999 ) ; q. zhao , b. saghai and j. s. al - khalili , phys . lett . b * 509 * , 231 ( 2001 ) . particle data group , k. hagiwara _ et al . d * 66 * , 010001 ( 2002 ) . h. et al . _ , nucl . b * 70 * , 257 ( 1974 ) . clas collaboration , e. anciant _ et al . lett . * 85 * , 4682 ( 2000 ) . s. ozaki , a. hosaka , h. nagahiro , and o. scholten , phys . c * 80 * , 035201 ( 2009 ) . a. hosaka , private communication . durham hep database ( http://www.slac.stanford.edu/spires/hepdata ) . chiang , s. n. yang , m. vanderhaeghen , and d. drechsel , nucl . a * 723 * , 205 ( 2003 ) . v. pascalutsa , m. vanderhaeghen , and s. n. yang , phys . rept . * 437 * , 125 ( 2007 ) . t. feuster and u. mosel , nucl . a * 612 * , 375 ( 1997 ) . g. penner and u. mosel , phys . c * 66 * , 055211 ( 2002 ) ; ibid . c * 66 * , 055212 ( 2002 ) ; + v. shklyar , h. lenske and u. mosel , phys . c * 72 * , 015210 ( 2005 ) . t. feuster and u. mosel , phys . c * 58 * , 457 ( 1998 ) ; * 59 * , 460 ( 1999 ) . y. oh , a. i. titov , and t .- s.h . lee , phys . c * 63 * , 025201 ( 2001 ) . k. schilling , k. seyboth , and g. wolf , nucl . b * 15 * , 397 ( 1970 ) . s. capstick and w. roberts , phys . d * 49 * no . 9 s. capstick , phys . d * 46 * no . 7 m. williams et.al . c * 80 * , 065208 ( 2009 ) .	we study whether the nonmonotonic behavior found in the differential cross section of the @xmath0-meson photoproduction near threshold can be described by a resonance . namely , we add a resonance to a model consisting of pomeron and @xmath1 exchange by fiat and see if , with a suitable assignment of spin and parity , mass and width , as well as the coupling constants , one would be able to obtain a good description to all the data reported by the leps collaboration in the low - energy region . the resonant contribution is evaluated by using an effective lagrangian approach . we find that , with the assumption of a @xmath2 resonance with mass of @xmath3 gev and width of @xmath4 gev , leps data can indeed be well described . the ratio of the helicity amplitudes @xmath5 calculated from the resulting coupling constants differs in sign from that of the known @xmath6 . we further find that the addition of this postulated resonance can substantially improve the agreement between the existing theoretical predictions and the recent @xmath7 photoproduction data if a large value of the ozi evading parameter @xmath8 is assumed for the resonance . photoproduction , @xmath0 meson , nucleon resonance , pomeron 13.60.le , 25.20.lj , 14.20.gk a well - established feature in the @xmath0-meson photoproduction reaction at high energies is that it is dominated by the diffractive processes , which are conveniently described by the @xmath9-channel pomeron @xmath10 exchange @xcite . in the low - energy region , the nondiffractive processes of the pseudoscalar @xmath11-meson exchange are known to contribute @xcite . in addition , many other processes , including nucleon and nucleon - resonance exchanges , second pomeron exchange , @xmath9-channel scalar meson and glueball exchanges , and @xmath12-cluster knockout have also been extensively studied @xcite . however , no definite conclusion has been reached because of the limited experimental data . recently , a local maximum in the differential cross sections of @xmath0 photoproduction on protons at forward angles at around @xmath13 gev , has been observed by the leps collaboration @xcite . models which consist of @xmath9-channel exchanges @xcite have not been able to account for such a nonmonotonic behavior . typically , local maxima in the cross sections are often associated with resonances . effects of the resonances in @xmath14- and @xmath15-channels up to mass 2 gev have been investigated in refs . @xcite . ref . @xcite used a constituent quark model with @xmath16 symmetry and included explicitly excited resonances with quantum numbers @xmath17 , while ref . @xcite considered all the known 12 resonances below 2 gev listed in particle data group @xcite , with coupling constants determined by the available data @xcite at large momentum transfers . the resonances are found to play non - negligible role , especially in polarization observables . however , no local maximum as observed in ref . @xcite was obtained . in this letter , we study whether the nonmonotonic behavior found in ref . @xcite can be described by a resonance . namely , we will add a resonance to a model consisting of pomeron and @xmath1 exchange by fiat and see if , with a suitable assignment of spin and parity , mass and width , as well as the coupling constants , one would be able to obtain a good description of all the data reported by the leps collaboration , which include the angular and energy dependence of the differential cross section and decay angular distributions in the gottfried - jackson frame , in the low - energy region from threshold to @xmath18 gev . since the local maximum appears quite close to the threshold , we will investigate , as a first step , the possibility of the spin of the resonance being either @xmath19 or @xmath20 . similar analysis was carried out in a coupled - channel model @xcite . however , the analysis was marred by a confusion in the phase of the pomeron - exchange amplitude @xcite . we first define the kinematical variables @xmath21 , @xmath22 , @xmath23 , and @xmath24 as the four - momenta of the incoming photon , initial proton , outgoing @xmath0-meson , and final proton , respectively ; and @xmath25 , @xmath26 , and @xmath27 . the full amplitude in our model consists of pomeron - exchange , @xmath9-channel @xmath11-exchange , and the @xmath14- and @xmath15-channel @xmath28-exchange amplitudes . the pomeron - exchange amplitude can be expressed as @xcite , @xmath29 where @xmath30 and @xmath31 are the polarization vectors of the @xmath0-meson and photon with helicities @xmath32 and @xmath33 , respectively ; and @xmath34 the dirac spinor of the nucleon with momentum @xmath35 and helicity @xmath36 . the explicit form for the transition operator @xmath37 can be found in refs . @xcite and the scalar function @xmath38 is given by the reggeon parametrization , @xmath39,\end{aligned}\ ] ] where we introduce an additional threshold factor @xmath40 as also done in refs . @xcite to adjust the shape of the energy dependence of the pomeron amplitude near the threshold . also , @xmath41 and @xmath42 are the isoscalar electromagnetic form factor of the nucleon and the form factor of @xmath43 coupling , respectively , and are taken to be of the form given in refs . @xcite . as in @xcite , we take @xmath44 , @xmath45 @xmath46 , and @xmath47 which is obtained by fitting to the total cross sections data at high energy . we choose @xmath48 @xmath46 by matching the forward differential cross sections data at around @xmath49 @xmath50 @xcite . the contribution of the @xmath9-channel @xmath11 exchange to the @xmath0 photoproduction is rather well understood . we use the same set of parameters for the pseudoscalar - exchange amplitude as adopted in ref . @xcite except , in their notation , @xmath51 @xcite and @xmath52 gev , the cutoff in the form factor . we will consider the cases where the spin of the resonance is either @xmath19 or @xmath20 . the interaction lagrangian densities which describe the coupling of spin-@xmath19 and @xmath20 particles to @xmath53 and @xmath54 , can in general be written as @xcite , @xmath55 where @xmath56 represents the @xmath0-meson field tensor and @xmath57 with @xmath58 . the operator @xmath59 are given by @xmath60 and @xmath61 , depending on the parity of the resonance @xmath28 . for the @xmath62 vertices , one simply changes @xmath63 and @xmath64 . however , current conservation consideration fixes @xmath65 for @xmath66 resonances to be zero . in addition , the term proportional to @xmath67 in the lagrangian densities of eq . ( [ phinnstar ] ) vanishes in the case of real photon . the form factor for the vertices used in the @xmath14- and @xmath15-channel diagrams , @xmath68 , is taken as @xmath69 $ ] @xcite , with @xmath70 the cutoff parameter for the virtual @xmath28 . we choose @xmath71 gev for all resonances . the effect of the width is taken into account in a breit - wigner form by replacing the usual denominator @xmath72 , with @xmath73 is the total decay width of @xmath28 . since @xmath74 , we take @xmath75 mev for the @xmath15-channel propagator . with the interaction lagrangian densities given in eqs . ( [ phinnstaronehalf],[phinnstar ] ) , it is straightforward to write down the invariant amplitudes of the @xmath14- and @xmath15-channel exchange diagrams of the corresponding @xmath28 . in tree - level approximation , only the products like @xmath76 , enter in the invariant amplitudes . they are determined with the use of minuit , by fitting to the experimental data @xcite , including differential cross section at forward angle as a function of photon energy and differential cross section as a function of @xmath9 at different photon energies , as well as to five decay angular distributions at two photon energies . we find that with the assignments of @xmath66 to the resonance , it is not possible to produce the nonmonotonic behavior near threshold , in contrast to the finding of refs . @xcite . at forward direction as a function of photon energy @xmath77 . the dotted , dashed , and solid lines denote contributions from nonresonant , resonance with @xmath78 , and their sum , respectively . data are from refs . @xcite . ] for the assignments of @xmath79 , we find that both parities can describe the differential cross section at forward angle well and can also describe other observables with comparable quality . the resulting @xmath80 , and ( mass , width ) in unit of gev , for the case of @xmath81 and @xmath82 are 1.066 and @xmath83 , and 0.983 and @xmath84 , respectively . this leads us to the problem of determining the parity of the resonance . to resolve this question , we perform a stability check against changes in pomeron contribution , whose low - energy behavior is not yet fully understood . it turns out that the extracted properties of the resonances are more sensitive with respect to the variation in the pomeron parameters if the positive parity is chosen . therefore , we prefer the choice of @xmath78 . the coupling constants and the extracted mass and width of the @xmath78 resonance are given in table [ tab : nstar- ] . resonance : ( a ) decay angular distributions @xmath85 ( b ) @xmath86 , and ( c ) @xmath87 , @xmath88 , and @xmath89 . all the decay angular distributions are given in two photon lab energies , @xmath90 gev ( upper panel ) and @xmath91 gev ( lower panel ) . data is taken from ref . @xcite . the notation is the same as in fig . [ d13_1 ] . ] resonance : ( a ) decay angular distributions @xmath85 ( b ) @xmath86 , and ( c ) @xmath87 , @xmath88 , and @xmath89 . all the decay angular distributions are given in two photon lab energies , @xmath90 gev ( upper panel ) and @xmath91 gev ( lower panel ) . data is taken from ref . @xcite . the notation is the same as in fig . [ d13_1 ] . ] our best fits with the choice of @xmath78 to the experimental energy dependence of the differential cross section at forward angle and angular dependence of the differential cross section @xcite are shown in figs . [ d13_1 ] and [ d13_2 ] , respectively . the dotted , dashed , and solid lines correspond to contributions from nonresonant , _ i.e. _ , pomeron plus @xmath11-exchange , resonant , and the full results , respectively . we find that no matter how the pomeron parameters are varied , it is not possible to describe the nonmonotonic behavior of the differential cross section at forward direction as a function of photon energy with only the nonresonant contribution . one also sees from fig . [ d13_2 ] that the addition of a resonance markedly improves the agreement with the data on angular dependence . .the results for @xmath28 parameters with @xmath78 . [ cols="^,^ " , ] [ tab : nstar- ] our results for the decay angular distributions of the @xmath0-meson in its rest frame ( or the gottfried - jackson system , hereafter , called gj - frame ) , which can be expressed in terms of the spin - density matrix elements @xmath92 @xcite , are shown in fig . [ d13_3 ] , where the contributions from nonresonant , resonant , and the full results are again denoted by dotted , dashed , and solid lines , respectively . we see that the data in @xmath93 at both energies of @xmath94 gev and @xmath95 gev , @xmath96 at @xmath95 gev , and @xmath87 again at @xmath95 gev can already be described relatively well by the nonresonant contribution only and do not need strong modification from a resonance . however , the rest of the distributions show some discrepancies between nonresonant contribution and experimental data and the inclusion of resonant contribution does help to reduce the discrepancies . this is especially true for @xmath96 at @xmath95 gev and @xmath87 at @xmath97 gev , where the nonresonant contribution does not describe satisfactorily the experimental data . for both @xmath98 and @xmath89 , our model still fail to give adequate agreement with the data which are of rather poor quality with large error bars . one might be tempted to identify the @xmath82 as the @xmath6 as listed in pdg @xcite . the coupling constants given table [ tab : nstar- ] can be used to calculate the ratio of the helicity amplitudes @xmath99 and @xmath100 , though not their magnitudes since only the product of the coupling constants for @xmath101 and @xmath102 are determined . we obtain a value of @xmath103 , while it is @xmath104 for @xmath6 . even though their magnitudes are similar , the relative sign is , however , different . for @xmath105 , we find that the value of @xmath106 , again with positive sign . since the resonance proposed here is obtained by fitting to the existing data , a critical check would be to see whether additional data would substantiate our interpretation . accordingly , we also calculate the predictions of our model with and without the inclusion of the proposed resonance for all the polarization observables @xcite . in general , we find that the effects of the resonance are substantial in many of the polarization observables @xcite . we show in figs . [ polarization ] our predictions for some of them like the single polarization observables @xmath107 in the upper panel , and , in the lower panel , the double polarization observables @xmath108 and @xmath109 at @xmath110 2 gev . it is seen that the effects of the proposed resonance are huge in these polarization observables . in the same figure , results that would be obtained if the @xmath81 resonance determined in our best fitting is adopted , are also shown by dash - dotted curve . we see that measurements of these polarization observables would help to resolve the question of the parity of the resonance . , @xmath111 , @xmath112 , and @xmath113 taken at photon laboratory energy @xmath114 gev . the solid and dash - dotted lines correspond to our results with the choices of @xmath78 and @xmath105 , respectively , while the dotted lines denote the nonresonant contribution . ] from the @xmath115 mixing , one would expect that a resonance in @xmath54 channel would also appear in @xmath116 channel . the only question is their relative decay strength . the conventional `` minimal '' parametrization relating @xmath102 and @xmath117 is @xmath118 with @xmath119 corresponds to the deviation from the ideal @xmath115 mixing angle . here , @xmath120 is called the ozi - evading parameter and the larger value of @xmath120 would indicate larger strangeness content of the resonance . photoproduction as a function of @xmath121 at @xmath122 gev . solid and dashed lines represent the model predictions of ref . @xcite without and with the addition of our resonance with @xmath123 . data are from ref . @xcite . ] in order to study the effects of the resonance postulated here in the @xmath7 photoproduction , we adopt the study of ref . @xcite which employs the nucleon resonances predicted by refs . @xcite . in fig . [ omega ] , it is seen that the prediction of their model for the @xmath9-dependence of differential cross section at @xmath122 gev , given in solid lines , still exhibits substantial discrepancy with the most recent experimental data @xcite for @xmath124 gev@xmath125 . by adding resonance postulated here to the model of ref . @xcite with @xmath126 , whose prediction is given in the dashed line in fig . [ omega ] , we see that the differential cross section at @xmath122 gev can be reproduced with roughly the correct strength . the large value of @xmath127 would imply that the resonance we propose here contains a considerable amount of strangeness content . in summary , we have explored the possibility of accounting for the nonmonotonic behavior as observed by the leps collaboration at energies close to threshold as a manifestation of a resonance . we carry out calculations using a model with a nonresonant contribution which consists of pomeron plus @xmath9-channel @xmath128-exchange amplitudes , and a resonant contribution . with resonance mass and width , and coupling constants as parameters , we perform a best fit to all the leps data at low energies with possible assignments of @xmath129 and @xmath130 . we confirm that nonresonant contribution alone can not describe the nonmonotonic behavior of the forward differential cross section near threshold and the @xmath9-dependence of the differential cross section @xcite . we find that the addition of a resonance with @xmath131 of either positive or negative parity can not explain the local maximum at around @xmath13 gev . however , with an assignment of @xmath132 , a nice agreement with most of the leps data can be achieved . we prefer the choice of @xmath133 as the best fit to the data since its results are more stable with respect to changes in the low - energy pomeron parameters . the obtained resonance mass and width are @xmath3 and @xmath4 gev , respectively . the resulting coupling constants give rise to a ratio of the helicity amplitudes @xmath134 , which differs from that of the known @xmath6 in sign . furthermore , we find that the postulated resonance gives substantial contribution to the polarization observables , which can also be used to determine the parity of the resonance if it indeed exists . the possible effects of this postulated resonance in the @xmath7 photoproduction are investigated by incorporating it within a recent calculation @xcite for this reaction . it turns out that the addition of our resonance , with a choice of a large value of ozi - evading parameter @xmath8 , could indeed considerably improve the agreement of the model prediction with the most recent data . that would imply the resonance postulated here does contain considerable amount of strangeness content . there are a few caveats in our study . the first concerns the low - energy pomeron parameters which are not presently very precisely determined . if the postulated resonance contains considerable amount of strangeness , then it could couple strongly to , say , @xmath135 channel . question would then arise on how the coupled - channel effects would modify the low - energy behavior of the nonresonant amplitude employed in this investigation . this can be answered only with a full coupled - channel calculations as carried out in ref . @xcite . another question is the validity of our assumption to account for the local maximum with just one resonance . as seen in the calculation of the effects of our postulated resonance , some discrepancies with the recent data still persist after the addition of this resonance . accordingly , our study may have raised more questions than it answers . clearly , further studies , both experimentally and theoretically , are needed on the @xmath0-meson photoproduction at low energies .
rare b meson decays are one of the most promising research area in particle physics and lie on the focus of theoretical and experimental physicists . in the standard model ( sm ) , they are induced by flavor changing neutral currents ( fcnc ) at loop level and therefore sensitive to the gauge structure of the theory . from the experimental point of view , they play an outstanding role in the precise determination of the fundamental parameters of the sm , such as cabbibo - kobayashi - maskawa ( ckm ) matrix elements , leptonic decay constants , etc . furthermore , these decays provide a sensitive test to the new physics beyond the sm , such as two higgs doublet model ( 2hdm ) , minimal supersymmetric extension of the sm ( mssm ) @xcite , etc . among the rare b decays , @xmath0 has received considerable interest since the branching ratios ( br ) of the inclusive @xmath5 @xcite and exclusive @xmath6 @xcite have been already measured experimentally . recently , the new experimental results for the inclusive @xmath0 decay are announced by cleo and aleph collaborations @xcite . therefore , the @xmath0 decay is under an extensive investigation in the framework of various extensions of the sm , in order to get information about the model parameters or improve the existing restrictions . it is well known that the fcnc are forbidden at the tree level in the sm . this restriction is achieved in the extended model with the additional conditions . 2hdm is one of the simplest extensions of the sm , obtained by the addition of a new scalar @xmath7 doublet . the yukawa lagrangian causes that the model possesses phenemologicaly dangerous fcnc s at the tree level . to protect the model from such terms , the ad hoc discrete symmetry @xcite on the 2hdm scalar potential and the yukawa interaction is proposed and there appear two different versions of the 2hdm depending on whether up and down quarks couple to the same or different scalar doublets . in model i , the up and down quarks get mass via vacuum expectation value ( v.e.v . ) of only one higgs field . in model ii , which coincides with the mssm in the higgs sector , the up and down quarks get mass via v.e.v . of the higgs fields @xmath8 and @xmath9 respectively where @xmath10 corresponds to first ( second ) higgs doublet of 2hdm @xcite . in the absence of the mentioned discrete symmetry , fcnc appears at the tree level and this model is called as model iii in current literature @xcite . a comprehensive phenemological analysis of the model iii was done in series of works @xcite . in particular , from a purely phenomenological point of view , low energy experiments involving @xmath11 , @xmath12 , @xmath13 , etc , place strong constraints on the existence of tree level flavor changing ( fc ) transitions , existing in the model iii . in the present work , we examine the @xmath0 decay in the model iii , taking the next to leading ( nlo ) qcd corrections into account , in a more detailed analysis compared to one existing in literature ( see @xcite ) . further , we obtain the constraints for the neutral couplings @xmath14 , @xmath15 and @xmath16 with the assumption that @xmath17 , @xmath18 , @xmath19 and the other couplings which include the first generation indices are negligible compared to former ones ( for the definition of @xmath20 see section 2 ) . our predictions are based on the cleo measurement @xmath21 and the restrictions coming from the @xmath2 ( @xmath3 ) mixing and the @xmath4 parameter @xcite . note that nlo qcd corrections to the @xmath0 decay in 2hdm ( for model i and ii ) were calculated in @xcite . the paper is organized as follows : in section 2 , we present the nlo qcd corrected hamiltonian responsible for the @xmath0 decay in the model iii and discuss the effects of the additional @xmath22 flipped operators to the decay rate . section 3 is devoted to the constraint analysis , more precisely to the the ratios @xmath23 , @xmath24 and our conclusions . before presenting the nlo qcd corrections to the @xmath25 decay amplitude in the 2hdm ( model iii ) , we would like to remind briefly the main features of the 2hdm . the yukawa interaction for the general case is @xmath26 where @xmath27 and @xmath28 denote chiral projections @xmath29 , @xmath30 for @xmath31 , are the two scalar doublets , @xmath32 and @xmath33 are the matrices of the yukawa couplings . for convenience we choose @xmath34 and @xmath35 in the following basis : @xmath36\ , ; \phi_{2}=\frac{1}{\sqrt{2}}\left(\begin{array}{c c } \sqrt{2 } h^{+}\\ h_1+i h_2 \end{array}\right ) \,\ , , \label{choice}\end{aligned}\ ] ] where the vacuum expectation values are , @xmath37 this choice permits us to write the fc part of the interaction as @xmath38 with the following advantages : * doublet @xmath39 corresponds to the scalar doublet of the sm and @xmath40 to the sm higgs field . this part of the yukawa lagrangian is responsible for the generation of the fermion masses with the couplings @xmath41 . * all new scalar fields belong to the @xmath42 scalar doublet . the couplings @xmath43 are the open window for the tree level fcnc s and can be expressed for the fc charged interactions as @xmath44 where @xmath45 as @xmath46 . ] is defined by the expression @xmath47 here the charged couplings appear as a linear combinations of neutral couplings multiplied by @xmath48 matrix elements . this gives an important distinction between model iii and model ii ( i ) . after this preliminary remark , let us discuss the nlo qcd corrections to the @xmath0 decay in the 2hdm for the general case . the appropriate framework is that of an effective theory obtained by integrating out the heavy degrees of freedom , which are , in this context , @xmath49 quark , @xmath50 , and @xmath9 bosons , where @xmath51 denote charged , @xmath8 and @xmath9 denote neutral higgs bosons . the llog qcd corrections are done through matching the full theory with the effective low energy theory at the high scale @xmath52 and evaluating the wilson coefficients from @xmath53 down to the lower scale @xmath54 . note that we choose the higher scale as @xmath52 since the evaluation from the scale @xmath55 to @xmath52 gives negligible contribution to the wilson coefficients . here we assume that the charged higgs boson is heavy due to theoretical analysis of the @xmath0 decay ( see @xcite ) . the effective hamiltonian relevant for @xmath56 decay is @xmath57 where the @xmath58 are operators given in eq . ( [ o7 ] ) and the @xmath59 are wilson coefficients renormalized at the scale @xmath60 . the coefficients are calculated perturbatively and expressed as functions of the heavy particle masses in the theory . the operator basis depends on the model used and the conventional choice in the case of sm , 2hdm model ii ( i ) and mssm is @xmath61 where @xmath62 and @xmath63 are @xmath64 colour indices and @xmath65 and @xmath66 are the field strength tensors of the electromagnetic and strong interactions , respectively . in our case , however , new operators with different chirality structures can arise since the general yukawa lagrangian includes both @xmath27 and @xmath28 chiral interactions . the conventional operator set is extended first adding two new operators which are left - right analogues of @xmath67 and @xmath68 , namely @xmath69 further we need the second operator set @xmath70 which are flipped chirality partners of @xmath71 : @xmath72 this extended basis is the same as the basis for @xmath73 extensions of sm @xcite . note that in the sm , model ii ( i ) 2hdm and the mssm , the absence of @xmath74 and @xmath75 are a consequence of assumption @xmath76 . in the calculations , we take only the charged higgs contributions into account and neglect the effects of neutral higgs bosons for the reasons given below : the neutral bosons @xmath77 , @xmath78 and @xmath79 are defined in terms of the mass eigenstates @xmath80 , @xmath81 and @xmath82 as @xmath83 where @xmath62 is the mixing angle and @xmath84 is proportional to the vacuum expectation value of the doublet @xmath34 ( eq . ( [ choice2 ] ) ) . here we assume that the massess of neutral higgs bosons @xmath81 and @xmath82 are heavy compared to the b - quark mass . the neutral higgs scalar @xmath81 and pseduscalar @xmath82 give contribution only to @xmath85 for @xmath0 decay . with the choice of @xmath86 , @xmath87 and @xmath88 can be expressed at @xmath89 level as @xmath90 where @xmath91 and @xmath92 are the masses and charges of the down quarks ( @xmath93 ) respectively . here we used the redefinition @xmath94 eq . ( [ c7a0h0 ] ) shows that neutral higgs bosons can give a large contribution to @xmath85 , which does not respect the cleo and aleph data @xcite . at this stage we make an assumption that the couplings @xmath95(@xmath96 and @xmath97 are negligible to be able to reach the conditions @xmath98 and @xmath99 . these choices permit us to neglect the neutral higgs effects . now , for the evaluation of wilson coefficients , we need their initial values with standard matching computations . denoting the wilson coefficients for the additional charged higgs contribution with @xmath100 , we have the initial values of the wilson coefficients for the first set of operators ( eqs.([o7 ] ) , ( [ newop1 ] ) ) @xmath101 the explicit forms of the wilson coefficients in the sm ( @xmath102 ) is presented in the literature @xcite . for the primed wison coefficients we get , @xmath103 @xmath104 where @xmath105 and @xmath106 . the functions @xmath107 , @xmath108 , @xmath109 and @xmath110 are given as @xmath111 in calculations we neglect the small contributions of the internal @xmath112 and @xmath113 quarks compared to one due to the internal @xmath49 quark . for the initial values of the mentioned wilson coefficients in the model iii ( eqs . ( [ coeffh ] ) , ( [ coeffsm2 ] ) and ( [ coeffh2 ] ) ) , we have @xmath114 using these initial values , we can calculate the coefficients @xmath115 and @xmath116 at any lower scale with five quark effective theory where large logarithims can be summed using the renormalization group . since the strong interactions preserve chirality , the operators in eqs . ( [ o7 ] , [ newop1 ] ) can not mix with their chirality flipped counterparts eq . ( [ newop2 ] ) and the anomalous dimension matrices of two separate set of operators are the same and do not overlap . with the above choosen initial values of wilson coefficients , their evaluations are similar to the sm case . for completeness , note that , the operators @xmath117,@xmath118,@xmath119 and @xmath120 ( @xmath121,@xmath122 , @xmath123 and @xmath124 ) give contributions to the matrix element of @xmath0 and in the ndr scheme which we use here , the effective magnetic moment type wilson coefficients are redefined as @xmath125 where @xmath126 is the color factor and @xmath127 ( @xmath128 ) is the charge of up ( down ) quarks . there is still another mixing in the operator set @xmath129 ( @xmath130 ) @xcite and we do not take into account since the initial values of the wilson coefficients @xmath131 and @xmath132 are zero in our case . the nlo corrected coefficients @xmath133 and @xmath134 are given as @xmath135 here @xmath136 , @xmath137 and @xmath138 are the numbers which appear during the evaluation @xcite . the functions @xmath139 @xcite and @xmath140)are the leading order qcd corrected wilson coefficients : @xmath141 and @xmath142 is the @xmath143 correction to the leading order result that its explicit form can be found in @xcite . @xmath144 can be obtained by replacing the wilson coefficients in @xmath142 with their primed counterparts . the nlo corrected coefficients @xmath145 , @xmath146 and @xmath147 , @xmath148 are numerically small at @xmath149 scale therefore we neglect them in our calculations . finally , the nlo qcd corrected @xmath0 decay rate in model iii is obtained as @xmath150 where @xmath151 is the fine structure constant , and @xmath149 is b - quark mass . @xmath152 is given in @xcite @xmath153 the functions @xmath154 and a are @xcite @xmath155 the explicite expressions for @xmath156 , @xmath157 , @xmath158 and @xmath159 can be found in @xcite . at this point we would like to note that the expressions for unprimed wilson coefficients in our case can be obtained from the results in @xcite by the following replacements : @xmath160 to obtain @xmath161 , it is enough to use the primed wilson coefficients at @xmath89 level + ( eq . [ coeffh2 ] ) since the evaluation of @xmath162 from @xmath163 to @xmath164 is the same as that of @xmath165 . note that , for model ii ( model i ) @xmath166 and @xmath167 are @xmath168 now let us turn our attention to the constraint analysis . restrictions to the free parameters , namely , the masses of the charged and neutral higgs bosons and the ratio of the v.e.v . of the two higgs fields , denoted by @xmath169 in the framework of model i and ii , have been predicted in series of works @xcite . recently the constraint which connects masses of the charged higgs bosons , @xmath170 and @xmath171 , is obtained by using the qcd corrected values in the llo approximation and it is shown that the constraint region is sensitive to the renormalization scale , @xmath60 @xcite . usually , the stronger restrictions to the new couplings are obtained from the analysis of the @xmath2 ( here @xmath172 ) decays , the @xmath4 parameter and the @xmath21 decay . in @xcite , all these processes have been analysed and two possible scenarios are obtained depending on the choice whether the constraint from @xmath173 is enforced or not . although the new experimental results are near the sm result , @xmath174 , the world average value for @xmath175 is still almost one standard deviations higher than the sm one . this brings the possibility that an enhancement to the sm value is still necessary to get the correct experimental one . such an enhancement is reached under the conditions @xmath176 and @xmath177 @xcite , where @xmath178 is a model iii parameter ( see section 2 ) and @xmath84 is the vacuum expectation value of the higgs field responsible for the generation of fermion masses . first , the constraints for the fc couplings from @xmath2 processes for the model iii were investigated withouth qcd corrections , under the following assumptions @xcite @xmath179 where @xmath180 is up ( down ) quark and @xmath181 are the generation numbers . using the constraint coming from @xmath184 , the measurement @xmath185 , @xmath2 mixing and the result coming from the analysis of the @xmath4 parameter , the following restrictions are obtained @xcite : @xmath186 since the experimental results for @xmath173 are still unclear , we disregard the constraint coming from @xmath173 and we analyse the restrictions for the couplings @xmath187 , @xmath188 and @xmath189 in the nlo aproximation , respecting the constraints due to the @xmath2 mixing , the @xmath4 parameter and using the measurement by the cleo @xcite collaboration : @xmath190 here , we explain why we use only the cleo data in our analysis but not aleph one ( @xmath191 ) . the aleph data has a larger error compared to cleo data and it leads to a wide restriction region for @xmath192 , which includes the one coming from the cleo data . therefore , the cleo data allows us to get more stringest constraints on the model parameters . the idea in this calculation is to take @xmath193 and @xmath194 , where the indices @xmath181 denote @xmath195 and @xmath196 quarks . this choice permit us to neglect the neutral higgs contributions because the yukawa vertices are the combinations of @xmath197 and @xmath198 . to reduce the b - quark mass dependence let us consider the ratio @xmath199 where @xmath200 is the phase space factor in semileptonic b - decay , @xmath201 is the qcd correction to the semileptonic decay width @xcite , @xmath202 and @xmath203 . in fig . ( [ tb45 ] ) , we plot the parameter @xmath205 with respect to @xmath206 at @xmath207 and @xmath208 . we see , that there are two different restriction regions , where the upper one corresponds to the positive @xmath209 value , however the lower one to the negative @xmath209 value . increasing @xmath206 causes @xmath210 to decrease in both regions . with the given value of @xmath211 , the condition @xmath212 is obtained . in the lower region it is possible that the ratio becomes negative , i.e. @xmath213 . further , increasing @xmath170 causes to increase @xmath214 and the area of the restriction region . ( [ tb45b ] ) is devoted the same dependence as in fig . ( [ tb45 ] ) and shows that the third region , which is almost a straight line , appears . in this region the ratio @xmath215 and increases with increasing @xmath170 similar to the previous regions . finally , we consider @xmath205 dependence of @xmath216 , which is a neutral fc coupling . in fig . ( [ cttt564 ] ) we plot the @xmath205 dependence of @xmath216 for fixed @xmath217 , at @xmath218 , and charged higgs mass @xmath208 . here the selected region for @xmath205 is @xmath219 . increasing @xmath205 forces the ratio @xmath220 to be negative . it is realized that the ratio @xmath221 becomes smaller when @xmath170 is larger . in conclusion , we find the constraints for the yukawa couplings @xmath205 , @xmath206 and @xmath216 using the cleo measurement @xmath224 and respecting the restrictions due to the @xmath2 mixing and the @xmath4 parameter ( see @xcite for details ) . the constraints for the other parameters of the model iii from the existing experimental results require more detailed new analysis . 1 j. l. hewett , in proc . of the @xmath225 annual slac summer institute , ed . l. de porcel and c. dunwoode , slac - pub6521 . m. s. alam et . al , cleo collaboration , phys . lett . b74 ( 1995 ) 2885 r. ammar et al . , cleo collaboration , phys . lett * 71 * ( 1993 ) 674 . m. s. alam , cleo collaboration , to appear in ichep98 conference ( 1998 ) , + r. barate et.al . , aleph collaboration , phys . lett . b429 ( 1998 ) 169 s. l. glashow and s. weinberg , phys . d15 ( 1977 ) 1958 . h. haber , c. kane , and t. sterlig , nucl . lett.b * 161 * ( 1979 ) 493 ; d. atwood , l. reina and a. soni , phys . d53 ( 1996 ) 1199 . d. atwood , l. reina and a. soni , phys . * 75 * ( 1995 ) 3800 . t. p. cheng and m. sher , phys . d * 35 * ( 1987 ) 3484 ; phys . d * 44 * ( 1991 ) 1461 ; + a. antaramian , l. j. halland a. rasin , phys . * 69 * ( 1992 ) 1871 ; + l. j. hall and s. weinberg , phys . d * 48 * ( 1993 ) 979 ; + m. savage , phys . b * 266 * ( 1991 ) 135 ; + m. luke and m. g. savage , phys . lett b307 ( 1993 ) 387 . d. atwood , l. reina and a. soni , phys . d * 55 * ( 1997 ) 3156 . f. m. borzumati and c. greub , phys . d * 58 * ( 1998 ) 0784004 . m. ciuchini , g. degrassi , p. gambino and g. i. giudice , nucl . b * 527 * ( 1998 ) 21 . t. m. aliev , g. hiller and e. o. iltan , nucl . b * 515 * ( 1998 ) 321 . p. cho and misiak , phys . d49 ( 1994 ) 5894 . a. j. buras et al . b * 424 * ( 1994 ) 374 . a. j. buras , m. misiak , m. mnz and s. pokorski , nucl . b424 ( 1994 ) 374 . a. j. buras , m. mnz , phys . d52 ( 1995 ) 186 ; + m. misiak , nucl . phys . b393 ( 1993 ) 23 , erratum b439 ( 1995 ) 461 . abreu et al . , z. phys . c * 64 * ( 1994 ) 183 ; + g. alexander et al . b * 370 * ( 1996 ) 17 ; + m. veltman , acta phys . polon . b * 8 * ( 1977 ) 475 ; + b. w. lee , c. quigg , h. b. thacker , phys . rev . d * 16 * ( 1977 ) 253 ; + m. veltman , phys . b * 70 * ( 1977 ) 253 ; + a. k. grant , phys . rev . d * 51 * ( 1995 ) 207 ; + m. acciari et al . d * 396 * ( 1997 ) 327 ; + k. kiers and a. soni , phys . rev d * 56 * ( 1997 ) 5786 . n. cabibbo and l. maiani , phys . b * 79 * ( 1978 ) 109	we study the @xmath0 decay including the next to leading qcd corrections in the two higgs doublet model with flavor changing neutral currents at the tree level . we find the constraints to the flavor changing parameters of the model , using the experimental results on the branching ratio of @xmath1 decay , provided by the cleo collaboration , the restrictions coming from the @xmath2 ( @xmath3 ) mixing and the @xmath4 parameter .
multifragmentation ( mf ) is the dominant decay mode in heavy ion reactions when the excitation energy is comparable to the nuclear binding energy . the recent use of 4@xmath22 detectors capable of observing a substantial fraction of the particles and fragments emitted in a given interaction has been invaluable in furthering the elucidation of this complex phenomenon @xcite . several review articles of mf have been published in recent years @xcite . one of the most complete mf experiments to date has been that performed by the eos collaboration for the @xmath23au on c system @xcite . the use of seamless detectors such as the eos time projection chamber coupled with the multiple sampling ionization chamber ( music ii ) permitted the observation of practically all the charged particles and fragments emitted in each event , ranging from protons to heavy fragments . full reconstruction was therefore possible for a large fraction of the events . the high energy asymmetric collision of a 1a gev projectile on a light target is uniquely favorable for the kinematic separation of the initial nuclear collision and the subsequent mf transition . this data set permitted us to establish that mf occurs following the expansion of a remnant formed with charge @xmath1 , mass @xmath0 and excitation energy @xmath2 after the prompt ejection of high momentum light particles @xcite . we have applied model independent methods used in the study of critical phenomena to extract the value of several critical exponents @xcite . the critical scaling function , which describes the behavior of the nuclear remnant near a critical point , was determined @xcite . in performing this analysis we were aided by percolation calculations @xcite , which served as a quantitative guide to the application of the methods developed for the study of critical exponents in small systems . the eos experiment was an outgrowth of earlier inclusive studies of mf in the interaction of xe and kr with high - energy protons @xcite . in this work high precision counter techniques were used to obtain accurate information about the kinetic energy spectra of _ isotopically _ resolved nuclear fragments . the reduced coulomb barrier seen in the fragment spectra indicated that nuclear fragments are emitted from an expanded nuclear system . systematics of the fragment kinetic energy spectra also showed that the remnant was lighter than the target nucleus implying that _ the nuclear remnant existed for a time _ after the initial collision . the relative fragment yields of 63 isotopes could be understood using a thermal droplet model @xcite with a free energy parameterization based on the semi - empirical mass formula . for energy depositions of @xmath248 mev per nucleon this multi - isotope thermometer gave a freezeout temperature of @xmath245 mev . the systematic variation of the `` kinetic temperature '' ( exponential slope parameter extracted from fragment kinetic energy spectra ) as a function of the fragment mass also indicated that fermi momentum in the remnant must play a significant role in the fragment formation process . more speculative was the suggestion that the power law yield of the fragment masses y=@xmath25 with @xmath262.6 observed in 50 - 400 gev p on xe collisions showed that nuclear fragmentation might give information about a possible `` liquid - gas '' thermal phase transition in charged nuclear matter @xcite . important support for this view came from the exclusive emulsion data of waddington and freier @xcite , which were analyzed by campi to show that the conditional moments of the individual fragment events exhibited features characteristic of a critical transition @xcite . the occurrence of thermal equilibrium in the mf process was strongly supported by the experiments of the aladin collaboration , which showed that fragment yields were independent of the entrance channel when the data was scaled for projectile or target mass @xcite . the above cited evidence for the formation of a substantially equilibrated remnant which expands prior to mf suggests that it would be appropriate to compare the eos results with a thermal model in which mf occurs from an expanded state . two widely used thermal multifragmentation models are the copenhagen ( smm ) @xcite and berlin ( mmmc ) @xcite statistical treatments . these models differ in their parameterization of the expansion and in technical details but are similar in their thermodynamic approach . for very small systems both these microcanonical models predict the onset of a very inhomogeneous state at a definite excitation energy . this inhomogeneous excited state consists of a number of normal density fragments accompanied by a statistically insignificant number of nucleons . subsequently these fragments cool by light particle emission . we present here a comparison of the eos data with smm . several data and model comparisons have been made previously @xcite . these comparisons seriously suffered from the fact that the experiments did not determine the @xmath1 , @xmath0 , and @xmath27 of the remnant . instead , these @xmath1 , @xmath0 , @xmath2 values , which constitute the proper input to test models of mf , were obtained either by use of a dynamical first stage model @xcite , by using a comparison of some of the data with smm to constrain them @xcite , or by more complex backtracing from fragment data @xcite . a great advantage afforded by asymmetric reverse kinematics collisions is that they permit an accurate separation of the initial reaction phase from the subsequent decay of the excited remnant . the use of smm also permits us to examine several aspects of the eos critical exponent analysis that are potentially problematic . the eos results were obtained for fragments in their final , cold state . however , the fragments are presumably formed in an expanded hot state . in smm the remnant is equilibrated , the fragments are formed in the hot system , then separate under the influence of the coulomb force and undergo deexcitation . as has been noted elsewhere @xcite , the distribution of the hot , primary fragments may be affected by secondary decay . the difference between the two distributions could , in principle , affect the values of the critical exponents . since smm gives separate results for hot and cold fragments , the effects of secondary decay can be probed . the eos results indicate that the first prompt step leads to a distribution of remnants . the analysis groups these remnants according to the total charged particle multiplicity , @xmath28 , which serves as the control parameter . events characterized by a given multiplicity will generally include a range of remnant @xmath29 values . it has been noted that such event mixing can affect the values of the critical exponents @xcite . furthermore , the use of multiplicity instead of temperature or excitation energy as the control parameter has been questioned @xcite . comparison of smm results obtained for the experimental distribution of remnants with those obtained for a single remnant permits us to probe the importance of these effects . this paper is organized in the following manner . section ii gives a brief summary of smm and shows how the data determine the variable volume of smm . only a single adjustable parameter remains . section iii summarizes the properties of the experimental remnant distribution , which serves as the input data to smm . the occurrence of radial expansion energy as part of the excitation energy is discussed in this section . the comparison between experimental and model results is presented in section iv . the various factors that can affect the fragment yields , the extraction of critical exponents , and the critical multiplicity are considered in section v. in sec . vi , the physics of mf and the nature of the thermal phase transition in smm as a function of the remnant mass and charge is explored . the identification of the order of a thermal phase transition for the mf of au on c requires the use of the microcanonical equation of state . a summary of the results and our conclusions are given in section vii . smm is a statistical description of the simultaneous breakup of an expanded excited nucleus into nucleons and hot fragments @xcite . individual fragments at normal nuclear density are described with a charged liquid drop parameterization . the free energy of a fragment @xmath30 is given by @xmath31 and is used to determine the fragment formation probability . this solution explicitly assumes the inhomogeneous nature of the hot mf final state . light fragments @xmath32 may also be present in the hot mf final state . for the @xmath33 fragments , a quantum mechanical description is used for the temperature dependent volume , surface , and translational free energy of the fragments . the temperature independent parameters are based on the coefficients of the semiempirical mass formula . the critical temperature , at which the surface tension of neutral nuclear matter droplets would go to zero , is in the range suggested by infinite neutral nuclear matter calculations @xcite . in smm the translational free energy depends on the _ free _ volume . the free volume , @xmath34 , can be expressed in terms of the volume of the multifragmenting system at normal nuclear density , @xmath35 , @xmath36 where the free volume parameter @xmath37 depends on the smm fragment multiplicity according to the relation : @xmath38 ^ 3 -1\ ] ] where @xmath39=1.17 @xmath40 fm and @xmath41 is the _ charged plus neutral _ hot fragment multiplicity . the crack width parameter , @xmath42 , scales the magnitude of the multiplicity dependent free volume . the breakup volume @xmath43 , which includes the volume of the fragments , is @xmath44 , where @xmath45 is the coulomb reduction parameter@xcite . we have previously shown that energy deposition in the au on c reaction is proportional to nucleons knocked out of the au nucleus by quasi - elastic nucleon - nucleon scattering @xcite . if we assume that the excited remnant initially is produced in the au volume , then the experimentally determined _ initial free volume _ is given by : @xmath46 and the remnant volume is given by : @xmath47 proton - proton correlation experiments for the multifragmentation of 1a gev au on au show that the freeze out volume is @xmath242 v@xmath48 , nearly independent of excitation energy and remnant mass and charge @xcite . = 8.5 cm fig.1a shows the smm free volume @xmath34 , and the initial free volume @xmath49 as a function of the multiplicity . as expected , the smm free volume @xmath34 is about twice the initial free volume , consistent with the expansion from @xmath50 to @xmath51 . note that the slopes of the @xmath34 and @xmath49 versus multiplicity curves track over the mf region . 1b compares the breakup volume @xmath52 with the freeze out volume from @xmath53-@xmath53 correlations as a function of the multiplicity . the breakup volume @xmath43 tracks the freezeout volume from the @xmath53-@xmath53 correlation experiments . the freeze out volume calculated in our earlier publication @xcite is different as it was obtained using initial volume of au nucleus and not the remnant volume . thus , experiment confirms the structure ( m dependence ) and scale ( crack width parameter @xmath42 ) of the volume parameterization of smm for the 1a gev au on c experiment . in addition to these experimental arguments , theoretical arguments based on buu modeling of the nuclear collision @xcite also suggest that the expansion of the remnant is energy dependent . the remaining parameter is @xmath54 , the inverse level density parameter . the @xmath54 values were determined by comparison of smm with the various experimental fragment yield distributions and it was found that @xmath55 mev . thus , the so - called standard values of all these parameters turn out to give the best agreement with the data . ( see table i. ) in smm the primary fragments are propagated in their mutual coulomb field and then undergo secondary decay . in the model successive particle emission from hot fragments with a@xmath5616 is assumed the deexcitation mechanism . the deexcitation of these fragments is treated by means of the standard weisskopf evaporation model . light fragments ( a @xmath57 16 ) deexcite via fermi breakup . the lightest particles ( a@xmath574 ) can be formed only in their ground states and undergo no secondary decay . we have used a version of the model that incorporates only thermal degrees of freedom . consequently , radial expansion or angular momentum are not included in this comparison between data and smm . the reverse kinematic eos experiment permitted the identification of charged projectile fragments @xmath58 in 1a gev @xmath59au on c interactions with high efficiency . the momenta of these fragments were measured and used to decompose the reaction into a prompt stage , in which high momentum z=1 , z=2 fragments and neutrons are emitted , and a second stage involving the decay of the remnant left after this stage @xcite . the analysis presented here is based on about 32,000 fully reconstructed mf events for which the total charge of the reconstructed au system was found to be 79@xmath604 . average fragment mass values for a given z were determined and used to reconstruct the mf final state mass @xmath61 of the charged fragments . the number of free neutrons in the mf final state is used in the determination of @xmath2 by energy balance . the remnant resulting from the prompt stage can be characterized by z , a , and @xmath27 . we follow previous practice and report @xmath27 on a per nucleon basis . the determination of these quantities has been described in detail elsewhere @xcite . fig.2 shows the mass distribution of the remnants . = 8.5 cm the most probable mass is a@xmath24190 . however , the distribution is broad and extends down to @xmath62 . the @xmath27 distribution is shown in fig.3a . the distribution peaks at @xmath242 mev / nucleon but extends beyond 16 mev / nucleon . it should be noted that the data were obtained with a minimum bias trigger that eliminated some events with very low @xmath27 . these events do not lead to mf . 4 shows the variation of @xmath27 and a with the total charged particle multiplicity @xmath28 . these quantities vary in the opposite way with @xmath28 , with @xmath27 increasing with @xmath28 , and a showing a concomitant decrease . at a given @xmath28 there is a distribution in the values of a and @xmath27 . for a given @xmath28 , the width of the distribution in a increases from @xmath63 to @xmath64 while that of the @xmath27 distribution is @xmath65 over most of the multiplicity range @xcite . = 8.5 cm we have previously shown that some of the excitation energy of the remnant actually consists of nonthermal expansion energy , @xmath66 @xcite . the model input energy @xmath67 is obtained by subtracting @xmath68 from the excitation energy @xmath2 . standard smm does not include radial expansion and so only @xmath69 must be used in the input data . fig.4 shows the dependence of @xmath66 and @xmath69 on the multiplicity @xmath28 . the expansion energy is very small for @xmath70 but then increases sharply , becoming comparable to the input energy for the largest observed multiplicity , @xmath71 . the spectrum of @xmath69 values is plotted in fig . 3b . the distribution peaks at slightly less than 2 mev / nucleon and extends to @xmath2410 mev / nucleon . = 8.5 cm the expansion energy was obtained as the difference between the sum of the measured charged fragment mean kinetic energies and the translational thermal and coulomb energies of the fragments @xcite . this procedure involves the use of the albergo double isotope ratio thermometer @xcite to determine the _ isotopic temperature _ @xmath72 . this temperature determines the translational thermal energy per particle , which is 3/2 @xmath72 . as shown in the next section , smm then independently predicts the observed double isotope ratio . this self consistency supports the validity of the @xmath69 determination and permits a combined definitive test of the isotope ratio thermometer @xmath73 and smm . additional independent evidence for the presence of expansion energy in the data can be seen in a comparison of the mean transverse kinetic energies of fragments with the smm predictions@xcite . 5 shows the results for li - n fragments for five multiplicity bins : 1 - 9 , 10 - 19 , 20 - 29 , 30 - 39 , and 40 - 59 . generally good agreement is obtained for low multiplicities , as expected , because of the small contribution of the radial expansion energy for small @xmath28 ( see fig.4 ) . the smm transverse kinetic energy values decrease with increasing @xmath28 . = 8.5 cm in general , the smm transverse energies are smaller than the experimental values . the trend in the smm values is the result of two factors , both of which lead to a decrease in the translational kinetic energy with increasing multiplicity : ( 1 ) the increase in the volume occupied by the fragments @xcite and ( 2 ) the decrease in the average charge of the fragments . in particular , see fig.11 , which shows the variation of the calculated and experimental charges of the largest fragment with @xmath28 . smm is in excellent agreement with experiment indicating that the reason for the discrepancy shown in fig.5 does not lie in the determination of the mutual coulomb energy . rather , the increasing contribution of non - thermal expansion to the experimental energies dominates the change in coulomb energy . a similar discrepancy has been observed when another data set was used to compare smm with the berlin statistical model @xcite . here too , the discrepancy was attributed to radial expansion energy @xcite . a single smm calculation was performed for each of the 32,000 eos events . the input data consisted of the z , a , and @xmath69 values of the remnants . the output of each smm calculation gives the z and a values of each fragment in its asymptotic cold state . the distribution of these mf products is used to make the comparisons presented in this section . = 8.5 cm owing to the important role of multiplicity in the eos experiment , we compare the smm and experimental multiplicities at the outset . this comparison is made as a function of @xmath69 . see fig . 6 . since smm does not include a prompt first stage , the smm charged particle multiplicity is compared to the experimental second stage multiplicity , @xmath74 . this quantity is obtained from @xmath28 by subtracting event - by - event the prompt first stage multiplicity , @xmath75 @xcite . the calculated and experimental @xmath76 distributions are in close agreement . because of this agreement , smm results will be plotted as a function of @xmath28 in many subsequent comparisons with data . here @xmath28 will be the sum of smm @xmath74 and experimental @xmath75 values . = 8.5 cm the fragment yield distributions offer the most direct comparison between the data and smm . fig.7 shows the number of z=3 - 16 fragments per event averaged over all multiplicities . good agreement is obtained for nearly all the fragments and the overall trend of decreasing yield with increasing charge is well reproduced . both data and smm show that the yield of fluorine is suppressed relative to that of neighboring fragments . smm indicates that this low yield reflects the influence of final state interactions on the primary fragments . the individual fragment yields are plotted as a function of @xmath28 in figs . 8 and 9 . smm generally does an excellent job of reproducing the data except for the lightest fragments , where the model predicts too many fragments for large @xmath28 . both data and smm show the characteristic rise and fall of the individual fragment yields as a function of @xmath28 or @xmath69 . both of these distributions were used in the statistical analyses which extract the power law and critical scaling behavior . the average value of the total number of intermediate mass fragments @xmath77 as a function of @xmath28 is shown in fig . 10 . = 8.5 cm the yield of imf s increases to @xmath244.4/event at @xmath78 and decreases for larger @xmath28 . this behavior is reproduced by smm , although the number of imf s at the peak is somewhat overestimated . this difference is a consequence of the above discrepancy between data and smm for li and be fragments at large @xmath28 . 11 shows the average value of the charge of the largest fragment in the distribution , @xmath79 . smm gives the best agreement with the data using @xmath55 mev . = 8.5 cm we have previously @xcite obtained from the data the isotopic temperature @xmath73 @xcite on the basis of @xmath80h/@xmath81h and @xmath81he/@xmath82he double isotopic yield ratios . a value of @xmath72 was also obtained from the @xmath83li/@xmath84li and @xmath81he/@xmath82he yield ratios . although the two values of @xmath72 are nearly equal , the @xmath85 values were found to be more robust @xcite . these temperatures are compared with the smm values in fig excellent agreement in the mf region is observed further confirming the self - consistent nature of data and the predictions of smm . it should be noted that the determination of freeze - out temperatures obtained by the double isotope yield ratios is subject to correction due to formation of these isotopes in secondary decay . this subject has been investigated by a number of workers @xcite . for the @xmath85 thermometer the correction has been reported @xmath24 10% below @xmath67 @xmath24 7 mev / nucleon @xcite . = 8.5 cm = 8.5 cm the asymptotic caloric curve , which is a plot of fragment isotopic temperatures versus input energy , was first obtained by pochodzalla et al . for 600 a mev au + au collisions @xcite . it was found that the temperature had the essentially constant value of @xmath245 mev for excitation energies between 3 and 10 mev per nucleon , a result that was interpreted as evidence for a first - order phase transition . a similar analysis of the eos data , in which the excitation energy included the expansion energy , showed that the temperature increased continuously but slowly with @xmath27 over the above range , i.e. from @xmath244 to @xmath246 mev @xcite . having established the presence of non - thermal expansion , we can redetermine our caloric curve and compare it with smm . = 8.5 cm these smm and eos caloric curves are compared in fig . 13 , where both smm isotopic temperatures and experimental isotopic temperatures are plotted versus the experimental input energy @xmath67 per nucleon . the smm and eos curves are in close agreement and show a very sharp increase in the temperature t over the experimental energy range . included for comparison is the caloric curve previously obtained with the inclusion of the expansion energy , which shows a much slower variation of t with @xmath27 . the determination of a caloric curve in the manner proposed in ref . @xcite has already been shown to be problematic @xcite . the recent reanalysis by the aladin group is in close agreement with our data@xcite . it must be noted that experimental caloric curves are actually inadequate measures of the thermodynamic caloric curve , which involves the breakup temperature rather than the isotope ratio temperature . this has been discussed both by bondorf et al . @xcite and in a recent publication from our group @xcite . = 8.5 cm the eos collaboration has analyzed the 1a gev @xmath23au data in terms of the theory of critical phenomena , according to which mf is viewed as a continuous phase transition . in this section we subject the smm events to this same analysis and compare the results with those obtained from the data . = 8.5 cm in order to extract the various critical exponents from the data the location of the critical point , which can be characterized by the critical multiplicity , @xmath88 , must be determined . we have used the method presented in reference @xcite , in which @xmath88 , the critical exponents @xmath89 and @xmath90 , and the scaling function were obtained from the eos data . we briefly summarize the procedure below . the fragment mass yield distribution , @xmath91 , where @xmath92 , is normalized to the mass of the remnant , @xmath93 . the normalized fragment distribution can be written as @xcite @xmath94 where @xmath95 is the scaling function and the scaling variable @xmath96 . _ if we assume that scaling is valid for clusters of all sizes _ , then @xmath97 is a function of only @xmath87 : @xmath98 where @xmath99 is the riemann zeta function and @xmath100 @xcite . at the critical point @xmath101 and a pure power law is obtained for the fragment mass yield distribution . the power law behavior is modified by the scaling function away from the critical point . = 8.5 cm in order to determine @xmath88 and @xmath87 one performs power law fits to @xmath102 versus @xmath103 for each value of @xmath28 over a particular range of fragments , e.g. @xmath104 . the best fit , as determined by the minimum value of @xmath105 , gives the value of @xmath106 . the results of this analysis for the eos data @xcite are shown in fig . the fitting procedure yields a deep minimum in @xmath105 at @xmath107 , and thus accurately determines the location of the critical point . the power law fit to the data at this multiplicity is also shown . the value of @xmath87 is [email protected] . the above analysis was also performed for the smm events for @xmath108 and the results are shown in fig . the dependence of @xmath105 and @xmath87 on @xmath28 and the quality of the power law fit at @xmath88 are similar to that exhibited by the data . the results are summarized in table [ tab2 ] and are in excellent agreement with the data . once @xmath88 is known , it is possible to determine the value of @xmath90 . we have used the percolation procedure where the largest piece @xmath109 is removed only on the liquid side @xcite . since the scaling function must have a single maximum @xcite , we define @xmath110 as the value of the scaling variable where the maximum for this fragment mass occurs . the previously given relation between @xmath111 and @xmath90 can then be written for this maximum as @xmath112 where @xmath113 , with @xmath114 being the multiplicity for which the maximum yield of fragments of mass @xmath103 is obtained . ( 7 ) leads to a power law , @xmath115 , from which @xmath90 can be found . applying this analysis to smm , we first determine the values of @xmath114 . typical results have been shown in figs . 8 and 9 . 16 shows the power law plot , @xmath116 versus @xmath117 , from which we obtain @xmath118 . applying the same analysis to eos data , we obtain @xmath119 , much smaller than the smm value@xcite . the analysis in sec . v suggests that the @xmath90 values in data could be affected by the hot fragment cooling process , and that the @xmath90 value adjusted for cooling would be @xmath120 ( see table ii ) this value is used in constructing the scaling function for the eos data . = 8.5 cm knowing the values of @xmath121 , and @xmath90 for data and smm , it is possible to evaluate the scaling function by rewriting eq . ( 5 ) as @xmath122 the unscaled data is shown in fig . 17 . scaling @xmath102 according to eq . ( 8) collapses the multifragmentation data from a broad range of fragments yields onto a narrower band for both data and smm , as shown in figs . 18 and 19 . both scaling functions show a comparable scatter of the fragments . by definition , both have a value of unity at the critical point , and both have a maximum of comparable magnitude for virtually the same value of the scaling variable . = 8.5 cm = 8.5 cm the exponent @xmath123 has been determined by means of the @xmath123-matching technique as applied to the second moment , @xmath124 , of the fragment yield distribution @xcite . the second moment is defined as @xmath125 and a similar expression may be used if the fragments are characterized by their nuclear charge instead of mass . again , following the percolation theory procedure , we omit the largest fragment from the summation in eq . ( 9 ) only on the `` liquid '' side of the critical point @xmath126 . all fragments are included on the `` gas '' side @xmath127 . the exponent @xmath123 is obtained from the power law @xmath128 by searching for various multiplicity regions that yield values of @xmath129 and @xmath130 that agree with each other . the value of @xmath88 is again determined in the same fit . = 8.5 cm the results for smm are shown in fig.20 . the procedure has also been described in ref . @xcite and the results obtained from an analysis of the eos data are summarized in table [ tab2 ] . the same analysis can be applied to the smm events and the results are also summarized in table [ tab2 ] . the experimental and calculated @xmath123 values are consistent within the limits of error . we have also evaluated several other quantities that have been associated with critical behavior @xcite . they include the fluctuations in the size of the largest fragment and in @xmath131 , the magnitude of the peak in the combination of moments @xmath132 , and the determination of @xmath87 from a plot of ln@xmath133 versus ln@xmath131 . these quantities , too , are in good agreement with experiment . = 8.5 cm the fluctuations of the size of the largest fragment and of the moments provide an independent method to locate the critical multiplicity @xmath86 of the mf transition . the first moment @xmath134 can also indicate how the nucleons are distributed into light particles , imfs and the largest piece in mf . this distribution can identify the phases present in mf and if the coexistence of liquid and gas phases occurs in mf . ( see sec . the smm calculations described in the preceding sections give results for deexcited secondary fragments formed from a distribution of remnants . these remnants are grouped according to multiplicity and we have noted that at a given @xmath28 there is a distribution of remnant charges , masses , and excitation energies . an ideal statistical analysis of multifragmentation would involve primary fragments formed from the breakup of a unique remnant @xmath135 as a function of remnant excitation energy or temperature . in this section we use smm to examine the extent to which the departures from this ideal situation affect the values of the critical exponents and related fragment properties . we introduce the terms smm@xmath136 and smm@xmath137 to designate results obtained from smm for primary fragments and deexcited fragments , respectively . the most direct view of the effect of deexcitation is provided by a comparison of the smm@xmath136 and smm@xmath137 fragment yield distributions . for simplicity , we evaluate these distributions for the decay of a single remnant as a function of @xmath69 . we have chosen the remnant formed in the au on c interaction corresponding to the critical multiplicity in the data , @xmath138 , and evaluated its breakup for a range of @xmath69 values . at a given @xmath139 the total rms width of the remnant distribution is @xmath140 5 - 7 % . in smm the average z / a ratio of the hot `` pre - fragments '' is the same as the remnant z / a ratio , i.e. , no particles or energy leaves the remnant system during the @xmath24 100 fm / c time frame for mf . = 8.5 cm the increase in the charged particle multiplicity @xmath76 from smm@xmath136 to smm@xmath137 is shown in fig . this figure shows that cooling is important . typical results for fragments with @xmath141 are shown in figs . 22 and 23 . the fragment multiplicities for smm@xmath136 and smm@xmath137 are in close agreement up to @xmath142 mev / nucleon . however , at higher excitation energies the primary fragment yields are significantly larger than those of the corresponding cold fragments . at these high @xmath69 , fragments undergo substantial secondary decay , which according to smm , occurs by evaporation for fragments with @xmath143 . for @xmath144 , secondary decay occurs by fermi breakup . these mechanisms tend to form the lightest particles and thereby reduce the yields of fragments with @xmath1453 . we emphasize the fact that the net effect of these secondary processes on the fragment distributions is rather small below 7 mev / nucleon , the energy range where most of the mf occurs . the critical point corresponds to a much lower excitation energy , @xmath146 mev / nucleon @xcite , and therefore quantities determined at these low energies should be nearly unaffected . = 8.5 cm fig . 24 shows the results for z=1 and z=2 particles . contrary to the results shown in figs . 22 and 23 , most of these particles are formed in the cooling step . averaging over the excitation energy range important for mf , 3 - 7 mev / nucleon , we find that the lightest hot fragments have a higher internal energy per nucleon than the heavier hot fragments , @xmath244 mev / nucleon versus @xmath243 mev / nucleon . consequently the main source of these particles is the fermi breakup of the lighter fragments . = 8.5 cm the analysis described in section ivc has been applied to the products of smm@xmath136 , using the experimental remnants as input . the resulting values of the various exponents are tabulated in table [ tab3 ] . the calculated value of @xmath87 for hot fragments is @xmath147 smaller than that for cold fragments while that of @xmath123 is @xmath148 smaller . in contrast , the determination of the exponent @xmath90 differs significantly for the two distributions , with the smm@xmath136 value being @xmath2440% larger than that from smm@xmath137 . the large difference in @xmath90 values is understandable given that @xmath90 is determined in part by the multiplicities for which the various fragments attain their highest yields . as shown in figs . 8 and 9 , the yields of the lighter fragments peak at large values of @xmath28 , corresponding to @xmath69 values of 7 mev / nucleon and higher . it is precisely at these energies that secondary decay reduces the yield of these fragments and thereby leads to a smaller @xmath90 . = 8.5 cm in a recent publication mastinu et al . @xcite used the copenhagen model to examine the effect of secondary fragment decay in the mf of @xmath23au . in particular , they compared the shape of the second moment , @xmath131 , of the fragment charge distribution for smm@xmath136 and smm@xmath137 . they found that the shape of the @xmath131 distribution in the vicinity of its peak is unaffected by secondary decay when plotted against @xmath27 whereas there is a change in shape when @xmath131 is plotted against @xmath28 . on this basis they concluded that secondary decay affects the values of exponents such as @xmath123 , which is obtained from @xmath131 , when charged particle multiplicity is used as the control parameter . unfortunately , they did not calculate the magnitude of this effect . we have already evaluated the effect on @xmath123 of secondary decay and , as shown in table [ tab3 ] , find that it is a relatively small 10@xmath149 correction . we note in passing that the @xmath131 plot shown by mastinu et al . @xcite is not appropriate for the extraction of @xmath123 because the largest fragment has been removed from _ both _ the liquid and gas sides instead of only on the liquid side , as was done in fig . furthermore , they did not turn off the fission channel in their smm calculation and the contribution of this mechanism disproportionally affects the analysis . the analysis described in section ivc is applied to the breakup of a single remnant , using the _ input energy _ as the control parameter . we use the critical remnant a = 160 , z = 64 as the system size for smm . 25 shows the smm@xmath136 results for the determination of the critical point by means of the one - parameter power law fit to the fragment mass yield distribution . again , there is a deep minimum in @xmath105 which determines the location of the critical point . this should be compared with the same results for smm@xmath137 using the experimental remnant distribution and @xmath28 as the control parameter ( fig . = 8.5 cm the value of @xmath87 obtained from the fit in fig . 25 is given in table [ tab3 ] . also included are the values of the other exponents . a comparison with the corresponding exponents obtained for smm@xmath137 for the experimental remnant distribution indicates that the difference in @xmath87 and @xmath123 values is less than 10@xmath149 . thus the combined effect of secondary decay , the remnant distribution for a given @xmath28 , and the use of multiplicity as the control parameter have only a minor effect on the values of the @xmath87 and @xmath123 exponents . however , the exponent @xmath90 for smm@xmath136 is much larger than the corresponding value of smm@xmath137 just as was the case for the experimental remnant distribution . again , this difference reflects the effect of cooling . we note here that the use of a single remnant and excitation energy as the control parameter , which is most readily seen in a comparison of the two smm@xmath136 columns , has essentially no effect on the critical exponents . = 8.5 cm fig . 26 shows the unscaled smm@xmath136 fragments for a = 5 to 30 for the single remnant system . 27 shows the scaling function obtained from these data and the resulting scaling collapse of the data into a very narrow band . this scaling function should be compared with the scaling function from smm@xmath137 ( fig . the more extensive scatter in fig . 19 can be attributed in comparable measure to the effects of fragment cooling and to the presence of the remnant distribution , using multiplicity as the control parameter . this indicates that the scatter seen in the experimental data at the ends of the scaling curves , fig . 18 , may also be due to these effects . we have shown that smm can reproduce the various features of the eos data , including the values of most of the critical exponents , critical multiplicity , and critical scaling . if smm did not predict the occurrence of a continuous phase transition then the exponents would merely constitute a particular parameterization of the data and would have no ulterior significance . it is therefore important to determine if smm predicts such a thermal phase transition . we follow the approach of gross @xcite and hller @xcite , who show that the nature of the transition can be determined by means of the microcanonical equation of state , using a plot of the reciprocal partition temperature @xmath151 versus @xmath69 . we assume @xmath152 is the best estimate of the average value of the fluctuating event - by - event smm microcanonical temperature . a first order transition is identified by backbending in the @xmath153 vs @xmath67 curve and hence has a region of negative system specific heat , @xmath154 . a continuous phase transition will not exhibit backbending and will have a positive system specific heat that peaks at the critical energy . = 8.5 cm recall that for infinite neutral nuclear matter a critical phase transition is expected at a temperature @xmath241518 mev @xcite . at lower temperatures , infinite neutral nuclear matter would exhibit a first order liquid - gas phase transition with possible coexistence of constant density phases . smm , however , describes the transition of a highly excited _ finite charged _ nuclear drop into a number of hot intermediate mass fragments followed by a de - excitation process . the cold state liquid and gas consists of smaller droplets and nucleons emitted in the cooling process . in the hot stage , the so - called pre - fragments are produced in thermodynamic equilibrium where the volume , surface , coulomb , symmetry and translational terms determine the fragment yields @xcite . to investigate the interplay of these energy terms we study smm@xmath136 as a function of the remnant mass and charge @xcite . fig.28 shows @xmath153 versus @xmath69 curves for a = 70 , z = 30 ; a = 100 , z = 40 ; and a = 130 , z = 53 . a characteristic backbending is observed for these systems . following the microcanonical prescription , we make the `` maxwell construction '' @xcite , which determines the average transition temperature and the magnitude of the transition energy or effective latent heat . for @xmath0=100 we obtain a transition energy of 3.2 mev / nucleon as given by the width of the backbend region . figure 29 shows that the system specific heat for a = 100 is negative indicating the possibility of a first order phase transition . we have computed the effect of the variable volume on the system specific heat and found that it is a very small fraction of the total . the translational energy contribution is also very small . thus @xmath155 describes the change in the internal energy of the mf system . the transition energy plays the role of an effective latent heat for mf . smm@xmath136 indicates that the effective latent heat is reduced for heavier remnants . for the a = 70 , 100 , and 130 systems , there is a progressive decrease in the transition energy as shown in fig.30 . a linear extrapolation suggests that the transition energy as an effective latent heat could vanish for a @xmath56 170 . we can gain further insight about these transitions by studying the distributions of light particles and fragments prior to the de - excitation or cooling process . for a = 100 the first moment of the hot mass distribution as a function of the input energy is shown in fig . 31 . for @xmath156 mev / nucleon only @xmath24 5 out of 100 nucleons are not part of the a@xmath564 droplet distribution . thus hot smm describes a transition where a large droplet is multifragmented into smaller droplets . the subsequent deexcitation processes are responsible for most of the light particles with a @xmath575 seen in the asymptotic mass distributions . = 8.5 cm in this smm model backbending can be a relatively large effect . the extra surface energy required for the mf transition cools the system and the temperature falls . for a larger system the fractional change in the surface to volume ratio due to mf is reduced and the backbending is smaller . in smm the total coulomb energy is reduced by the factor @xmath157 due to expansion prior to clusterization . since the total energy is measured with respect to the ground state of the remnant system @xmath0 , @xmath1 at normal density , this introduces a @xmath1 dependence into the @xmath153 versus @xmath67 equation of state . 30b shows the average transition temperature , which decreases as the remnant mass and charge increases . this decrease clearly reflects the increase in coulomb energy for the heavier remnants @xcite . in contrast , for a liquid - gas phase transition in _ finite neutral matter _ we would expect the transition temperature to increase with the system size , because the binding energy per nucleon increases . for a larger charged nuclear remnant , the transition energy will vanish if enough coulomb energy is converted to heat energy in the expansion of the remnant prior to the mf transition . = 8.5 cm for the remnant mass @xmath0 = 160 the reciprocal partition temperature @xmath153 is plotted versus @xmath69 in fig . 32a and the fluctuating event - by - event reciprocal microcanonical temperatures @xmath153 , are shown in fig . the area under the @xmath153 vs @xmath69 curve is the system entropy . in fig . 33a , we expand the @xmath153 vs @xmath67 plot . the system specific heat per particle is computed by taking differences between adjacent points . @xmath155 is shown in fig . the specific heat is positive and peaks near the critical energy , indicating a continuous phase transition . hller characterizes such behavior as a continuous phase transition with an anomalous specific heat!@xcite = 8.5 cm a method to determine the order of the phase transition based on canonical kinetic energy fluctuations has recently been proposed@xcite . to test the conclusions based on the smm@xmath136 microcanonical equation of state , we have done this type of analysis on the au on c data . for a single remnant system size @xmath158 this analysis compares the canonical specific heat @xmath159 with the thermally scaled kinetic energy fluctuations @xmath160 . following the prescription used on the 35 mev / nucleon au on au data we find that the energy fluctuations from the au on c data are dramatically reduced when a @xmath161 remnant mass cut is made , which assures that we have a single system size , the necessary constraint for the fluctuation analysis . thus for the au on c data , we find that @xmath162 in the whole mf region . the value of @xmath159 still remains positive even for a larger remnant mass cut e. g. @xmath163 @xmath24 10% . this result argues for a continuous phase transition . a report of this work is in preparation . very recently,(since the submission of the present article ) theoretical analyses of the nature of mf phase transition and its effect on the caloric curve have been published @xcite , indicating the occurrence of a first order phase transition . both of these calculations neglect the coulomb interaction , which has been shown to play an important role in mf @xcite . it is obvious that there will be a first order phase transition in absence of coulomb energy . = 8.5 cm it has been suggested on the basis of a caloric curve using the smm partition temperature versus an experimentally determined input energy , that mf of gold is a first order liquid - gas phase transition @xcite . this curve , computed for comparison with the aladin data showed that the temperature first increases with excitation energy , then remains nearly constant at 5 - 6 mev as @xmath69 increases between 3 and 10 mev / nucleon , and then again increases . the first regime has been interpreted as the liquid phase , i.e. the compound nucleus , the second as the mf coexistence phase , and the third as the gas phase , consisting of a mixture of nucleons and a few of the lightest fragments @xcite . this viewpoint can be probed by examining the first moment of the _ hot fragment _ yield distribution for a heavy remnant . 34a shows a plot of the dependence on @xmath69 of the first moment of the hot fragment yield distribution for the experimental remnant distribution . = 8.5 cm the figure shows that even for @xmath164 mev / nucleon , well above the mf transition , only about 8% of the remnant mass ends up in particles with @xmath165 and that the other 92% are in intermediate mass fragments with @xmath1664 . this argues against the coexistence of the constant density liquid and gas phases in smm . for the experimental remnant distribution , we compare the first moment @xmath134 , for smm@xmath136 fig . 34a with @xmath134 for smm@xmath137 as shown in fig . the cooling of the hot fragments which carry the multifragmentation signal produces a large number of final state nucleons and light composites in agreement with the data . the imfs survive the cooling process and identify the mf transition . this is due in part to the low @xmath24 4.3 mev / nucleon critical energy for mf . for an a = 70 , z = 30 remnant , the predicted center of the backbend region energy is @xmath24 7.5 mev / nucleon . here , the hot fragments will have to emit a significantly higher fraction of their fragment mass in the cooling process and the mf signal could be severely attenuated . thus the coulomb energy , which lowers the mf threshold and reduces the effective latent heat , plays a central role in mf . here the vanishing transition energy reflects the changes in the surface , volume , and coulomb energies associated with mf rather than the traditional latent heat in a liquid - gas phase transition , which primarily involves the transfer of individual nucleons from the liquid to the gas , where both phases are at constant density @xmath167 and @xmath168 . we have compared the eos multifragmentation results for 1a gev au on c with the smm model where the volume parameterization is determined by experiment . we have found that the standard smm parameters produced excellent agreement . the input to smm consisted of the @xmath169 , and @xmath69 of the individual remnants produced in some 32,000 fully reconstructed events . thus , in contrast to earlier experiments which have been compared with smm , the @xmath170 event - by - event input data were obtained from experiment rather than by using smm to constrain average input conditions , or by the use of a theoretical first stage transport model . we find that smm is in very good agreement with the observed fragment charge yield distribution , fragment multiplicity distributions , total charged particle multiplicity , isotope ratio temperatures and caloric curve . this agreement is obtained when the expansion energy is subtracted from the experimentally determined excitation energy of the remnant . the agreement provides further confirmation that multifragmentation can be described as an equilibrium process . smm also predicts a power law for the fragment mass yield distribution at essentially the same multiplicity or excitation energy as is observed in the data . other features characteristic of a continuous phase transition , such as the scaling function , are also reproduced . the critical exponents @xmath87 , @xmath123 , and @xmath90 were obtained from smm for comparison with the experimental values . the agreement is excellent for @xmath87 , which is based on results obtained just at the critical point , fair for @xmath123 which is based on results obtained above and below the critical point , and poor for @xmath90 , which is based on results well above the critical point . the effect of cooling of the primary hot fragments produced by smm has been evaluated . the primary hot fragment yield distributions for @xmath1453 are only minimally affected by the cooling process below @xmath69 = 7 mev / nucleon . the smm @xmath90 values are substantially affected by cooling , while @xmath87 and @xmath123 are unaffected by cooling . the scaling function obtained for smm@xmath136 is fully collapsed into a very narrow band . in contrast to fragments with z@xmath1713 , the yield of particles with z@xmath42 is substantially increased by cooling , primarily as a result of the fermi breakup of light fragments . we have also performed smm calculations in which the experimental remnant distribution has been replaced by a single average remnant . the results are virtually unchanged indicating that the mixing of different remnants in the experimental data does not affect the results . we have used both multiplicity and excitation energy as the control parameter in the determination of critical exponents and related quantities and find no difference . the nature of the phase transition predicted by smm has been examined using the microcanonical equation of state . for lighter remnants we find evidence of backbending in the caloric curve and a negative system specific heat , which are the signatures of a first - order phase transition . we estimate the transition energy by means of a maxwell construction . the transition energy decreases with increasing remnant mass and charge and may extrapolate to zero just above @xmath172 , which might suggest a continuous phase transition in the breakup of this remnant . here the positive system specific heat peaks at the same @xmath69 value for which both data and smm fragment mass yield distributions obey a power law . for both the first order and continuous phase transition cases , smm@xmath136 indicates that the multifragmentation final state consists of droplets with @xmath173 and that particles having @xmath174 account for only @xmath175 of the mass in the mf region . the smm results agree with the theoretical expectations for a small ( @xmath24 150 constituents ) isolated system . @xcite . the most probable equilibrium state of a highly excited small isolated system with short range interactions is an inhomogeneous state . the addition of the long range coulomb force lowers the mf transition temperature significantly for heavier remnants and can influence the order of the thermal phase transition . both the microcanonical equation of state and thermally scaled kinetic energy fluctuation arguments favor a continuous phase transition . we thank dr . s. botvina for giving us the smm code and for valuable discussions concerning its use . this work was supported by the u.s . department of energy under contract numbers ( de - fg02 - 88 er 40412c ) . j. hbele et al . a * 340 * , 263 ( 1991 ) . jeong et al . , phys . lett * 72 * , 3468 ( 1994 ) . p. desesquelles et al . , phys . c * 48 * , 1828 ( 1993 ) j. c. steckmeyer et al . lett . * 76 * , 4895 ( 1996 ) . k. b. morley et al . , phys . c * 54 * , 737 ( 1996 ) . g. f. peaslee et al . , phys . c * 49 * , r2271 ( 1994 ) . w. g. lynch , ann . nucl . part . * 37 * , 493 ( 1987 ) . l. g. moretto and g. j. wozniak , annu . nucl . part . * 43 * , 379 ( 1993 ) . g. peilert , h. stocker , and w. greiner , rep . . phys . * 57 * , 533 ( 1994 ) . m. l. gilkes et al . rev . lett . * 73 * , 1590 ( 1994 ) . j. a. hauger et al . , phys . lett . * 77 * , 235 ( 1996 ) . j. b. elliott et al . , phys . lett . b * 381 * , 35 ( 1996 ) . j. a. hauger et al . , phys . c * 57 * , 764 ( 1998 ) . j. b. elliott et al . , b * 418 * , 34 ( 1998 ) . j. lauret et al . , phys . c * 57 * , r1051 ( 1998 ) . b. k. srivastava et al . , phys . c * 60 * , 0604606 ( 1999 ) . j. a. hauger et al . , phys . c * 62 * , 024616 ( 2000 ) . j. b. elliott et al . , phys . c * 49 * , 3185 ( 1994 ) . j. b. elliott et al . , phys . c * 55 * , 1319 ( 1997 ) . j. a. gaidos et al . , phys . rev . * 42 * , 82 ( 1979 ) . j. e. finn et al . , phys . lett . * 49 * , 1321 ( 1982 ) . r. w. minich et al . rev . lett . b * 118 * , 458 ( 1982 ) . a. s. hirsch et al . , phys . c * 29 * , 508 ( 1984 ) . t. c. sangster et al . b * 188 * , 29 ( 1987 ) . m. mahi et al . , phys . * 60 * , 1936 ( 1988 ) . n. t. porile et al . , phys . c * 39 * , 1914 ( 1989 ) . m. e. fisher , physics ( n.y . ) * 3 * , 255 ( 1967 ) . c. j. waddington and p. s. freier , phys . c * 31 * , 888 ( 1985 ) . x. campi , j. phys . a * 19 * , l917 ( 1986 ) . x. campi , phys . b * 206 * , 351 ( 1988 ) . a. schttauf et al . , phys . a * 607 * , 457 ( 1996 ) . j. bondorf et al . , phys . b * 150 * , 57 ( 1985 ) ; nucl . a * 443 * , 321 ( 1985 ) ; nucl . a * 444 * , 460 ( 1985 ) . phys . rep . * 257 * , 133 ( 1995 ) ; d. h. e. gross , rep . phys . * 53 * , 605 ( 1990 ) . j. hbele et al . , phys . c * 46 * , r1577 ( 1992 ) . h. w. barz et al . , a * 561 * , 466 ( 1993 ) . p. kreutz et al . , a * 556 * , 672 ( 1993 ) . a. botvina et al . , nucl . phys . a * 584 * , 737 ( 1995 ) . p. desquelles et al . , nucl . phys . a * 604 * , 183 ( 1996 ) . m. dagostino et al . , phys . b * 371 * , 175 ( 1996 ) . h. xi et al . , z. phys . a * 359 * , 397 ( 1997 ) . c. williams et al . c * 55 * , r2132 ( 1997 ) . x. campi , h. krivine , and e. plagnol , phys . c * 50 * , r2680 ( 1994 ) . w. bauer and a. botvina , phys . c * 52 * , r1760 ( 1995 ) . w. bauer and w. a. friedman , phys . lett . * 75 * , 767 ( 1995 ) . p. f. mastinu et al . , phys . c * 57 * , 831 ( 1998 ) . d. g. ravenhall , c. j. pethick , and j. m. lattimer , nucl . a * 407 * , 572 ( 1983 ) . s. fritz , ph.d . thesis , university of frankfurt ( 1997 ) . p. danielewicz , phys . c * 51 * , 716 ( 1995 ) . p. g. warren , ph.d . thesis , purdue university , 1996 . s. albergo et al . , nuovo cimento a * 89 * , 1 ( 1985 ) . b390 , 41 ( 1997 ) . m. j. huang et al . , phys . lett . * 78 * , 1648 ( 1997 ) . m. b. tsang et al . , phys . * 78 * , 3836 ( 1997 ) . h. xi et al . , phys . b431 , 8 ( 1998 ) . h. f. xi et al . c * 59 * , 1567 ( 1999 ) . h. xi et al . , phys . . c * 58 * , r2636 ( 1998 ) . j. pochodzalla et al . , phys . lett . * 75 * , 1040 ( 1995 ) . j. b. natowitz et al . , phys . c * 52 * , r2322 ( 1995 ) . x. campi , h. krivine , and e. plagnol , phys . b * 385 * , 1 ( 1996 ) . l. g. moretto et al . , phys . 76 * , 2822 ( 1996 ) . v. serfling et al . , phys . lett . * 80 * , 3928 ( 1998 ) . w. f. j. muller , _ proceedings of the international workshop xxvii on gross properties of nuclei and nuclear excitations _ , hirschegg , austria , p 200 , gsi 1999 . j. p. bondorf , a. s. botvina , and i. n. mishustin , phys . c * 58 * , r27 ( 1998 ) . d. stauffer and a. aharony , _ introduction to percolation theory _ , 2@xmath176 ed . , taylor and francis , london , 1992 . h. nakanishi and h. e. stanley , phys . b * 22 * , 2466 ( 1980 ) . d. stauffer , phys . 54 * , 1 ( 1979 ) . j. b. elliott et al . , submitted to phys . rev . c. b. k. srivastava , _ proceedings of the international workshop xxvii on gross properties of nuclei and nuclear excitations _ , hirschegg , austria , p 247 , gsi 1999 . d. h. e. gross , phys . 279 * , 119 ( 1997 ) ; _ proceedings of cris 96 _ , world scientific , p. 296 . a. hller , z. phys . b * 93 * , 401 ( 1994 ) . r. p. scharenberg , _ proceedings of the international workshop xxvii on gross properties of nuclei and nuclear excitations _ , hirschegg , austria , p 237 , gsi 1999 . s. levit and b. bonche , nucl . a * 437 * , 426 ( 1985 ) . m. dagostino et al . , nuclear physics a * 647 * , 153 ( 1999 ) . chomaz and f. gulminelli , nuclear physics a * 650 * , 329 ( 1999 ) . m. dagostino et al . 473b * , 219 ( 2000 ) . ph chomaz et al . , phys . lett . * 85 * , 3587 ( 2000 ) . o. mulken and p. borrmann , phys . c * 63 * , 024306 ( 2001 ) . gross , m. e. madjet , o. schapiro , z. phys . d * 39 * , 75 ( 1997 ) . lll + + @xmath177&volume binding energy of cold nuclear matter & 16 mev + @xmath178&surface tension of the cold nucleus & 18 mev + @xmath179&neutral matter critical temperature&18 mev + @xmath180&symmetry energy coefficient & 25 mev + d&crack width parameter&1.4 fm + @xmath45&coulomb reduction parameter & 2 + + + + @xmath181 & inverse density parameter & 16 mev + cccccc parameter&data&smm + + @xmath88&22@xmath601&20@xmath602 + @xmath87&[email protected]@xmath182&[email protected] + @xmath123&[email protected]@xmath182&[email protected] + @xmath90&[email protected]@xmath183&[email protected]@xmath184 + & [email protected]@xmath183 \b . the first value is obtained by including the largest fragment on the `` gas '' side of @xmath88 . the second value was obtained by adjusting @xmath90 for the effect of cooling . the experimental value for the scaling function in fig . 18a was obtained using the second value of @xmath90 . ccccc parameter & & + & & + & smm@xmath137 & smm@xmath136 & smm@xmath137 & smm@xmath136 + + @xmath88&12@xmath602&5@xmath602&12@xmath602&5@xmath602 + @xmath87&[email protected]&[email protected]&[email protected]&[email protected] + @xmath123&[email protected]&[email protected]&[email protected]&0.97@xmath600 .15 + @xmath90&[email protected]&[email protected]&[email protected]&[email protected] +	multifragmentation ( mf ) results from 1a gev au on c have been compared with the copenhagen statistical multifragmentation model ( smm ) . the complete charge , mass and momentum reconstruction of the au projectile was used to identify high momentum ejectiles leaving an excited remnant of mass @xmath0 , charge @xmath1 and excitation energy @xmath2 which subsequently multifragments . measurement of the magnitude and multiplicity ( energy ) dependence of the initial free volume and the breakup volume determines the variable volume parameterization of smm . very good agreement is obtained using smm with the standard values of the smm parameters . a large number of observables , including the fragment charge yield distributions , fragment multiplicity distributions , caloric curve , critical exponents , and the critical scaling function are explored in this comparison . the two stage structure of smm is used to determine the effect of cooling of the primary hot fragments . average fragment yields with @xmath3 are essentially unaffected when the excitation energy is @xmath4 7 mev / nucleon . smm studies suggest that the experimental critical exponents are largely unaffected by cooling and event mixing . the nature of the phase transition in smm is studied as a function of the remnant mass and charge using the microcanonical equation of state . for light remnants @xmath5 100 , backbending is observed indicating negative specific heat , while for @xmath6 170 the effective latent heat approaches zero . thus for heavier systems this transition can be identified as a continuous thermal phase transition where a large nucleus breaks up into a number of smaller nuclei with only a minimal release of constituent nucleons . @xmath7 2 particles are primarily emitted in the initial collision and after mf in the fragment deexcitation process . @xmath8 present address : space systems / loral , palo alto , ca 94202 . @xmath9 present address sun kwun kwan university , suwon 440 - 746 , republic of korea . @xmath10 present address : nuclear science division , lawrence berkeley national laboratory , berkeley , ca 94720 . @xmath11 present address : mission research corp . , santa barbara , ca 93102 . @xmath12 present address : augusta state college , augusta , ga 30910 . @xmath13 present address : brookhaven national laboratory , upton , ny 11793 . @xmath14 present address : st mary s college , morage , ca 94575 . @xmath15 universitt heidelberg , germany . @xmath16 present address : ohio state university , columbus , oh 43210 . @xmath17 present address : the svedberg laboratory , university of uppsala , s751 - 21 , uppsala , sweden . @xmath18 present address : crump institute for biological imaging , ucla , los angeles , ca 91776 . @xmath19 mit lincoln laboratory , lexington , ma 02420 . @xmath20 present address : naval research laboratory , washington , d.c . 20375 . @xmath21 deceased .
in view of einstein s theory of general relativity , the accelerating expansion of the universe was interpreted as being driven by a tiny positive cosmological constant . even though this assumption is consistent with observational data , the physical origin of the cosmological constant , its theoretical expectation values which are exceeding the observational limits and its presence , in all stages of the evolution of the universe , leading to an eternally acceleration , seem to favor alternative scenarios to nearly flat universe @xcite . recently , in order to explain the current cmb anisotropy , supernovae and lss data , and the age of the universe , the spatially open friedmann robertson walker ( frw ) spacetime fueled with different types of matter sources has been seen as a possible candidate @xcite . since an open frw universe with a sufficiently small curvature can approximate a flat frw background , in the context of recently proposed nonlinear massive gravity , open frw universes driven by arbitrary matter sources have been thoroughly investigated @xcite . another reason for dealing with negatively curved frw models is related to the origin and evolution of the large - scale magnetic field which has been detected not only in galaxies , galaxy clusters and remote protogalactic structures , but also in empty intergalactic space @xcite . in contrast to the spatially flat frw cosmological models , where the magnetic field is drastically diluted by the universal expansion , in the case of open universes , the cosmological magnetic field can experience a superadiabatic amplification , irrespective of their matter content and one may assume that this has led to the strong residual magnetic fields which seeded the galactic dynamo @xcite . as for the matter content , besides the ordinary constituents ( matter , radiation etc ) , an exotic fluid called _ dark energy _ , with a large negative pressure , has been added for explaining the expansion history and growth of structure of the universe @xcite . in this respect , the cosmological constant is undoubtedly the simplest and the most appealing candidate and the standard @xmath1 ( cosmological constant + cold dark matter ) model has been remarkably successful in describing the real universe , from the matter - dominated phase to the recent accelerated expansion . nevertheless , this theory is facing small scale controversies coming from recent observations @xcite . in order to solve this tension , viable alternative models have been formulated , based on various phenomenological non - constant equations of states ( eos ) , with the parameter around the value @xmath2 @xcite . in the present work , we deal with the friedmann equation with a cosmological constant and two interacting fluids , as sources of a spatially open @xmath3frw spacetime . the first fluid is characterized by a dust - type eos ( @xmath4 ) while the second one has a general eos where the dependence on the hubble function is manifest , like in the case of viscous fluids . the form of the interaction term is not unique and it is usually postulated from different physical arguments or derived as a solution of some dynamical equations describing the required properties of the matter fields @xcite . the case of the dust coupled with dark energy is particularly interesting , leading to an effective eos and predicting a future evolution of the universe different from the de sitter one . the non - constant values of @xmath5 are supported by measurements of cosmic microwave background anisotropies , supernovae luminosity distances and baryonic acoustic oscillations . recently , by analyzing a large sample of cosmological datasets , a clear indication for @xmath6 , at low redshift , has been found @xcite . let us consider the 4-dimensional @xmath0frw metric written as @xmath7-dt^2,\ ] ] where @xmath8 is the scale factor . with the ricci tensor components @xmath9 \delta_{\alpha \alpha } \ ; ; \nonumber \\ & & r_{44 } \ , = -3 \frac{\ddot{a}}{a } \ , , \end{aligned}\ ] ] and the ricci scalar @xmath10 , \ ] ] the einstein tensor components , defined by @xmath11 , where @xmath12 , read @xmath13 \delta_{\alpha \alpha } \ ; ; \nonumber \\ * & & g_{44}= 3 \left [ \left ( \frac { \dot{a}}{a } \right)^2 - \frac{1}{a^2 } \right ] , \end{aligned}\ ] ] where the overdot means the derivative with respect to the cosmic time @xmath14 . in view of homogeneity and isotropy of the universe , we assume the energy - momentum tensor of a perfect fluid , characterized by its proper density and the pressure @xmath15 and @xmath16 . the einstein s system of equations , @xmath17 gets the explicit form @xmath18 where @xmath19 is the hubble s rate and @xmath20 . by combining the two equations in ( 5 ) , we get the well - known relation @xmath21 where the quantities @xmath22 and @xmath23 are satisfying the equation of state @xmath24 . let us assume the existance of two ideal fluids , with @xmath25 and @xmath26 , with an interaction term between them @xmath27 , described by the relations @xmath28 for the pressureless matter source characterized by @xmath25 and @xmath29 , and @xmath30 for the other source , so that the continuity equation ( 6 ) is the summation of ( 7 ) and ( 8) . one of the most common form of the interaction term @xmath27 can be found in @xcite , where the authors are building a model which allows energy inputs and outflows proportional to the densities of the two fluids , in an expanding frw universe with zero spatial curvature . the total energy being conserved , they have started with the following expression of the exchanged energy , written in terms of the hubble parameter and the two energy densities as @xmath31 where @xmath32 and @xmath33 are two positive arbitrary dimensionless constants . in the followings , we are using the relation ( 9 ) and we are postulating an eos of the form @xmath34 where , for the moment , @xmath35 ( expressed in units of @xmath36 ) is assumed to be positive , @xmath5 is a constant that will be determined a bit later and @xmath37 is given by the friedmann s equation @xmath38 simple corrections to the net pressure of the form @xmath39 , where @xmath40 is a constant parameter , have been proposed for the frw universe filled with _ stiff fluid _ and the frw universes have been classified according to the numerical values of the bulk viscous coefficient @xmath40 @xcite . for a motivation of the necessity of using generalized inhomogeneous eos fluids , in the context of singular inflation , we recommend @xcite . by switching to the scale function as variable , the equations ( 7 ) and ( 8) become @xmath41 @xmath42 where @xmath43 . these are leading to the following differential equation satisfied by @xmath25 , @xmath44 \nonumber \\ * & & + \left [ 3(w+1)(\alpha + 3 ) + 3 \gamma + \beta \kappa \left ( \alpha + \gamma + 3 \right ) \right ] \rho_m + \beta \gamma \lambda + \frac{3 \beta \gamma}{a^2 } = 0 \nonumber \\\end{aligned}\ ] ] while @xmath26 is given by ( 12 ) as @xmath45 .\ ] ] in order to get a dust matter contribution in the energy density @xmath25 , one should impose the following relation between the model parameters @xmath46 so that the solution of ( 14 ) is given by the expression @xmath47 } \ , , \ ] ] where the first term corresponds to the dust - type source ( with zero pressure ) , the second term comes as a curvature one , while the last term is proportional to the cosmological constant . we notice that , for large @xmath48s , one is only left with a negative matter energy density , which does actually mean exotic matter , so that , the whole universe becomes a giant wormhole . the same type of contributions are in the expression of @xmath26 , computed with ( 15 ) , as @xmath49 } , \ ] ] while the total energy density reads @xmath50 } \ ; .\ ] ] putting everything together in ( 10 ) , one is able to derive the pressure @xmath51 } , \ ] ] which is a positive quantity given by the curvature term and the cosmological constant , which keeps open the wormhole throat ( pressing against it ) . finally , from the hubble function written from ( 10 ) as @xmath52 , \ ] ] one gets the relation @xmath53 } \lambda \ , , \end{aligned}\ ] ] pointing out the special value @xmath54 for which @xmath37 gets the simplified expression @xmath55}{9 \left [ 2 \gamma + \alpha ( \alpha + \gamma + 1 ) ( \alpha + \gamma + 3 ) \right ] } \lambda \ , .\ ] ] by introducing the notations @xmath56}{9 \left [ 2 \gamma + \alpha ( \alpha + \gamma + 1 ) ( \alpha + \gamma + 3 ) \right ] } \lambda \ ; \ ; { \rm and } \ ; \ ; d= \frac{c(\alpha + \gamma ) \kappa}{3 \gamma}\ ] ] one ends up with the equation @xmath57 which leads , by integration , to the following scale function @xmath58^{2/3 } \ , .\ ] ] the acceleration parameter @xmath59\ ] ] turns from negative values , for @xmath60 to positive values afterwards , suggesting an initial decelerating stage followed by an accelerating one . as expected , in view of the previous results , in the very distant future , the expression ( 25 ) can be approximated to the de sitter solution @xmath61 for small @xmath62s , the scale function is given by @xmath63 pointing out that this model behaves like a cold - matter - dominated universe with @xmath64 being initiated by a standard _ big bang _ process . the relations ( 19 ) and ( 20 ) , with the condition ( 22 ) lead to the following generalized eos characterizing the universe with the scale function ( 25 ) @xmath65 } \lambda \nonumber \\ & = & - \frac{1}{3 } \rho_t + \frac{1}{\kappa } h^2 - \frac{2 \gamma \lambda}{3 \kappa \left [ 2 \gamma + \alpha ( \alpha + \gamma + 1 ) ( \alpha + \gamma + 3 ) \right ] } \ , . \end{aligned}\ ] ] let us turn now to the other possible situation , characterized by a negative @xmath35 value . for simplicity , let us just change the @xmath35 sign in ( 10 ) , i.e. @xmath66 now , @xmath5 is a positive constant given by @xmath67 and the energy densities ( 17 ) and ( 18 ) are both positive , given by the relations @xmath68}\ ] ] and @xmath69 } .\ ] ] these are leading , by summation , to the ( positive ) total energy density @xmath70 } , \ ] ] where the constant term can be identified with the cosmological density @xmath71 . the pressure , computed with ( 27 ) , corresponds to a dark energy component , @xmath72 } \ , , \ ] ] coming from the curvature and the cosmological constant contributions . the hubble function coming from ( 27 ) , i.e. @xmath73 \nonumber \\ * & = & \frac{c(\alpha + \gamma ) \kappa}{3 \gamma a^3 } + \frac{\alpha ( \alpha + \gamma + 1 ) + ( \alpha + 1)(\alpha+\gamma)\beta \kappa}{\alpha ( \alpha + 1 ) + \gamma ( \alpha + \beta \kappa ) } \frac{1}{a^2 } \nonumber \\ * & & + \frac{3 \alpha ( \alpha + \gamma + 3 ) + ( \alpha + 3)(\alpha + \gamma ) \beta \kappa}{9 \left[\alpha ( \alpha + 3 ) + \gamma ( \alpha + \beta \kappa ) \right ] } \lambda \equiv \frac{d}{a^3 } + \frac{f}{a^2 } + \lambda_0 \ , , \end{aligned}\ ] ] leads , by integration , to an elliptic function from which the dependence @xmath74 is impossible to derive analytically . however , in the particular case of zero cosmological constant , the relation ( 33 ) gets the simple form @xmath75 leading to the transcendental equation @xmath76 with @xmath77 an integration constant corresponding to a special universal ( cosmic ) moment . based on the substitution @xmath78 we can derive from ( 34 ) the following representation @xmath79 .\end{aligned}\ ] ] from the system above , we notice that for @xmath80 and @xmath81 we get @xmath82 , pointing out the singular cosmic event , at @xmath77 . also , for small @xmath83 , when the density of matter term dominates the curvature term , the behavior is @xmath84 , as in the previous case . on the other hand , as the time evolves , the curvature of space starts to dominate the matter density , which is more and more diluted and we are recovering the _ milne model _ of the universe , with @xmath85 . the equation of state parameter @xmath86 , computed with ( 32 ) and ( 30 ) for @xmath87 , is a function of @xmath83 , decreasing from zero , at @xmath88 , and tending to the negative constant value @xmath89 , \ ] ] in the far future . as in @xcite , let us consider the particular case @xmath90 so that the interaction term is @xmath91 . such a choice for an open universe has been proved to be consistent with the observed universe expansion epochs , i.e. first deceleration and then acceleration . the same expression of @xmath27 has been used in the analysis of generalized ghost pilgrim dark energy in non - flat frw universe @xcite , the negative sign meaning that the cdm decays into de . the particular form of the system ( 12 , 13 ) , i.e. @xmath92 @xmath93 with @xmath94 given in ( 10 ) , is satisfied by the energy densities @xmath95 and @xmath96 so that the total energy density reads @xmath97 and the total pressure is given by @xmath98 one may notice that , for the particular value @xmath99 , the energy density @xmath25 is corresponding to radiation , while the relations ( 40 ) and ( 41 ) become @xmath100 and @xmath101 for the relation ( 16 ) between the model parameters , which in this particular case is @xmath102 , one deals with an universe filled with a mixed source made of dust and radiation , with the total energy @xmath103 the corresponding pressure , @xmath104 , \ ] ] is negative for negative values of @xmath35 , pointing out a dark energy component , as in the general analysis developed in section 3 . the other particular case with @xmath105 is characterizing an universe with stiff matter and radiation with the total energy density @xmath106 and pressure @xmath107 for negative @xmath35 , the energy density is always positive and the eos parameter @xmath108 is decreasing from the maximum value @xmath109 , for @xmath110 to @xmath111 , in the far future , gaining all the values inbetween . in the present paper , we have considered a @xmath0frw universe filled with two interacting ideal fluids , characterized by @xmath112 and a general eos of the form ( 10 ) respectively , and we have derived the total energy density , the pressure and the corresponding scale function . the main constraint on the model parameters , ( 16 ) , is generated by imposing the existence of a pressureless matter term in the energy density ( 17 ) , besides the curvature and the cosmological constant contributions . one may notice that , with the special value of the coupling parameter ( 22 ) , the simplified relation ( 24 ) can be written , in terms of the redshift , as @xmath113 this is of the form @xmath114 , \ ] ] corresponding to the flat @xmath1 model , with @xmath115 where the index `` 0 is for the present time values . in order to determine the values of the model s parameters @xmath116 and @xmath117 , one may use the current fractional density of matter estimated from nine - year wmap observations , @xmath118 , for the present hubble parameter @xmath119 = 0.705 $ ] @xcite . even though , in the last two decades , the @xmath1 model has been remarkably successful in describing the large - scale structure of the universe , the predictions at small scales are controversial and disagree with recent observations @xcite . among the attempts to relieve the tension , we mention the ones which are changing the nature of spacetime to a non - flat background @xcite or the nature of the dark matter , by adding a hot component @xcite or a viscous one @xcite . nevertheless , for values of @xmath35 which are not fulfilling the condition ( 22 ) , the form of the hubble function is @xmath120 } { [ \alpha ( \alpha + \gamma + 1 ) - \gamma \beta \kappa ] a_0 ^ 2 } ( 1+z)^2 + \lambda_0 \nonumber \\ * & = & \frac{d}{a_0 ^ 3 } ( 1+z)^3 + f(z ) + \lambda_0 \ , , \nonumber\end{aligned}\ ] ] and the lower values of @xmath121 at higher redshifts can be achieved when the dynamically evolving term @xmath122 is negative . among the numerous forms of this screening term , the curvature one has been considered as a natural and plausible choice @xcite , to satisfy the results from boss experiment @xcite . on the other hand , the eos ( 26 ) is similar to the one of a inhomogeneous viscous dark fluid coupled with dark matter in the frw universe , @xmath123 , where @xmath124 is the bulk viscosity . in the simplest case , @xmath40 is taken as a positive constant , @xmath125 , while in other theories , it is proportional to @xmath121 as @xmath126 , with @xmath127 a positive constant @xcite . the form of the eos parameter ( 36 ) is sustained by a number of scalar field models which have proposed a dynamical ( time - dependent ) @xmath5 parameter , such as quintessence ( with @xmath128 ) , phantom ( @xmath5 lies below -1 ) @xcite and quintom matter which crosses over the cosmological constant boundary @xmath2 @xcite . in the final part of the paper , we have concentrated on the special case of one - direction outflow , from the postulated dust - type component characterized by @xmath25 and @xmath29 to the second component with the general eos ( 10 ) . quite intriguing , if one assumes @xmath99 , the first component is behaving radiation - like , until it starts to earn energy , according to ( 9 ) . caldwell , r. r. , dave , r. and steinhardt , p.j . 1998 , , 80 , 1582 liddle , a.r . and scherrer , r.j . 1999 , , 59 , 023509 dvali , g.r . , gabadadze , g. and porrati , m. 2000 , phys . lett b , 484 , 112 sahni , v. and shtanov , y. 2003 , jcap , 0311 , 014 aurich , r. , steiner , f. 2002 , , 334 , 735 gron , o. , johannesen , s. 2011 , eur . phys . j. plus , 126 , 29 gumrukcuoglu , a.e . , lin , c. l. and mukohyama , s. 2011 , jcap , 1111 , 030 tavecchio , f. _ et al . _ 2010 , , 406 , l70 ando , s. and kusenko , a. 2010 , , 722 , l39 neronov , a. and vovk , l. 2010 , science , 328 , 73 essey , w. , ando s. and kusenko , a. 2011 , , 35 , 135 barrow , j. d. , tsagas , c. g. and yamamoto , k. 2012 , , 86 , 023533 chiba , t . , takahashi , r. and sugiyama , n. 2005 , class . , 22 , 3745 de vega , h. j. and sanchez , n. g. 2011 , preprint arxiv:1109.3187 weinberg , d. h. _ et al . _ 2013 , preprint arxiv:1306.0913 nojiri , s . , odintsov , s. d. and oikonomou , v. k. 2015 , phys . b , 747 , 310 timoshkin , a. v. 2009 , the open astronomy journal , 2 , 39 jamila , m. , and rashidb , m. a. 2008 , eur . phys . j. c , 56 , 429 said , n. _ et al . _ 2013 , preprint arxiv:1303.4353 [ astro - ph ] barrow , j. d. , and clifton , t. 2006 , , 73 , 103520 mathew , t.k . , aswathy , m.b . and manoj , m. 2014 , eur . phys . j. c , 74 , 3188 bahrehbakhsh , a. f. 2016 , preprint arxiv:1601.00310 jawad , a. 2014 , eur . j. c , 74 , 3215 wmap collaboration . 2013 , suppl . , 208 , 19 boss collaboration ( delubac et al . ) . 2015 , astron . astrophys . , 574 , a 59 kumar , s. 2015 , 92 , 103512 jeong , k. s. , kawasaki , m. and takahashi , f. 2014 , jcap , 1402 , 046 velten , h. 2014 , int . j. geom . , 11 , 1460013 elizalde , e. , obukhov , v. v. and timoshkin , a. v. 2014 , mod . lett . a , 29 , 1450132 caldwell , r.r . 2002 , phys . b , 545 , 23 caldwell , r. r. , kamionkowski , m. and weinberg , n. n. 2003 , , 91 , 071301 feng , b. , wang , x. l. and zhang , x. m. 2005 , phys . b , 607 , 35 cai y.f . _ 2010 , phys . rept . , 493 , 1 .	in the present paper , we are analyzing the @xmath0frw universe with two interacting ideal fluids . the energy densities and the pressures are used to derive the corresponding equations of state . for a particular form of the interaction term between the two fluids , it turns out that the universe is filled with a mixture of dust and radiation .
global monopoles have been investigated for years as possible seeds for structure formation in the universe @xcite . although they appear to be ruled out by the latest cosmological data @xcite , their appearance in condensed matter and other systems and their peculiar properties make them worthy of investigation . these objects have divergent energy , due to the slow fall - off of angular gradients in the fields , which has to be cut - off at a certain distance r ( in practice , the distance to the nearest monopole or antimonopole ) and has two important consequences , in particular for cosmology . first , the evolution of a network of global monopoles is very different from that of gauged monopoles , as long - range interactions enhance annihilation to the extent of eliminating the overabundance problem altogether @xcite . second , their gravitational properties include a deficit solid angle @xcite , which makes them rather exotic . the stability of global @xmath0 monopoles has been the subject of some debate in the literature @xcite . in this paper we try to settle the issue by : a ) analysing the axial perturbation equations in the limit where the cut - off is taken to infinity , and b ) proving that the energy barrier between the monopole and the vacuum ( meaning the _ extra _ energy required by the monopole to reach an unstable configuration that decays to the vacuum ) is _ finite_. it is somewhat surprising for different topological sectors to be separated by finite energy barriers , but in this case it is a consequence of the scale invariance of gradient energy on two dimensional surfaces ( @xmath1 constant ) , and therefore independent of the details of the scalar potential . we consider the simplest model that gives rise to global monopoles , the @xmath0 model with lagrangian : @xmath2 @xmath3 is a scalar triplet , @xmath4 and @xmath5 . the @xmath0 symmetry is spontaneously broken to @xmath6 , leading to two goldstone bosons and one scalar excitation with mass @xmath7 . the set of ground states is the two - sphere @xmath8 and , since @xmath9 , there are field configurations with non - trivial topological charge . one such configuration with unit winding is the spherically symmetric monopole , @xmath10 where @xmath11 and @xmath12 . its asymptotic behaviour is @xmath13 , @xmath14 and @xmath15 , as can be seen from the e.o.m . of @xmath16 @xmath17 the two parameters ( @xmath18 ) appearing in the lagrangian can be absorbed by the rescaling @xmath19 which amounts to choosing @xmath20 as the unit of energy and the inverse scalar mass as the unit of length ( up to a numerical factor ) . note , however , that the energy of a configuration with non - trivial winding such as ( [ mon ] ) is ( linearly ) divergent with radius , due to the slow fall - off of angular gradients and has to be cut off at @xmath21 , say . unlike @xmath22 and @xmath23 , the ( rescaled ) cut - off is an important parameter which could affect the dynamics of solutions with non - trivial topology . dropping tildes : @xmath24 since the energy diverges , derrick s theorem does not apply in this case , and in @xcite it was shown that the global monopole is stable towards radial rescalings . on the other hand the question of stability with respect to angular perturbations has led to some discussion in the literature after goldhaber @xcite pointed out that the ansatz @xmath25 which describes axially symmetric deformations of the spherical monopole ( [ mon ] ) , leads to the following expression for the energy after a change of variables @xmath26 ) : @xmath27 \nonumber\\ & \rho_1 = f_y^2+f^2 [ \sin^2({\bar \theta } ) + ( { \bar \theta}_y)^2 ] \nonumber\\ & \rho_2 = f_r^2 + f^2 ( { \bar \theta}_r)^2 + { 1 \over 2 } ( f^2 -1 ) ^2 \label{energy}\end{aligned}\ ] ] with @xmath28 , etc . the term in brackets in @xmath29 is identical to the energy of the sine - gordon soliton , so translational invariance in @xmath30 implies that configurations with @xmath31 have the same energy as ( [ mon ] ) ( which corresponds to @xmath32 ) . on a given @xmath33const . shell , the effect of taking @xmath34 is to concentrate the angular gradients in an arbitrarily small region around the north pole . when the gradient energy is inside a region of size comparable to the inverse scalar mass , it is energetically favourable to `` undo the knot '' by reducing the modulus of the scalar field to zero and climbing over the top of the mexican hat potential . unwinding is estimated to occur at a critical value of @xmath35 ( say @xmath36 ) whose dependence with @xmath37 far from the core is logarithmic , @xmath38 numerical simulation on individual shells closer to the core gives @xmath39 with slowly varying @xmath40 and @xmath41 . unwinding is expected if @xmath42 . + as explained in @xcite , the shift ( [ globalshift ] ) creates a tension pulling the monopole core , and the apparent unwinding ( which starts in the inner shells ) is only a manifestation of the core s translation . in order to stop the motion of the core , we consider a hybrid configuration such that the monopole core remain unperturbed and the unwinding occur in the outer shells . this is achieved by taking @xmath43 for , say , @xmath44 , a string - like configuration for @xmath45 , and some continuous interpolation in between . one such configuration ( see figure [ image]a ) would be : @xmath46 with @xmath47 , @xmath48 , and @xmath49 given by continuity of @xmath50 . if the energy ( i.e. mass ) inside @xmath51 is large enough there will be no appreciable motion of the monopole core ( because the tension of the string is constant @xmath52 ) . for instance , simulations with @xmath53 , @xmath54 , @xmath55 show different behaviour depending on @xmath56 : for @xmath57 the core translates , but for @xmath58 , the unwinding happens far from the core , which remains fixed . in the latter case , the decay of the string can also be understood as a monopole - antimonopole pair creation with the new monopole appearing at @xmath21 and the antimonopole appearing near @xmath59 . the equations solved in the simulations were @xmath60 with a dissipative term added to make the integration faster . simulations using different @xmath61 show no appreciable difference . the equations were integrated using cylindrical coordinates ( @xmath62 ) in a @xmath63 grid using explicit runge - kutta method with step size control @xcite . the output can be seen in fig . [ image]b , where the potential energy is plotted at various times , confirming that the configuration ( [ init ] ) is unstable and decays to the vacuum . we will now show that the extra energy required to reach this unstable configuration ( [ init ] ) from the spherical monopole configuration ( [ mon ] ) is _ finite_. consider the set of configurations with @xmath64 and @xmath65 . from ( [ energy ] ) the difference between the energy of any such configuration ( @xmath66 $ ] ) and the energy of the spherically symmetric unperturbed monopole ( @xmath67 $ ] ) is : @xmath68-e[0 ] & = & \left [ \int_{0}^{r _ * } + \int_{r_}^r \right]\ , \rd r\,\left [ { { r^2 } \over 2 } f^2(r ) \xi_r^2(r ) i[\xi(r ) ] \right]\nonumber\\ i[\xi]&=&\int_{-\infty}^{+\infty } \rd y \ , \sech^2(y ) \ , \sech^2(y+\xi(r))\ \ . \label{ixi}\end{aligned}\ ] ] @xmath69 $ ] is bell - shaped : from a maximum value @xmath70=4/3 $ ] , it rapidly falls to zero for @xmath71 as @xmath72 . @xmath73 is the radius at which @xmath74 . the first integral is clearly finite . the second can be estimated using the asymptotic form of @xmath69 $ ] and is negligible for @xmath75 even in the limit @xmath76 , where one gets @xmath77 . moreover , there is a continuous path in configuration space connecting the configurations ( [ mon ] ) and ( [ init ] ) such that @xmath66 -e[0]$ ] remains finite along the whole path : first increase @xmath35 in the outer shells @xmath78 using goldhaber s deformation ( @xmath79 ) until @xmath35 reaches @xmath80 ; then adjust the radial dependence to match ( [ init ] ) . thus , we have shown that the extra energy required by the monopole ( [ mon ] ) to go over the energy barrier and decay to the vacuum is finite even as @xmath81 ; moreover , since the monopole energy grows with r , the ratio as @xmath82 . we now turn to the classical stability of ( [ mon ] ) by considering small perturbations parametrised by @xmath83 neglecting quadratic terms gives @xmath84 , @xmath85 . introducing @xmath86 , @xmath87 which shows that the correct boundary conditions are that @xmath88 should vanish on the @xmath89 axis . note that @xmath90 and @xmath91 need not vanish . an infinitesimal translation of the monopole in the @xmath89 direction corresponds to @xmath92 , and both @xmath93 and @xmath94 tend to @xmath95 as @xmath96 ( see fig [ pot]a ) . there is no zero mode associated with global rotations since these have been factored out in the ansatz ( [ ansatz ] ) . + as usual , the perturbation equations @xmath97\delta + 2 x_y - 4 x \tanh y & \nonumber\\ & 0= \sech^2y \ r^2 \box x + \sech^2y [ r^2(f^2 - 1 ) + 2 ] x + 2 \tanh y \ , x_y - 2\delta_y & \label{perteqs1}\end{aligned}\ ] ] reduce to an eigenvalue problem in @xmath98 for perturbations of the form @xmath99 , @xmath100 . eigenfunctions @xmath101 with negative eigenvalues @xmath102 correspond to instabilities . dropping hats and defining @xmath103 @xmath104 -2 \partial_u [ ( u^2 - 1 ) x ] = \omega^2 r^2 \delta & \nonumber \\ & r_2 x + \partial_u^2 [ ( u^2 - 1 ) x ] -2\partial_u\delta = \omega^2 r^2 x & , \label{perteqs2}\end{aligned}\ ] ] where @xmath105 , @xmath106 are radial operators ( see fig [ pot]b ) : @xmath107 using legendre polynomials and changing variables : @xmath108 = \sum \chi_l ( r ) p_l(u ) , \ \ \delta = \sum \delta_l(r)p_l(u ) $ ] the equations for different values of @xmath109 decouple , giving @xmath110 where we introduced @xmath111 and @xmath112 . in order to get ( [ eom2 ] ) we multiplied the @xmath113 eqn . in ( [ perteqs2 ] ) by @xmath114 and differentiated w.r.t . @xmath115 , so there may be spureous solutions ; in particular , @xmath116 corresponds to angular perturbations that are singular on the @xmath89 axis . these are not physical , and will be discarded . but if there is no solution of ( [ eom1],[eom2 ] ) with negative @xmath98 there will be no instability in the original problem ( [ perteqs1 ] ) . our task is to find the solution to ( [ eom1],[eom2 ] ) with the minimum value of @xmath98 over all admissible perturbations and all @xmath117 . we know one @xmath118 solution , the translational zero mode ( @xmath119 , @xmath120 ) . goldhaber s deformation ( @xmath121 , @xmath122 ) is also @xmath118 but is _ not _ a solution of ( [ eom1 ] , [ eom2 ] ) , and it can be shown that there are no instabilities with @xmath123 . let us first consider normalizable perturbations . note that , for each @xmath109 , the equations ( [ eom1],[eom2 ] ) can be obtained by functional variation from @xmath124 = \omega^2 \int \rd r r^2 [ \delta^2 + \chi^2 ] \qquad . \label{el}\ ] ] the lowest value of @xmath125 can be found minimising @xmath126 over normalized functions , @xmath127 = 1 $ ] and over all @xmath117 . however the minimum must be in the @xmath118 sector since , for all @xmath128 and for given @xmath129,@xmath130 , @xmath131 is a sum of squares with positive coefficients : @xmath132 \ , \ ] ] where @xmath133 /2 \ > \ 0 $ ] for @xmath128 . in order to investigate the @xmath118 sector , we rewrite @xmath134 using arbitrary functions @xmath135 , @xmath136 and @xmath137 : @xmath138 - \left . g(r)\delta^2(r ) \right\|_0^\infty + \left . h(r)\chi^2(r ) \right\|_0^\infty & \qquad . \label{big}\end{aligned}\ ] ] choosing @xmath139 , @xmath140 , @xmath141 , the coefficients of @xmath142 and @xmath143 in ( [ big ] ) vanish identically by virtue of ( [ eom ] ) . for large r , @xmath144 and @xmath145 , thus , @xmath146 for all functions ( @xmath147 , @xmath148 ) that decay faster than @xmath149 as @xmath150 . this proves that all normalizable perturbations have @xmath151 . the above argument and a host of numerical simulations strongly suggest that non - normalizable perturbations are at best compatible with @xmath152 , as can be verified directly from the equations for perturbations that fall to zero like a power of r , but in this case we have no analytic proof . in this paper we have derived and analysed the axial perturbation equations of @xmath0 monopoles and proved that , contrary to statements in the literature , @xmath0 monopoles are perturbatively stable ( or neutrally stable ) to infinitesimal , axially symmetric , normalizable ( or power - law decay ) perturbations . we have also proved that the energy barrier between topological sectors is finite , irrespective of the details of the scalar potential . this feature is specific of global monopoles ; global vortices in two dimensions , whose energy grows logarithmically with radius , and gauge monopoles , whose energy is finite , are separated from the vacuum by an energy barrier growing linearly with the cut - off . but the energy of global monopoles is dominated by two - dimensional ( angular ) gradients far from the core , and these can be deformed with no energy cost . one would naively expect thermal fluctuations with @xmath153-e[0]$ ] to cause the monopole to decay . as far as we know , this effect is not seen in `` cosmological '' numerical simulations of global monopole networks , but perhaps this is not so surprising . first , we worked in flat space . second , the range of scales introduced by the expansion of the universe forces drastic approximations on cosmological simulations ; in particular , the sigma model approximation , which is widely used , sets the field on the vacuum manifold everywhere , so unwinding and pair creation events are not resolved by the grid . finally , cosmological simulations do not include thermal effects , and it has been shown that full thermal simulation across the phase transition @xcite gives qualitatively different results . our results provide further evidence that a more careful analysis of global monopole networks may be required . ' '' '' we thank m. j. esteban for pointing out an error in an earlier version of this paper , and m. escobedo , r. durrer , r. gregory , a. vilenkin , k. kuijken , i.l . egusquiza , j.m . aguirregabiria , l. perivolaropoulos and a. lande for conversations . we acknowledge support from grants cicyt aen99 - 0315 and upv 063.310-eb225/95 . a. vilenkin , e. p. s. shellard , _ cosmic strings and other topological defects _ , cambridge university press , cambridge ( 1994 ) , and references therein . see , e.g. j. magueijo , r.h . brandenberger , preprint , astro - ph/0002030 , and references therein . r. durrer , m. kunz , a. melchiorri , preprint , astro - ph/9901377 ; r. durrer , in `` energy densities in the universe '' , proceedings of the xxxvth rencontres de moriond , edited by jacques dumarchez , to be published ( astro - ph/0003363 ) . m. barriola , a. vilenkin , phys . 63 * ( 1989 ) 341 . a. s. goldhaber , phys . rev . lett . * 63 * ( 1989 ) 2158 . s. h. rhie , d. p. bennett , phys . * 67 * ( 1991 ) 1173 . l. perivolaropoulos nucl b375 * ( 1992 ) 665 . the code `` _ dopri5 _ '' , by dormand&prince , is described in e. hairer , s.p . norsett and g. wanner , solving ordinary differential equations i , nonstiff problems , springer series in computational mathematics , springer - verlag ( 1993 ) . inequalities , @xmath154 edition ( p. 244 ) g. hardy , j.e . littlewood & g.plya , cambridge university press 1973 . n. d. antunes , l. m. a. bettencourt , a. yates , preprint , hep - ph/9901391 .	we analyse the stability of global o(3 ) monopoles in the infinite cut - off ( or scalar mass ) limit . we obtain the perturbation equations and prove that the spherically symmetric solution is classically _ stable _ ( or _ neutrally stable _ ) to axially symmetric , square integrable or power - law decay perturbations . moreover we show that , in spite of the existence of a conserved topological charge , the energy barrier between the monopole and the vacuum is _ finite _ even in the limit where the cut - off is taken to infinity . this feature is specific of global monopoles and independent of the details of the scalar potential .
qsos first appear at high redshifts @xmath1 and their peak activities are reached at @xmath2 ( e.g. peterson 1997 ) . the general evolutionary trend is that the number of bright qsos rapidly decrease at @xmath3 . such an observationally established fact has been interpreted in terms of the two drastically different theoretical models . in the pure luminosity evolution ( ple ) model ( e.g. mathez 1976 , boyle et al . 1993 , mathez et al . 1996 , page et al . 1996 ) , a single generation of long - lived ( e.g. @xmath4 yr ) qsos form at high z and evolve mainly by decreasing their luminosities . in the other model ( density evolution model ) , a multiple of qso populations rise and fall successively and conspire to result in the observed evolutionary trend . in the latter , the evolution is largely determined by the qso number density evolution ( haehnelt et al . 1997 and references therein ) . in the ple model , due to the longevity of the qsos , the massive bh remnants as massive as @xmath5 are expected in a small number of massive galaxies with quiescent nuclei . the density evolution model suggests that a massive bh is a normal feature of the central region of essentially every nearby galaxy ( e.g. magorrian et al . 1998 , salucci et al . 1998 ) . it is to be further studied whether both the luminosity evolution and the density evolution co - exist in the observed qso evolution ( e.g. miyaji et al . 1998 ) . it is also unclear in the density evolution model why qsos forming at low @xmath6 s have lower luminosities than their high @xmath6 counterparts . therefore , it is still interesting to see whether the ple model can be further probed using some testable predictions . if qsos are powered by massive accreting bhs , the qso luminosities are essentially determined by the black hole masses and their mass accretion rates ( e.g. frank et al . 1992 ) . in this sense , the qso luminosity evolution has to be related to the evolution of bh masses or the evolution of the mass accretion rates or both . in most of the previous studies in the context of the ple model , the qso luminosities have been interpreted loosely as bolometric luminosities or some band luminosities simply proportional to the bolometric luminosities ( e.g. yi 1996 and references therein ) . however , as observed in some galactic black hole candidates ( gbhcs ) , luminosities are strongly correlated with spectral states ( e.g. rutledge et al . 1998 ) . that is , as luminosities vary , emission spectra change so much that some energy bands luminosities no longer reflect the bolometric luminosity changes ( cf . narayan et al . 1998 , esin et al . 1998 ) . in the case of qsos , direct spectral changes correlated with luminosities in an individual qso have not been established . however , given the striking similarities in accretion flows emission properties between qsos ( or active galactic nuclei ( agn ) in general ) , and gbhcs , a correlation between luminosity and spectrum in qsos is quite plausible . the seyferts x - ray spectral photon index @xmath7 is similar to that of the hard state in gbhcs ( rutledge et al . 1998 ) . the soft x - ray emission in bright gbhcs ( e.g. mcclintock 1998 ) is reminiscent of the so - called optical / uv / soft x - ray `` big blue bumps '' in bright active galactic nuclei ( frank et al . it is quite reassuring that the recently observed galactic superluminal sources appear to be a scaled - down version of the superluminal radio sources in extragalactic nuclei ( mirabel and rodriguez 1998 ) . therefore , we assume that the qso luminosity evolution is accompanied by a strong spectral evolution . the main effect of the spectral evolution is that the luminosity evolution in different energy bands differ significantly . we therefore investigate the luminosity evolution trend in various energy bands ( choi et al . our approach can eliminate some potentially serious uncertainties in the qso luminosity evolution models when only the bolometric luminosity evolution is considered ( cf . yi 1996 ) . in this paper , we attempt to explain the cosmological evolution of qsos using a model in which the lf of the qso population is constructed from one generation of long - lived qsos with an evolutionary time - scale of @xmath8yr for a flat universe with no cosmological constant ( yi 1996 ) . we consider a single qso population in terms of the lf which evolves in @xmath6 . each qso is turned on at an initial @xmath6 with an initial bh mass @xmath9 and an initial mass accretion rate @xmath10 . an ensemble of qsos evolve individually while their luminosities and spectra evolve in a correlated manner ( choi et al . 1999 ) , which is directly reflected in the lf evolution . our approach to the qso evolution problem hinges on the hypothesis that the luminosity - correlated qso emission spectra are physically similar to those of the gbhcs which show spectral transitions closely connected to luminosity levels ( e.g. rutledge et al . 1998 , narayan et al . 1998 and references therein ) . for instance , optical / uv / soft x - ray emission from geometrically thin , optically thick accretion disks around supermassive bhs contributes to the big blue bump postulated in qsos ( frank et al . based on the x - ray emission spectra observed in the gbhcs , qsos spectral states are assumed to be directly related to luminosities . in this picture , hard x - ray emission generally requires a scattering corona with energetic electrons or advection - dominated accretion flows ( adafs ) ( e.g. zdziarski 1998 ) . our main assumptions are as follows . ( i ) the qsos are powered by massive accreting bhs with accretion flows . therefore , emission properties directly reflect the underlying accretion flow structure which is mainly determined by the mass accretion rate . we take the dimensionless accretion rate , @xmath11 , which is defined as a physical accretion rate scaled by the mass - dependent eddington accretion rate @xmath12 where @xmath13 and the 10% radiative efficiency has been assumed . ( ii ) the qso emission is largely composed of three spectral states , defined by the x - ray spectral hardness ; `` soft / high '' state ( hs ) , `` hard / low '' state ( ls ) , in addition , an `` off '' state ( os which is included in ls ) , and `` very high '' state ( vhs ) deduced from the observed spectral states of gbhcs ( rutledge et al . although direct evidence for qso spectral changes in a single qso does not exist ( e.g. in part due to long physical time scales appropriate for qsos and observational flux limits ) , it is plausible that qsos have undergone considerable luminosity and spectral changes in the cosmological context . we explain the cosmological evolution of qsos using the spectral states discussed above , which are mainly determined by the dimensionless accretion rate , @xmath14 ( yi 1996 , narayan et al . ( i ) at high accretion rates , @xmath15 , the spectral state of qso corresponds to vhs . while hs , ls , and os are well recorded in the gbhcs , the existence and the exact nature of vhs is less clear ( e.g. narayan et al . 1998 ) . we assume that the x - rays emitted during vhs originate in the scattering hot corona which is thought to exist above the disk at high accretion rates . the main accretion flow itself contributes to thermal soft x - ray / uv / optical emission . we adopt a fraction @xmath16 of the total luminosity for the x - ray luminosity while the bolometric luminosity , @xmath17 ( with @xmath18 , i.e. a high efficiency ) is assumed . we identify the x - ray spectrum of this state with a power law photon index @xmath19 ( rutledge et al 1998 ) and the optical / uv continuum with the thermal disk emission determined by a set of @xmath9 and @xmath10 ( frank et al . 1992 ) . however , since vhs is poorly understood , the x - ray and bolometric efficiencies remain largely unconstrained . this uncertainty is significant for the qso evolution near the initial birth epoch near birth . theoretically , the accretion flow could be in the form of the so - called slim disk with super - eddington accretion rates ( szuszkiewicz et al . ( ii ) at lower accretion rates , @xmath20 , the accretion flow corresponds to the geometrically thin disk radiating optically thick thermal emission ( e.g. frank et al . 1992 ) with a high efficiency ( @xmath18 ) . the corresponding spectral state is hs . x - ray luminosity of this state is also calculated in the same manner as in vhs ( i.e. 10% efficiency for the bolometric luminosity , @xmath17 with @xmath18 ) . the x - ray spectrum is identified with a single power law of photon index , @xmath21 , deduced from the observed hs of the gbhcs ( rutledge et al . ( iii ) as the accretion fuel is gradually exhausted , the accretion rate falls below the critical rate for the advection - dominated accretion flows ( adafs ) , @xmath22 . the corresponding spectral state is ls . when the accretion rate falls further , the os , which is seen in nearby , x - ray bright galactic nuclei appears ( yi & boughn 1998 , 1999 and references therein ) . we assume that the os is a direct extension of ls to lower luminosities for @xmath23 . during ls / os , the bolometric luminosity is adequately described by @xmath24 with @xmath25 ( e.g. narayan & yi 1995 , yi 1996 , yi & boughn 1998 , 1999 ) . in this low efficiency regime , the accretion flow becomes optically thin and radiates very weakly in optical / uv . the nonthermal emission dominates from optical / uv to x - rays ( narayan & yi 1995 ) . the spectral energy distribution roughly follows a power law with photon index @xmath26 which is observed in gbhcs in ls . the simplified spectral shapes of the above - discussed states are displayed in figure 2(a ) . the similar spectral shapes for the gbhcs have been applied to several systems ( narayan et al . 1998 , esin et al . 1998 ) . we consider a population of qsos which are born randomly within the prescribed parameter range . the initial birth occurs within such a narrow range of redshifts that the subsequent qso evolution is largely synchronous . ( i ) first , we generate a qso population which is large enough to avoid the small - number statistics problems . the total number of qsos in the sample is @xmath27 . the lf evolution is obtained under the assumption that qsos evolve as a long - lived single population with the cosmological evolution time scale , @xmath28yr ( yi 1996 ) . qsos initial birth redshifts are randomly distributed following a gaussian distribution with its center at @xmath29 and width of 1 . the initial bh masses are also randomly chosen from an initial mf which is defined as a single power - law slope of 2.5 in the range @xmath30 . the bhs are initially accreting at randomly chosen accretion rates around their respective mass - dependent eddington rates , @xmath31 . the accretion rates are assumed to be distributed as a gaussian with its center at @xmath32 and width of 0.1 . ( ii ) each qso in the sample evolves as the mass accretion in the qso decreases and the bh mass grows through mass accretion starting with the initial conditions . the evolutionary time - scale , @xmath33 is taken as a fraction of the cosmic time @xmath34 for a flat universe ( @xmath35 ) with no cosmological constant : @xmath36 , @xmath37 , with @xmath38 km s@xmath39/mpc . the mass accretion rate decreases exponentially with the characteristic e - folding time scale @xmath33 ( e.g. yi 1996 ) . the remarkably synchronous birth and evolution of qsos could indirectly support the global mass accretion evolution ( e.g. small & blandford 1992 , turner 1991 , and references therein for related discussions ) assumed in the present study ( yi 1996 and references therein ) . as @xmath40 decreases during the course of the cosmological evolution ( in part due to decreasing @xmath10 and more importantly due to the growth of the bh mass ) , the luminosity and spectral state changes as described above . ( iii ) we then construct the differential lfs in various energy bands at different @xmath6 s . one of our major goals is to derive and compare lfs at various energy bands . the lfs are shown in the redshift range , @xmath41 , and the luminosity range , @xmath42erg / s . the lf and its evolution contains several free parameters : a slope of initial bh mass function , bh mass range , @xmath33 , and initial @xmath43 for a given cosmological model . the best fit values for these parameters are chosen for qualitative and quantitative comparisons . the evolutionary timescale @xmath33 controls the overall evolution . the evolution of the lfs in different energy bands is shown in figure 1 . \(i ) in the observable luminosity range ( @xmath44erg / s ) , the qso lfs show an apparent break from a simple power - law ( reflecting the initial bh mass function ) . that is , even for our simple single power - law initial mass function ( and hence initial lf ) , the lfs at late times show that the steep power law slope at high luminosity end turns over to a much flatter slope at low luminosity end , which is observed in the qso samples ( e.g. boyle et al . 1993 , jones et al . this particular feature is surprisingly in good agreement with the trend of the observed lfs ( e.g. the combined rosat and emss samples by jones et al . ( 1997 ) and rosat samples by miyaji et al . our evolution model shows that lfs may appear to deviate from the pure luminosity evolution ( in which only bolometric luminosity is considered ) even when the qso evolution is essentially driven by the luminosity evolution . this effect is obviously caused by the spectral changes accompanying the luminosity changes . we point out that this particular result implies there is no strong evidence for cosmological evolution of the space density of low - luminosity agn in contradiction to miyaji et al . ( 1998 ) . \(ii ) the luminosity functions in x - ray bands are differently affected by the spectral changes caused by the transition of accretion flows . most notably , the most significant spectral change occurs when the qso spectrum changes from hs ( thin disk , @xmath21 ) to ls ( adaf , @xmath45 ) , which to the lowest order induces a sudden luminosity decrease . however , since adafs are relatively hard x - ray bright ( yi & boughn 1998 , 1999 ) , qsos hard x - ray lfs show a trend very different from other energy bands as shown in figure 1 ( also see below ) . the qsos undergo an accretion flow transition at a critical redshift @xmath46 , which corresponds to an epoch where @xmath47 ( narayan & yi 1995 ) as discussed above . although the exact value of @xmath48 depends on various model parameters ( e.g. yi 1996 ) , it is plausible to identify @xmath48 with the observed sudden decline of bright qsos . for instance , @xmath49yr could drive much faster evolution than the observed one . our model @xmath48 is comfortably close to the observed break near @xmath50 ( hewett et al . 1993 , page et al . 1996 ) . \(iii ) optical / uv ( @xmath51 ) lfs evolve much faster than x - ray lfs at low @xmath6 s . at high @xmath52 , however , optical / uv luminosity evolution is very slow , which is in excellent agreement with the observed trend . there also exists a brief period of evolution during which the lf slightly moves to higher luminosity range although such a phenomenon is hard to detect observationally . in the case of soft x - ray lf , the slow evolution is also seen at @xmath52 . this is also in agreement with the results of page et al . ( 1996 ) . \(iv ) at hard x - ray energies ( @xmath53kev ) , a dramatic turnaround in the lf evolution is identified at @xmath54 . that is , the hard x - ray lf as a whole gradually shifts to lower luminosities until @xmath55 is reached . below @xmath48 , the relatively x - ray bright ( especially in hard x - rays ) adafs drive an increasing number of qsos into higher hard x - ray luminosities . this results in reversal in the direction of the hard x - ray lf evolution as shown in figure 1(d ) . this rather surprising result is explained by the fact that qsos have undergone an accretion flow transition from a thin disk ( hs ) to an adaf ( ls ) as @xmath56 decreases . observationally , neither transition in hard x - ray luminosity nor luminosity evolution itself has been clearly seen . however , once @xmath57kev x - ray lfs become available ( e.g. chandra ) , this predicted trend should be testable . \(v ) at soft x - ray energies ( @xmath58kev ) , there is no apparent dramatic transition . the soft x - ray luminosity evolves roughly as @xmath59 with @xmath60 , up to @xmath48 . this value is similar to but larger than those obtained by boyle et al . ( 1994 ) and page et al . ( 1996 ) based on the pure ( bolometric ) luminosity evolution model . the discrepancy could be resolved if spectral evolution is taken into account . for a quantitative comparison between model lfs and observed ones , we convert the space density per comoving volume in the rosat agn soft x - ray lf estimated by miyaji et al . ( 1998 ) to qso lf assuming @xmath61 , @xmath62 , and @xmath35 . on the whole , our model and the miyaji et al . ( 1998 ) result agree quite well with each other . we consider this qualitative agreement as a support for the qso luminosity evolution models . the mf of bhs and its evolution can be naturally derived from our evolutionary calculation . as long as the model qso lf agrees with the observed lf , the resulting mf has immediate consequences on the fate of the qso remnants at the present epoch . figure 2(b ) shows the differential mf of bhs which have grown during the qso evolution . the initial mf with a slope of power law , @xmath63 has been assumed . the bh mf suggests that the qso remnants are likely to have bh masses in the range @xmath64 to @xmath65 . on average , remnant bh masses are bigger than the initial masses by a factor @xmath66 . bh remnants in the mass range , @xmath67 are most likely to be found in some present - day galaxies centers . this finding obviously suggests that it is difficult for our model to explain small bhs with masses in the range @xmath68 which are found in nearby , typical spiral @xmath69 ( e.g. peebles 1993 ) galaxies with the comoving space density @xmath70 ( magorrian et al . 1998 ) . in the present single population scenario , the integrated comoving density of very massive qso remnants is @xmath71 , which suggests that they are likely to be found in rare , giant elliptical galaxies . our model inevitably points to massive bhs with masses @xmath72 ( e.g. fabian & canizares 1988 , mahadevan 1997 ) in elliptical galaxies as qso remnants . it is possible that smaller mass black holes in galactic nuclei might have grown without experiencing the qso phase . our model is largely consistent with the observed qso evolution trend , which supports the possibility that the qso evolution is accounted for by a single , long - lived qso population . the evolution of lf shows an apparent transition at a critical redshift caused by the accretion flow transition ( yi 1996 ) . as far as we know , there is no other convincing explanation for this apparent break in the qso evolution . the overall agreement between the observed lf evolution and our model implies that spectral states of qsos may indeed be similar to those of gbhcs . the accretion flow has undergone a transition as @xmath14 has declined from @xmath73 to @xmath74 while the initial bh masses have evolved by a factor @xmath66 . the mf of bhs indicates that qso remnants are likely to be found as massive bhs with masses in the range @xmath75 , which are likely to be found as massive bhs of @xmath72 residing in very rare elliptical galaxies . the distinguishable evolutionary signature at hard x - ray energies is significant and also in principle testable . future confirmation or rejection of our prediction by hard x - ray luminosity functions ( e.g. by chandra ) could provide a crucial piece of information . it remains to be seen whether an alternative scenario , in which qso evolution is composed of many short - lived populations , is also affected by the luminosity - spectrum correlation we have looked into . the qso spectral evolution could result in some interesting consequences for the x - ray background although the detailed analyses require rather sophisticated spectral calculations ( e.g. yi & boughn 1998 and references therein ) . jy acknowledges the partial support from krf 1998 - 015-d00129 and 1998 - 001-d00365 . iy acknowledges some useful discussions with r. narayan on spectral states during the early stage of this work and support from krf 1998 - 001-d00365 , most project 98-n6 - 01 - 01-a-06 , and ewha faculty reserach fund 1998 . iy thanks suam foundation for its support during the early stage of this work performed at ias . boyle , b. j. , griffiths , r. e. , shanks , t. , stewart , g. c. , & georgrantopoulos , i. 1993 , mnras , 260 , 49 choi , y. , yang , j. , & yi , i. 1999 , jkps , in press esin , a. a. et al . 1998 , apj , 505 , 854 fabian , a. c. & canizares , c. r. 1988 , nature , 333 , 829 frank , j. , king , a. r. , & raine , d. 1992 , accretion power in astrophysics , cambridge : cambridge university press haehnelt , m. g. , natarajan , p. & rees , m. j. 1997 , mnras in press ( astro - ph/9712259 ) haehnelt , m. g. & rees , m. j. 1993 , mnras , 263 , 168 hewett , p. c. , foltz , c. b. , & chaffee , f. h. 1993 , apj , 406 , l43 jones , l. r. et al . 1997 , mnras , 285 , 547 mcclintock , j. e. 1998 , preprint ( astro - ph/9802080 ) magorrian , j. et al . 1998 , aj , 115 , 2285 mahadevan , r. 1997 , apj , 477 , 585 mathez , g. 1976 , a&a , 53 , 15 mathez , g. , waerbeke , v. , mellier , y. , bonnet , h. , & lachieze - rey , m. 1996 , a & a , 316 , 19 mirabel , i. f. & rodriguez , l. f. 1998 , nature , 392 , 673 miyaji , t. , hasinger , g. & schmidt , m. 1998 , preprint ( astro - ph/9809398 ) narayan , r. , mahadevan , r. , & quataert , e. 1998 , preprint ( astro - ph/9803141 ) narayan , r. & yi , i. 1995 , apj , 452 , 710 page , m. j. et al . 1996 , mnras , 281 , 579 peebles , p. j. e. 1993 , principles of physical cosmology , princeton : princeton university press peterson , b. m. 1997 , an introduction to active galactic nuclei , cambridge : cambridge university press rutledge , r. e. et al . 1998 , preprint ( astro - ph/9807235 ) salucci , p. , szuszkiewicz , e. , monaco , p. , & danese , l. 1998 , mnras , submitted ( astro - ph/9811102 ) small , t. a. & blandford , r. d. 1992 , mnras , 259 , 725 szuszkiewicz , e. , malkan , m. a. , & abramowicz , m. a. 1996 , apj , 458 , 474 turner , e. l. 1991 , aj , 101 , 5 yi , i. 1996 , apj , 473 , 645 yi , i. & boughn , s. p. 1998 , apj , 499 , 198 yi , i. & boughn , s. p. 1999 , apj , in press zdziarski , a. a. 1998 , mnras , 296 , l51	we apply the observed optical / x - ray spectral states of the galactic black hole candidates ( gbhcs ) to the cosmological qso luminosity evolution under the assumptions that qsos and gbhcs are powered by similar accretion processes and that their emission mechanisms are also similar . the qso luminosity function ( lf ) evolution in various energy bands is strongly affected by the spectral evolution which is tightly correlated with the luminosity evolution . we generate a random sample of qsos born nearly synchronously by allowing the qsos to have redshifts in a narrow range around an initial high redshift , black hole masses according to a power - law , and mass accretion rates near eddington rates . the qsos evolve as a single long - lived population on the cosmological time scale . the pure luminosity evolution results in distinct luminosity evolution features due to the strong spectral evolution . most notably , different energy bands ( optical / uv , soft x - ray , and hard x - ray ) show different evolutionary trends and the hard x - ray lf in particular shows an apparent reversal of the luminosity evolution ( from decreasing to increasing luminosity ) at low redshifts , which is not seen in the conventional pure luminosity evolution scenario without spectral evolution . the resulting mass function of black holes ( bhs ) , which is qualitatively consistent with the observed qso lf evolution , shows that qso remnants are likely to be found as bhs with masses in the range @xmath0 . the long - lived single population of qsos are expected to leave their remnants as supermassive bhs residing in rare , giant elliptical galaxies .
let @xmath0 denote the space of _ square _ @xmath1 matrices with real - valued coefficients , and @xmath2 the matrix vector space of _ symmetric _ matrices . a matrix @xmath3 is said _ symmetric positive definite _ @xcite ( spd , denoted by @xmath4 ) iff . @xmath5 and only _ symmetric positive semi - definite_. ] ( spsd , denoted by @xmath6 ) when we relax the strict inequality ( @xmath7 ) . let @xmath8 denote the space of positive semi - definite matrices , and @xmath9 denote the space of positive definite matrices . a matrix @xmath10 is defined by @xmath11 real coefficients , and so is a spd or a spsd matrix . although @xmath12 is a _ vector space _ , the spsd matrix space does not have the vector space structure but is rather an abstract _ pointed convex cone _ with _ apex _ the zero matrix @xmath13 since @xmath14 . symmetric matrices can be _ partially _ ordered using the _ lwner ordering _ : @xmath15 and @xmath16 . when @xmath17 , matrix @xmath18 is said to _ dominate _ matrix @xmath19 , or equivalently that matrix @xmath19 is dominated by matrix @xmath18 . note that the difference of two spsd matrices may not be a spsd matrix . and @xmath20 then @xmath21 and @xmath22 . ] a non - spsd symmetric matrix @xmath23 can be dominated by a spsd matrix @xmath18 when @xmath24 . is dominated by @xmath25 ( by taking the absolute values of the eigenvalues of @xmath23 ) . ] the _ supremum _ operator is defined on @xmath26 symmetric matrices @xmath27 ( not necessarily spsds ) as follows : @xmath28 , x\succeq s_i \},\ ] ] where @xmath29=\{1 , ... , n\}$ ] . this matrix @xmath30 is indeed the `` smallest '' , meaning the _ tightest upper bound _ , since by definition there does not exist another symmetric matrix @xmath31 dominating all the @xmath32 s and dominated by @xmath33 . trivially , when there exists a matrix @xmath34 that dominates all others of a set @xmath27 , then the supremum of that set is matrix @xmath34 similarly , we define the _ minimal / infimum matrix _ @xmath35 as the tightest lower bound . since matrix inversion reverses the lwner ordering ( @xmath36 ) , we link those extremal supremum / infimum matrices when considering sets of invertible symmetric matrices as follows : @xmath37 . extremal matrices are _ rotational invariant _ @xmath38 , where @xmath39 is any orthogonal matrix ( @xmath40 ) . this property is important in dt - mri processing that should be invariant to the chosen reference frame . computing lwner extremal matrices are useful in many applications : for example , in matrix - valued imaging @xcite ( morphological operations , filtering , denoising or image pyramid representations ) , in formal software verification @xcite , in statistical inference with domain constraints @xcite , in structure tensor of computer vision @xcite ( frstner - like operators ) , etc . this letter is organized as follows : section [ sec : covering ] explains how to transform the extremal matrix problem into an equivalent geometric minimum enclosing ball of balls . section [ sec : mebb ] presents a fast iterative approximation algorithm that scales well in high - dimensions . section [ sec : concl ] concludes by hinting at further perspectives . we build on top of @xcite to prove that solving the @xmath41-dimensional lwner maximal matrix amounts to either find ( 1 ) the minimal covering lwner matrix cone ( wrt . set containment @xmath42 ) of a corresponding sets of @xmath43-dimensional cones ( with @xmath11 ) , or ( 2 ) the minimal enclosing ball of a set of corresponding @xmath44-dimensional `` matrix balls '' that we cast into a geometric _ vector _ ball covering problem for amenable computations . let @xmath45 denote the _ lwner ordering cone _ , and @xmath46 the reverted and translated _ dominance cone _ ( termed the penumbra cone in @xcite ) with apex @xmath32 embedded in the space of symmetric matrices that represents all the symmetric matrices dominated by @xmath32 : @xmath47 , where @xmath48 denotes the _ minkowski set subtraction operator _ : @xmath49 ( hence , @xmath50 ) . a matrix @xmath23 dominates @xmath27 iff . @xmath51 , \call(s_i)\subseteq \call(s)$ ] . in plain words , @xmath23 dominates a set of matrices iff . its associated dominance cone @xmath52 covers all the dominance cones @xmath46 for @xmath53 $ ] . the dominance cones are `` abstract '' cones defined in the @xmath1 symmetric matrix space that can be `` visualized '' as equivalent _ vector _ cones in dimension @xmath11 using _ half - vectorization _ : for a symmetric matrix @xmath23 , we stack the elements of the lower - triangular matrix part of @xmath54 $ ] ( with @xmath55 ) : @xmath56^\top\in \bbr^(\frac{d(d+1)}{2})$ ] . note that this is not the unique way to half - vectorize symmetric matrices but it is enough for _ geometric containment _ purposes . later , we shall enforce that the @xmath57-norm of vectors @xmath58 matches the frbenius matrix norm @xmath59 . let @xmath60 denotes the vectorized matrix lwner ordering cone : @xmath61 , and @xmath62 denote the _ vector dominance cone _ : @xmath63 . next , we further transform this minimum @xmath43-dimensional matrix / vector cone covering problems as equivalent _ minimum enclosing ball _ ( meb ) problems of @xmath44-dimensional matrix / vector balls . a _ basis _ @xmath64 of a convex cone @xmath65 anchored at the origin @xmath66 is a convex subset @xmath67 so that @xmath68 there exists a _ unique decomposition _ : @xmath69 with @xmath70 and @xmath71 . for example , @xmath72 is a basis of the lwner cone @xmath73 . informally speaking , a basis of a cone can be interpreted as a _ compact _ cross - section of the cone . the lwner cone @xmath74 is a smooth convex cone with its interior @xmath75 denoting the space of positive definite matrices @xmath76 ( full rank matrices ) , and its border @xmath77 the _ rank - deficient _ symmetric positive semi - definite matrices ( with apex the zero matrix @xmath66 of rank @xmath66 ) . a point @xmath78 is an _ extreme element _ of a convex set @xmath23 iff . @xmath79 remains convex . it follows from minkowski theorem that every compact convex set @xmath80 in a finite - dimensional vector space can be reconstructed as convex combinations of its extreme points @xmath81 : that is , the compact convex set is the closed convex hull of its extreme points . associated with matrix @xmath23 has apex @xmath23 and base @xmath82 , a ball centered at matrix @xmath83 of radius @xmath84 . the cone @xmath52 has an equivalent representation @xmath85 provided that @xmath86 . [ fig : coneprojection ] ] a face @xmath87 of a closed cone @xmath65 is a subcone such that @xmath88 . the @xmath89-dimensional faces are the _ extremal rays _ of the cone . the basis of the lwner ordering cone is @xcite @xmath90 . other rank - deficient or full rank matrices can be constructed by convex combinations of these rank-@xmath89 matrices , the extremal rays . for any square matrix @xmath91 $ ] , the _ trace operator _ is defined by @xmath92 , the sum of the diagonal elements of the matrix . the trace also amounts to the sum of the eigenvalues @xmath93 of matrix @xmath94 : @xmath95 . the basis @xmath96 of a dominance cone @xmath46 is @xmath97 . note that all the basis of the dominance cones lie in the _ subspace _ @xmath98 of symmetric matrices with zero trace . let @xmath99 denote the _ matrix inner product _ and @xmath100 the matrix _ frbenius norm_. two matrices @xmath94 and @xmath101 are orthogonal ( or perpendicular ) iff . it can be checked that the identity matrix @xmath103 is perpendicular to any zero - trace matrix @xmath94 since @xmath104 . the center of the ball basis of the dominance cone @xmath105 is obtained as the _ orthogonal projection _ of @xmath23 onto the zero - trace subspace @xmath98 : @xmath106 . the dominance cone basis is a _ matrix ball _ since for any rank-@xmath89 matrix @xmath107 with @xmath108 ( an extreme point ) , we have the radius : @xmath109 that is non - negative since we assumed that @xmath86 . reciprocally , to a basis ball @xmath110 , we can associate the apex of its corresponding dominance cone @xmath111 : @xmath112 . figure [ fig : coneprojection ] illustrates the notations and the representation of a cone by its corresponding basis and apex . thus we associate to each dominance cone @xmath46 its corresponding ball basis @xmath113 on the subspace @xmath98 of zero trace matrices : @xmath114 , @xmath115 . we have the following containment relationships : @xmath116 and @xmath117 finally , we transform this minimum enclosing _ matrix _ ball problem into a minimum enclosing _ vector _ ball problem using a half - vectorization that preserves the notion of distances , _ i.e. _ , using an isomorphism between the space of symmetric matrices and the space of half - vectorized matrices . the @xmath57-norm of the vectorized matrix should match the matrix frbenius norm : @xmath118 . since @xmath119 , it follows that @xmath120^\top \in\bbr^{\frac{d(d+1)}{2}}$ ] . we can convert back a vector @xmath121 into a corresponding symmetric matrix . since we have considered all dominance cones with basis rooted on @xmath122 in order to compute the ball basis as orthogonal projections , we need to _ pre - process _ the symmetric matrices to ensure that property as follows : let @xmath123 denote the minimal trace of the input set of symmetric matrices @xmath27 , and define @xmath124 for @xmath53 $ ] where @xmath103 denotes the identity matrix . recall that @xmath125 . by construction , the transformed input set satisfies @xmath126 $ ] . furthermore , observe that @xmath127 iff . @xmath128 where @xmath129 , so that @xmath130 . as a side note , let us point out that the reverse basis - sphere - to - cone mapping has been used to compute the convex hull of @xmath41-dimensional spheres ( convex homothets ) from the convex hull of @xmath131-dimensional equivalent points @xcite . finally , let us notice that there are severals ways to majorize / minorize matrices : for example , once can seek extremal matrices that are invariant up to an _ invertible transformation _ @xcite , a stronger requirement than the invariance by orthogonal transformation . in the latter case , it amounts to geometrically compute the minimum volume enclosing ellipsoid of ellipsoids ( mveee ) @xcite . first , let us summarize the algorithm for computing the lwner maximal matrix of a set of @xmath26 symmetric matrices @xmath27 as follows : 1 . normalize matrices so that they have all non - negative traces : @xmath134 2 . compute the vector ball representations of the dominance cones : @xmath135 with @xmath136 and @xmath137 3 . compute the small(est ) enclosing ball @xmath138 of basis balls ( either exactly or an approximation ) : @xmath139 4 . convert back the small(est ) enclosing ball @xmath140 to the dominance cone , and recover its apex @xmath141 : @xmath142 5 . adjust back the matrix trace : @xmath143 computing _ exactly _ the extremal lwner matrices suffer from the _ curse of dimensionality _ of computing mebs @xcite . in @xcite , burgeth et al . proceed by discretizing the basis spheres by sampling^\top$ ] for @xmath144 . in 3d , we use spherical coordinates @xmath145^\top$ ] for @xmath144 and @xmath146 . ] the extreme x points @xmath147 for @xmath108 . this yields an approximation term , requires more computation , and even worse the method does not scale @xcite in high - dimensions . thus in order to handle high - dimensional matrices met in software formal verification @xcite or in computer vision ( structure tensor @xcite ) , we consider @xmath132-approximation of the extremal lwner matrices . the notion of tightness of approximation of @xmath133 ( the epsilon ) is imported straightforwardly from the definition of the tightness of the geometric covering problems . a @xmath132-approximation @xmath148 of @xmath33 is a matrix @xmath149 such that : @xmath150 . it follows from eq . [ eq : matrad ] that a @xmath132-approximation satisfies @xmath151 . we present a fast guaranteed approximation algorithm for approximating the minimum enclosing ball of a set of balls ( or more generally , for sets of compact geometric objects ) . [ cols= " < , < " , ] our novel extremal matrix approximation method allows one to leverage further related results related to core - sets @xcite for dealing with high - dimensional extremal matrices . for example , we may consider clustering psd matrices with respect to lwner order and use the @xmath152-center clustering technique with guaranteed approximation @xcite . a java code of our method is available for reproducible research . this work was carried out during the matrix information geometry ( mig ) workshop @xcite , organized at cole polytechnique , france in february 2011 ( https://www.sonycsl.co.jp/person/nielsen/infogeo/mig/ ) . frank nielsen dedicates this work to the memory of his late father gudmund liebach nielsen who passed away during the last day of the workshop . x. allamigeon , s. gaubert , e. goubault , s. putot , and n. stott , `` a scalable algebraic method to infer quadratic invariants of switched systems , '' in _ embedded software ( emsoft ) , 2015 international conference on _ , oct 2015 , pp . 7584 . j. a. calvin and r. l. dykstra , `` maximum likelihood estimation of a set of covariance matrices under lwner order restrictions with applications to balanced multivariate variance components models , '' _ the annals of statistics _ , pp . 850869 , 1991 . boissonnat , a. crzo , o. devillers , j. duquesne , and m. yvinec , `` an algorithm for constructing the convex hull of a set of spheres in dimension @xmath41 , '' _ computational geometry _ , vol . 6 , no . 2 , pp . 123130 , 1996 . boissonnat and m. i. karavelas , `` on the combinatorial complexity of euclidean voronoi cells and convex hulls of @xmath41-dimensional spheres , '' in _ proceedings of the fourteenth annual acm - siam symposium on discrete algorithms_.1em plus 0.5em minus 0.4emsociety for industrial and applied mathematics , 2003 , pp . 305312 . s. jambawalikar and p. kumar , `` a note on approximate minimum volume enclosing ellipsoid of ellipsoids , '' in _ computational sciences and its applications , 2008 . international conference on_.1em plus 0.5em minus 0.4emieee , 2008 , pp . 478487 . k. fischer and b. grtner , `` the smallest enclosing ball of balls : combinatorial structure and algorithms , '' _ international journal of computational geometry & applications _ , vol . 04n05 , pp . 341378 , 2004 . , `` smaller core - sets for balls , '' in _ proceedings of the fourteenth annual acm - siam symposium on discrete algorithms _ , ser . soda 03.1em plus 0.5em minus 0.4emphiladelphia , pa , usa : society for industrial and applied mathematics , 2003 , pp . 801802 . [ online ] . available : http://dl.acm.org/citation.cfm?id=644108.644240 j. mihelic and b. robic , `` approximation algorithms for the @xmath152-center problem : an experimental evaluation , '' in _ selected papers of the international conference on operations research ( sor 2002)_.1em plus 0.5em minus 0.4emspringer , 2003 , p. 371 .	matrix data sets are common nowadays like in biomedical imaging where the diffusion tensor magnetic resonance imaging ( dt - mri ) modality produces data sets of 3d symmetric positive definite matrices anchored at voxel positions capturing the anisotropic diffusion properties of water molecules in biological tissues . the space of symmetric matrices can be partially ordered using the lwner ordering , and computing extremal matrices dominating a given set of matrices is a basic primitive used in matrix - valued signal processing . in this letter , we design a fast and easy - to - implement iterative algorithm to approximate arbitrarily finely these extremal matrices . finally , we discuss on extensions to matrix clustering . * keywords * : positive semi - definite matrices , lwner ordering cone , extremal matrices , geometric covering problems , core - sets , clustering .
double - peaked emitters are active galactic nuclei ( agns ) whose broad balmer - line profiles have a characteristic `` double - humped '' shape , with one peak blueward of the rest - frame wavelength of the line and another one redward of it . the best explanation for the origin of the double - peaked shape involves emission from a region in the agn accretion disk . since the first examples of double - peaked emission lines in agns were published in the 1980s ( e.g. , * ? ? ? * ; * ? ? ? * ) , more than 100 double - peaked emitters have been found . some of these objects have also been studied in the x - ray band ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , but most of these studies focused on individual objects or on serendipitous x - ray detections of limited quality . ccccc [email protected] & @xmath3 & 0.260 & 15200 & @xmath4 + [email protected] & @xmath6 & 0.199 & 16000 & @xmath7 + [email protected] & @xmath8 & 0.348 & 15400 & @xmath9 + [email protected] & @xmath10 & 0.309 & 20100 & @xmath4 + [email protected] & @xmath11 & 0.625 & 14000 & @xmath12 + [ tabs ] ccccccccc j0132@xmath10952 & 6770 & acis-7 & 2006 - 06 - 24 & 8151 & 727 & @xmath13 & @xmath14 & @xmath15 + j1004@xmath54801 & 6771 & acis-235678 & 2006 - 01 - 31 & 8628 & 359 & @xmath16 & @xmath17 & @xmath18 + j1238@xmath55325 & 6773 & acis-7 & 2006 - 05 - 31 & 4061 & 1133 & @xmath19 & @xmath20 & @xmath21 + j1407@xmath50235 & 6772 & acis-235678 & 2005 - 12 - 07 & 5471 & 1348 * & @xmath22 * & @xmath23 & @xmath24 + j2125@xmath10813 & 6774 & acis-7 & 2006 - 04 - 02 & 4110 & 155 & @xmath25 & @xmath23 & @xmath26 + [ tab1 ] the small sample of double - peaked emitters studied in detail in the x - ray band to date consists of predominantly low - luminosity radio - loud=0.4(i - t)>1 $ ] , where @xmath27 is the sloan digital sky survey @xmath27-band magnitude and @xmath28 is the 20 cm ab radio magnitude . ] ( rl ) objects broad line radio galaxies ( blrgs , eracleous et al . 1994 , 2003 , hereafter e03 ) . based on the prototype disk emitter ( arp 102b ) and the small original blrg sample , a model was developed to explain the existence of double - peaked emission in agns . @xcite proposed that blrgs and other low - accretion - rate , low - luminosity , where @xmath29 is the black hole mass in units of @xmath30@xmath31 , @xmath32 is the accretion efficiency , and @xmath33 is the eddington luminosity . efficient accretors can convert up to @xmath34 ( for a maximally rotating black hole ; * ? ? ? * ) of the gravitational energy into radiation . ] agns which do not accrete through a standard thin disk could exhibit low - ionization lines with double - peaked profiles . the line profiles are produced as the outer portions of the disk are illuminated by an x - ray - emitting elevated structure in the center a structure that is characteristic of radiatively inefficient accretion flows ( riafs ; e.g , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) and replaces the standard thin disk . the illumination provided by the central elevated structure in agns with riafs is necessary , but not sufficient , for the production of double - peaked lines of the observed strength : all blrgs with double - peaked lines require external illumination ( since the viscous power available locally in the disk is insufficient to explain the line strength ) , but not all blrgs or other riaf agns have double - peaked lines . the association of rl agns with riafs could explain the higher frequency of double - peaked emitters observed among blrgs ( 20% ; e03 ) , compared to 3% of low - redshift sloan digital sky survey ( sdss ) agns . recent work by @xcite , however , has shown that not all blrg double - peaked emitters with low optical luminosities are inefficient emitters : pictor a , for example , emits a bolometric luminosity of more than 8% of the eddington luminosity , which is too high to support a riaf ( riaf solutions exist for @xmath35 , for a viscosity parameter @xmath36 , and the radiated energy is further reduced by the accretion efficiency ; * ? ? ? the systematic search for agns with double - peaked balmer lines in the sdss revealed that as many as 3% of @xmath37 objects are double - peaked emitters , regardless of their radio loudness ( strateva et al . 2003 ; hereafter , s03 ) . moreover , a handful of luminous higher - redshift double - peaked emitters were discovered serendipitously based on their h@xmath38 and line profiles . the expansion of the parameter space of the double - peaked emission - line agns to include radio - quiet ( rq , @xmath39 ) and higher luminosity objects , as well as the discovery of low optical luminosity objects like pictor a that are efficient accretors , require a critical evaluation of the existing models of illumination and the conditions necessary for the existence of double - peaked emission lines . x - ray observations of the new extended sample are particularly important , since they allow us to obtain more accurate estimates of the bolometric luminosities of the sources and to gauge the need for extra illumination of the accretion disk . double - peaked emitters which are efficient accretors likely require an alternative mode of illumination . the development of new scenarios can be guided by observations in the x - ray band . basic x - ray absorption constraints , for example , provide sensitive probes of the nuclear conditions : ( 1 ) they allow us to see any intervening material right up to the agn core , and ( 2 ) they are sensitive to molecular , atomic , or ionized gas and dust . the absorption constraints could provide sensitive tests of specific models proposed to explain the expanded population of double - peaked emitters . a recent model by @xcite , for example , aims to explain the disk illumination necessary for the existence of the double - peaked lines with the presence of jets in rl , and slower , less - collimated outflows in rq double - peaked emitters . the predictions of this model can be tested directly by obtaining x - ray spectroscopy of double - peaked emitters and looking for signs of these outflows in absorption . in addition to testing specific models of illumination , comparison of the x - ray emission of double - peaked emitters to that of typical agns is essential for uncovering the necessary and sufficient conditions for the dominance of accretion - disk balmer - line emission in this type of agn . such findings , which will also elucidate the physical conditions in agns with the broadest observed balmer lines , will have implications for the ( still uncertain ) structure of the broad line region ( blr ) in agns in general . it is conceivable that the accretion disk is part of the blr in all agns , but its contribution dominates the balmer - line emission only in the small subsample of agns with the broadest lines ( see also * ? ? ? * for some tantalizing ideas about the structure of the blr ) . in this paper we present the first results of an exploratory campaign to map the x - ray properties of a representative sample of agns with the broadest double - peaked balmer - lines . using the _ chandra _ x - ray observatory we obtained high - quality x - ray images and spectra for five sources , complemented by contemporaneous optical spectra from the ground . our five targets span the redshift and optical luminosity plane of the sdss double - peaked sample presented in s03 . the paper is organized as follows : in [ sample ] we present the current sample and its selection criteria , followed by the analysis of the x - ray and optical observations of each individual source in [ analysis ] and [ optspec ] , and the discussion and conclusions in [ conclusions ] . throughout this paper we use the @xcite concordance cosmology and quote all flux and luminosity errors as 90% confidence limits ; all other errors quoted are 1@xmath40 gaussian equivalent . the sample of five sdss double - peaked emitters observed with _ chandra _ was selected to span the part of the luminosity - redshift - rl space which was not covered by the initial blrg sample . figure [ lzplot ] shows the positions of the five targets on the redshift vs. 2500 monochromatic - luminosity plane , using the rest - frame flux densities measured during the sdss epoch . the full sdss sample of double - peaked emitters from s03 is also shown for comparison , as well as the sample of sdss double - peaked emitters serendipitously detected with _ rosat _ and presented in strateva et al . ( 2006 ; hereafter , s06 ) . sample is representative of the full sdss double - peaked emitter sample with regard to optical luminosity , redshift , and radio - loudness ( to the extent that most sdss double - peaked emitters are rq ) . the prototype disk emitter , arp102b , is fainter at 2500 than most sdss double - peaked emitters , as is the optically - brighter , efficient accretor , pictora . the initial blrg sample of @xcite has been extended with serendipitously discovered double - peaked emitters in a handful of low - ionization nuclear emission - line region galaxies ( liners ) and quasars . a small number of sources from this original sample of double - peaked emitters have serendipitous or pointed x - ray coverage ; in all but 9 cases only soft - band coverage by _ rosat _ is available , with too few counts for spectral fits . among the 9 double - peaked emitters from the original sample with _ xmm - newton _ and _ chandra _ observations , 4 are liners ( ngc1097 , m81 , ngc4203 , ngc4579 ; s06 and references therein ) and the remaining 5 are blrgs ( arp102b , 3c390.3 , pictora , pks0921@xmath1213 , and 3c382 ; s06 , grandi et al . 2006 ) . the x - ray data for these 9 double - peaked emitters are summarized in @xcite . these are predominantly low uv luminosity and redshift ( @xmath41 ) objects , as shown in figure [ lzplot ] . the new _ chandra _ targets were selected to be predominantly rq ( four out of the five ) , higher luminosity and redshift ( ) counterparts to the original - sample sources studied in the hard x - ray band , forming an ideal comparison sample of double - peaked emitters . the selection criteria used only optical data , and the sample is therefore unbiased with respect to the x - ray properties of the selected targets . we present the five targets in table [ tabs ] , listing the sdss @xmath27-band magnitudes , redshifts and radio - loudness , as well as the full - widths at half maximum ( fwhms ) of the h@xmath42 ( or h@xmath38 ) lines . the five double - peaked emitters observed with _ are among the 16 agns with the broadest published balmer lines : . among the parent sample of over 3000 sdss agns from which the sdss double - peaked emitters were selected , virtually all agns with such extremely broad balmer lines showed line profiles indicative of a disk origin . thus , with regard to their balmer line widths , the five _ chandra _ targets are not representative of the full sample of sdss double - peaked emitters . they were selected because the widths of their balmer lines suggest an origin very close to the center of the disk . as such they are the most suitable objects for addressing the question of the energy budget of the line - emitting portion of the disk . by extension , this set of objects can also help us address the important question of why only a small fraction of agns show double - peaked balmer line profiles , while the majority of the profiles are single peaked . table [ tab1 ] summarizes the observations carried out during _ chandra _ cycle 8 . the exposure times were selected to yield enough counts for a basic spectroscopic fit in each case ( @xmath43100 counts ) . three of the targets ( sdssj1004@xmath54801 , sdssj1238@xmath55325 , and sdssj2125@xmath10813 ) had prior serendipitous _ rosat _ detections , while the remaining two have _ rosat _ all sky survey ( rass ) upper limits . we used ciao v.3.4 and xspec v.12.3.0 to analyze the data , starting with the event1 files . we fit all spectra with more than 200 counts in the 0.310kev band using @xmath44-minimization and use the cash statistic ( @xmath45-stat ) for the one object with less than 200 counts . the results for each of the five double - peaked emitters are summarized below , where we also discuss their x - ray variability . sdssj0132@xmath10952 is a rq double - peaked emitter which was observed with _ chandra _ for 8.2ks . the x - ray image of the target is a point source with a point spread function ( psf ) fwhm @xmath46 ( table [ tab1 ] ) . based on the sdss image , our target lies in a cluster at identified by @xcite ; it is situated at a projected distance of @xmath47kpc from the brightest cluster galaxy . the spectrum of sdssj0132@xmath10952 can not be modeled by a simple power law absorbed by neutral hydrogen with a column density equal to the galactic value ( @xmath48 , where @xmath49 stands for the number of degrees of freedom ) . since intrinsic absorption is common among agns , we attempted to model the spectrum by adding an intrinsic absorber . the addition of a neutral intrinsic absorber does not improve the fit significantly ( @xmath50 ) , because the spectrum is too soft ; allowing for partial coverage improves the fit , but the model is only marginally acceptable ( @xmath51 ) . the fit residuals are suggestive of an absorption edge around , and we attempted to fit the spectrum with a ionized absorber model ( _ absori _ in xspec ; * ? ? ? the spectral fit is shown in fig . [ xspec ] . the fit is improved at @xmath52% confidence ( @xmath53 ) ; the unabsorbed spectrum is a power law with @xmath54 , typical for similar x - ray luminosity agns , which have an average photon index of @xmath55 ( e.g. , * ? ? ? the column density of the ionized absorber is constrained to @xmath56@xmath57 with an ionization parameter , with @xmath58 , where @xmath59 denotes the luminosity of the source , @xmath60 is the distance to the source , and @xmath61 is the hydrogen column density @xcite . ] @xmath62ergs@xmath63 cm . the _ absori _ fit parameters are given in table [ tab2 ] and the estimated fluxes and monochromatic luminosities in table [ tab3 ] . we show the @xmath64 contours for two interesting parameters in figure [ xspec ] . since the xspec _ absori _ model is not fully self consistent ( it treats the temperature and ionization parameter as independent parameters , for example ) , we used the _ etable _ xstar model @xcite , as implemented in xspec , to verify the _ absori _ model parameters . the ionization parameter of about a factor of two , with @xmath65 . ] and absorber column density are both consistent : @xmath66@xmath57 and @xmath67ergs@xmath63 cm , but the xstar model constraints are much looser than those obtained with the xspec _ absori _ model . this is a result of the limited signal - to - noise ratio of the current data , for which the _ absori _ model , which has no sophisticated line information , is sufficiently detailed . sdssj0132@xmath10952 falls within rass observation rs931905n00 ( june 1990 ) , but is not detected in 440s . we estimate a count - rate upper limit of @xmath680.011s@xmath63 in the band ( see table [ tab1 ] ) , corresponding to an expected _ chandra _ acis - s count rate of @xmath680.036s@xmath63 , fainter than the observed 0.52kev acis - s count rate of @xmath69s@xmath63 . this suggests that the source is variable , brightening by at least @xmath7060% in the last 17 years ( corresponding to @xmath71years in the source rest - frame ) . ccccccc j0132@xmath10952 & @xmath72 & @xmath73 & @xmath74 & @xmath75 & 37/30 & _ wabs(absorizpow ) _ + & & & @xmath62 & & & + j1004@xmath54801 & @xmath76 & @xmath77 & ... & @xmath78 & 21/32 & _ wabszpow _ + j1238@xmath55325 & @xmath79 & @xmath80 & ... & @xmath81 & 54/46 & _ wabszpow _ + j1407@xmath50235 & @xmath82 & @xmath83 & ... & ... & 49/58 & _ jdpileup(abspower ) _ + j2125@xmath10813 & @xmath84 & @xmath85 & ... & @xmath86 & 123/118 & _ wabszpow _ + [ tab2 ] sdssj1004@xmath54801 is a rq double - peaked emitter that was observed with _ chandra _ for 8.6ks ( table [ tab1 ] ) . the spectrum has 359 counts and can be modeled with a simple power law modified by absorption equal to the galactic value ( @xmath87 ; see table [ tab2 ] and figure [ xspec ] ) . we measure a power - law photon index of @xmath88 , which is rather flat , but within the range observed for rq agns ; the soft - band flux is @xmath89ergs@xmath57s@xmath63 . considering the small number of counts in the spectrum , we can alternatively fix the power - law slope of the spectral fit to the average value observed in similar redshift and luminosity agns , @xmath90 ( e.g. , ? ? ? * ) , and allow for some intrinsic absorption . the model fit including a neutral absorber is acceptable ( @xmath91 ) , but the inferred intrinsic absorbing column density is small , @xmath92@xmath57 . sdssj1004@xmath54801 was detected in the rass in october 1990 , with a 0.52kev count rate of @xmath18s@xmath63 and a hardness ratio , where @xmath93 refers to the _ rosat _ pspc count rate in the hard ( 0.52.0kev ) and b in the soft ( 0.10.4kev ) band . ] of 0.01 ( corresponding to @xmath94 with absorption equal to the galactic value ) . assuming the _ chandra _ power - law model slope and absorption equal to the galactic value , the _ rosat _ soft - band flux was @xmath95ergs@xmath57s@xmath63 . this is significantly brighter than the 2006 _ chandra _ soft - band flux given above . sdssj1238@xmath55325 ( table [ tab1 ] ) is a rl double - peaked emitter which we observed with _ chandra _ using the acis - s 1/8 subarray . the 0.310kev spectrum can be represented by a simple power law with photon index @xmath96 modified by galactic absorption ( see table [ tab2 ] and figure [ xspec ] ) . any intrinsic neutral absorption is constrained to be @xmath97@xmath57 . sdssj1238@xmath55325 was previously detected with _ rosat _ in december 1990 . assuming that the _ chandra _ power - law model modified by galactic absorption applies , the soft - band flux was @xmath98ergs@xmath57s@xmath63 during the _ rosat _ observation , essentially unchanged from the 2006 _ chandra _ observation . sdssj1407@xmath10235 is a rq double - peaked emitter observed by _ chandra _ for 5.5ks ( table [ tab1 ] ) . the source falls within two rass @xmath70390s pointings : rs931738n00 and rs931638n00 observed on 1991 august 13 . the estimated count - rate upper limits , @xmath24s@xmath63 , are consistent for the two observations , corresponding to an expected _ chandra _ acis - s count rate of @xmath99s@xmath63 ( ) . the actual observed count rate ( using the 1/2 subarray ) was @xmath100s@xmath63 ( a brightening by at least a factor of 4 ) and the source is piled up . we used sherpa s _ jdpileup _ model to estimate the pile - up and model the x - ray spectrum . the estimated pile - up fraction is 12% , and the spectrum can be modeled by a simple power law with photon index @xmath101 modified by galactic absorption . the pile - up and absorption corrected fluxes in the hard and soft bands are @xmath102erg@xmath57s@xmath63 and @xmath103erg@xmath57s@xmath63 , respectively ( see table [ tab3 ] ) . sdssj2125@xmath10813 is the highest optical luminosity sdss double - peaked emitter ( , table [ tabs ] ) and one of only three high - redshift ( @xmath104 ) double - peaked emitters discovered serendipitously based on their h@xmath38-line profiles . its inclusion in the current sample allows us to extend substantially the redshift and luminosity coverage of the sample . a simple power - law model modified by galactic absorption in the band results in a rather flat power - law slope of @xmath105 , with large residuals around , corresponding to in the agn rest frame ( @xmath45-stat=123 for 118@xmath49 ; see figure [ xspec ] ) . the inclusion of neutral intrinsic absorption does not improve the fit . the addition of a gaussian emission line centered at 6.7kev ( rest - frame ) with a width of 0.3kev improves the fit ( @xmath45-stat=118 for 117@xmath49 ) , but the small number of counts prevents us from reliably constraining the line parameters . the equivalent width of the gaussian line is @xmath106kev , and if confirmed , this line is consistent with k@xmath42 emission . note that the putative fek@xmath42 line ew is too high relative to the observed hard - band x - ray luminosity and would require unusually high fe abundance ; typical agns with kev luminosities similar to that of sdssj2125@xmath10813 ( @xmath107ergs@xmath63 ) are expected to have ew@xmath108ev @xcite . more complex models including a partial or ionized absorber or an exponentially cut - off power - law model reflected by ionized material in addition to a relativistic fe line provide better fits to the data , but can not be well constrained with only 155 counts . for example , a model including an ionized absorber ( @xmath109ergs@xmath63 cm ) with a large column density ( @xmath110@xmath57 ) provides a statistically acceptable fit , suggesting the slow wind outflow model of @xcite might provide extra illumination to the h@xmath42 emission region . a longer observation of this source is necessary to characterize better its x - ray emitting region . sdssj2125@xmath10813 was detected in the rass in november 1990 , with a flux of @xmath111erg@xmath57s@xmath63 and a hardness ratio of 0.86 , which corresponds to a power - law photon index of @xmath112 ( assuming absorption equal to the galactic value ) . for comparison , the best - fit ( figure [ xspec ] ) model flux in the _ chandra _ observation was @xmath113erg@xmath57s@xmath63 , a soft x - ray band decrease in flux by a factor of @xmath114 over 15.5 years . the observed change in the soft - band flux is significant and robust to changes in the power - law model ( e.g. , assuming intrinsic absorption ) or the x - ray band ( e.g. , restricting the analysis to the band ) used to obtain the flux during the _ chandra _ epoch . ccccccccc j0132@xmath10952 & @xmath115 & @xmath116 & 26.24 & 29.01 & @xmath11.06 & 0.28 & 0.17 & @xmath117 + j1004@xmath54801 & @xmath118 & @xmath119 & 25.45 & 28.22 & @xmath11.06 & 0.17 & 0.19 & @xmath120 + j1238@xmath55325 & @xmath121 & @xmath122 & 26.91 & 29.53 & @xmath11.00 & 0.41 & ... & @xmath123 + j1407@xmath50235 & @xmath124 & @xmath125 & 26.66 & 29.30 & @xmath11.01 & 0.37 & 0.17 & @xmath126 + j2125@xmath10813 & @xmath127 & @xmath128 & 26.77 & 30.23 & @xmath11.33 & 0.18 & 0.20 & @xmath129 + [ tab3 ] cccccccccccc j0132@xmath10952 & 2006 - 08 - 23 & 1800 & 32 & 2 & 340 & 850 & 2400 & 0.1 & 50 & @xmath130 & 0.4 + j1004@xmath54801 & 2006 - 03 - 29 & 2500 & 36 & 3 & 750 & 1470 & 2200 & @xmath131 & @xmath131 & @xmath132 & 1.8 + j1238@xmath55325 & 2006 - 06 - 17 & 900 & 48 & 2 & 400 & 1400 & 1800 & 0.3 & 30 & @xmath133 & 0.3 + j1407@xmath50235 & 2006 - 03 - 05 & 3200 & 20 & 2 & 180 & 210 & 1400 & 0.2 & 58 & @xmath134 & 0.4 + j2125@xmath10813 & 2006 - 06 - 23 & 1100 & 15 & 2 & 150 & 340 & 2000 & @xmath131 & @xmath131 & @xmath135 & 0.1 + [ tabfits ] cccccccc j0132@xmath10952 & 2001 - 08 - 26 & 20 & 5 & @xmath136 & 43004600 & 1.15 & 0.9 + ... & 2001 - 09 - 26 & 20 & 5 & @xmath136 & 43004600 & 0.95 & 0.85 + j1004@xmath54801 & 2002 - 03 - 06 & 70 & 4 & @xmath137 & 45007200 & 0.7 & 0.7 + j1238@xmath55325 & 2002 - 04 - 15 & @xmath13810 & 4.2 & @xmath139 & 40006200 & 0.53 & 0.63 + j1407@xmath50235 & 2001 - 03 - 25 & 50 & 5 & @xmath140 & 40004500 & 3 & 2.4 + j2125@xmath10813 & 2001 - 09 - 01 & @xmath13810 & 4.7 & @xmath141 & 33004200 & 0.45 & 0.4 + [ tabvar ] we re - observed all five double - peaked emitters with the marcario low - resolution spectrograph ( lrs ) of the hobby - eberly telescope ( het ) in the optical band , within a few months of the _ chandra _ observations . double - peaked emitters are known to vary on timescales of months to years ( e.g. , * ? ? ? * ; * ? ? ? * ) , and thus contemporaneous observations are necessary to ensure proper comparison of the optical and the x - ray monochromatic luminosities . in addition to flux variations , the line profile shapes also vary on timescales of months to years and a contemporaneous spectrum allows us to estimate the emission - region size during the _ chandra _ epoch . we used the g1 grism and 1@xmath142 slit to cover the wavelength range . our spectra have a spectral resolution of 8.4 , corresponding to a velocity resolution of 350kms@xmath63 at 7200 . the data were reduced in a standard manner , as described in @xcite , for example . we used the [ ] line flux measured from the sdss spectra to determine the flux calibration of the het spectra . the [ ] line emission is a good tracer of nuclear activity in agns ( e.g. * ? ? ? * and references therein ) and provides a suitable and relatively robust measure ( we estimate @xmath143% uncertainty ) of the spectral normalization . we isolated the h@xmath42-line profiles ( or the h@xmath38 profile in the case of sdss j2125@xmath10813 ) in the het / lrs spectra by subtracting a local power - law continuum . we fitted the resulting line profiles with the relativistic keplerian disk models of @xcite and @xcite in order to determine the parameters that characterize the line - emitting region of the disk . for these fits we ignore the regions around the narrow h@xmath42 line as well as the s , n , and o forbidden lines . we used the sdss line - profile fits , which include modeling of the narrow lines , to verify that the results obtained from the het spectra are robust . the line - profile fits allow us to estimate the following parameters : ( 1 ) the inner and outer radii of the emission region , @xmath144 and @xmath145 , in units of @xmath146 , where @xmath147 is the mass of the black hole , ( 2 ) the inclination of the axis of the disk to the line of sight , @xmath27 , ( 3 ) the emissivity power - law index , @xmath148 , with @xmath149 , where @xmath150 represents a point source high above the disk on the disk axis , while @xmath151 represents a constant - density illuminator above the emission region ( see * ? ? ? ] under the convention that @xmath152 , and ( 4 ) the broadening parameter , @xmath40 ( in km s@xmath63 ) . the line profiles of the five objects of our sample are shown in figure [ ospec ] , together with the best - fitting disk models . table [ tabfits ] lists the model parameters . two of the objects are well fit by an axisymmetric disk model , and all disk parameter estimates are fairly robust ( in the case of sdssj2125@xmath10813 , for example , the inner radius of the emission region is constrained to lie between @xmath153 ) . the remaining three objects require non - axisymmetric disk models , and we model them with an elliptical disk . the elliptical disk model @xcite is the simplest non - axisymmetric accretion - disk model which we use to obtain estimates of the inner and outer radii of the h@xmath42-line emission region ; besides the five disk parameters listed above , this model requires fitting for the disk ellipticity , @xmath154 , and the angle of the semi - major axis with respect to our line of sight , @xmath155 . the inner and outer radii of the emitting region are less well constrained in these three cases , but choosing models with fixed emissivity power - law indices , we can estimate the inner radius to about @xmath15630% . we start by assuming either a point - source illumination ( @xmath150 ) or illumination by scattering ( @xmath151 ) , whenever the point - source model fit is inadequate . we were able to constrain the inclination angles to within a few degrees in all cases , which is important since this parameter is degenerate with the inner radius . the inner radius of the emitting region is a crucial parameter for the determination of the viscous energy available locally . the ratio of the line luminosity to the locally available energy ( listed in table [ tabfits ] in all cases ) detemines the need for external illumination as discussed in [ diskp ] below . we use the bluest parts of the het spectra ( 3300 for sdssj2125@xmath10813 , and for the remaining targets ) to estimate the change in flux between the het and sdss epochs and list the results in table [ tabvar ] . table [ tabvar ] summarizes the optical variability time scales and amplitudes for the five _ chandra _ double - peaked emitters . the changes in the continuum fluxes over the wavelength ranges specified in column 6 are given in column 7 , with the changes in the h@xmath42-line fluxes ( h@xmath38 in the case of sdssj2125@xmath10813 ) listed in column 8 . the continuum flux varied by factors of 23 for the three higher redshift sources ( sdssj1238@xmath55325 , sdssj1407@xmath10235 , and sdssj2125@xmath10813 ) in the 34year ( rest - frame ) period between the sdss and het observations . in addition to the reported continuum - flux variability , the line profile shapes also varied between the sdss and het observations . the ratio of the blue - to - red peak height varied from @xmath700.9 to @xmath701 in the case of sdssj0132@xmath10952 , and @xmath157 to @xmath158 in the case of sdssj1407@xmath10235 ( see the last panel of figure [ ospec ] ) , as has been observed for other double - peaked emitters ( e.g. , * ? ? ? * ; * ? ? ? * ) . in order to estimate the host - galaxy corrections , we fit the sdss spectra , which have broader wavelength coverage and provide better constraints on the agn host - galaxy decomposition , with a combination of host - galaxy and agn components . the results of these fits , for the three double - peaked emitters with detectable host - galaxy contribution in the 3@xmath142-diameter sdss fibers , are shown in figure [ ospecfit ] . the host - galaxy and agn components for the fit are constructed using agn and galaxy eigenspectra obtained from large sdss samples by @xcite . we subtract the host - galaxy spectra obtained in this fashion from the three sdss spectra with substantial host - galaxy contributions : sdssj0132@xmath10952 , sdssj1004@xmath54801 , and sdssj1407@xmath10235 ( see table [ tabvar ] for the respective host - galaxy contributions ) . in order to extrapolate the optical observations to the uv and obtain an estimate of the monochromatic luminosities at 2500 , @xmath159 , we scale the mean quasar spectral energy distribution ( sed ) of @xcite to match the observed agn continua in the optical . we use the het spectra of sdssj1238@xmath55325 and sdssj2125@xmath10813 for this purpose , and the sdss agn components scaled to the het epoch for the three sources with significant galaxy contributions . this is necessary as the het coverage is insufficient to provide good constraints on the host - galaxy contributions . we caution , however , that this solution is not optimal , since the 3@xmath142-diameter sdss fibers are larger than the rectangular @xmath160 lrs slits , and the host - galaxy contribution could be smaller in the het spectra . under the seeing conditions at which the het spectra were obtained , however , the agn spectrum is further diluted by starlight and this representation is adequate . the @xcite quasar sed was constructed to cover the infrared ( ir ) to the x - ray region using _ spitzer space telescope _ , sdss , galex ( galaxy evolution explorer ) , and _ rosat _ data for a sample of unobscured sdss quasars . the slopes of the host - galaxy corrected spectra for all of our sources agree well with the @xcite quasar continuum , as discussed further in [ bollum ] . cccccc j0132@xmath10952 & 2007 - 10 - 22 & @xmath161 & @xmath162 & 0.028 & 0.22 + j1004@xmath54801 & 2007 - 02 - 10 & @xmath163 & @xmath164 & 0.0092 & 0.073 + j1238@xmath55325 & 2005 - 03 - 02 & @xmath165 & @xmath166 & 0.015 & 0.12 + j1407@xmath50235 & 2004 - 05 - 17 & @xmath167 & @xmath168 & 0.037 & 0.29 + j2125@xmath10813 & 2003 - 08 - 27 & @xmath169 & @xmath170 & 0.056 & 0.44 + [ tabgalex ] ccccc j0132@xmath10952 & 1999 - 08 - 09 & @xmath171 & @xmath172 & @xmath173 + j1004@xmath54801 & 1999 - 01 - 19 & @xmath174 & @xmath175 & @xmath176 + j1238@xmath55325 & 2000 - 04 - 25 & @xmath177 & @xmath178 & @xmath179 + j1407@xmath50235 & 2000 - 03 - 28 & @xmath180 & @xmath181 & @xmath182 + j2125@xmath10813 & 2000 - 08 - 27 & @xmath183 & @xmath184 & @xmath185 + [ tab2mass ] four of the five _ chandra _ double - peaked emitters are well fit by power - law models with spectral indices in the range @xmath186 , in good agreement with the continua of similar luminosity agns , @xmath187 , studied by @xcite . a well - known correlation exists between @xmath188 and the fwhm the h@xmath38 line in agns ( e.g. , * ? ? ? @xcite show that the @xmath188fwhm@xmath189 relation is driven by the relation between the hard x - ray band photon index and the accretion rate ( parameterized by @xmath190 ) , since fwhm@xmath189 is an accretion - rate indicator for unobscured agns . there is no evidence that the balmer - line widths can be used as an accretion - rate indicator for double - peaked emitters , especially for the double - peaked emitters with extremely large line widths considered here . however , if the x - ray power - law photon index can be used as an indicator of the accretion rate , the double - peaked emitters studied here would have @xmath191 ( with a @xmath192 range of @xmath193 ) and at least four of the five would be considered efficient accretors , a result that is supported further by the existence of big blue bumps " a characteristic feature in the spectral energy distributions of quasars ( see [ bollum ] ) . four of the five double - peaked emitters observed with _ chandra _ are x - ray variable on long timescales . both sdssj0132@xmath10952 and sdssj1407@xmath50235 have rass upper limits lower than the observed acis - s fluxes in the band , suggesting that these sources have brightened on decade timescales by at least a factor of @xmath194 and @xmath195 , respectively . sdssj1238@xmath55325 has the same x - ray flux during the two observations separated by 12 years in the agn rest - frame . the remaining two sources are observed in a fainter state with _ chandra _ than their corresponding early 1990s _ rosat _ pspc detections by factors of about 3 ( sdssj1004@xmath54801 ) and 6 ( sdssj2125@xmath10813 ) . for comparison , @xcite find that the typical flux - variability amplitude of similar luminosity agns on timescales of decades is @xmath196 . @xcite compute the mean flux - variability amplitude , @xmath197 ( for a description of this quantity see * ? ? ? * ) , of the unobscured agns from the lockman hole sample . they find that the average agn has a @xmath198 , with a range of @xmath199 for the agns with confirmed variability . the corresponding @xmath197 values for our four variable sources are @xmath430.2 , 0.5 , @xmath430.6 , and 0.7 for sdssj0132@xmath10952 , sdssj1004@xmath54801 , sdssj1407@xmath10235 , and sdssj2125@xmath10813 , respectively . based on this small sample , it appears that most double - peaked emitters are more variable on decade timescales than the typical agn . we tested the _ light curves for short - term variability using the xronos package . it enables us to compute the rms fractional variation ( @xmath200 ) of the count rate within each exposure after binning the data in intervals ranging from half a minute to about 1530 minutes ( allowing for at least 5 bins in each case ) . no short - term variability was detected . the @xmath201 gaussian limits on the variability range between @xmath2020.3 for the higher signal - to - noise low - redshift objects and @xmath2030.6 for sdssj2125@xmath10813 , depending on the length of the time interval probed . the non - detection of short - term variability is expected , as our objects are relatively luminous in the band ( with luminosities ranging between a few times @xmath204ergs@xmath63 and @xmath107ergs@xmath63 ) , and agn variability amplitudes on short timescales are known to decrease with increasing luminosities ( e.g. , * ? ? ? @xcite show that for agns with luminosities in the @xmath205ergs@xmath63 range , timescales longer than @xmath206s are typically necessary to detect rms variability of more than 10% ; these timescales are 210 times longer than the exposure times in our snapshot survey . the optical variability of the five double - peaked emitters is also significant on long timescales ( see table [ tabvar ] ) , with flux - density changes of up to a factor of 3 over 34yrs in the agn rest frame in our sample , accompanied by h@xmath42-line flux changes of up to a factor of 2.5 . figure [ ospecfit ] shows the het and sdss spectra of three of the five _ chandra _ sources . only two of the double - peaked emitters ( sdssj0132@xmath10952 and sdssj1407@xmath50235 ) show clear h@xmath42 line - profile variations which are similar to those observed in double - peaked emitters with long - term variability monitoring ( e.g. , * ? ? ? * ; * ? ? ? the three higher redshift agns in our sample ( sdssj1238@xmath55325 , sdssj1407@xmath50235 , and sdssj2125@xmath10813 ) vary by factors of in the 4000 continuum in a few years . for comparison , the expected agn variability at 4000 for a similar time lag in the sdss is @xmath207mag , or about 60% @xcite . we conclude that the double - peaked emitters studied here are typically more variable on long timescales ( years ) in the optical ( a variability range of 10300% , see table [ tabvar ] ) and the x - ray ( factors of @xmath194 , 3 , 1 , @xmath195 , and 6 , in order of increasing ra ) than the average agn . in s06 we compared the width of the @xmath208 residual distribution to that of the luminosity - matched @xcite subsample and found that it was broader at the 1@xmath40 level . this finding , which could be interpreted as an indication of larger x - ray and/or optical / uv variability for double - peaked emitters , is confirmed with the small sample studied here . we use the uv - to - x - ray index , @xmath208 , defined in terms of the logarithm of the ratio of the rest - frame 2500 and 2kev monochromatic fluxes , @xmath209 $ ] , to compare the x - ray emission of our sample to that of more typical agns with similar optical / uv monochromatic luminosities . the uv - to - x - ray indices of the five _ chandra _ targets are shown as a function of their 2500 monochromatic luminosities in figure [ aox ] and listed in table [ tab3 ] . there is a tendency for the double - peaked emitters to have flatter @xmath208 values than typical rq agns . this is expected for the one rl double - peaked emitter , sdssj1238@xmath15325 , but is also apparent in the four rq targets . we can characterize this discrepancy by subtracting from each observed value of @xmath208 the value expected based on the @xmath210 relation found by @xcite for rq agns , @xmath211 , and compare this to the spread , @xmath212 , observed around the relation for normal rq agns . the @xmath208 residual values ( @xmath213 ) , together with the values of @xmath212 for similar luminosity rq agns are given in columns 7 and 8 of table [ tab3 ] . the @xmath208 residuals of the four rq double - peaked emitters are larger than or about equal to the spread observed around the @xmath210 relation for normal rq agns . the @xmath208 values are @xmath214 too flat in the cases of sdssj0132@xmath10952 and sdssj1407@xmath50235 , and @xmath215 too flat in the cases of sdssj1004@xmath54801 and sdssj2125@xmath10813 . the remaining double - peaked emitter , sdssj1238@xmath55325 , has an uv - to - x - ray index which is 0.41dex flatter than the value expected for similar - luminosity rq agn . this difference is typical for rl agns , which are known to have 2kev monochromatic luminosities of a factor of @xmath216 higher than those of rq agns with similar uv emission ( e.g. , * ? ? ? the combined optical and x - ray variability of the double - peaked emitters together with the uncertainties in the determination of @xmath217 is large enough to account at least partially for the @xmath213 values observed for sdssj1004@xmath54801 ( up to 0.2dex ) , sdssj1407@xmath50235 ( up to 0.3dex ) , and sdssj2125@xmath10813 ( up to 0.3dex ) . for x - ray or optical variability to be the cause of the anomalous @xmath208 values observed for these three rq agns , the variability amplitude over the 24 months separating the _ chandra _ and het observations would have to reach the amplitude observed over years ( optical ) or decades ( x - ray ) . since the amplitude of agn variability typically increases with time , this is not likely . using alternative methods to estimate @xmath217 , by ( 1 ) extrapolating the measured flux density at 3700 to 2500 using a uniform power - law of the form , @xmath218 , with slope @xmath219 @xcite , or ( 2 ) using the @xcite sed template scaled to the observed fluxes in the extreme uv ( see [ bollum ] ) , we find results consistent with those reported in table [ tab3 ] within 0.2dex ( typically @xmath1380.1dex ) in all cases this is equivalent to an uncertainty of @xmath1380.1dex in @xmath208 , which is inadequate to explain the observed flattening of the uv - to - x - ray index . in addition , although variability and measurement uncertainty could be responsible for the flattening of the @xmath208 values in individual cases , it is unlikely that all four rq double - peaked emitters would be affected in the same way . the mean value of the @xmath208 residuals is @xmath220 for the four rq double - peaked emitters observed with _ , compared to @xmath221 , for the luminosity - matched subsample of rq agns from @xcite . both the gehan and logrank tests ( @xmath222 , @xmath223% and @xmath224 , @xmath225% , respectively ; see s06 for the use of these sample - comparison tests ) confirm that the current sample of 4 rq double - peaked emitters has a statistically different distribution of @xmath208 residuals from the one found for all agns with similar uv luminosities and single - peaked lines presented in @xcite . this tendency was discernible in the s06 sample ( left panel of figure 9 in s06 ) , but not statistically significant ; it was also observed in the preliminary analysis of the five _ chandra _ targets in combination with a sample of double - peaked emitters observed with _ swift _ @xcite . figure [ sed ] shows the spectral energy distributions ( seds ) of the five _ chandra _ sources from the near - infrared ( nir ) to the x - ray band . the x - ray spectra are represented as power laws with photon indices given in table [ tab2 ] . for the two agn with negligible host - galaxy contamination ( sdssj1238@xmath55325 and sdssj2125@xmath10813 ) we show the het - epoch spectra in the optical region ; for the remaining three objects we plot the sdss agn components scaled to the flux level during the het epoch . in all five cases we also plot the sdss - epoch agn components . comparison of the het - epoch and sdss - epoch spectra gives an idea of the optical variability on long timescales ( 45years ) . the near - uv ( nuv , @xmath226 ) and far - uv ( fuv , @xmath227 ) fluxes , obtained by the galex mission ( see table [ tabgalex ] ) and corrected for galactic extinction , are also shown . the extinction corrections , @xmath228 and @xmath229 , are from @xcite , who used the @xcite parameterization of the galactic extinction law . the nir fluxes ( obtained from the 2mass magnitudes reported in table [ tab2mass ] ) contain the host - galaxy contributions in all cases , and are well above the template seds in figure [ sed ] . if we subtract an elliptical - galaxy sed ( scaled to the galaxy contribution in the sdss spectra and corrected for the sdss fiber apertures ; see fioc & rocca - volmerange 1997 for the galaxy sed ) from the nir fluxes , we can account for part of the discrepancy between the observed nir fluxes and the sed templates for sdssj0132@xmath10952 and sdssj1004@xmath54801 . in both cases a standard elliptical - galaxy template can account for @xmath680.1dex in the nir ; a dusty starburst would be necessary to explain the high observed nir luminosities . in the cases of sdssj1238@xmath55325 and sdssj2125@xmath10813 , the nir colors are more consistent with an agn sed and the high observed luminosities are in line with the higher optical / uv fluxes . the average quasar sed from @xcite , scaled to the agn continuum flux in the optical in each case , is also shown in figure [ sed ] . we used the sdss agn component , scaled to the het - epoch flux in the three cases with significant host - galaxy contribution , and the het spectra in the remaining two cases . the @xcite average seds are similar from the infrared ( ir ) to the uv region to the standard @xcite median seds ( both rq sed templates are shown in the second panel of figure [ sed ] for one of our sources , sdssj1407@xmath50235 ) , but they differ substantially in the x - ray band . the @xcite sample was biased toward agns with good s / n _ einstein _ data ( hence x - ray brighter than the typical agn ) and bright in the extreme uv , and @xcite argue that typical agns are redder and have lower x - ray - to - optical ratios . the @xcite rq sed has an @xmath230 and the @xcite rq sed an @xmath231 , compared to @xmath232 and @xmath233 for their respective @xmath217 from the @xcite relation ; both seds have total ( including the ir bump ) bolometric luminosities of @xmath234ergs@xmath63 . note that the galex observations agree well with the optically - scaled @xcite template seds for the four rq _ chandra _ agn ( within 0.11dex , or @xmath235% ) . the _ nuv _ band and the optical bracket rest - frame 2500 in all four cases , suggesting that the @xmath217 estimates based on the optically - scaled teplate seds and the uv - to - x - ray indices are fairly robust ( changes in @xmath208 of less than 0.05 ) . the rl agn , sdssj1238@xmath55325 , is 0.4dex more luminous in the _ nuv _ band during the galex epoch relative to the het - epoch optically - scaled rl template sed ; a similar variation at 2500 in the 17 days ( 13 in the agn rest frame ) between the _ chandra _ and het observations would result in a change in @xmath208 of 0.16 , and can not fully account for the observed flatter @xmath208 . the arp102b sed , discussed in more detail by @xcite , is shown in the last panel of figure [ sed ] for comparison . we use the _ advanced satellite for cosmology and astrophysics _ ( _ asca _ ) x - ray spectrum presented by eracleous et al . ( 2003a ; see also lewis & eracleous 2006 ) , the _ hubble space telescope _ ( _ hst _ ) uv spectrum given in @xcite , and the nir and mid - infrared ( mir ) spectra published by @xcite and @xcite , respectively . @xcite note strong stellar features in the nir spectrum of arp102b , which was integrated over a 700pc region centered on the active nucleus . the mir source extent , studied by @xcite , was around 250pc ; for comparison , the _ hst _ aperture diameter is @xmath70400pc . in the same panel of figure [ sed ] , we show the average quasar sed from @xcite in the infrared - to - uv region ( scaled to the reddest uv wavelength of the arp102b _ hst _ spectrum , @xmath703200 ) together with the _ asca _ power law in the x - ray region . in addition , the nir emission of arp102b includes a strong galaxy contribution , while in the mir variability by a factor of 2.5 over a five year period was noted ; both variability and varying host - galaxy contributions are likely responsible for part of the discrepancy between the quasar sed and the arp102b ir data . the arp102b sed is characterized by a lack of a ` big blue bump ' ( bbb ) , the excess emission between the x - ray and uv bands ( @xmath236hz ) normally associated with the innermost parts of the accretion disk ( which is likely replaced by a hot radiatively inefficient central structure in the case of arp102b and other agns with riafs ) . in contrast to the arp102b sed , the galex observations of the five double - peaked emitters presented here show that the bbb is present in all five seds , with the possible exception of sdssj1004@xmath54801 . the presence of bbbs is further confirmation for the relatively high efficiency of the accretion flows in these agns , excluding riaf models for them . using the template seds from figure [ sed ] , we can estimate the bolometric luminosities of the double - peaked emitters which we list in the last column of table [ tab3 ] . the @xcite sed does not extend to the radio domain ( @xmath237 ) , but the @xmath238hz region is energetically unimportant and does not affect the bolometric luminosity estimates . the template sed bolometric luminosities will overestimate the true luminosities if the ir bump consists of fully reprocessed nuclear continuum emission and should be excluded when computing the total luminosity . such ` double - counting ' of the ir emission can result in overestimates of the bolometric luminosity of up to 2030% . we can obtain upper limits to the bolometric luminosities and an estimate of possible uncertainties by normalizing the ir - to - extreme - uv part of the sed template to the nir ( in all cases but sdssj1004@xmath54801 ) , the extreme uv ( in the case of sdssj1238@xmath55325 ) , and/or the sdss - epoch spectra ( in the case of sdssj2125@xmath10813 , for example , where the sdss - epoch flux was higher ) . in all cases , we find that the bolometric luminosities can be at most a factor of 23 higher than those reported in table [ tab3 ] . on average about 30% of the bolometric luminosity is emitted between kev in our sample ( with a range of 1636% ; or between 2% and 12% in the 0.52kev x - ray band ) , compared to @xmath239% emitted between kev for the agn composites of @xcite and @xmath240% for the median rq agn composite of @xcite . this comparison is not entirely fair , since the @xcite and @xcite seds are constructed using more luminous agns ( @xmath241 and @xmath242 , respectively ) and @xmath208 is luminosity dependent . if we renormalize the @xcite sed to the average value of @xmath217 in our sample , @xmath243 , and scale the x - ray part of the seds to the level inferred from the @xcite @xmath210 relation for typical rq agns , we find that even in this case only 78% of the bolometric luminosity is emitted between kev by typical rq agns . this restates the conclusion from [ saox ] : the average rq double - peaked emitter from our _ chandra _ sample has an @xmath208 value flatter by 0.25dex relative to that of a typical rq agn and is 0.65dex more luminous at 2kev , which is equivalent to emitting about 30% of its bolometric luminosity between kev . to compute theoretically the emerging spectrum of an accretion disk , both an accurate representation of the vertical structure of the disk and a suitable approximation for the radiation transfer are necessary . @xcite , for example , computes the balmer - line emission from the outer , optically thin parts of the accretion disks of cataclysmic variables , assuming a vertically homogeneous disk and local termodynamic equilibrium ( lte ) . this paper finds that the outer regions of the disk cool through line emission ( with @xmath244% of the locally available viscous energy emitted as h@xmath42 ) , producing the balmer - line series with small balmer decrements ( @xmath245 ) . for comparison , the observed balmer - line decrements range between 3 and 4 for the four low - redshift _ chandra _ targets ( for which both h@xmath42 and h@xmath38 fall within the sdss wavelength coverage ) similar to the balmer decrement of 3.2 which is considered typical for the blr gas . assuming a homogeneous photoionized disk , @xcite and @xcite argued that standard accretion disks in agns can produce low - ionization lines with characteristics similar to those observed in the blrs of agns ( including steeper balmer - line decrements ) if the disk is externally illuminated . the lines arise in the radiatively excited skin of the inner ( @xmath246 ) accretion disk or the radiatively heated body of the outer ( @xmath247 ) disk . the need for external illumination of the accretion disk is an important constraint on models aiming to explain the existence of double - peaked balmer lines in agns . we can use the ratio of the h@xmath42 luminosity to the viscous power available locally in the line - emitting region of the disk , @xmath248 , in comparison with the @xcite computations , to estimate the need for external illumination . the viscous power available locally in the line - emitting region of the disk , @xmath249 , can be estimated using an estimate of the bolometric luminosity and the extent of the emission - line region as described in , e.g. , e03 and s06 . in this work we use the bolometric luminosity estimates based on the observed spectral energy distributions ( seds ) presented in [ bollum ] . e03 and s06 have found that for most double - peaked emitters @xmath248 ranges between 0.1 and 3 . the @xmath248 ratios for the five double - peaked emitters studied here are shown in figure [ power ] in comparison to the 20 double - peaked emitters from table 7 of e03 . note that the apparent anti - correlation seen in this figure is just a consequence of the fact that @xmath249 is derived from the bolometric luminosity which correlates with the x - ray flux . for @xmath250 , the energy released locally by viscous dissipation can not power the observed line even in principle . this is the case for one of our _ chandra _ targets , sdssj1004@xmath54801 , and 7 out of the 20 ( 35% ) double - peaked emitters presented by e03 . three of the remaining four _ chandra _ double - peaked emitters ( as well as 9 out of the 20 e03 sources ) have @xmath251 . thus , the observed h@xmath42 luminosities are uncomfortably close to the total energy available locally , and larger than the 20% upper limit of @xcite , suggesting that external illumination plays a role . the only exception is the most luminous sdss double - peaked emitter , sdssj2125@xmath10813 ; for a standard h@xmath42 to h@xmath38 luminosity ratio of 3.2 , the measured @xmath252 corresponds to @xmath253 , and no external illumination is necessary . the total soft - band x - ray luminosity is a factor of 1030 higher than the h@xmath42 luminosities for all five double - peaked emitters ( and ranges between 0.7 and 100 for the e03 sources ) , showing that enough energy is released within a few @xmath254 of the black hole to illuminate the balmer - line - emitting portion of the disk and produce line - emission with the observed strength . assuming that the disk is photoionized and a thin skin emits the lines , the h@xmath42 power depends on the illuminating flux and the reprocessing efficiency ( i.e. , what fraction of the cooling comes out as h@xmath42 ) . @xcite show that , in the case of a photoionized disk , about 20% of the effective continuum power " , defined as @xmath255 , could be emitted as h@xmath42 . in all five cases studied here , the h@xmath42-line luminosities are equal ( in the case of sdssj1238@xmath55325 ) or much smaller ( by factors of 7 , 6 , 16 , and 66 , respectively , for the remaining four cases in order of increasing ra ) than 20% of the effective continuum power . we conclude that external illumination is necessary in at least four of the five cases studied here , since the locally available viscous energy is insufficient and the balmer - line decrements too steep , compared to the @xcite predictions . a photoionized disk as described in e.g. , @xcite , can produce h@xmath42-line emission of the observed strength with the proper balmer - line decrements . based on the x - ray spectral fits presented in [ analysis ] , large column density neutral absorbers are ruled out in all cases . as a simple test for the presence of ionized absorbers , we added one or more edges at @xmath256kev to all model fits ; this failed to improve the fits at significance levels @xmath257% in all cases , with the exception of sdssj0132@xmath10952 . sdssj0132@xmath10952 ( see [ spec6770 ] ) , requires ionized intrinsic absorption with @xmath258@xmath57 and @xmath259ergs@xmath63 cm , but both quantities are poorly constrained . silimarly , although not required by the data , the limited signal - to - noise ratio in the case of sdssj2125@xmath10813 allows for the addition of a @xmath260@xmath57 , @xmath261ergs@xmath63 cm ionized intrinsic absorber . recent work by @xcite casts doubt on the suitability of a riaf to illuminate efficiently the outer portions of the accretion disk , suggesting instead that the required illumination can be provided by low lorentz factor ( @xmath262 ) jets in rl objects or slow outflows in the case of rq double - peaked emitters . if slow - moving outflows in the immediate vicinity of the black holes are responsible for the disk illumination , then x - ray observations are expected to show large x - ray absorbing columns : @xmath263@xmath57 , providing a suitable scattering depth of @xmath264 , according to @xcite . as an example of the prevalence of suitable ionized absorbers in agns , @xcite cite the work of @xcite , who found that 20 out of the 45 agns in their sample show signs of ionized absorption ( parameterized by @xmath265kev absorption edges in 16 of the 20 cases with @xmath266 , and xspec _ absori _ fits in the remaining 4 ) . the x - ray spectra of the four rq double - peaked emitters studied here do not require large column density ionized absorbers , with the exception of sdssj0132@xmath10952 , where a lower column - density absorber is present , and sdssj2125@xmath10813 , where a high column - density and ionization absorber could not be ruled out . snapshot survey presented here has yielded spectra with a limited numbers of counts and was not designed to provide stringent constraints on the column densities and ionization states of possible ionized absorbers , but the presence of such absorbers appears unlikely in at least two of the four rq objects from the current sample . if high column density absorbers were common in double - peaked emitters as a class , we might also expect that the uv - to - x - ray indices measured based on x - ray photometry ( e.g. , s06 ) should reflect this by showing steeper @xmath208 distributions ( since the measured x - ray fluxes would be depressed by absorption ) . in fact , as argued above , the double - peaked emitters tend to have flatter @xmath208 values than similar optical / uv luminosity agns , suggesting excess x - ray emission relative to the uv emission . we conclude that the slow - outflow model of @xcite could provide the illumination necessary to power the strong double - peaked lines observed in some double - peaked emitters , but is unlikely to be responsible for the necessary extra power in all cases . the five double - peaked emitters observed with _ chandra _ show diverse properties in the x - ray band . their kev emission is well fit by simple power laws ; the power - law emission is modified by intrinsic absorption in the case of sdssj0132@xmath10952 . the measured power - law photon indices , @xmath186 , are typical for similar luminosity agns , with the exception of the most optically luminous object , sdssj2125@xmath10813 . the kev spectrum of sdssj2125@xmath10813 is either absorbed , or has an unusually flat slope of @xmath2 , and possible k@xmath42 emission . the soft x - ray emission is strongly variable on timescales of years , but no short term ( minutes ) variability is detected , as expected for snapshot observations of similar x - ray luminosity agns . the optical fluxes and h@xmath42-line fluxes also vary strongly on long timescales , with stronger variability amplitudes than typical agn observed in the optical . the four rq double - peaked emitters have strong x - ray emission ( with about 30% of the total flux emitted in the kev band ) relative to their estimated monochromatic luminosities at 2500 , and statistically flatter uv - to - x - ray spectral indices . they all show a classical bbb in the extreme uv , suggesting the absence of riafs . external illumination of the accretion disk is necessary to produce balmer - line emission of the observed strength in at least four of the five cases ( with the exception of the most luminous sdss double - peaked emitter sdssj2125@xmath10813 ) , and the excess x - ray emission could supply the extra energy although it is still unclear how the accretion disk will intercept a sizable fraction of the soft x - ray emission . the hobby - eberly telescope ( het ) is a joint project of the university of texas at austin , the pennsylvania state university , stanford university , ludwig - maximillians - universitt mnchen , and georg - august - universitt gttingen . the het is named in honor of its principal benefactors , william p. hobby and robert e. eberly . the marcario low - resolution spectrograph is named for mike marcario of high lonesome optics , who fabricated several optics for the instrument but died before its completion ; it is a joint project of the hobby - eberly telescope partnership and the instituto de astronoma de la universidad nacional autnoma de mxico . dum abramowicz , m. a. , igumenshchev , i. v. , quataert , e. , & narayan , r. 2002 , , 565 , 1101 almaini , o. , et al . 2000 , , 315 , 325 bianchi , s. , guainazzi , m. , matt , g. , & fonseca bonilla , n. 2007 , , 467 , l19 brandt , w. n. , mathur , s. , & elvis , m. 1997 , , 285 , l25 cao , x. , & wang , t .- 2006 , , 652 , 112 cardelli , j. a. , clayton , g. c. , & mathis , j. s. 1989 , , 345 , 245 chary , r .- 2007 , the central engine of active galactic nuclei , 373 , 443 chen , k. , halpern , j. p. , & filippenko , a. v. 1989 , , 339 , 742 chen , k. , & halpern , j. p. 1989 , , 344 , 115 collin - souffrin , s. 1987 , , 179 , 60 collin - souffrin , s. , & dumont , a. m. 1989 , , 213 , 29 done , c. , mulchaey , j. s. , mushotzky , r. f. , & arnaud , k. a. 1992 , , 395 , 275 dumont , a. m. , & collin - souffrin , s. 1990 , , 229 , 313 elvis , m. , et al . 1994 , , 95 , 1 eracleous , m. , & halpern , j. p. 1994 , , 90 , 1 eracleous , m. , livio , m. , halpern , j. p. , & storchi - bergmann , t. 1995 , , 438 , 610 eracleous , m. , halpern , j. p. , & livio , m. 1996 , , 459 , 89 eracleous , m. , & halpern , j. p. 1998 , , 505 , 577 eracleous , m. , sambruna , r. , & mushotzky , r. f. 2000 , , 537 , 654 eracleous , m. , shields , j. c. , chartas , g. , & moran , e. c. 2002 , , 565 , 108 eracleous , m. , halpern , j. p. , & charlton , j. c. 2003 , , 582 , 633 eracleous , m. , & halpern , j. p. 2003 , , 599 , 886 eracleous , m. , lewis , k. t. , bogdanovi , t. , gezari , s. , & halpern , j. p. 2006 , astronomical society of the pacific conference series , 360 , 227 fioc , m. , & rocca - volmerange , b. 1997 , , 326 , 950 gezari , s. , halpern , j. p. , & eracleous , m. 2007 , , 169 , 167 gil de paz , a. , et al . 2007 , , 173 , 185 grandi , p. , malaguti , g. , & fiocchi , m. 2006 , , 642 , 113 halpern , j. p. , eracleous , m. , filippenko , a. v. , & chen , k. 1996 , , 464 , 704 ivezi , . , et al . 2002 , , 124 , 2364 ivezi , . , et al . 2004 , the interplay among black holes , stars and ism in galactic nuclei , 222 , 525 kallman , t. r. , & mccray , r. 1982 , , 50 , 263 kallman , t. r. , palmeri , p. , bautista , m. a. , mendoza , c. , & krolik , j. h. 2004 , , 155 , 675 kauffmann , g. , et al . 2003 , , 346 , 1055 koester , b. p. , et al . 2007 , , 660 , 239 laor , a. 2007 , the central engine of active galactic nuclei , 373 , 384 lawrence , a. , & papadakis , i. 1993 , , 414 , l85 lewis , k. t. , & eracleous , m. 2006 , , 642 , 711 magdziarz & zdziarski 1995 , , 273 , 837 mateos , s. , barcons , x. , carrera , f. j. , page , m. j. , ceballos , m. t. , hasinger , g. , & fabian , a. c. 2007 , , 473 , 105 narayan , r. , & yi , i. 1994 , , 428 , l13 narayan , r. , & yi , i. 1995 , , 452 , 710 oke , j. b. 1987 , superluminal radio sources , 267 quataert , e. , & narayan , r. 1999 , , 520 , 298 piconcelli , e. , cappi , m. , bassani , l. , fiore , f. , di cocco , g. , & stephen , j. b. 2002 , , 394 , 835 piconcelli , e. , jimenez - bailn , e. , guainazzi , m. , schartel , n. , rodrguez - pascual , p. m. , & santos - lle , m. 2005 , , 432 , 15 rees , m. j. , begelman , m. c. , blandford , r. d. , & phinney , e. s. 1982 , , 295 , 17 richards , g. t. , et al . 2006 , , 166 , 470 riffel , r. , rodrguez - ardila , a. , & pastoriza , m. g. 2006 , , 457 , 61 schlegel , d. j. , finkbeiner , d. p. , & davis , m. 1998 , , 500 , 525 shakura , n.i , sunyaev , r.a . 1973 , astron . & astrophys . , 24 , 337 shemmer , o. , brandt , w. n. , netzer , h. , maiolino , r. , & kaspi , s. 2006 , , 646 , l29 shemmer , o. , brandt , w. n. , netzer , h. , maiolino , r. , & kaspi , s. 2008 , arxiv e - prints , 804 , arxiv:0804.0803 shinozaki , k. , miyaji , t. , ishisaki , y. , ueda , y. , & ogasaka , y. 2006 , , 131 , 2843 spergel , d. n. , et al . 2007 , , 170 , 377 steffen , a. t. , strateva , i. , brandt , w. n. , alexander , d. m. , koekemoer , a. m. , lehmer , b. d. , schneider , d. p. , & vignali , c. 2006 , , 131 , 2826 strateva , i. v. , brandt , w. n. , eracleous , m. , schneider , d. p. , & chartas , g. 2006 , , 651 , 749 strateva , i. , brandt , w. n. , eracleous , m. , garmire , g. , & komossa , s. 2007 , the central engine of active galactic nuclei , 373 , 399 strateva , i. v. , et al . 2003 , , 126 , 1720 nandra , k. , george , i. m. , mushotzky , r. f. , turner , t. j. , & yaqoob , t. 1997 , , 476 , 70 tarter , c. b. , tucker , w. h. , & salpeter , e. e. 1969 , , 156 , 943 thorne , k. s. 1974 , , 191 , 507 vanden berk , d. e. , et al . 2001 , , 122 , 549 wilkes , b. j. , tananbaum , h. , worrall , d. m. , avni , y. , oey , m. s. , & flanagan , j. 1994 , , 92 , 53 williams , r. e. 1980 , , 235 , 939 yip , c. w. , et al . 2004 , , 128 , 585 yip , c. w. , et al . 2004 , , 128 , 2603	we study the 0.510kev emission of a sample of five of the broadest double - peaked balmer - line emitters with _ the balmer lines of these objects originate close ( within a few hundred gravitational radii ) to the central black holes of the active galactic nuclei ( agns ) , and their double - peaked profiles suggest an origin in the agn accretion disk . we find that four of the five targets can be modeled by simple power - law continua with photon indices ( @xmath0 ) typical of similar luminosity agns . one object , sdssj0132@xmath10952 , shows evidence of ionized intrinsic absorption . the most - luminous sdss double - peaked emitter , sdssj2125@xmath10813 , has either an unusual flat spectrum ( @xmath2 ) or is also highly absorbed . it is the only double - peaked emitter for which no external illumination is necessary to account for the balmer line emission . the strength of the balmer - line emission in the remaining four objects suggests that the total line flux likely exceeds the viscous energy that can be extracted locally from the accretion disk and external illumination is necessary . all five double - peaked emitters have unusually strong x - ray emission relative to their uv / optical emission , which is the likely source of the external illumination necessary for the production of the observed strong broad lines . on average about 30% of their bolometric luminosities are emitted between 0.510kev . the spectral energy distributions of the five double - peaked emitters show the big blue bumps characteristic of radiatively efficient accretion flows . the balmer line profiles , as well as the optical and x - ray fluxes of the double - peaked emitters , are highly variable on timescales of months to years in the agn rest frame .
in this paper we study the existence of positive solutions for the nonlinear system @xmath5 \dots\\[.2pc ] ( q(t)\phi(p(t ) u'_{n}(t)))'+f^{n}(t,\textbf{u})=0,\quad 0<t<1 , \end{array}\right.\ ] ] with the following boundary condition @xmath6 here @xmath7 , @xmath1 . we always make the following assumptions : * @xmath8\times\r^{n}_{+ } \rightarrow(0,\oo)$ ] is continuous , @xmath9 * @xmath10\rightarrow ( 0,\oo)$ ] are continuous functions and @xmath11 is nondecreasing on [ 0,1 ] . problems and often arise from the study of positive radial solutions for the nonlinear elliptic system of the form @xmath12 \dots\\[.2pc ] \textmd{div}(\|\nabla u_{n}\|^{p-2}\nabla u_{n})+k_{n}(\|x\|)g^{n}(\textbf{u})=0,\end{array}\right.\ ] ] in the domain @xmath13 with the following boundary condition : @xmath14 has at least one positive radial solution under the assumption that @xmath15 is either superlinear or sublinear . the one - dimensional @xmath16-laplacian boundary value problem has also attracted considerable attention @xcite . in @xcite , it was proved that @xmath17 with boundary condition @xmath18 having at least one positive solution for certain finite intervals of @xmath19 if one of @xmath20 and @xmath21 is large enough and the other one is small enough . in this paper , we also study the following eigenvalue problem @xmath22 \dots\\[.2pc ] ( q(t)\phi(p(t)u'_{n}(t)))'+\lambda h_{n}(t)g^{n}(\textbf{u})=0,\quad 0<t<1.\end{array}\right.\ ] ] we prove that and have at least one positive solution for each @xmath19 in an explicit eigenvalue interval . recently , several eigenvalue characterizations for different kinds of boundary value problems have appeared and we refer the reader to @xcite . in this paper , we will show how our method allows to improve the range of eigenvalue intervals . the new results are easily derived from a general result , stated as theorem [ main ] , which gives sufficient conditions to guarantee the existence of at least one positive solution for systems and . our arguments are based on a well - known fixed point theorem in cones . many authors @xcite have used this fixed point theorem to discuss the existence of positive solutions for different boundary value problems . in @xcite , the existence , multiplicity and nonexistence of positive solutions for nonlinear systems of ordinary differential equations were considered using fixed point index . the same problems for the quasilinear elliptic system were studied in @xcite . the novelty of this paper is a choice of a cone different from that used in @xcite . such a choice of cones ( see ) is well fit with the systems we are considering . finally , it is worth remarking here that , we can also deal with the nonlinear system with one of the following two sets of boundary conditions @xmath23 \rm{\textbf{u}}(0)&=0,\qquad \rm{\textbf{u}}'(1)=0.\end{aligned}\ ] ] however since the arguments are essentially the same ( in fact easier ) , we will restrict our discussion to boundary data . the notation used is as follows : @xmath24 for @xmath25 the remaining part of the paper is organized as follows . in 2 , some preliminary results are given and in 3 , the main results are proved . the proof of the main results is based on a well - known fixed point theorem in cones . we recall the statement of this result below , after introducing the definition of a cone . @xmath26 let @xmath27 be a banach space and @xmath28 be a closed , nonempty subset of @xmath27 . @xmath28 is a cone if 1 . @xmath29 for all @xmath30 and all @xmath31 , 2 . @xmath32 implies @xmath33 we also recall that a completely continuous operator means a continuous operator which transforms every bounded set into a relatively compact set . if @xmath34 is a subset @xmath27 , we write @xmath35 and @xmath36 let @xmath27 be a banach space and @xmath37 be a cone . assume that @xmath38 are open subsets of @xmath27 with @xmath39 let @xmath40 be a continuous , completely continuous operator such that either 1 . @xmath41 and @xmath42 ; or 2 . @xmath43 and @xmath44 . then @xmath45 has a fixed point in @xmath46 as usual , we denote by @xmath47 $ ] the space of continuous functions from [ 0,1 ] to @xmath48 . in @xmath47 $ ] we shall consider the norm @xmath49 . in order to apply theorem [ cone ] below , we take @xmath50\times c[0,1]\times\dots\times c[0,1]$ ] ( @xmath51 times ) with the norm @xmath52 for @xmath53 . then @xmath27 is a banach space . define @xmath54~\mbox{and}~\min_{\frac{1}{4}\leq t\leq \frac{3}{4 } } u_{i}(t)\geq \rho\|u_{i}\|_{0},~i=1 , 2 , \dots , n\right\},\ ] ] where @xmath55 is given by @xmath56^{-1 } \min \left\ { \int_{0}^{\frac{1}{4}}\frac{1 } { p(s)}{\rm d}s,~\int^{1}_{\frac{3}{4}}\frac{1}{p(s)}{\rm d}s\right\}.\ ] ] one can easily verify that @xmath28 is a cone in @xmath27 . let @xmath57 be a map with components @xmath58 where @xmath59 is defined by @xmath60 \dots\\[.2pc ] \int_{t}^{1}\frac{1}{p(s)}\phi^{-1 } \left(\frac{1}{q(s)}\int_{\sigma_{i}}^{s}f^{i}(\tau,\textbf{u}(\tau ) ) { \rm d}\tau\right){\rm d}s,\quad \sigma_{i}\leq t\leq 1;\end{array}\right.\ ] ] here @xmath61 is a solution of the equation @xmath62 and the map @xmath63 $ ] is defined by @xmath64 & \quad\ , -\int_{t}^{1}\frac{1}{p(s)}\phi^{-1 } \left(\frac{1}{q(s)}\int_{t}^{s}f^{i}(\tau,\textbf{u}(\tau ) ) { \rm d}\tau\right){\rm d}s,\quad 0<t<1.\end{aligned}\ ] ] -.3pc @xcite . assume @xmath65 and @xmath66 hold . then , for any @xmath67 and @xmath68 , @xmath69 has at least one solution in @xmath70 and @xmath45 is well - defined . -.3pc @xcite . assume @xmath71 holds . let @xmath72 $ ] be a nonnegative function and @xmath73 is nonincreasing on @xmath74 $ ] . then @xmath75^{-1}\min \left\{\int_{0}^{t}\frac{1 } { p(s)}{\rm d}s,~\int_{t}^{1}\frac{1}{p(s)}{\rm d}s\right\}\|u\|_{0},~0\leq t\leq 1.\end{aligned}\ ] ] in particular , @xmath76 if @xmath77 for all @xmath78 then @xmath79 , and this was used in @xcite . -.3pc@xcite . assume @xmath80 and @xmath66 hold . then @xmath81 . moreover , @xmath45 is continuous and completely continuous . lemma [ har ] implies that @xmath81 . it is easy to see that @xmath45 is continuous and completely continuous since @xmath65 and @xmath66 hold . see @xcite for a proof.@xmath82 in this section we establish the existence of positive solutions for and . moreover , we characterize the eigenvalues for problems and . for the given function @xmath83,$ ] let @xmath84 & \quad\ , \left . + \int^{\frac{3}{4}}_{t}\frac{1}{p(s ) } \phi^{-1}\left ( \frac{1}{q(s)}\int^{s}_{t}a(\tau){\rm d}\tau\right){\rm d}s\right],\quad \frac{1}{4}\leq t\leq \frac{3}{4}.\end{aligned}\ ] ] here @xmath55 is given as in . assume @xmath80 and @xmath66 hold . furthermore , it is assumed that for all @xmath85 the following hypotheses hold : 1 . there exist a constant @xmath86 and a continuous function @xmath87\rightarrow ( 0,\oo)$ ] such that @xmath88 for all @xmath89 and @xmath90 and @xmath91 2 . there exist a constant @xmath92 and a continuous function @xmath93\rightarrow ( 0,\oo)$ ] such that @xmath94 and @xmath95 then problems and have at least one positive solution @xmath96 satisfying @xmath97 we assume that @xmath98 the case @xmath99 is analogous . define the sets @xmath100 we claim that @xmath101 and @xmath102 first we shall prove ( i ) . note , from the definition of @xmath103 , that @xmath104 is the maximum of @xmath105 on @xmath74 $ ] . since @xmath106 , there exists @xmath107 such that @xmath108 and @xmath109 . we consider three cases . @xmath110.$ ] then @xmath111 & \quad\ , \left . + \int^{\frac{3}{4}}_{\sigma_{i}}\frac{1}{p(s ) } \phi^{-1 } \left(\frac{1}{q(s)}\int^{s}_{\sigma_{i}}f^{i } ( \tau,{\bf{u}}(\tau)){\rm d}\tau\right){\rm d}s\right]\nn\\[.5pc ] & \ge \frac{\rho\alpha}{2 } \left[\int_{\frac{1}{4}}^{\sigma_{i } } \frac{1}{p(s)}\phi^{-1 } \left(\frac{1}{q(s)}\int_{s}^{\sigma_{i } } \psi_{i}(\tau){\rm d}\tau\right){\rm d}s \right.\\[.5pc ] & \quad\ , \left . + \int^{\frac{3}{4}}_{\sigma_{i}}\frac{1}{p(s ) } \phi^{-1}\left(\frac{1}{q(s)}\int^{s}_{\sigma_{i}}\psi_{i}(\tau){\rm d}\tau\right){\rm d}s\right]\nn\\[.5pc ] & = \alpha\gamma_{\psi_{i}}(\sigma_{i})\geq \alpha \min_{\frac{1}{4 } \leq t\leq \frac{3}{4}}\gamma_{\psi_{i}}(t)\geq\alpha.\nn\end{aligned}\ ] ] @xmath112 then @xmath113 & \ge \rho\alpha\int_{\frac{1}{4}}^{\frac{3}{4 } } \frac{1}{p(s)}\phi^{-1 } \left(\frac{1}{q(s ) } \int_{s}^{\frac{3}{4}}\psi_{i}(\tau){\rm d}\tau \right){\rm d}s\nn\\[.5pc ] & = 2\alpha\gamma_{\psi_{i}}\left(\frac{3}{4 } \right)\geq\alpha\min_{\frac{1}{4}\leq t\leq \frac{3}{4}}\gamma_{\psi_{i}}(t)\geq\alpha.\nn\end{aligned}\ ] ] @xmath114 then @xmath115 & \ge \rho\alpha\int_{\frac{1}{4}}^{\frac{3}{4 } } \frac{1}{p(s)}\phi^{-1 } \left(\frac{1}{q(s)}\int^{s}_{\frac{1}{4 } } \psi_{i}(\tau){\rm d}\tau\right){\rm d}s\nn\\[.5pc ] & = 2\alpha\gamma_{\psi_{i } } \left(\frac{1}{4}\right)\geq\alpha\min_{\frac{1}{4}\leq t\leq \frac{3}{4}}\gamma_{\psi_{i}}(t)\geq\alpha.\nn\end{aligned}\ ] ] hence @xmath116 this implies that ( i ) holds . next we prove ( ii ) . in fact , for any @xmath117 we have @xmath118 for each @xmath119 . fix @xmath120 . then @xmath121 @xmath122 & \le \beta.\nn\end{aligned}\ ] ] therefore , @xmath123 for each @xmath124 this implies that ( ii ) holds . now theorem [ cone ] guarantees that @xmath45 has a fixed point @xmath125 thus @xmath126 . clearly , @xmath127 is a positive solution of and .@xmath82 next we employ theorem [ main ] to establish the existence of positive solutions of the following problem : @xmath128 \dots\\[.2pc ] ( q(t)\phi(p(t)u'_{n}(t)))'+h_{n}(t)g^{n}(\textbf{u})=0,\quad 0<t<1.\end{array}\right.\ ] ] we assume the following conditions : * @xmath129 is continuous with @xmath130 for @xmath131 * @xmath132\rightarrow\r_{+}$ ] is continuous and @xmath133 on any subinterval of [ 0,1 ] , @xmath9 in order to state the result , we introduce the notation @xmath134 a_{i } & = \int_{0}^{1}\frac{1}{p(s ) } \phi^{-1}\left(\frac{1}{q(s)}\int_{0}^{1}h_{i}(\tau){\rm d}\tau\right){\rm d}s,\quad b_{i}=\min_{\frac{1}{4}\leq t\leq \frac{3}{4}}\gamma_{h_{i}}(t)\\[.5pc ] a & = \max\{a_{i } , i=1,2,\dots , n\},\quad b=\min\{b_{i } , i=1,2,\dots , n\}.\end{aligned}\ ] ] suppose that conditions @xmath135 and @xmath136 hold . then problems and have at least one positive solution @xmath137 with @xmath138 for @xmath139 if one of the following conditions holds . * @xmath140 and @xmath141 ; * @xmath142 and @xmath143 . to see this , we will apply theorem [ main ] with @xmath144 we assume that @xmath145 holds . the case when @xmath146 holds is similar . from the first part of @xmath145 , there exists @xmath147 such that @xmath148 for @xmath149 choose @xmath150 for @xmath124 fix @xmath151 . then @xmath152 & \quad\ , \hbox { if } \,\ , 0\leq t\leq 1 \,\,\hbox { and } \,\,~0<u_{j}\leq\beta \,\,\hbox { for } \,\,j\in \{1,\dots , n\}\end{aligned}\ ] ] and @xmath153 & \quad\ , = \frac{1}{a_{i } } \int_{0}^{1}\frac{1}{p(s ) } \phi^{-1}\left(\frac{1}{q(s ) } \int_{0}^{1}h_{i}(\tau){\rm d}\tau\right){\rm d}s=1.\end{aligned}\ ] ] thus hypothesis @xmath154 holds . from the second part of @xmath145 , there exists @xmath86 such that @xmath155 and @xmath156 for @xmath157 thus @xmath158 for @xmath159 choose @xmath160 for @xmath124 then @xmath161 & \quad\ , \frac{1}{4}\leq t\leq \frac{3}{4},~u_{i}\geq\rho\alpha,\end{aligned}\ ] ] ( so in particular for @xmath162 and for @xmath163 , @xmath164 & \quad\ , \left . + \int^{\frac{3}{4}}_{t}\frac{1}{p(s ) } \phi^{-1}\left(\frac{1}{q(s)}\int^{s}_{t}\psi_{i}(\tau){\rm d } \tau\right){\rm d}s\right]\nn\\[.5pc ] & = \frac{1}{b_{i}}\gamma_{h_{i}}(t).\nn\end{aligned}\ ] ] thus @xmath165 this implies that hypothesis @xmath166 holds . the result now follows from theorem [ main].@xmath82 next we consider the nonlinear eigenvalue problems and . by applying theorem [ sep ] , we easily get the following result . suppose that conditions @xmath135 and @xmath136 hold . then problems and have at least one positive solution for each @xmath167 if @xmath168 . the same result remains valid for each @xmath169 if @xmath170 . here we write @xmath171 if @xmath172 and @xmath173 if @xmath174 where @xmath175 we consider the case . the case is similar . if @xmath19 satisfies , then @xmath176 and @xmath177 so theorem [ sep ] applies directly.@xmath82 @xmath26 problems and have at least one positive solution for each @xmath178 if one of the following two conditions holds : 1 . @xmath179 and @xmath180 @xmath181 2 . @xmath182 and @xmath183 @xmath9 finally , it is worth remarking here that , we can apply theorems [ sep ] and [ eig ] to study the existence of positive radial solutions for the nonlinear elliptic systems and . in fact , a radial solution of and can be considered as a solution of the system @xmath184 \dots\\[.2pc ] ( r^{n-1}\phi(u'_{n } ( r)))'+r^{n-1}k_{n}(r)g^{n}(\textbf{u})=0,\end{array}\right.\ ] ] @xmath185 with the following boundary condition @xmath186 applying the change of variables , @xmath187 we can transform and into the form @xmath188 \dots\\[.2pc ] ( q(t)\phi(\xi u'_{n}(t)))'+h_{n}(t)g^{n}(\textbf{u})=0,\quad 0<t<1,\end{array}\right.\ ] ] with boundary condition , where @xmath189 and @xmath190 which correspond to problems and with @xmath191 and @xmath192 thus , we only need to consider problems and and we can obtain results similar to those in theorems [ sep ] and [ eig ] . here we omit the details . this work is supported by the national natural science foundation of china ( item no . 10325102 and no . 10531010 ) .	this paper deals with the existence of positive solutions for the nonlinear system @xmath0 this system often arises in the study of positive radial solutions of nonlinear elliptic system . here @xmath1 and @xmath2 are continuous and nonnegative functions , @xmath3\rightarrow ( 0,\oo)$ ] are continuous functions . moreover , we characterize the eigenvalue intervals for @xmath4 the proof is based on a well - known fixed point theorem in cones . [ theore]theorem [ theore]_lemma _ [ theore]proposition [ theore]corollary [ theore]definition [ theore]_example _ [ theore]_remark _ [ theoree]_case _
the authors acknowledge support from alberta innovates technology futures ( atif ) , the natural sciences and engineering council of canada ( nserc ) , the canadian institute for advanced research ( cifar ) of which w.t . is a senior fellow , and the national science foundation of the united states ( nsf ) under award no . phy-1212462 , no . phy-1415628 , and no . 10 r. m. macfarlane and r. m. shelby , in _ modern problems in condensed matter sciences _ , vol . 21 of _ spectroscopy of solids containing rare earth ions _ , a. a. k. a. r. m. macfarlane , ed . ( elsevier , 1987 ) , pp . 51184 . n. sinclair , e. saglamyurek , m. george , r. ricken , c. l. mela , w. sohler , and w. tittel , journal of luminescence * 130 * , 1586 ( 2010 ) . special issue based on the proceedings of the tenth international meeting on hole burning , single molecule , and related spectroscopies : science and applications ( hbsm 2009 ) - issue dedicated to ivan lorgere and oliver guillot - noel .	nano - structuring impurity - doped crystals affects the phonon density of states and thereby modifies the atomic dynamics induced by interaction with phonons . we propose the use of nano - structured materials in the form of powders or phononic bandgap crystals to enable or improve persistent spectral hole - burning and coherence for inhomogeneously broadened absorption lines in rare - earth - ion - doped crystals . this is crucial for applications such as ultra - precise radio - frequency spectrum analyzers and optical quantum memories . as an example , we discuss how phonon engineering can enable spectral hole burning in erbium - doped materials operating in the convenient telecommunication band , and present simulations for density of states of nano - sized powders and phononic crystals for the case of y@xmath0sio@xmath1 , a widely - used material in current quantum memory research . rare - earth - ion - doped crystals have been studied for decades because of their unique spectroscopic properties @xcite that arise since their 4@xmath2-electrons do not participate in chemical bonding . at cryogenic temperatures they can offer narrow linewidths @xcite together with the possibility to spectrally tailor their broad inhomogeneous absorption lines @xcite . these properties have led to many applications , including optical quantum memories @xcite , signal processing @xcite , laser stabilization @xcite , as well as ultra - precise radio frequency spectrum analyzers @xcite . quantum memory implementations in rare - earth - ion - doped ( rei - doped ) crystals , such as those based on electromagnetically - induced transparency @xcite ) , atomic frequency comb @xcite , or controlled reversible inhomogeneous broadening @xcite , crucially rely on long coherence and spin - state lifetimes to achieve high efficiency and long storage time . + operating rei - doped crystals at low temperatures generally improves material properties . yet , even at temperatures below 2 k , spin - lattice relaxation , i.e. thermalization of spins via interaction with phonons , still restricts lifetimes of spin states , reducing the ability to spectrally tailor the material . furthermore , by contributing to spectral diffusion @xcite and two - phonon elastic scattering processes @xcite , lattice vibrations limit coherence times . we propose the use of nano - structured materials to overcome these limitations . this is achieved by tailoring the phonon density of states @xcite to restrict phonon processes , an approach not limited to rei - doped materials but also applicable to other impurity - doped materials such as color centers in diamond @xcite . the result is an improved performance for all applications based on spectral hole burning , or that require long coherence times , by providing long - lived spin states while simultaneously suppressing phonon - driven decoherence . we note that the modification of population dynamics between rei crystal field levels in small powders , possibly related to phonon restriction , has been observed in the context of understanding luminescence dynamics @xcite - but not yet to improve spectral hole burning or optical coherence properties . + in this paper , we propose a general procedure to identify the frequency range of detrimental lattice vibrations and restrict the relevant phonon processes in impurity - doped solids . as an example , we apply this approach to erbium - doped materials . first , we recall the mechanisms causing detrimental electronic spin flips and discuss how the suppression of the direct phonon process influences spin - state lifetimes and spectral diffusion . then , two methods for nano - structuring rei - doped crystals to suppress the direct phonon process affecting electronic spin states in er@xmath3:y@xmath0sio@xmath1 are analyzed : powder materials , which introduce a cutoff frequency in the phonon density of states , and phononic band gap crystals with a structure optimized for the targeted material and application . numerical simulations for er@xmath3:y@xmath0sio@xmath1 are presented . + relaxation of optically pumped spin populations back to thermal equilibrium by spin flips is caused by phonons primarily via three different processes @xcite : the direct process where phonons resonant with the spin transition induce a spin flip , the orbach process where two resonant phonons produce a spin flip via excitation to , and relaxation from , excited crystal field levels , and the off - resonant raman process where two phonons lead to a spin flip via inelastic scattering . levels involved in spectral hole burning , typically hyperfine or zeeman levels , are mostly separated by mhz to ghz . the rates of phonon processes are proportional to the phonon density of states in the material as discussed in @xcite . interestingly , the debye approximation for the density of states @xmath4 with v the volume of the particle and @xmath5 the frequency , breaks down for small structures . in this regime , the density of states becomes discrete @xcite and acquires a cutoff frequency below which no vibrational modes are possible and thus no phonons can exist @xcite . we predict that an ensemble of sufficiently small rei - doped crystals , or an appropriate rei phononic band gap material , will allow suppression of the direct phonon process responsible for relaxation between the spin states . + , the total rate @xmath6 ( solid red line ) is composed of a constant term @xmath7 ( dotted blue line ) , a spin flip - flop term ( green dashed line ) and a direct phonon term ( orange dash - dotted line ) , dominant above 1 t. we assume @xmath8 k , @xmath9 , @xmath10 khz , @xmath11 khz / t@xmath12 and @xmath13 khz . ] as an example , we consider extending lifetimes of the ground - state zeeman levels for persistent spectral hole burning and reducing spectral diffusion in erbium - doped crystals , particularly appealing for quantum memory applications due to the @xmath14 transition of er@xmath3 ions around 1530 nm and long coherence times . three important cases are er@xmath3:y@xmath0sio@xmath1 , for which optical coherence lifetimes of up to 4.4 ms have been observed @xcite , er@xmath3:linbo@xmath15 , exhibiting a large oscillator strength accompanied by a very broad inhomogenous linewidth @xcite , and er@xmath3:ktiopo@xmath16 , which has shown narrow homogeneous linewidths down to 1.6 khz @xcite . all er - doped crystals so far eluded efficient quantum - state storage @xcite since no efficient persistent hole burning involving electronic zeeman levels @xcite has been achieved . the only material that has shown long spin - state lifetimes is an er - doped glass fiber @xcite , in which disorder broadens the spin transition and effectively weakens the coupling between er@xmath3 ions . this reduces the relaxation rate at low magnetic field . introducing controlled disorder in crystals is potentially an alternative approach to improve spin lifetimes however , disorder can be extremely detrimental for the coherence properties , as it is the case in impurity - doped glasses . indeed , the latter are restricted in their applications since coherence times are intrinsically limited to a few @xmath17s or less by low frequency disorder mode dynamics that are present in any amorphous material @xcite . the goal of our study is to propose a method to simultaneously achieve long coherence and spin lifetimes , which excludes , e.g. , er - doped fibers - at least given current knowledge . + phonons can limit the lifetime of persistent spectral holes , but other processes can also contribute , for example paramagnetic interactions of rei with neighbouring ions or with other impurities in the host . for erbium - doped materials such as er@xmath3:y@xmath0sio@xmath1 , er@xmath3:linbo@xmath15 or er@xmath3:ktiopo@xmath16 , the total spin - flip rate at temperatures below 4 k can be approximated by @xcite @xmath18 the second term represents the average rate of mutual spin flip - flops , i.e. exchange of spin states between two er ions , with @xmath19 a constant , @xmath20 the @xmath20-factor of the ions , @xmath21 the bohr magneton and @xmath22 the boltzmann constant . the last term corresponds to the direct phonon process and is proportional to @xmath23 for @xmath24 , where the parameter @xmath25 characterizes the strength of the phonon coupling @xcite . the constant @xmath7 includes all residual relaxation processes . the raman and orbach rates are not considered here since they are strongly suppressed at low temperature @xcite , particularly when the crystal - field splitting is much larger than @xmath26 as is the case for the materials considered here . the magnetic field dependence of the spin - lattice relaxation is illustrated in figure [ fig : eryso_rate ] , which shows the total rate and its different components . values of the parameters vary for the three materials cited above . the @xmath7 term is negligible in er : linbo@xmath15 @xcite , whereas the spin flip - flop rate is very small for er : y@xmath0sio@xmath1 @xcite . er : ktiopo@xmath16 has unusually small @xmath7 and weak spin flip - flop process , so that only the direct phonon process significantly contributes to the field - dependent spectral diffusion , even at low field strengths @xcite . however , in all cases the direct phonon process becomes dominant for magnetic field strengths of 1 t and higher . + , @xmath27 after @xmath28s is shown as a function of the magnetic field strength in the presence ( blue dashed line ) and in the absence ( red solid line ) of the direct phonon process . we assume the same parameters as in fig . [ fig : eryso_rate ] and , in addition , @xmath29 hz , @xmath30 khz . ] from the above discussion , the optimum operating point would seem to be at magnetic fields below 0.5 t ; however , to reduce spectral diffusion and improve coherence properties , erbium - ion - doped - crystals often require magnetic fields of a few tesla . it has been shown that the maximum linewidth due to spectral diffusion can be described by @xcite @xmath31 with @xmath32 the full - width at half - maximum frequency broadening caused by magnetic dipole - dipole interactions between er ions . one can see that @xmath33 decreases with magnetic field and spectral diffusion can thus be suppressed at high fields due to the increased magnetic order in the system . this leads to narrow homogeneous linewidths at high fields . however , high fields lead to increased spin - lattice relaxation ( see figure [ fig : eryso_rate ] ) . in addition to being detrimental to persistent spectral hole burning , this results in faster broadening of the spectral line to its maximum value , as shown in the effective linewidth after a time delay t @xcite @xmath34 with @xmath35 the homogeneous linewidth without spectral diffusion . therefore , removing the direct phonon process will not only improve persistent spectral hole burning , but also inhibit spectral diffusion , especially at magnetic fields around 1 t ( see figure [ fig : sd ] ) . + we now identify the range of frequencies at which phonons must be suppressed . in our example of erbium - doped materials , the goal is to eliminate phonons below the frequency corresponding to the maximum zeeman splitting under an external magnetic field of around 1 t. in er : y@xmath0sio@xmath1 the @xmath20-factor along the @xmath36-axis is 13.6 @xcite , leading to a splitting of 190 ghz / t . as we will describe next , inhibiting phonons at such high frequency requires small structures . for other types of paramagnetic centers or nuclear hyperfine states , the energy splitting of the spin states is usually smaller , so that larger structures can inhibit the detrimental phonon processes . furthermore , it is possible to suppress phonons resonant with transitions between crystal - field levels , depending on the energy level splitting . in consequence , two - phonon processes such as raman or orbach processes can be suppressed . this would also lead to improved coherence lifetimes for materials in which two - phonon elastic scattering is responsible for dephasing @xcite . more generally , suppressing phonons will enable to increase the range of working temperatures for all impurity - doped materials that have properties limited by phonon processes . + we first discuss the case of small crystals . to suppress phonon processes we use the fact that the density of states of a particle features a cutoff at @xmath37 with @xmath38 the sound velocity , @xmath39 the diameter of the crystal , and @xmath40 a numerical constant that is 2.05 for a spherical particle @xcite . below this cutoff no phonons are allowed . in the region above the cutoff , the phonon density of states exhibits gaps before it transitions into a continuous function described by equation [ eq : debye ] . + to simulate the phonon density of states for a spherical particle , the eigenmodes must first be calculated . they can be separated into spheroidal - transverse , spheroidal - longitudinal , and torsional modes . the characteristic equations needed to calculate their frequencies have been described in a quantized form by takagahara @xcite . each eigenmode is identified using three quantum numbers : the radial quantum number @xmath41 , the angular momentum @xmath42 , and @xmath43 , which specifies the type of mode . using those eigenmodes , the phonon density of states can be written as @xcite @xmath44^{-1 } \ , . \label{eq : pdos}\ ] ] each eigenmode is @xmath45 times degenerate . we assume that the density of states is composed of lorentzian shaped lines at the frequency of the respective eigenmode @xmath46 and linewidth @xmath47 . in the work of meltzer and hong @xcite , the lorentzian width was assumed to increase with the frequency squared , an empirical assumption that described their particular experimental data adequately . however , since there is no theoretical framework to support that assumption , the width is assumed to be constant here . + sio@xmath1 nanoparticle with a diameter of 12 nm . each eigenmode of the spherical particle is represented by a lorentzian with a width of 1 ghz . ] in the case of er@xmath3:y@xmath0sio@xmath1 around 1 t , we estimated from eq . [ eq : cutoff ] that nanoparticles of 12 nm diameter are suitable to suppress the direct phonon process . the phonon density of states for a y@xmath0sio@xmath1 nanoparticle of 12 nm diameter , calculated according to equation [ eq : pdos ] , is shown in figure [ fig : pdos ] . phonon processes at frequencies below 200 ghz are fully restricted . this assumes that there is no phonon propagation between the particles . simulating the effect of touching particles on phonon propagation is not trivial and subject further theoretical and experimental research . however , for a lightly packed powder immersed in cold helium gas we expect properties approaching those of single , isolated particles . note that particles separated by electrostatic interaction can be obtained by coating the powder with the appropriate surfactant or by specific syntheses @xcite . + in powders , averaging of anisotropic properties over the random distribution of crystallite orientations generally causes issues in applications such as quantum memories and spectroscopy . however , if the material has a dielectric or magnetic anisotropy , the crystals can be preferentially oriented using external fields @xcite . this allows the control of parameters , such as zeeman splittings and the direction of the ion s transition dipole moment . scattering of laser light by powder samples can be reduced or even eliminated if the size of the crystals is smaller than the wavelength of the interacting light , since the powder can be treated as a bulk material with an effective ( average ) refractive index . + crystal powders can be produced using several approaches and spectroscopic properties comparable to bulk crystals have been observed in some cases @xcite . a thorough review of different growth methods can be found in @xcite and references therein . growing nano / microcrystals has the advantage of being relatively inexpensive . an alternative way to produce nanocrystals is to grind a bulk crystal into powder using techniques such as ball milling . with this approach , sizes on the order of 10 nm are achievable by using very small grinding balls @xcite . however , reducing dimensions below around 10 nm may result in additional ( detrimental ) decoherence @xcite and lifetime limiting effects @xcite . + a second possibility to avoid detrimental phonons is to create a phononic band gap material @xcite engineered to produce a gap in the density of states around the undesired frequencies . such structures are employed in the field of optomechanics @xcite , and are also used for thermal phonon engineering i.e. for thermal metamaterials and thermocrystals @xcite . by changing the geometry and overall size of phononic crystal structures , it is possible to optimize the band gaps for a specific application @xcite . + phononic crystal materials have the advantage of employing an oriented single crystal . in addition , the problem of contact between nanoparticles would be avoided . the fabrication of phononic band gap materials has already been explored for materials such as silicon or gallium arsenide @xcite and phononic structures with a bandgap at 33 ghz and dimensions around 100 nm have been demonstrated @xcite . to create phononic crystal structures in currently - used host materials such as y@xmath0sio@xmath1 , y@xmath15al@xmath1o@xmath48 or linbo@xmath15 , thin films of those materials would be a prerequisite . thin films of linbo@xmath15 are commercially available and there have been demonstrations of thin single - crystal film fabrication for other rei host materials @xcite . finally , structuring these materials with nanometer precision has been shown to be feasible using focused ion beam milling @xcite . + to illustrate this approach for our example of er : ysio , we designed and optimized a phononic structure using finite element simulations ( comsol ) . the structure is inspired by designs of safavi - naeini et al.@xcite . we engineered our structure such that it produces a band gap around 178 ghz with a width of 3.5 ghz . figure [ fig : pnc ] shows the proposed geometry of the phononic crystal unit cell together with its band structure . even though it is still challenging , the rapid progress in the field of nano - fabrication together with the discovery of helium focused ion beam milling allows structures with these dimensions to be fabricated out of a thin film @xcite . through additional etching it is possible to remove part of the substrate below the phononic crystal such that the structure suspended and only connected to the substrate at its borders . this ensures complete phonon suppression . + in summary , we propose to restrict phonon processes in rei - doped materials through nano - tailoring in order to improve persistent spectral hole burning and reduce spectral diffusion . we briefly discussed two methods towards this end the use of nano - crystals and engineered phononic band gap structures and described how our concept can be applied to well - studied er - doped crystals . in turn , this would allow realizing optical quantum memories operating at convenient telecommunication wavelengths and increase the performance of other applications that are based on persistent spectral hole burning or that rely on narrow homogeneous line widths .
quantum entanglement @xcite is a quantum mechanical phenomenon created by the separation of interacting quantum particles . this fundamental concept underpins a wide range of research in physics , such as quantum information , quantum computing , and quantum gravity , etc . when one considers this phenomenon in a study , it is often quantified as an _ entanglement entropy_. in the field of condensed - matter physics , the von neumann entanglement entropy has been frequently used as an order parameter to investigate quantum phase transition or topological order in the quantum many - body problems . in particular , for one - dimensional ( 1d ) quantum systems the entropy can be directly related to the _ central charge _ of the conformal field theory ( cft ) . @xcite to date , a method with this relation has been developed as a powerful tool to study the universal properties of the tomonaga - luttinger ( tl ) liquid @xcite , in combined with numerical methods such as exact diagonalization ( ed ) or density - matrix renormalization group ( dmrg ) @xcite techniques . let us consider a quantum 1d periodic system with length @xmath0 . the von neumann entanglement entropy of its subsystem with length @xmath1 is given as @xmath2 , where @xmath3 is the reduced density matrix of the subsystem and @xmath4 is the full density matrix of the whole system . using the cft , the entropy of the subsystem with length @xmath1 for a fixed system length @xmath0 has been derived : @xcite @xmath5+s_1 \label{entropy}\end{aligned}\ ] ] where @xmath6 is the central charge of the associated cft and @xmath7 is a non - universal constant . a prime objective of using this formula is to estimate the central charge , @xcite which provides definitive information concerning the universality class of @xmath8 dimensional system . @xcite this estimation has been applied to a variety of fermionic systems ; e.g. , with impurity @xcite , coupled to bosons , @xcite and under the magnetic field , @xcite etc . although the central charge can be also obtained by examining the finite - size correction of ground - state energy or the low - temperature behavior of specific heat , @xcite excited states have to be taken into account and considerable difficulty is involved in the ed and dmrg calculations . but , if one uses eq.([entropy ] ) , it is only necessary to consider the ground state . so far , two kinds of numerical methods have been principally adopted to extract the value of @xmath6 from eq.([entropy ] ) : one is direct fitting of @xmath9 as a function of @xmath1 for a fixed system length and the other is scaling analysis using a relation @xmath10 , considering two systems with lengths @xmath0 and @xmath11 . however , one would encounter difficulty attributable to a finite - size effect in their practical use . more specifically , in the former method the equation ( [ entropy ] ) is exact only in the thermodynamic limit @xmath12 and some ambiguity remains in the fitting of finite - size result ; @xcite whereas , in the latter method the system length to be studied must be strongly restricted due to the incommensurability ( or frustration ) of spin and/or charge fluctuations as well as the ( open-)shell problem in itinerant systems . therefore , the aim of this paper is to propose a new method that overcomes those problems . we first derive an efficient formula for calculating the central charge . and then , to ascertain the validity of the formula we apply it to a couple of 1d quantum systems ; ( i ) frustrated spin model and ( ii ) itinerant model with strong charge fluctuation . we demonstrate that a very accurate estimation of critical point between tl - liquid and gapped ( or ordered ) states is enabled with this method . furthermore , we suggest that the tl parameter can be calculated in a like manner as the central charge . let us now derive the formula for estimating the central charge . we first prepare @xmath13 and @xmath14 from eq.([entropy ] ) . then , by subtracting the one from the other a simple expression is obtained : @xmath15 } { \ln\left[\cos\left(\frac{\pi}{l}\right)\right ] } \equiv c_1 . \label{cc1}\end{aligned}\ ] ] using this expression we can easily calculate the central charge for a 1d system with a fixed system length . the main advantages of eq.([cc1 ] ) are ; ( i ) the non - universal constant @xmath7 does not appear explicitly , ( ii ) only two values of @xmath9 at the middle of system , where the finite - size correction to eq.([entropy ] ) is the smallest @xcite and the dmrg calculation is the most accurate , are used , and ( iii ) a unique value of @xmath6 is estimated for each system length and it enables us to perform a systematic extrapolation to the thermodynamic limit . also , an alternative expression can be derived from the second derivative of eq.([entropy ] ) with respect to @xmath1 . since @xmath9 is symmetric about @xmath16 , i.e. , @xmath17 , we obtain @xmath18 \equiv c_2 . \label{cc2}\end{aligned}\ ] ] note that eq.([cc2 ] ) can be also derived by applying @xmath19\simeq-\frac{\pi^2}{2l^2}$ ] into eq.([cc1 ] ) . .system - size dependence of @xmath20 and @xmath21 obtained from eqs.([cc1 ] ) and ( [ cc2 ] ) for the 1d @xmath22 frustrated heisenberg model . [ cols="^,^,^,^,^,^ " , ] [ krho ] as a function of @xmath23 for @xmath24 . the system lengths are taken at intervals of @xmath25 sites from @xmath26 to @xmath27 . the red dashed line shows a simple linear fitting . inset : finite - size - scaling analysis of @xmath28 as a function of @xmath29 for @xmath30 and @xmath31 . the blue and green dashed lines are fitting with function @xmath32 . through all the calculations @xmath33 is kept to be odd , and therefore , @xmath34 ( see text ) . ] lastly , we mention that the tl parameter @xmath35 , which is one of the most notable quantity to study the tl - liquid properties , can be estimated in a manner similar to above . the tl parameter is related to the charge - density fluctuation of subsystem with length @xmath1 , @xmath36 [ @xmath37 subsystem ] , like @xcite @xmath38-\frac{(-1)^la}{\left[\frac{l}{\pi}\sin\left(\frac{\pi l}{l}\right)\right]^{2k_\rho}}+f_1 , \label{fluctuation}\end{aligned}\ ] ] where @xmath39 and @xmath40 are non - universal constants . the function @xmath41 behaves similarly to @xmath9 with @xmath1 . then , in the same way as eq.([cc1 ] ) is derived , we easily obtain for @xmath42 @xmath43 \\ & & = \frac{\pi^2\left[f_l\left(\frac{l}{2}-2\right)-f_l\left(\frac{l}{2}\right)\right ] } { \ln\left[\cos\left(\frac{2\pi}{l}\right)\right ] } \equiv k_\rho(l ) . \label{krhol}\end{aligned}\ ] ] since the @xmath44 correction in eq.([fluctuation ] ) oscillates on alternate sites , the set of @xmath45 and @xmath46 is a more proper choice than that of @xmath45 and @xmath47 . the values of @xmath28 estimated from eq.([krhol ] ) at @xmath48 as well as near the critical point are shown in table [ krho ] . with increasing @xmath0 , @xmath28 seems to approach the exact value @xmath49}. \label{krhoexact}\end{aligned}\ ] ] we can easily perform the finite - size - scaling analysis with a fitting function @xmath32 ( @xmath50 ) where @xmath51 and @xmath52 are fitting parameters . from eq.([krhol ] ) , we know @xmath53 ( @xmath34 ) for @xmath54even ( odd ) and expect a simple relation @xmath55 if the logarithmic corrections are small . the fitting analyses are shown in fig . as expected , @xmath56 is obtained at @xmath24 . near the critical point , we have @xmath57 ( @xmath58 ) for @xmath30 ( @xmath31 ) and they have a kind of deviation from the exact values of @xmath35 due to the existence of logarithmic corrections . nonetheless , the extrapolated values are reasonably close to the exact ones ( see table [ krho ] ) . the remaining error in the thermodynamic limit is only about @xmath59 ( @xmath60 ) for @xmath30 ( @xmath31 ) . here , we briefly comment on applying the open boundary conditions ( obc ) . thus far the periodic boundary conditions are assumed ; however , in general the obc are more preferable in the dmrg study for accurate and convenient calculations . in this regard , with applying the obc the function @xmath9 oscillates like eq.([fluctuation ] ) due to the friedel oscillation and the analysis with eq.([cc1 ] ) is unreasonable . if one can suppress the friedel oscillation sufficiently at the center of the system , for example , by using the @xmath61-deformed obc technique , @xcite the central charge might be calculated by taking @xmath62 and @xmath63 as eq.([krhol ] ) even with the obc . a numerical approach to calculate the central charge for 1d quantum system using the entanglement entropy is proposed . this approach is applied to the frustrated spin chain and the half - filled spinless fermions with nearest - neighbor repulsion to check the validity . it is confirmed that the central charge is estimated very accurately for both the models even with short system length . we also suggest a new method for determining the critical point between tl - liquid and gapped ( or ordered ) phases by using the proposed numerical approach . furthermore , we demonstrate that the tl parameter can be calculated in a like manner as the central charge .	the von neumann entanglement entropy is studied with the density - matrix renormalization group technique . we propose a simple approach to calculate the central charge using the entanglement entropy for one - dimensional ( 1d ) quantum system . this approach is applied to a couple of quantum systems : ( i ) 1d frustrated spin model and ( ii ) 1d half - filled spinless fermions with nearest - neighbor repulsion ; and , it is confirmed that the central charge is estimated very accurately for the both systems . also , a new method to determine the critical point between tl - liquid and gapped ( or ordered ) phases from the proposed approach is suggested . furthermore , we mention that the tomonaga - luttinger parameter can be obtained in a like manner as the central charge , using the charge - density fluctuation of a part of the 1d system .
it is a common belief that general relativity , despite its fabulous success in describing our universe at middle and large scale , has to be corrected at short distance . in particular , the apparent tension between einstein s theory and quantum field theory supports the idea that general relativity is merely an effective model that would be replaced in the uv regime by a different theory , and such a _ new _ theory would ultimately permit us to make sense of what we call quantum gravity . the natural scale at which one expects such short distance corrections to manifestly appear is the planck scale @xmath4 , determined by the newton s coupling constant @xmath5 . at present , the most successful candidate to represent a quantum theory of gravity is string theory ( or its mother theory , m - theory ) . in fact , one of the predictions of string theory is the existence of a massless particle of spin @xmath6 whose dynamics at classical level is governed by einstein equations@xmath7 in addition , string theory also predicts next - to - leading corrections to ( baseepsilon ) , which would be relevant at distances comparable with the typical length scale of the theory @xmath8 . these short - distance corrections are typically described by supplementing einstein - hilbert action by adding higher - curvature terms @xcite , correcting general relativity in the uv regime . as a result , the stringy spin @xmath6 interaction turns out to be _ finite _ , and this raises the hope to finally have access to a consistent theory of quantum gravity . to investigate black hole physics in higher - curvature gravity theories , the first question we have to answer is whether such theories actually induce short - distance modifications to the black hole geometry or not . despite expectations that the inclusion of higher - curvature terms in the gravitational action yields modifications to general relativity , it is not necessarily the case that such modifications manifestly appear in the static spherically symmetric sector of the space of solutions . in fact , as we will see below , schwarzschild geometry usually resists modifications . in turn , first it is important to identify the theories of gravity that yield modifications to the spherically symmetric solution . to warm up , let us start by considering a very simple example of higher - curvature term . consider the action@xmath9which corresponds to einstein - hilbert action in four dimensions augmented with the square of the curvature scalar , where @xmath10 is a coupling constant with dimensions of @xmath11=$]length@xmath12 . this action is a particular case of the so - called @xmath13-gravity theories , which are defined by adding to the einstein - hilbert lagrangian a function of the ricci scalar @xmath14 . it is well known that @xmath13-gravity theories are equivalent ( after field redefinition that involves a conformal transformation ) to general relativity coupled to a scalar field @xmath15 , provided a suitable self - interaction potential @xmath16 that depends on the function @xmath17 ( see @xcite and references therein ) . in this sense , these theories are not different from particular models of quintessence . here , we are interested in less simple models ; however , let us consider ( [ quinta ] ) as the starting point of our discussion . a remarkable point is that the theory defined by action ( [ quinta ] ) admits ( anti- ) de sitter - schwarzschild metric as its static spherically symmetric solution . in particular , when @xmath18 the theory still admits the schwarzschild solution even for @xmath19 , and it is due to the property @xmath20 . the theory defined by action ( [ quinta ] ) is not the only theory of gravity that admits schwarzschild metric as a persistent solution . actually , this is a rather common feature of theories with higher - curvature terms . in the case of quadratic terms in four dimensions this is an indirect consequence of the gauss - bonnet theorem , and because the kretschmann invariant @xmath21 is independent from the quadratic scalars @xmath22 and @xmath23 , quadratic deformation of einstein gravity may exclude schwarzschild - tangherlini solution . ] . a second example we can consider is einstein gravity coupled to conformally invariant gravity ; namely @xmath24where @xmath25 is a coupling constant and @xmath26 is the weyl tensor , whose quadratic contraction reads @xmath27 the equations of motion associated to this action read @xmath28where @xmath29 is the bach tensor , @xmath30 it is easy to show that , when @xmath18 , scwarzschild metric solves equations ( [ weeqs ] ) as well . this follows from the fact that bach tensor ( [ eq : bach ] ) vanishes if ricci tensor vanishes , and thus all solutions to general relativity are also solutions of ( [ weeqs ] ) . another example of a modified theory that admits schwarzschild metric as a solution is the jackiw - pi theory @xcite . this theory has recently attracted much attention @xcite . it is defined by the action @xmath31where the function @xmath32 is a lagrange multiplier that couples to the pontryagin density @xmath33 , constructed via the dual curvature tensor @xmath34where @xmath35 is the volume 4-form . the inclusion of the non - dynamical field @xmath32 comes from the fact that the pontryagin density @xmath33 is a total derivative . action ( [ eq : jp ] ) is often called chern - simons modified gravity ; however , this has to be distinguished from the chern - simons gravitational theories we will discuss in the section 2 . it is not hard to see that the equations of motion derived from action ( eq : jp ) are solved by the schwarzschild metric . actually , this is because the pontryagin density @xmath36 of schwarzschild metric vanishes . in contrast , kerr metric has non - vanishing pontryagin form , and thus it is not a solution of jackiw - pi theory . in fact , the rotating solution of this theory has not yet been found , and this represents an interesting open problem as the jackiw - pi theory is considered as a phenomenologically viable correction to einstein theory . summarizing , there are several models that , while representing short distance corrections to general relativity , still admit the schwarzschild metric as an exact solution . in particular , this implies that such models can not be the solution to problems like the issue of the black hole singularity . on the other hand , there are other models which , still being integrable , do yield deviations from general relativity solutions even in the static spherically symmetric sector . in this paper we will be concerned with one of such models . we will study a very special case of higher - curvature corrections to einstein gravity in higher dimensions , and we will see that substantial modifications to schwarzschild solution are found at short distances . in addition to higher - curvature corrections to einstein theory , string theory makes other strong predictions about nature . probably , the most important ones are the existence of supersymmetry and the existence of extra dimensions . in fact , one of the requirements for superstring theory to be consistent is the space - time to have @xmath37 dimensions ; and we learn from our daily experience that six of these extra dimensions have to be hidden somehow . this digression convinces us that studying higher - curvature modification of general relativity in higher dimensions seems to be important to address the problem of quantum gravity , at least within the context of string theory . this is precisely the subject we will study here . more precisely , in this paper we will investigate how the string inspired higher - curvature corrections to einstein - hilbert action modify the black hole physics in the uv regime . this turns out to be a very important question since the black holes are known to be a fruitful arena to explore gravitational phenomena beyond the classical level . the prototypical example we will analyze is @xmath38-dimensional quadratic lovelock lagrangian . but , first , before introducing this theory , let us begin by considering a much more general example . consider the action @xmath39 where the constants @xmath40 @xmath41 and @xmath42 are the coupling constants for each quadratic term . the field equations obtained by varying the action ( [ eq : squad ] ) with respect to the metric read@xmath43 action ( [ eq : squad ] ) is the most general quadratic action one can write down in @xmath38-dimensions . for @xmath44 , the gauss - bonnet theorem permits to fix @xmath45 without loss of generality . in @xmath46 , however , three quadratic invariants are needed to describe the most general lagrangian of this type . for generic values of the coupling constants @xmath40 @xmath47 and @xmath48 , the equations of motion ( [ ewem ] ) are fourth - order differential equations for the metric ( i.e. there are terms prportional to @xmath49 , @xmath50 and @xmath51 ) . nevertheless , a remarkable property of ( [ eq : squad ] ) is that there exists one particular choice of the coupling constants @xmath10 , @xmath47 and @xmath52 that results in the cancellation of all these higher order terms , yielding second order differential equations . it is easy to see that this choice is @xmath53 , which only gives a non - trivial modification to einstein theory for @xmath46 . actually , in @xmath54 and @xmath55 this choice corresponds to lovelock theory ( see ( [ l ] ) below ) ; namely @xmath56 it is worth emphasizing that this choice of coupling constants that yields second order equations of motion is _ unique _ ( up to a free parameter @xmath10 ) . this feature also holds in @xmath38 dimensions , and is a consequence of a more general result known as _ the lovelock theorem _ @xcite . in this paper we will be mainly concerned with the theory defined by action ( [ eq : squad ] ) with @xmath53 . besides the _ uniqueness _ of the choice @xmath53 , which already makes this model interesting in its own right , let us say that this is exactly the effective lagrangian that appears in the low energy action of heterotic string theories in the appropriate frame ( and in m - theory compactifications too ) . now , let us sketch how the five - dimensional lovelock theory arises in the low energy limit of m - theory ( and , consequently , of string theory ) when the theory is compactified from @xmath57 ( resp . @xmath58 ) to @xmath59 . m - theory is supposed to be an extension of string theory ; a fundamental theory that , in certain regime , would flow to string theory @xcite . this mother - theory , if it exists , is yet to be found ; nevertheless , we do know what it has to look like in certain low energy limit : it has to look like eleven - dimensional supergravity augmented with higher - curvature terms . that is , the bosonic sector of the m - theory effective action is given by the graviton @xmath60 ( i.e. the metric ) and the 3-form gauge field @xmath61 ( with field strength @xmath62}$ ] ) . including the pure gravitational fourth order corrections @xmath63(@xmath64 ) , this effective action takes the form . ] @xcite @xmath65 + \text { } ... \notag\end{aligned}\]]where the ellipses stand for the fermionic content and higher - order contributions . these higher order contributions include terms like @xmath66(@xmath67 ) and also couplings of the form @xmath63(@xmath68 @xmath64 ) ; we will not consider these terms here . however , let us mention that the existence of the terms @xmath69 in the action above are related to the terms @xmath63(@xmath68 @xmath64 ) through supersymmetry , although indirectly . the tensor @xmath70 in the third line of ( baseasterisco ) is defined in terms of the way it acts on antisymmetric tensors of second rank , namely @xmath71where tr(@xmath72 ) refers to the trace of @xmath72 . the term in the fourth line in ( [ baseasterisco ] ) is actually one of the terms that appear in the lagrangian of lovelock theory ( see ( [ l ] ) below , where this term is expressed in an alternative way ) . in contrast , the term in the third line , which is of the same order , does not correspond to a term in the lovelock theory . a string theory @xmath69 contribution similar to that of the third and fourth lines of ( [ baseasterisco ] ) also appears in ten dimensions @xcite . this can be written as follows ] see @xcite for a recent discussion on these quartic terms in four dimensions . ] @xmath73 now , let us analyze what happens when the m - theory effective action we discussed above ( including the higher - curvature terms @xmath74 ) is compactified to five dimensions . let us assume we reduce from @xmath57 to @xmath75 by compactifying six of the eleven dimensions in compact calabi - yau ( cy@xmath76 ) threefold . in that case , the effective action of the five - dimensional theory takes the form @xcite @xmath77where we used units such that @xmath78 , and where the coupling @xmath79 is a quantity that depends on the _ details _ of the internal cy@xmath76 manifold are the components of the second chern - class of the @xmath80 calabi - yau space , while @xmath81 are the so - called scalar components of the vector multiplet , which are proportional to the khler moduli of the calabi - yau ; see also @xcite . the quantity @xmath82 is given by the integral of the 6-dimensional extension of the 4-dimensional euler characteristic over cy@xmath83 , namely @xmath84 . in addition , the dimensional reduction of terms @xmath69 gives raise to other corrections , like the shifting of the coefficient of the einstein - hilbert term . ] . in turn , we see that quadratic terms in ( [ ewemm ] ) come from the @xmath74 termstr@xmath85tr @xmath86 . ] of ( [ baseasterisco ] ) . we observe that action ( [ ewemm ] ) resembles a particular case of ( [ eq : squad ] ) , namely the case @xmath54 with @xmath87 , identifying @xmath88 . this is precisely the theory we will study in this paper : the most general quadratic theory of gravity with equations of motion of second order , which , as we have just seen , arises as calabi - yau compactifications of m - theory . we already mentioned that a quadratic action similar to ( [ ewemm ] ) also appears in the 1-loop corrected effective action of heterotic string theory . written in the einstein frame , the coupling of higher - curvature terms in the heterotic effective action is given by @xmath89 , where the dilaton field @xmath15 clearly contributes . black holes solutions in dilatonic einstein - gauss - bonnet theory were studied in refs . @xcite . because an action like ( [ ewemm ] ) also appears in the effective action of the heterotic string , it is also usually referred to as `` string inspired higher - curvature corrections '' . in turn , it represents a nice model to explore the effects of next - to - lading contributions of string theory to gravitational physics . in particular , this five - dimensional ( lovelock ) model of gravity was recently considered in the context of ads / cft holographic correspondence @xcite . actually , one of the applications of the lovelock theory to ads / cft that has attracted attention recently was that of showing that the so - called kovtun - son - starinets bound @xcite may be violated in a theory that contains higher - curvature corrections . the kovtun - son - starinets bound ( kss ) is the conjecture that states : the ratio between the shear viscosity @xmath90 to the entropy @xmath91 of all the materials obey the universal relation@xmath92 in refs . @xcite it was observed that when action ( eq : squad ) with @xmath53 and @xmath93 is considered in asymptotically locally ads@xmath94 space , then the conformal field theory ( cft ) that would be dual to such a theory of gravity would satisfy@xmath95what then would violate ( [ thekss ] ) for @xmath96 . therefore , the kss bound would be violated for all the cfts with a einstein - gauss - bonnet gravity duals with positive @xmath10 , and this is precisely the sign of @xmath97 that comes from string theory . the consideration of five - dimensional lovelock theory as a working example to study the effects of including higher - curvature terms in ads / cft has been an active line of research in the last years . just recently , very interesting papers discussing the interplay between causality and higher - curvature terms in the context of ads / cft appeared @xcite ; ref . @xcite considers the case of @xmath54 lovelock theory . causality in the dual cft constrains the value of the coupling of the quadratic gauss - bonnet term . in terms of this parameter , the permitted range reads @xmath98 . ] . the causality bound comes from demanding that the _ recidivist gravitons _ that hit back the boundary after a bulk excursion do not spoil locality in the cft . the permitted range for the coupling @xmath10 turns out to be @xmath99 the value @xmath100 ( i.e. @xmath101 ) , which is not in this range , corresponds to the chern - simons theory of gravity , which we will discuss in section 2 . at this value , the ration @xmath102 would vanish ( if it were the case that the theory at @xmath101 has a dual description too . other works discussing higher - curvature actions in the context of ads / cft appeared recently . see for instance @xcite , where holographic superconductors in five - dimensional lovelock gravity are considered , showing that higher - curvature corrections affect the condensation phenomenon . besides , the @xmath103 corrections in the non - relativistic version of ads / cft @xcite were also studied , and the @xmath54 lovelock theory is also used in ref . @xcite as the working example for illustrating the renormalization of the dynamical exponent @xmath104 . let us now move to discuss lovelock theory in detail . lovelock theory is the most general metric theory of gravity yielding conserved second order equations of motion in arbitrary number of dimensions @xmath38 . in turn , it is the natural generalization of einstein s general relativity ( gr ) to higher dimensions @xcite . in three and four dimensions lovelock theory coincides with einstein theory lanczos , but in higher dimensions both theories are actually different . in fact , for @xmath46 einstein gravity can be thought of as a _ particular _ case of lovelock gravity since the einstein - hilbert term is one of several terms that constitute the lovelock action . besides , lovelock theory also admits other quoted models as particular cases ; for instance , this is the case of the so called chern - simons gravity theories , which in a sense are actual gauge theories of gravity . on the other hand , lovelock theory resembles also string inspired models of gravity as its action contains , among others , the quadratic gauss - bonnet term , which is the dimensionally extended version of the four - dimensional euler density . this quadratic term is present in the low energy effective action of heterotic string theory @xcite , and it also appears in six - dimensional calabi - yau compactifications of m - theory ; see strominger and references therein . in @xcite zwiebach earlier discussed the quadratic gauss - bonnet term within the context of string theory , with particular attention on its property of being free of ghost about the minkowski space . besides , the theory is known to be free of ghosts about other exact backgrounds @xcite . for a nice and concise review on stringy corrections to gravity actions @xcite see the introduction of @xcite and references therein . for interesting recent discussions on higher order curvature terms see @xcite and related works . the lovelock theory represents a very interesting scenario to study how the physics of gravity results corrected at short distance due to the presence of higher order curvature terms in the action . in this paper we will be concerned with the black hole solutions of this theory , and we will discuss how short distance corrections to black hole physics substantially change the qualitative features we know from our experience with black holes in gr . so , let us introduce the lovelock theory . the lagrangian of the theory is given as a sum of dimensionally extended euler densities , and it can be written as follows @xcite@xmath105where the generalized kronecker @xmath106-function is defined as the antisymmetric product @xmath107}^{\nu _ { n}}.\ ] ] each term @xmath108 in ( [ l ] ) corresponds to the dimensional extension of the euler density in @xmath109 dimensions - dimensional euler density @xmath110 is given by @xmath111m@xmath112 , where , again , we are not considering the boundary terms . ] , so that these only contribute to the equations of motion for @xmath113 . consequently , without lack of generality , @xmath114 in ( [ l ] ) can be taken to be @xmath115 for even dimensions and @xmath116 for odd dimensions . the coupling constants @xmath117 in ( [ l ] ) have dimensions of [ length]@xmath118 , although it is convenient to normalize the lagrangian density in units of the planck scale @xmath119 . expanding the product in ( [ l ] ) the lagrangian takes the familiar form@xmath120where we see that coupling @xmath121 corresponds to the cosmological constant @xmath122 , while @xmath117 with @xmath123 are coupling constants of additional terms that represent ultraviolet corrections to einstein theory , involving higher order contractions of the riemann tensor @xmath124 . in particular , the second order term @xmath125 is precisely the gauss - bonnet term discussed above . the cubic term still has a moderate form @xcite , namely @xmath126the fourth order term @xmath127 coincides with the pure gravitational term in the last line of ( [ baseasterisco ] ) . even though the way of writing lovelock action in its tensorial form ( l2)-([cubo ] ) may result clear to introduce the theory , it is not the most efficient way for most of the calculations one usually deal with . a more convenient way of working out these expressions is to resort to the so - called first - order formalism , which turns out to be useful both for formal purposes and for practical ones . nevertheless , it is important to point out that the first - order formalism is not necessarily equivalent to the second - order formalism , so it should not be regarded merely as a different nomenclature . in the first - order formalism , both the vielbein @xmath128 and the spin connection @xmath129 are considered as independent degrees of freedom , and the torsion acquires in general propagating degrees of freedom @xcite . it is only in the torsion - free sector where both formulations are equivalent ; notice that the vanishing torsion condition is always allowed by the equations of motion ; see @xcite , see also @xcite . we will make use of the first - order formalism at the end of section 2 , as it is almost unavoidable in the discussion of chern - simons theory . however , with the intention to make the exposition as friendly as possible , we will avoid abstruse notation in the rest of the paper . in any case , since we could not afford to give all the definitions necessary to introduce the subject , we will assume the reader is familiarized with basic notions of the theory of gravity and with the standard nomenclature . the paper is organized as follows . in section 2 , we analyze the spherically symmetric black hole solutions in lovelock theory @xcite . in five - dimensions this is given by the boulware - deser solution boulwaredeser , whose most important properties we review . the special properties of electrically charged black holes @xcite are also briefly discussed . in one of the subsections of section 2 , we extend the analysis to those black objects whose horizon geometries correspond to more general spaces of constant ( but not necessarily positive ) curvature @xcite . these are the so - called topological black holes , which can be thought of as black brane solutions of the theory . also , we briefly review the most relevant features of the lovelock black hole thermodynamics @xcite , focusing our attention on the qualitative features that have no analogue in gr . throughout the discussion , the five - dimensional black hole of the einstein - gauss - bonnet theory will serve as prototypical example . in section 3 , we discuss the role of boundary terms @xcite and the junction conditions these yield @xcite . we show how solutions with non - trivial topology can be constructed by a method of a geometric surgery . particular attention is focussed on vacuum wormhole solutions recently found gravaniswillison3 , gggw . finally , we study the spherically symmetric solutions that develop naked curvature singularities . we study these naked singularities with quantum probes and show that , in spite of the divergence in the curvature , these spaces are well - behaved within a quantum mechanical context . let us first consider the theory in five dimensions . since in @xmath130 the @xmath131 term does not contribute to the equations of motion , the five - dimensional lovelock theory basically corresponds to einstein gravity coupled to the dimensional extension of the four dimensional euler density , i.e. the theory that is usually referred as einstein - gauss - bonnet theory ( egb ) . the spherically symmetric static solution of egb theory was obtained by boulware and deser in ref . the metric takes the simple form @xcite@xmath132where @xmath133 is the metric of a unitary @xmath134-sphere , and where the metric function @xmath135 is given by@xmath136with @xmath137 . here we used the standard convention @xmath138 , @xmath139 , and , besides , we have set the newton constant to a specific value for short . from ( [ sol ] ) we notice that there exist two different branches of solutions to the spherically symmetric ansatz ( [ intervalo ] ) , namely @xmath140 and @xmath141 , and this reflects the fact that the equations of motion give a differential equation quadratic in the metric function @xmath135 . as usual , the parameter @xmath142 arises here as an integration constant , and it corresponds to the mass of the solution , up to the factor we absorbed we absorbed a factor @xmath143 where @xmath144 is the surface of the @xmath145-sphere , and where @xmath146 is the newton constant , given by @xmath147 , which has been fixed to a specific values such that @xmath148 . ] in @xmath142 . it is worth mentioning that ( [ intervalo])-([sol ] ) is the most general spherically symmetric solution to egb theory , provided the fact that the metric is smooth everywhere and that the parameters @xmath122 and @xmath97 are generic enough . in turn , a birkhoff theorem holds for this model @xcite . it is important to emphasize that for very particular choices of the set of parameters @xmath117 , degeneracy in the space of solutions can appear , and in those special cases the birkhoff s theorem can be circumvented ; see @xcite for a very interesting discussion . to our knowledge , the most complete analysis of the egb analogue of birkhoff s theorem was performed in @xcite , where the nariai - type solutions @xcite where also discussed . if @xmath96 , the solution corresponding to @xmath141 in ( [ sol ] ) may represent a black hole solution whose horizon , in the case @xmath18 , is located at @xmath149 . on the other hand , as long as @xmath150 and @xmath151 , the branch @xmath140 has no horizon but presents a naked singularity at @xmath152 . solutions @xmath141 and @xmath140 have substantially different behaviors , and only one of them tends to the gr solution in the small @xmath97 limit . in fact , in the limit @xmath153 the branch @xmath154 looks like @xmath155where we see it approaches the five - dimensional ( anti)-de sitter - schwarzschild - tangherlini solution @xcite . on the other hand , in the @xmath153 limit the solution corresponding to the branch @xmath140 behaves like @xmath156and we see it acquires a large effective cosmological constant term @xmath157 . in particular , this implies that microscopic ( a)ds space - time is a solution of the theory even for @xmath18 . this feature was expressed by boulware and deser @xcite by saying that egb theory has its own cosmological constant problem , with @xmath158 . in a sense , the branch @xmath140 is commonly believed to be a false vacuum of the theory , and it is known to present ghost instabilities @xcite ; see also @xcite . the branch @xmath141 , on the other hand , is well - behaved , and it represents short distance corrections to gr black holes ( [ sol1 ] ) . while at short distances the black hole solutions of both theories are substantially different due to the effects of the gauss - bonnet term , in the large distance regime @xmath159 the lovelock black hole ( intervalo ) with @xmath141 behaves like a gr black hole whose parameters @xmath142 and @xmath122 get corrected by finite-@xmath10 subleading contributions @xmath160 ; for instance , the parameter of the mass term gets corrected yielding the effective mass @xmath161 . in the large @xmath162 limit , the next - to - leading @xmath162-dependent contribution to ( [ sol1 ] ) goes like @xmath163 . the damping of this additional term , which in @xmath38 dimensions goes like @xmath164 , is actually strong , and , for distance large enough , it is negligible even in comparison with semiclassical corrections to the metric due to field theory backreaction , which typically go like @xmath165 ( for instance , see @xcite ) . all these features are essentially due to the nature of the gauss - bonnet term , and also hold in higher dimensions . in fact , it is straightforward to generalize solution ( [ intervalo ] ) to the case of egb gravity in @xmath166 dimensions , and the metric is seen to adopt a very similar form @xcite . actually , it is given by simply replacing the element of the @xmath134-sphere in ( [ intervalo ] ) by the element of the unitary @xmath167-sphere @xmath168 , and by replacing the piece @xmath169 in ( [ sol ] ) by @xmath170 . in spite of the non - polynomial form of ( [ sol ] ) , the horizon structure of boulware - deser solution is quite simple , and in @xmath38 dimensions the horizon location is given by the roots of the polynomial @xmath171 where @xmath122 has been appropriately rescaled by a @xmath38-dimensional constant factor . from ( [ fromesta ] ) we observe that the five - dimensional case is actually a remarkable example since , among other special features , it allows to have massive solutions with naked singularities . we mentioned above that if @xmath54 and @xmath18 the black hole horizon is located at @xmath172 , and this implies a lower bound for the spherical solution not to develop a naked singularity , namely @xmath173 . that is , for @xmath174 we do find naked singularities even for the well - behaved branch @xmath141 with positive @xmath142 . for the model with a second order term @xmath175 this only occurs in @xmath54 . in seven dimensions , for instance , the boulware - deser solution with @xmath18 develops horizons at @xmath176 and then the horizon always exists provided @xmath96 , @xmath151 . naked singularities in @xmath177 dimensions usually arise when a term of order @xmath108 is present in the action . so , for the egb theory this only occurs for @xmath54 . another special feature of the ( uncharged ) five - dimensional case is that the metric ( [ intervalo ] ) turns out to be finite at the origin , namely @xmath178 . nevertheless , the curvature still diverges at the origin , although not in a dramatic way . we will return to this point in the last section where we will discuss naked singularities . it could be important to mention that the analysis of the dynamical stability of egb black holes is also special for @xmath54 . the stability analysis under tensor mode perturbations has been explored recently , and it has been shown that the egb theory exhibits some differences with respect to einstein theory ; at least , it seems to be the case for sufficiently small values of mass in five and six dimensions @xcite where instabilities arise ; see also refs . in this sense , the cases @xmath54 and @xmath55 are special ones . @xcite for an interesting recent discussion . on the other hand , let us be reminded of the fact that in @xmath179 dimensions the lovelock action ( [ l ] ) presents also additional terms of higher order @xmath180 , so that in @xmath181 the boulware - deser black hole geometry ( [ intervalo])-([sol ] ) only corresponds to a very special example of lovelock black hole . spherically symmetric solutions in higher dimensions containing an arbitrary higher order terms @xmath108 in ( [ l ] ) can be implicitly found by solving a polynomial equation of degree @xmath145 whose solutions give the metric function @xmath135 ; this was originally noticed by wheeler in wheeler , wheeler2 . moreover , several explicit examples containing arbitrary amount of terms @xmath182 @xmath183 are also known . these correspond to particular choices of the couplings @xmath117 in ( [ l ] ) . one of these explicitly solvable cases corresponds to the chern - simons theory , which exists in odd dimensions . we will briefly discuss this special case below . a remarkable fact is that in the case a term @xmath108 of the lovelock expansion ( [ l ] ) is considered in the action , then the spherically symmetric solution may still take a very simple expression , and , depending on the coupling constants @xmath117 , it may merely correspond to replacing the square root in ( [ sol ] ) by a power @xmath184 ; see @xcite for explicit examples . on the other hand , it is quite remarkable that electrically charged black hole solutions in lovelock theory also present a very simple form . the solutions charged under both maxwell and born - infeld electrodynamics have been known for long time @xcite , and these solutions were reconsidered recently @xcite . in general , the metric function of a charged solution takes the form ( [ sol ] ) but replacing the mass parameter @xmath142 by a mass function @xmath185 that depends on the radial coordinate @xmath162 . function @xmath185 depends on the particular electromagnetic lagrangian one considers . in the case of maxwell theory , and in five dimensions , this function is given by the energy contribution @xmath186 , where @xmath187 represents the electric charge of the black hole , and where the uv cut - off in the integral is absorbed in the definition of the additive constant @xmath188 . more precisely , for charged black holes in einstein - gauss - bonnet - maxwell theory we have @xmath189 , as it was originally noticed by wiltshire @xcite . on the other hand , in the case of black holes charged under born - infeld theory , the function @xmath185 is given by@xmath190where the @xmath191 is the born - infeld parameter , according to the standard form of the lagrangian @xmath192 . in the large @xmath47 limit @xmath193 , and then the metric approaches the charged solution for the maxwell - einstein - gauss - bonnet theory , @xmath194 as expected , the five - dimensional reissner - nordstrm black hole is recovered in the large @xmath162 regime for the case @xmath141 . charged solutions of lovelock theory coupled to born - infled electrodynamics present curious features that are not present in the case of einstein - maxwell theory . perhaps the most relevant one is the existence of single - horizon charged solutions @xcite . besides , lovelock black holes charged under maxwell electrodynamics , and for certain values of the coupling constants @xmath117 , can develop curvature singularities at fixed values of the radial coordinate @xcite , making necessary to exclude a region of the space . this kind of divergence is usually called branch singularity , and it can also be present in uncharged solutions , as it happens for solutions of egb gravity with @xmath195 and @xmath96 , r12,r13 . as in the case of hoffmann s solution in born - infeld - einstein @xcite theory , the lovelock black holes charged under born - infled theory induce a contribution to the mass coming from the finite concentration of electromagnetic energy around the singularity . of course , this happens because both theories coincides at large distances . for finite values of @xmath196 , @xmath185 has a large distance behavior that induces a mass contribution @xmath197 . in particular , this implies that , for certain range of @xmath47 and @xmath187 , naked singularities in five dimensions may arise even for values of the effective mass @xmath198 grater than @xmath10 . notice that the cosmological constant term also acquires a @xmath47-dependent contribution @xmath199 . in the next subsection we will consider a generalization of the black hole solutions reviewed here . we will discuss extended black objects in egb theory . one of the interesting aspects of lovelock theory is that it admits another class of black objects , whose horizons are not necessarily positive curvature hypersurfaces @xcite . these solutions are usually called topological black holes , and their metric are obtained by replacing the @xmath200-sphere @xmath168 in ( [ intervalo ] ) by a base manifold @xmath201 of constant ( but not necessarily positive ) curvature , provided a suitable shifting in the metric function @xmath135 . namely , these solutions read hypersurfaces , correspond to fibrations of ( closed ) base manifolds @xmath202 with non - trivial topology . ] @xmath203where the metric function is now given by @xmath204 with @xmath205 , being its sign that of the curvature of the horizon hypersurface , whose line element is @xmath206 . for @xmath207 the boulware - deser solution ( intervalo)-([sol ] ) is recovered . in general , the base manifold @xmath201 here may be given by a more general constant curvature space : for instance , it can be given by the product of hyperbolic spaces @xmath208 for the case of negative curvature @xmath209 , or merely by a flat space piece @xmath210 . in turn , solutions ( [ siete ] ) correspond to black brane type geometries . such black objects represent fibrations over constant curvature @xmath211-dimensional hypersurfaces , implying that the event horizon , in the cases it exists , is not necessarily a compact simply connected manifold . consider for example the five - dimensional egb theory with negative cosmological constant @xmath212 , and its black brane solution of the form@xmath213with @xmath214where @xmath215 . these objects ( brane - like configurations and topological black holes ) have attracted some attention recently due to their curious properties , and , more recently , these were considered in applications to ads / cft ; see for instance @xcite . in @xcite , an exhaustive classification of static topological black hole solutions of five - dimensional lovelock theory was presented . the authors considered an ansatz such that spacelike sections are given by warped product of the radial coordinate @xmath162 and an arbitrary base manifold @xmath216 , and they showed that , for values of the coupling constant @xmath217 generic enough , the base manifold must be necessarily of constant curvature , and then the solutions of the theory reduce to the topological extension of the boulware - deser metric of the form ( [ siete ] ) . in addition , they showed that for the special case where the coupling @xmath217 is appropriately tuned in terms of the cosmological constant @xmath218 , then the base manifold could admit a wider class of geometries , and such enhancement of the freedom in choosing @xmath216 allows to construct very curious solutions with non - trivial topology . we will return to this point in section 2 . the existence of black holes with generic horizon structure was also analyzed in @xcite , where selection criteria for the base manifold @xmath201 were discussed - dimensional einstein - gauss - bonnet theory . they proved that such a @xmath219-manifold has to obey the equation @xmath220 , where @xmath221 is the weyl tensor in @xmath219 dimensions . ] , and the authors concluded that sensible physical models strongly restrict most of the examples of exotic black holes with non - constant curvature horizons . moreover , the different horizon structures were also studied in @xcite together with its relation to the asymptotic behavior of the corresponding solutions ; see also @xcite . recently , the electrically charged topological black hole solutions were also analyzed , both for the case of the second order lovelock theory in @xcite and for the case of the third order lovelock theory in @xcite . one of the most interesting aspects of these objects with non - trivial horizon geometries is that they enable us to construct a very simple class of kaluza - klein black holes with interesting properties from the four - dimensional viewpoint . for instance , such a solution was recently studied by maeda and dadhich in ref . these kaluza - klein black holes are given by a product m@xmath222h@xmath223 between a four - dimensional manifold m@xmath224 and a @xmath225-dimensional hyperbolic space h@xmath223 . it turns out that the four - dimensional piece of the geometry asymptotically approaches the charged black hole in locally ads@xmath224 space . in turn , the gauss - bonnet term acts by emulating the reissner - nordstrm term for large @xmath162 , while it changes the geometry at short distances r15,r16,r17 . in addition to these solutions , other exotic kaluza - klein lovelock black hole solutions with arbitrary order terms of the form @xmath226 and for a specific values of the coefficients @xmath117 were studied in @xcite . these black holes are different from those studied in @xcite , and are obtained by considering black @xmath227-brane geometries of the form m@xmath228t@xmath229 in the lovelock theory with @xmath230 and @xmath231 . these solutions exist for @xmath232 even , and , in addition , the horizon structure also depends on @xmath145 . analogous toric compactifications of the form m@xmath228t@xmath229 were studied in @xcite , and warped brane - like configurations were also discussed in both @xcite and @xcite . it was shown in @xcite that , in spite of the difference between lovelock theory and einstein theory , the qualitative features of thermodynamic stability of brane - like configurations in both theories are considerable similar , although the higher order terms @xmath108 can be seen to contribute . for example , the thermodynamical analogue of gregory - laflamme transition between black hole and black string configurations was discussed in @xcite . extended string - like objects in lovelock theory and their thermodynamics were also discussed in @xcite . we discuss black hole thermodynamics in the next subsection . the purpose of this section is to describe the general aspects of black hole thermodynamics in lovelock theory . in fact , one of the most interesting features of the lovelock theory regards the thermodynamics of its black hole solutions . this is because it is in the analysis of the black hole thermodynamics where the substantial differences between lovelock theory and einstein theory manifest themselves . pioneer works where the lovelock black hole thermodynamics was discussed in detail are references @xcite ; see also @xcite . in @xcite , jacobson and myers derived a close expression for the entropy of these solutions in @xmath38 dimensions , and they showed that the entropy of these black holes does not satisfy the area law , but contains additional terms that are given by a sum of intrinsic curvature invariants integrated over the horizon . the thermodynamics of charged solutions was originally studied by wiltshire in refs . @xcite , while the thermodynamics of topological black holes was studied more recently , in refs . the study of charged topological black holes in presence of cosmological constant was addressed in @xcite , where the most general solution of this type in egb theory was obtained . references @xcite also analyze topological black holes and their thermodynamics ; see also @xcite . the aim of this section is to discuss the more relevant thermodynamical features of lovelock solutions . to do this , we will consider again the five - dimensional case ( [ intervalo])-([sol ] ) . actually , besides it represents a simple instructive example , the five - dimensional case is also special in what concerns thermodynamical properties . it is the best example to see that substantial differences between lovelock gravity and einstein gravity exist . it is easy to verify that the hawking temperature associated to the solution in @xmath54 with @xmath18 is given by @xmath233 then , we see that , as expected , ( [ t ] ) behaves like the hawking temperature of a gr solution if the black hole is large enough , @xmath234 , going like @xmath235 . on the other hand , temperature tends to zero for small values of @xmath236 , going like @xmath237 . this implies that the specific heat changes its sign at length scales of order @xmath238 , and a direct consequence of this phenomenon is that five - dimensional lovelock black holes turn out to be thermodynamically stable , as they yield eternal remnants . this can be easily verified by considering the rate of thermal radiation which goes like @xmath239 , behaving like @xmath240 at short distances . nevertheless , it is worth pointing out that for dimension @xmath166 the functional form of the temperature is substantially different from the case @xmath241 , as it includes an additional term which is actually proportional to @xmath242 . the general formula reads @xmath243 which implies that , in @xmath166 , the short distance limit is given by @xmath244 , and the specific heat is then negative . this is the reason why the thermodynamic behavior of higher dimensional einstein - gauss - bonnet black holes turns out to be more similar to that in einstein theory if @xmath245 . in general , eternal black holes arise in @xmath246 dimensions if an @xmath247-order term @xmath108 is present in the action . so , let us return to our instructive example of five dimensions . the entropy associated to ( [ t ] ) is given by @xmath248 from what we observe that black holes of lovelock theory do not in general obey the bekenstein - hawking area law . actually , some particular solutions , corresponding to topological black holes with flat horizon geometry @xmath249 , do obey the area law @xcite , but it is not the case for spherically symmetric static solutions . a very interesting discussion on the area law is that of ref . @xcite , where a version of the area law for symmetric dynamical black holes defined by a future outer trapping horizon was derived . there , the authors discussed the differences between the branches of solutions with gr limit and those without it , and argue how for the latter one still can define a concept of increasing dynamical entropy . notice that the second term in the right hand side of ( [ s ] ) implies that if @xmath250 then the entropy turns out to be negative for sufficiently small black holes . this was discussed in @xcite , where it was argued there that an additive ambiguity in the definition of the entropy could be a solution for the negative entropy contributions ; see also the related discussion in @xcite . in any case , the theory for negative values of the coupling constant @xmath10 is somehow pathological in several respects . it not only gives negative contributions to the entropy , but also ghost instabilities and strange causal structure arise if @xmath250 . we will not consider the negative values of @xmath10 here . because of the current interest in black hole thermodynamics of higher order theories , we consider convenient to mention that the entropy function formalism , recently proposed by a. sen @xcite within the context of the attractor mechanism , works nicely for the case of lovelock black holes . in particular , this was recently studied in @xcite for the case of egb black holes , and it was explicitly shown that ( [ s ] ) is recovered by analyzing the near horizon geometry . a rather general analysis was presented in ref . very interesting discussions are those of refs . qwert9,qwert10 . the thermodynamic properties of topological black holes are also very interesting ; see for instance @xcite . as we already mentioned , it can be shown that those black objects whose horizons are of zero curvature do obey the area law for the entropy density . for instance , consider the black brane geometry ( [ brane ] ) , which is solution of the theory with negative cosmological constant , @xmath212 . it is straightforward to check that the hawking temperature of this solution is given by @xmath251 and that the area formula for the entropy density does hold in this special case . remarkably , identical expression for the temperature is obtained in the particular case of the chern - simons theories of gravity , which we discuss in the next subsection . now , let us move on , and analyze a very particular case of lovelock theory which exist in odd dimensions . this is the so - called chern - simons gravity ( cs ) , and can be thought of as a higher - dimensional generalization of the chern - simons description of three - dimensional einstein gravity @xcite . basically , these theories are those particular cases of lovelock lagrangian ( [ l ] ) that admit a formulation in terms of a chern - simons action . as we will discuss , these models are given by a very precise choice of the set of coefficients @xmath117 . to discuss cs gravity theories it is convenient to resort to the first - order formalism which , in spite of its advantage , it is paradoxically avoided in physics discussions . so , let us first review some basic notions : consider the vielbein @xmath252 , which defines the metric as @xmath253 , where we are using the standard notation such that the greek indices @xmath254 correspond to the space - time while the latin indices @xmath255 are reserved for the tangent space . now , consider the 1-form associated to the vielbein , defined by @xmath256 , and the corresponding 1-form associated to the spin connection @xmath129 , defined by @xmath257 . these quantities enable us to define the so - called curvature 2-form , which is given by @xmath258and is related to the riemann tensor by @xmath259 . the torsion - free condition is then given by @xmath260 in this language , local lorentz invariance of the theory is expressed in terms of the covariant derivative @xmath261where @xmath262 represent the parameters of the transformation . the remarkable fact is that , for particular cases of the action ( [ l ] ) , if the coupling constants are chosen appropriately , the theory exhibits an additional local symmetry . for instance , if we consider the case @xmath18 , such additional symmetry turns out to be given by the invariance of the lagrangian density under the gauge transformation@xmath263 that is , the cs theory possesses a local symmetry under gauge transformation @xmath264 , with @xmath265 being a parameter . this is actually an off - shell local gauge symmetry of the theory ( [ l ] ) that arises for special choices of the coupling constants @xmath117 , as far as the boundary conditions are also chosen in the appropriate way . besides , it can be easily verified that transformation ( [ dos ] ) , once considered together with ( [ uno ] ) , satisfies the poincar algebra @xmath266 , and this is why these theories are usually referred as poincar - chern - simons gravitational theories @xcite ; see also @xcite for an excellent introduction to chern - simons gravity . so , let us specify which are the theories that possess the gauge symmetry like ) is the transofrmation that corresponds to the case @xmath18 . the analogous tranformation for the case @xmath267 takes a slightly different form , see @xcite . ] ( [ uno])-([dos ] ) , namely the cs theories . to do this , first it is convenient to rewrite the lovelock lagrangian . in the first - order formalism , the lovelock action corresponding to ( [ l ] ) in @xmath116 dimensions can be written as @xmath268 where @xmath269 correspond to @xmath114 independent coefficients that are a rearrangement of the coefficients @xmath117 . in ( [ mn ] ) , the convention is such that the @xmath270 coupling @xmath271 has been set to @xmath272 ( or , alternatively speaking , it has been absorbed in the definition of the curvature @xmath273 ) , so that in this notation we have @xmath274 . it is worth noticing that , in order to represent the most general form of ( [ l ] ) , the coefficients @xmath269 in ( [ mn ] ) should be allowed to take complex values . in fact , lovelock action ( [ l ] ) with real coefficients @xmath117 can correspond to ( [ mn ] ) with imaginary @xmath275 . an example is given by the five - dimensional theory whose action reads @xmath276 , which leads to the particular form of ( [ l ] ) where no einstein - hilbert contribution is present , but only the cosmological constant and the gauss - bonnet term appear , with @xmath277 for a real @xmath47 . the cs gravity theories , however , are given by real values of @xmath269 . more precisely , cs theory correspond to the special case where the coupling @xmath278 in ( [ mn ] ) combine to give only one value for the effective cosmological constant @xmath279 . in terms of the lagrangian density ( [ l ] ) this corresponds to taking the coupling constants @xmath117 to be @xmath280 for@xmath281 , while @xmath121 is given by the cosmological constant @xmath282 . it is important to mention that ( [ mn ] ) corresponds to the case of negative cosmological constant , which yields the cs theory with the ads@xmath283 group ( i.e. the group @xmath284 ) as the one that generates the gauge symmetry . the case of positive @xmath122 is simply obtained by changing @xmath285 , while the poincar invariant theory is obtained through the inonu - winger contraction of ( a)ds group ; see zanelli for details . an example of poincar invariant cs is given by the lagrangian containing only the quadratic gauss - bonnet @xmath286 term in five dimensions , without the einstein - hilbert term and with @xmath287 . as it is well known , an example of the cs gravity theory is given by three - dimensional einstein theory , whose action.],@xmath288admits to be formulated as a cs theory . to see this , and then extend the construction to higher dimensional cases , let us first point out that ( quince ) can be written as follows,@xmath289with @xmath290 . it turns out that ( [ quince])-([secasa ] ) admits to be formulated as a cs theory @xcite for the groups @xmath291 , @xmath292 and @xmath266 , depending on whether the cosmological constant @xmath122 is negative , positive or zero , respectively . to make contact with the usual form of the cs action , let us introduce a ( @xmath293)-dimensional 1-form @xmath294 whose indices run over @xmath295 ( recall @xmath116 ) , and its strength field @xmath296 , which are given by@xmath297that is , @xmath298 for @xmath299 , while @xmath300 for @xmath301 . analogously , @xmath302 for @xmath303 , while @xmath304 for @xmath305 . then , making use of these definitions , ( [ quince])-([secasa ] ) can be alternatively expressed in its chern - simons form@xmath306where the trace is over the indices @xmath307 that run from @xmath308 to @xmath134 ( corresponding to @xmath309 , i.e. @xmath310 ) . local symmetry under ( [ uno ] ) and ( [ dos ] ) is then gathered by gauge symmetry of ( [ scs ] ) . the next example we could consider is the five - dimensional one , which corresponds to the lovelock theory ( [ l ] ) for the particular case @xmath311 ( i.e. @xmath312 ) . then , the action reads @xmath313 where @xmath314 and @xmath315 . this can be also written as@xmath316and , again , it admits to be written in its chern - simons form @xmath317 actually , this structure goes on as @xmath38 increases , and it expands a whole family of theories which , still being particular cases of lovelock theory ( [ l ] ) , represent odd - dimensional field theories with local off - shell symmetry under the ( a)ds ( or poincar ) group . now , once we have introduced the theories , let us analyze their black hole solutions . going back to solution ( [ intervalo ] ) , and considering again the five - dimensional case as an example , we observe that replacing the chern - simons condition and @xmath212 . ] @xmath318 in the metric function ( [ sol ] ) leads to a rather different geometry , given by @xmath319 this solution still may represent a black hole , provided @xmath320 , with the horizon located at @xmath321 . however , this is a black hole of a different sort . in particular , it does not present a limit where gr is recovered , and this can be understood in terms of the condition @xmath322 in the following way : while the cosmological constant @xmath122 introduces an infrared cut - off ( the length scale @xmath323 ) where the cosmological term dominates over the einstein - hilbert term , the gauss - bonnet term introduces an ultraviolet cut - off ( the length scale @xmath324 ) where the quadratic terms dominate . therefore , the condition @xmath322 basically states that in chern - simons theory both length scales are of the same order , and consequently there is no range where the einstein - hilbert term is the leading one . this explains why there is no range where ( [ sol2 ] ) approaches schwarzschild - tangherlini solution . this asphyxia of the einstein - hilbert term is a typical feature of chern - simons theories for @xmath325 , where a unique free parameter @xmath326 appears in the action . the hawking temperature associated to black hole solution ( [ sol2 ] ) is given by@xmath327which in turn agrees with ( [ agree ] ) , although now it corresponds to a spherically symmetric solution . as it is well known @xcite in @xmath328 formula ( [ naitin ] ) agrees with the area law . certainly , solution ( [ sol2 ] ) is reminiscent of the baados - teitelboim - zanelli three - dimensional black hole ( btz ) , which , after all , also corresponds to a cs black hole . in fact , this is not a coincidence , and regarding this , let us make a historical remark : it turns out that , even though one could imagine that cs black holes ( [ sol2 ] ) were discovered as higher - dimensional extensions of the btz , the story was precisely the opposite : in 1992 , baados , teitelboim and zanelli discovered the btz as a particular case of a family of lovelock black holes they were studying at that time @xcite . the analogy between the btz black hole and those solutions for higher - dimensional cs theories was discussed in detail in @xcite . in particular , it was emphasized there that five - dimensional solution ( sol2 ) shares several properties with its three - dimensional analogue . for instance , it is the case of their thermodynamics properties , which , after all , are actually encoded in the function @xmath135 . this is also why all cs black holes have infinite lifetime . notice that the parameter @xmath329 in eq . ( [ sol2 ] ) plays the role that the mass @xmath142 plays in the btz solution . also , as in the three - dimensional case , the anti - de sitter space is obtained for a particular value of this parameter , namely @xmath330 , and a naked singularity is developed for the range @xmath331 . in @xcite the cs black holes and their dimensional extensions were exhaustively studied , together with their topological and charged extensions . there , a very interesting class of black holes was found by considering the particular choice of coefficients that leads to the @xmath332-dimensional cs theory , but dimensionally extending the action from @xmath333to @xmath334 . the metrics of such solutions are given by replacing the constant @xmath329 in ( [ sol2 ] ) by the quantity @xmath335 . a further generalization of the solutions of @xcite would be given by adding a volume term to the gravitational action , which in turn corresponds to shifting the coupling @xmath336 but keeping the rest of @xmath337 tuned as they are in the ( @xmath338)-dimensional cs theory , given in terms of the length scale @xmath326 . the solution for this case is given by replacing the constant @xmath329 in ( [ sol2 ] ) by a term @xmath339 , where @xmath340 . these black holes do have a gr limit since now the cosmological length scale can be pushed away by choosing @xmath341 appropriately . it is also important to mention that black hole solution ( [ sol2 ] ) is also a solution of the cs theory with torsion giacomini , giacomini2,qwert11 . the solutions of chern - simons theory are very special ones , and this is due to the fact that for that specific choice of the coupling constants @xmath342 the equations of motion of lovelock theory somehow degenerate . in particular , it is remarkable that the obstruction imposed by birkhoff - like theorems does not hold for cs theories . before concluding this section , a word on the spinning black hole case : the problem of finding a rotating solution in einstein - gauss - bonnet gravity , which would generalize the kerr s spinning black hole of gr , is a hard and still unsolved problem . recently , it was proven in @xcite that the kerr - schild ansatz does not work in lovelock theory ( except for very special cases as einstein theory and chern - simons theory ) , and this manifestly shows how difficult this classical problem is . nevertheless , despite the difficulty , some advances in this area were recently achieved : in @xcite an exact analytic rotating solution was found for chern - simons gravity in five dimensions . this einstein - gauss - bonnet solution , however , does not present a horizon , and thus it does not represent a black hole . nevertheless , the numerical analysis of @xcite supports the idea that rotating solutions actually exist . besides , approximated analytic solutions at first order in the angular momentum parameter were found in @xcite . other solutions are known which represent rotating flat branes ; these are a simple extension of topological black holes with @xmath343 . despite these recent advances , the problem of finding an exact analytic rotating black hole solution in lovelock theory still remains an open problem . in this section , we will discuss other constructions which , locally , coincide with the deser - boulware spherically symmetry metric . the next class of solutions we would like to discuss is a class of vacuum solutions of lovelock theory which represents wormhole geometries that connect two disconnected asymptotic regions of the space - time . recently , several examples of such solutions were found @xcite , describing vacuum wormholes with different asymptotic behaviors , and in different number of dimensions . so , the first question we might ask is : why do wormholes exist in lovelock theory ? the main reason why vacuum wormholes exist in a theory like ( [ l ] ) is actually simple , and it can be heuristically explained as follows : consider the equations of motion corresponding to lagrangian ( [ l ] ) , which can be always written as@xmath344where the higher order terms act as an effective stress tensor that here we denoted @xmath345 . in the case of egb theory it reads@xmath346where , as usual , @xmath347 , @xmath348 . the key point is that this effective stress tensor @xmath349 , thought of as a kind of matter contribution , can be shown to violate the energy conditions for @xmath10 large enough . actually , this does not represent an actual problem from the conceptual point of view since this `` matter '' is actually made of pure gravity . however , a consequence of this violation of the energy conditions is that eqs . ( [ eco ] ) allow the existence of vacuum wormhole solutions at scales of order @xmath324 , unlike the case of gr , where solutions of this sort require the consideration of _ exotic matter_. furthermore , there is a second reason for such curious solutions to exist in lovelock theory . as mentioned above , when the coefficients @xmath117 in ( [ l ] ) correspond to the cs theory , the space of solutions experiments an unusual enhancement , which translates into a large degeneracy of the metric of spaces with enough symmetry . roughly speaking , for such particular cases , lovelock theory is somehow degenerated enough to admit metric with very special properties , and wormholes are some of them . nevertheless , here we will focus our attention on wormhole solutions that exist in five - dimensional egb theory without requiring the coefficients @xmath350 and @xmath10 to be those that correspond to cs theory . therefore , the existence of such solutions , regarded as an anomaly , is ultimately attributed to the issue of the energy conditions mentioned above . the particular configurations we will consider are the so - called thin - shell wormholes , which correspond to connecting two regions of the space through a codimension - one hypersurface that plays the role of the wormhole throat . for such a geometry to be constructed , we have to make use of the junction conditions of the egb theory @xcite . in particular , we will consider the configuration of two boulware - deser spaces connected through a hypersurface on which the induced stress - tensor vanishes . such geometries are not possible in gr , where wormholes require the energy conditions to be violated on the thin - shell . however , in lovelock theory , and because of the higher order terms , spherically symmetric vacuum wormholes with positive mass can be constructed , as shown by gravanis and willison in @xcite . let us review the procedure here . let @xmath351 be a four - dimensional timelike orientable hypersurface of codimension one , whose normal vector is denoted by @xmath352 . suppose @xmath353 separates two regions of the space , which we call @xmath354 and @xmath355 . then , junction conditions read@xmath356where @xmath357 denotes the jump of the quantity @xmath358 across the hypersurface @xmath351 , which means @xmath359 , where the sign @xmath360 depends on the relative orientation of the regions . above , tensor @xmath361 represents the induced stress - tensor on the hypersurface @xmath351 , in complete analogy with the israel junction conditions in einstein theory . in fact , we see that the first two terms in ( [ jc ] ) actually correspond to the israel junction conditions constructed with the extrinsic curvature @xmath362 and its trace @xmath363 . in addition , the junction conditions corresponding to the egb theory contains contributions cubic in the extrinsic curvature,@xmath364and also contributions that involve the riemann curvature tensor of the hypersurface @xmath365 the notation used here is such that latin indices @xmath366 refer to coordinates on the four - dimensional hypersurface that separate the two five - dimensional regions of the space . the induced metric is denoted by @xmath367 . it is worth mentioning that in ref . @xcite the junction conditions were studied in the most general case , including the case of space - like junctures , which corresponds to a cosmological - type geometries that experiment a change of behavior at a given time characterized by the hypersurface @xmath351 . it was pointed out by h. maeda that this kind of space - like junction conditions could be used to construct regular black hole solutions by means of geometric surgery procedure inside the black hole horizon . here we will be mainly concerned with static spherically symmetric geometries , and , besides , with spherically symmetric boundary conditions . that is , we will consider the time - like hypersurface @xmath351 that separates the two regions of the space to be located at fixed radial coordinate @xmath368 , and the system of coordinates we will parameterize the three angular directions @xmath369 of the junction hypersurface , and the proper time @xmath370 of an observer on @xmath351 . then , we introduce the metrics @xmath371on each region @xmath354 and @xmath355 , and the two regions join at @xmath368 . since here we consider vacuum solutions , @xmath372 and @xmath372 are given by ( [ siete ] ) ( or by ( [ sol ] ) in the case @xmath373 ) . in general , there is no reason for the mass parameters @xmath374 of the two regions to be equal , and the same happens with the choice of the branches @xmath375 . moreover , the orientation of @xmath376 and that of @xmath355 with respect to the normal vector @xmath377 are also independent one on each other , and we will take this degree of freedom into account by introducing the variables @xmath378 and @xmath379 which indicate whether in each region the radial coordinate @xmath380 is parallel ( @xmath381 ) or anti - parallel ( @xmath382 ) to @xmath352 . therefore , wormhole - like geometry corresponds to the orientation @xmath383 , while the standard shell - likegeometry corresponds to the case @xmath384 . the freedom in choosing the parameters @xmath142 , @xmath385 , @xmath90 independently in each region allows for a wide class of solutions . the whole catalog was recently studied in @xcite . the metric on @xmath351 induced from region @xmath354 is the same as the one induced from region @xmath355 , and is given by@xmath386where , according to ( [ siete ] ) , @xmath387 will be chosen to be the line element of a 3-manifold with ( intrinsic ) curvature @xmath388 , i.e. it is a unit sphere , a hyperboloid or flat space respectively . the hypersurface @xmath351 is the world - volume of the juncture where regions @xmath354 and @xmath355 join . to see whether such a wormhole - like ( or shell - like ) configuration is possible in vacuum , we have to solve junction conditions ( [ jc ] ) with @xmath389 . to do this we first need to compute the components of the intrinsic curvature . these are given by from these equations we see that the radius of the throat of the wormhole is given by@xmath400and from this we can also calculate the mass of the wormhole easily . eq . ( [ mmm ] ) implies that , in the case of the spherically symmetric wormhole ( @xmath401 ) , we need @xmath402 for the wormhole to exist . then , provided @xmath403 is of order one , the radius of the wormhole throat is or order @xmath404 . besides , we should ask the throat to be located outside the horizon , namely @xmath405 . it is remarkable that all these conditions can be satisfied @xcite for positive values of @xmath10 , @xmath406 , @xmath142 and @xmath122 . however , it is worth mentioning that spherically symmetric wormhole solutions only exist if at least one of the two regions @xmath407 corresponds to the branch @xmath408 in ( [ sol ] ) . more remarkable is the fact that the analysis of the dynamic case @xmath409 follows straightforward . when @xmath410 , equations @xmath411 and @xmath412 are not linearly independent , and it is sufficient to solve the first of them . considering @xmath401 , we get@xmath413which has the form of a one - dimensional dynamic equation of motion constrained by the vanishing energy condition . notice that ( [ rr ] ) is recovered by demanding @xmath414 . notice also that the effective potential @xmath415 has negative derivative , and for large values of @xmath416 it goes like @xmath417 . the effective potential @xmath418 can be positive and of positive derivative for non - symmetric wormhole configurations . summarizing , we have just seen that spherically symmetric ( microscopic ) thin - shell wormholes in vacuum are admitted as solutions of the five - dimensional lovelock theory . these solutions are allowed by additional terms arising in the junction conditions of the egb theory . it is worth mentioning that the one we discussed here is not the only class of wormhole - like solutions that exists in lovelock theory . for instance , in @xcite a static wormhole solution for gravity in vacuum was found for cs gravity in arbitrary ( odd ) dimensions @xmath419 . this wormhole connects two asymptotic regions whose respective boundaries are locally given by r @xmath420 @xmath421 h@xmath422 . besides , @xmath38-dimensional static wormhole solutions of the egb theory were also studied in @xcite , and explicit wormhole solutions respecting the energy conditions in the whole spacetime were found for the case @xmath96 . the asymptotic behavior of these solutions is given by r @xmath420 h@xmath423 . as we have seen in the previous sections , there are many features of lovelock solutions that are not present in gr . eternal black holes and wormholes are remarkable examples . another example is the existence of positive mass solutions with naked singularities . in fact , naked singularities appear in all the catalog of solutions , for both spherically symmetric and extended objects , for both solutions with a suitable gr limit and solutions without it . but , what kind of naked singularities are these ? for instance , we could ask whether these are stable under gravitational perturbations @xcite ; or whether these turn out to be `` bad '' singularities when probed with wave functions @xcite . regarding the question about the stability , this issue was studied recently within the framework of the kodama - ishibashi formalism , and some evidence of instabilities was found @xcite . on the other hand , here we will address the second question , the one about how these naked singularities look like when analyzed with quantum probes . to do this we will employ the method developed by horowitz and marolf in ref . @xcite , based on the pioneer work of wald @xcite . the basic idea es the following : unlike what happens in the classical regime , where a singular space is defined by the concept of geodesic incompleteness , in the quantum mechanical regime the singular character of the space - time is defined in terms of the ambiguity in the definition of the hamiltonian evolution of wave functions on it @xcite . more specifically , the singular nature of a given space is determined in terms of the ambiguity when trying to find a self - adjoint extension of the hamiltonian operator to the whole space . when such self - adjoint extension exists and is unique , then it is said that the space is quantum mechanically regular , in spite of the singularities it might present at classical level . notice that this is not matter of deforming the space or somehow resolving it , but it is rather a reconsideration of what is the relevant physical dynamics on it . in fact , a space can be classically singular but still regular when it is analyzed with quantum probes . here , we will apply the concept of quantum probes to the singular solutions of lovelock theory discussed above . but , first , let us review the method developed in @xcite . consider the quantum dynamics of a scalar field @xmath424 on the spherically symmetric space ( [ intervalo ] ) , which is governed by the klein - gordon equation@xmath425this equation can be written as follows@xmath426where @xmath427 is the covariant derivative on the spacelike hypersurfaces defined by constant @xmath114 foliations , and where the metric function @xmath135 is given by ( [ sol ] ) . the piece @xmath428 in ( [ adove ] ) involves the laplacian operator on the unitary 3-sphere , whose eigenvalues are known to be given by @xmath429 with positive integers @xmath430 now , equation ( [ adove ] ) can be written in its schrdinger - like form , schematically , @xmath431and then the problem to deal with is to decide whether the hamiltonian operator @xmath432 admits a unique self - adjoint extension in spite of the fact the space is singular at the origin @xmath152 . as mentioned , in the quantum mechanical context the existence of singularity is associated to the non - existence of a unique self - adjoint extension of the hamiltonian operator rather than to a geodesical completeness . then , the problem of determining whether the space is regular is translated into the problem of verifying whether @xmath433 admits a unique self - adjoint extension @xmath434 . if such extended operator exists , then the hamiltonian evolution of the wave function in this space would be given by@xmath435and it would be well - defined . fails to be of finite norm near the origin for any value of @xmath438 and for any of the two possible signs @xmath360 in ( [ differential ] ) ; see horowitzmarolf for details . in other words , for the space to be considered regular quantum mechanically it is necessary to see that at least one solution @xmath15 to ( [ differential ] ) is non - normalizable around the origin . this criterion strongly depends on which norm @xmath439 is considered . the well - posedness of an initial value problem requires not only the existence and uniticity of conditions , but also continuous dependence of solutions on initial data . then , the norm @xmath439 to be considered should select a the function space that fulfills these requirements . a sensitive norm in this sense is the sobolev norm @xcite . to see how the method works in the case we are interested in , let us consider again the five - dimensional boulware - deser space ( [ intervalo])-([sol ] ) . the branch @xmath140 of this space presents a naked singularity for all positive values of @xmath142 , while the branch @xmath141 only presents naked singularities within the range @xmath174 . then , let us solve the wave equation for these spaces . to analyze the solutions of ( [ differential ] ) near the singular point @xmath152 it is convenient to write this equation as @xmath440 , with @xmath441 and @xmath442 being two functions analytic at the origin . this is a fuchsian equation and so it admits solutions with the form @xmath443 for certain analytic function @xmath444 and a complex number @xmath90 that is known to solve the indicial equation @xmath445 . then , replacing ( [ sol ] ) in ( [ differential ] ) we find @xmath446 , @xmath447 , and two independent solutions to ( [ differential ] ) are then given by the two values of @xmath90 that solve @xmath448 . therefore , we find that one of the solutions to ( [ differential ] ) always diverges at least as rapidly as @xmath449 , and so it fails to be integrable with respect to the sobolev norm . summarizing , there exists a unique self - adjoint extension @xmath434 , from what we conclude that five - dimensional boulware - deser metric turns out to be regular when tested by quantum probes . it is remarkable that the positive ( but small ) mass solutions of five - dimensional black holes are in a sense regular quantum mechanically , despite the naked curvature singularity they exhibit at the origin . before concluding , we wish to make a remark about the consistency of studying naked singularities in this way actually , one could wonder whether probing naked singularities in a theory with a finite higher curvature expansion makes sense or not . for instance , in string inspired models , as soon as one approaches the singularity , neglecting higher order corrections seems to be impossible since higher and higher order terms start to dominate as we go close enough to the singularity . however , let us argue here that , even though this is true , this is not necessarily an obstruction for testing singularities with quantum probes up to certain order in the higher curvature expansion . let us be reminded of what we do when we solve the schrdinger equation for the coulombian potential ( e.g. in fact , the analogy with the hydrogen atom in quantum mechanics is quite good since such problem also corresponds to solving a wave function equation in presence of a central potential whose classical counterpart breaks down at the origin ) . in quantum mechanics , even though the coulombian potential diverges at the origin , we know that the quantum problem still makes sense , and we do solve the wave equation without complaining about the fact that other corrections to the potential ( e.g. effective screening due to quantum effects , or short distance corrections to the coulombian potential ) could in principle appear at very short distances . heuristically speaking , what one really has to do to make sure the whole procedure makes sense is comparing the typical size of the wave packet with the length scale where the terms that were neglected would dominate . for example , above we were dealing with the egb action , and the terms @xmath450 were certainly neglected , and so the analysis carried out could still make sense as long as the compton length of the wave packet is small enough in comparison with the length scale imposed by the coupling constant @xmath342 with @xmath180 , and provided the fact higher curvature terms act as a perturbation . for some particular models where the couplings @xmath117 are given in terms of the same fundamental scale ( like the models inspired in string theory where the scale is given by @xmath451 ) the story could be a little more subtle , and so the argument above would not be valid in general . nevertheless , it is likely the case that higher order terms would contribute by smoothing out the singularity even more , although not necessarily resolving it in a classical sense . this work was supported by uba , conicet , and anpcyt , through grants ubacyt x861 , pip6160 , pict34557 . conversations with m. aiello , a. anabaln , e. ayn - beato , m. baados , c. bunster , f. canfora , g. dotti , r. ferraro , a. garbarz , m. hassane , d. hofman , a. giacomini , j. giribet , m. kleban , d. mazzitelli , r. olea , j. oliva , j. saavedra , and d. tempo are acknowledged . the authors specially thank e. gravanis and s. willison for collaborations in the subject , and they are grateful to j. edelstein , m. leston , h. maeda , r. troncoso , and j. zanelli for reading the manuscript and making very important remarks . the authors also thank the members of the ccpp at new york university for the hospitality . thanks the people of the brandeis theory group of brandeis university . f. moura and r. schiappa , _ higher - derivative corrected black holes : perturbative stability and absorption cross - section in heterotic string theory _ , class . . grav . * 24 * ( 2007 ) 361 , [ arxiv : hep - th/0605001 ] . c. sahabandu , p. suranyi , c. vaz and l. wijewardhana , _ thermodynamics of static black objects in d dimensional einstein - gauss - bonnet gravity with d-4 compact dimensions _ , phys * d73 * ( 2006 ) 044009 , [ arxiv : gr - qc/0509102 ] . m. cvetic , s. nojiri and s. odintsov , _ black hole thermodynamics and negative entropy in de sitter and anti - de sitter einstein - gauss - bonnet gravity _ , b628 * ( 2002 ) 295 , [ arxiv : hep - th/0112045 ] . s. nojiri , s. odintsov and s. ogushi , _ friedmann - robertson - walker brane cosmological equations from the five - dimensional bulk ( a)ds black hole _ , int a17 * ( 2002 ) 4809,[arxiv : hep - th 0205187 ] . f. canfora , a. giacomini and r. troncoso , _ black holes , parallelizable horizons and half - bps states for the einstein - gauss - bonnet theory in five dimensions _ , to appear in phys . rev . * d * , [ arxiv:0707.1056 ] . a. anabalon , n. deruelle , y. morisawa , j. oliva , m. sasaki , d. tempo and r. troncoso , _ kerr - schild ansatz in einstein - gauss - bonnet gravity : an exact vacuum solution in five dimensions _ , to appear in class . ( 2008 ) , arxiv:0812.3194 [ hep - th ] . m. nozawa and h. maeda , _ effects of lovelock terms on the final fate of gravitational collapse : analysis in dimensionally continued gravity _ , class . * 23 * ( 2006 ) 1779 , [ arxiv : gr - qc/0510070 ] .	lovelock theory is a natural extension of einstein theory of gravity to higher dimensions , and it is of great interest in theoretical physics as it describes a wide class of models . in particular , it describes string theory inspired ultraviolet corrections to einstein - hilbert action , while admits the einstein general relativiy and the so called chern - simons theories of gravity as particular cases . recently , five - dimensional lovelock theory has been considered in the literature as a working example to illustrate the effects of including higher - curvature terms in the context of ads / cft correspondence . here , we give an introduction to the black hole solutions of lovelock theory and analyze their most important properties . these solutions can be regarded as generalizations of the boulware - deser solution of einstein - gauss - bonnet gravity , which we discuss in detail here . we briefly discuss some recent progress in understading these and other solutions , like topological black holes that represent black branes of the theory , and vacuum thin - shell wormhole - like geometries that connect two different asymptotically de - sitter spaces . we also make some comments on solutions with time - like naked singularities . brxth596 + @xmath0 brandeis theory group , martin fisher school of physics _ brandeis university , waltham , ma 02454 - 9110 . _ @xmath1 instituto de astronoma y fsica del espacio , conicet _ ciudad universitaria , c.c . 67 suc . 28 , 1428 , buenos aires , argentina . _ @xmath2 department of physics , universidad de buenos aires and conicet _ ciudad universitaria , pabelln i , 1428 . buenos aires , argentina . _ @xmath3 centro de estudios cientficos , cecs , valdivia , chile _ arturo prat 514 , valdivia , chile .
the principal objects of interest in this paper are * dephased complex hadamard matrices * : matrices @xmath1 such that @xmath2_{i,1 } & = [ h]_{1,i } = 1 \text { for } i=1,\ldots , d.\end{aligned}\ ] ] here @xmath3 is the @xmath4 identity matrix , @xmath5 denotes conjugate transpose , and @xmath6 is the complex unit circle . we will adhere to the common practice of declaring two hadamards @xmath0 and @xmath7 equivalent if there exist unitary diagonal matrices @xmath8 and @xmath9 and permutation matrices @xmath10 and @xmath11 such that @xmath12 complex hadamards are natural generalizations of their real analogue : square matrices with entries in @xmath13 with mutually orthogonal rows and columns , discovered to have the largest determinant among all real matrices whose entries have absolute values bounded by one @xcite . while both real and complex hadamards are mathematically interesting objects in their own right , they also have a wide range of applications including uses in coding theory @xcite , the design of statistical experiments @xcite , numerous constructions in theoretical physics @xcite and quantum information theory @xcite . the focus of much of the current mathematical research is aimed at complete classification of equivalence classes of complex hadamards , at least for small dimensions @xcite . in dimensions @xmath14 , complex hadamards have been completely described @xcite , while the classification of @xmath15 hadamards remains open if not on the precipice of completion @xcite . towards understanding the spaces of complex hadamards , numerous methods of construction have been developed which give explicit families of matrices @xcite . often one imposes additional assumptions on the structure of the matrices , thereby simplifying the problem and allowing algebraic conditions to be solved explicitly . to study the dimension of the space of complex hadamards near a fixed matrix @xmath0 , the defect @xmath16 can be computed @xcite . this quantity ( defined in section [ sec : theory ] ) bounds the dimension of affine orbits stemming from a dephased hadamard @xcite . notably , if @xmath17 , then there exists a neighborhood of @xmath0 which does not contain any other dephased hadamards , and @xmath0 is said to be * isolated. the main contributions of this paper include 1 ) introducing an equivalent definition of the defect using classical dynamical systems theory , 2 ) demonstrating through examples how this new framework can be used to investigate the local structure of complex hadamards and , in particular , 3 ) constructing several new affine families stemming from known hadamards . this paper is organized as follows : in section [ sec : theory ] we motivate and define our methodology . section [ sec:4x4 ] serves to justify and explicate the technique by applying it to the space of @xmath18 hadamard matrices . in section [ sec : examples ] we use our approach on a host of hadamard matrices where the local structure is not fully understood to demonstrate both its versatility and limitations . we end with some discussion of the scope of this new perspective . fix an integer @xmath19 and consider a dephased matrix of the form @xmath20 depending on @xmath21 $ ] with @xmath22 . if @xmath23 is a dephased hadamard , then @xmath24 . naturally this imposes conditions on the allowed phases in the entries of @xmath23 . in particular , @xmath25 must be chosen to satisfy @xmath26 equations stemming from the requirement that the off - diagonal entries of @xmath27 must be identically 0 : @xmath28_{i , j } = \begin{cases } 0 , & \mbox{if } i \neq j \\ d , & \mbox{if } i = j \end{cases}. \label{eq : potentialdef}\ ] ] using the equations in ( [ eq : potentialdef ] ) we define a scalar potential @xmath29 , which can be thought of as measuring the extent of the failure of a matrix to be hadamard , by @xmath30_{i , j}}^2 . \label{eq : gradientdef}\ ] ] observe , @xmath31 vanishes exactly when @xmath23 is a complex hadamard matrix . by computing the negative gradient of @xmath32 , we define a gradient system of ordinary differential equations , @xmath33 whose stationary points are the dephased complex hadamard matrices . equation ( [ eq : gradsys ] ) can be thought of as defining a flow on the @xmath34-torus , @xmath35 , since only the core * the @xmath36 lower - right submatrix is allowed to vary ( i.e. we insist that the matrices @xmath23 be dephased ) . note that this system does not take into account permutation equivalences , and so the set to which @xmath37 converges is not the space of inequivalent hadamards , but rather the space of inequivalent hadamards together with all copies of this space derived from permutations of the cores of its members . let @xmath38 be a fixed point of the nonlinear system @xmath39 where @xmath40 is an @xmath41 matrix and @xmath42 is @xmath43 in a neighborhood of @xmath38 . then the eigenvalues of @xmath40 can be used to determine the local dynamics of the system near the origin . in particular , there exist stable ( @xmath44 ) , unstable ( @xmath45 ) , and center ( @xmath46 ) subspaces spanned by the generalized eigenvectors corresponding to the eigenvalues of @xmath40 with negative , positive , and zero real - parts respectively . the generalized eigenvectors spanning the stable and unstable subspaces are tangent at the origin to the stable and unstable manifolds ( @xmath47 and @xmath48 ) : invariant sets with consistent asymptotic behavior with respect to the origin . similarly , the center subspace is tangent to every center manifold , @xmath49 : invariant sets on which @xmath50 ( the non - linear part of the vector field ) governs the dynamics as given by the center manifold theorem @xcite . assume that @xmath40 has @xmath51 eigenvalues with real part equal to zero and @xmath52 eigenvalues with negative real part . then the system defined by equation ( [ eq : nonlingensys ] ) can be written in diagonal form @xmath53 where @xmath54 , @xmath55 is a square matrix whose eigenvalues all have 0 real part , @xmath56 is a square matrix whose eigenvalues have negative real part , and @xmath57 . furthermore , for some @xmath58 there exists @xmath59 that defines the local center manifold @xmath60 \in \mathbb{r}^n\| \textbf{y } = h(\textbf{x } ) \text { for } \|\textbf{x}\| < \delta\}$ ] and satisfies @xmath61 = s h(\textbf{x})+h(\textbf{x},h(\textbf{x}))\ ] ] for @xmath62 . the flow on the center manifold is governed by the system @xmath63 for all @xmath64 with @xmath65 . [ thm : cmthm ] note that the linearized system given by the center manifold theorem has no eigenvalues with positive real part , and thus no unstable subspace or manifold at @xmath38 . this assumption is not required but is appropriate for us since hadamard matrices are global minima of the potential ( [ eq : gradientdef ] ) . the center manifold theorem provides a means to bound the local dimension of the space of degenerate fixed points , @xmath66 , containing @xmath38 since the center manifold at @xmath38 must contain @xmath66 . thus , by applying the theorem to the gradient system ( [ eq : gradsys ] ) , we can estimate the local dimension of the space of dephased complex hadamards at @xmath0 by computing the dimension of the center manifold at @xmath0 . this construction is reminiscent of the definition of the * defect * of a @xmath4 hadamard matrix @xmath0 as the dimension of the solution space of the real linear system @xmath67_{i , k}[h^]_{j , k}\left([r]_{i , k}-[r]_{j , k}\right ) & = 0 , \text { for } 1 \leq i < j \leq d,\end{aligned}\ ] ] where @xmath68 is a matrix of variables . the linear system ( [ eq : lindef ] ) is derived by considering a matrix @xmath69 , @xmath70_{i , j } = e^{{\mathrm{i}}[r]_{i , j}}\right)$ ] , and computing the jacobian of the non - linear system @xmath71_{i , k}[h^]_{j , k } e^{{\mathrm{i}}\left([r]_{i , k}-[r]_{j , k}\right ) } & = 0 , \text { for } 1 \leq i < j \leq d.\end{aligned}\ ] ] the @xmath72 equations in ( [ eq : nonlindef ] ) follow from the unitary condition , together with the dephased property which must be satisfied if @xmath69 is a dephased hadamard . the dimension of the center manifold of system ( [ eq : gradsys ] ) at a hadamard @xmath0 is actually equal to the defect @xmath73 since we are linearizing the dephased hadamard conditions in both cases . in particular , recall that if the defect of a hadamard matrix is 0 , then it must be isolated . likewise , if there does not exist a center subspace of @xmath74 at @xmath23 , then the stable subspace is @xmath34-dimensional and all points sufficiently close to @xmath23 must flow to it . since a center manifold is not necessarily ( and not usually ) comprised entirely of fixed points , the center manifold theorem could improve upon an overestimate of the local dimension of the space of complex hadamards suggested by the the dimension of the center subspace and the defect . if @xmath75 , then @xmath23 is not a complex hadamard matrix and therefore any flow on the center manifold , as slow as it may be , will shrink the bound given by the defect . a center manifold reduction is typically accomplished by performing a change of coordinates on the system into eigen - coordinates so that the center manifold can be written as a graph over the center subspace . this becomes impractical for a high - dimensional systems since the process first requires diagonalization of ( large ) symbolic matrices . we have adopted an alternative method in which the center manifold , @xmath49 , is written as an embedding over the @xmath51-dimensional center subspace , @xmath76 span(@xmath77 ) @xcite . one begins by writing @xmath78 and expands @xmath79 in a taylor expansion by repeated differentiation of the vector field . by observing that the center manifold is an invariant set ( i.e. the flow at a point on the @xmath49 is tangent to @xmath49 there ) one derives @xmath80 for some real - valued functions @xmath81 , which can be shown to be the time - rates - of - change of the embedding parameters , @xmath82 . more precisely , @xmath83 for @xmath84 . as one expands @xmath85 , they also approximate each @xmath86 and thus the flow on the center manifold . notice , if flow is nowhere present on the center manifold , then @xmath87 represents a local embedding of dephased hadamards and so this framework yields a series approximation of the manifold of interest . the principal computational limitation is the memory requirements of storing high - order tensors , which are needed if one wishes to expand the center manifold to high orders and which grow exponentially with order . this is exaggerated for large matrices , since the base of the exponential growth depends on the order of the hadamard . although identification of flow on the center manifold may not require expansion to high orders as shown by example in section [ sec:4x4 ] it should be noted that the _ absence _ of flow at any finite order can not guarantee that flow is not present at some higher order . it is known that every @xmath18 hadamard matrix is equivalent to a member of the continuous one - parameter family of inequivalent hadamards @xcite , @xmath88 moreover , for all but one choice of parameter @xmath89 $ ] , the defect of @xmath90 coincides with the dimension of this topological circle of inequivalent hadamards . however , for @xmath91 , when the hadamard is real , the defect jumps from 1 to 3 . thus , this is an example where we know that the defect overestimates the local freedom of inequivalent hadamards . this section serves as a illustrative example of the methodology developed in section [ sec : theory ] . in particular we use the center manifold reduction to show what is already known : the space of inequivalent hadamards near @xmath92 is one - dimensional , despite what the defect there might suggest . for the remainder of this section we will denote the one - parameter family of inequivalent @xmath18 hadamards as @xmath93 . we will interchangeably refer to elements of this space as either the matrix @xmath94 , for a particular @xmath95 $ ] , or the vector @xmath96.\ ] ] derivation of the gradient system @xmath97 ( equation [ eq : gradsys ] ) begins with the matrices @xmath98 with the aid of a computer algebra system , we compute explicit formulae for the eigenvalues of the jacobian matrix , @xmath99 , as functions of the parameter @xmath89 $ ] : @xmath100 the characteristic polynomial of @xmath99 is @xmath101 & [ \lambda^2 + ( 32 - 8\sin a ) \lambda + 128(1-\sin a)]\\ & [ \lambda^2 + ( 32 + 8\sin a ) \lambda + 128(1+\sin a ) ] \\ & [ \lambda^3 + 36\lambda^2 + ( 352- 64\sin^2 a)\lambda + 512(1 - \sin^2 a)].\end{aligned}\ ] ] let @xmath102 , @xmath103 , and @xmath104 , @xmath105 and @xmath106 equal the three real roots of the cubic factor @xmath107 @xmath108 and @xmath109 the roots of the quadratic @xmath110 and @xmath111 and @xmath112 the roots of the quadratic @xmath113 the constant terms of @xmath114 and @xmath115 vanish precisely at @xmath116 , increasing the multiplicity of the root @xmath117 , of @xmath118 , to three there . a plot of the eigenvalues of @xmath99 is given in figure [ fig : eigenvaluesdim4 ] . at @xmath119 for @xmath89 $ ] . @xmath120 ( blue ) , @xmath104 ( green ) , and @xmath108 ( red ) simultaneously vanish at @xmath91 , while all other eigenvalues ( gray ) are strictly negative for all @xmath89 $ ] . ] as stated , for every parameter value ( other than @xmath91 ) there are exactly 8 negative eigenvalues and 1 eigenvalue equal to 0 . the latter corresponds to the one - dimensional manifold of fixed points parametrized by @xmath121 , as its eigenvector is the tangent vector to the manifold @xmath122 embedded in @xmath123 , @xmath124.\ ] ] at @xmath116 , @xmath125 and @xmath126 also vanish , giving rise to a three - dimensional center manifold spanned by @xmath127 and the eigenvectors @xmath128 \text { and } \textbf{v}_3 \doteq \left[0,1,1,0,1,1,0,0,0\right].\ ] ] we anticipate nonlinear flow at all nearby points off of the lines spanned by each @xmath127 , @xmath129 and @xmath130 since the space of inequivalent hadamards is one - dimensional at @xmath131 . upon expanding the center manifold as an embedding over @xmath132 , @xmath133 we derive cubic approximations of the time - rates - of - change of the embedding parameters : @xmath134 thus , the motion of the point @xmath135 on @xmath49 is governed by ( [ eqn : timerates ] ) . because every nonzero cubic - term in ( [ eqn : timerates ] ) is a mixed monomial , setting any two embedding parameters to zero will result in no flow . for example , setting @xmath136 amounts to choosing a point @xmath137 on the manifold of fixed points ( since moving in the direction of @xmath127 amounts to varying the parameter @xmath121 ) . any solution trajectory of ( [ eqn : timerates ] ) converges to a point on one of the axes , where two of the embedding parameters vanish . the important point is that there _ is _ flow for any choice of embedded point @xmath135 not directly over the lines spanned by @xmath127 , @xmath129 or @xmath138 ( i.e. on an axis in @xmath139 space ) . therefore , the local dimension of the space of inequivalent hadamards at @xmath131 can not be greater than one . recall that @xmath140 does not converge to the space of inequivalent hadamards , but rather to the superset containing all row and column permutations of the core of dephased @xmath18 hadamards . let @xmath141 and @xmath142 be the @xmath18 permutation matrices which act to swap rows @xmath143 and @xmath144 and columns @xmath143 and @xmath144 respectively . there are exactly five unique row and column permutations of @xmath94 which , for some choice of parameter @xmath121 , are again equal to the matrix @xmath131 . in particular , @xmath145 any such core - permutation amounts to a permutation of the the coordinates @xmath146 $ ] of @xmath123 . for example , @xmath147 corresponds to the permutation @xmath148 \mapsto [ \theta_4,\theta_6,\theta_5,\theta_1,\theta_3,\theta_2,\theta_7,\theta_9,\theta_8 ] . \end{aligned}\ ] ] let @xmath149 ; the permutation of the coordinates resulting from the action of @xmath150 and @xmath151 . these permutations act on the vector @xmath127 , which we recall is tangent to @xmath94 at @xmath152 , in the following ways : @xmath153 & \xmapsto{\sigma_2 } \left[0,0,0,1,1,0,1,1,0\right ] = \textbf{v}_2 , \text { and } \\ \textbf{v}_1 = \left[1,0,1,0,0,0,1,0,1\right ] & \xmapsto{\sigma_3 } \left[0,1,1,0,1,1,0,0,0\right ] = \textbf{v}_3,\end{aligned}\ ] ] while @xmath154 remains fixed under @xmath155 and @xmath156 . we see that there are five total ( three distinct ) directions of fixed - point degeneracy at @xmath131 caused by the permutations of its core . we conclude that the three - dimensional center manifold emerges at the real hadamard to account for three copies of the space of dephased , permutation - equivalent hadamards intersecting here . the local geometric consequence of this result can be seen quite convincingly by directly visualizing the flow of core phases taken from a neighborhood of @xmath131 . figure [ fig : flowf41 ] shows snapshots of the evolution of a point cloud of 500 initial conditions drawn uniformly at random from a neighborhood of @xmath131 . the point cloud appears to be converging to three lines intersecting at @xmath131 , as expected in light of ( [ eqn : timerates ] ) . an ancillary animation reflecting the flow of @xmath97 in @xmath157 near @xmath131 is provided . uniformly at random from a neighborhood of the core phases corresponding to @xmath131 as they evolve under the flow defined by @xmath97 , at times ( i ) 5 , ( ii ) 20 , ( iii ) 70 , and ( iv ) 500 . each point cloud has been projected onto its top three principal components , and each point @xmath25 is colored by @xmath158 of the magnitude of the vector field @xmath97.,scaledwidth=100.0% ] [ sec:6x6 ] the smallest order for which classification of complex hadamards is incomplete is order six . numerical searches , analysis of known families , and a general method of construction due to szllsi , which depends on four free parameters , all support the conjecture that szllsi s generic four - parameter family and the isolated matrix @xmath159 capture all inequivalent @xmath15 hadamards @xcite . this is evidenced by calculation of the defect at both exact and numerical hadamards generated from known families and by random searches : save for @xmath159 the defect is always found to be four . we further support the conjecture by expanding the flow on the center manifold of @xmath160 at selected hadamards without ever encountering non - zero coefficients in the taylor expansion of the time - rates - of - change of the embedding parameters . * the affine fourier family * stemming from the fourier matrix @xmath161 ( where @xmath162 ) are two , two - parameter affine orbits of hadamards @xmath163 where @xmath164 note that @xmath165 denotes the hadamard ( entrywise ) product of the matrices @xmath0 and @xmath7 , and the numeric value 0 has been replaced with @xmath166 to improve readability . the defect of @xmath167 is four , and numerical evidence supports the conjecture that there is a four - dimensional non - affine family stemming from it @xcite . in kind the center subspace of @xmath160 at @xmath167 ( i.e. the kernel of the symmetric matrix @xmath168 ) is four dimensional and is spanned by the vectors @xmath169\\ \textbf{v}_2 & = [ 0 , 1 , 0 , 0 , 1 , 0 , 0 , 0 , 0 , 0 , 0 , 1 , 0 , 0 , 1 , 0 , 0 , 0 , 0 , 0 , 0 , 1 , 0 , 0 , 1 ] \\ \textbf{v}_3 & = [ 1 , 0 , 1 , 0 , 1 , 0 , 0 , 0 , 0 , 0 , 0 , 0 , 0 , 0 , 0 , 1 , 0 , 1 , 0 , 1 , 0 , 0 , 0 , 0 , 0]\\ \textbf{v}_4 & = [ 0 , 0 , 0 , 0 , 0 , 1 , 0 , 1 , 0 , 1 , 0 , 0 , 0 , 0 , 0 , 0 , 0 , 0 , 0 , 0 , 1 , 0 , 1 , 0 , 1 ] .\end{aligned}\ ] ] recalling the coordinates of @xmath170 in which @xmath160 is expressed , it is clear that @xmath127 and @xmath129 span @xmath171 and @xmath130 and @xmath172 span @xmath173 . we compute the center manifold @xmath174 as an embedding and , as before , expand the functions @xmath175 in power series . in support of the conjecture that there exists a four - dimensional manifold of complex hadamards passing through @xmath167 , all partial derivatives of each @xmath176 through fifth order are found to vanish . this means that if the four - dimensional center manifold stemming from @xmath167 does not consist entirely of fixed points , the flow near @xmath167 must be very slow . this may be interpreted in the following way : points on the local center manifold are very close to being hadamard , although they may not be . if the local manifold is four - dimensional , our expansion of @xmath177 gives a series approximation of the space of dephased hadamards near @xmath167 . again we can not rule out the possibility that a higher - order expansion might reveal flow on the center manifold . * the affine family * another maximal affine order-6 family , this one found by di @xcite , stems from the symmetric matrix @xmath178 one representative of the five , permutation - equivalent families stemming from @xmath179 is @xmath180 , where @xmath181 computing @xmath182 gives the linearized flow explicitly in terms of @xmath51 . the characteristic polynomial of @xmath182 is found to be @xmath183 where @xmath184 the factor @xmath185 in @xmath186 guarantees that the center subspace at @xmath187 is at least four dimensional for all values of @xmath51 . a plot of the eigenvalues of @xmath182 given in figure [ fig : eigenvaluesdim6 ] suggests that the stable subspace is 21 dimensional for all choices of @xmath51 . in fact , careful consideration of the factors that depend on @xmath51 reveals that the center subspace is exactly four dimensional for all @xmath51 since @xmath188 is never a root of @xmath189 , @xmath190 , or @xmath191 . this follows from the observation that @xmath192 is not a divisor of @xmath189 or @xmath190 for any @xmath51 since the constant terms in these factors do not depend on @xmath51 . in @xmath191 , the constant ( in @xmath192 ) term is @xmath193 which can not be made to vanish with @xmath194 . this proves that @xmath195 for every member of this affine family . for @xmath196 $ ] . the zero eigenvalue ( blue ) has multiplicity four , the roots of @xmath197 ( green ) have multiplicity two , and all other eigenvalues ( gray ) are simple . ] permutations of the core of @xmath187 give four other families which are permutation equivalent . these permutations act on the coordinates of the vector tangent to the curve @xmath187 at @xmath198 , @xmath199,\ ] ] to give @xmath200\\ \textbf{v}_3 & = [ 0 , -1 , 0 , -1 , 0 , 1 , 0 , 0 , 0 , 1 , 0 , 0 , 0 , 0 , 0 , 1 , 0 , 0 , 0 , 1 , 0 , -1 , 0 , -1 , 0]\\ \textbf{v}_4 & = [ 0 , 1 , 1 , 0 , 0 , -1 , 0 , 0 , -1 , 0 , -1 , 0 , 0 , -1 , 0 , 0 , 1 , 1 , 0 , 0 , 0 , 0 , 0 , 0 , 0]\\ \textbf{v}_5 & = [ 0 , 0 , -1 , 0 , -1 , 0 , 0 , -1 , 0 , -1 , 1 , 1 , 0 , 0 , 0 , 0 , 0 , 0 , 0 , 0 , 1 , 1 , 0 , 0 , 0].\end{aligned}\ ] ] since @xmath201 , we take @xmath202 as a natural choice of basis for the center subspace at @xmath179 . if flow exists on the center manifold , our choice of basis ensures that it will present itself as nonzero coefficients of the mixed monomials in the taylor expansions of @xmath203 unsurprisingly , we did not encounter any nonzero coefficients in these expansions , reinforcing the conjecture that the space of dephased hadamards near @xmath179 is four dimensional . * beauchamp and nicoara s * often one is able to obtain explicit eigenvectors and eigenvalues for the linearizations of @xmath204 and explicit coefficients in the taylor expansions of the center manifold and the time - rates - of - change of the embedding parameters . if numerical computations are undertaken , numerical approximations of the time - rates - of - change of the embedding parameters are determined . if these approximate coefficients are bounded away from 0 , one can deduce that there exists non - linear flow on parts of the center manifold , even without exact knowledge of that flow . it is known that the defect of the matrix @xmath205 where @xmath206 , is @xmath207 @xcite . since no family was known that contained @xmath208 , it was a longstanding open problem to determine if this matrix was actually isolated , despite having a positive defect . recently , it was shown that , in fact , the defect arises from a non - affine 2-dimensional family @xcite containing it . prior to learning of these results , we computed to 200 decimal places of accuracy numerical approximations of the basis vectors , @xmath127 and @xmath129 , which span the center subspace . we then expanded the center manifold @xmath209 as a numerical embedding over the center subspace . naturally , the numerical derivatives of the embedding parameters remain zero , as they must at all orders . visualizing the flow of @xmath210 in a neighborhood of @xmath211 , which appears to be drawn to a 2-dimensional plane ( see figure [ fig : flowd90 ] and ancillary files ) , provides another type of evidence of the conclusions which were proven definitively in @xcite . thus , this serves as an example where the numerical series expansion and numerical integration of the flow was only able to provide evidence for the existence of a center manifold of fixed points . uniformly at random from a neighborhood of the core phases corresponding to @xmath211 as they evolve under the flow defined by @xmath210 , at times ( i ) 5 , ( ii ) 20 , ( iii ) 70 , and ( iv , v ) 500 . each point cloud has been projected onto its top three principal components , and each point @xmath25 is colored by @xmath158 of the magnitude of the vector field @xmath210.,scaledwidth=100.0% ] * golay sequence affine family * @xmath212 * & non - di - type * @xmath213 in this section we derive two @xmath214 affine families stemming from a member of the one parameter family @xmath215 presented here in dephased form , discovered by lampio _ et al_. @xcite . the defect of @xmath216 and the dimension of the center subspace of @xmath217 there is 8 , so it may be possible to introduce additional affine parameters . as we have seen , the center subspace at a hadamard , @xmath0 , must contain all vectors tangent to dephased families stemming from @xmath0 , affine or otherwise . for example , the vector @xmath218 with 1 s in the coordinates corresponding to the core entries of @xmath219 whose phases depend on @xmath121 ( and 0 s elsewhere ) will be in the center subspace because it is tangent to the line of hadamards , @xmath219 . thus , one can search for affine families by considering linear combinations of vectors in a basis for the center subspace and introducing appropriately - scaled parameters in the phases corresponding to nonzero coordinates . contained in the kernel of @xmath220 is the vector @xmath221 shown here as the lower - right @xmath222 submatrix of @xmath223 to establish its relationship with the core of @xmath224 . by computing @xmath225 ( where @xmath121 is a free parameter ) we uncover a one - parameter affine family ; explicitly @xmath226 similarly , the null vectors of @xmath220 , expressed in the cores of @xmath227 together reveal the two - parameter family @xmath228 @xmath229 @xmath230 and @xmath231 were verified to be hadamard by symbolic computation . interestingly , these families and @xmath232 are independent in the sense that they do not , together , form a three - parameter affine family . that being said , a center manifold reduction did not rule out the possibility that they belong to some unifying family that has yet to be discovered . we applied the same approach to the non - di - type matrix @xmath233 whose membership in a three - parameter affine family @xmath234 was already established by szllsi @xcite . notice that the defect of @xmath235 is 16 , and so there may be more parameters which can be introduced to the family @xmath236 . towards identifying these additional parameters , we searched for and found a basis , @xmath237 , for the kernel of @xmath238 whose elements each describe a one - parameter affine family stemming from @xmath235 . in table [ tab:16basis ] we list the nonzero coordinates of each @xmath239 each of which happens to be 1 or -1 by identifying the coordinates having either value . .nonzero coordinates of the vectors in the basis @xmath240 for @xmath238 . a nonzero coordinate has value 1 or -1 , indicated by the subcolumn to which it belongs . [ cols="^,>,<,<,<,<,<,<,<,<,^,^ , < , < , < , < , < , < , < , < " , ] [ tab:16basis ] for each @xmath241 , the collection of matrices @xmath242 where @xmath243 is a parameter is an affine family . this does not say that there is a 16-dimensional affine family stemming from @xmath235 ; in fact , simultaneous inclusion of all @xmath243 destroys the hadamard property for most choices of parameters . however , many combinations of inclusion of parameters do give two- and three - dimensional families . for example , @xmath244 another example of a three - parameter family found among these vectors is @xmath245 note that no combination of four ( or more ) of these vectors give an affine family , although this does not guarantee one does not exist . the existence of this particular basis for the kernel of @xmath238 does provide one possibility for the genesis of a 16-dimensional center subspace : certainly the defect must be at least 16 to account for the intersection of 16 , possibly permutation - equivalent , one - parameter affine ( sub-)families passing through @xmath235 . in this paper we have given a new interpretation of the defect of a hadamard matrix as the dimension of a center subspace of a gradient flow whose fixed points are exactly the dephased hadamard matrices . we have applied this technique to a simple example , using dynamical systems theory to explain why the defect of the real @xmath18 hadamard is larger than the complex members of the affine family @xmath122 . we have used tools from dynamical systems theory to prove that the @xmath246 for all values of @xmath51 and have presented a new type of evidence in support of existing conjectures concerning @xmath15 hadamards . finally we have used our perspective to uncover new @xmath214 affine families . it is a virtue of the formalism built in section [ sec : theory ] that we need not have explicit values for the taylor coefficients of the functions @xmath86 to conclude that flow on the center manifold exists . it is a shortcoming of center manifold reduction , in general , that it can not prove flow does _ not _ exist on some part of a center manifold . this limitation is a consequence of the fact that , _ a priori _ , one has no knowledge of the smallest order in the taylor expansion where one might first encounter nonzero contribution to the time - rates - of - change of the embedding parameters . we are not certain that our application is bound with this deficiency and are hopeful that a deeper understanding of the derivatives of the vector field @xmath74 may be exploited to further expand the use of dynamical systems theory to attack questions about complex hadamards . in particular , if one could show that all coefficients in the expansion of the time - rates - of - change of the embedding parameters vanish in the center manifold reduction of @xmath74 at a hadamard @xmath0 , then one would prove the existence of a positive - dimensional family of complex hadamards stemming from @xmath0 . finally , vectors in a basis for the center subspace of @xmath247 at an order-@xmath248 matrix @xmath0 may be tangent to affine families stemming from @xmath0 . this fact can be exploited to find new affine families , in the manner of section [ sec : examples ] . m. reck m , a. zeilinger , h.j . bernstein and p. bertani , http://link.aps.org/doi/10.1103/physrevlett.73.58[_experimental realization of any discrete unitary operator _ ] , phys . lett . , * 73 * ( 1994 ) , 5861 . p.h.j lampio , f. szllsi and p.r.j stergrd , http://dx.doi.org/10.1016/j.disc.2012.10.001[_the quaternary complex hadamard matrices of orders 10 , 12 , and 14 _ ] , discrete mathematics , * 313 * ( 2013 ) , 189206 . m. matolcsi , j. rffy and f. szllsi , http://chaos.if.uj.edu.pl/~karol/hadamard/chm_papers/mrs06.pdf[_constructions of complex hadamard matrices via tiling abelian groups _ ] , arxiv : quant - ph/0607073v2 , ( 2006 ) .	the defect of a complex hadamard matrix @xmath0 is an upper bound for the dimension of a continuous hadamard orbit stemming from @xmath0 . we provide a new interpretation of the defect as the dimension of the center subspace of a gradient flow and apply the center manifold theorem of dynamical systems theory to study local structure in spaces of complex hadamard matrices . through examples , we provide several applications of our methodology including the construction of affine families of hadamard matrices .
the photometric analysis of large galaxies is a double edged sword . while increased resolution , compared to distant galaxies , provides avenues for more detailed analysis of galaxy properties ( to name a few : examination of star formation regions , examination of core parameters to search for massive blackholes , spiral arm analysis , isophotal irregularities that may signal rings , bars or other secular evolution processes ) , it is a data reduction fact that the larger number of pixels complicates the extraction of simple parameters , such as total magnitude , mean surface brightness and isophotal radius . for example , the fact that the galaxy is spread over a larger area of the sky means that the outer pixels have more sky luminosity than galaxy luminosity , increasing the error in any galaxy value . in addition , increased angular size has frequently prevented a fair comparison between distant and nearby galaxy samples simply because the techniques used to extract parameters from nearby galaxies differ from those used on small galaxies . in general , the analysis of large galaxies ( i.e. ones with many pixels ) requires a full surface photometry study of the isophotes and their shape . for most extragalactic systems , the shape of choice is an ellipse . the astrophysics behind this assumption is that galaxy light traces stellar mass , and stellar mass follows elliptical orbits as given by kepler s 1st law . certainly , this is true the case of early - type galaxies ( elliptical and s0 s ) as demonstrated by studies that examined the residuals from elliptical isophotes ( jedrzejewski 1987 ) . this is also mostly true for disk galaxies , although the lumpiness of their luminosity distribution due to recent star formation increases the noise around each ellipse ( see the discussion in 2.4 ) . for dwarf irregular systems , any regular contours are poor describers of their shape , thus an ellipse is used because it is the simplest shape ( aside from a circle ) to describe an irregular isophote . therefore , the analysis of large galaxies begins with the reduction of their 2d images into 1d surface photometry as described by elliptical isophotes . in turn , the 1d profiles can be fit to various functions in order to extract characteristic luminosities ( stellar mass ) , scale lengths and standard surface brightnesses ( luminosity density ) . when combined with kinematic information , these three parameters form the fundamental plane for galaxies , a key relationship for understanding the formation and evolution of galaxies . aside from direct relevance to the fundamental plane , the need for better galaxy photometric tools has also increased with the influx of quality hst imaging . before hst , distant galaxies were mere point sources , but now with wfpc2 , acs and nicmos data , there is the need to perform full surface photometric studies on a much larger volume of the universe . the sizes of our database on the photometric structure of galaxies has increased a thousandfold in the last 10 years , but most of the tools used to reduce this new data are up to 20 years out - of - date . thus , the analysis of high resolution space imaging data is far behind spectroscopic and high energy data , not due to lack of interest , but due to the inadequacy of our 2d analysis tools . the goal of this software project ( called the archangel project for obtuse historical reasons ) has been to produce a series of proto - nvo type tools , related to surface photometry , and develop a computing environment which will extend the capability of individual observers ( or entire data centers ) to perform virtual observatory science . there is no attempt herein to replace existing analysis packages ( i.e. pyiraf ) , but rather our goal is to supplement existing tools and provide new avenues for data reduction as it relates to galaxy photometry . we hope that the fundamental components in this package will provide the community with new methods to which they can add their own ideas and techniques as well as provide learning environment for new researchers . in addition , there is growing amount of data by non - optical astronomers as new space missions provide imaging data in wavelength regions previously unexplored . thus , there is a new and growing community of non - optical astronomers with 2d analysis needs that we hope to serve . the tools described herein are not intended to be a complete data reduction package per say , but rather a set of basic modules that allows the user to 1 ) learn the procedures of galaxy photometry , 2 ) tailor the tools to their particular needs , 3 ) begin an advanced learning curve of combining basic modules to produce new and more sophisticated tools . turning raw data ( level 1 or 2 data ) from the telescope ( space or ground ) into calibrated , flattened images is the job of other , more powerful packages such as pyraf . the tools presented herein bridge the intermediate step between calibrated data and astrophysically meaningful values . specifically , we are concerned with the analysis of 2d images into 1d values , such as a surface brightness profile , and further tabulation into final values , such as total luminosity or scale length . with respect to galaxy images , the numbers most often valued are luminosity , scale length and luminosity density . unfortunately , due to the extended nature of galaxies , the quality and accuracy of these values can varying depending on the type of value desired . for example , luminosities can be extracted in metric form ( luminosity within 16 kpc ) or isophotal ( luminosity inside the 26.5 mag arcsecs@xmath1 isophote or the total luminosity , an extrapolation to infinite radius . scale length can be expressed as the radius of half - light or a formula fit to the luminosity distribution ( e.g. seric function ) . luminosity density can be described through a detailed surface photometry profile , or integrated as a mean surface brightness within an isophote , or again a fitted curve such as an exponential disk . the tools provided by this project allow an inexperienced user the capability to explore their dataset and extract meaningful results while outlining the limitations to that data . for the experienced researcher , these tools enhance their previous background in data reduction and provide new , and hopefully , faster avenues of analysis . to this end , the tools provided by this package provide a user with most basic of descriptions of a galaxy s light , then allowing the option to select any meaningful parameter by toggling a switch . for most parameters , such as aperture magnitudes , the switch is simple and automatic . for more complicated parameters , such as a profile fit or an asymptotic magnitude , the switch is understandably more sophisticated and needing more explanation to the user for accurate use . this paper is divided into five sections describing the major components of the reduction package : 1 ) sky determination , 2 ) 2d profile fitting , 3 ) aperture photometry , 4 ) extraction of 1d parameters from surface brightness profiles and 5 ) extracting total magnitudes . each section contains examples of the reduction of galaxy images from the 2mass archive . the fastest way to introduce the techniques and tools used in our package is to walk through the analysis of a several different types of galaxy images . a more non - linear reader can refer to the appendix for a listing of the major tools . a script titled @xmath2 is included which outlines the usage of the routines described below . for a majority of galaxy images , this script will produce a usable surface brightness profile , and this script forms the core of the client / server version of this package ( see 5 ) . but , a sharper understanding of the data requires more interaction with the techniques , the user is encouraged to run through the examples given in the package . to illustrate our tools , we have selected 2mass @xmath3 images of several galaxies found in the revised shapley - ames catalog with the characteristics of smooth elliptical shape ( ngc 3193 ) , disk shape ( ic 5271 ) , spiral features ( ngc 157 ) and low in surface brightness / low contrast to sky ( ngc 2082 ) . the analysis procedure for each galaxy is divided into five basic parts ; 1 ) sky determination , 2 ) cleaning , 3 ) ellipse fitting , 4 ) aperture photometry and 5 ) profile fitting . before starting it is assumed that the data frame has be initially processed for flatfielding , dark subtraction and masking of chip defects . small defects , such as cosmic rays , are cleaned by the analysis routines . but the errors are always reduced if known features are removed before analysis . the following routines work on poorly flattened data ( e.g. gradients or large - scale features ) , and will signal the poorness by the resulting errors , but the removal of large - scale flattening problems requires more interaction then acceptable for this package and remains the responsibility of the user . any galaxy photometry analysis process begins with an estimate of the image s sky value . while this is not critical for isophote fitting , it is key for actually finding targets , cleaning the frame of stars and smaller galaxies , plus determination of the photometry zeropoints . accurate sky determination will , in the end , serve as the final limit to the quality of surface photometry data since a majority of a galaxy s luminosity distribution is near the sky value . for this reason , sky determination has probably received as much attention in astronomical data literature as any other comparable reduction or analysis problem . the difficulty in sky determination ranges from too few photons to know the behavior of the instrumental response ( e.g. , high energy data ) to a high temporal varying flux of sky photons that overwhelms the galaxy signal ( e.g. , near - ir data ) . surface fitting , drift scans , sky flats and super flats are all procedures used to minimize the sky contribution to the noise levels of the final data . several clever , but not technically challenging , algorithms were included in the noao iraf system to handle time averaged flats and data , median co - adding and cosmic ray subtraction . in the end , improved ccd quality lowered the demands of sky subtraction as the production of linear , good charge transfer and uniform sensitivity chips replaced the earlier generations and their wildly inaccurate backgrounds . frames for ngc 3193 , an elliptical selected from the rsa sample ( schombert 2007 ) . note the proper cleaning of contaminating stars , even a object near the galaxy core . ] for a cosmetically smooth image , an efficient , but crude , sky fit is one that simply examines the border of the frame and does an iterated average , clipping pixels more than 4@xmath4 from the mean . a border sky fit is often sufficient to find the starting center of the galaxy ( for the ellipse fitting routines ) , clean the frame of stars / galaxies external to the object of interest ( the ellipse fitting routines will clean along the isophotes , see below ) and provide a preliminary error estimate to the photometry . this error estimate is preliminary in that the true limiting error in the surface ( and aperture ) photometry of large galaxies is not the rms of an isophote , but how well the sky value is know . once the number of pixels involved in a calculation ( be it an isophote or an aperture ) becomes large ( greater than 50 for typical readout noises ) , then the error is dominated by the precision of the sky value . the disadvantage to a border sky fit is the occasional inconvenient occurrence of stars or bright galaxies on the edge of the frame . an iterated mean calculation will remove small objects . and large objects will be signaled with large @xmath4 s in an iterative mean search . in an automated procedure , more than likely , the task will have to halt and request human intervention to find a starting sky value . after years of experimentation , the method of choice for accurate sky determination for extended galaxies is to evaluate sky boxes . this is a procedure where boxes of a set sized are placed semi - randomly ( semi in the sense of avoiding stars and other galaxies ) in the frame . an algorithm calculates an iterative mean and @xmath4 for each box . these means ( and @xmath4 s ) are then super - summed to find the value of the sky as the mean of the means ( and likewise , the error on the sky value is the @xmath4 on this mean ) . from an analysis point of view , there are several advantages to this technique . one is that each box exists as a measurement independent of the other boxes . thus , the statistical comparison of the boxes is a real measure of the quality of the sky determination for the frame in terms of its accuracy and any gross trends with the frame . another advantage is that contaminating objects are relatively easy to avoid ( visual choice of sky boxes ) or to sort by the higher @xmath4 per box . lastly , sky boxes are the easiest method of finding regions for sky determination outside the galaxy itself , particularly where an irregular object may fill a majority of the data frame . the most difficult decision in sky determination by boxes is , of course , where to place the boxes . when done visually , the user selects region ( usually with a cursor ) that are free of stars and sufficiently far away from the target galaxy to be clear of its envelope light . for an automated process , the procedure returns for a final sky estimate after the ellipse fitting process is completed and when all the stars / galaxies are cleaned ( set to values of not - a - number , nan ) . then , the outer edge of the large galaxy is determined and an iterative analysis of sky boxes outside this radius is used to determine the true sky and , most importantly , the variation on the mean of those boxes as a measure of how well the sky is known . this procedure is the role of _ sky_box _ , see the appendix for a more detailed description of its options . reduction of a 2d image into a 1d run of intensity versus radius in a galaxy assumes some shape to the isophote . very early work on galaxies used circles since the data was obtained through circular apertures in photoelectric photometers . for early type galaxies , the ellipse is the shape that most closely follows the shape of the isophotes . this would confirm that the luminosity being traced by an isophote is due to stellar mass , which follow elliptical orbits ( kepler s 1st law ) . as one moves to along the hubble sequence to later type galaxies , the approximation of an ellipse to the isophotes begins to break down due to recent star formation producing enhancements in luminosity density at semi - random positions . however , no consistent shape describes the isophotes of irregular galaxies , so an ellipse is the best shape , to first order , and provides a common baseline for comparison to more regular galaxies . image of ic 5271 , a sb(rs ) galaxy selected from the rsa catalog . typical of the isophotes for a disk / bulge galaxy , there is a cross over point as one transitions from a more spherical bulge to a flattened disk . while this is flagged by the reduction software , it is astrophysically real and signals the lens morphology often seen in surface brightness profiles of disk galaxies . ] fitting a best ellipse to a set intensity values in a 2d image is a relatively straight forward technique that has been pioneered by cawson ( 1987 ) and refined by jedrzejewski ( 1987 ) ( see also an excellent review by jensen & jorgensen 1999 ) . the core routine from these techniques ( prof ) was eventually adopted by stsdas iraf ( i.e. ellipse ) . the primary fitting routine in this package follows the same techniques ( in fact , uses much of the identical fortran code from the original gasp package of cawson ) with some notable additions . these codes start with an estimated x - y center , position angle and eccentricity to sample the pixel data around the given ellipse . the variation in intensity values around the ellipse can be expressed as a fourier series with small second order terms . then , an iterative least - squares procedure adjusts the ellipse parameters searching for a best fit , i.e. minimized coefficients . there are several halting factors , such as maximum number of iterations or minimal change in the coefficients , which then moves the ellipse outward for another round of iterations . once a stopping condition is met ( edge of the frame or sufficiently small change in the isophote intensity ) , the routine ends . a side benefit to above procedure is that the cos(4@xmath5 ) components to each isophote fit are easily extracted , which provides a direct measure of the geometry of the isophote ( i.e. boxy versus disk - like , jedrzejewski 1987 ) . one new addition , from the original routines , is the ability to clean ( i.e. mask ) pixels along an isophote . basically , this routine first allows a few iterations to determine a mean intensity and rms around the ellipse . any pixels above ( or below ) a multiple of the rms ( i.e. 3@xmath4 ) are set to not - a - number ( nan ) and ignored by further processing . due to the fact that all objects , stars and galaxies , have faint wings , a growth factor is applied to the masked regions . while this process is efficient in early - type galaxies with well defined isophotes , it may be incorrect in late - type galaxies with bumpy spiral arms and hii regions . the fitting will be smoother , but the resulting photometry will be underestimated . this process can be controlled early in the analysis pipeline by the user with an initial guess of the galaxy s hubble type . also , the erased pixels are only temporary stored until an adequate fit is found . once a satisfactory ellipse is encountered , only then are the pixels masked for later ellipse fitting . the masked data is written to disk at the end of the routine as a record of the cleaning . the ellipse fitting is the function of _ efit _ as described in the appendix . for early - type galaxies , lacking any irregular features , the cleaning process is highly efficient . the pipeline first identifies the galaxy and its approximate size by moment analysis . it then cleans off stars / galaxies outside the primary galaxy by moment identification and radius growth for masking . stars / galaxies inside the primary galaxy are removed by the ellipse fitting routine . the resulting ellipses are inspected for crossover ( isophotes that crossover are assumed to be due to errors or embedded stars / galaxies and removed by averaging nearby ellipse isophotes , this is not true for disk galaxies ) . the smoothed ellipses are used by a more robust cleaning algorithm and the whole ellipse fitting process is repeated on the cleaned frame . image of ngc 157 , a face - on sc(s ) galaxy selected from the rsa catalog . similar to ic 5271 , there are several crossover points in the fitted ellipses . the fitting program does a good job of following the spiral arms in the inner regions , then a large jump from bulge to disk region . ] an example of the analysis of an elliptical is found in figure 1 , a 2mass @xmath3 image of ngc 3193 . the top panel is the raw 2mass image , the bottom panel is the resulting cleaned image output at the end of the reduction process . the cleaning process efficiently removed all the stars on the frame , including the brighter object on the northern edge of the frame and its diffraction spikes . the star closest to the galaxy core is a problem in two arenas . the first is in the calculation of ellipse , as the inner star would drag the calculated moments off center . the isophote erasing routine has handled this as can be seen in figure 2 , where the fitting ellipses are shown and are not deflected by the erased star . second , is that calculated total magnitudes would either be over estimated ( if the star is not masked ) or under estimated ( if the star is masked and the galaxy light from those pixels is not replaced ) . this problem will be discussed in 2.6 . as one goes towards later type galaxies , there is an increase in the non - elliptical nature to their isophotes and an increase in luminosity density enhancements ( hii regions , stellar clusters , spiral features ) which are legitimate components to the galaxy s light distribution and should not be cleaned . the user can specify the galaxy type and the cleaning restrictions will be tightened ( only to stellar objects and at a higher cleaning threshold ) plus the restrictions on overlapping ellipses is loosened ( e.g. the transition from a round bulge to a flat disk ) . most importantly , while some galaxy features are cleaned for the sake of a harmonious ellipse fit , those pixels need to be filled for later aperture photometry . image of ngc 2082 , a lsb disk galaxy . note , that the ellipse fitting routine expanded the annulus size to increase the s / n . ] an example of this behavior can be found in figure 3 , the @xmath3 image of ic 5271 . the red ellipses indicate isophote fits that crossover . while flagged as an error , this is in fact the real behavior of the isophotes as one transitions from bulge to disk . the resulting intensities are probably overestimated due to the crossover effect , but this error will be minor compared to the errors that would result from an off - center or overly round ellipse . the quality of the fitting procedure can be judged by the behavior of the ellipse parameters such as eccentricity , position angle and center . if there are large jumps in any of the parameters that determine shape , then this may signal a feature in the galaxy that needs to be cleaned ( a buried star for example ) . slightly less abrupt changes may signal an astrophysically interesting features , such as a bar or lens morphology . under the assumption that the isophotes of a typical galaxy are a smooth function with radius , the ellipse fitting algorithm checks for ellipse parameters that indicate a crossing of the isophotal lines . these ellipses are smoothed and flagged ( the mean of the inner and outer ellipse parameters is used ) . in certain scenarios , crossing isophotes are to be expected , for example the transition region from a bulge to a disk ( see figure 2 ) , and the smoothing criteria is relaxed . this is the function of @xmath6 as described in the appendix . an example of @xmath6 s corrections can be seen in figure 4 , the @xmath3 image of ngc 157 . several interior ellipses display erratic behavior , but @xmath6 took the mean average of nearby ellipses ( in green ) to produce a more rational fit . the resulting intensities were also more stable , although the rms is going to be highly than the typical isophotes found in an elliptical . an example of a lsb galaxy fit is found in figure 5 . the ellipse fitting routine , recognizing that the target is low in contrast with respect to sky , widened the annulus for collecting pixel values . this increases the s / n at some loss of spatial resolution . since resolution is usually not important in a galaxy s halo region , this is an acceptable trade off . lsb galaxies are susceptible to fitting instability , the fitting routines are tightened against rapid changes in eccentricity and centering to prevent this behavior . lsb galaxies also demonstrate a key point in determining errors from surface photometry . there are two sources of error per isophote , the rms around the ellipse and the error in the sky value . the rms value is a simple calculation using the difference between the mean and the individual pixel values . this rms then reflects into an observable error as the @xmath7 . however , as the isophote intensity approaches the sky value , the number of pixels increases and the error due to rms becomes an artificially low value . in fact , at low intensities , the knowledge of the sky value dominates and the error in the isophote is reflected by the sky error ( preferably as given by the @xmath4 on the means of a large number of sky boxes ) . with a file of isophotal intensities versus radius in hand , it is a simple step to producing a surface brightness profile for the galaxy . there are a few tools are in the package to examine the quality of the ellipse fitting ( e.g. @xmath8 , an interactive comparison of the image and the ellipses ) . at the very least , a quick visual inspection of the ellipses seems required as a bad mismatch leads to strongly biased results ( see cautionary tale in 4 ) . a user can either step through a directory of data files ( e.g. using the @xmath9 tool ) or a user can automatically produce a group of gif images with a corresponding html page , then use a browser to skim through a large number of files . calibration from image data numbers ( dn ) to fluxes ( or magnitudes ) is usually obtained through standard stars with corrections for airmass and instrumental absorption . if these values are in the fits headers , then they are automatically added to the object s xml file . additional corrections for galactic absorption , k - corrections and surface brightness dimming are well documented in the literature and can be assigned automatically by grabbing xml data from ned . a chosen cosmology converts radius in pixel units into astrophysically meaningful values of kiloparsecs . a python command line script ( @xmath10 ) based on ned wright s cosmology calculator is included in the package all these values can be added to the xml file for automatic incorporation to the analysis programs . if they do nt exists , then instrumental mags will be used , which can easily be converted to real units later on . fits to disk and bulge . the solid line is the addition of the two curves . ] analysis of a 1d surface brightness profile ( the job for the @xmath11 tool ) depends on the scientific goals of the user . for example , early - type galaxies are typically fit to a de vaucouleurs r@xmath12 curve to extract a scale length ( effective radius ) and characteristic surface brightness ( effective surface brightness ) . irregular and dwarf galaxies are well fit by exponential profiles which provide a disk scale length and central surface brightness . disk galaxies can be fit with a combination of bulge and disk fits , to extract b / d ratios and disk scale lengths . due to this combination of @xmath13 and exponential curves for large bulge spirals , it is computationally impossible to correctly determine which function , or combination of functions , best fits a particular galaxy s profile . in the past , one would examine the 2d image of the galaxy and obvious disk - like galaxies would be fit to @xmath13 plus exponential . objects with elliptical appearance were fit to a strict @xmath13 shape . this produces a problem for large bulge s0 s which are difficult to detect visually unless nearly edge - on . the simplest solution to this problem , using only the 1d surface photometry , is to examine the profiles in a plot of mag arcsecs@xmath14 versus linear radius . with this plot , exponential disks appear as straight lines , see figure 6 as an example of a pure disk in ngc 2403 . bulge plus disk components are also straight forward in this mag / linear radius space , see figure 7 a good example of a bulge plus disk fit in ngc 3983 . if a profile displays too much curvature , with no clear linear disk portion , then it is a good candidate for a pure @xmath13 fit ( see figure 8 , ngc 3193 ) . this option is easily checked by plotting the profile in mag arcsecs@xmath14 versus r@xmath12 space as shown in figure 8 . most r@xmath12 profiles only have a linear region in the middle of the surface brightness profile , typically with a flattened core and fall - off at large radii ( see schombert 1987 ) . fit is shown . ] the seric function is also popular for fitting surface brightness profiles ( graham & driver 2005 ) , although not currently supported by this package , any fitting function is easy to add to the reduction routines as the core search routine is a grid search @xmath15 minimization technique . however , there are issues with surface photometric data where the inner regions have the highest s / n but the outer regions better define a galaxy s structure ( schombert & bothun 1987 ) . with user guidance , this grid search works well for any user defined function . also , since there are a sufficient number of packages for fitting 1d data in the community , this package only provides a simple graphic plotting function . more sophisticated analysis needs guidance by the user , but this package provides the framework for just such additions . often the scientific goal of a galaxy project is to extract a total luminosity for the system ( and colors for multiple filters ) . for small galaxies , a metric aperture or isophotal magnitude is suitable for comparison to other samples ( certainly the dominate source of error will not be the aperture size ) . however , for galaxies with large angular size ( i.e. many pixels ) , their very size makes total luminosity determination problematic . natively , one would think that a glut of pixels would make the problem of determining a galaxies luminosity easier , not more difficult . however , the problem here arises with the question of where does the galaxy stop ? or , even if you guess an outer radius , does your data contain all the galaxy s light ? the solution proposed by de vaucouleurs decades ago is to use a curve of growth ( de vaucouleurs 1977 ) . almost all galaxies follow a particular luminosity distribution such that the total light of a galaxy can be estimated by using a standard growth curve to estimate the amount of light outside your largest aperture . for a vast majority of galaxies , selecting either an exponential or r@xmath12 curve of growth is sufficient to adequately describe their total luminosities ( burstein 1987 ) . however , for modern large scale ccd imaging , the entire galaxy can easily fit onto a single frame and there is no need for a curve of growth as all the data exists in the frame . with adequate s / n , it would seem to be a simple task to place a large aperture around the galaxy and sum the total amount of light ( minus the sky contribution ) . however , in practice , a galaxy s luminosity distribution decreases as one goes to larger radii , when means the sky contribution ( and , thus , error ) increases . in most cases , larger and larger apertures simply introduce more sky noise ( plus faint stars and other galaxies ) . and , to further complicate matters , the breakover point in the optical and near - ir , where the galaxy light is stronger than the sky contribution will not contain a majority of the galaxy s light . so the choice of a safe , inner radius will underestimate the total light . the procedure selected in this package , after some numerical experimentation , is to plot the aperture luminosity as a function of radius and attempt to determine a solution to an asymptotic limit of the galaxy s light . this procedure begins by summing the pixel intensities inside the various ellipses determined by @xmath16 . for small radii , a partial pixel algorithm is used to determine aperture luminosity ( using the surveyors technique to determine each pixel s contribution to the aperture ) . at larger radii , a simply sum of the pixels , and the number used , is output . in addition , the intensity of the annulus based on the ellipse isophote and one based on the fit to the surface photometric profile are also outputted at these radii ( see below ) . note that a correct aperture luminosity calculation requires that both a ellipse fit and a 1d fit to the resulting surface photometry has be made . the ellipse fit information is required as these ellipses will define the apertures , and masked pixels are filled with intensities given by the closest ellipse . a surface photometric fit allows the aperture routine to use a simple fit to the outer regions as a quick method to converge the curve of growth . once the aperture luminosities are calculated , there are two additional challenges to this procedure . the first is that an asymptotic fit is a difficult calculation to make as the smallest errors at large radii reflect into large errors for the fit . two possible solutions are used to solve this dilemma . the first solution is to fit a 2nd or 3rd order polynomial to the outer radii in a luminosity versus radius plot . most importantly for this fit , the error assigned the outer data points is the error on the knowledge of the sky , i.e. the rms of the mean of the sky boxes . this is the dominant source of error in large apertures and the use of this error value results in a fast convergence for the asymptotic fit . the resulting values from the fit will be the total magnitude and total isophotal size , determined from the point where the fit has a slope of zero . a second solution is to use an obscure technique involving rational functions . a rational function is the ratio of two polynomial functions of the form @xmath17 where @xmath18 and @xmath19 are the degree of the function . rational functions have a wide range in shape and have better interpolating properties than polynomial functions , particularly suited for fits to data where an asymptotic behavior is expected . a disadvantage is that rational functions are non - linear and , when unconstrained , produce vertical asymptotes due to roots in the denominator polynomial . a small amount of experimentation found that the best rational function for aperture luminosities is the quadratic / quadratic form , meaning a degree of 2 in the numerator and denominator . this is the simplest rational function and has the advantage that the asymptotic magnitude is simply @xmath20 , although is best evaluated at some radii in the halo of the galaxy under study . usually the aperture luminosity values will not converge at the outer edges of a galaxy . this is the second challenge to aperture photometry , correct determination of the luminosity due to the faint galaxy halo . this is where the surface photometry profile comes in handy . contained in that data is the relationship between isophotal luminosity and radius , using all the pixels around the galaxy . this is often a more accurate number than attempting to determine the integrated luminosity in an annulus at the same radius . this information can be used to constraint the curve of growth in two ways . one , we can use the actual surface brightness intensities and convert them to a luminosity for each annulus at large radii . then , this value can be compared to the aperture value and a user ( or script ) can flag where the two begin to radically deviate . often even the isophotal intensities will vary at large radii and , thus , a second , more stable method is to make a linear fit of an exponential , r@xmath12 or combined function to the outer radii and interpolate / extrapolate that fit to correct the aperture numbers . figure 9 display the results for all three techniques for the galaxy ngc 1003 . the black symbols are the raw intensities summed from the image file . the blue symbols are the intensities determined from the surface photometry . the orange symbols are the intensities determined from the fits to the surface photometric profile . this was one of the worst case scenarios due to the fact the original image is very lsb ( in the near - ir @xmath3 band ) . due to noise in the image and surface photometry , the outer intensities grow out of proportional to the light visible in the greyscale figure . a fit to the raw data does not converge ( blue line ) . a 2nd order fit to the profile fit ( orange line ) also fails to capture the asymptotic nature . the rational function fit ( pink line ) does converge to an accurate value . if similar types of galaxies are being analyzed , it is a simple procedure to automate this process . in the past , when disk space was at a premium and i / o rates were slow , astronomical data was stored in machine specific formats . however , today disk space is plentiful and file access times are similar to processing times on most desktop systems . thus , a majority of simple astronomical databases are stored in flat file format , also called plain text or ascii files ( note : this is an interesting throwback to the original data methods from the end of the 19th century , where information is stored as a system of data and delimiters , such as spaces or commas ) . the endproduct data files for most packages rarely exceed a few kilobytes or a few hundred lines . the simplest access to these types of files is an editor such as vi or emacs . sufficient documentation ( i.e. header files ) makes understanding the data , and writing applications for further analysis these data files , a relatively simple task . however , there is a strong driver to migrate output files into xml format for even the simplest data files . extensible markup language ( xml ) is a w3c - recommended general - purpose markup language that is designed store and process data . the core of xml is the use of tags , similar to html tags ( i.e. ` < tag > data</tag > ` ) , to delineate data values and assign attributes to those values ( for example , units of measure ) . xml is not terse and , therefore , somewhat human - legible ( see figure 1 for xml example of astronomical data ) . scans . the red ellipses are fits from the 2mass galaxy pipeline . the blue ellipses are the resulting fits from this package . the 2mass fits clearly fail to follow the flatter isophotes in the outer regions . this results in an underestimate of these isophote s intensity values , as seen in the surface brightness profile in the bottom panel . ] xml has several key advantages over plain file formats . for one , xml format allows an endless amount of additional information to be stored in each file that would not have fit into the standard data plus delimiter style . for example , calibrating data , such as redshift or photometric zeropoint , can be stored in each file along with the raw data with very little increase in file size overhead as the tags handle the separation . there is no need to reserve space for these quantities nor is there any problem adding future parameters to the xml format . using xml format puts all the reduction data into a single file for compactness and , in addition , since xml files are plain text files , there is no problem with machine to machine transfer . the reading of xml files is not a complication for either compiled or interpreted languages . a disadvantage to xml format is that it s clumsy to read . however , there exist a number of excellent xml editors on the market ( for example , http://www.oxygenxml.com ) . these allow a gui interface with an efficient query system to interact with the xml files . while many users would prefer to interact with the raw data files in plain text form , in fact , even a simple editor is gui window into the bytes and bits of the actual data on machine hardware . a gui xml editor is simply a more sophisticated version of vi or emacs . scans . the red ellipses are fits from the 2mass galaxy pipeline . the blue ellipses are the resulting fits from this package . the 2mass galaxy pipeline failed to follow the isophote twists ( i.e. changes in position angle ) . this results in an underestimate of these isophote s intensity values , as seen in the surface brightness profile in the bottom panel . ] an additional reason to migrate to xml is a new power that xml data files bring to data analysis . many interpreted languages ( i.e. python and perl ) have an _ eval _ or _ exec _ function , a method to convert xml data into actual variables within the code at runtime ( i.e. dynamically typed ) . this has a powerful aspect to analysis programs as one does not have to worry about formats or the type of data entries , this is handled in the code itself . dynamical typing introduces a high level of flexibility to code . in python , one can convert xml data ( using python s own xml modules to read the data ) into lists that contain the variable name and value , then transform these lists into actual code variables using an _ exec _ command . for example , .... for var , value in xml_vars : exec(var+'='+value ) .... produces a set of new variables in the running code . and python s unique try / except processing traps missing variables without aborting the routine . for example , if the variable redshift exists in the xml data file then .... try : distance = redshift / h_o except : print ' redshift undefined ' distance = std_distance .... this same try / except processing also traps overflows and other security flaws that might be used by a malicious user attempting to penetrate your server using the xml files . thus , xml brings a level of security as well as enhancing your code . lastly , another advantage to xml format is the fact that all of the reduction data ( ellipse fitting , aperture photometry , calibration information , surface photometry ) can be combined into a single file , e.g. galaxy_name.xml , which can be interrogated by any analysis routine that understands xml . a simple switch at the end of the reduction process integrates the data into an xml file for transport , or access by plotting packages , etc . if you have read this far , and are still awake , this section walks through the reduction of part of the revised shapley - ames sample ( schombert 2007 ) taken from the 2mass database that overlaps the 2mass large galaxy atlas ( jarrett 2003 ) . as a cautionary tale to the importance to doing large galaxy photometry with care , we also offer in this section a comparison of our technique with the results from an automated , but much cruder reduction pipeline from the 2mass project . allowing the ellipses to vary from isophote to isophote , not only in eccentricity , but also in position angle and ellipse center , are critical to obtaining an accurate description of a galaxy s luminosity profile . shown in figures 10 , 11 and 12 are examples of @xmath3 images extracted from the 2mass archives that were part of the large galaxy atlas ( jarrett 2003 ) . in each case , the 2mass pipeline calculates a luminosity profile based on the isophotes around an ellipse from the mean moments of the whole galaxy . thus , the fitted ellipses do not change in axial ratio or position angle and , for most spirals , this technique will result in an ellipse that is too flat in the core and , often , too round in the halo regions . if the galaxy has a bar , this technique will also underestimate the bar contribution , spreading its light into larger radii ellipses . the red ellipses are fits from the 2mass galaxy pipeline . the blue ellipses are the resulting fits from this package . the 2mass galaxy pipeline , for some strange reason , assumed a circular shape to was is clearly an elongated galaxy . this results in an underestimate of these isophote s intensity values , as seen in the surface brightness profile in the bottom panel . ] given that the light is averaged around the ellipse , this effect may be minor if the galaxy is fairly smooth and uniform . however , galaxies that are smooth and regular are a minority in the local universe . for the three examples , shown in figure 10 , 11 and 12 , the 2mass fits consistently underestimate the amount of disk light per isophote , as seen in comparison to the luminosity profile determined from the raw data using the archangel routines . this , in turn , results in fitted central surface brightnesses that are too bright in central surface brightness , and fitted disk scale lengths that are too shallow . in fact , for 49 galaxies in common between the near - ir rsa sample ( schombert 2007 ) and the 2mass large galaxy atlas , figure 13 displays the difference between the fitted disk scale lengths ( @xmath21 ) and the difference between the fitted disk central surface brightness ( @xmath22 ) . given the typical @xmath21 s , the error in 2mass fits corresponds to a 50% error in a galaxy s size . likely , errors in the central surface brightness fits averages around 0.5 mags . thus , not using the proper reduction technique not only increases the noise in the measured parameters , but produces a biased result . ) from fits of an exponential law to 2mass surface photometry versus this package s surface brightness reduction . the bottom panel displays the difference in fitted center surface brightnesses between the 2mass profiles and the new rsa sample . typically , the 2mass pipeline overestimates the eccentricity ( too round ) which results in smaller scale lengths and brighter central surface brightnesses . , width=642 ] one of the more powerful modules to the python language is the @xmath23 , the module that allows python scripts to download any url address . if address is a web page , there also exist several addition modules that parse html and convert html tables into arrays . this means a simple script can be written to pull down a web page , parse it html and extract a data into table format . and , on top of this procedure , the information could be then be used into a standard get / post web form used by many data archives . as an example , the package contains _ dss_read _ , a script that takes the standard name for a galaxy , queries ned for its coordinates and then goes to the dss website and extracts the pss - ii image of the galaxy . while this sounds like a computationally intense task , in fact the script is composed of 49 lines . the downside to this network power is , of course , the possibility of abuse . unrestricted application of such scripts will overload websites and given network speeds , the typical user does nt need their own personal digital sky at their home installation . lastly , various archives , in order to slow massive downloads , have an id / password interface . to penetrate these sites requires the _ mechanize _ module which simulates the actions of a brower , following links , parsing i d s and passwords and handling cookies . while these avatars are simple to build , the wise usage of them remains a key challenge for the future . the fastest way to learn a data reduction process is to jump in and try it . to this end , the tarball contains all the images discussed in this document , and several test images with known output . this allows the user to practice on images where the final results are known . thus , we encourage the readers to download , compile and run ! tarballs are found at http://abyss.uoregon.edu/@xmath0js/archangel . another option , for the user who does nt wish to set - up the package on their own system ( or perhaps only has a handful of galaxies to reduce ) , is the client / server version of this package available at http://http://abyss.uoregon.edu/@xmath0js / nexus ( see figure 14 ) . although more limited in its options , the web version has the advantage of speed ( it s run on a solaris sun blade ) and a fast learning curve . as to the future , a number of tools need to be added to this package . for example , quantitative morphology uses the concentration and asymmetry indices to parameterize a galaxy s global structure . while these values are easy to extract from small angular size objects , they are a challenge for large systems . yet , a detailed comparison of these values to visual morphology is a key step in understanding quantitative morphology at higher redshifts . however , in order get the current tools out to the community , the package is frozen . additional tools will be added to the package website as , in order of priority , 1 ) needed by the pi to meet various science goals , 2 ) requested by outside users to obtain their science goals , and 3 ) requested by outside users as possible new computational areas to explore . as with all evolving software , an interested user should contact the author to see where future directions lie ( [email protected] ) . this project was funded by joe bredekamp s incredible nasa s airs program . i am grateful to all the suggestions i have gotten from airs pi s at various workshops and panel reviews . the program is a mixed of technology plus science types and is one of nasa s true gems for innovative research ideas . burstein , d. , davies , r. l. , dressler , a. , faber , s. m. , stone , r. p. s. , lynden - bell , d. , terlevich , r. j. , & wegner , g. 1987 , , 64 , 601 cawson , m. g. m. , kibblewhite , e. j. , disney , m. j. , & phillipps , s. 1987 , , 224 , 557 graham , a. w. , & driver , s. p. 2005 , publications of the astronomical society of australia , 22 , 118 jarrett , t. h. , chester , t. , cutri , r. , schneider , s. e. , & huchra , j. p. 2003 , , 125 , 525 jedrzejewski , r. i. 1987 , , 226 , 747 schombert , j. m. , & bothun , g. d. 1987 , , 93 , 60 de vaucouleurs , g. 1977 , , 33 , 211 this package is a combination of fortran and python routines . the choice of these languages was not arbitrary . python is well suited for high level command processing and decision making . it is a clear and expressive language for text processing . therefore , its style is well suited to handling file names and data structures . since it is a scripting language , it is extremely portable between os s . currently , every flavor of unix ( linux , mac os x and solaris ) comes packaged with python . in addition , there is a hook between the traditional astronomy plotting package ( pgplot ) and python ( called ppgplot ) , which allows for easy gui interfaces that do not need to be compiled . the use of fortran is driven by the fact that many of the original routines for this package were written in fortran . for processing large arrays of numbers , c++ provides a faster routine , but current processor speeds are such that even a 2048x2048 image can be analyzed with a fortran program on a dual processor architecture faster than the user can type the next command . stsci provides a hook to fits formats and arrays ( called pyfits and numarray ) , but python is a factor of 100 slower than fortran for array processing . currently there are three fortran compilers in the wild , g77 , gfortran and g95 . the routines in this package can use any of these compilers plus a version of python greater than 2.3 . cfitsio is required and avaliable for all os s from its gsfc website . the python libaries pyfits and numarray are found at stsci s pyraf website . for any graphics routines , the user will need a verson of pgplot and install ppgplot as a python library . the ppgplot source is avaliable at the same website as this package . the graphics routines are only needed for data inspection , the user should probably develop their own high - level graphics to match their specifics . in the directory /util one can find all the python subroutines to fit 1d data surface or aperture photometry . the examples in this manual will guide you in constructing your own interface . lastly , the output data files for this package are all set in xml format . this format is extremely cumbersome and difficult to read ( it is basically an extension of the html format that web browsers use ) . however , a simple command line routine is offered ( @xmath24 ) that will dump or add any parameter or array out of or into a xml file . to go from a raw data frame containing a galaxy image to a final stage containing ellipse fits , surface photometry , profile fits and aperture values requires three simple scripts , _ profile _ , _ bdd _ and _ el_. the scripts _ profile _ and _ el _ are automatic and can be run as batch jobs . _ bdd _ is an interactive routine to fit the surface photometry and is a good mid - point to study the results of the ellipse fitting . in a majority of cases , the user simply needs to run those three scripts with default options to achieve their science results . .... usage : sky_box option file_name box_size prf_file options : -h = this message -f = first guess of border -r = full search , needs box_size and prf_file -t = full search , needs box_size -c = find sky for inner region ( flats ) needs x1,x2,y1,y2 boundarys output : 1st mean , 1st sig , it mean , it sig npts , iterations .... ellipse fitting routine , needs a standard fits file , output in .prf file format ( xml_archange converts this format into xml ) options : -h = this mesage -v= output each iteration -q = quiet -xy = use new xc and yc -rx = max radius for fit -sg = deletion sigma ( 0=no dets ) -ms = min slope ( -0.5 ) -rs = stopping radius -st = starting radius when deleting , output fits file called file_name.jedsub .... cursor commands : / = abort q = move to next frame c = contrast r = reset zoom z = zoom t = toggle ellipse plot p = peek at values a,1 - 9 = delete circle b = delete box .... window # 2 cursor commands : x = erase point d = disk fit only m = erase all min pts f = do bulge+disk fit u = erase all max pts e = do r**1/4 fit only b = redo boundaries p = toggle 3fit/4fit q = abort r = reset graphics / = write .xml file and exit .... add or delete data into xml format -o = output element value or array -d = delete element or array -a = replace or add array , array header and data is cat'ed into routine -e = replace or add element -c = create xml file with root element -k = list elements , attributes , children ( no data ) ....	photometry of galaxies has typically focused on small , faint systems due to their interest for cosmological studies . large angular size galaxies , on the other hand , offer a more detailed view into the properties of galaxies , but bring a series of computational and technical difficulties that inhibit the general astronomer from extracting all the information found in a detailed galaxy image . to this end , a new galaxy photometry system has been developed ( mostly building on tools and techniques that have existed in the community for decades ) that combines ease of usage with a mixture of pre - built scripts . the audience for this system is a new user ( graduate student or non - optical astronomer ) with a fast , built - in learning curve to offer any astronomer , with imaging data , a suite of tools to quickly extract meaningful parameters from decent data . the tools are available either by a client / server web site or by tarball for personal installation . the tools also provide simple scripts to interface with various on - line datasets ( e.g. 2mass , sloan , dss ) for data mining capability of imaged data . as a proof of concept , we preform a re - analysis of the 2mass large galaxy atlas to demonstrate the differences in an automated pipeline , with its emphasis on speed , versus this package with an emphasis on accuracy . this comparison finds the structural parameters extracted from the 2mass pipeline is seriously flawed with scale lengths that are too small by 50% and central surface brightness that are , on average , 1 to 0.5 mags too bright . a cautionary tale on how to reduce information - rich data such as surface brightness profiles . this document and software can be found at http://abyss.uoregon.edu/@xmath0js/archangel .
neutron stars and galactic black holes have very similar gravitational potentials , so the accretion flow in these systems might also be expected to be similar . the key difference is that neutron stars have a solid surface . therefore , they can have a boundary layer and thermal emission from the surface . additionally , where there is a strong magnetic field the accretion flow is collimated along the field lines . when the spin and magnetic axes are misaligned the surface emission should be coherently pulsed at the spin period . this is seen in the x - ray pulsars high mass x - ray binary systems where young , rapidly rotating neutron stars with a high magnetic field ( @xmath3 g ) collimate the accreting material onto the magnetic poles . however , up till 1998 no _ millisecond _ pulses were seen from the low mass x - ray binaries ( lmxbs ) , so there was no direct measure of the spin and magnetic field in such systems . this would be an important test of models of the formation of the millisecond pulsars . currently the most probable scenario for these rapidly rotating , low magnetic field ( @xmath4 g ) neutron stars is that their progenitors are the lmxbs , where the neutron star is spun up through accretion torques , which also might dissipate the magnetic field ( e.g. the review by bhattacharya & srinivasan 1995 ) . this situation changed dramatically with the discovery of the coherent x - ray pulsations with 2.5 ms period from the transient x - ray source sax j1808.43658 ( wijnands & van der klis 1998a ) . its rapid rotation and inferred magnetic field of a few @xmath4 g is in excellent agreement with the predictions of the millisecond pulsar progenitor models ( psaltis & chakrabarty 1999 ) . the key question is then what makes sax j1808.43658 so different from the rest of the lmxbs ? there are only a rather limited number of fundamental parameters which determine the accretion flow the mass of the neutron star , mass accretion rate , magnetic field , spin of the neutron star and angle between the spin and magnetic axes . these could be coupled with the age and/or evolutionary state of the system , and a further variable which could alter the appearance of the system is its inclination angle to the line of sight . one way in which to look for differences is to do a detailed comparison of the spectrum and variability of sax j1808.43658 with other neutron star lmxbs . these fall into two main categories , atolls and z sources , named after their different behaviours on a colour colour diagram ( hasinger & van der klis 1989 ) . atolls can show rather hard spectra dominated by a power law component , similar to that seen in the galactic black holes ( e.g. mitsuda et al . 1989 ; yoshida et al . 1993 ; barret et al . this hard or island state is generally seen only at mass accretion rates below a few per cent of eddington ( ford et al . 2000 ) . above this the source spectrum makes a rapid transition to a much softer spectrum , termed the banana branch , so tracing out a c ( or atoll ) shape on a colour colour diagram . conversely the z sources move from a horizontal branch , through a normal branch to the flaring branch as their mass accretion rate increases , tracing out a z shape on a colour colour diagram . the power spectral properties also correlate with atoll or z source classification , and with position on the colour colour diagram ( see e.g. the review by van der klis 1995 ) . the broad band spectrum of sax j1808.43658 is entirely typical of an atoll system in the island state , with a power law hard x - ray spectrum ( heindl & smith 1998 ; gilfanov et al . the variability power spectrum of sax j1808.43658 is also very similar to island state atoll systems ( apart from the coherent signal at the spin period ) : the broad band noise properties ( wijnands & van der klis 1998b ) and even the correlation between the noise break frequency and the qpo frequency matches onto that from other lmxbs ( wijnands & van der klis 1999 ) . thus it seems highly unlikely that there is any large scale difference between the properties of the accretion flow in sax j1808.43658 compared with other neutron star lmxb systems . yet there is a coherent spin pulse in sax j1808.43658 which is definitely not present in the other atoll systems ( vaughan et al . 1994 ; chandler & rutledge 2000 ) . there were three observed outbursts of sax j1808.43658 : in september 1996 ( in t zand et al . 1998 ) , in april 1998 ( wijnands & van der klis 1998a ) and in january 2001 ( wijnands et al . 2001 ) . here we examine the _ rxte _ spectra of sax j1808.43658 from 1998 outburst in detail , looking at both the phase averaged and phase resolved spectra at differing mass accretion rates as the outburst declines . we derive a source geometry which is consistent with the observational constraints , and finally speculate that evolutionary effects are responsible for visibility of the spin pulse . [ cols="^,<,^,^ , > , > , > , > " , ] [ tab : power_law ] we replace the phenomenological model above with a more physically motivated description of the spectrum . instead of the power law we use an approximate solution of the kompaneets ( 1956 ) equation ( zdziarski , johnson & magdziarz 1996 ) to model the thermal comptonization hard component in the spectrum . the model thcomp is parameterized by the asymptotic power - law photon index , @xmath5 , electron temperature , @xmath6 and blackbody seed photons temperature , @xmath7 . though the seed photons temperature _ can _ be independent of the soft component temperature , the simplest solution is to equalize them . later on we will argue that indeed the observed blackbody is the source of seed photons . we calculate the reflected spectrum from this continuum using a more sophisticated reflection model where both the continuum reflection and the line are calculated self - consistently for a given ionization state ( ycki , done & smith 1998 ) . the reflected spectrum ( continuum and line ) is relativistically smeared for a given inner disc radius , @xmath8 . for the soft component we use a single - temperature blackbody . again , the inclusion of the soft component is _ required _ by the data : using just a comptonization model and its reflection gives @xmath9 . allowing the absorption to be free reduces this only to @xmath10 , whereas with a blackbody the fit is @xmath11 . the same is if we use more conservative 1 per cent systematic errors in the pca spectra instead of 0.5 per cent . a fit without the soft component gives @xmath12 while adding a blackbody reduces @xmath13 to 58.2/80 . the lack of data below 3 kev in the pca means that we see only the high - energy part of the soft component , so its overall spectral shape is poorly constrained . a multi - colour disc spectrum can equally well fit the soft excess , but we choose to use a blackbody as its variability implies that it arises from a hot spot on the neutron star surface ( see section 3.4 ) . due to poor statistics , the electron temperature is poorly constrained by the hexte data , particularly for later observations when the source is fainter . from the fits we obtain lower limits on @xmath14 only , typically @xmath1530 kev . similarly , the reflector ionisation state and amount of relativistic smearing are also poorly constrained . to get the maximum possible signal to noise add observations 312 and fit this total spectrum . this gives an electron temperature of @xmath16 kev . the reflector ionisation state is @xmath17 erg cm s@xmath18 , and there is still only an upper limit on the amount of relativistic smearing . hereafter we fix the electron temperature and disk ionisation at these best fit values , and fix the inner disc radius at 50@xmath0 ( where @xmath19 ) . we fit all the phase averaged spectra using this model consisting of the thermal comptonization , compton reflection and the blackbody . fit results are presented in fig . [ fig : fit_params ] , and an example of the spectral decomposition is shown in fig . [ fig : spectrum ] . = 8.5 cm despite an order of magnitude change in the observed flux during the declining part of the outburst , the energy spectrum did not change very much ( see gilfanov et al . there is a slight softening in the hard component , starting from 24@xmath20 april onwards . the amount of reflection is always low , @xmath21 , except for the first observation , where it was about a factor of 2 times higher . the soft component decreased both its temperature and apparent area , though this did not affect the overall spectral shape very much . we use the unabsorbed best - fitting model ( without reflection ) to compute the bolometric fluxes from the hard and soft components ( fig.[fig : fluxes ] ) . the ratio of @xmath22 decreased roughly by factor two , though the errors of measurement are substantial and , to some extent , model dependent . it is however clear that the evolution of @xmath23 closely followed that of @xmath24 . thus , both components must be closely linked in some way . = 8.5 cm the comptonization model we use here is only an analytical approximation . therefore , we check it s validity and how it affects our fitting results , using compps comptonization code ( poutanen & svensson 1996 ) , which finds a numerical solution of the comptonization problem explicitly considering successive scattering orders . we fit the co - added spectrum assuming that the comptonizing cloud has a shape of a cylinder with height to radius ratio @xmath25 and the seed photons come from the bottom . we find the best - fitting electron temperature @xmath26 kev and vertical optical depth of the cylinder of @xmath27 . while the compton parameters change , the spectral shape is very similar , and the inferred soft excess is almost identical ( blackbody parameters of @xmath28 kev and @xmath29 km@xmath30 with thcomp compared to @xmath31 kev and @xmath32 km@xmath30 for compps ) . therefore , we conclude that the detailed choice of the comptonization model does not affect the fit results of the soft component . we have assumed solar abundances of the reflector , though we expect the companion star to be highly evolved and helium rich ( see e.g. ergma & antipova 1999 ) . however , the shape of the reflected spectrum in x - rays is determined mostly by heavy elements , so variation in the abundance of hydrogen and helium would affect the reflected spectrum only very little ( e.g. george & fabian 1991 ) . additionally , the amount of reflection is rather small , so the effect of abundances on the total spectrum is even weaker . the spectral resolution and response uncertainties of the pca do not allow us to investigate this issue in detail . next , we check for the presence of the compton reflection from the neutron star surface . depending on the equation of state we should expect this reflection to be red - shifted by @xmath33 = 0.070.77 ( miller , lamb & psaltis 1998 ) . this means that the most prominent reflection features , the iron k line and edge , can appear between 3.6 and 6.0 kev . unfortunately , this coincides with the pca instrumental xenon edge at 4.8 kev , therefore the following results should be treated with caution . we add another reflection component in our model with redshift allowed to be free between the above limits and fit the data . if the neutron star surface is fairly highly ionized , with @xmath34 erg cm s@xmath18 , we find an upper limit for @xmath35 for a redshift range of @xmath33 = 0.070.77 . however , if the neutron star surface is completely ionized , the reflected spectrum is simply a featureless power law with high - energy cutoff around 30 kev and the same spectral index as irradiating comptonization . in such a case we can not disentangle the irradiating and reflected continuum , so there are no constraints on the amplitude of the reflection from the neutron star . the phase - resolved spectra with the lowest and highest flux are shown in fig . [ fig : minimax ] . there is a slight softening of the highest flux spectrum at energies below @xmath36 5 kev , but otherwise the spectral shape is remarkably constant . to investigate the details of spectral evolution with the pulse phase we fit the phase - resolved spectra by the blackbody plus thermal comptonization model ( thcomp ) , the same we used in section [ sec : phase - averaged ] to fit the phase - averaged spectra . we fit each spectrum of the 16 phase bins and find the fit parameters and their errors . the results are presented in fig . [ fig : phase_free ] , with an average @xmath37 per spectrum . = 8.5 cm we clearly see that some parameters vary more than the others , in particular reflection ( @xmath38 ) is consistent with a constant while @xmath39 exhibits strong sinusoidal pulsations . we want to distinguish between the parameters that vary _ significantly _ and those in which variability can be due to statistical noise . therefore , we quantify the variability of a given parameter by an _ intrinsic _ variance . let @xmath40 and @xmath41 be a given parameter s best - fitting value and variance , in the @xmath42-th phase bin ( @xmath43 ; here @xmath44 ) . then , the intrinsic variance of this parameter is estimated by @xmath45 where @xmath46 the reduced @xmath47 results from fitting a series of @xmath40 ( for @xmath43 ) by the constant . when @xmath48 we assume @xmath49 . we estimate error of @xmath50 by propagating data errors , @xmath51 when @xmath50 is zero , then any variability is completely consistent with having a statistical origin so that there is no significant intrinsic variability in the parameter . in such a case we decide that the parameter is intrinsically constant , and fix it at its average value for the next series of fits . using this we found that only two parameters exhibited significant intrinsic variability : @xmath39 and @xmath52 ( see fig.[fig : phase_free ] ) . we fixed @xmath5 and @xmath38 ( but not @xmath53 for a moment ) at their average values and refit the spectra ( average @xmath54 , so a difference of @xmath55 for 2 fewer free parameters ) . this showed an anti - correlation between the pulse profiles of @xmath52 and @xmath53 , which is most probably a sign of degeneracy in the soft component parameterization . when we additionally fix @xmath56 the average @xmath47 rises only to 48.2/47 , however fixing @xmath52 instead causes significant worsening of the fits with average @xmath47 = 54.1/47 , which means that temperature pulsation is preferred over normalization pulsation . however , from the physical point of view we should expect more variability in apparent area ( normalization ) then in temperature ( see discussion in section [ sec : hotspot ] ) . we see only wien part of the blackbody spectrum in the pca , so there is some degeneracy in the model , namely in the soft component temperature , normalization and the hard component low - energy turnover . this creates model - dependent uncertainty . therefore , we decide to chose the model with worse @xmath13 but better physical foundation , and to fix @xmath52 , letting only the normalizations of the soft and the hard components ( @xmath39 and @xmath53 ) to be free , and fit the data in each phase bin again . the final result is presented in fig . [ fig : phase_param]@xmath57 . the pulse profiles of @xmath39 and @xmath53 are shifted in phase in a way that the soft component lags the hard one . this will manifest itself as an energy dependent time lag in which soft x - rays lag hard x - rays , as reported in previous work ( cui et al . 1998 ) . in fig . [ fig : lags ] we show the time lags taken from cui et al . ( 1998 ) and the prediction of our two - component model in which only the normalizations @xmath39 and @xmath53 are free to vary . up to about 7 kev , as the contribution from the soft component decreases , the soft time lag quickly increases . above this energy , where the soft component is negligible , the time lag flattens , and shows the @xmath36 8 per cent shift in phase ( about 200 @xmath63s ) between the soft and hard components . finally , we use the model with only @xmath53 and @xmath39 free to fit the phase - resolved spectra for each of the observation from 1 to 12 . we measure the intrinsic variance of both of the fit parameters , and show the result in fig . [ fig : rms ] . the intrinsic variance of @xmath39 increases monotonically from @xmath36 4.5 to @xmath36 6 per cent during the declining part of the outburst . the behaviour of the @xmath53 variance is more complex : it slowly rises to its maximum around 18 april , then declines only to rise again after 26 april . our approach contrasts with that of ford ( 2000 ) in that we use standard background subtraction techniques to extract the phase resolved spectra whereas he used the phase minimum spectrum as the persistent emission and subtracted this from the phase - resolved spectra to get the spectrum of the pulsed component . however , our results clearly show that there are _ two _ pulsed components , the hard and the soft , and that these _ do not _ vary in phase . ford ( 2000 ) fit the pulsed emission by a power law , and noticed significant phase variability of the spectral index . however , in this approach a single power law tries to follow the overall spectral shape , which is actually more complex . additionally , poor statistics of the persistent - emission - subtracted spectra make fitting more difficult . we have reproduced this approach but fitted these pulsed spectra with a model consisting of a power law _ and _ a blackbody . interestingly , the power - law index was consistent with remaining constant with phase , while the blackbody and power - law normalizations were shifted in phase , as in our analysis . the properties of the power - density spectrum of sax j1808.43658 were described in details by wijnands & van der klis ( 1998b ) . here we re - fit the pds selected from the observations 112 ( table [ tab : obslog ] ) , using similar model , in order to obtain break and qpo frequencies , corresponding to the exact times at which the data were taken . we show the results in fig . [ fig : pow_param ] , for completeness . none of the two frequencies is correlated with the x - ray flux ( fig . [ fig : fluxes ] ) , with any of the spectral parameters ( fig . [ fig : fit_params ] ) , nor with spectral colour . while the spectral parameters either remain constant or change monotonically , both break and qpo frequencies reach their minimum around 18 april . after 24 april we do not detect qpo , but the break frequency reverses its rise around 26 april . a similar behaviour ( though in opposite direction ) can be seen in the soft component variance ( fig . [ fig : rms ] ) . thus , there is a clear anti - correlation between @xmath64 and @xmath65 . we now have many interrelated constraints , which can be used to determine the geometry of the system , and how this changes as the outburst declines . since both hard and soft x - rays pulsate then they must be affected by the spin of the neutron star . we assume that the underlying geometry involves a hotspot on the surface of the neutron star to produce the soft pulsed component , and a hard x - ray shock above the magnetic poles to produce the pulsed hard x - ray component . the constancy of the hard x - ray spectral shape with spin phase implies that the hard x - rays are from a single type of emission region , rather than being the sum of a pulsed and unpulsed component . the sinusoidal light - curve for the soft component as a function of spin phase strongly implies that we see only one hotspot . the shape of the hard component light - curve probably implies that we do nt see the second shock . more constraints come from the variability characteristics . it is commonly accepted that the qpo frequency is related to the inner disc radius in some way . the inner disc radius , together with the shock height and the neutron star radius , determines the solid angle subtended by the disc , and so the amount of reflection and seed photons from the disc . seed photons can also arise from the hotspot on neutron star surface , and their contribution is determined mostly by the shock height . but the spectral shape gives us observational constraints on the amount of reflection from the disc , and the amount of the seed photons , so there should be a unique solution to the geometry of the system . in practice of course , there are significant observational and theoretical uncertainties , but we use these observational constraints to develop a consistent geometrical picture for the emission from sax j1808.43658 . the 3200 kev spectrum _ requires _ a two component model . the soft component is consistent with a blackbody at @xmath36 0.50.7 kev . the hard component can be described by thermal comptonization of these blackbody seed photons in plasma of optical depth of @xmath36 2 and temperature of @xmath36 40 kev . the soft and hard component fluxes are very closely correlated ( see fig . [ fig : fluxes ] ) . thus the soft component is most probably from reprocessing of the hard component . we further discuss relation between both components in appendix [ sec : energy_balance ] . we consider the source of the soft component as a simple circular hotspot on the surface of the neutron star . we take into account doppler shifts and geometry in the rotating system of the neutron star . we demonstrate , how in the frame of this simple model the data can constrain inclination angle of the system and the size of the hotspot . the pulse profile of the soft component is consistent with a single sinusoid . this implies that we _ do not _ see the second spot , since it would have given rise to a secondary maximum in the profile . this , in turn , yields a constrain on the spot radius , @xmath66 where @xmath42 is the inclination angle of the observer from the rotation axis , @xmath67 is the angle from the rotation axis to the spot at the magnetic pole and @xmath68 is the neutron star radius . the inferred luminosity of a spot depends on both its temperature and projected area . both of these change as a function of phase . the doppler shift gives a ratio between the maximum and minimum temperature of @xmath69 where @xmath70 is the equatorial rotation speed of the neutron star . the ratio of @xmath71 is only weakly affected by the spot size . the maximum to minimum ratio of the projected area is @xmath72 where the maximum area is offset in phase by @xmath1 from the maximum doppler shift . this formula is exact even for a large circular spot , again assuming that we do not see the spot from the second pole . when the temperature of the soft component is a free parameter in the fits to the phase - resolved spectra ( section [ sec : phase - resolved ] ) then it yields large modulation in this temperature , with @xmath73 ( see fig . [ fig : phase_free ] ) . however , physically we expect a rather small variation in temperature from the doppler shift , since the offset angle between the magnetic spin and the rotation axis is probably small ( ruderman 1991 ) . assuming an inclination angle of @xmath74 and offset angle of @xmath75 we find @xmath76 , much less then observed . furthermore , this would give an inferred luminosity modulation of about 1.08 , much less than the observed minimum to maximum ratio of 1.7 . therefore , doppler effects alone are _ unlikely _ to be the predominant cause of the variability , contrary to ford ( 2000 ) . conversely , for these parameters the projected area variability results in a maximum to minimum luminosity ratio of 1.9 . thus it seems likely that the majority of the luminosity change should come from the change in projected area while the temperature remains mostly constant . to check consistency of these predictions with the data , we fit the phase - resolved spectra leaving only @xmath39 and @xmath53 as free parameters , but additionally enforcing sinusoidal variation of @xmath52 with maximum to minimum ratio of 1.02 , as predicted using equation ( [ eq : t_maxmin ] ) . as a result , maximum to minimum ratio of @xmath53 is 1.56 . we use these numbers to constrain the inclination and offset angles . we solve equation ( [ eq : area ] ) with @xmath77 . this gives us one to one relation between @xmath42 and @xmath67 ( fig . [ fig : iarh]@xmath57 ) . for the inclination angle of @xmath78 the corresponding offset is @xmath79 . equations ( [ eq : rhs ] ) and ( [ eq : area ] ) then give an upper limit for the hotspot radius ( fig . [ fig : iarh]@xmath80 ) . we find @xmath81 , or @xmath82 km , assuming @xmath68 of 10 km . the hotspot emission gives us an independent way to estimate its size . from the blackbody normalization we estimate its average projected area as @xmath36 30 km@xmath30 ( see fig . [ fig : fit_params ] ) . this is a lower limit as some of the hotspot could be obscured by the shock , and we have not included gravitational redshift . nonetheless , for an inclination of @xmath83 this yields @xmath84 km . given that we then have both an upper and lower limit for the spot size of @xmath85 km , we take this as our estimate of the spot size . for the maximum possible inclination angle of @xmath86 the required offset is small , @xmath87 . this yields in turn @xmath88 , which is not consistent with our lower limit on the hotspot size from the normalization of the blackbody . therefore , we find the inclination angle of @xmath89 the most likely . our discussion neglects general relativistic effects . these will change the quantitative results , but for our model then the effects of light bending are not particularly strong . full relativistic calculations by psaltis , zel & dedeo ( 2000 ) show that the inferred area of the spot is within 10 per cent of its intrinsic area for our model geometry , where the spot is large and not eclipsed . thus the qualitative constraints derived here on the geometry should be valid . we consider the source of the hard component as an x - ray emitting shock in the accretion column . we again take into account doppler shifts and geometry . the hard component pulse profile , like the soft one , lacks the secondary maximum , which implies that we _ do not _ see the second shock . this immediately yields strong constraints on the shock height : @xmath90 the solution to this equation , solved simultaneously with the offset - inclination relation ( [ eq : area ] ) derived above is shown in fig . [ fig : iarh]@xmath91 . for a small offset angle of @xmath92 the shock must be short , @xmath93 . the hard x - ray spectrum can be described as a power law with photon spectral index @xmath94 . we can expect it to be almost isotropic . then , in the non - relativistic limit the doppler shift gives the ratio of maximum to minimum observed hard flux , as @xmath95 for @xmath79 and @xmath78 this ratio is 1.07 , while the observed maximum to minimum is 1.16 . an additional source of variability could be reflection from the neutron star surface and/or anisotropy of the shock emission . the reflection from the fully ionized surface would be spectrally indistinguishable from the irradiating continuum and so is consistent with our spectral fits ( see appendix [ sec : ns_reflection ] ) . for the above inclination and offset angles the reflection component should have maximum to minimum flux ratio of about 1.2 . _ unlike _ the soft component pulse profile , the hard one is not symmetric : it is skewed . in section [ sec : phase - resolved ] we have fitted this profile with two sinusoids : a dominating one with the spin frequency and a weaker one , with double - spin frequency . this can be explained if the cross - section of the shock is elongated , with bi - axial symmetry , which would create a double - spin frequency on top of the main , single - spin frequency pulse profile . in reality , magnetic - field induced column accretion onto the neutron star surface might create a latitudaly - elongated , crescent - shaped shock cross - section . there are two main sources of phase - dependent variability : doppler shifts and variation in the projected area . as the spot goes around the spin pole , the projected area is maximized when the spot is closest to the observer , while the doppler shifts are maximized when the spot is midway between its closest and furthest points , as this is where its velocity along the line of sight is largest . both doppler and projected area effects should affect the source spectrum . however , the spin pulse variability of the soft component is probably dominated by the projected area terms ( see section [ sec : hotspot ] ) . if the hard component were dominated by doppler shift then the two components should be shifted in phase by 90@xmath96 . however , should be some projected area effects in the hard emission , both from observing the translucent column at differing angles and from the presence of highly ionized reflected continuum emission ( see section [ sec : shock ] ) . this means that the hard spectrum varies as the weighted sum of these components , one which peaks as the doppler emission and one which peaks as the effective area . thus the maximum of the hard emission moves closer to the soft emission . while detailed modelling depends on the unknown geometry of the accretion column , its seems possible that this can result in the observed phase lag of @xmath97 ( fig . [ fig : phase_param ] ) . this is a very different interpretation to the observed energy dependent phase lag than that of ford ( 2000 ) . ford ( 2000 ) used only a single blackbody component at @xmath98 kev as the pulsing spectrum this does give energy dependent phase lags , but of course contributes very little to the hard spectrum . although he does get 10 per cent variability at 20 kev , this is measured with respect to the blackbody rather than to the total spectrum . since the blackbody flux contributes only a tiny fraction ( @xmath99 ) of the total flux at 20 kev , his 10 per cent variation of the blackbody flux predicts only a @xmath100 per cent modulation at 20 kev , so can not explain the observed phase lags in the total spectrum . by contrast , our model can explain the observed time lags , as shown in fig.[fig : lags ] . in particular , it predicts flattening of the time lag above @xmath36 7 kev , where the soft component is negligible . the key assumption required by our model is its two - component nature . there is the soft and hard spectral component and they pulsate _ shifted in phase_. whatever is the actual geometry , this phase shift is required to see the time lags . the 250 kev luminosity of sax j1808.33658 changed by an order of magnitude during the decline from outburst , yet the spectral shape remained remarkably constant . however , neutron star binaries in general show a dramatic transition from hard to soft at a 250 kev luminosity @xmath36 0.01 of @xmath101 ( ford et al . 2000 ) , corresponding to a bolometric luminosity of a few per cent of eddington . this spectral switch is startlingly like that seen in gbhc at low bolometric luminosities ( e.g. nowak 1995 ) . this similarity makes it likely that the underlying mechanism for the transition is the same irrespective of the nature of the compact object , so implying a connection to the accretion flow itself rather than to anything to do with a magnetosphere or solid surface . assuming the distance of 2.5 kpc ( in t zand et al . 2001 ) the 250 kev flux during the outburst changed from 0.01 to 0.001 of @xmath102 ( where @xmath103 erg s@xmath18 ) , so it is likely that sax j1808.33658 never became quite bright enough to change from the hard ( island ) state to the soft ( banana ) state . the reason for the low peak outburst luminosity is easy to find . the binary is a very small system with an orbital period of only 2.1 hours ( chakrabarty & morgan 1998 ) , so the disc outer radius ( set by tidal truncation ) is also small ( @xmath104 cm ) . this limits the total disc mass @xmath105 which can build up in quiescence , so when the outburst is triggered and all this mass flows in on the viscous time - scale @xmath106 , then the mass accretion rate onto the central object is @xmath107 ( king & ritter 1998 ; shahbaz , charles & king 1998 ) . as the outburst declines there is a rather steep drop in the light - curve , after 26@xmath20 of april ( fig.[fig : lightcurve ] ) . this has been identified with the centrifugal barrier ( gilfanov et al . however , pulsed emission is _ always _ seen , so the magnetic field is not inhibiting accretion . given that the system is transient we suggest instead that the drop is associated with the onset of the cooling wave , switching the disc back into quiescence ( e.g. king & ritter 1998 ; dubus , hameury & lasota 2001 ) . the variance of the soft and hard components change in a rather different way as a function of time ( see fig . [ fig : rms ] ) . the hard variance monotonically increases , while the soft variance follows the inner disc radius ( assuming that this relates to the qpo frequency ) . if the disc is truncated by the magnetic field then the point at which the material connects on is the same as the inner disc radius . a larger disc radius means that the material connects onto field lines which are closer to the magnetic pole , so the projected area effects decrease , so the soft variance decreases ( e.g. frank , king & raine 1992 ) . this is exactly _ opposite _ to the observed behaviour ( figs . [ fig : rms ] and [ fig : pow_param ] ) , strongly suggesting that the inner disc radius is truncated by some means other than that of the magnetic field . since other atolls ( and galactic black holes ) show similar qpo behaviour then the mechanism must be an intrinsic property of the accretion flow . if the optically thick disc truncates at some radius into an inner , optically thin flow ( e.g. raska 1999 ; meyer , liu & meyer - hofmeister 2000 ) , then the point at which this optically thin flow connects onto the magnetic field is determined by the ratio of its ram pressure to the magnetic pressure ( e.g. frank , king & raine 1992 ) . but if the inner disc retreat is connected with increased evaporation into the hot flow ( raska 1999 ) then its density , and so its ram pressure , will be inversely related to the inner disc radius . as the inner edge of the optically disc recedes then the ram pressure of the hot flow increases so it penetrates further into the magnetosphere before connecting onto the field lines . the material then collimates onto a spot which is further from the magnetic pole , increasing the effective angle between the spot and the spin axis , so increasing the soft variability . the hard component should then also increase its variance . however , as the mass accretion rate goes down , the shock optical depth probably decreases , while its height probably increases . the amount of variance apparently depends on the geometry and optical depth of the shock . in principle it is possible that these effects could contribute to the observed variance more then the position of the accretion column , and so explain the observed monotonic increase in variance . a geometry which is consistent with all the above constraints is one in which a neutron star with magnetic field @xmath4 g accretes from a disc which evaporates into an inner hot flow at some radius which is set by the overall evolution of the disc . this inner hot flow falls inward until its ram pressure is comparable with that of the magnetic field , after which point it is collimated by the field . it then free falls onto the magnetic poles , forming an accretion column ( see fig . [ fig : ns ] ) . all the other source properties are consistent with this geometry ( see appendices ) . the seed photons for comptonization in the shock are probably from the surface of the neutron star , while the disc emission is negligible ( appendix [ sec : seed_photons ] ) . energy balance shows that these seed photons can not be only due to irradiation of the surface by the hard x - rays from the shock . either reflection from the surface and/or re - processing of the kinetic energy of the in - falling material is necessary to support the sufficient amount of the seed photons ( appendix [ sec : energy_balance ] ) . the reflection of the hard x - rays from the neutron star surface , though not detected , can be present in the data as a featureless cutoff power law ( appendix [ sec : ns_reflection ] ) . and finally , we applied one of the qpo models to the data and estimated the inner disc radius to vary between @xmath36 15 and @xmath108 , consistent with the observed amount of the disc reflection ( appendix [ sec : disc ] ) . we speculate that the true distinction between this and other lmxbs is purely an evolutionary effect . sax j1808.43658 is highly evolved , the companion star is mostly evaporated / accreted . the mass transfer rate is very low , and has been low for a long time ( king 2000 ) . models of accretion onto a magnetized neutron star suggest that the magnetic fields of up to @xmath109 g ) are buried by inflow where the accretion rate is a few per cent of eddington . the field can diffuse back out , but only on time - scales of 1001000 years ( cumming , zweibel & bildsten 2001 ) . only systems where the time averaged mass transfer rate has been lower than this for a few thousand years can have a magnetic field which is high enough to collimate the accretion flow . we analyse spectra from sax j1808.43658 from the whole _ rxte _ campaign of 1998 . this follows the april 1998 outburst as the luminosity declines by a factor of over 10 . we fit physically motivated models to the phase - averaged spectrum from each observation , showing that the continuum spectrum can be described by thermal compton scattering of soft seed photons . this spectrum is reflected from the accretion disc , producing a weak reflected continuum and iron k@xmath110 line as has been seen in other hard state ( low mass accretion rate ) lmxbs ( yoshida et al . 1993 ; barret et al . 2000 ) . however , there is also an additional component in the spectrum which can be fit by a blackbody . in phase - resolved spectroscopy it is this component which is most variable . the intensity of the soft component changes as the outburst progresses in the way expected for it to be produced by reprocessing of in - falling material energy on the neutron star surface . we show that the spectrum is consistent with a geometry where the disc is truncated at fairly large radii , the hard x - ray emission is from a shocked accretion column where the magnetically collimated flow impacts onto the neutron star surface , and the blackbody component is from the hotspot on the neutron star surface ( see fig . [ fig : ns ] ) the disc can not extend down to the last stable orbit as plainly the inner flow is collimated by the magnetic field . the truncation radius of the optically thick disc can be estimated by the qpo and break frequency seen in the variability power spectra assuming that these correspond to vertical perturbations at the inner disc edge ( psaltis & norman 2001 ) . the amount of reflection can also give an independent estimate for the inner disc edge . the two results are consistent , although the uncertainties are large , supporting the relativistic orbits interpretation of the qpos . the inner radius evolution ( as measured by the qpo frequency ) is not at all consistent with the magnetic field truncating the disc . it seems much more likely that the accretion disc truncates spontaneously , and that the resulting inner hot accretion flow is then collimated by the magnetic field at rather smaller radii . this has obvious application to other relativistic disc accreting systems , both neutron stars and black holes . these have very similar variability power spectra at low mass accretion rates to that of the millisecond pulsar . if the qpo / break frequency in the millisecond pulsar indicates the inner disc radius then it also shows that the disc is generally truncated in all the other disc accreting systems at low mass accretion rate . given that we argue above that the disc in sax j1808.43658 spontaneously truncates , then it seems likely that the same mechanism ( whatever that is ) operates generically in all low mass accretion rate discs . finally , we speculate that the reason sax j1808.43658 is currently unique in showing spin pulsations is due to its evolutionary state . all the well known lmxb neutron star systems have a higher time averaged mass accretion rate which can bury the magnetic field below the surface . summarizing the results from this paper , we find as follows : * the x - ray spectrum of sax j1808.43658 consists of the two components . * the soft component is consistent with the blackbody . we find that most likely this emission originates from the hotspot on the neutron star surface , with radius @xmath85 km . * the hard component is consistent with comptonization of the seed photons from the hotspot in a plasma of intermediate optical depth and temperature of @xmath111 kev . this is probably a short ( @xmath112 ) shock in the accretion column where the accreting material collimated by the magnetic field impacts onto the neutron star surface . * the hard component includes also weak compton reflection from the accretion disc . the amount of reflection implies that the disc is truncated at fairly large radii ( 2040 @xmath0 ) , consistent with the lack of relativistic smearing of the spectral features . * both components pulsate shifted in phase , while the neutron star rotates . this creates the energy - dependent time lags in which the soft photons lag the hard ones . * the inner disc radius derived from the break and qpo frequencies is consistent with the inner disc radius derived from the solid angle subtended by the reflected spectrum . it evolves independently of the instantaneous accretion rate , implying that the inner disc radius is independent of and larger then the magnetospheric radius . the magnetic field does not truncate the disc in sax j1808.43658 . instead , the inner disc radius is set by some much longer time - scale process , most probably connected to the overall evolution of the accretion disc . * the disc truncation mechanism is probably generic to low mass accretion rate flows both in disc accreting neutron stars and black hole systems . * we do not detect compton reflection from the neutron star surface , though we predict that it should be present in the x - ray spectrum . this would give an unambiguous observational measure of @xmath2 if there is any iron on the neutron star surface which is not completely ionized . this research has been supported in part by the polish kbn grants 2p03d00514 , 2p03d00614 , by the foundation for polish science fellowship 15,4/99 and by a polish - french exchange program . arnaud k. a. , 1996 , in jacoby g. h. , barnes j. , eds . , astronomical data analysis software and systems v. asp conf . series vol . 101 , san francisco , p. 17 barret d. , olive j. f. , boirin l. , done c. , skinner g. k. , grindlay j. e. , 2000 , apj , 533 , 329 bhattacharya d. , srinivasan g. , 1995 , in lewin w. h. g. , van paradijs j. , van den heuvel e. p. j. , eds . , x - ray binaries , cambridge university press , cambridge , p. 495 bildsten l. , chakrabarty d. , 2001 , astro - ph/0104153 chakrabarty d. , morgan e. h. , 1998 , nature , 394 , 346 chandler a. m. , rutledge r. e. , 2000 , apj , 545 , 1000 churazov e. , gilfanov m. , revnivtsev m. , 2001 , mnras , 321 , 759 cook g. b. , shapiro s. l. , teukolsky s. , 1994 , apj , 424 , 823 cui w. , morgan e. h. , titarchuk l. g. , 1998 , apj , 504 , l27 cumming a. , zweibel e. g. , bildsten l , 2001 , astro - ph/0102178 dubus g. , hameury j .- m . , lasota j .- p . , 2001 , a&a , 373 , 251 ergma e. , antipova j. , 1999 , a&a , 343 , 45l friedman j. l. , ipser j. r. , parker l. , 1986 , apj , 304 , 115 ford e. c. , 2000 , apj , 535 , l119 ford e. c. , van der klis m. , mndez m. , wijnands r. , homan j. , jonker p. g. , van paradijs j. , 2000 , apj , 537 , 368 frank j. , king a. , raine d. , 1992 , accretion power in astrophysics , cambridge university press , cambridge friedman j. l. , parker l. , ipser j. r. , 1986 , apj , 304 , 115 george i. m. , fabian a. c. , 1991 , mnras , 249 , 352 gilfanov m. , revnivtsev m. , sunayev r. , churazov e. , 1998 , a&a , 338 , 83 hasinger g. , van der klis m. , 1989 , a&a , 225 , 79 heindl w. a. , smith d. m. , 1998 , apj , 506 , 35 in t zand j. j. m. , heise j. , muller j. m. bazzano a. , cocchi m. natalucci l. , ubertini p. , 1998 , a&a , 331 , l25 in t zand j. j. m. , et al . , 2001 , astro - ph/0104285 king a. r. , 2000 , mnras , 315 , l33 king a. r. , ritter h. , 1998 , mnras , 293 , l42 kompaneets a. s. , 1956 , soviet phys . , jetp 31 , 876 magdziarz p. , zdziarski a. a. , 1995 , mnras , 273 , 837 meyer f. , liu b. f. , meyer - hofmeister e. , 2000 , a&a , 361 , 175 mndez m. , van der klis m. , ford e. c. , wijnands r. , van paradijs j. , 1999 , apj , 511 , l49 miller m. c. , lamb f. k. , psaltis d. , 1998 , apj , 508 , 791 mitsuda k. , inoue h. , nakamura n. , tanaka y. , 1989 , pasj , 41 , 97 morrison r. , mccammon d. , 1983 , apj , 270 , 119 nowak m. a. , 1995 , pasp , 107 , 1207 poutanen j. , svensson r. , 1996 , apj , 470 , 249 psaltis d. , chakrabarty d. , 1999 , apj , 521 , 332 psaltis d. , norman c. , 2001 , astro - ph/0001391 psaltis d. , zel f. , dedeo s. , 2000 , apj , 544 , 390 raska a. , 1999 , mnras , 308 , 751 ruderman m. , 1991 , apj , 366 , 261 shahbaz t. , charles p. a. , king a. r. , 1998 , mnras , 301 , 382 stella l. , vietri m. , 1998 , apj , 492 , l59 van der klis m. , 1995 , in lewin w. h. g. , van paradijs j. , van den heuvel e. p. j. , eds . , x - ray binaries , cambridge university press , cambridge , p. 252 vaughan b. a. , et al . , 1994 , apj , 435 , 362 wardziski g. , zdziarski a. a. , 2000 , mnras , 314 , 183 wijnands r. , van der klis m. , 1998a , nature , 394 , 344 wijnands r. , van der klis m. , 1998b , apj , 507 , l63 wijnands r. , van der klis m. , 1999 , apj , 514 , 939 wijnands r. , mndez m. , markwardt c. , van der klis m. , chakrabarty d. , morgan e. , 2001 , astro - ph/0105446 wilson c. d. , done c. , 2001 , mnras , 325 , 167 yoshida k. , mitsuda k. , ebisawa k. , ueda y. , fujimoto r. , yaqoob t. , done c. , 1993 , pasj , 45 , 605 zdziarski a. a. , johnson w. n. , magdziarz p. , 1996 , mnras , 283 , 193 ycki p. , done c. , smith d. a. , 1998 , apj , 496 , l25 here we quickly discuss the possible sources of the seed photons for comptonization . let us consider a comptonizing region eleveted at height @xmath113 above the pole of the neutron star of radius @xmath68 . it can receive the seed photons from the accretion disc and from the hotspot on the surface . we consider the shakura - sunyaev disc extending from @xmath8 to infinity , around the neutron star of mass @xmath114 and with accretion rate @xmath115 . the circular hotspot of radius @xmath116 and temperature @xmath117 is located on the spin pole ( for these estimations only ) . we assume @xmath114 = 1.4m@xmath118 and @xmath119 g s@xmath18 . the energy density of the photons from the disc in the comptonizing region is @xmath120,\ ] ] where @xmath121^{1/2}$ ] . the energy density of the photons from the hotspot in the comptonizing region is @xmath122.\ ] ] for a short shock , @xmath123 , the hotspot of radius @xmath124 and the inner disc radius of @xmath125 , @xmath126 . the two quantities are comparable only if the shock is tall ( @xmath127 ) , which we find unlikely ( see section [ sec : shock ] ) . the energy density of the self - absorbed synchrotron seed photons in plasma of temperature @xmath128 and magnetic field @xmath129 g is @xmath130 where @xmath131 hz is the turnover frequency ( wardziski & zdziarski 2000 ) . for @xmath132 = 0.1 , @xmath133 and @xmath134 g we find @xmath135 . therefore , we conclude that the dominating source of the seed photons for comptonization is the hotspot on the neutron star surface , which is at the same time a source of the soft component ( blackbody ) seen in the x - ray spectrum . let us consider a source of the hard component located above the neutron star surface and irradiating it . a fraction of the irradiating luminosity is then re - processed and re - emitted as seed photons for comptonization . here we check whether our assumed geometry is consistent with the energy balance in such a comptonizing closed cycle . the luminosity of the hard component is @xmath136 . a fraction @xmath137 of it is intercepted by the surface , where @xmath138 is the solid angle subtended by the surface to the hard x - ray source , and @xmath139 accounts for intrinsic anisotropy of comptonization . the surface has albedo @xmath140 , so the power in the re - processed photons is @xmath141 . on the other hand the comptonizing cloud amplifies the seed photons , giving rise to the hard component luminosity , @xmath142 , where @xmath143 is the solid angle subtended by the hard x - ray source to the hotspot and @xmath144 is compton amplification factor . hence , @xmath145 the maximum albedo here is @xmath146 because the hard @xmath147 kev photons are compton down - scattered and can not be completely reflected . however , there is also a minimum albedo of @xmath148 since the observed temperature is @xmath36 0.50.7 kev , and so the material is probably highly ionized from collisional ionization . we have performed monte carlo simulations of the observed spectrum and found @xmath149 and @xmath150 ( in the slab geometry ) . then , the photons re - processed from irradiation can account only for about @xmath151 of the seed photons . both solid angles are less than @xmath152 ( though not much less for a short shock ) , so 0.4 constitutes a firm upper limit on @xmath153 . therefore , there must be an additional source of the seed photons . since they must come from the neutron star surface ( appendix [ sec : seed_photons ] ) , we sugguest the following two possibilities . first , the additional seed photons can originate from reflection from the neutron star surface ( see appendix [ sec : ns_reflection ] ) . due to large albedo the luminosity carried by the photons compton reflected from the surface is roughly four times luminosity in the re - processed photons . second , the in - falling material kinetic energy can be directly re - processed into heating of the hotspot , additionally to the irradiation , and this process can give rise to the blackbody seed photons . we note that an anisotropy of the hard x - ray source ( e.g. due to bulk motion comptonization ) can not boost the re - processed emission enough to reach the equilibrium . even if all of the hard x - ray luminosity was directed towards the neutron star ( @xmath154 ) , then @xmath153 would be still significantly less then 1 . in section [ sec : phase - averaged ] we have derived an upper limit on the reflection from a mildly ionized star , @xmath155 . such a small solid angle subtended by the star would require an irradiating source at a distance @xmath156 from the surface , implying a _ very tall _ shock , which we find unlikely . a more likely short shock of @xmath157 subtends an angle @xmath158 . this can be in agreement with the observations only if the reflected spectrum is featureless , which means that either the reflecting spectrum is completely ionized or the heavy elements , including iron , sank under the surface in the strong gravitational field , so the surface could act as a mirror . the component reflected from the surface would then contribute to the total spectrum as a power - law of the same index as the irradiating comptonization and a high - energy cutoff above @xmath36 30 kev . such a component would be impossible to disentangle from the comptonization in _ rxte _ data . here we consider a particular qpo model and check whever the results obtained from it are consistent with the our assumed geometry of sax j1808.43658 . if the ( low frequency ) qpo is connected to the precession time - scale of a vertical perturbation in the disc ( stella & vietri 1998 ; psaltis & norman 2001 ) then its frequency is uniquely linked to some transition radius ( we assume it is the inner radius @xmath159 ) in the disc , as ( psaltis & norman 2001 ) @xmath160 where @xmath161m@xmath118 is the mass and @xmath162 is the spin on the neutron star . all current equations of state yield @xmath163 ( friedman , parker & ipser 1986 ; cook , shapiro & teukolsky 1994 ) , and we assume @xmath164 . since we do not observe the qpo in all of our observations , we use the empirical relation between the break and qpo frequencies from the fit to the data , @xmath165 to estimate all the qpo frequencies . this , together with equation ( [ eq : qpo ] ) implies the inner disc radius varying between 1729@xmath0 ( @xmath166 ) or 2645@xmath0 ( @xmath167 ) . these rather large radii are consistent with the rather small amount of relativistic smearing of the reflected spectral features which require that the disc is always larger than @xmath168 ( see section [ sec : phase - averaged ] ) . the solid angle subtended by the disc as measured by the reflected spectrum gives us an independent way to estimate the inner disc radius . the hard x - ray emission can only illuminate half the disc in our geometry where most of the hard x - ray emission comes from a short shock above the magnetic pole . then we derive the inner disc radius as shown in fig . [ fig : radref ] , which is consistent with the radii derived from the qpo frequencies . the behaviour of the inner disc radius is rather complex . it evolves on the long time - scale of days . it first recedes as the outburst declines , then comes in again and then recedes ( fig . [ fig : radref ] ) , rather than being uniquely related to the instantaneous mass accretion rate onto the compact object . this lack of correlation with the mass accretion rate is generally seen in other lmxbs , e.g. 4u 160852 ( mndez et al . 1999 ) , showing that the key determinant of the behaviour of the system is tied instead to something with a longer time - scale , most probably the overall evolution of the disc .	we present new results based on _ rxte _ observations of the millisecond pulsar sax j1808.43658 carried out during the decay of the april 1998 outburst . the x - ray spectrum can be fitted by a two - component model . we interpret the soft component as blackbody emission from a heated spot on the neutron star , and the hard component as coming from comptonization in plasma heated by the accretion shock as the material collimated by the magnetic field impacts onto the neutron star surface . the hotspot is probably the source of seed photons for comptonization . the hard component illuminates the disc , giving rise to a reflected spectrum . the amount of reflection indicates that the disc is truncated at fairly large radii ( 2040 @xmath0 ) , consistent with the lack of relativistic smearing of the spectral features . the inferred evolution of the inner radius is not consistent with the magnetic field truncating the disc . instead it seems more likely that the inner disc radius is set by some much longer time - scale process , most probably connected to the overall evolution of the accretion disc . this disc truncation mechanism would then have to be generic in all low mass accretion rate flows both in disc accreting neutron stars and black hole systems . the phase resolved spectra show clearly that the blackbody and hard comptonized spectra pulse independently . this obviously gives an energy dependent phase lag . full general relativistic effects are not required to explain this . the soft blackbody component is optically thick , so its variability is dominated by its changing projected area , while the doppler shifts ( which are maximized @xmath1 before the maximum in projected area ) are somewhat stronger for the translucent column . we do not detect compton reflection from the neutron star surface , though we predict that it should be present in the x - ray spectrum . this would give an unambiguous observational measure of @xmath2 if there is any iron on the neutron star surface which is not completely ionized . = -0.5 cm [ firstpage ] accretion , accretion discs pulsars : individual ( sax j1808.43658 ) x - rays : binaries
the study of few - boson systems close to the unitary limit is an intense subject of research nowadays . for identical bosons , the unitary limit is defined when the two - body scattering length @xmath8 . in this limit the three - boson system shows the efimov effect @xcite . moreover , close to the unitary limit the system manifests universal behavior : the details of the two - body interaction are not important and its spectrum is determined essentially by @xmath1 and the three - body parameter @xmath0 which defines the energy @xmath9 of level @xmath10 at the unitary limit , here @xmath11 is the boson mass ( for a recent review see ref . @xcite ) . at the unitary limit the spectrum shows a discrete scaling invariance ( dsi ) : an infinite series of bound states appears distributed geometrically and accumulates at zero energy . the ratio of two consecutive energy states is constant , @xmath12 , with the universal number @xmath13 . the universal characteristics of the system can be exploited studying the dynamics using potential models constructed in such a way that the control parameters of the spectrum are reproduced . for example a two - parameter potential as a gaussian can be used to this aim @xcite . universal behavior manifests in a simple dependence of many observables on the angle defined by the ratio @xmath4 . this is the case of the product @xmath14 and , in the case in which the two bosons forms a dimer , the ratio @xmath5 between the boson - dimer scattering length @xmath6 and the two - body scattering length @xmath1 . these relations are exactly fulfilled in the zero - range limit and , as we will show , range corrections can be introduced using the potential models . in fact the gaussian potentials can be used to determine the dependence on the angle @xmath3 of the observables defined above . the first one , @xmath14 , is used to define the gaussian level function @xcite . moreover a gaussian potential verifying @xmath4 can be used to compute the ratio @xmath5 at that angle . after introducing range corrections we compute this ratio and , as an application , we determine @xmath6 for the case of a system composed by three @xmath7he atoms . in the case of a zero - range interaction the @xmath15 spectrum of three bosons is determined by the efimov radial law @xmath16 [ eq : energyzr ] where @xmath1 is the two - body scattering length and @xmath17 is the energy of level @xmath18 . the binding momentum @xmath0 gives the energy of the system at the unitary limit . the spectrum is determined by the knowledge of the universal function @xmath19 which is equal for all levels ( a parametrization of the universal function is given in @xcite ) . in fact knowing one energy value the complete spectrum is determined . the zero - range theory is not always sufficiently accurate to describe real systems and range corrections have to be introduced at some level . since close to the unitary limit the details of the interaction are not important , it is possible to construct a potential model in order to capture the essential ingredients of the dynamics . following this strategy a minimal information that preserves universal behavior is encoded in the effective range expansion for two particles , @xmath20 . with @xmath21 the relative momentum and @xmath22 the effective range . at low energies this perturbative expansion is well fulfilled and , in the case of shallow states , it can be extended to negative energies relating @xmath1 , @xmath22 and the two - body energy @xmath23 : @xmath24 where we have introduced the energy length from the relation @xmath25 . therefore a two - parameter potential describing @xmath23 and @xmath1 will also describe @xmath22 . accordingly we define a local and a non local gaussian potential @xmath26 with the strengths @xmath27 , @xmath28 and the ranges @xmath29 , @xmath30 determined to describe particular values of @xmath1 and @xmath23 of a two - boson system . if these values are experimental values we call this set of values a physical point . once this point is fixed the strength of the potential can be varied to reach the unitary limit . with the potentials defined above the lengths , momenta and energies scale with @xmath29 , @xmath31 and @xmath32 ( or @xmath33 ) , respectively . therefore the local gaussian ( lg ) and the nonlocal gaussian ( nlg ) potentials define a particular path to the unitary limit that encompasses all the local and nonlocal gassian potentials . in particular , for the ground state , the values of the effective range and strength at unitary are @xmath34 and @xmath35 ( local gaussian ) and @xmath36 and @xmath37 ( nonlocal gaussian ) . the potentials defined in eqs.([eq : pot1],[eq : pot2 ] ) can be used to describe the three - boson system close to the unitary limit . the efimov law of eq.([eq : energyzr ] ) suggests the following representation of the gaussian @xmath15 spectrum of three - bosons @xmath38 [ eq : energyfr ] where @xmath39 defines the energy of level @xmath18 at the unitary limit , @xmath40 , and the gaussian level function is defined as @xmath41 the scaling properties of the gaussian potentials are such that the level function is the same for all local gaussian and for all nonlocal gaussian , being the local and nonlocal level functions slightly different for the ground state level @xmath42 . as @xmath43 both level functions tend to be equal and tend to the zero - range function , @xmath44 . this behavior is show in fig.1 in which the lg and nlg level functions @xmath45 are shown for the ground and first excited state levels @xmath46 . the universal zero - range function is shown as well and completely overlap with the nlg level function calculated for the third excited state @xmath47 . in particular the ground state level function @xmath48 incorporate range corrections and can be used to determine the corresponding three - body parameter @xmath49 @xcite . are shown for the ground state and first excited state levels @xmath46 . in the nlg case the third excited state ( @xmath50 ) level function @xmath47 ( red circles ) completely overlap with the zero - range universal function @xmath19 ( red solid curve ) . ] we now discuss the @xmath3 dependence of the ratio @xmath5 . as discussed by efimov @xcite ( see also ref . @xcite ) , in the zero - range limit , the ratio @xmath5 has the form @xmath51\ ] ] with @xmath52 , @xmath53 and @xmath54 universal numbers ( numerical values are given in refs . @xcite ) . due to the @xmath3 dependence of the product @xmath2 , the above ratio depends on this angle too . accordingly it is possible to study this ratio using potentials models . following ref . @xcite we introduce range corrections by studying the ratio @xmath55 . we calculate it for different values of @xmath3 and , in particular , we consider valid the following relation at the same @xmath3 value : @xmath56 where the first ratio refers to the values predicted by any realistic interaction or experimental values whereas @xmath57 refers to the values calculated with the lg or nlg potentials . as an application we calculate the value of the atom - dimer scattering length @xmath6 for a system composed by three @xmath7he atoms . this system is well described by the realistic lm2m2 potential of aziz @xcite for which @xmath58mk . using this potential the three - boson ground state energy is @xmath59mk and , accordingly , the angle @xmath3 is obtained from @xmath60 this ratio can be reproduced using a gaussian potential and with that potential the value of @xmath6 can be calculated using standard methods . after this straightforward procedure we obtain @xmath61 to be compared to @xmath62a@xmath63 obtained from a direct calculation using the lm2m2 interaction @xcite . as can we see the simple dependence of the ratio @xmath55 on the angle @xmath3 is well fulfilled and , furthermore , the gaussian description reproduces this dependence with good approximation . we have studied the three - boson system using potential models of the gaussian form . this simple representation of the interaction seems to capture the main aspects of the low energy dynamics close to the unitary limit . we have looked at the ratio @xmath64 defining the angle @xmath3 . at the unitary limit it corresponds to @xmath65 . the dsi manifests at fixed values of @xmath3 , this means that the infinite series of states observed at @xmath65 can be observed also along the line in which @xmath3 remains constant and the energy ratios along that line verify the same relation @xmath12 . it should be noticed that the case @xmath65 corresponds to a single value of @xmath23 ( equal to zero ) , whereas @xmath23 varies along the line in which @xmath3 is constant and therefore the values of @xmath23 at which @xmath66 and @xmath67 are calculate are different . this property allows to study dsi at finite values of @xmath23 . in the present work we have exploited the property that some observables close to the unitary limit have a simple dependence , they are functions of the angle @xmath3 . first we have analysed the product @xmath68 from which we have defined the gaussian level function @xmath45 . in second place we have analysed the ratio @xmath55 and the gaussian values for this ratio have been used to determine @xmath6 for the case of atom - dimer collision of three @xmath7he atoms at zero energy . from the results we observe that @xmath3 dependence is well verified and , the gaussian model reproduces well the lm2m2 value for @xmath6 . further investigations along this line are under way . this work was partly supported by ministerio de economa y competitividad ( spain ) under contracts mtm2015 - 63914-p and fpa2015 - 65035-p . part of the calculations of this work were performed in the high capacity cluster for physics , funded in part by universidad complutense de madrid and in part with feder funds as a contribution to the campus of international excellence of moncloa , cei moncloa . r.a.r thanks ministerio de educacin , cultura y deporte ( spain ) for the `` jos castillejo '' fellowship in the framework of plan estatal de investigacin cientfica y tcnica y de innovacin 2013 - 2016 . 3 v. efimov , phys . lett . b * 33 * , 563 ( 1970 ) v. efimov , sov.j . * 12 * , 589 ( 1971 ) e. braaten and h .- w . hammer , phys . rep . * 428 * , 259 ( 2006 ) a. kievsky and m. gattobigio , phys . a * 92 * , 062715 ( 2015 ) r. lvarez - rodrguez , a. deltuva , m. gattobigio , and m. viviani , phys . a * 93 * , 062701 ( 2016 ) v. efimov , sov.j . * 29 * , 546 ( 1979 ) a. kievsky and m. gattobigio , phys . a * 87 * , 052719 ( 2013 ) r.a . aziz and m.j . slaman , j. chem . phys . * 94 * , 8047 ( 1991 ) e.a . kolganova , a.k . motovilov , w. sandhas , phys . nucl . * 40 * , 206 ( 2009 ) a. kievsky , e. garrido , c. romero - redondo and p. barletta , few - body syst . , * 51 * , 259 ( 2011 )	the universal behavior of a three - boson system close to the unitary limit is encoded in a simple dependence of many observables in terms of few parameters . for example the product of the three - body parameter @xmath0 and the two - body scattering length @xmath1 , @xmath2 depends on the angle @xmath3 defined by @xmath4 . a similar dependence is observed in the ratio @xmath5 with @xmath6 the boson - dimer scattering length . we use a two - parameter potential to determine this simple behavior and , as an application , to compute @xmath6 for the case of three @xmath7he atoms . example.eps gsave newpath 20 20 moveto 20 220 lineto 220 220 lineto 220 20 lineto closepath 2 setlinewidth gsave .4 setgray fill grestore stroke grestore
in spite of real - world qcd being neither supersymmetric nor conformal invariant , it shares many features with @xmath0 sym . both theories are identical when loops of quarks and scalars do not appear , i.e. at tree - level , and in the leading @xmath1 term of the high - energy limit , where the scattering is dominated by exchange of a gluon ladder in the @xmath2-channel . the infrared structure of both theories is similar at the level of soft divergences . it is interesting to look for regimes where qcd and @xmath0 give similar predictions . with this target in mind we will study both theories in the regge limit @xmath3 , with @xmath4 and @xmath2 the usual mandelstam invariants . to define the observables of interest , we consider the kinematic configuration of fig . [ dijet ] , the well - known mueller - navelet jets @xcite . two forward jets with similar transverse scales @xmath5 are tagged at a large rapidity separation @xmath6 and relative azimuthal angle @xmath7 , where @xmath8 are the fractions of longitudinal momenta of the parent hadrons carried by the jet . such a configuration is particularly suited to unveil qcd dynamics in the high - energy limit , specifically through the study of dijet azimuthal correlation @xcite . the large value of @xmath9 calls for a resummation of high - energy logarithms of the form @xmath10 . this resummation is performed in the bfkl approach @xcite , and the differential cross section for dijet production at the parton level . ] is given by @xmath11 in terms of the mellin transform of the solution to the bfkl equation @xmath12 in the leading log approximation , this equation enjoys conformal sl(2,@xmath13 invariance in the plane transverse to the colliding partons @xcite , and the integral kernel @xmath14 is diagonalised by the eigenfunctions @xmath15 . the discrete quantum number @xmath16 controls the azimuthal behaviour in the transverse plane and corresponds to a conformal spin , since it carries a representation of sl(2,@xmath13 . it turns out @xcite that the asympotic intercepts of the kernel corresponding to @xmath17 are very similar at ll and nll , not having the weak convergence problem of the @xmath18 case corresponding to the pomeron intercept which needs an all - orders collinear resummation @xcite . this motivates us to look for observables insensitive to the conformal spin @xmath18 . + we consider the fourier expansion of the differential cross section in the azimuthal angle @xmath19}{1/4+\nu^2}. \end{gathered}\ ] ] the label @xmath20 for each of the moments is the conformal spin . the functions @xmath21 in the definition of the coefficients @xmath22 are the building blocks of the nll bfkl kernel @xcite . a similar formula for @xmath22 holds in the supersymmetric case with @xmath23 and @xmath24 replaced by the corresponding t hooft coupling @xmath25 , see @xcite for details . different observables can be built out of the coefficients @xmath22 . the total ( averaged over @xmath26 ) cross - section is @xmath27 , while contributions of higher conformal spins are projected in the moments @xmath28 . the dependence on @xmath18 cancels when constructing the ratios @xmath29 . these ratios @xmath30 have a good perturbative convergence , and are clean observables to test the properties of the regge limit , in particular conformal invariance , without interference of collinear contributions . they were computed in @xcite in qcd and @xmath0 sym for different renormalisation prescriptions . the results found in @xcite are summarised in fig . [ ratio ] . @xmath0 results are showed in a yellow band since the non - running coupling was allowed to take values in a given range . one can see that the predictions of the two theories for @xmath30 lie very close , in particular when choosing the brodsky - lepage - mackenzie ( blm ) procedure @xcite . the blm procedure effectively resums the effects of a non - zero @xmath31-function and has a conformal behaviour like the one exhibited by @xmath0 theory . in @xcite it was shown to give a more sensible result than @xmath32 scheme for the pomeron intercept . the appearance of an unnaturally high blm scale for the coupling in this case is relaxed when studying observables insensitive to @xmath18 @xcite . it is interesting to note that only when considering clean and perturbatively convergent observables for the high - energy limit , the blm prescription is systematically closer than the other renormalization schemes to the supersymmetric result . a. h. mueller and h. navelet , nucl . b * 282 * ( 1987 ) 727 . m. angioni , g. chachamis , j. d. madrigal and a. sabio vera , phys . lett . * 107 * ( 2011 ) 191601 [ arxiv:1106.6172 [ hep - th ] ] . v. del duca and c. r. schmidt , phys . d * 49 * ( 1994 ) 4510 [ hep - ph/9311290 ] ; w. j. stirling , nucl . b * 423 * ( 1994 ) 56 [ hep - ph/9401266 ] ; a. sabio vera and f. schwennsen , nucl . b * 776 * ( 2007 ) 170 [ hep - ph/0702158 ] ; c. marquet and c. royon , phys . rev . d * 79 * ( 2009 ) 034028 [ arxiv:0704.3409 [ hep - ph ] ] ; d. colferai , f. schwennsen , l. szymanowski and s. wallon , jhep * 1012 * ( 2010 ) 026 [ arxiv:1002.1365 [ hep - ph ] ] ; f. caporale , d. yu . ivanov , b. murdaca , a. papa and a. perri , jhep * 1202 * ( 2012 ) 101 [ arxiv:1112.3752 [ hep - ph ] ] . l. goerlich , these proceedings . l. n. lipatov , sov . j. nucl . * 23 * ( 1976 ) 338 [ yad . * 23 * ( 1976 ) 642 ] ; e. a. kuraev , l. n. lipatov and v. s. fadin , sov . jetp * 45 * ( 1977 ) 199 [ zh . * 72 * ( 1977 ) 377 ] ; i. i. balitsky and l. n. lipatov , sov . j. nucl . phys . * 28 * ( 1978 ) 822 [ yad . fiz . * 28 * ( 1978 ) 1597 ] . l. n. lipatov , sov . jetp * 63 * ( 1986 ) 904 [ zh . * 90 * ( 1986 ) 1536 ] . a. v. kotikov and l. n. lipatov , nucl . b * 582 * ( 2000 ) 19 [ hep - ph/0004008 ] ; a. sabio vera , nucl . b * 746 * ( 2006 ) 1 [ hep - ph/0602250 ] . g. p. salam , jhep * 9807 * ( 1998 ) 019 [ hep - ph/9806482 ] ; a. sabio vera , nucl . b * 722 * ( 2005 ) 65 [ hep - ph/0505128 ] . v. s. fadin and l. n. lipatov , phys . b * 429 * ( 1998 ) 127 [ hep - ph/9802290 ] ; m. ciafaloni and g. camici , phys . b * 430 * ( 1998 ) 349 [ hep - ph/9803389 ] . s. j. brodsky , g. p. lepage and p. b. mackenzie , phys . d * 28 * ( 1983 ) 228 . s. j. brodsky , v. s. fadin , v. t. kim , l. n. lipatov and g. b. pivovarov , hep - ph/0111390 . g. chachamis , j. d. madrigal and a. sabio vera , arxiv:1110.5830 [ hep - ph ] .	we investigate dijet production at large rapidity separation in qcd and @xmath0 sym , showing that both theories give similar predictions for observables only sensitive to conformal properties of the scattering such as ratios of azimuthal angle correlations . renormalization prescriptions are important in this comparison .
a _ graph _ @xmath10 consists of a vertex - set @xmath11 and an edge - set @xmath12 , such that either one or two vertices are assigned to each edge as its ends . a graph is said to be _ simple _ if no edge has only one end - vertex and no two edges have identical end - vertices . a _ plane _ graph is a particular drawing of a graph in the euclidean plane using smooth curves that cross each other only at the vertices of the graph . a graph , which has at least one such drawing , is called _ planar_. we only deal with simple connected plane graphs , whose vertex - sets and edge - sets are finite . by a _ we mean a convex @xmath0-dimensional polytope . the vertices and the edges of a polyhedron form a simple planar @xmath0-connected graph called _ skeleton_. such a graph has a unique drawing ( up to diffeomorphisms of the euclidean plane ) , but there exist many distinct polyhedra having this graph as their skeleton ; all such polyhedra are of the same _ combinatorial type_. the group of isometries of a polyhedron ( so - called _ point group _ ) is a subgroup of the algebraic symmetry group of the graph . by a theorem of mani ( @xcite ) , there is , for each combinatorial type , at least one polyhedron , for which this inclusion becomes equality . so , we can identify the polyhedron and its graph , as well as the algebraic symmetry group and the point group . the _ @xmath13-vector _ @xmath14 of a polyhedron @xmath10 enumerates the numbers @xmath15 of vertices of degree @xmath16 . a polyhedron is _ simple _ if @xmath17 for @xmath18 . the _ @xmath19-vector _ @xmath20 of a polyhedron @xmath10 counts the numbers @xmath21 of faces having @xmath16 sides . for a connected plane graph @xmath10 , we denote its _ plane dual graph _ by @xmath22 and define it on the set of faces of @xmath10 with two faces being adjacent if they share an edge . clearly , @xmath23 and @xmath24 . let @xmath10 be a plane graph , such that every of its vertices has at least degree @xmath0 . then all edges , incident to a vertex @xmath25 , can be labeled in counter - clockwise order as @xmath26 , where @xmath27 is the degree of @xmath25 in @xmath10 . for any edge @xmath28 , @xmath29 , the edges @xmath30 , and @xmath31 ( with @xmath32 and @xmath33 being addition modulo @xmath27 ) are called , respectively , the _ left _ and the _ right_. a circuit of edges of @xmath10 is called a _ zigzag _ ( or a _ petrie path @xcite _ , _ geodesic @xcite _ , _ left - right path @xcite _ ) , if , in tracing the circuit , we alternately select , as the next edge , the left neighbor and the right neighbor ( see figure [ typei - typeii ] ) . in a @xmath0-valent plane graph , any pair of edges sharing a vertex define a zigzag . given an edge , there are two possible directions for extending it to a zigzag . hence , each edge is covered exactly twice by zigzags . a zigzag @xmath34 is called _ simple _ if it has no self - intersection . otherwise , it has at least one edge , say , @xmath35 , in which it self - intersects . let us choose an orientation on @xmath34 ; call the edge @xmath35 of _ type i _ or _ type ii _ @xcite if @xmath34 traverses it twice in the opposite or in the same direction , respectively ( see figure [ typei - typeii ] ) . clearly , the type of edge does not depend on the choice of orientation . edges of type i and type ii correspond to edges of _ cocycle _ and _ cycle character _ in @xcite . the _ signature _ of a zigzag is the pair @xmath36 , where @xmath37 and @xmath38 are the numbers of its edges of type i and type ii , respectively . @xmath39-vector _ of a graph @xmath10 is the vector enumerating the lengths of all its zigzags with their signature as subscript . the simple zigzags are put in the beginning , in increasing order of length , without their signature @xmath40 , and separated by a semicolon from others . self - intersecting zigzags are also ordered by increasing lengths . if there are @xmath41 zigzags of the same length @xmath42 and the same signature @xmath43 , then we write @xmath44 . for a zigzag @xmath34 , its _ intersection vector _ @xmath45 is such that @xmath46 is an increasing sequence of sizes @xmath47 of its intersection with all others zigzags , and @xmath48 denote respective multiplicities . the length of each zigzag is even , because we consecutively take left and right turns . but the length of intersection of two simple zigzags can be odd ( see klein and dyck maps on the table [ tab : t1 ] ; they both have oreinted genus @xmath0 and come as quotients of regular triangulations of the hyperbolic plane having valency @xmath49 and @xmath50 , respectively ) . returning to our plane case , the number of intersection can be even only . a graph is called _ @xmath39-uniform _ if all zigzags have the same length and signature , _ @xmath39-transitive _ if its group acts transitively on zigzags . clearly , @xmath39-transitivity implies @xmath39-uniformity . a graph is called _ @xmath39-knotted _ if it has only one zigzag , _ @xmath39-balanced _ if all its zigzags of the same length and same signature , have identical intersection vectors . clearly , @xmath39-transitivity implies @xmath39-balancedness . let @xmath51 , , @xmath52 be all zigzags of @xmath10 . on every zigzag choose an orientation ; there are @xmath53 possibilities . every edge , being covered twice by zigzags ( one or two ) , can be , with respect to the choosen orientation of all zigzags , oriented in the opposite or same direction . we call such an edge of _ type i _ or _ type ii with respect to the fixed orientation_. if an edge is a self - intersection edge of some zigzag , then the notion of type is independent of the fixed orientation and coincides with the notion of type i and ii , which was introduced above . while for dual graphs @xmath13- and @xmath19-vectors are interchanged , the @xmath39-vector remains the same , except that type i and type ii in the signature are interchanged . other kinds of dualities ( for example , interchanging @xmath39- and @xmath19-vector , but preserving @xmath13-vector ) are considered in @xcite . the _ medial _ graph @xmath54 of @xmath10 is defined on the set of edges of @xmath10 with two edges being adjacent if they share a common vertex and if they belong to a common face . the medial graph is @xmath55-valent and @xmath56 . zigzags of @xmath10 and @xmath22 correspond to central circuits ( see below ) of @xmath54 . any @xmath55-valent plane graph @xmath57 is the medial graph for a pair of mutually dual plane graphs : one can assign two colors to the faces of @xmath57 in the `` chess way '' , such that no two adjacent faces of @xmath57 have the same color . two faces of the same color are said to be adjacent if they share a vertex . in general , a _ central circuit _ of an _ eulerian _ ( i.e. , having only vertices of even degree ) plane graph is a circuit , which is obtained by starting with an edge and continuing at each vertex by the edge opposite to the entering one . link _ is one or more circles ( components of a link ) embedded into the space @xmath58 ; a link with one component is called _ knot_. a _ projection _ of a link is a drawing of it on the plane with gaps representing underpass and solid line representing overpass . if a drawing of a link has alternating underpasses and overpasses , then it is called _ alternating _ ; so , we can see it just as a @xmath55-valent plane graph . we will consider only _ minimal _ projections , i.e. , those without @xmath59-gons ( loops ) . clearly , each edge belongs to exactly one central circuit and any @xmath55-valent graph without @xmath59-gons can be seen as a minimal projection of an alternating link with components corresponding to its central circuits . since zigzags in @xmath10 correspond to central circuits in @xmath54 , each zigzag corresponds to a component of the corresponding alternating link . the bipartition of edges of a @xmath39-knotted graph into type i and type ii corresponds to a bipartition of the vertices of its medial graph ( which is a projection of an alternating knot ) . call a polyhedron _ two - faced _ if it has only @xmath1- and @xmath2-gonal faces with @xmath60 . in particular , denote @xmath0-valent two - faced @xmath61-vertex polyhedron by @xmath62 , @xmath63 , @xmath64 if @xmath7=@xmath8 , @xmath5 , @xmath6 , respectively . given two circuits @xmath65 and @xmath66 , an _ @xmath67-sided prism _ @xmath68 is formed when every @xmath69 is joined to @xmath15 by an edge . now , an _ @xmath67-sided antiprism _ @xmath70 is formed by adding the cycle @xmath71,@xmath72,@xmath73,@xmath74 , , @xmath75,@xmath76,@xmath77,@xmath71 . point group _ is a finite subgroup of the group @xmath78 of isometries of the space @xmath58 , fixing the origin . the connection with plane graphs come from representing them on the sphere centered at the origin . the list of point groups is splitted into two classes : the infinite families and the sporadic cases . every point group @xmath10 contains a normal subgroup formed by the rotations of @xmath10 . the group , denoted @xmath79 , is the cyclic group of rotations by angle @xmath80 with @xmath81 around a fixed axis @xmath82 . both groups @xmath83 and @xmath84 contains @xmath85 as normal subgroups of rotations . the group @xmath84 ( respectively , @xmath83 ) is the group , generated by @xmath85 and a symmetry of plane @xmath86 , with @xmath86 being orthogonal to @xmath82 ( respectively , containing @xmath82 ) . the group @xmath87 is the group , generated by @xmath79 and a rotation by angle @xmath88 , whose axis is orthogonal to @xmath82 . the point group @xmath89 ( respectively , @xmath90 ) is generated by @xmath83 and a rotation by angle @xmath88 , whose axis is orthogonal to @xmath82 and contained in a plane of symmetry ( respectively , going between two planes of symmetry ) . both , @xmath89 and @xmath90 , contain @xmath91 as a normal subgroup . if @xmath92 is even , one defines the point group @xmath93 to be the cyclic group generated by the composition of a rotation by angle @xmath94 with axis @xmath82 and a symmetry of plane @xmath86 with @xmath86 being orthogonal to @xmath82 . the particular cases @xmath95 , @xmath96 and @xmath97 correspond to the trivial group , the plane symmetry group and the central symmetry inversion group . the point groups @xmath98 , @xmath99 and @xmath100 are the symmetry group of the tetrahedron , cube and icosahedron ; the point groups @xmath101 , @xmath102 and @xmath103 are their normal subgroup of rotations . the point group @xmath104 is formed by all @xmath105 and @xmath106 with @xmath107 . more detailed description of point groups are available , for example , from @xcite and @xcite . the list of all symmetry groups of fullerenes @xmath64 is known ( see @xcite ) ; they are ( with their first appearance ) : @xmath111 , @xmath112 , @xmath113 , @xmath114 , @xmath115 , @xmath116 , @xmath117 , @xmath118 , @xmath119 , @xmath120 , @xmath121 , @xmath122 , @xmath123 , @xmath124 , @xmath125 , @xmath126 , @xmath127 , @xmath128 , @xmath129 , @xmath130 , @xmath131 , @xmath132 , @xmath133 , @xmath134 , @xmath135 , @xmath136 , @xmath137 , @xmath138 . there exist fullerenes admitting railroads with _ triple _ self - intersection ; see , for example , figure [ fig : constructedbyhand].a ) . more generally , we believe that any curve , which represents a railroad in a graph of type @xmath63 , also appears as a curve representing a railroad in some @xmath139 . [ theorembigdutourone ] for any even number @xmath140 , there exists a fullerene @xmath64 , with @xmath141 , having two simple zigzags intersecting in exactly @xmath142 edges . it has symmetry @xmath98 if @xmath143 and @xmath144 , @xmath145 if @xmath146 is even , odd respectively . * proof . * for any even @xmath140 there exist a unique @xmath142-vertex @xmath55-valent plane graph @xmath57 , whose faces are four @xmath147-gons ( in two pairs of adjacent ones ) and @xmath148 @xmath55-gons only , and having two simple central circuits ( see @xcite ) . this graph has symmetry @xmath149 for @xmath150 and , for larger values , symmetry @xmath144 , @xmath145 if @xmath146 is even , odd respectively . we will identify each of the two central circuits of @xmath57 with simple zigzags , @xmath51 and @xmath151 , and each vertex with an edge of intersection between them . every face of @xmath57 can be seen as a patch in the graph @xmath64 which we will construct , and so , the local euler formula ( [ local - euler - formula ] ) can be applied . fix a face @xmath152 of @xmath57 ; one can assign to every angle of @xmath152 an angle ( obtuse or acute ) , so that every @xmath147-gon has one acute and one obtuse angle , while every @xmath55-gon has two obtuse and two acute angles . see below the graph for the first values of @xmath142 . the obtained graph has @xmath153 hexagonal faces , and so , @xmath141 vertices . the symmetry group is the same as the one of @xmath57 , except for the first values @xmath154 . see figure [ fig : constructedbyhand].b ) for the corresponding graph with @xmath155.@xmath156 h.s.m . coxeter , _ virus macromolecules and geodesic domes _ , in _ a spectrum of mathematics _ ; ed . by j.c.butcher , oxford university press / auckland university press : oxford , u.k./auckland new - zealand , ( 1971 ) 98107 . m. deza and v.p . grishukhin , _ face - regular bifaced polyhedra _ , report liens 97 - 10 , cole normale suprieure ( 1997 ) . j. of statistical planning and inference , * 95/1 - 2 * , special issue in honor of s.s . shrikhande ( 2001 ) 175195 . m. deza , m.dutour and m. shtogrin , _ @xmath55-valent plane graphs with @xmath147- , @xmath0- and @xmath55-gonal faces _ , satellite conference ( of congress of ism in beijing ) on algebra and combinatorics , 73 - 97 . h. shank , _ the theory of left - right paths _ , in combinatorial mathematics iii , proceedings of 3rd australian conference , st . lucia 1974 , lecture notes in mathematics 452 , springer - verlag , new york , ( 1975 ) 4254 . thurston , _ shapes of polyhedra and triangulations of the sphere _ , in geometry and topology monographs 1 , the epstein birthday schrift , ( j. rivin , c. rourke and c. series , eds . ) geom . coventry , 1998 , 511549 .	a _ zigzag _ in a plane graph is a circuit of edges , such that any two , but no three , consecutive edges belong to the same face . a _ railroad _ in a plane graph is a circuit of hexagonal faces , such that any hexagon is adjacent to its neighbors on opposite edges . a graph without a railroad is called _ tight_. we consider the zigzag and railroad structures of general @xmath0-valent plane graph and , especially , of simple two - faced polyhedra , i.e. , @xmath0-valent @xmath0-polytopes with only @xmath1-gonal and @xmath2-gonal faces , where @xmath3 ; the main cases are @xmath4 , @xmath5 and @xmath6 ( the _ fullerenes _ ) . we completely describe the zigzag structure for the case @xmath7=@xmath8 . for the case @xmath7=@xmath5 we describe symmetry groups , classify all tight graphs with simple zigzags and give the upper bound @xmath9 for the number of zigzags in general tight graphs . for the remaining case @xmath7=@xmath6 we give a construction realizing a prescribed zigzag structure . [ section ] [ theorem]definition [ theorem]conjecture [ theorem]corollary [ theorem]lemma [ theorem]claim [ theorem]remark [ theorem]proposition . primary 52b05 , 52b10 ; secondary 05c30 , + 05c10 . _ key words_. two - faced polyhedra , plane graphs , zigzags , point groups .
to extract the src pair ratios ( @xmath109 , @xmath87 , and @xmath110 ) from the measured cross - section ratios ( @xmath111 , @xmath112 , @xmath113 ) we assumed factorization and used the equations a.1-a.3 listed below :	we studied simultaneously the @xmath0 , @xmath1 , and @xmath2 reactions at + @xmath3 and @xmath4 , for an ( e , e^p ) @xmath5 missing - momentum range of 400 to 830 @xmath6 . the knocked - out proton was detected in coincidence with a proton or neutron recoiling almost back to back to the missing momentum , leaving the residual @xmath7 system at low excitation energy . these data were used to identify two - nucleon short - range correlated pairs and to deduce their isospin structure as a function of missing momentum , in a region where the nucleon - nucleon ( @xmath8 ) force is expected to change from predominantly tensor to repulsive . the abundance of neutron - proton pairs is reduced as the nucleon momentum increases beyond @xmath9 @xmath10 . the extracted fraction of proton - proton pairs is small and almost independent of the missing momentum . our data are compared with calculations of two - nucleon momentum distributions in @xmath11 and discussed in the context of probing the elusive repulsive @xmath8 force . [ dead ] the stability of atomic nuclei is due to a delicate interplay between the long - range attraction that binds nucleons and the short - range repulsion that prevents the collapse of the system . in between , the dominant scalar part of the nucleon - nucleon force almost vanishes and the interaction is dominated by the tensor force , which depends on the spin orientations and the relative orbital angular momentum of the nucleons . recent high - momentum - transfer triple - coincidence @xmath12(e , e^pn ) @xmath13 and @xmath12(p,2pn ) @xmath14 measurements @xcite have shown that nucleons in the nuclear ground state form pairs with large relative momentum and small center - of - mass ( cm ) momentum , where large and small are relative to the fermi momentum of the nucleus . we refer to these pairs as short - range correlated ( src ) pairs @xcite . in the missing - momentum ( the knocked - out proton initial momentum in the absence of final state interactions ) range of @xmath15 , these pairs were found to dominate the high - momentum tails of the nuclear wave functions , with neutron - proton ( @xmath16 ) pairs nearly @xmath17 times more prevalent than proton - proton ( @xmath18 ) pairs , and by inference neutron - neutron ( @xmath19 ) pairs . this is due to the strong dominance of the @xmath8 tensor interaction at the probed sub - fermi distances @xcite . the association of the small @xmath12(e , e^pp ) @xmath20 / @xmath12(e , e^pn ) @xmath21 ratio , at ( e , e^p ) @xmath5 missing momenta of @xmath15 , with dominance of the @xmath8 tensor force , leads naturally to the quest for increasing missing momenta . this allows the search for pairs at distances in which the nuclear force changes from being predominantly tensor to the essentially unexplored repulsive interaction . we report here on a simultaneous measurement of the @xmath11(e , e^p ) @xmath5 , @xmath11(e , e^pp ) @xmath20 and @xmath11(e , e^pn ) @xmath21 reactions at ( e , e^p ) @xmath5 missing momenta from 400 to 830 @xmath6 . the observed changes in the isospin composition of the src pairs as a function of the missing momentum are presented , discussed , and compared to calculations . the experiment was performed in hall a of jefferson laboratory ( jlab ) using a 4 @xmath22a electron beam with an energy of 4.454 gev incident on a 20-cm long high pressure ( 13 atm , 20 k , 0.033 @xmath23 ) @xmath11 gas target contained in a 8 cm diameter 20 cm - long aluminum cylinder . the two hall a high resolution spectrometers ( hrs ) @xcite were used to identify @xmath11(e , e^p ) @xmath5 events . scattered electrons were detected in the left hrs ( l - hrs ) at a central scattering angle of @xmath24 and momentum of @xmath25 . this setup corresponds to the quasi - free knockout of a single proton with transferred three - momentum @xmath26 , transferred energy @xmath27 gev , the negative four - momentum transfer squared @xmath28 , and @xmath29 , where @xmath30 is the proton mass . knocked - out protons were detected using the right hrs ( r - hrs ) , which was set at three different central angles and momenta : ( 33.5@xmath31 , @xmath32 ) , ( @xmath33 , @xmath34 ) , and ( @xmath35 , @xmath36 ) . these kinematical settings correspond to ( e , e^p ) @xmath5 central missing momenta ( @xmath37 ) values of @xmath38 , @xmath39 , and @xmath40 , respectively , covering a missing - momentum range of @xmath41 with overlap between the three different settings . the @xmath11(e , e^p ) @xmath5 events were selected by placing a @xmath42 cut around the @xmath43 ns coincidence timing peak . the fraction of random events increased from @xmath44 at the lowest missing momentum measurement to @xmath45 at the highest . the other cuts on the ( e , e^p ) @xmath5 data were the nominal hrs phase - space cuts on momentum ( @xmath46 ) and angles ( @xmath47 mrad vertical , @xmath48 mrad horizontal ) . to reduce the random - coincidence background , a cut on the target - reconstructed vertex ensured that both the electron and the proton emerged from the same place within @xmath49 cm . the @xmath50 excitation was excluded by a cut on the quasi elastic ( e , e^p ) @xmath5 peak , as in ref . @xcite . for highly correlated pairs , the missing momentum of the @xmath51 reaction is expected to be balanced almost entirely by a single recoiling nucleon . a large acceptance spectrometer ( bigbite ) followed by a neutron detector ( hand ) with a matching solid angle was used to detect correlated recoiling protons or neutrons . the experiment triggered on @xmath52 coincidences between the hrs spectrometers , with the bigbite and hand detectors read out for every trigger . the recoiling protons were detected by the bigbite spectrometer @xcite centered at an angle of @xmath53 , for the @xmath54 and @xmath39 measurements , and @xmath55 for the @xmath40 measurement . the angle between @xmath56 and the recoil nucleon was @xmath57 . the angular acceptance was about @xmath58 msr and the detected momenta accepted ranged from @xmath59 to @xmath60 . the momentum resolution of bigbite , determined from elastic electron - proton scattering , was @xmath61 . the overall proton detection efficiency was @xmath62 . hand consists of several elements : a 2.4-cm thick lead shield ( to block low - energy photons and most of the charged particles coming from the target ) , followed by 64 2-cm thick scintillators ( to identify and veto charged particles ) , and 112 plastic scintillator bars arranged in six 10-cm thick layers covering an area of 1 @xmath63 3 m@xmath64 ( to detect the neutrons ) . hand was placed six meters from the target , just behind bigbite , covering a similar solid angle as bigbite . the pattern of hits in sequential layers of hand was used to identify neutrons @xcite . a time resolution of @xmath65 ns allowed determination of the neutron momentum with an accuracy that varied from @xmath66 ( at @xmath67 ) to @xmath68 ( at @xmath69 ) . the detection efficiency was @xmath70 for @xmath41 neutrons . this determination is based on the efficiency measured up to @xmath71 using the @xmath72 reaction , and extrapolated using a simulation that reproduces well the measured efficiency at lower momenta @xcite . the picture of 2n - src pair breakup with the other two nucleons in @xmath11 being essentially spectators is supported by fig . [ fig2 ] . the figure shows the distribution of the cosine of the angle between the missing momentum and the recoiling neutrons ( @xmath73 ) . we also show the angular correlation for the random background ( dashed - dotted ) as defined by a time window off the coincidence peak . while the placement of the neutron detector opposite to the nominal missing momentum defined by the central rays of the high resolution spectrometers leads to a geometrical angular correlation in the random background , the real triple coincidence events show a clear back - to - back peak above this background . the curve is a result of a simulation of the scattering of a moving pair as discussed below . similar back - to - back correlations were observed for the recoiling protons . the inserts to fig . [ fig2 ] shows the missing - mass for the @xmath11(e , e^pp ) @xmath20 and @xmath11(e , e^pn ) @xmath21 reactions corresponding to a two - nucleon residual system with a low excitation energy . between the @xmath74 and @xmath75 for the @xmath11(e , e^pn ) @xmath21reaction ( @xmath76 and 750 @xmath6 kinematics combined ) . the solid curve is a simulation of scattering off a moving pair with a cm momentum having a width of 100 @xmath6 . the inserts show the missing - mass distributions . in both the main figure and the inserts the data is shown with no random background subtraction . the random background is shown as dashed dotted ( red online ) curves.,scaledwidth=45.0%,height=268 ] software cuts were applied to both bigbite and hand that limited their acceptances to @xmath77 in the vertical direction , @xmath78 in the horizontal direction , and @xmath79 in momentum . a simulation based on the measurements was used to correct the yield of the @xmath11(e , e^pn ) @xmath13 events for the finite acceptances of the recoiling protons and neutrons in bigbite and hand . following ref . @xcite , the simulations assumed that an electron scatters off a moving src pair with a center - of - mass ( cm ) momentum relative to the @xmath80 spectator system described by a gaussian distribution as in ref . @xcite . we assumed an isotropic @xmath81-dimensional motion of the pair and varied the width of the gaussian equally in each direction until the best agreement with the data was obtained . the nine measured distributions ( three components in each of the three kinematic settings for @xmath16 pairs ) yield , within the uncertainties , the same width with a weighted average of @xmath82 . this is in good agreement with the cm momentum distribution calculated in ref . @xcite . fig . [ fig2 ] compares the simulated and measured distributions of the opening angle between the knocked - out and recoiling nucleons . the fraction of events detected within the finite acceptance was used to correct the measured yield . the uncertainty in this correction was typically @xmath83 , which dominates the systematic uncertainties of the @xmath11(e , e^pn ) @xmath13 yield . the measured @xmath84 ratios are given by the number of events in the background - subtracted triple - coincidence tof peak corrected for the finite acceptance and detection efficiency of the recoiling nucleons , divided by the number of random - subtracted ( double - coincidence ) @xmath11(e , e^p ) @xmath5 events . these ratios , as a function of @xmath85 in the @xmath11(e , e^p ) @xmath5 reaction , are displayed as full symbols in the two upper panels of fig . [ fig : wide ] . because the electron can scatter from either proton of a @xmath18 pair ( but only from the single proton of an @xmath16 pair ) , we divided the @xmath11(e , e^pp ) @xmath20 yield by two . also displayed in fig . [ fig : wide ] , as empty symbols with dashed bars , are similar ratios for @xmath12 obtained from previous electron scattering @xcite and proton scattering @xcite measurements . in comparing the @xmath12 and @xmath11 data , it is noted that the measured ratios are about equal and very different from the ratios of naive pair counting in these nuclei . the horizontal bars show the overlapping momentum acceptance ranges of the various kinematic settings . the vertical bars are the uncertainties , which are predominantly statistical . because we obtained the @xmath11(e , e^pp ) @xmath20 and @xmath11(e , e^pn ) @xmath21 data simultaneously and with the same solid angles and momentum acceptances , we could also directly determine the ratio of @xmath11(e , e^pp ) @xmath20 to @xmath11(e , e^pn ) @xmath21 . in this ratio , many of the systematic factors needed to compare the triple - coincidence yields cancel out , and we need to correct only for the detector efficiencies . this ratio as a function of the missing momentum is displayed in the lower panel of fig . [ fig : wide ] together with the previously measured ratio for @xmath12 @xcite . ( e , e^pp ) @xmath20/ @xmath11(e , e^pn ) @xmath21 shown as solid symbols , as a function of the @xmath86 missing momentum . each point is the result of a different setting of the detectors . the bands represent the data corrected for fsi to obtain the pair ratios , see text for details . also shown are calculations using the momentum distribution of ref . @xcite for pairs with weighted - average cm momentum assuming arbitrary angles between the cm and the relative momenta in the pair ( solid black line ) . the middle panel shows the measured @xmath11(e , e^pp ) @xmath20/ @xmath11(e , e^p ) @xmath5 and extracted @xmath87 ratios . the upper panel shows the measured @xmath11(e , e^pn ) @xmath21 / @xmath11(e , e^p ) @xmath5 and extracted @xmath88 ratios . the unphysical region above @xmath89 obtained due to systematic uncertainties and statistical fluctuations is marked by white strips . ratios for @xmath12 are shown as empty symbols with dashed bars . the empty star in the upper panel is the bnl result @xcite for @xmath12(p,2pn ) @xmath14/ @xmath12(p,2p ) @xmath90 . , scaledwidth=45.0%,height=326 ] to extract from the measured cross section ratios the underlying pair ratios , corrections for final - state interactions ( fsi ) were calculated using the glauber approximation @xcite . the glauber corrections ( @xmath91 and @xmath92 ) , with @xmath93 and @xmath94 the leading and recoil transparencies , were calculated by the ghent group @xcite . we assumed the uncertainties to be @xmath95 of these values . the single charge exchange ( scx ) probability ( @xmath96 ) was assumed to be @xmath97 based on the scx total cross section of @xmath98 mb @xcite . the pair fraction extracted from the measured ratios with the fsi calculated corrections are shown in fig . [ fig : wide ] as bands ( see appendix for details ) . the statistical and systematic uncertainties were treated as independent and combined by simulation to create the width of the one standard deviation bands shown in fig . [ fig : wide ] . the systematic uncertainties in the correction factor ( @xmath99 due to finite detector acceptance , @xmath100 due to fsi ) and the statistical fluctuation can explain the extention of the band beyond @xmath89 . the correction to the ratios due to attenuation of the leading - proton is small . the attenuation of the recoiling nucleon decreases the measured triple / double coincidence ratios . because the measured @xmath11(e , e^pn ) @xmath21 rate is about an order of magnitude larger than the @xmath11(e , e^pp ) @xmath20 rate , @xmath11(e , e^pn ) @xmath21 reactions followed by a single charge exchange ( and hence detected as @xmath11(e , e^pp ) @xmath20 ) increase the @xmath11(e , e^pp ) @xmath20/ @xmath11(e , e^pn ) @xmath21 and the @xmath11(e , e^pp ) @xmath20/ @xmath11(e , e^p ) @xmath5 measured ratios . the two - nucleon momentum distributions were calculated for the ground states of @xmath11 using variational monte - carlo wave functions derived from a realistic hamiltonian with argonne v18 and urbana x potentials @xcite . the solid ( black ) curve in fig . [ fig : wide ] was obtained using the calculations @xcite weighted average over arbitrary angles between @xmath101 and @xmath102 , the cm momentum of the pair . the calculation with @xmath103 , which agrees quantitatively with the perugia group calculation @xcite , differs little from the average shown in the figure . to compare the calculations to the data in fig . [ fig : wide ] we assumed that the virtual photon hits the leading proton and @xmath104 ( plane wave impulse approximation ) . the measurements reported here were motivated by the attempt to study the isospin decomposition of 2n - src as a proxy to a transition from primarily tensor to the short range repulsive , presumably scalar , nucleon - nucleon force . in the ground state of @xmath11 @xcite , the number of @xmath18-src pairs is much smaller than @xmath16-src pairs for values of the relative nucleon momentum @xmath105 . this is because the correlations induced by the tensor force are strongly suppressed for @xmath18 pairs which are predominantly in @xmath106 state @xcite . as the relative momenta increase , the tensor force becomes less dominant , the role played by the short - range repulsive force increases and with it the ratio of @xmath107 pairs . in our measurement , as the missing momenta is increased beyond 500 @xmath6 , the triple coincidence @xmath11(e , e^pp ) @xmath20/ @xmath11(e , e^pn ) @xmath21 ratio increases , in good agreement with the prediction based on the ratio of @xmath18-src/@xmath16-src pairs in the @xmath11 ground state @xcite . the measured triple / double coincidence ratios shed further light on the dynamics . the measured @xmath11(e , e^pp ) @xmath20/ @xmath11(e , e^p ) @xmath5 ratio reflects a small contribution from @xmath18-src pairs . these pairs are likely dominated by a scalar repulsive short - range force which is relatively constant over the reported momentum range . the @xmath11(e , e^pn ) @xmath21/ @xmath11(e , e^p ) @xmath5 ratio clearly shows that the reduction in the @xmath16/@xmath18 ratio with increasing @xmath85 is due to a drop in @xmath16-src pairs with increasing @xmath108 . while @xmath16-pairs still dominate two nucleon src , even at missing momentum of 800 @xmath6 the total fraction of the ( e , e^p ) @xmath5 cross section associated with scattering from 2n - src pairs drops with increasing missing momentum . this is likely due to an increase of more complex mechanisms , such as stronger fsi and the onset of src involving more than two nucleons @xcite . a definitive understanding of the relative importance of these effects requires exclusive measurements at large missing momentum on heavier nuclei , and a more detailed theoretical study . to summarize , the short range part of the @xmath8 force is empirically known to be repulsive , it is essential to describe @xmath8 scattering and stability of nuclei , but it is difficult to explore and poorly known both theoretically and experimentally . the measurements reported here probe a transition from an attractive to a repulsive @xmath8 force . the data set , interpreted as changes in the isospin decomposition of the src pairs , is consistent with a reduced contribution from a tensor component and a constant contribution from a scalar component of the @xmath8 force over the probed missing momentum range . it confirms the phenomenological description of the @xmath8 force in this range . one should question to what level the naive interpretation of the data in terms of the ground state nuclear properties is appropriate . comprehensive calculations , which take into account the full reaction mechanism in a relativistic treatment , as well as additional data with better statistics will allow a more detailed determination of the role played by the elusive repulsive short - range nucleon - nucleon interaction . we acknowledge the contribution of the hall a collaboration and technical staff . we thank c. colle , w. cosyn and j. ryckebusch for the glauber calculations . we also want to thank r.b . wiringa , r. schiavilla , s. steven , and j. carlson for the calculations presented in ref . @xcite that were provided specifically for this paper . useful discussions with j. alster , c. ciofi degli atti , w. cosyn , a. gal , l. frankfurt , j. ryckebusch , m. strikman , and m. sargsian , are gratefully acknowledged . this work was supported by the israel science foundation , the u.s . national science foundation , the u.s . department of energy grants de - ac02 - 06ch11357 , de - fg02 - 94er40818 , and u.s . doe contract de - ac05 - 060r23177 under which jefferson science associates operates the thomas jefferson national accelerator facility .
this work was supported by the national natural science foundation of china under grand no.10334020 and in part by the national minister of education program for changjiang scholars and innovative research team in university , and hong kong central allocation fund hkust3/06c . + @xmath36correspondence should be addressed to : [email protected] m. rahm , d. schurig , d. a. roberts , s. a. cummer , d. r. smith , and j. b. pendry , `` design of electromagnetic cloaks and concentrators using form - invariant coordinate transformations of maxwell s equations , '' photon . nanostruct . : fundam . applic . * 6 , * 87 - 95 ( 2008 ) .	a general method is proposed to design the cylindrical cloak , concentrator and superscatterer with arbitrary cross section . the method is demonstrated by the design of a perfect electrical conductor ( pec ) reshaper which is able to reshape a pec cylinder arbitrarily by combining the concept of cloak , concentrator and superscatterer together . numerical simulations are performed to demonstrate its properties . the pioneer work on transformation optics and cloaking [ 1 - 4 ] has suggested a new kind of methodology to manipulate the electromagnetic ( em ) waves by means of the metamaterials . based on the transformation optics , various exciting functional devices have been designed both in theory [ 5 - 15 ] and in experiment [ 16 ] . chen _ et al . _ has recently proposed the `` imperfect cloak '' [ 7 ] which can reduce the scattering cross section of an object . start from this imperfect cloak , they also suggested a design of `` dispersive cloak '' . in contrast to the imperfect cloak , a concept of `` superscatterer '' [ 12 ] was also proposed which can obtain a giant scattering cross section beyond the device itself . if the inner core of the concentrator [ 11 ] is replaced with a pec cylinder , the concentrator can employed as a device to change the scattering cross section of the pec cylinder . and its functionality is between the imperfect cloak and the superscatterer . in addition , a cylindrical cloak of arbitrary cross section [ 10 ] has been given , which can be easily extended to the imperfect cloak , concentrator , and superscatterer . in this paper , we will show the method of such an extension . and based on our methodology , we will propose a kind of transformation media , which we shall call the `` pec reshaper '' . an explicit design will be given . the pec reshaper can reshape the pec cylinder at will . for example , the effective pec cylinder can be partially inside the device and partially outside the device . we will demonstrate the properties of the pec reshaper by using the finite - element methods . the general coordinate transformation is given by the following relation , @xmath0 where @xmath1 which are three functions that specify the inner and outer boundaries of the transformation media and the boundary of an imaginary cylinder which will be useful later , respectively . we note that all the boundaries are smooth and non - convex . this transformation maps the field in the domain @xmath2 onto another domain @xmath3 . from the transformation optics , one can obtain the parameters inside the transformation media in @xmath4 . we consider the transverse electric ( te ) mode for simplicity and suppose that the inner cylinder ( @xmath5 is a pec cylinder . in the view of transformation optics , after coated with the transformation media in @xmath6 , the inner pec cylinder ( @xmath5 looks like another effective pec cylinder in the domain @xmath7 . if @xmath8 , the transformation media is an imperfect cylindrical cloak [ 7 , 8 ] with arbitrary cross section , which can reduce the cross section of the inner pec cylinder ( see in fig . 1(a ) ) . if @xmath9 , the imperfect cylindrical cloak becomes perfect [ 1 , 10 ] . if @xmath10 , the transformation media is the outer shell of a cylindrical concentrator [ 11 ] with arbitrary cross section , which can enhance the cross section of the inner pec cylinder ( see in fig . 1(b ) ) . we shall still call the device `` concentrator '' for simplicity . however , the effective pec cylinder is still within the domain @xmath11 . with the help of the concept of `` superscatterer '' [ 12 ] , we can reshape the inner pec cylinder very versatilely . in the case of superscatterer ( i.e. @xmath12 , the transformation maps the field in the domain @xmath13 onto another domain @xmath14 through the mirror of @xmath15 . the superscatterer can enhance the cross section of the inner pec cylinder to that of an effective pec cylinder which can be much larger than the device ( see in fig . 1(c ) ) . now we consider an extraordinary case shown in fig . 1(d ) , where the effective pec cylinder is partially inside the inner cylinder , partially within the domain of transformation media and partially outside the device for different angles . such a device combines cloak , concentrator and superscatterer all together into one case , which we shall call the `` pec reshaper '' as it could reshape the pec boundary more or less arbitrarily . before demonstrating the properties of the pec reshaper , we have to work out the required parameters for it from the transformation optics . the jacobian is @xmath16,\ ] ] where @xmath17 , @xmath18 , @xmath19 , and @xmath20 with @xmath21 , and @xmath22 . the corresponding permittivity and permeability tensors of the transformation media on the circular cylindrical coordinate are [ 13 ] , @xmath23,\ ] ] where @xmath24 @xmath25 we note that the permittivity and permeability tensors here should be written as the functions of positions in the form of @xmath26 . consider the te mode , we have , @xmath27 and @xmath28 = \left [ { { \begin{array}{{20}c } { \cos \theta ' } \hfill & { - \sin \theta ' } \hfill \\ { \sin \theta ' } \hfill & { \cos \theta ' } \hfill \\ \end{array } } } \right]\left [ { { \begin{array}{{20}c } { \mu _ { rr } } \hfill & { \mu _ { r\theta } } \hfill \\ { \mu _ { r\theta } } \hfill & { \mu _ { \theta \theta } } \hfill \\ \end{array } } } \right]\left [ { { \begin{array}{{20}c } { \cos \theta ' } \hfill & { \sin \theta ' } \hfill \\ { - \sin \theta ' } \hfill & { \cos \theta ' } \hfill \\ \end{array } } } \right ] \\ = \left [ { { \begin{array}{{20}c } { \mu _ { rr } \cos ^2\theta ' - 2\mu _ { r\theta } \sin \theta ' \cos \theta ' + \mu _ { \theta \theta } \sin ^2\theta ' } \hfill & { ( \mu _ { rr } - \mu _ { \theta \theta } ) \sin \theta ' \cos \theta ' + \mu _ { r\theta } ( \cos ^2\theta ' - \sin ^2\theta ' ) } \hfill \\ { ( \mu _ { rr } - \mu _ { \theta \theta } ) \sin \theta ' \cos \theta ' + \mu _ { r\theta } ( \cos ^2\theta ' - \sin ^2\theta ' ) } \hfill & { \mu _ { rr } \sin ^2\theta ' + 2\mu _ { r\theta } \sin \theta ' \cos \theta ' + \mu _ { \theta \theta } \cos ^2\theta ' } \hfill \\ \end{array } } } \right ] , \\ \end{array}\ ] ] on cartesian coordinate [ 14 ] . we note that the above parameters are also valid for cloak , concentrator and superscatterer with arbitrary cross section . + with the explicit form of the required parameters , the properties of the pec reshaper can be demonstrated by numerical simulations with the comsol multiphysics finite element - based electromagnetics solver . let @xmath29 , and @xmath30 where @xmath31 . we choose @xmath32 as a circle for simplicity . a plane wave is normal incident from left to right with unit amplitude and a frequency of 0.1 ghz . fig . 2(a ) and ( b ) show the snapshots of the total electric and scattering field caused by a pec cylinder with its outer boundary depicted by @xmath33 , respectively . fig . 2(c ) and ( d ) show the total electric and scattering field induced by the pec reshaper which is a concentric cylindrical shell . comparing the similar far - field pattern from fig . 2(a ) and ( c ) , or fig . 2(b ) and ( d ) , we can conclude that the pec reshaper reshapes the pec cylinder with @xmath34 to an effective pec cylinder with @xmath33 . the large overvalued fields in fig . 2(c ) and ( d ) are caused by the surface mode resonances excited at @xmath33 , which are replaced with white flecks [ 12 ] . in conclusion , we have proposed the concept of pec reshaper which combines the concept of cloak , concentrator and superscatterer all together into one case . the pec reshaper can reshape the pec boundary arbitrarily if the objective effective pec cylinder shares a domain with the original pec cylinder ( i.e. there exists @xmath35 . the properties of the pec reshaper are demonstrated by numerical simulations from finite - element methods . the design can be extended to three dimensions straitforwardly .
a random walk on @xmath0 is a stochastic process @xmath1 defined as @xmath2 where @xmath3 are independent and identically distributed random variables with a common distribution @xmath4 $ ] for @xmath5 . the relationship between the behavior of @xmath6 as @xmath7 and properties of @xmath8 is quite well - understood and is easy to analyse , using often fourier analysis as the main tool . for instance if @xmath9 , the law of large numbers states that @xmath10 if @xmath11 and @xmath12 , then the central limit theorem asserts that the distribution of @xmath13 is asymptotically gaussian on @xmath14 with mean @xmath15 and covariance @xmath16 given by @xmath17 cramr s theorem on large deviations provides exponential rates of convergence of @xmath18\ ] ] for sets @xmath19 excluding the mean @xmath20 . assuming that @xmath21 for @xmath22 , for a wide class of sets @xmath19 , @xmath23\ ] ] exists and is given by @xmath24 where @xmath25.\ ] ] the situation is quite different if we drop the hypothesis of independence of increments of @xmath6 and replace it with the assumption that @xmath6 is a markov process . in other words , there exists @xmath26 such that @xmath27=p(s_n , z).\ ] ] the random walk case considered earlier is a special case where @xmath28 takes the form @xmath29 . sometimes it is more convenient to denote @xmath30 by @xmath31 so that the independent case is when @xmath32 is independent of @xmath33 . in this generality not much can be said . however if @xmath34 is periodic in @xmath33 , the phase gets averaged out and some homogenization takes place with @xmath35 getting effectively replaced by an averaged @xmath36 . results on law of large numbers , central limit theorem and large deviation estimates can be established , with some extra work , quite similar to the independent case . we will investigate the case when @xmath34 are random but spatially homogeneous in @xmath33 . this can viewed as a limiting version of the periodic case when the period becomes large . we will start with a probability space @xmath37 on which @xmath0 acts ergodically as a family @xmath38 of measure preserving transformations . we are given @xmath39 which is a probability distribution on @xmath0 for each @xmath40 , that are measurable functions of @xmath40 . one can then generate random transition probabilities @xmath41 by defining @xmath42 for each @xmath40 , @xmath41 can serve as the transition probability of a markov process on @xmath0 and the measure corresponding to this process , starting from @xmath15 , is denoted by @xmath43 . this is of course random ( depends on @xmath40 ) and is called the random walk in the random environment @xmath40 . one can ask the same questions about this random walk and in some form they may be true for almost all omega with respect to @xmath44 . the law of large numbers , if valid , will take the form @xmath45 = 1.\ ] ] such statement concerning the almost sure behavior of @xmath43 for almost all @xmath40 with respect to @xmath44 are said to deal with the quenched version . sometimes one wishes to study the behavior of the averaged measure @xmath46 the law of large numbers is the same because ( [ lln ] ) is equivalent to @xmath47 = 1.\ ] ] on the other hand , questions on the asymptotic behavior of probabilities , like for instance , the central limit theorem or large deviations could be different for the quenched and the averaged cases . a special environment called the product environment is one in which @xmath41 are independent for different @xmath33 and have a common distribution @xmath48 which is a probability measure on the space @xmath49 of all probability measures on @xmath0 . in this case the canonical choice for @xmath50 is the countable product of @xmath49 . the product measure having marginals @xmath48 is of course the choice for @xmath44 . one can consider the continuous versions of these problems . for instance instead of the action of @xmath0 we can have @xmath14 acting on @xmath51 ergodically and consider a diffusion on @xmath14 with a random infinitesimal generator @xmath52 acting on smooth function on @xmath14 . here @xmath53 is generated from a map @xmath54 by the action @xmath55 of @xmath14 , @xmath56 again there is the quenched measure @xmath43 that corresponds to the diffusion with generator @xmath57 that starts from @xmath15 at time @xmath15 and the averaged measure that is given by the same formula ( [ average ] ) . this model is referred to as diffusion with a random drift . exactly the same questions can be asked in this context . we shall describe the results in the one - dimensional case , which is quite well - understood . a very good reference with a full bibliography is a survey by zeitouni @xcite at the beijing congress as well as his notes from st . flour @xcite . for simplicity we will take the special case of the product environment in which only @xmath58 are non - zero and are random variables @xmath59 , @xmath60 that are independent for different values of @xmath61 , @xmath48 being the distribution on @xmath62 $ ] of @xmath63 . to avoid some technical details we shall assume that @xmath48 has no mass near the edges i.e. is supported on a closed subinterval of @xmath64 . the question of when @xmath6 is transient was answered in @xcite . in order that @xmath65=1\ ] ] a.e . @xmath44 it is necessary and sufficient that @xmath66 = \int\log\frac{p}{1-p } \beta({\rm d}p ) > 0.\ ] ] in the original random walk where @xmath67 is a constant this is equivalent to @xmath68 or @xmath69 . in such a case the law of large numbers asserts that @xmath70 in our context the law of large numbers , also proved in @xcite , states the following theorem . [ lln2 ] @xmath71 a.e . @xmath43 for almost all @xmath40 with respect to @xmath44 where @xmath72 is given by @xmath73}{1+e[\frac{q}{p } ] } , & if\ e[\frac{q}{p}]<1,\\[.3pc ] 0 , & otherwise . \end{cases}\ ] ] it is therefore possible that @xmath74>0 $ ] while @xmath75{\geqslant}1 $ ] , in which case @xmath76 but @xmath77 . in the original random walk this could not happen . in the one - dimensional case further detailed analysis can be carried out in this case . the problem comes from traps where the environment conspires to hold the particle for a long time . as the particle moves to @xmath78 , it encounters increasingly deeper traps that hold it even longer producing a critical slowing down . this phenomenon discovered by sinai @xcite is well - understood . in one - dimension the traps can not be avoided . therefore the one - dimensional case is strikingly different from the higher dimensional ones . assuming that @xmath79 , one can study the behavior of @xmath80 and ask for a central limit theorem . it turns out that the fluctuations of @xmath81 have two sources ; from the environment @xmath82 and fluctuations of the walk in this environment . under @xmath43 , the distribution of @xmath81 is gaussian with a random mean which depends on the environment and has an asymptotic gaussian distribution by itself . hence while there is a central limit theorem for the averaged @xmath83 there is none for @xmath43 . this phenomenon is not expected to persist in higher dimensions . there are large deviation results regarding the limits @xmath84 = i(a)\ ] ] and @xmath85 = \bar{i } ( a).\ ] ] the difference between @xmath86 and @xmath87 has a natural explanation in terms of the large deviation behavior of the environment . it is related to the following question . if at time @xmath88 , a particle is near @xmath89 instead of being around @xmath90 which is what we should expect , what does it say about the environment in @xmath91 $ ] ? did the particle behave strangely in a normal environment or did it encounter a strange environment ? it is in fact a combination of both as is seen in the relation of @xmath87 to @xmath86 which is essentially bayes rule . see for instance @xcite and @xcite . if @xmath92 the law of large numbers and central limit theorems are not that well - understood . for instance the questions of recurrence , transience and the existence of a non - zero limit @xmath93 have only partial answers with sufficient conditions that are not that easy to check . even when we limit ourselves to a product environment there is no clear analog of the condition ( [ rec.cond ] ) . however if @xmath94 , under some symmetry conditions that ensure that @xmath95 , a central limit theorem was proved by bricmont and kupiainen @xcite , provided the randomness in the environment is small . a recent preprint by sznitman and zeitouni @xcite provides an alternate proof . large deviation results however exist in much wider generality , both in the quenched and the averaged cases . the large deviation principle is essentially the existence of the limits @xmath96=\psi(\theta).\ ] ] the expectation is with respect to @xmath43 or @xmath83 , which could produce different limits for @xmath97 . the law of large numbers and central limit theorem involve the differentiability of @xmath97 at @xmath98 , which is harder . in fact large deviation results have been proved by general sub - additivity arguments for the quenched case . roughly speaking fixing @xmath40 , @xmath99{\geqslant}q^\omega[s_k\simeq ka]\times q^{\tau_{ka}\omega}[s_\ell\simeq \ell a].\ ] ] one can then use some version of the sub - additive ergodic theorem . this idea has to be cleaned up a bit , but is not hard . see for instance @xcite or @xcite . we will however explore , in the next two sections a proof of the large deviation principle for the averaged case and an alternate proof for the quenched case . we want to prove that the limits @xmath100 = -\bar{i } ( a)\ ] ] or equivalently @xmath101 = \bar{\psi } ( \theta)\ ] ] exist . the problem is that the measure @xmath83 is not very nice . as the random walk explores @xmath0 it learns about the environment and in the case of the product environment , when it returns to a site that it has visited before , the experience has not been forgotten and leads to long term correlations . however , if we are interested in the behavior @xmath102 with @xmath103 , the same site is not visited too often and the correlations should decay fast . this can be exploited to provide a proof of ( [ ldpa ] ) . one can use bayes rule to calculate the conditional distribution @xmath104=q(z\|w),\ ] ] where @xmath105 is the past history of the walk . before we do that it is more convenient to shift the origin as we go along so that the current position of the random walk is always the origin and the current time is always @xmath15 . then an @xmath88 step walk looks like @xmath106 . we pick a @xmath61 with probability @xmath107 . we get a new walk of @xmath108 steps @xmath109 given by @xmath110 for @xmath111 . we can now calculate @xmath107 . we need to know all the numbers @xmath112 of the number of times the walk has visited @xmath113 in the past and jumped from @xmath113 to @xmath114 . it is not hard to see that the _ a posteriori _ probability can be calculated as @xmath115 while this makes sense initially only for walks of finite length it can clearly be extended to all transient paths . note that although we only use @xmath116 , in order to obtain the new @xmath117 we would need to know @xmath118 . let us suppose that @xmath119 is a process with stationary increments @xmath120 . if the increments process were ergodic and had a non - zero mean @xmath121 we can again make the current position the origin and the process will be transient . @xmath122 would exist a.e . @xmath119 and can be compared to the corresponding conditional probabilities @xmath123 under @xmath119 . the relative entropy @xmath124\ ] ] is then well - defined . the function @xmath125 defined for @xmath103 extends as a convex function to all of @xmath126 , and with this @xmath87 , ( [ ldpa ] ) is satisfied . the proof , which uses basic large deviation techniques , can be found in @xcite . it is interesting to note that the rate functions @xmath127 and @xmath128 coincide at @xmath15 and can be calculated explicitly . let @xmath129 be the convex hull of the support of @xmath48 in @xmath130 . then @xmath131 in particular @xmath132 if and only if @xmath15 is not in the range of @xmath133 as @xmath134 varies over @xmath135 . this is referred to in the literature as the non - nestling case . although a proof using the subadditive ergodic theorem exists , we will provide an alternate approach that is more appealing . we will illustrate this in the context of brownian motion with a random drift ( [ randomdiff ] ) . we can define a diffusion on @xmath50 with generator @xmath136 where @xmath137 are the generators of the translation group @xmath138 . this is essentially the image of lifting the paths @xmath139 of the diffusion on @xmath14 corresponding to @xmath140 to @xmath50 by @xmath141 while there is no possibility of having an invariant measure on @xmath14 , on @xmath50 one can hope to find an invariant density @xmath142 provided we can find @xmath143 in @xmath144 , that solves @xmath145 if such a @xmath146 exists , then we have an ergodic theorem for the diffusion process @xmath43 corresponding @xmath147 on @xmath50 , @xmath148 this translates to an ergodic theorem on @xmath126 as well @xmath149 where now @xmath43 is the quenched process in the random environment . since @xmath150 it is clear that @xmath151 providing a law of large numbers for @xmath139 . while we can not be sure of finding @xmath146 for a given @xmath152 it is easy to find a @xmath152 for a given @xmath146 . for instance , we could take @xmath153 . or more generally @xmath154 with @xmath155 . if we change @xmath152 to @xmath156 with @xmath157 , the new process will have relative entropy @xmath158.\ ] ] moreover , for almost all @xmath40 with respect to @xmath44 , almost surely with respect to @xmath159 , @xmath160{\varphi}{\rm d}p=\int c { \rm d}p.\ ] ] if we fix @xmath161 , the bound @xmath162 { \geqslant}-\frac{1}{2 } \int \left\\| b - \frac{\nabla{\varphi}}{2{\varphi } } - \frac{c}{{\varphi } } \right\\|^2 { \varphi}{\rm d}p\ ] ] is easily obtained . if we define @xmath163 then @xmath164 { \geqslant}- i(a).\ ] ] of course these statements are valid a.e . @xmath43 a.e . @xmath44 . one can check that @xmath86 is convex and the upper bound amounts to proving the dual estimate @xmath165{\leqslant}\psi(\theta),\ ] ] where @xmath166.\ ] ] we need a bound on the solution of @xmath167 with @xmath168 $ ] . by hopf - cole transformation @xmath169 this reduces to estimating @xmath170 with @xmath171 . this can be done if we can construct a subsolution @xmath172 on @xmath50 , where @xmath173 satisfies @xmath174 and @xmath105 is closed in the sense that @xmath175 . the existence of the subsolution comes from convex analysis . @xmath176\\[.3pc ] & = \sup_{{\varphi}}\sup_c \inf_{u}\left[\int { \langle}c,\theta+\nabla u { \rangle}{\rm d}p -\frac{1}{2}\int \left\\|b-\frac{\nabla{\varphi}}{2{\varphi}}-\frac{c}{{\varphi } } \right\\|^2 { \varphi}{\rm d}p \right]\\[.3pc ] & = \sup_{{\varphi } } \inf_{u}\sup_c\left[\int { \langle}c,\theta+\nabla u{\rangle}{\rm d}p -\frac{1}{2}\int \left\\|b-\frac{\nabla{\varphi}}{2{\varphi}}-\frac{c}{{\varphi}}\right\\|^2 { \varphi}{\rm d}p\right]\\[.3pc ] & = \sup_{{\varphi } } \inf_{u}\int \sup_c\left[{\langle}c,\theta+\nabla u { \rangle}-\frac{1}{2 } \left\\|b-\frac{\nabla{\varphi}}{2{\varphi}}-\frac{c}{{\varphi}}\right\\|^2 { \varphi}\right ] { \rm d}p\ ] ] @xmath177{\varphi}{\rm d}p\\[.3pc ] & = \sup_{{\varphi } } \inf_{u}\int \left [ \left[{\langle}b,\theta + \nabla u{\rangle}+\frac{1}{2}\left\\|\theta+\nabla u\right\\|^2\right]{\varphi}-\frac{1}{2}{\langle}\nabla u,\nabla { \varphi}{\rangle}\right]{\rm d}p\\[.3pc ] & = \sup_{{\varphi } } \inf_{u}\int \left[\frac{1}{2}\delta u + { \langle}b,\theta + \nabla u{\rangle}+\frac{1}{2}\left\\|\theta+\nabla u\right\\|^2\right]{\varphi}{\rm d}p\\[.3pc ] & = \sup_{{\varphi } } \inf_{\substack{w\ { \rm closed}\\ \int w { \rm d}p=\theta}}\int \left[\frac{1}{2}\nabla\cdot w+ { \langle}b , w{\rangle}+\frac{1}{2}\\|w\\|^2\right]{\varphi}{\rm d}p\\[.3pc ] & = \inf_{\substack{w\ { \rm closed}\\ \int w { \rm d}p=\theta}}\sup_{{\varphi}}\int \left[\frac{1}{2}\nabla\cdot w+ { \langle}b , w{\rangle}+\frac{1}{2}\\|w\\|^2\right]{\varphi}{\rm d}p\\[.3pc ] & = \inf_{\substack{w\ { \rm closed}\\ \int w { \rm d}p=\theta}}\sup_{\omega } \left[\frac{1}{2}\nabla\cdot w+ { \langle}b , w{\rangle}+\frac{1}{2}\\|w\\|^2\right]\end{aligned}\ ] ] which proves the existence of the subsolution . this can be viewed as showing the existence of a limit as @xmath178 ( homogenization ) of the solution of @xmath179 with @xmath180 . the limit satisfies @xmath181 with @xmath182 . this can be generalized to equations of the form @xmath183 we have left out the details as well as the hypothesis needed to prove the results . they will appear in @xcite . this research was supported by a grant from the national science foundation dms-0104343 . comets francis , gantert nina and zeitouni ofer , quenched , annealed and functional large deviations for one - dimensional random walk in random environment , _ probab . theory related fields _ * 118(1 ) * ( 2000 ) 65114	random walks as well as diffusions in random media are considered . methods are developed that allow one to establish large deviation results for both the ` quenched ' and the ` averaged ' case . [ theore]theorem [ theore]remark [ theore]proposition [ theore]lemma [ theore]definition [ theore]corollary [ theore]example
this section provides some background on the distinct forms of the dzyaloshinskii moriya interaction ( dmi ) referred to in the main text . for the dmi to be noticeable at the continuum level , it is necessary that the central inversion symmetry of the system is broken . this may be the case because the crystal structure itself is noncentrosymmetric ( chiral or polar ) or because the geometry of the system breaks the inversion symmetry , for instance near a surface or interface . a variety of continuum models for the dmi exist , which all arise from the same microscopic definition . the difference lies in the orientation of the dzyaloshinskii moriya vectors @xmath104 that define the interaction between atoms @xmath105 . we consider two high - symmetry special cases of the dmi . the continuum form @xmath106 describes a dmi that is _ chiral _ ( it changes sign under any reflection ) but which is otherwise completely isotropic . an alternative form @xmath107 pertains to a system that is _ polar _ : the inversion symmetry is broken by a director @xmath108 , which changes sign under central reflection ; however , the system is achiral in that it remains invariant under any reflections that leave @xmath108 intact . the system is isotropic only in the plane normal to @xmath108 . the interface of a magnetic layer with another material defines a polar axis @xmath108 ( the interface normal ) , which gives rise to a dmi of the polar form ( if we neglect any in - plane anisotropies ) . for this reason , the polar form is usually referred to as the interfacial dmi . the chiral but isotropic form of the dmi is usually referred to as the bulk dmi , since it is the simplest expression that models the dmi in a chiral crystal in bulk ( again neglecting any anisotropies ) . a crystal structure is chiral if it can not be superimposed onto its mirror image in any way . we remark , for completeness , that a bulk crystal may also show an ` interfacial ' dmi effect if it has a polar crystal structure @xcite . the distinct effects of the bulk dmi and the interfacial dmi are well known from the theory of magnetic skyrmions , where it is found that the ` bulk ' model predicts bloch skyrmions while the ` interfacial ' model predicts nel skyrmions @xcite . similarly , it is found that a strong interfacial dmi tends to push a domain wall into a nel configuration with a well - defined chirality , while a strong bulk dmi stabilizes one particular chirality of the bloch domain wall . in sec . [ sec : microdmi ] , we link the cases of bulk dmi and interfacial dmi to atomistic toy models . in sec . [ sec : macrodmi ] , we explain that the preference of each type of dmi for one or the other type of domain wall ( bloch _ vs. _ nel ) is a direct consequence of its symmetries . directions of the dzyaloshinskii moriya vectors @xmath104 and the lattice displacement vectors @xmath109 between atoms @xmath105 . ( a ) for a bulk dmi , the @xmath104 roughly point into the direction of @xmath109 . the @xmath104 `` diverge away from '' ( or `` converge to '' ) a given atom @xmath110 . ( b ) for an interfacial dmi , the @xmath104 `` curl around '' a given atom @xmath110 , with a sense determined by the normal @xmath108 . ] this section shows how the continuum energy functionals for bulk and interfacial dmi can be derived from the atomistic definition eq . , given appropriate choices for the dzyaloshinskii moriya vectors @xmath104 . at the level of individual magnetic moments , the dmi is by definition any interaction that can be written as @xmath111 where @xmath112 is the microscopic magnetic moment on site @xmath110 and @xmath104 is the dzyaloshinskii moriya interaction vector between sites @xmath110 and @xmath113 , which without loss of generality satisfies @xmath114 . for simplicity , let us assume that the magnetic atoms of the crystal are arranged as a bravais lattice , where we use the symbol @xmath115 to represent any given ( near - neighbor ) lattice vector . ( notice that , for obvious reasons , the overall crystal structure of a noncentrosymmetric crystal must be more complicated than just a single bravais lattice . ) by translation invariance in the bulk , the magnitude and direction of @xmath104 must be a function of @xmath116 only . for simplicity , we shall assume that only nearest - neighbor interactions are important and that @xmath116 represents a nearest - neighbor lattice vector . if the material is isotropic , the only reasonable choice for the dzyaloshinskii moriya vector is @xmath117 , where @xmath14 is a ( positive or negative ) interaction strength . the dzyaloshinskii moriya vectors @xmath104 seem to `` diverge '' from any given site @xmath110 , as shown schematically in fig . [ fig : dmvec](a ) . near a surface or interface , by contrast , the interface normal @xmath108 introduces a preferential direction , which we could give a definite sense be defining it to point from material @xmath118 into material @xmath119 . by symmetry arguments @xcite , it is found that the @xmath104 vectors `` curl around '' a given site @xmath110 with a well - defined sense induced by the direction of @xmath108 , as shown schematically in fig . [ fig : dmvec](b ) . considering only the highest layer of magnetic atoms below the interface and only their nearest - neighbor interactions , the only natural choice in the absence of in - plane anisotropy is @xmath120 , since @xmath104 must be a vector - valued function linear both in @xmath121 ( because of the rule @xmath114 ) and in @xmath108 ( because the polar axis @xmath108 is the only element that breaks inversion symmetry ) . passing to a continuum theory , eq . becomes @xmath122 \ , \omega^{-1 } d^3 r \text{,}\ ] ] where @xmath123 represents a unit - cell volume and @xmath124 sums over all relevant near - neighbor lattice vectors . in the case of isotropic bulk dmi with only nearest - neighbor interactions , we substitute @xmath125 . equation becomes , in tensor notation , @xmath126 where @xmath127 . assuming @xmath128 for some scalar @xmath129 ( isotropy ) , we get @xmath130 for the interfacial dmi , on the other hand , we have @xmath131 , where @xmath108 is the interface normal . for simplicity , we take @xmath132 . we get @xmath133 notice that we now integrate over the interface plane only and assume that @xmath134 does not depend on the perpendicular coordinate @xmath31 ( here @xmath123 represents a unit area ) . assuming @xmath128 , we get @xmath135\,\omega^{-1 } d^2 r \text{.}\end{aligned}\ ] ] using the divergence theorem , we get @xmath136 where @xmath137 . the boundary term is irrelevant if we integrate over all space ( @xmath138 ) . ( color online ) . symmetries of the two types of domain wall . ( a ) the two distinct states of the bloch wall are related by a reflection that leaves the surface normal @xmath108 invariant . the bulk dmi selects one bloch state as the most favorable and the other bloch state as the least favorable orientation . the interfacial dmi does not discriminate between the two . ( b ) the two distinct states of the nel wall are related by @xmath10 rotation that reverses the surface normal @xmath108 . the interfacial dmi selects one nel state as the most favorable and the other nel state as the least favorable orientation . the bulk dmi does not discriminate between the two . ] while the bulk dmi is chiral but isotropic , the interfacial dmi is polar but achiral ( and isotropic under rotations around the @xmath1 axis ) . the two forms considered here are the only possibilities that satisfy these respective constraints , regardless of the details of the microscopic model . the same symmetry considerations determine how the two forms of the dmi respond to bloch or nel domain walls in a thin film with perpendicular anisotropy . in fig . [ fig : dwsym ] , we compare the two chirality states that exist of either type of wall . the bulk dmi selects one of the two chiralities of the bloch wall as its most favorable orientation ( energy minimum ) , and the other bloch chirality as the least favorable orientation ( energy maximum ) . on the other hand , the interfacial dmi selects one of the two chiralities of the nel wall as its most favorable orientation , and the other nel chirality as the least favorable orientation . we can interpret this behavior as follows . the two chirality states of the bloch wall are related by a reflection in the @xmath139 plane . since such a reflection leaves the @xmath31-director unchanged , it follows that the interfacial dmi can not discriminate between the two chiralities of the bloch wall . the bulk dmi can , because the bulk dmi is not invariant under any spatial reflections . the two chiralities of the nel wall , on the other hand , are related by a @xmath10 rotation around the @xmath29 axis . such a rotation is a symmetry of the bulk dmi ( isotropy ) but not of the interfacial dmi ( reversal of @xmath108 ) . as a consequence , only the interfacial dmi discriminates between the two chiralities of the nel wall . the orientation of a bloch domain wall will not be affected by a small bulk dmi , as both chiralities represent extrema of the bulk dmi . a small interfacial dmi , by contrast , causes a change in orientation of the bloch domain wall , reorienting it slightly towards the favorable nel state . this section specifies the energy functionals used to describe the dipolar ( magnetostatic ) interaction in our micromagnetic simulations and numerical calculations . throughout this work , we assume that the magnetization is homogeneous in @xmath31 inside the ferromagnetic film ( uniform - mode approximation ) . in other words , we assume that the magnetization may be written as @xmath140 , where @xmath30 is the film thickness and where @xmath141 for @xmath142 and @xmath143 otherwise . this simplification can be justified in the regime that @xmath144 , where @xmath145 is the exchange length . it can be derived ( see , for example , ref . @xcite ) that the total dipolar energy ( defined per unit thickness ) is given , in reciprocal space , by @xmath146 with [ eq : greci ] @xmath147 where @xmath148 and where the indices @xmath149 represent the in - plane coordinates @xmath150 . the function @xmath151 describes the @xmath152 component of the normalized magnetization field . its fourier transform @xmath153 is defined as @xmath154 the demagnetizting factor @xmath155 is given by @xmath156 we define @xmath157 by continuity . notice that @xmath158 . we have omitted from eq . some contributions of the form @xmath159 , which are constant . for a system with periodic boundary conditions that is defined on a rectangular grid , we evaluate the dipolar interaction in @xmath160 time using a fast fourier transform ( fft ) . finite magnetic elements , such as a waveguides with a finite width , might also , somewhat approximately , be simulated in this way , provided that sufficient zero padding is inserted around the object to minimize the influence of the other periodic copies . for a one - dimensional magnetization profile @xmath0 , assumed to extend infinitely and uniformly in the @xmath29 direction ( @xmath161 ) , we may write the dipolar energy in real space as @xmath162 where @xmath163 these expressions can be useful when evaluating the dipolar interaction for a system that is not periodic in @xmath28 , such as a domain wall , where we have @xmath164 while @xmath165 . we remark that an efficient numerical implementation of eq . might still evaluate the convolution through an fft , but special care must be taken to correctly take into account the asymptotic behavior of @xmath0 for @xmath166 and @xmath167 . in this section , we provide a derivation of our expression for the relative part @xmath169 of the phase shift in the @xmath168 limit , presented in the main text as eq . ( 4 ) . in fig . [ fig : plotssup ] , we evaluate the geometric phase shift @xmath71 and the expression ( 4 ) for @xmath169 for numerically calculated equilibrium domain - wall profiles @xmath0 , confirming the accuracy of eqs . ( 3 ) and ( 5 ) of the main text . an equilibrium magnetization profile @xmath0 is found by minimization of total energy @xmath23 under the constraint that @xmath170 for all @xmath28 . it satisfies @xmath171 where the expression on the left - hand size denotes a functional derivative . the scalar - valued function @xmath172 is a lagrange multiplier . the equilibrium profile @xmath0 suffices to calculate the geometric phase induced by parallel transport of the basis vectors . however , to determine the relative phase we must solve the normal - mode equation , which can be expressed in coordinate - free form [ 23 ] as @xmath173 = \lambda(x ) \mathbf{m}(x)\text{,}\end{gathered}\ ] ] where @xmath174 is the gyromagnetic ratio , @xmath172 is fixed by eq . , and where @xmath175 is a lagrange multiplier , under the constraint that @xmath176 at all @xmath28 . the frequency @xmath177 is fixed by the wavenumber @xmath99 of the spin wave far away from the domain wall . equation , when written out explicitly in terms of @xmath178 , becomes a complicated integro - differential equation . ( nonlocality arises from the dipolar interaction . ) here we find an approximate solution using the wentzel kramers brillouin ( wkb ) approximation , which is exact in the short - wavelength limit . we assume that the solution @xmath179 locally resembles a plane wave and replace any differential operator @xmath180 with @xmath181 ( analogously for convolution operators ) in order to solve for @xmath182 independently for each @xmath28 . for the interactions specified in the main text , including interfacial dmi , we get @xmath183 where , anticipating the limit @xmath184 , we have written only terms that are of at least first order in @xmath182 or @xmath99 . the exchange interaction acts as a scalar @xmath185 because we define the basis @xmath90 according to parallel transport . we substitute the dispersion relation far away from the domain wall , @xmath186 , again up to corrections of constant order . the characteristic equation of the matrix in eq . has four solutions . we are interested in the solution @xmath187 closest to @xmath99 , which is given by @xmath188 the phase induced by the domain wall on top of the phase factor @xmath189 is now given in the wkb approximation by @xmath190 { \ , \mathrm{d}}{}x$ ] . we find @xmath191 up to a correction of order @xmath86 . phase - shift contributions @xmath20 and @xmath95 ( for @xmath64 ) obtained from the minimum - energy domain - wall profiles , which are numerically calculated as a function of the parameters @xmath118 , @xmath192 , @xmath33 , @xmath30 , and @xmath14 taking the dipolar interaction into account . the variables @xmath73 and @xmath92 are also evaluated . ( a ) the geometric part @xmath20 follows eq . ( 3 ) of the main text ( solid line ) , with a deviation of at most a few degrees even in extreme cases ( dashed line , @xmath193 ) . ( b ) the relative part @xmath95 depends on the dmi strength @xmath14 ( as compared to exchange ) and on the domain - wall angle @xmath73 ; it follows approximately eq . ( 5 ) of the main text . for given @xmath194 and @xmath73 , the dependence on the third dimensionless parameter is negligible . the dashed line indicates the relation between @xmath14 and @xmath73 if we fix @xmath195 and @xmath196 . ] figure [ fig : plotssup ] shows a numerical evaluation of the two contributions @xmath20 and @xmath95 to the phase - shift difference @xmath102 between domain walls of opposite chirality ( _ e.g. _ , @xmath12 and @xmath11 in the main text ) . to describe the behavior of @xmath95 , we introduce the characteristic domain - wall width @xmath197 , which we evaluate from the equilibrium profile @xmath0 . the variable @xmath92 is then the value of @xmath198 for an equivalent domain wall with @xmath41 and the other parameters ( @xmath118 , @xmath192 , @xmath33 , and @xmath30 ) the same . for completeness , we derive that the equivalent of eq . for a bulk dmi is @xmath199 \int_{-\infty}^\infty m_x(x ) { \ , \mathrm{d}}{}x$ ] . notice that the latter expression vanishes for any bloch domain wall . in this section , we derive analytically , under certain approximations , the effect of an interfacial dmi on the equilibrium profile and phase shift of a bloch domain wall . in particular , we present a closed - form expression for the phase shift that is valid for arbitrary wavenumber @xmath99 . the expression suggests that the difference in phase shift between the two chiralities is almost independent of the wavelength of the spin wave . our analytical treatment complements the wkb approach of sec . [ sec : phwkb ] in two ways . first , it allows one to obtain a semianalytical expression for the equilibrium domain - wall profile , which in the wkb approach is taken as given ( ie , calculated numerically ) . second , it provides an expression for the phase shift for arbitrary @xmath99 , where the wkb approach considers only the @xmath64 limit . the two approximations made are that we take into account the main effect of the dipolar interaction as an effective local interaction , and that we treat the effect of the dmi perturbatively ( small @xmath14 ) . the results presented here are consistent with the wkb expression ( sec . [ sec : phwkb ] ) if we substitute into the equilibrium profile @xmath0 calculated for the approximated dipolar interaction [ but notice that we need to add the geometric part @xmath71 to eq . to obtain the total phase shift ] . in our analytical treatment , we follow ref . [ ] in including the dipolar effects ( shape anisotropy ) into a local anisotropy energy . we argue that , on scales much smaller than the film thickness @xmath30 , only the @xmath200-function part of eq . is important and the effect of the dipolar interaction reduces to a local anisotropy @xmath201 with @xmath202 . notice that , even though the dipolar interaction is isotropic , the @xmath28 coordinate plays a special role because we assume that @xmath203 is the domain - wall normal ( magnetization is a function of @xmath28 only ) . the total anisotropy energy is now given by @xmath204 where @xmath205 are positive constants . the in - plane anisotropy @xmath206 models the dipolar interaction in introducing a preference for bloch domain walls ( flux closure ) . this localized approximation is formally valid in the limit that the magnetization profile @xmath134 depends only on @xmath28 and extends infinitely not only in the @xmath29 but also in the @xmath31 direction . in practice , this means that we assume normal incidence of spin waves and a wavelength and exchange length much shorter than film thickness @xmath30 . while these assumptions are unrealistic in most practical cases , they allow us to obtain some analytical results that are qualitatively correct . in addition to the easy - axis and in - plane anisotropies , we take into account the interfacial dmi @xmath210 and the usual exchange term @xmath211 , where a prime denotes a derivative with respect to @xmath28 . for a system with only exchange @xmath118 and uniaxial anisotropy @xmath192 ( @xmath215 ) , it is well known that the equilibrium profile , assuming @xmath164 and @xmath165 , is given by @xmath216 \text{,}\ ] ] where @xmath38 is the exchange length , while the function @xmath217 takes an arbitrary constant value . if we set @xmath218 , we get the well - known bloch magnetization profile the sign ( @xmath220 ) defines the chirality of the domain wall . both chiralities represent equivalent stable energy minima . the positive sign corresponds to the configuration @xmath9 as defined in the main text ; the negative sign to @xmath8 . as a result of the competition with @xmath206 , the interfacial dmi modifies the domain - wall profile , as calculated numerically in ref . , we treat the dmi as a small perturbation ( @xmath221 ) . it can be derived that we get , to first order in @xmath14 , a minimum - energy configuration where @xmath222 . perpendicular magnetization implies @xmath223 . the function @xmath207 , shown in fig . [ fig : funcg ] , is uniquely defined as the solution of @xmath224 that is even and vanishes at infinity ( particular part ) . it is useful to compare eq . to the equilibrium profile that is obtained if a bulk dmi instead of the interfacial dmi is present . in that case , we obtain the same equilibrium profile as in eq . ; in other words , the bulk dmi has no effect on the profile of the bloch domain wall . the reason is that , given _ any _ bloch profile , where @xmath225 and @xmath226 is arbitrary , the functional derivatives @xmath227 and @xmath228 of the bulk dmi energy @xmath229 with respect to the profile functions @xmath212 vanish . the dynamics is described by the landau lifshitz gilbert ( llg ) equation without damping @xmath230 where @xmath33 is the saturation magnetization , @xmath174 is the gyromagnetic ratio , and @xmath231 is the total interaction energy . let us consider small variations @xmath232 and @xmath233 around eq . . we get a linearized equation of motion @xmath234\partial_t \mp ( 2d / l ) \hat{f } \\ { [ } m_\text{s}/(2\|\gamma\|)]\partial_t \mp ( 2d / l ) \hat{f}^\dagger & - a\left ( \partial_x^2 + \partial_y^2 \right ) + k(1 - 2\operatorname{sech}^2 x / l ) + k_\perp \end{array } \right ) \left ( \begin{array}{c c } \delta \theta \\ \operatorname{sech}(x / l)\delta \phi \end{array } \right ) = 0\ ] ] where @xmath235 with @xmath236 is the perturbation caused by the dm interaction , to first order in @xmath14 . notice that @xmath237 is odd in @xmath28 and breaks reflection symmetry . ( the system remains invariant under a simultaneous reflection and change in polarity , which is equivalent to a rotation around @xmath1 . ) we assume that the solutions are periodic in @xmath238 and @xmath29 . equation becomes a hamiltonian normal - mode problem @xcite . away from the domain wall ( @xmath239 ) , the spin waves take the form @xmath240 with @xmath241 , where @xmath242 . we now calculate the reflection ( @xmath243 ) and transmission ( @xmath238 ) amplitudes of an incoming spin wave with wavenumber @xmath99 that propagates in the positive-@xmath28 direction . the amplitudes @xmath244 refer to an incoming spin wave wave that propagates in the negative-@xmath28 direction ( wavenumber @xmath245 ) and approaches the domain wall from the other side . equation defines the scattering states . for @xmath246 , the analytic solutions of eq . are well known @xcite . the propagating - wave solutions take the form @xmath247 where @xcite @xmath248 it follows immediately that @xmath249 where @xmath250 notice that the domain wall shows total transmission in the localized model . [ however , if the full dipolar interaction is taken into account , we find a nonzero reflection for long - wavelength spin waves ( see also , for example , @xcite ) . ] we now calculate the effect of the interfacial dmi on the transmission phase and amplitude . as above , we take into account the dmi to first order in the interaction strength @xmath14 . with some algebraic work , we obtain @xmath251 where @xmath252 is again given by eq . , and where @xmath253 notice that only @xmath254 depends on chirality ( @xmath220 ) . since defines the scattering problem relative to the basis vectors @xmath255 , which have a fixed orientation at @xmath256 and at @xmath257 , the expression includes both the geometric and relative parts of the phase shift . dependence of the phase shifts @xmath252 [ given by eq . ] and @xmath254 [ given by eq . ] on wavenumber @xmath258 . for a spin wave propagating in the potitive-@xmath28 direction , the total phase shift induced by the domain wall is the sum @xmath259 . notice that @xmath254 is almost constant in @xmath258 . since only the part @xmath254 depends on domain - wall chirality ( @xmath220 ) , the difference in phase shift between the two chiralities in other words , the phase difference obtained on the right - hand side of the interferometer shown in fig . 2(d ) of the main text is also almost independent of @xmath258 . ] unlike @xmath252 , the chirality - dependent part @xmath254 depends only very weakly on wavenumber @xmath258 ; it is , in fact , almost constant in @xmath258 , as shown in fig . [ fig : kxdep ] . we find numerically that this conclusion even holds if @xmath14 is not small or if the full dipolar interaction is taken into account in the scattering problem , at least in the regime where the wavelength is comparable to the domain - wall width or shorter . this justifies our approach of taking the @xmath64 limit in the calculation of @xmath169 in sec . [ sec : phwkb ] ( the geometric part @xmath71 is independent of @xmath99 by definition ) .	spin - wave technology ( magnonics ) has the potential to further reduce the size and energy consumption of information processing devices . in the submicrometer regime ( exchange spin waves ) , topological defects such as domain walls may constitute active elements to manipulate spin waves and perform logic operations . we predict that spin waves that pass through a domain wall in an ultrathin perpendicular - anisotropy film experience a phase shift that depends on the orientation of the domain wall ( chirality ) . the effect , which is absent in bulk materials , originates from the interfacial dzyaloshinskii moriya interaction and can be interpreted as a geometric phase . we demonstrate analytically and by means of micromagnetic simulations that the phase shift is strong enough to switch between constructive and destructive interference . the two chirality states of the domain wall may serve as a memory bit or spin - wave switch in magnonic devices . motivated by the aim to reduce energy dissipation in electronic devices , spin waves are considered as an alternative information carrier in the field of magnon spintronics @xcite . a spin wave acquires a phase shift when it passes through a magnetic domain wall ( dw ) @xcite . in this letter , we show that the dzyaloshinskii moriya interaction ( dmi ) in ultrathin ferromagnetic films makes the phase shift dependent on the dw chirality , leading to constructive / destructive interference in a two - branch interferometer . the mechanism we identify raises the prospect of magnonic devices in which dw chirality acts as a spin - wave switch . ( color online ) . effect of the interfacial dmi on the magnetization profile @xmath0 of a dw in a thin film with perpendicular anisotropy . ( a ) away from the dw , magnetization points out of the film ( @xmath1 or @xmath2 ) . near the dw , the dmi creates an effective field @xmath3 in the @xmath4 direction . depending on the competition between the dipolar and dmi interactions , the equilibrium configurations @xmath5 , @xmath6 , and @xmath7 , shown in ( b)(d ) , are possible . ( b ) without dmi , the minimum - energy configurations ( flux closure ) are two equivalent bloch dws ( @xmath8 , in dark colors , and @xmath9 , in light colors ) , whose in - plane orientations differ by @xmath10 . ( c ) for intermediate dmi , the minimum - energy configurations are intermediate between bloch and nel . there are two equivalent minimum - energy states ( @xmath11 and @xmath12 ) , whose in - plane orientations differ by approx . @xmath13 for an appropriately tuned dmi strength @xmath14 . ( d ) for strong dmi , a single minimum - energy configuration @xmath7 exists : a nel dw with magnetization in the center pointing in the @xmath4 direction . ] it is by now clear that the dmi plays a crucial role in the magnetization dynamics of ultrathin films @xcite , due to the broken inversion symmetry at the interfaces . the interfacial dmi favors , in a perpendicular - anisotropy film , nel dws with a fixed chirality [ fig . [ fig : groundstate](d ) ] @xcite , in competition with the dipolar interaction , which tends to favor bloch dws [ fig . [ fig : groundstate](b ) ] . the most interesting regime is when the two interactions have a comparable strength , yielding a dw intermediate between bloch and nel @xcite with two stable minimum - energy configurations ( chirality states ) whose in - plane orientations differ by @xmath15 , as shown in fig . [ fig : groundstate](c ) . recent experiments demonstrated that dws can be brought into the intermediate regime , and that the dmi strength can be fine - tuned by modifying the thicknesses of the adjacent nonmagnetic layers @xcite . the internal orientation might be also tuned by an adjacent layer of a topological insulator ; its surface states induce in the magnetic layer an interfacial - dmi - like effect that depends on chemical potential and applied electric field @xcite . ( color online ) . interferometer setup in a thin film with perpendicular anisotropy . the two dws may have ( a ) identical or ( b ) opposite chiralities . spin waves enter the device from the left . if the chiralities are identical , constructive interference is always obtained on the right - hand side . for opposite chiralities , the phase difference depends on the dmi strength , as shown in ( c)(e ) . ( c ) without dmi , spin waves interfere constructively even if the chiralities are opposite ( @xmath16 ) . ( d ) for intermediate dmi , we find a phase difference @xmath17 of up to @xmath10 ( destructive interference ) for opposite chirality states @xmath6 . ( e ) for strong dmi , the configurations in both branches are the same ( @xmath7 ) , trivially resulting in constructive interference . in ( c)(e ) , large arrows represent the equilibrium magnetization direction @xmath0 . on the left , spin - wave basis vectors @xmath18 are defined identically for all configurations @xmath19 . their orientation after parallel transportation , shown on the right , depends on the dw configuration . in ( d ) , notice that the transported basis vectors for @xmath11 and @xmath12 are rotated by @xmath10 . this geometric phase difference @xmath20 is the dominant contribution to @xmath17 . ] our main result is summarized in fig . [ fig : partrans ] , where we consider an interferometer in which incoming spin waves are divided between two identical waveguides , each containing a dw . the two dws are identical in every respect except possibly their chirality . when the spin waves rejoin , they are transmitted or reflected depending on the phase difference . while it is obvious that spin waves interfere constructively if the two dws have the same chirality [ fig . [ fig : partrans](a ) ] , we ask if it is possible to achieve destructive interference ( spin wave blocked ) by reversing the chirality of one dw [ fig . [ fig : partrans](b ) ] . without dmi , the two chirality states of the bloch dw induce identical phase shifts , leading to constructive interference [ fig . [ fig : partrans](c ) ] . for strong dmi , the dw has a single stable ( nel ) configuration and the phase shifts are obviously also identical [ fig . [ fig : partrans](e ) ] . however , for intermediate dmi , the two equilibrium orientations induce geometric phase shifts differing by as much as @xmath10 [ fig . [ fig : partrans](d ) ] . in this regime , the interferometer can switch between constructive and destructive interference transmission or reflection depending on whether the chiralities are identical or opposite . the two chirality states are separated by an energy barrier @xmath21 ( an unfavorable nel configuration ) . if @xmath21 is high enough , spontaneous reversals of chirality due to thermal fluctuations are very rare ( for the system in fig . [ fig : sims ] , we obtain @xmath22 ) . we could consider the intermediate - dmi interferometer as a two - state memory device where the transmission of spin waves serves as readout mechanism ( ` open ' or ` closed ' ) . switching does not require modifications of the material parameters , nor to insert or remove dws @xcite , but only to reverse the chirality of one dw , for instance by a field pulse normal to the plane of the film . ( the ` field pulse ' might alternatively be generated through optomagnetic effects @xcite , provided the light can be focused onto a single branch . ) the field causes the dw magnetization to precess as shown in fig . [ fig : precess ] until , when it is switched off , the dw relaxes to the nearest chirality state . we have tested the results of fig . [ fig : partrans ] , which we derive analytically below , by means of explicit micromagnetic simulations . the total energy @xmath23 is given by the sum of the usual micromagnetic energy functionals for exchange @xmath24 , uniaxial anisotropy @xmath25 , and dipolar energy @xcite , plus a functional @xmath26 describing the dmi induced near the interfaces of the ultrathin film @xcite . the dmi strength @xmath14 can be positive or negative ( we take @xmath27 in figs . [ fig : groundstate ] and [ fig : partrans ] ) . we treat the film as effectively two - dimensional ( magnetization is a function of @xmath28 and @xmath29 only ) , but we do consider the finite film thickness @xmath30 in the @xmath31 direction for the dipolar interactions @xcite . here we consider a waveguide made of a long strip of ultrathin film with perpendicular magnetization ( @xmath32 , where @xmath33 is saturation magnetization ) . the waveguide width @xmath34 is at least so large that the dipolar interactions , in the absence of dmi , favor a bloch dw . the interfacial dmi is qualitatively different from a dmi @xmath35 present in isotropic bulk materials with a chiral crystal structure @xcite . the effect of a bulk dmi on the interaction of spin waves with dws was considered in refs . [ ] and [ ] . since a bulk dmi favors the bloch dw ( @xmath8 in figs . [ fig : groundstate ] and [ fig : partrans ] ) , it does not , in the geometry considered here , provide the competition with dipolar interactions that is essential to obtain the intermediate dw with two equivalent minimum - energy orientations @xmath6 differing by approximately @xmath13 . ( color online ) . micromagnetic simulations of the propagation of spin waves in a ferromagnetic thin - film waveguide with perpendicular magnetization ( @xmath36 , @xmath37 , where @xmath38 ) through dws of the indicated chiralities ( @xmath39 _ vs. _ @xmath40 ) . ( a ) without dmi ( @xmath41 ) , spin waves experience the same phase shift regardless of dw chirality , leading to constructive interference ( avg = average ) . ( b ) for intermediate dmi ( @xmath42 ) , there is a phase difference of almost @xmath10 between spin waves that passed through dws of different chirality , leading to destructive interference . we remark that the attenuation on the right - hand side is not the result of gilbert damping ( we take @xmath43 ) , but merely represents the present location of the wavefront [ @xmath44 . ] figure [ fig : sims ] shows how spin waves , generated on the left - hand side of the strip , pass through a dw . we solve the landau lifshitz gilbert ( llg ) equation , for relaxed initial states , on a square grid ( @xmath45 cells ) using a self - developed c++ code @xcite with implicit - midpoint time integration @xcite . spin waves are generated by a space - local , time - periodic in - plane applied field ( @xmath46 ) , switched on at @xmath47 . each waveguide strip ( @xmath48 ) is simulated in a vacuum - padded periodic box ( @xmath49 ) . we calculate the difference @xmath50 in phase shift between the two chiralities at the right - hand side of the interferometer , comparing intermediate dmi to the case without dmi . a phase difference of up to @xmath10 ( destructive interference ) is obtained for intermediate dmi . fixing @xmath51 and @xmath52 @xcite , the other parameters of fig . [ fig : sims ] become @xmath53 ( frequency ) , @xmath54 ( waveguide width @xmath55 wavelength ) , @xmath56 , and @xmath57 , assuming the free - electron gyromagnetic ratio . spin waves of such frequencies and wavelengths can be experimentally generated , observed , and visualized @xcite . since the dmi is an interfacial effect , @xmath14 is inversely proportional to film thickness @xmath30 @xcite . extrapolation of the values in ref . @xcite ( @xmath58 , @xmath59 ) suggests that @xmath60 is realistic for @xmath61 . the phase difference @xmath62 between spin waves traveling along the two paths ( @xmath11 and @xmath12 ) has a geometric @xcite origin . it is convenient to define @xmath63 . a spin wave causes magnetization to precess around its local equilibrium direction @xmath0 . direction . in the @xmath31 direction , we consider a finite @xmath30 in the calculation of the dipolar interactions but neglect the @xmath31 dependence of the magnetization inside the film , making the magnetization a function of @xmath28 only . the two - dimensional simulations of fig . [ fig : sims ] show that the essence of our analytical results carries over to waveguides of finite width @xmath34 . ] in the limit of exchange spin waves ( @xmath64 ) , the dynamics induced by the wave is given by the real part of @xmath65 } [ \hat{\mathbf{a}}(x ) - i \hat{\mathbf{b}}(x ) ] \text{,}\ ] ] where @xmath66 is the infinitesimal amplitude of the spin wave ( linear regime ) . the orthonormal basis vectors @xmath67 must be perpendicular to @xmath0 for all @xmath28 , so that their orientation continually changes across the dw . a natural choice is to define @xmath67 according to parallel transport , @xmath68 , by which the basis vectors , at any given point @xmath28 , match their orientation in an infinitesimal neighborhood of @xmath28 as closely as possible . the function @xmath69 in eq . [ eq : swdef ] determines the phase of the spin wave relative to the basis @xmath70 . however , the orientation of the basis @xmath70 after parallel transportation across the dw strongly depends on the dw configuration ( @xmath8 , @xmath9 , @xmath11 , @xmath12 , or @xmath7 ) , as shown in fig . [ fig : partrans](c)(e ) . this reorientation of @xmath70 implies an additional phase shift @xmath71 , which is purely geometric in nature . it is apparent from fig . [ fig : partrans](c)(e ) that the geometric contribution is approximately given by @xmath72 where @xmath73 is the in - plane angle of the magnetization at the dw center , as shown in fig . [ fig : partrans](d ) . for example , we have a geometric phase difference @xmath74 for intermediate dws with @xmath75 . the value of @xmath73 is determined by the competition between the dmi and the dipolar interaction , as shown in fig . [ fig : plotsmain](a ) . while in principle @xmath76 depends on the exact shape of the equilibrium profile @xmath0 , we find that the deviation from eq . is at most a few degrees @xcite . ( a ) dw angle @xmath73 as a function of the ratio of dmi strength @xmath14 and dipolar interaction , for three film thicknesses @xmath30 , taking @xmath77 @xcite . the angle @xmath73 determines whether intermediate dws ( @xmath78 ) are closer to a bloch or a nel configuration . more `` nel - like '' dws ( larger @xmath73 ) are obtained for stronger @xmath14 . conversely , the dipolar interaction penalizes the nel configuration ( this effect is weaker in thinner films ) . ( b ) phase difference @xmath17 between spin waves passing through dws of opposite chiralities ( @xmath12 _ vs. _ @xmath11 ) , as a function of @xmath73 , in the @xmath64 limit , for two values of @xmath79 , taking @xmath80 . the dominant contribution @xmath76 is separated out . the remaining contribution @xmath81 lowers the value of @xmath73 needed for destructive interference ( @xmath82 ) . for @xmath83 , @xmath81 is larger than for @xmath84 because a relatively strong dmi @xmath85 is then needed to obtain a given @xmath73 . ] we derive @xcite the relative contribution @xmath81 for @xmath64 , up to a correction of order @xmath86 , as @xmath87 where @xmath88 and @xmath89 are the magnetization profiles @xmath12,@xmath11 calculated numerically , taking into account dmi and dipolar interactions . notice that the exchange interaction does not contribute directly to eq . because the basis @xmath90 ( parallel transport ) absorbs such a contribution into @xmath76 . equation gives , approximately , @xmath91 where @xmath92 is a characteristic dw width ( @xmath93 for @xmath94 ) . notice that @xmath95 vanishes for @xmath41 ( @xmath8 , @xmath9 ) and for large @xmath14 ( @xmath7 ) , where @xmath96 ( nel wall ) . as shown in fig . [ fig : plotsmain](b ) , the contribution of @xmath81 enhances the effect of @xmath76 and merely shifts the critical internal angle @xmath73 for perfect destructive interference ( @xmath97 ) to a somewhat lower value ( more bloch - like dw ) . therefore the concept of the interferometer spin - wave switch is robust : we can always find a dw angle @xmath98 such that the phase difference @xmath17 between opposite chiralities is exactly @xmath10 . the desired value @xmath73 could then be realized by fine - tuning the dmi strength @xmath14 [ fig . [ fig : plotsmain](a ) ] . while eq . is derived in the short - wavelength limit , we have numerically solved the spin - wave normal - mode problem @xcite for incoming waves of arbitrary wavenumber @xmath99 . the phase shifts @xmath100 depend significantly on @xmath99 , as in the case without dmi ( @xmath101 @xcite ) , but the difference @xmath102 between the two chiralities , which is the relevant quantity in our interferometer , is weakly wavelength dependent for wavelengths comparable to ( or shorter than ) the dw width . the weak dependence of @xmath17 on @xmath99 can , under certain approximations , also be derived analytically @xcite . this observation justifies our approach @xmath103 . in summary , we have shown that the interfacial dzyaloshinskii moriya interaction in ultrathin magnetic films provides a new way of manipulation of spin waves . with this interaction , spin waves experience a different phase shift when passing through dws of different chiralities , leading to either constructive or destructive interference in a two - branch interferometer . one can open or close the transmission of spin waves through the device by changing the dw chirality in one of the two branches . this opens the possibility of developing a memory element or transistor based on the manipulation of magnonic currents without charge transport . the authors thank k. s. novoselov for useful discussions . a. qaiumzadeh provided some additional literature . this work is part of the research programme of the foundation for fundamental research on matter ( fom ) , which is part of the netherlands organisation for scientific research ( nwo ) . y.f . acknowledges support from the spanish mecd grant no . fis2011 - 23713 . 33ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ] + 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) _ _ , @noop ph.d . thesis , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) _ _ , @noop ph.d . thesis , ( ) @noop `` , '' @noop * * , ( ) @noop * * , ( ) @noop @noop * * , ( ) @noop * * , ( ) @noop `` , '' @noop * * , ( ) @noop `` , '' ( ) , @noop * * , ( ) @noop * * , ( )
this work is supported by the fundamental research funds for the central universities , program for key science and technology innovative research team of shaanxi province ( 2013kct-05 ) , and the national natural science foundation of china ( grant nos . 11374008 , 11374238 , 11374239 and 11534008 ) .	it is known that the cross - correlation function ( ccf ) of a partially coherent vortex ( pcv ) beam shows a robust link with the radial and azimuthal mode indices . however , the previous proposals are difficult to measure the ccf in practical system , especially in the case of astronomical objects . in this letter , we demonstrate experimentally that the hanbury brown and twiss effect can be used to measure the mode indices of the original vortex beam and investigate the relationship between the spatial coherent width and the characterization of ccf of a pcv beam . the technique we exploit is quite efficient and robust , and it may be useful in the field of free space communication and astronomy which are related to the photon s orbital angular momentum . in 1992 , l. allen _ et al . _ @xcite pointed out that beam with spiral phase distribution of @xmath0 carries an orbital angular momentum ( oam ) of @xmath1 , where @xmath2 is an integer which denotes the azimuthal mode index ( topological charge ) , and @xmath3 is the azimuthal coordinate . beam with this kind of phase structure is also called vortex beam . vortex beam has been widely studied in the last two decades and found a lot of applications , such as optical tweezers @xcite , spiral phase contrast microscopy@xcite . since the values of @xmath2 is theoretically unlimited , oam states construct an infinite dimensional hilbert space , and it exhibits great potential for applications in the field of quantum information process@xcite , free - space information transfer and communications@xcite . despite the extensive applications , determining the topological charge @xmath2 of oam state remains an intriguing problem , and a lot of methods has been proposed . such as mach - zehnder interferometer@xcite , diffraction pattern with specific masks@xcite , image reformatting@xcite , intensity analysis@xcite . recently , more and more researches have focused on the partially coherent vortex ( pcv ) beam @xcite . in 2004 , palacios _ et al . _ firstly verified that a robust phase singularity exists in the spatial coherence function when a vortex is presented @xcite in the original beam . meanwhile , in refs . @xcite , the authors revealed the linkage between the mode indices and the cross correlation function ( ccf ) of a pcv beam , and they also discussed the spatial coherence on determining the mode indices of a pcv beam . also , there has been interest in the use of spatial modes , such beams carrying oam , as an additional degree of freedom to increase the available information bandwidth for free - space communication . because a large hilbert space is helpful to improve the security of cryptographic keys transmitted with a quantum key distribution system@xcite . since the important application of free space communication , the effects of propagation through random aberrations ( atmospheric turbulence ) on coherence for single - photon communication systems based on orbital angular momentum states are also investigated @xcite . however , determining the mode indices of a pcv beam is still difficult . in refs . @xcite , the authors experimentally used a wavefront folding interferometer to measure the mode indices of a pcv beam . when the size of pcv beam is large compared to the measuring instrument , their scheme is difficult be used to measure the cross - correlation function . meanwhile , the interference pattern in their results is complex and the visibility is low , which is difficult to characterize the mode indices from the experimental results . in this letter , we experimentally prove that the mode indices of a pcv beam can be measured efficiently through a hanbury brown and twiss ( hbt ) type experiment which was firstly used in the field of astronomy @xcite . an incoherent or partially coherent source both tend to emit photons together ( bunching ) with enhanced photon number fluctuations relative to classical expectations ( super - poisson statistics ) . this kind of bunching effect was firstly measured in the experiment of observing the second - order temporal and spatial intensity correlations of star light by hanbury brown and twiss @xcite . it was believed that the nontrivial hbt correlation is caused by the measured intensity fluctuations of the thermal light . in the hbt experiment , the second - order correlation function is expressed as @xmath4 where @xmath5 denotes the ensemble average , @xmath6 and @xmath7 are the intensity at spatial point @xmath8 and @xmath9 , respectively . @xmath10 is the mutual coherence function ( mcf ) of the light source . since the second - order correlation function in eq . ( [ 00 ] ) can be easily obtained by the coincidence measurement of two detectors , we may get the mode indices of pcv beam from the measured magnitude square of mcf , @xmath11 . laguerre - gaussian ( lg ) modes are circularly symmetric and structurally stable solutions of the paraxial wave equation . the electromagnetic field amplitude of lg modes in @xmath12 is given by @xmath13 where @xmath14 and @xmath15 are the radial and azimuthal coordinates , respectively . @xmath16 is the radial mode index , and @xmath17 is the azimuthal mode index which describes the phase structure . @xmath18 denotes the waist width and @xmath19 is the generalised laguerre polynomial . and @xmath20 , respectively . mode indices for @xmath21 , and ( a ) @xmath22 ; ( b ) @xmath23 ; ( c ) @xmath24.,width=302 ] a partially coherent lg beam can be generated by propagating the coherent lg beam through a rotating ground glass disk ( ggd ) . it is well known that the statistical distribution @xcite of the generated beam corresponds to a gaussian - schell correlator @xmath25 $ ] , where @xmath26 represents the transverse spatial coherence width . combining eq . ( [ 01 ] ) with @xmath27 , the mcf of the partially coherent lg at source plane @xmath12 can be expressed as @xcite @xmath28 after propagating a distance @xmath29 in the far field , the mcf can be written as @xmath30 , \label{03}\end{aligned}\ ] ] where @xmath31 is the wave number . as we known , the far field ccf , @xmath32 , shows a closed relationship with the topological charge of the original vortex beam , and it has been proved that the ccf will maintain the dislocation rings . the blue - dash line in figs . [ 1d ] ( a)-(c ) show some simulation results of the far field one dimensional ccf for mode indices @xmath33 , @xmath34 lg beams . in the simulation , the waist width to spatial coherence width ratio is set as @xmath35 . it is straightforward to find that the number of dislocation rings in the @xmath32 is equal to the value of azimuthal mode index @xmath17 of the lg mode . since the direct measurement of the ccf is troublesome , we will discuss and measure the @xmath20 in the following . as shown in fig . [ 1d ] , the orange - solid lines represent the magnitude square of the ccf , @xmath20 , which have the same number of dislocation rings as the original far field ccf . it means that the hbt effect of the partially coherent beam could be used to reveal the features of the ccf indirectly . and l@xmath36 are lenses . p@xmath37 and p@xmath36 are pinholes . pbs is polarization beam splitter and slm is the phase spatial light modulate . ccd represents charge - coupled device camera . ( b ) and ( c ) are the the examples of holograms to generate lg beams with waist width @xmath18 , and their mode indices are ( @xmath38 ) and ( @xmath39 ) , respectively . , width=264 ] the experimental setup which uses the hbt effect to measure the mode indices of a pcv beam is shown in fig . [ setup ] ( a ) . a coherent gaussian beam is generated from the he - ne laser with wavelength @xmath40 nm , then it is focused on a ggd by lens l@xmath37 . the angular velocity of the ggd is kept at @xmath41 rad / s , and the spatial coherence width @xmath42 can be controlled by changing the distance between lens l@xmath37 and ggd . lens l@xmath36 and pinhole p@xmath37 are used to collimated the partially coherent beam after the ggd . then the beam is filtered by a polarization beam splitter ( pbs ) to the horizontal polarization to match the requirement of phase spatial light modulator ( slm ) . figs . [ setup ] ( b ) and ( c ) display the example computer - generated hologram gratings @xcite which are imprint on the phase slm ( hamamatsu model x10468 ) to generate the partially coherent lg beam . the gratings correspond to mode indices ( @xmath38 ) and ( @xmath39 ) , respectively . the waist width @xmath18 of partially coherent lg can also be handily controlled by changing the gratings on computer , and the average full width at half maximum of the speckles is 0.34 mm . the first - order diffracted beam is selected out by a pinhole p@xmath36 , and the intensity pattern is recorded by a charge - coupled device ( ccd ) . the ccd is a @xmath43 array of @xmath44m@xmath45 pixels , and the measurement is made with an exposure time of @xmath46 ms to guarantee that the acquired images are temporally incoherent . the field on the plane of the ccd can be regarded as the partially coherent lg beam . instead of using two point detectors for the coincidence measurement of the second - order correlation function in eq . ( [ 00 ] ) , a ccd is used here to recorded the intensity pattern and then analyse the second - order correlation on computer @xcite . ) modes . ( a)-(c ) and ( d)-(f ) are simulated and experimental results , respectively . the results correspond to waist width @xmath47.,width=245 ] to acquire the @xmath20 of the partially coherent lg beam with the setup in fig . [ setup ] ( a ) , we record @xmath48 frames of images in each experiment . according to definition of intensity correlation in eq . ( [ 00 ] ) , the recorded data for the magnitude square of ccf are processed as follows @xmath49 where @xmath50 is the @xmath51th frame of intensity distribution on the ccd plane and @xmath52 denotes the pixel coordinate of the ccd . before recording the data for @xmath20 , the ccd is placed beforehand on the plane of the slm in fig . [ setup ] ( a ) to measure spatial coherent width @xmath26 . ) mode with different waist width to spatial coherence width ratio . ( a)-(d ) and ( e)-(h ) are corresponding to simulated and experimental results , respectively.,width=302 ] fig . [ result1 ] shows the theoretically and experimentally obtained two dimensional @xmath20 of the far - field partially coherent lg beams with mode indices @xmath53 and waist width @xmath47 . here we see that the number of dislocation rings in @xmath20 matches with the azimuthal mode index @xmath17 of the original lg beam . the results are consistent with the earlier theoretical predication in ref . @xcite , and it means that one can measure the azimuthal index of a pcv beam through the hbt type measurement . furthermore , we investigate the character of @xmath20 under different waist width to spatial coherence width ratios . figs . [ result2 ] ( a)-(d ) and ( e)-(f ) show the simulated and experimental results with lg mode @xmath38 , and they are corresponded to 4 typical waist width to spatial coherence width ratios . the experimental patterns agree well with the theoretical distributions . from fig . [ result2 ] we can see that the bright spot in the center of the @xmath20 will gradually diminish in radius as the increase of the spatial coherent width , and eventually vanish in the situation of a coherent lg beam . on the contrary , the outside rings also will be too blurry to be invisible as the spatial coherence of the beam become quite low . the high and low coherence cases are shown in figs . [ result2 ] ( e ) and ( h ) , respectively . in both of these two situations , it might be difficult to count the number of dislocation rings for low contrast ratio . . to show the picture clearly , we reduce the contrast of the picture by setting the values of @xmath54 where @xmath55.,width=283 ] in ref . @xcite , the authors investigated the linkage between the number of dislocation rings @xmath56 in the far - field cross correlation function and the mode indices of a partially coherent lg beam , and they showed that @xmath57 . their experiments demonstrated that the spatial correlation singularity still exist even for a non - vortex ( @xmath58 ) partially coherent beam if the radial index @xmath59 is non - zero . likewise , we experimentally prove the relationship @xmath57 for some higher - order partially coherent lg beams with waist width to spatial coherence width ratios to be @xmath35 , and the results are shown in fig . [ result3 ] which have high visibility . figs . [ result3 ] ( a ) and ( d ) show the @xmath20 of non - vortex partially coherent lg beam with radial index @xmath60 and @xmath61 , respectively . it is obvious that the spatial correlation singularity still exists and the number of dislocation rings is @xmath62 . we also measured the @xmath20 of higher - order partially coherent lg modes which are shown in figs . [ result3 ] ( b ) , ( c ) , ( e ) , ( f ) . a desirable results in the experiment are achieved , and it is easy to count the number of dislocation rings as @xmath57 . in conclusion , we experimentally show the relationship between the number of dislocation rings of the magnitude square of the ccf and the radial and azimuthal mode indices ( @xmath63 ) of a pcv beam through a hbt type experiment , and our results prove that the number of dislocation rings is identical to @xmath64 . we also investigate the effect of spatial coherent width on determining the mode indices of the pcv beams . our method offers a powerful tool to investigate the character of the mcf of a pcv beam and determine the mode indices of the pcv beam , which is efficient and robust in practical measurement system because only the hbt type setup is used to get the second - order intensity correlation . it has potential applications in the field of free space communication and astronomical research with the photon s oam degree of freedom , especially in the situation with non - ignorable effect of atmospheric turbulence scattering medium .
semiconductor spintronics , which aims at utilizing or incorporating the spin degree of freedom in electronics , has attracted much interest.@xcite during the last decade , the fast developments of techniques of coherent manipulation of electron spins via optical or electrical methods@xcite have intrigued a lot of studies.@xcite many of the important findings are developed in bulk gaas or gaas epilayers , where the spin relaxation time ( srt ) was found to be as long as 130 ns,@xcite and the spin diffusion length was reported as large as 100 @xmath2m.@xcite remarkably , it has been found that the srt can vary by more than three orders of magnitude with temperature or electron density.@xcite the relevant spin relaxation mechanisms for electron system in the metallic regime have been recognized for a long time as : ( i ) the elliot - yafet ( ey ) mechanism in which electron spins have a small chance to flip during each scattering due to spin mixing in the conduction band;@xcite ( ii ) the dyakonov - perel ( dp ) mechanism in which the electron spins decay due to their precession around the * k-dependent spin - orbit fields ( inhomogeneous broadening@xcite ) during the free flight between adjacent scattering events;@xcite ( iii ) the bir - aronov - pikus ( bap ) mechanism in which electrons exchange their spins with holes.@xcite the hyperfine interaction is another mechanism which is usually important for spin relaxation of localized electrons and ineffective in metallic regime where most of the carriers are in extended states.@xcite despite decades of study , a detailed theoretical investigation from a fully microscopic approach in bulk system has not yet been performed . although song and kim have investigated electron spin relaxation due to all the three relevant mechanisms for various conditions in both @xmath0- and @xmath1-type semiconductors,@xcite in their work they use the analytical expressions based on single - particle approach which are only applicable for the non - degenerate electron system , and hence make the discussion in low temperature and/or high density regime questionable . more importantly , the carrier - carrier coulomb scattering , which has been shown to be very important for spin relaxation in two - dimensional systems,@xcite has not yet been well studied in bulk system . also , at finite spin polarization , the coulomb hartree - fock ( hf ) term acts as an effective longitudinal magnetic field , which has been demonstrated to be able to increase the srt by more than one order of magnitude when the initial spin polarization is high in two - dimensional electron system ( 2des).@xcite however , the effect of the coulomb hf term in bulk system has not been investigated . another issue is that the commonly used analytical formula for spin relaxation due to the bap mechanism is based on the elastic scattering approximation , which has been proved to be invalid for low temperature due to pretermission of the pauli blocking of electrons very recently by zhou and wu.@xcite consequently , the bap mechanism has been demonstrated to be unimportant in two - dimensional system at moderate and high excitation density first theoretically@xcite and then experimentally,@xcite which is in stark contrast with the common belief in the literature . whether it is still true in bulk system remains unchecked . furthermore , in most works only the short - range electron - hole exchange interaction was considered , while the long - range part was ignored.@xcite all these questions suggest that a detailed fully microscopic investigation is needed . in this work , we perform such a study from the fully microscopic kinetic spin bloch equation ( ksbe ) approach.@xcite we focus on the metallic regime where most of the carriers are in extended states . we restrict ourselves to the zero magnetic field case.@xcite previously , the ksbe approach has been applied extensively to study spin dynamics in semiconductor nanostructures in both markovian and non - markovian limits and in systems both near and far away from the equilibrium ( under strong static or terahertz electric field or with high spin polarization).@xcite the ksbe approach has been demonstrated to be successful in the study of spin relaxation in semiconductor quantum wells where good agreements with experiments have been achieved and many predictions have been confirmed by experiments.@xcite this approach has also been applied in the investigation of spin relaxation / dephasing in bulk gaas many years ago by wu and ning , where only the electron nonmagnetic - impurity and electron - phonon scatterings are included.@xcite in this work , we include all the scatterings , especially the electron - electron coulomb , electron - hole coulomb , and electron - hole exchange scatterings together with the ey mechanism which were not considered in ref . . an important goal of this work is to find the dominant mechanism in different parameter regimes and for different materials from the fully microscopic ksbe approach . previous investigations in the literature indicate that for @xmath0-type iii - v semiconductors , the spin relaxation is mostly dominated by the dp mechanism , except at low temperature where the ey mechanism is most important ; for intrinsic and @xmath1-type iii - v semiconductors , the bap mechanism dominates at low temperature when the hole density is high , whereas the dp mechanism dominates in other regimes.@xcite in the present work , from the fully microscopic ksbe approach , however , we find that the ey mechanism is less important than the dp mechanism in @xmath0-type iii - v semiconductors even for narrow band - gap semiconductors , such as inas and insb . for @xmath1-type iii - v semiconductors , we find that the bap mechanism dominates spin relaxation in the low temperature regime only when the photo - excitation density is low enough . however , when the photo - excitation density is high , the bap mechanism can be comparable with the dp mechanism only in the moderate temperature regime roughly around the fermi temperature of electrons , and for higher or lower temperature it is unimportant . the relative importance of the bap mechanism decreases with the photo - excitation density and eventually becomes negligible for sufficiently high photo - excitation density . for intrinsic iii - v semiconductors , the bap mechanism is negligible . an important method of spin injection is the hot - electron spin injection where high electric field is applied.@xcite moreover , in 2des , the spin relaxation can be effectively manipulated by the high in - plane electric field.@xcite in this work , we also study spin relaxation in @xmath0-type bulk semiconductors under high electric field . we show that there is some essential difference of the electric field effect on spin dynamics between 2des and the bulk system . using gaas as an example , we demonstrate that the electric field dependence of spin lifetime can be nonmonotonic or monotonic depending on the lattice temperature and the densities of impurities and electrons . the underlying physics is analyzed . the results indicate that the spin lifetime can be effectively controlled by electric field . this paper is organized as follows : in sec . ii , we introduce the ksbes . in sec . iii , we study the spin relaxation in @xmath0-type iii - v semiconductors . in secs . iv and v , we investigate the spin relaxations in intrinsic and @xmath1-type iii - v semiconductors , respectively . we study the effects of electric field on spin relaxation in @xmath0-type iii - v semiconductors in sec . finally , we conclude in sec . the spin dynamics is studied by solving the microscopic ksbes derived via the nonequilibrium green function method,@xcite @xmath3 here @xmath4 is the single particle density matrix with the diagonal terms representing the distributions of each spin band , and the off - diagonal terms denoting the correlation of the two spin bands . the coherent term is given by @xmath5.\ ] ] here @xmath6 $ ] is the commutator and @xmath7 with @xmath8 due to the dresselhaus spin - orbit coupling ( soc)@xcite and @xmath9 due to the strain induced soc.@xcite @xmath10 is the coulomb hf term of the electron - electron interaction . previously , it was found that the coulomb hf term serves as a longitudinal effective magnetic field which increases with the initial spin polarization . the effective magnetic field can be as large as 40 t , which blocks the inhomogeneous broadening of the @xmath11-dependent spin - orbit field and reduces the spin relaxation due to the dp mechanism in 2des.@xcite however , the effect of the hf effective magnetic field on spin relaxation in bulk system is still unknown , which is one of the goal of this paper . the drift term is given by,@xcite @xmath12 where @xmath13 is the electron charge ( @xmath14 ) and @xmath15 is the electric field . as the hole spin relaxation is very fast ( @xmath16100 fs),@xcite one can assume that the hole system is kept in the thermal equilibrium state where the hole distribution ( @xmath17 ) is described by the fermi distribution . the scattering term @xmath18 contains the contributions from the electron - impurity scattering @xmath19 , the electron - phonon scattering @xmath20 , the electron - electron scattering @xmath21 , the electron - hole coulomb scattering @xmath22 , and the electron - hole exchange scattering @xmath23 , @xmath24 these terms read @xmath25 @xmath26 @xmath27 + { \rm h.c . } , \label{scat_ee}\end{aligned}\ ] ] @xmath28 + { \rm h.c . } , \label{scat_eh}\end{aligned}\ ] ] @xmath29 + { \rm h.c.}. \label{scat_ex}\end{aligned}\ ] ] in these equations , @xmath30 describes the spin - mixing due to the conduction - valence band mixing which originates from the ey spin relaxation mechanism.@xcite here @xmath31 with @xmath32 . @xmath33 and @xmath34 are the band - gap and the spin - orbit splitting of the valence band , respectively.@xcite @xmath35 is the impurity density . @xmath36 is the charge number of the impurity which is taken to be @xmath37 throughout the paper.@xcite @xmath38 with @xmath39 being the conduction band effective mass . @xmath40 is the screened coulomb potential where the screening is treated within the random phase approximation ( rpa),@xcite @xmath41 where @xmath42 here @xmath43 is the bare coulomb potential with @xmath44 and @xmath45 representing the vacuum permittivity and the static dielectric constant , respectively ; @xmath46 is the electron distribution on @xmath47-spin band ; @xmath17 is the hole distribution on the hole @xmath48-spin band . @xmath49 stands for the hole energy dispersion with @xmath50 being the hole effective mass . for heavy - hole ( @xmath51 ) , @xmath52 ; while for light - hole ( @xmath53 ) , @xmath54 . here @xmath55 and @xmath56 are the parameters of the luttinger - kohn hamiltonian in the spherical approximation@xcite and @xmath57 is the free electron mass . note that in these bands the spins are mixed.@xcite consequently , there are form factors in the electron - hole coulomb scattering : @xmath58 s are the eigen - states of the hole hamiltonian which can be found in ref . . @xmath59 is the matrix element of electron - phonon interaction with @xmath60 being the phonon branch index , which is further composed of the electron longitudinal - optical(lo)-phonon and electron acoustic - phonon interactions . the expressions of @xmath59 can be found in ref . . @xmath61 is the phonon energy spectrum . @xmath62 with @xmath63 being the phonon number at lattice temperature . for electron - hole exchange scattering , the form factor @xmath64 in eq . ( [ scat_ex ] ) comes from the electron - hole exchange interaction . the hamiltonian for the electron - hole exchange interaction consists of two parts : the short - range part @xmath65 and the long - range part @xmath66.@xcite the short - range part contributes a term proportional to @xmath67 , where @xmath68 is the exchange splitting of the exciton ground state and @xmath69 with @xmath70 being the exciton bohr radius . @xmath71 and @xmath72 are the hole and electron spin operators respectively . the long - range part gives a term proportional to @xmath73 where @xmath74 is the longitudinal - transverse splitting ; @xmath75 , @xmath76 , @xmath77 , and @xmath78 are operators in hole spin space . @xmath79 are the electron spin ladder operators . the expressions for @xmath80 s can be obtained from the expressions of @xmath66 in ref . . specifically , the spin - flip matrix @xmath77 in the spin unmixed base can be written as ( in the order of @xmath81 , @xmath82 , @xmath83 , @xmath84 ) , @xmath85 - \frac{2}{\sqrt{3 } } k_{\parallel}^2 & \frac{4}{3 } k_z k_{- } & \frac{2}{3 } k_{-}^2 & 0 \\[8pt ] \frac{4}{\sqrt{3}}k_z k_{+ } & -\frac{8}{3 } k_z^2 & -\frac{4}{3 } k_z k_{- } & 0 \\[8pt ] 2 k_{+}^2 & -\frac{4}{\sqrt{3 } } k_z k_{+ } & -\frac{2}{\sqrt{3 } } k_{\parallel}^2 & 0 \end{bmatrix}.\end{aligned}\ ] ] here @xmath86 is the center - of - mass momentum of the interacting electron - hole pair . @xmath87 and @xmath88 . @xmath89 corresponds to the spin - conserving process which is irrelevant in the study of spin relaxation . summing up the contribution from the short - range and long - range parts and keeping only the spin - flip terms , one obtains @xmath90 with @xmath91 $ ] . @xmath92 are the hole spin ladder operators . it should be noted that in gaas , @xmath93 mev is four times as large as @xmath94 mev.@xcite thus in gaas the long - range part of electron - hole exchange interaction should be more important than the short - range part . later in this paper , we will compare the contributions from the short - range and long - range parts to show that the long - range part is much more important than the short - range part in gaas . we will also compare the results from the fully microscopic ksbe approach with those from the analytical formula widely used in the literature.@xcite in summary , all relevant spin relaxation mechanisms have been fully incorporated in our ksbe approach . by solving the ksbes [ eq . ( [ ksbe ] ) ] , one obtains the temporal evolution of spin density matrix @xmath4 . after that , the time evolution of macroscopic quantities , such as , the electron spin density along the @xmath95-axis , @xmath96 $ ] , are obtained . by fitting the decay of @xmath97 with an exponential decay , one obtains the srt . the srts under various conditions are studied with the underlying physics discussed . the material parameters used are listed in table i. the parameter for strain - induced soc @xmath98 is always taken to be zero unless otherwise specified . the numerical scheme is laid out in appendix a. the error of our computation is less than @xmath99 according to our test . we first compare our calculation with experiments . in appendix b , we compare the srts calculated from the ksbes with those from experiments in refs . , and . we find good agreement with experimental data in almost the whole temperature or photo - excitation density range with only one fitting parameter @xmath100.@xcite this demonstrates that our calculation has achieved _ quantitative accuracy _ in modeling the spin relaxation in metallic regime . .material parameters used in the calculation ( from ref . unless otherwise specified ) [ cols="<,<,<,<,<,<,<,<,<,<",options="header " , ] + @xmath101 the soc parameter @xmath100 is from ref . , + other parameters are from ref . . in this work , we reexamine the problem from the fully microscopic ksbe approach . let us first examine the case of insb . insb is a narrow band - gap semiconductor where the spin relaxation is believed to be dominated by the ey mechanism at low temperature previously.@xcite the factor @xmath102 is much smaller than that of gaas , which indicates the importance of the ey mechanism according to eq . ( [ eydp2 ] ) ( a list of the factor @xmath102 for different materials is given in table ii ) . in fig . [ fig : eyinsb](a ) , we plot the ratio of the srt due to the ey mechanism , @xmath103 , to that due to the dp mechanism , @xmath104 , calculated from the ksbes , as function of temperature for various electron densities . in the calculation , @xmath105 . remarkably , it is noted that the ratio @xmath106 is always larger than 1 , and in most cases it is even larger than 10 , i.e. , the spin relaxation in @xmath0-type insb is _ not _ dominated by the ey mechanism . moreover , the temperature dependence is not monotonic , which is different from the intuition given by eq . ( [ eydp2 ] ) . this can be understood as following : the increase of the ratio comes from the increase of @xmath107 which is understood easily . the decrease of the ratio comes from the decrease of @xmath108 due to the increase of the electron lo - phonon scattering with temperature which becomes important for @xmath109 k in insb . therefore , after a crossover regime , the ratio eventually decreases with temperature . moreover , for large electron density @xmath110 @xmath111 , the electron system is in the degenerate regime and @xmath107 changes slowly with temperature which further facilitates the decrease of the ratio . to elucidate it clearly , we plot the srt due to the ey mechanism and dp mechanism in fig . [ fig : eyinsb](b ) and ( c ) respectively . it is seen from fig . [ fig : eyinsb](b ) that the srt due to the ey mechanism always decreases with temperature due to both the increase of @xmath107 and the enhancement of the scattering . from fig . [ fig : eyinsb](c ) , it is further noted that the temperature dependence of the srt due to the dp mechanism is , however , different for low density and high density regimes : in low density regime , the srt decreases with the temperature , whereas in high density regime it increases . the decreases of srt with temperature at low density is consistent with previous studies in the literature.@xcite the increases of srt with temperature at high density , however , is because that the electron lo - phonon scattering increases with temperature faster than the inhomogeneous broadening @xmath112 in the degenerate regime . note that the increase of the srt with temperature at such high temperature ( @xmath113 k ) without any magnetic field has neither been observed in experiment nor been predicted in theory . note that the results for high density cases are only given at high temperatures due to the limitation of our computation power . however , according to the above analysis , the ratio in the low temperature regime should be larger than its value at 300 k for high density cases . for low density cases with temperature lower than our calculation range , both the momentum scattering and the inhomogeneous broadening change little with temperature as the electron system is degenerate . therefore the ratio @xmath106 changes slowly . consequently , it does not change the conclusion that the ey mechanism is less efficient than the dp mechanism in @xmath0-type insb.@xcite in contrast , based on eq . ( [ eydp2 ] ) , song and kim reported that the ey mechanism is more important than the dp mechanism for temperature lower than 5 k.@xcite what makes our conclusion different from that in the literature is that most of the previous investigations use eqs . ( [ dpapp ] ) and ( [ eyapp ] ) to calculate @xmath104 and @xmath103.@xcite however , these equations are only applicable for non - degenerate electron system . if it is used in the degenerate electron system , it _ exaggerates _ the relative efficiency of the ey mechanism according to eq . ( [ eydp1 ] ) . ratio of the srt due to the ey mechanism @xmath103 to that due to the dp mechanism @xmath104 ( a ) , @xmath103 ( b ) and @xmath104 ( c ) as function of temperature for various electron densities . @xmath114 @xmath111 ( curve with @xmath115 ) , @xmath116 @xmath111 ( curve with @xmath117 ) , @xmath118 @xmath111 ( curve with @xmath119 ) , @xmath120 @xmath111 ( curve with @xmath121 ) . the corresponding fermi temperatures for those densities are @xmath122 , 300 , 670 , and 1390 k respectively . @xmath123.,title="fig : " ] -insb . ratio of the srt due to the ey mechanism @xmath103 to that due to the dp mechanism @xmath104 ( a ) , @xmath103 ( b ) and @xmath104 ( c ) as function of temperature for various electron densities . @xmath114 @xmath111 ( curve with @xmath115 ) , @xmath116 @xmath111 ( curve with @xmath117 ) , @xmath118 @xmath111 ( curve with @xmath119 ) , @xmath120 @xmath111 ( curve with @xmath121 ) . the corresponding fermi temperatures for those densities are @xmath122 , 300 , 670 , and 1390 k respectively . @xmath123.,title="fig : " ] -insb . ratio of the srt due to the ey mechanism @xmath103 to that due to the dp mechanism @xmath104 ( a ) , @xmath103 ( b ) and @xmath104 ( c ) as function of temperature for various electron densities . @xmath114 @xmath111 ( curve with @xmath115 ) , @xmath116 @xmath111 ( curve with @xmath117 ) , @xmath118 @xmath111 ( curve with @xmath119 ) , @xmath120 @xmath111 ( curve with @xmath121 ) . the corresponding fermi temperatures for those densities are @xmath122 , 300 , 670 , and 1390 k respectively . @xmath123.,title="fig : " ] we further examine other two iii - v semiconductors : inas and gaas . figure [ fig : eyinas ] shows the ratio @xmath124 for different electron densities as function of temperature for both inas and gaas . it is seen that the ratio is always larger than 1 for both inas@xcite and gaas,@xcite which is different from previous results in the literature.@xcite especially , for gaas the ratio is larger than 100 , which indicates that the ey mechanism is _ irrelevant _ for @xmath0-type gaas . the temperature dependence of the ratio is similar to that in insb and the underlying physics is also the same . in summary , we find that _ the ey mechanism is less efficient than the dp mechanism in @xmath0-type gaas , inas , and insb_. according to the data listed in table ii , we believe that the same conclusion holds for other @xmath0-type iii - v semiconductors . -inas ( solid curves ) and @xmath0-gaas ( dashed curve ) . ratio of the srt due to the ey mechanism @xmath103 to that due to the dp mechanism @xmath104 as function of temperature for various electron densities : @xmath114 @xmath111 ( curve with @xmath115 ) , @xmath125 @xmath111 ( curve with @xmath117 ) , @xmath126 @xmath111 ( curve with @xmath119 ) for inas , and @xmath114 @xmath111 ( curve with @xmath121 ) for gaas ( note that the value in the figure has been rescaled by a factor of 0.02 ) . the fermi temperature @xmath127 is 85 , 629 , and 1840 k for inas as well as 29 k for gaas respectively . ] as the bap mechanism and the ey mechanism are unimportant for spin relaxation in @xmath0-type iii - v semiconductors , in this subsection we focus on spin relaxation due to the dp mechanism . the effect of the electron - electron coulomb scattering on spin relaxation is studied . previously , it was found that the electron - electron scattering plays an important role in two - dimensional system.@xcite especially , it becomes the dominant scattering mechanism in high mobility samples at low temperature , where other scattering mechanisms are relatively weak.@xcite consequently @xmath128 in eq . ( [ tau - taup ] ) should be replaced by @xmath129 with @xmath130 including the electron - electron scattering @xmath131 [ later we will use the symbols @xmath132 and @xmath133 to denote the ensemble averaged value].@xcite remarkably , the srt due to the electron - electron scattering has _ nonmonotonic _ temperature dependence : in the low temperature ( degenerate ) regime the srt increases with temperature as the electron - electron scattering does ; in the high temperature ( non - degenerate ) regime the electron - electron scattering decreases with the temperature , so does the srt.@xcite thus there is a peak @xmath134 in the temperature dependence of the srt which is comparable with the fermi temperature.@xcite this prediction was confirmed by experiments very recently.@xcite however , in bulk semiconductors , the role of electron - electron scattering in spin relaxation is still unclear . although glazov and ivchenko have discussed the problem , they only gave an approximate expression of the srt due to the electron - electron scattering in the non - degenerate regime , while the relative importance of the electron - electron scattering compared with other scattering mechanisms is not addressed.@xcite in this subsection , we present a close study on the effect of the electron - electron scattering on spin relaxation in bulk semiconductors . we use gaas as an example , where the behavior is applicable to all iii - v semiconductors given that the system is in the motional narrowing regime . -gaas at @xmath135 k. ( a ) srt @xmath136 as function of electron density @xmath137 ( @xmath123 ) from full calculation ( curve with @xmath115 ) , from calculation with only the electron - electron scattering ( curve with @xmath117 ) , with only the electron - impurity scattering ( curve with @xmath119 ) , and with only the electron - phonon scattering ( curve with @xmath121 ) ; ( b ) srt @xmath136 as function of electron density @xmath137 ( @xmath123 ) for the case with strain ( curve with @xmath115 : with both the linear and the cubic soc ; curve with @xmath119 : with only the linear soc ) and the case without strain ( curve with @xmath117).,title="fig : " ] -gaas at @xmath135 k. ( a ) srt @xmath136 as function of electron density @xmath137 ( @xmath123 ) from full calculation ( curve with @xmath115 ) , from calculation with only the electron - electron scattering ( curve with @xmath117 ) , with only the electron - impurity scattering ( curve with @xmath119 ) , and with only the electron - phonon scattering ( curve with @xmath121 ) ; ( b ) srt @xmath136 as function of electron density @xmath137 ( @xmath123 ) for the case with strain ( curve with @xmath115 : with both the linear and the cubic soc ; curve with @xmath119 : with only the linear soc ) and the case without strain ( curve with @xmath117).,title="fig : " ] we first discuss the electron density dependence of the srt . it is noted that during the variation of electron density , the impurity density also varies as @xmath123 . this is different from the situation in 2des , where the impurity density can be different from the electron density due to the modulation doping . in fig . [ fig : ne](a ) , we plot the srt as function of electron density for @xmath135 k. it is noted that , remarkably , the density dependence of the srt is _ nonmonotonic _ and there is a peak in the @xmath136-@xmath137 curve . previously , the nonmonotonic density dependence of the srt was observed in low temperature ( @xmath138 k ) measurements , where the localized electrons play an important role and the electron system is in the mott metal - insulator transition area.@xcite the localized electrons have different spin relaxation mechanisms and the scatterings of the localized electrons and free electrons give rise to the nonmonotonic density dependence.@xcite it is noted that , up till now , there is no report on the nonmonotonic density dependence in the metallic regime in @xmath0-type bulk iii - v semiconductors . it should be further pointed out that the nonmonotonic density dependence and the appearance of the peak is a _ universal _ behavior in the metallic regime in @xmath0-type bulk iii - v semiconductors . we also plot the srts calculated with only the electron - electron scattering ( curve with @xmath117 ) , with only the electron - impurity scattering ( curve with @xmath119 ) , and with only the electron - phonon scattering ( curve with @xmath121 ) in the figure to elucidate the role of these scatterings in spin dynamics . it is noted that the electron - impurity and electron - electron scatterings are the relevant scattering mechanisms , while the electron - phonon scattering is much weaker than the two as the temperature is low . interestingly , both the electron - electron scattering and the electron - impurity scattering lead to nonmonotonic density dependence of srt . for the srt due to the electron - electron scattering , the nonmonotonic behavior comes from the nonmonotonic density dependence of the electron - electron scattering . the density and temperature dependences of the electron - electron scattering have been investigated in spin - unrelated problems.@xcite from these works , after some approximation , the asymptotic density and temperature dependences of the electron - electron scattering time @xmath133 in the degenerate and non - degenerate regimes are given by,@xcite @xmath139 from above equations , one notices that the electron - electron scattering in the non - degenerate and degenerate regimes has different density and temperature dependence . in the non - degenerate ( low density ) regime the electron - electron scattering increases with electron density , while the inhomogeneous broadening @xmath140 changes slowly as the distribution function is close to the boltzmann distribution . the srt thus increases with density . in degenerate ( high density ) regime , both @xmath133 and the inhomogeneous broadening increases with electron density . thus , the srt , which can be estimated as @xmath141$],@xcite decreases with density . for the srt due to the electron - impurity scattering , the scenario is similar : in the non - degenerate regime , the distribution function is close to the boltzmann distribution . the inhomogeneous broadening hence changes slowly with density . the electron - impurity scattering increases as @xmath142 , which increases with the impurity density because @xmath143 changes little with the density when the distribution is close to the boltzmann distribution . the srt hence increases with density . in the degenerate regime , the inhomogeneous broadening increases as @xmath144 . on the other hand , the electron - impurity scattering decreases with density because it is proportional to @xmath145 . consequently , the srt decreases with density . therefore , both the electron - electron and electron - impurity scatterings contribute to the nonmonotonic density dependence of the srt . the peak density @xmath146 appears in the crossover regime , where the corresponding fermi temperature is comparable with the lattice temperature . a careful calculation gives the peak density as @xmath147 @xmath111 , with the corresponding fermi temperature being 37 k , close to the lattice temperature of 40 k. we further discuss the electron density dependence of the srt with strain - induced soc . in fig . [ fig : ne](b ) , we plot the density dependence of the srt for @xmath135 k under strain with @xmath123 . we choose @xmath148 mev@xmath149.@xcite for this value , at low density the soc is dominated by the linear term due to strain . however , in high density regime ( @xmath150 @xmath111 ) , the cubic dresselhaus term can surpass the linear term . it is noted that the density dependence is also nonmonotonic and exhibits a peak . the underlying physics is similar : in the non - degenerate regime , the srt increases with the density because both electron - electron and electron - impurity scattering increase with density and the inhomogeneous broadening changes little . in the degenerate regime , the inhomogeneous broadening increases as @xmath151 , whereas both the electron - electron and electron - impurity scatterings decrease with the electron density , which thus leads to the decrease of the srt with density . it is further noted that the peak is at @xmath152 @xmath111 which is larger than the peak density in the case without strain . this is because that the inhomogeneous broadening here increases as @xmath153 , which is much slower than that of @xmath154 in the strain - free case . nevertheless , the increase of the scattering with density remains the same , therefore the peak shows up at a larger electron density . this is also confirmed in the figure that the srt with only the linear term decreases slower than the one with only the cubic dresselhaus term in the high density regime . however , in the low density regime , the srt in the case with strain increases as fast as the one in the strain - free case . this is because here the increase of the srt is due to the increase of the scattering , whereas the inhomogeneous broadening changes little . we now study the temperature dependence of the srt . in fig . [ fig : t ] , the srt as function of temperature is plotted for @xmath155 @xmath111 . from the figure , it is seen that the srt decreases with temperature monotonically , which coincides with previous experimental results.@xcite this trend is also the same as that in the 2des with high impurity density.@xcite however , for high mobility 2des , which can be achieved by modulation doping , the temperature dependence of the srt is nonmonotonic and there is a peak @xmath134 around the fermi temperature due to the electron - electron scattering.@xcite in @xmath0-type bulk materials , as the impurity density is always equal to or larger than the electron density , this peak disappears . nevertheless , similar effect can be obtained if one artificially reduces the impurity density . for example when @xmath156 , it is seen from fig . [ fig : t ] that the srt shows nonmonotonic behavior with a peak around 40 k. -gaas . srt @xmath136 as function of @xmath157 for @xmath123 ( solid curve with @xmath115 ) and @xmath158 ( dashed curve with @xmath117 ) . @xmath155 @xmath111 . ] we now turn to the dependence of initial spin polarization on spin relaxation . previously , it was discovered that at finite spin polarization , the coulomb hf term serves as an effective magnetic field along the direction of the spin polarization.@xcite the effective magnetic field can be as large as 40 t at high spin polarization in 2des which suppresses the dp spin relaxation.@xcite this effect was first predicted by weng and wu@xcite and then confirmed by experiments very recently.@xcite however , the effect of the coulomb hf term on spin relaxation in bulk system still needs to be evaluated . here , we present such an investigation . dependence of srt @xmath136 on initial spin polarization @xmath159 for @xmath123 with ( without ) the coulomb hf term [ curve with @xmath115 ( @xmath117 ) ] and for @xmath160 with ( without ) the coulomb hf term [ curve with @xmath119 ( @xmath121 ) ] . @xmath155 @xmath111 and @xmath161 k. ] the coulomb hf term for spin polarization along , e.g. , the @xmath95-direction can be written as @xmath162 the corresponding effective magnetic field is along the @xmath95-axis , @xmath163 under this effective magnetic field , the spin precession is blocked and the srt is elongated.@xcite recently , this effective magnetic field has been probed experimentally,@xcite and its effect on spin accumulation in 2des was also discussed theoretically.@xcite in fig . [ fig : diffp1 ] , we plot the srt as function of the initial spin polarization @xmath159 for @xmath155 @xmath111 and @xmath161 k. it is seen that the srt increases with the initial spin polarization . to elucidate the effect of the coulomb hf term , we also plot the results from the calculation without the coulomb hf term . the results indicate that the increase of the srt with initial spin polarization is due to the coulomb hf term . however , the increment is less than @xmath164 . previously , it was shown in 2des that the srt can increase over 30 times for low impurity case at 120 k , while less than 3 times for high impurity or high temperature case where the scattering is strong.@xcite the results can be understood as follows : the srt under the hf effective magnetic field can be estimated as@xcite @xmath165 \label{hfeq}\ ] ] where @xmath166 is the averaged effective magnetic field . thus the effect of the hf effective magnetic field increases with @xmath132 , i.e. , the effect is more pronounced for weak scattering case , such as the low impurity density case . however , in bulk system the impurity density is always equal to or larger than the electron density @xmath167 , therefore the effect of the coulomb hf term is suppressed . this can be seen from the calculation with @xmath160 ( the artificial case ) . the results are also plotted in the figure . one finds that the coulomb hf term effectively enhances the srt at low impurity density by 40 times . therefore , due to the large impurity density ( @xmath167 ) , the srt is insensitive to the initial spin polarization in @xmath0-type bulk iii - v semiconductors . we have checked that the conclusion holds for other cases with different temperatures , electron densities and materials . @xmath111 . ( a ) srt @xmath136 due to the bap and dp mechanisms as function of temperature . ( b ) srt due to the bap mechanism calculated from eq . ( [ bapapp1 ] ) ( dotted curve with @xmath119 ) , from the ksbe approach with both long - range and short - range exchange scatterings ( solid curve with @xmath115 ) as well as from the ksbe approach with only the short - range exchange scattering ( solid curve with @xmath117 ) . ( c ) srt due to the dp mechanism from full calculation ( solid curve with @xmath115 ) , from the calculation without the coulomb hf term ( dashed curve with @xmath117 ) , and from the calculation without the electron - hole coulomb scattering ( dotted curve with @xmath119 ) . for electrons @xmath168 k , and for holes @xmath169 k.,title="fig : " ] @xmath111 . ( a ) srt @xmath136 due to the bap and dp mechanisms as function of temperature . ( b ) srt due to the bap mechanism calculated from eq . ( [ bapapp1 ] ) ( dotted curve with @xmath119 ) , from the ksbe approach with both long - range and short - range exchange scatterings ( solid curve with @xmath115 ) as well as from the ksbe approach with only the short - range exchange scattering ( solid curve with @xmath117 ) . ( c ) srt due to the dp mechanism from full calculation ( solid curve with @xmath115 ) , from the calculation without the coulomb hf term ( dashed curve with @xmath117 ) , and from the calculation without the electron - hole coulomb scattering ( dotted curve with @xmath119 ) . for electrons @xmath168 k , and for holes @xmath169 k.,title="fig : " ] @xmath111 . ( a ) srt @xmath136 due to the bap and dp mechanisms as function of temperature . ( b ) srt due to the bap mechanism calculated from eq . ( [ bapapp1 ] ) ( dotted curve with @xmath119 ) , from the ksbe approach with both long - range and short - range exchange scatterings ( solid curve with @xmath115 ) as well as from the ksbe approach with only the short - range exchange scattering ( solid curve with @xmath117 ) . ( c ) srt due to the dp mechanism from full calculation ( solid curve with @xmath115 ) , from the calculation without the coulomb hf term ( dashed curve with @xmath117 ) , and from the calculation without the electron - hole coulomb scattering ( dotted curve with @xmath119 ) . for electrons @xmath168 k , and for holes @xmath169 k.,title="fig : " ] in this section we study spin relaxation in intrinsic iii - v semiconductors . for intrinsic semiconductors , the electrons and holes are created by optical excitation , and their numbers are equal . compared to @xmath0-type semiconductors , there are two new scattering mechanisms : the electron - hole coulomb and electron - hole exchange scatterings , where the latter corresponds to the bap mechanism . another important property of intrinsic semiconductors is that the impurity density is very low ( we take @xmath170 ) , which offers a good platform for demonstrating the effect of the many - body carrier - carrier scattering on spin relaxation . moreover , the effect of the coulomb hf term would be enhanced as the electron - impurity scattering can be eliminated . our first goal is to compare the relative efficiency of the dp and bap mechanisms . after that we also study the temperature and photo - excitation density @xmath171 dependences of the srt . the role of electron - hole coulomb scattering as well as the effect of the coulomb hf term are also addressed . we further compare the results from the ksbes with those from the widely used analytical formulae [ eqs . ( [ bapapp1 ] ) and ( [ bapapp2 ] ) ] . the initial spin polarization is chosen to be @xmath164 which corresponds to circularly polarized optical excitation . we focus on gaas , while the situation is similar for other iii - v semiconductors.@xcite in fig . [ fig : t_int](a ) , we plot the srts due to the dp and bap mechanisms as function of temperature for @xmath172 @xmath111 . the srt due to the bap mechanism alone is calculated by removing the spin precession due to the soc , but keeping all the scattering terms . it is noted that the srt due to the bap mechanism is larger than that due to the dp mechanism by more than one order of magnitude , which indicates that the bap mechanism is negligible for intrinsic gaas.@xcite moreover , the spin relaxation due to the dp mechanism increases with temperature more rapidly than that due to the bap mechanism at high temperature . this is because the increase of spin relaxation due to the dp mechanism mainly comes from the increase of the inhomogeneous broadening which is proportional to @xmath173 in high temperature ( non - degenerate ) regime . meanwhile , according to eq . ( [ bapapp1 ] ) , the increase of spin relaxation due to the bap mechanism mainly comes from the increase of @xmath174 which is proportional to @xmath175 in that regime . for a close examination of the bap mechanism , we also plot the srt limited by the the short - range electron - hole exchange scattering calculated from the ksbes in fig . [ fig : t_int](b ) . it is seen that the srt limited by the short - range electron - hole exchange scattering is much larger than that limited by both long- and short - range exchange scattering . this confirms that the long - range scattering is more important than the short - range one in gaas as @xmath74 is four times larger than @xmath68 . therefore , previous investigations@xcite with only the short - range exchange scattering included are questionable . moreover , to check the validity of the widely used elastic scattering approximation , we also compare the results from the ksbes with those from the elastic scattering approximation . under the elastic scattering approximation,@xcite @xmath176 the srt is then obtained by averaging over the electron distribution . ( [ bapapp1 ] ) and ( [ bapapp2 ] ) are derived from the above equation under some approximations . for example , by including only the short - range exchange scattering and ignoring the light - hole contribution as well as the term of @xmath177 , eq . ( [ bapapp1 ] ) is obtained . to show that the elastic scattering approximation fails in the degenerate regime , we compare our results with the results from eq . ( [ bapelastapp1 ] ) . for simplicity , we include only the short - range exchange scattering . in fig . [ fig : t_int](b ) , we plot the srt obtained from eq . ( [ bapelastapp1 ] ) as the dotted curve . it is seen that the result from eq . ( [ bapelastapp1 ] ) agrees well with our result from the ksbes at high temperature , but deviates at low temperature . the deviation is due to the pauli blocking of electrons in the degenerate regime , which is neglected in the elastic scattering approximation ( @xmath168 k).@xcite @xmath111 . srt @xmath136 as function of temperature for @xmath178 ( curve with @xmath117 ) and the srt as function of initial spin polarization @xmath159 for @xmath161 k ( curve with @xmath115 ) ( note that the scale of @xmath159 is on the top of the frame).,height=207 ] we also discuss the effect of the electron - hole coulomb scattering and the coulomb hf term on spin relaxation due to the dp mechanism . in fig . [ fig : t_int](c ) , we plot the srts due to the dp mechanism obtained from the full calculation , from the calculation without the electron - hole coulomb scattering , and from the calculation without the coulomb hf term . let us first examine the effect of the coulomb hf term on spin relaxation . it is seen that the coulomb hf term has important effect on spin relaxation only for low temperature case@xcite ( @xmath179 k ) which is consistent with the results in 2des.@xcite we then turn to the effect of the electron - hole coulomb scattering . it is seen that without the electron - hole coulomb scattering the srt is larger for @xmath179 k but smaller for @xmath180 k compared with that from the full calculation . this behavior can be understood as following : for @xmath181 k , the coulomb hf term has important effect on spin relaxation . the hf effective magnetic field elongates the srt . according to eq . ( [ hfeq ] ) , this effect increases with the momentum scattering time . without the electron - hole coulomb scattering the momentum scattering time is elongated , which enhances the effect and leads to longer srt . for higher temperature ( @xmath182 k ) , the effect of the coulomb hf term is weak,@xcite and the system returns back to the motional narrowing regime . the srt thus decreases when the electron - hole coulomb scattering is removed . the results indicate that the electron - hole coulomb scattering is comparable with the electron - electron and electron lo - phonon scatterings . in other words , besides the screening from holes , the main contribution of the hole system to electron spin relaxation comes from the electron - hole coulomb scattering in intrinsic semiconductors . in fig . [ fig : tp_int ] , we plot the srt as function of temperature for @xmath183 @xmath111 with @xmath178 . in sec . iii b2 , we showed that there is a peak in the temperature dependence of srt due to the electron - electron coulomb scattering when the impurity density is low . in intrinsic semiconductors , as the impurity density is very low , the peak may appear . indeed , we find that the srt has a peak at @xmath184 k. the peak temperature @xmath134 is comparable with the fermi temperature ( @xmath185 k ) [ actually , our calculation indicates that the peak temperature @xmath134 is around @xmath186 and lies in the range of @xmath187 depending on the carrier density . ] nevertheless , at high spin polarization , such as @xmath188 , the peak disappears as indicated in fig . [ fig : t_int](c ) . this peak can be observed within current technology of optical orientation . however , up till now , no such experimental investigation has been performed . in fig . [ fig : tp_int ] , we also plot the srt as function of initial spin polarization at @xmath161 k. it is seen that the srt is elongated by 9 times when @xmath159 is tuned from @xmath189 to @xmath164 . therefore , the effect of the coulomb hf term can also be observed in intrinsic materials and is more pronounced compared to the @xmath0-type case . due to the bap mechanism and that due to the dp mechanism as function of photo - excitation density @xmath171 at : @xmath135 k ( a ) and @xmath190 k ( b ) ( note that the value of @xmath191 in figure ( b ) has been rescaled by a factor of 0.01 ) . ( c ) : the srt due to the dp mechanism from the full calculation ( curve with @xmath115 ) , from the calculation without the coulomb hf term ( curve with @xmath117 ) , and from the calculation without the electron - hole coulomb scattering ( curve with @xmath119 ) for @xmath135 k.,title="fig : " ] due to the bap mechanism and that due to the dp mechanism as function of photo - excitation density @xmath171 at : @xmath135 k ( a ) and @xmath190 k ( b ) ( note that the value of @xmath191 in figure ( b ) has been rescaled by a factor of 0.01 ) . ( c ) : the srt due to the dp mechanism from the full calculation ( curve with @xmath115 ) , from the calculation without the coulomb hf term ( curve with @xmath117 ) , and from the calculation without the electron - hole coulomb scattering ( curve with @xmath119 ) for @xmath135 k.,title="fig : " ] due to the bap mechanism and that due to the dp mechanism as function of photo - excitation density @xmath171 at : @xmath135 k ( a ) and @xmath190 k ( b ) ( note that the value of @xmath191 in figure ( b ) has been rescaled by a factor of 0.01 ) . ( c ) : the srt due to the dp mechanism from the full calculation ( curve with @xmath115 ) , from the calculation without the coulomb hf term ( curve with @xmath117 ) , and from the calculation without the electron - hole coulomb scattering ( curve with @xmath119 ) for @xmath135 k.,title="fig : " ] we plot the density dependence of the srt in fig . [ fig : ne_int ] for both low temperature ( @xmath135 k ) and room temperature ( @xmath190 k ) cases . it is seen that for both cases the bap mechanism is far less efficient than the dp mechanism . another remarkable feature is that the srt shows a _ nonmonotonic _ photo - excitation density dependence with a peak at some density @xmath146 which resembles that in @xmath0-type materials . further calculation gives @xmath192 @xmath111 ( @xmath193 k ) for @xmath135 k case and @xmath194 @xmath111 ( @xmath195 k ) for @xmath190 k case . interestingly , two recent experiments give different photo - excitation density dependences of srt at room temperature : in ref . the srt decreases with @xmath171 where @xmath196 @xmath111 , while in ref . the srt increases with @xmath171 where the photo - excitation density is lower . these observations are consistent with our results . however , the peak has not been reported in the literature . we then discuss the effects of the electron - hole coulomb scattering and the coulomb hf term on the spin relaxation for @xmath135 k as function of photo - excitation density . in fig . [ fig : ne_int](c ) , we plot the srts obtained from the full calculation , from the calculation without the electron - hole coulomb scattering , and from the calculation without the coulomb hf term . it is seen that the coulomb hf term plays a visible role only for high densities , as the hf effective magnetic field increases with electron density.@xcite similar to the temperature dependence [ fig . [ fig : t_int](c ) ] , without the electron - hole coulomb scattering , the srt is larger for @xmath197 @xmath111 where the coulomb hf term plays a prominent role , while it is smaller for lower photo - excitation densities where the coulomb hf term is unimportant . notably , the peak of srt still exists when the coulomb hf term is removed , which implies that the degree of initial spin polarization is irrelevant for the existence of the peak . in this section , we study spin relaxation in @xmath1-type iii - v semiconductors . the main sources of spin relaxation have been recognized as the bap mechanism and the dp mechanism.@xcite we first compare the relative efficiency of the two mechanisms for various hole densities and temperatures . after that , the hole density and the photo - excitation density dependences of the srt at given temperature are also discussed . we first address the relative importance of the bap and dp mechanisms for various hole densities and temperatures in gaas . the electrons are created by photo - excitation with @xmath188 ( i.e. , we assume ideal optical orientation by circularly polarized light).@xcite in order to avoid exaggerating the dp mechanism , we use the soc parameter fitted from the experimental data in ref . , i.e. , @xmath198 ev@xmath149@xmath199 ( see appendix b ) throughout this section , which is smaller than the value from the @xmath200 calculation @xmath201 ev@xmath149@xmath199.@xcite we first concentrate on low photo - excitation density regime , where we choose @xmath202 @xmath111 . the ratio of the srt due to the bap mechanism to that due to the dp mechanism is plotted in fig . [ fig : t_p](a ) for various hole densities . it is seen that the dp mechanism dominates at high temperature , whereas the bap mechanism dominates at low temperature , which is consistent with previous investigations.@xcite an interesting feature is that the ratio first decreases rapidly , then slowly , and then again rapidly with decreasing temperature . a typical case is shown in fig . [ fig : t_p](b ) for @xmath203 @xmath111 . it is noted that the `` plateau '' is around the hole fermi temperature @xmath204 k which is given by @xmath205^{2/3}}.\ ] ] the underlying physics is that : on one hand , the pauli blocking of holes becomes important when @xmath206 , which slows down the bap spin relaxation effectively [ see eq . ( [ bapapp2 ] ) ] ; on the other hand , the increase of the screening ( mainly from holes ) with decreasing temperature weakens the electron - impurity and carrier - carrier scatterings and thus enhances the dp spin relaxation . consequently , the decrease of the ratio with decreasing temperature slows down and the `` plateau '' is formed around @xmath207 . however , after the hole system enters the degenerate regime , the screening changes little with temperature . the ratio thus decreases rapidly with decreasing temperature again [ see appendix c ] . we also plot the srt due to the bap mechanism without the pauli blocking of holes as dotted curve in fig . [ fig : t_p](b ) , which indicates that the pauli blocking of holes effectively suppresses the bap spin relaxation at low temperature ( @xmath206 ) . it is also seen from fig . [ fig : t_p](b ) that the total srt increases with decreasing temperature as both @xmath104 and @xmath191 do . ratio of the srt due to the bap mechanism to that due to the dp mechanism as function of temperature for various hole densities with @xmath202 @xmath111 and @xmath208 . ( a ) : @xmath209 @xmath111 ( curve with @xmath115 ) , @xmath116 @xmath111 ( curve with @xmath117 ) , @xmath120 @xmath111 ( curve with @xmath119 ) , and @xmath210 @xmath111 ( curve with @xmath121 ) . the hole fermi temperatures for these densities are @xmath211 , 7.3 , 34 , and 156 k , respectively . the electron fermi temperature is @xmath212 k. ( b ) : the srts due to the bap and dp mechanisms , the total srt , together with the ratio @xmath213 ( curve with @xmath119 ) versus the temperature for @xmath214 @xmath111 . the dotted curve represents the srt due to the bap mechanism without the pauli blocking of holes . note the scale of @xmath213 is on the right hand side of the frame.,title="fig:",height=188 ] -gaas . ratio of the srt due to the bap mechanism to that due to the dp mechanism as function of temperature for various hole densities with @xmath202 @xmath111 and @xmath208 . ( a ) : @xmath209 @xmath111 ( curve with @xmath115 ) , @xmath116 @xmath111 ( curve with @xmath117 ) , @xmath120 @xmath111 ( curve with @xmath119 ) , and @xmath210 @xmath111 ( curve with @xmath121 ) . the hole fermi temperatures for these densities are @xmath211 , 7.3 , 34 , and 156 k , respectively . the electron fermi temperature is @xmath212 k. ( b ) : the srts due to the bap and dp mechanisms , the total srt , together with the ratio @xmath213 ( curve with @xmath119 ) versus the temperature for @xmath214 @xmath111 . the dotted curve represents the srt due to the bap mechanism without the pauli blocking of holes . note the scale of @xmath213 is on the right hand side of the frame.,title="fig:",height=188 ] -gaas . ratio of the srt due to the bap mechanism to that due to the dp mechanism as function of temperature for various hole densities with @xmath215 and @xmath208 . ( a ) : @xmath209 @xmath111 ( curve with @xmath115 ) , @xmath116 @xmath111 ( curve with @xmath117 ) , @xmath120 @xmath111 ( curve with @xmath119 ) , and @xmath210 @xmath111 ( curve with @xmath121 ) . the hole fermi temperatures for these densities are @xmath216 , 7.7 , 36 , and 167 k , respectively . the electron fermi temperatures are @xmath217 , 13 , 61 , and 283 k , respectively . ( b ) : the srts due to the bap and dp mechanisms , the total srt , together with the ratio @xmath213 ( curve with @xmath119 ) versus the temperature for @xmath214 @xmath111 . the dotted ( dashed ) curve represents the srt due to the bap mechanism without the pauli blocking of electrons ( holes ) . note the scale of @xmath213 is on the right hand side of the frame.,title="fig:",height=188 ] -gaas . ratio of the srt due to the bap mechanism to that due to the dp mechanism as function of temperature for various hole densities with @xmath215 and @xmath208 . ( a ) : @xmath209 @xmath111 ( curve with @xmath115 ) , @xmath116 @xmath111 ( curve with @xmath117 ) , @xmath120 @xmath111 ( curve with @xmath119 ) , and @xmath210 @xmath111 ( curve with @xmath121 ) . the hole fermi temperatures for these densities are @xmath216 , 7.7 , 36 , and 167 k , respectively . the electron fermi temperatures are @xmath217 , 13 , 61 , and 283 k , respectively . ( b ) : the srts due to the bap and dp mechanisms , the total srt , together with the ratio @xmath213 ( curve with @xmath119 ) versus the temperature for @xmath214 @xmath111 . the dotted ( dashed ) curve represents the srt due to the bap mechanism without the pauli blocking of electrons ( holes ) . note the scale of @xmath213 is on the right hand side of the frame.,title="fig:",height=188 ] we then discuss the case with high photo - excitation density , where we choose @xmath215 . the ratio of the srt due to the bap mechanism to that due to the dp mechanism is plotted in fig . [ fig : t_p2](a ) for various hole densities . it is seen that , interestingly , the ratio is nonmonotonic and has a minimum roughly around the fermi temperature of electrons , @xmath218 . the bap mechanism is comparable with the dp mechanism only in the moderate temperature regime roughly around @xmath127 , whereas for higher or lower temperature it becomes unimportant . to explore the underlying physics , we plot the srts due to the bap and dp mechanisms in fig . [ fig : t_p2](b ) for @xmath219 @xmath111 . it is seen that the pauli blocking of electrons and holes largely suppresses the bap spin relaxation in the low temperature regime and hence makes @xmath191 always increase with decreasing temperature . on the other hand , @xmath104 first increases with decreasing temperature , then saturates at low temperature ( @xmath220 ) due to the fact that both the inhomogeneous broadening and the momentum scattering change little in the degenerate regime . therefore , the ratio @xmath213 first decreases then increases with decreasing temperature and shows a minimum roughly around the electron fermi temperature . this scenario holds for arbitrary excitation density , and the temperature where the ratio @xmath213 reaches its minimum increases with excitation density . finally , it is seen that the total srt saturates at low temperature as @xmath104 does . srts @xmath136 due to the bap and dp mechanisms together with the total srt versus the photo - excitation density @xmath171 . the ratio of the two is plotted as dashed curve ( note that the scale is on the right hand side of the frame ) . ( a ) : @xmath221 @xmath111 . ( b ) : @xmath222 @xmath111 . @xmath223 k.,title="fig : " ] -gaas . srts @xmath136 due to the bap and dp mechanisms together with the total srt versus the photo - excitation density @xmath171 . the ratio of the two is plotted as dashed curve ( note that the scale is on the right hand side of the frame ) . ( a ) : @xmath221 @xmath111 . ( b ) : @xmath222 @xmath111 . @xmath223 k.,title="fig : " ] we now turn to the photo - excitation density dependence of the srt . in fig . [ fig : nex_p](a ) , we plot the srt due to the dp mechanism and that due to the bap mechanism as function of photo - excitation density for @xmath224 @xmath111 with @xmath223 k. it is seen that the srt due to the dp mechanism decreases with the photo - excitation density monotonically . specifically , it first decreases slowly , then ( @xmath225 @xmath111 ) rapidly with the photo - excitation density . the scenario is as follows : in @xmath1-type semiconductors at low temperature , the dominant scattering mechanisms are the electron - hole and electron - impurity scatterings . the momentum scattering due to these two mechanism changes little with photo - excitation ( electron ) density ( @xmath226 ) for @xmath227 . in the low density regime ( @xmath228 @xmath111 , or @xmath229 k ) , where the electron system is non - degenerate , the increase of density affects the inhomogeneous broadening very little . thus the srt changes slowly with the photo - excitation density . in the high density regime ( @xmath230 @xmath111 , or @xmath231 k ) , the electron system is degenerate where the inhomogeneous broadening increases fast with density . consequently the srt decreases rapidly with the photo - excitation density . for the srt due to the bap mechanism , it decreases slowly with the photo - excitation density in the low density regime , but rapidly in the high density regime . the decrease is mainly due to the increase of the averaged electron velocity @xmath232 [ see eq . ( [ bapapp2 ] ) ] , which is determined by the temperature and is insensitive to density in the non - degenerate regime , but increases rapidly in the degenerate regime . however , the increase of the spin relaxation due to the bap mechanism is slower than that due to the dp mechanism , because the inhomogeneous broadening increases as @xmath233 while @xmath232 increases as @xmath234 . consequently , the bap mechanism becomes even less important in the high photo - excitation density regime . similar situation also happens for other hole densities and temperatures . in fig . [ fig : nex_p](b ) , we plot the case for a larger hole density @xmath219 @xmath111 . it is seen that under low photo - excitation , the bap mechanism is more important than the dp mechanism . however , the bap mechanism becomes less important than the dp mechanism in the high photo - excitation density regime . the crossover of the low photo - excitation density regime to the high photo - excitation density regime takes place around @xmath235 . this leads to the conclusion that the bap mechanism is not important at high photo - excitation density in @xmath1-type materials . it is seen from fig . [ fig : nex_p ] that the total srt decreases with photo - excitation density as both @xmath104 and @xmath191 do . this behavior is also consistent with what observed in experiments in ref . . ( a ) : srts @xmath136 due to the bap and dp mechanisms together with the total srt against hole density @xmath236 . @xmath202 @xmath111 , @xmath208 , and @xmath237 k. the dotted curve denotes a fitting of the curve with @xmath115 using @xmath238 scale . the curve with @xmath119 denotes the ratio @xmath213 ( note that the scale is on the right hand side of the frame ) . ( b ) : srts due to the dp mechanism with the dh ( curve with @xmath117 ) , tf ( curve with @xmath119 ) , and the rpa ( curve with @xmath115 ) screenings . the ratio @xmath239 is plotted as curve with @xmath121 ( note that the scale is on the right hand side of the frame).,title="fig : " ] -gaas . ( a ) : srts @xmath136 due to the bap and dp mechanisms together with the total srt against hole density @xmath236 . @xmath202 @xmath111 , @xmath208 , and @xmath237 k. the dotted curve denotes a fitting of the curve with @xmath115 using @xmath238 scale . the curve with @xmath119 denotes the ratio @xmath213 ( note that the scale is on the right hand side of the frame ) . ( b ) : srts due to the dp mechanism with the dh ( curve with @xmath117 ) , tf ( curve with @xmath119 ) , and the rpa ( curve with @xmath115 ) screenings . the ratio @xmath239 is plotted as curve with @xmath121 ( note that the scale is on the right hand side of the frame).,title="fig : " ] we also study the hole density dependence of spin relaxation due to the bap and dp mechanisms . in fig . [ fig : nh_p](a ) , we plot the srt due to the bap mechanism and that due to the dp mechanism as function of hole density for @xmath237 k and @xmath202 @xmath111 . it is seen that the srt due to the bap mechanism decreases as @xmath238 at low hole density , which is consistent with eq . ( [ bapapp1 ] ) , i.e. , for non - degenerate holes @xmath240 . at high hole density , @xmath191 decreases slower than @xmath238 due to the pauli blocking of holes . however , for the srt due to the dp mechanism , the dependence is not so obvious : the srt first increases , then decreases and again increases with the hole density . as the electron distribution , and hence the inhomogeneous broadening , does not change with the hole density , the variation of the srt due to the dp mechanism solely comes from the momentum scattering ( mainly from the electron - impurity scattering ) . to elucidate the underlying physics , we plot the srt due to the dp mechanism calculated with the rpa screening together with those calculated with the thomas - fermi ( tf ) screening@xcite [ which applies in the degenerate ( high density ) regime ] , the debye - huckle ( dh ) screening@xcite [ which applies in the non - degenerate ( low density ) regime ] in fig . [ fig : nh_p](b ) . from the figure it is seen that the first increase and the decrease is connected with the dh screening , whereas the second increase is connected with the tf screening . the underlying physics is as follows : in the low hole density regime , the screening from the holes is small and the coulomb potential , which is proportional to @xmath241 , changes slowly with the screening constant @xmath242 . hence the electron - impurity scattering increases with @xmath236 as it is proportional to @xmath243 ( as @xmath208 ) . for higher hole density ( @xmath244 @xmath111 ) , the screening constant @xmath242 becomes larger than the transfered momentum @xmath245 . [ to elucidate the relative ratio of the two , we plot the ratio of the average of the square of the transfered momentum @xmath246 to the square of the screening constant @xmath247 as curve with @xmath121 in fig . [ fig : nh_p](b ) . ] hence the electron - impurity scattering decreases with @xmath236 because it is proportional to @xmath248 as @xmath249 for the dh screening . as the hole density increases , the hole system enters into the degenerate regime , where the tf screening applies and @xmath250 . hence , the electron - impurity scattering increases with the hole density as @xmath251 . consequently , the srt first increases , then decreases and again increases with the hole density as the momentum scattering does . it should be mentioned that this behavior is different from that in the @xmath1-type ( 001 ) quantum wells where @xmath104 increases with @xmath35 monotonically@xcite as the screening from holes is much weaker in that case due to lower - dimension in phase - space and smaller hole effective mass [ in ( 001 ) gaas quantum wells , the in - plane effective mass of the heavy - hole is @xmath252 compared to @xmath253 in bulk ] . it is also noted in fig . [ fig : nh_p](a ) that the ratio @xmath254 first decreases rapidly , then slowly and again rapidly with the hole density @xmath236 . the first decrease is because that @xmath191 decreases with @xmath236 , whereas @xmath104 increases with it . in the crossover regime ( @xmath255 @xmath111 ) , where @xmath235 , the srt due to the dp mechanism varies slowly with hole density . as the srt due to the bap mechanism also varies slowly with hole density due to the pauli blocking of holes , the ratio @xmath213 changes slowly with hole density in this regime and a `` plateau '' is formed at @xmath235 . in higher hole density regime , however , @xmath104 increases with @xmath236 , whereas @xmath191 decreases with it . the ratio @xmath213 hence decreases rapidly with @xmath236 again and the bap mechanism becomes more and more important . it is seen from fig . [ fig : nh_p](a ) that the total srt first increases then decreases in the low hole density regime as @xmath104 does . consequently , the hole density dependence of the srt exhibits a peak which has never been reported . nevertheless , in the regime of higher hole density , the total srt changes slowly with @xmath236 as the bap and dp mechanisms compete with each other . ratio of the srt due to the bap mechanism to that due to the dp mechanism as function of temperature for @xmath209 @xmath111 ( curve with @xmath115 ) , @xmath256 @xmath111 ( curve with @xmath117 ) , @xmath120 @xmath111 ( curve with @xmath119 ) , and @xmath210 @xmath111 ( curve with @xmath121 ) . ( a ) : @xmath202 @xmath111 . ( b ) : @xmath215.,title="fig : " ] -gasb . ratio of the srt due to the bap mechanism to that due to the dp mechanism as function of temperature for @xmath209 @xmath111 ( curve with @xmath115 ) , @xmath256 @xmath111 ( curve with @xmath117 ) , @xmath120 @xmath111 ( curve with @xmath119 ) , and @xmath210 @xmath111 ( curve with @xmath121 ) . ( a ) : @xmath202 @xmath111 . ( b ) : @xmath215.,title="fig : " ] although the above conclusions are obtained from gaas , they also hold for other iii - v semiconductors . to demonstrate that , we also investigate the problem in gasb . gasb is a narrow band gap iii - v semiconductors of which the values of @xmath74 and @xmath257 can be found in literature.@xcite in fig . [ fig : t_pgasb ] , we plot the ratio of the srt due to the bap mechanism to that due to the dp mechanism as function of temperature for various hole densities . it is seen from the figures that the features are similar to those in gaas , whereas the ratio is much larger than that in gaas under the same condition . that is , the relative importance of the bap mechanism in gasb is smaller than that in gaas . this is because the soc in gasb is much larger than that in gaas while the longitudinal - transversal splitting @xmath74 in gasb is smaller than that in gaas . -gaas . ratio of the srt under electric field to the electric - field free one @xmath258 ( curve with @xmath115 ) and the ratio of the hot - electron temperature to the lattice temperature @xmath259 ( curve with @xmath119 ) as function of electric field for @xmath260 @xmath111 at @xmath261150 k. the mobility is also plotted as curve with @xmath117 ( note that the scale is on the right hand side of the frame ) . ] in this section , we study the effects of electric field on spin relaxation in @xmath0-type iii - v semiconductors . previous studies have demonstrated that in quantum wells a relatively high in - plane electric field can effectively manipulate the srt.@xcite the underlying physics is that the high electric field induces two effects : the drift of the electron ensemble which enhances the inhomogeneous broadening ( as electrons distribute on larger @xmath11 states where the soc is larger ) , as well as the hot - electron effect which enhances the momentum scattering . the former tends to suppress while the latter tends to enhance the srt . thus the srt has nonmonotonic electric field dependence : it first increases due to the hot - electron effect then decreases due to the enhancement of inhomogeneous broadening . in bulk semiconductors , the electric field dependence of spin lifetime has not been investigated . in this section , we present such a study . using @xmath0-type gaas as an example , we demonstrate that the electric field dependence of spin lifetime can be nonmonotonic ( first increasing then decreasing ) or monotonic ( decreasing ) depending on the lattice temperature and the densities of impurities and electrons . the underlying physics is analyzed . the study indicates that the spin lifetime can be effectively controlled by electric field . ratio of the srt under electric field to the electric - field - free one @xmath258 ( solid curves ) and the ratio of the hot - electron temperature to the lattice temperature @xmath259 ( dashed curves ) as function of electric field for ( a ) : @xmath262 k with @xmath114 @xmath111 ( curve with @xmath115 : @xmath123 ; curve with @xmath117 : @xmath263 ) and ( b ) : @xmath190 k with @xmath264 ( curve with @xmath115 : @xmath114 @xmath111 ; curve with @xmath117 : @xmath265 @xmath111).,title="fig : " ] -gaas . ratio of the srt under electric field to the electric - field - free one @xmath258 ( solid curves ) and the ratio of the hot - electron temperature to the lattice temperature @xmath259 ( dashed curves ) as function of electric field for ( a ) : @xmath262 k with @xmath114 @xmath111 ( curve with @xmath115 : @xmath123 ; curve with @xmath117 : @xmath263 ) and ( b ) : @xmath190 k with @xmath264 ( curve with @xmath115 : @xmath114 @xmath111 ; curve with @xmath117 : @xmath265 @xmath111).,title="fig : " ] in fig . [ fig : e1 ] , we plot the ratio of the srt under electric field to the electric - field - free one as function of electric field for @xmath265 @xmath111 with @xmath178 at @xmath266 k and @xmath123 . the electric field is chosen to be along the @xmath267-axis and the initial spin polarization is along the @xmath95-axis . due to the cubic form of the soc , the average of the spin - orbit field is negligible even in the presence of finite center - of - mass drift velocity , which is different from the case in quantum wells where the linear @xmath268-term gives a large effective magnetic field in the presence of electric field thanks to the strong well confinement.@xcite it is seen that the ratio first increases a little and than decreases rapidly with the electric field . at @xmath269 kv / cm , the ratio drops to 0.6 which demonstrates that the electric field can effectively affect the srt . to understand these effects , we also plot the hot - electron temperature in the figure . it is noted that the electron temperature increases first slowly then ( @xmath270 kv / cm ) rapidly with the electric field , indicating clearly the hot - electron effect . for the 2des , where the soc is dominated by the linear term , the hot - electron effect mainly leads to the enhancement of scattering , whereas the enhancement of inhomogeneous broadening due to the hot - electron effect is marginal.@xcite differently , the hot - electron effect also has important effect on inhomogeneous broadening in bulk system , as the soc is cubic @xmath268-dependent . as both the drift effect and the hot - electron effect increase the inhomogeneous broadening , the enhancement of the inhomogeneous broadening is faster than the increase of momentum scattering . consequently , the srt is easier to decrease with the electric field , which is different from the case of 2des . we speculate that the electric field dependence of the srt in wurtzite semiconductors and strained iii - v semiconductors resembles that in the gaas quantum wells , when the soc is dominated by the linear @xmath268 term.@xcite the mobility of the electron system is also plotted in the figure . the variation of the mobility indicates the nonlinear effects of the electric field in the kinetics of electron system . for lower and higher temperature cases , we plot the results in fig . [ fig : e2 ] . the electron density is @xmath114 @xmath111 in fig . [ fig : e2](a ) . it is seen that the ratio for the case with low impurity density @xmath271 decreases faster than that for the case with @xmath123 . this is because both the hot - electron effect and the drift effect are more pronounced in cleaner system,@xcite which thus leads to a faster decrease of the srt due to the enhancement of inhomogeneous broadening . in the background of hot - electron spin injection under high bias , our results indicate that the manipulation of the srt by the electric field is more pronounced for samples with high mobility . for high temperature case ( @xmath190 k ) , we plot the srt and hot - electron temperature as function of electric field for two different electron densities @xmath114 @xmath111 and @xmath265 @xmath111 with @xmath123 in fig . [ fig : e2](b ) . it is seen that both the srt and the electron temperature differ marginally for the two cases even though their impurity and electron densities differ by 20 times . this is because that at 300 k the electron lo - phonon scattering is more important than the electron - impurity scattering . thus both the drift of the electron system and the hot - electron effect is mainly determined by the electron lo - phonon scattering , and the variation of the srt with electric field is insensitive to impurity density . for electron density , as the electron system is in the non - degenerate regime for both cases , the electron density dependence is hence also weak . in conclusion , we have applied an efficient scheme , the fully microscopic ksbes , to study the spin dynamics in bulk iii - v semiconductors , with all scatterings explicitly included . this approach takes full account of the spin relaxation due to the dp , ey , and bap mechanisms in a fully microscopic fashion , and enables us to find important results missing in the previous simplified approaches in the literature . from the ksbe approach , we study the electron spin relaxation in @xmath0-type , intrinsic , and @xmath1-type iii - v semiconductors . we also investigate the effects of electric field on spin relaxation in @xmath0-type iii - v semiconductors . for @xmath0-type iii - v semiconductors , differing from the previous conclusions , we find that the spin relaxation due to the ey mechanism is less important than that due to the dp mechanism even in narrow band - gap semiconductors , such as inas and insb . we then focus on the spin relaxation due to the dp mechanism . we find that the density dependence of the srt is nonmonotonic and we predict a peak which appears in the metallic regime . this behavior is due to the different density dependences of the inhomogeneous broadening and the momentum scattering in the degenerate and non - degenerate regimes . in the non - degenerate regime , as the electron distribution is close to the boltzmann distribution , the inhomogeneous broadening changes little with the density but the electron - electron and electron - impurity scatterings increase with the electron density . as a result , the srt increases with the density . in the degenerate regime , the inhomogeneous broadening increases with electron density , whereas the momentum scatterings decrease with it . consequently , the srt decreases with the electron density in the degenerate regime . the peak of the srt is hence formed in the crossover regime , where the corresponding fermi temperature is close to the lattice temperature , @xmath235 . our results show that the electron - electron scattering plays an important role in the spin relaxation . we also study the density dependence for the case with strain - induced soc , where the density dependence of the inhomogeneous broadening is different due to the linear @xmath268-dependence of the strain - induced soc . however , the srt still has a peak but at a larger density compared to the strain - free case . we further study the temperature dependence of the srt . we find that the srt decreases monotonically with the temperature which is consistent with experimental findings . after we artificially lower the impurity density , we find a peak in the srt which is due to the different temperature dependence of the electron - electron scattering in the degenerate and non - degenerate regimes . this is consistent with the results in 2des where the peak in the srt due to the electron - electron scattering appears only when the impurity density is low ( e.g. , @xmath272).@xcite we also study the initial spin polarization dependence of the srt where the effect of the coulomb hf term is discussed . we find that the dependence is quite weak in bulk system compared to that in the 2des , which is again due to the large impurity density @xmath167 in bulk system . for intrinsic iii - v semiconductors , we first compare the bap mechanism and the dp mechanism . we find that the bap mechanism is far less efficient than the dp mechanism . we further compare our results from the fully microscopic ksbe approach with those from the approach widely used in the literature . we find that the previous approach deviates in the low temperature regime due to the pretermission of the pauli blocking . also , the previous approach ignores the long - range electron - hole exchange scattering which is shown to be dominant in gaas . we find that the electron - hole coulomb scattering plays an important role in spin relaxation . the coulomb hf term is found to have important effects on spin relaxation at low temperature and high photo - excitation density , as the impurity density is very low in intrinsic semiconductors ( we choose @xmath170 ) . due to the same reason , the peak in the temperature dependence of the srt due to the electron - electron scattering also appears at small spin polarization . we further discuss the photo - excitation density dependence of the srt . we find that the srt first increases then decreases with the density which resembles the case in @xmath0-type samples where the underlying physics is also similar . for @xmath1-type iii - v semiconductors , we first examine the relative importance of the bap mechanism . we find that the bap mechanism dominates the spin relaxation in the low temperature regime only when the photo - excitation density is low . however , when the photo - excitation density is high , the bap mechanism can be comparable with the dp mechanism only in the moderate temperature regime roughly around the fermi temperature of electrons , whereas for higher or lower temperature it is unimportant . the photo - excitation density dependences of srts due to the bap and dp mechanisms are also discussed . we find that the relative importance of the bap mechanism decreases with photo - excitation density and eventually becomes negligible at sufficiently high photo - excitation density . for hole density dependence at small photo - excitation density , we find that the spin relaxation due to the bap mechanism increases with hole density linearly in low hole density regime but the increase becomes slower in high hole density regime where the pauli blocking of holes becomes important . interestingly , the srt due to the dp mechanism first increases , then decreases and again increases with hole density . the underlying physics is that the momentum scattering ( mainly from the electron - impurity scattering ) first increases with hole ( impurity ) ( @xmath208 ) density , then decreases with hole density due to the increase of the screening . however , at high hole density when the hole system is degenerate , the screening increases slower with the hole density and the momentum scattering again increases with the hole ( impurity ) density . on the other hand , the inhomogeneous broadening does not change with hole density as the electron density is solely determined by the photo - excitation density . consequently , the srt due to the dp mechanism first increases , then decreases and again increases with the hole density . this behavior makes the ratio @xmath254 first decreases rapidly , then slowly and again rapidly with the hole density . the bap mechanism is more important than the dp one for high hole density . the relative importance of the bap mechanism in gasb is found to be much less than that in gaas due to both the weaker electron - hole exchange interaction and the larger soc in gasb . finally , we study the effect of electric field on the spin relaxation in @xmath0-type gaas . we find that the srt can be largely affected by the electric field . the underlying physics is that the electric field induces two effects : the center - of - mass drift which enhances the inhomogeneous broadening and the hot - electron effect which increases both the momentum scattering and the inhomogeneous broadening . the electric field dependence of srt thus can be nonmonotonic : it first increases due to the increase of scattering then decreases due to the enhancement of inhomogeneous broadening . however , we find that differing from the 2des , the srt is easier to decrease with the electric field . this is because that the inhomogeneous broadening increases faster when the soc is cubic compared to the 2des where the soc is dominated by linear @xmath268 term . we expect that the electric field dependence of the srt resembles the 2des for wurtzite semiconductors or strained semiconductors , where the soc can be dominated by the linear @xmath268 term . we also find that the effect of the electric field becomes more significant for low impurity density samples at low temperature as both the drift effect and the hot - electron effect are more pronounced . however , at room temperature , the effect of the electric field is insensitive to the impurity density as the electron lo - phonon scattering is more important than the electron - impurity scattering . the electron density dependence is also weak as long as the system is in the non - degenerate regime . note added : after we submitted this manuscript , the peak we predicted in the doping density dependence of the srt of @xmath0-gaas in the metallic regime has been realized experimentally in a subsequent paper.@xcite also an independent theoretical calculation from the ksbe approach well reproduced the peak in the same paper.@xcite this work was supported by the natural science foundation of china under grant no . 10725417 , the national basic research program of china under grant no . 2006cb922005 , and the knowledge innovation project of chinese academy of sciences . one of the authors ( j.h.j . ) would like to thank m. q. weng and k. shen for helpful discussions . our numerical scheme is based on the discretization of the @xmath268-space similar to that in ref . . here we extend it to the three dimensional case . our technique greatly reduces the calculation complexity and makes the quantitatively accurate calculation possible . the @xmath268-space is divided into @xmath273 control regions where the @xmath268-grid points are chosen to be @xmath274 . to facilitate the evaluation of the @xmath275-functions in the scattering terms , we set @xmath276 , where the energy span in each control region is @xmath277 with @xmath278 being an integer number and @xmath279 denoting the lo - phonon frequency . the electron - impurity and electron - phonon scatterings are then solved easily since the @xmath275-functions can be integrated out directly . for the electron - electron coulomb , electron - hole coulomb and electron - hole exchange scatterings , the situation is much more complex . the electron - electron coulomb scattering term [ eq . ( [ scat_ee ] ) ] can be rewritten as , @xmath280 + { \rm h.c.},\end{aligned}\ ] ] where @xmath281 substituting @xmath282 and @xmath283 , one has @xmath284 now the @xmath275 function can be simplified as @xmath285 where @xmath286 is the angle between @xmath287 and @xmath288 , and @xmath289 . to evaluate the @xmath275 function , it is helpful to rotate to the new coordinate system with @xmath288 being along the @xmath95-axis . in this coordinate system , @xmath290 and the @xmath275 function can be evaluated readily . the result is @xmath291 with @xmath292 $ ] . note that the integration over @xmath293 is restrained by the condition @xmath294 ^ 2/2m_c$ ] according to @xmath295 . now the electron - electron coulomb scattering is easily integrated out . note that the calculation of @xmath296 can be done before the calculation of the electron - electron coulomb scattering terms , which thus reduces the whole calculation complexity from @xmath297 to @xmath298 . this method , first developed by cheng in two - dimensional system,@xcite greatly reduces the calculation complexity . for the electron - hole coulomb scattering , the idea is similar , but the technique is more complex . denoting @xmath299 , the @xmath275-function in eq . ( [ scat_eh ] ) can be written as @xmath300 the @xmath275-function is then integrated out similarly , @xmath301 the integration over @xmath302 is restrained by the condition @xmath303 , where @xmath304 the restriction can be simplified as @xmath305 where @xmath306 and @xmath307 . the above inequality can be worked out readily for given @xmath308 , and the restriction condition for @xmath309 is then obtained . the electron - hole exchange scattering is solved similarly by substituting the matrix element @xmath310 with @xmath311 . finally , the drift term is solved with similar method of that in ref . . the differential equations are solved by the fourth - order runge - kutta method . the computation is carried out in a parallel manner by using openmp . for a typical calculation with the partition of @xmath312 grid points in the @xmath268-space , it takes about ten hours to evolute 50 ps on a quad - core amd phenom 9750 . -gaas . srts @xmath136 from the experiment in ref . ( @xmath115 ) and from the calculation via the ksbe approach with only the dp mechanism ( solid curve ) as well as that with only the ey mechanism ( dashed curve ) . @xmath114 @xmath111 , @xmath313 and @xmath202 @xmath111 . @xmath314 ev@xmath149@xmath315 . ( b ) @xmath1-gaas . srts @xmath136 from the experiment in ref . ( @xmath115 ) and from the calculation via the ksbe approach ( solid curve ) . @xmath316 @xmath111 , @xmath208 and @xmath317 k. @xmath198 ev@xmath149@xmath315 . ( c ) @xmath1-gaas . srts @xmath136 from the experiment in ref . ( @xmath115 ) and from the calculation via the ksbe approach ( solid curve ) . @xmath318 @xmath111 , @xmath208 and @xmath202 @xmath111 . @xmath319 ev@xmath149@xmath315.,title="fig : " ] -gaas . srts @xmath136 from the experiment in ref . ( @xmath115 ) and from the calculation via the ksbe approach with only the dp mechanism ( solid curve ) as well as that with only the ey mechanism ( dashed curve ) . @xmath114 @xmath111 , @xmath313 and @xmath202 @xmath111 . @xmath314 ev@xmath149@xmath315 . ( b ) @xmath1-gaas . srts @xmath136 from the experiment in ref . ( @xmath115 ) and from the calculation via the ksbe approach ( solid curve ) . @xmath316 @xmath111 , @xmath208 and @xmath317 k. @xmath198 ev@xmath149@xmath315 . ( c ) @xmath1-gaas . srts @xmath136 from the experiment in ref . ( @xmath115 ) and from the calculation via the ksbe approach ( solid curve ) . @xmath318 @xmath111 , @xmath208 and @xmath202 @xmath111 . @xmath319 ev@xmath149@xmath315.,title="fig : " ] -gaas . srts @xmath136 from the experiment in ref . ( @xmath115 ) and from the calculation via the ksbe approach with only the dp mechanism ( solid curve ) as well as that with only the ey mechanism ( dashed curve ) . @xmath114 @xmath111 , @xmath313 and @xmath202 @xmath111 . @xmath314 ev@xmath149@xmath315 . ( b ) @xmath1-gaas . srts @xmath136 from the experiment in ref . ( @xmath115 ) and from the calculation via the ksbe approach ( solid curve ) . @xmath316 @xmath111 , @xmath208 and @xmath317 k. @xmath198 ev@xmath149@xmath315 . ( c ) @xmath1-gaas . srts @xmath136 from the experiment in ref . ( @xmath115 ) and from the calculation via the ksbe approach ( solid curve ) . @xmath318 @xmath111 , @xmath208 and @xmath202 @xmath111 . @xmath319 ev@xmath149@xmath315.,title="fig:",height=196 ] we compare the calculation from the fully microscopic ksbe approach with the experimental results in refs . and . these experiments were carried out in @xmath0- and @xmath1-type gaas respectively . in fig . [ fig : exp](a ) , we plot the srt as function of temperature calculated from the ksbes , together with the experimental data in ref . for @xmath0-gaas with @xmath114 @xmath111 , @xmath313 and @xmath320 it is seen that the calculation agrees well with the experimental results in @xmath0-gaas for @xmath321 k. the deviation in the lower temperature regime is due to the rising of the localization of electrons.@xcite the srt due to the ey mechanism is also plotted in the figure , which is much larger than the experimental data , indicating the irrelevance of the ey mechanism . the calculation gives a fit of the soc parameter as @xmath198 ev@xmath149@xmath315 which is different from the value @xmath201 ev@xmath149@xmath315 calculated from the tight - binding or @xmath200 parametric theories.@xcite however , the value is still in the reasonable range of @xmath100 [ for lists of @xmath100 calculated and measured via various methods , see ref . ] . our fitting , though with only one fitting parameter @xmath100 , agrees well with the experimental data in almost the whole temperature range , which is much better than the fittings with the same experimental data in refs . - . in fig . [ fig : exp](b ) , we plot the photo - excitation density dependence of the srts from our calculation and from the experiment in ref . for @xmath1-gaas with @xmath316 @xmath111 , @xmath208 and @xmath317 k. it is seen that with the same @xmath100 , the calculation again agrees well with the experimental data . the srt due to the bap mechanism is about 20 times larger , which is consistent with our conclusion that the bap mechanism is unimportant at high photo - excitation density . in fig . [ fig : exp](c ) , we plot the temperature dependence of the srts from our calculation and from the experiment in ref . for @xmath1-gaas with @xmath318 @xmath111 , @xmath208 and @xmath322 @xmath111 . the best fitting gives a slightly larger @xmath323 ev@xmath149@xmath315 . the srt due to the bap mechanism is @xmath16100 times larger than that due to the dp mechanism . this is consistent with our conclusion that the bap mechanism is unimportant at high temperature for low doping density . throughout the paper , the srt in gaas due to the dp mechanism is calculated with @xmath201 ev@xmath149@xmath315 unless in sec . v where we use @xmath198 ev@xmath149@xmath315 in order to avoid possible exaggeration of the dp mechanism by using a `` larger '' soc parameter . however , the srt due to the dp mechanism is proportional to @xmath324 in motional narrowing regime , thus the results presented in the paper can be easily converted to the results for another @xmath100 . the ratio for the srts in the two cases is 8.5 . -gaas . the srts due to the dp mechanism with the tf ( curve with @xmath117 ) , dh ( curve with @xmath119 ) , and the rpa ( curve with @xmath115 ) screenings for excitation density ( a ) @xmath202 @xmath111 and ( b ) @xmath325 @xmath111 . hole density @xmath214 @xmath111 and @xmath208.,title="fig : " ] -gaas . the srts due to the dp mechanism with the tf ( curve with @xmath117 ) , dh ( curve with @xmath119 ) , and the rpa ( curve with @xmath115 ) screenings for excitation density ( a ) @xmath202 @xmath111 and ( b ) @xmath325 @xmath111 . hole density @xmath214 @xmath111 and @xmath208.,title="fig : " ] we study the effect of screening on dp spin relaxation in @xmath1-type gaas . we focus on the temperature dependence of the srt which corresponds to the discussions on the results in figs . [ fig : t_p ] and [ fig : t_p2 ] . as in sec . v , we discuss both the low and high excitation density cases . we first study the low excitation case . to elucidate the role of screening on the dp spin relaxation , we plot the srt due to the dp mechanism with the tf screening@xcite [ which applies in the degenerate ( low temperature ) regime ] , the dh screening@xcite [ which applies in the non - degenerate ( high temperature ) regime ] , and the rpa screening@xcite ( which applies in the whole temperature regime ) in fig . [ fig : pscreen](a ) . the crossover from the dh screening to the tf one is clearly seen in the figure . also , compared with the case of temperature - independent screening ( i.e. , the tf screening ) , one can see that the increase of the srt with decreasing temperature is indeed slowed down around @xmath207 ( where the screening effect is prominent ) due to the increase of screening . for high excitation case , we find a peak in the temperature dependence of the srt due to the dp mechanism . this peak is roughly around @xmath127 . to explore the underlying physics , we also plot the srt due to the dp mechanism with the tf and dh screenings . it is seen that with the tf screening the peak disappears , while with the dh screening the peak remains . this elucidates that the appearance of the peak is due to the increase of screening at low temperature . what is different for the high excitation case is that the electron fermi temperature is higher than the hole fermi temperature . thus at low temperature when the electron system is in degenerate regime ( hence inhomogeneous broadening changes slowly ) and the hole system has not yet entered into the degenerate regime ( hence the increase of screening with decreasing temperature is prominent ) , the srt due to the dp mechanism @xmath104 decreases with decreasing temperature . on the other hand , at high temperature , @xmath104 increases with decreasing temperature . consequently , the peak is formed . it should be noted that this peak is different from the peak found in intrinsic semiconductors in sec . iv where the peak is due to the nonmonotonic temperature dependence of the electron - electron coulomb scattering , whereas the main scattering mechanism here is the electron - impurity scattering . 0 f. meier and b. p. zakharchenya , _ optical orientation _ ( north - holland , amsterdam , 1984 ) . s. a. wolf , d. d. awschalom , r. a. buhrman , j. m. daughton , s. von molnr , m. l. roukes , a. y. chtchelkanova , and d. m. treger , science 294 * , 1488 ( 2001 ) . , ed . by d. d. awschalom , d. loss , and n. samarth ( springer - verlag , berlin , 2002 ) ; i. uti , j. fabian , and s. das sarma , rev . * 76 * , 323 ( 2004 ) ; j. fabian , a. matos - abiague , c. ertler , p. stano , and i. uti , acta phys . slov . * 57 * , 565 ( 2007 ) ; _ spin physics in semiconductors _ , ed . by m. i. dyakonov ( springer , berlin , 2008 ) , and references therein . j. m. kikkawa and d. d. awschalom , phys . rev . lett . * 80 * , 4313 ( 1998 ) . long spin relaxation time was also observed in r. i. dzhioev , v. l. korenev , i. a. merkulov , b. p. zakharchenya , d. gammon , al . l. efros , and d. s. katzer , phys . * 88 * , 256801 ( 2002 ) ; j. s. colton , t. a. kennedy , a. s. bracker , d. gammon , and j. b. miller , phys . b * 67 * , 165315 ( 2003 ) ; m. oestreich , m. rmer , r. j. haug , and d. hgele , phys . * 95 * , 216603 ( 2005 ) . j. m. kikkawa and d. d. awschalom , nature * 397 * , 139 ( 1999 ) . j. m. kikkawa and d. d. awschalom , science * 287 * , 473 ( 2000 ) . m. oestreich , m. bender , j. hbner , d. hgele , w. w. rhle , t. hartmann , p. j. klar , w. heimbrodt , m. lampalzer , k. volz , and w. stolz , semicond . * 17 * , 285 ( 2002 ) . i. malajovich , j. m. kikkawa , d. d. awschalom , j. j. berry , and n. samarth , phys . rev * 84 * , 1015 ( 2000 ) . j. a. gupta , r. knobel , n. samarth , and d. d. awschalom , science * 292 * , 2458 ( 2001 ) . y. kato , r. c. myers , d. c. driscoll , a. c. gossard , j. levy , and d. d. awschalom , science * 299 * , 1201 ( 2003 ) . y. kato , r. c. myers , a. c. gossard , and d. d. awschalom , nature * 427 * , 50 ( 2004 ) . l. meier , g. salis , i. shorubalko , e. gini , s. schn , and k. ensslin , nat . phys . * 3 * , 650 ( 2007 ) . y. yafet , phys . rev . * 85 * , 478 ( 1952 ) ; r. j. elliot , phys . rev . * 96 * , 266 ( 1954 ) . m. w. wu and c. z. ning , eur . j. b * 18 * , 373 ( 2000 ) ; m. w. wu and h. metiu , phys . rev . b * 61 * , 2945 ( 2000 ) ; m. w. wu , j. phys jpn . * 70 * , 2195 ( 2001 ) . m. i. dyakonov and v. i. perel , zh . . fiz . * 60 * , 1954 ( 1971 ) [ sov . jetp * 33 * , 1053 ( 1971 ) ] ; fiz tela ( leningrad ) * 13 * , 3581 ( 1971 ) [ sov . solid state * 13 * , 3023 ( 1972 ) ] . g. l. bir , a. g. aronov , and g. e. pikus , zh . 69 * , 1382 ( 1975 ) [ sov . jetp * 42 * , 705 ( 1976 ) ] . r. i. dzhioev , k. v. kavokin , v. l. korenev , m. v. lazarev , b. y. meltser , m. n. stepanova , b. p. zakharchenya , d. gammon , and d. s. katzer , phys . b * 66 * , 245204 ( 2002 ) . z. chen , s. g. carter , r. bratschitsch , p. dawson , and s. t. cundiff , nat . phys . * 3 * , 265 ( 2007 ) . y. v. pershin and v. privman , nano lett . * 3 * , 695 ( 2003 ) . for localized electrons , besides the hyperfine interaction , the antisymmetric dzyaloshinskii - moriya interaction also leads to spin relaxation . this mechanism was studied recently with renewed interest [ see , k. v. kavokin , phys . b * 64 * 075305 ( 2001 ) ; phys . ( a ) * 190 * , 221 ( 2002 ) ; arxiv:0806.4753 ; l. p. gorkov and p. l. krotkov , phys . b * 67 * , 033203 ( 2003 ) ; w. o. putikka and r. joynt , phys . b * 70 * , 113201 ( 2004 ) ; p. i. tamborenea , d. weinmann , and r. a. jalabert , phys . rev . b * 76 * , 085209 ( 2007 ) ] . other spin relaxation mechanisms for localized electrons are proposed in a recent work [ putikka and joynt , phys . rev . b * 70 * , 113201 ( 2004 ) ] , where both the localized and itinerant electrons are considered with the spin dynamics treated in a phenomenological way . p. h. song and k. w. kim , phys . b * 66 * , 035207 ( 2002 ) . m. q. weng and m. w. wu , phys . b * 68 * , 075312 ( 2003 ) . m. q. weng , m. w. wu , and l. jiang , phys . b * 69 * , 245320 ( 2004 ) . m. q. weng and m. w. wu , phys . rev . b * 70 * , 195318 ( 2004 ) . c. l , j. l. cheng , and m. w. wu , phys . b * 73 * , 125314 ( 2006 ) . j. zhou , j. l. cheng , and m. w. wu , phys . b * 75 * , 045305 ( 2007 ) . j. zhou and m. w. wu , phys . rev . b * 77 * , 075318 ( 2008 ) . p. zhang , j. zhou , and m. w. wu , phys . b * 77 * , 235323 ( 2008 ) . j. h. jiang , m. w. wu , and y. zhou , phys . b * 78 * , 125309 ( 2008 ) . m. m. glazov and e. l. ivchenko , pisma zh . fiz . * 75 * , 476 ( 2002 ) [ jetp lett . * 75 * , 403 ( 2002 ) ] ; zh . . fiz . * 126 * , 1465 ( 2004 ) [ jetp * 99 * , 1279 ( 2004 ) ] . w. j. leyland , g. h. john , r. t. harley , m. m. glazov , e. l. ivchenko , d. a. ritchie , i. farrer , a. j. shields , and m. henini , phys . b * 75 * , 165309 ( 2007 ) . x. z. ruan , h. h. luo , y. ji , z. y. xu , and v. umansky , phys . b * 77 * , 193307 ( 2008 ) . l. h. teng , p. zhang , t. s. lai , and m. w. wu , europhys . lett . * 84 * , 27006 ( 2008 ) . for brief review , see m. w. wu , m. q. weng , and j. l. cheng , in _ physics , chemistry and application of nanostructures : reviews and short notes to nanomeeting 2007 _ , eds . v. e. borisenko , v. s. gurin , and s. v. gaponenko ( world scientific , singapore , 2007 ) p. 14 , and references therein . d. stich , j. zhou , t. korn , r. schulz , d. schuh , w. wegscheider , m. w. wu , and c. schller , phys . rev . lett . * 98 * , 176401 ( 2007 ) ; phys . b * 76 * , 205301 ( 2007 ) . t. korn , d. stich , r. schulz , d. schuh , w. wegscheider , and c. schller , arxiv:0811.0720 . c. yang , x. cui , s .- q . shen , z. xu , and w. ge , arxiv:0902.0484 . g. fishman and g. lampel , phys . b * 16 * , 820 ( 1977 ) . k. zerrouati , f. fabre , g. bacquet , j. bandet , j. frandon , g. lampel , and d. paget , phys . b * 37 * , 1334 ( 1988 ) . p. zhang and m. w. wu , phys . b * 76 * , 193312 ( 2007 ) . the magnetic field dependence of the spin relaxation has been studied in previous works [ e. l. ivchenko , sov . solid state * 15 * , 1048 ( 1973 ) ; f. x. bronold , i. martin , a. saxena , and d. l. smith , phys . b * 66 * , 233206 ( 2002 ) ] . moreover , the magnetic field could enable the spin relaxation due to the inhomogeneous broadened @xmath326-factor , i.e. , the energy dependent of @xmath326-factor provides another inhomogeneous broadening besides the @xmath268-dependent soc [ see , m. w. wu and c. z. ning , eur . j. b * 18 * , 373 ( 2000 ) ; f. x. bronold , i. martin , a. saxena , and d. l. smith , phys . b * 66 * , 233206 ( 2002 ) ] . also note that a recent experimental result reveals anomalous magneitc field dependence of the spin relaxation time in the insulator phase [ j. s. colton , m. e. heeb , p. schroeder , a. stokes , l. r. wienkes , and a. s. bracker , phys . b * 75 * , 205201 ( 2007 ) ] . l. jiang and m. w. wu , phys . b * 72 * , 033311 ( 2005 ) . d. stich , j. h. jiang , t. korn , r. schulz , d. schuh , w. wegscheider , m. w. wu , and c. schller , phys . b * 76 * , 073309 ( 2007 ) . a. w. holleitner , v. sih , r. c. myers , a. c. gossard , and d. d. awschalom , new j. phys . * 9 * , 342 ( 2007 ) . f. zhang , h. z. zheng , y. ji , j. liu , and g. r. li , europhys . lett . * 83 * , 47006 ( 2008 ) . f. zhang , h. z. zheng , y. ji , j. liu , and g. r. li , europhys . lett . * 83 * , 47007 ( 2008 ) . m. w. wu and c. z. ning , phys . ( b ) * 222 * , 523 ( 2000 ) . a. g. aronov , g. e. pikus , and a. n. titkov , zh . . fiz . * 84 * , 1170 ( 1983 ) [ sov . jetp * 57 * , 680 ( 1983 ) ] . i. malajovich , j. j. berry , n. samarth , and d. d. awschalom , nature * 411 * , 770 ( 2000 ) . f. x. bronold , a. saxena , and d. l. smith , phys . b * 70 * , 245210 ( 2004 ) . h. haug and a. p. jauho , _ quantum kinetics in transport and optics of semiconductors _ ( springer , berlin , 1996 ) . g. dresselhaus , phys . rev . * 100 * , 580 ( 1955 ) . d. j. hilton and c. l. tang , phys . rev . * 89 * , 146601 ( 2002 ) . m. krau , m. aeschlimann , and h. c. schneider , phys . 100 * , 256601 ( 2008 ) . we assume that the potential fluctuation mainly comes from the ionized impurities , and neglect the scattering by dislocation . the effect of the electron - dislocation scattering on spin relaxation has been studied in [ d. jena , phys . b * 70 * , 245203 ( 2004 ) ] . g. d. mahan , _ many - particle physics _ ( plenum press , new york , 1990 ) . j. schliemann , phys . rev . b * 74 * , 045214 ( 2006 ) . m. z. maialle , e. a. de andrada e silva , and l. j. sham , phys . b * 47 * , 15776 ( 1993 ) . , landolt - brnstein , new series , vol . 17a , ed . by o. madelung ( springer - verlag , berlin , 1987 ) . w. ekardt , k. lsch , and d. bimberg , phys . b * 20 * , 3303 ( 1979 ) . j. y. fu , m. q. weng , and m. w. wu , physica e * 40 * , 2890 ( 2008 ) . jancu , r. scholz , e. a. de andrada e silva , and g. c. la rocca , phys . b * 72 * , 193201 ( 2005 ) . u. rssler and h .- trebin , phys . b * 23 * , 1961 ( 1981 ) . lists of the soc parameters @xmath100 in gaas calculated and measured via various methods are given in the supplementary information of [ j. j. krich and b. i. halperin , phys . lett . * 98 * , 226802 ( 2007 ) ] and in [ a. n. chantis , m. van schilfgaarde , and t. kotani , phys . lett . * 96 * , 086405 ( 2006 ) ] . our fitting value of @xmath100 is in the reasonable range in these lists . the ey spin relaxation due to the electron - electron coulomb scattering was studied in [ p. boguslawski , solid state commun . * 33 * , 389 ( 1980 ) ; p. i. tamborenea , m. a. kuroda , and f. l. bottesi , phys . b * 68 * , 245205 ( 2003 ) ] . the latter also included the electron - impurity scattering , showing that the ey mechanism is ineffective in bulk gaas which is also consistent with our results . a. amo , l. via , p. lugli , c. tejedor , a. i. toropov , and k. s. zhuravlev , phys . b * 75 * , 085202 ( 2007 ) . m. a. brand , a. malinowski , o. z. karimov , p. a. marsden , r. t. harley , a. j. shields , d. sanvitto , d. a. ritchie , and m. y. simmons , phys lett . * 89 * , 236601 ( 2002 ) . w. j. h. leyland , r. t. harley , m. henini , a. j. shields , i. farrer , and d. a. ritchie , phys . b * 76 * , 195305 ( 2007 ) . w. j. h. leyland , r. t. harley , m. henini , a. j. shields , i. farrer , and d. a. ritchie , phys . b * 77 * , 205321 ( 2008 ) . s. krishnamurthya , m. van schilfgaarde , and n. newman , appl . lett . * 83 * , 1761 ( 2003 ) . w. h. lau , j. t. olesberg , and m. e. flatt , phys . b * 64 * , 161301 ( 2001 ) . z. g. yu , s. krishnamurthy , m. van schilfgaarde , and n. newman , phys . rev . b * 71 * , 245312 ( 2005 ) . a. dyson and b. k. ridley , phys . rev . b * 69 * , 125211 ( 2004 ) . m. z. maialle , phys . b * 54 * , 1967 ( 1996 ) . r. t. harley , in _ spin physics in semiconductors _ , ed . by m. i. dyakonov ( springer , berlin , 2008 ) . j. n. chazalviel , phys . b * 11 * , 1555 ( 1975 ) . the spin - orbit parameter is another quantity which may affect the ratio . to check the parameter we used , we calculate the srt corresponding to that measured in a recent experiment [ p. murzyn , c. r. pidgeon , p. j. phillips , j - p . wells , n. t. gordon , t. ashley , j. h. jefferson , t. m. burke , j. giess , m. merrick , b. n. murdin , and c. d. maxey , phys . b * 67 * , 235202 ( 2003 ) ] , where @xmath327 @xmath111 and @xmath328 @xmath111 at 150 k. we get a srt of 15.2 ps which is close to the measured value of 16 ps . the spin - orbit parameter is also checked by comparing with the experimental results in ref . . a best fitting gives @xmath329 ev@xmath149@xmath330 in inas which is 2.3 times as large as that in ref . . this only enhances the dp spin relaxation by 5.5 times and strengthens our conclusion that the dp mechanism is more efficient ( @xmath331 times ) than the ey mechanism in inas . it is noted that the fitted @xmath329 ev@xmath149@xmath330 is comparable with the value @xmath332 ev@xmath149@xmath330 calculated in [ t. f. boggess , j. t. olesberg , c. yu , m. e. flatt , and w. h. lau , appl . phys . lett . * 77 * , 1333 ( 2000 ) ] . the spin - orbit parameter @xmath100 in gaas fitted from experiments is 2.9 times smaller than that in table i. even for such a @xmath100 , the spin relaxation due to the dp mechanism is still much more efficient ( @xmath333 times ) than that due to the ey mechanism [ see also appendix b ] . k. l. litvinenko , l. nikzad , j. allam , b. n. murdin , c. r. pidgeon , j. j. harris , and l. f. cohen , j. supercon . * 20 * , 461 ( 2007 ) . g. f. giulianni and g. vignale , _ quantum theory of the electron liquid _ ( cambridge university press , cambridge , 2005 ) . b. n. murdin , k. litvinenko , j. allam , c. r. pidgeon , m. bird , k. morrison , t. zhang , s. k. clowes , w. r. branford , j. harris , and l. f. cohen , phys . rev . b * 72 * , 085346 ( 2005 ) . k. l. litvinenko , b. n. murdin , j. allam , c. r. pidgeon , t. zhang , j. j. harris , l. f. cohen , d. a. eustace , and d. w. mccomb , phys . b * 74 * , 075331 ( 2006 ) . p. e. hohage , g. bacher , d. reuter , and a. d. wieck , appl . . lett . * 89 * , 231101 ( 2006 ) . b. beschoten , e. johnston - halperin , d. k. young , m. poggio , j. e. grimaldi , s. keller , s. p. denbaars , u. k. mishra , e. l. hu , and d. d. awschalom , phys . b * 363 * , 121202 ( 2001 ) . h. saarikoski and g. e. w. bauer , arxiv:0810.3386 . as demonstrated in sec . v , the bap mechanism is more important in gaas than in gasb , thus in gasb the bap mechanism is also negligible . also a recent experimental investigation indicates that the bap mechanism is unimportant in intrinsic insb [ b. n. murdin , k. litvinenko , d. g. clarke , c. r. pidgeon , p. murzyn , p. j. phillips , d. carder , g. berden , b. redlich , a. f. g. van der meer , s. clowes , j. j. harris , l. f. cohen , t. ashley , and l. buckle , phys . * 96 * , 096603 ( 2006 ) ] . note that the soc parameter we used in this section ( @xmath201 ev@xmath149@xmath315 ) is larger than the value ( @xmath198 ev@xmath149@xmath315 ) fitted from experimental data ( see appendix b ) . however , as the dp spin relaxation is more than two orders of magnitude larger than the bap spin relaxation , the choice of the @xmath100 does not change the conclusion that the bap mechanism is unimportant . further calculation indicates that the results from eq . ( [ bapelastapp1 ] ) deviate more from the calculation via the ksbes at small spin polarization as the pauli blocking of spin - flip scattering is more pronounced when both spin bands are nearly equally populated . according to eq . ( [ hfeq ] ) , the effect of the coulomb hf term can be estimated by @xmath334 $ ] where @xmath335 describes the average coulomb interaction with @xmath336 , @xmath137 , @xmath159 being the average momentum , the electron density , and the initial spin polarization , respectively . thus the effect of the coulomb hf term is more pronounced for larger @xmath137 , @xmath159 and weaker scattering ( i.e. , larger @xmath132 ) . moreover , increase of the temperature would reduce the effect as @xmath336 becomes larger and @xmath335 becomes smaller . s. oertel , j. hbner , and m. oestreich , appl . . lett . * 93 * , 132112 ( 2008 ) . the ey mechanism is checked to be unimportant in the parameter region we studied for both @xmath1-gaas and @xmath1-gasb . the initial spin polarization dependence of the srt is weak due to the fact that the impurity density is larger than the electron density according to the analysis in sec . r. j. seymour , m. r. junnarkar , and r. r. alfano , phys . b * 24 * , 3623 ( 1981 ) . h. haug and s. w. koch , _ quantum theory of the optical and electronic properties of semiconductors _ ( world scientific , singapore , 2004 ) , 4th ed . j. h. jiang , y. zhou , t. korn , c. schller , and m. w. wu , arxiv:0901.0061 . j. y. fu and m. w. wu , j. appl . phys . * 104 * , 093712 ( 2008 ) . spin relaxation in zno has attracted much interest recently . experiment indicates very long spin relaxation time [ s. ghosh , v. sih , w. h. lau , d. d. awschalom , s .- y . bae , s. wang , s. vaidya , and g. chapline , appl . . lett . * 86 * , 232507 ( 2005 ) ] and the electrical control of spin lifetime was also achieved [ s. ghosh , d. w. steuerman , b. maertz , k. ohtani , h. xu , h. ohno , and d. d. awschalom , appl . . lett . * 92 * , 162109 ( 2008 ) ] . a theoretical study including both the localized and conduction electrons with the wurzite - structure - induced rashba - like soc considered was performed [ n. j. harmon , w. o. putikka , and r. joynt , arxiv:0808.2913 ] . m. krau , r. bratschitsch , z. chen , s. t. cundiff , and h. c. schneider , arxiv:0902.0270 . j. l. cheng , ph.d . dissertation , university of science and technology of china , 2007 .	electron spin relaxation in bulk iii - v semiconductors is investigated from a fully microscopic kinetic spin bloch equation approach where all relevant scatterings , such as , the electron nonmagnetic - impurity , electron - phonon , electron - electron , electron - hole , and electron - hole exchange ( the bir - aronov - pikus mechanism ) scatterings are explicitly included . the elliot - yafet mechanism is also fully incorporated . this approach offers a way toward thorough understanding of electron spin relaxation both near and far away from the equilibrium in the metallic regime . the dependences of the spin relaxation time on electron density , temperature , initial spin polarization , photo - excitation density , and hole density are studied thoroughly with the underlying physics analyzed . we find that these dependences are usually _ qualitatively _ different in the non - degenerate and degenerate regimes . in contrast to the previous investigations in the literature , we find that : ( i ) in @xmath0-type materials , the elliot - yafet mechanism is _ less _ important than the dyakonov - perel mechanism , even for the narrow band - gap semiconductors such as insb and inas . ( ii ) the density dependence of the spin relaxation time is nonmonotonic and we predict a _ peak _ in the metallic regime in both @xmath0-type and intrinsic materials . ( iii ) in intrinsic materials , the bir - aronov - pikus mechanism is found to be negligible compared with the dyakonov - perel mechanism . we also predict a peak in the temperature dependence of spin relaxation time which is due to the nonmonotonic temperature dependence of the electron - electron coulomb scattering in intrinsic materials with small initial spin polarization . ( iv ) in @xmath1-type iii - v semiconductors , the bir - aronov - pikus mechanism dominates spin relaxation in the low temperature regime only when the photo - excitation density is low . when the photo - excitation density is high , the bir - aronov - pikus mechanism can be comparable with the dyakonov - perel mechanism only in the moderate temperature regime roughly around the fermi temperature of electrons , whereas for higher or lower temperature it is unimportant . the relative importance of the bir - aronov - pikus mechanism decreases with the photo - excitation density and eventually becomes negligible at sufficiently high photo - excitation density . the effect of electric field on spin relaxation in @xmath0-type iii - v semiconductors is also studied with behaviors very different from those in the two - dimensional case reported . finally , we find good agreement of our calculation with the experimental results .
our main aim in this paper is to develop a method to calculate the nron - tate heights @xmath0 of certain tautological integral cycles @xmath1 on jacobians @xmath2 of curves @xmath3 defined over a number field . the cycles @xmath1 are obtained as images inside @xmath2 of small cartesian powers @xmath4 of the curve . here @xmath5 is an @xmath6-tuple of non - zero integers , and @xmath7 is a chosen base divisor of degree one on @xmath3 . we obtain closed expressions for the height of the difference surface inside @xmath2 , the abel - jacobi images of the square of the curve , and of any symmetric theta divisor . in many cases , including the case of the abel - jacobi images of the curve @xmath3 itself , we are able to derive a positive lower bound for the nron - tate height of @xmath1 . we recall that such lower bounds yield effective versions of the ( generalized ) bogomolov conjecture for @xmath1 . the closed expression that we obtain for the nron - tate height of a symmetric theta divisor leads to a proof , in the case of jacobians , of a formula , proposed around ten years ago by p. autissier @xcite , relating the faltings height of a principally polarized abelian variety with the nron - tate height of a symmetric theta divisor . let @xmath8 be an abelian variety defined over @xmath9 , and let @xmath10 be a symmetric ample line bundle on @xmath8 defining a principal polarization on @xmath8 . then to each closed subvariety @xmath11 of @xmath8 one has associated its nron - tate height @xmath12 , measuring in an intrinsic way the arithmetic complexity of @xmath11 . the case of points @xmath11 being classical , approaches to define @xmath13 for higher dimensional @xmath11 were only developed much more recently in work of p. philippon @xcite , w. gubler @xcite , j .- b . bost , h. gillet and c. soul @xcite @xcite , and s. zhang @xcite . two important properties of the nron - tate height are : for all integers @xmath14 and all subvarieties @xmath15 we have a quadratic relation @xmath16_z)=n^2\,\h'_\ll(z)$ ] , and moreover we always have @xmath17 . one well known application of the nron - tate height has been in the study of the ( generalized ) bogomolov conjecture for subvarieties of @xmath8 . let @xmath18 be a closed subvariety and let @xmath19 be the so - called ( first ) essential minimum of @xmath11 . then zhang showed in @xcite the fundamental inequalities @xmath20 between @xmath21 and the nron - tate height @xmath13 of @xmath11 . in particular we have the equivalence @xmath22 the generalized bogomolov conjecture , proved by e. ullmo and zhang in @xcite @xcite using equidistribution techniques , states that both conditions in ( [ equiv ] ) are satisfied if and only if @xmath11 is not a translate by a torsion point of an abelian subvariety of @xmath8 . an effective positive lower bound for @xmath13 or @xmath21 is usually called an `` effective bogomolov - type result '' for @xmath11 . the methods of @xcite @xcite do not provide such lower bounds , but later on such lower bounds have been given for general @xmath11 , starting with a fundamental work of s. david and p. philippon @xcite @xcite @xcite @xcite . in a rough form , these lower bounds are of the shape @xmath23 where @xmath24 are positive constants depending only on @xmath8 and @xmath10 . in this paper we are interested in the particular case of jacobians . thus , let @xmath3 be a smooth projective geometrically connected curve of genus @xmath25 defined over @xmath26 , and let @xmath27 denote the jacobian of @xmath3 , endowed with its canonical principal polarization . choose a divisor @xmath7 of degree one on @xmath3 . we consider integral cycles @xmath1 on @xmath2 obtained as the images of maps @xmath28 \ , , \ ] ] where @xmath29 and where @xmath30 is an @xmath6-tuple of non - zero integers with @xmath31 . we note that the maps @xmath32 are always generically finite . if @xmath33 we usually just write @xmath34 and @xmath35 , and for all @xmath36 we simply write @xmath37 and @xmath38 in the fundamental case where @xmath5 is the all-@xmath39 tuple of length @xmath6 . let @xmath10 be a symmetric ample line bundle on @xmath2 defining @xmath40 . let @xmath41 be a number field such that @xmath3 is defined and has semistable reduction over @xmath42 . we note that such a number field exists by the semistable reduction theorem of deligne - mumford . our aim is to express the nron - tate height @xmath0 of @xmath1 in terms of admissible arithmetic intersection theory on models of @xmath3 over the ring of integers of @xmath42 in the sense of zhang @xcite @xcite . for the cycle @xmath43 such an expression is already well known . namely , let @xmath44 be the relative dualizing sheaf of @xmath3 equipped with its canonical admissible metric ( ? ? ? * section 4 ) . assume @xmath45 and put @xmath46 , where @xmath47 is a canonical divisor on @xmath3 . then the identity @xmath48 } { \langle\hat{\omega},\hat{\omega}\rangle } + \frac{g-1}{g}\h'_\ll(x_\alpha)\ ] ] holds in @xmath49 , as is proved in for instance ( * ? ? ? * theorem 5.6 ) or ( * ? ? ? * theorem 3.9 ) . here @xmath50 denotes the self - intersection of the relative dualizing sheaf on @xmath3 in the admissible sense ( * ? ? ? * section 5.4 ) . we can further express the nron - tate height @xmath51 of the degree zero divisor @xmath52 in terms of admissible intersection theory on @xmath3 by the celebrated faltings - hriljac formula , or hodge index theorem ( * ? ? ? * section 5.4 ) . our approach is inspired by zhang s work @xcite on the height of gross - schoen cycles . we start by briefly reviewing this work . let @xmath53 denote the set of finite places of @xmath42 , let @xmath54 denote the set of complex embeddings of @xmath42 , and let @xmath55 . an important role in @xcite is played by a new real - valued local invariant @xmath56 of @xmath3 for all @xmath57 . we refer to section [ phi_explained ] below for a precise definition of @xmath56 . for now , we mention that we have @xmath58 for all @xmath57 if @xmath59 , and that we have @xmath58 if @xmath60 is non - archimedean and @xmath3 has good reduction at @xmath60 . for @xmath61 the invariant @xmath56 has independently been introduced and studied by n. kawazumi in the papers @xcite @xcite . for @xmath62 the invariant @xmath56 can be calculated from the polarized dual graph of the reduction of @xmath3 at @xmath60 . for @xmath62 we put @xmath63 , where @xmath64 is the residue field at @xmath60 , and for @xmath61 we put @xmath65 . we then have a well - defined global invariant @xmath66 of @xmath3 . this global invariant shows up in @xcite in a formula for the height @xmath67 in the sense of beilinson - bloch @xcite @xcite of the gross - schoen cycle @xmath68 associated to @xmath3 and @xmath7 ( see @xcite or @xcite for a definition ) . zhang s formula ( cf . * theorem 1.3.1 ) ) reads in front of @xmath69\,\h'_\ll(x_\alpha)$ ] . the difference is due to a different normalization of the nron - tate height of a point . ] @xmath70\,\h'_\ll(x_\alpha ) \ , .\ ] ] the arithmetic analogue of grothendieck s standard conjecture of hodge index type proposed by gillet - soul @xcite predicts that @xmath71 , which is however not known to be true in general . by ( [ htgs ] ) the arithmetic standard conjecture implies the lower bound @xmath72 for @xmath50 . as it turns out , the right hand side of this inequality is strictly positive if @xmath45 . indeed , if @xmath45 we have @xmath73 for each @xmath61 , and moreover by a deep result due to z. cinkir @xcite we have if @xmath74 for each @xmath62 the inequality @xmath75 } \frac{2i(g - i)}{g}\delta_i(x_v ) \ , , \ ] ] where @xmath76 is some positive constant depending only on @xmath77 and where , following traditional notation , we denote by @xmath78 the number of non - separating geometric double points on the reduction of @xmath3 at @xmath60 , and by @xmath79 for @xmath80 $ ] the number of geometric double points on the reduction of @xmath3 at @xmath60 whose local normalization has two connected components , one of arithmetic genus @xmath81 , and one of arithmetic genus @xmath82 . it follows that the conjectural inequality in ( [ conjlowerbound ] ) states an effective positive lower bound for @xmath50 , and therefore , by equation ( [ heightcurve ] ) , an effective bogomolov - type result for @xmath43 . we will show in corollary [ effectivebog ] below that the lower bound in ( [ conjlowerbound ] ) actually holds with the coefficient @xmath83 replaced by @xmath84 . hence , we unconditionally have an effective bogomolov - type result for @xmath43 . if @xmath3 is hyperelliptic and @xmath85 , the gross - schoen cycle @xmath86 is rationally equivalent to zero ( * ? ? ? * proposition 4.8 ) , and hence has vanishing beilinson - bloch height . we conclude that for hyperelliptic curves @xmath3 the equality @xmath87 holds . in this paper we will elaborate on zhang s method in @xcite that leads to his formula ( [ htgs ] ) for the height of the gross - schoen cycle . we will show that it allows more generally to compute the height of any of the cycles @xmath1 introduced above . more precisely , we will show the following result . [ mainintro ] assume @xmath45 . let @xmath29 and let @xmath30 be an @xmath6-tuple of non - zero integers . there exist rational numbers @xmath88 depending only on @xmath5 and @xmath77 such that for all number fields @xmath42 , all curves @xmath3 of genus @xmath77 with semistable reduction over @xmath42 , and all @xmath89 the equality @xmath90 } \left ( a \ , { \langle\hat{\omega},\hat{\omega}\rangle } + b \ , \sum_{v \in m(k ) } \varphi(x_v ) \log n v \right ) + c \ , \h'_\ll(x_\alpha)\ ] ] holds . here @xmath46 . let @xmath3 be a curve over @xmath9 of genus @xmath45 , and let @xmath89 . let @xmath91 be the @xmath92-span inside @xmath49 of all @xmath0 where @xmath5 is an @xmath6-tuple of non - zero integers with @xmath93 . then @xmath94 . if @xmath3 is hyperelliptic , we have @xmath95 . apart from @xcite , our proof of theorem [ mainintro ] is heavily inspired by a method developed recently by r. wilms in the context of the moduli space of curves @xcite . in particular , our proof provides an algorithm to compute @xmath96 , @xmath97 and @xmath98 from @xmath5 and @xmath77 . upon hearing about our proof , d. holmes implemented the algorithm in ` sage ` , and kindly made it publicly available . from running the algorithm it seems not straightforward in general to give a closed expression for @xmath96 , @xmath97 or @xmath98 in terms of @xmath5 and @xmath77 . we will illustrate theorem [ mainintro ] by exhibiting some particular examples , and deducing some consequences , in the cases @xmath99 and @xmath100 , the case @xmath101 being essentially dealt with by ( [ heightcurve ] ) . here is first of all what we find if we set @xmath99 . [ squareintro ] let @xmath3 be a smooth projective geometrically connected curve of genus @xmath45 with semistable reduction over the number field @xmath42 . let @xmath102 be the image of @xmath103 in @xmath2 under the map @xmath104 given by @xmath105 $ ] . then the equality @xmath106 \ , \h'_\ll(f ) = ( 3g-1)\,{\langle\hat{\omega},\hat{\omega}\rangle } - 2 \sum_{v \in m(k ) } \varphi(x_v ) \log nv\ ] ] holds . next , let @xmath107 , and let @xmath108 be the image of @xmath103 in @xmath2 under the map @xmath104 given by @xmath109 $ ] . then the equality @xmath110 \ , \h'_\ll(z_{2,\alpha } ) = \frac{3g^2 - 8g-1}{g-1}{\langle\hat{\omega},\hat{\omega}\rangle } \\ & + 2 \sum_{v \in m(k ) } \varphi(x_v)\log nv + 48(g-1)(g-2 ) \,\h'_\ll(x_\alpha ) \end{split}\ ] ] holds . as @xmath111 we immediately deduce from the first equation the following result . [ effectivebog ] let @xmath3 be a smooth projective geometrically connected curve of genus @xmath45 with semistable reduction over the number field @xmath42 . then the inequalities @xmath112 hold . by equation ( [ heightcurve ] ) we obtain from corollary [ effectivebog ] an effective bogomolov - type result for @xmath43 . more generally , whenever one has @xmath113 , @xmath114 and @xmath115 in the identity of theorem [ mainintro ] , one finds by corollary [ effectivebog ] an effective bogomolov - type result for @xmath1 . experimenting with the ` sage ` implementation of our algorithm has so far not led to an improvement of the lower bound in corollary [ effectivebog ] , and we believe that the lower bound in corollary [ effectivebog ] is in fact the best possible one that one can obtain by the methods described in this paper . upper bounds for @xmath116 and @xmath117 play an important role in a recent work @xcite by p. parent . for example , a combination of theorem [ squareintro ] and corollary [ effectivebog ] yields the estimate @xmath118 } \frac{g-2}{2(g-1)^2 } { \langle\hat{\omega},\hat{\omega}\rangle } + \frac{4(g-2)}{g } \h'_\ll(x_\alpha ) \ , , \ ] ] and then using equation ( [ heightcurve ] ) we find @xmath119 this estimate may be used to improve some of the estimates in ( * ? ? ? * section 5.1 ) . the next example that we consider is the case of a symmetric theta divisor inside @xmath2 . we note that such theta divisors can be identified with cycles @xmath120 where @xmath107 is such that @xmath85 , in particular we are dealing with the case that @xmath100 . we find the following result . [ thetaheight ] let @xmath3 be a smooth projective geometrically connected curve of genus @xmath121 with semistable reduction over a number field @xmath42 , and denote by @xmath27 its jacobian . let @xmath122 be an effective symmetric ample divisor on @xmath2 that defines the principal polarization @xmath40 , and put @xmath123 . then the identity @xmath124 \ , \h'_\ll(\theta ) = \frac{1}{12 } { \langle\hat{\omega},\hat{\omega}\rangle } + \frac{1}{6 } \sum_{v \in m(k ) } \varphi(x_v ) \log nv\ ] ] holds . more generally , let @xmath125 be a principally polarized abelian variety over @xmath9 of dimension @xmath126 . then one has naturally associated to @xmath125 the nron - tate height @xmath127 of any symmetric effective divisor @xmath122 defining @xmath40 on @xmath8 , where @xmath128 . another natural invariant associated to @xmath125 is the stable faltings height @xmath129 as introduced in @xcite . it is natural to ask in general how @xmath127 and @xmath129 are related . this question has been investigated by p. autissier in his paper @xcite . before we can state autissier s main result , we need to introduce some more notation . assume that @xmath8 and @xmath10 are defined over a number field @xmath41 , and assume that an admissible adelic metric @xmath130 has been chosen on @xmath10 ( see section [ admissible ] for definitions ) . let @xmath131 be any non - zero global section of @xmath10 on @xmath8 . for each @xmath61 we then put @xmath132 where @xmath133 denotes the haar measure on the complex torus @xmath134 . the real - valued local invariant @xmath135 is independent of the choice of @xmath131 and @xmath10 , and an application of jensen s inequality yields the strict inequality @xmath136 . the main result of @xcite is then the following . as above let @xmath129 denote the stable faltings height of @xmath8 , and put @xmath137 . assume that @xmath8 has good reduction over the number field @xmath42 . then one has the identity @xmath138 } \sum_{v \in m(k)_\infty } i(a_v,\lambda_v ) \log nv\ ] ] in @xmath49 . autissier s result is based on the `` key formula '' for principally polarized abelian schemes . naturally , one wants to remove the condition that @xmath8 has ( potentially ) everywhere good reduction . in fact , autissier proposes in ( * ? ? ? * question ) that for _ any _ principally polarized abelian variety @xmath125 of dimension @xmath126 with semistable reduction over some number field @xmath42 one should have a natural identity of the type @xmath139 } \sum_{v \in m(k)_0 } \alpha_v \log nv \\ & - \kappa_0 g + \frac{2}{[k:\qq ] } \sum_{v \in m(k)_\infty } i(a_v,\lambda_v ) \log nv \end{split}\ ] ] where @xmath140 is a non - negative rational number that can be calculated from the reduction of @xmath8 at @xmath60 , with @xmath141 if @xmath8 has good reduction at @xmath60 . in @xcite autissier proved his proposed formula ( [ autissconj ] ) for all principally polarized abelian varieties of dimension one or two , and for arbitrary products of such . in these cases he was also able to give an explicit description for the local non - archimedean factors @xmath140 . our theorem [ autforjac ] below will give identity ( [ autissconj ] ) for ( products of ) arbitrary jacobians , with an explicit description of the @xmath140 in terms of the reduction graph of the underlying curves at @xmath60 . let @xmath57 be a finite place . we recall @xcite that the dual graph of the reduction of @xmath3 at @xmath60 is naturally a metrized graph . let @xmath142 be any metrized graph . then viewing @xmath142 as an electric circuit we have for each pair @xmath143 , @xmath144 of points of @xmath142 the effective resistance @xmath145 between @xmath143 and @xmath144 . taking @xmath143 as a base point the function @xmath145 is piecewise quadratic in @xmath146 and has a well - defined laplacian @xmath147 in the sense of ( * ? ? ? * appendix ) . we put @xmath148 then @xmath149 is a signed measure on @xmath142 , independent of the choice of @xmath150 . the canonical measure @xmath149 was first introduced and studied by t. chinburg and r. rumely @xcite who also give an explicit formula for it . we write @xmath151 for the natural real - valued invariant @xmath152 of @xmath142 , which can also be shown @xcite to be independent of the choice of @xmath150 . the invariant @xmath151 and its applications in arithmetic geometry have been studied in various papers , including @xcite @xcite @xcite @xcite @xcite . we note that @xmath151 is rational if in some model of @xmath142 all edges have rational lengths . when @xmath3 is a curve with semistable reduction over a number field @xmath42 and @xmath57 is a finite place , we simply write @xmath153 for the tau - invariant of the dual graph of the reduction of @xmath3 at @xmath60 . in a similar vein we write @xmath154 } \delta_i(x_v)$ ] for the total length of the dual graph at @xmath60 , or equivalently the number of geometric singular points on the reduction of @xmath3 at @xmath60 . both @xmath153 and @xmath155 are non - negative rational numbers , and vanish if @xmath3 has good reduction at @xmath60 . with these definitions we have the following result , answering autissier s question in the affirmative for jacobians . [ autforjac ] let @xmath3 be a smooth projective geometrically connected curve of genus @xmath126 with semistable reduction over a number field @xmath42 . let @xmath27 be the jacobian of @xmath3 . for each @xmath62 we define the rational number @xmath156 then for each @xmath62 we have @xmath157 . moreover we have @xmath141 if and only if the jacobian @xmath2 has good reduction at @xmath60 . finally the identity @xmath158 } \sum_{v \in m(k)_0 } \alpha_v \log nv \\ & - \kappa_0 g + \frac{2}{[k:\qq ] } \sum_{v \in m(k)_\infty } i(j_v,\lambda_v ) \log nv \end{split}\ ] ] holds in @xmath49 . we note that the inequality @xmath159 for metrized graphs @xmath142 is due to rumely and has been known for some time ( * ? ? ? * equation ( 14.3 ) ) ( * ? ? ? * equation ( 10 ) ) . our proofs of theorems [ thetaheight ] and [ autforjac ] rely on autissier s result ( [ aut ] ) in the potentially everywhere good reduction case , the noether formula for arithmetic surfaces , and the remarkable identity @xmath160 for @xmath61 found recently by r. wilms in the already quoted paper @xcite , expressing the faltings delta - invariant @xmath161 from @xcite in terms of the invariants @xmath162 and @xmath56 . + the set - up of this paper is as follows . in section [ prelims ] we review the necessary notions and results from intersection theory of integrable adelic line bundles from zhang s paper @xcite . we will assume that the reader is familiar with arithmetic intersection theory in the style of h. gillet and c. soul @xcite @xcite . in section [ sec : projection ] we show how several projection formulae from @xcite @xcite generalize to the context of integrable line bundles . these results seem to be of independent interest . we include a study of the deligne pairing of integrable line bundles along flat projective morphisms . in section [ admissible ] we consider a special class of integrable line bundles on curves and polarized abelian varieties , namely the admissible bundles . on curves , the admissible line bundles are precisely those constructed in @xcite by considering green s functions on the reduction graph of a curve . we will see how the nron - tate height of a cycle on a polarized abelian variety can be expressed in terms of admissible intersection theory . in section [ phi_explained ] we review from @xcite the @xmath163-invariant and some of its properties . from section [ ntheightspecial ] we start with actual computations . we first derive in theorem [ main ] a general formula for the nron - tate height of a tautological integral cycle in terms of deligne pairings of suitable integrable line bundles . in section [ proofmain ] we then deduce theorem [ mainintro ] from theorem [ main ] . we focus on the particular case of two - dimensional cycles in section [ surfaces ] , leading to a proof of theorem [ squareintro ] . finally in section [ proofautforjac ] we prove theorems [ thetaheight ] and [ autforjac ] about the symmetric theta divisor . in this paper , a variety is a reduced scheme of finite type over a field , and a curve is a variety of dimension one . + acknowledgements i would like to thank david holmes for valuable remarks and discussions , and for implementing the method described in this paper in ` sage ` . we recall the notion of integrable adelic line bundle from @xcite and discuss the arithmetic intersection number of a tuple of integrable line bundles . in particular this leads to the notion of height of a cycle with respect to an ample integrable line bundle , cf . definition [ height ] . although we work mainly from @xcite we also recommend @xcite @xcite for a different perspective on the material of this section ( and the next ones ) , making a more consistent use of berkovich analytic spaces to deal with the metrics at the finite places . we assume throughout that the reader is familiar with the basics of arithmetic intersection theory in the style of gillet - soul @xcite @xcite . a word on terminology : what is here and in @xcite called an algebraic line bundle corresponds to what is called a smooth line bundle in @xcite , and what is here and in @xcite called an integrable line bundle corresponds to what is called an admissible line bundle in @xcite . let @xmath164 be a local field , and let @xmath165 be an algebraic closure of @xmath166 . let @xmath3 be a projective geometrically irreducible variety over @xmath166 , and let @xmath10 be a line bundle on @xmath3 . a continuous metric @xmath167 on @xmath10 is the datum of a continuously varying family of @xmath165-norms @xmath168 on the fibers @xmath169 of @xmath10 , where @xmath143 moves through the topological space @xmath170 . if @xmath166 is archimedean , we always assume that @xmath171 is the pullback , along a closed immersion @xmath172 of @xmath3 into a smooth projective geometrically irreducible variety @xmath173 , of a smooth hermitian line bundle on the complex manifold @xmath174 . in particular , we have a curvature current @xmath175 on @xmath170 . if @xmath166 is non - archimedean , an important class of continuously metrized line bundles on @xmath3 can be defined via models of @xmath3 over the ring of integers @xmath176 of @xmath166 as follows . in general , when @xmath177 is a locally free @xmath176-module of rank one , we have a natural @xmath165-norm @xmath178 on the @xmath165-vector space @xmath179 by putting @xmath180 for all @xmath181 . now let @xmath182 , and let @xmath183 be a projective flat morphism with generic fiber @xmath3 . let @xmath10 be a line bundle on @xmath3 , and let @xmath184 be a line bundle on @xmath185 whose restriction to @xmath3 is equal to @xmath186 , for some @xmath187 . we call @xmath188 a model for @xmath189 . for @xmath190 let @xmath191 be the unique section of @xmath192 that extends @xmath143 ( recall that @xmath192 is projective ) . then @xmath193 is a locally free @xmath176-module of rank one satisfying @xmath194 and for @xmath195 we then put @xmath196 . the definition extends naturally to points @xmath197 . we say that the continuous metric @xmath198 on @xmath10 is algebraic induced by the model @xmath188 of @xmath189 . next , let @xmath42 be a number field . let @xmath53 denote the set of finite places of @xmath42 , @xmath54 the set of complex embeddings of @xmath42 , and write @xmath55 . for @xmath57 let @xmath199 denote the completion of @xmath42 at @xmath60 , in particular we have for each @xmath57 a canonical embedding @xmath200 . for @xmath62 let @xmath201 be the unique prime of @xmath202 dividing @xmath60 . then we endow @xmath199 with the unique absolute value @xmath203 such that @xmath204}$ ] . for @xmath61 we put the usual euclidean norm on @xmath199 . the collection of absolute values thus obtained for all @xmath57 satisfies the product formula : for all @xmath205 we have @xmath206 . let @xmath3 be a projective geometrically irreducible variety over @xmath42 , and let @xmath10 be a line bundle on @xmath3 . for each @xmath57 we denote @xmath207 and @xmath208 . an adelic metric on @xmath10 is a collection of continuous metrics @xmath209 on @xmath210 on @xmath211 indexed by @xmath57 . let @xmath187 . let @xmath176 be the ring of integers of @xmath42 . by a model of @xmath189 we mean a pair @xmath212 where @xmath185 is a projective flat model of @xmath3 over @xmath176 and @xmath213 is a smooth hermitian line bundle on @xmath185 whose underlying line bundle @xmath184 coincides with @xmath186 over @xmath3 . a model of @xmath189 gives naturally , using for @xmath62 the construction described before , an adelic metric @xmath130 on @xmath10 . we say that an adelic metrized line bundle @xmath214 on @xmath3 is approximated by models @xmath215 of @xmath216 if for all but finitely many @xmath57 the metric @xmath217 induced by @xmath215 is independent of @xmath81 and if for all @xmath218 the functions @xmath219 on @xmath220 converge uniformly to zero . let @xmath215 be a sequence of models approximating the adelic line bundle @xmath221 , and assume that for each @xmath222 we have a birational morphism @xmath223 of models which is an isomorphism over @xmath42 . then the sequence of models @xmath224 also approximates @xmath221 . indeed , each section @xmath225 lifts uniquely to a section @xmath226 by properness , and then @xmath227 so that for each @xmath222 the algebraic metrics defined by @xmath228 and @xmath229 on @xmath3 are the same . in particular , we may assume the models @xmath230 to be normal . let @xmath231 for @xmath232 be sequences of models approximating the adelic line bundles @xmath233 . put @xmath234 for @xmath235 , and denote by @xmath236 the pullbacks of the @xmath237 along the projections @xmath238 . then the sequences of models @xmath239 also approximate @xmath233 . let @xmath185 be a projective flat scheme over @xmath176 . we say that a smooth hermitian line bundle @xmath213 on @xmath185 is semipositive ( resp . ample ) if @xmath184 is relatively semipositive ( resp . relatively ample ) , and each metric @xmath209 for @xmath61 is the restriction of a smooth hermitian metric with semipositive ( resp . positive ) curvature form . let @xmath3 be a projective geometrically irreducible variety over @xmath42 , then an adelic metrized line bundle @xmath240 on @xmath3 is said to be semipositive ( resp . ample ) if @xmath221 is approximated by models @xmath215 of @xmath216 with each @xmath241 semipositive ( resp . finally we say that the adelic line bundle @xmath221 on @xmath3 is integrable if there exist two semipositive adelic line bundles @xmath242 , @xmath243 on @xmath3 such that @xmath244 . let @xmath245 be a morphism of projective geometrically irreducible varieties over @xmath42 , and let @xmath221 be an integrable line bundle on @xmath3 . then @xmath246 is an integrable line bundle on @xmath173 . indeed , for each model @xmath185 of @xmath3 there exists a model @xmath247 of @xmath173 together with a morphism @xmath248 extending @xmath32 . then if @xmath249 is a semipositive smooth hermitian line bundle on @xmath185 , we have that @xmath250 is a semipositive smooth hermitian line bundle on @xmath247 . for a projective flat scheme @xmath185 over @xmath176 with smooth generic fiber we denote by @xmath251 the arithmetic chow group of @xmath185 in degree @xmath81 , cf . * section 3.3 ) . we write @xmath182 . by ( * section 2.1 ) we have canonical maps @xmath252 the arithmetic chow groups @xmath253 vanish for @xmath254 . let @xmath255 be an integral closed subscheme of dimension @xmath256 on @xmath185 . then for each collection @xmath257 of smooth hermitian line bundles on @xmath185 we have by ( * section 2.3 ) an element @xmath258 in @xmath259 . if @xmath260 we obtain , by taking @xmath261 or @xmath262 , an element of @xmath49 or @xmath202 , for which we use the same notation . if @xmath185 is regular , the element @xmath263 in @xmath259 is constructed in ( * ? ? ? * section 2.3 ) using arithmetic intersection theory on @xmath185 . if @xmath185 is not necessarily regular , one way of obtaining @xmath263 is to use the existence of a smooth morphism @xmath264 , a closed immersion @xmath265 over @xmath266 , and smooth hermitian line bundles @xmath267 on @xmath268 together with isometries @xmath269 for @xmath270 , and to put @xmath271 . it can be shown that the result is independent of the choices of @xmath272 and @xmath273 . we refer to ( * ? ? ? * section 2.3.1 , remark ( ii ) ) and ( * ? ? ? * section 3.2.1 , remark ) for details . the intersection number @xmath263 is symmetric and multilinear in @xmath274 . moreover , when @xmath275 we have @xmath276 \in \zz \ , , \ ] ] where @xmath277 denote the generic fibers of @xmath278 . we can extend these arithmetic intersection numbers to the integrable adelic case as follows . let @xmath233 be semipositive adelic line bundles on @xmath3 approximated by semipositive models @xmath279 of @xmath280 . let @xmath281 be an integral cycle of dimension @xmath282 on @xmath3 and let @xmath283 be the zariski closure of @xmath11 in @xmath230 . by ( * ? * theorem 1.4 ) the sequence @xmath284 converges , with limit independent of choice of models . we denote the limit by @xmath285 . by writing an arbitrary integrable line bundle as a quotient of two semipositive line bundles , the definition of @xmath286 is extended to arbitrary tuples @xmath287 of integrable line bundles on @xmath3 . when @xmath288 we often simply write @xmath289 , and when @xmath290 we often write @xmath291 . the intersection number @xmath292 is symmetric and multilinear in @xmath293 , and vanishes unless @xmath294 . in case @xmath275 we have @xmath295 \in \zz \ , .\ ] ] [ height ] let @xmath296 be a closed subvariety of dimension @xmath6 . let @xmath221 be an integrable ample line bundle on @xmath3 . then one puts @xmath297 } \frac { \langle \hat{\ll}^{r+1 } \| z \rangle}{\langle \hat{\ll}^r\|z \rangle ( r+1 ) } = \frac{1}{[k:\qq ] } \frac { \langle \hat{\ll}^{r+1 } \| z \rangle}{\deg_{\ll}(z ) ( r+1 ) } \ ] ] for the ( normalized or absolute ) height of @xmath11 with respect to @xmath221 . the purpose of this section is to state and prove some projection formulae for integrable line bundles . recall that an integrable line bundle is essentially determined by a sequence of models . at each level in the sequence , the usual arithmetic intersection numbers satisfy a number of projection formulae as shown in @xcite @xcite . these formulae yield analogues for integrable line bundles by passage to the limit . along the way we include a discussion of the deligne pairing of integrable bundles . the first projection formula we state is also observed in ( * ? ? ? * section 3.1.2 ) . [ projection ] let @xmath298 be projective geometrically irreducible varieties over the number field @xmath42 . let @xmath245 be a morphism . let @xmath299 be an integral cycle on @xmath173 . let @xmath233 be integrable adelic line bundles on @xmath3 . then the adelic line bundles @xmath300 are integrable on @xmath173 , and the equality @xmath301 holds . without loss of generality we may assume that the @xmath302 are semipositive and approximated by semipositive models @xmath303 of @xmath304 . for each @xmath305 there exists a model @xmath306 of @xmath173 together with a morphism @xmath307 extending the morphism @xmath32 . then each @xmath308 is semipositive , approximated by semipositive models @xmath309 . for each @xmath305 let @xmath283 denote the zariski closure of @xmath11 inside @xmath306 . by the projection formula from ( * proposition 3.2.1 ) and the remark following ( * ? ? ? * proposition 3.2.1 ) we have for each @xmath305 an equality of arithmetic intersection numbers @xmath310 in @xmath49 . dividing both sides by @xmath311 and taking the limit as @xmath312 we obtain the equality from the proposition . [ externalproducts ] let @xmath298 be projective geometrically irreducible varieties over the number field @xmath42 . let @xmath313 and @xmath314 denote the canonical projections . let @xmath233 be integrable adelic line bundles on @xmath3 , and let @xmath315 be integrable adelic line bundles on @xmath173 . then the adelic line bundles @xmath316 are integrable on @xmath317 , and the equality @xmath318 holds . without loss of generality we may assume that the @xmath302 resp . @xmath319 are semipositive and approximated by semipositive models @xmath303 of @xmath304 resp . @xmath320 of @xmath321 . denote by @xmath322 and @xmath323 the canonical projections extending @xmath201 and @xmath324 . we claim that for each @xmath222 the equality @xmath325 holds . upon dividing both sides of the equality by @xmath326 and taking limits as @xmath312 we then obtain the equality from the proposition . to prove the claim , recall that @xmath327 and @xmath306 can be realized as integral closed subschemes of projective schemes @xmath328 and @xmath329 that are smooth over @xmath266 , and that the hermitian line bundles @xmath237 and @xmath330 can be assumed to be restrictions to @xmath327 and @xmath306 of hermitian line bundles on @xmath328 and @xmath329 . the required equality then follows by an application of ( * ? ? ? * proposition 2.3.3 ) . both sides of the equality in proposition [ externalproducts ] vanish unless @xmath331 and @xmath332 , or @xmath333 and @xmath332 , or @xmath331 and @xmath334 . our next goal is to introduce and study the deligne pairing for integrable adelic line bundles . we refer to ( * ? ? ? * section 1.1 ) and ( * ? ? ? * section 2.1 ) for more details . let @xmath245 be a flat projective morphism of integral schemes of relative dimension @xmath335 , and let @xmath336 be @xmath337 line bundles on @xmath173 . then we have a canonical line bundle @xmath338 on @xmath3 locally generated by sections @xmath339 , where @xmath340 are local sections of @xmath336 with empty common zero locus , and with relations @xmath341 for all @xmath342 and all regular functions @xmath343 such that @xmath344 is finite flat over @xmath3 and has empty intersection with @xmath345 . here @xmath346 denotes the norm of @xmath347 along the finite flat map @xmath348 . the formation of the deligne pairing @xmath338 is symmetric and multilinear in @xmath336 and commutes with arbitrary base change . now let @xmath245 be a flat morphism of relative dimension @xmath335 between two smooth projective geometrically connected varieties over @xmath42 , let @xmath349 and @xmath350 be projective flat models of @xmath3 and @xmath173 over @xmath176 , and assume that @xmath32 extends into a flat morphism @xmath248 . let @xmath351 be @xmath337 smooth hermitian line bundles on @xmath350 with underlying line bundles @xmath352 and generic fibers @xmath336 . then for each @xmath61 the deligne pairing @xmath353 along @xmath354 is equipped with a canonical hermitian metric @xmath209 which is given recursively as follows . let @xmath340 be non - zero local sections of @xmath336 over @xmath173 with empty common zero locus . assume that @xmath355 is a prime divisor on @xmath173 and that @xmath356 is non - zero for @xmath357 . then for all @xmath61 we have @xmath358 by ( * ? ? ? * theorem a ) the metric @xmath209 is indeed continuous on @xmath359 . if @xmath32 is smooth then the metric @xmath209 is actually smooth on @xmath359 by ( * ? ? ? * proposition 8.5 ) and ( * ? ? ? * thorme i.1.1 ) . [ pushforwarddeligne ] let @xmath3 , @xmath173 be smooth projective geometrically connected varieties over @xmath42 , and let @xmath245 be a smooth morphism of relative dimension @xmath335 . let @xmath349 and @xmath350 be projective flat models of @xmath3 and @xmath173 over @xmath176 , and assume that @xmath32 extends into a flat morphism @xmath248 . then the deligne pairing @xmath353 along @xmath354 is a smooth hermitian line bundle on @xmath349 . moreover , we have a natural induced map @xmath360 , and the equality @xmath361 holds in @xmath362 . the first assertion rephrases what was said just before the proposition . the existence of @xmath360 follows from ( * ? ? ? * theorem 3.6.1 ) . to prove the equality , let @xmath363 be sufficiently general rational sections of @xmath352 over @xmath350 , and write @xmath340 for their restrictions to @xmath173 . for each @xmath342 and each @xmath61 we set @xmath364 . then for each @xmath342 the class @xmath365 in @xmath366 is represented by the arakelov divisor @xmath367 , and the class @xmath368 in @xmath362 is represented by the arakelov divisor @xmath369 using the star - product of green s currents as in ( * ? ? ? * section 2.1 ) . on the other hand the class @xmath370 in @xmath362 is represented by the arakelov divisor @xmath371 where @xmath372 for each @xmath61 . now we have @xmath373 and the recursive relations ( [ local_recursive_infinite ] ) imply that @xmath374 the proposition follows . let @xmath375 be @xmath337 semipositive adelic line bundles on @xmath173 , approximated by semipositive models @xmath320 of @xmath376 for @xmath222 . let @xmath230 be any sequence of models of @xmath3 . possibly after replacing each @xmath306 by a birational model @xmath377 we may assume that there exist morphisms @xmath307 extending @xmath32 . by an elementary argument using hilbert schemes ( cf . * section 5.2 ) ) we may even assume , after replacing @xmath230 by a suitable blow - up and replacing @xmath378 by its strict transform along that blow - up , that each @xmath379 is flat . we then find a sequence of deligne pairings @xmath380 along @xmath379 on the models @xmath327 . it turns out that this sequence defines an adelic line bundle @xmath381 on @xmath3 , independent of choices . assuming that the morphism @xmath32 is smooth , the metrics @xmath209 for @xmath61 are smooth by proposition [ pushforwarddeligne ] , and the adelic line bundle @xmath382 is in fact integrable . the metrics @xmath383 can be described recursively as follows . assume without loss of generality that @xmath384 and that the models @xmath306 are normal . let @xmath340 be non - zero local sections of @xmath336 with empty common zero locus . assume that @xmath355 is a prime divisor on @xmath173 and that @xmath356 is non - zero for @xmath357 . then for all @xmath57 we have @xmath385 for @xmath386 this formula boils down to ( [ local_recursive_infinite ] ) . assume therefore that @xmath57 is a finite place . then we have to make sense of the integral in ( [ local_recursive ] ) . we can do that using the models @xmath320 of @xmath376 as follows . for all @xmath222 let @xmath387 be the rational section of @xmath388 extending the section @xmath389 of @xmath390 . then we have on each @xmath306 an equality of weil divisors @xmath391 where @xmath392 is a weil divisor supported in the closed fiber of @xmath306 over @xmath393 . here @xmath394 denotes the zariski closure of @xmath395 in @xmath306 . then we define @xmath396 \log nv /e_i^n\ , . \end{split}\ ] ] we see that the definition of @xmath382 extends to any collection of integrable line bundles @xmath375 on @xmath173 . also we note that in the context of the berkovich analytic space @xmath397 associated to @xmath398 , where @xmath62 , the measure @xmath399 would be the chambert - loir measure @xcite associated to the integrable line bundles @xmath400 . continuing with the case that @xmath384 , we note the global equalities @xmath401 in @xmath49 , where @xmath340 can be any non - zero rational sections of @xmath336 whose supports have empty common intersection on @xmath173 . assuming that @xmath355 is a prime divisor on @xmath173 , we then obtain from the local recursive formula ( [ local_recursive ] ) the global recursive formula @xmath402 [ projii ] let @xmath298 be smooth projective geometrically connected varieties over @xmath42 and let @xmath245 be a smooth morphism of relative dimension @xmath335 . let @xmath233 be integrable adelic line bundles on @xmath3 , and let @xmath403 be integrable adelic line bundles on @xmath173 . then @xmath404 are integrable adelic line bundles on @xmath173 , and we have that @xmath405 \langle \hat{\ll}_0 , \ldots , \hat{\ll}_r \| x \rangle \ , .\ ] ] here @xmath406 $ ] denotes the multidegree of the generic fiber of @xmath245 with respect to the line bundles @xmath407 . we also have that @xmath408 is an integrable adelic line bundle on @xmath3 , and the identity @xmath409 holds in @xmath49 . without loss of generality we may assume that the @xmath302 and @xmath319 are semipositive , approximated by semipositive models @xmath303 of @xmath410 and @xmath411 of @xmath412 . we may assume in addition that there exist flat morphisms @xmath307 for each @xmath222 extending the morphism @xmath32 . since the morphism @xmath32 is smooth by assumption we have by ( * ? ? ? * theorem 3.6.1 ) natural induced pushforward maps @xmath413 for @xmath414 and @xmath222 . then by ( * ? ? ? * theorem 4.4.3 ) we have for @xmath414 and @xmath222 an identity @xmath415 in the arithmetic chow group @xmath416 . we note that the condition in ( * ? ? ? * theorem 4.4.3 ) that the models @xmath327 and @xmath306 should be regular can be removed in our setting by the argument in ( * ? ? ? * section 2.3.1 , remark ( ii ) ) . taking @xmath417 in ( [ bigformula ] ) we find @xmath418 \langle \bar{\ll}_{i0 } \cdots \bar{\ll}_{ir}\|\xx_i \rangle\ ] ] in @xmath49 . by proposition [ pushforwarddeligne ] the class of the deligne pairing @xmath419 along @xmath379 in @xmath420 is equal to @xmath421 . hence by taking @xmath422 in ( [ bigformula ] ) we obtain the identity @xmath423 in @xmath49 . we obtain the identities from the proposition by dividing both sides in both identities above by @xmath424 and taking limits as @xmath312 . we continue to work with the number field @xmath42 . we recall that we have fixed a collection of absolute values @xmath203 on all @xmath199 for @xmath57 satisfying the product formula . let @xmath8 be an abelian variety over @xmath42 and let @xmath10 be a symmetric line bundle on @xmath8 . the following proposition follows directly from ( * ? ? ? * theorem 2.2 ) . [ basicadmissible ] let @xmath425 be an integer . let @xmath426^\ll$ ] be an isomorphism of line bundles . then there exists a unique integrable adelic metric @xmath427 on @xmath10 such that for each @xmath57 the isomorphism @xmath428 is an isometry . this adelic metric is independent of the choice of @xmath335 . if @xmath428 is changed into @xmath429 for some @xmath430 then for all @xmath57 the metric @xmath209 changes into @xmath431 . we call @xmath427 the admissible metric associated to @xmath428 . we deduce from proposition [ basicadmissible ] that if @xmath10 is ample , the associated height @xmath432 is independent of the choice of @xmath335 and @xmath428 . we therefore just denote this height by @xmath433 . the height @xmath13 coincides with the nron - tate height of @xmath11 as defined in @xcite and @xcite as explained in ( ? ? ? * section 3.1 ) . in particular , for all @xmath434 and all integral cycles @xmath11 on @xmath8 we have @xmath16_z)=n^2\h'_\ll(z)$ ] , and moreover we have the fundamental inequality @xmath17 . we mention that for @xmath61 the smooth hermitian line bundles @xmath435 have a translation - invariant curvature form . for the remainder of this section , let @xmath3 be a smooth projective geometrically connected curve defined over a number field @xmath42 , and let @xmath27 be the jacobian of @xmath3 . we will work with a symmetric line bundle @xmath10 on @xmath2 defining the principal polarization @xmath40 . we will assume moreover that @xmath10 is rigidified at the origin . the rigidification determines a unique choice of isomorphism @xmath436^\ll$ ] and hence a canonical structure of admissible adelic line bundle @xmath221 on @xmath10 . the construction ( depending on a choice of semi - canonical divisor @xmath437 on @xmath3 ) goes as follows . for every @xmath438 we first consider the line bundle @xmath439}^\theta ) \otimes e^\oo_j(t_{[d]}^\theta)^{\otimes -1}\ ] ] on @xmath2 , where @xmath440 is the canonical theta divisor of @xmath3 consisting of the classes of effective divisors , where @xmath441 } \colon j=\mathrm{pic}^0 x \to \mathrm{pic}^{g-1 } x$ ] is the translation along the divisor class @xmath442 $ ] , and @xmath443 is the origin of @xmath2 . note that each line bundle @xmath444 is canonically rigidified at the origin . moreover each @xmath444 is ample and induces the principal polarization @xmath445 of @xmath2 . assume that @xmath437 is a semi - canonical divisor on @xmath3 . then the associated @xmath444 is symmetric , and we will take @xmath10 to be @xmath444 . for each @xmath107 recall the morphism @xmath446 given by sending @xmath447 to the class of @xmath448 . for every line bundle @xmath449 on @xmath3 there exist ( possibly after replacing @xmath42 by a finite extension ) an integer @xmath450 and a divisor @xmath107 such that @xmath451 is isomorphic , as a line bundle , to @xmath452 ( in fact , note that @xmath451 is isomorphic to the line bundle @xmath453 on @xmath3 ) . given such @xmath450 and @xmath7 together with an isomorphism @xmath454 , pullback along @xmath455 and transporting structure through the isomorphism yields a structure of integrable adelic metric on @xmath449 . varying @xmath450 , @xmath7 and the isomorphism @xmath454 , we call the resulting metrics on @xmath449 admissible metrics on @xmath449 . if @xmath456 and @xmath457 are admissible line bundles on @xmath3 , then so is their tensor product @xmath458 . moreover , the admissible metrics on the trivial line bundle @xmath459 are all constant . assume that @xmath3 has semistable reduction over @xmath42 . an alternative way to obtain adelic metrics on all line bundles @xmath449 on @xmath3 is described in the paper @xcite . for example , for each @xmath460 the line bundle @xmath461 is endowed with a canonical adelic metric @xmath462 obtained by using green s functions . more precisely , if @xmath62 let @xmath463 denote the reduction map to the dual graph @xmath464 of the reduction of @xmath3 at @xmath60 , and let @xmath465 denote zhang s green s function on @xmath464 . then for all @xmath466 we have @xmath467 . if @xmath61 let @xmath465 denote the arakelov - green s function on the compact riemann surface @xmath359 . then for all @xmath468 we have @xmath469 . we shortly denote the resulting adelic line bundle on @xmath3 by @xmath470 . the dualizing sheaf @xmath471 is also endowed with a canonical adelic metric , denoted in this paper by @xmath44 . importantly , for each @xmath460 the canonical residue map @xmath472 is an isometry of adelic line bundles on @xmath473 . it is stated in ( ? ? ? * section 4.7 ) and worked out in more detail in @xcite that the adelic metrics on line bundles on @xmath3 obtained by this second method are precisely the admissible metrics on line bundles on @xmath3 described above by using pullbacks of @xmath221 from the jacobian . let @xmath474 be the group law of @xmath2 , and let @xmath475 for @xmath476 denote the projections onto the two factors . we set @xmath477 equipped with the integrable adelic metric determined by the pullback admissible adelic metrics from @xmath221 along @xmath5 and the @xmath478 . the line bundle @xmath479 is canonically rigidified , and the rigidification is an adelic isometry . [ delpairingbiext ] let @xmath173 be a projective geometrically connected variety over the number field @xmath42 . let @xmath480 denote integrable adelic line bundles on @xmath317 that are of relative degree zero for the canonical projection @xmath481 . then the deligne pairings @xmath482 and @xmath483 are integrable adelic line bundles on @xmath173 . assume that @xmath480 are fiberwise admissible . let @xmath484 for @xmath485 be the canonical map determined by the line bundle @xmath486 and write @xmath487 . then there exist isometries @xmath488 of integrable adelic line bundles on @xmath173 . the first statement follows from proposition [ projii ] . as to the isometries , without loss of generality we may assume that @xmath3 has a rational point @xmath489 . it suffices to consider the universal case where @xmath490 and @xmath32 is the identity map , and the line bundles @xmath491 are the universal line bundles on @xmath492 rigidified along @xmath96 . let @xmath493 be the poincar bundle on @xmath494 , and consider @xmath495 as the functor of rigidified line bundles on @xmath2 with vanishing chern class . let @xmath496 be a test scheme over @xmath42 and let @xmath497 . then @xmath498)^\pp \\ & = x^ t_y^\ll \otimes x^\ll^{\otimes -1 } \otimes y^ * \ll^{\otimes -1 } \\ & = ( x+y)^ * \ll \otimes x^\ll^{\otimes -1 } \otimes y^ \ll^{\otimes -1 } \\ & = ( x , y)^\bb \ , , \end{split}\ ] ] so that we have a canonical isomorphism of line bundles @xmath499 on @xmath500 . from ( section 2.9 ) we obtain a canonical isomorphism of line bundles @xmath501 over @xmath500 . we claim that the induced isomorphism @xmath502 can be taken to be an isometry of adelic line bundles . note that the induced metrics on the squares @xmath503 and @xmath504 are again fiberwise admissible . it follows that we have isometries @xmath505^ * { \langle \hat{\mm}_0 , \hat{\mm}_1 \rangle } { \xrightarrow{\sim}}{\langle \hat{\mm}_0^{\otimes 2 } , \hat{\mm}_1^{\otimes 2 } \rangle } { \xrightarrow{\sim}}{\langle \hat{\mm}_0 , \hat{\mm}_1 \rangle}^{\otimes 4 } \ , .\ ] ] we also have an isometry @xmath506^ * \hat{\bb } { \xrightarrow{\sim}}\hat{\bb}^{\otimes 4}$ ] , hence by uniqueness we obtain an isometry @xmath507 . to obtain an isometry @xmath508 we specialize the isometry just established to the diagonal in @xmath500 . let @xmath509 denote the projection onto the first two factors , and let @xmath510 denote the tautological sections of this projection . these give rise to a canonical relative degree zero line bundle @xmath511 on @xmath512 over @xmath103 . we have canonical isomorphisms of line bundles @xmath513 where @xmath514 denotes the difference map @xmath105 $ ] . from the given admissible structure on @xmath10 and the canonical admissible line bundle @xmath44 we then obtain an integrable structure @xmath515 on @xmath516 . as it turns out , this structure does not depend on the choice of semi - canonical divisor @xmath437 , and hence is canonical . we refer to ( * ? ? ? * section 3.5 ) for a slightly different approach . by restriction to horizontal or vertical slices @xmath517 or @xmath518 of @xmath103 , where @xmath460 , the integrable line bundle @xmath515 yields the canonical admissible line bundle @xmath470 . finally we have that the canonical isomorphism @xmath519 given by adjunction induces an isometry @xmath520 of adelic line bundles on @xmath3 . in this section we define the local @xmath163-invariants appearing in theorem [ mainintro ] , following @xcite , and discuss some of their properties . we start by discussing the non - archimedean @xmath163-invariant . we say that a metrized graph is a compact connected metric space @xmath142 such that for each @xmath150 there exist @xmath434 and @xmath521 such that @xmath143 has an open neighborhood isometric to the star - shaped set @xmath522 endowed with the path metric . let @xmath142 be a metrized graph . a divisor on @xmath142 is an element of @xmath523 ; note that a divisor has naturally a degree in @xmath202 . let @xmath126 be an integer . a polarization on @xmath142 is a divisor on @xmath142 of degree @xmath524 . we then call @xmath77 the genus of the polarized metrized graph @xmath525 . when @xmath142 is a metrized graph , then from ( [ crmeasure ] ) we have a canonical measure @xmath149 on @xmath142 . assume that a polarization @xmath166 is given so that @xmath525 has genus @xmath77 . then , following @xcite , one considers the admissible measure @xmath526 on @xmath527 given by the identity @xmath528 the green s function @xmath529 associated to @xmath526 is the unique piecewise quadratic function on @xmath530 uniquely determined by the two conditons @xmath531 for all @xmath150 . let @xmath532 denote the total length of @xmath142 . then we set @xmath533 the invariant @xmath534 is closely related to the epsilon - invariant @xcite @xcite of @xmath527 , @xmath535 now let @xmath3 be a smooth projective geometrically connected curve of genus @xmath126 with semistable reduction over a number field @xmath42 . let @xmath62 and let @xmath464 be the dual graph of the reduction of @xmath3 at @xmath60 . then by ( * ? ? ? * section 2.2 ) ( * ? ? ? * section 2.1 ) the metrized graph @xmath464 comes with a canonical polarization , resulting in a polarized metrized graph @xmath536 of genus @xmath77 . we then set @xmath537 . likewise we set @xmath538 . we have @xmath58 for @xmath59 , and also @xmath58 if @xmath3 has good reduction at @xmath60 . in general we have @xmath539 . in fact , we have a more precise lower bound . let @xmath78 denote the number of non - separating geometric double points on the reduction of @xmath3 at @xmath60 , and let @xmath79 for @xmath80 $ ] denote the number of geometric double points on the reduction of @xmath3 at @xmath60 whose local normalization has two connected components , one of arithmetic genus @xmath81 , and one of arithmetic genus @xmath82 . in @xcite zhang conjectured , and proved in the special case of so - called elementary graphs , for @xmath45 a lower bound @xmath540 } \frac{2i(g - i)}{g}\delta_i(x_v ) \ , , \ ] ] where @xmath76 is some positive constant depending only on @xmath77 . zhang s conjecture was proved in low genera @xmath541 by x. faber ( * ? ? ? * theorem 3.4 ) , and by z. cinkir ( * ? ? ? * theorem 2.11 ) in general . cinkir showed that one can even take @xmath542 and @xmath543 for @xmath544 , but these values can most likely be improved . a list of the non - archimedean @xmath163-invariants of all polarized metrized graphs in genus two is provided in @xcite , and for all polarized metrized graphs in genus three in @xcite . the case of polarized metrized graphs arising as reduction graphs of semistable hyperelliptic curves is completely understood by yamaki s work @xcite . we next discuss the archimedean @xmath163-invariant . this invariant has been independently introduced and studied by n. kawazumi in @xcite @xcite ( for the invariant @xmath96 defined in @xcite @xcite we have @xmath545 ) . let @xmath546 be a compact and connected riemann surface of genus @xmath126 . let @xmath547 for @xmath548 be the normalized real eigenforms of the arakelov laplacian @xcite on @xmath546 and let @xmath549 be its positive eigenvalues . let @xmath550 be an orthonormal basis for the hermitian inner product @xmath551 on the space of holomorphic @xmath39-forms on @xmath546 . then we set @xmath552 this invariant is well - defined and independent of the choice of @xmath550 . if @xmath3 is a smooth projective geometrically connected curve of genus @xmath126 defined over a number field @xmath42 , and @xmath61 , we shortly write @xmath56 for @xmath553 . note that we clearly have @xmath554 . in fact we have @xmath555 if @xmath59 , and @xmath556 else , cf . * corollary 1.2 ) or ( * ? ? ? * remark following proposition 2.5.3 ) . a fast method to compute @xmath557 when @xmath558 is described in @xcite . let @xmath27 denote the jacobian of @xmath546 . let @xmath10 be a symmetric ample line bundle on @xmath2 inducing the polarization @xmath40 , endowed with an admissible metric @xmath559 ( that is , a smooth hermitian metric whose curvature form is translation - invariant ) . let @xmath131 be any non - zero global section of @xmath10 . then following @xcite we put @xmath560 where @xmath561 is the haar measure on @xmath2 . we note that @xmath562 is independent of the choice of @xmath10 and of @xmath131 , moreover we have @xmath563 by jensen s inequality . a fundamental identity relating @xmath557 with the faltings delta - invariant @xmath564 from @xcite and the invariant @xmath562 has recently been found by r. wilms in his paper @xcite . [ wilmsrewrite ] ( r. wilms ) let @xmath564 be the faltings delta - invariant of the compact riemann surface @xmath546 . let @xmath565 . then the identity @xmath566 holds . this is the main result of @xcite , but slightly rewritten . let @xmath567 be the riemann theta - function of @xmath27 , seen as a global section of a suitable symmetric ample line bundle @xmath10 representing @xmath40 . endow @xmath10 with its standard admissible metric giving @xmath567 the norm @xmath568 defined on @xcite . let then @xmath569 . then ( * ? ? ? * theorem 1 ) states that @xmath570 from ( * ? ? ? * proposition 2.5.6 ) it follows that @xmath571 . this gives @xmath572 the required identity follows by combining ( [ wilmss ] ) and ( [ iandh ] ) . now , let @xmath3 be a smooth projective geometrically connected curve of genus @xmath126 with semistable reduction over a number field @xmath42 . let @xmath27 denote its jacobian . we have the following arithmetic consequence of wilms s result . [ corofwilms ] let @xmath573 denote the stable faltings height of @xmath2 , and let @xmath50 denote the self - intersection of the canonical admissible relative dualizing sheaf on @xmath3 . then the identity @xmath574 \ , \h_f(j ) = & \,\,\frac{1}{12 } { \langle\hat{\omega},\hat{\omega}\rangle } + \frac{1}{12 } \sum_{v \in m(k)_0 } ( \delta(x_v)+\vareps(x_v))\log nv - \kappa_0 g \ , [ k:\qq]\\ & + 2 \sum_{v \in m(k)_\infty } i(j_v,\lambda_v ) + \frac{1}{6 } \sum_{v \in m(k)_\infty } \varphi(x_v ) \end{split}\ ] ] holds . let @xmath575 be the usual self - intersection of the relative dualizing sheaf of the minimal regular model of @xmath3 over the ring of integers of @xmath42 . the noether formula ( * ? ? ? * theorem 6 ) ( * ? ? ? * thorme 2.5 ) then reads @xmath576 \ , \h_f(j ) = & \,\ , { \langle\bar{\omega},\bar{\omega}\rangle } + \sum_{v \in m(k)_0 } \delta(x_v)\log nv \\ & + \sum_{v \in m(k)_\infty } \delta_f(x_v ) -4g\,[k:\qq]\,\log(2\pi ) \ , . \end{split}\ ] ] by ( * ? ? ? * theorem 5.5 ) the non - archimedean epsilon - invariants give the difference between @xmath575 and @xmath50 , more precisely we have @xmath577 we find that @xmath576 \ , \h_f(j ) = & \,\ , { \langle\hat{\omega},\hat{\omega}\rangle } + \sum_{v \in m(k)_0 } ( \delta(x_v)+\vareps(x_v))\log nv \\ & + \sum_{v \in m(k)_\infty } \delta_f(x_v ) -4g\,[k:\qq]\,\log(2\pi ) \ , . \end{split}\ ] ] combining with theorem [ wilmsrewrite ] we obtain the identity stated in the corollary . we obtain theorem [ thetaheight ] in the case of potentially everywhere good reduction . [ realcatalyzer ] assume that @xmath2 has everywhere good reduction . then the identity @xmath124 \ , \h'_\ll(\theta ) = \frac{1}{12 } { \langle\hat{\omega},\hat{\omega}\rangle } + \frac{1}{6 } \sum_{v \in m(k ) } \varphi(x_v ) \log nv\ ] ] holds . recall that autissier s result ( [ aut ] ) states that @xmath578 } \sum_{v \in m(k)_\infty } i(j_v,\lambda_v ) \log nv \ , .\ ] ] on the other hand , applying corollary [ corofwilms ] we have @xmath579 \ , \h_f(j ) = \frac{1}{12 } { \langle\hat{\omega},\hat{\omega}\rangle } - \kappa_0 g\,[k:\qq ] + 2 \sum_{v \in m(k)_\infty } i(j_v,\lambda_v ) + \frac{1}{6 } \sum_{v \in m(k ) } \varphi(x_v)\log nv \ , .\ ] ] we find the stated identity by combining these two equations . in section [ proofautforjac ] we will use this special case to prove theorem [ thetaheight ] in general . in this section we derive from proposition [ delpairingbiext ] an explicit formula for the nron - tate height of a tautological cycle @xmath1 on a jacobian , in terms of admissible adelic intersection theory . let @xmath126 be an integer and let @xmath3 be a smooth projective geometrically connected curve of genus @xmath77 with semistable reduction over a number field @xmath42 . let @xmath6 be an integer satisfying @xmath29 , and let @xmath30 be an @xmath6-tuple of non - zero integers and let @xmath437 be a divisor on @xmath3 with @xmath580 . let @xmath27 denote the jacobian of @xmath3 . consider the map @xmath581 \ , , \ ] ] and denote by @xmath582 the image of @xmath32 . we note that the map @xmath32 is generically finite . for @xmath583 let @xmath584 denote the projection on the @xmath81-th factor , and let @xmath585 denote the projection on the @xmath81-th and @xmath272-th factor . let @xmath44 denote the canonical line bundle @xmath471 of @xmath3 equipped with its canonical admissible adelic metric , and let @xmath515 denote the line bundle associated to the diagonal @xmath586 on @xmath103 equipped with its canonical integrable metric , both as discussed at the end of section [ admissible ] . denote by @xmath587 the line bundle @xmath588 on @xmath3 endowed with its canonical admissible metric . write @xmath589 , @xmath590 and @xmath591 . let @xmath10 be a symmetric ample line bundle on @xmath2 defining the principal polarization @xmath40 . [ main ] each of @xmath592 , @xmath593 and @xmath594 is an integrable adelic line bundle on @xmath4 , and the identity @xmath595}\ ] ] holds , where @xmath596 is the degree of @xmath32 . denote by @xmath597 the projection on the first @xmath6 coordinates , with tautological sections @xmath598 given by @xmath599 for @xmath600 . we denote by @xmath601 the adelic metrized line bundle determined by the image of @xmath602 obtained by taking the fiberwise canonical admissible metric along @xmath192 . note that @xmath603 , which implies that the @xmath601 are integrable . let @xmath604 denote projection on the last coordinate . consider the deligne pairing @xmath605 along the projection @xmath597 . note that @xmath606 , and that @xmath607 and @xmath608 , canonically . thus , by expanding brackets we find @xmath609 canonically , as integrable adelic line bundles on @xmath4 . let @xmath221 denote the line bundle @xmath10 , endowed with any admissible adelic metric . then by proposition [ delpairingbiext ] we have an isometry @xmath610 of adelic line bundles . by the projection formula from proposition [ projection ] we have @xmath611\h'_\ll(z ) \ , . \end{split}\ ] ] we obtain the required formula by combining ( [ longm ] ) and ( [ rewriteheight ] ) . we illustrate ( the proof of ) theorem [ main ] with a few examples that already exist in the literature . first , take @xmath612 and @xmath437 any divisor of degree zero . then theorem [ main ] gives for the point @xmath442 \in j$ ] that @xmath613 \,\h'_\ll([d ] ) = -{\langle\hat{d},\hat{d}\rangle } \ , .\ ] ] this is precisely the well - known faltings - hriljac formula or hodge index theorem ( * ? ? ? * theorem 4 ) ( * ? ? ? * theorem 3.1 ) ( * ? ? ? * section 5.4 ) . here the adelic intersection product @xmath614 is taken on the curve @xmath3 . assume that @xmath45 . let @xmath89 , and write @xmath615 . let @xmath47 denote the canonical divisor class of @xmath3 . taking @xmath616 a representative of the class @xmath46 we find from ( [ faltingshr ] ) , with all adelic intersection products taken on the curve @xmath3 , @xmath617\,\h'_\ll(x_\alpha ) & = -{\langle\hat{\alpha},\hat{\alpha}\rangle } + \frac{1}{g-1}{\langle\hat{\alpha},\hat{\omega}\rangle } - \frac{1}{(2g-2)^2}{\langle\hat{\omega},\hat{\omega}\rangle } \\ & = \frac{1}{2g-2}{\langle\hat{\beta},\hat{\omega}\rangle } - \frac{1}{(2g-2)^2}{\langle\hat{\omega},\hat{\omega}\rangle } \ , . \end{split}\ ] ] denote by @xmath131 the section @xmath618 of the projection on the first coordinate @xmath619 . then for @xmath620 we have @xmath621 canonically , as integrable adelic line bundles on @xmath3 . we compute , noting that @xmath622 and using ( [ htpoint ] ) , @xmath623\,\h_\ll'(x_\alpha ) \\ & = \frac{g}{g-1}{\langle\hat{\omega},\hat{\omega}\rangle } + 8(g-1)\,[k:\qq]\ , \h'_\ll(x_\alpha ) \ , . \end{split}\ ] ] theorem [ main ] then gives , noting that @xmath624 , @xmath625}{\langle\hat{\omega},\hat{\omega}\rangle } + 8(g-1 ) \h'_\ll(x_\alpha)\right ) \\ & = \frac{1}{8(g-1)[k:\qq ] } { \langle\hat{\omega},\hat{\omega}\rangle } + \frac{g-1}{g}\h'_\ll(x_\alpha ) \ , . \end{split}\ ] ] this recovers ( * ? ? ? * theorem 3.9 ) . [ degrees ] in general we have @xmath626 . indeed , let @xmath122 be a divisor with @xmath627 , then in cohomology we have @xmath628 = \frac{1}{(g - r ) ! } \bigwedge^{g - r } [ \theta]\ ] ] by poincar s formula ( see for instance ( * ? ? ? * section 11.2 ) ) . it follows that @xmath629 we conclude by noting that @xmath630 by the riemann - roch theorem . we continue with the notation from the previous section , and assume throughout that @xmath45 . in order to prove theorem [ mainintro ] we will apply theorem [ main ] with @xmath631 , where @xmath632 . as above , write @xmath615 . let @xmath633 . following theorem [ main ] , our aim is to write @xmath634 as a rational linear combination of @xmath50 , @xmath635 and @xmath69\h'_\ll(x_\alpha)$ ] with coefficients that only depend on @xmath5 and @xmath77 . by equation ( [ htpoint ] ) we may as well aim for a rational linear combination of @xmath50 , @xmath635 and @xmath636 with coefficients that only depend on @xmath5 and @xmath77 . by the following proposition , we may then as well aim for a rational linear combination of @xmath50 , @xmath637 and @xmath636 with coefficients that only depend on @xmath5 and @xmath77 , where @xmath637 is the triple self - product of the canonical integrable bundle @xmath515 on @xmath103 . [ triplediagonal ] let @xmath637 denote the triple self - product of the integrable line bundle @xmath515 on @xmath103 . then the identity @xmath638 holds in @xmath49 . we apply the global recursive formula ( [ global_recursive ] ) to the morphism @xmath639 and the section @xmath39 of @xmath516 on @xmath103 . we find @xmath640 now we recall from ( [ adjunction ] ) that @xmath641 so that @xmath642 . by ( * proposition 2.5.3 ) for @xmath61 and ( * ? ? ? * lemma 3.5.4 ) for @xmath62 we have @xmath643 the result follows . consider the set @xmath644 of integrable adelic line bundles on @xmath4 . we define a map @xmath645 by putting @xmath646 for @xmath647 , @xmath648 . we obtain @xmath649 we claim that each @xmath650 where @xmath651 is a rational multiple of @xmath50 , @xmath637 or @xmath636 with a coefficient that only depends on @xmath77 . this suffices to prove theorem [ mainintro ] . in order to prove our claim , we will use a graphical calculus that is very similar to a method used in r. wilms s article @xcite . we refer to @xcite @xcite for variants of this calculus in the context of the tautological ring on the moduli space of curves . first of all , it is convenient to slightly generalize the set @xmath652 as defined above . let @xmath653 be a finite set . we have natural projections @xmath654 for @xmath655 and @xmath656 for @xmath657 , @xmath658 . we define the set @xmath659 consisting of integrable line bundles on @xmath660 obtained by pulling back from @xmath3 , @xmath103 along @xmath478 , @xmath661 . we say that an undirected graph @xmath662 on @xmath91 ( loops and multiple edges allowed ) is loop - labelled if each loop @xmath450 in @xmath663 has been assigned an element @xmath664 . we say that @xmath142 is edge - ordered if some linear ordering of @xmath663 has been chosen . let @xmath662 be a loop - labelled edge - ordered graph with edges @xmath665 . then to @xmath142 we associate a tuple @xmath666 as follows : for each @xmath667 we define the integrable line bundle @xmath668 to be the integrable line bundle @xmath669 if @xmath670 is a loop based at @xmath655 and @xmath671 , and to be the integrable line bundle @xmath590 if @xmath670 is an edge between @xmath81 and @xmath272 . conversely , given a tuple @xmath672 one has naturally associated to it a loop - labelled edge - ordered graph with @xmath337 edges on @xmath91 , that we denote by @xmath673 . we define the intersection number @xmath674 of a loop - labelled edge - ordered graph @xmath662 to be the arithmetic intersection number @xmath675 if @xmath666 is the tuple associated to @xmath142 . note that the intersection number does not depend on the actual ordering of the edges , and hence yields an invariant of merely loop - labelled graphs . [ multiplicative ] the intersection number of a loop - labelled graph @xmath142 is multiplicative over connected components of @xmath142 . this follows immediately from proposition [ externalproducts ] . let @xmath142 be a connected loop - labelled graph on a vertex set @xmath653 . a contraction of @xmath142 is the connected loop - labelled graph obtained by performing one of the following operations on @xmath142 : ( i ) contract an edge emanating from a vertex of degree one ; ( ii ) if vertex @xmath272 of @xmath142 has degree @xmath676 with emanating edges @xmath677 and @xmath678 with @xmath679 , then replace the edges @xmath677 , @xmath678 by one edge @xmath680 ; ( iii ) if vertex @xmath81 of @xmath142 has degree @xmath676 with both emanating edges connecting @xmath81 and @xmath272 , and @xmath681 , then replace the two edges by one loop attached at @xmath272 and labelled @xmath682 . [ contractions ] the intersection number of a connected loop - labelled graph is invariant under contractions . we consider each of the three types of contraction separately . after relabeling the vertices we may assume that @xmath683 . in case ( i ) , after reordering the vertices we may assume that vertex @xmath39 has degree one and that the unique edge emanating from it is @xmath684 . the following observation then proves the assertion in this case : let @xmath685 be the projection on the first two coordinates , and let @xmath686 be the projection forgetting the first coordinate . let @xmath687 be integrable adelic line bundles on @xmath688 . then as @xmath689 we find by proposition [ projii ] that @xmath690 in case ( ii ) it suffices by proposition [ projii ] to verify the following : let @xmath691 be the projection on the first and third coordinate . then for the deligne pairing of @xmath692 and @xmath693 along @xmath694 one has an isometry @xmath695 of adelic line bundles . this assertion follows from the recursive formula ( [ global_recursive ] ) applied to @xmath694 and the divisor @xmath696 on @xmath512 . for each @xmath57 the integral in ( [ global_recursive ] ) vanishes . in case ( iii ) , after reordering the vertices we may assume that vertex @xmath39 has degree two and that both edges emanating from it end at vertex @xmath676 . the following observation then proves the assertion in this case : let @xmath685 be the projection on the first two coordinates , and let @xmath686 be the projection forgetting the first coordinate . let @xmath697 be integrable adelic line bundles on @xmath688 . then by proposition [ projii ] we have @xmath698 where the deligne pairing is along @xmath699 . next we have @xmath700 by the recursive formula ( [ global_recursive ] ) applied to a projection @xmath701 and the rational section @xmath39 of @xmath516 on @xmath103 . note that for each @xmath57 the integral in ( [ global_recursive ] ) vanishes . from the adjunction formula ( [ adjunction ] ) we then finally find that @xmath702 . observe that under a contraction both the number of edges and the number of vertices of a connected loop - labelled graph decrease by one . in particular , the euler characteristic @xmath703 remains preserved , and after finitely many steps no more contractions are possible . a connected loop - labelled graph on which no contractions are possible is called minimal . a minimal graph either consists of one vertex ( these are the minimal graphs with @xmath704 ) , or of one vertex with one loop attached to it ( these are the minimal graphs with @xmath705 ) , or has each vertex of degree at least @xmath706 ( and @xmath707 ) . in the latter case , if we let @xmath60 denote the number of vertices of @xmath142 , and @xmath450 the number of edges , then @xmath708 and hence @xmath709 . in particular , if @xmath710 , then @xmath711 . we call a loop - labelled graph @xmath142 relevant if all connected components of @xmath142 have their euler characteristic @xmath712 , and @xmath710 occurs for precisely one connected component . recall that our task is to show that for each tuple @xmath713 the intersection number @xmath714 is a rational multiple of @xmath50 , @xmath637 or @xmath636 with a coefficient that only depends on @xmath77 . note that the associated graph @xmath142 of such a tuple @xmath715 has @xmath6 vertices and @xmath716 edges , hence has @xmath717 . our task would thus be accomplished once we show that the intersection number of a loop - labelled graph @xmath142 with @xmath717 is a rational multiple of @xmath50 , @xmath637 or @xmath636 with a coefficient that only depends on @xmath77 . hence , we are done by the proposition below . let @xmath142 be a loop - labelled graph with @xmath717 . then the intersection number @xmath674 vanishes unless @xmath142 is relevant . assume @xmath142 is relevant . then @xmath674 is a product of powers of @xmath718 , @xmath676 and @xmath524 times one of the intersection numbers @xmath719 . assume that @xmath142 is not relevant , that is , some component of @xmath142 has @xmath720 . as @xmath717 there is in fact a component with @xmath704 . this component is a tree , which is contractible to a point , and hence has vanishing intersection product . by proposition [ multiplicative ] , the intersection product of @xmath142 vanishes . assume that @xmath142 is relevant , and let @xmath721 be a connected component of @xmath142 . by proposition [ multiplicative ] , it suffices to prove that @xmath722 is either @xmath723 , @xmath676 , @xmath524 or plus or minus one of the intersection numbers @xmath724 . by proposition [ contractions ] , we may assume without loss of generality that @xmath721 is minimal . if @xmath725 , then @xmath721 consists of one vertex with one loop attached to it . if @xmath726 is the label of the unique loop of @xmath721 we have @xmath727 , in particular @xmath728 . if @xmath729 then as we saw above @xmath721 has at most two vertices and hence @xmath721 is one of the following three loop - labelled graphs : ( -1,0 ) circle [ radius=0.5 ] ; ( 1,0 ) circle [ radius=0.5 ] ; ( -0.5,0 ) node[circle , draw , fill = black , inner sep=0pt , minimum width=4pt ] ( 0.5,0 ) node[circle , draw , fill = black , inner sep=0pt , minimum width=4pt ] ; at ( -1.75,0 ) @xmath726 ; at ( 1.75,0 ) @xmath730 ; at ( 0,0 ) ; or ( -0.5,0 ) circle [ radius=0.5 ] ; ( 0.5,0 ) circle [ radius=0.5 ] ; ( 0,0 ) node[circle , draw , fill = black , inner sep=0pt , minimum width=4pt ] ; at ( -1.25,0 ) @xmath726 ; at ( 1.25,0 ) @xmath730 ; or ( 0,0 ) circle [ radius=0.75 ] ; ( -0.75,0 ) node[circle , draw , fill = black , inner sep=0pt , minimum width=4pt ] ( 0.75,0 ) node[circle , draw , fill = black , inner sep=0pt , minimum width=4pt ] ; at ( 0.75,0.75 ) ; at ( 0,0 ) ; at ( 0.75,-0.75 ) ; where @xmath731 . in the first two cases , we have @xmath732 so that @xmath733 , up to sign . in the third case , we have @xmath734 . this proves the proposition . let @xmath3 be a smooth projective geometrically connected curve of genus @xmath45 with semistable reduction over a number field @xmath42 . let @xmath27 denote its jacobian . in this section we illustrate theorem [ mainintro ] by considering the abel - jacobi images of @xmath103 as well as the difference surface @xmath735 inside @xmath2 . let @xmath107 , and as before let @xmath736 be shorthand for the line bundle @xmath737 on @xmath3 endowed with its canonical admissible adelic metric . recall that we put @xmath615 . we shall frequently make use of the following observations , that easily follow from the graphical calculus discussed in the previous section . let @xmath738 be any of the admissible line bundles @xmath44 , @xmath739 on @xmath3 , and let @xmath740 denote the two projections . for @xmath741 let @xmath742 denote the pullbacks of @xmath738 along @xmath478 . then we have the relations @xmath743 and @xmath744 next , a triple intersection @xmath745 vanishes if @xmath746 . on the other hand if @xmath747 and @xmath42 is different from @xmath81 and @xmath272 then we have @xmath748 let @xmath108 be the image of @xmath103 in @xmath2 under the map @xmath749 . by remark [ degrees ] we have @xmath750 . let @xmath751 be the image of @xmath103 under @xmath752 . the homology classes @xmath753 $ ] and @xmath754 $ ] in @xmath755 are equal , and hence @xmath756 as well . [ heightdifferencesurface ] the equality @xmath757\,\h'_\ll(f ) = ( 3g-1){\langle\hat{\omega},\hat{\omega}\rangle } - 2 \sum_{v \in m(k ) } \varphi(x_v ) \log nv\ ] ] holds . let @xmath758 be the adelic line bundle @xmath759 on @xmath103 . using that @xmath756 we obtain from theorem [ main ] that @xmath760 \,\h'_\ll(f ) = \mhat^3 \ , .\ ] ] we calculate , using proposition [ triplediagonal ] and the rules stated above , @xmath761 the required formula follows . [ square ] the equality @xmath762\,\h'_\ll(z_{2,\alpha } ) \\ & = \frac{3g^2 - 8g-1}{g-1}{\langle\hat{\omega},\hat{\omega}\rangle } + 2 \sum_{v \in m(k ) } \varphi(x_v)\log nv + 48(g-1)(g-2 ) \,\h'_\ll(x_\alpha ) \end{split}\ ] ] holds . let @xmath758 be the adelic line bundle @xmath763 on @xmath103 . using that @xmath764 we obtain from theorem [ main ] ( with @xmath765 ) that @xmath766 \,\h'_\ll(z_{2,\alpha } ) = \mhat^3 \ , .\ ] ] it thus suffices to show that @xmath767\,\h'_\ll(x_\alpha ) \ , . \end{split}\ ] ] we recall from equation ( [ htpoint ] ) that @xmath768 \,\h'_\ll(x_\alpha)=\frac{1}{2g-2}{\langle\hat{\beta},\hat{\omega}\rangle } - \frac{1}{(2g-2)^2}{\langle\hat{\omega},\hat{\omega}\rangle } \ , .\ ] ] inserting this in ( [ aux1 ] ) we are reduced to proving that @xmath769 write @xmath770 then we have @xmath771 we compute each of the terms in this expression . from proposition [ triplediagonal ] we recall that @xmath772 next we have @xmath773 and , using that @xmath774 , @xmath775 finally we have @xmath776 summing all contributions with weights as in ( [ mhatexpanded ] ) we find ( [ aux2 ] ) as required . we finish this section by showing that we can write the height of the gross - schoen cycle essentially as the difference of the nron - tate heights of @xmath735 and @xmath108 . let @xmath107 . then we have @xmath777\,(\h'_\ll(f)-\h'_\ll(z_{2,\alpha } ) ) + 12(g-1)^2\,[k:\qq]\,\h'_\ll(x_\alpha ) \ , .\ ] ] theorem [ square ] and theorem [ heightdifferencesurface ] together give @xmath778(\h'_\ll(f)-\h'_\ll(z_{2,\alpha } ) ) \\ & = 4\left ( \frac{2g+1}{2g-2}{\langle\hat{\omega},\hat{\omega}\rangle } - \sum_{v \in m(k ) } \varphi(x_v)\log nv \right ) - 48(g-1)(g-2)\ , [ k:\qq]\,\h'_\ll(x_\alpha ) \ , . \end{split}\ ] ] using zhang s formula ( [ htgs ] ) we find @xmath779(\h'_\ll(f)-\h'_\ll(z_{2,\alpha } ) ) = 4 { \langle\delta_\alpha,\delta_\alpha\rangle } - 48(g-1)^2 \,[k:\qq]\,\h'_\ll(x_\alpha ) \ , .\ ] ] the result follows . assume that @xmath780 is a canonical divisor . then @xmath777\,(\h'_\ll(f)-\h'_\ll(z_{2,\alpha } ) ) \ , .\ ] ] in particular , the equivalence @xmath781 holds . when @xmath3 is hyperelliptic , we have @xmath782 and we obtain another proof that @xmath783 in this case . let @xmath3 be a smooth projective geometrically connected curve of genus @xmath126 with semistable reduction over a number field @xmath42 . let @xmath27 be the jacobian of @xmath3 , and let @xmath122 be a symmetric effective divisor on @xmath2 such that @xmath123 is an ample line bundle defining the principal polarization @xmath40 . note that we can write @xmath784 for a suitable @xmath107 such that @xmath780 is a canonical divisor on @xmath3 . for @xmath785 let @xmath786 denote the projection on the @xmath81-th factor , and let @xmath787 denote the projection on the @xmath81-th and @xmath272-th factor . write @xmath589 , @xmath590 and @xmath788 . we note that @xmath789 by our choice of @xmath107 we can write @xmath790 where @xmath791 is a suitable admissible line bundle on @xmath473 . then , applying theorem [ main ] with @xmath792 we find @xmath793 \ , \h'_\ll(\theta ) \\ & = \left\langle \left(\sum_{i=1}^{g-1 } \hat{\omega}_i -2\sum_{i < j}^{g-1 } \hat{\delta}_{ij } + ( 2g-2 ) \sum_{i=1}^{g-1 } \hat{\alpha}_i -(g-1)^2{\langle\hat{\alpha},\hat{\alpha}\rangle}\right)^g \bigg\| x^{g-1 } \right\rangle \\ & = \left\langle \left(2\sum_{i=1}^{g-1 } \hat{\omega}_i -2\sum_{i < j}^{g-1 } \hat{\delta}_{ij } + ( g-1 ) \hat{\mu } -(g-1)^2{\langle\hat{\alpha},\hat{\alpha}\rangle}\right)^g \bigg\| x^{g-1 } \right\rangle \\ & = 2^g \left\langle \left(\sum_{i=1}^{g-1 } \hat{\omega}_i -\sum_{i < j}^{g-1 } \hat{\delta}_{ij } \right)^g \bigg\| x^{g-1 } \right\rangle + c ' \left ( \widehat{\deg } \,\hat{\mu } - ( g-1){\langle \hat{\alpha},\hat{\alpha}\rangle } \right ) \end{split}\ ] ] for some rational number @xmath794 depending only on @xmath77 . the equation @xmath790 leads by squaring both sides to the identity @xmath795 , and we conclude that @xmath796 \ , \h'_\ll(\theta ) = \left\langle \left ( \sum_{i=1}^{g-1 } \hat{\omega}_i -\sum_{i < j}^{g-1 } \hat{\delta}_{ij}\right)^g \bigg\| x^{g-1 } \right\rangle + c \ , { \langle\hat{\omega},\hat{\omega}\rangle}\ ] ] for some rational number @xmath98 depending only on @xmath77 . now , our considerations in section [ proofmain ] leading to the proof of theorem [ mainintro ] show that there exist rational numbers @xmath797 , @xmath798 depending only on @xmath77 such that @xmath799 as it turns out , in @xcite precisely the right amount of combinatorics is carried out in order to determine the universal constant @xmath798 . more precisely , combining ( * ? ? ? * lemma 5.3.1(b ) ) and ( * ? ? ? * lemma 5.3.2 ) we find that @xmath800 ( our @xmath798 is @xmath801 times the @xmath802 from ( * ? ? ? * lemma 5.3.2 ) ) . as @xmath803 by proposition [ triplediagonal ] we find that there exists a rational number @xmath96 depending only on @xmath77 such that @xmath804 \ , \h'_\ll(\theta ) = a \ , { \langle\hat{\omega},\hat{\omega}\rangle } + \frac{1}{6 } \sum_{v \in m(k ) } \varphi(x_v ) \log nv \ , .\ ] ] specializing to a curve of genus @xmath77 over some number field @xmath42 such that the associated jacobian has everywhere good reduction ( such curves exist by ( * ? ? ? * thorme 0.7 , exemple 0.9 ) ) , and using that @xmath805 is always positive , we deduce from theorem [ realcatalyzer ] that @xmath806 . our next goal is to prove theorem [ autforjac ] . by combining corollary [ corofwilms ] and theorem [ catalyzer ] we find that @xmath807 } \sum_{v \in m(k)_0 } \frac{1}{12}\left(\delta(x_v ) + \vareps(x_v ) - 2 \,\varphi(x_v)\right ) \log nv \\ & - \kappa_0 g + \frac{2}{[k:\qq ] } \sum_{v \in m(k)_\infty } i(a_v,\lambda_v ) \ , . \end{split}\ ] ] the issue is therefore to control the contribution from the finite places in this formula . this is taken care of by the next two propositions . [ graph1 ] let @xmath525 be a polarized metrized graph . let @xmath808 be the total length and let @xmath809 be the tau - invariant ( [ tauinv ] ) of @xmath142 . write @xmath810 and @xmath811 for the @xmath163- and @xmath812-invariant of @xmath527 , cf . ( [ phiinv ] ) and ( [ epsinv ] ) . then the equality @xmath813 holds . let @xmath77 be the genus of @xmath527 . let @xmath526 be the admissible measure ( [ canandadmmeasure ] ) on @xmath142 determined by @xmath166 and let @xmath529 denote the associated green s function ( [ green ] ) . we will write the invariant @xmath814 in two different ways , and compare the outcomes . we recall from ( [ phiinv ] ) and ( [ epsinv ] ) that @xmath815 and @xmath816 we find @xmath817 let @xmath145 denote the effective resistance between points @xmath818 on @xmath142 . then we have @xmath819 by ( * ? ? ? * equation ( 3.5.1 ) ) and this yields , since @xmath820 by ( [ green ] ) , @xmath821 by ( * ? ? ? * lemma 4.1 ) we have for all @xmath150 that @xmath822 and hence upon integrating both sides against @xmath526 we find @xmath823 recall from ( [ canandadmmeasure ] ) that @xmath824 and from ( [ tauinv ] ) that @xmath825 independently of the choice of @xmath150 . combining ( [ formulac ] ) and ( [ formulaeps ] ) then gives @xmath826 combining ( [ firstsum ] ) and ( [ secondsum ] ) we find the required identity . we refer to ( * ? ? ? * equation ( 14.3 ) ) or ( * ? ? ? * equation ( 10 ) ) for the statement that @xmath829 , with equality if and only if @xmath142 is a tree . this statement can be deduced for example from the explicit formula for @xmath809 in ( * ? ? ? * proposition 2.9 ) . [ symmetric ] for each @xmath62 let @xmath830 then for each @xmath62 we have @xmath831 , we have @xmath832 if and only if @xmath2 has good reduction at @xmath60 , and the identity @xmath833 } \sum_{v \in m(k)_0 } \alpha_v \log nv- \kappa_0 g + \frac{2}{[k:\qq ] } \sum_{v \in m(k)_\infty } i(j_v,\lambda_v)\ ] ] holds in @xmath49 . a. chambert - loir , _ heights and measures on analytic spaces . a survey of recent results , and some remarks_. in : r. cluckers , j. nicaise , j. sebag ( eds . ) , motivic integration and its interactions with model theory and non - archimedean geometry , volume ii . cambridge university press 2011 . s. david , p. philippon , _ minorations des hauteurs normalises des sous - varits de varits abliennes_. in : number theory ( tiruchirapalli , 1996 ) , contemp . 210 , amer . soc . , providence , ri , 1998 , 333364 .	we develop a method to calculate the nron - tate height of tautological integral cycles on jacobians of curves defined over number fields . as examples we obtain closed expressions for the nron - tate height of the difference surface , the abel - jacobi images of the square of the curve , and of any symmetric theta divisor . as applications we obtain a new effective positive lower bound for the essential minimum of any abel - jacobi image of the curve and a proof , in the case of jacobians , of a formula proposed by autissier relating the faltings height of a principally polarized abelian variety with the nron - tate height of a symmetric theta divisor .
explaining the origin of the stellar initial mass function ( imf ) is an outstanding question before the theory of star - formation . ever since the seminal work by edwin salpeter ( salpeter 1955 ) on this subject , first suggesting a power - law mass - distribution of stellar masses for stars more massive than @xmath11 m@xmath2 , astronomers have toiled hard to explain its origin . early work on the subject attempted to explain it by invoking the possibility of successive fragmentation of unstable clumps ( larson 1973 ) . though it was later shown that such fragmentation can not continue unbridled and must be arrested at some point ; the minimum mass of a fragment was calculated by low & lynden - bell ( 1976 ) . one of the first propositions for a universal imf after larson ( 1973 ) , was made by miller & scalo ( 1979 ) , who having studied a number of nearby star - forming regions proposed a lognormal distribution . the question has been receiving greater attention in recent years , especially in view of the reported similarities between the distribution of core masses and the stellar imf ; see for example , motte _ et al . _ 1998 , nutter & ward - thompson 2007 , and simpson _ et al . _ 2011 , amongst a number of other authors . while molecular clouds(mcs ) contract under self - gravity , the presence of a turbulent velocity field fragments it into a number of clumps and filaments . with growing evidence pointing towards the turbulent nature of mcs , a number of authors have attempted to simulate the effect of turbulence . numerical models by for instance , vzquez - semadeni ( 1994 ) , and padoan & nordlund ( 1999 ) showed that turbulence generates a lognormal density distribution in the volume of gas . though turbulence is dissipative and decays relatively quickly , usually on a timescale comparable to the sound - crossing time , it delays the onset of gravitational instability ( see for e.g. , scalo & pumphrey 1982 , maclow _ et al . _ 1998 , stone _ et al . _ 1998 , and klessen _ et al . some of the more massive clumps once dominated by self - gravity , begin to collapse , where - as other smaller clumps are more likely to be sheared apart by turbulent flows . according to the fragmentation model introduced by padoan & nordlund ( 2002 ; pn ) , the mass distribution of clumps forming out of a turbulent density field reflects the distribution of gas density . in recent years two variants of this theory have been introduced . these theories have a subtle variation in the density threshold at which a volume of gas is likely to support fragmentation . in the pn theory this threshold is fixed by the thickness of the shocked layer resulting from collisions between turbulent flows . on the other hand , in the krumholz & mckee ( 2005 ; km ) theory , the density threshold is set by equating the fastest growing unstable mode with the classical jeans length . finally , in the hennebelle & chabrier ( 2008 ; hc ) theory , there is no density threshold but instead , density perturbations are assumed to condense out as unstable clumps . numerical simulations by padoan & nordlund ( 2007 ) have demonstrated the power - law nature of clumps resulting from turbulent fragmentation . more recent simulations such as those by federrath _ et al._(2010 ) , argue about the possible effects of the modes of driven turbulence , compressive against solenoidal . the resulting density distribution according to these latter authors depends on the driven mode . a possible short - coming of the theory of turbulent fragmentation is its failure to predict beyond the mass - spectrum for gas clumps , and therefore the stellar imf itself , since not all clumps actually spawn stars . while sophisticated numerical algorithms can possibly make the simulations more realistic by including effects of energy feedback , tracing the evolution of gas over several orders of magnitude , up to stellar densities , exacts enormous computational costs . and so , a semi - analytic scheme like the one presented here can shed some light on the problem . veltchev _ et al._(2011 ) for example , recently presented one such calculation where the mass function for unstable clumps , cores , and stars was obtained . having derived the mass function for clumps and cores using respectively , the turbulent fragmentation theory , and the scaling relations due to larson ( 1981 ) , their work suffers on 2 counts : ( i ) the stellar masses were calculated by applying the bondi - hoyle(bh ) accretion theory to collapsing cores . the bh theory as is well - known , assumes that the fluid being accreted originates from a non - self gravitating background . this assumption breaks down in a collapsing core . ( ii ) the calculation of the velocity with which gas is accreted by the protostar is unclear . despite this , their model provides a reasonable estimate of the protostellar accretion rate using which a power - law imf has been derived . in a slightly earlier work ( anathpindika 2011 ) , we had argued on similar lines where accounting for protostellar feedback , and treating the protostellar accretion rate as a free parameter , power - law imfs were derived starting with different choices of core mass distribution that also included a uniform distribution . we had then sought to underline the importance of physical processes during the protostellar phase and the protostellar multiplicity in determining the final imf . the problem of deriving the stellar imf from first principles is indeed non - trivial and so a semi - analytic approach is useful to ascertain the effects of some crucial parameters . cooling processes in star - forming clouds most likely hold the key to sub - fragmentation of unstable clumps into prestellar cores . however , owing to computational expenses , simulations about turbulent fragmentation of gas often assume a cold , isothermal gas , which though reasonable to some extent , fails to account for heating due to protostellar feedback and from other possible external sources . thus regions within a star - forming cloud exhibit substantial variation in gas temperature and therefore , an equally variable dynamical state . possible effects of variations in gas temperature were neglected by veltchev _ _ in their derivation of the core mass function . in the present work these effects are included by generating multiple realisations of clump fragmentation by subjecting them to a number of random choices of gas temperature . having so derived a distribution of core masses , the stellar imf is derived by applying a typical core - to - star conversion efficiency , @xmath0 , between 5% - 15% to each bound core . observations of star - forming clouds have yielded comparable values for the star - formation efficiency ( e.g. , motte _ et al . _ 1998 ; ward - thompson _ _ et al.__2007 ) . the paper is organised as follows . the process of generating the mass - spectra is described in 2 , and the spectra are then discussed in 3 and 4 . we conclude in 5 . we begin our calculations by assuming a volume of molecular gas composed of the usual cosmic mixture where neutral hydrogen makes up about 9 parts out of 10 , while the remainder is made up by helium and other elements like carbon , nitrogen and oxygen . the average mass of a gas molecule , @xmath3 , is taken to be @xmath4 gms . the gas within this volume is characterised by its average density , @xmath5 g @xmath6 , temperature , @xmath7 k , and the magnitude of turbulent velocity , @xmath8 . the size , @xmath9 , of the turbulent region was fixed using the velocity dispersion relation due to larson ( 1981 ) so that the total mass within the volume , @xmath10 m@xmath2 . in a second realisation of the problem , the gas was maintained at a lower temperature , @xmath11 = 15 k , and @xmath12 m@xmath2 . the probability distribution of gas density , @xmath13 , in the turbulent field is @xmath14,\ ] ] where @xmath15 , and @xmath16 with @xmath17 , as suggested by padoan & nordlund ( 2002 ) . @xmath18 , is the mach number of the turbulent field superposed on the gas maintained at a uniform temperature , @xmath11 , and sound speed , @xmath19 . finally , @xmath20 . the turbulent field is defined by its energy - spectrum , @xmath21 , for @xmath22 , so that the turbulent velocity , @xmath23 , is simply @xmath24 where the integration and/or summation on the right - hand side extends over the driven wave - modes , @xmath25 . turbulence was driven over spatial scales between 10 pc and 200 pc , corresponding to wavenumbers between 1 and 8 . the length of the fastest growing mode , @xmath26 , is given by the following equation , @xmath27 @xmath28 , where turbulence - induced shocks are assumed to be isothermal . the effective sound - speed , @xmath29 , so that a slight manipulation of eqn . ( 3 ) after substituting for @xmath30 and @xmath31 , leads us to @xmath32 where @xmath33 is the thermal jeans length for sound - speed , @xmath19 . for a highly supersonic gas , the left - hand side of the above expression may be approximated to @xmath34^{1/2},\ ] ] or in terms of only the wave - vector as @xmath35 the scale of fragmentation is set by @xmath26 , so that the mass of a clump , @xmath36 , condensing out is of the order , @xmath37 or @xmath38 the number of clumps , @xmath39 , condensing out is @xmath40 , so that the clump - mass function ( mf ) , is @xmath41 the mf was generated by monte - carlo integrating the above equation with about @xmath42 choices of random perturbations in the density field . we note that clump - formation was terminated once @xmath43 , so that mass is conserved . the dynamical stability of clumps , it is well - known , is determined by the balance between heating and cooling processes . possible effects of various cooling processes such as molecular line emission at densities below 10@xmath44 @xmath6 , and the gas - dust coupling at much higher densities , were introduced in the calculations by subjecting the gas within these clumps to 10@xmath45 random choices of temperature distributed uniformly between 7 k- 20k , and then 6 k- 15 k , in each of the two realisations . the temperature distribution is characterised by a probability density , @xmath46 , and the core mass function ( cmf ) was obtained by convolving it with the clump mass function , so that @xmath47 alternatively , @xmath48 , so that @xmath49 . the number of cores , @xmath50 , is then , @xmath51 . finally , the initial mass function , ( imf ) , was derived by applying a random core - to - star conversion efficiency , @xmath0 , between 5% to 15 % , to each bound core . , and the index of turbulent power - spectrum , @xmath52 , as marked on each plot . the spectra are roughly lognormal in nature , though with a slight power - law extension at the high - mass end.,title="fig : " ] , and the index of turbulent power - spectrum , @xmath52 , as marked on each plot . the spectra are roughly lognormal in nature , though with a slight power - law extension at the high - mass end.,title="fig : " ] using the basic turbulent fragmentation hypothesis we have derived mass functions ( mfs ) for dense clumps , cores and stars . fragmentation of the volume of gas considered here is purely due to the growth of density perturbations that are random in nature . while our interest lies in the clump - formation via fragmentation , the dissipation of turbulent energy is neglected . results of this exercise are summarised by the plots in figs . 1 through to 5 . the distribution of clump masses derived here has been shown in fig . 1 . irrespective of the choice of driven wave - modes , the mass - functions exhibit remarkable similarities . the distributions in each case are roughly lognormal , though with a slight power - law extension at the high - mass end . the spectra , however , are significantly steeper than that for typical clumps , which have a slope of @xmath53 ( e.g. kramer _ et al . _ 1998 ) . also , it follows from eqn . ( 5 ) , a relatively shallow energy spectrum ( @xmath52=1.2 ) shows a greater tendency of producing larger clumps . this can indeed be seen from the plots in fig . 1 the core mass function(cmf ) , generated for different choices of the spectral index , @xmath52 , and driven modes , @xmath25 , have been shown in fig . 2 . overlaid on the plot shown in the upper panel of fig . 2 is the power - law cmf defined as , @xmath54 with , @xmath55 the power - law with these exponents fits the derived cmf reasonably well , acceptable at 5% significance level of the @xmath56-test , performed with 5 degrees of freedom , viz . , the gas temperature , density perturbations , turbulence mach number , @xmath57 , spectral - index , @xmath52 , and the wavemodes driven , @xmath25 . evidently the power - law defined by eqn . ( 10 ) is considerably steeper than that for cores found in typical star - forming regions such as the orion mc ( e.g. , nutter & ward - thompson 2007 ) , and the ophiuchus mc ( e.g. , johnstone _ 2001 ) . increasing the gas temperature and the mach number , @xmath57 , raises the fragmentation length and therefore the mass of resulting cores which results in a marginally shallow cmf ( @xmath58 , for @xmath59 ) , as can be seen in the central panel of fig . likewise , a shallow cmf also results from driving longer wavemodes , @xmath60 , when comparatively lesser power reaches smaller spatial scales . this latter cmf is shown in the lower panel of fig . a unique feature of the cmfs derived here is the appearance of a plateau for intermediate masses in the range ( 0.2,1 ) m@xmath2 , which is probably due to the uniform distribution of temperature used in generating the cmf . , and the power - spectrum index , @xmath52 . plotted with a continuous black line are the power - laws discussed in the text above . , title="fig : " ] , and the power - spectrum index , @xmath52 . plotted with a continuous black line are the power - laws discussed in the text above . , title="fig : " ] , and the power - spectrum index , @xmath52 . plotted with a continuous black line are the power - laws discussed in the text above . , title="fig : " ] the core - to - star conversion efficiency , @xmath0 , is the sixth parameter in this scheme apart from the five listed previously . the derived imf has been shown in the two panels of fig . 3 , where the one in the upper panel was derived from the cmf defined by eqn . ( 10 ) . it is significantly steep for @xmath61 . on the other hand , the imf derived from the relatively shallow cmf obtained by driving the @xmath60 modes , has been shown in the lower - panel of fig . 3 . it has been fitted by a power - law defined below , @xmath62 acceptable at a 10% significance level of the @xmath56-test with 6-degrees of freedom . more interestingly , the power - law exponents are within the error - bars on the universal imf suggested by kroupa ( 2002 ) . it can be seen that like the cmf , this imf also has a plateau in the sub - solar mass range which most likely reflects a similar feature of the cmf . + + * lowering the initial gas temperature ( @xmath11 = 15 k , @xmath57=15 ) * + retaining all other parameters , the mass function for clumps and cores was re - derived in this case , the result of which has been plotted in figs . 4 and 5 . lowering the initial temperature of the gas , @xmath11 , by a factor of 2 to 15 k , does not much alter the overall nature of the two distributions , though massive clumps with mass in excess of @xmath63 m@xmath2 now do not form as can be seen in fig . 4 . as before , gas within these clumps was subjected to a random choice of temperatures in the range 6 k- 15k , and then allowed to fragment . the resulting distribution of core masses has been shown in the upper - panel of fig . the distribution of core - masses in the lower - panel was derived for a slightly higher initial mach number ( @xmath57 = 20 ) . not only is the cmf in this latter case relatively shallow ( @xmath64 m@xmath2 ) , but the plateau at intermediate masses is also less wide . has been shown in the two panels here . the power - law fit to the derived spectrum in the lower panel is defined by eqn . ( 11 ) in the main text . , title="fig : " ] has been shown in the two panels here . the power - law fit to the derived spectrum in the lower panel is defined by eqn . ( 11 ) in the main text . , title="fig : " ] = 15 k. apart from a broader full - width at half - maximum , these spectra do not differ much from those plotted in the upper panel . ] in the previous section , assuming a set of demonstrative parameters we derived mass distributions for clumps , prestellar cores and stars by invoking the theory of turbulent fragmentation . analogous to the pn02 model , we commenced with a lognormal pdf while setting the density threshold as dictated by the jump condition for an isothermal shock . the derived clump mass function largely reflects the density pdf , though it also has a power - law tail at the high - mass end . this tail is relatively shallow in the case with @xmath11 = 15 k. steepening the power - spectrum index , on the other hand , simply shifts the mass distributions towards lower masses , closer to those of typical clumps . the core mass function , irrespective of the choice of wavenumbers driven , the temperature of gas within clumps or indeed , the mach number , has two parts : a salpeter - like power - law for @xmath65 , and a plateau at lower masses . we have also seen that the power - law for @xmath65 is relatively steep , and lies between 3.5 - 4 which is probably the result of an analogously steep mass - function for clumps which precede the cmfs . interestingly though , these cmfs extend into the regime of low mass cores ( @xmath66 m@xmath2 ) . the plateau observed in the cmfs also appears in the stellar imfs as can be easily noticed through a comparison between figs . 2 and 3 . the cmfs and the imfs derived here both show a significant over - population in the ( 0.5,1.0 ) m@xmath2 range . the plateau in the cmfs derived here merely reflect the distribution , @xmath46 , of gas temperature within clumps which , we remind , was assumed to be distributed uniformly between 7 k - 30 k , and then between 6 k- 15 k in the next realisation . the plateau can be avoided by confining the gas temperature to a fairly narrow range implying that , the temperature of gas across star - forming clumps is unlikely to show much variation . of course , we are discounting regions which may exhibit extreme conditions . fragmentation in a volume of turbulent gas proceeds along a hierarchy that begins with clumps and filamentary clouds . some of which fragment further to produce prestellar cores . indeed , some recent maps of star - forming regions in various sub - millimetre wavebands have revealed extensive networks of filamentary clouds with a number of prestellar cores ( e.g. andr ' e _ et al . _ 2010 ; menshchikov _ et al . _ 2010 ) . from the preceding analysis in 2 it is clear that the prevalent physical conditions within gas clumps are crucial in determining the proclivity towards further fragmentation . to the first order , the minimum mass of fragmentation is set by the thermal jeans mass , @xmath67 . in reality of course , the magnetic field will revise this mass upward , however , in the present exercise we do not include the effects of magnetic field . gas temperature and density are locked in a reciprocally causative relationship . it is in this respect that a density threshold assumes critical significance , for cooling mechanisms are density dependent . atomic cooling for instance , dominates in the relatively low - density gas up - to a few hundred particles per cubic centimetre , then the coupling between gas and dust dominates up - to densities of a few thousand particles per cubic centimetre and at even higher densities such as those in typical cores ( @xmath68 @xmath6 ) , molecular species freeze out . similarly , effects of magnetic flux diffusion are likely to become significant at densities upward of several thousand particles per cubic centimetre . taken cognitively , these effects can raise the minimum mass of fragmentation considerably ( e.g. larson 2005 ) . here we have only included the effects of gas thermodynamics by accounting for the gas temperature and the velocity dispersion . the contribution of the velocity dispersion to the fragmentation length scale is determined by the energy - spectrum , @xmath69 . in general , the impact of the energy - spectrum is readily visible on the mass - function for clumps where relatively smaller clumps form with steepening energy - spectrum ; see plots in fig . the cmfs on the other hand , while exhibiting immunity to changes in @xmath69 , are significantly affected by changes in the thermal state of the gas , as demonstrated by plots shown in the central panel of fig . this is further corroborated by results from the exercise where the temperature of the pre - fragmentation gas was lowered by a factor of 2 ; see also figs . 4 and 5 . in either case , temperature of the gas within a clump was treated as a random variable to study the effects of its variation . the gas temperature thus , was allowed to take values within a range typically found in star - forming clouds such as taurus ( e.g. , falgarone _ et al . _ 1998 ) , ophiuchus ( e.g. , johnstone _ et al . _ 2000 ) , and orion ( e.g. , yu _ 1997 ) . prestellar cores collapse under favourable conditions to form stars and dwarfs , though the collapse phase itself is associated with competing processes such as feedback from young stellar objects , competition between accreting stars to acquire mass and the interaction between a young star and its attendant disk . these processes , though highly non - linear in nature , are critical in determining the stellar imf . in literature , these processes have sometimes been grouped together as the core fragmentation mass function ( e.g. elmegreen 2000 ; 2011 ) . clearly , stellar multiplicity is an essential component of the stellar imf ( e.g. , goodwin & kouwenhowen 2009 ) . in an earlier work ( anathpindika 2011 ) , we had presented a model of core fragmentation that included the details listed above , and therefore the prospect of forming brown - dwarfs via core - fragmentation . in view of the non - triviality about estimating the mass of a protostellar disk , that model failed to predict any objects possibly resulting from disk - fragmentation . however , assuming the rate of protostellar accretion as a free parameter , we had then demonstrated that the processes during the collapse of a core hold the key to determining the nature of the imf . that model had also demonstrated the robustness of the imf to changes in the preceding cmf . in the present work though , we have derived the imf by applying a typical core - to - star conversion efficiency , @xmath0 , to each bound core . in the present case it is therefore not possible to comment on the companion stars . however , it is clear that the power - law tail in the cmf reappears in the imf , albeit with a slightly shallower slope and with significant changes to the plateau at sub - solar masses . the number of parameters used in the current exercise could invite some criticism , but these parameters are essential components of the turbulent fragmentation model and adopted from recent reports about star - forming clouds . for a typical choice of parameter values , the derived mass functions are broadly similar to those determined observationally . the agreement was best seen for results obtained with - ( i ) a steep power - spectrum , typically steeper than the kolmogorov - type , up to about @xmath70 , and ( ii ) by driving longer wavemodes , @xmath60 . in view of ( i ) , we suggest that the power - spectrum index for gas in mcs could indeed be steeper than kolmogorov - type . second , turbulence is likely to be driven on larger scales which is consistent with observations and has also demonstrated by brunt _ et al . _ ( 2009 ) through hydrodynamic simulations . furthermore , studies of the @xmath71co emission line for the cygnus star - forming region ( schneider _ _ et al.__2011 ) , and a number of other star - forming clouds in the galactic ring survey ( roman - duval _ et al . _ 2011 ) , have shown that turbulence is indeed driven on larger spatial scales , typically on the scale of the the cloud size . a point that has not been addressed in the present work is about the likely effects of changes in the gas density . however , from eqns . ( 6 ) it is quite clear that lower density may very well produce predominantly larger clumps , which might possibly explain the relatively steep cmfs observed in low - density clouds . it might also explain the steep stellar mass distributions in some spiral galaxies with peculiar properties , such as those with low surface brightness ( e.g. lee _ et al . thus , if indeed the stellar imf has a universal nature then the underlying processes which generate it could also be similar . gravoturbulent fragmentation , it appears , could play a key role in assembling the initial conditions for star - formation . this theory also admits the possibility of rapid star - formation which makes it more attractive . using the concept of turbulent fragmentation we have derived the mass spectra for clumps , prestellar cores and stars via a monte carlo integration of the underlying probability distribution . we first calculated the clump mass spectrum by allowing the random density perturbations to condense out as clumps , while the cmf was derived through the convolution of this mass spectrum with a random distribution of gas temperature for the dense clumps . irrespective of the initial gas temperature , we observed that best agreement with observational findings resulted from driving longer wavemodes and with a power - spectrum steeper than kolmogorov - type . thus , while fragmentation of the original volume of gas produces clumps which themselves may break - up further , chemical processes that determine gas temperature in the dense phase are likely to play a crucial part , leading up to the distribution of core masses . in fact , we also suggest that the characteristic knee seen in the cmf and the stellar imf could be a pointer to the temperature of the gas within star - forming clumps , and therefore , the efficiency of cooling processes . for the theory of fragmentation to predict the core masses more accurately , particularly at the low - mass end , it must be coupled with appropriate thermodynamic details of the density field . arguably , this is a non - trivial task , and so , in the present work temperature has been substituted as a random variable that takes values similar to those found in typical star - forming clouds . the imf was also derived in a similar fashion by treating the star - formation efficiency as a random variable . the imf so derived was found to be in reasonably good agreement with the universal imf , acceptable at a 10% significance level of the @xmath56-test with 6 degrees of freedom , viz . the gas temperature , density perturbations , the mach number and spectral - index of the turbulent field , the wavemodes driven , and the core - to - star conversion efficiency . it is therefore clear that apart from the physical properties of the gas , the history of protostellar collapse is critical towards determining the nature of the imf . our work shows that a typical cmf , and therefore an imf , emerge quite naturally from the application of the theory of turbulent fragmentation by driving longer wavemodes . this claim is corroborated by numerical and observational results presented by brunt _ ( 2009 ) and schneider _ et al . _ while isolated regions within star - forming clouds may exhibit extreme physical conditions leading to corresponding variations in the imf , results derived here suggest that most star - formation in the nearby clouds is unlikely to vary much ; see also the review by bastian _ et al._(2010 ) . the author acknowledges support via a post doctoral fellowship of the department of science & technology , government of india , held at the indian institute of astrophysics . useful suggestions from an anonymous referee towards the improvement of the original manuscript are gratefully acknowledged . andr ' e , ph . , menshchikov , a. , bontemps , s. , koenyves , v. , motte , f. _ et al . _ 2010 , a&a , 518 , l102 anathpindika , s. , 2011 , new astronomy , 16 , 477 bastian , n. , covey , k. r. & meyer , m. r. , 2010 , ann . rev . a&a , 48 , 339 bergin , e & tafalla , m. , 2007 , ann . a&a , 45 , 339 brunt , c. m. , heyer , m. h. & maclow , m. m. , 2009 , a&a , 504 , 883 elmegreen , b. , 2000 , mnras , 311 , l5 elmegreen , b. , 2011 , apj , 731 , 61 flagrone , e. , pavis , j. f. , heithhausen , a. , perault , m. , stutzki , j _ et al . _ , 1998 , a&a , 331 , 669 goodwin , s & kouwenhowen , m.b.m . , 2009 , mnras , 397 , l36 hennebelle , p & chabrier , g. , 2005 , apj , 684 , 395 ( hc ) johnstone , d. , wilson , c.d . , mariarty - schieven , g. , joncas , g. _ et al . _ , 2000 , apj , 545 , 327 gilman r.c . , 1972 , apj , 178 , 423 johnstone , d. , fich , m. , mitchell , gf . , moriarty - schieven , g. , 2001 , apj , 559 , 307 klessen , r. , heitsch , f. & maclow , m. m. , 2000 , apj , 535 , 887 kroupa , p. , 2002 , _ astronomical soc . of pacific _ , p. 86 krumholz , m & mckee , c. , 2005 , apj , 630 , 250 ( km ) larson , r. b. , 1973 , mnras , 161 , 133 larson , r. b. , 1981 , mnras , 194 , 809 larson , r. b. , 2005 , mnras , 359 , 211 low , c & lynden - bell , d. , 1976 , mnras , 176 , 367 lee , h. , gibson , b. k. , flynn , c. , kawata , d. , beasely , m. a. , 2004 , mnras , 353 , 113 maclow , m.m . , klessen , r. , burkert , a & smith , m.d . , 1998 , lttrs . , 80 , 2754 miller , g & scalo , j.m . , 1979 , apjss , 41 , 253 menshchikov , a. , andr ' e , ph . , didelon , p. , koenyves , v. , schneider , n. _ et al . _ , 2010 , a&a , 518 , l103 motte , f. , andr , ph & neri , r. , 1998 , a&a , 336 , 150 nutter , d & ward - thompson , d. , 2007 , mnras , 374 , 1413 padoan , p & nordlund , , 1999 , apj , 526 , 279 padoan , p & nordlund , , 2002 , apj , 576 , 870 ( pn ) padoan , p. , nordlund , , kritsuk , a. g. , norman , m. l & li , p. s. , 2007 , apj , 661 , 972 roman - duval , j . , federrath , c. , brunt , c. , heyer , m. , jackson , j & klessen , r. , 2011 , apj , 740 , 120 salpeter , e. , 1955 , apj , 121 , 161 scalo , j.m & pumphrey , w.a . , 1982 , apj , 258 , l29 schneider , n. , bontemps , s. , simon , r. , ossenkopf , v. , federrath , c. , klessen , r. , motte , f. , andr , ph _ et al . _ , 2011 , a&a , 529 , a1 stone , j.m . , ostriker , e.c & gammie , c.f . , 1998 , apj , 508 , l99 vzquez - semadeni , e. , 1994 , apj , 423 , 681 ward - thompson , d. , andr ' e , ph . , crutcher , r. , johnstone , d. , onsihi , t & wilson , c. , 2007 , in protostars & planets , ed . by reipurth , b , jewitt & k. keil ( tuscon , az . ) ; university of arizona press , 117 yu , k.c . , bally , j & devine , d. , 1997 , apj , 485 , l45	two varieties of the universal stellar initial mass function ( imf ) viz . , the kroupa and the chabrier imf , have emerged over the last decade to explain the observed distribution of stellar masses . the possibility of the universal nature of the stellar imf leads us to the interesting prospect of a universal mode of star - formation . it is well - known that turbulent fragmentation of gas in the interstellar medium produces a lognormal distribution of density which is further reflected by the mass - function for clumps at low and intermediate masses . stars condense out of unstable clumps through a complex interplay between a number of dynamic processes which must be accounted for when tracing the origin of the stellar imf . in the present work , applying the theory of gravitational fragmentation we first derive the mass function ( mf ) for clumps . then a core mass function ( cmf ) is derived by allowing the clumps to fragment , having subjected each one to a random choice of gas temperature . finally , the stellar imf is derived by applying a random core - to - star conversion efficiency , @xmath0 , in the range of 5%-15% to each cmf . we obtain a power - law imf that has exponents within the error - bars on the kropua imf . this derived imf is preceded by a similar core mass function which suggests , gravoturbulent fragmentation plays a key role in assembling necessary conditions that relate the two mass - functions . in this sense the star - formation process , at least at low redshifts where gas cooling is efficient , is likely to be universal . we argue that the observed knee in the cmf and the stellar imf may alternatively be interpreted in terms of the characteristic temperature at which gas in potential star - forming clouds is likely to be found . our results also show that turbulence in star - forming clouds is probably driven on large spatial scales with a power - spectrum steeper than kolmogorov - type . star - formation gravoturbulent fragmentation mass functions
we thank walter giele for providing a copy of the vecbos program for @xmath84jet production and thank tao han for discussions . jo would like to thank the rutherford appleton laboratory theory group and the desy theory group for hospitality during the course of this work . jo is supported by the uk science and engineering research council . the work of vb is supported in part by the u.s . department of energy under contract no . de - ac02 - 76er00881 , in part by the texas national laboratory research commission under grant no . rgfy93 - 221 , and in part by the university of wisconsin research committee with funds granted by the wisconsin alumni research foundation . 00 cdf collaboration : a. barbaro - galtieri , report to eps conference , marseilles , july 1993 . d0 collaboration : m. narain , xxvii@xmath85 recontres de moriond , march 1993 , to be published . p. langacker , pennsylvania preprint upr-0555-t ( 1993 ) . s. protopopescu and g.p . yeh , reports to top physics symposium , madison , november 1992 . h. baer , v. barger , j. ohnemus , and r.j.n . phillips , phys . rev . * d42 * , 54 ( 1990 ) and references therein . t. han and s. parke , fermilab - pub-93/105-t . h. baer , v. barger , and r.j.n . phillips , phys . rev . * d39 * , 3310 ( 1989 ) . berends , h. kuif , b. tausk , and w.t . giele , nucl . phys . * b357 * , 32 ( 1991 ) ; f.a . berends , j.b . tausk , and w.t . giele , phys . rev . * d47 * , 2746 ( 1993 ) . benlloch , n. wainer , and w.t . giele , fermilab - pub-93/060-t . v. barger , j. ohnemus , and r.j.n . phillips , int . phys . * a4 * , 617 ( 1989 ) . v. barger , t. han , j. ohnemus , and d. zeppenfeld , phys . rev . * d41 * , 2782 ( 1990 ) . a.d . martin , r.g . roberts , and w.j . stirling , phys . * d47 * , 867 ( 1993 ) . r.k . ellis , phys . b259 * , 492 ( 1991 ) . e. laenen , fermilab - pub-93/155-t . cdf collaboration : f. abe _ et al . 70 * , 4042 ( 1993 ) . cdf collaboration : f. abe _ et al . _ , phys . rev . * d45 * , 3921 ( 1992 ) . 1 . [ s , a dist ] sphericity @xmath40 and aplanarity @xmath41 distributions for @xmath0 signal ( @xmath65 gev ) and @xmath49jets background , for @xmath86jet events at the tevatron . [ sigma min ] integrated signal and background cross sections for ( a ) @xmath87 and ( b ) @xmath88 , versus @xmath69 and @xmath70 , respectively . [ tilde m ] reconstructed top - quark mass @xmath81 distributions for tevatron @xmath0 signals with @xmath89 , 150 , 170 gev , and @xmath90jets background . upper curves are for @xmath91 ; lower curves are for @xmath92 .	single - lepton plus jets signals from @xmath0 production at hadron colliders generally give more spherically symmetrical events than the principal backgrounds from @xmath1 production . we show that sphericity and aplanarity criteria , applied to lepton plus neutrino plus four jet final states at the fermilab tevatron @xmath2 collider , help to discriminate ; they can be used either to validate an eventual top signal or simply to reduce background . an alternative circularity criterion in the transverse plane is less successful . 6.5 in # 1#20=1=to0@xmath30=1=to0#2#1 - 01 = cmssbx10 scaled 2 to * event shape criteria for single - lepton top signals * + [ .4 in ] v. barger@xmath4 , j. ohnemus@xmath5 and r.j.n . phillips@xmath6 + _ @xmath7physics dept . , university of wisconsin , madison wi 53706 , usa + @xmath8physics dept . , university of durham , durham , dh1 3le , england + @xmath9rutherford appleton laboratory , chilton , ox11 0qx , england _ the predicted top quark , an essential component of the standard model ( sm ) , is apparently too heavy to have been discovered yet at high energy colliders ; the present experimental limit is @xmath10 gev from the cdf detector @xcite and independently @xmath11 gev from the d0 detector @xcite . the consistency of radiative corrections to all electroweak data however indicates that @xmath12 gev @xcite , and hence that the top quark can probably be discovered at the fermilab tevatron @xmath2 collider sooner or later @xcite . here the sm predicts mainly @xmath0 pair production via qcd with @xmath13 decays ; the smallest and cleanest signals are expected in two - lepton channels @xcite , but larger single - lepton @xcite and all - jet @xcite signals can also in principle be separated from backgrounds . since the top signals will initially ( and perhaps for a considerable time ) be based on a small number of events , it will be important to suppress the backgrounds as far as possible in ways that do not seriously reduce the signals . in this note we propose the use of event - shape criteria to distinguish the top signal and reduce backgrounds in the relatively copious single - lepton - plus-4jet channel . the underlying idea is that tevatron @xmath0 production will be mainly near threshold in the cm frame of the lowest - order qcd subprocesses @xmath14 or @xmath15 , and hence will mostly lead to rather spherical event shapes in this frame . phase - space factors favor spherical configurations ; the @xmath0 spin correlations @xcite and decay matrix elements introduce substructure but do not drastically change this spherical tendency , which can help to distinguish the signal from backgrounds . this idea is very familiar in the context of heavy quark searches at @xmath16 colliders and has recently been applied to 6-jet top signals at the tevatron @xcite ; we propose here to apply it to lepton - neutrino-4jet signals from @xmath17 decays ( where the neutrino is partially measured via transverse energy balance ) . backgrounds in this channel from @xmath18- and @xmath19-quark semileptonic decays can be removed by lepton @xmath20 and isolation cuts @xcite . the major background then comes from @xmath1 production plus 4 qcd jets , which remains comparable with the signal after all the usual cuts on transverse momenta @xmath20 , pseudo - rapidities @xmath21 , and missing transverse energy @xmath22 @xcite ; we neglect contributions from @xmath23 or @xmath24 plus 2 qcd jets @xcite that are an order of magnitude smaller . this background can be further suppressed by requiring a @xmath18-jet - tag , costing the signal a tagging - efficiency factor of order 0.3 at present @xcite . alternatively , we may expect the inherent collinear singularities to cause spatial correlations of qcd jets with each other and with the beam axis , giving less spherical configurations and hence a discrimination between background and signal via an appropriate choice of event - shape variable as follows . first we restrict ourselves to momenta in the transverse plane , in order to use the most direct neutrino information contained in its transverse momentum defined by @xmath25 . we consider the usual circularity " event shape variable @xmath26 defined by @xmath27 summed over the lepton , neutrino , and 4 jet transverse momenta and minimized with respect to the choice of the unit vector @xmath28 . this quadratic variable is unstable against the splitting or combining of jets ; although @xmath26 is well defined for any given jet algorithm , it would be theoretically preferable to use a linearized variable such as @xmath29 both @xmath26 and @xmath30 are normalized to 1 for an ideal circular event and vanish for a linear configuration . next we consider event shapes in three dimensions . the longitudinal neutrino momentum @xmath31 is not measured ; it can be inferred approximately by requiring that the lepton - neutrino invariant mass @xmath32 , but this leaves a two - fold ambiguity . we choose the solution that allows the most consistent @xmath0 reconstruction ( details below ) and form the conventional normalized @xmath33 momentum tensor @xmath34 from the lepton plus neutrino plus 4jet momenta @xmath35 , in the cm frame of these momenta . @xmath36 has three eigenvalues @xmath37 with @xmath38 and @xmath39 ; the sphericity @xmath40 and aplanarity @xmath41 are then defined by @xmath42 and @xmath43 . ( it would be theoretically preferable to use linearized variables such as thrust and acoplanarity , but for ease of computation we use @xmath40 and @xmath41 here ) . thus @xmath44 for ideal spherical events , @xmath45 and @xmath46 for plane circular events , and @xmath47 for linear events . we therefore expect the tevatron top signal to give typically larger @xmath48 and @xmath41 than the @xmath49jets background , and have made quantitative parton - level calculations at @xmath50 tev to confirm this . for illustration we compute the @xmath51 production and decay distributions in lowest order with full spin correlation effects @xcite , using the mrs set @xmath52 parton distributions of ref . @xcite evaluated at scale @xmath53 , neglecting possible additional qcd jets and taking @xmath54jet final states . the total @xmath0 production rate is normalized conservatively to the central value of ref . @xcite , although values about 20% higher have recently been proposed @xcite . we calculate the @xmath55jet background from the full tree - level matrix elements following ref . @xcite , evaluating the parton distributions at scale @xmath56 in accord with tevatron results @xcite . we sum lepton flavors and charges @xmath57 . we simulate calorimeter errors by gaussian uncertainties on jet and lepton energies , following the cdf values tabulated in ref . @xcite , and calculate the apparent neutrino @xmath20 from overall @xmath58 imbalance . we impose the following acceptance cuts on transverse momenta , pseudorapidities , and jet multiplicity @xmath59 , selecting regions where the top signal should be strong , @xmath60 we also set minimum jet - jet and lepton - jet separations @xmath61 where @xmath62 , to mimic some effects of jet - finding algorithms and lepton isolation cuts @xcite . we note that these @xmath63 cuts already introduce some event - shape discrimination . we first compare the shapes of the @xmath48 and @xmath41 distributions for @xmath0 signal and @xmath64 jets background , for the case @xmath65 gev . the means @xmath66 and standard deviations @xmath67 for the signal ( background ) are @xmath68 from which we see that the circularity variables offer little discrimination in this case , whereas the @xmath40 and @xmath41 variables show appreciable displacements between signal and background . the @xmath40 and @xmath41 distributions are shown in fig . [ s , a dist ] . we see that the signal has significantly different distributions from the background , expecially for @xmath41 ; this could help to validate an eventual top signal ( extracted by @xmath18-tagging say ) . alternatively , cutting out events below a minimum value @xmath69 or @xmath70 would reduce the background more than the signal ; integrated cross sections for @xmath71 and @xmath72 are shown in fig . [ sigma min ] . we shall illustrate below the effects of a choice @xmath73 ( not specifically optimized ) , similar to cuts suggested for all - jet signals @xcite ; this reduces the signal ( background ) by 15% ( 36% ) , giving a cleaner result and an increase in its statistical significance ( signal)/@xmath74 . to extract a top quark mass peak , we introduce a constrained event parameterization as follows . we first compute the two solutions for @xmath31 by requiring @xmath32 ; if they are formally complex , we set the imaginary parts to zero , getting the single nearest - to - onshell solution . we then select the pair of jets with invariant mass closest to @xmath75 , and identify them as the @xmath76 decay candidates . the remaining two jets are candidates for @xmath18 and @xmath77 jets , if this is a @xmath0 event . from the four possible ways of pairing the latter two jets with our unique @xmath76 choice and the two - fold @xmath78 solution , we select whichever pairing gives the closest agreement between the resultant @xmath79 invariant masses . the mean of these two closest @xmath80 values is defined as the reconstructed top mass @xmath81 ( following the notation of ref . @xcite ) ; taking the mean gives a narrower signal peak than either the semileptonic or the hadronic @xmath80 value separately . our @xmath81 variable differs from those proposed in refs . @xcite . figure [ tilde m ] presents the calculated @xmath81 distribution for signal and background , for @xmath82 , 150 , and 170 gev , first with the standard acceptance cuts of eqs . ( [ accept])([sep ] ) ( upper curves ) , and also with the additional event - shape cuts @xmath83 ( lower curves ) . hadronization effects have not been included , except in their contributions to calorimeter resolution ; we expect they will somewhat further smear the @xmath81 peak . our results show that 1 . @xmath40 and @xmath41 distributions of top signals differ significantly from background in the lepton + neutrino + 4jets channel ; 2 . these differences could help to validate an eventual top signal or alternatively reduce background and improve significance ; 3 . the signal would peak in our reconstructed top mass variable @xmath81 . event shape criteria have previously proved promising in the 6jet channel @xcite ; they may also have value in lepton - plus-3 jet and dilepton channels .
giant molecular clouds ( gmcs ) play a fundamental role in the evolution of a spiral galaxy . they represent the end result of the continual collection of diffuse gas into molecular form in the interstellar medium , some of it having been recycled from past generations of stars . gmcs are also the sites where the bulk of ongoing star formation in a galaxy then occurs . they can form from the atomic substrate , as evident in global views of other galaxies ( e.g. m33 * ? ? ? * ) , or perhaps from the assembly of smaller molecular clouds . how this occurs within our own galaxy remains unclear as the process has not yet been seen . models range from quasi - static self - gravitational collapse , for instance along magnetic field lines ( e.g. * ? ? ? * ) , to transient phenomena such as the coalescence of clouds in converging flows in a turbulent medium . since stars appear to form almost as soon as molecular gas is present , then the rate of formation of gmcs is also instrumental in determining the star formation rate . furthermore , despite being molecular , forming clouds will not necessarily be evident through co line emission , normally the tracer used to delineate their locations . this is because in the low molecular column densities of clouds forming from the diffuse atomic medium far uv radiation will dissociate the co , whereas self - shielding can allow h@xmath4 to exist ; i.e. a pure h@xmath4 molecular cloud ( sometimes also known as dark h@xmath4 due to the absence of an emitting molecular species ; * ? ? ? * ) . the carbon in such clouds can not be hidden , however , and will exist primarily in neutral form where it can emit through the [ ci ] lines at 492 and 809ghz ( 610 & @xmath5 m , respectively ) , or in singly - ionized form , emitting as [ cii ] 1.90thz ( @xmath6 m ) , or in both . in either case , its detection is difficult , for observations at these frequencies require exceedingly dry atmospheric conditions or airborne / space platforms . facilities able to map these species at good spectral and spatial resolution , and over the large areas of sky that may encompass the environs of forming giant molecular clouds , have not previously been available . we present the first data from a new facility at the summit of the antarctic plateau , ridge a @xcite . the 62 cm aperture high elevation antarctic terahertz ( heat ) telescope ( kulesa et al . 2014 , in preparation ) currently provides spectroscopic imaging in the 810ghz lines of co j=76 and [ ci ] j=21 , with upcoming capabilities at 1.46thz [ nii ] and 1.90thz [ cii ] . it can already map [ ci ] emission over significant regions of the galactic plane , and so be able to search for forming molecular clouds and `` dark '' h@xmath4 . the exceedingly low and stable columns of precipitable water vapor at ridge a ( falling below 100@xmath7 m for more than 25% of winter ; @xcite ) make this possible . we have identified a filament in [ ci ] in the g328 region that is also evident in molecular ( co ) and atomic ( hi ) survey data taken with similar resolution . the results of our mapping of this filament are the subject of this paper . [ cii ] emission has also been identified along a sightline near to this filament by the got c+ program using the _ herschel _ telescope . the observations are summarised in [ sec : obs ] and the results presented in [ sec : results ] . they are then discussed in [ sec : discuss ] with reference to the physical characteristics of the g328 filament , a pdr model for its line emission , and the carbon abundance within it . the data presented here comes from three separate spectral line surveys of sub - mm [ ci ] , mm co and cm hi emission in the southern galactic plane . heat is imaging the [ ci ] 809.3ghz line ( @xmath8 ) from ridge a , antarctica , yielding data cubes with @xmath9 and 0.7km / s resolution . @xmath10co , @xmath11co and c@xmath12o j=10 emission lines ( 115.3 , 110.2 & 109.8ghz , respectively ) were imaged as part of the 22 m mopra southern galactic plane co survey @xcite , with @xmath13 and 0.1km / s resolution . the 1.420ghz hi line comes from archival data in the parkes atca southern galactic plane survey ( the sgps ; * ? ? ? * ) and provides @xmath9 and 3km / s resolution . the contiguous data set for these species covers @xmath14@xmath15 to @xmath16 , although the co and hi images presented here in fact cover a larger region , of @xmath17 to @xmath18 to @xmath19 . [ ci ] and co emission extend over a velocity range from @xmath20 to 0km / s @xmath21 in the data cube , the hi from @xmath22 to @xmath23km / s . spatially , the emission from the three species across these velocity ranges covers the entire region mapped . a sinuous emission feature was identified in the co data cubes , extending @xmath24 in angular size and centred at @xmath25km / s , with a narrow spectral range of just 4km / s . this filament is analysed in this paper , with an analysis across the full spectral emission range of the cubes presented in kulesa et al . 2014 ( in preparation ) . figure [ fig : cicohi ] presents integrated flux images of the [ ci ] , @xmath10co and hi for the g328 filament , together with a 250@xmath7 m continuum image of the dust emission from the _ herschel _ higal program . the @xmath11co image ( not shown ) looks essentially the same as @xmath10co ( albeit at @xmath26 the flux ) , and is used to determine the co optical depth . c@xmath12o is also detected at the brightest portions of the filament at the @xmath27 level , indicative of high optical depths . it is also notable that the hi is evident as a depression in the level of the extended atomic emission found across the entire g328 region ( i.e. hi self - absorption or hisa ) , whereas the co and [ ci ] emission are primarily confined to just the filament itself in this velocity range . also notable is that the filament is not apparent in the dust emission mapm images this level is @xmath281,200 , 600 & 200mjy / sr , respectively , in g328 . ] . there also do not appear to be any ir sources that are clearly associated with the filament , as might be expected if star formation were active in it . figure [ fig : contours ] shows the hi image overlaid with contours of the co and [ ci ] . it is clear that these features are closely related . the heat [ ci ] dataset so far spans only the nw portion of the filament ( the `` head '' ) . however , given the similarity of the co and [ ci ] here , both spectrally and in morphology , it seems likely that the [ ci ] emission will be co - extensive with the co along the rest of the filament . figure [ fig : profiles ] presents profiles for the lines averaged over the head of the filament . we show both the entire spectrum along the sight line as well as zoomed into the velocity range for the g328 filament , for clarity . the co and [ ci ] are clearly similar for the filament ( i.e. same line centre and narrow fwhm ) , but for hi this is evident as a dip in the profile , which extends considerably beyond the spectral limits shown . figure [ fig : cuts ] shows a spatial cut perpendicular to the filament in the integrated flux images . the filament is narrow and shows similar structure in co and [ ci ] . then their profiles along this cut are virtually the same . ] as for the velocity profiles , there is a dip in the hi flux across the filament . the reduction amounts to approx 40% of the flux level found either side of the filament . line parameters are then calculated for 3 different apertures associated with the filament . the first is for the integrated flux for its head , and the other two are for the peak pixels seen in the [ ci ] image . these apertures are defined in table [ tab : apertures ] and their line fluxes are listed in table [ tab : fluxes ] . @xmath10co optical depths are also calculated from the @xmath10co / @xmath11co flux ratio , @xmath29 ( i.e. @xmath30 with @xmath31 \sim 50 $ ] appropriate to a galactocentric radius of @xmath32kpc ; the derived values are only mildly sensitive to its likely variation see @xcite ) . the co optical depths are significant(@xmath33 ) and greatest at the brightest pixels . column densities for co , c and h are then given in table [ tab : coldens ] . these are calculated using a standard excitation analysis from the upper level column density given by radiative transfer ( e.g. * ? ? ? * ; * ? ? ? * ) , with standard notation : @xmath34 the column density for the species is then determined from @xmath35 here @xmath36 is the partition function , taken to be @xmath37k ) for co , 4 for hi , and from fitting to the tabulated values in the cologne database for molecular spectroscopy ( cdms ) for [ ci ] . the principal assumptions made in these calculations are of a beam filling factor of unity and in the value of the excitation temperature , here taken to be @xmath38k . @xmath39 is determined for co as described above and the optically thin limit is assumed for c. can be calculated and is found to be @xmath40 , so this assumption is self - consistent . ] motivating the temperature assumption , in the pdr model we discuss in [ sec : pdr ] the [ ci ] and co lines arise in regions of similar temperature of @xmath41k . also , combining the [ ci ] 809ghz flux with a low s / n spectrum for the 492ghz @xmath42[ci ] line obtained with the nanten2 telescope at the filament s peak ( glck et al . 2014 , in preparation ) yields a similar value . changing @xmath43 by @xmath44k likewise changes @xmath45 by about one - third . the boltzmann factor applied to the 60k energy level of the 809 ghz [ ci ] line makes @xmath46 particularly sensitive to lower values of @xmath43 . for 10k it would increase by an order of magnitude ; however for 30k @xmath46 falls by only a half . on the other hand , if @xmath43 were significantly less than 20k then the nanten2 492ghz flux would have been much higher , given the measured value for the 809ghz line . the range in the derived columns for @xmath45 and @xmath46 , considering all such likely variations in these parameters , does not alter the essential results that are given below . the column density of cold atomic gas is estimated from the hi self - absorption profiles following the method of @xcite and @xcite to determine the absorption optical depth . we use the difference in fluxes between the bottom of the absorption profile ( @xmath47 ) and its value if no absorption had occurred ( @xmath48 , obtained by interpolating across the line profile from the values either side ) to determine @xmath49 $ ] ( see table [ tab : fluxes ] ) , where @xmath50 is the spin temperature and @xmath51 is the fraction of the hi emission assumed to originate from behind the filament ( we have also implicitly assumed that the background continuum brightness temperature is very much less than the line brightness temperature in deriving this formula ) . we take @xmath52k , intermediate between the value in the cold molecular gas ( @xmath41k ) and an upper limit equal to the lowest value found in the hi data cube over the entire filament ( @xmath53k ) . following the discussion in @xcite , @xmath51 is likely to be close to 1 for hisa to be prominent ( but note also that it has minimum value given by @xmath54 in the limit that @xmath55 ; for the values used here this is @xmath56 ) . the column of cold atomic gas in the filament is then given by @xmath57 and is listed in table [ tab : coldens ] . for @xmath58(20k , 50k ) these columns should be multiplied by @xmath59 times , respectively . if only two - thirds of the hi emission was from behind the filament ( i.e. @xmath60 ) then the hi columns should be multiplied by a further factor of @xmath61 . for c and co the derived columns are of order a few @xmath62 . it is clear that the average c and co abundances are comparable . their ratio is a few times higher than that found in the dense pdr of carina , where nanten2 measurements found @xmath63 \sim 0.2 $ ] . for h the total column in emission is found to be @xmath64 in the velocity range of the filament ( and so assumed to be at its distance ) . however this gas is extended over the whole field seen in fig . [ fig : cicohi ] and so this column can not be directly associated with atomic gas around the filament , although its projected distance is only 1020pc away . the cold atomic gas is evident as the hisa and has a column estimated to be @xmath65 , though this figure could readily be increased by @xmath61 times , dependent upon the assumed value for @xmath50 , the background continuum and particularly the foreground hi contribution , as discussed above . along a sightline at @xmath66 the got c+ program with _ herschel _ has detected [ cii ] emission in the same narrow velocity range as the co , [ ci ] and hi discussed above . this sightline does not intersect the g328 filament , and lies @xmath67 away from its head , in a region where [ ci ] emission is absent and the co emission diffuse and weak ( see fig . [ fig : cicohi ] ) . the hi here , however , is seen in emission rather than absorption . thus , while this [ cii ] emission can not be used to determine the c@xmath68 column density associated with the filament , it does suggest that c@xmath68 emission is widespread and extends beyond the molecular cloud . the c@xmath68 column density at this position is estimated by langer et al . to be several times @xmath69 , so comparable in magnitude to that of the co and c we have derived for the head . striking in the images is the long , sinuous nature of the filament ( @xmath70 in all the spectral features ) , together with its narrow 4 km / s line width along this entire length . the central emission velocity ( @xmath71km / s ) yields a kinematic distance of ( 5.2 , 9.2)kpc ( near , far - solutions ) and a galactocentric radius of 5.0kpc . based on the dip in the hi profile along its length we assume the near - distance . the line centre is constant to within 1km / s along the filament s length . this yields a size of @xmath72pc , with any velocity gradient or shear along it constrained to be @xmath73km / s / pc . this latter limit is comparable to the shear that would produced by galactic rotation over this distance . the mean co line flux over the length of the filament ( @xmath74 over @xmath75 ) is similar to that over its head ( see table [ tab : apertures ] ) . applying the empirical conversion between flux and column ( @xmath76 $ ] ; ) to the integrated flux over the entire filament , and converting to a molecular mass yields @xmath77 ; i.e. comparable to that expected for a low - mass gmc . this implies an average number density @xmath78 in the filament , given its size ( and assuming a depth comparable to its width ) . the co luminosity is @xmath79 . the fwhm of the co line of 4 km / s is greater than the thermal width for cold gas ( @xmath80 ) so presumably it is mildly turbulent . however any shear along its length is low , being @xmath81km / s over its 75pc . the absence of far ir dust emission in the _ herschel _ images associated with the filament implies that the dust must be cold . assuming grain emissivities , @xmath82 , as given by , then for @xmath83k the fluxes would be , applying @xmath84 1000 , 400 & 150mjy / sr at 250 , 350 & 500@xmath7 m , respectively . these fluxes are comparable with the galactic background and so should be discernible in the _ herschel _ images . however , for @xmath85k they fall to 5030mjy / sr , and so would not readily be distinguished above the background . this implies an upper limit to the dust temperature of @xmath86k . the galactic background in the _ herschel _ images can also be used to obtain an upper limit for the dust luminosity ; we obtain @xmath87 , and correspondingly @xmath88 using the estimated molecular mass . these are strong limits ; if the dust emission from the filament was 20% of the background levels it would likely be discernible , tracing the outline of the filament in the _ herschel _ images . comparing to the values derived by @xcite and @xcite for @xmath89 for ir quiet and non - h ii region clouds ( i.e. those lacking in massive star formation ) , this places the upper limit for the g328 filament at the lower end of the range these authors found , and similarly for the corresponding range found in @xmath90 vs. @xmath91 . the g328 filament is thus underluminous compared to ir - quiet gmcs , suggesting that active star formation has not yet commenced . we note that these previous studies used clouds that were bright enough to be seen by _ iras _ against the galactic background and thus our @xmath89 ratio represents an unbiased and perhaps more representative value for a quiescent gmc . we have examined a range of cloud properties using a modified form of the @xcite photodissociation region code to calculate column densities and line emission to compare with the data . the model consists of a slab of gas illuminated by the interstellar radiation field on two sides . the gas temperature and the abundances of atomic / molecular species are calculated as a function of optical depth , @xmath92 , under the assumptions of thermal balance , chemical equilibrium and constant thermal pressure . we have made a simple approximation to include @xmath93 line transfer by assuming a constant @xmath94 abundance ratio of 50 . for details on the chemistry and thermal processes we refer to @xcite , @xcite , @xcite and @xcite . while not intended as a comprehensive attempt to fit the data , it has provided a representative model which approximates the observations for aperture # 2 . the parameters and results are shown in table [ tab : pdrmodel ] . this model adopts an interstellar radiation field of @xmath95 ( with @xmath96 incident on each side ; @xmath97 is the interstellar field strength in units of habing fields ; @xcite ) . this field is roughly consistent with interstellar fields expected at a galactocentric radius of 5kpc @xcite . somewhat surprisingly , the co and [ ci ] flux results are not strongly dependent on @xmath97 , as discussed in @xcite , because stronger fields drive the c to co transition deeper into the cloud , where the temperature is lower than at the surface . higher fields will increase the c@xmath68 1.9thz flux , however . the model shown in table [ tab : pdrmodel ] adopts a primary cosmic ray ionization rate of @xmath98 , consistent with rates for low column density diffuse clouds determined from observations of h@xmath99 @xcite , oh@xmath68 and h@xmath100o@xmath68 @xcite . there is some evidence that the ionization rate is depth dependent in clouds , varying from @xmath101 in low @xmath92 clouds to values of @xmath102 in the centers of molecular clouds . we present in table [ tab : pdrmodel ] a model with the high cosmic ray rate , but an equally good fit to the data can be obtained with the low cosmic ray rate . we find that the low cosmic ray rate produces a higher fraction of the total co cooling in the j=10 line , because the co gas is cooler . the observed co j=10 luminosity of the whole filament is of order 0.1 l@xmath103 . in the pdr models , this corresponds to a total co luminosity of the cloud of 36l@xmath103 , depending on the cosmic ray rate . for the low cosmic ray rate , the total co cooling is 3l@xmath103 and most of the heating of the co is provided by the fuv radiation at intermediate depths into the cloud . for the high cosmic ray rate the total co cooling is 6l@xmath103 , of which about half the heating is provided by the fuv radiation at intermediate depths and half by cosmic rays distributed through the entire co region . the pdr models that fit the data have a total cloud optical depth of @xmath104 mags . and the interior gas temperature where co 10 becomes optically thick is @xmath105 k with an @xmath106 density of 650 @xmath107 . note , however , that the model line fluxes are approximately twice that observed indicating that the beam filling factor is @xmath108 . for this beam filling factor , the estimated columns in table [ tab : coldens ] would be multiplied by a factor of @xmath109 . with this correction , the model column densities and line fluxes agree with observations to within a factor of 2 . we also provide estimates for the c 492 ghz and @xmath110 1.90 thz transitions . the models predict overlapping c@xmath68 , c and co emission zones but their peaks are separated by @xmath111 mags . , equivalent to @xmath108pc , which is not resolvable at the resolution of the heat telescope at 809ghz ( fig . [ fig : cuts ] ) , but should begin to be discernible were it at the resolution of the mopra dataset . we also note that the model incident field is @xmath96 . this field produces dust temperatures at the surface of @xmath112 k for silicates and @xmath113 k for carbonaceous grains @xcite . however , the the bulk of radiation field is absorbed , and the dust emission is produced at greater depth , where the field strength is lower . at @xmath114 the grain temperatures drop to 15 k and 20 k for silicates and carbonaceous grains respectively . the model dust temperatures are marginally within the limits set by the _ herschel _ observations . applying the empirical @xmath115 factor to the peak co flux and comparing to the columns for co and c yields [ c / h ] @xmath116 , about half of the measured gas - phase abundance of @xmath117 @xcite in diffuse gas . the @xmath110 column density is expected to contribute only an additional 10% to the total . this discrepancy between the estimated carbon abundances from our observations , and those measured in the diffuse gas and used in models , might simply arise from the inadequacy of using the x factor conversion to provide the total column . on the other hand , low carbon abundances ( e.g. [ co / h ] @xmath118 ) have been previously adopted in analysis of co observations ( e.g. , * ? ? ? * ; * ? ? ? * ) to match the data , suggesting that most of the non - refractory carbon is hidden is some form as yet undetected . the model in table [ tab : pdrmodel ] is shown only for representational purposes and is not a best fit from an exhaustive search of model parameter space . however it demonstrates that reasonable fits can be obtained without depleting the gas phase carbon from the observed diffuse cloud values . there is no `` missing carbon '' in these models , with the gas phase carbon ( @xmath117 in abundance relative to h nuclei ) contained in c@xmath68 at the surface , in c at intermediate depth , and then as co from about @xmath119 to the center of the cloud . most of the h nuclei are in the co zone , so the previous observations which ratio the co column to the h nuclei column and find a few @xmath120 can not be explained by additional h column in the non - co zone . our model explains the observed co and ci intensities and predicts c@xmath68 1.9thz intensities comparable to those observed by got c+ in a nearby line of sight . the models include freeze - out of atomic carbon and carbon molecules onto grain surfaces using the formalism discussed in @xcite . it is possible that some of the co may be lost from the gas phase due to co ice formation . for the model presented in table [ tab : pdrmodel ] , at cloud center about 40% of the carbon is in co ice . however , the freeze - out zone extends for only a short column in @xmath92 and the column density of co ice amounts to only @xmath121% of the co gas - phase column density . it is likely that there exists and extended , low density atomic halo around the g328 molecular filament , as evident through the hi emission that surrounds it , together with the c@xmath68 emission detected along the got c+ sightline . the filament would be expected to connect to the ambient ism . however the presence of such an extended halo will have little affect on the pdr model we have presented , as it would not contribute either co or [ ci ] line emission . it would , however , provide low density c@xmath68 gas that extends much further from the filament than the constant thermal pressure assumption inherent in our pdr model allows for . mapping of the extended [ cii ] emission will be necessary to address such questions regarding the connection of the molecular to atomic mediums , as well as the possible presence of `` dark '' molecular gas , where co is not present and the carbon predominantly in the form of c@xmath68 . we have found a quiescent , filamentary molecular cloud in the g328 region , situated about 5kpc from the sun and extending @xmath72pc , with a mass @xmath2 , comparable to that of a low mass gmc . the filament is evident in co , [ ci ] and in hi self - absorption . the line width is small , @xmath122km / s along its length , and no velocity gradient is discernible . over the sightline through the cloud the average abundance of atomic carbon is similar to that of carbon in its molecular form ( i.e. co ) . the filament is not discernible in far ir continuum images , and the flux limits imply that the dust must be cold , @xmath86k . the limits on its luminosity place it at the low end of the range in @xmath89 previously found for ir - quiet gmcs . no clear evidence for star formation in the filament can be seen in the ir images , consistent with no significant internal heating . the molecular cloud is thus cold and has limited internal turbulence . we speculate that this filament represents a gmc that has recently formed , or indeed may still be in the process of forming . star formation has not yet progressed far within it , and there are no , or limited , sources of turbulent energy injection into the gas . this hypothesis can be tested through further wide - field thz spectral imaging , in both the [ cii ] 158@xmath7 m ( 1.90 thz ) line and the two [ ci ] lines . [ cii ] would be expected to also show a similar structure to the co and [ ci ] . as the gas transitions from atomic to molecular form a high ratio of c@xmath68 and c to co is expected , for the external far uv radiation field will limit the amount of gas that can exist as co until sufficient column density has been built up to shield it from dissociating radiation . it would be of particular interest to determine whether the c and c@xmath68 associated with the filament has a more extended distribution than the co as this scenario would suggest . it is first planned to upgrade heat to operate at 1.9thz to measure the extended c@xmath68 . then a 5 m diameter thz telescope , such as the proposed date5 for dome a in antarctica , and the 0.8 m balloon - borne sto-2 , to be launched from mcmurdo station in antarctica , in late 2015 , would provide similar resolution to the co in the [ ci ] and [ cii ] lines , accordingly , allowing one to examine this question further . funding for the heat telescope is provided by the national science foundation under grant number plr-0944335 . plato - r was funded by astronomy australia limited , as well as the university of new south wales , as an initiative of the australian government being conducted as part of the super science initiative and financed from the education investment fund . logistical support for heat and plato - r is provided by the united states antarctic program . the mopra radio telescope is part of the australia telescope national facility . operations support was provided by the university of new south wales and the university of adelaide . the unsw digital filter bank ( the mops ) used for the observations with mopra was provided with financial support from the australian research council ( arc ) , unsw , sydney and monash universities . we also acknowledge arc support through discovery project dp120101585 . ; k km / s ) for the g328 region across @xmath123 to @xmath71 km / s velocity range of the filament . from top - left , going anti - clockwise : the heat [ ci ] 809 ghz , mopra @xmath10co j=10 115ghz and parkes atca 21 cm hi images . the final image overlays contours of co on the _ herschel _ 250@xmath7 m image . coordinates are galactic . [ fig : cicohi ] ] integrated flux ( green ) and @xmath10co ( red ) on an image of the hi ( greyscale ; note that white is low and black is high ) for the @xmath123 to @xmath71 km / s velocity range of the g328 filament . the data for [ ci ] only extends for @xmath124 , as indicated by the green box . the flux scale bar in k km / s is for the hi . [ ci ] contours are drawn at 1 , 2 & 3 k km / s ; @xmath10co contours at 10 , 20 & 30 k km / s ( all @xmath125 ) . [ fig : contours ] ] . the lines shown are @xmath10co ( solid ) , @xmath11co ( dotted ) , [ ci ] ( dash - dot ) and hi/50 ( dash - dot - dot ) . the top plot shows the entire spectrum at this location , and the bottom plot is zoomed in to show the emission around the -76 to -80kms@xmath126 velocity range analysed in this paper . in the bottom plot the @xmath11co and [ ci ] lines are also multiplied by a factor 5 for clarity , whereas the hi is divided by 50 in each plot . the hisa is evident as the dip in the hi profile.,title="fig : " ] + . the lines shown are @xmath10co ( solid ) , @xmath11co ( dotted ) , [ ci ] ( dash - dot ) and hi/50 ( dash - dot - dot ) . the top plot shows the entire spectrum at this location , and the bottom plot is zoomed in to show the emission around the -76 to -80kms@xmath126 velocity range analysed in this paper . in the bottom plot the @xmath11co and [ ci ] lines are also multiplied by a factor 5 for clarity , whereas the hi is divided by 50 in each plot . the hisa is evident as the dip in the hi profile.,title="fig : " ] co ( solid ) , @xmath11co ( dotted ) , [ ci ] ( dash - dot ) and hi ( dash - dot - dot ) . the cut runs from @xmath127 to @xmath128 , with the offset given in arcminutes on the bottom axis and in parsecs on the top ( assuming a source distance of 5.2kpc ) . the triangles show the @xmath129 line ratios across the cut , as indicated on the right - hand axis . [ fig : cuts ] ] cccccccccc 1 & @xmath130 & @xmath131 & @xmath132 & @xmath133 & @xmath134 & 13 & 58 & 105 & 1.0 + 2 & @xmath135 & @xmath136 & @xmath137 & @xmath138 & @xmath139 & 18 & 67 & 118 & 0.9 + 3 & @xmath140 & @xmath141 & @xmath142 & @xmath143 & @xmath144 & 15 & 63 & 115 & 1.0 + lc @xmath92 & @xmath154 mags . + @xmath155 & @xmath156 k @xmath107 + g@xmath157 & 3 habings + @xmath158 & 2.4 km s@xmath126 + @xmath159 & @xmath160 @xmath161 + @xmath162 $ ] abundance & @xmath163 + n(h ) & @xmath164 + n(c@xmath68 ) & @xmath165 + n(c ) & @xmath166 + n(co ) & @xmath167 + i(@xmath10co 10 115 ghz ) & @xmath168 + i(@xmath93 10 110 ghz ) & @xmath169 + i([ci ] 492 ghz ) & @xmath170 + i([ci ] 810 ghz ) & @xmath171 + i([cii ] 1.90 thz ) & @xmath172 + @xmath173 & 25 k + @xmath174 & 650 @xmath107	we present spectral line images of [ ci ] 809ghz , co j=10 115 ghz and hi 1.4ghz line emission , and calculate the corresponding c , co and h column densities , for a sinuous , quiescent giant molecular cloud about 5kpc distant along the @xmath0 sightline ( hereafter g328 ) in our galaxy . the [ ci ] data comes from the high elevation antarctic terahertz ( heat ) telescope , a new facility on the summit of the antarctic plateau where the precipitable water vapor falls to the lowest values found on the surface of the earth . the co and hi datasets come from the mopra and parkes / atca telescopes , respectively . we identify a filamentary molecular cloud , @xmath1pc long with mass @xmath2 and a narrow velocity emission range of just 4km / s . the morphology and kinematics of this filament are similar in co , [ ci ] and hi , though in the latter appears as self - absorption . we calculate line fluxes and column densities for the three emitting species , which are broadly consistent with a pdr model for a gmc exposed to the average interstellar radiation field . the [ c / co ] abundance ratio averaged through the filament is found to be approximately unity . the g328 filament is constrained to be cold ( @xmath3k ) by the lack of far ir emission , to show no clear signs of star formation , and to only be mildly turbulent from the narrow line width . we suggest that it may represent a gmc shortly after formation , or perhaps still be in the process of formation .
correlations play a key role both in classical and quantum physics . in particular the study of correlations is crucial in many - body physics and classical and quantum statistical physics . recently , it turned out that correlations play prominent role in quantum information theory and many modern applications of quantum technologies and there are dozens of papers dealing with this problem ( for the recent review see e.g. @xcite ) . the aim of this paper is to analyze classical and quantum correlations encoded in the bi - partite quantum states . beside quantum entanglement we analyze a new measure so called @xmath0-correlations and the quantum discord . we propose to compare correlations of different bi - partite states with the same reduces states , i.e. locally they contain the same information . it is shown that surprisingly a separable state may be more correlated that an entangled one . analyzing simple examples of bell diagonal states we illustrate the behavior of various measures of correlations . we also provide an introduction to bi - partite states and entanglement mappings introduced by belavkin and ohya and recall basic notions from classical and quantum information theory . an entanglement mapping encodes the entire information about a bi - partite quantum state and hence it provides an interesting way to deal with entanglement theory . interestingly , it may be applied in infinite - dimensional case and in the abstract @xmath1-algebraic settings . therefore , in a sense , it provides a universal tool in entanglement theory . the paper is organized as follows : in the next section we recall basic facts from the theory of composite quantum systems and introduce the notion of entanglement mappings . moreover , we recall an interesting construction of quantum conditional probability operators . section [ e ] recall classical and quantum entropic quantities and collects basic facts from classical and quantum information theory . in particular it contains the new measure of correlation called @xmath0-correlation . section [ discord ] recalls the notion of _ quantum discord _ which was intensively analyzed recently in the literature . in section [ circ ] we recall the notion of a circulant state and provide several examples of states for which one is able to compute various measures of correlations . final conclusions are collected in the last section . throughout the paper , we use standard notation : @xmath2 for complex separable hilbert spaces and denote the set of the bounded operators and the set of all states on @xmath3 by @xmath4 and @xmath5 , respectively . in the @xmath6-dimensional hilbert space , the standard basis is denoted by @xmath7 and the inner product is denoted by @xmath8 . we write @xmath9 for @xmath10 . given any state @xmath11 on the tensor product hilbert space @xmath12 , we denote by @xmath13 the partial trace of @xmath11 with respect to @xmath14 . consider a quantum system living in the hilbert space @xmath3 . in this paper we consider only finite dimensional case . however , as we shall see several results may be nicely generalized to the infinite - dimensional setting . denote by @xmath15 a set of trace class operators in @xmath3 , meaning that @xmath16 if @xmath17 and @xmath18 , which is always true in finite - dimensional case . finally , let @xmath19 consider now a composite system living in @xmath20 and denote by @xmath21 a convex subset of separable states in @xmath22 . recall that @xmath23 is separable if @xmath24 , where @xmath25 and @xmath26 , and @xmath27 denotes probability distribution : @xmath28 an @xmath29 . a state @xmath30 is called positive partial transpose ( ppt ) if its partial transpose satisfies @xmath31 , where @xmath32 denotes an identity map in @xmath33 . it means that @xmath34 is ppt if @xmath35 . denote by @xmath36 a convex subset of ppt states . it is well known @xcite that @xmath37 . in general , the ppt condition is not sufficient for separability . interestingly , due to the well known duality between states living in @xmath20 and linear maps @xmath38 , one may translate the above setting in terms of linear maps . let us recall basic facts concerning completely positive maps @xcite . a linear map @xmath39 is said to be completely positive ( cp ) if , for any @xmath40 , the map @xmath41 is positive , where @xmath4 denotes bounded operators in @xmath3 and @xmath42 stands for @xmath43 matrices with entries in @xmath44 . a linear map @xmath45 is said to be completely copositive ( ccp ) if composed with transposition @xmath46 , i.e. @xmath47 , is cp . consider now a state @xmath48 and let @xmath49 be a linear map defined by @xmath50\ , \ ] ] for any @xmath51 . the dual map @xmath52 reads @xmath53 \ , \ ] ] for any @xmath54 . it should be stressed that the above construction is perfectly well defined also in the infinite - dimensional case if wew assume that @xmath11 is a normal state , that is , it is represented by the density operator . note , that a state @xmath11 and the linear map @xmath55 give rise a linear functional @xmath56 @xmath57 for any @xmath58 . this formula may be equivalently rewritten as follows @xmath59 it is clear that the marginal states read @xmath60 belavkin and ohya observed @xcite that if @xmath48 , then both @xmath55 and its dual @xmath52 are ccp . we denote by @xmath61 the dual space to the algebra @xmath61 . a ccp map @xmath62 normalized as @xmath63 is called the entanglement map from @xmath64 to @xmath65 . a density operator @xmath66 corresponding to the entanglement map @xmath55 with its marginals @xmath67 and @xmath68 can be represented as follows : let @xmath69 denotes a maximally entangled state in @xmath70 . then @xmath71 with @xmath72 , where @xmath73 . if @xmath74 stands for an orthonormal basis in @xmath14 , then @xmath75 with @xmath76 , and hence @xmath77 the map assigning @xmath78 to @xmath55 is usually called a choi - jamiokowski isomorphism . it should be stressed that @xmath78 does not depend upon the choice of @xmath79 . [ w - criterion ] a linear map @xmath80 is ccp if and only if @xmath81 . clearly , @xmath82 is cp if and only if @xmath83 is ccp . due to lemma [ w - criterion ] , we have the following criterion . @xcite a state @xmath66 is a ppt state if and only if its entanglement map @xmath82 is cp . recently , kossakowski et al.@xcite proposed the following construction : for @xmath84 one defines the bounded operator @xmath85 where @xmath86 . it is verified that @xmath87 satisfies @xmath88 in what follows we assume that @xmath34 is a faithful state , i.e. @xmath89 . it follows from ( [ con1 ] ) and ( [ con2 ] ) that the operator @xmath87 is the quantum analogue of a classical conditional probability . indeed , if @xmath90 is replaced by commutative algebra , then @xmath87 coincides with a classical conditional probability . an operator @xmath91 is called the quantum conditional probability operator ( qcpo , for short ) if @xmath92 satisfies condition ( [ con1 ] ) and ( [ con2 ] ) . it is easy to verify@xcite that for any cp unital map @xmath93 and an orthonormal basis in @xmath14 the following operator @xmath94 defines qcpo . from lemma [ w - criterion ] and unitality of @xmath95 , it follows that @xmath96 satisfies conditions ( [ con1 ] ) and ( con2 ) . for a given @xmath97 and any faithful marginal state @xmath98 , one can construct a state @xmath11 of the composite system @xmath99 it is clear that @xmath100 is a ppt state if and only if the map @xmath95 is a ccp . there exists a simple relation between the density operator @xmath78 in ( [ def1 ] ) and the qcpo @xmath101 in ( def2 ) due to the following decomposition of the entanglement map @xmath82 . @xcite every entanglement map @xmath82 with @xmath102 has a decomposition @xmath103 where @xmath104 is a cp unital map to be found as a unique solution to @xmath105 @xcite if a composite state @xmath66 given by ( [ def1 ] ) has a faithful marginal state @xmath106 , then @xmath78 is represented by @xmath107 where @xmath108 . in classical description of a physical composite system its correlation can be represented by a joint probability measure or a conditional probability measure . in classical information theory we have proper criteria to estimate such correlation , which are so - called the mutual entropy and the conditional entropy given by shannon @xcite . here we review shannon s entropies briefly . let @xmath109 and @xmath110 be random variables with probability distributions @xmath111 and @xmath112 , respectively , and let @xmath113 denotes conditional probability @xmath114 . the joint probability @xmath115 is given by @xmath116let us recall definitions of mutual entropy @xmath117 and conditional entropies @xmath118 : @xmath119and @xmath120 using ( [ fundac ] ) , we can easily check that the following relations @xmath121 and @xmath122 where @xmath123 , and @xmath124 . note , that @xmath113 gives rise to a stochastic matrix @xmath125 and hence it defines a classical channel @xmath126 note , that data provided by @xmath127 are the same as those provided by @xmath128 and @xmath129 . hence one may instead of @xmath117 use the following notation @xmath130 , where @xmath131 represent an _ input _ state and @xmath132 the classical channel . one interprets @xmath130 as a information transmitted _ via _ a channel @xmath132 . the fundamental shannon inequality @xmath133 gives the obvious bounds upon the transmitted information . now , we extend the classical mutual entropy to the quantum system using the umegaki relative entropy.@xcite let @xmath134 with marginal states @xmath135 and @xmath136 . one defines quantum mutual entropy as a relative entropy between @xmath11 and the product of marginals @xmath137 : @xmath138 \bigr ) \}\ .\ ] ] as in the classical case one shows that @xmath139 introducing quantum conditional entropy @xmath140 one finds @xmath141 or , equivalently @xmath142 @xcite for any entanglement map @xmath143 with @xmath144 and @xmath145 , the quantum mutual entropy @xmath146 is defined by @xmath147 \bigr ) \}\ , \ ] ] where @xmath148 is the umegaki relative entropy . one easily finds @xmath149 the above relation ( [ relation4 ] ) is a quantum analog of ( [ relation1 ] ) . one defines the quantum conditional entropies as generalizations of ( [ relation2 ] ) , ( [ relation3 ] ) @xcite : @xmath150 it is usually assumed that @xmath151 measures all correlations encoded into the bipartite state @xmath78 with marginals @xmath34 and @xmath152 . for the entanglement map @xmath153 one finds @xmath154 , and hence @xmath155 which recover well known relations for a product state @xmath156 . [ example : pure entangled ] let @xmath157 be the sequence of complex numbers satisfying @xmath158 . for entanglement mappings @xmath159 where @xmath160 and @xmath161 are orthonormal basis in @xmath3 and @xmath14 , respectively , the state @xmath66 can be written in the following form @xmath162 where @xmath163 note , that @xmath164 equals to the schmidt rank of @xmath165 . one finds for the reduced states @xmath166 and hence @xmath167 together with @xmath168 where @xmath169 . as is mentioned in section 2 , the classical mutual entropy always satisfies the shannon s fundamental inequality , i.e. it is always smaller than its marginal entropies , and the conditional entropy is always positive . note that separable state has the same property . it is no longer true for pure entangled states . now we introduce another measure for correlation of composite states.bo1,bo2,ckmo,mo for the entanglement map @xmath170 , we define the @xmath0-correlation @xmath171 of @xmath11 as @xmath172 note that the @xmath0-correlation with the opposite convention @xmath173 is called the degree of entanglement.@xcite one proves the following : [ section4:purecase ] @xcite if @xmath78 is a pure state , then the following statements hold : 1 . 2 . it is well - known that if @xmath11 is a ppt state , then @xmath174 where @xmath175 and @xmath176 are the marginal states of @xmath11.@xcite [ section4:mixedcase ] if @xmath11 is a ppt state , then @xmath177 suppose now that we have two entanglement mappings @xmath178 such that @xmath179 and @xmath180 . let @xmath181 be the corresponding states . we propose the following : @xmath182 is said to have stronger @xmath0-correlations than @xmath183 if @xmath184 several measures of correlation based on entropic quantities were already discussed by cerf and adami@xcite , horodecki@xcite , henderson and vedral@xcite , groisman et al.@xcite . let us briefly recall the definition of quantum discord @xcite . recall , that mutual information may be rewritten as follows @xmath185 an alternative way to compute the conditional entropy @xmath186 goes as follows : one introduces a measurement on @xmath3-party defined by the collection of one - dimensional projectors @xmath187 in @xmath3 satisfying @xmath188 . the label ` @xmath189 ' distinguishes different outcomes of this measurement . the state after the measurement when the outcome corresponding to @xmath190 has been detected is given by @xmath191 where @xmath192 is a probability that @xmath3-party observes @xmath189th result , i.e. @xmath193 , and @xmath194 is the ( collapsed ) state in @xmath20 , after @xmath3-party has observed @xmath189th result in her measurement . the entropies @xmath195 weighted by probabilities @xmath192 yield the conditional entropy of part @xmath14 given the complete measurement @xmath187 on the part @xmath3 @xmath196 finally , let @xmath197 be the corresponding measurement induced mutual information . the quantity @xmath198 is interpreted @xcite as a measure of classical correlations . now , these two quantities @xmath199 and @xmath200 may differ and the difference @xmath201 is called a quantum discord . evidently , the above definition is not symmetric with respect to parties @xmath3 and @xmath14 . however , one can easily swap the role of @xmath3 and @xmath14 to get @xmath202 where @xmath203 and @xmath204 is a collection of one - dimensional projectors in @xmath14 satisfying @xmath205 . for a general mixed state @xmath206 . however , it turns out that @xmath207 . moreover , on pure states , quantum discord coincides with the von neumann entropy of entanglement @xmath208 . states with zero quantum discord so called classical - quantum states represent essentially a classical probability distribution @xmath192 embedded in a quantum system . one shows that @xmath209 if and only if there exists an orthonormal basis @xmath210 in @xmath3 such that @xmath211 where @xmath212 are density matrices in @xmath14 . similarly , @xmath213 if and only if there exists an orthonormal basis @xmath214 in @xmath14 such that @xmath215 where @xmath216 are density matrices in @xmath3 . it is clear that if @xmath217 , then @xmath11 is diagonal in the product basis @xmath218 and hence @xmath219 . finally , let us introduce a symmetrized quantum discord @xmath220 \ .\ ] ] let us observe that there is an intriguing relation between ( [ d - symm ] ) and ( [ den ] ) . one has @xmath221 \ , \ ] ] whereas @xmath222 note , that @xmath223 but @xmath171 can be negative ( for ppt states ) . it is assumed that @xmath224 measures perfectly quantum correlations encoded into @xmath11 . [ example : separable ] for the entanglement map given by @xmath225 the corresponding state @xmath11 can be written in the form @xmath226 with @xmath227 and @xmath228 . then , we have the following inequalities.@xcite @xmath229 let @xmath230 and @xmath231 be the complete orthonormal systems in @xmath3 and @xmath14 , respectively . for the entanglement map given by @xmath232 the corresponding state @xmath11 can be written in the form @xmath233 with @xmath234 . it is clear that @xmath235 . moreover , one obtains @xmath236 where @xmath237 . this correlation corresponds to a perfect correlation in the classical scheme . in this section , we analyze correlations encoded into the special family of so called _ circulant states_. we start this section by recalling the definition of a circulant state introduced in @xcite ( see also @xcite ) . consider the finite dimensional hilbert space @xmath238 with the standard basis @xmath239 . let @xmath240 be the subspace of @xmath241 generated by @xmath242 @xmath243 define a shift operator @xmath244 by @xmath245 and let @xmath246 it turns out that @xmath247 and @xmath248 are mutually orthogonal and one has the following direct sum decomposition @xmath249 this decomposition is called a circulant decomposition.@xcite let @xmath250 be positive @xmath251 matrices with entries in @xmath44 such that @xmath216 is supported on @xmath247 . moreover , let @xmath252 now , for each @xmath253 one defines a positive operator in @xmath254 be the following formula @xmath255 finally , let us introduce @xmath256 one proves@xcite that @xmath34 defines a legitimate density operators in @xmath254 . one calls it a _ circulant state_. for further details of circulant states we refer to refs . ckcir07,art . now , let consider a partial transposition of the circulant state . it turns out that @xmath257 is again circulant but it corresponds to another cyclic decomposition of the original hilbert space @xmath254 . let us introduce the following permutation @xmath92 from the symmetric group @xmath258 : it permutes elements @xmath259 as follows @xmath260 we use @xmath92 to introduce @xmath261 and @xmath262 it is clear that @xmath263 and @xmath264 are mutually orthogonal ( for @xmath265 ) . moreover , @xmath266 and hence it defines another circulant decomposition . now , the partially transformed state @xmath267 has again a circulant structure but with respect to the new decomposition ( [ d - new ] ) : @xmath268 where @xmath269 and the new @xmath270 matrices @xmath271 $ ] are given by the following formulae : @xmath272 where @xmath273 " denotes the hadamard product , matrices @xmath274 $ ] and @xmath275 $ ] is defined by @xmath276 ] and @xmath277 being a @xmath270 permutation matrix corresponding to @xmath92 , i.e. @xmath278 . it is therefore clear that our original circulant state is ppt iff all @xmath6 matrices @xmath279 satisfy @xmath280 the most important example of circulant states is provided by bell diagonal states @xcite defined by @xmath281 where @xmath282 , @xmath283 and @xmath284 with @xmath285 being the collection of @xmath286 unitary matrices defined as follows @xmath287 with @xmath288 the matrices @xmath285 define an orthonormal basis in the space @xmath289 of complex @xmath270 matrices . one easily shows @xmath290 some authors call @xmath285 generalized spin matrices since for @xmath291 they reproduce standard pauli matrices : @xmath292 let us observe that bell diagonal states ( [ bell ] ) are circulant states in @xmath254 . indeed , maximally entangled projectors @xmath293 are supported on @xmath294 , that is , @xmath295 defines a projector onto @xmath294 , i.e. @xmath296 one easily shows that the corresponding matrices @xmath297 are given by @xmath298 where @xmath299 is a unitary @xmath251 matrix defined by @xmath300 and @xmath301 is a collection of diagonal matrices defined by @xmath302 one has @xmath303 and hence it defines a circulant matrix @xmath304 where the vector @xmath305 is the inverse of the discrete fourier transform of @xmath306 ( @xmath307 is fixed ) . let @xmath308 . for any @xmath309 $ ] , one defines@xcite the following state @xmath310 the eigenvalues of @xmath311 are calculated as @xmath312 and @xmath313 and hence one obtains for the @xmath0-correlations @xmath314 [ section5:thm1 ] @xcite the family @xmath315 satisfies : 1 . @xmath315 is ppt if and only @xmath316 $ ] 2 . @xmath315 is separable if and only if @xmath317 $ ] 3 . @xmath315 is both entangled and ppt if and only if @xmath318 $ ] 4 . @xmath315 is npt if and only if @xmath319 $ ] . due to this theorem , one can find that the @xmath320 does admit a natural order . that is , the @xmath0-correlation for any entangled state is always stronger than @xmath0-correlation for an arbitrary separable state . similarly , one observes that @xmath0-correlation for any npt state is always stronger than @xmath0-correlation for an arbitrary ppt state . the graph of @xmath321 is shown in fig . [ section5:fig1 ] . actually , one finds that the minimal value of @xmath0-correlations corresponds to @xmath322 , that is , it lies in the middle of the separable region . on the other hand , we can also compute the symmetrized discord @xmath323 and have obtained fig . [ section5:fig1 ] . it is easy to find that the graph is symmetric with respect to @xmath324 . as in fig . 2 , the value of the symmetrized discord satisfies the following inequality ; @xmath325 where @xmath326,\,\beta \in \left[1,2\right ] \cup \left[3,4\right]$ ] and @xmath327 \cup \left[4,5\right]$ ] . the family of @xmath328 has the quantum correlation even in separable states corresponding to @xmath326 $ ] in the sense of discord . we know that the above two types of criteria give the similar order of correlation . notice that @xmath329 is always negative even in npt sates and the positivity of @xmath0-correlation represents a true quantum property ( see example 3.3 and proposition 3.5 ) . in this sense the quantum correlation of @xmath328 is not so strong . with @xmath330 $ ] . the minimal value of @xmath0 corresponds to @xmath331 . right the graph of @xmath332 . , title="fig:",width=226 ] with @xmath330 $ ] . the minimal value of @xmath0 corresponds to @xmath331 . right the graph of @xmath332 . , title="fig:",width=226 ] this family may be generalized to @xmath254 as follows : consider the following family of circulat 2-qudit states @xmath333 with @xmath334 , and @xmath335 . let us take the following special case corresponding to @xmath336 and @xmath337 , with @xmath338 one may prove the following@xcite the family @xmath339 satisfies : 1 . @xmath339 is ppt if and only @xmath340 $ ] 2 . @xmath339 is separable if and only if @xmath341 $ ] 3 . @xmath315 is both entangled and ppt if and only if @xmath342 $ ] 4 . @xmath315 is npt if and only if @xmath343 $ ] . for example if @xmath344 one obtains the following picture of @xmath345 ( see fig . 4 ) with @xmath346 $ ] . the minimal value of @xmath0 corresponds to @xmath347.,width=226 ] again , one finds that the @xmath348 does admit a natural order . that is , the @xmath0-correlation for any entangled state is always stronger than @xmath0-correlation for an arbitrary separable state . similarly , one observes that @xmath0-correlation for any npt state is always stronger than @xmath0-correlation for an arbitrary ppt state . consider the following class of bell - diagonal states in @xmath349 : @xmath350 with @xmath351 . one easily finds for its @xmath352-correlations @xmath353 the following theorem gives us a useful characterization of @xmath354 @xcite . [ section3:thm2 ] the states of @xmath355 are classified by @xmath356 as follows : 1 . @xmath357 is separable if @xmath358 2 . @xmath357 is both ppt and entangled for @xmath359 . the graph of @xmath360 is shown in fig . [ section3:fig2 ] . @xmath360 is rapidly decreasing with @xmath356 approaching @xmath361 from @xmath362 and increases when @xmath356 is over @xmath361 . that is , @xmath360 takes the minimal value at @xmath358 and it is approximated about @xmath363 . as is the case of @xmath315 , the @xmath0-correlation @xmath364 for an entangled state is always stronger than the one for a separable state . as @xmath365 or @xmath366 , @xmath367 converges to a separable perfectly correlated state which can be recognized as a classical state @xmath368 @xmath369 and for every @xmath370 , @xmath371it shows that a correlation of a ppt entangled state @xmath372 is weaker than that of the ( classical ) separable perfectly correlated states in the sense of ( [ order ] ) . now , since @xmath311 and @xmath373 have common marginal states , we can compare the order of quantum correlations for them . one has , for example , @xmath374accordingly theorem [ section5:thm1 ] and [ section3:thm2 ] , however , @xmath375 is separable while @xmath376 is entangled state . incidentally , this means that the correlation for the separable state @xmath375 is stronger than the entangled state @xmath376 in the sense of ( [ order ] ) . . note that @xmath0 is minimal for @xmath377 which correspond to the separable state . right the graph of @xmath378 for @xmath379 $ ] . note that @xmath380 . , title="fig:",width=226 ] . note that @xmath0 is minimal for @xmath377 which correspond to the separable state . right the graph of @xmath378 for @xmath379 $ ] . note that @xmath380 . , title="fig:",width=226 ] on the other hand one finds the following plot of the quantum discord fig . it is clear that @xmath381 since both @xmath382 and @xmath383 are perfectly classical states . note , that @xmath384 which shows that separable state @xmath385 does contain quantum correlations . we provided several examples of bi - partite quantum states and computed two types of correlations for them . it turned out that the correlation for a separable state can be stronger than the one for an entangled state in the sense of ( [ order ] ) . this observation is inconsistent with the conventional understanding of quantum entanglement . however , we also showed that the discord of such separable states might strictly positive . this means that these states have a non - classical correlation . from this point of view , it is no longer unusual that the correlation for a separable state is stronger than the one for an entangled state . t.m . is grateful to v.p.belavkin for fruitful discussions on mutual entropy and entanglement maps . d.c . was partially supported by the national science centre project dec-2011/03/b / st2/00136 . we would like to acknowledge the supports of qbic ( quantum bio - informatics center ) grant of tokyo university of science .	we study the correlations of classical and quantum systems from the information theoretical points of view . we analyze a simple measure of correlations based on entropy ( such measure was already investigated as _ the degree of entanglement _ by belavkin , matsuoka and ohya ) . contrary to naive expectation , it is shown that separable state might possesses stronger correlation than an entangled state .
broad emission - line regions ( blrs ) in active galactic nuclei ( agns ) have been the subject of extensive studies for more than two decades . these regions are not spatially resolved and information about their size , and the gas distribution , can only be obtained via reverberation mapping . understanding such regions requires a combination of detailed observations with sophisticated theoretical models . the extensive monitoring campaigns , of the last decade , focused on the broad line emission , variability , and response of the broad emission lines to the continuum variations in more than a dozen agns ( see peterson 1993 and netzer & peterson 1997 , for reviews ) . a major result is that in all well studied cases , the blr is stratified with a typical size of order of light weeks in seyfert galaxies and light months in quasars . present day modeling of such stratified blrs leaves much to be desired . while time - averaged spectra of agns can reasonably be explained by photoionization models ( netzer 1990 and references therein ) , there is a lack of detailed models fitting to the actual gas distribution , and motion , as deduced from the reverberation studies . this paper investigates the physical conditions in the line emitting gas of a well studied agn by incorporating modern photoionization calculations into a detailed geometrical model . our approach , that requires the simultaneous fit of several emission line light curves , enables us to constraint several important properties of the blr , such as the gas density , column density and level of ionization . the model is rather complex and is , therefore , limited to a spherical blr with no attempt to reconstruct the velocity field . in 2 we present the model , in 3 we apply it to the seyfert 1 galaxy ngc 5548 , and in 4 we discuss the new results in view of past studies . years of intensive agn monitoring resulted in a wealth of information on the blr response to continuum variations . excellent data sets are now available for half a dozen sources and less complete sets on a dozen or so more . unfortunately , theoretical understanding lags behind and there are few , if any , systematic attempts to produce complete models for the objects in questions . most theoretical efforts have been limited to modeling of only one or two emission lines . they did not consider real limitations such as the gas density and column density . our goal is to reconstruct _ all the observed _ light curves . this will enable us to set stringent limits on the physical parameters of the gas . more specifically , the aim is to deduce the run of density , column density and cloud distribution across the blr . we will demonstrate that the time dependent relative and absolute line intensities , and their relationship to the continuum fluctuations , leave little freedom in such models . in particular , the density and column density are restricted to a rather limited range . we wish to use as many observational constraints as possible and , therefore , limit ourselves to the best available data sets . we need both recombination lines , to constraint the covering factor , and collisionally excited lines to set limits on the run of density and temperature . data sets including both high and low ionization lines are preferred since , in this way , we can set limits on the run of ionization parameter as a function of distance from the continuum source . most importantly , we chose to rely as little as possible on balmer line intensities since , as explained below , their calculated intensities are subjected to large theoretical uncertainties . thus , the chosen data set must include a large enough number of uv lines and a well sampled , preferably large amplitude variable continuum . the only information that will not be used in this work is line profiles . given the above limitations , there are only 4 data sets that we consider suitable : 1 . ngc 4151 - see crenshaw et al . ( 1996 ) and references therein 2 . ngc 5548 ( clavel et al . 1991 ) 3 . ngc 3783 ( reichert et al . 1994 ) 4 . ngc 7469 ( wanders et al . 1997 ) . out of those we chose to model the 1989 uv data of ngc 5548 . this data set is unique in terms of its relatively large line and continuum variations , the _ combination _ of long duration with intensive regular 4-day sampling , and complete information on all strong uv lines ( see however comments about @xmath02798 in the discussion section ) . other advantages of this light curve are discussed below . in modeling the gas distribution in the nucleus , we distinguish between direct and indirect methods . direct methods are those where one guesses a gas distribution and calculates the resulting line light curves by putting , through this geometry , the observed ( or deduced ) continuum light curve . the result is compared with the observed emission - line light curves . indirect methods attempt to obtain the gas distribution by computing transfer functions . this procedure is somewhat ill - defined since it produces responsivity maps , rather than mass - distribution maps . it therefore requires the additional confirmation that the so - obtained responsivity maps are consistent with photoionization calculations . there have been several attempts to produce transfer functions for various lines in ngc 5548 , we are not aware of any successful mapping that is consistent with full photoionization calculations ( see also maoz 1994 ) . several previous attempts to model the blr in ngc 5548 differ from the present study in several important ways . the first systematic attempt to reconstruct the blr is by krolik et al . these authors used photoionization calculations to find the parameters that best fit the mean observed line rations . in parallel , they have used the maximum entropy method to reconstruct several transfer functions . based on these transfer functions , they modeled a spherically symmetric blr with locally isotropic line emitting clouds . the blr is divided into two distinct zones : one emitting the high - ionization lines ( column density of @xmath9 @xmath6 and ionization parameter of 0.3 ) and the other emitting the low - ionization lines ( column density of @xmath10 @xmath6 and ionization parameter of 0.1 ) . obrien , goad , & gondhalekar ( 1994 ) have combined photoionization and reverberation - mapping calculations in a way similar to the present study ( [ formalism ] ) . their work , and also the one by prez , robinson & de la funte ( 1992 ) , focused on the shape of the transfer function and the line responsivity in a few generic models . they did not attempt a detailed reconstruction of a specific data set , and the comparison with the observations includes _ only time lags _ , and not the detailed shape of the light curves . shields , ferland , & peterson ( 1995 ) used photoionization calculations to examine the @xmath01549/ line ratio in ngc 5548 and other objects . they concluded that an optically thin non - variable component contributes to the blr high ionization lines . bottorff et al . ( 1997 ) presented a detailed kinematic model which they combined with photoionization calculations . they applied this model to _ one emission line _ ( @xmath01549 ) in the spectrum of ngc 5548 . dumont , collin - suffrin , & nazarova , ( 1998 ) modeled the blr in ngc 5548 as a 3-zone region where the location of the various components were determined by the _ mean _ line - to - continuum lags . these authors assumed high column density ( @xmath11 @xmath6 ) and photon density ( the ionization parameter multiplied by the hydrogen density gives 10@xmath12 to 10@xmath13 @xmath8 ) . much of their conclusion regarding the density and column density is based on the relative strength of the balmer lines which we consider to be unreliable for reasons we describe below . like most others , they did not compare the observed shape of the emission line light curves with the prediction of the model . finally , in a recent work , goad and koratkar ( 1998 ) re - examined the 1989 ngc 5548 data set and deduced a strong radial dependence for the gas density ( @xmath14 , with density in the range of 10@xmath15 @xmath8 ) . here again , the main assumption is of a two - zone model and there is no attempt to use the observed light curves , only the measured time lags . evidently , there is a definite lack of modeling complete data sets , i.e. the information provided by half a dozen emission line light curves . the published works focus on either detailed photoionization calculations or detailed geometrical modeling ( i.e. using transfer functions ) and no study have combined the two in a satisfactory way . thus , no previous work has demonstrated that the assumed gas distribution , and the calculated physical conditions , are indeed compatible . global consistency checks are certainly required . our work relies heavily on the direct approach . we prefer to avoid , as much as possible , ill - defined transfer functions and unreliable mass distributions . instead we make a large number of initial assumptions ( `` guesses '' ) and check them , one by one , against the complete data set of ngc 5548 . this makes the numerical procedure more time - consuming but is more robust because of the direct relationship between the assumed geometry and the resulting light - curves . it also allow us to assess , in a more meaningful way , the statistical significance of our results . we follow the approach described by netzer ( 1990 ) for a blr consisting of numerous small clouds and a point - like ionizing source . the contribution of each cloud to the various emission lines is determined by the physical conditions within the cloud and its distance from the center . the main assumption is that the cloud pressure is controlled by an external medium ( magnetic field , hot gas , etc . ) whose pressure is a simple function of the radius , @xmath16 . for completeness , we summarize below the main ingredients of the model . consider a spherical system of clouds around a central radiation source and assume pure radial dependence of all properties . the hydrogen number density , @xmath17 , assumed to be constant within each cloud , is controlled by equilibrium with the external pressure . neglecting the small electron temperature variation we take this to be @xmath18 next , we define the cloud column density , @xmath19 , by considering spherical clouds , of radius @xmath20 . the mass of the individual clouds is conserved , as they move in or out , but it is not necessarily the same for all clouds , thus , @xmath21 . the cloud mean ( over the sphere ) column density is thus @xmath22 and the geometrical cross - section is @xmath23 ( note that the mass conservation assumption is only essential if we are to avoid an additional free parameter that specifies the run of @xmath19 with distance ) . finally , the number of clouds per unit volume is @xmath24 in the numerical integration we consider only one type of clouds ( i.e. same size at the same @xmath2 for all clouds ) . simple generalization of this scheme can involve a local population of clouds having some size distribution and the same @xmath25 for all sub - classes . this requires extra normalization ( see below ) which we decided to avoid at this stage . the clouds are illuminated by a central source whose ionizing luminosity , @xmath26 , varies in time . designating @xmath27 as the flux emitted by the cloud in a certain emission - line , @xmath28 , per unit projected surface area ( @xmath29 ) , we get the following relation for the emission of a single cloud : @xmath30 ( where we note that in the actual calculations , we use a thin slab geometry , hence the geometrical cross section is somewhat different from @xmath31 ) . assuming the system of clouds extends from @xmath32 to @xmath33 we obtain cumulative line fluxes : @xmath34 having determined the properties of the emission line clouds , and having assumed a spectral energy distribution ( sed ) for the ionizing source , we now calculate @xmath27 using a photoionization code . we then follow the above formalism to obtain @xmath35 . since all parameters have been fixed , there is no need to independently specify the ionization parameter , @xmath36 which , in our model , is given by @xmath37 where q is the number of hydrogen ionizing photons per second , and @xmath38 is the speed of light . agn emission is variable , causing the local emission - line flux to be a function of time . we take this into account by calculating @xmath39 for the entire range of continuum luminosity applicable to the source under discussion . the calculated _ in situ _ line fluxes are , therefore , the result of integrating eq . [ el ] , using , at each distance @xmath2 the ionizing flux obtained with the ionizing luminosity @xmath40 . observed fluxes further require the accounting for the light travel time between the cloud and the observer . a model is specified by the source luminosity and sed , the radial parameter @xmath41 , and the normalization of the various free parameters . these include @xmath42 and the density and column density at a fiducial distance which we take to be one light - day ( 1 ld ) , i.e. outside the range of interest ( in all the cases presented here @xmath43 ld ) . the comparison with observations further requires the normalization of the total line fluxes and hence the integrated number of clouds . we prefer to use the total covering factor @xmath44 which is obtained by integrating @xmath45 between @xmath32 and @xmath33 . the normalization of @xmath44 is achieved by comparing the observed and calculated flux of the emission lines . the numbers quoted below refer to the total covering factor as determined by taking the mean covering factor derived for the individual lines . we do not consider cloud obscuration and the calculations are limited to values of @xmath44 smaller than about 0.3 . considering all this , the total number of free parameters in the model is seven . computing a single model is not a trivial task . the reason is that , as explained , @xmath39 is a function of both distance and time . we have accomplished this by calculating large grids of photoionization models covering the required range of density , column density , and incident flux . the calculations were performed using ion97 , the 1997 version of the code ion ( see netzer 1996 , and references therein ) . the code includes a large number of atomic transitions and solves for the level of ionization , and the gas temperature , at each location , taking into account all optical depth effects . the line and continuum transfer is calculated by using the local escape probability formalism . the cloud geometry assumed is of a thin shell . this is somewhat different from the spherical - cloud assumption made here , but the uncertainty introduced is well below the uncertainty of the escape probability method . a serious problem in ion , and in similar codes like _ cloudy _ by g. ferland ( ferland et al . 1998 ) , is the treatment of the optically thick balmer lines . the local escape probability approximation used in such codes is inadequate for these lines ( netzer 1990 ) . unfortunately , there is no known cure for this problem within that formalism and we have decided , therefore , not to consider blamer lines in this work . the calculated intensity of other optically thick lines produced in the large , extended , low ionization zone of the cloud ( e.g. , @xmath02798 and many feii lines ) is also considered uncertain . a further limitation is the anisotropy of line emission which can be important when emission line light - curves are concerned ( see ferland et al . 1992 and references therein ) . we have decided not to include this effect since the local escape probability method can not accurately take it into account and since it depends on the cloud shape , which is unknown . we caution that while codes like ion and _ cloudy _ calculate this anisotropy as part of their standard output , the fundamental limitation of the local escape probability method makes such numbers highly uncertain . we have adopted a goodness of fit criteria to evaluate the model results . this is based on a simple @xmath46 score which is evaluated by comparing the observed and theoretical light cures of _ all _ chosen emission lines , giving each line an equal weight . thus , an acceptable model must reproduce the correct line variability as well as the observed line ratios . to the best of our knowledge , no previous agn model have posed such stringent conditions on the goodness of fit . the seyfert 1 galaxy ngc 5548 is a subject of an intensive study for more then a decade now . this includes optical spectroscopic monitoring for 8 years ( peterson et al . 1999 , and references therein ) , several satellite monitoring campaigns in different wavelengths each with a duration of several months ( e.g. , nandra et al . 1993 , korista et al . 1995 , tagliaferri et al . 1996 , marshal et al . 1997 ) , and many complementary studies of those data sets ( e.g. , clavel et al . 1992 , maoz et al . 1993 , wanders & peterson 1996 ) . this , together with the reasons mentioned in [ target ] , makes ngc 5548 the best object for testing our model . in this study we have concentrated on the 1989 @xmath47 campaign ( clavel et al . 1991 ) and used the @xmath01337 continuum and five emission lines ; ly@xmath481216 , @xmath01549 , @xmath01909 , @xmath01640 , and @xmath02798 . on top of the general considerations described in the previous section , there are several constraints pertaining to ngc 5548 : 1 . the 1989 @xmath49 spectra , and the following hst observations ( korista et al . 1995 ) , clearly show that the @xmath01909 line is blended with @xmath01895 . clavel et al . ( 1991 ) have measured the combined flux of both lines , without deblending ( this is the flux listed in their tables under the @xmath01909 header ) . this is also the case in the korista et al . ( 1995 ) paper . the observed ratio of @xmath01909 to @xmath01895 in the mean spectrum of ngc 5548 , is about 3:1 . the line ratio , which is density dependent , provides an important constraint on the density . we calculate the intensities of the silicon and carbon lines and restrict ourselves to those models producing a ratio consistent with the observations . the diagrams and tables below gives the summed intensity under the title @xmath01909 . 2 . galactic and intrinsic reddening can affect the observed line ratios and the assumed ionizing continuum . we address this in [ reddening ] . another important consideration is the shape of the ionizing continuum . the observed sed of ngc 5548 is reviewed by dumont et al . ( 1998 ) who discuss also various seds assumed in previous works . we have chosen a range of possible seds , all in agreement with observations of typical seyfert 1 galaxies . our optical - uv continuum is combined with the observed asca continuum of the source assuming @xmath50=1.06 ( see george et al . 1998 for a discussion of the x - ray spectrum of ngc 5548 ) . we have varied the assumed sed , as described in [ sedch ] , and tested the effect on the emission line spectrum . all chosen seds are rather similar to the ones discussed in dumont et al . our fiducial case is presented in fig . [ sed ] on the same scale as in dumont et al . 4 . using the observed @xmath01337 continuum flux , a chosen cosmology ( @xmath51=@xmath52 km s@xmath53mpc@xmath53 and @xmath54 ) , and the assumed sed , we have estimated the time averaged ionizing luminosity of ngc 5548 to be @xmath55 erg s@xmath53 . the continuum flux at @xmath01337 varied by a factor of @xmath56 , during the @xmath49 campaign . hence we choose our grid to cover the ionizing luminosity range of @xmath57 to @xmath58 erg s@xmath53 , with a step size corresponding to 0.1 in @xmath59 . the observed @xmath01337 light curve was transformed into ionizing - luminosity light curve covering the entire range between these values - see fig . [ con ] . considering the blr geometry , we have defined our grid of distances to cover the range of 1 to 100 lds , with a step size of 0.02 in @xmath60 . we have adopted solar abundances for our standard model ( see also [ gt ] below ) . model calculation proceed in two stages . first we produce two - dimensional grids of @xmath39 by calculating , for each radial distance , all line emissivities for all ionizing luminosities . each grid represents a particular choice of the density parameter @xmath41 , and the density and column density at the fiducial distance . we consider three density laws : @xmath61 , 1.5 , and 2 , a density at 1 ld of @xmath62 to @xmath63 ( varying by a factor of @xmath642 from one grid to the next ) and column densities at 1 ld of @xmath65 , @xmath66 , and @xmath67 . the choice of density , column density , @xmath41 and the time - variable luminosity , completely specify the ionization parameter at each location . in the second stage we calculate theoretical light curves by integrating eq . [ el ] . for this we need to define @xmath25 , @xmath32 and @xmath33 . we have considered models with @xmath68 , 1.5 and 2 . for each of these we have varied @xmath32 and @xmath33 in order to minimize the calculated @xmath46 . an important consideration is the error bar size for the first few points . obviously , we have no knowledge of the continuum behavior prior to the beginning of the campaign . we have therefore decided to extrapolated the continuum light curve , at a constant level , for all jd @xmath69 2447510 . this lack of knowledge introduces a large uncertainty in the first part of the line light - curves which is likely to be much larger than the pure observational error . to partly cure this effect , we have artificially increased all experimental error bars during the first 100 days . we used a multiplication correction factor which is 3.0 for day 1 and decreases linearly with time , to 1.0 , on day 100 ( jd=2447600 ) . to illustrate our method , we present the results of a model with @xmath70 ( i.e. @xmath71 and @xmath72 ) , @xmath73 , @xmath74 @xmath6 , and various densities . we note , again , that what is referred to here as a _ single model _ covers , in fact , a large range of density and column density . thus , choosing @xmath75 @xmath8 and @xmath74 @xmath6 we get at 10 lds , @xmath76 @xmath8 and @xmath77 @xmath6 , and at 100 lds , @xmath78 @xmath8 and @xmath79 @xmath6 . examining this case ( fig . [ la_m2 ] ) we note several obvious features . first , for @xmath80 @xmath8 , the emission line response is reversed , i.e. , increasing continuum luminosity results in decreasing line luminosity . for example , while the ionizing luminosity is increasing from jd@xmath642447600 to jd@xmath642447650 ( fig . [ con ] ) , the ly@xmath81 luminosity is decreasing ( doted line in fig . [ la_m2 ] ) , and the calculated light curve is a negative image of the observed one . further demonstration of this effect is shown in fig . [ eps_m2 ] : here we plot the line emission per unit solid angle ( @xmath82@xmath83@xmath84 ) versus @xmath2 for four ionizing luminosities . for the lowest density @xmath85 @xmath8 ( fig . [ eps_m2]a ) @xmath82@xmath83@xmath84 is larger for lower ionizing continuum for @xmath86 lds , i.e , the emitted ly@xmath81 is getting weaker when the ionizing luminosity is getting stronger . this effect is seen also in all other emission lines . the reason for this behavior is that for low density , the fast column density drop with distance causes the clouds to become optically thin a few lds away from the center . for lines other than ly@xmath81 , the increase in ionizing flux causes the gas to become too ionized to emit a significant amount in the line under question . the decrease in ly@xmath81 is of different origin . this line is mostly due to recombination and its intensity depends exponentially on the lyman continuum optical depth ( which , in this case , is of order unity ) and linearly on the ionizing flux . hence , the decrease in the optical depth affects the line more than the increase in ionizing flux . as evident from figs . [ la_m2 ] and [ eps_m2 ] , the reversed response disappears for @xmath87 @xmath8 , resulting in a much better fit to the observations . this shows that careful analysis of a single light - curve is already enough to put limits on the blr density . while the simulated ly@xmath81 light curve , for the @xmath70 grid , nicely fit the observed light curve with certain choice of density , @xmath32 and @xmath33 , this geometry seem to do a poorer job for the other emission lines this is illustrated in figs . [ allr74 ] and [ r74 ] that show that while calculated light curves of ly@xmath81 , and reasonably fit the observations , those for and badly disagree with the data . [ r74 ] shows the reason for this discrepancy which is the positive response for ly@xmath81 , and up to about a distance of 30 lds , and the negative response for and . furthermore , the apparent good fit shown by a solid line in fig . [ allr74 ] is misleading since it was obtained by allowing different normalizations of eq . [ el ] for the different lines . when requiring the same covering factor for all lines , there is no satisfactory solution ( see the dashed line in that diagram ) . the disagreement amounts to a factor 1.4 for and a factor 12 ( ! ) for . for that model ( dashed line in fig . [ allr74 ] ) the reduced @xmath88 score ( for the 4 emission lines excluding the line as explained in [ gt ] ) is 9.8 . the above example demonstrates the difficulty in fitting , simultaneously , line responses and line ratios . most previous work considered one line at a time and did not attempt full solution . hence , their seemingly satisfactory solution do not represent realistic spectra . our simulations show that no normalization or a choice of @xmath25 can cure this problem in the @xmath70 case . this is in contradiction to the result of goad & koratkar ( 1998 ) who favor an @xmath70 blr . ( see further discussion of their model in [ comp ] ) . the underlying reason for the failure of the @xmath14 family of models is that , in such cases the ionization parameter is constant throughout the blr . this is in contradiction to the observations that show different lags for lines representing different levels of ionization . we have made a careful investigation of a large number of models . we have considered cases with @xmath41 @xmath89 1 , 1.5 , and 2 , with @xmath90 in the range of @xmath91 to @xmath92 @xmath8 , and three values of @xmath93 , @xmath65 , @xmath66 and @xmath67 @xmath6 . below is a summary of the most important results . we first note that as the column density increases from @xmath94 to @xmath67 @xmath6 , so does the agreement between the model and the data . thus , models with @xmath94 @xmath6 give very poor fit ( the line rations are wrong and there is a reversed response in some lines ) while models with @xmath5 @xmath6 fit the data much better . current photoionization codes are limited to compton thin gas , i.e. they can not reliably calculate cases with column densities larger than about @xmath95 . thus we are unable to put upper limits on the column density of the gas . while large column density clouds are not sensitive to the exact column , we have noted some changes that suggest that even the @xmath96 @xmath6 models are not thick enough . for example , in some models with this column there is a noticeable improvement in @xmath46 when increasing the blr size ( i.e. @xmath33 ) which is not seen in the @xmath74 @xmath6 models . it indicates that the reversed response of optically thin cloud is present even in @xmath96 @xmath6 cases . as shown in [ example ] , @xmath70 models are practically ruled out . regarding cases with @xmath97 and @xmath61 , we have found that in both cases there are lower and upper limits on the gas density . for @xmath98 @xmath8 , the @xmath97 models result in reverse response ( like the @xmath70 models ) and the @xmath61 models produce wrong line ratios ( in particular in these low density models the @xmath01909 to @xmath01895 line ratio does not agree with the observations , see [ consider ] ) . for @xmath99 @xmath8 , the line ratios ( e.g. , / ) in both @xmath41 cases do not agree with the observations . hence , our study shows that the blr density in ngc 5548 , is in the range of @xmath7 @xmath8 . this is true for all models with @xmath5 @xmath6 . a general trend found for all values of @xmath41 is that as @xmath25 increases from 1 to 2 ( i.e. , more weight is given to clouds closer to the central source ) the amplitude of the modeled light curves is in better agreement with the observations and so are most line ratios ( an exception is the ly@xmath81/ line ratio - see discussion below ) . hence , models with higher @xmath25 are preferred . a common problem in all models is the too weak mgii@xmath02798 line . while we do not have a complete explanation for this , we suspect that it is due to one of two reasons : either the transfer of this line is inaccurately treated , similar to the case of the balmer lines ( discussed in [ formalism ] ) , or else it is caused by the thousands of highly broadened feii lines , in that part of the spectrum , that make the measured line intensity highly uncertain ( see for example the discussion in wills , netzer and wills , 1985 ; maoz et al . below we discuss the possibility of enhanced metallicity which also does not cure this problem . this is a definite failure of our model and is the reason for excluding this line when computing the @xmath46 and the normalization factor for eq . [ el ] . another common trend is the improved @xmath88 score for larger @xmath33 . in general , models with @xmath33 in the range 70 to 100 ld , agree better with the observations compared with models with @xmath100 lds . we also found that the models are not very sensitive to the exact value of @xmath32 and the best value is 3 lds . considering the above trends and limitations , and using the @xmath46 score for the four chosen emission lines , we find that models with @xmath61 best fit the observed spectra and models with @xmath97 give somewhat inferior fits . an example of one of the best cases is a model with @xmath61 , @xmath74 @xmath6 , @xmath101 @xmath8 , @xmath102 lds and @xmath103 lds . the light curves produced in this case are shown in fig . [ r53n_4 ] for various values of @xmath25 . the trend of improved fit with increasing @xmath25 is evident for all lines except , as explained , for . the reduced @xmath88 scores for the set of four emission lines are 4.5 for the @xmath68 model , 3.1 for the @xmath73 model , and 2.2 for the @xmath104 model . a note about the @xmath88 score is in order . our best models are characterized by @xmath105 which , statistically , is not very significant . this is somewhat arbitrary since enlarging the estimated errors by 50% ( which can be attributed to to systematic measurement errors or model uncertainty ) will result in @xmath106 . as noted earlier , almost all previous studies used solar composition material . in order to test the sensitivity of the results to the metal abundance , we run our best fitted model ( fig . [ r53n_4 ] ) with all metal abundances increased by a factor two . the results are generally similar to the solar composition model , with somewhat larger @xmath46 score ( slightly stronger and slightly weaker ly@xmath81 and ) . in view of this we have not investigated other compositions . reddening can affect line ratios as well as the apparent shape of the continuum . here we investigate the possibility that the intrinsic continuum differs , in luminosity and sed , from the observed one and the intrinsic line luminosities are larger than those observed . this would have an effect on the gas distribution in the source since the amount of continuum energy reprocessed by this gas , and the level of ionization , are different under these conditions . galactic reddening toward ngc 5548 is discussed by wamsteker et al . ( 1990 ) who estimated it to be @xmath107 . these authors noted the possibility of additional internal extinction which they have estimated to be of the same order of the galactic reddening . following this , we have assumed an internal extinction identical to the galactic extinction , i.e. , total extinction of @xmath108 . we have investigated the effect of reddening by modifying the sed in our best fit model ( section [ gt ] as shown in fig . [ r53n_4 ] ) assuming the above @xmath109 . new line luminosities were obtained by using extinction correction factors from seaton ( 1979 ) . dereddening the observed sed is more complicated since our adopted continuum is made of a collection of power - laws . the adopted sed is presented in fig . [ seds]e ( dot - dashed line ) . it was normalized to fit the reddened @xmath01337 light curve , in order to be consistent with the observations ( see explanation in [ consider ] ) . the results of this model are essentially identical to those of the unreddened case and are , therefore , not presented . this is not surprising in view of the further tests that we have conducted that are discussed below . several theoretical ideas ( e.g. binette et al . 1989 , clavel & santos - lleo 1990 ) as well as observational evidence ( e.g. , romano & peterson 1998 and references therein ) point to a change in the sed during continuum variation . this is expected since the shape of the optical - uv continuum is known to undergo large changes ( e.g. wamsteker et al . 1990 , clavel et al . 1991 ) . regarding the x - ray continuum variations , these are of a different time scale , and different amplitude , compared with the uv variations . it is therefore likely that the shape of the uv - x - ray ionizing continuum is a complicated function of the source luminosity . unfortunately , there is little observational study of such changes and we can only estimate the effect by testing various possible seds and their effect on changes in the emission line spectrum . we have carried out calculations with various possible schemes shown in fig . [ seds ] . the first case ( fig . [ seds]a ) is the standard sed ( the one used in all previous models , see fig . [ sed ] and solid line in fig . [ seds]e ) shown here at 11 pre - chosen fluxes . next we have tried different shape seds . in doing so we are not interested in the radio - fir part of the continuum hence we did not alter the shape and level of the spectrum below a frequency of 0.01 ryd . we also keep the flux above 20 ryd constant since the _ average _ x - ray and @xmath110-ray flux does not change on time scales similar to the optical uv time scale ( although it is changing very strongly on shorter time scales ) . keeping in mind the observed variation in the optical - uv continuum slope , we assume that the flux change at 3 ryd is a factor 2 larger than the flux change at 0.12 ryd ( the 0.12 and 3 ryd are two of the energies used to define our sed ) . this is causing the optical - uv continuum to become harder as the continuum luminosity increases and is consistent with observations ( clavel et al . we have normalized the seds such that the variable flux at 1337 remains consistent with the observations . the different seds for this scheme are presented in fig . we have used them to compute a new @xmath39 grid for @xmath61 model and to produce light curves for @xmath68,1.5 and 2 . the agreement of these results with the observations is not as good as our best fit , constant shape sed model ( [ gt ] and fig . [ r53n_4 ] ) the @xmath46 score is about a factor 2 worse . the reason for the inferior fit is related to the relative number of low and high energy ionizing photons . in particular , we attribute it to the relative number of ionizing photons above 4 ryd . in our original model ( fig . [ seds]a ) the relative number of those photons is higher than in the variable sed model ( fig . [ seds]c ) . to verify this , we have tested two other schemes . in the first scheme ( fig . [ seds]b ) , as the continuum flux increases the break point at 3 ryd moves , gradually , to _ lower _ frequencies ( to 1.5 ryd when the continuum flux is highest ) . thus , there are even fewer ionizing photons above 4 ryd compared with the scheme shown in fig . indeed , the resulting @xmath46 score is much worse ( by about a factor 3.5 ) compared with the constant sed scheme . in the second scheme ( fig . [ seds]d ) , the break point moves to _ higher _ frequencies as the continuum flux increases ( from 3 ryd when the flux is lowest to 5 ryd when the flux is highest ) . the relative number of ionizing photons with energy larger than 4 ryd increases with continuum luminosity . the resulting @xmath88 is indeed better - see fig . [ sedsv ] . in fact , the last scheme gives the very best fit of all our models , to the line light curve . however , the fits to the other lines ( l@xmath81 , , and ) are somewhat poorer and the overall @xmath46 slightly worse ( a factor 1.5 ) compared with the constant sed case . the above examples show that our model results are , indeed sensitive to the assumed sed and its changes in time . interestingly , the light curve fits point to a change which is opposite to the one suggested by binette et al . we suspect that by scanning a large grid of parameters with such variable sed we can improve our best result . this is beyond of the scope of the present work . we have computed the covering factor by integrating eq . [ cr ] as described in [ formalism ] . for models with @xmath70 we find a covering factors between 0.5 ( for @xmath104 ) to 0.95 ( for @xmath68 ) . this is much higher than predicted for typical agns but is in agreement with the results of goad & koratkar ( 1998 ) who find for their @xmath111 model a covering factor of @xmath640.7 . as explained , fitting the light curves of several lines simultaneously , we can rule out models with @xmath70 . the too high covering factor seems to confirm our conclusion . models with @xmath97 and @xmath61 result in smaller covering factors . for our best model with @xmath61 we find a covering factor of 0.25 for @xmath68 , 0.28 for @xmath73 , and 0.30 for @xmath104 . this is in good agreement with previous estimates and is consistent with the no - obscuration assumption of our model . several past models are detailed enough to allow meaningful comparison with our calculations . 1 . our models are inconsistent with the @xmath1 density law . this is in disagreement with the study of goad and koratkar ( 1998 ) despite of the good agreement with their range of acceptable densities and the lower limit on the column density . goad and koratkar ( 1998 ) did not explore a large parameter space , in particular , they only considered a limited range of density laws . we suspect that their conclusion regarding @xmath70 models is related to the use of mean time lags , rather than detailed fits to the observed light curves . 2 . dumont et al . ( 1998 ) used a three zone model and reached several conclusions based on time - averaged properties . they noted three problems arising from their modeling : an energy budget problem , a line ratio problem , and line variability problem . the first two are most probably related to the use of balmer line intensities as prime elements in evaluating their model results . as explained , we have not included the balmer lines in our fit since we consider present day photoionization models to be inadequate in this respect . regarding the line ratio , we find good agreement with observations for both high and low ionization lines . we suggest that even a three - zone model , like the one used by dumont et al . , is too simplified to account for the continues gas distribution in ngc 5548 . `` locally optimally - emitting clouds '' ( locs ) models ( baldwin et al . 1995 , korista et al . 1997 ) have been suggested to explain the broad line spectrum of agns . the models assume clouds with a range of density and column density at every point in the blr . this is very different from our assumption of a specific density and column density at each radius . locs must be put into a real test by checking whether they result in light curves that are in better agreement with the observations , compared with the simpler models tested here . + there are two obvious ways to compare the two models . first , in loc models the incident central source flux `` selects '' clouds at certain radii to be efficient emitters of certain lines while at other radii , other lines are more efficiently produced . thus , successful loc models of ngc 5548 should resemble ours , in their run of density and column density for the most efficient emitting component . second , future work along the lines suggested here should test multi - component models ( i.e. various densities at each location ) against the observed light - curve . this will make them similar , in some respect , to loc models . finally , we must comment that present day loc models are too general and do not contain full treatment of shielding and mixing of the various coexisting components . alexander and netzer ( 1994 ; 1997 ) suggested that blr clouds may be bloated stars ( bss ) with extended envelopes . in their work they fitted the emission - line intensities , profiles and variability to the mean observe agn spectrum . one of the conclusion is that the density at the edge of the bss ( the part emitting the lines ) falls off like r@xmath112 , where r is the distance from the central source to the bs , and the number density of the bss falls off like r@xmath113 . these two trends are in good agreement with our preferred values of @xmath41 and @xmath25 . while the bs model is consistent with the mean time - lags measured in intermediate luminosity agns , it remains to be seen whether it can fit , in detail , the time dependent spectrum of objects like ngc 5548 . we have presented a new scheme to model the blr in agns . it combines photoionization calculations with reverberation mapping and assumes gradual changes of density and column density as a function of distance from the central source . using this scheme we were able to constraint the physical conditions in the blr and put limits on the important physical parameters . the present work addresses only spherical blrs . future work will hopefully generalize it to other geometries . when applying the scheme to ngc 5548 , we were able to reconstruct four out of the five observed uv emission - line light curves . we found that a large population of optically thin - clouds ( due to low density and/or column density ) results in a reversed response to continuum variations . such models are therefore excluded . we also excluded models where @xmath14 and favor models where the density scales as @xmath4 or @xmath3 . we have placed a lower limit of @xmath5 @xmath6 on the column density of the clouds . we have used the results to place limits on the density of the blr . for ngc 5548 , this must be in the range of @xmath7 @xmath8 . we have studied a range of possible seds , including one which resulted from applying a reddening correction with @xmath109=0.1 . the one that produces the best agreement with the observations is an sed where the relative number of @xmath114 ryd photons increases with the agn luminosity . there are obvious limitations to this method and several ways to improve it . for example , line beaming ( anisotropy in the emission line radiation pattern ) must be considered . unfortunately , similar to the balmer line problem , the radiation pattern can not be accurately calculated in present - day escape - probability - based codes . multi - component models , with a range of density and column density at each radius ( not necessarily similar to loc models ) , must be tested too . finally , testing different blr geometries and a denser parameter space is highly desirable . we are grateful for constructive suggestions by the referee , b.m . we acknowledge financial support by the israel science foundation , and the jake adler chair of extragalactic astronomy . acknowledge financial support by the colton scholarships . romano , p. , & peterson , b. m. 1998 , in structure and kinematics of quasars broad line regions , eds . gaskell , c. m. , brandt , w. n. , dietrich , m. , dultzin - hacyan , d. , and eracleous , m. ( san francisco : asp )	we present a new scheme for modeling the broad line region in active galactic nuclei ( agns ) . it involves photoionization calculations of a large number of clouds , in several pre - determined geometries , and a comparison of the calculated line intensities with observed emission line light curves . fitting several observed light curves simultaneously provides strong constraints on model parameters such as the run of density and column density across the nucleus , the shape of the ionizing continuum , and the radial distribution of the emission line clouds . when applying the model to the seyfert 1 galaxy ngc 5548 , we were able to reconstruct the light curves of four ultraviolet emission - lines , in time and in absolute flux . this has not been achieved by any previous work . we argue that the balmer lines light curves , and possibly also the @xmath02798 light curve , can not be tested in this scheme because of the limitations of present - day photoionization codes . our fit procedure can be used to rule out models where the particle density scales as @xmath1 , where @xmath2 is the distance from the central source . the best models are those where the density scales as @xmath3 or @xmath4 . we can place a lower limit on the column density at a distance of 1 ld , of @xmath5 @xmath6 and limit the particle density to be in the range of @xmath7 @xmath8 . we have also tested the idea that the spectral energy distribution ( sed ) of the ionizing continuum is changing with continuum luminosity . none of the variable - shape sed tried resulted in real improvement over a constant sed case although models with harder continuum during phases of higher luminosity seem to fit better the observed spectrum . reddening and/or different composition seem to play a minor role , at least to the extent tested in this work .
let @xmath1 be the moduli of degree @xmath2 genus one stable maps @xcite to @xmath0 and let @xmath3 be the primary component that is the closure of the open subset of all stable morphisms @xmath4\in \mpd$ ] with smooth domains . in @xcite , vakil - zinger found a canonical desingularzation @xmath5 of @xmath6 by performing `` virtual '' blowing - ups on @xmath1 . they also showed that for @xmath7 and @xmath8 , the pull - back universal family over @xmath9 , the direct image sheaf @xmath10 is locally free . those desingularizations are useful for applying atiyah - bott localization formula to the hyperplane relation proved by li - zinger in @xcite . while the desingularization result in @xcite is algebro - geometric , its proof is analytic in nature . this paper provides an algebro - geometric approach to these desingularization results . it will be a part of an algebro - geometric approach of the relation between ordinary and reduced genus one gw - invariants of complete intersection in products of projective spaces @xcite . it will also serve as our first step to generalize the structure results on moduli spaces of genus one stable maps to higher genera . our proof consists of two stages . at first , we use the classical method of studying special linear series on curves to give an algebro - geometric proof of the local equations of @xmath1 , obtained by zinger in @xcite . after that , we `` modular '' blow up @xmath1 and prove that the resulting stack has smooth irreducible components . the blowup construction used in this paper follows that in @xcite . we now briefly outline our approach . since a stable map @xmath4 $ ] to a projective space @xmath0 is given by @xmath11 : c\lra\pn,\quad u_i\in h^0(u\sta \so(1)),\ ] ] its deformation is determined by the combined deformation of the curve @xmath12 and the sections @xmath13 . since moduli spaces of curves is smooth , the singularity of @xmath1 is caused by the non - locally freeness of the direct image sheaf @xmath14 of the universal family @xmath15 of @xmath1 . to study the non - locally freeness of the direct image sheaf , by assigning to each stable map @xmath4 $ ] the divisor @xmath16 , locally we can view @xmath1 as a stack over the artin stack @xmath17 of pairs @xmath18 of genus one nodal curves @xmath12 and effective divisors @xmath19 . over each chart @xmath20 , by picking an auxiliary section of the universal curve @xmath21 with @xmath22 the tautological divisor , we construct explicitly a complex @xmath23 $ ] whose sheaf cohomology gives the cohomology @xmath24 . we then apply the deformation theory of nodal curves to derive a simple explicit form of the homomorphism @xmath25 in @xmath26 . under a suitable trivialization , @xmath27 where each @xmath28 is a suitable product of the pull back of regular functions whose vanishing loci are irreducible components of nodal curves in @xmath17 . this description enables us to derive explicit local equations of @xmath1 ( see theorem [ thm : localequtree ] ) . they are analogous to the equations described by zinger @xcite . to construct the desingularization of the moduli space , we introduce the artin stack @xmath29 of pairs @xmath30 of genus one nodal curves @xmath12 with non - negative weights @xmath31 , meaning that @xmath32 for all irreducible @xmath33 . the stack @xmath29 is smooth and contains closed substacks @xmath34 whose general point is a pair @xmath30 such that @xmath12 consists of a smooth elliptic curve @xmath35 with @xmath36 rational curves attached to @xmath35 and the restriction of @xmath37 to @xmath35 is zero , while its restriction to the other components of @xmath12 in non - zero . we then blow up @xmath29 successively along @xmath38 , @xmath39 , @xmath40 , etc . , to obtain @xmath41 . the desingularization of @xmath1 is @xmath42 after a detailed study of the lifting of the local equations of @xmath1 mentioned earlier , we prove that the irreducible components of @xmath43 are smooth and intersect transversally . we also prove that for each irreducible component @xmath44 with @xmath45 the pull - back universal family on @xmath46 , the direct image sheaf @xmath47 is locally free . it is of rank @xmath48 over the desingularization of the primary component and of rank @xmath49 elsewhere . we comment that the results for the primary component were proved by vakil - zinger @xcite . weighted graph was also used to study stable maps to @xmath50 by pandharipande @xcite . this paper is organized as follows . in 2 , we outline our approach , stating the main desingularization theorems [ thm : m1 ] and [ thm : m2 ] , and the main local structure theorems [ thm : equ1 ] and [ thm : equ3 ] . in 3 , we introduce the notion of weighted rooted trees of weighted nodal curves . in the following section , we state and prove the main structural result of the direct image sheaf @xmath51 . finally , in 5 we prove the theorems stated in 2 . we d like to thank zinger for numerous suggestions on improving the presentation of the paper and for pointing out several oversights . the second author also thanks him for valuable discussion during their collaboration . the first author was partially supported by nsa ; the second author was partially supported by nsf dms-0601002 . throughout the paper , we fix an arbitrary algebraically closed base field @xmath52 . all schemes in this paper are assumed to be noetherian over @xmath52 . in this section , we state our main results . let @xmath12 be a curve . we set @xmath53 a weighted nodal curve is a pair @xmath30 consisting of a connected nodal curve @xmath12 and a weight assignment @xmath54 . for any scheme @xmath55 and flat family of nodal curves @xmath56 , a weight assignment of @xmath57 is a section @xmath37 of the sheaf @xmath58 , where @xmath59 is the subsheaf consisting of all sections @xmath37 whose restrictions @xmath60 to any closed point @xmath61 lies in @xmath62 . here @xmath60 is the image of @xmath37 under the pullback homomorphism @xmath63 . a flat family of weighted nodal curves over @xmath55 is a pair @xmath64 of a flat family of nodal curves over @xmath55 together with a weight assignment over @xmath57 . we say that two families of weighted nodal curves @xmath64 and @xmath65 are equivalent if there is an isomorphism @xmath66 such that @xmath67 . a weighted nodal curve is called stable if every smooth ghost ( weight 0 ) rational curve @xmath68 , @xmath69 contains at least three nodes of @xmath12 . fixing a genus @xmath70 , we form a groupoid @xmath71 that sends any scheme @xmath55 to the set of all equivalence classes of flat families of stable weighted nodal curves of genus @xmath72 over @xmath55 . mimicking the proof that the stack @xmath73 of nodal curves of genus @xmath72 is an artin stack , we obtain the groupoid @xmath74 is a smooth artin stack of dimension 3g-3 . the projection @xmath75 by forgetting the weight assignment is tale . note that @xmath74 is the disjoint union of infinitely many smooth components , each of which is indexed by the total weight of the weighted curves . because of the stablity requirement , each component is of finite type . @xmath76 to describe the desingularization of @xmath1 , the notion of core curve is pivotal . the core of a connected genus - one curve @xmath12 is the unique smallest ( by inclusion ) subcurve of arithmetic genus one . the core of a weighted curve @xmath77 is called ghost if the induced weight on the core is zero . the stack @xmath29 contains an open substack @xmath78 consisting of weighted curves with non - ghost cores . the complement @xmath79 admits a natural partition according to the number of rational trees attached to the ghost core curves : @xmath80 is the subset of pairs @xmath30 such that @xmath12 can be obtained from the ghost core @xmath81 by attaching @xmath36 ( connected ) trees of rational curves to the core @xmath35 at @xmath36 distinct smooth points of @xmath35 . then @xmath82 . we let @xmath34 be the closure of @xmath80 . we can successively blow up @xmath29 along the loci @xmath34 . the locus @xmath38 is a cartier divisor ; blowing up along @xmath38 does nothing . we start by blowing up @xmath29 along the locus @xmath39 , which is a smooth codimension 2 closed substack of @xmath29 ; we denote the resulting stack by @xmath83}$ ] , which is smooth . inductively , after obtaining @xmath84}$ ] , we blow it up along the proper transform @xmath85,k } \subset \fm\uwt_{1,[k-1]}$ ] of the closed substack @xmath86 . since @xmath85,k}$ ] is a smooth closed substack of @xmath84}$ ] of codimension @xmath36 , the new stack is smooth . we continue this process for all @xmath87 . since each connected component of @xmath88 is of finite type , the blowup process on this component will terminate after finitely many steps . therefore , the limit stack , which is the blowup of @xmath88 along the proper transforms of @xmath34 in @xmath89}$ ] for all @xmath90 , is a well - defined smooth artin stack ; we denote this stack by @xmath91 . to induce a blowup on @xmath1 , we form the fiber product @xmath92 here the morphism @xmath93 is defined as follows . the first chern class @xmath94 gives a weight assignment to the domain curves of the universal family @xmath95 of @xmath96 , making @xmath97 a family of weighted nodal elliptic curves . this family then defines a tautological morphism @xmath98 that is a lift of the tautological morphism @xmath99 . note that since @xmath94 has degree @xmath2 , defining @xmath100 requires blowing up @xmath88 along the proper transforms of @xmath34 from @xmath101 to @xmath102 . that is , @xmath103}.\ ] ] we can now succinctly reformulate the end result of the virtual blowup construction of @xcite . @xmath100 is a dm - stack with normal crossing singularities . for a refined version of this theorem with explicit local equations , see theorem [ thm : inductioncase = d ] . desingularizations of the sheaves @xmath104 over @xmath100 are essential ingredients in computing genus one gw - invariants of complete intersections @xcite . the blowup @xmath100 contains a primary irreducible component whose generic points are stable maps with non - ghost elliptic core . we denote this component by @xmath105 . the other irreducible components are indexed by the set of all partitions of @xmath2 . for @xmath106 either @xmath107 or a partition of @xmath2 , let @xmath108 be the pull back of the universal map over @xmath1 . the following theorem is due to vakil - zinger @xcite . for every @xmath109 , the direct image sheaf @xmath110 is a locally free sheaf over @xmath111 with @xmath106 either @xmath107 or a partition of @xmath2 . it is of rank @xmath112 when @xmath113 and of rank @xmath114 otherwise . we will treat the two theorems in the reverse order from what was done in @xcite . specifically , we will first prove a structure result for the direct image sheaf @xmath115 for all positive integers @xmath36 . we then derive local defining equations for @xmath1 and @xmath100 , and obtain theorems [ thm : m1 ] and [ thm : m2 ] as corollaries . for later use , we form the artin stack of stable pairs @xmath18 of ( connected ) nodal elliptic curves @xmath12 with effective divisors @xmath116 supported on the smooth loci of @xmath12 . here @xmath18 is stable if any smooth rational curve in @xmath12 disjoint from @xmath117 contains at least three nodes of @xmath12 . we denote this stack by @xmath17 . it maps to @xmath88 by sending @xmath18 to @xmath118 . the morphism @xmath119 is smooth and has connected fibers . in particular , the connected components of @xmath17 are indexed by the degree of the effective divisors . for any closed point @xmath4 \in \mpd$ ] , we let @xmath120 : c \lra \pn , \quad u_i \in \gamma ( c , u^ * \so_{\pn}(1)),\ ] ] be the associated stable morphism . by a change of homogeneous coordinates on @xmath0 , we can assume that @xmath121 is a smooth simple divisor @xmath117 of degree @xmath2 . once @xmath122 is fixed , we can choose @xmath123 to be the constant section @xmath124 . under this convention , the remaining sections @xmath125 are uniquely determined by the morphism @xmath126 . consequently , deforming @xmath127 is equivalent to deforming the pair @xmath128 and the sections @xmath129 . we next choose a small open neighborhood @xmath4\in u$ ] in @xmath1 . let @xmath130 be the the universal family over @xmath131 with @xmath132 the fiber over @xmath4 $ ] . let @xmath133 $ ] ( where @xmath134 $ ] are the homogeneous coordinates of @xmath0 ) . by shrinking @xmath131 if necessary , we can assume that @xmath135 is away from the singular points of the fibers @xmath136 . this way , we obtain a morphism @xmath137 \in u \longmapsto ( c ' , c ' \cap\cs).\ ] ] we now construct the deformation space of the data @xmath138 . we let @xmath139 be a smooth chart of @xmath140 that contains the image of @xmath141 ; let @xmath142 be the tautological family over @xmath143 with @xmath144 for some point @xmath145 , and let @xmath146 be the projection . we set @xmath147 . set theoretically , our deformation space is the union @xmath148 . the deformation is singular at points where the core curve of @xmath149 is ghost due to @xmath150 . the algebraic construction of the deformation space is via a standard trick . by shrinking @xmath143 if necessary , we can find a section @xmath151 of @xmath152 , away from @xmath153 , such that it passes through smooth parts of the core curves of all fibers of @xmath152 . because @xmath154 is effective and has positive degree on the core curve of every fiber of @xmath152 , we have r^1 _ * ( ) = 0 . we let @xmath155 be the total space of the vector bundle @xmath156 and let @xmath157 be the projection . then the tautological restriction homomorphism @xmath158 lifts to a section f ( _ , p^ * _ * ( ( ) ^n\| _ ) ) . let @xmath159 . then there is a canonical open immersion @xmath160(@xmath161 ) @xmath162 . these local equations are made explicit in theorem [ thm : localequtree ] . we let @xmath163 be the blowup described in [ inductive - blowups ] . we form the fiber product @xmath164 let @xmath165 be the projection . then @xmath166 is the total space of the pull back bundle @xmath167 . the immersion @xmath168 of theorem [ thm : equ1 ] induces an immersion @xmath169 . let @xmath170 be the projection and let @xmath171 be the pull back section in @xmath172 . after shrinking @xmath143 if necessary and fixing a suitable trivialization @xmath173 , we can find @xmath174 regular functions @xmath175 over @xmath176 such that @xmath177 further , each @xmath178 and @xmath179 has smooth vanishing locus and the vanishing locus of their product @xmath180 has normal crossing singularities . note that some @xmath181 may be invertible . for more explicit local equations for @xmath100 ; see theorem [ thm : inductioncase = d ] . in this section , we discuss the combibatorics of the dual graphs of nodal curves and introduce terminally weighted trees for weighted elliptic nodal curves . this combibatorics is not strictly necessary for our presentation , but they do make our exposition more intuitive and formulas more elegant . let @xmath182 be a connected rooted tree with @xmath183 its root . the root @xmath183 defines a unique partial ordering on the set of all vertices of @xmath182 according to the descendant relation so that the root is the unique _ minimal _ element ; all non - root vertices are descendants of the root @xmath183 . we call a vertex terminal if it has no descendants ; in the combinatorial world , this is also called a leaf . equivalently , a non - root terminal vertex is a non - root vertex that has exactly one edge connecting to it . following this convention , the root is terminal only when @xmath182 consists of a single vertex . in the following , given a rooted tree @xmath182 we denote by @xmath184 the set of its vertices , by @xmath185 its non - root vertices , and by @xmath186 its terminal vertices . note that for any vertex @xmath187 in @xmath182 , there is a unique ( directed ) path between @xmath183 and @xmath187 ; this is the maximal chain of vertices @xmath188 next we consider weighted rooted trees . a weighted rooted tree is a pair @xmath189 consisting of a rooted tree @xmath182 together with a function on its vertices @xmath190 , called the weight function . a vertex is positive if its weight is positive ; a ghost vertex is a vertex with zero weight . the total weight of the tree is the sum of all its weights . in this paper , we will consider only _ terminally weighted _ rooted trees . a weighted rooted tree @xmath189 is terminal if all terminal vertices are positive and all positive vertices are terminal ; it is called stable ( resp . semistable ) if every ghost non - root vertex has at least three ( resp . two ) edges attached to it . when the weight @xmath37 is understood , we will use @xmath182 to denote the weighted rooted tree @xmath191 as well as its underlying rooted tree @xmath182 with weights removed . 4.5 cm ( 3 , 15 ) ( 45,2 ) .5 cm figure 1 . two trees for positive @xmath2 , we let @xmath192 ( resp . @xmath193 ) be the set of all stable ( resp . semistable ) terminally weighted rooted trees of total weight @xmath2 . the set @xmath192 is finite , while @xmath194 is infinite . the set @xmath194 admits the following geometrical operations that will be useful for our discussion . the first operation is _ pruning _ a tree . given @xmath195 , to prune @xmath182 from a vertex @xmath187 , we simply remove all descendants ( i.e. those @xmath196 with @xmath197 ) and the edges connecting the removed vertices . after pruning @xmath182 from @xmath187 , the new graph has @xmath187 as its terminal vertex . if @xmath182 is a weighted tree , we define the weight of @xmath187 in the pruned tree to be the sum of the original weight of @xmath187 and the weight of the removed vertices ; we keep the weights of the other vertices unchanged . note that the resulting pruned tree is terminally weighted as well . the second operation is _ collapsing _ a vertex . collapsing a vertex @xmath187 in @xmath182 is a two - step process : first merge @xmath187 with its unique ascendent , removing the edge between them and assigning the sum of their weights to the merged vertex , and then prune the resulting tree along all positively weighted non - terminal vertices , repeating the process as long as possible . the third operation is _ specialization _ : it is the inverse operation of a collapsing . the fourth operation is _ advancing _ a vertex . let @xmath187 be a vertex in @xmath182 and let @xmath198 be its direct ascendant . to advance @xmath187 , replace every edge connecting @xmath199 to a direct descendant @xmath200 other than @xmath187 by an edge connected @xmath200 to @xmath187 and then prune the resulting tree along all positively weighted non - terminal vertices , repeating the process as long as possible . ( 3 , 15 ) ( 20,2 ) .5 cm figure 2 . operations on weighted trees instead of drawing a picture , a weighted rooted tree can also be described compactly as follows . here is an example . let @xmath201.\ ] ] this is a weighted rooted tree whose root is @xmath183 ; the other vertices are labelled by @xmath202 . the vertices inside a square bracket are the descendants of the vertex immediately proceeding the bracket ; the weights of the terminal vertices @xmath203 , @xmath204 , and @xmath2 are indicated in the following parenthesis . collapsing @xmath205 , we obtain @xmath206 $ ] . advancing @xmath205 , we get @xmath207 $ ] . advancing @xmath203 , we obtain @xmath208 $ ] . when the weight function is irrelevant to the discussion , we will drop any reference to weights . for example , the above tree @xmath182 would then be written as @xmath209 $ ] in this compact representation of a tree , pruning a tree from a vertex @xmath187 means removing the bracket immediately after @xmath187 and assigning it the total weight of the vertices inside the bracket . collapsing a ghost vertex @xmath187 is removing the vertex @xmath187 as well as the closest brackets @xmath210[,]@xmath211 associated to it . advancing a ghost vertex @xmath187 is moving all the other vertices located inside of the same square bracket as @xmath187 into the square bracket following @xmath187 . observe that advancing can make a stable tree semistable . for example , consider the stable tree @xmath212 , c]]$ ] . if we advance @xmath205 , we obtain @xmath213]$ ] , which is semistable but not stable , since the vertex @xmath203 has only two edges attached to it . @xmath76 to keep track of the changes of the strata of @xmath1 after blowups , we need the notion of monoidal transformations of weighted trees . we begin with the following . let @xmath182 be a semi - stable terminally weighted tree with root @xmath183 and at least one non - root vertex . the trunk of @xmath182 is the maximal chain @xmath214 of vertices in @xmath182 such that each vertex @xmath200 with @xmath215 has exactly one immediate descendant . in this case , we abbreviate the trunk by @xmath216 ; we call @xmath217 the branch vertex of @xmath182 if it is not a _ terminal _ vertex . otherwise we call @xmath182 a path tree . when @xmath218 , we say the tree has no trunk . figure 1 shows two trees : the first one has no trunk ; the second one has a trunk . let @xmath182 be a tree with trunk @xmath219 . then @xmath182 can be obtained by attaching @xmath220 rooted trees , called branches , @xmath221 , to the trunk so that the roots of @xmath222 are direct descendants of @xmath217 . according to our convention , @xmath182 can be expressed as @xmath223:=o[v_1[\cdots[v_r[\gamma_1',\cdots,\gamma_\ell']]]].\ ] ] the tree @xmath182 has no branches if and only if it is a path - tree . let @xmath224 denote the number of branches of @xmath182 . we call @xmath182 simple if all of the branches are stable . a monoidal transform of a simple terminally weighted tree @xmath225 is a tree obtained by advancing one of the immediate descendants of the branch vertex of @xmath225 , if @xmath225 has a branch vertex . we denote the set of monoidal transforms of @xmath182 by @xmath226 . note that every tree in @xmath192 is simple , and if @xmath225 is a simple tree , so is every monoidal transform of @xmath225 . let @xmath182 be a simple terminally weighted tree and @xmath227 . then either @xmath228 , which is when @xmath229 is a path - tree , or @xmath230 . the same conclusion holds when @xmath229 is a collapsing of @xmath182 at a direct descendant of the branch vertex . if @xmath231 is the result of advancing a direct descendant @xmath187 of the branch vertex @xmath199 of @xmath225 and @xmath187 is not terminal , then the direct descendants of @xmath187 in @xmath231 are the direct descendants of @xmath187 in @xmath225 and the direct descendants of @xmath199 in @xmath225 other than @xmath187 . furthermore , @xmath187 is the branch vertex of @xmath231 in this case ; thus , @xmath232 . on the other hand , if @xmath187 is terminal in @xmath225 , then @xmath231 is the path from @xmath183 to @xmath187 in @xmath225 . the proof of the second statement is similar . to index the strata of the various blowups of @xmath1 , we introduce the following . we set @xmath233}=\lambda_d$ ] and define @xmath234}$ ] inductively for @xmath235 : @xmath236}=\{\gamma \in \lambda_{d,[k-1 ] } \| { \rm br } ( \gamma ) \ge k+1\ } \cup \{\gamma \in { \rm mon } ( \gamma ' ) \| \gamma ' \in \lambda_{d,[k-1 ] } , { \rm br } ( \gamma ' ) = k \}.\ ] ] for any @xmath237}$ ] with @xmath238 , either @xmath239 or @xmath240 . in particular , @xmath241}$ ] consists of path trees only . first , the fact the total weight of a terminally weighted tree @xmath182 is @xmath2 implies that @xmath242 . the assertion of the lemma holds for @xmath243 . on the other hand , lemma [ mon1 ] implies that if the assertion holds for @xmath36 , then it also holds for @xmath244 . let @xmath12 be a connected nodal genus one curve . we associate to @xmath12 the dual @xmath245 , with vertices corresponding to the irreducible components of @xmath12 and the edges to the nodes of @xmath12 . since the arithmetic genus of @xmath12 is @xmath246 , either @xmath247 has a unique vertex corresponding to the genus one irreducible component of @xmath12 or @xmath247 contains a unique loop . in the first case , we designate that vertex the _ root _ of @xmath248 ; in the latter case we shall contract the whole loop to a single vertex and designate it the _ root _ of the resulting tree . we will call the resulting rooted tree the _ reduced dual tree _ of @xmath12 and denoted it by @xmath249 . next we consider a weighted nodal genus one curve @xmath250 . the weight @xmath37 induces weights on the vertices of the dual graph @xmath247 of @xmath12 . if @xmath247 contains a loop , then upon contracting the loop , we assign the resulting root the total weight of the loop . this way , we obtain a natural weighted tree @xmath251 associated to the weighted curve @xmath250 . in general , the @xmath249 with this weight function may not be terminally weighted . if not , we can prune @xmath249 along all positively weighted non - terminal vertices to obtain a terminally weighted rooted tree . we call the result the terminally weighted rooted tree of @xmath250 and denoted it by @xmath252 . in the case that @xmath250 is understood , we shall write it simply as @xmath182 . following this construction , each terminal vertex @xmath187 of @xmath182 is a positively weighted vertex of @xmath247 . we let @xmath253 be the union of all irreducible components of @xmath12 whose associated vertices @xmath254 ( in @xmath247 ) satisfy @xmath255 . then , @xmath253 is a tree of rational curves . further , if @xmath182 has zero weight root and @xmath256 are its terminal vertices , then @xmath257 is the maximal weight zero connected subcurve of @xmath12 that contains the core elliptic curve . any stable map @xmath258 \in \mpd$ ] naturally gives rise to a weighted nodal genus one curve @xmath250 , where the weight of an irreducible component of @xmath12 is the degree of the map @xmath196 on that component . we will then call the associated terminally weighted rooted tree of @xmath250 the terminally weighted rooted tree of the stable map @xmath259 $ ] . we denote it by @xmath260 } , w)$ ] . it is stable . for any @xmath261 , we define @xmath262 be the subset of all @xmath263}\in\mpd$ ] whose associated terminally weighted rooted trees is @xmath182 . each @xmath262 is a smooth , locally closed substack of @xmath1 ; together they form a stratification of @xmath1 . suppose @xmath264 has @xmath265 terminal vertices , indexed by @xmath266 and of weights @xmath267 with @xmath268 . without the weight , @xmath182 is the _ reduced dual graph _ of some genus 1 nodal curve . we denote by @xmath269 the stratum in @xmath270 consisting of stable genus 1 curves whose _ reduced dual graph _ is @xmath182 with terminal vertices replaced by the corresponding marked points . then @xmath262 is ( up to equivalence by automorphisms ) @xmath271 , [ { u_i},c_i , q_i]_1^\ell)\in m_{\gamma}\times\prod_{i=1}^\ell\overline m_{0,1}(\pn , d_i ) \mid { u_1}(p_1)=\cdots={u_\ell}(p_\ell)\bigr\}.\ ] ] since @xmath0 has ample tangent bundle , @xmath272 smooth and the evaluation morphisms @xmath273 are submersions . hence @xmath262 is smooth . in this section , we state and prove structure results of the direct image sheaf @xmath51 . let @xmath12 be a proper nodal curve with arithmetic genus @xmath274 . we call a node @xmath275 of @xmath12 a separating node if @xmath276 is disconnected . similarly , we call an irreducible component @xmath33 a separating component if @xmath277 is disconnected . along the same line , we introduce an inseparable curve is a connected curve with no separating node ; an inseparable component of @xmath12 is an inseparable subcurve of @xmath12 that is not a proper subcurve of another inseparable subcurve of @xmath12 . we say that a ( separating ) node @xmath275 separates @xmath278 and @xmath279 if @xmath278 and @xmath280 lie in different connected components of @xmath276 ; we say that the node @xmath275 lies between @xmath278 and @xmath280 in this case . we let @xmath281}$ ] be the collection of all nodes that lie between @xmath278 and @xmath280 . this notion extends beyond nodes : for any smooth point @xmath282 , we denote by @xmath283 the inseparable component of @xmath12 that contains @xmath284 ; we say @xmath284 ( or @xmath283 ) separates or lies between @xmath278 and @xmath279 if @xmath278 and @xmath280 lie in different connected components of @xmath285 . for a nodal elliptic curve @xmath12 and two distinct smooth points @xmath203 and @xmath205 on the core of @xmath12 , we have h^0 ( c , _ c(a - b))= 0 h^1 ( c , _ c(a - b))= 0 ; for any point @xmath286 of @xmath12 distinct from @xmath287 , we have h^0 ( c , _ c(+ a - b))= 1 h^1 ( c , _ c(+ a - b))= 0 . let @xmath288 be a scheme , @xmath117 a cartier divisor of @xmath288 , and @xmath289 a closed subscheme of @xmath288 . we will write @xmath290 for the restriction @xmath291 . our aim is to describe the structure of @xmath292 ; we let @xmath293 in the rest of this section . when we investigate the structure of @xmath1 via @xmath294 , we will specialize to @xmath295 . consider the substack @xmath296 of the artin stack @xmath17 of pairs @xmath18 of nodal elliptic curves with effective degree @xmath2 divisors @xmath19 . let @xmath297 be a point with the divisor @xmath117 simple and supported on the smooth locus of @xmath12 . we let @xmath298 be a smooth chart containing @xmath128 . again , let @xmath142 be the tautological family over @xmath143 with @xmath299 for some point @xmath145 and @xmath300 be the projection ; set @xmath147 . as in [ a ] , we choose a general section @xmath301 of @xmath152 and this time around also an additional general section @xmath302 of @xmath152 such that they are disjoint from @xmath153 and pass through the core of every fiber of @xmath152 . this is possible after shrinking @xmath143 if necessary . by the mittag - leffler exact sequence , the sheaf @xmath303 over @xmath143 is the kernel sheaf of : _ * ( ) _ * _ ( ) . the complex of locally free sheaves of @xmath304-modules @xmath305=[\rho _ * \sl(\ca ) \mapright{\psi } \rho_\so_{\ca}(\ca)]\ ] ] has sheaf cohomology @xmath306 $ ] . further , for any scheme @xmath307 with the induced family @xmath308 since @xmath309 , by cohomology and base change , @xmath310).\ ] ] to get hold of the sheaf @xmath51 , we shall study the local structure of the homomorphism . as in [ simpldivisor]-[deform ] , we only need to consider the case that @xmath117 is a smooth simple divisor on @xmath12 ; we will assume that this holds . we may also assume that @xmath143 is affine . after shrinking @xmath143 and an tale base change if necessary , we may assume that @xmath311 , where @xmath312 are disjoint sections of the family @xmath152 . to the sheaf @xmath313 , the standard inclusion @xmath314 provides us a section @xmath315 , called the obvious section . to capture other sections , we consider the inclusion of sheaves @xmath316 and the induced inclusions @xmath317 both are locally free since @xmath318 and @xmath319 by . by riemann - roch , @xmath320 is invertible and the rank of @xmath321 is @xmath322 . we then let @xmath323 and @xmath324 be the evaluation homomorphisms . obviously , @xmath325 . since @xmath143 is assumed affine , the sheaf @xmath326 . we have 1 . @xmath327 ; 2 . @xmath328 ; 3 . @xmath329 is an isomorphism , and @xmath330 consequently , the sheaf @xmath331 is a direct sum of @xmath304 with the kernel of the homomorphism @xmath332 taking the direct image of the exact sequence @xmath333 we obtain the exact sequence @xmath334 clearly , @xmath335 . also , @xmath336 is surjective because the obvious section @xmath337 maps surjectively onto @xmath338 . finally , because @xmath143 is affine , @xmath339 . therefore , the sequence @xmath340 is exact and splits . this proves ( 1 ) . the second is obvious . we now obtain the third . since both sheaves @xmath341 and @xmath342 are locally free , we only need to show that for any closed @xmath343 , @xmath344 is an isomorphism . because @xmath345 of @xmath346 and @xmath347 are zero , by base change , this is equivalent to that the tautological homomorphism @xmath348 is an isomorphism . because both sides are of equal dimensions , it suffices to show that it is injective . for this , we look at the composite of @xmath349 with @xmath350 obviously , @xmath351 for @xmath352 and is an isomorphism for @xmath353 . this shows that @xmath354 is injective . this proves the last claim of the lemma . the homomorphism @xmath355 then is completely determined by the homomorphism _ i _ i : _ i _ _ i _ ( + - ) _ * _ ( ) . the homomorphism _ i : _ * _ i _ * _ ( ) will be our focus in the next subsection . our strategy is to find an explicit expression for @xmath356 so that its vanishing locus has precise geometric meaning . for this , we need some regular functions associated to the smoothing of nodes . by the deformation theory of nodal curves , for each separating node @xmath357 there is a regular function @xmath358 so that @xmath359 is the locus where the node @xmath275 is not smoothed ; the divisor @xmath360 is an irreducible smooth cartier divisor . for any @xmath361 , we introduce @xmath362 we then collect all the nodes @xmath275 that lie between @xmath363 and @xmath203 ( cf . [ nodes ] ) and form the product of their associated functions @xmath364 : _ [ , a]=_qn_[_i , a]_q . in case @xmath365}=\emptyset$ ] , we set @xmath366}=1 $ ] . there are trivializations @xmath367 and @xmath368 such that the homomorphism @xmath356 is given by _ i= _ [ _ i , a ] : _ * _ i _ * _ ( ) . since @xmath143 is affine , we fix a trivialization @xmath369 , and keep the trivialization @xmath370 mentioned before . this way , @xmath371 then the proposition is equivalent to that as divisors , _ i ( 0)=_q n_[_i , a]_q(0)= _ q n_[_i , a ] _ q. we next let @xmath372 be either a point or a smooth affine curve , we let @xmath373 be @xmath374 over @xmath375 , and let @xmath376 be the corresponding pull back divisors . since @xmath377 , by cohomology and base change , the natural homomorphism @xmath378 is an isomorphism . finally , let @xmath379 be the composite @xmath380 then , if @xmath375 is a smooth curve not contained in @xmath381 , as divisors , @xmath382 we now prove the claim . first let @xmath375 be a smooth point away from @xmath383 } } \sigma_q$ ] . then @xmath384 , @xmath385 , and @xmath386 lie in the same inseparable component of @xmath387 . therefore @xmath388 and @xmath389 because @xmath390 if @xmath375 is a smooth point in @xmath360 for some @xmath391}$ ] , then @xmath385 and @xmath386 lie in different inseparable components of @xmath387 . this time , @xmath392 . therefore , for the same reason as above , @xmath393 . this proves that @xmath356 vanishes exactly along @xmath383 } } \sigma_q$ ] . it remains to show that @xmath356 vanishes at first order only along @xmath360 , @xmath394}$ ] . to prove this we only need to study @xmath356 near a general point @xmath395 . we let @xmath396 be an affine curve passing through @xmath397 and transversal to @xmath360 at @xmath398 . after shrinking @xmath399 if necessary , the family @xmath400 is the blowup of a family of smooth elliptic curves @xmath401 at a point @xmath402 . we let @xmath403 be the projection and let @xmath404 be the rational component , which is also the exceptional divisor of @xmath387 . let @xmath405 , @xmath406 , and @xmath407 be the image divisors in @xmath408 . then @xmath409 further , since @xmath410 , the cokernel of the inclusion _ r__r(_r , i-_r+_r)_r__r(_r , i+_0-_r+_r ) is @xmath411 , which is isomorphic to @xmath412 . therefore , since @xmath413 , @xmath379 factors through a homomorphism @xmath414 as shown in the commutative diagram @xmath415 since the lower horizontal arrow is an isomorphism , @xmath414 is an isomorphism . combined with that the cokernel of is @xmath412 , this proves that @xmath379 has precisely order one vanishing at @xmath399 . for the convenience of reference , we record an immediate consequence of proposition [ keyprop ] . there are trivializations @xmath367 and @xmath416 such that the homomorphism @xmath25 is given by _ i=1^m _ i : _ i=1^m _ * _ i _ * _ ( ) , _ [ _ i , a ] . the homomorphism can be further simplified . recall that @xmath143 is a neighborhood of @xmath417 . the pair @xmath128 induces a weighted curve @xmath250 with @xmath418 . we let @xmath182 be the terminally weighted tree of @xmath250 with terminal vertices @xmath419 according to our convention , each non - root vertex @xmath420 corresponds to a connected subcurve @xmath421 ( c.f . [ reducedgraph ] ) ; on the subcurve @xmath253 there is a unique separating node @xmath275 of @xmath12 that separates @xmath253 and the remainder part @xmath422 . we call this node @xmath275 the node associated to @xmath187 . with such node identified , for each vertex @xmath187 we define @xmath423 where @xmath275 the associated node of @xmath187 . for any terminal vertex @xmath424 , we let @xmath425 } = \prod_{i \succeq v\succ o}\zeta_v.\ ] ] the direct image sheaf @xmath426 is a direct sum of @xmath427 with the kernel sheaf of the homomorphism _ i=1^_i:_^ _ , _ i=_[i , o ] . we express @xmath117 as @xmath428 , and continue to denote by @xmath429 the terminal vertices of @xmath182 . by our construction of the terminally weighted tree @xmath182 of the weighted curve @xmath250 of the pair @xmath18 ( cf . [ reducedgraph ] ) , each @xmath200 is associates to a connected tree @xmath430 of rational curves ; there is a unique irreducible component @xmath431 of @xmath430 closest to the core of @xmath12 . again , by the construction of @xmath182 , at least one of @xmath432 lies on @xmath431 ; we pick one and index it by @xmath363 . further , for every @xmath433 , there is @xmath434 , such that @xmath363 is between the point @xmath203 ( of the core curve of @xmath12 ) and @xmath435 ( @xmath363 and @xmath435 can be on the same irreducible component ) . this shows that @xmath366 } \| \zeta_{[\delta_j , a]}$ ] ; thus , every @xmath436}$ ] is divisible by one of @xmath437},\cdots,\zeta_{[\delta_\ell , a]}$ ] . thus in the expression , we can choose a new basis for @xmath438 so that with respect to the new trivialization @xmath439_j=1^m _ j_^m , @xmath439 the homomorphism @xmath440 has the form @xmath441 together with lemma [ lemm : usefulfacts ] and corollary [ maincoro ] , this proves the theorem . in this section , we prove the theorems stated in 2 . in the meantime , we describe local defining equations for @xmath1 in terms of weighted trees . recall that in theorem [ thm : equ1 ] , for any @xmath4\in\mpd$ ] we first pick a small open subset @xmath4\in u\sub \mpd$ ] and a homogeneous coordinate @xmath134 $ ] of @xmath0 so that the pull back divisor @xmath442 $ ] is a family of simple divisors on the domain family @xmath97 of the universal family @xmath443 of @xmath131 ; the family @xmath97 coupled with the divisor @xmath444 $ ] defines a tautological morphism @xmath445 . we next pick a smooth chart @xmath446 so that its image contains the image of @xmath141 . let @xmath447 . we now construct the promised open immersion : _ . we let @xmath455 be the composition of the projection @xmath456 with @xmath457 ; let @xmath458 , @xmath459 , be the pull back sections in @xmath460 . according to our convention , @xmath461 is the section @xmath246 induced by the inclusion @xmath462 . this way , all other @xmath463 , @xmath464 , are canonically defined . as mentioned in [ a ] , we have a section @xmath301 of @xmath152 . its pull back section in @xmath465 is @xmath466 . because @xmath467 , by the cohomology and base change theorem , we have canonical isomorphism ^ * _ * ( ) _ * ( ) . we let @xmath468 be the inclusion . then @xmath469 is a section of @xmath470 . on the other hand , since @xmath471 is the vector bundle @xmath472 , defining a @xmath143-morphism @xmath473 is equivalent to giving a section of the pull back sheaf @xmath474 . we define the morphism @xmath106 in to be the one induced by the homomorphism ( s)=((s_1),,(s_n)):_()^n . to complete the proof , we need to show that @xmath106 factors through @xmath475 and the factored morphism @xmath476 is an open immersion . we first check that @xmath106 factors . by definition , @xmath106 factors if the pull back @xmath477 . let @xmath478 be the projection . by definition , @xmath479 is the composite @xmath480 where @xmath481 is the tautological section and @xmath482 is the restriction homomorphism . therefore , @xmath483 is the composite @xmath484 since @xmath485 , @xmath486 is the composite @xmath487 therefore @xmath483 is the composite @xmath488 since @xmath489 , we get @xmath490 . this proves that @xmath106 factors through @xmath491 we next prove that @xmath492 is an open immersion . we let @xmath493 and let @xmath494 be the tautological immersion . because @xmath471 is the total space of the vector bundle @xmath472 on @xmath143 , the morphism @xmath495 is equivalent to giving a section ( homomorphism ) @xmath496 at the same time , @xmath497 is equivalent to the vanishing of the composite of the homomorphisms : _ z p()^n p(()^n\| _ ) . we remark that because @xmath498 , by the universality property of morphisms to @xmath471 , ^(s_)=(s ) : _ ^p()^n ( ) ^n . to continue , we will show that the vanishing provides us a family of stable morphisms parameterized by an open subset of @xmath289 that contains @xmath499 . we let @xmath500 because @xmath501 , by the cohomology and base change theorem , we have the canonical identity @xmath502 this identity transforms into _ _ z ( _ z)^n ( _ z)^n\|__z . because @xmath503 is zero and the kernel of the second arrow is @xmath504 , @xmath505 factors through a unique homomorphism s=(s_1,,s_n ) : _ _ z^n . let @xmath506 be the section @xmath246 of @xmath507 . the @xmath508 sections @xmath509 considered as sections of @xmath510 on @xmath511 define a morphism [ s_0,,s_n ] : _ z \{s_0== s_n=0}. to analyze the domain of this morphism , we notice that due to , the morphism [ s_0,,s_n]=[s_0,,s_n] : =_u=_u , where @xmath512 is the lift of @xmath492 to @xmath465 . therefore @xmath513 because @xmath514 is proper , there is an open @xmath515 containing @xmath499 so that @xmath516 we let @xmath517 be the restriction of to @xmath518 . finally , because restricting to @xmath499 this morphism is a family of stable morphisms , possibly after shrinking @xmath519 if necessary , @xmath520 is a family of stable morphisms . we let @xmath521 be the tautological morphism induced by the family @xmath520 . because of the identity , the composite of @xmath522 with @xmath523 is identical to the projection @xmath524 . therefore , if we let @xmath525 @xmath526 is open and @xmath492 factor through @xmath527 we claim that , with @xmath526 endowed with the open subscheme structure of @xmath289 , the morphism @xmath528 is an isomorphism . to prove this , we will construct the inverse of @xmath528 . let @xmath529 be the morphism induced by @xmath530 . because the composite @xmath529 with the tautological @xmath141 is identical to the composite of @xmath531 with @xmath446 , the pair @xmath532 lifts to a morphism @xmath533 because of the identity , the composite @xmath534 is identical to the projection @xmath535 . this implies that @xmath536 . on the other hand , @xmath537 is exactly the inclusion @xmath538 , again due to the identity , therefore @xmath539 . thus @xmath528 is an isomorphism . this proves the theorem . the first is in a direct form : to every non - root vertex @xmath541 , we associate the coordinate function of @xmath542 indexed by @xmath203 : @xmath543 . to a terminal vertex @xmath544 , we associate @xmath545 coordinate functions @xmath546 . we then set @xmath547}w_{b , e},\quad z_{[b , o]}= \prod_{b \succeq a \succ o } \zetaz_a.\ ] ] we make a convention that if @xmath548 , we define @xmath549 and hence @xmath550 . the second is by induction on @xmath551 . _ if @xmath203 is a terminal vertex , we set @xmath552 with @xmath553 as before . if @xmath554 $ ] with @xmath555 having roots @xmath200 , then set @xmath556 _ the third is in terms of the bracket representation of @xmath182 . each @xmath557 is derived from @xmath182 by dropping the root @xmath183 , replacing each ghost vertex @xmath203 by its associated function @xmath558 , replacing each terminal vertex @xmath205 by @xmath559 , replacing `` , '' by `` @xmath560 '' , and replacing `` [ '' `` ] '' by `` ( '' `` ) '' . the resulting expression @xmath557 is identical to the one from the second method . take @xmath209 $ ] ( see the first tree in figure 1 ) . the domain of a generic @xmath561 in @xmath262 has a genus-1 ghost component labelled by the root @xmath183 , a genus zero ghost component labelled by @xmath205 , and three rational tails labelled by @xmath203 , @xmath204 and @xmath2 , respectively . the tails @xmath203 are attached to the genus 1 ghost component . tails @xmath204 and @xmath2 are attached to the rational component @xmath205 . by the first approach , @xmath562 by the second , @xmath563 } = \zetaz_aw_{a , e}+\zetaz_b(\zetaz_cw_{c , e}+\zetaz_dw_{d , e}).\ ] ] by the third , @xmath564 we now describe the local model of the singularity type of @xmath1 near @xmath262 . we let @xmath565 where @xmath566 ( resp . @xmath265 ) is the cardinality of @xmath185 ( resp . @xmath186 ) . the expressions @xmath557 then become regular functions on @xmath567 after we identify @xmath568 with the coordinate function of the @xmath203-th copy of @xmath569 and identity @xmath570 with the coordinate function of the @xmath205-th copy of @xmath571 in the @xmath572-th component of the product @xmath573 . let @xmath4\in\mpd$ ] be a closed point with associated terminally weighted rooted tree @xmath182 . we let @xmath582 be as in theorem [ thm : equ1 ] . theorem [ thm : key ] provides trivializations @xmath583 so that the restriction homomorphism @xmath584 is given by @xmath585}\oplus 0 : \bl\oplus_{b\in \vg^t}\so_\cv\br \oplus \so_\cv^{\oplus ( d-\ell+1 ) } \lra\so_\cv.\ ] ] the composite homomorphism @xmath586 induces a morphism @xmath587 the regular functions @xmath588 define a morphism = _ a_a : _ a = v_. together , they define a morphism : _ _ b^t^ne_= v_(_b^t^n . we comment that since deformations of nodal curves are unobstructed , the morphisms @xmath414 and @xmath589 are smooth . finally , because @xmath591 is smooth and @xmath592 is an open immersion , and thus smooth , the composite ( f=0)_z_e_is smooth . also , it is clear that a lift @xmath593 of @xmath259\in\mpd_\gamma$ ] is mapped to a point in @xmath594 . this proves that @xmath1 has singularity type @xmath182 at @xmath258 $ ] . for the purposes of keeping track of blowups of @xmath1 , we need to classify the singularity types of the blowups of @xmath574 . such types will be classified by simple weighted rooted trees that are monoidal transformations and collapsings of @xmath182 . we first classify the singularity types of the space @xmath574 for a simple terminally weighted tree @xmath182 . the singularity type of a @xmath595 is defined by its associated tree @xmath596 . let @xmath595 ( resp . @xmath597 ) and let @xmath598 ( resp . @xmath599 ) be its coordinate representation . the non - vanishing of @xmath568 identifies a subset of @xmath185 : @xmath600 we let @xmath596 be the collapsing of @xmath182 at vertices in @xmath601 . we now investigate the blowing up of @xmath574 . let @xmath602 it is clear that @xmath603 is smooth and locally closed . in general , the closure @xmath604 of @xmath605 in @xmath606 is quite complicated . however , in the case when @xmath607 , @xmath604 is smooth and bears a simple description . let @xmath616 be the trunk of @xmath182 and let @xmath617 be the direct - descendants of the branch point @xmath217 . then @xmath605 consists of those @xmath599 so that @xmath618 its closure is given by the vanishing of all @xmath619 : _ , k=\{_a_1==_a_k=0}. by the definition of @xmath596 , @xmath620 if and only if @xmath621 . the second claim is clear . we now consider the blowup of @xmath606 along @xmath604 in the case @xmath622 . we let @xmath623= o[v_1[\cdots[v_r[\gamma_1,\cdots,\gamma_k]]\cdots]],\ ] ] and let @xmath624 be the root of @xmath625 ; thus , @xmath626 are the direct descendants of the branch point @xmath217 of @xmath182 . accordingly , _ , e(z , w)=(z_v_1z_v_r)(z_a_1__1,e + + z_a_k__k , e ) . we denote the blowup of @xmath606 along @xmath604 by @xmath627}$ ] and define @xmath628}=z_\gamma\times_{v_\gamma } v_{\gamma,[k]}.\ ] ] the scheme @xmath629}$ ] is a subscheme of @xmath630 defined by the equations @xmath631 for @xmath632 . ( here @xmath633 $ ] is the homogeneous coordinate on @xmath634 . ) in the affine open subset @xmath635 , we have @xmath636 for @xmath637 . thus , over this chart , @xmath629}$ ] is defined by ( z_v_1z_v_r)_a_i ( _ _ i , e + _ j i u_j _ _ j , e ) = 0 , 1 e n. there are two cases . if @xmath625 is a single - vertex tree , then becomes ( z_v_1z_v_r_a_i ) ( w_a_i , e + _ j i u_j _ _ j , e ) = 0 , 1 e n. after introducing @xmath638 , becomes z_v_1z_v_r_i w_a_i , e=0 , 1 e n. this is the system associated to the tree @xmath639 $ ] , a path tree containing vertices @xmath640 . it is also the advancing @xmath624 in @xmath182 and thus is in @xmath226 . this shows that @xmath641 } \cap \{u_i=1\ } \cong z_{\gamma'}$ ] . if @xmath625 is a nontrivial rooted tree and is of the form @xmath642 $ ] , where @xmath643 because @xmath182 is simple by assumption , the system becomes ( z_v_1z_v_r)_a_i ( _ s=1^m _ b_s _ _ s,e+ _ j i u_j _ _ j , e ) = 0,1en , where @xmath644 is the root of @xmath645 . after replacing @xmath646 by @xmath647 , ( noticing that the system does not contain the variables @xmath619 , ) this is the system associated to the tree @xmath648 that is obtained from @xmath182 by advancing the vertex @xmath624 . this shows that @xmath641 } \cap \{u_i=1\ } \cong z_{\gamma'}$ ] . consider @xmath209 $ ] . then we have @xmath654 we blow up @xmath574 along the locus @xmath655 the blown - up is a subspace of @xmath656 defined by equations @xmath657 , where @xmath658 $ ] are the homogeneous coordinates of @xmath659 . in the affine open subset @xmath660 , the equation @xmath661 becomes @xmath662 it has normal crossing singularities , and the resulting system is associated to the tree @xmath663 $ ] , the advancing of @xmath203 of @xmath664 $ ] . in the affine open subset @xmath665 , the equation @xmath661 becomes @xmath666 this is the system associated to the tree @xmath667 $ ] , the advancing of @xmath205 of @xmath664 $ ] . we begin with recalling the notations and facts about the blown - up @xmath100 . in [ theta ] , we introduced @xmath80 that is a smooth locally closed substack of @xmath88 of ghost core elliptic curves attached on @xmath36 distinct smooth points with @xmath36-rational tails ; @xmath34 is the closure of @xmath80 . in [ inductive - blowups ] , we successively blow up @xmath88 along proper transforms of @xmath34 . inductively , after obtaining @xmath84}$ ] , we blow it up along the proper transform @xmath668 } \subset \fm\uwt_{1,[k-1]}$ ] of @xmath669 . we denote by @xmath670 } \sub \fm\uwt_{1,[k]}$ ] the proper transform of @xmath671 . we let @xmath672}\sub \fm\uwt_{1,[k]}$ ] be the exceptional divisor of the @xmath673}\to \fm\uwt_{1,[k-1]}$ ] . finally , we denote @xmath91 the resulting limit stack . as the image of @xmath674 is disjoint from @xmath34 for @xmath675 , the fiber product @xmath100 can be defined after @xmath2-th blowing up of @xmath88 : @xmath103}.\ ] ] to pave a way for our proof , we also need to record the intermediate blowup spaces . for this , we introduce for @xmath676 @xmath677 } = \mpd \times_{\fmw } \fm\uwt_{1,[k]}.\ ] ] to each closed point @xmath679}$ ] we can find a graph @xmath680}$ ] and a smooth morphism @xmath681}$ ] whose image contains @xmath682 of which the following holds : + ( i ) . there are smooth chart @xmath683}\uwt$ ] and smooth morphisms @xmath684 , @xmath685 , @xmath686 , and @xmath687 shown below , making the diagram commutative @xmath688}\\ @vv{\psi_\gamma}v @vv{p_\gamma}v @vvv\\ v_\gamma @<{\phi_\gamma } < < \cv_\gamma @ > > > \fm_{1,[k]}\uwt ; \end{cd}\ ] ] ( ii ) for any @xmath689 , @xmath690}\uwt}\mathring\theta_i = w_{\ga}\times_{v_{\ga}}\mathring\pi_{\ga , i}.\ ] ] the proof is by induction on @xmath36 . in the case @xmath243 , the desired morpisms are provided in the proof of theorem [ thm : localequtree ] . suppose @xmath608 and the lemma holds for a closed point @xmath691}$ ] . let @xmath692 etc . , be the corresponding data provided by the statement of the lemma . if @xmath613 or @xmath614 , then @xmath693}$ ] , while by ( ii ) with @xmath36 replaced by @xmath694 and lemma 5.12 @xmath695}\uwt}\theta_k= \eset.\ ] ] thus , @xmath682 does not lies over the blowing up center of @xmath89}\uwt$ ] , and is in @xmath696}$ ] . by shrinking @xmath692 if necessary , the morphism @xmath697 ( provided by the inductive assumption ) lifts to @xmath698}$ ] and the @xmath36-version of the statements ( i ) and ( ii ) are identical to its @xmath699-version . if @xmath609 , we define @xmath700}$ ] to be the fiber product using the right square of ( see below ) . by the property ( ii ) and the construction of the blowups @xmath701}\uwt\to\fm_{1,[k-1]}\uwt$ ] and @xmath627}\to v_\gamma$ ] , there is a unique @xmath702}$ ] making the left square a fiber product : this shows that @xmath703}\uwt } \fm_{1,[k]}\uwt = w_\gamma\times_{v_\gamma } v_{\gamma,[k]},\ ] ] which is smooth over @xmath629}=z_\gamma\times_{v_\gamma } v_{\gamma,[k]}$ ] . therefore , by lemma 5.14 @xmath704}\uwt } \fm_{1,[k]}\uwt$ ] is covered by smooth morphisms @xmath705}\uwt } \fm_{1,[k]}\uwt\ ] ] with @xmath706}$ ] that satisfy the statements of the lemma . finally , because @xmath707}=\widetilde{m}_1(\pn , d)_{[k-1]}\times_{\fm_{1,[k-1]}\uwt } \fm_{1,[k]}\uwt,\ ] ] the lemma follows . we now describe a stratification of @xmath100 . we define an equivalence relation on @xmath708}$ ] by demanding that @xmath709 if @xmath182 is isomorphic to @xmath182 as rooted but unweighted trees . note that @xmath692 and @xmath710 has isomorphic germs at their origin if and only if @xmath711 . for @xmath680}$ ] , we denote by @xmath712 $ ] the equivalence class of @xmath182 in @xmath204 . for any point @xmath713}$ ] , we let @xmath714 } ) \and ( \eta , w_\ga ) \lra ( x , z_\gamma)\ ] ] be the smooth morphisms provided by the previous lemma . we define @xmath715= [ \gamma_x]$ ] . by the comment in the previous paragraph , the equivalence class @xmath715 $ ] is independent of the choice of the chart covering @xmath716 . therefore , we can define @xmath717 , [ k ] } = \{\xi \in \wmpd_{[k ] } \mid [ \gamma_\xi]= [ \gamma]\}.\ ] ] in general , @xmath718 , [ k]}$ ] is a disjoint union of closed substacks . for each @xmath724}$ ] , it follows from the proof of lemma [ adhoc ] that to each @xmath725 , we can find an tale @xmath726 , a scheme @xmath727 and a smooth morphism @xmath728 with an open embedding @xmath729 . let @xmath730 be induced by . by the same reasoning as in the proof of theorem [ thm : localequtree ] , if we let @xmath731 and @xmath732 be the natural morphism ( similarly defined as ) , then @xmath733 since @xmath734}$ ] , @xmath182 is of the form @xmath735]$ ] with @xmath736 . hence each @xmath737 this proves the theorem . finally , since the primary component @xmath738 of @xmath739 is irreducible and of dimension @xmath740 , thus the primary component @xmath105 of @xmath100 , which is the proper transform of @xmath738 , is smooth and is defined by @xmath741 in each chart . * theorem [ thm : m2 ] . * _ for any @xmath742 , the direct image sheaf @xmath743 is locally free over every component @xmath111 where @xmath106 is either 0 or a partition of @xmath2 . it is of rank @xmath744 when @xmath113 and of rank @xmath745 otherwise . _ for @xmath746 , we set @xmath747 and follow the notations introduced earlier . we let @xmath748 be any point and let @xmath749 its image . we pick a local chart @xmath159 of @xmath750 , and introduce @xmath751 then @xmath752 is a chart of @xmath100 . continue to write @xmath142 the tautological family over @xmath143 , we set @xmath753 here the second equality in each set of identities follows from the universality of @xmath17 . we let @xmath754 and @xmath755 . the square for @xmath25 , by it is the direct sum of the zero homomorphism on @xmath304 with @xmath759}.\ ] ] therefore , @xmath757 is given by a direct sum of the zero homomorphism with the pullbacks of @xmath760}$ ] . we now let @xmath761\cdots ] ] \in \lambda_{d , [ d]}$ ] be the tree associated to @xmath762 and let @xmath182 be the tree associated to @xmath716 . by lemma [ adhoc ] , to @xmath763}$ ] we can find an tale neighborhood @xmath764 and a scheme @xmath765 with smooth morphisms @xmath766 ( we can assume @xmath767 . ) by the inductive construction of [ induc - blowup ] , we know that the tree @xmath229 is obtained from @xmath182 by successively advancing vertices until a terminal vertex is advanced . assume that @xmath768 is the terminal vertex ( in @xmath182 ) that is advanced in the last step ( here we adopt the indexing scheme as arranged in the proof of theorem [ thm : key ] ) . then by following the same inductive construction of [ induc - blowup ] step by step , we obtain @xmath439^_i= z_v_1 z_v_r ^_i \| ^_j .@xmath769 _ ( ) = ^ * _ * ( ) = _ ^__i _ w_^m+1,@xmath439 we see that the kernel of @xmath757 is a direct sum of @xmath770 with the kernel of the homomorphism @xmath771 now if @xmath772 , then @xmath773 does not vanish at general points of @xmath774 . hence the kernel sheaf of @xmath757 is locally free of rank @xmath322 over @xmath774 . if @xmath775 for @xmath106 a partition of @xmath2 , then , one of @xmath776 vanishes along @xmath777 . hence the kernel sheaf of @xmath757 is locally free of rank @xmath778 over @xmath777 . this proves the theorem . finally , we point out that all the main results in this paper generalize directly to moduli space of genus one stable maps with marked points . this extension requires introducing blowup loci involving the marked points analogous to those described in @xcite . m. kontsevich , _ enumeration of rational curves via torus actions . _ in : the moduli space of curves , ed . by r. dijkgraaf , c. faber , g. van der geer , progress in math . 129 , birkhuser , 1995 , 335368 .	we describe an algebro - geometric approach to vakil - zinger s desingularization of the main component of the moduli of genus one stable maps to @xmath0 @xcite . the new approach provides complete local structural results for this moduli space as well as for the desingularization of the entire moduli space and should fully extend to higher genera .
in hierarchical theories of structure formation , structures of progressively increasing size separate out from the hubble expansion , recollapse and virialise . groups of galaxies are the characteristic structures which have formed by the present epoch , and are believed to contain the bulk of the matter in the universe @xcite . as such , an understanding of the universe requires an understanding of galaxy groups . the depth of the potential wells of groups is similar to that of individual galaxies , and galaxy velocities within groups are only a few hundred km s@xmath1 . under these circumstances , galaxies can interact strongly with one another , and with the group potential ( e.g. @xcite ) . hence , not only are groups the most common environment for galaxies , but they provide an environment which has a strong effect on galaxy properties , and which is itself evolving , as groups collapse , virialise and grow through mergers and accretion . despite the crucial role played by groups in cosmic structure formation and evolution , they have received relatively little attention compared to larger clusters . this is at least partly due to the fact that typical groups may contain only a handful of bright galaxies in their inner regions . in optical data , groups are therefore often difficult to detect with confidence in projection . x - ray emission from a hot intragroup medium ( igm ) provides a much more reliable method of detecting the potential well of a virialised group . some earlier works based on pointed _ rosat _ x - ray observations utilized this to study the global properties of groups and those of their member galaxies in a coherent manner @xcite . however , although care had been taken to include groups spanning a wide range of properties , as in the study of @xcite , from a statistical viewpoint these works all suffered from the fact that _ rosat _ only targeted a heterogeneous sample of hand - picked groups . x - ray investigations of optically selected group samples include those of @xcite , and later @xcite , who invoked data from the _ rosat _ all - sky survey ( rass ) to study a redshift - selected , unbiased sample of groups in x - rays . however , no detailed information on the galaxy dynamics within these groups was extracted . @xcite also employed rass data to investigate the x - ray properties of a sample of groups , drawn from the optically selected and statistically complete poor cluster catalogue of @xcite and @xcite . this work furthermore included a detailed investigation of the galaxy dynamics within the groups , though the systems were originally selected by photometric enhancements and were not based on velocity information . rass data also formed the basis for the search for x - ray gas in the optically selected @xcite catalogue of groups conducted by @xcite , a work later augmented by @xcite , who included pointed rosat data . even in the latter study , however , the majority ( roughly two - thirds ) of the groups only had rass coverage . due to the shallow rass exposures of typically a few hundred seconds , the x - ray properties of groups with @xmath2 erg @xmath1 could generally not be investigated in these studies . moreover , in many cases the data only allowed for a detection of hot intragroup gas , whereas no detailed characterization of gas properties could be carried out . a further concern for rass based results for the diffuse x - ray luminosity @xmath3 from groups is the potential danger of contamination from active galactic nuclei due to the broad point spread function ( psf ) of the _ rosat _ pspc at large off - axis angles . with the exception of dedicated large - scale structure surveys ( e.g. , @xcite ) , more recent observations with _ xmm - newton _ and _ chandra _ have largely been targeting groups already studied by _ rosat _ in pointed observations or serendipitously detected in x - rays as part of the rass . it is not clear to what extent these groups , most of which are x - ray selected , can be seen as representative of the overall group population at low redshift . studies of clusters of galaxies indicate that x - ray selection could incorporate a serious bias , with x - ray and optically selected samples showing highly disparate x - ray properties and in some cases @xcite being largely non - overlapping . @xcite found that a significant fraction of optically selected clusters do not have a clear x - ray counterpart in _ rosat _ data , even though the physical reality of the x - ray faint systems were confirmed by spectroscopic follow - up . a similar conclusion , although based on optical multi - colour imaging rather than spectroscopy , was reached by @xcite using _ chandra _ data . the optically selected high - redshift clusters of @xcite were found to be x - ray underluminous for their velocity dispersion @xmath4 , deviating from the @xmath3@xmath4 relation derived for rich , nearby x - ray clusters . a similar result was shown for colour - selected clusters by @xcite , while @xcite found that a subset of their abell clusters were x - ray underluminous with respect to their virial mass . these results clearly suggest that x - ray selection could provide a biased picture of the cluster population , and there is little reason to assert that the situation for groups should be significantly different . given the lower x - ray luminosities of groups , such a bias is likely to be even stronger for these systems . as a consequence of the fact that most x - ray studies of groups have either been concentrating on x - ray selected targets picked from a variety of different group catalogues , or have employed the shallow rass data to investigate group x - ray emission , we currently have no _ unbiased _ census of the properties of hot gas and their relation to the dynamics of galaxies within galaxy groups . the advent of large redshift surveys , coupled with the high sensitivity , spatial resolution , and spectral capability of _ xmm - newton _ and _ chandra _ , has made it possible to remedy this situation . groups can be selected in redshift space , the properties of any hot igm can be assessed using deep x - ray data , and group dynamics can be studied with multi - slit spectrographs . we have launched a project to take advantage of this situation , by targeting a statistically representative sample of groups with both _ xmm - newton _ and the imacs multi - object camera and spectrograph @xcite , installed on the 6.5-m baade / magellan telescope at las campanas . the primary aim of the _ xi _ project is to understand the nature and evolution of the galaxy group population , and to study the way in which the properties of groups are related to those of their member galaxies . by performing a sensitive test for the presence of any hot igm in the selected groups , one of the key outcomes of this project will be a reliable estimate of the fraction of optically selected groups which actually contain a hot igm . another issue to be addressed is whether the properties of any hot x - ray gas can be used as an evolutionary tracer of the dynamical state of a group . the thrust of the present paper is to provide an initial attack on these questions , using the results obtained for the first four groups observed by _ xmm _ as part of the project . the large collecting area and field of view of _ xmm _ makes it the ideal x - ray instrument for this study . imacs has a field of view perfectly matched to that of _ xmm _ ( @xmath5 arcmin in both cases ) , and allows us to obtain several hundred spectra over this entire field in a single exposure , down to @xmath6 at the selected sample redshift of @xmath7 . we assume @xmath8 , @xmath9 , and a hubble constant of @xmath10 km s@xmath1 mpc . the virial radius @xmath11 of each group can then be identified with @xmath12 , the radius enclosing a mean density of 100 times the critical density @xmath13 @xcite . if the dark matter density of the groups follows an nfw profile @xcite with concentration parameters in the plausible range @xmath14 @xcite , we then have @xmath15 . in the adopted cosmology , the characteristic redshift of our sample of @xmath16 corresponds to a luminosity distance @xmath17 mpc , and 1 arcmin to @xmath18 kpc . a redshift - selected sample provides the best basis for a study of this type . x - ray selection is impractical , due to the lack of wide - area high - sensitivity data , and would in any case fail to detect groups which are collapsing but do not yet have virialised cores . we therefore selected our sample from the catalogue of 2209 groups derived by ( * ? ? ? * hereafter mz ) from a friends - of - friends ( fof ) clustering analysis ( optimised by reference to cosmological simulations ) on the first ( 100-k ) data release from the 2df galaxy redshift survey @xcite . from this large group catalogue , we selected a sample of 25 systems which satisfied the following additional criteria : ( i ) redshift @xmath19 : this narrow slice puts all groups on an equal observational footing at a redshift where the characteristic radius @xmath20 of groups of @xmath21 mpc corresponds to 13 arcmin and so is well matched to both the _ xmm _ and imacs fields of view . using a narrow redshift interval further has the advantage of circumventing a typical problem associated with fof - catalogues , namely that the fof linking length grows with @xmath22 , implying that the properties of fof - selected groups may vary systematically with redshift . ( ii ) velocity dispersion @xmath23 km s@xmath1 : limiting our study to poor systems , which are the most common ( containing more than half of all galaxies , see @xcite and @xcite ) and provide the environment in which dynamical evolution is most rapid , and the dispersion in observed properties is greatest . ( iii ) number of spectroscopically confirmed 2df member galaxies @xmath24 : reducing the danger of including groups which are not real physical associations , by allowing only systems which incorporate at least five 2df galaxies . @xmath25-body simulations show that fof - groups with @xmath26 almost always correspond to gravitationally bound structures rather than being unbound density fluctuations ( e.g. , @xcite ) . ( iv ) redshift completeness : we avoid the edges of the 2df survey area , and regions with poor completeness in the 100-k data release . within these constraints , we selected our sample of 25 groups entirely at random . since our aim is to explore the full population of optically selected systems , we deliberately did not attempt to apply any further constraints of regularity , optical luminosity or dynamical status . hence our sample contains groups spanning a wide range of properties , subject to having a number density contrast @xmath27 with respect to the mean galaxy number density @xmath28 , which distinguishes them in the fof analysis of mz . to illustrate this diversity in group properties , fig . [ fig , sigmarv ] shows the distribution of the full sample in the ( @xmath29)plane , with virial radii from @xcite based on the projected distribution of galaxies . highlighted are the four groups targeted for the initial study described in this paper . basic optical properties of these four groups are provided in table [ tab , opt ] . note that the subsample studied here includes the two groups with the largest values of @xmath30 in the full _ xi _ sample , and that , within the constraints imposed by our selection criteria , the _ xi _ sample covers most of the available parameter space . also note that , despite the fact that some of the groups are rather compact as judged from the mz virial radius , we find that none of the _ xi _ groups would satisfy the compact group criteria introduced by @xcite . this is in all cases because they fail to meet the compactness criterion @xmath31 , where @xmath32 is the total @xmath33-band magnitude per arcsec@xmath34 of the galaxies within 3 mag of the brightest one , averaged over the ( circular ) angular extent of the subgroup defined by these galaxies . [ cols="<,^,^,^,^,^,^ " , ] summarizing the x - ray properties of the hot igm in the four groups , we find that two groups , mz 5383 and mz 9307 , show no detectable diffuse emission , whereas mz 4577 shows some evidence of igm emission though this is not significant at the @xmath35 level . only in one system , mz 9014 , do we detect some irregular emission at @xmath36 significance . the x - ray luminosity of this system is among the lowest found for any x - ray detected group , and since its hot gas is clearly not in hydrostatic equilibrium , no x - ray mass analysis was attempted . on the basis of their velocity dispersions , all four groups are remarkably x - ray underluminous with respect to typical x - ray selected groups , as illustrated in fig . [ fig , lxsig ] . this plot shows the corresponding results from the gems sample of @xcite , along with several observationally derived @xmath3@xmath37 relations spanning a large range of derived slopes , from @xmath38 @xcite to @xmath39 @xcite ( all normalized to our adopted value of @xmath40 ) . even when comparing to the combined results from these highly disparate data sets , the _ xi _ groups emerge as significantly x - ray underluminous for their velocity dispersions . this is particularly true in light of the fact that , apart from the @xcite relation , the x - ray luminosities underlying the @xmath41@xmath37 relations shown in fig . [ fig , lxsig ] have generally not been extrapolated to @xmath42 . doing so would raise the normalization of the relations , thus aggravating the discrepancy with respect to the _ xi _ groups . it is perhaps also surprising that it is the two _ xi _ groups with the largest values of @xmath4 that remain undetected in the x - ray . these results indicate already that the full _ xi _ sample will have properties quite different from the heterogeneous group samples studied in the past , and that these earlier samples provided a biased view of the group population . given the importance of this conclusion , it seems worth addressing the robustness of the result shown in fig . [ fig , lxsig ] , before proceeding to investigate its source of origin . one potential concern is that @xmath3 of our groups has been evaluated inside a fiducial radius of @xmath43 ( assumed roughly equal to @xmath42 ) , for physical consistency and straightforward comparison to the results of @xcite . the reliability of the virial radii adopted from mz for this purpose could be questioned , as these are in some cases based on only five galaxies . this issue will be addressed in more detail when improved membership statistics become available for a larger sample of _ xi _ groups , but we note for now that the main conclusion that our four groups are x - ray underluminous for their velocity dispersions is clearly robust to reasonable changes in @xmath11 . another issue is whether the velocity dispersions could be overestimated in our analysis . we tested this for the three groups with available imacs velocities , by estimating errors on galaxy velocities both from the template fitting code and from the variation in derived redshifts resulting from manually fitting multiple line features . reassuringly , the resulting values of @xmath4 were entirely consistent . the argument that @xmath4 could be artificially boosted by contamination from interlopers could possibly be justified for a single group , but hardly for all four , and certainly not by a factor of 23 required to bring mz 5383 , mz 9014 , and mz 9307 into formal agreement with the @xmath3@xmath30 relation of @xcite . for mz 9307 , the velocity dispersion adopted from mz could potentially be unreliable due to the small number of galaxies involved . however , mz derive @xmath30 for groups with @xmath44 using the gapper estimator ( see e.g. @xcite ) , which should improve robustness in @xmath4estimates for poorly sampled systems . the consistency of the imacs results for @xmath4 with the corresponding mz values for the three other groups also suggests that the adopted velocity dispersion of mz 9307 can not be seriously biased . with the robustness of the result shown in fig . [ fig , lxsig ] reasonably well established , the question remains how to interpret the lack of significant x - ray emission in our groups , and whether this lack of emission is coupled to their dynamical state . it is most unlikely that the two undetected groups are not gravitationally bound , since ( as mentioned in section [ sec , sample ] ) only groups with number density contrasts @xmath45 have been included in the mz catalogue . this leaves at least three possible explanations for the lack of significant igm emission in our sample . \(i ) the groups could be in the process of collapsing for the first time , in which case the present velocity dispersion could be a poor proxy of the depth of the gravitational potential of the group or the temperature of its x - ray gas . this scenario allows for significant amounts of intragroup gas , which , however , would be more uniformly distributed than in collapsed groups and may not yet have been shock - heated to the virial temperature of the final system . in this case one might also expect to see evidence of dynamical substructure in the galaxy distribution . to investigate this , we plot velocity histograms of the groups in fig . [ fig , velhist ] . there is generally some indication , particularly for mz 5383 , of a central gap in the histogram . the result for mz 4577 also suggests a bimodal velocity distribution , with a subgroup of four galaxies separated from the remaining nine galaxies by @xmath46 km s@xmath1 . these features could be associated with the merging of two subgroups , but there is no convincing evidence that any apparent velocity substructures are localized on the sky in any of the groups . however , given the fairly modest number of galaxies found in each group , it is likely that general conclusions about velocity substructure in these four groups can not be reached on the basis of individual velocity histograms . to verify this , we confirmed , using the data from @xcite , that apparently virialised systems with a comparable number of confirmed members , such as hcg 42 and ngc 4325 , can show velocity histograms with features similar to those seen in fig . [ fig , velhist ] . potentially , some insight into the dynamics of the groups can still be gained by combining all the velocity measurements . to this end , we constructed a pseudo - group containing all confirmed member galaxies , by normalizing the galaxy velocities shown in fig . [ fig , velhist ] to the velocity dispersion of the relevant group . fig . [ fig , velhist_norm ] shows the resulting stacked histogram of the 65 normalized galaxy velocities . it is clear that the central deficit of galaxy velocities persists in this representation . using a k s test , we find that the probability that the normalized velocities have been drawn from a gaussian with unit variance @xmath47 ^ 2 $ ] , is @xmath48 for all 65 galaxies , decreasing to @xmath49 if including only the 58 galaxies with imacs velocity measurements . although still dealing with a relatively small number of galaxies , this result may suggest that the groups , as a class , are still in the process of dynamical relaxation . a more detailed dynamical analysis of the _ xi _ sample will be undertaken when results for the full sample become available . within the framework of the present analysis , we note that a picture in which the groups are still collapsing also has support from studies of other group samples . these include both purely optical analyses ( e.g.@xcite ; @xcite and references therein ) and x - ray / optical comparisons ( e.g. @xcite ) . in particular , @xcite infer a limiting number density contrast of @xmath50 for their p group subsample , comparable to that of the mz catalogue , and estimate that most of these groups are in the phase of collapse . @xcite investigated rass data for a sample of spectroscopically confirmed abell clusters and found that 50 out of 138 clusters were x - ray underluminous for their optically derived virial mass . although detecting no obvious optical substructure in their individual x - ray faint clusters , their superior statistics enabled these authors to conclude , on the basis of the position and velocity distributions of the cluster members , that these systems as a class are most likely still collapsing . further support for ongoing collapse of a considerable fraction of the group population comes from cosmological simulations , which suggest that 3050 per cent of all groups selected via standard fof - algorithms could be collapsing for the first time at the present epoch ( j. sommer - larsen , priv . comm . ) . \(ii ) a second explanation for the general lack of significant x - ray emission in the groups is that the gravitational potentials of these groups are too shallow to heat the intragroup gas to x - ray temperatures . as in scenario ( i ) above , there could be plenty of intragroup gas , but in this case its temperature would largely remain too low ( @xmath51 k ) to render it detectable in these observations . however , given the high values measured for the velocity dispersions of our groups , this explanation seems unattractive , and , as has already been discussed , the velocity dispersions of our groups are unlikely to be substantially overestimated . the fact that we _ do _ detect hot gas in two of the groups also suggests that this scenario can not offer an exhaustive explanation , even if assuming that the measured velocity dispersions are poor proxies of the total group mass . the long mean cooling times derived for the x - ray detected gas indicate that the gas is emitting inefficiently in the x - ray band due to its density , rather than temperature , being surprisingly low . the fact that mz 9014 is underluminous relative to the expectation from the @xmath3@xmath52 relation but is consistent with the @xmath53@xmath52 relation ( cf . section [ sec , mz9014 ] ) , seems to support this interpretation . \(iii ) alternatively , the groups could be x - ray faint because many collapsed groups simply contain very little intragroup gas . our unbiased selection could thus be picking up such systems because they are more numerous than x - ray bright groups . one mechanism which could give rise to this situation is strong galactic feedback ejecting a significant fraction of the original igm from the group potential . however , that would leave the challenge of explaining why feedback would be so much stronger in some groups as to reduce the x - ray detectable hot gas mass by 12 orders of magnitude relative to other systems with potential wells of similar depth . moreover , since the formation of elliptical galaxies is expected to generate more feedback than that of spirals @xcite , strong group - wide feedback would probably require a galaxy population with a low spiral fraction @xmath54 . if tentatively identifying those 2df group galaxies with values of the spectral type parameter @xmath55 as spirals ( see @xcite ) , then the mean spiral fraction of our four groups is @xmath56 per cent . such a large value of @xmath57 may not be easily reconciled with the notion that galactic feedback has ejected much of the igm from the groups . we note that @xcite detected x - ray emission in rass data from only 42 out of a statistically complete sample of 260 groups , suggesting that @xmath58 per cent of their groups do not contain a hot igm . however , due to the very shallow rass exposures , these authors could not study x - ray emission from groups with @xmath59 erg s@xmath1 , which would be expected to be more common . in fact , based on the expectation from x - ray bright groups , more than half of the 25 _ xi _ groups are expected to show x - ray emission around or below this limit ( cf . [ fig , sigmarv ] and [ fig , lxsig ] ) . until a larger sample of such groups has been studied using more sensitive x - ray data , it seems premature to accept scenario ( iii ) without considering viable alternatives . in summary , we can not at this stage exclude the possibility that the surprisingly low diffuse x - ray luminosity of our groups is due to any igm being either largely absent or too cold to produce copious x - ray emission . in order to help constrain the _ total _ gas content of the _ xi _ groups , we have commenced a programme of hi imaging of the groups using the giant metrewave radio telescope . when h i results for a significant number of groups become available , we will be in a better position to assess the validity of scenarios ii ) and iii ) above . until then , we find the idea that the groups could still be in the process of virialisation more attractive . in contrast to the other scenarios discussed , this explanation draws support from both cosmological simulations and other observational group studies , without facing any immediate challenges . we note that the morphological composition and the likely dynamical status of these groups suggest an analogy to the local group ( lg ) of galaxies , which consists of three spirals brighter than @xmath60 , with a varied assortment of early and late - type dwarfs , and is collapsing for the first time . although having a low velocity dispersion of @xmath61 km s@xmath1 , comparable to those of the lowest-@xmath4 groups in the _ xi _ sample ( fig . [ fig , sigmarv ] ) , the total lg mass of @xmath62 m@xmath63 ( e.g. @xcite ) suggests a mass overdensity with respect to the critical density of @xmath64 , within the region occupied by the milky way and m31 subgroups . the equivalent galaxy number density contrast in the adopted cosmology of @xmath65 suggests that the lg would in fact meet the overdensity criterion for inclusion in the mz catalogue . though no x - ray emitting hot igm has yet been detected in the lg , which is unlikely to have a diffuse x - ray luminosity of more than a few times @xmath66 erg s@xmath1 @xcite , recent detections of zero - redshift ovi and ovii absorption lines in _ fuse _ and _ chandra _ observations of quasars suggest the presence of an lg intergalactic medium of electron density @xmath67 @xmath68 and temperature @xmath69 k ( e.g.@xcite ) . this is consistent with the low fluxes we observe for our groups . our hi studies of the _ xi _ groups will help us compare their cold gas content with that of the lg . as will be described in more detail in a subsequent paper , a substantial fraction of the _ xi _ group members are found from our imacs spectroscopy to be emission - line galaxies . in particular , strong optical emission lines are found in roughly half of the confirmed member galaxies in the three groups with imacs data discussed here . hence , star - forming galaxies and active galactic nuclei are seen in all of these groups . given this , it is perhaps surprising that out of a total of 58 confirmed group galaxies , we detect x - ray emission above a few times @xmath66 erg s@xmath1 in only three of them ( cf.tables [ tab , opt ] and [ tab , results ] ) . these three galaxies are all ellipticals , displaying no evidence for significant optical emission lines . their x - ray luminosities are fairly typical of moderately x - ray bright ellipticals , and their measured luminosities and temperatures compare well to the expectations from the @xmath41@xmath52 relation derived for ellipticals by @xcite . the galaxies that show agn - like optical emission in our groups must have x - ray emission below our detection limit of @xmath70 erg s@xmath1 , and their nuclei could be related to the class of low - luminosity agn seen in some nearby galaxies ( see , e.g. , @xcite ) . from an optically selected sample of groups having pointed _ rosat _ pspc coverage , @xcite found that groups with detectable x - ray emission typically have a bright ( @xmath71 ) elliptical at the centre of the x - ray emission . in addition to the unexpectedly faint igm emission of the four _ xi _ groups studied here , another potential difference between these and more x - ray luminous systems is therefore that none of our groups hosts a dominant central elliptical galaxy . in two of the groups , including the x - ray detected mz 9014 , the optically most luminous galaxy is in fact a spiral . none of the ellipticals is 1 mag brighter than @xmath72 , and we note in particular that the brightest 2df galaxy in mz 9307 is @xmath73 mag _ fainter _ than @xmath72 . the absence of a dominant elliptical in the groups also seems to support the hypothesis that the groups are still collapsing , since such galaxies are probably formed via mergers in dense environments . we should also note , though , that the two x - ray detected groups , mz 4577 and mz 9014 , are not the only systems known to show diffuse intragroup x - ray emission which is not peaking on a central , bright early - type galaxy . other examples include the spiral - dominated hickson compact groups hcg 16 @xcite , hcg 57 @xcite , and the well - studied hcg 92 ( stephan s quintet ; @xcite ) , as well as the early - type dominated hcg 37 @xcite . these groups all exhibit low diffuse x - ray luminosities in the range @xmath74 15@xmath75 erg s@xmath1 , comparable to @xmath3 of the x - ray detected _ xi _ groups . systems like the hickson groups , likely to represent groups close to maximum collapse , are not very common in the nearby universe , however . for example , we find that only about @xmath76 per cent of the groups in the mz catalogue ( and , as mentioned in section [ sec , sample ] , none of the _ xi _ groups ) satisfy the @xcite criteria . outside hickson s catalogue , other examples are ngc 7777 and shk 202 @xcite . however , the number of such systems in the literature is still very low , placing mz 4577 and mz 9014 in an exclusive club . coupled with the results described in section [ sec , comparison ] , this is a further indication that the _ xi _ project is targeting a class of groups not previously studied in much detail in x - rays . as part of an ongoing effort to investigate the x - ray and optical properties of a substantial , statistically unbiased , and kinematically selected sample of galaxy groups , for the first time using deep x - ray data , we have performed _ xmm - newton _ observations of the first four groups in this sample . in two of the groups , we detect an x - ray emitting intragroup medium , with luminosities among the lowest found for x - ray detected groups so far . the two other groups observed here remain undetected in the x - ray , and all four groups are found to be x - ray underluminous for their velocity dispersions when compared to expectations from x - ray bright groups . furthermore , none of the groups hosts a dominant elliptical galaxy at the centre of the x - ray emission . our results therefore suggest that the nature of the igm in these optically selected groups may be very different from that seen in standard x - ray selected group samples . the fact that we are finding some of the faintest igm yet seen in groups in our first four observations clearly demonstrates that the _ xi _ project is exploring new parameter space . it also establishes that the predominantly x - ray selected group samples studied in the past with e.g. _ rosat _ were not representative of the overall group population . in contrast , the unbiased redshift selection employed in the _ xi _ project appears to be targeting a different class of groups not previously studied in much detail . the low levels of igm emission in the groups could be an indication that i ) the groups are in the process of collapsing for the first time , ii ) the gravitational potentials are too shallow for the gas to emit much x - ray emission ( i.e. the gas is essentially too cool to produce x - rays ) , or iii ) there is simply little or no intragroup gas . we find the first explanation to be the more attractive , at least until the validity of the other scenarios can be more firmly addressed by subjecting a larger sample of optically selected groups to deep x - ray observations . the idea that many groups are collapsing for the first time at the present epoch is also consistent with expectations from other optical group studies and with results of cosmological simulations . in contrast to the more well - studied class of x - ray bright groups , the type of groups studied in this paper truly represents the most common galaxy environments in the universe . investigating the detailed properties of a larger sample of such groups is therefore crucial if one is to obtain an unbiased understanding of the nature of the group population and a census of the distribution of baryons in the universe . we thank gus oemler , dan kelson , and greg walth for their role in developing the imacs reduction pipeline , gary mamon , andrea biviano , and jesper sommer - larsen for useful discussions , and the referee for comments which helped improve this paper . jr acknowledges the support of the european community through a marie curie intra - european fellowship under contract no . meif - ct-2005 - 011171 . jsm acknowledges partial support from nasa grants nng04gf78 g and nng04gc846 . arnaud m. , rothenflug r. , boulade o. , vigroux l. , vangioni - flam e. , 1992 , a&a , 254 , 49 arnaud m. et al . , 2002 , a&a , 390 , 27 awaki h. , koyama k. , matsumoto h. , tomida h. , tsuru t. , ueno s. , 1997 , pasj , 49 , 445 barkhouse w.a . , et al . , 2006 , apj , 645 , 955 beers t.c . , flynn k. , gebhardt k. , 1990 , aj , 100 , 32 belsole e. , sauvageot j .- , ponman t.j . , bourdin h. , 2003 , a&a , 398 , 1 bertin e. , arnouts s. , 1996 , a&as , 117 , 393 bigelow b.c . , dressler a.m. , 2003 , proc . spie , 4841 , 1727 bullock j.s . , kolatt t.s . , sigad y. , somerville r.s . , kravtsov a.v . , klypin a.a . , primack j.r . , dekel a. , 2001 , mnras , 321 , 559 burns j.o . , ledlow m.j . , loken c. , klypin a. , voges w. , bryan g.l . , norman m.l . , white r.a . , 1996 , apjl , 467 , l49 colless m. et al . , 2001 , mnras , 328 , 1039 de propris r. et al . , 2002 , mnras , 329 , 87 donahue m. et al . , 2002 , apj , 569 , 689 dos santos s. , mamon g.a . , 1999 , a&a , 352 , 1 ebeling h. , voges w. , bhringer h. , 1994 , apj , 436 , 44 eke v.r . , cole s. , frenk , c.s . , 1996 , mnras , 282 , 263 eke v.r . et al . , 2004 , mnras , 348 , 866 fukazawa y. , kawano n. , ohto a. , mizusawa h. , 2002 , pasj , 54 , 527 fukugita m. , hogan c.j . , peebles p.j.e . , 1998 , apj , 503 , 518 gilbank , d.g . , bower r.g . , castander f.j . , ziegler b.l . , 2004 , mnras , 348 , 551 girardi m. , giuricin g. , 2000 , apj , 540 , 45 helsdon s.f . , ponman , t.j . , 2000 , mnras , 315 , 356 helsdon s.f . , ponman , t.j . , 2003 , mnras , 340 , 485 hicks a. , ellingson e. , bautz m. , yee h. , gladders m. , garmire g. , 2004 , 35th cospar scientific assembly , 1624 hickson p. , 1982 , apj , 255 , 382 kelson d.d . , 2003 , pasp , 115 , 688 ledlow m.j . , loken c. , burns j.o . , hill j.m . , white r.a . , 1996 , aj , 112 , 388 lubin l.m . , mulchaey j.s . , postman m. , 2004 , apjl , 601 , l9 madgwick d.s . et al . , 2002 , mnras , 333 , 133 mahdavi a. , bhringer h. , geller m.j . , ramella m. , 1997 , apj , 483 , 68 mahdavi a. , bhringer h. , geller m.j . , ramella m. , 2000 , apj , 534 , 114 mendes de oliveira c. , amram p. , plana h. , balkowski c. , 2003 , aj , 126 , 2635 menon t.k . , 1992 , mnras , 255 , 41 merchn m. , zandivarez a. , 2002 , mnras , 335 , 216 miles t.a . , raychaudhury s. , forbes d.a . , goudfrooij p. , ponman t.j . , kozhurina - platais v. , 2004 , mnras , 355 , 785 mulchaey j.s . , zabludoff a.i . , 1998 , apj , 496 , 73 mulchaey j.s . , davis d.s . , mushotzky r.f . , burstein d. , 1996 , apj , 456 , 80 mulchaey j.s . , davis d.s . , mushotzky r.f . , burstein d. , 2003 , apjs , 145 , 39 navarro j.f . , frenk c.s . , white s.d.m . , 1995 , mnras , 275 , 720 nicastro f. et al . , 2003 , nature , 421 , 719 osullivan e. , ponman t.j . , collins r.s . , 2003 , mnras , 340 , 1375 osmond j. p. f. , ponman t. j. , 2004 , mnras , 350 , 1511 pietsch w. , trinchieri g. , arp h. , sulentic j.w . , 1997 , a&a , 322 , 89 pierre m. et al . , 2004 , jcap , 9 , 11 ponman t.j . , bourner p.d.j . , ebeling h. , bhringer h. , 1996 , mnras , 283 , 690 popesso p. , biviano a. , bhringer h. , romaniello m. , 2006 , a&a , in press ( astro - ph/0606191 ) ramella m. , pisani a. , geller , m.j . , 1997 , aj , 113 , 483 rasmussen j. , pedersen k. , 2001 , apj , 559 , 892 rasmussen j. , ponman t.j . , 2004 , mnras , 349 , 722 read a.m. , ponman t.j . , 2003 , a&a , 409 , 395 srsic j.l . , 1974 , ap&ss , 28 , 365 sulentic , j.w . , pietsch w. , arp h. , 1995 , a&a , 298 , 420 sutherland r.s . , dopita m.a . , 1993 , apjs , 88 , 253 terashima y. , iyomoto n. , ho l.c , ptak a.f . , 2002 , apjs , 139 , 1 trinchieri g. , sulentic j. , breitschwerdt d. , pietsch w. , 2003 , a&a , 401 , 173 trinchieri g. , sulentic j. , pietsch w. , breitschwerdt d. , 2005 , a&a , 444 , 697 tully , r.b . , 1987 , apj , 321 , 280 van den bergh s. , 1999 , a&ar , 9 , 273 verdes - montenegro l. , yun m.s . , perea j. , del olmo a. , ho p.t.p . , 1998 , apj , 497 , 89 white r.a . , bliton m. , bhavsar s.p . , bornmann p. , burns j.o . , ledlow m.j . , loken c. , 1999 , aj , 118 , 2014 zabludoff a.i . , mulchaey j.s . , 1998 , apj , 496 , 39	x - ray observations of hot , intergalactic gas in galaxy groups provide a useful means of characterizing the global properties of groups . however , x - ray studies of large group samples have typically involved very shallow x - ray exposures or have been based on rather heterogeneous samples . here we present the first results of the _ xi _ ( xmm / imacs ) groups project , a study targeting , for the first time , a redshift - selected , statistically unbiased sample of galaxy groups using deep x - ray data . combining this with radio observations of cold gas and optical imaging and spectroscopy of the galaxy population , the project aims to advance the understanding of how the properties and dynamics of group galaxies relate to global group properties . here , x - ray and optical data of the first four galaxy groups observed as part of the project are presented . in two of the groups we detect diffuse emission with a luminosity of @xmath0 erg s@xmath1 , among the lowest found for any x - ray detected group thus far , with a comparable upper limit for the other two . compared to typical x - ray selected groups of similar velocity dispersion , these four systems are all surprisingly x - ray faint . we discuss possible explanations for the lack of significant x - ray emission in the groups , concluding that these systems are most likely collapsing for the first time . our results strongly suggest that , unlike our current optically selected sample , previous x - ray selected group samples represented a biased picture of the group population . this underlines the necessity of a study of this kind , if one is to reach an unbiased census of the properties of galaxy groups and the distribution of baryons in the universe . [ firstpage ] galaxies : clusters : general galaxies : distances and redshifts x - rays : galaxies x - rays : galaxies : clusters
throughout , @xmath0 is a polish space and @xmath13 a sequence of @xmath0-valued random variables on the probability space @xmath14 . we let @xmath15 denote the borel @xmath7-field on @xmath0 and @xmath16 the set of probability measures on @xmath15 . a random probability measure on @xmath0 is a map @xmath17 such that @xmath18 , where @xmath19 is the @xmath7-field on @xmath20 generated by @xmath21 for all @xmath22 . for each @xmath9 , let @xmath23 be the @xmath24th _ predictive measure_. thus , @xmath23 is a random probability measure on @xmath0 , and @xmath25 is a version of @xmath26 for all @xmath27 . define also @xmath28 . if @xmath29 is _ exchangeable _ , as assumed in this section , there is a random probability measure @xmath3 on @xmath0 such that @xmath30 such an @xmath3 can also be viewed as @xmath31 where @xmath32 is the empirical measure . further , @xmath3 grants the usual representation @xmath33 where @xmath34 . let @xmath6 be a @xmath7-finite measure on @xmath15 . our _ first problem _ is to give conditions for @xmath35 the conditions should concern the finite dimensional distributions @xmath36 , @xmath9 , only . while investigating ( [ goal ] ) , one meets another problem , of possible independent interest . let @xmath11 denote total variation norm on @xmath37 . our _ second problem _ is to give conditions for @xmath38 again , let @xmath39 be exchangeable . reasonable conditions for ( [ goal ] ) look of theoretical interest . they are of practical interest as well thanks to bayesian nonparametrics . in this framework , the starting point is a prior @xmath40 on @xmath16 . since @xmath41 , condition ( [ goal ] ) is equivalent to @xmath42 this is a basic information for the subsequent statistical analysis . roughly speaking , it means that the `` underlying statistical model '' consists of absolutely continuous laws . notwithstanding the significance of ( [ goal ] ) , however , there is a growing literature which gets around the first problem of this paper . indeed , in a plenty of bayesian nonparametric problems , condition ( [ goal ] ) is just a crude _ assumption _ and the prior @xmath40 is directly assessed on a set of densities ( with respect to @xmath6 ) . see , for example , @xcite and references therein . instead , it seems reasonable to get ( [ goal ] ) as a consequence of explicit assumptions on the finite dimensional distributions @xmath43 , . from a foundational point of view , in fact , only assumptions on observable facts make sense . this attitude is strongly supported by de finetti , among others . when dealing with the sequence @xmath29 , the observable facts are events of the type @xmath44 for some @xmath9 and @xmath45 . this is why , in this paper , the conditions for ( [ goal ] ) are requested to concern @xmath43 , @xmath9 , only . some references related to the above remarks are @xcite and @xcite . in particular , in @xcite and @xcite , diaconis and freedman have an exchangeable sequence of indicators and give conditions for the mixing measure ( i.e. , the prior @xmath40 ) to be absolutely continuous with respect to lebesgue measure . the present paper is much in the spirit of @xcite and @xcite . the main difference is that we give conditions for the mixands @xmath46 , and not for the mixing measure @xmath40 , to be absolutely continuous . next , a necessary condition for ( [ goal ] ) is @xmath47 where @xmath48 . condition ( [ nec ] ) clearly involves the finite dimensional distributions only . thus , a ( natural ) question is whether ( [ nec ] ) suffices for ( [ goal ] ) as well . the answer is yes provided @xmath3 can be approximated by the predictive measures @xmath23 in some stronger sense . in fact , condition ( [ nec ] ) can be written as @xmath49 hence , if ( [ nec ] ) holds and @xmath50 , the set @xmath51 has probability 1 . and , for each @xmath52 , one obtains @xmath53 therefore , ( [ goal ] ) follows from ( [ nec ] ) and @xmath54 . in addition , a martingale argument implies the converse implication , that is , @xmath55 see theorem [ p09nh5 t ] . thus , our first problem turns into the second one . the question of whether @xmath56 is of independent interest . among other things , it is connected to bayesian consistency . surprisingly , however , this question seems not answered so far . to the best of our knowledge , @xmath54 in every example known so far . and in fact , for some time , we conjectured that @xmath54 under condition ( [ nec ] ) . but this is not true . as shown in example [ mainexamp ] , when @xmath57 and @xmath58 lebesgue measure , it may be that @xmath59 for all @xmath24 , and yet @xmath3 is singular continuous a.s . indeed , the ( topological ) support of @xmath60 has hausdorff dimension 0 for almost all @xmath61 . thus , ( [ nec ] ) does not suffice for ( [ goal ] ) . to get ( [ goal ] ) , in addition to ( [ nec ] ) , one needs some growth conditions on the conditional densities . we refer to forthcoming theorem [ rfg76yu8 ] for such conditions . here , we mention a result on the second problem . actually , for @xmath50 , it suffices that @xmath62 where @xmath63 denotes the continuous part of @xmath64 ; see theorem [ gt67fc4 ] . finally , most results mentioned above do not need exchangeability of @xmath29 , but the weaker assumption @xmath65 those sequences @xmath29 satisfying the above condition , investigated in @xcite , are called _ conditionally identically distributed _ ( c.i.d . ) . in this section , @xmath66 , @xmath67 for @xmath9 and @xmath68 . if @xmath69 is a random probability measure on @xmath0 , we write @xmath70 to denote the real random variable @xmath71 , @xmath27 . similarly , if @xmath72 is a borel function , integrable with respect to @xmath73 for almost all @xmath61 , we write @xmath74 to denote @xmath75 . let @xmath39 be c.i.d . , as defined in section [ mot ] . since @xmath29 needs not be exchangeable , the representation @xmath76 can fail for any @xmath77 . however , there is a random probability measure @xmath3 on @xmath0 such that @xmath78 for all @xmath27 . in particular , @xmath4 a.s . also , letting @xmath79 be the empirical measure , one obtains @xmath80 a.s . such an @xmath3 is of interest for one more reason . there is an exchangeable sequence @xmath81 of @xmath0-valued random variables on @xmath14 such that @xmath82 see @xcite for details . we next recall some known facts about vector - valued martingales ; see @xcite . let @xmath83 be a separable banach space . also , let @xmath84 be a filtration and @xmath85 a sequence of @xmath86-valued random variables on @xmath87 such that @xmath88 for all @xmath24 . then , @xmath85 is an @xmath89-martingale in case @xmath90 is an @xmath89-martingale for each linear continuous functional @xmath91 . if @xmath85 is an @xmath92-martingale , @xmath93 is a real - valued @xmath92-submartingale . so , doob s maximal inequality yields @xmath94 the following martingale convergence theorem is available as well . let @xmath95 be @xmath96-measurable and such that @xmath97 , where @xmath98 . then , @xmath99 provided @xmath100 a.s . for all @xmath24 and all linear continuous functionals @xmath101 . in the sequel , @xmath6 is a @xmath7-finite measure on @xmath15 . when @xmath57 , it may be natural to think of @xmath6 as the lebesgue measure , but this is only a particular case . indeed , @xmath6 could be singular continuous or concentrated on any borel subset . in addition , @xmath29 is c.i.d . ( in particular , exchangeable ) , and @xmath3 is a random probability measure on @xmath0 such that @xmath4 a.s . equivalently , @xmath3 can be obtained as @xmath102 a.s . it can ( and will ) be assumed @xmath103 . let @xmath109 be such that @xmath110 for all @xmath61 . since @xmath15 is countably generated , @xmath111 can be taken to be @xmath112-measurable ( see @xcite , v.5.58 , page 52 ) so that @xmath113 thus , given @xmath114 , @xmath115 is well defined for @xmath6-almost all @xmath116 . since @xmath29 is c.i.d . , condition ( [ cideq23vg ] ) also implies @xmath117 since @xmath15 is countably generated , the previous equality yields @xmath118 this proves that @xmath59 for all @xmath24 . in particular , up to modifying @xmath23 on a @xmath119-null set , it can be assumed @xmath120 for all @xmath114 , all @xmath61 , and suitable functions @xmath121 . regard @xmath122 as @xmath123-valued random variables . then , @xmath124 is @xmath125-measurable for @xmath126 is @xmath127-measurable for all @xmath128 . clearly , @xmath129 for all @xmath24 and @xmath130 . finally , @xmath29 c.i.d . implies @xmath131 by the martingale convergence theorem ( see section [ prel ] ) @xmath132 in the space @xmath123 , that is , @xmath133 in the exchangeable case , the argument of the previous proof yields a little bit more . indeed , if @xmath29 is exchangeable and @xmath5 a.s . , then @xmath134 where @xmath135 is any integer and @xmath136 . the next result deals with the second problem of section [ intro ] . for each @xmath137 , let @xmath138 and @xmath139 denote the continuous and discrete parts of @xmath140 , that is , @xmath141 for all @xmath27 and @xmath142 . [ gt67fc4 ] suppose @xmath39 is c.i.d . and @xmath143 . then , @xmath144 if and only if @xmath145 ( recall that @xmath146 denotes the basic @xmath7-field on @xmath147 ) . moreover , condition ( [ mo87xqml9 ] ) is automatically true if @xmath29 is exchangeable , so that @xmath148 provided @xmath29 is exchangeable and @xmath149 a.s . the `` only if '' part is trivial . suppose condition ( [ mo87xqml9 ] ) holds . for each @xmath114 , take functions @xmath150 and @xmath151 on @xmath20 such that @xmath152 and @xmath153 are measures on @xmath15 for all @xmath61 and @xmath154 for all @xmath27 . since @xmath29 is c.i.d . , condition ( [ cideq23vg ] ) yields @xmath155 a.s . we first prove @xmath156 . it can be assumed @xmath157 for all @xmath158 , so that @xmath159 for all @xmath61 and some function @xmath160 . for fixed @xmath27 , arguing as in the proof of theorem [ p09nh5 t ] , one has @xmath161 by standard arguments , it follows that @xmath162 a.s . again , it can be assumed @xmath163 for all @xmath61 and some function @xmath164 . define @xmath165 and regard @xmath166 as @xmath123-valued random variables . by the same martingale argument used for theorem [ p09nh5 t ] , one obtains @xmath132 in the space @xmath123 . that is , @xmath167 . we next prove @xmath168 . take @xmath169 as in condition ( [ mo87xqml9 ] ) , and define @xmath170 then , @xmath171 and @xmath172 for all @xmath173 and @xmath116 . define also @xmath174 since @xmath175 , then @xmath176 . fix @xmath52 and let @xmath177 . then @xmath178 since @xmath179 is countable and @xmath180 for all @xmath181 . further , @xmath182 therefore , @xmath183 is an immediate consequence of @xmath184 finally , suppose @xmath29 is exchangeable . we have to prove condition ( [ mo87xqml9 ] ) . if @xmath0 is countable , condition ( [ mo87xqml9 ] ) is trivial for @xmath185 for fixed @xmath27 . if @xmath57 , the glivenko cantelli theorem yields @xmath186 , where @xmath187 $ ] and @xmath188 is the empirical measure hence , ( [ mo87xqml9 ] ) follows from @xmath189 see corollary 3.2 of @xcite . if @xmath0 is any uncountable polish space , take a borel isomorphism @xmath190 . ( thus @xmath191 is bijective with @xmath191 and @xmath192 borel measurable ) . then @xmath193 is an exchangeable sequence of real random variables , and condition ( [ mo87xqml9 ] ) is a straightforward consequence of @xmath194 for each borel set @xmath195 . this concludes the proof . let @xmath85 and @xmath197 be independent sequences of independent real random variables such that @xmath198 and @xmath199 , where @xmath200 and @xmath201 . as shown in example 1.2 of @xcite , @xmath202 is c.i.d . and @xmath203 for some real random variable @xmath204 . since @xmath205 a.s . , then @xmath206 and @xmath207 a.s . ( in fact , @xmath208 a.s . ) . however , condition ( [ mo87xqml9 ] ) fails . in fact , @xmath59 for all @xmath24 , where @xmath6 is lebesgue measure . hence , @xmath209 while @xmath210 for all @xmath24 and almost all @xmath158 . we now turn to the first problem of section [ intro ] . recall that condition ( [ nec ] ) amounts to @xmath211 a.s . for all @xmath114 . therefore , up to modifying @xmath23 on a @xmath119-null set , under condition ( [ nec ] ) one can write @xmath212 for each @xmath61 , each @xmath114 and some function @xmath121 . we also let @xmath213 [ rfg76yu8 ] suppose @xmath39 is c.i.d . and @xmath214 for all @xmath24 . then @xmath5 a.s . if and only if , for each @xmath215 , @xmath216 in particular , @xmath5 a.s . provided , for each @xmath217 , there is @xmath218 such that @xmath219 moreover , for condition ( [ mer45vk01q ] ) to be true , it suffices that @xmath220 if @xmath5 a.s . , theorem [ p09nh5 t ] yields @xmath221 . thus , @xmath222 converges in @xmath223 , for almost all @xmath61 , and this implies condition ( [ serre123 ] ) . conversely , we now prove that @xmath5 a.s . under condition ( [ serre123 ] ) . fix a nondecreasing sequence @xmath224 such that @xmath225 , @xmath226 , and @xmath227 . since @xmath228 and @xmath0 is polish , there is @xmath229 satisfying @xmath230 and @xmath231 . by induction , for each @xmath232 , there is @xmath233 such that @xmath234 and @xmath235 . since @xmath29 is c.i.d . , condition ( [ cideq23vg ] ) implies @xmath236 define @xmath237 and @xmath238 . if @xmath239 , then @xmath240 moreover , @xmath241 . in fact , @xmath242 and @xmath243 a.s . for all @xmath24 , so that @xmath244 thus , to prove @xmath5 a.s . , it suffices to see that @xmath245 a.s . for all @xmath246 . suppose ( [ serre123 ] ) holds . fix @xmath247 , define @xmath248 and take a set @xmath249 such that @xmath176 and , for each @xmath52 , @xmath250 let @xmath52 . since @xmath251 and @xmath252 is uniformly integrable under @xmath253 , there is a subsequence @xmath254 and a function @xmath255 such that @xmath256 in the weak - topology of @xmath257 . this means that @xmath258 therefore , @xmath259 by standard arguments , the previous two relations yield @xmath260 thus , @xmath261 . this proves that condition ( [ serre123 ] ) implies @xmath5 a.s . fix @xmath27 , @xmath218 , and suppose @xmath264 . let @xmath265 for all @xmath266 . it can be assumed @xmath267 for all @xmath61 and @xmath9 . thus , each @xmath268 can be seen as an @xmath269-valued random variable such that @xmath270 further , @xmath271 is @xmath272-measurable for all @xmath273 , where @xmath274 . since @xmath29 is c.i.d . , condition ( [ cideq23vg ] ) also implies @xmath275 hence , @xmath276 is a @xmath277-martingale . by doob s maximal inequality , @xmath278 in particular , @xmath279 a.s . , and this completes the proof . third , suppose @xmath29 exchangeable , and fix _ any _ random probability measure @xmath285 on @xmath0 such that @xmath286 . then @xmath287 a.s . under the assumptions of theorem [ rfg76yu8 ] . in fact , @xmath3 and @xmath285 have the same probability distribution , when regarded as @xmath288-valued random variables . [ mainexamp ] let @xmath57 and @xmath58 lebesgue measure . all random variables are defined on the probability space @xmath289 . we now exhibit an exchangeable sequence @xmath29 such that @xmath59 for all @xmath9 and yet @xmath290 . in fact , the support of @xmath60 has hausdorff dimension 0 for almost all @xmath61 . two known facts are to be recalled . first , if @xmath291 and @xmath292 are independent @xmath293-valued random variables , then @xmath294 where @xmath45 and @xmath295 is the distribution of @xmath292 . hence , @xmath296 provided @xmath297 . the second fact is the following : define @xmath309 and @xmath310 then , @xmath39 is conditionally i.i.d . given @xmath311 . precisely , for @xmath61 and @xmath22 , define @xmath312 then , @xmath313 is a version of @xmath314 and @xmath315 . in particular , @xmath29 is exchangeable . moreover , @xmath80 a.s . for @xmath316 the ( topological ) support of @xmath60 has hausdorff dimension 0 for almost all @xmath61 . define in fact @xmath317 and @xmath318 . by theorem [ omygod47 ] , it suffices to verify that @xmath319 and condition ( [ m78yuhg5 ] ) follows immediately from @xmath320 we finally prove that @xmath59 for all @xmath9 . given the array @xmath321 , with @xmath322 for all @xmath323 , define @xmath324 fix @xmath9 and denote @xmath325 the @xmath326 identity matrix . if @xmath327 satisfies @xmath328 then @xmath329 in this case , since @xmath330 , then @xmath331 . hence , letting @xmath332 , the conditional distribution of @xmath333 given @xmath334 is absolutely continuous with respect to @xmath283 as far as @xmath327 satisfies ( [ uffa34ed5 t ] ) . to complete the proof , it suffices to note that @xmath335	let @xmath0 be a polish space and @xmath1 an exchangeable sequence of @xmath0-valued random variables . let @xmath2 be the predictive measure and @xmath3 a random probability measure on @xmath0 such that @xmath4 a.s . two ( related ) problems are addressed . one is to give conditions for @xmath5 a.s . , where @xmath6 is a ( nonrandom ) @xmath7-finite borel measure on @xmath0 . such conditions should concern the finite dimensional distributions @xmath8 , @xmath9 , only . the other problem is to investigate whether @xmath10 , where @xmath11 is total variation norm . various results are obtained . some of them do not require exchangeability , but hold under the weaker assumption that @xmath12 is conditionally identically distributed , in the sense of [ _ ann . probab . _ * 32 * ( 2004 ) 20292052 ] . ,
the study of ultracold atomic gases is one of the most actively developed areas of the physics of quantum many - body systems . initiated by the pioneering experiments with synthetic gauge fields in both bose gases @xcite and fermi gases @xcite , theoretical physicists took over the research for providing various schemes to synthesize certain extensions to rashba dresselhaus @xcite spin - orbit coupling for cold atoms @xcite . as a result , one derives a single - particle hamiltonian of the form @xmath1 , where @xmath2 is the laplacian , @xmath3 is the identity operator in @xmath4 ( or @xmath5 ) , and @xmath6 is the atom - light coupling containing the spin - orbit interaction of the rashba or dresselhaus form and the zeeman field . in a one - dimensional atomic center - of - mass motion , the simplified hamiltonian of a particle with mass @xmath7 ( in @xmath8 units ) accedes to a formal differential expression in the configuration space @xmath9 , @xmath10 ( @xmath11 ) , where @xmath12 labels the spin - orbit - coupling strength , @xmath13 results from the zeeman field and is named by the raman - coupling strength ; @xmath14 , @xmath15 are the pauli matrices . in ( [ eq : hamilt ] ) , @xmath0 obeys the meaning of a short - range disorder localized in the neighborhood of @xmath16 . it seems to be the first time when the spectral properties and in particular bound states of the hamiltonian realized through ( [ eq : hamilt ] ) are considered in detail . for the most part , our attempt to provide the analysis of the spectral characteristics for the spin - orbit hamiltonian is motivated by the work of @xcite , where the authors examined the free hamiltonian @xmath17 in @xmath18 , with @xmath19 in @xmath20 , and calculated , particularly , the dispersion relation . in a recent report of @xcite ( see also @xcite ) such a dispersion was shown to had been measured in @xmath21li . a straightforward calculation shows that the atom - light coupling @xmath6 is unitarily equivalent to @xmath22 ( @xmath23 , @xmath15 are the pauli matrices ) , and the associated unitary transformation is @xmath24 , where @xmath25 . the operator @xmath26 , provided @xmath27 , is nothing more than the free one - dimensional dirac operator for the particle with spin one - half and mass @xmath28 ( in @xmath8 units ) ; see @xcite for the analysis of this operator . it turns out that @xmath29 in ( [ eq : hamilt ] ) can also be interpreted as being equivalent to the ( operator ) sum of the free dirac operator plus a schrdinger operator @xmath30 . in particular , this means that , as the spin - orbit - coupling strength @xmath12 increases , @xmath31 approaches the one - dimensional massless dirac operator in weyl s form . for arbitrary @xmath27 , however , one can show that @xmath32 , with @xmath33 defined on a suitable domain ( sec . [ sec : fermi ] ) , is unitarily equivalent to @xmath34 , the one - dimensional dirac operator for the particle moving in fermi pseudopotential ( see ( [ eq : vf ] ) ) . this particular feature enables us to show that @xmath29 admits both continuous and discontinuous functions at a zero point . throughout , by a ( dis)continuous function @xmath35 , one accounts for the property whether @xmath36 ( continuity ) or not ( discontinuity ) , though @xmath35 is assumed to be defined on any subset of @xmath37 . originally , one would naturally conjecture that the disorder @xmath0 is prescribed by a potential well with its minimum at @xmath16 . a good survey of approximations by smooth potentials can be found , for example , in @xcite . also , there are numerous works concerning the generalized point - interactions in one - dimension ; see eg the papers of @xcite , and also the citations therein . in the present paper , we assume that @xmath0 is approximated by the square - well of width @xmath38 and depth @xmath39 for some arbitrarily small @xmath40 ; the coupling strength of interaction is @xmath41 . evidently , this is a familiar @xmath42-interaction . the one - dimensional schrdinger and dirac operators with @xmath42-interaction are known to be well - defined via the boundary conditions for everywhere continuous functions . in our case we have a mixture , to some extent , of schrdinger - like and dirac - like operators . in sec . [ sec : extra ] we argue that in such a case there is a possibility that discontinuous eigenfunctions would appear . to avoid the difficulties concerning the uniqueness of self - adjoint extensions of the operators on intervals @xmath43 , @xmath44 $ ] and @xmath45 , we consider two distinct representations of @xmath29 in the hilbert space @xmath46 . the first one , denoted @xmath47 , is obtained by integrating @xmath29 in the interval @xmath48\ni 0 $ ] and then taking the limit @xmath49 ; this gives the required boundary condition in defining the domain @xmath50 of @xmath47 . the second representation of @xmath29 is a distribution @xmath51 on @xmath52 , with @xmath42 the delta - function . here and elsewhere , @xmath53 , with @xmath54 , is the closure of @xmath55 in @xmath56 , the sobolev space of functions whose ( weak ) derivatives of order @xmath57 are in @xmath58 ( * ? ? ? 3 ) ; we also use the notation @xmath59 . by default , we take into account the isomorphism from @xmath46 to @xmath60 by ( * ? ? ? * theorem ii.10 ) . to demonstrate that representatives @xmath47 and @xmath61 are proper realizations of @xmath29 we explore the method developed by @xcite . as a result , we establish that @xmath62=0 $ ] in a strict ( classical ) sense , and that @xmath63=0 $ ] in a weak ( distributional ) sense . here @xmath64 . the commutator predetermines a nonempty set of common eigenfunctions of @xmath47 and @xmath65 , provided @xmath66 ( theorem [ thm : ea ] ) . the latter inequality shows that extra states in @xmath67 can be observed only for nonzero spin - orbit and raman coupling , and that their appearance in the spectrum is essentially dependent on the location of the dressed spin states @xcite in the dispersion curve . although @xmath47 and @xmath61 are equivalent representations for providing the spectral characteristics for @xmath29 in @xmath46 , we explore both of them . the main reason for such a choice is because the interaction is drawn in @xmath61 explicitly , and thus one can easier attach the physical meaning to @xmath61 , rather than @xmath47 ; the same applies to @xmath68 and @xmath65 , respectively . on the other hand , equivalence classes of functions in @xmath69 , with @xmath70 , are in a one - to - one correspondence with functions in @xmath71 , with the same @xmath72 , if and only if one imposes certain conditions on the normalization constant and the eigenfunction itself ( sec . [ sec : spectrum ] ) . this agrees with ( * ? ? ? v.4 ) , which in our case says that weak solutions @xmath69 are equal to the classical solutions @xmath71 if and only if the classical solutions exist . the paper is organized as follows . in sec . [ sec : operators ] , we give basic definitions of potential @xmath0 and the representatives @xmath47 , @xmath61 , and examine their correctness . [ sec : fermi ] deals mainly with operator @xmath65 and its distributional version @xmath68 . as a result , the fermi pseudopotential @xmath73 is introduced . in sec . [ sec : extra ] , we provide spin - orbit induced states for @xmath47 , as well as compute the essential spectrum . finally , we compute the remaining part of the discrete spectrum of @xmath47 ( @xmath61 ) in sec . [ sec : spectrum ] , and summarize the results in sec . [ sec : summary ] . throughout , we define @xmath59 , @xmath74 , @xmath75 for @xmath54 , @xmath76 for some @xmath77 , @xmath78 $ ] for some @xmath40 . given function @xmath0 which is defined as the limit of a sequence of rectangles @xmath79 then @xmath80 is supported in @xmath81 , and it approaches @xmath42 , the delta - function , in the usual sense of distributions , with the property @xmath82 . as a matter of fact , @xmath80 has a wider meaning than @xmath42 in the sense that ( * ? ? ? * eq . ( 7 ) ) [ eq : v - delta ] @xmath83 ( @xmath84 is the @xmath85th derivative of @xmath35 with respect to @xmath20 at a given point ) . as a functional , @xmath86 if and only if @xmath87 for @xmath88 for @xmath89 in particular , ( [ eq : v - delta-1 ] ) yields @xmath90 equation ( [ eq : v - delta-2 ] ) serves for the criterion in establishing whether the delta - function approximation of ( [ eq : v ] ) is a proper one . this is done by calculating @xmath84 at @xmath16 for all @xmath91 , where function @xmath35 is in the kernel of the operator that involves @xmath0 as in ( [ eq : v ] ) . afterward , one needs to verify under what circumstances the infinite series in ( [ eq : v - delta-2 ] ) converges . for the analysis of specific operator classes , the reader is referred to @xcite . the application of ( [ eq : v - delta-2 ] ) to @xmath29 in ( [ eq : hamilt ] ) is examined below . let @xmath92 in @xmath81 . the solutions @xmath93 ( @xmath94 ) are found by solving the characteristic equation for @xmath29 : @xmath95=0 $ ] ( @xmath96 ) or explicitly , @xmath97 the solutions with respect to @xmath98 read @xmath99 and so @xmath100 the upper , @xmath101 , and lower , @xmath102 , components of @xmath35 are then of the form @xmath103 for some @xmath104 , @xmath105 . clearly , @xmath106 as @xmath49 . hence @xmath107 , @xmath92 is continuous at @xmath16 . the @xmath85th derivative ( @xmath91 ) of @xmath35 at @xmath16 is found by differentiating @xmath108 @xmath85 times with respect to @xmath109 and then setting @xmath16 , @xmath110 as seen , @xmath111 with @xmath112 for @xmath89 . but then @xmath113 as @xmath49 , and the infinite series in ( [ eq : v - delta-2 ] ) vanishes . this proves that , as a functional , @xmath86 makes sense for functions in certain domains of @xmath29 . as a result , at least two possibilities are valid to construct these domains . the first one is obtained by integrating @xmath114 in @xmath81 and then taking the limit @xmath49 . in agreement with ( [ eq : v - delta-2 ] ) and the discussion above , this gives the operator @xmath115 ( @xmath116 ) where @xmath117 is of the form in ( [ eq : v - delta-1 ] ) . it appears from ( [ eq : a ] ) that for zero spin - orbit coupling @xmath118 , or continuous functions at @xmath16 , the boundary condition in @xmath50 is a familiar relation valid for the operators with @xmath42-interaction . this suggests the second realization of @xmath29 in @xmath119 , namely , @xmath120 with @xmath42 the delta - function . here we recall that although @xmath61 is a distribution , operator @xmath47 can be interpreted in the classical sense due to the fact ( * ? ? ? * theorem 3.17 ) that distributional and classical derivatives coincide whenever the latter exist ( and certainly are continuous on @xmath121 ) . if , however , we start from the pure point - interaction ( that is , @xmath42-interaction ) and integrate @xmath61 in @xmath81 , we derive that the property @xmath107 is only the ( additional , though reasonable ) assumption , as also discussed by @xcite . moreover , the operator @xmath122 is not self - adjoint , and it has deficiency indices , d.i . , ( 2,2 ) as @xmath49 . this means that additional boundary conditions at @xmath123 are required , and so again , @xmath124 is not necessarily equal to @xmath125 , in general . this is our motive to inspect the boundary condition in @xmath50 in its most general form . to this end , let us comment on the self - adjointness of operator @xmath47 ( @xmath61 ) . let us solve @xmath126 for some @xmath127 . the solutions @xmath128 are of the form ( [ eq : fss ] ) , with @xmath129 in ( [ eq : kss ] ) replaced by @xmath130 for @xmath131 , one requires @xmath132 in order to make solutions square integrable . this yields @xmath133 and @xmath134 . for @xmath135 , however , @xmath136 , and possible values are @xmath137 and @xmath138 . evidently , the intersection of possible solutions which are square integrable in the whole @xmath5 is the empty set . in terms of deficiency indices , operator @xmath47 has d.i . ( 0,0 ) , hence self - adjoint . a general solution to @xmath139 for @xmath127 can be written in the form @xmath140 where @xmath141 is the fourier transform of @xmath6 . to see this , one simply needs to solve @xmath142 ( @xmath143 ) by noting that @xmath144 , in agreement with ( [ eq : b ] ) ; here @xmath145 . it follows from ( [ eq : fp ] ) that the fourier transform @xmath146 of @xmath128 is proportional to @xmath147 . as a result , @xmath148 is not in @xmath119 ( * ? ? ? ix.6 ) , hence @xmath61 has d.i . similarly to the case for the dirac operator , one can also construct the quadratic form @xmath149 and show that it satisfies the klmn theorem ( * ? ? ? * theorem x.17 ) with respect to @xmath150 . in the present section we consider the operator @xmath151 ( @xmath116 ) . as discussed in sec . [ sec : intro ] of the present paper , @xmath6 has a meaning of the atom - light coupling originated from the synthetic gauge fields ( for more details , the reader is referred to @xcite ) . now we wish to examine the properties of its representative @xmath65 . the arguments of self - adjointness are similar to those for operator @xmath47 in the previous section . one solves @xmath152 with respect to @xmath153 for @xmath127 , and gets that @xmath154 ( @xmath155 ) . clearly , @xmath128 is not in @xmath119 , hence @xmath65 has d.i . [ alternatively , one can explore the weyl s criterion by noting from ( [ eq : f12 ] ) that there is one solution in @xmath58 as @xmath156 , and one solution as @xmath157 . ] the boundary condition in ( [ eq : a0 ] ) suggests that , similarly to the case of operator @xmath47 and its distributional version @xmath61 , there should be some weak form , @xmath68 , of @xmath65 as well . given @xmath158 on @xmath159 for some distribution @xmath73 . let us integrate @xmath160 in @xmath81 for @xmath161 , and then take the limit @xmath49 , @xmath162 in @xcite , the authors have defined the modified @xmath163-interaction to which we refer as the @xmath164-interaction , @xmath165 the reason for modifying the original @xmath163-interaction is that it is not applicable to discontinuous functions , as pointed out by @xcite . the integral ( * ? ? ? * eq . ( 44 ) ) @xmath166 diverges for discontinuous functions , as @xmath49 , because of the last term . on the other hand ( see also ( * ? ? ? * eq . ( 24 ) ) ) , the integral @xmath167 is convergent . below we show that the divergent term can be canceled in the following manner : [ prop:1 ] let @xmath168 . let @xmath164 be as in ( [ eq : deltap ] ) . then for @xmath49 , @xmath169 where @xmath170 for @xmath171 , and the same for @xmath172 . to prove the statement we only need the definition of @xmath164 , ( [ eq : deltap ] ) , and that of @xmath163 , @xcite , @xmath173 let @xmath174 and @xmath175 for @xmath40 arbitrarily small . by ( [ eq : delta ] ) , @xmath176f(x)dx = \frac{1}{2\beta}[(f(\alpha-\beta)-f(\alpha+\beta ) ) \\ & -(f(-\alpha-\beta)-f(-\alpha+\beta))]=-\frac{f(\alpha+\beta)-f(\alpha-\beta)}{2\beta}+ \frac{f(-\alpha+\beta)-f(-\alpha-\beta)}{2\beta } \\ & = -f^{\prime}(\alpha)+f^{\prime}(-\alpha).\end{aligned}\ ] ] in the limit @xmath177 , this gives ( [ eq : vfderiv2 ] ) . by ( [ eq : deltap ] ) and ( [ eq : delta ] ) , @xmath178\tilde{f}(x)dx \\ & = \frac{1}{2\beta}[(\tilde{f}(\alpha-\beta)-\tilde{f}(\alpha+\beta))- ( \tilde{f}(-\alpha-\beta)-\tilde{f}(-\alpha+\beta))]=\frac{1}{2\beta}\biggl[\biggl(f(\alpha-\beta ) \\ & -\frac{f(0_{+})-f(0_{-})}{2}-f(\alpha+\beta)+\frac{f(0_{+})-f(0_{-})}{2}\biggr)-\biggl(f(-\alpha-\beta)+\frac{f(0_{+})-f(0_{-})}{2 } \\ & -f(-\alpha+\beta)-\frac{f(0_{+})-f(0_{-})}{2}\biggr)\biggr]= -\frac{f(\alpha+\beta)-f(\alpha-\beta)}{2\beta}+\frac{f(-\alpha+\beta)-f(-\alpha-\beta)}{2\beta } \\ & = -f^{\prime}(\alpha)+f^{\prime}(-\alpha).\end{aligned}\ ] ] in the limit @xmath177 , we again derive ( [ eq : vfderiv2 ] ) . the proof is accomplished . we apply proposition [ prop:1 ] to functions in @xmath179 . then the substitution of the left - hand side of ( [ eq : vfderiv2 ] ) in ( [ eq : vfderiv ] ) along with @xmath180 ( @xmath117 as in ( [ eq : v - delta-1 ] ) ) yields @xmath181 ( @xmath182 ) , with @xmath183 relevant to proposition [ prop:1 ] . by virtue of ( [ eq : vf ] ) we have found that suitably rotated in spin space ( recall the unitary operator @xmath24 , with @xmath184 , discussed in sec . [ sec : intro ] ) , the operator @xmath32 , with @xmath65 as in ( [ eq : a0 ] ) and the spin - orbit coupling @xmath27 , describes the dirac - like ( or weyl dirac ) particle of spin one - half and mass @xmath28 moving in the fermi pseudopotential @xmath185 . we close the present section with the spectral properties of @xmath65 ( @xmath68 ) . [ thm : a0 ] 1 . the resolvent of @xmath65 is given by + @xmath186 \\ \intertext{($x\neq x^{\prime};x , x^{\prime}\in{\varmathbb{r}};z\in{\varmathbb{c}}\backslash\sigma(a_{0});\omega,\eta>0 ; { \mathrm{re}}\:\omega_{z}\neq0;\gamma\in{\varmathbb{r}_{0}}$ ) , where $ a_{0}^{0}=u\upharpoonright w_{0}^{1}({\varmathbb{r}_{0}})^{2}$ and } ( a_{0}^{0}-z{\text{\large $ \otimes$}}i)^{-1}(x - x^{\prime})= & \frac{e^{-\omega_{z}\vert x - x^{\prime}\vert}{\text{\large $ \otimes$}}i}{2\eta^{2}\omega_{z}}\ : ( i\eta\omega_{z}\:{\mathrm{sgn}}(x - x^{\prime}){\text{\large $ \otimes$}}\sigma_{2}+(\omega/2){\text{\large $ \otimes$}}\sigma_{3}+z{\text{\large $ \otimes$}}i)\end{aligned}\ ] ] + ( @xmath187 ) , where @xmath188 is as in ( [ eq : fz12 ] ) ; 2 . @xmath189 + ( @xmath190 ) , where @xmath191 denotes the heaviside theta function , and @xmath192 ( @xmath193 ) for the upper ( lower ) sign in @xmath194 ; 3 . @xmath195 , with @xmath196 ( @xmath197 ) containing equivalence classes of functions @xmath198 ( @xmath199 ) ; 4 . @xmath200\cup[\omega/2,\infty)$ ] ( @xmath201 ) ; 5 . there are no eigenvalues embedded into the essential spectrum : @xmath202 . ( see eq . ( [ eq : a0 ] ) and theorem [ thm : a0 ] ) for the point - interaction strength @xmath203 and the spin - orbit - coupling strength @xmath204 ( in @xmath8 units ) . the eigenvalues divided by @xmath27 are those of the one - dimensional dirac - like operator for the particle of spin one - half and mass @xmath28 moving in the fermi pseudopotential ( [ eq : vf ] ) . in figure , red lines show the border of the essential spectrum of @xmath65 , which is @xmath205 . the blue @xmath206 ( green @xmath207 ) line , showing the bound state as a function of the raman coupling @xmath208 , corresponds to the eigenfunction with a zero - valued lower ( upper ) component at the origin @xmath16 . ] \(1 ) in order to find the eigenvalues @xmath209 explicitly , one needs to solve the cubic equation with respect to @xmath210 , as it is seen from theorem [ thm : a0]-(ii ) . the solutions to such type of equations are well known for a long time . however , their general form is rather complicated and we did not find it valuable here . instead of that we displayed the spectrum of @xmath65 versus the raman coupling @xmath208 in fig . [ fig : eigenvaluesa0 ] . \(2 ) we also note that , unlike in theorem [ thm : a0]-(iii ) , where @xmath117 is undetermined because of the delta - function , @xmath117 in theorem [ thm : a0]-(ii ) obeys the form as in ( [ eq : v - delta-1 ] ) . the solutions in @xmath211 are strict so that @xmath117 can be replaced by any constant ( @xmath212 , say ) . \(i ) the integral kernel @xmath213 ( for simplicity , we replace @xmath214 by @xmath109 ) is defined through the formal differential equation @xmath215 in agreement with ( [ eq : vf ] ) , @xmath216 is of the form @xmath217 and @xmath218 . as one would have noticed , @xmath219 is the fourier transform of @xmath220 . recalling that the integrals @xmath221 , @xmath222 for @xmath223 and @xmath224 , we derive the expression @xmath225 with the integral kernel @xmath220 as in the theorem . by using this equation , calculate @xmath226 and @xmath227 , and get the equation for @xmath228 , @xmath229\phi(\gamma;z)= 2\eta^{2}\omega_{z}{\text{\large $ \otimes$}}i.\ ] ] substitute obtained expression of @xmath228 in ( [ eq : gr0 ] ) , replace @xmath109 by @xmath214 back again and get ( i ) , as required . note that @xmath230 is because of @xmath231 ( in the sense of distributions ) , that is , the resolvent of @xmath65 ( @xmath68 ) is a distribution , and hence the equation @xmath232 is meaningless in the classical sense . \(ii ) the discrete spectrum is easily recovered by setting the denominator of the resolvent of @xmath65 equal to zero . as for the eigenfunctions , we begin with ( [ eq : fz12 ] ) by letting @xmath233 and @xmath234 . we rewrite ( [ eq : fz12 ] ) in the following form @xmath235 ( @xmath236 ) , where @xmath237 , and @xmath238 for @xmath239 . by ( [ eq : f12 ] ) , with @xmath117 as in ( [ eq : v - delta-1 ] ) , @xmath240 but @xmath161 , ( [ eq : a0 ] ) , and so it must hold @xmath241 ( @xmath236 ) , with @xmath242 for @xmath239 . after some elementary simplifications , equations ( [ eq : f12 ] ) and ( [ eq : vareps ] ) lead to ( ii ) . \(iii ) let @xmath243 for some @xmath244 . combining the fourier transform of ( [ eq : vf ] ) with ( [ eq : green ] ) we get that @xmath245 now , if we calculate @xmath246 by taking @xmath35 from the left side of the above expressions , we get that @xmath247 and that @xmath248 thus yielding ( iii ) . \(iv ) the essential spectrum of @xmath65 is found from the dispersion curve @xmath249 which in turn is found by taking the fourier transform of @xmath6 and solving the eigenvalue equation , namely , @xmath250 the result reads @xmath251 for all @xmath143 . the essential spectrum of @xmath68 is found from the integral kernel of the resolvent of @xmath65 , by virtue of ( iii ) . this is exactly the case as for deriving the spectrum of @xmath252 . then one needs to solve @xmath253 with respect to @xmath254 ( @xmath143 ) . the solutions are those as above , and hence ( iv ) holds . \(v ) the present item immediately follows from ( iv ) and from the requirement that , for @xmath209 , it holds @xmath255 . [ lem : commutator ] we have : 1 . @xmath62=0 $ ] on @xmath50 strictly ; 2 . @xmath63=0 $ ] almost everywhere in @xmath9 . we note that @xmath256 for @xmath257 ; see eg @xcite . by ( [ eq : a ] ) and ( [ eq : a0 ] ) , @xmath258 . by ( [ eq : b ] ) and ( [ eq : vf ] ) , @xmath259 . thus @xmath62 $ ] makes sense since @xmath260 , @xmath261 , and the same for @xmath63 $ ] ( @xmath262 is the range ) . item ( 1 ) is easy to perform : @xmath62 $ ] on @xmath50 is given by @xmath263=[-\delta{\text{\large $ \otimes$}}i , u]=0 $ ] . the same applies to the resolvents @xmath264 , @xmath265 ( @xmath266 ) and to the exponents @xmath267 , @xmath268 ( @xmath269 ) in consonance with ( * ? ? ? * theorem viii.13 ) . the fact that the exponents commute follows from the commutation relation of resolvents . this can be seen by noting eg @xmath270 ( @xmath271 ) . that the resolvents commute ( weakly ) , the easiest way to see this is to apply ( [ eq : green ] ) and ( [ eq : green2 ] ) , where one concludes that the integral @xmath272f)(x)dx$ ] is equal to @xmath273f(x)dx=0 $ ] , provided @xmath230 . in order to prove ( 2 ) , we integrate @xmath63 $ ] in the interval @xmath274 because @xmath259 contains functions which are well - defined for @xmath223 . in this case , all integrands containing @xmath42 or @xmath163 ( see ( [ eq : delta ] ) ) vanish because the argument of @xmath42 ( @xmath163 ) is nonzero for all @xmath275 . the remaining terms , that is , those which do not include deltas , commute with each other . finally , we extend @xmath274 to the whole @xmath5 by setting @xmath276 as @xmath49 , and we have ( 2 ) . we already know from theorem [ thm : a0]-(ii ) that @xmath277 is a nonempty set for @xmath209 . now , we assume that @xmath278 , and let @xmath279 . then by lemma [ lem : commutator ] , @xmath280 for some @xmath281 . we say that the set @xmath282 contains spin - orbit coupling induced states @xmath283 . this is because @xmath282 is nonempty only for nonzero spin - orbit coupling @xmath27 , in agreement with theorem [ thm : a0 ] . here , our main goal is to establish @xmath282 . for that reason we prove that : [ lem : ess ] let @xmath47 and @xmath61 be as in ( [ eq : a ] ) and ( [ eq : b ] ) . then : 1 . the resolvent of @xmath47 is given by + @xmath284 + ( @xmath285 ) , @xmath286 ( @xmath287 ) ; 2 . @xmath288 , where @xmath289 is equal to @xmath290/4 $ ] for @xmath291 , and to @xmath292 for @xmath293 . \(i ) the proof is pretty much similar to that of ( [ eq : fp ] ) and theorem [ thm : a0]-(i ) . the integral kernel @xmath294 ( for simplicity , we replace @xmath214 by @xmath109 ) is defined through the formal differential equation @xmath295 then @xmath296 with @xmath297 as in ( [ eq : fp ] ) and @xmath298 . as one would have noticed , @xmath299 is the fourier transform of @xmath300 . for more convenience , we rewrite the denominator by @xmath301 , with @xmath302 ( @xmath239 ) as in lemma [ lem : ess]-(i ) . without loss of generality , we assume that @xmath303 ( @xmath239 ) . then the integration over @xmath143 can be performed in two distinct ways . consider @xmath304 and integrate it around the contour @xmath305 oriented counterclockwise , with the poles @xmath306 , @xmath307 . this implies that the integral exists for @xmath131 . similarly , integrate @xmath308 around the contour @xmath309 oriented counterclockwise but with the poles @xmath310 , @xmath311 , and get @xmath135 for the existence of the integral . [ we note that these two contours of integration are not unique . one can choose , for example , the contour with poles @xmath306 , @xmath311 ( @xmath312 ) so that the integral exists for @xmath131 , and the contour with poles @xmath310 , @xmath307 ( again , @xmath312 ) so that the integral exists for @xmath135 . ] by the residue theorem , @xmath313 where the contour integration is performed over @xmath314 ( @xmath315 $ ] ) in the first contour , and over @xmath314 ( @xmath316 $ ] ) in the second contour . in the limit @xmath317 , function @xmath318 for @xmath131 in the first integral , and for @xmath135 in the second one . the residues are easy to calculate by noting that @xmath319 after some elementary simplifications , and replacing @xmath109 with @xmath214 , we obtain the integral kernel of the resolvent of @xmath320 as in lemma [ lem : ess]-(i ) . following ( [ eq : green2 ] ) , @xmath321 by using this equation , calculate @xmath322 and get the equation for @xmath323 , @xmath324\psi(\gamma;z)= 2(p_{1}+p_{2}){\text{\large $ \otimes$}}i.\ ] ] substitute obtained expression of @xmath323 in ( [ eq : gr ] ) , replace @xmath109 with @xmath214 and get the resolvent of @xmath47 as required . that @xmath230 , the arguments are those as in the proof of theorem [ thm : a0]-(i ) . ) for the spin - orbit - coupling strength @xmath204 ( in @xmath8 units ) , for a range of raman couplings @xmath201 as @xmath13 increases ( @xmath325 ) , the two dressed spin states @xcite are merged into a single minimum @xmath292 at @xmath326 . this is a regime when the spin - orbit coupling induced states @xmath282 , theorem [ thm : ea ] , are observed below the continuous spectrum as well as above it . for @xmath327 , the spin states have two minima @xmath328/4 $ ] at @xmath329 , and the spin - orbit induced states are embedded into the essential spectrum of @xmath47 . ] \(ii ) the essential spectrum of @xmath47 as well as the spectrum of @xmath320 is found from ( [ eq : green2 ] ) by solving @xmath330 ( @xmath143 ) with respect to @xmath331 , whereas for @xmath61 , one needs to solve the same equation due to ( [ eq : fp ] ) . the solutions read @xmath332 the lower bound of @xmath333 is found by differentiating @xmath334 with respect to @xmath143 . one finds three critical points : @xmath335 , @xmath336 and @xmath337 . as seen , @xmath307 and @xmath338 are in @xmath5 only for @xmath327 . hence it holds @xmath339/4 $ ] . if , however , @xmath325 , only @xmath306 is valid . then @xmath340 . this proves that @xmath341 , hence ( ii ) , and the proof of the statement is accomplished . for the illustrative and comparison purposes ( see ( * ? ? ? 1b ) and ( * ? ? 2c ) ) , we displayed the dispersion relation @xmath334 , ( [ eq : dispersion ] ) , in fig . [ fig : dispersion ] . we are now in a position to establish the properties of spin - orbit coupling induced states . [ thm : ea ] given @xmath47 as in ( [ eq : a ] ) and @xmath65 as in ( [ eq : a0 ] ) . then : 1 . @xmath342 ; 2 . @xmath343 ; 3 . @xmath344 ; 4 . @xmath345 ; 5 . @xmath346 ; 6 . @xmath347 for @xmath348 ; 7 . @xmath349 for @xmath350 ; 8 . the equivalence classes of functions from the kernel @xmath352 , for @xmath353 , are of the form given in theorem [ thm : a0]-(iii ) . the eigenfunctions that correspond to @xmath354 are as in theorem [ thm : a0]-(ii ) . the proof is essentially based on the combination of theorem [ thm : a0 ] with lemmas [ lem : commutator][lem : ess ] . \(i ) in agreement with lemma [ lem : commutator]-(1 ) , and in particular ( [ eq : raa0 ] ) , substitute @xmath355 ( refer to theorem [ thm : a0]-(ii ) ) in @xmath356 for some @xmath357 . then @xmath358 ( @xmath359 as in theorem [ thm : a0 ] ) , where the upper sign corresponds to @xmath131 , and the lower one to @xmath135 . it appears from above that for either @xmath192 or @xmath193 , the following holds , @xmath360 the solution @xmath283 satisfying the above system of equations is given by @xmath361 or explicitly , @xmath362 . in order to accomplish the proof of ( i ) , it remains to establish valid eigenvalues @xmath210 from @xmath194 thus generating proper eigenvalues @xmath283 from @xmath282 . by a straightforward inspection , @xmath363 for all @xmath66 , where @xmath364 is as in lemma [ lem : ess]-(ii ) . the lower bound is obtained at @xmath365 ( the solution to @xmath366 ) . on the other hand , @xmath367 and @xmath368 at @xmath369 ( @xmath370 is improper due to theorem [ thm : a0]-(ii ) ) . therefore , the points @xmath365 and @xmath371 , which hold whenever @xmath350 , must be excluded as the resonant states , by theorem [ thm : a0]-(i ) ( inspect solutions to @xmath372 with respect to @xmath373 given by @xmath205 ) and by lemma [ lem : ess]-(i ) ( inspect solutions to @xmath374 with respect to @xmath373 given by @xmath364 , and solutions to @xmath375 , @xmath239 , given by @xmath205 ) . item ( i ) holds . ( ii)(v ) the reason for extracting @xmath282 into subsets is in different behavior of the involved eigenvalues : @xmath376 and @xmath377 . this is easy to verify by considering @xmath283 and @xmath289 : for @xmath348 , one finds that @xmath378 , which is @xmath379 . for @xmath350 , @xmath380 for @xmath381 , thus yielding @xmath382 , and @xmath378 for @xmath383 , thus yielding @xmath384 . the values @xmath385 are excluded due to the previous discussion ( these are resonant states ) . \(vi ) since @xmath386 for @xmath348 , we have that @xmath387 in this regime . but @xmath388 , and hence ( vi ) holds . \(vii ) for @xmath350 , @xmath389 . in the present regime we have that @xmath390 with @xmath391 . this gives ( vii ) . \(viii ) following lemma [ lem : commutator]-(2 ) , we need to show that ( weak ) solutions in @xmath352 yield eigenvalues @xmath392 . by theorem [ thm : a0]-(iii ) , [ eq:12 ] @xmath393 where we have explored the integral @xmath394 for @xmath395 ( recall @xmath230 in theorem [ thm : a0]-(i ) and lemma [ lem : ess]-(i ) ) . but @xmath396 and hence the combination of ( [ eq:12 ] ) yields @xmath397 equation ( [ eq:3 ] ) has solutions with respect to @xmath357 only if either @xmath192 or @xmath193 ( recall theorem [ thm : a0 ] ) . then it holds @xmath398 , where the upper sign is for @xmath192 , and the lower one for @xmath193 . recalling that @xmath399 , we recover @xmath282 . this accomplishes the proof of the theorem . the points in @xmath400 are illustrated in fig . [ fig : eigenvaluesea ] . ( refer to theorem [ thm : ea ] ) for the point - interaction strength @xmath203 and the spin - orbit - coupling strength @xmath204 ( in @xmath8 units ) . in figure , red line shows the border @xmath401 of the essential spectrum of @xmath47 ( lemma [ lem : ess ] ) . the eigenvalues @xmath402 ( @xmath209 ) , as functions of the raman coupling @xmath208 , are drawn by the blue ( @xmath403 ) and green ( @xmath382 ) lines . resonant states of @xmath47 are drawn by yellow curves ( r ) . ] as yet , we have established the part of @xmath67 which is associated with discontinuous eigenfunctions at @xmath16 . these states originate from the property that @xmath47 commutes with @xmath65 , where @xmath32 ( @xmath27 ) is unitarily equivalent to the one - dimensional dirac operator for the particle in fermi pseudopotential . in this section , our main goal is to determine the remaining part of @xmath67 , namely , @xmath404 , thus recovering all discrete states of the spin - orbit hamiltonian , and to show that the associated eigenfunctions are continuous in the whole @xmath5 . [ thm : a ] let @xmath47 and @xmath61 be as in ( [ eq : a ] ) and ( [ eq : b ] ) , respectively . then : 1 . @xmath405 + where @xmath282 is given in theorem [ thm : ea ] , the @xmath302 ( @xmath239 ) and @xmath364 are as in lemma [ lem : ess ] , with @xmath406 , @xmath407 , @xmath408 ; 2 . the equivalence classes of functions from @xmath69 ( with @xmath409 ) are of the form @xmath410 ( with @xmath411 ) , with the integral kernel , for @xmath408 , as in lemma [ lem : ess]-(i ) ; 3 . the ( strict ) solutions @xmath71 associated with @xmath72 from @xmath404 are of the form : 1 . for @xmath412 with the upper sign , + [ eq : aaaa ] @xmath413\quad(x>0 ) , \label{eq : aaaa - a } \\ = & c\left[\frac{e^{-ip_{1}x}}{p_{1}}\left(\begin{matrix}\lambda+\omega/2-p_{1}^{2 } \\ -i\eta p_{1}\end{matrix}\right)-\frac{e^{-ip_{2}x}}{p_{2}}\left(\begin{matrix}\lambda+\omega/2-p_{2}^{2 } \\ -i\eta p_{2}\end{matrix}\right ) \right]\quad(x<0 ) \label{eq : aaaa - b}\end{aligned}\ ] ] + for any @xmath414 , @xmath27 ; 2 . for @xmath412 with the lower sign , + [ eq : aaaa1 ] @xmath415 \nonumber \\ & ( x>0 ) , \label{eq : aaaa1-a } \\ = & c\left[-\frac{e^{-ip_{1}x}}{\lambda+\omega/2-p_{1}^{2}}\left(\begin{matrix}\lambda+\omega/2-p_{1}^{2 } \\ -i\eta p_{1}\end{matrix}\right)+\frac{e^{-ip_{2}x}}{\lambda+\omega/2-p_{2}^{2}}\left(\begin{matrix}\lambda+\omega/2-p_{2}^{2 } \\ -i\eta p_{2}\end{matrix}\right ) \right ] \nonumber \\ & ( x<0 ) \label{eq : aaaa1-b}\end{aligned}\ ] ] + for any @xmath414 , @xmath27 ; 3 . for @xmath118 , we have that the discrete spectrum is given by the union @xmath416 ; the associated eigenfunctions are @xmath417 , with @xmath418 ( @xmath419 ) ; 4 . there are no eigenvalues from @xmath404 embedded into the essential spectrum of @xmath47 : @xmath420 . \(1 ) as is seen from the theorem , the eigenfunctions of @xmath47 and @xmath61 , which correspond to the upper sign for @xmath72 in @xmath404 , coincide if and only if [ eq : unique ] @xmath421 ( @xmath422 ) . the eigenfunctions of @xmath47 and @xmath61 , which correspond to the lower sign for @xmath72 in @xmath404 , coincide if and only if @xmath423 ( @xmath424 ) . therefore , equations ( [ eq : unique ] ) provide unique solutions ( up to the constant @xmath425 ) for functions @xmath426 ( @xmath239 ) which are undetermined in @xmath69 ; see theorem [ thm : a]-(2 ) . \(2 ) it is interesting to compare the eigenfunctions at @xmath16 ( having the meaning as in ( [ eq : v - delta-1 ] ) ) , which correspond to the spin - orbit coupling induced states ( theorem [ thm : ea ] ) , with those given above . for @xmath427 with the upper sign , @xmath428 yields @xmath192 ; in comparison , @xmath429 for @xmath430 with the upper sign . hence in both cases , the < < total > > lower component @xmath192 . similarly , there is also another case but with the upper component @xmath193 . \(3 ) as in theorem [ thm : ea ] , the eigenvalues @xmath72 in @xmath404 can be written in an explicit form by solving the cubic equation . we chose not to do that , but displayed @xmath72 graphically instead ; see fig . [ fig : eigenvaluesa ] . associated with everywhere continuous eigenfunctions . the point - interaction strength @xmath203 and the spin - orbit - coupling strength @xmath204 ( in @xmath8 units ) . in figure , red line shows the border @xmath401 of the essential spectrum of @xmath47 ( lemma [ lem : ess ] ) . the blue @xmath431 ( green @xmath432 ) line , showing the bound state as a function of the raman coupling @xmath201 , corresponds to the eigenfunction with a zero - valued lower ( upper ) component at the origin @xmath16 ( theorem [ thm : a ] ) . the eigenvalue @xmath431 approaches @xmath433 at @xmath434 and then disappears ( for details , refer to remark [ rem : fig ] ) . resonant states of @xmath47 are drawn by the yellow curve ( r ) . ] first off , we note that , for @xmath279 , @xmath435 due to lemma [ lem : ess]-(i ) . next , combining ( [ eq : fp ] ) with lemma [ lem : ess]-(i ) we immediately infer ( see also the proof of lemma [ lem : ess]-(i ) and in particular ( [ eq : green2 ] ) ) item ( 2 ) of the theorem . but then , it holds @xmath436 . by solving @xmath437 , we recover @xmath438 ( @xmath351 by theorem [ thm : ea]-(viii ) ) . in order to accomplish the proof of ( 1 ) , it therefore remains to establish @xmath71 ( @xmath430 ) thus proving that items ( 3a)(3b ) yield @xmath439 , which in turn is found by computing the poles of @xmath265 in lemma [ lem : ess]-(i ) . we solve the characteristic equation for @xmath440 ; see ( [ eq : kss2 ] ) . then @xmath441 where @xmath442 ( @xmath443 ) . the condition @xmath444 ( @xmath239 ) is due to @xmath445 ( recall ( [ eq : a ] ) ) . the boundary condition in @xmath50 , provided @xmath107 , yields @xmath446 we now substitute obtained functions @xmath35 in @xmath440 and find that @xmath447 we need to solve the system of equations ( [ eq : f1])([eq : f2 ] ) . in particular , one finds from ( [ eq : f2 ] ) , @xmath448 where @xmath449 and @xmath450 , @xmath451 , @xmath452 for all @xmath453 . hence @xmath454 for @xmath455 . for example , let @xmath456 , @xmath457 . from the first and third equations in ( [ eq : f2 ] ) one gets that @xmath458 and similarly for the remaining @xmath459 . by ( [ eq : c34])([eq : yjs ] ) , there are @xmath460 possible solutions with respect to @xmath461 and @xmath462 for @xmath453 . these are tabulated in tab . [ tab:1 ] . the number of distributions in tab . [ tab:1 ] must be reduced with the help of ( [ eq : f1 ] ) . by ( [ eq : f1 ] ) , one can express @xmath463 in terms of @xmath464 ( @xmath453 ) . namely , .all possible solutions of ( [ eq : f2 ] ) with respect to @xmath465 for @xmath461 , @xmath452 for @xmath453 given in ( [ eq : c34])([eq : yjs ] ) . [ cols=">,^,^,^,^,^,^,^,^,>,>,>,>",options="header " , ] [ tab:1 ] [ eq : further ] @xmath466 by ( [ eq : c34 ] ) , substitute @xmath467 , @xmath468 and @xmath469 in the first equation of ( [ eq : further - b ] ) and get @xmath470 now multiply the first equation of ( [ eq : further - a ] ) by @xmath471 and subtract both obtained equations so that @xmath472 is eliminated , [ eq : c12y ] @xmath473 similarly , by using ( [ eq : c34 ] ) , substitute @xmath474 , @xmath468 and @xmath469 in the second equation of ( [ eq : further - b ] ) and get @xmath475 multiply the second equation of ( [ eq : further - a ] ) by @xmath476 and subtract both obtained equations so that @xmath477 is eliminated , @xmath478 by using ( [ eq : yjs ] ) , equations ( [ eq : c12y ] ) can be rewritten explicitly as follows @xmath479 by noting that @xmath480 and @xmath481 are two independent constants , we can subtract both equations and separate the expressions at @xmath480 and @xmath481 one from another . then @xmath482 where @xmath483+b_{4}k_{1}(2k_{1}-\gamma ) [ \omega^{2}-(2\eta k_{2})^{2}]^{\frac{1}{2 } } \\ & + k_{2}[b_{3}(k_{1}+k_{2}-\gamma)-b_{1}(3k_{1}+k_{2}-2\gamma)][\omega^{2}-(2\eta k_{1})^{2}]^{\frac{1}{2}},\end{aligned}\ ] ] with a one - to - one map @xmath484 , @xmath485 , @xmath486 , @xmath487 , @xmath488 , and @xmath489 . then @xmath490 ( identity ) for @xmath491 , and @xmath492 for @xmath493 equation @xmath494 holds for the distributions ( tab . [ tab:1 ] ) numbered by @xmath495 , @xmath496 , @xmath497 , @xmath498 and @xmath499 , @xmath500 , @xmath501 , @xmath502 . on the other hand , @xmath503 with @xmath208 is well defined for @xmath495 , @xmath497 and @xmath500 , @xmath502 . therefore , we deduce that for @xmath201 , @xmath503 makes sense if @xmath495 , @xmath497 and @xmath500 , @xmath502 . expression @xmath503 can be represented by the sum of @xmath504 and @xmath505 , where both @xmath504 and @xmath505 are invariant under the action of @xmath506 , namely , @xmath507,\quad \varphi f_{\omega}(k_{1},k_{2})=f_{\omega}(k_{1},k_{2}),\ ] ] and @xmath505 is defined by @xmath508^{\frac{1}{2 } } \\ & + k_{2}[b_{3}(k_{1}+k_{2}-\gamma)-b_{1}(3k_{1}+k_{2}-2\gamma)][\omega^{2}-(2\eta k_{1})^{2}]^{\frac{1}{2}}.\end{aligned}\ ] ] then @xmath505 satisfies @xmath509 and @xmath510 then @xmath511 yields @xmath512 \nonumber \\ & \times [ \omega^{2}-(2\eta k_{1})^{2}]^{\frac{1}{2}}-k_{1}[b_{2}(3k_{2}+k_{1}-2\gamma ) \nonumber \\ & + b_{4}(k_{1}-k_{2})][\omega^{2}-(2\eta k_{2})^{2}]^{\frac{1}{2}}=0 . \label{eq : solvek2}\end{aligned}\ ] ] equation ( [ eq : solvek2 ] ) shows that , depending on @xmath513 distributions in tab . [ tab:1 ] , four distinct classes can be considered . [ eq : distr4 ] @xmath514^{\frac{1}{2 } } -b_{2}k_{1}[\omega^{2}-(2\eta k_{2})^{2}]^{\frac{1}{2}}\bigr ) \label{eq : distr4-a } \\ \intertext{($b_{1}=b_{3}$ , $ b_{2}=b_{4}$ ) , } & ( ii):\:e_{\omega}^{(2)}(k_{1},k_{2})\equiv 0,\;\text{with}\;\ ; e_{\omega}^{(2)}(k_{1},k_{2 } ) = \gamma\omega(k_{1}-k_{2 } ) \nonumber \\ & + b_{1}k_{2}(k_{1}+k_{2}-\gamma)[\omega^{2}-(2\eta k_{1})^{2}]^{\frac{1}{2 } } -b_{2}k_{1}(2k_{2}-\gamma)[\omega^{2}-(2\eta k_{2})^{2}]^{\frac{1}{2 } } \label{eq : distr4-b } \\ \intertext{($b_{1}=b_{3}$ , $ b_{2}=-b_{4}$ ) , } & ( iii):\:e_{\omega}^{(3)}(k_{1},k_{2})\equiv 0,\;\text{with}\;\ ; e_{\omega}^{(3)}(k_{1},k_{2})=-\varphi_{1}e_{\omega}^{(2)}(k_{1},k_{2 } ) \label{eq : distr4-c } \\ \intertext{($b_{1}=-b_{3}$ , $ b_{2}=b_{4}$ and $ \varphi_{1}{\!\colon\thinspace}k_{1}\mapsto k_{2}$ , $ k_{2}\mapsto k_{1}$ , $ b_{1}\mapsto b_{2}$ , $ b_{2}\mapsto b_{1}$ ) , } & ( iv):\:e_{\omega}^{(4)}(k_{1},k_{2})\equiv 0,\;\text{with}\;\ ; e_{\omega}^{(4)}(k_{1},k_{2 } ) = \gamma\omega(k_{1}-k_{2 } ) \nonumber \\ & + b_{1}k_{2}(2k_{1}-\gamma)[\omega^{2}-(2\eta k_{1})^{2}]^{\frac{1}{2 } } -b_{2}k_{1}(2k_{2}-\gamma)[\omega^{2}-(2\eta k_{2})^{2}]^{\frac{1}{2 } } \label{eq : distr4-d}\end{aligned}\ ] ] ( @xmath515 , @xmath516 ) . by the isomorphism in ( [ eq : distr4-c ] ) , it suffices to consider three classes : @xmath517 , @xmath518 , @xmath519 . class @xmath517 . given @xmath208 , the equation @xmath520 ( [ eq : distr4-a ] ) holds for the distributions numbered by @xmath521 , @xmath497 , @xmath500 and @xmath513 . if , however , @xmath455 , then @xmath522 holds for all @xmath523 , @xmath524 , which is inconsistent with the point spectrum of @xmath47 . subsequently , class @xmath517 is improper . class @xmath518 . for @xmath208 , @xmath525 ( [ eq : distr4-b ] ) holds for the distributions numbered by @xmath495 , @xmath526 , @xmath527 and @xmath502 . due to the isomorphism @xmath528 , the number of distributions decreases to @xmath495 , @xmath529 , @xmath526 , @xmath498 , @xmath499 , @xmath527 , @xmath530 and @xmath502 . but @xmath531 yields @xmath532 which is satisfied only for @xmath533 , hence improper due to @xmath435 . class @xmath519 . for @xmath208 , @xmath534 ( [ eq : distr4-d ] ) holds for the distributions numbered by @xmath535 , @xmath536 , @xmath537 and @xmath501 . for @xmath455 , @xmath538 yields a correct relation @xmath539 . possible distributions are numbered by @xmath540 and @xmath541 . as a result , we have found that @xmath534 is the only one correct equation which holds for all @xmath201 . the associated distributions in tab . [ tab:1 ] are numbered by @xmath540 and @xmath537 . by solving ( [ eq : distr4-d ] ) , we find that @xmath542 where @xmath543^{\frac{1}{2}}\bigr ) } { 2[(2\eta k_{1}k_{2})^{2}+(\gamma\omega)^{2 } ] } \label{eq : maineq2}\ ] ] ( @xmath544 ) , @xmath545 and @xmath546 . as it should be by ( [ eq : solvek2 ] ) , equation ( [ eq : maineq ] ) is invariant under the action of @xmath506 as well as @xmath528 . recalling that @xmath547^{1/2}$ ] ( @xmath548 ) , one can construct the equation for the eigenvalues @xmath72 . by ( [ eq : maineq ] ) , @xmath72 satisfies the following cubic equation @xmath549\lambda^{2 } \nonumber \\ & \pm 8\omega[2\omega^{2}+(\gamma^{2}+2\eta^{2})(\eta^{2}\pm\omega)]\lambda+ \omega^{2}[4\eta^{4}+(\gamma^{2}\pm 2\omega)^{2}]=0 \label{eq : lambdaobt}\end{aligned}\ ] ] ( @xmath201 ) , provided @xmath444 for @xmath239 . note that the sign @xmath550 corresponds to that in ( [ eq : maineq2 ] ) . now , it is necessary to show that the eigenvalues @xmath72 , which satisfy ( [ eq : lambdaobt ] ) , are also in @xmath438 , thus accomplishing the proof of theorem [ thm : a]-(1 ) , and that the eigenfunctions of @xmath404 are as in ( [ eq : aaaa])([eq : aaaa1 ] ) , thus giving theorem [ thm : a]-(3a ) and ( 3b ) . we solve ( [ eq : f2 ] ) with respect to @xmath468 , @xmath469 and @xmath467 , @xmath474 , by assuming that @xmath27 , @xmath551 we note that each equality in every row can be chosen arbitrarily ; we choose the first one . substitute obtained expressions in ( [ eq : fxstart ] ) and find by ( [ eq : f1 ] ) , @xmath552 these functions , with @xmath107 , are in @xmath50 . hence the boundary condition given by @xmath553 yields [ eq : b12 ] @xmath554 by ( [ eq : b12 ] ) , four possible cases are then considered , provided @xmath27 : case ( 1 ) . @xmath555 and @xmath556 . by ( [ eq : fxstart ] ) , hence @xmath558 . by ( [ eq : f1 ] ) , @xmath559 ( see also ( [ eq : further - a ] ) ) . hence @xmath560 . if @xmath561 , then @xmath562 , hence trivial . if @xmath563 , then @xmath564 , by ( [ eq : kssp ] ) , and @xmath562 , by ( [ eq : fxstart ] ) ; hence improper again . case ( 2 ) . @xmath565 if we expand the latter equation by using ( [ eq : kssp ] ) , this agrees with ( [ eq : lambdaobt ] ) for the upper sign . by noting that @xmath566 for @xmath239 , and @xmath302 as in the theorem , we find that the correspondence is one - to - one with the eigenvalues in @xmath438 obtained by setting the lower sign . by ( [ eq : fxstart ] ) , @xmath557 , and thus @xmath567 . then ( [ eq : further - a ] ) yields @xmath568 and @xmath569 . the substitution of these coefficients in ( [ eq : fxstart ] ) gives ( [ eq : aaaa1 ] ) , with @xmath566 ( @xmath239 ) , @xmath570 . case ( 3 ) . @xmath571 similarly to the previous case , by expanding the former equation with the help of ( [ eq : kssp ] ) , we establish ( [ eq : lambdaobt ] ) with the lower sign . subsequently , this corresponds to the upper sign in @xmath438 . the latter equation , @xmath556 , along with ( [ eq : fxstart ] ) yields @xmath572 substitute obtained coefficients in ( [ eq : fxstart ] ) and get ( [ eq : aaaa ] ) , with @xmath566 ( @xmath239 ) and the coefficient @xmath573 ( note that the denominator is nonzero unless @xmath72 is in the essential spectrum ) . case ( 4 ) . @xmath574 the combination of both equations yields @xmath575 . if @xmath576 , then , recalling that ( refer to ( [ eq : kssp ] ) ) @xmath577 ( @xmath548 ) , it holds @xmath578 , hence improper . if , however , @xmath579 , then @xmath564 , by ( [ eq : kssp ] ) , hence improper again . as a result , cases ( 2)(3 ) accomplish the proof of items ( 1 ) and ( 3a)(3b ) of theorem [ thm : a ] . we now concentrate on ( 3c ) . for @xmath118 , equation @xmath440 , @xmath445 , is easy to deal with since the components @xmath101 and @xmath102 are separated and thus can be solved independently one from another : @xmath580 , @xmath581 . by substituting obtained exponents in the boundary condition we get ( 3c ) . moreover , the condition @xmath582 is obtained from the inspection of the resolvent in lemma [ lem : ess]-(i ) , where one requires @xmath303 for @xmath239 . for @xmath118 , @xmath583 , and hence @xmath584 thus yielding @xmath582 . otherwise , only one eigenvalue @xmath585 remains . in particular , this also proves that @xmath420 ( see item ( 4 ) of the theorem ) for @xmath118 , since @xmath586 . for arbitrary spin - orbit coupling @xmath27 , let us examine the conditions @xmath303 for @xmath239 . it suffices to show the converse for at least one @xmath302 . let @xmath456 and @xmath348 . then @xmath386 . assume that the eigenvalue @xmath587 for some real @xmath588 . then it holds @xmath589 . but @xmath590 for all @xmath348 . hence @xmath591 , which is invalid . let @xmath456 and @xmath350 . then @xmath389 . let @xmath592 for some @xmath588 . then we have that @xmath593 , where @xmath594 for all @xmath350 . as seen , @xmath591 for all @xmath595 . next , let @xmath596 for some @xmath597 , and substitute @xmath598 in ( [ eq : lambdaobt ] ) . one gets that @xmath599 for the upper one . it is evident that the above equations do not have real solutions for all @xmath597 for all @xmath66 ( @xmath546 ) , since all the terms on the right - hand side are positive , whereas the left - hand side is zero . therefore , @xmath591 for @xmath600 as well . subsequently , item ( 4 ) holds , and this accomplishes the proof of the theorem . [ rem : fig ] in fig . [ fig : eigenvaluesa ] , one finds that @xmath431 vanishes for @xmath601 , by substituting @xmath578 in @xmath404 ( theorem [ thm : a]-(1 ) ) or in ( [ eq : lambdaobt ] ) and solving the obtained equation with respect to @xmath13 . the suffix < < + > > indicates that the eigenvalue is found from @xmath404 with the plus sign ( or from ( [ eq : lambdaobt ] ) with the minus sign ) . we also note that the condition @xmath602 alone is insufficient to derive proper bound states ; this must be implemented with the requirement @xmath303 for @xmath239 as well . in this paper , we solved the bound state problem for the spin - orbit coupled ultracold atom in a one - dimensional short - range potential describing the impurity scattering . the potential is assumed to be approximated by the @xmath42-interaction . as a result , two distinct realizations of the original differential expression , @xmath29 , were proposed . the first one , @xmath47 , is implemented through the boundary condition defining the domain of the operator . the second realization , @xmath61 , has a meaning of distribution . although both representatives provide identical spectra , the eigenfunctions differ in their form : equivalence classes of functions of @xmath61 supply with insufficient information concerning the ( classical ) behavior of eigenfunctions . based on the property that @xmath29 contains both the spin - orbit and the raman coupling , we showed that , for nonzero spin - orbit and raman coupling , the spectrum is implemented with some extra states , in addition to those which are found by solving the eigenvalue equation directly . extra states , called the spin - orbit coupling induced states , have a peculiarity that the associated eigenfunctions are discontinuous at the origin @xmath16 , and that there might be a point embedded into the essential spectrum . by ( dis)continuity we assume that , although functions are defined on any subset of @xmath37 , their left ( @xmath603 ) and right ( @xmath604 ) representatives either coincide ( continuity ) or not ( discontinuity ) . such states originate from the fact that the spin - orbit hamiltonian is not purely dirac - like or schrdinger - like operator but rather their one - dimensional mixture . it turns out that @xmath47 ( @xmath61 ) commutes with the operator which is unitarily equivalent to the one - dimensional dirac operator ( in weyl s form ) for the particle with spin one - half moving in the fermi pseudopotential @xmath73 . in turn , we showed that @xmath73 is a combination of both @xmath42- and @xmath163-interactions , where the latter accounts for the divergent terms occurring if dealt with discontinuous functions ( one has the so - called @xmath164-interaction ) . finally , we established the remaining part of the discrete spectrum of @xmath47 ( @xmath61 ) and showed that the eigenvalues under consideration are found by solving the cubic equation . depending on the regime of the raman coupling , that is to say , on the strength of the zeeman field , one observes either two or a single point in the spectrum . the associated eigenfunctions are everywhere continuous but with zero - valued component ( either upper or lower one ) at the origin . it is worth noting that the ( self - adjoint ) representatives @xmath65 and @xmath47 of the atom - light coupling @xmath6 and the hamiltonian @xmath29 could serve for a tool to recover other self - adjoint extensions thus corresponding to modified point - interactions . this could be done with the help of krein s formula ( * ? ? ? * eq . ( 6.10 ) ) ( see also ( * ? ? ? * appendix a ) ) . for that purpose one needs to apply the resolvents of @xmath65 and @xmath47 given in theorem [ thm : a0]-(i ) and lemma [ lem : ess]-(i ) , respectively . following eg @xcite , one constructs operators on the intervals @xmath605 and @xmath606 , and finds the orthonormal bases relevant to deficiency subspaces . so defined , the operators have d.i . . the entries of the associated unitary matrix from @xmath607 group thus determine all self - adjoint extensions . the authors gratefully acknowledge dr . g. juzelinas who bears much of the credit for the genesis of the present paper . it is a pleasure to thank dr . i. spielman for useful discussions . r.j . acknowledges dr . i. spielman for warm hospitality extended to him during his visit at the university of maryland , where the part of this work has been done . the present work was supported by the research council of lithuania ( no . vp1 - 3.1-mm-01-v-02 - 004 ) .	we solve the bound state problem for the hamiltonian with the spin - orbit and the raman coupling included . the hamiltonian is perturbed by a one - dimensional short - range potential @xmath0 which describes the impurity scattering . in addition to the bound states obtained by considering weak solutions through the fourier transform or by solving the eigenvalue equation on a suitable domain directly , it is shown that ordinary point - interaction representations of @xmath0 lead to spin - orbit induced extra states . * pacs(2010 ) : * 03.65.ge , 67.85.-d , 71.70.ej
the layered transition metal oxide @xmath3 combines high thermopower and metallic conductivity , making it a promising candidate for thermoelectric energy conversion @xcite . the discovery of superconductivity in hydrated compounds @xcite and other types of ordered electronic groundstates for higher sodium content @xcite have intensified the interest and raised questions related to fermi surface ( fs ) instabilities . itinerant electrons moving on the @xmath4 triangular lattice and the ability to control the filling of the narrow co - o derived bands through the sodium content @xmath0 render it an appealing model system in the field of correlated electron physics . it is especially interesting to address the low energy electronic excitation density as it is both a measure of correlation effects and an indicator of collective groundstates . in order to explore the magnetic and thermodynamic properties of the electronic system in this context in detail , we studied the heat capacity of a large number of both polycrystalline and single crystalline samples and performed muon spin rotation ( @xmath1 ) measurements . our data , combined with previous results , provide a clear picture across the whole sodium concentration range : in addition to the charge - ordered insulator at @xmath5 we observe a strong enhancement of the linear specific heat coefficient @xmath6 for @xmath7 which might be taken as evidence for either further enhanced correlation effects or for additional bands crossing @xmath8 . collective magnetic groundstates of spin - density wave ( sdw ) type with various ordering temperatures are observed in the high @xmath0 range accompanied by pronounced fermi surface gapping . additional evidence for the loss of dos stems from the low - temperature ( low-@xmath9 ) nuclear heat capacity . the magnetic instabilites are foreshadowed by spin fluctuations without magnetic long - range order at slightly lower @xmath0 , which manifest themselves in an even more strongly enhanced and field - dependent @xmath6 at lowest temperatures ( @xmath10 ) . polycrystals were prepared employing the conventional synthesis method involving either slow heating or the rapid heat - up technique @xcite . a stoichiometric mixture of high - purity @xmath11 nanopowder ( aldrich , @xmath12 ) and waterfree @xmath13 ( aldrich , @xmath12 ) was pressed into pellets and annealed overnight at @xmath14 in air . the samples were confirmed to be a single phase of hexagonal by x - ray powder diffraction . the na content of the polycrystals was determined from the unit - cell parameters , in particular by the @xmath15 ratio . as the @xmath15 vs. @xmath0 curve becomes flat for @xmath16 , this method of sodium content determination is inaccurate in this range . thus , @xmath0 values higher than @xmath17 have to be considered with caution and are in fact most likely not representative of the true na content . this will be taken into account in the interpretation of the data . single crystals were grown applying a standard floating zone technique : first , powder was prepared by direct reaction of @xmath11 and @xmath13 ( the same starting materials as mentioned above ) in an alumina boat under oxygen flow at @xmath18 . rapid heating of the furnace ( @xmath19 ) to the desired sintering temperature reduced na evaporation . after @xmath20 of sintering the resulting polycrystalline samples were ground under argon atmosphere and heated again for @xmath21 under oxygen flow at @xmath22 . the feeding material for single crystal growth were pressed pellets sintered at @xmath22 for @xmath23 . using a four - mirror optical floating - zone furnace ( crystal systems corp . ) we grew single crystals under flow of a mixture of argon and oxygen at @xmath24 . a more detailed preparation procedure can be found in the literature @xcite . single crystals were analyzed using single crystal x - ray diffractometry and von - laue backscattering which generally confirmed the samples to be single - crystalline , the mosaic spread being @xmath25 . some samples showed twinning , but in all cases the c - axis and the ab - planes were well - defined . the na content of the crystal batches was determined by inductively coupled plasma atomic emission spectroscopy ( icp - aes ) . the study comprised more than 30 samples from different batches . the specific heat was measured in a quantum design ppms-14 and some measurements made use of the ppms he-3 option . addenda measurements and field - dependent thermometer calibration ensured reliability of the @xmath26 measurements . @xmath1 measurements were performed at the @xmath27 beamline at the paul - scherrer - institut ( psi ) at villigen , switzerland . this beamline provides @xmath28 polarized muons . in the following we contrast the linear specific heat coefficient @xmath6 , which is a measure for the low energy electronic excitation density , with @xmath6 as calculated from results of numerous previous angle - resolved photoemission spectroscopy ( arpes ) studies . only in a few reports the value for the density of states @xmath6 is given explicitly , but it can readily be determined from the fermi velocity and wave vector assuming a 2d fermi surface topology . deviations from an exact 2d nature of the electronic structure have been discussed for higher @xmath0 as e.g. a 3d magnetic structure was observed @xcite or tight - binding calculations yielded a significant @xmath29-dispersion @xcite . other theoretical studies challenge these findings based on the assumption of strong correlations or an afm groundstate @xcite . to calculate a @xmath6 value from the fermi surface observed in arpes we follow the approach of a quasi-2d electronic structure which is straightforward because of the quasi-2d fs topology observed in almost all arpes studies and which has previously been established to be applicable for @xcite : @xmath30 thus we can calculate @xmath6 corresponding to the measured fermi velocity @xmath31 and the fermi wave vector @xmath32 . in some measurements @xcite an anisotropy of the fermi velocity by factor 2 was noted . we take this into account by stating the maximum and minimum dos that would result from both extremes in these cases ( those values are linked by broken lines in the phase diagram , fig . [ fig : phasediag ] ) . as will be discussed later , @xmath6 as obtained from arpes is significantly lower than the thermodynamically determined @xmath6 which can be interpreted as evidence for the existence of fs parts and/or a pronounced three - dimensionality of the fs not seen in arpes . typical specific heat data for a slowly cooled single crystal with two transitions into a long - range ordered magnetic state ( at @xmath33 and @xmath34 ) . ( a ) the electronic density of states at intermediate temperatures @xmath35 ( above any fermi surface instabilities ) is determined by a linear fit to the @xmath36 vs. @xmath37 data . the residual low-@xmath9 dos ( @xmath38 ) is determined from a linear extrapolation of the low-@xmath9 data down to @xmath39 ( panel ( b ) ) . , scaledwidth=45.0% ] in this study we distinguish between the intermediate temperature dos @xmath35 and the low temperature dos @xmath38 , both measured via the specific heat . the first characterizes the dos at t above any magnetic order or spin fluctuation effects ( @xmath40 ) and is obtained either through a fit of the @xmath26 data using the standard debye law @xmath41 for @xmath42 or a model comprising an einstein and a debye contribution @xmath43 to take into account the two prominent contributions to the phonon spectrum for @xmath44 @xcite . we have analyzed the data of several samples using the two models and found excellent agreement with both . the resulting value for @xmath35 is essentially unaffected by the choice of the phonon model . thus , the debye law fit is sufficient for the present purpose and it was therefore generally applied . @xmath38 is a measure for the excitation density in a @xmath9 range where magnetic order and spin fluctuations contribute to @xmath26 , and @xmath38 is obtained through extrapolation of the low-@xmath9 @xmath26 data down to @xmath39 . [ fig : cphighlow ] shows an example of both extrapolation procedures for a single crystal ( low temperature details in the inset ) . muon spin rotation ( @xmath1 ) data taken on poly- and single crystals contrasting the low - t magnetic state for a sample with spin fluctuations and two with magnetic order . ( a ) the absence of any oscillatory contribution in the positron decay asymmetry corresponds to the absence of static magnetic lro for a @xmath45 polycrystal . ( b ) in the @xmath46 polycrystal , the entire sample volume is magnetically ordered ( within experimental limits of @xmath47 ) . ( c ) the same is true for the @xmath17 single crystal . note that the saturation towards @xmath48 of the initial asymmetry value in panel ( b ) is due to the powder average but _ not _ due to only partial magnetic order ( statistically , @xmath48 of the internal fields have the same orientation as the muon spin polarisation in the polycrystals and thus do not force a muon precession).,scaledwidth=45.0% ] the results reveal several distinct regions in the phase diagram which will be discussed separately in the following . the central observation in this region of the phase diagram is the absence of any magnetic long - range order ( lro ) , which can be deduced from the heat capacity data @xcite in accordance with previous studies @xcite . here we also show zero field muon - spin rotation spectra taken on @xmath50 and @xmath45 polycrystals . these @xmath1 data to do not show any lro down to @xmath51 , which would result in an oscillatory part in the positron emission asymmetry ( fig . [ fig : usrmarkus1](a ) ) . this is consistent with @xmath1 results reported for similar compositions @xcite . for a comprehensive description of the @xmath1 technique applied to see e.g. sugiyama et al . @xcite . while @xmath1 experiments reveal important information on the static magnetism they can not elucidate on the possible presence of spin fluctuations with a time scale shorter than the order of @xmath52 . on the other hand , from specific heat we get strong indications of low - energy magnetic excitations as will be discussed in the following . we note the absence of any sign of a phase transition in @xmath26 for @xmath53 which confirms the @xmath1 results . however , the low temperature @xmath26 trend shows some peculiarity : @xmath26 distinctly increases below @xmath54 , not inconsistent with a @xmath55 bahaviour , indicating an increasing excitation density at lowest temperatures ( fig . [ fig : markus1](a ) ) . the contribution of the nuclei to the heat capacity has been subtracted for clarity ( the nuclear heat capacity is discussed in section iii.b.2 ) . dependent on @xmath0 , values as large as @xmath56 are reached for @xmath57 , consistent with other studies @xcite . in an external magnetic field , these additional excitations are significantly suppressed ( fig . [ fig : markus1](b ) ) . upon application of a @xmath58 external field , @xmath59 decreases to @xmath60 . in the @xmath61 range accessible with our magnet there is an indication of a saturation of @xmath38 towards the value of @xmath35 , which is @xmath62 for @xmath53 . here ( and also for higher x ) , @xmath35 is significantly higher than in the @xmath63 range . this remarkable high dos will be discussed in section iv . obviously , compared to @xmath35 a strong renormalization of the electronic excitation density is present only at a low excitation energy scale ( @xmath38 ) which can be suppressed by magnetic fields . this behaviour is consistent with a magnetic origin of the apparent mass enhancement for @xmath64 which is likely to be spin fluctuations that arise due to the proximity to magnetic ordering . this conclusion is in good agreement with nmr @xcite and neutron diffraction data @xcite that provide evidence for ferromagnetic in - plane fluctuations being present in with @xmath65 . this conclusion has also been invoked before based on heat capacity studies @xcite , however without the evidence from field dependent measurements as those shown here . additional support comes from theoretical studies yielding the onset of fm fluctuations for higher @xmath0 assisted by charge disproportionation @xcite or band structure effects @xcite . specific heat measured on polycrystals highlighting the two distinct magnetic regions in the phase diagram . ( a ) the low - t dos ( @xmath10 ) is enhanced due to spin fluctuations for @xmath66 . ( b ) suppression of @xmath67 in an external magnetic field of up to @xmath58 . ( c ) the onset of magnetic ordering for @xmath68 leads to a drastic reduction of the residual dos ( @xmath57 ) reflecting fermi surface gapping . the nuclear contribution to the specific heat has been subtracted ( see section iii.b.2 and fig . 4 for a discussion).,scaledwidth=45.0% ] for @xmath70 , develops a magnetically ordered groundstate as can be seen in the heat capacity and magnetization . for @xmath71 it affects the whole sample volume as shown by @xmath1 data for polycrystals ( fig . [ fig : usrmarkus1](b ) ) and single crystals ( fig . [ fig : usrmarkus1](c ) ) . for @xmath72 , only @xmath73 magnetically ordered volume fraction has been reported @xcite which was interpreted as being the result of a phase separation into one magnetically ordered and one non - ordered phase @xcite to explain the large non - magnetic volume fraction . in single crystals the actual phase separation and thus coexistance of two distinct magnetic phases at low temperatures has been directly observed recently : in addition to the well - known @xmath33 afm transition @xcite there is another transition at @xmath34 associated with a distinct second phase whose formation is dependent on the sample s cooling protocol @xcite and can be completely suppressed by thermal quenching . this points to sodium order as a driving force of magnetic order due to the patterned sodium coulomb potential breaking the crystal symmetry and highlights the importance to account for multiple phases when interpreting low - t data . in polycrystalline samples ( in which na ordering effects are less pronounced in general ) the @xmath33 sdw magnetic ordering can be observed and only a broad hump at lower temperatures suggests the @xmath34 magnetic ordering transition ( fig . [ fig : markus1](c ) and ref.@xcite ) . even as the details of the development of magnetic lro are somewhat different in single crystals , the @xmath57 behaviour is fully consistent with the polycrystals . heat capacity measurements on @xmath74 polycrystalline samples show a systematic variation with @xmath0 ( fig . [ fig : markus1](c ) ) : @xmath38 decreases monotonically with @xmath0 while the intensity of the discontinuity at @xmath33 grows . interestingly , within the experimental uncertainty of a few percent , @xmath38 stays constant in fields up to @xmath58 , in contrast to the @xmath53 range . the peak position of the discontinuities associated with the magnetic ordering transitions changes slightly with the applied magnetic field : the @xmath34 transition ( only present in single crystals ) moves up , the @xmath33 transition moves down , thus indicating a different nature of the magnetic state at @xmath33 and @xmath34 @xcite . comparing @xmath35 and @xmath38 we note that @xmath35 decreases slightly with @xmath0 while @xmath38 falls dramatically from the highly renormalized value at @xmath72 and drops below the low-@xmath0 value of @xmath35 of around @xmath75 at approximately @xmath76 ( fig . [ fig : phasediag ] ) . the reduction of @xmath77 from @xmath78 above the magnetic ordering to a value as low as @xmath79 for @xmath38 reflects a substantial fermi surface gapping ( @xmath80 ) , as expected for a sdw - like instability affecting a significant fraction of the fermi surface(s ) . thus our results are compatible with an at least partial fermi surface gapping due to the onset of magnetic order in the high @xmath0 range . nuclear specific heat analysis according to eqns . [ eq : cplowt ] and [ eq : cplowta ] . the main result is the reduction of the nuclear specific heat contribution ( reflected in the coefficient a ) in samples where ( partial or complete ) fermi surface gapping is observed . the correspondance between the nuclear specific heat coefficient a and the dos @xmath6 is established in panel ( d ) with the comparison of the zero - field low - t @xmath6 obtained from @xmath26 with a for several samples with different @xmath0 . the dashed red line is a guide to the eye . panel ( c ) shows the suppression of a for @xmath5 and @xmath68 compared to the expected value following eqn . [ eq : cplowta ] ( dashed red line ) . we consistently interpret this effect as fermi surface gapping which reduces the thermal coupling of the co and na nuclei to the heat pulse used to probe the specific heat . the arrow accounts for the uncertainty of the sodium content for the sample with the highest @xmath0 , which is likely to be sligthly smaller than the nominal @xmath0 ( see section ii ) . panels ( a ) and ( b ) are auxiliary panels showing the experimental determination of a. panel ( a ) is a @xmath36 vs. @xmath81 plot where @xmath82 can be extracted as the slope . plotting @xmath82 vs. @xmath83 yields a as the slope ( panel ( b ) ) . , scaledwidth=45.0% ] the reduction of dos expressing itself in the small value of @xmath38 as discussed above is consistently reflected in the low - temperature nuclear specific heat . the measured specific heat below @xmath84 is described by @xmath85 where the first term is the electronic heat capacity , the second term accounts for the nuclear heat capacity and the third and almost negligible term proportional to @xmath86 is an unknown contribution which might be attributed to an electric quadrupole moment effect stemming from some localized co @xmath87 electrons . the quality of the data lends itself to a quantitative analysis because all microscopic terms entering are known with the only free parameters being a and @xmath88 . the nuclear heat capacity originating from the @xmath89 na and @xmath90 co nuclei comprises the coefficient @xmath91 , \label{eq : cplowta}\ ] ] with the g - factors @xmath92 and @xmath93 , respectively . the lattice heat capacity is negligible at such low temperatures . [ eq : cplowt ] yields a linear relationship of @xmath36 vs. @xmath94 ( fig . [ fig : nuccp](a ) ) from which @xmath38 is determined as the ordinate intercept and @xmath82 as the slope . plotting the slope vs. @xmath95 ( fig . [ fig : nuccp](b ) ) gives a and its variation with the sodium content @xmath0 is shown in fig . [ fig : nuccp](c ) . the variation of a as a function of @xmath0 is noteworthy for two reasons . for @xmath96 , i.e. in the range without long - range order ( be it magnetic , or charge order at @xmath5 where @xmath97 ) , the value of a is very close to the expected value calculated from eqn . [ eq : cplowta ] ( dashed red line in fig . [ fig : nuccp](c ) ) . remarkebly , a is much depressed or even unmeasurably small ( for @xmath5 ) for these compositions where the fermi surface is known from the previous analysis to be partly or fully gapped due to instabilities . a natural interpretation of these data takes into account that the nuclear spins obviously do not couple to the short heat pulse applied to the crystal in the @xmath26 measurement . apparently , this coupling involves the dos at @xmath8 . indeed , the values of a systematically vary with @xmath98 ( fig . [ fig : nuccp](d ) ) . this fact is also generally reflected in the much higher spin - lattice relaxation times of insulating materials compared to metals . [ fig : nuccp ] therefore provides independent additional self - consistent evidence for the fs gapping observed in the sdw state in addition to the well - known gapping for the charge - ordered insulator at @xmath5 . the data just discussed are presented in an extended electronic excitation phase diagram ( fig . [ fig : phasediag ] ) . a highly consistent picture emerges when we combine our data with pertinent previously reported data . the open red and blue symbols represent the data from specific heat studies of bayrakci et al . ( bay04,@xcite ) , sales et al . ( sal04,@xcite ) , sakurai et al . ( sak04,@xcite ) , wooldridge et al . ( woo05,@xcite ) , yokoi et al . ( yok05,@xcite ) and yoshizumi et al . ( yos07,@xcite ) . the open triangles stand for the arpes data of yang et al . @xcite , hasan et al . @xcite , qian et al . @xcite and geck et al . @xcite while the broken line marks the approximate values of the dos as obtained by lda calculations @xcite . in the high @xmath0 range , the two regions discussed above are clearly identified by distinctly different behaviour in terms of thermodynamic and magnetic properties . inspecting the phase diagram , two aspects are worth noting : 1 . for @xmath7 the intermediate - temperature dos @xmath35 obtained at a low energy scale via the specific heat ( red arch in fig . [ fig : phasediag ] ) increases by a factor 2 .. 3 compared to the @xmath63 range and stays enhanced even for higher @xmath0 while both the lda calculated @xmath6 and the arpes - measured high - energy dos amount to around @xmath99 over the whole @xmath0 range ( dashed line and triangles in fig . [ fig : phasediag ] ) . 2 . the excitation density at lowest temperatures @xmath38 ( blue arch in fig . [ fig : phasediag ] ) is dominated by spin fluctuations for @xmath53 which lead to an additional enhancement compared to @xmath35 . however , due to magnetic order for @xmath68 the low-@xmath9 dos is strongly reduced by the onset of fermi surface gapping associated with the magnetic instabilities . the two distinct magnetic regions in the phase diagram , with spin fluctuations and with magnetic order , are separated by a transition region . the smooth transition between @xmath100 may well reflect the coexistence of more than one phase , as was first invoked by sakurai et al . based on magnetization and heat capacity studies @xcite . in the following it will be discussed in the light of extensive structural studies by huang et al.@xcite . structural transitions - particularly associated with the complex interplay between the temperature - dependent na mobility and the inherent propensity in to na ordering - were intensely studied . there is evidence @xcite that only some well - defined compositions result in stable crystallographic phases that differ in the na positioning in the unit cell and the occupation of the inequivalent na sites . in between those particular @xmath0 values , phase separation into distinct phases with different na content occurs @xcite . this effect has been invoked in the high-@xmath0 range also to explain the presence of paramagnetic volume in otherwise magnetically ordered samples @xcite . recently a particular na rearrangement process was shown to lead to distinct magnetically ordered states associated with different phases in the same crystal and the connection to na diffusion as its origin was established @xcite . in this context the distinct but gradual change observed in @xmath35 and @xmath38 for @xmath100 which marks the transition between the two adjacent magnetic ranges could well be the result of the coexistence of one high-@xmath0 phase hosting magnetic order and another phase with lower @xmath0 associated with the spin fluctuations . interestingly , in the studies by huang et al . @xcite a coexistance between the so called h1 and h2 phases is observed in this narrow @xmath0 range . thus it would be reasonable to associate the spin fluctuation phase with h1 and the sdw phase with h2 . it would be worthwile , however , to clarify in detail to what degree structural parameters , other than the ones distinguishing h1 , h2 and h3 , need to be included in order to capture the subtle differences of the various phases . as shown in the phase diagram above , in addition to the spin fluctuation contribution to @xmath6 that vanishes in high magnetic fields , there is another enhancement for @xmath7 that is not suppressed by magnetic fields and is also detectible at intermediate temperatures ( @xmath40 ) . in the following paragraph we concentrate our discussion on @xmath35 as it is not affected by spin fluctuations or magnetic order . we note that @xmath35 is enhanced by a factor of 2 .. 3 compared to the expected values from band structure calculations ( around @xmath102)@xcite and the arpes experimental data ( @xmath103 ) . in addition to our findings , two other extensive studies @xcite support our results . obviously the electronic dos is significantly higher for @xmath7 which is tentatively ascribed to an additional band crossing the fermi surface following yoshizumi et al . @xcite . this could most likely be the @xmath104 character band which has been calculated to develop a dip near the @xmath105 point for higher @xmath0 @xcite . this would result in the emergence of an electron pocket when @xmath8 touches the band from below upon filling of the co - o states with increasing x. consistent with our findings , the onset of small-@xmath106 magnetic fluctuations is predicted to accompany the changes of the fermi surface topology @xcite . a second explanation would involve the @xmath107 character hole pockets that were intensely sought for in numerous arpes studies . in a rigid band picture these should rather disappear with increasing fermi level which has also been predicted by numerous theoretical studies @xcite and recently been claimed to be observed experimentally @xcite . even if the hole pockets were stabilized also for higher @xmath0 as predicted by some models @xcite , the dos behaviour observed here could hardly be explained . further , another possibility would be ascribing the observed enhancement for higher @xmath0 purely to the onset of strong electron - electron correlations in which case a significant impact on the band structure was anticipated by theoretical studies @xcite leading to a band picture being more consistent with the arpes results . however , no origin for such a dramatic increase of correlations around @xmath108 could be identified so far rendering this option rather speculative . it is interesting to note that @xmath6 calculated from arpes - measured band dispersions involving one major fermi surface agrees well with the lda - calculated @xmath6 across the whole phase diagram although there are severe discrepancies between the arpes findings and the calculated dispersions : the bandwidth observed in the arpes studies is systematically reduced compared to bandstructure calculations , being @xmath109 instead of @xmath110 . this leads to the conclusion of the hopping integral @xmath111 being renormalized by a factor @xmath112 already at higher temperatures which should be reflected in a dramatically enhanced dos compared to the calculations . however , the absence of the @xmath107 character hole pockets or other additional fermi surface segments seem to exactly compensate this effect in the arpes measurements . thus it appears that the arpes results so far do not account for the highly enhanced dos values for @xmath7 . even though it is commonly noted that there must be strong correlations present to explain the small bandwidth , in combination with the experimentally observed fermi surface topology this does not suffice to explain the high @xmath6 values . interestingly , there is a discrepancy concerning the dimensionality of the electronic system in the high @xmath0 range among the arpes studies . one study sees a significant @xmath29-dispersion resulting in additional fs segments @xcite while in most cases the system seems to remain quasi-2d @xcite . if additional fermi surface caused by three - dimensionality was present , a higher dos is consequently to be expected . to decide which of these scenarios - or possibly others - realistically describe the electronic states in is an intriguing puzzle and awaits clarification by further studies . in the light of the recently observed coupling between the fermi surface instabilities and the degree of freedom imposed by sodium mobility and the plethora of possible ordering patterns @xcite , it is obvious that future theoretical studies should reach beyond the virtual crystal approximation and realistically model the interplay between na driven coulomb potential patterns , fermi surface topology & dimensionality and charge / magnetic order . the interpretation of our heat capacity and muon - spin rotation data provides a consistent picture of the low energy electronic excitation density for both intermediate and low temperatures over a large sodium concentration range and for both single and polycrystals . the comparison with previous data as compiled in fig . [ fig : phasediag ] provides a comprehensive summary of the current and previously published data and reflects the agreement among the results from various studies . some of the conclusions arrived at in the present study have been suggested before , such as the coexistence of a spin fluctuation and a magnetically ordered phase for @xmath113 @xcite . new insight , however , is gained from our in - depth analysis of both @xmath35 and @xmath38 : our field - dependent heat capacity measurements in quantitative comparison with arpes and lda results reveal a rich and complex picture which renders spin fluctuations to be only one out of two coexisting low energy excitation density enhancement mechanisms for @xmath7 . for the whole @xmath0 range the _ high energy _ band dispersion as measured by arpes yields a density of states consistent with lda band structure calculations ( dos around @xmath114 ) . however , the _ low energy _ excitation spectrum ( @xmath115 ) as measured by specific heat reveals distinctly richer phenomena : for @xmath7 the intermediate-@xmath9 dos is 2 ... 3 times higher ( @xmath116 ) . in the range @xmath66 the low - temperature excitation density is dominated by spin fluctuations that can be suppressed in a magnetic field and cause even stronger renormalization of @xmath6 as compared to the intermediate - temperature range , but no magnetic long - range order is present . a variety of magnetically ordered states is then observed for @xmath68 , including a @xmath34 ferromagnetic - type transition in addition to the previously studied sdw - like @xmath33 transition . in the light of the low dos derived from arpes and lda calculations and the existence of a second dos enhancement mechanism seen in the excitation density phase diagram ( be it a band structure or renormalization effect ) , the question whether fermi surface gapping caused by magnetic ordering is present in the high-@xmath0 range is not trivially answered by a loss of low-@xmath9 dos as compared to intermediate-@xmath9 dos . here , our nuclear heat capacity analysis provides important additional evidence based upon which it can be concluded that the magnetic order involving the itinerant electrons indeed causes the disappearance of part of the fermi surface . thus , the @xmath117 range of the phase diagram is highlighted to be most interesting by the presence of a high dos and fermi surface loss through magnetic instabilities . the revisited phase diagram illustrates the rich physics of the correlated electrons moving on the triangular lattice in the presence of a controllable potential landscape defined by ordering patterns formed by highly mobile na ions .	in this study an extended low energy phase diagram for is experimentally established with emphasis on the high @xmath0 range . it is based on systematic heat capacity studies on both polycrystalline and single crystalline samples and on @xmath1 measurements . main features are the existence of mass enhancement , spin fluctuations without long - range order , and magnetic order with associated fermi surface gapping . the latter is seen in the electronic density of states ( dos ) and suppression of nuclear specific heat . while there is agreement between the band structure and the low energy dos in the low @xmath0 range , in the high @xmath0 range ( @xmath2 ) the thermodynamically determined dos is approximately three times that deduced from the angle - resolved photoemission spectroscopy ( arpes)-measured band dispersion or local - density approximation ( lda ) calculations . strae 5 , d-12489 berlin , germany ]
it is well - known from general theory on semigroups that the infinitesimal generator of any lvy process always exists for all functions in the schwartz space @xmath3 . from the celebrated lvy - khintchine formula , we can derive that the infinitesimal generator is given by an operator of the general form @xmath4\,\dd{{\mathfrak m}}(y ) . \end{split}\ ] ] the first term corresponds to the diffusion , the second to the drift , and the third to the jump part . in this paper , we shall focus on the operators which are obtained in purely jump processes , i.e. processes without diffusion or drift part . in particular , we are mainly concerned with the operators with the general form @xmath5\,\dd{{\mathfrak m}}(y),\,\,t>0,\ ] ] where @xmath6 is a positive measure satisfying @xmath7 where @xmath8 . the value of @xmath9 is well - defined when @xmath10 is bounded on @xmath11 and @xmath12 at @xmath13 . these concepts shall be defined more precisely later . the operator @xmath14 described above is called a _ linear integro - differential operator_. the operator was introduced with too much generality . so we shall restrict our attention to the operators @xmath14 where the measure @xmath6 is given by a positive kernel @xmath15 which is not necessarily symmetric . that is to say , the operators @xmath14 are given by @xmath16k(y)\,dy,\,t>0.\ ] ] then we see that @xmath9 is well - defined provided that @xmath17 ( refer to definition [ def-2.3 ] for the definition of @xmath18 ) where @xmath19 denotes _ the family of all real - valued bounded functions defined on @xmath11_. if @xmath15 is symmetric , then an odd function @xmath20k(y)$ ] will be canceled in the integral , and so we have that @xmath21k(y)\,dy,\,t>0.\ ] ] on the other hand , if @xmath15 is not symmetric , the effect of @xmath22k(y)$ ] persists and we can actually observe that the influence of this gradient term becomes stronger as we try to get an estimate in smaller regions . the aim of this work is to obtain regularity results for fully nonlinear integro - differential equations with possible nonsymmetric kernels . this kind of equations are often obtained in stochastic control problems @xcite . if a player in a stochastic game is permitted to choose different strategies at every step in order to maximize the expected value of some function at the first exit point of a domain , then a convex nonlinear operator @xmath23 appears in the game . in a competitive game with two or more players , more complicated operators of the form @xmath24 can be obtained . the different aspect between and is convexity . also an operator like @xmath25 can be considered . characteristic properties of these operators can easily be derived as follows ; @xmath26(x)-{{\mathcal i}}^{\pm } u(x)\le\sup_{t\ge 1/2}\sup_{{\alpha}{\beta}}{{\mathcal l}}^t_{{\alpha}{\beta } } v(x),\end{split}\ ] ] where @xmath27 are given by @xmath28 and @xmath29 ( here we use @xmath30 on @xmath31 instead of @xmath32 with certain technical reasons which can be found in the proof of [ thm-7.9 ] below ) . a more general and better description of the fully nonlinear operators we want to deal with is the operator @xmath33 for which holds for some family of linear integro - differential operators @xmath34 . the idea is that the concept of ellipticity can be replaced by an estimate @xmath35(x)-{{\mathcal j}}u(x)\le{{\mathcal m}}^+_{{\lambda},{\lambda}}v(x)$ ] , where @xmath36 and @xmath37 are the pucci extremal operators @xcite . then it is easy to see that @xmath33 must be an elliptic second order differentiable operator . if instead we compare with suitable nonlocal extremal operators , we will have a concept of ellipticity for nonlocal operators @xmath33 . we shall give a precise definition in section [ sec-3 ] ( see definition [ def-3.1 ] ) . we now explain the natural dirichlet problem for such nonlocal operators @xmath38 . let @xmath39 be an open domain in @xmath11 . given a function @xmath40 defined on @xmath41 , we want to find a function @xmath10 such that @xmath42 note that the boundary condition is given not only on @xmath43 but also on the whole complement of @xmath39 . this is because of the nonlocal character of the operators @xmath38 . from the stochastic point of view , it corresponds to the fact that a discontinuous lvy process can exit the domain @xmath39 for the first time jumping to any point in @xmath41 . in this paper , we shall concentrate mainly upon the regularity properties of viscosity solutions to an equation @xmath44 . we shall briefly give a very general comparison principle from which existence of the solutions can be obtained in smooth domains . since kernels of integro - differential operators are comparable to the kernel of the fractional laplace operator @xmath45 , the theory we want to develop can be understood as a theory of viscosity solutions for fully nonlinear operators of fractional order . most of all , let us emphasize the main differences between local and nonlocal equations in terms of harnack estimate and hlder continuity . @xmath46 first , the nonnegativity of the solution of a local uniformly elliptic equation in a ball @xmath47 is enough to get a harnack inequality in a smaller ball @xmath48 . for the nonlocal equation , there is a counterexample @xcite that is nonnegative in @xmath47 but that has zero value in @xmath48 . it is due to the fact that there are influence for the values outside @xmath47 , which should be controlled to have a harnack inequality . so we impose that the solution is nonnegative in @xmath11 . @xmath46 second , by applying harnack inequalities on @xmath49 and @xmath50 , we are able to show the oscillation lemma and hlder regularity of the solution of a local equation . however such method can not be directly applicable due to the fact that we need nonnegativity of the solution in @xmath11 , not only @xmath47 . @xmath46 the last interesting fact is that there is a discontinuous solution if the kernel is allowed to oscillate between two different exponents @xmath51 and @xmath52 , @xcite . the authors show that there is still a harnack estimate if the radius of the ball has a positive lower bound , for example one . there are some known results about harnack inequalities and hlder estimates for integro - differential operators with positive symmetric kernels ( see @xcite for analytical proofs and @xcite , @xcite , @xcite,@xcite , @xcite , @xcite for probabilistic proofs ) . the estimates in all these previous results blow up as the index @xmath53 of the operator approaches @xmath54 . in this respect , they do not generalize to elliptic partial differential equations . however there is some known result on regularity results for fully nonlinear integro - differential equations associated with nonlinear integro - differential operators with positive symmetric kernels which remain uniform as the index @xmath53 of the operator approaches @xmath54 ( see @xcite ) . therefore these results make the theory of integro - differential operators and elliptic differential operators become somewhat unified . there has been serious consideration on the concept of viscosity solution of fully nonlinear integro - differential equations and their properties , @xcite . in this paper , we extend the important regularity results of caffarelli and silvestre @xcite on positive symmetric kernels with certain decay to those on certain positive ( not necessarily symmetric ) kernels including such symmetric kernels which remain uniform as the index @xmath53 of the operator approaches @xmath54 . in this occasion we can not expect any cancelation on the estimates contrary to the case of the symmetric kernel , which stirs up some difficulties in this problem . throughout this paper we would like to briefly present the necessary definitions and then prove some regularity estimates . our results in this paper are the following . @xmath46 we introduce more general new class of operators to consider nonsymmetric case . it is invariant under translation and scaling , which are crucial properties used at @xcite . still we are able to show standard porperties for viscosity solutions of the general nonlinear integro - differential equations . @xmath46 we show a new version of the nonlocal alexandroff - backelman - pucci estimate for fully nonlinear integro - differential equations with possible nonsymmetric kernel . it contains the extra term caused by the nonsymmetry of kernel . @xmath46 we also give new proof for the construction of special functions to handle the nonsymmetry of the kernel . and then we show a harnack inequality , hlder regularity , and an interior @xmath55-regularity result for certain fully nonlinear integro - differential equations . nonsymmetric case developes the following differences from the symmetric case and nontrivial difficulties . key observations are the following : @xmath46 for the nonsymmetric case , @xmath56 and @xmath57 can be chosen any of @xmath58 or @xmath59 . therefore there could be an extra term @xmath60 . @xmath46 the equation is not scaling invariant due to @xmath61 . @xmath46 somehow the equation has a drift term , not only the diffusion term . the case @xmath62 and the case @xmath63 require different technique due to the difference of the blow rate as @xmath64 approaches to zero and the decay rate as @xmath64 approaches to infinity . when @xmath62 , a controllable decay rate of kernel allows hlder regularity in a larger class , which is invariant under an one - sided scaling i.e. if @xmath10 is a solution of the homogeneous equation , then so is @xmath65 for @xmath66 . critical case ( @xmath67 ) and supercritical case ( @xmath68 ) have been studied in @xcite with different techniques due to the slow decay rate of the kernel as @xmath69 . it is noticeable that a gradient term has be considered as a lower order term ( @xmath70 ) , @xcite , while our gradient effect comes from the diffusion term which is a main order term . the outline of the paper is as follows . in section [ sec-2 ] , the appropriate definitions of subsolutions and supersolutions of fully nonlinear integro - differential equation in the viscosity sense shall be given . in the definitions , we shall allow some kind of discontinuities outside of the domain of the equation . in section [ sec-3 ] , we introduce more general concept of fully nonlinear integro - differential equation which is invariant under one - sided scaling . we define a nonlocal elliptic operator by comparing its increments with a suitable maximal operator . this definition is more general than . in section [ sec-4 ] , we study the stability properties of viscosity solutions given in the definition . a comparison principle is proven in section [ sec-5 ] under very mild assumptions . we guess that one of the most nontrivial results in the paper is a nonlocal version of the alexandroff - backelman - pucci estimate to be shown in section [ sec-6 ] . it has an extra term caused by the nonsymmetry of the kernel . it leads to regularity results for certain fully nonlinear integro - differential equations . in section [ sec-7 ] , we construct a special function , considering nonsymmetry of kernel , and obtain some pointwise estimates which shall be helpful in proving hlder estimates in section [ sec-9 ] . in section [ sec-8 ] , we prove a harnack inequality which plays an important role in analysis . we obtain the hlder estimates in section [ sec-9 ] . finally we show an interior @xmath55-estimates in section [ sec-10 ] . for our purpose , we shall restrict our attention to the operators @xmath14 where the measure @xmath6 is given by a positive kernel @xmath15 which is not necessarily symmetric . that is to say , the operators @xmath14 are given by @xmath71k(y)\,dy,\,\,t>0.\ ] ] then we see that @xmath9 is well - defined provided that @xmath72 ( refer to definition [ def-2.3 ] for the definition of @xmath18 ) where @xmath19 denotes _ the family of all real - valued bounded functions defined on @xmath11_. to simplify the notation , we write @xmath73 then the expression for @xmath14 may shortly be written as @xmath74 in particular , for @xmath75 , we consider the class @xmath76 of the operators @xmath14 associated with the measures @xmath6 given by positive kernels @xmath77 satisfying that @xmath78 in what follows , our main concern shall be on the nonlinear integro - differential operators which have the form like where we think that each @xmath79 has a kernel @xmath80 satisfying . the minimum assumption so that @xmath27 are well - defined is that every kernel @xmath81 must satisfy the following integrability condition in a uniform way ; more precisely , if we set @xmath82 , then @xmath83 we say that @xmath84 is a _ paraboloid of opening @xmath85 _ if @xmath86 where @xmath85 is a positive constant , @xmath87 is real constant and @xmath88 is a linear function . then @xmath84 is called _ convex _ when we have @xmath89 in and _ concave _ when we have @xmath90 in . let @xmath91 be a bounded domain . given two semicontinuous functions @xmath92 defined on an open subset @xmath93 and a point @xmath94 , we say that _ @xmath95 touches @xmath10 by above at @xmath94 _ if @xmath96 and @xmath97 for any @xmath98 . similarly , we say that _ @xmath95 touches @xmath10 by below at @xmath94 _ if @xmath96 and @xmath99 for any @xmath98 . for a semicontinuous function @xmath10 on @xmath39 and an open subset @xmath100 of @xmath39 , we define @xmath101 to be the infimum of all positive constants @xmath85 for which there is a convex paraboloid of opening @xmath85 that touches @xmath10 by above at @xmath94 . also we define @xmath102 if no such constant @xmath85 exists . similarly , we define @xmath103 to be the infimum of all positive constants @xmath85 for which there is a concave paraboloid of opening @xmath85 that touches @xmath10 by below at @xmath94 , and also we define @xmath104 if no such constant @xmath85 exists . finally we set @xmath105 for these definitions , the readers can refer to @xcite . [ def-2.1 ] let @xmath91 be a bounded domain and let @xmath106 be a semicontinuous function . given @xmath107 , we say that @xmath10 is @xmath108 at @xmath109 @xmath110 resp . @xmath12 at @xmath109 @xmath111 if @xmath112 @xmath110 resp . @xmath111 for some open neighborhood @xmath100 of @xmath109 and we write @xmath114 $ ] if @xmath115 . given a fixed @xmath116 , we set @xmath117 for @xmath118 and we write @xmath119 $ ] if @xmath120<{\infty}$ ] . [ rem-2.2 ] 1 . we note that if @xmath114 $ ] then @xmath10 is differentiable at @xmath109 because @xmath10 lies between two tangent paraboloids in an open neighborhood of @xmath109 . 2 . in fact , the number @xmath121 $ ] in definition [ def-2.1 ] depends upon @xmath122 . but existence of @xmath122 so that @xmath121<{\infty}$ ] is enough , because @xmath121 $ ] decreases as @xmath122 approaches @xmath123 . [ def-2.3 ] a function @xmath124 is said to be @xmath125 @xmath110we write @xmath126@xmath111 , if there exist some vector @xmath127 , @xmath128 and @xmath129 such that @xmath130 we write @xmath131 if @xmath132 for any @xmath98 and the constant @xmath85 in is independent of @xmath133 , where @xmath100 is an open subset of @xmath11 . [ rem-2.4 ] 1 . such vector @xmath95 exists uniquely and moreover @xmath134 . if @xmath135 $ ] for @xmath118 , then we easily see that @xmath126 . moreover , it is easy to show that the converse holds . thus we conclude that @xmath136=\rc^{1,1}({\omega})$ ] . for @xmath118 and a function @xmath124 which is semicontinuous on @xmath137 , we say that @xmath138 belongs to the function class @xmath139 ( resp . @xmath140 ) and we write @xmath141 ( resp . @xmath142 ) if there are an open neighborhood @xmath93 of @xmath13 and @xmath143 such that @xmath144 and @xmath145 ( resp . @xmath146 ) on @xmath147 . we note that geometrically @xmath148 having a local maximum at @xmath13 in @xmath39 is equivalent to @xmath141 and @xmath148 having a local minimum at @xmath13 in @xmath39 is equivalent to @xmath142 . for @xmath118 and @xmath149 , we write @xmath150 and the expression for @xmath151 and @xmath152 may be written as @xmath153 where @xmath80 . we set @xmath154 and @xmath155 . we note that if @xmath126 , then @xmath156 , @xmath157 and @xmath158 for @xmath118 and @xmath159 . [ def-2.5 ] let @xmath160 be a function . then a function @xmath124 which is upper @xmath110lower@xmath111 semicontinuous on @xmath137 is said to be a viscosity subsolution ( viscosity supersolution ) of an equation @xmath161 on @xmath39 and we write @xmath162 @xmath110@xmath163@xmath111 on @xmath39 in the viscosity sense , if for any @xmath118 there is some open neighborhood @xmath100 of @xmath13 with @xmath93 such that @xmath164 is well - defined and @xmath165 @xmath110@xmath166@xmath111 for @xmath167 whenever @xmath143 satisfying @xmath144 and @xmath145 @xmath110@xmath146@xmath111 on @xmath147 exists . also a function @xmath10 which is both a viscosity subsolution and a viscosity supersolution of the equation @xmath161 on @xmath39 is said to be a @xmath168 to @xmath161 on @xmath39 . [ rem-2.6 ] definition [ def-2.5 ] is essentially the same as definition 2 in @xcite . instead of test functions @xmath169 given in the above , functions @xmath138 which are @xmath12 only at the contact point @xmath13 could be used . this is a larger set of test functions , so that a priori it may provide a stronger concept of solution . in section [ sec-4 ] , we shall show that the two approaches are actually equivalent . [ thm-2.7 ] let @xmath160 be a function . then we have the followings : @xmath170 if @xmath124 is a function which is upper semicontinuous on @xmath137 , then @xmath171 on @xmath39 in the viscosity sense if and only if @xmath172 is well - defined and @xmath173 @xmath174 if @xmath124 is a function which is lower semicontinuous on @xmath137 , then @xmath175 on @xmath39 in the viscosity sense if and only if @xmath172 is well - defined and @xmath176 @xmath177 if @xmath124 is a function which is continuous on @xmath137 , then @xmath10 is a viscosity solution to @xmath178 on @xmath39 if and only if it satisfies both and . since ( b ) and ( c ) can be obtained by the similar way to the proof of ( a ) , we have only to prove the equivalence ( a ) for @xmath179 ; similarly for @xmath180 . assume that @xmath181 on @xmath39 in the viscosity sense . fix any @xmath118 and take any @xmath141 . then there is an open neighborhood @xmath93 of @xmath13 such that @xmath182 and @xmath145 on @xmath147 . for @xmath183 , we set @xmath184 and @xmath185 . then , for any @xmath186 and any @xmath187 , there is some @xmath188 such that @xmath189 by the lebesgue s dominated convergence theorem , taking @xmath190 we conclude that for any @xmath186 and any @xmath187 there is some @xmath188 such that @xmath191 . therefore @xmath192 is well - defined and @xmath193 . conversely , suppose that holds . let @xmath124 be a function which is upper semicontinuous on @xmath137 . fix any @xmath118 and take a function @xmath143 satisfying that @xmath144 and @xmath145 on @xmath147 where @xmath100 is an open neighborhood @xmath93 of @xmath13 . then if we set @xmath185 , then for any @xmath186 and any @xmath187 there is some @xmath188 such that @xmath191 , and thus @xmath194 . hence we conclude that @xmath181 on @xmath39 in the viscosity sense . in and , we considered the supremum or an inf - sup of a collection of linear integro - differential operators . let us consider a class @xmath195 of linear operators which includes the class @xmath196 given in section [ sec-2 ] . the maximal operator and the minimal operator with respect to @xmath195 are defined by @xmath197 for a function @xmath124 semicontinuous on @xmath137 and @xmath159 , we set @xmath198 where @xmath199 and @xmath200 then we see that @xmath201 , and @xmath202 have the following simple forms ; @xmath203 where @xmath204 , @xmath205 , @xmath206 and @xmath207 are given by @xmath208 we note that if @xmath126 , then @xmath209 . we shall use these maximal and minimal operators to obtain regularity estimates . the factor @xmath210 is important when @xmath211 , because we need such factor if we want to obtain second order partial differential equations as limits of linear integro - differential equations . in terms of the regularity , we need the factor @xmath210 for the estimates not to blow up as @xmath211 . let @xmath212 where @xmath213 s are all the kernels of all operators in @xmath195 . instead of , for any class @xmath195 we shall assume that @xmath214 using the extremal operators , we provide a general definition of ellipticity for nonlocal equations . the following is a kind of operators of which the regularity result shall be obtained in this paper . [ def-3.1 ] let @xmath195 be a class of linear integro - differential operators . assume that holds for @xmath195 . then we say that an operator @xmath33 is @xmath215 , if it satisfies the following properties : @xmath170 @xmath216 is well - defined for any @xmath217\cap \rb({{\mathbb r}}^n)$ ] . @xmath174 @xmath218 is continuous on an open set @xmath91 , whenever @xmath119\cap\rb({{\mathbb r}}^n)$ ] . @xmath177 if @xmath219\cap\rb({{\mathbb r}}^n)$ ] , then we have that @xmath220(x)\le{{\mathcal j}}u(x)-{{\mathcal j}}v(x)\le{{\mathcal m}}^+_{{{\mathfrak l}}}[u - v](x).\ ] ] we shall show that any operator as in is elliptic with respect to any class containing all the operators @xmath34 as long as the condition is satisfied ( see lemma [ lem-3.2 ] and lemma [ lem-4.2 ] ) . [ lem-3.2 ] let @xmath38 be the operators as in so that holds for every @xmath81 and let @xmath195 be any collection of linear integro - differential operators . if @xmath221 where @xmath196 , then we have that @xmath222(x)&\le{{\mathcal i}}^{\pm } u(x)-{{\mathcal i}}^{\pm } v(x)\le{{\mathcal m}}^+_{{{\mathfrak l}}}[u - v](x ) \end{split}\ ] ] for any @xmath223\cap\rb({{\mathbb r}}^n)$ ] . it can be shown in a similar way as in @xcite . definition [ def-2.3 ] is not set up to evaluate the operator @xmath33 in the original function @xmath10 . whenever a smooth function @xmath138 touches @xmath10 from above , we can always construct a test function @xmath224 $ ] to evaluate @xmath33 . it is very interesting that if @xmath33 is any nonlinear operator which is formulated by an inf - sup ( or a sup - inf ) of linear operators satisfying and , then without constructing such a test function @xmath33 can be evaluated classically in @xmath10 at those points @xmath13 where @xmath10 can be touched by above with a @xmath225 function . this interesting fact is proved in the next lemma . [ lem-3.3 ] let @xmath38 be the operator as in so that holds for every @xmath81 and let @xmath226 be a viscosity subsolution to @xmath178 on @xmath39 . if @xmath135 $ ] for a point @xmath118 , then @xmath227 is defined in the classical sense and @xmath228 . it can be obtained in a similar way as in @xcite . in the next theorem , we shall obtain a result on the operators @xmath229 which is similar to theorem [ thm-2.7 ] . the proof is almost the same as that of theorem [ thm-2.7 ] , and so we will just write out the statement without detailed proof . [ thm-3.4 ] let @xmath160 be a function . then we have the followings : @xmath170 if @xmath124 is a function which is upper semicontinuous on @xmath137 , then @xmath10 is a viscosity subsolution to @xmath230 on @xmath39 if and only if @xmath231 is well - defined and @xmath232 @xmath174 if @xmath124 is a function which is lower semicontinuous on @xmath137 , then @xmath10 is a viscosity supersolution to @xmath233 on @xmath39 if and only if @xmath234 is well - defined and @xmath235 in this section , we obtain a few technical properties of the operators @xmath38 as in . first we shall show that if @xmath236\cap\rb({{\mathbb r}}^n)$ ] , then @xmath27 are continuous on @xmath39 . as mentioned in the previous sections , it is necessary to justify that the operators of the form and satisfy the conditions of definition [ def-3.1 ] . next we shall show that our notion of viscosity solutions allows to touch with solutions which are only punctually @xmath12 instead of @xmath225 in a neighborhood of the point . then we shall show the important stability property of viscosity solutions given in definition [ def-2.3 ] . we now start with several technical lemmas . @xcite[lem-4.1 ] let @xmath237 be a family of functions such that @xmath238 for @xmath239 . if @xmath240 , then the family @xmath241 is uniformly equicontinuous on every compact subsets of @xmath11 . when we gave the definition of viscosity solutions in section [ sec-2 ] , we used @xmath225 test functions . now we show that it is equivalent to use punctually @xmath12 functions . [ lem-4.2 ] let @xmath242 be the family of operators as in and satisfying and let @xmath243 be a bounded domain . if @xmath119\cap \rb({{\mathbb r}}^n)$ ] , then @xmath27 are continuous on @xmath39 . fix any @xmath75 . then we have only to prove that the family @xmath244 is equicontinuous on @xmath39 . we set @xmath245 as in . take any @xmath246 and @xmath107 . since @xmath121<{\infty}$ ] , we have that @xmath247\,\|y\|^2 $ ] , whenever @xmath118 and @xmath248 . then choose some sufficiently small @xmath249 so that @xmath250\,\|y\|^2 k(y)\,dy<{\epsilon}/3.\ ] ] now we have that @xmath251 from , we easily obtain that @xmath252\,\|y\|^2 k(y)\,dy<{\epsilon}/3 $ ] for any @xmath253 , whenever @xmath118 . we also write @xmath254\,h_{{\alpha}{\beta}}(y)\,dy\\&{\fallingdotseq}{{\mathcal l}}^{t,2}_{{\alpha}{\beta}}u(x)+{{\mathcal l}}^{t,3}_{{\alpha}{\beta}}u(x)+{{\mathcal l}}^{t,4}_{{\alpha}{\beta}}u(x ) \end{split}\ ] ] where @xmath255 . since @xmath121<{\infty}$ ] and @xmath256 is lipschitz continuous on @xmath39 by standard analysis , we have that @xmath257\,\,\|x - x_0\| \end{split}\ ] ] whenever @xmath258 and @xmath259 . since @xmath260 ( by ) and @xmath119 $ ] , by lemma [ lem-4.1 ] , and there exists some sufficiently small @xmath261 such that @xmath262 for any @xmath253 , and @xmath186 , whenever @xmath118 and @xmath263 . thus it follows from and that @xmath264 for any @xmath253 , and @xmath186 , whenever @xmath118 and @xmath263 . hence this implies that @xmath265 , whenever @xmath118 and @xmath263 . therefore we complete the proof . when we gave the definition of viscosity solutions in section [ sec-2 ] , we used @xmath225 test functions . we show it is equivalent to use punctually @xmath12 functions . [ lem-4.3 ] let @xmath33 be elliptic with respect to a class @xmath195 in the sense of definition [ def-3.1 ] . assume that @xmath124 is a viscosity subsolution to @xmath161 on @xmath39 and @xmath266\cap\rb({{\mathbb r}}^n)$ ] for @xmath118 . if @xmath267 touches globally @xmath10 from above at @xmath13 , then @xmath268 is defined in the classical sense and @xmath269 . it can be shown in a similar way as in @xcite . one of the most useful properties of viscosity solutions is their stability property under uniform limits on compact sets . we shall prove a slightly stronger result that the notion of viscosity subsolution ( supersolution ) is stable with respect to the natural limits for upper ( lower ) semicontinuous functions . this type of limit is well - known and usually called _ @xmath270-limit_. it was originally called as the celebrated half relaxed limit techniques by barles and perthame , but we are going to follow the similar definitions at @xcite . [ def-4.4 ] a sequence @xmath271 of@xmath272 lower semicontinuous functions is said to @xmath273 on a set @xmath91 if @xmath170 for any sequence @xmath274 with @xmath275 , @xmath276 and @xmath174 for any @xmath118 , there is a sequence @xmath274 with @xmath275 such that @xmath277 . _ remark . _ ( a ) a uniformly convergent sequence @xmath271 converges in the @xmath270 sense . \(b ) if @xmath271 @xmath270-converges to @xmath10 on @xmath39 and @xmath10 has a strict local minimum at @xmath13 then there is a sequence @xmath278 with @xmath275 such that each @xmath279 has a local minimum at @xmath280 ( see @xcite ) . \(c ) if @xmath271 @xmath270-converges to @xmath10 on @xmath39 , then @xmath281 @xmath270-converges to @xmath148 on @xmath39 where @xmath282 . \(d ) from ( b ) and ( c ) , we can get that if @xmath271 @xmath270-converges to @xmath10 on @xmath39 and @xmath148 has a strict local minimum at @xmath13 where @xmath282 , then there is a sequence @xmath278 with @xmath275 such that each @xmath283 has a local minimum at @xmath280 . [ lem-4.5 ] let @xmath33 be elliptic in the sense of definition [ def-3.1 ] . if @xmath284 is a sequence of viscosity supersolutions @xmath279 to @xmath285 on @xmath39 such that @xmath170 @xmath271 @xmath270-converges to @xmath10 in @xmath39 , @xmath174 @xmath271 converges to @xmath10 @xmath286 on @xmath11 and @xmath177 @xmath287 converges to @xmath288 locally uniformly on @xmath39 , then @xmath10 is a viscosity supersolution to @xmath161 on @xmath39 . it can be done with minor changes in a similar way as in @xcite . we just obtained the stability property of supersolutions under @xmath270-limits . for the corresponding result for subsolutions as in the following lemma , we would also consider the natural limit in the space of upper semicontinuous functions which is the same as the @xmath270-convergence of @xmath289 to @xmath290 . [ lem-4.6 ] let @xmath33 be elliptic in the sense of definition [ def-3.1 ] . if @xmath284 is a sequence of viscosity subsolutions to @xmath285 on @xmath39 such that @xmath170 @xmath291 @xmath270-converges to @xmath290 in @xmath39 , @xmath174 @xmath271 converges to @xmath10 @xmath286 on @xmath11 and @xmath177 @xmath287 converges to @xmath288 locally uniformly on @xmath39 , then @xmath10 is a viscosity subsolution to @xmath161 on @xmath39 . as a corollary , we also obtain the stability property under uniform limits . [ cor-4.7 ] let @xmath33 be elliptic in the sense of definition [ def-3.1 ] . if @xmath284 is a sequence of viscosity solutions to @xmath285 on @xmath39 such that @xmath292 @xmath271 and @xmath287 converge to @xmath10 and @xmath288 locally uniformly on @xmath39 , respectively , @xmath293 @xmath271 converges to @xmath10 @xmath286 on @xmath11 , then @xmath10 is a viscosity solution to @xmath161 on @xmath39 . since @xmath294 locally uniformly on @xmath39 , we see that @xmath271 @xmath270-converges to @xmath10 in @xmath39 . thus the required result follows from lemma [ lem-4.5 ] and [ lem-4.6 ] . the comparison principle for viscosity solutions can be shown by very standard ideas in nonlinear analysis , which originated from the idea of jensen @xcite using sup - convolutions and inf - convolutions . the method has been succesfully adapted to integro - differential equations @xcite and a more general proof can be found in @xcite in case that the viscosity solutions have an arbitrary growth at infinity . our definitions do not quite fit in with the previous frameworks because we consider mainly the general class of operators given by definition [ def-3.1 ] and we allow discontinuities outside of the domain @xmath39 of the equation . however the similar techniques can be applied to our equations by the stability property on viscosity subsolutions and supersolutions . the key result of this section that is crucial for our regularity theory is theorem [ thm-5.4 ] , because we can apply it to incremental quotients of viscosity solutions to fully nonlinear integro - differential equations to get its @xmath55-estimates in section [ sec-10 ] . in order to obtain a comparison principle for a nonlinear operator @xmath33 , we need to impose a minimal ellipticity condition to our collection @xmath195 of linear operators as follows ( see also @xcite ) . [ ass-5.1 ] there is a constant @xmath295 so that for each @xmath296 and @xmath297 there exists some @xmath298 such that for any @xmath299 we have that @xmath300 where @xmath138 is a function given by @xmath301 assumption [ ass-5.1 ] is enough for the comparison principle . we note that assumption [ ass-5.1 ] is very mild . in fact , we shall show in the next lemma that the class @xmath302 satisfies assumption [ ass-5.1 ] . it just says that , given the particular function @xmath303 , the value of the operator will be strictly positive on @xmath304 but it does not require any uniform estimate on how that happens . assumption [ ass-5.1 ] is pretty mild . indeed , the following lemma can be shown by simple computation . so we just state it without detailed proof . [ lem-5.2 ] if @xmath305 , then @xmath196 satisfies assumption [ ass-5.1 ] . [ thm-5.3 ] let @xmath33 be elliptic with respect to @xmath195 in the sense of definition [ def-3.1 ] where @xmath195 is some class satisfying assumption [ ass-5.1 ] and let @xmath91 be a bounded domain . if @xmath306 is a viscosity subsolution to @xmath162 on @xmath39 , @xmath307 is a viscosity supersolution to @xmath308 on @xmath39 and @xmath309 on @xmath41 , then @xmath309 on @xmath39 . we obtain in theorem [ thm-5.4 ] the result which allows functions @xmath10 and @xmath95 to be discontinuous on @xmath41 and is useful in proving the comparison principle . its proof can be done by a nonsymmetric adaptation of the proof in @xcite . [ thm-5.4 ] let @xmath33 be elliptic with respect to some class @xmath195 in the sense of definition [ def-3.1 ] and let @xmath91 be a bounded domain . if @xmath306 is a viscosity subsolution to @xmath162 on @xmath39 and @xmath310 is a viscosity supersolution to @xmath311 on @xmath39 , then @xmath312\ge f - g$ ] on @xmath39 in the viscosity sense . [ lem-5.5 ] let @xmath313 be a class of linear integro - differential operators which satisfies assumption [ ass-5.1 ] . if @xmath314 is a viscosity subsolution to @xmath315 on @xmath39 , then @xmath316 . given @xmath317 , take a sufficiently large @xmath318 so that @xmath319 . for @xmath246 and @xmath320 , let @xmath321 then @xmath322 for any @xmath323 . since @xmath324 for all @xmath299 , by assumption [ ass-5.1 ] there is a @xmath325 so that @xmath326(x)\le-{\epsilon}{\delta}/r^5 $ ] for any @xmath327 . then we can complete the proof by applying a similar method as in @xcite . _ @xmath328proof of theorem [ thm-5.3]@xmath329 $ ] _ by theorem [ thm-5.4 ] , we see that @xmath312\ge 0 $ ] on @xmath39 in the viscosity sense . applying lemma [ lem-5.5 ] , we obtain that @xmath330\le\sup_{{{\mathbb r}}^n{\setminus}{\omega}}[u - v].$ ] hence this completes the proof . once we obtain the comparison principle for viscosity subsolutions and supersolutions which is semicontinuous on @xmath137 , existence of the solutions of the dirichlet problem that we mentioned in the introduction follows from the perron s method @xcite as long as we can construct suitable barriers . for a domain @xmath39 which has the exterior ball condition and prescribed boundary data in @xmath41 being continuous , the function @xmath331 to be constructed in section [ sec-7 ] can be used as the barriers . the alexandroff - bakelman - pucci ( a - b - p ) estimate plays an important role in krylov and sofonov theory @xcite , which is an essential tool in the proof of harnack inequality for linear uniformly elliptic equations with measurable coefficients . in this section , the influence of the gradient term has been addressed in our proof of a - b - p estimate to which converges as @xmath53 is getting close to @xmath54 . in a later section , we shall use this nonlocal version of the a - b - p estimte to prove hlder estimates for @xmath53 close to @xmath54 . let @xmath124 be a function which is not positive outside the ball @xmath332 and is upper semicontinuous on @xmath333 . we consider its _ concave envelope _ @xmath270 in @xmath334 defined as @xmath335 in @xmath336 and @xmath123 in @xmath337 , where @xmath338 is the family of all the hyperplanes in @xmath11 . also we denote the _ contact set _ of @xmath10 and @xmath270 in @xmath339 by @xmath340 . [ lem-6.1 ] let us assume @xmath70 . let @xmath341 in @xmath342 and let @xmath270 be its concave envelope in @xmath336 . if @xmath314 is a viscosity subsolution to @xmath343 on @xmath339 where @xmath160 is a function with @xmath344 on @xmath345 , then there exists some constant @xmath346 depending only on @xmath347 and @xmath348 @xmath110but not on @xmath53@xmath111 such that for any @xmath349 and any @xmath129 there is some @xmath350 such that @xmath351 where @xmath352 for @xmath353 and @xmath354 . here , @xmath355 denotes any element of the superdifferential @xmath356 of @xmath357 at @xmath13 . we note that @xmath358 for @xmath359 if @xmath270 and @xmath10 are differentiable at @xmath359 . in this case , @xmath356 is a singleton set with element @xmath360 . _ @xmath328proof of lemma [ lem-6.1]@xmath329 $ ] _ take any @xmath349 . since @xmath10 can be touched by a hyperplane from above at @xmath13 , we see @xmath361 for a @xmath141 . thus it follows from theorem [ thm-3.4 ] that @xmath362 is well - defined . we observe that @xmath363 for any @xmath364 and @xmath186 , by the definition of concave envelope of @xmath10 in @xmath336 . since @xmath365 for any @xmath186 , we obtain that @xmath366 then the constant @xmath367 depending only on @xmath368 and @xmath348 is finite for @xmath369 . thus by theorem [ thm-3.4 ] we have that @xmath370 where @xmath371 . splitting the above integral in the rings @xmath372 , we have that @xmath373 assume that the conclusion does not hold , i.e. for any @xmath346 there are some @xmath374 and @xmath375 such that @xmath376 for all @xmath350 . since @xmath377 and @xmath378 remains bounded below for @xmath297 , it follows from that @xmath379 for any @xmath346 . taking @xmath380 large enough , we obtain a contradiction . hence we are done . _ lemma [ lem-6.1 ] would hold for any particular choice of @xmath381 by modifying @xmath380 accordingly . the particular choice @xmath354 is convenient for the proofs in section [ sec-7 ] . @xcite[lem-6.2 ] let @xmath270 be a concave function on @xmath382 where @xmath323 and let @xmath383 . if @xmath384 for any small @xmath246 where @xmath385 , then we have @xmath386 for any @xmath387 . [ cor-6.3 ] for any @xmath246 , there is a constant @xmath346 such that for any function @xmath10 with the same hypothesis as lemma [ lem-6.1 ] , there is some @xmath388 such that @xmath389 ^ -\,dy\le c\sup_{y\in\overline q}(1+{\eta}^{-n } \|f(y)\|^n)\ , \end{split}\ ] ] for any @xmath390 and any cube @xmath391 with diameter @xmath392 such that @xmath393 and @xmath394 , where @xmath354 and @xmath395 . the first part can be obtained by choosing @xmath396 in lemma [ lem-6.1 ] . also the second part follows as a consequence of lemma [ lem-6.2 ] and concavity ; @xmath397 ^ -\le c ( f(x)+\|{\nabla}{\gamma}(x)\|)^n\le 4^n c\frac{(1+{\eta}^{-n}\|f(x)\|^n)}{g_{\eta}({\nabla}{\gamma}(x))}. \end{split}\ ] ] thus we have that @xmath398 ^ -\le 4^n c ( 1+{\eta}^{-n}\|f(x)\|^n).$ ] take any @xmath399 where @xmath391 is a cube with diameter @xmath392 such that @xmath400 and @xmath394 . similarly to the above , we can obtain that @xmath401 ^ -\le 4^n c ( 1+{\eta}^{-n}\|f(\cdot)\|^n)$ ] @xmath286 on @xmath402 because @xmath403 ^ -=0 $ ] @xmath286 on @xmath404 as in @xcite . hence this implies the second part . we obtain a nonlocal version of alexandroff - bakelman - pucci estimate in the following theorem . [ thm-6.4 ] let @xmath405 and @xmath270 be functions as in lemma [ lem-6.1 ] . then there exist a finite family @xmath406 of open cubes @xmath407 with diameters @xmath408 such that @xmath170 any two cubes @xmath409 and @xmath407 do not intersect , @xmath174 @xmath410 , @xmath177 @xmath411 for any @xmath407 , @xmath412 @xmath413 where @xmath354 , @xmath414 @xmath415 ^ -\,dy\le c\sup_{\overline q_j}(1+{\eta}^{-n}\|f\|^n ) \|q_j\|$ ] , @xmath416 @xmath417 , where the constants @xmath346 and @xmath418 depend on @xmath419 and @xmath420 @xmath110 but not on @xmath53@xmath111 . in order to obtain such a family , we start by covering @xmath339 with a tiling of cubes of diameter @xmath421 . then discard all those that do not intersect @xmath345 . whenever a cube does not satisfy ( e ) and ( f ) , we split it into @xmath422 cubes of half diameter and discard those whose closure does not intersect @xmath345 . now our goal is to prove that eventually all cubes satisfy ( e ) and ( f ) and this process ends after a finite number of steps . assume that the process does not finish in a finite number of steps . then we can have an infinite nested sequence of cubes . the intersection of their closures will be a point @xmath423 . so we may choose a sequence @xmath424 with @xmath425 . since @xmath426 for all @xmath427 , by the upper semicontinuity of @xmath10 on @xmath333 we have that @xmath428 . also we have that @xmath429 because @xmath430 on @xmath336 by the definition of the concave envelope @xmath270 in @xmath334 . thus we obtain that @xmath431 . we will now get a contradiction by showing that eventually one of these cubes containing @xmath423 will not split . take any @xmath246 . by corollary [ cor-6.3 ] there is some @xmath432 such that @xmath433 ^ -\,dy\le c\sup_{y\in\overline q_j}(1+{\eta}^{-n } \|f(y)\|^n)\ , \|q_j\| \end{split}\ ] ] for any @xmath390 and a cube @xmath434 with diameter @xmath408 such that @xmath435 and @xmath436 . so we easily see that @xmath437 and @xmath438 . we recall that @xmath439 for any @xmath440 because @xmath270 is concave on @xmath336 and @xmath441 . since @xmath408 is comparable to @xmath442 , it thus follows that @xmath443 thus we proved ( f ) . moreover , ( e ) holds for @xmath407 because @xmath444 and @xmath445 . hence the cube @xmath407 would not split and the process must stop there . note that the upper bound for the diameters @xmath446 becomes very small as @xmath53 is getting close to @xmath54 . adding @xmath447 and taking @xmath211 , we obtain the classical alexandroff estimate as the limit of the riemann sums . for each @xmath448 , we have that @xmath449 ^ -\,dy\le c\sum_j \sup_{y\in\overline q_j}(1+{\eta}^{-n } of @xmath345 is getting close to the contact set @xmath345 and the above becomes the following estimate @xmath110refer to ch . 9.1 of @xcite@xmath111 ; @xmath450 for any @xmath390 , where @xmath451 . since @xmath452 , we have @xmath453 if we set @xmath454 , then we obtain that @xmath455 in this section , we are going to apply the a - b - p estimate to get the geometric decay rate of upper level sets for a nonnegative subsolution @xmath10 . to do this , we need a special function @xmath331 so that @xmath456 meets the conditions of the a - b - p estimate . it is based on the method used in @xcite , but nontrivial computations have been done to create positive terms so that it absorbs the influence of the gradient term . [ lem-7.1 ] there exist some @xmath457 and @xmath458 such that the function @xmath459 for any @xmath460 and @xmath461 . it is enough to show that there is some @xmath462 so that @xmath463 for @xmath464 ; for every other @xmath13 with @xmath465 , the above inequality follows by rotation . if @xmath466 , then we consider @xmath467 . note that @xmath468 for @xmath469 and @xmath470 for @xmath471 . then we can derive that @xmath472 for any @xmath473 , @xmath474 , @xmath475 and @xmath476 for @xmath364 . thus we see that @xmath477 for any @xmath186 . we denote by @xmath478 the class of all positive kernels satisfying and . for @xmath186 and @xmath77 , we define the map @xmath479 given by @xmath480 . then it is easy to check that the mapping @xmath481 given by @xmath482 is an isometry because @xmath483 is its inverse mapping . since @xmath484 , it follows from and the change of variables that @xmath485k_{\|x\|}(y)\,dy\\ & { \fallingdotseq}\|x\|^{{\sigma}+p}{{\mathcal l}}^{t\|x\|}_{\|x\|}f(x ) \end{split}\ ] ] for any @xmath486 and @xmath460 . since @xmath487 and the mapping @xmath481 given by @xmath488 is an isometry , by taking the infimum of both sides on @xmath302 in the above inequality we obtain that @xmath489 for any @xmath460 . by , and , we conclude that @xmath490 for any @xmath460 . in order to prove , we use the following elementary inequality that holds for any @xmath491 and @xmath458 ; @xmath492 using this inequality , we have that @xmath493 for any @xmath186 and @xmath364 . we now choose some sufficiently large @xmath458 so that @xmath494 for any sufficiently small @xmath495 , where @xmath496 is the surface measure of @xmath497 . since @xmath498 and @xmath499 for any @xmath186 and @xmath500 , it follows from , , and that @xmath501 for @xmath502 . thus we may take some sufficiently small @xmath503 and take some @xmath457 close enough to @xmath54 in the above so that @xmath504 for any @xmath505 . hence we complete the proof . [ cor-7.2 ] given any @xmath506 , there exist some @xmath325 and @xmath507 such that the function @xmath508 is a subsolution to @xmath509 for any @xmath510 , @xmath511 . let @xmath457 be the number of lemma [ lem-7.1 ] . without loss of generality , we may assume that @xmath512 . lemma [ lem-7.1 ] implies that the result of this corollary always holds for @xmath460 , when @xmath513 . if @xmath514 , then the result still holds for @xmath460 because @xmath515 for any @xmath473 , @xmath511 and @xmath186 . we shall select @xmath516 so small that the result holds also for @xmath517 $ ] . now we let @xmath518 as in the proof of lemma [ lem-7.1 ] . . then we write @xmath520 if we take some @xmath516 small enough so that @xmath521 for any @xmath522 and @xmath186 , from simple geometric observation it is easy to check that @xmath523 for any @xmath524 and @xmath186 . so we see that @xmath525 since @xmath526 , we have that @xmath527 for a constant @xmath367 depending on @xmath528 and the dimension @xmath529 . on the other hand , since @xmath530 and @xmath531 for any @xmath532 with @xmath533 , we have that @xmath534 if we select some @xmath535 sufficiently small so that @xmath536 then we can complete the proof . [ lem-7.3 ] given any @xmath537 , there exists a function @xmath538 such that @xmath170 @xmath331 is continuous on @xmath11 , @xmath174 @xmath539 on @xmath540 , @xmath177 @xmath541 on @xmath542 , @xmath412 @xmath543 is continuous on @xmath544 , @xmath414 @xmath545 on @xmath11 where @xmath546 is a positive bounded function on @xmath11 which is supported in @xmath547 , for any @xmath510 . let @xmath548 be the number of corollary [ cor-7.2 ] . we consider the function @xmath331 given by @xmath539 in @xmath549 , @xmath550 in @xmath551 , and @xmath552 in @xmath553 . where @xmath84 is a quadratic paraboloid chosen so that @xmath331 is @xmath12 across @xmath554 . we now choose the constant @xmath555 so that @xmath556 for @xmath557 ( recall that @xmath558 ) . since @xmath559 , @xmath543 is continuous on @xmath560 . also by corollary [ cor-7.2 ] we see that @xmath561 on @xmath562 . hence this completes the proof . the main tool that shall be useful in proving hlder estimates is a lemma that connects a pointwise estimate with an estimate in measure . the corresponding lemma in our context is the following . [ lem-7.4 ] let @xmath537 be given . if @xmath563 , then there exist some constants @xmath564 , @xmath565 and @xmath566 @xmath110depending only on @xmath567 and the dimension @xmath529@xmath111 for which if @xmath314 is a viscosity supersolution to @xmath568 on @xmath569 such that @xmath570 on @xmath11 and @xmath571 , then @xmath572 . _ remark . _ ( a ) we denote by @xmath573 an open cube @xmath574 and @xmath575 . if we set @xmath576 , then we denote by @xmath577 for @xmath578 . \(b ) if we assume that @xmath579 , then there is a simpler proof of lemma [ lem-7.3 ] using the ideas from @xcite . the result here is more complicated as in @xcite because we want to get an estimate that remains uniform as @xmath211 . _ @xmath328proof of lemma [ lem-7.4]@xmath329 $ ] _ we consider the function @xmath580 where @xmath331 is the special function constructed in lemma [ lem-7.3 ] . then we easily see that @xmath95 is upper semicontinuous on @xmath581 and @xmath95 is not positive on @xmath549 . moreover , @xmath95 is a viscosity subsolution to @xmath582 on @xmath569 . so we want to apply theorem [ thm-6.4 ] ( rescaled ) to @xmath95 . let @xmath270 be the concave envelope of @xmath95 in @xmath583 . since @xmath571 , @xmath584 and @xmath585 , we easily see that @xmath586 for some @xmath587 . we consider the function @xmath40 whose graph is the cone in @xmath588 with vertex @xmath589 and base @xmath590 . for any @xmath591 with @xmath592 , the hyperplane @xmath593 is a supporting hyperplane for @xmath40 at @xmath109 in @xmath594 . then @xmath595 has a parallel hyperplane @xmath596 which is a supporting hyperplane for @xmath95 in @xmath583 at some point @xmath597 . by the definition of concave envelope , we see that @xmath596 is also the hyperplane tangent to the graph of @xmath270 at @xmath598 , so that @xmath599 . this implies that @xmath600 . thus we have that @xmath601 ^ -\,dy,\ ] ] where @xmath602 is the function given in corollary [ cor-6.3 ] . we also observe as shown in @xcite that @xmath603 let @xmath604 be the finite family of cubes given by theorem [ thm-6.4 ] ( rescaled on @xmath569 ) . then it follows from , and theorem [ thm-6.4 ] that @xmath605^n}{\eta^n}+1\biggr)&\le c\,\int_{{{\mathcal c}}(u,{\gamma},b_1 ) } g_{{\eta}}({\nabla}{\gamma}(y))\det[d^2{\gamma}(y)]^-\,dy\\ & \le c\biggl(\sum_j\sup_{\overline q_j}\bigl(1+\eta^{-n}(\psi+{\varepsilon}_0)^n\bigr ) \|q_j\|\biggr)\\ & \le c\biggl(\sum_j\|q_j\|+{\eta}^{-n}\sum_j\sup_{\overline q_j}(\psi+{\varepsilon}_0)^n \|q_j\|\biggr).\end{split}\ ] ] if we set @xmath606 since @xmath571 and @xmath584 , we see that @xmath607 . if we choose @xmath608 and @xmath609 small enough , the above inequality implies that @xmath610 we recall from the proof of lemma [ lem-7.3 ] that @xmath546 is supported on @xmath611 and bounded on @xmath11 . thus the above inequality becomes latexmath:[$1/2\le c(\,\,\sum_{q_j\cap b_{1/4}\neq\phi } volumes of the cubes @xmath407 intersecting @xmath613 as follows ; @xmath614 since @xmath615 for any @xmath616 , the cube @xmath617 is contained in @xmath332 for any @xmath407 with @xmath618 . set @xmath619 . then by theorem [ thm-6.4 ] we have that @xmath620 and @xmath621 . then the family @xmath622 is an open covering of the union @xmath623 now we may take a subcovering of @xmath624 with finite overlapping number ( depending only on the dimension @xmath529 ) which covers the set @xmath625 . thus it follows from and that @xmath626 so we have that @xmath627 taking @xmath628 , we conclude that @xmath629 because @xmath630 . hence we complete the proof . let @xmath542 be the unit cube . then we split it into @xmath422 cubes of half side . we do the same splitting step with each one of these @xmath422 cubes and we continue this process . the cubes obtained in this way are called _ dyadic cubes . _ if @xmath402 is a dyadic cube different from @xmath542 , then we say that @xmath631 is the _ predecessor _ of @xmath402 if @xmath402 is one of @xmath422 cubes obtained from splitting @xmath631 . @xcite[lem-7.5 ] let @xmath632 and @xmath633 be measurable subsets of @xmath11 with @xmath634 . if @xmath635 is some number such that @xmath170 @xmath636 and @xmath174 @xmath637 for any dyadic cube @xmath402 with @xmath638 , then @xmath639 . the following lemma is a consequence of lemma [ lem-5.2 ] , lemma [ lem-7.4 ] and lemma [ lem-7.5 ] . [ lem-7.6 ] let @xmath564 be the constant in lemma [ lem-7.4 ] . if @xmath306 is a viscosity supersolution to @xmath640 on @xmath569 such that @xmath570 on @xmath11 and @xmath571 , then there exist universal constants @xmath346 and @xmath641 such that @xmath642 for any @xmath643 we note that @xmath644 note that @xmath645 first , we shall prove that @xmath646 where @xmath647 and @xmath566 are the constants chosen as in lemma [ lem-7.4 ] . if @xmath648 , then it has been done in lemma [ lem-7.4 ] assume the result holds for @xmath649 ( @xmath650 ) , and let @xmath651 and @xmath652 if we can show that @xmath653 , then can be obtained for @xmath654 . to show this , we apply lemma [ lem-7.6 ] . by lemma [ lem-7.4 ] , it is clear that @xmath655 and @xmath656 . so it remain only to prove ( b ) of lemma [ lem-7.5 ] ; that is , we need to show that if @xmath657 is a dyadic cube satisfying @xmath658 then @xmath637 . indeed , we suppose that @xmath659 and take @xmath660 such that @xmath661 we now consider the transformation @xmath662 and the function @xmath663 . if we can show that @xmath95 satisfies the hypothesis of lemma [ lem-7.4 ] , then we have that @xmath664 , and thus @xmath665 and the function @xmath666 . it now remains to show that @xmath95 satisfies the hypothesis of lemma [ lem-7.4 ] . we now take any @xmath667 . if we set @xmath668 , then we observe that @xmath669 if @xmath77 , then we note that @xmath670 where @xmath671 , and moreover the mapping @xmath481 given by @xmath672 is an isometry . since @xmath673 , we have that @xmath674 for any @xmath675 and any @xmath676 . taking the infimum of the right - hand side in the above inequality , we obtain that @xmath677 thus @xmath678 because @xmath640 on @xmath679 . by theorem [ thm-3.4 ] , we see that @xmath680 on @xmath681 . also it is obvious that @xmath682 on @xmath11 and we see from that @xmath683 . finally the result follows immediately from by taking @xmath684 and @xmath641 so that @xmath685 . hence we complete the proof . by a standard covering argument we obtain the following theorem . [ thm-7.8 ] for any @xmath537 , let @xmath616 be given . if @xmath306 is a viscosity supersolution to @xmath568 on @xmath336 such that @xmath686 on @xmath11 and @xmath687 where @xmath609 is the constant given in lemma [ lem-7.4 ] , then there are universal constants @xmath346 and @xmath641 such that @xmath688 for any @xmath643 in contrast to symmetric cases , we can not obtain the following theorem by rescaling the above theorem because our cases are not scaling invariant . we note that theorem [ thm-7.9 ] on @xmath689 shall be applied to obtain a harnack inequality , and also theorem [ thm-7.9 ] on @xmath690 $ ] will be used to prove an interior @xmath55-regularity . [ thm-7.9 ] for any @xmath537 , let @xmath616 be given , and let @xmath323 and @xmath691 $ ] . if @xmath306 is a viscosity supersolution to @xmath692 on @xmath693 such that @xmath570 on @xmath11 , then there are universal constants @xmath641 and @xmath346 such that @xmath694 let @xmath323 and set @xmath695 for @xmath696 where @xmath697 take any @xmath698 . if we set @xmath699 , then we see that @xmath700 . thus by the change of variables we have that @xmath701 for any @xmath675 and any @xmath702 , where @xmath703 for @xmath691 $ ] . taking the infimum of the right - hand side in the above inequality , we get @xmath704 thus by theorem 3.4 we have that @xmath680 on @xmath336 . applying theorem [ thm-7.8 ] to the function @xmath95 , we complete the proof . harnack inequality plays an important role in analysis . in this section , we obtain the harnack inequality for integro - differential equations whose associated kernel is not necessarily symmetric . our estimate depends only on a lower bound @xmath537 for @xmath705 and also remains uniform as @xmath211 . in this respect , we can look upon this estimate as a generalization of krylov - safonov harnack inequality . the proof is an adaptation of the method used in @xcite to the case whose associated kernel is not necessarily symmetric . [ thm-8.1 ] for @xmath537 , let @xmath706 . if @xmath314 is a positive function such that @xmath707 in the viscosity sense , then there is some constant @xmath346 depending only on @xmath708 and @xmath709 such that @xmath710 let @xmath711 be a point so that @xmath712 . then it is enough to show that @xmath713 without loss of generality , we may assume that @xmath714 and @xmath715 by dividing @xmath10 by @xmath716 . let @xmath641 be the number given in theorem [ thm-7.9 ] and let @xmath717 . we now set @xmath718 then we see that @xmath719 because @xmath10 is positive on @xmath11 . also there is some @xmath720 such that @xmath721 where @xmath722 . to finish the proof , we have only to show that @xmath723 can not be too large because @xmath724 for any @xmath725 if @xmath726 is some constant with @xmath727 . assume that @xmath723 is very large . then by theorem [ thm-7.8 ] we have that @xmath728 since @xmath729 for @xmath730 , we easily obtain that @xmath731 in order to get a contradiction , we estimate @xmath732 for some very small @xmath325 ( to be determined later ) . for any @xmath733 , we have that @xmath734 for @xmath325 so that @xmath735 is close to @xmath32 . we consider the function @xmath736 then we see that @xmath682 on @xmath737 , and also @xmath738 on @xmath739 because @xmath740 on @xmath739 . we now want to apply theorem [ thm-7.9 ] to @xmath95 . however @xmath95 is not positive on @xmath11 but only on @xmath739 . to apply theorem [ thm-7.9 ] , we consider @xmath741 instead of @xmath95 . since @xmath742 , we have that @xmath743 on @xmath739 . since @xmath744 on @xmath737 , if @xmath745 then we have that @xmath746 for any @xmath186 , @xmath747 and @xmath748 . take any @xmath749 and any @xmath745 . since @xmath750 , we thus have that @xmath751 set @xmath752 for @xmath753 and @xmath754 then there is some @xmath755 such that @xmath756 and we see that @xmath757 because @xmath714 . since @xmath758 , we have that @xmath759 for some constant @xmath346 which is independent of @xmath53 , and so we have that @xmath760 since @xmath761 on @xmath336 , by we have that @xmath762 thus we obtain that @xmath763 for a constant @xmath346 which is independent of @xmath53 , and so there is some @xmath764 such that @xmath765 since @xmath766 for any @xmath186 and @xmath767 , and @xmath768,\ ] ] we may assume that @xmath764 must be finite . since @xmath769 for any @xmath473 , by we obtain that @xmath770 we now may assume that @xmath771 because @xmath723 was very large and @xmath735 was close to @xmath32 . since @xmath772 , by and the change of variables we have that @xmath773 for any @xmath774 . thus by and theorem [ thm-3.4 ] we obtain that @xmath775 since @xmath776 and @xmath777 , applying theorem [ thm-7.9 ] we have @xmath778^{{\varepsilon}_}\bigl[u(\check x)((1-{\delta})^{-{\beta}}-1/2)\bigr]^{-{\varepsilon}_}\\ & \qquad\le c({\delta}r)^n\bigl[((1-{\delta})^{-{\beta}}-1)^{{\varepsilon}_}+{\delta}^{-n{\varepsilon}_}s_0^{-{\varepsilon}_}\bigr].\end{split}\ ] ] we now choose @xmath325 so small enough that @xmath779 since @xmath548 was chosen independently of @xmath723 , if @xmath723 is large enough for such fixed @xmath548 then we get that @xmath780 therefore we obtain that @xmath781 thus we conclude that @xmath782 which contradicts if @xmath723 is large enough . thus we complete the proof . the purpose of this section is to prove the following hlder regularity result ( see theorem [ thm-9.2 ] ) . before doing this , we obtain a technical lemma which shall be useful in proving theorem [ thm-9.2 ] . [ lem-9.1 ] for any @xmath537 , let @xmath705 be given . if @xmath405 is a bounded function with @xmath783 on @xmath11 such that @xmath784 in the viscosity sense where @xmath564 is some sufficiently small constant , then there is some universal constant @xmath785 @xmath110depending only on @xmath708 and @xmath709@xmath111 such that @xmath786 at the origin . more precisely , @xmath787 for some universal constant @xmath346 depending only on @xmath188 . we shall prove the following theorem using only theorem [ thm-7.9 ] . theorem [ thm-9.2 ] easily follows from lemma [ lem-9.1 ] by a simple rescaling argument . thus we have only to prove lemma [ lem-9.1 ] to establish theorem [ thm-9.2 ] . [ thm-9.2 ] for any @xmath537 , let @xmath705 be given . if @xmath10 is a bounded function on @xmath11 such that @xmath788 in the viscosity sense , then there is a constant @xmath785 @xmath110depending only on @xmath708 and @xmath709@xmath111 such that @xmath789 where @xmath346 is some universal constant depending only on @xmath188 . _ @xmath328proof of lemma [ lem-9.1]@xmath329 $ ] _ take any @xmath790 and choose an @xmath791 so large that @xmath792 then it suffices to show that there exist a nondecreasing sequence @xmath793 and a nonincreasing sequence @xmath794 such that @xmath795 in @xmath796 and @xmath797 so that the theorem holds with @xmath798 ; for , if @xmath799 for @xmath350 , then @xmath800 we construct @xmath801 and @xmath802 by induction . for @xmath803 , we can take @xmath804 and @xmath805 because @xmath806 . assume that we have the sequences up to @xmath801 and @xmath802 for @xmath807 . then we want to show that we can continue the sequences by finding @xmath808 and @xmath809 . fix any @xmath810 where @xmath811 . take any @xmath812 where @xmath813 if we set @xmath814 , then we see that @xmath815 and moreover we have that @xmath816 in @xmath817 , either @xmath818 in at least half of the points ( in measure ) or @xmath819 in at least half of the points . first , we deal with the case @xmath820 then we note that @xmath821 on @xmath822 and @xmath823 . then we have @xmath824 for any @xmath675 and any @xmath825 . taking the infimum of the right - hand side in the above inequality , we obtain that @xmath826 since @xmath568 on @xmath339 in the viscosity sense , we have that @xmath827 thus this implies that @xmath828 it also follows from the inductive hypothesis that if @xmath829 for any @xmath830 . that is to say , @xmath831 outside @xmath339 . if we set @xmath832 , then we have that @xmath833 on @xmath834 ; for , if @xmath835 and @xmath836 , then @xmath837 for any @xmath186 because @xmath682 on @xmath822 ; if @xmath835 and @xmath838 , then @xmath839 for any @xmath186 because @xmath682 on @xmath822 . since @xmath742 , we see that @xmath840 . thus it follows from and that if @xmath841 is given , then @xmath842 for any @xmath486 , whenever @xmath790 and @xmath616 . so we have that @xmath843 on @xmath834 . thus by we conclude that @xmath844 on @xmath834 where @xmath845 . take any point @xmath359 . since @xmath846 , we can apply theorem [ thm-7.9 ] on @xmath847 to obtain that @xmath848 thus we have that @xmath849 for any @xmath359 . if we choose @xmath609 sufficiently small , then this implies that @xmath850 on @xmath339 for some @xmath851 . if we set @xmath852 and @xmath853 , then we have that @xmath854 in @xmath817 . moreover , @xmath855 . then we may choose some small @xmath785 and @xmath851 so that @xmath856 , so that we obtain that @xmath857 . in this section , we prove an interior @xmath55-regularity result for viscosity solutions to a general class of fully nonlinear integro - differential equations . the key idea of proof is to apply the hlder estimates of theorem [ thm-9.2 ] to incremental quotients of the solution . there being no uniform bound in @xmath860 for the incremental quotients outside the domain may cause a technical difficulty since we are dealing with nonlocal equations . we shall solve it by assuming some extra regularity of the family of linear integro - differential operators @xmath14 . the extra assumption , added to the growth condition for the kernel , is a modulus of continuity of @xmath15 in measure so that far away oscillations tend to cancel out . if @xmath15 is a radial function satisfying , then it is interesting that the condition is not required . indeed , if @xmath864 for @xmath865 , then it follows from the mean value theorem and schwartz inequality that @xmath866 for any @xmath867 and @xmath868 , because @xmath869 for such @xmath870 and @xmath871 for @xmath872 $ ] . [ thm-10.1 ] for any @xmath537 , let @xmath705 be given . then there is some @xmath862 @xmath110depending on @xmath873 and the dimension @xmath529@xmath111 so that if @xmath874 is a nonlocal elliptic operator with respect to @xmath861 in the sense of definition [ def-3.1 ] and @xmath306 is a viscosity solution to @xmath875 on @xmath339 , then there is a universal constant @xmath785 @xmath110depending only on @xmath873 and the dimension @xmath529@xmath111 such that @xmath876 for some constant @xmath346 depending on @xmath877 and the constant given in @xmath110@xmath878@xmath111 @xmath110where we denote by @xmath879 the value we obtain when we apply @xmath38 to the constant function that is equal to zero@xmath111 . since @xmath875 on @xmath339 , it follows from lemma [ lem-3.2 ] that @xmath880 on @xmath339 . similarly we have that @xmath881 on @xmath339 . thus by theorem [ thm-9.2 ] we see that @xmath882 for any @xmath883 and @xmath884 now we want to improve the obtained regularity iteratively by applying theorem [ thm-9.2 ] again until we obtain lipschitz regularity in a finite number of steps . assume that we have proven that @xmath885 for some @xmath886 $ ] and @xmath887 . then we apply theorem [ thm-9.2 ] for the difference quotient @xmath888 where @xmath889 is a translation operator given by @xmath890 for @xmath891 . since @xmath892 $ ] , we see from theorem [ thm-5.4 ] that @xmath893 and @xmath894 on @xmath895 for any @xmath896 . since @xmath885 , we see that the family @xmath897 is uniformly bounded on @xmath895 . but the functions @xmath898 is not uniformly bounded outside the ball @xmath895 , and so we can not apply theorem [ thm-9.2 ] directly . however we observe from that @xmath898 has oscillations that give cancelations in the integrals . let @xmath899 be a smooth cutoff function supported in @xmath895 such that @xmath900 in @xmath901 where @xmath325 is some small positive number which shall be determined later . we write @xmath902 where @xmath903 and @xmath904 . take any @xmath905 and @xmath906 . then @xmath907 and @xmath908 . we prove that @xmath909 for some @xmath785 with @xmath910 . we note that @xmath911\ge 0-{{\mathcal m}}^+_{{{\mathfrak l}}^1_0 } w_2^h,\\{{\mathcal m}}^-_{{{\mathfrak l}}_0 } w_1^h\le { { \mathcal m}}^-_{{{\mathfrak l}}^1_0 } w_1^h&={{\mathcal m}}^-_{{{\mathfrak l}}^1_0 } [ w^h - w_2^h]\le 0-{{\mathcal m}}^-_{{{\mathfrak l}}^1_0 } w_2^h.\end{split}\ ] ] in order to apply theorem [ thm-9.2 ] , we have only to show that @xmath912 and @xmath913 are bounded on @xmath914 by @xmath915 for some universal constant . to show this , we prove that it is true for any operator @xmath916 . take any @xmath916 . since @xmath917 on @xmath914 , the jump part of @xmath918 vanishes on @xmath914 . thus by the change of variable and the mean value theorem we have that @xmath919\,d\tau\bigr)\tau_h u(x+y)}{\|h\|^{{\beta}}}\,k(y)\,dy\\&{\fallingdotseq}{{\mathcal l}}^t_1 w_2^h(x)+{{\mathcal l}}^t_2 w_2^h(x)\end{split}\ ] ] for any @xmath905 . set @xmath920 . since @xmath921 for any @xmath922 and @xmath923 , by we obtain that @xmath924 for any @xmath922 and @xmath906 . since @xmath925 for every @xmath922 , @xmath923 and @xmath906 , we have that @xmath926 for any @xmath922 and @xmath906 . then it follows from , and that @xmath927 and @xmath928 on @xmath914 for any @xmath929 . thus by theorem [ thm-9.2 ] we have that @xmath930 for any @xmath931 . by the standard telescopic sum argument @xcite , we obtain that @xmath932 repeating the above argument , after @xmath933$]-steps we have that @xmath934 for any unit vector @xmath935 , we consider the following incremental quotients with the same reasoning @xmath936 if we choose @xmath937 , @xmath938 and @xmath939 in , then by and we obtain that @xmath940 for any unit vector @xmath935 and for any @xmath31 with @xmath941 . from this , taking @xmath942 we conclude that @xmath943 hence we complete the proof . acknowledgement . * this work had been started during yong - cheol kim was visiting to university of texas at austin in the fall semester 2008 for his sabbatical year . he would like to thank for kind hospitality and deep concern of professor luis a. caffarelli during that time . ki - ahm lee was supported by the korea research foundation grant funded by the korean government(moehrd , basic research promotion fund ) ( krf-2008 - 314-c00023 ) . david gilbarg and neil s. trudinger , _ elliptic partial differential equations of second order , second edition . _ grundlehren der mathematischen wissenschaften [ fundamental principles of mathematical sciences ] , 224 . springer - verlag , berlin , 1983 . xiii+513 pp . e. r. jakobsen and k. h. karlsen , _ a maximum principle for semicontinuous functions applicable to integro - partial differential equations , _ nodea nonlinear differential equations appl . , * 13 * ( 2006 ) , 137165 .	we introduce a new class of fully nonlinear integro - differential operators with possible nonsymmetric kernels . for the index @xmath0 of the operator in @xmath1 ( subcritical case ) , we introduce very general class of fully nonlinear integro - differential operators and obtain a comparison principle , a nonlocal version of the alexandroff - backelman - pucci estimate , a harnack inequality , a hlder regularity , and an interior @xmath2-regularity for equations associated with such a class .
the following analyses were performed using data samples collected at the @xmath13 resonance with the belle detector operating at the kekb asymmetric - energy @xmath14 collider . the belle detector is a large - solid - angle magnetic spectrometer that consists of a silicon vertex detector ( svd ) , a 50-layer central drift chamber ( cdc ) , an array of aerogel cherenkov counters ( acc ) , a barrel - like arrangement of time - of - flight scintillation counters ( tof ) and an electromagnetic calorimeter comprised of csi(tl ) crystals ( ecl ) located inside a superconducting solenoid coil that provides a 1.5 t magnetic field . an iron flux - return located outside the coil is instrumented to detect @xmath15 mesons and to identify muons ( klm ) . the detector is described in detail in ref . one of the two @xmath17 mesons is reconstructed in a hadronic decay cascade ( `` full reconstruction '' ) , yielding the so - called tag momentum , @xmath18 . the direction of the signal @xmath17 in the c.m . frame is assumed to be opposite of the tag meson . the energy is set to equal the beam energy , @xmath19 . the magnitude of the four momentum is defined via the nominal @xmath17 mass , @xmath20 . the remaining event is then used to reconstruct both a light lepton @xmath21 , either @xmath22 or @xmath23 , and a meson @xmath24 , which is one of the five light mesons @xmath25 , @xmath26 , @xmath27 , @xmath28 or @xmath29 . .partial branching fractions in 3 bins of @xmath1 , obtained from the fit . these are summed to give the full branching fraction quoted in the `` sum '' column . errors are statistical and systematic . the numbers are premliniary . [ cols="<,^,^,^,^",options="header " , ] for signal the missing momentum @xmath30 corresponds to the momentum of the neutrino . a fit to the missing mass distribution , using the root ( @xcite ) ` tfractionfitter ` class , which uses the algorithm described in ref . @xcite , is used to extract the signal yield . three subsamples , corresponding to the phase space regions @xmath31 , @xmath32 and @xmath33 , are investigated seperately , where @xmath34 is the momentum transfer . the results of the fit can be seen in fig . [ fig : kevinmissingmass ] , where the three subsamples have been summed . the partial branching fractions , which are obtained from the fit , can be seen in tab . [ tab : kevinresults ] . the error is dominated statistically . is shown.,title="fig : " ] is shown.,title="fig : " ] is shown.,title="fig : " ] + is shown.,title="fig : " ] is shown.,title="fig : " ] the analysis aims at determining the quadruple differential decay width of the process @xmath6 , which can be found e.g. in ref . @xcite or @xcite . this decay width is a four dimensional function depending on the four kinematic variables @xmath35 and the three angles @xmath36 , @xmath37 and @xmath38 , which are defined in fig . [ fig : meinekinematicvariablesreconstruction ] . we use the parametrization defined in ref . @xcite , which introduces the three free parameters @xmath3 , @xmath39 and @xmath40 to govern the shape of the form factors of the decay . , @xmath36 , @xmath37 and @xmath38 and a sketch of the reconstruction of the signal @xmath17 momentum using momentum conservation.,title="fig : " ] , @xmath36 , @xmath37 and @xmath38 and a sketch of the reconstruction of the signal @xmath17 momentum using momentum conservation.,title="fig : " ] the decay cascade @xmath41 , @xmath42 and @xmath43 or @xmath44 is reconstructed . the light lepton @xmath21 is either an electron or a muon . due to momentum conservation , the spatial momentum of the signal @xmath17 has to lie on a cone around the spatial momentum of the @xmath45 system . the inclusive sum of the entire remaining event is used to obtain the best @xmath17 candiate by orthogonal projection , as sketched in the right hand plot of fig . [ fig : meinekinematicvariablesreconstruction ] . off - resonance data is used to investigate background from @xmath46 decays and mc for a set of five additional background components . the parameters @xmath47 , @xmath3 , @xmath39 and @xmath40 are obtained by a binned least squares fit to the four one - dimensional marginal distributions of the decay width . the bin - to - bin correlations between these one dimensional histograms have to be considered . only the branching ratio of the mode @xmath43 is used as an external parameter , the branching ratio of the mode @xmath44 is determined by fitting the ratio between the two @xmath48 channels , @xmath49 . a @xmath50 function is formed for each of the four channels seperately and the sum of these four @xmath50 s is minimized numerically using the minuit package @xcite . the preliminary results of the fit are @xmath51 , @xmath52 , @xmath53 , @xmath54 , @xmath55 and @xmath56 , where the first error is the statistical error reported by minuit and the second ( where shown ) is the preliminary systematic error . the @xmath57 of the fit gives @xmath58 . the systematic error of the hqet parameters is dominated by background uncertainty , for @xmath59 the error of the track reconstruction efficiency is dominating . the results of the fit are shown graphically in fig . [ fig : meineresultplot ] . the obtained value for @xmath47 is one of the lowest obtained by recent measurements , but in acceptable agreement with measurements by babar @xcite . according to hqet there are two doublets of orbitally excited charmed mesons , labeled @xmath61 @xmath62 , @xmath11 @xmath63 , @xmath10 @xmath64 and @xmath65 @xmath66 . the entire set of these four states is labeled @xmath67 . these states decay into either @xmath68 ( @xmath69 , @xmath70 ) or @xmath71 ( @xmath72 , @xmath73 , @xmath70 ) . this analysis investigates @xmath74 decays , for both charged and neutral @xmath17 mesons , and measures the excited @xmath75 contributions to the @xmath76 final state . the signal @xmath17 meson is reconstructed in the semileptonic mode of interest ( @xmath77 ) and the remaining event is then combined into either a @xmath78 , @xmath79 , or a @xmath80 combination ( @xmath18 ) . the missing mass squared spectrum , @xmath81 , where @xmath82 is the total four - momentum of the beams , is used to identify signal events . the signal window is defined as @xmath83 . background is estimated using mc simulation . the @xmath67 signals are extracted via simultaneous unbinned likelihood fits to the signal and background @xmath84 mass spectra . the results of the fit can be seen in fig . [ fig : dssmassdistributions ] . the signal functions , shown as the solid lines in the figure , describe each @xmath67 state by a relativistic breit - wigner function . a non - resonant part is modeled by the goity - roberts model @xcite , its contribution is consistent with zero in all cases . to further investigate the @xmath68 mass spectrum , also a @xmath85 hypothesis is tested . the @xmath86 contribution is described by a tail of the breit - wigner function . fit results for this combination are shown as a dashed line . the signal yields and branching fractions obtained from this fit can be seen in tab . [ tab : dssresults ] . , @xmath87 , @xmath88 and @xmath89 . insets show the distributions before background subtraction . the hatched histograms show the background . ] mode & yield & @xmath90 $ ] + @xmath91 & @xmath92 & @xmath93 + @xmath94 & @xmath95 & @xmath96 + @xmath97 & @xmath98 & @xmath99 + @xmath100 & @xmath101 & @xmath102 + @xmath103 & @xmath104 & @xmath105 + @xmath106 & @xmath107 & @xmath108 + mode & yield & @xmath90 $ ] + @xmath109 & @xmath110 & @xmath111 + @xmath112 & @xmath113 & @xmath114 + @xmath115 & @xmath116 & @xmath117 + @xmath106 & @xmath118 & @xmath119 + we presented three analyses based on data samples collected by the belle detector at the kekb @xmath0 asymmetric energy collider : the measurement of the branching fractions and @xmath1 distributions of exclusive charmless semileptonic b decays , the determination of @xmath47 and the form factor parameters @xmath3 , @xmath4 and @xmath5 in the exclusive decay @xmath6 and the investigation of @xmath8 decays . we thank the kekb group for the excellent operation of the accelerator , the kek cryogenics group for the efficient operation of the solenoid , and the kek computer group and the national institute of informatics for valuable computing and super - sinet network support .	we present three analyses based on data samples collected by the belle detector at the kekb @xmath0 asymmetric energy collider . the first one contains the measurement of the branching fractions and @xmath1 distributions of exclusive charmless semileptonic b decays , which are used to extract the cabibbo - kobayashi - maskawa matrix element @xmath2 . the sample of events is obtained by fully reconstructing one of the b mesons in selected hadronic modes . the second analysis determines the branching fraction and the form factors @xmath3 , @xmath4 and @xmath5 for the exclusive decay @xmath6 . these measurements are used to extract the cabibbo - kobayashi - maskawa matrix element @xmath7 . the third analysis investigates @xmath8 decays and determines branching ratios for the four modes @xmath9 , @xmath10 , @xmath11 and @xmath12 for both charged and neutral b decays .
the synthesis of graphene has boosted the research in atomically thin two - dimensional ( 2d ) materials@xcite . the ability to manipulate single atomic layers and reassemble them to form heterostructures opens novel routes for applications . in this field , 2d semiconducting dichalcogenides mx@xmath0 ( m = mo , w x = s , se , te ) are promising compounds since they can be easily exfoliated and present a suitable gap for electronic devices.@xcite the dichalcogenides systems has recently gained attention for combining an electron mobility comparable to that of graphene devices with a direct energy gap in monolayer and indirect gap in multilayers . another interesting feature is that the electronic properties and the gap size are highly sensitive to external pressure , strain and temperature.@xcite in addition , the lack of lattice inversion symmetry together with spin - orbit coupling ( soc ) leads to coupled spin and valley physics in monolayers of mx@xmath0 , making it possible to control spin and valley in these materials . . the fermi level is set to zero . panel b ) . top view of the mos@xmath0 single layer . mo and s atoms are shown as grey and yellow balls respectively . we show the @xmath2-like wannier function where red and blue contours are for isosurfaces of identical absolute values but opposite signs . panel c ) . charge density of the valence band at the k point . panel d ) . charge density of the valence band at the @xmath4 point . , width=328,height=313 ] while for graphenic materials the theoretical work has been based on tight binding - like approaches , this is not so easy for the mx@xmath0 because of the larger number of bands . numerous tight binding ( tb ) models for mx@xmath0 materials have been proposed in the last years . most of them were derived using the slater - koster parametrization @xcite . the difficulty of this approach to reproduce accurately the dft band structure with a limited set of orbitals leads to the necessity of a trustful tb model . even the crystal field is not accurately reproduced @xcite . a tb model based on the lattice symmetry just for the reflection - symmetric m-@xmath2 orbitals was derived @xcite leading to a good match with the dft energy bands . however including the x-@xmath6 orbitals is crucial in order to explain the electronic properties of mx@xmath0 dichalcogenides . without a good tight - binding model that accounts both for the m-@xmath2 and x-@xmath6 orbitals it is hazardous to deduce effective @xmath7 models from a mere analogy with monolayer graphene @xcite . it is also not clear whether an analogy with bilayer graphene is relevant @xcite . in this work , we derive a tb model from group theory for the m-@xmath2 and x-@xmath6 orbitals and fix its free parameters using the matrix elements obtained from the mlwfs . this sets the basis for a future derivation of a full and accurate tight binding model and the corresponding effective @xmath7 models . from the tb parameters obtained from density functional theory , we will study how the interplay between the crystal field and the hybridizations opens the gap and produces different spin - orbit ( so ) splitting in different regions of the brillouin zone . we will show how the lack of inversion symmetry produces the large so splitting in the monolayer and the coexistence of the spin orbit and hybridization contribution to the splitting in the bilayer . we have performed first - principles density functional theory ( dft ) calculations by using the vasp @xcite package based on plane wave basis set and projector augmented wave method.@xcite a plane - wave energy cut - off of 400 ev has been used . for the treatment of exchange - correlation , perdew - burke - ernzerhof @xcite generalized gradient approximation has been considered . after obtaining the bloch wave functions in density functional theory , the maximally localized wannier functions@xcite ( mlwf ) are constructed using the wannier90 code.@xcite a tb model based on group theory is derived . we want to mention that the mismatch between our tb model and the dft bands comes from two sources . the first is the truncation of the range of hopping parameters . the second source comes from a residual mixing between m and x orbitals inherent to wannier90 ( this is the consequence of the fact that the algorithm does nt use any symmetry constraints ) while in our model we work with disentangled orbitals . this issue will be addressed elsewhere . however since the m - x mixing of the mlwfs is quantitatively small , the incorporation of the matrix elements obtained by wannier90 into our tb model leads to a good qualitative picture . we define the last occupied band as the valence band ( v ) and the first unoccupied band as the conduction band ( c ) . we calculate the mlwfs for the last 7 occupied bands of the dichalcogenide materials mx@xmath0 and find that the general features are similar for all of them . our results show the presence of 6 equivalent @xmath6-like mlwfs centred on the x atoms and 1 @xmath2-like in the basal plane . therefore the other 4 @xmath2-bands are above the fermi level . basically from ionic charge point of view we have a m@xmath1-like and two x@xmath3-like atoms . this ionic picture it is more difficult to be understood by the local density of states.@xcite the presence of the ionic bond was already found in some compounds@xcite . the x@xmath3-like atoms are less electronegative than oxygen , therefore the chemical bonds are slightly more covalent respect to oxides . the 6 equivalent @xmath6-like mlwfs have the lobe of the orbital along the m - x direction.@xcite the only @xmath2-like mlwf shown in fig . [ mlwf]-b ) is more unusual . it is a trivalent bond centered among three m atoms and it extends in the basal plane with c@xmath8 symmetry . one of the components of the gap in these systems is the crystal field splitting induced by x@xmath3-like atoms . the @xmath2-like mlwf avoids the @xmath6-like out of plane orbitals and it is lower in energy than the other @xmath2-orbitals . going down the group in the periodic table , the electronegativity of the x element decreases . however the effective electronic charge on the x atom is mostly unmodified and the crystal field ( cf ) persists . now , we want to investigate in more detail the @xmath2-character of the v band and we try do disentangle the v band from the @xmath6-bands . we repeat the mlwf procedure just for the v band . a tight binding ( tb ) model for the v band is presented in the supplementary materials.@xcite the v band presents an almost pure 3z@xmath9-r@xmath9 character at @xmath4 ( see fig . [ mlwf]-d ) and a linear combination of x@xmath9-y@xmath9 and xy character at @xmath5 ( see fig . [ mlwf]-c ) . on the contrary the m point has a mixed character with non negligible @xmath6-orbital contribution . the corresponding mlwf lies mainly in the basal plane and it is basically the same of fig . [ mlwf]-b ) . . the fermi level is set to zero . , width=325,height=257 ] . the fermi level is set to zero . , width=325,height=257 ] we have found a perfect disentanglement between the v band and the other bands for the s / se compounds while it is not possible to disentangle it for the te - compounds as we can see in fig . [ mlwfmos2 ] and [ mlwfmote2 ] . this is due to the strong @xmath6-character at the m point for the te systems . due to the reduced electronegativity the te electrons hybridize so strongly with the v band that is not possible to disentangle them . as a result , it is possible to derive an effective 3 bands model for the s / se compounds as already proposed@xcite but not for the te compounds . the different behaviour of the te compounds was already addressed regarding the optical properties and the formation energies.@xcite a tb model based on group theory is derived for the basal plane reflection - symmetric orbitals only , including m - m , m - x and x - x nearest - neighbor and next - nearest - neighbor hopping terms without the inclusion of soc . the lattice of monolayer mx@xmath0 can be obtained from the combination of the translations of the 2d hexagonal lattice together with the rotations and reflections of the point group @xmath10=@xmath11 . it has no inversion center , hence the parity is not a good quantum number . instead the reflection symmetry with respect to the basal plane , @xmath12 , plays an important role . we restrict ourselves to the @xmath2 electrons of m and the @xmath6 electrons of x atoms . since there is in the unit cell the bottom ( b ) and the top ( t ) x atoms , we introduce the bonding ( @xmath13 ) and antibonding ( @xmath14 ) basis @xmath15 . at the @xmath4-point , the little group is @xmath10 and the electronic orbitals split under the cf as ( we use the notation of ref . ) @xmath16 with , in the left column , the states that are symmetric ( @xmath17 ) under @xmath12 , and , in the right column , the states that are antisymmetric ( @xmath18 ) under @xmath12 . since translational symmetry here only involves the basal plane , the coupling between @xmath17- and @xmath18-states is not allowed . on the contrary , the coupling among @xmath17- or @xmath18-states is allowed . below we focus on the @xmath17-bands since they are the most relevant in order to understand the formation of the gap . the comparison between the tb model and the dft calculation for mos@xmath0 is reported in fig . [ tb ] ( top panel ) . -bands obtained using the tb model ( blue ) for mos@xmath0 ( top panel ) . band gap as a function of the m - x , m@xmath19-m@xmath20 and m - m hybridization ( bottom panel ) . in the inset we have the @xmath17-bands obtained using the tb model with m - x hyvridization set to zero . , title="fig:",width=321,height=207 ] -bands obtained using the tb model ( blue ) for mos@xmath0 ( top panel ) . band gap as a function of the m - x , m@xmath19-m@xmath20 and m - m hybridization ( bottom panel ) . in the inset we have the @xmath17-bands obtained using the tb model with m - x hyvridization set to zero . , title="fig:",width=321,height=215 ] we note the presence of twofold degeneracies of the band structure at the @xmath4-point which correspond to the two - dimensional representations introduced in eq . ( [ sym ] ) , @xmath21 from the @xmath17-bands and @xmath22 from the @xmath18-bands . as we go away from the @xmath4-point , the little group has a lower symmetry ( @xmath23 at the @xmath5-point , and @xmath24 at the @xmath25-point ) such that the two - dimensional representations split into two distinct one - dimensional representations . we extract the hopping parameters for the model with the x - p and m - d orbitals . we analyze what happens at the @xmath4 point for the d - orbitals @xmath17-bands . the orbital for @xmath26=1,2,3 are respectively x@xmath9-y@xmath9 , xy and 3z@xmath9-r@xmath9 . because of the orthogonality condition , the d - subspace of the hamiltonian is diagonal.@xcite it is determined by the on - site energy and the m@xmath27-m@xmath27 hybridization that are given in table [ tabcf ] . .the on - site energies and sum of hopping integrals between the m atoms for the nearest - neighbor ( nn ) , for the next - nearest - neighbor ( nnn ) as the selected mlwfs of mos@xmath0 respectively in the first , second and third line . the on - site energy of the 3z@xmath9-r@xmath9-like wf is set to zero . in the fourth and fifth lines , we report the diagonal matrix elements at @xmath4 and @xmath5 point up to the nn@xcite . hopping integrals @xmath28 from the site 0 with orbital @xmath26 to neighboring site i with orbital @xmath26 . @xmath29 is the on - site energy for the orbital @xmath26 . the unit is mev . [ cols="^,^,^,^,^,^,^,^,^,^ " , ] [ tabgapbilayer ] the bonding - antibonding scenario created by the m@xmath30-m@xmath31 hybridization is not modified for the bilayer but the adding of the hybridization between different layers contributes to the splitting at the k point . though the x - x hybridization is dominant , we can have an analytic formula for the splitting if we set it to zero . the bilayer splitting at the k point in the limit of @xmath32 is@xcite @xmath33 and we can observe how the splitting has a dependence both on soc and hybridizations . the splitting for the bilayer is always larger than the splitting for the monolayer . the splitting is non zero also when the soc vanishes and it is always larger than the splitting for the monolayer . at @xmath34=0 using the numerical values@xcite we get @xmath35=23.7 mev for the mos@xmath0 . this is a small contribution compared with the spin - orbit and the hybridization between the in plane x - p orbitals of the different monolayers . thus , we have demonstrated that the splitting of the valence band in the case of the bilayer does not come just from the spin - orbit coupling . in conclusion , the model presented here provides a basis for tight - binding calculations for mx@xmath0 systems with an _ ab - initio _ accuracy . the crystal field splitting induced by the x@xmath3-like atoms is barely affected while changing the x element . on the contrary , the m - m and the m - x hybridization are reduced going down the group because of the increase in the lattice parameter . the cf , the m - m and m - x hybridization play in the same direction to open the gap at the @xmath4 point . along the @xmath4-@xmath5 line the m@xmath27-m@xmath27 tends to create a crossing of the @xmath17-band closing the gap but the m - x hybridization opens it . at the @xmath5 point , the m@xmath27-m@xmath27 hybridization inverts the @xmath2-bands respect to the cf and the gap is enhanced by the m - x and the m@xmath30-m@xmath31 hybridization . we show that we need both the m - x and m - m hybridization to open the gap . the large so splitting observed in the v band at the @xmath5 point is assisted by the strong m@xmath30-m@xmath31 hybridization and the absence of the inversion symmetry . the change from direct to indirect gap and the splitting in the valence bands in the multilayers is attributed to the hybridization between out of plane orbitals of different layers . the splitting at the k point is attributed both to the soc and to the hybridization between in plane orbitals . recently , a similar paper has been published , see ref . . though the basic computation details in the initial part of the papers are the same , the physical properties under consideration differ in the rest of the papers . we thank c. noce and s. haldar for useful discussions . the simulations were performed on resources provided by the swedish national infrastructure ( snic ) at national supercomputer centre at linkping university ( nsc ) . k. s. novoselov , a. k. geim , s. v. morozov , d. jiang , y. zhang , s. v. dubonos , i. v. gregorieva and a. a. firsov , science * 306 * , 666 ( 2004 ) . k. s. novoselov , d. jiang , f. schedin , t. j. booth , v. v. khotkivich , s. v. morozov and a. k. geim , proc . usa * 102 * , 10451 ( 2004 ) . f. zahid , l. liu , y. zhu , j. wang and h. guo , aip advances * 3 * , 052111 ( 2013 ) . h. rostami , a. g. moghaddam , and r. asgari , phys . b * 88 * , 085440 ( 2013 ) . e. ridolfi , d. le , t. s. rahman , e. r. mucciolo , and c. h. lewenkopf , arxiv:1507.00356 [ cond-mat.mes-hall ] . k. v. shanavas and s. satpathy , phys . b * 91 * , 235145 ( 2015 ) .	by means of first - principles density functional calculations , we study the maximally localized wannier functions for the 2d transition metal dichalcogenides mx@xmath0 ( m = mo , w x = s , se , te ) . we found a m@xmath1-like ionic charge and a single occupied @xmath2-band . the center of the @xmath2-like maximally localized wannier function associated with this band is distributed among three m sites . part of the energy gap is opened by the crystal field splitting induced by the x@xmath3-like atoms . we extract the hopping parameters for the wannier functions and provide a perspective on tight - binding model . from the analysis of the tight binding model , we have found an inversion of the band character between the @xmath4 and the @xmath5 points of the brillouin zone due to the m - m hybridization . the consequence of this inversion is the closure of the gap . the m - x hybridization is the only one that tends to open the gap at every k - point , the change in the m - x and m - m hybridization is the main responsible for the difference in the gap between the different dichalcogenide materials . the inversion of the bands gives rise to different spin - orbit splitting at @xmath4 and @xmath5 point in the valence band . the different character of the gap at @xmath4 and @xmath5 point offers the chance to manipulate the semiconductive properties of these compounds . for a bilayer system , the hybridizations between the out of plane orbitals and the hybridizations between the in plane orbitals split the valence band respectively at the @xmath4 and k point . the splitting in the valence band is opened also without spin - orbit coupling and comes from the m - m and x - x hybridization between the two monolayers .
current research in music - information - retrieval(mir ) is largely limited to western music cultures and it does not address the north - indian - music - system hereafter nims , cultures in general . nims raises a big challenge to current rhythm analysis techniques , with a significantly sophisticated rhythmic framework . we should consider a knowledge - based approach to create the computational model for nims rhythm . tools developed for rhythm analysis can be useful in a lot of applications such as intelligent music archival , enhanced navigation through music collections , content based music retrieval , for an enriched and informed appreciation of the subtleties of music and for pedagogy . most of these applications deal with music compositions of polyphonic kind in the context of blending of various signals arising from different sources . apart from the singing voice , different instruments are also included . as per @xcite rhythm relates to the _ patterns of duration _ that are phenomenally present in the music . it should be noted that that these _ patterns of duration _ are not based on the actual duration of each musical event but on the inter onset interval(ioi ) between the attack points of successive events . as per [ @xcite ] , an accent or a stimulus is marked for consciousness in some way . accents may be phenomenal , i.e. changes in intensity or changes in register , timbre , duration , or simultaneous note density or structural like arrival or departure of a cadence which causes a note to be perceived as accented . it may be metrical accent which is perceived as accented due to its metrical position [ @xcite . percussion instruments are normally used to create accents in the rhythmic composition . the percussion family which normally includes timpani , snare drum , bass drum , cymbals , triangle , is believed to include the oldest musical instruments , following the human voice [ @xcite ] . the rhythm information in music is mainly and popularly provided by the percussion instruments . one simple way of analyzing rhythm of a composite or polyphonic music signal having some percussive component , may be to extract the percussive component from it using some source separation techniques based on frequency based filtering . various attempts have been made in western music to develop applications for re - synthesizing the drum track of a composite music signal , identification of type of drums played in the composite signal[ex . the works of @xcite etc . , described in section [ past ] in detail ] . human listeners are able to perceive individual sound events in complex compositions , even while listening to a polyphonic music recording , which might include unknown timbres or musical instruments . however designing an automated system for rhythm detection from a polyphonic music composition is very difficult . in the context of nims rhythm popularly known as _ tla _ , _ tabl _ is the most popular percussive instrument . its right hand drum-_dayan _ and left hand drum-_bayan _ are played together and amplitude - peaks spaced at regular time intervals , are created by playing every stroke . one way of rhythm information retrieval from polyphonic composition having _ tabl _ as one of the percussive instruments , may be to extract the _ tabl _ signal from it and analyze it separately . the _ dayan _ has a frequency overlap with other instruments and mostly human - voice for polyphonic music , so if we extract the whole range of frequencies for both _ bayan _ and _ dayan _ components , by existing frequency based filtering methods , the resultant signal will be a noisy version of original song as it will still have part of other instruments , human voice components along with _ tabl_. also conventional source separation methods lead to substantial loss of information or sometimes addition of unwanted noise . this is the an area of challenge in _ tla _ analysis for nims . although , nims _ tla _ functions in many ways like western meter , as a periodic , hierarchic framework for rhythmic design , it is composed of a sequence of unequal time intervals and has longer time cycles . moreover _ tla _ in nims is distinctively cyclical and much more complex compared to western meter [ @xcite ] . this complexity is another challenge for _ tla _ analysis . due to the above reasons defining a computational framework for automatic rhythm information retrieval for north indian polyphonic compositions is a challenging task . very less work has been done for rhythmic information retrieval from a polyphonic composition in nims context . in western music , quite a few approaches are followed for this purpose , mostly in the areas of beat - tracking , tempo analysis , annotation of strokes / pulses from the separated percussive signal . we have described these systems in the section [ past ] . for nims , very few works of rhythm analysis are done by adopting western drum - event retrieval system . these works result in finding out meter or speed which are not very significant in the context of nims . hence this is an unexplored area of research for nims . in this work we have proposed a completely new approach , i.e.instead of extracting both _ bayan _ and _ dayan _ signal , we have extracted the _ bayan _ signal from the polyphonic composition by using band - pass filter . this filter extracts lower frequency part which normally does not overlap with the frequency of human voice and other instruments in a polyphonic composition . most of the _ tla_-s start with a _ bol _ or stroke which has a _ bayan _ component(either played with _ bayan _ alone or both _ bayan _ and _ dayan _ together ) and also the some consequent section(_vibhga _ in nims terminology ) boundary-_bol_-s have similar _ bayan _ component . hence these strokes would be captured in the extracted _ bayan _ signal . for a polyphonic composition , its _ tla _ is rendered with cyclically recurring patterns of fixed time - lengths . this is the cyclic property of nims , discussed in detail in section [ defn ] . so after extracting the starting _ bol_-s and the section boundary strokes from the _ bayan _ signal , we can exploit the cyclic property of a _ tla _ and the pattern of strokes appearing in a single cycle and can detect important rhythm information from a polyphonic composition . this would be a positive step towards rhythm information retrieval from huge collection of music recordings for both film music and live performances of various genres of _ hindi _ music . here , we consider the _ tla _ detection of different single - channel , polyphonic clips of _ hindi _ vocal songs of devotional , semi - classical and movie soundtracks from nims , having variety of tempo and _ mtr_-s . the rest of the paper is organized as follows . a review of past work is presented in section [ past ] . some definitions are provided in section [ defn ] . in section [ meth ] the proposed methodology is elaborated . experimental results are placed in section [ exp ] and the paper ends with concluding remarks in section [ con ] . the basic identifying features of rhythm or _ tla _ in nims are described as follows . * * _ tla _ and its cyclicity : * north indian music is metrically organized and it is called _ nibaddh_(bound by rhythm ) music . this kind of music is set to a metric framework called _ tla_. each _ tla _ is uniquely represented as cyclically recurring patterns of fixed time - lengths . * * _ vart _ : * this recurring cycle of time - lengths in a _ tla _ is called _ vart_. _ vart _ is used to specify the number of cycles played in a composition , while annotating the composition . * * _ mtra : _ * the overall time - span of each cycle or _ vart _ is made up of a certain number of smaller time units called _ mtra_-s . the number of _ mtra_-s for the nims _ tla_-s , usually varies from @xmath0 to @xmath1 . * * _ vibhga : _ * the _ mtra_-s of a _ tla _ are grouped into sections , sometimes with unequal time - spans , called _ vibhga_-s . + * * _ bol : _ * in the _ tla _ system of north indian music , the actual articulation of _ tla _ is done by certain syllables which are the mnemonic names of different strokes / pulses corresponding to each _ mtra_. these syllables are called _ bol_-s . there are four types of _ bol_-s as defined below . 1 . * _ sam : _ * the first _ mtr _ of an _ vart _ is referred as _ sam _ which is mandatorily stressed [ @xcite ] . _ tl - bol _ : * _ tl - bol_-s are usually stressed , whereas _ khali_-s are not . _ tl - bol_-s are gestured by the _ tabl _ player with claps of the hands , hence are called _ sasabda kriya_. the _ sam _ is almost always a _ tl - bol _ for most of the _ tla_-s , with only exception of _ rupak _ _ tla _ which designates the _ sam _ with a moderately stressed _ bol _ called _ khali_(as explained below ) [ @xcite ] . + highly stressed _ vibhga _ boundaries are indicated through the _ _ tl - bol__s[@xcite ] . _ tl - sam _ is indicated with a ( @xmath2 ) in the rhythm notation of nims . consequent _ tl - vibhga_-boundaries are indicated with @xmath3 . 3 . * _ khali - bol _ : * _ khali _ literally means empty and for nims it implies wave of the hand or _ nisabda kriya_. moderately stressed _ vibhga _ boundaries are indicated through the _ _ khali - bol__s so we almost never find the _ khali _ applied to strongly stressed _ bol_-s like _ sam _ [ @xcite ] . + _ khali - sam _ is indicated with a ( @xmath4 ) in the rhythm notation of nims and consequent _ khali - vibhga_-boundaries are indicated also with @xmath4 . * absent-_bol _ : * sometimes while playing _ tabl _ , certain _ bol_-s are dropped maintaining the perception of rhythm intact . they are called rests and they have equal duration as a _ bol_. we have termed them as absent strokes/_bol_-s . bol_-s are denoted by @xmath5 in the rhythm notation of a nims composition http://www.ancient-future.com/pronuind.html[ancient-future ] . in the figure [ absent ] , the waveform of absent _ bol _ , denoted by @xmath5 , is shown just after another _ bol ta _ , played in a _ tabl_-solo . + normally in a nims composition there may be many absent _ bol_-s in the _ thek _ played for the _ tla_. in these cases other percussive instruments(other than _ tabl _ ) and vocal emphasis might generate percussive peaks for the time positions of the absent strokes , depending on the composition , the lyrics being sung and thus the rhythm of the composition is maintained . * * _ thek _ : * for _ tabl _ , the basic characteristic pattern of _ bol_-s that repeats itself cyclically along the progression of the rendering of _ tla _ in a composition , is called _ thek_. in other words it s the most basic cyclic form of the _ tla _ [ @xcite ] . naturally _ thek _ corresponds to the basic pattern of _ bol_-s in an _ vart_. the strong starting _ sam _ along with the _ tl_-_vibhga_-boundaries in a _ _ carries the main accent and creates the sensation of cadence and cyclicity . * description of the definitions with an example : * the details of these theories are shown in the structure of a _ tla _ , called _ jhaptal _ in the table [ jhap ] and figure [ fig_jhap ] . the hierarchy of the features and their interdependence are shown in the figure [ heir ] . the cyclic property of _ tla _ is evident here . [ cols="<,<,<,<,<,<,<,<,<,<,<",options="header " , ] overall _ tla _ and tempo detection performance is shown in table [ gross ] . it is clear that the proposed methodology performs satisfactorily and that too with wide variety of data . ll + _ mtr _ detection&tempo detection + 81.59&78.60 + 1 . this paper presents the results of analysis of _ tabl _ signal of north indian polyphonic composition , with the help of new technique by extracting the _ bayan _ signal . the justification of using _ bayan _ signal as the guiding signal in case of north indian polyphonic music and detecting _ tla _ using the parameters of nims rhythm , has been clearly discussed . a large number of polyphonic music samples from _ hindi _ vocal songs from _ bhajan _ or devotional , semi - classical and filmy genres were analyzed for studying the effectiveness of the proposed new method . the experimental result of the present investigation clearly supports the pronounced effectiveness of the proposed technique . 5 . we would extend this methodology for studying other features(both stationary and non - stationary ) of the all the relevant _ tla_-s of nims and designing an automated rhythm - wise categorization system for polyphonic compositions . this system may be used for content - based music retrieval in nims . also a potential tool in the area of music research and training is expected to come out of it . limitations of the method is that it can not distinguish between _ tla_-s of same _ matr_. for example _ deepchandi _ and _ dhamar _ _ tla_-s have @xmath6 number of _ matr_-s , textitbol - s and beats in a cycle . we plan to extend this elementary model of _ tla_-detection system for all the nims _ tla_-s , by including other properties like timbral information and nonlinear properties of different kinds of _ tabl _ strokes/_bol_-s . we may also attempt to transcript the _ tla_-_bol_-s in a polyphonic composition . this extended version of the model may address the nims _ tla_-s which share same _ matr _ and also have variety of _ lay_-s . we thank the * department of higher education and rabindra bharati university , govt . of west bengal , india * for logistics support of computational analysis . we also thank renowned musician http://www.subhranilsarkar.com[subhranil sarkar ] for helping us to annotate test data , validate test results and shiraz ray(http://www.dgfoundation.in[deepa ghosh research foundation ] ) for extending help in editing the manuscript to enhance its understandability . nobutaka ono , kenichi miyamoto , j.l.r . , kameoka , h. , sagayama , s. : separation of a monaural audio signal into harmonic / percussive components by complementary diffusion on spectrogram . in : proc . of the eusico european signal processing conf . ( 2008 ) b.schuller , eyben , f. , rigoll , g. : fast and robust meter and tempo recognition for the automatic discrimination of ballroom dance styles . in : acoustics , speech and signal processing , 2007 . icassp 2007 . ieee international conference . ( 2007 ) 217220 bhaduri , s. , saha , s.k . , mazumdar , c. : _ matra _ and _ tempo _ detection for indic _ tala_-s . in : advanced computing and informatics proceedings of the second international conference on advanced computing , networking and informatics ( icacni-2014 ) . volume 1 . ( 2014 ) 213 220 bhaduri , s. , saha , s.k . , mazumdar , c. : a novel method for tempo detection of indic _ tala_-s . in : emerging applications of information technology ( eait ) , 2014 fourth international conference of ieee . ( 2014 ) 222227 maity , a.k . , pratihar , r. , mitra , a. , dey , s. , agrawal , v. , sanyal , s. , banerjeeb , a. , sengupta , r. , ghosh , d. : multifractal detrended fluctuation analysis of alpha and theta eeg rhythms with musical stimuli . chaos , solitons and fractals * 81(a ) * ( 2015 ) 5267 uhle , c. , dittmar , c. , sporer , t. : extraction of drum tracks from polyphonic music using independent subspace analysis . in : in proceedings of the 4th international symposium on independent component analysis and blind signal separation ( ica2003 ) . ( 2003 ) 843848 yoshii , k. , goto , m. , okuno , h.g . : automatic drum sound description for real - world music using template adaptation and matching methods . in : in proceedings of the international conference on music information retrieval ( ismir ) . ( 2004 ) 184191	in north - indian - music - system(nims),_tabl _ is mostly used as percussive accompaniment for vocal - music in polyphonic - compositions . the human auditory system uses perceptual grouping of musical - elements and easily filters the _ tabl _ component , thereby decoding prominent rhythmic features like _ tla _ , tempo from a polyphonic - composition . for western music , lots of work have been reported for automated drum analysis of polyphonic - composition . however , attempts at computational analysis of _ tla _ by separating the _ tabl_-signal from mixed signal in nims have not been successful . _ tabl _ is played with two components - right and left . the right - hand component has frequency overlap with voice and other instruments . so , _ tla _ analysis of polyphonic - composition , by accurately separating the _ tabl_-signal from the mixture is a baffling task , therefore an area of challenge . in this work we propose a novel technique for successfully detecting _ tla _ using left-_tabl _ signal , producing meaningful results because the left-_tabl _ normally does nt have frequency overlap with voice and other instruments . north - indian - rhythm follows complex cyclic pattern , against linear approach of western - rhythm . we have exploited this cyclic property along with stressed and non - stressed methods of playing _ tabl_-strokes to extract a characteristic pattern from the left-_tabl _ strokes , which , after matching with the grammar of _ tla_-system , determines the _ tla _ and tempo of the composition . a large number of polyphonic(vocal+__tabl__+other - instruments ) compositions has been analyzed with the methodology and the result clearly reveals the effectiveness of proposed techniques . * * keywords:**left-_tabl _ drum , _ tla _ detection , tempo detection , polyphonic composition , cyclic pattern , north indian music system
let @xmath0 be given statements . we deal with a generalization of the formal logical rule @xcite @xmath1 in @xcite and @xcite we prove , clarify and use the equivalence @xmath2 it gives an algorithm for composition of inverse problems with a given logical structure that is based on the steps below . * formulating and proving _ generating problems _ with logical structures of the statements as those at the left hand side of ( * ) . * formulating a problem with a logical structure of the statement @xmath3 * formulating an _ inverse problem _ with a logical structure @xmath4 . a technology for composition of equivalent and inverse problems is appropriate for training of mathematics students and teachers . in what follows we illustrate the above algorithm by two groups of problems . the logical statements used for the formulation of the problems in this group are * @xmath5 _ in the plane are given a @xmath6 , a straight line @xmath7 and an arbitrary point _ @xmath8 * @xmath9 _ the straight line @xmath10 is a median in _ @xmath11 * @xmath12 * @xmath13 let in a @xmath6 the straight line @xmath10 passes through @xmath14 and is a median . prove that for any point @xmath15 on @xmath10 @xmath16 the triangles @xmath17 and @xmath18 are equal in area . this problem has a logical structure @xmath19 . _ since the straight line @xmath10 passes through @xmath14 and is a median in a @xmath6 , then @xmath20 . let @xmath21 ( fig . 1 ) . hence @xmath22 . let @xmath23 and @xmath24 be the altitudes in @xmath25 and @xmath26 through the vertices @xmath27 and @xmath28 respectively . then @xmath29 let in @xmath6 the straight line @xmath10 passes through @xmath14 and is parallel to the straight line @xmath30 . prove that for any point @xmath15 on @xmath10 @xmath31 the triangles @xmath32 and @xmath18 are equal in area . this problem has a logical structure @xmath33 . the proof follows immediately from figure 2 . according to ( * ) and the _ generating problems _ 2.1 and 2.2 , we can formulate a problem with a logical structure @xmath3 since the statements @xmath34 and @xmath35 are mutually exclusive , then @xmath36 . under this condition problems with structures @xmath37 and @xmath38 are equivalent . we formulate and solve the following inverse problem with a logical structure @xmath39 let the point @xmath15 in the plane of @xmath6 is such that @xmath17 and @xmath18 exist and are equal in area . prove that the straight line @xmath40 either cuts the side @xmath30 of @xmath6 at its middle point or is parallel to @xmath30 . _ the triangles @xmath17 and @xmath18 have a common side @xmath40 and are equal in area . hence the altitudes to their common side are equal , i. e. the vertexes @xmath28 and @xmath27 are equidistant from @xmath40 . there are two possibilities for the location of the points @xmath28 and @xmath27 with respect to the straight line @xmath40 . * if @xmath28 and @xmath27 are situated at opposite sides of @xmath40 ( fig . 1 ) , then the point @xmath41 is the middle point of the side @xmath30 ( @xmath42 ) and the straight line @xmath40 is the median of @xmath6 through @xmath14 . * if @xmath28 and @xmath27 are situated at one and the same side of @xmath40 ( fig . 2 ) , then they lie on a straight line parallel to @xmath40 , i. e. @xmath43 . the logical statements used for the formulation of the problems in this group are	by suitable examples we illustrate an algorithm for composition of inverse problems .
recent solar @xcite , atmospheric @xcite , reactor @xcite and accelerator @xcite neutrino experiments have convincingly verified the hypothesis of neutrino oscillation , a quantum phenomenon which can naturally happen if neutrinos are slightly massive and lepton flavors are not conserved . the mixing of lepton flavors is described by a @xmath8 unitary matrix @xmath9 , whose nine elements are commonly parameterized in terms of three rotation angles ( @xmath10 , @xmath11 , @xmath12 ) and three cp - violating phases ( @xmath13 , @xmath14 , @xmath15 ) @xcite . the phase parameters @xmath14 and @xmath15 , which have nothing to do with cp violation in neutrino oscillations , are usually referred as to the majorana phases . a global analysis of current neutrino oscillation data yields @xmath16 , @xmath17 and @xmath18 at the @xmath19 confidence level @xcite , but three phases of @xmath9 remain entirely unconstrained . while the absolute mass scale of three neutrinos is not yet fixed , their two mass - squared differences have already been determined to a good degree of accuracy @xcite : @xmath20 and @xmath21 . the on - going and forthcoming neutrino oscillation experiments will shed light on the sign of @xmath22 , the magnitude of @xmath12 and even the cp - violating phase @xmath13 . from a phenomenological point of view , at least two lessons can be learnt from current experimental data : * the lightest neutrino is allowed to be massless ; i.e. , either @xmath0 ( normal neutrino mass hierarchy ) or @xmath1 ( inverted neutrino mass hierarchy ) has no conflict with the present neutrino oscillation measurements . in both cases , the non - vanishing neutrino masses can be determined in terms of @xmath23 and @xmath24 : @xmath25 whether one of the above two neutrino mass spectra is true or essentially true remains an open question . but we stress that some interesting neutrino models , such as the minimal seesaw model @xcite , are actually able to predict the neutrino mass spectrum with either @xmath0 or @xmath1 . * a special neutrino mixing pattern , the so - called tri - bimaximal mixing @xcite , @xmath26 is particularly favored . it yields @xmath27 ( or @xmath28 ) for the large - mixing - angle msw solution @xcite to the solar neutrino problem , @xmath29 ( or @xmath30 ) for the atmospheric neutrino oscillation , and @xmath31 . as a direct consequence of @xmath32 , the cp - violating phase @xmath13 is not well defined . this interesting neutrino mixing pattern is in general expected to result from an underlying flavor symmetry ( e.g. , the discrete @xmath33 @xcite , @xmath34 @xcite or @xmath35-@xmath36 @xcite symmetry ) in the lepton sector . such a symmetry must be broken spontaneously or explicitly , in order to account for both the observed lepton mass spectra and the realistic neutrino mixing pattern . one purpose of this paper is just to combine both lessons and reconstruct the simplest neutrino mass texture for either @xmath37 or @xmath1 . looking back to eqs . ( 1 ) and ( 2 ) , we find that @xmath38 and @xmath39 hold in the @xmath0 and @xmath1 cases , respectively . this observation implies that a realistic neutrino mass texture can be established by starting from the symmetry limit ( a ) @xmath2 and @xmath40 or ( b ) @xmath41 and @xmath1 . we shall show that it is possible to introduce a particular perturbation , which ensures the resultant neutrino mass matrix @xmath42 to reproduce the tri - bimaximal mixing pattern or its viable variations with all entries being formed from small integers and their square roots . the second purpose of this paper is to incorporate the texture of @xmath42 in the minimal type - ii seesaw model @xcite , an economical extension of the standard model with only one heavy right - handed majorana neutrino @xmath5 in addition to the @xmath6 higgs triplet @xmath7 . we shall focus our interest on the @xmath1 case , so as to obtain a non - vanishing cp - violating asymmetry in the lepton - number - violating decays of @xmath5 . such an asymmetry arises from the interference between the tree - level amplitude of @xmath5 decays and the one - loop vertex correction mediated by @xmath7 . following the idea of baryogenesis via leptogenesis @xcite and taking account of the flavor - dependent effects @xcite , we shall show that it is possible to interpret the observed baryon number asymmetry of the universe ( i.e. , @xmath43 @xcite ) via thermal leptogenesis in our model , in which only the electron flavor plays a role in the lepton - to - baryon conversion . the remaining part of this paper is organized as follows . in section ii we describe a purely phenomenological way to get viable variations of the tri - bimaximal neutrino mixing pattern from two simple textures of the neutrino mass matrix @xmath42 , one with @xmath0 and the other with @xmath1 . section iii is devoted to incorporating the texture of @xmath42 with @xmath1 and a non - trivial cp - violating phase in the minimal type - ii seesaw model , and to calculating the flavor - dependent leptogenesis in order to account for the cosmological baryon number asymmetry @xmath44 . a brief summary of our main results is presented in section iv . let us work in the basis where the flavor eigenstates of three charged leptons are identified with their mass eigenstates ( i.e. , the charged - lepton mass matrix @xmath45 is diagonal , real and positive ) . then the mass eigenstates of three neutrinos ( @xmath46 , @xmath47 , @xmath48 ) are directly linked to their flavor eigenstates ( @xmath49 , @xmath50 , @xmath51 ) through the neutrino mixing matrix @xmath9 . if @xmath9 is of the tri - bimaximal mixing pattern as given in eq . ( 3 ) , it can be decomposed into a product of two euler rotation matrices : @xmath52 , where @xmath53 with @xmath54 . allowing for small deviations of @xmath55 from unity , we are then left with some variations of the tri - bimaximal neutrino mixing pattern which can fit current or future neutrino oscillation data to a better degree of accuracy . our strategy of reconstructing the neutrino mass matrix @xmath42 is three - fold : ( 1 ) we take a proper symmetry limit of @xmath42 , denoted as @xmath56 , which can be diagonalized by the orthogonal transformation @xmath57 ; ( 2 ) we introduce a particular perturbation to @xmath56 , denoted as @xmath58 , which can be diagonalized by the orthogonal transformation @xmath59 ; ( 3 ) we require that @xmath60 should also be diagonalized by the transformation @xmath61 . of course , the texture of @xmath42 ought to guarantee either @xmath0 or @xmath1 . in the @xmath0 case , we observe from eq . ( 1 ) that @xmath62 holds . hence a reasonable symmetry limit of @xmath42 is expected to be @xmath63 where @xmath64 is assumed to be real and positive . the @xmath65 permutation symmetry in the @xmath66 sector of @xmath56 assures that this mass matrix can be diagonalized by the @xmath57 transformation : @xmath67 in other words , @xmath68 and @xmath69 hold in the chosen symmetry limit . a non - vanishing value of @xmath70 and a generalized tri - bimaximal neutrino mixing pattern can result from the perturbation @xmath71 where @xmath72 is a small dimensionless quantity , and @xmath55 is a positive number of @xmath73 . when @xmath74 holds , @xmath75 has the @xmath65 permutation symmetry in its @xmath76 sector . given the orthogonal transformations in eq . ( 4 ) , the diagonalization @xmath77 works . the neutrino mass matrix @xmath78 \ ; % ( 9)\ ] ] can then be diagonalized by the unitary matrix @xmath79 : @xmath80 three neutrino mass eigenvalues of @xmath42 turn out to be @xmath0 , @xmath81 and @xmath68 . taking account of eq . ( 1 ) , we immediately obtain the results @xmath82 ev and @xmath83 \approx 0.35/\left ( 2+x^2 \right ) $ ] . in the @xmath1 case , we observe from eq . ( 2 ) that @xmath84 holds . thus a reasonable symmetry limit of @xmath42 is expected to be @xmath85 where @xmath64 is also assumed to be real and positive . this neutrino mass matrix can similarly be diagonalized by the @xmath57 transformation : @xmath86 namely , @xmath87 and @xmath1 hold in the chosen symmetry limit . to break the degeneracy of @xmath88 and @xmath70 , we may introduce the same perturbation to @xmath56 as that given in eq . ( 7 ) , which can be diagonalized by the same transformation as that shown in eq . it is then possible to diagonalize the neutrino mass matrix @xmath89 \ ; % ( 13)\ ] ] by using the orthogonal matrix @xmath90 : @xmath91 in this case , three neutrino mass eigenvalues of @xmath42 are @xmath92 , @xmath93 c$ ] and @xmath94 . with the help of eq . ( 2 ) , one may easily arrive at @xmath95 ev and @xmath96 \approx 3.25 \times 10^{-2}/\left ( 2+x^2 \right ) $ ] . although the tri - bimaximal neutrino mixing pattern is of great interest , it is by no means unique in describing current neutrino oscillation data . hence we have gone beyond this pattern by allowing for @xmath97 in the above discussions . in both case ( a ) and case ( b ) , the neutrino mixing matrix @xmath90 reads @xmath98 it is obvious that @xmath9 takes the exact tri - bimaximal mixing pattern for @xmath54 . the allowed range of @xmath55 can be determined from that of @xmath10 through the relationship @xmath99 . in view of @xmath100 , which is obtained from a global analysis of current neutrino oscillation data @xcite , we easily arrive at @xmath101 . we see that the possibility of @xmath102 , which leads @xmath9 to the bimaximal neutrino mixing , has clearly been excluded . on the other hand , @xmath54 seems to be the simplest and most favored possibility . within the allowed range of @xmath55 , it is not difficult to find out some viable variations of the tri - bimaximal neutrino mixing pattern . in particular , we pay interest to such a category of neutrino mixing matrices @xmath9 : the entries of @xmath9 are all formed from small integers and their square roots , which are often suggestive of a certain flavor symmetry in the language of group theories . below are three examples : * @xmath103 , corresponding to @xmath104 and @xmath105 * @xmath106 , corresponding to @xmath107 and @xmath108 * @xmath109 , corresponding to @xmath110 and @xmath111 the pattern of @xmath9 in eq . ( 16 ) is especially interesting , because all of its nine elements are formed from four smallest integers @xmath112 , @xmath113 , @xmath114 , @xmath115 and their square roots . this pattern has actually been conjectured in ref . @xcite , but here we illustrate how it can be obtained from @xmath42 . now let us consider how to derive the neutrino mass matrix @xmath42 in eq . ( 9 ) or eq . ( 13 ) from a specific seesaw model . one may naively expect that the minimal seesaw model with two heavy right - handed majorana neutrinos @xcite is a good candidate , because it naturally assures that @xmath42 is of rank 2 and has a vanishing mass eigenvalue ( either @xmath0 or @xmath1 ) . taking account of the fact that @xmath42 is composed of two mass matrices @xmath56 and @xmath58 , however , we find that it is more natural to incorporate @xmath42 in the minimal type - ii seesaw model with only one heavy right - handed majorana neutrino @xmath5 in addition to the @xmath6 higgs triplet @xcite . in this case , the neutrino mass term can be written as @xmath116 where @xmath117 denotes the column vector of @xmath118 fields , @xmath119 is a @xmath120 matrix arising from the leptonic yukawa interaction induced by the higgs triplet @xmath7 , @xmath121 is a @xmath122 matrix arising from the leptonic yukawa interaction induced by the higgs doublet @xmath123 , and @xmath124 is just the mass of the right - handed majorana neutrino @xmath5 . provided @xmath125 is considerably higher than the mass scale of @xmath121 , one may obtain the effective ( left - handed ) majorana neutrino mass matrix @xmath42 from eq . ( 19 ) via the well - known type - ii seesaw mechanism @xcite : @xmath126 . comparing this formula with @xmath127 , we arrive at @xmath128 the texture of @xmath129 given in eq . ( 5 ) or eq . ( 11 ) may easily be obtained from certain flavor symmetries ( such as the discrete @xmath35-@xmath36 @xcite or @xmath65 @xcite symmetry ) . on the other hand , the texture of @xmath58 in eq . ( 7 ) can be derived from eq . ( 20 ) with a unique form of @xmath121 , @xmath130 together with @xmath131 . we remark that such a seesaw realization of the texture of @xmath42 does not involve any parameter fine - tuning or cancellation , and thus it is quite natural . more interestingly , the minimal type - ii seesaw model under consideration can offer a possibility of understanding the cosmological baryon number asymmetry via thermal leptogenesis @xcite . for simplicity , we assume the mass of @xmath7 is much higher than that of @xmath5 such that the cp - violating asymmetry in the out - of - equilibrium decays of @xmath5 is in practice the only source of leptogenesis . we allow @xmath55 to be complex in @xmath121 and its imaginary part is just responsible for cp violation in the model . in the @xmath0 case , a straightforward analysis shows that @xmath132 holds due to the special textures of @xmath119 in eq . ( 5 ) and @xmath121 in eq . ( 21 ) , implying the absence of cp violation in the decays of @xmath5 . hence we shall focus our interest on the @xmath1 case in the following . as @xmath55 is now taken to be a complex parameter , the diagonalization of @xmath42 in eq . ( 13 ) turns out to be quite non - trivial . we need a unitary matrix @xmath9 to make the transformation @xmath133 . we obtain two non - vanishing mass eigenvalues as @xmath134 where @xmath135 in addition , @xmath136 is just the neutrino mixing matrix , where @xmath137 with @xmath138 and @xmath139 \ ; , \nonumber \\ \sigma & = & \frac{1}{2 } \arg \left [ \left ( 2 + \varepsilon x^2 \right ) { z_{2}^{}}^2 + 4 \varepsilon x z_{2}^ { } + 4 \left ( 1 + \varepsilon \right ) \right ] \ ; . % ( 27)\end{aligned}\ ] ] three neutrino mixing angles are @xmath140 $ ] , @xmath141 and @xmath142 . because of @xmath1 , only the difference between @xmath14 and @xmath15 is a physical majorana cp - violating phase . if @xmath55 is real , then there will be no cp violation and eq . ( 24 ) will be simplified to eq . ( 15 ) . the cp - violating asymmetry @xmath154 can give rise to a net lepton number asymmetry in the universe , and this lepton number asymmetry can partially be converted into a net baryon number asymmetry due to non - perturbative sphaleron interactions @xcite @xmath155 where the efficiency factors @xmath156 ( for @xmath157 ) measure the flavor - dependent washout effects associated with the out - of - equilibrium decays of @xmath5 . to evaluate the size of @xmath158 in eq . ( 30 ) , one may introduce a parameter @xmath159 , where @xmath160 with @xmath161 being the effective neutrino mass and @xmath162 being the equilibrium neutrino mass @xcite . explicitly , @xmath163 and @xmath164 hold . since the relationship between @xmath158 and @xmath165 is rather complicated , we do not write it out here but refer the reader to ref . @xcite for details . we just mention that a numerical analysis yields @xmath166 and @xmath167 in our model . hence @xmath168 should be of @xmath169 and have a minus sign , in order to correctly reproduce @xmath170 . let us count the number of free parameters relevant to neutrino masses , flavor mixing and leptogenesis . they are @xmath64 , @xmath72 , @xmath171 , @xmath172 and @xmath125 . on the other hand , we have four observable quantities @xmath88 , @xmath70 , @xmath10 and @xmath44 which depend on the magnitudes of those five parameters . although five free parameters can not be fully determined from four measured quantities , it is possible to constrain the former by using the latter . we shall carry out a numerical calculation and illustrate the viable parameter space of this minimal type - ii seesaw model in the next subsection . given @xmath1 , @xmath173 and @xmath174 in our model , the inputs of our numerical calculation include @xmath175 , @xmath176 , @xmath177 @xcite and @xmath178 @xcite . since @xmath171 is expected to be of @xmath73 , we typically take @xmath179 as a reasonable upper limit . then it is straightforward to obtain the allowed ranges of @xmath64 , @xmath72 , @xmath171 , @xmath172 and @xmath125 with the help of those analytical expressions given in eqs . ( 22)(30 ) . we demonstrate that this minimal type - ii seesaw model can simultaneously account for current neutrino oscillation data and the cosmological baryon number asymmetry . some results and discussions are in order . * first of all , we find that @xmath64 gradually ranges between @xmath180 and @xmath181 . there exist a lower bound on @xmath171 and an upper bound on @xmath72 ; namely , @xmath182 and @xmath183 . the correlated parameter space of @xmath171 and @xmath72 is shown in fig . 1 , from which one can get a number of nearly tri - bimaximal neutrino mixing patterns . * 2 illustrates the correlated parameter space of @xmath125 and @xmath172 , which can roughly be understood if one takes into account @xmath184 $ ] as indicated in eq . ( 29 ) . neglecting the influence of @xmath171 on @xmath168 , we find that the lower bound on @xmath125 comes out around @xmath185 ( or equivalently , @xmath186 \sim 1 $ ] ) : @xmath187 gev . when @xmath172 approaches zero , @xmath125 has to approach infinity in order to assure @xmath188 . in our model with @xmath189 , the favored range of @xmath125 is actually between @xmath190 gev and @xmath191 gev . * note that @xmath192 is a particularly interesting possibility , corresponding to @xmath193 ev , @xmath194 , @xmath195 and @xmath196 gev . in this special case , the model is simplified to a unique version which only contains four free parameters and they can all be determined from the experimental values of @xmath88 , @xmath70 , @xmath10 and @xmath44 . more accurate data will impose much narrower constraints on the model parameters @xmath64 , @xmath72 , @xmath172 and @xmath125 . finally , it is worthwhile to point out that our neutrino mixing pattern is stable against radiative corrections . running from the seesaw scale @xmath197 down to the electroweak scale @xmath198 , @xmath199 keeps unchanged while other two neutrino masses and three mixing angles can only receive tiny corrections in the standard model @xcite . although the so - called dirac cp - violating phase @xmath13 can be generated together with @xmath12 from radiative corrections @xcite , the resultant cp - violating effect in neutrino oscillations ( characterized by the jarlskog invariant of @xmath169 or smaller in this model ) is too small to be observable . in summary , we have proposed a new category of neutrino mass ans@xmath200tze by starting from a combination of two phenomenological observations : ( 1 ) the lightest neutrino mass might be zero or vanishingly small , and ( 2 ) the neutrino mixing matrix might be the tri - bimaximal mixing pattern or a pattern close to it . we have shown that a realistic neutrino mass matrix @xmath42 can be established either in the limit of @xmath201 and @xmath40 or in the limit of @xmath202 and @xmath1 , corresponding to the possibility of @xmath0 or @xmath1 . in both cases , it is possible to introduce a particular perturbation which ensures the resultant neutrino mixing matrix to be the tri - bimaximal mixing pattern or its viable variations with all entries being formed from small integers and their square roots . we have incorporated the texture of @xmath42 in the minimal type - ii seesaw model with only one heavy right - handed majorana neutrino @xmath5 in addition to the @xmath6 higgs triplet @xmath7 . the @xmath1 case has been discussed in detail to accommodate cp violation in the lepton - number - violating decays of @xmath5 . we have demonstrated that our model can simultaneously interpret current neutrino oscillation data and the cosmological baryon number asymmetry via thermal leptogenesis , in which only the electron flavor plays a role in the lepton - to - baryon conversion . finally let us remark that both the neutrino mass spectrum and the flavor mixing angles are well fixed in the proposed model . it is therefore easy to test them in the near future , when more accurate experimental data are available . _ acknowledgments : _ this work was supported by the nus research grant no wbs : r-144 - 000 - 178 - 112 ( a.h.c . ) and by the national natural science foundation of china ( z.z.x . ) . z.z.x . is indebted to a.h . chan and c.h . oh for warm hospitality at the nus . s.l . and z.z.x . are grateful to s. zhou for useful discussions . kamland collaboration , k. eguchi _ et al . _ , lett . * 90 * , 021802 ( 2003 ) ; chooz collaboration , m. apollonio _ et al . _ , b * 420 * , 397 ( 1998 ) ; palo verde collaboration , f. boehm _ et al . _ , lett . * 84 * , 3764 ( 2000 ) . harrison , d.h . perkins , and w.g . scott , phys . b * 530 * , 167 ( 2002 ) ; z.z . xing , phys . b * 533 * , 85 ( 2002 ) ; p.f . harrison and w.g . scott , phys . b * 535 * , 163 ( 2002 ) ; x.g . he and a. zee , phys . b * 560 * , 87 ( 2003 ) . see , e.g. , e. ma , phys . rev . d * 70 * , 031901 ( 2004 ) ; new j. phys . * 6 * , 104 ( 2004 ) ; g. altarelli and f. feruglio , nucl . b * 720 * , 64 ( 2005 ) ; nucl . b * 741 * , 215 ( 2006 ) ; hep - ph/0610165 ; k.s . babu and x.g . he , hep - ph/0507217 ; a. zee , phys . b * 630 * , 58 ( 2005 ) ; x.g . keum , and r.r . volkas , jhep * 0604 * , 039 ( 2006 ) ; e. ma , h. sawanaka , and m. tanimoto , phys . b * 641 * , 301 ( 2006 ) ; y. koide , hep - ph/0701018 ; f. feruglio , c. hagedorn , y. lin , and l. merlo , hep - ph/0702194 . see , e.g. , p.f . harrison and w.g . scott , phys . b * 557 * , 76 ( 2003 ) ; w. grimus and l. lavoura , jhep * 0508 * , 013 ( 2005 ) ; j.e . kim and j.c . park , jhep * 0605 * , 017 ( 2006 ) ; r.n . mohapatra , s. nasri , and h.b . yu , phys . lett . b * 639 * , 318 ( 2006 ) ; n. haba , a. watanabe , and k. yoshioka , phys . lett . * 97 * , 041601 ( 2006 ) ; y. koide , phys . rev . d * 73 * , 057901 ( 2006 ) ; hep - ph/0605074 ; hep - ph/0612058 ; s. kaneko , h. sawanaka , t. shingai , m. tanimoto , and k. yoshioka , prog . . phys . * 117 * , 161 ( 2007 ) . see , e.g. , t. fukuyama and h. nishiura , hep - ph/9702253 ; r.n . mohapatra and s. nussinov , phys . d * 60 * , 013002 ( 1999 ) ; z.z . xing , phys . d * 61 * , 057301 ( 2000 ) ; phys . d * 64 * , 093013 ( 2001 ) ; e. ma and m. raidal , phys . rev . lett . * 87 * , 011802 ( 2001 ) ; c.s . lam , phys . b * 507 * , 214 ( 2001 ) ; p.f . harrison and w.g . scott , phys . b * 547 * , 219 ( 2002 ) ; t. ohlsson and g. seidl , nucl . b * 643 * , 247 ( 2002 ) ; t. kitabayashi and m. yasu , phys . d * 67 * , 015006 ( 2003 ) ; w. grimus and l. lavoura , phys . b * 572 * , 189 ( 2003 ) ; j. phys . g * 30 * , 73 ( 2004 ) ; y. koide , phys . rev . d * 69 * , 093001 ( 2004 ) ; r.n . mohapatra , jhep * 0410 * , 027 ( 2004 ) ; a. de gouvea , phys . d * 69 * , 093007 ( 2004 ) ; a. ghosal , mod . lett a * 19 * , 2579 ( 2004 ) ; w. grimus , a.s . joshipura , s. kaneko , l. lavoura , h. sawanaka , and m. tanimoto , nucl . b * 713 * , 151 ( 2005 ) ; r.n . mohapatra and w. rodejohann , phys . d * 72 * , 053001 ( 2005 ) ; t. kitabayashi and m. yasu , phys . b * 621 * , 133 ( 2005 ) ; r.n . mohapatra , s. nasri , and h.b . yu , phys . b * 615 * , 231 ( 2005 ) ; f. plentinger and w. rodejohann , phys . b * 625 * , 264 ( 2005 ) ; i. aizawa , t. kitabayashi , and m. yasu , nucl . phys . b * 728 * , 220 ( 2005 ) ; k. matsuda and h. nishiura , phys . d * 73 * , 013008 ( 2006 ) ; y.h . ahn , s.k . kang , c.s . kim , and j. lee , phys . d * 73 * , 093005 ( 2006 ) ; r.n . mohapatra , s. nasri , and h.b . yu , phys . b * 636 * , 114 ( 2006 ) ; z.z . xing , phys . d * 74 * , 013010 ( 2006 ) ; k. fuki and m. yasue , phys . d * 73 * , 055014 ( 2006 ) . r. friedberg and t.d . lee , hep@xmath203np * 30 * , 591 ( 2006 ) ; z.z . xing , h. zhang , and s. zhou , phys . b * 641 * , 189 ( 2006 ) ; s. luo and z.z . xing , phys . b * 646 * , 242 ( 2007 ) ; z.z . xing , hep - ph/0703007 ; y.h . ahn , s.k . kang , c.s . kim , and j. lee , phys . d * 75 * , 013012 ( 2007 ) . r. barbieri , p. creminelli , a. strumia , and n. tetradis , nucl . b * 575 * , 61 ( 2000 ) ; t. endoh , t. morozumi , and z.h . xiong , prog . phys . * 111 * , 123 ( 2004 ) ; t. fujihara , s. kaneko , s. kang , d. kimura , t. morozumi , and m. tanimoto , phys . d * 72 * , 016006 ( 2005 ) ; a. pilaftsis and t.e.j . underwood , phys . d * 72 * , 113001 ( 2005 ) ; o. vives , phys . d * 73 * , 073006 ( 2006 ) ; a. abada , s. davidson , f.x . josse - michaux , m. losada , and a. riotto , jcap * 0604 * , 004 ( 2006 ) ; e. nardi , y. nir , e. roulet , and j. racker , jhep * 0601 * , 164 ( 2006 ) ; a. abada , s. davidson , f.x . josse - michaux , m. losada , and a. riotto , hep - ph/0605281 ; z.z . xing and s. zhou , hep - ph/0607302 ; s. blanchet and p. di bari , jcap * 0606 * , 023 ( 2006 ) ; s. antusch , s.f . king , and a. riotto , jcap * 0611 * , 011 ( 2006 ) ; s. pascoli , s.t . petcov , and a. riotto , hep - ph/0609125 ; g.c . branco , r. gonzalez felipe , and f.r . joaquim , phys . b * 645 * , 432 ( 2007 ) ; s. antusch and a.m. teixeira , jcap * 0702 * , 024 ( 2007 ) . mohapatra and g. senjanovic , phys . lett . * 44 * , 912 ( 1980 ) ; j. schechterm and j.w.f . valle , phys . d * 22 * , 2227 ( 1980 ) ; m. magg and c. wetterich , phys . b 94 , 61 ( 1980 ) ; g. lazarides , q. shafi and c. wetterich , nucl . phys . b * 181 * , 287 ( 1981 ) . see , e.g. , s. antusch , j. kersten , m. lindner , and m. ratz , nucl . b * 674 * , 401 ( 2003 ) ; j.w . mei and z.z . xing , phys . d * 69 * , 073003 ( 2004 ) ; s. luo and z.z . xing , phys . b * 632 * , 341 ( 2006 ) .	current experimental data allow the zero value for one neutrino mass , either @xmath0 or @xmath1 . this observation implies that a realistic neutrino mass texture can be established by starting from the limit ( a ) @xmath2 and @xmath3 or ( b ) @xmath4 and @xmath1 . in both cases , we may introduce a particular perturbation which ensures the resultant neutrino mixing matrix to be the tri - bimaximal mixing pattern or its viable variations with all entries being formed from small integers and their square roots . we find that it is natural to incorporate this kind of neutrino mass matrix in the minimal type - ii seesaw model with only one heavy right - handed majorana neutrino @xmath5 in addition to the @xmath6 higgs triplet @xmath7 . we show that it is possible to account for the cosmological baryon number asymmetry in the @xmath1 case via thermal leptogenesis , in which the one - loop vertex correction to @xmath5 decays is mediated by @xmath7 and the cp - violating asymmetry of @xmath5 decays is attributed to the electron flavor .
our starting point is a theorem of bishop and jones , stated below , which roughly says that a connected subset of @xmath9 that is uniformly non - flat in every ball centered upon it ( or in other words , is very wiggly " ) , must have large dimension . we measure flatness with jones @xmath10-numbers : if @xmath11 is a subset of a hilbert space @xmath12 , @xmath13 and @xmath14 , we define @xmath15 where the infimum is taken over all lines @xmath16 . ( ( * ? ? ? * theorem 1.1 ) ) there is a constant @xmath4 such that the following holds . let @xmath17 be a compact connected set and suppose that there is @xmath18 such that for all @xmath19 and all @xmath13 , @xmath20 . then the hausdorff dimension for the definition of hausdorff dimension and other definitions and notation . ] of @xmath11 satisfies @xmath21 . [ t : bj ] there are also analogues of theorem [ t : bj ] for surfaces of higher topological dimension , see for example @xcite . our main theorem extends this result to the metric space setting using an alternate definition of @xmath10 . before stating our results , however , we discuss the techniques and steps involved in proving theorem [ t : bj ] to elucidate why the original methods do nt immediately carry over , and to discuss how they must be altered for the metric space setting . + the main tool in proving theorem [ t : bj ] is the _ analyst s traveling salesman theorem _ , which we state below . first recall that for a metric space @xmath22 , a _ maximal @xmath23-net _ is a maximal collection of points @xmath24 such that @xmath25 for all @xmath26 . ( ( * ? ? ? * theorem 1.1 ) ) let @xmath27 , @xmath11 be a compact subset of a hilbert space @xmath12 , and @xmath28 be a nested sequence of maximal @xmath29-nets in @xmath11 . for @xmath27 , define @xmath30 there is @xmath31 such that for @xmath32 there is @xmath33 ( depending only on @xmath34 ) so that for any @xmath11 , @xmath35 implies there is a connected set @xmath36 such that @xmath37 and @xmath38 conversely , if @xmath36 is connected and @xmath39 , then for any @xmath27 , @xmath40 [ t : tst ] at the time of @xcite , this was only known for the case @xmath41 , due to jones @xcite . this was subsequently generalized to @xmath42 by okikiolu @xcite and then to hilbert space by schul @xcite . the proof of theorem [ t : bj ] goes roughly as follows : one constructs a _ frostmann measure _ @xmath43 supported on @xmath11 satisfying @xmath44 for some @xmath45 , @xmath46 and for all @xmath13 and @xmath14 . this easily implies that the hausdorff dimension of @xmath11 is at least @xmath47 ( see ( * ? ? ? * theorem 8.8 ) and that section for a discussion on frostmann measures ) . one builds such a measure on @xmath11 inductively by deciding the values @xmath48 for each dyadic cube @xmath49 intersecting @xmath11 and for each @xmath50-th generation descendant @xmath51 intersecting @xmath11 , where @xmath50 is some large number that will depend on @xmath52 . if the number of such @xmath50-th generation descendants is large enough , we can choose the ratios and hence disseminate the mass @xmath53 amongst the descendants @xmath54 in such a way that the ratios will be very small and will be satisfied . to show that there are enough descendants , one looks at the skeletons of the @xmath50-th generation descendants of @xmath49 and uses the second half of theorem [ t : tst ] coupled with the non - flatness condition in the satement of theorem [ t : bj ] to guarantee that the total length of this skeleton ( and hence the number of cubes ) will be large . in the metric space setting , however , no such complete analogue of theorem [ t : tst ] exists , and it is not even clear what the appropriate analogue of a @xmath10-number should be . note , for example , that it does not make sense to estimate the length of a metric curve @xmath36 using the original @xmath10-number , even if we consider @xmath36 as lying in some banach space . a simple counter example is if @xmath55)$ ] is the image of @xmath56\rightarrow l^{1}([0,1])$ ] defined by @xmath57}$ ] . this a geodesic , so in particular , it is a rectifiable curve of finite length . however , @xmath58 ( i.e. the width of the smallest tube containing @xmath59 in @xmath60 , rescaled by a factor @xmath61 ) is uniformly bounded away from zero , and in particular , @xmath62 . in @xcite , hahlomaa gives a good candidate for a @xmath10-number for a general metric space @xmath2 using menger curvature and uses it to show that if the sum in is finite for @xmath63 ( using his definition of @xmath64 ) , then it can be contained in the lipschitz image of a subset of the real line ( analogous to the first half of theorem [ t : tst ] ) . an example of schul @xcite , however , shows that the converse of theorem [ t : tst ] is false in general : with hahlomaa s @xmath64 does not hold with the same constant for all curves in @xmath65 . we refer to @xcite for a good summary on the analyst s traveling salesman problem . to generalize theorem [ t : bj ] , we use a @xmath10-type quantity that differs from both jones and hahlomaa s definitions . it is inspired by one defined by bishop and tyson in @xcite that measures the deviation of a set from a geodesic in a metric space : if @xmath2 is a metric space , @xmath66 , and @xmath67 an ordered sequence , define @xmath68 and define @xmath69 where the infimum is over all finite ordered sequences in @xmath70 of any length @xmath50 . in @xcite , bishop and tyson ask whether , for a compact connected metric space @xmath2 , being uniformly larger than zero is enough to guarantee that @xmath71 . we answer this in the affirmative . there is @xmath72 such that the following holds . if @xmath2 is a compact connected metric space and @xmath73 for all @xmath6 and @xmath74 for some @xmath18 , then @xmath75 . [ t : bt - answer ] instead of @xmath76 , however , we work with a different quantity , which we define here for a general compact metric space @xmath2 . first , by kuratowski embedding theorem , we may assume @xmath2 is a subset of @xmath77 , whose norm we denote by @xmath78 . let @xmath79 and define @xmath80))}{\|s(0)-s(1)\|}\ ] ] where the infimum is over all curves @xmath56\rightarrow b(x , r)\subseteq \ell^{\infty}$ ] and @xmath81 is the length of @xmath47 , where the supremum is over all partitions @xmath82 . in general , if @xmath47 is defined on a union of disjoint open intervals @xmath83 , we set @xmath84 the case in which @xmath47 is just a straight line segment through the center of the ball with length @xmath85 gives the estimate @xmath86 . the quantity @xmath1 measures how well @xmath87 may be approximated by a geodesic . to see this , note that if , for some @xmath56\rightarrow\ell^{\infty}$ ] , the @xmath88-neighborhood of @xmath89)$ ] contains @xmath87 , then the length of @xmath47 must be at least @xmath90 , which is @xmath88 more than the length of any geodesic connecting @xmath91 and @xmath92 . the quantity @xmath93 similarly measures how well the portion of @xmath87 may be approximated by a geodesic polygonal path with vertices in @xmath2 . in figure [ f : betas ] , we compare the meanings of @xmath94 and @xmath95 . we will refer to the quantities @xmath96 and @xmath97 as the _ geodesic deviation _ of @xmath47 and @xmath98 respectively . we will also say @xmath99 and @xmath100 measure the _ geodesic deviation _ of @xmath2 inside the ball @xmath3 . ( 100,300)(130,0 ) ( 0,0 ) containing a portion of a curve @xmath2 . in the first picture , @xmath101 is the width of the smallest tube containing @xmath87 . in the second , we see that @xmath102 is such that for @xmath103 , there are @xmath104 with vertices in @xmath105 so that balls centered on the @xmath106 of radius @xmath107 cover @xmath105 , and so that the geodesic deviation ( that is , its length minus @xmath108 is at most @xmath107 . in the last , we show that if @xmath109 , there is @xmath56\rightarrow \ell^{\infty}$ ] whose geodesic deviation and whose distance from any point in @xmath105 are both at most @xmath110.,title="fig : " ] ( 55,200 ) @xmath101 ( 45,105)@xmath111 ( 85,20)@xmath112 ( 240,15)@xmath113 ( 220,85)@xmath114 ( 285,38)@xmath115)}$ ] ( 275,200)@xmath116 ( 90,165)@xmath2 note that for the image of @xmath117}\in l^{1}([0,1])$ ] described earlier , it is easy to check that @xmath118 for all @xmath6 and @xmath14 , even though @xmath119 is bounded away from zero . this , of course , makes the terminology wiggly " rather misleading in metric spaces , since there are certainly non - flat or highly wiggly " geodesics in @xmath60 ; we use this terminology only to be consistent with the literature . later on in proposition [ p : bb ] , however , we will show that in a hilbert space we have for some @xmath45 , @xmath120 that the two should be correlated in this setting seems natural as @xmath121 is measuring how far @xmath2 is deviating from a straight line , which are the only geodesics in hilbert space . in lemma [ l : beta-bhat ] below , we will also show that for some @xmath45 , @xmath122 so that theorem [ t : bt - answer ] follows from the following theorem , which is our main result . there is @xmath123 such that the following holds . if @xmath2 is a compact connected metric space and @xmath124 for all @xmath6 and @xmath74 for some @xmath18 , then @xmath125 . [ t : main ] we warn the reader , however , that the quadratic dependence on @xmath10 appears in theorem [ t : main ] and theorem [ t : bj ] for completely different reasons . in theorem [ t : bj ] , it comes from using theorem [ t : tst ] , or ultimately from the pythagorean theorem , which of course does no hold in general metric spaces ; in theorem [ t : main ] , it seems to be an artifact of the construction and can perhaps be improved . our approach to proving theorem [ t : main ] follows the original proof of theorem [ t : bj ] described earlier : to show that a metric curve @xmath126 has large dimension , we approximate it by a polygonal curve , estimate its length from below and use this estimate to construct a frostmann measure , but in lieu of a traveling salesman theorem . ( in fact , taking @xmath127 instead of @xmath128 in theorem [ t : tst ] does not lead to a metric version of theorem [ t : tst ] for a similar reason that hahlomaa s @xmath10-number does nt work ; one need only consider schul s example ( * ? ? ? * section 3.3.1 ) . ) + the original context of bishop and tyson s conjecture , and the motivation for theorem [ t : main ] , concerned conformal dimension . recall that a _ quasisymmetric map _ @xmath129 between two metric spaces is a map for which there is an increasing homeomorphism @xmath130 such that for any distinct @xmath131 , @xmath132 the _ conformal dimension _ of a metric space @xmath2 is @xmath133 where the infimum ranges over all quasisymmetric maps @xmath134 . for more information , references , and recent work on conformal dimension , see for example @xcite . in @xcite , it is shown that the antenna set has conformal dimension one yet every quasisymmetric image of it into any metric space has dimension strictly larger than one . the _ antenna set _ is a self similar fractal lying in @xmath135 whose similarities are the following : @xmath136 where @xmath137 is some fixed angle ( see figure [ f : antenna ] ) . . ] to show the conformal dimension @xmath138 is never attained under any quasisymmetric image of the antenna set , the authors show by hand that any quasisymmetic map of the antenna set naturally induces a frostmann measure of dimension larger than one . at the end of the paper , however , the authors suggested another way of showing the same result by proving an analogue of theorem [ t : bj ] for a @xmath10-number which is uniformly large for the antenna set as well as any quasisymmetric image of it . theorem [ t : main ] does nt just give a much longer proof of bishop and tyson s result , but it lends itself to more general sets lacking any self - similar structure . let @xmath4 , @xmath139\cup [ 0,e_{2}]\cup [ 0,e_{3}]\subseteq \br^{3}$ ] , where @xmath140 is the @xmath141th standard basis vector in @xmath142 , and let @xmath2 be a compact connected metric space . for @xmath6 , @xmath14 , we say @xmath70 has a _ @xmath143-antenna _ if there is a homeomorphism @xmath144 such that the distance between @xmath145 and @xmath146\cup [ 0,e_{k}]))$ ] is at least @xmath147 for all permutations @xmath148 of @xmath149 . we say @xmath2 is _ @xmath143-antenna - like _ if @xmath70 has a @xmath143-antenna for every @xmath6 and @xmath150 , clearly , the classical antenna set in @xmath9 is antenna - like . let @xmath126 be a compact connected metric space in @xmath77 . 1 . if @xmath70 has a @xmath143-antenna , then @xmath151 . hence , if @xmath126 is @xmath143-antenna - like , we have @xmath152 . any quasisymmetric image of an antenna - like set into any metric space is also antenna - like and hence has dimension strictly larger than one . [ t : antenna - like ] note that this result does nt say the conformal dimension of an antenna - like set is larger than one , only that no quasisymmetric image of it has dimension equal to one . however , see @xcite , where the author bounds the conformal dimension of a set from below using a different quantity . in section [ s : prelims ] , we go over some necessary notation and tools before proceeding to the proof of theorem [ t : main ] in section [ s : proof ] . in section [ s : antenna ] , we prove theorem [ t : antenna - like ] , and in section [ s : betas ] we compare @xmath153 and @xmath10 . the author would like to thank steffen rohde , tatiana toro , and jeremy tyson for their helpful discussions , and to matthew badger , john garnett , raanan schul , and the anonymous referee for their helpful comments on the manuscript . part of this manuscript was written while the author was at the ipam long program interactions between analysis and geometry , spring 2013 . since we are only dealing with compact metric spaces , by the kuratowski embedding theorem , we will implicitly assume that all our metric spaces are contained in @xmath77 , whose norm we will denote @xmath154 . for @xmath155 and @xmath14 , we will write @xmath156 if @xmath157 and @xmath158 , we write @xmath159 for @xmath160 . for a set @xmath161 and @xmath162 , define @xmath163 where @xmath164 for a set @xmath165 , let @xmath166 denote its lebesgue measure . for an interval @xmath167 , we will write @xmath168 and @xmath169 for its left and right endpoints respectively . for @xmath170 , @xmath171 $ ] and @xmath161 , define @xmath172 @xmath173 the _ hausdorff dimension _ of a set @xmath34 is @xmath174 in this section , we construct a family of subsets of @xmath77 , tailored to a metric space @xmath2 , that have properties similar to dyadic cubes in euclidean space . these cubes appeared in @xcite ( where they were alternatively called cores " ) and are similar to the so - called christ - david cubes ( @xcite ) in some respects , although they are not derived from them . fix @xmath175 and @xmath176 . let @xmath177 be a nested sequence of maximal @xmath178-nets in @xmath2 . let @xmath179 for @xmath180 , define @xmath181 basically , @xmath182 is the union of all balls @xmath183 that may be connected to @xmath184 by a chain @xmath185 with @xmath186 , @xmath187 , and @xmath188 for all @xmath141 . for such a cube @xmath49 constructed from @xmath189 , we let @xmath190 and @xmath191 . let @xmath192 note that , for @xmath193 , @xmath194 . if @xmath195 , then for @xmath2 and @xmath196 as above , the family of cubes @xmath196 satisfy the following properties . 1 . if @xmath197 and @xmath198 , then @xmath199 or @xmath200 . 2 . for @xmath201 , @xmath202 [ l : cubes ] the proof is essentially in @xcite , but with slightly different parameters . so that the reader need not perform the needed modifications , we provide a proof here . part 1 follows from the definition of the cubes @xmath49 . to prove part 2 , we first claim that if @xmath203 is a chain of balls with centers @xmath204 for which @xmath205 , then for @xmath206 , @xmath207 we prove by induction . let @xmath204 denote the center of @xmath208 if @xmath209 , @xmath210 , and @xmath211 and @xmath212 are the centers of @xmath213 and @xmath214 respectively , then @xmath215 since otherwise @xmath216 for some @xmath217 and @xmath218 since @xmath195 , which is a contradiction . hence , @xmath219 now suppose @xmath220 . let @xmath221 and @xmath217 be an integer so that @xmath222 recall that all balls in @xmath223 have radii that are powers of @xmath224 , so there exists an @xmath217 so that the above happens . note that @xmath225 and @xmath226 can not have the same diameter ( which follows from the @xmath209 case we proved earlier ) . since @xmath226 has the maximum diameter of all the @xmath208 , we in fact know that @xmath227 ( again , recall that all balls have radii that are powers of @xmath224 ) . let @xmath228 be the minimal integer for which @xmath229 ( which exists by the previous discussion ) and let @xmath230 be the maximal integer such that @xmath231 . by the induction hypothesis , @xmath232 and @xmath233 so that @xmath234 @xmath235 . note that if @xmath236 , then @xmath237 for @xmath238 and @xmath239 ( this makes @xmath240 ) . since @xmath241 and points in @xmath242 are @xmath243-separated , we must have @xmath244 , hence @xmath245 . thus , @xmath246 which contradicts the minimality of @xmath247 , hence @xmath235 . we can prove similarly that @xmath248 , and this with proves . this in turn implies that for any @xmath249 , if @xmath250 , then @xmath251 , hence @xmath252 for @xmath217 large enough , this means we can pick our cubes so that they do nt differ much from balls . we will set @xmath253 for some @xmath254 to be determined later , so that @xmath255 there are a few different constructions of families of metric subsets with properties similar to dyadic cubes , see @xcite , @xcite , and @xcite for example , and the references therein . readers familiar with any of these references will see that schul s cores " we have just constructed are very different from the cubes constructed in the aforementioned references . in particular , each @xmath256 does not partition any metric space in the same way that dyadic cubes ( half - open or otherwise ) would partition euclidean space , not even up to set of measure zero ) . however , for each @xmath50 we do have @xmath257 and we still have the familiar intersection properties in lemma [ l : cubes ] . the reason for the ad hoc construction is the crucial roundness " property . let @xmath258\rightarrow \ell^{\infty}$ ] be a piecewise linear function and set @xmath259)$ ] , whose image is a finite union of line segments , and let @xmath196 be the cubes from lemma [ l : cubes ] tailored to @xmath2 . then for any @xmath201 , @xmath260 and @xmath261 . [ l : zero - boundary ] note that since @xmath36 is a finite polynomial curve , @xmath262 is _ doubling _ on @xmath36 , meaning there is a constant @xmath263 so that @xmath264 for all @xmath265 and @xmath14 . if @xmath266 for some @xmath201 , then there is a sequence @xmath267 such that @xmath268 since the @xmath269 are maximal @xmath178-nets . to each @xmath270 corresponds a ball @xmath271 . let @xmath217 be such that @xmath250 . since @xmath272 , we have by lemma [ l : cubes ] that either @xmath273 ( if @xmath274 ) or @xmath275 for some @xmath276 with @xmath277 . in either case , since cubes do nt contain their boundaries ( since they are open ) , we have that @xmath278 . this implies that @xmath49 is porous , and it is well known that such sets have doubling measure zero . more precisely , the doubling condition on @xmath43 guarantees that @xmath279 @xmath43-a.e . @xmath265 ( see ( * ? ? ? * theorem 1.8 ) ) , but if @xmath280 and @xmath281 is as above , then one can show using the doubling property of @xmath43 that @xmath282 and thus @xmath283 . the last part of the theorem follows easily since @xmath284 is piecewise affine . the following lemma will be used frequently . let @xmath167 be an interval , @xmath285 be continuous and @xmath286 a subinterval . then @xmath287 [ l : subarc ] we may assume @xmath288 , otherwise is trivial . we estimate @xmath289 for this section , we fix a compact connected set @xmath2 satisfying the conditions of theorem [ t : main ] . the main tool is the following lemma , which can be seen as a very weak substitute for theorem [ t : tst ] . let @xmath290 . we can pick @xmath291 large enough ( by picking @xmath292 small enough ) and pick @xmath293 such that , for any @xmath2 satisfying the conditions of theorem [ t : main ] for some @xmath294 , the following holds . if @xmath269 is any nested sequence of @xmath178-nets in @xmath2 , there is @xmath295 such that for @xmath296 with @xmath297 , @xmath298 [ l : lemma - main ] we will prove this in section [ s : lemma - main ] , but first , we ll explain why this proves theorem [ t : main ] . without loss of generality , we may assume @xmath299 by scaling @xmath2 if necessary . we first consider the case that @xmath300 . let @xmath196 be the cubes from lemma [ l : cubes ] tailored to the metric space @xmath2 with @xmath301 and define inductively , @xmath302 by lemma [ l : lemma - main ] , for any @xmath303 , if @xmath304 , then @xmath305 and moreover , since @xmath290 , @xmath306 define a probability measure @xmath43 inductively by picking @xmath307 , setting @xmath308 and for @xmath303 and @xmath309 , @xmath200 @xmath310 let @xmath6 , @xmath311 . pick @xmath50 so that @xmath312 there is at most one @xmath313 such that @xmath314 indeed , if there were another such @xmath315 with @xmath316 , then @xmath317 since @xmath290 and we can pick @xmath318 so that @xmath319 , which gives a contradiction and proves the claim . now , assuming we have @xmath313 satisfying , @xmath320 for @xmath291 large enough ( that is , for @xmath321 , which is possible by picking @xmath322 ) . if @xmath323 , then implies @xmath324 for all @xmath325 , and so @xmath326 otherwise , if @xmath327 , then @xmath328 , so that @xmath329 thus @xmath43 is a @xmath330-frostmann measure supported on @xmath2 , which implies @xmath331 ( c.f . * theorem 8.8 ) ) . now we consider the case when @xmath332 . trivially , @xmath333 for all @xmath6 and @xmath7 , and our previous work gives @xmath334 for all @xmath335 , hence @xmath336 . since @xmath337 , we must have @xmath338 , and so @xmath339 and the theorem follows with @xmath340 . to show lemma [ l : lemma - main ] , we will approximate @xmath2 by a tree containing a sufficiently dense net in @xmath2 and estimate its length from below . the following lemma relates the length of this tree to the number of net points in @xmath2 . let @xmath341 be a maximal @xmath342-net for a connected metric space @xmath2 where @xmath343 is so that @xmath344 . then we may embed @xmath2 into @xmath77 so that there is a connected union of finitely many line segments @xmath345 containing @xmath341 such that for any @xmath346 and @xmath347 , @xmath348 [ l : tree ] embed @xmath2 isometrically into @xmath349 so that for any @xmath6 , the first @xmath350 coordinates are all zero . construct a sequence of trees @xmath351 as follows . enumerate the elements of @xmath352 . for two points @xmath353 and @xmath354 , let @xmath355\}\ ] ] where @xmath356 is the standard basis vector in @xmath349 ( i.e. it is equal to @xmath138 in the @xmath357th coordinate and zero in every other coordinate ) . now construct a sequence of trees @xmath351 in @xmath358 inductively by setting @xmath359 and @xmath360 equal to @xmath351 united with @xmath361 , where @xmath362 and @xmath363 are such that @xmath364 since @xmath126 is connected , @xmath365 , so that @xmath366 then @xmath367 is a tree contained in @xmath349 containing @xmath341 ( the reason we made the arcs @xmath368 reach into an alternate dimension is to guarantee that the branches of the tree do nt intersect except at the points @xmath341 ) . to prove , note that since @xmath369 and @xmath370 we have @xmath371 we now dedicate ourselves to the proof of lemma [ l : lemma - main ] . again , let @xmath126 be a connected metric space satisfying the conditions of theorem [ t : main ] . without loss of generality , @xmath372 , so that @xmath373 . embed @xmath126 into @xmath77 as in lemma [ l : tree ] . fix @xmath374 . let @xmath375 be the tree from lemma [ l : tree ] containing the @xmath342-net @xmath376 . since @xmath375 is a tree of finite length that is a union of finitely many line segments , it is not hard to show that there is a piecewise linear arc length parametrized path @xmath377\rightarrow \gamma_{n_{0}}$ ] that traverses almost every point in @xmath375 at most twice ( except at the discrete set of points @xmath341 ) . the proof is similar to that of its graph theoretic analogue . let @xmath196 be the cubes from lemma [ l : cubes ] tailored to @xmath375 and fix @xmath378 . we will adjust the values of @xmath4 in lemma [ l : cubes ] and the value @xmath292 in the definition of @xmath291 as we go along the proof . note that @xmath373 implies @xmath379 if @xmath195 , and so @xmath380 . for @xmath197 , write @xmath381 if @xmath382 is a maximal cube in @xmath196 properly contained in @xmath49 . for @xmath383 and @xmath201 , define @xmath384 @xmath385 @xmath386 for @xmath201 , let @xmath387 : ( a , b)\mbox { is a connected component of } \gamma^{-1}(q)\}\ ] ] and for @xmath388 , define @xmath389 to be the continuous function such that for all @xmath390 and @xmath391\in \lambda(q)$ ] , @xmath392}(at+(1-t)b)=t\gamma(a)+(1-t)\gamma(b)\mbox { for } t\in [ 0,1],\ ] ] that is , @xmath389 is linear in all cubes in @xmath256 and agrees with @xmath284 on the boundaries of the cubes ( see figure [ f : cubes ] ) . ( 100,200)(60,0 ) ( 0,0 ) ( 65,105)(a ) ( 0,0)(b ) ( 125,0)(c ) ( 45,170)@xmath49 ( 170,175)@xmath393 ( 195,105)@xmath394 ( 20,110)@xmath395 ( 105,15)@xmath396 lemma [ l : lemma - main ] will follow from the following two lemmas : there is @xmath397 and @xmath398 ( independent of @xmath343 above ) such that if @xmath294 , @xmath399 , and @xmath400 , either @xmath401 or @xmath402 , where @xmath403 [ l : good - or - bad ] with @xmath404 defined as above , we have @xmath405 [ l : bad ] we ll prove these in sections [ s : good - or - bad ] and [ s : lemma - bad ] respectively , but first let us finish the proof of lemma [ l : lemma - main ] . + for @xmath406 , let @xmath407 be such that @xmath408 and define @xmath409 by telescoping sums and lemma [ l : zero - boundary ] , we have @xmath410 note that @xmath411 since @xmath378 , @xmath412 , and @xmath375 is connected . this , lemma [ l : good - or - bad ] , and lemma [ l : bad ] imply @xmath413 so that @xmath414 by lemma [ l : tree ] , and since @xmath415 has radius @xmath143 , @xmath416 combining these two estimates we have , for @xmath417 that @xmath418 pick @xmath419 . since @xmath420 , we get @xmath421 since @xmath422 , and this proves lemma [ l : lemma - main ] with @xmath423 . + by inspecting the proof of lemma [ l : good - or - bad ] below , one can solve for explicit values of @xmath424 , and @xmath11 . in particular , one can choose @xmath425 , @xmath426 , @xmath427 , and @xmath428 , so that the supremum of permissible values of @xmath429 is at least @xmath430 , and is by no means tight . in the next two subsections , we prove lemma [ l : good - or - bad ] and lemma [ l : bad ] . fix @xmath49 as in the statement of the lemma . for any @xmath431 , @xmath432 hence , to prove the lemma , it suffices to show that either @xmath433 or there is an interval @xmath431 for which @xmath434 fix @xmath217 so that @xmath250 . let @xmath435 be such that @xmath436 and pick @xmath437 such that @xmath438 . note that @xmath439 is a segment with endpoints the same as @xmath394 , hence @xmath440 before proceeding , we ll give a rough idea of how the proof will go . we will consider a few cases , which are illustrated in figure [ f : cases ] below . + ( 100,240)(80,0 ) ( 0,0 ) ( 55,145)@xmath441 ( 70,175)@xmath442 ( 40,190)@xmath443 ( 10,220)@xmath49 ( 52,85)@xmath444 ( 35,95 ) @xmath126 ( 200,80)@xmath445 ( 0,125)case 1 ( 125,125)case 2a ( 0,0)case 2b ( 185,85)@xmath444 ( 125,0)case 2b cont . in the first case , we assume the diameter of @xmath441 is small with respect to @xmath49 ; since @xmath394 has the same endpoints as @xmath396 and intersects the center cube @xmath443 , there must be a large difference in length between @xmath442 and @xmath441 since the former must enter @xmath49 , hit @xmath443 , and then exit @xmath49 , and so will hold . for the next two cases , we assume @xmath441 has large diameter . the second case ( 2a ) assumes that @xmath442 contributes more length than @xmath441 , again implying trivially . ( it is possible to combine this case with ( 1 ) , but we found this split to be somewhat convenient . ) in the final case ( 2b ) we assume the difference in length between @xmath442 and @xmath441 is small . since @xmath446 , we can show this implies the existence of @xmath447 far away from @xmath442 ( since @xmath394 has small geodesic deviation , so it ca nt approximate all of @xmath126 in @xmath448 ) . since @xmath375 approximates @xmath126 , we can find a large curve @xmath449 entering @xmath448 , approaching @xmath444 , and then leaving @xmath448 . the presence of both @xmath450 and @xmath445 inside @xmath49 implies that the total length of @xmath451 must be large , which means @xmath402 . + now we proceed with the actual proof . * case 1 : * suppose @xmath452 . since @xmath394 is a path entering @xmath49 , hitting @xmath443 , and then leaving @xmath49 , we can estimate @xmath453 thus , @xmath454 if @xmath455 , which implies the lemma in this case . + * case 2 : * suppose @xmath456 we again split into two cases . + * case 2a : * suppose @xmath457 then @xmath458 * case 2b : * now suppose @xmath459 note that in this case , we have a better lower bound on @xmath460 , namely , @xmath461 let @xmath462 ( we will pick its value later ) . assuming the conditions in case 2b , let @xmath286 be the smallest interval with @xmath463 and @xmath464 . then @xmath465 since @xmath466 enters @xmath467 , hits @xmath443 , and then leaves @xmath468 , we have @xmath469 hence , @xmath470 thus , @xmath471 if we pick @xmath472 and @xmath473 ( recall @xmath474 ) . by lemma [ l : subarc ] , @xmath475 which proves . by the main assumption in theorem [ t : main ] , and because we re assuming @xmath372 so that @xmath476 , @xmath477 so there is @xmath478 with @xmath479 if @xmath480 . since @xmath481)$ ] hits every cube in @xmath482 , which all have diameter at most @xmath483 by ( recall @xmath217 was chosen so that @xmath250 ) , @xmath484))_{2(1+\ve\beta)cm^{-n-1 } } \subseteq ( \gamma_{n+1}([0,1]))_{4cm^{-n-1}}\ ] ] note that since @xmath400 , we have @xmath485 . since @xmath486 and @xmath485 , @xmath487))_{4cm^{-n-1}+ 2m^{-n_{0}}}\\ & \subseteq ( \gamma_{n+1}([0,1]))_{(4cm^{-n-1 } + 2m^{-n-1 } ) } = ( \gamma_{n+1}([0,1]))_{(2+\frac{1}{c})m^{-1}2cm^{-n}}\\ & = ( \gamma_{n+1}([0,1]))_{(2+\frac{1}{c})m^{-1}\diam b_{q } } \subseteq ( \gamma_{n+1}([0,1]))_{\frac{2}{c}m^{-1}\diam b_{q}}\end{aligned}\ ] ] since @xmath195 . since @xmath478 , there is @xmath488 $ ] such that @xmath489 and so @xmath490 for @xmath491 . also , since @xmath492 , we know that @xmath493 for @xmath494 . in particular , @xmath495 . note @xmath496),\diam \gamma([b_{i}',b_{i}])\ } \notag \\ & \geq \dist(\gamma_{n+1}(t),\gamma_{n+1}(i'))-\ell(\gamma\|_{i / i ' } ) \stackrel{{\eqref{e:3veb+cb } } \atop { \eqref{e : b/32}}}{\geq } \frac{\beta}{32}\diam q-(3\ve\beta+c\beta)\diam q \notag \\ & \geq \frac{\beta}{64}\diam q\end{aligned}\ ] ] for @xmath497 and @xmath498 . thus , since of course @xmath499 , we have @xmath500 in particular , @xmath501 , and so by construction , @xmath502 $ ] for some @xmath391\in \lambda(q)$ ] , where @xmath503 and @xmath504 are both in @xmath375 . in particular , @xmath505 is a line segment in a cube @xmath506 . if @xmath507 , then @xmath508 for @xmath509 , and so @xmath510 thus , since @xmath375 is connected and @xmath511 , we know there is a curve @xmath512 connecting @xmath513 to @xmath514 , and hence has diameter at least @xmath515 . hence , @xmath516 moreover , @xmath517 hence , since any cube in @xmath518 intersecting @xmath445 has diameter at most @xmath519 by , they are disjoint from those intersecting @xmath450 by if we choose @xmath520 ( since if they intersect @xmath450 , they also intersect @xmath442 by the definition of @xmath466 ) . thus , we have @xmath521 for @xmath522 . hence , by picking @xmath523 , we see that @xmath402 , which finishes the proof of lemma [ l : good - or - bad ] for @xmath201 , define @xmath524 to be the number of cubes in @xmath404 that properly contain @xmath49 , and set @xmath525 @xmath526 @xmath527 we will soon define , for each @xmath528 , a nonnegative weight function @xmath529 @xmath530-a.e . in a martingale fashion by defining it as a limit of a sequence @xmath531 . each @xmath531 will be constant on various subsets of @xmath375 that partition @xmath532 . we will actually decide the value of @xmath531 on an element @xmath34 of the partition , say , by declaring the value of @xmath533 then we will define @xmath534 to be constant on sets in a partition subordinate to the previous partition so that , on sets @xmath34 in the @xmath141th partition , @xmath535 , and so forth . we do this in such a way that we disseminate the mass of the weight function @xmath536 so that @xmath536 is supported in @xmath49 , has integral @xmath537 , and so that @xmath538 , where @xmath539 is the total number of bad cubes containing @xmath353 . by geometric series , this will mean that @xmath540 is a bounded function , so that its total integral is at most a constant times @xmath541 . however , the integral of each of these functions @xmath536 is @xmath537 , and so the integral is also equal to @xmath542 , which gives us . this method appears in @xcite . now we proceed with the proof . first set @xmath543 and construct @xmath534 from @xmath531 as follows : 1 . if @xmath544 for some @xmath141 , and @xmath545 , set @xmath534 to be constant in @xmath546 so that @xmath547 2 . set @xmath534 to be constant in @xmath548 so that @xmath549 3 . for points @xmath353 not in in any @xmath544 , set @xmath550 . like a martingale , we have by our construction that , if @xmath544 , then @xmath551 for all @xmath552 , and in particular , @xmath553 for all @xmath554 . + we will need the following inequality : @xmath555 the first inequality comes from the fact that if @xmath162 and @xmath556 is a cover of @xmath548 by sets so that @xmath557 , then @xmath558 is a cover of @xmath382 ( up to a set of @xmath559-measure zero by lemma [ l : zero - boundary ] ) , and so @xmath560 which gives the first inequality in by taking @xmath561 . the last inequality in is from the definition of @xmath404 . for @xmath545 and @xmath544 , by induction we have @xmath562 hence , since @xmath534 is constant in @xmath546 , for @xmath563 , @xmath564 moreover , if @xmath565 , @xmath566 since @xmath567 , and @xmath568 , almost every point @xmath569 is contained in at most finitely many cubes in @xmath404 , and hence the value of @xmath570 changes only finitely many times in @xmath141 , thus the limit @xmath571 is well defined almost everywhere . for @xmath572 , set @xmath573 where @xmath200 is the smallest cube in @xmath404 containing @xmath353 . then and imply @xmath574 and so @xmath575 since @xmath576 . hence , @xmath577 which finishes the proof of lemma [ l : bad ] . this section is devoted to the proof of theorem [ t : antenna - like ] . it is easy to verify using the definitions that being antenna - like is a quasisymmetric invariant quantitatively , so by theorem [ t : main ] , it suffices to verify that , if @xmath126 is @xmath143-antenna - like , then any ball @xmath3 with @xmath578 and @xmath579 has @xmath151 . fix such a ball , so there is a homeomorphism @xmath580\rightarrow x\cap b(x , r)$ ] so that @xmath581\cup [ 0,e_{k}]))\geq cr \label{e : hy}\ ] ] for all permutations @xmath148 of @xmath149 ( see figure [ f : antenna - beta ] ) . let @xmath56\rightarrow b(x , r)$ ] satisfy @xmath582})-\|s(0)-s(1)\|+\sup_{z\in x\cap b(x , r)}\dist(z , s([0,1]))<2\beta'(x , r)\|s_{0}-s_{1}\|=:\beta.\ ] ] set @xmath583 for @xmath584 and let @xmath585 this always exists since @xmath586))_{\beta}$ ] . without loss of generality , assume @xmath587 similarly , let @xmath588 and again , without loss of generality , assume @xmath589 . ( 100,180)(45,0 ) ( 0,0 ) ( 37,77)@xmath590 ( 60,60)@xmath444 ( 13,35)@xmath212 ( 2,17)@xmath591 ( 175,33)@xmath592 ( 75,30)@xmath593 ( 75,177)@xmath594 ( 175,65)@xmath595 ( 80,110)@xmath596 note that @xmath597\cup [ 0,e_{3}])$ ] is a path connecting @xmath212 to @xmath594 , where the latter point is not contained in @xmath598))_{\beta}$ ] by our choices of @xmath599 and @xmath600 , although the latter point is ; otherwise , there would be @xmath601 $ ] such that @xmath602 , contradicting the minimality of @xmath600 . since @xmath597\cup [ 0,e_{3}])$ ] is connected and @xmath598))_{\beta}$ ] contains @xmath212 but not @xmath594 , we can pick a point @xmath603\cup [ 0,e_{3}])$ ] so that @xmath604))=\beta$ ] . pick @xmath605 $ ] and @xmath606 $ ] so that @xmath607))=\beta \;\;\ ; \mbox { and } \;\;\ ; \|s(\zeta_{2})-z\|<\beta . \label{e : zeta12}\ ] ] then by lemma [ l : subarc ] , @xmath608})-\|s(0)-s(1)\| \geq \ell(s\|_{[\zeta_{1},\zeta_{2}]})-\|s(\zeta_{1})-s(\zeta_{2})\|\\ & \geq \ell(s\|_{[\zeta_{1},t_{2 } ] } ) + \ell(s\|_{[t_{2 } , \zeta_{2 } ] } ) -\|s(\zeta_{1})-z\|-\|z - s(\zeta_{2})\|\\ & \hspace{-3pt}\stackrel{{\eqref{e : zeta12 } } } { > } \|s(\zeta_{1})-s(t_{2})\|+\|s(t_{2})-s(\zeta_{2})\|-\beta-\beta\\ & \geq \|z - x_{2}\|-\|s(\zeta_{1})-z\|-\|x_{2}-s(t_{2})\| \\ & \hspace{13pt}+\|x_{2}-z\|-\|s(t_{2})-x\|-\|s(\zeta_{2})-z\|-2\beta\\ & \hspace{-13pt}\stackrel{{\eqref{e : hy}},{\eqref{e : zeta12 } } } { \geq } cr-\beta-\beta+cr-\beta-\beta-2\beta \\ & = 2cr-6\beta\geq c\|s(0)-s(1)\|-12\beta(x , r)\|s(0)-s(1)\|\end{aligned}\ ] ] which yields @xmath609 and completes the proof of theorem [ t : antenna - like ] for quantities @xmath34 and @xmath184 , we will write @xmath610 if there is a universal constant @xmath263 so that @xmath611 , and @xmath612 if @xmath613 . for the opposite inequality , let @xmath56\rightarrow\ell^{\infty}$ ] be such that @xmath617))}{\|s(0)-s(1)\|}\leq 2\beta'(x , r)=:\beta . \label{e : infs}\ ] ] let @xmath618))_{2\beta\|s_{0}-s_{1}\| } \right)}\ ] ] which is a relatively open subset of @xmath619 $ ] . let @xmath620 and define @xmath621 inductively by setting @xmath622 : \dist ( s(t),s([t_{0},t_{i}]))>\beta^{\frac{1}{2}}\|s(0)-s(1)\|\}.\ ] ] we claim that @xmath623 to see this , note that since @xmath624 , the sets @xmath625 are disjoint , so that @xmath626 which gives @xmath627 . on the other hand , the balls @xmath628 cover @xmath89)$ ] , and so @xmath629 which gives @xmath630 , and this proves . @xmath632})-\|s(t_{0})-s(t_{n})\|+c\beta^{\frac{1}{2}}\|s(0)-s(1)\|\\ \stackrel{{\eqref{e : infs}}}{\leq } \beta\|s_{0}-s_{1}\|+c\beta^{\frac{1}{2}}\|s(0)-s(1)\|\lec \beta^{\frac{1}{2}}\|s(0)-s(1)\|.\end{gathered}\ ] ] since @xmath634 is connected and @xmath150 , there is a path connecting @xmath353 to @xmath635 , which naturally must be of diameter at least @xmath61 , hence @xmath636 which , if @xmath637 is small enough , this implies @xmath638 so that the above estimates imply @xmath639 moreover , @xmath640))_{\beta\|s(0)-s(1)\| } \subseteq \bigcup_{i } b(s(t_{i}),(2\beta^{\frac{1}{2 } } + \beta)\|s(0)-s(1)\| ) \notag \\ & \subseteq \bigcup_{i}b(y_{i},(2\beta^{\frac{1}{2 } } + \beta+2\beta)\|s(0)-s(1)\| ) \notag \\ & \subseteq \bigcup_{i}b(y_{i},5\beta^{\frac{1}{2}}\|s(0)-s(1)\| ) \subseteq \bigcup_{i}b(y_{i},20\beta^{\frac{1}{2}}\|y_{0}-y_{n}\| ) \label{e : yb1/22}\end{aligned}\ ] ] thus and imply @xmath641 . note that is tight in the sense that if @xmath645 , @xmath646 , and @xmath647\cup [ 0,i\ve]$ ] , then by theorem [ t : antenna - like ] and , for all @xmath292 , @xmath648 however , if @xmath649\cup [ 0,e^{i\ve}]$ ] , then for all @xmath292 , again by ( and estimating @xmath650 by letting @xmath47 be the path traversing the segments @xmath651\cup [ 0,e^{i\ve}]$ ] ) , @xmath652 for the first inequality , simply let @xmath56\rightarrow \ch$ ] be the line segment spanning @xmath653 where @xmath654 is some line passing through @xmath655 . then @xmath656 and hence @xmath657))}{\|s(0)-s(1)\| } \leq \frac{\sup_{z\in b(x , r)\cap x}\dist ( z , l)}{r}.\ ] ] since @xmath578 , the range of admissible lines in the infimum in can be taken to be lines intersecting @xmath655 . using this fact and infimizing the above inequality over all such lines proves the first inequality in . for the opposite inequality , let @xmath47 satisfy @xmath658))}{\|s(0)-s(1)\| } \leq 2\beta''(b(x , r))=:\beta.\ ] ] let @xmath659}\dist ( s(t),[s(0),s(1)]).\ ] ] then by the pythagorean theorem , there is @xmath4 so that @xmath660 so that @xmath661 . hence , if @xmath654 is the line passing through @xmath91 and @xmath92 , @xmath662)\\ \leq \beta(s)+\sup_{z\in b(x , r)\cap x}\dist ( z , s([0,1 ] ) ) \leq c^{-1}\beta+\beta \lec \beta\end{gathered}\ ] ]	for a compact connected set @xmath0 , we define a quantity @xmath1 that measures how close @xmath2 may be approximated in a ball @xmath3 by a geodesic curve . we then show there is @xmath4 so that if @xmath5 for all @xmath6 and @xmath7 , then @xmath8 . this generalizes a theorem of bishop and jones and answers a question posed by bishop and tyson . + * mathematics subject classification ( 2010 ) : * 28a75 , 28a78 , 30l10 . + * keywords : * wiggly sets , traveling salesman , geodesic deviation , hausdorff dimension , conformal dimension .
the development environment that we have used for the spm simulator program is similar that used to develop our real - time scanning program rtspm.@xcite the details can be found in ref . , so we shall only give an outline here . we emphasize that all the software used is open - source . in order to achieve real - time control , we use a desktop computer running a linux kernel patched with the real time application interface ( rtai).@xcite communication with the data acquisition hardware is achieved through the open - source comedi drivers@xcite with appropriate real - time extensions that interface well with rtai . for this paper , we used a national instruments@xcite pci-6052e card for input and output , although any of the data cards supported by comedi can be used , so long as they meet the required data acquisition rates of the program . the pci-6052e card has analog input and output rates of 333 ks / sec , more than sufficient for our purposes , but we have also used other national instruments data acquisition cards . the real time components of the program are coded in a library as callable subroutines that are written in c using the integrated development environment code::blocks,@xcite which is also open - source . the graphical user interface ( gui ) is written in free pascal@xcite using the open - source integrated development environment lazarus;@xcite the real - time parts of the program are called by this main program . figure [ fig1 ] shows a screenshot of the gui of the program , which lets the user choose the channels for the data acquisition inputs and outputs . there are three data acquisition inputs to the program , corresponding to the @xmath1 , @xmath2 and @xmath3 voltage drives of a standard piezotube scanner , which can be selected independently from the gui . these inputs are provided by a spm control program : in our case , this control program is the real - time control program rtspm that we developed earlier.@xcite the voltages on the @xmath1 and @xmath2 channels are assumed to correspond directly to the @xmath1 and @xmath2 displacements of the scanner : of course , one can easily implement an appopriate scale factor for each channel independently to model specific hardware if needed . for the @xmath3 voltage , the user can input the scale factor directly . for the data shown in this paper , we use a scale factor of 155 nm / v , corresponding to the scan tube in our physical instrument . for testing purposes , reading of the @xmath1 and @xmath2 channels can be disabled , and the @xmath1 and @xmath2 positions of the scanner can be manually entered from the gui . as we noted earlier , the real - time part of the program is very simple in concept . once the simulator is started , a real - time loop runs continuously with a loop time of 50 @xmath0s . in the loop itself , the program first determines the @xmath1 and @xmath2 positions of the scanner . if external scanning is enabled , these positions are read directly from the @xmath1 and @xmath2 input channels , otherwise they are taken to be the manually entered values discussed above . depending on the @xmath1 and @xmath2 positions of the scanner , the program then determines what the corresponding topographical height should be . for this paper , we have taken the simplest structure , an array of squares of specified lateral dimension and specified height with a lattice constant of 1 @xmath0 m , although clearly more complicated topographic structures can easily be programmed . the program then reads the @xmath3 channel to determine the height of the scanner . since the program now knows the height of the @xmath3 scanner and the topographic height of any feature at that @xmath4 position directly below it , it can then calculate and output a voltage according to whatever model is being used to generate the tip - sample interaction . this output voltage , which corresponds to the feedback signal input to the spm controller program , is then read by the spm controller program to adjust the @xmath3 piezo voltage accordingly . the process is then repeated on the next 50 @xmath0s cycle . the voltage that the program generates depends on the model of the tip - sample interaction . we discuss below two models that we have implemented , although other tip - sample interactions can also be modeled . our own research is devoted to development of a low temperature scanning probe microscope , so it is natural for us to first try to implement a model for an atomic force microscope . our home - built scanning probe microscope is based on a tuning fork transducer,@xcite so the parameters in the model described below will refer to the measured parameters from this instrument , but the program allows these parameters to be modified to match any force transducer . we start by assuming that the interaction between the tip and sample is due to van der waals forces at larger distances with a strong repulsion at short distances . this interaction can be described by a lennard - jones type potential of the form @xmath5 where @xmath3 is the distance between the tip and the sample , and @xmath6 and @xmath7 are constants . the first term describes the strong repulsion between tip and sample at very short distances , and the second describes the relatively short range van der waals interaction . the two unknowns in the potential are the parameters @xmath6 and @xmath7 . we shall use the measured characteristics of the close - approach curve to determine these constants . the force between the tip and the sample corresponding to this potential is @xmath8 to eliminate one of these constants , we specify the position of the minimum of force @xmath9 as a function of @xmath3 as @xmath10 . by setting @xmath11 at @xmath12 , this allows us to express @xmath7 in terms of @xmath6 and @xmath10 , @xmath13 a tuning fork based afm is operated in non - contact mode . the tuning fork is oscillated at its resonance frequency @xmath14 , which for our tuning forks is close to 32768 hz . the tip - sample interaction modifies the resonant frequency . the shift in resonant frequency @xmath15 can be thought of as arising from a shift @xmath16 in the effective spring constant of the tuning fork @xmath17 where @xmath18 is the spring constant of the tuning fork far from the surface . this is because the resonance frequency is proportional to the square root of the spring constant . we shall use @xmath18=1800 n / m , corresponding to the spring constants of the tuning forks we normally use . now @xmath16 is given by the derivative of the tip - sample force @xmath19 . \label{eqn5}\ ] ] in non - contact mode , one can measure the amplitude or phase of the oscillation as a feedback signal . however , if the quality factor @xmath20 of the force transducer is large , as it is for the tuning forks , it may take a very long time for these signals to relax to their proper values . hence , it is common to use a phase - locked - loop ( pll ) to stay on the resonance and track the change in resonant frequency as a function of distance @xmath3 . thus , we will use the change in frequency as the feedback signal for our spm controller . the change in frequency is proportional to @xmath16 , which still has one unknown constant @xmath6 ( eqn . [ eqn5 ] ) . since one directly measures the frequency - distance curve during close approach , @xmath6 can be determined from this curve . let the total change in frequency between where the tip is far from the sample ( @xmath21 ) and the value of @xmath22 where the frequency shift @xmath15 has its minimum be @xmath23 . ( @xmath24 is related to @xmath10 by @xmath25 . ) then @xmath6 can be expressed as @xmath26 determining @xmath27 in a real experiment ( and hence the absolute values of @xmath10 or @xmath24 ) is difficult , since @xmath27 is the point at which the tip makes contact with the sample , and one prefers not to crash the tip into the sample . at distances far from the surface ( @xmath21 ) , @xmath15 is ideally 0 : in reality , due to experimental offsets , it may have a small finite value , which we denote @xmath28 . then the value of @xmath3 near the surface at which @xmath29 is by definition @xmath10 , and the minimum of @xmath15 as a function of @xmath3 occurs at @xmath30 . thus by measuring the value of @xmath3 at these two points , one can determine @xmath10 . consequently , by specifying the known or measured quantities @xmath23 , @xmath18 , @xmath14 and @xmath10 , one can determine all the parameters of the model . these quantities are entered in the main program by hand . the frequency shift , and hence the feedback signal that the program generates , can then be calculated using eqns . ( [ eqn4 ] ) , ( [ eqn5 ] ) and ( [ eqn6 ] ) . m etched tungsten wire acting as a tip . the spring constant of the tuning fork is 1800 n / m , and the curve was obtained with the rtspm program . red : approach curve obtained with the simulator program , using a lennard - jones type potential as described in the text . , width=321 ] the black trace in fig . [ fig2 ] shows an example of an experimental close - approach curve measured using our home - built tuning fork scanning probe microscope and the spm control program rtspm . from this curve , we determine @xmath31 hz and @xmath32 nm . the red trace in fig . [ fig2](b ) shows the corresponding curve generated using the spm simulator program , again taken with the spm control program rtspm . it can be seen that the experimental close - approach curve is broader than the simulation based on the lennard - jones potential . we do not know the origin of this discrepancy ; however , it should be noted that the experimental curve was taken with a conducting tip , and hence may be influenced by residual electrostatic interactions , which have a slower power - law dependence . m @xmath33 12 @xmath0 m forward topographic scan with a resolution of 256 x 156 pixels . each square in the image is 0.5 @xmath0 @xmath33 0.5 @xmath0 m and has a height of 5 nm . ( b ) line scan profiles corresponding to the line marked in ( a ) for the forward scan ( black trace ) and the reverse scan ( red trace ) . , width=321 ] figure [ fig3](a ) shows an image generated by the rtspm program coupled to the spm simulator program . ( this figure and other images in this paper are generated using the open source spm analysis program gwyddion.@xcite ) as described earlier , the `` sample '' is a computer generated array of squares of side 0.5 @xmath0 m and height 5 nm , separated by a distance of 1 @xmath0 m . the rtspm has a real - time proportional - integral - differential ( pid ) controller@xcite that controls the extension of the scan tube based on the desired set point , which is a fixed frequency deviation from the @xmath34 limit . for the image in fig . [ fig3](a ) , the pid parameters used in the rtspm program were @xmath35 , @xmath36 ms , and @xmath37 ms with a set point of 2 hz , which puts us in the hard repulsive region of the approach curve . in this region , a small change in @xmath3 gives rise to a large change in frequency . nevertheless , as can be seen from the line profiles shown in fig . [ fig3](b ) , the topography is accurately mapped by the rtspm program in both the forward and reverse scan directions , with a difference corresponding to about 1 pixel . the scan resolution is 256x256 pixels , and the rtspm program averages for 10 ms on each pixel to reduce the noise , so the image of fig . [ fig3](a ) took approximately 12 minutes to acquire . , but with a non - realtime simulator coded in labview on a computer running windows xp . ( b ) line scan profiles corresponding to the line marked in ( a ) for the forward scan ( black trace ) and the reverse scan ( red trace ) . , width=321 ] in order to illustrate the importance of having a real - time spm simulator program , we have also encoded the same program on a desktop computer running windows xp in national instruments labview without any real - time extensions . in this case , even on a relatively fast computer , the minimum loop time is at best 10 ms . figure [ fig4](a ) shows the image obtained with the rtspm program using the same pid parameters and setpoint as in fig . [ fig3 ] , but coupled to the labview simulator . it can be seen that the image quality is much worse , and the line profiles in fig , [ fig4](b ) show that this is because the @xmath3 position of the scanner does not follow the topography . since all parameters in the rtspm program are the same for figs . [ fig3 ] and [ fig4 ] , the difference is clearly due to the lack of real - time response in the labview program . the van der waals interaction for afm in principle can be extended to other types of tip - sample interactions . for example , for scanning tunneling microscopy , one can model a tunneling current that depends exponentially on the distance between the tip and the sample , and also on bias . for a more sophisticated model , one can think of a conducting `` sample '' with a spatially inhomogenous density of electronic states . even relatively inexpensive modern multicore desktop computers should be able to calculate the feedback response of most tip - sample interaction models in a loop time of 50 @xmath0s . for long range tip - sample interactions , the models are more complicated . for example , for magnetic force microscopy ( mfm ) on a ferromagnetic film , one must consider not only the magnetic interaction between the magnetic tip and the part of the sample just below the tip , but also the magnetic interaction with other parts of the ferromagnetic film that are much further away . in electrostatic force microscopy ( efm ) , the electrostatic interaction between the tip and the sample is also long range , and an accurate calculation would require numerical techniques . there have been a number of papers that have developed approximations for the tip sample interaction both in the context of efm and scanning capacitance microscopy.@xcite however , with some simplifiying assumptions , one can use a simple model that gives a realistic reproduction of an electrostatic image . the assumption that we shall make is that the conducting tip on the scan tube is very close to the conducting surface of the sample , so that the interaction between the tip and the sample can be approximated by the interaction between a small conducting sphere and an infinite conducting plane , where the radius @xmath38 of the small sphere is the same as the radius of the microscope tip . as can be expected , this assumption neglects fringe fields at the edges of a conducting region in the sample , but the result should be accurate beyond a few multiples of the tip - sample distance from any edge . consider then the force between a conducting sphere of radius @xmath38 and an infinite grounded conducting plane . this is a classic problem that can be solved by the method of images.@xcite the force between the sphere and conductor is given by @xmath39 where @xmath40 is the voltage difference between the sphere and plane ( which in the case of real metals will also include any contact potentials ) , and @xmath41 is the capacitance between them . since @xmath41 decreases with increasing distance @xmath3 , the force is always attractive . calculation of the force then reduces to calculation of the capacitance , which is given by the series@xcite @xmath42 where @xmath43 . the resulting expression for the force is @xmath44 since we would like to calculate the derivative of @xmath45 with respect to @xmath3 to calculate the shift in frequency along the lines of eqns . [ eqn4 ] and [ eqn5 ] , it would be nice to have a simpler equation to work with . _ et al.__@xcite have shown that the following expression closely approximates eqn . [ eqn9 ] , with a maximum error of a few percent@xcite @xmath46 v^2 . \label{eqn10}\ ] ] we shall use this equation in our calculations of the frequency shift @xmath47 due to the electrostatic interaction . this frequency shift can be calculated from eqn . [ eqn4 ] and the relation @xmath48 . we obtain @xmath49 ^ 2}\right ] . \label{eqn11}\ ] ] unlike the equation for the van der waals interaction , we note that there are no free parameters in this equation . figure [ fig5 ] shows the close - approach curves generated with the spm simulator program both with and without the electrostatic force . here the radius @xmath38 of the tip is taken to be 20 nm , and the potential difference @xmath40 between the tip and sample is 5 v. as is well known and can be seen from the two curves , the van der waals interaction is dominant close to the surface , at a distance of less than a few nm , while the electrostatic interaction , being longer range , contributes to the attractive potential at larger distances . this forms the basis of two techniques to extract the electrostatic force . in the first technique , for each line of the scan , a forward and reverse scan is made under feedback at a setpoint corresponding to a distance from the surface at which the van der waals force is dominant . the resulting @xmath3 positions of the piezo scan tube reflect primarily the topography of the sample along the scan line . the spm controller then takes the scan tube out of feedback mode , raises the scan tube in the @xmath3 direction by a distance @xmath50 , the lift height , and retraces the forward and reverse traces while recording the frequency shift . since the system is no longer under feedback , @xmath50 should be larger than any topographic features in the sample . the idea behind this constant height technique is that for sufficiently large @xmath50 , the primary contribution to the resulting signal will be from electrostatic forces . one can also retrace the stored @xmath3 positions of the topographic trace at the added height @xmath50 while recording the frequency shift . the hope is that this `` lift mode '' technique will subtract the contribution from the signal due to topography , leaving only the signal due to the electrostatic force . of course , one can also measure other signals such as the phase or amplitude of the oscillation when the system is not in feedback , but we have chosen here to measure the frequency shift as it is conceptually easier to understand . the benefit of modeling the electrostatic interaction is that one can determine immediately what appropriate value of @xmath50 to use from the close - approach curves in fig . if we use a set point of 2 hz for the topographic image , it appears that a lift height of about 8 nm should give the largest electrostatic signal . m x 12 @xmath0 m , and the topographic height of each square is 5 nm . the colour bars are for the @xmath3-scale ; for the topographic image the units are m , for the efm images the units are hz.,width=321 ] m x 12 @xmath0 m , and the topographic height of each square is 5 nm . the colour bars are for the @xmath3-scale ; for the topographic image the units are m , for the efm images the units are hz.,width=321 ] in order to model an electrostatic sample , we use the same array of squares described earlier , but assume each alternate square in the array has a potential @xmath40 applied , with the tip and the other squares being grounded . this approximates the standard samples frequently used for efm calibration , which consist of interdigitated metallic fingers which have a potential applied only to every alternate finger , so that one can immediately distinguish between the topographic image and the electrostatic image . figure [ fig6 ] shows three dimensional representations of the resulting topographic and electrostatic images , using constant height mode in the dc efm module of rtspm . for these images , we used a setpoint for the topographic image of 2 hz ( referring to fig . [ fig5 ] ) , a voltage @xmath51 v , and a lift height of @xmath50= 8 nm . it is clear from the figure that there is substantial leakage of the topographic image into the efm image . this is not surprising , since for a lift height of 8 nm , the height of the tip above each square is only 3 nm so the van der waals force would contribute considerably to the signal . and [ fig7 ] , but with a lift height of 20 nm . the lift mode and constant height mode scan were acquired separately . the colour bars are for the @xmath3-scale ; for the topographic image the units are m , for the efm images the units are hz.,width=321 ] for comparison , fig . [ fig7 ] shows the corresponding image ( at the same lift height of 8 nm ) obtained using lift mode . while there is some bleeding of the topographic image into the efm image , it is quite small : the ratio of the signal between squares with and without a potential is about a factor of 20 . the maximum frequency shift in fig . [ fig7](b ) is about 300 mhz . clearly , lift mode is the preferable mode of operation when the sample has any appreciable topographical relief . to eliminate any topographic signal in the efm image , one can use a larger lift height . figure [ fig8 ] shows the lift mode and constant height mode efm images for a lift height of 20 nm . while the constant height mode image still shows some hint of the squares without any potential , the lift mode image shows no hint of the topography , but only the electrostatic profile . the overall lift mode signal is reduced in comparison to fig . [ fig7 ] , but not by much ( 80 mhz peak frequency shift ) , a reflection of the long range nature of the electrostatic force . in summary , we have developed a real - time software simulator that models the response of a scanning probe microscope . the simulator is useful for testing scanning probe control software as well as different models for tip sample interactions . text the software is availably freely as a git repository at git://github.com / chandranorth / tfsimulator.git . v. chandrasekhar and m.m . mehta , rev . instrum . * 84 * , 013705 ( 2013 ) . _ rtai - the realtime application interface for linux from diapm _ , https://www.rtai.org . _ comedi : linux control and measurement device interface _ , http://www.comedi.org . _ code::blocks _ , http://www.codeblocks.org . _ free pascal : advanced open source pascal compiler for pascal and object pascal _ , http://www.freepascal.org . _ lazarus _ , http://www.lazarus.freepascal.org . y. seo , p. cadden - zimansky , and v. chandrasekhar , applied physics letters * 87 * , 103103 ( 2005 ) . _ gwyddion : free spm data analysis software _ , http://www.gwyddion.net . s. hudlet , m. saint jean , c. guthmann and j. berger , eur . j. b * 2 * , 5 ( 1998 ) . m. saint jean , s. hudlet , c. guthmann and j. berger , j. appl . phys . * 86 * , 5245 ( 1999 ) . s. gmez - moivas , j.j . senz , r. carminati and j.j . greffet , appl . . lett . * 76 * , 2955 ( 2000 ) . y. naitou , a. yasaka and n. ookubo , j. appl . phys . * 105 * , 044311 ( 2009 ) . see , for example , w.r . smythe , _ static and dynamic electricity _ , 2nd edition , page 121 [ mcgraw - hill , new york , 1950 ] .	we describe software that simulates the hardware of a scanning force microscope . the essential feature of the software is its real - time response , which is critical for mimicking the behavior of real scanning probe hardware . the simulator runs on an open - source real time linux kernel , and can be used to test scanning probe microscope control software as well as theoretical models of different types of scanning probe microscopes . we describe the implementation of a tuning - fork based atomic force microscope and a dc electrostatic force microscope , and present representative images obtained from these models . obtaining and interpreting images with a scanning probe microscope is a complicated task , even more so in the case of completely home - built microscopes with both custom electronics and custom software . this is because of the integral relationship between software and hardware in scanning probe microscopes , particularly in the case of more recent instruments , where many of the functions previously performed by analog electronics are now performed in software , either on the control computer itself or on a separate digital signal processor . during development , in the frequent case that something goes wrong , it is not easy to identify whether the problem lies with the software or the hardware . thus , it would be useful to have a diagnostic tool to test whether the control software is indeed performing as designed . the heart of such a software simulator is in principle very simple : one needs only to convert the position of the tip ( as output by the scanning probe control program , perhaps by means of a voltage ) to a voltage corresponding to a feedback signal that can be read back by the control program . this conversion occurs according to some model of the tip - sample interaction , which in turn depends on the type of scanning probe microscopy , be it atomic force microscopy ( afm ) , scanning tunneling microscopy ( stm ) , magnetic force microscopy ( mfm ) or electrostatic force microscopy ( efm ) , to name a few . the difficult part is to do this in a time - critical manner ; the response of the simulator must be as fast or faster than typical scanning probe hardware , which means the response must be in fractions of a millisecond , and the response must be deterministic , in that response loops must be executed when expected . this is usually beyond the capabilities of standard desktop computers running modern operating systems , as these operating systems have preemptive multitasking , making their response times at best on the order of 10 ms , and more critically , not deterministic . the software we describe here@xcite is based on an open - source , real - time version of linux running on a standard desktop computer that enables loop times as short as 25 @xmath0s ( more reliably around 50 @xmath0s ) , corresponding to an upper limit frequency response on the order of 20 khz , fast enough to model most commercial and home - built scanning probe electronics . another major advantage of such a software simulator is that one is able to easily test different models of tip - sample interactions as well as the influence of various model parameters on the resulting images . the remainder of the paper is organized as follows : section i gives an overview of the software and hardware development platform . section ii describes the model used to implement an atomic force microscope ( afm ) , and section iii describes the model used to implement an electrostatic force microscope . these are two modes that are important for our own research , and we emphasize that many other types of interactions can also be implemented relatively easily if appropriate theoretical models are available .
recent experimental studies of the mechanical properties of biological materials , ranging from individual dna molecules@xcite to single actin filaments and their networks@xcite , led to revival of interest in the worm - like chain model of semiflexible biopolymers . in particular , the force - extension curves predicted by this model in the long chain limit ( i.e. , when the length of the chain is much larger than the persistence length ) were shown to agree quantitatively with experimental results on dsdna at large elongations@xcite . however , many interesting questions concerning worm - like chains under tension are still open and there are delicate issues concerning even the linear response of the worm - like chain to stretching force@xcite . also , while the equilibrium distribution function of the extension was calculated for the worm - like chain@xcite and fluctuating ribbon@xcite models , only approximate ( using the gaussian approximation ) calculations of the distribution function for stretched chains were reported to datewinkler . such distributions have been measured in experiments on stretched dna molecules@xcite and provide a much more stringent test of the theory than the force - extension curves . finally , even though all mechanical experiments involve biopolymers attached to surfaces , none of the present theories incorporates wall effects on force - extension curves and on the corresponding distributions . while such effects are negligible for nearly fully stretched chains , they are expected to be important in the weak force regime . in previous work@xcite , we have discussed how to generate via monte - carlo equilibrium ensembles of fluctuating ribbons with spontaneous curvature and/or twist . this method was based on using the intrinsic geometry of the filament , in which representation the energy is purely quadratic . this enables a direct generation of independent samples distributed according to the gibbs equilibrium ensemble . when external forces act on the filaments , the energy is no longer a quadratic function of the curvatures , and the problem is now not only nonlinear but also nonlocal . in this paper we discuss a biasing method which succeeds in efficiently generating monte - carlo samples for the simplest case of an external force , namely that of the stretched filament . this problem was studied by marko and siggia , ( ms ) , @xcite and others @xcite who calculated the force - extension curve for long worm - like chains . we will reproduce these results as well as extend the analysis to the case of nearly stiff filaments with length comparable to or shorter than the persistence length , and in addition calculate the distribution of extensions for various forces . lastly , we will show how we can also easily incorporate steric constraints ; we will demonstrate this in the context of a restriction of the filament to a half - space by the presence of a wall . in sections 2 and 3 we review the general theory of fluctuating ribbons ( we define a ribbon as a filament with an asymmetric cross section ) and the frenet algorithm for computer simulations of such ribbons , respectively . in section 4 we describe the biased monte - carlo simulations of filaments under tension and present the resulting force - extension curves and distribution functions . wall effects are considered in section 5 . in section 6 we discuss our results and outline directions for future research . in appendix a we expand the euler angles in terms of the principal curvatures and then use this expansion in appendix b in order to calculate the partition function of a stretched filament . in part 1 of this appendix we derive an analytic expression for the partition function and obtain a force - extension relation for arbitrary force in the limit of large persistence length ( expansion about the rigid rod limit ) . in part 2 of the appendix we derive asymptotic expressions for the above quantities in the limit of large force ( for arbitrary rigidity ) . finally , in appendix 3 we present the calculation of the renormalization of the elastic moduli of the various fourier modes of the curvature . the linear elasticity of ribbons is described by the elastic energyhelixpre @xmath0 where @xmath1 is the distance measured along the contour of the ribbon , the coefficients @xmath2 and @xmath3 are the moduli associated with bending along the two principal axes of inertia of the cross section ( they differ if the cross section is not circular ) , and @xmath4 is the twist modulus . in this paper we treat @xmath5 as given material parameters of the ribbon . the functions @xmath6 and @xmath7 are the generalized curvatures in the deformed and the stress - free states of the ribbon , respectively . these curvatures completely determine the three dimensional conformation of the ribbon through the generalized frenet equations@xcite @xmath8 here @xmath9 is the unit tangent to the centerline and the unit vectors @xmath10 and @xmath11 are oriented along the principal axes of the cross section ( @xmath12 is the antisymmetric unit tensor ) . note that since these equations describe a pure rotation of the triad @xmath13 as one moves along the contour of the ribbon , the constraint @xmath14 is automatically satisfied in this intrinsic coordinate description . the physical frame of the ribbon @xmath13 can be related to the original frenet description of space curves in terms of a unit tangent ( which coincides with @xmath9 ) , normal @xmath15 and binormal @xmath16 , provided that in addition to the frenet triad @xmath17 associated with the centerline of the ribbon , one introduces an angle @xmath18 that describes the rotation of two of the principal axes of the cross section @xmath10 , @xmath11 with respect to the binormal @xmath16 and normal @xmath19 ( see fig . 1 in the second of refs . [ ] ) , @xmath20the three dimensional configuration of the centerline is determined by the frenet equations @xmath21where the curvature @xmath22 and torsion @xmath23 are treated as known functions of position along the contour . the three generalized curvatures @xmath24 can be expressed in terms of these parameters and the angle @xmath25@xmath26 since the energy is a quadratic form in the deviations @xmath27 of the generalized curvatures from their values in the stress free state , eq . ( [ energy ] ) is valid as long as the characteristic length scale of the deformation is much larger than the diameter of the ribbon@xcite . when a ribbon undergoes fluctuations in the presence of a thermal bath at temperature @xmath28 this energy determines the statistical weight of the configuration @xmath29 . the statistical average of any functional of the configuration @xmath30 is defined as the functional integral@xcite @xmath31 where @xmath32 is the boltzmann constant . from the form of the energy , eq . [ energy ] , it follows that the distribution of @xmath27 is gaussian , with zero mean ( @xmath33 ) and second moments given by @xmath34 in order to study the distribution functions of the various fluctuating quantities ( e.g. , the extension @xmath35 ) , we developed what we entitled the frenet algorithm , which is based on the combination of statistical methods and differential geometry and is described in the following . examination of eq . ( [ corr ] ) shows that the decoupling of the `` noises '' @xmath36 in the intrinsic coordinate representation permits an efficient numerical generation of independent samples drawn from the exact canonical distribution . the gaussian distribution of the @xmath27 means that each @xmath37 can be directly generated as one of a string of independent random numbers drawn from a distribution symmetric about the origin with width @xmath38 where @xmath39 is the discretization step length . the remaining task is to construct the curve using the frenet equations with @xmath40 . the frenet equations are best integrated by stepping the basic triad @xmath41 forward in @xmath1 through a suitable small rotation . in this way , the orthonormality of the triad is guaranteed to be preserved up to machine accuracy . to construct this rotation matrix , we begin with eqs . ( [ frenet ] ) and , defining the three vectors @xmath42 , and so forth , we can write this equation as @xmath43 where @xmath44 is an antisymmetric matrix with elements @xmath45 . we now discretize eq . ( [ aeq ] ) as @xmath46 where the orthogonal matrix @xmath47 is @xmath48 the choice of discretization step length ( @xmath49 is the number of points ) depends on the amplitude of the random noise , @xmath50 we found that @xmath51 points gave sufficient accuracy ( i.e. , was limited by the statistical noise determined by the number of independent samples ) for the calculation of the distribution function of the extension , when the smallest of the moduli satisfied @xmath52 this algorithm is optimal for studying the fluctuations of a free ribbon . however , if the ribbon is acted upon by external forces and/or constraints , the algorithm must be modified . this is described in the next section . consider a filament with a circular cross section ( @xmath53 ) which is characterized by a bending persistence length@xmath54the simplest example of a filament under tension is that of a constant force acting on one end and directed along the @xmath55-axis , with the second end constrained to lie at the origin but otherwise free to rotate about it . the energy of the filament thus picks up an additional contribution @xmath56 , where @xmath55 is the projection of the end to end vector on the direction of the force . the problem is that @xmath55 is a complicated nonlinear function of all the @xmath57 , and so the external force destroys the gaussian structure of the energy on which the generation of the @xmath57 depends . one can simply incorporate the extra contribution to the energy by treating it as a weighting factor , or bias , and suitably normalizing all averages . thus , for example , one can calculate the mean value of @xmath55 via @xmath58where the sum is over configurations generated via the frenet algorithm described above and @xmath59 is the force in units of inverse length . for small forces , this works well . however , as the force increases , the probability of generating highly extended configurations which are heavily favored by the bias , is so small that the sampling becomes ineffective . one therefore needs a way to bias the generation of configurations towards those that are extended , so they can be sampled appropriately . this task is not simple , since the connection between the @xmath57 s through which the filament is parameterized and the value of @xmath55 is very complicated . we can gain insight into this by transforming to fourier space and looking at the expectation value of the moments of the @xmath60th fourier component of the curvature , @xmath61 , using the above algorithm , eq . [ dumb - bias ] , for small to intermediate forces where it is still effective . for an extensional force , we do not expect spontaneous curvature to develop , and indeed the first moments of the fourier components all vanish . the second moments , however , do exhibit an interesting structure , as seen in fig . [ kmoments ] , where the expectation value of the squared amplitude of the low @xmath60 fourier modes are presented for the case @xmath62 , @xmath63 . for @xmath64 , the second moments are independent of @xmath60 , as expected . when the filament is subjected to a tensile force we see that the small @xmath60 modes are suppressed , with the degree of suppression increasing with @xmath65 and decreasing with @xmath60 . in the case of large @xmath66 , where the fluctuations are in any case small , we observe that only the @xmath67 mode is significantly affected by the force ( not shown ) . the suppression of the various curvature modes can be calculated analytically when fluctuations are small , namely either in the large @xmath68 limit , or for any @xmath65 provided @xmath69 ( see appendix ) . the calculation confirms the above - mentioned trends . we learn from this that the high fourier modes ( or equivalently the small - scale fluctuations ) of the filament are essentially unaffected by the force , whereas the long - wavelength fluctuations are significantly damped . for different forces @xmath70 , for @xmath71 , @xmath62 . the lines are added to guide the eye.,width=480 ] this suggests the possibility of biasing the monte - carlo by selectively dampening the small-@xmath60 modes . the idea is to generate the @xmath72 s in fourier space , with appropriate amplitudes for the various modes , and construct the real - space @xmath73 via a fast fourier transform . of course , this bias has to be corrected explicitly in the weighting in the ensemble average . we do this in a self - consistent fashion , choosing the amplitude of each fourier mode to reproduce the measured square magnitude of the mode . in practice , it is sufficient to do this only for the first few modes , which are the ones most severely impacted by the force . also , in order to reduce the noise in the measurement of the mode amplitudes , we only calculate the averages after every block of @xmath74 realizations and update accordingly , where @xmath75 is the total number of realizations we intend to generate for these parameters . the force dependence of the mode amplitudes is obtained by a stepwise increase of @xmath65 , where for each value of the force we use as a starting point the amplitudes measured at the previous value . there is one other degree of freedom of the problem that also must be treated appropriately , namely the initial orientation of the filament at its constrained end . this is because the force not only straightens the filament , but also orients it toward its direction . we have employed two different algorithms to deal with this problem . the simpler approach is to simply bias the choice of initial orientation , and correct for this in the weighting . the zero force distribution of the orientation angle , @xmath76 , of the tangent vector with respect to the direction of the force is @xmath77 . in the presence of the force we modified this to @xmath78 where one can choose @xmath79 to reproduce the measured mean of @xmath80 . a more sophisticated approach is to explicitly perform the integral over @xmath76 . we generate a given configuration and calculate its end - to - end distance @xmath35 and angle @xmath81 that the tangent to the end of the filament makes with the direction of the end - to - end vector @xmath82 . in order to evaluate the contribution of this realization to the average elongation , eq . [ dumb - bias ] , one can perform a rigid rotation of this configuration about the end point and calculate the contribution to the statistical sum , from each angle of rotation @xmath76 . the resulting integrals over @xmath76 can be expressed in terms of @xmath35 and @xmath81 as @xmath83where the curly braces indicate the integrand for the numerator and denominator of the ensemble sum , respectively . these integrals can easily be evaluated numerically , using standard techniques . this second method is more efficient , but more involved to implement . using these techniques , we have measured the force - extension curve @xmath84 vs. @xmath85 . we present the results for @xmath86 , @xmath63 in the left box of fig . [ extenfig ] together with the interpolation formula of ms @xmath87we see that for this value of @xmath66 ( @xmath88 the agreement is very good . in the right half of fig . [ extenfig ] we present the results for @xmath89 , again with @xmath63 . here the ms formula does not work for small and intermediate values of the force , since we are not in the long - chain limit ( @xmath90 ) for which it was derived . on the other hand , for this case analysis can also be done , as presented in the appendix , and the analytic formula derived there is in perfect agreement with the simulations ( see the right box of fig . [ extenfig ] ) . notice that the ms expression systematically overestimates the extension of the filament and approaches our results only in the infinite force limit . the physical basis of the deviation is the large orientational entropy which is not accounted for in the ms calculation . inspection of the right box of fig . [ extenfig ] suggests that both the simulation and our analytical approximation ( eq . [ frey ] ) yield a finite initial slope of the extension versus force curve in the rigid rod limit ( @xmath91)@xmath92 since the effective spring constant is proportional to the inverse of the initial slope of this curve this suggests that the linear response of the rigid rod to small stretching forces does not depend on the persistence length @xmath66 and is a function of its length @xmath93 only . the above expectation is confirmed by calculations reported in appendix b where we show that the corresponding spring constant is @xmath94 . obviously , this behavior differs from the flexible chain limit ( @xmath95 ) for which the spring constant scales as@xcite @xmath96 it also differs from the predictions of ref . [ ] where it was argued that the effective spring constant diverges in the rigid rod limit , as @xmath97 . notice that such a divergence is not expected for a single filament under tension since there is a finite contribution to the susceptibility ( and thus to the response to the external force ) due to orientational fluctuations , even in the rigid rod limit . the difference between our result and that of ref . [ ] stems from the fact that the latter considers a network in which filaments are not free to rotate and the only angular averaging is over the randomly distributed orientations of different filaments in the network . ) for @xmath71 . left : @xmath98 ; right : @xmath99,title="fig:",width=307 ] ) for @xmath71 . left : @xmath98 ; right : @xmath99,title="fig:",width=307 ] not only can we calculate the average @xmath84 , but we can also measure the entire distribution @xmath100 . we show this in fig . distribut - nowall , for @xmath98 and @xmath101 . in the former case , the persistence length is much shorter than length of the filament and the force - free distribution is nearly gaussian . the peak of the distribution moves to higher values of the extension and its width decreases with increasing force . in the latter case ( persistence length approximately equal to the length of the filament ) , the force - free distribution is nearly flat up to @xmath102 as expected for random sampling of orientations by a nearly rigid rod ( due to thermal fluctuations , the effective length of this rod is somewhat shorter than its contour length , @xmath103 ) . the application of small force ( in the range @xmath104 ) orients the filament and leads to dramatic change of the distribution . at larger forces , the distribution changes slowly with the force , in a manner similar to the case @xmath98 . the transition from orientation - dominated to stretching - dominated behavior of nearly rigid rods is clearly observed in the right box of fig . [ extenfig ] , where pronounced deviations from the marko - siggia formula are observed in the weak force regime ( orientational effects are not taken into account in the ms theory ) . , @xmath105 , @xmath106 and @xmath107 , with @xmath71 . left : @xmath98 ; right : @xmath108,title="fig:",width=307 ] , @xmath105 , @xmath106 and @xmath107 , with @xmath71 . left : @xmath98 ; right : @xmath108,title="fig:",width=307 ] it is fairly straightforward to incorporate steric constraints into our calculations . as an example , we consider the case where the filament is confined to the upper half space @xmath109 . after generating each configuration according to the procedure outlined in the previous section , we simply reject any configuration which violates the constraint . our method of integrating over the initial orientation angle @xmath76 can also be employed . we have simply to compute the limits of the integral over @xmath76 which can be done straightforwardly given the configuration . we present the results for the force - extension curve in fig . [ force - ext - wall ] and for the distribution @xmath100 in fig . [ xdist - wall ] . we see that , as one might expect , the effect of the constraint is most pronounced at small @xmath65 , where the straightening / aligning of the curve is smallest . for vanishing @xmath65 , the average extension is positive as a result of the constraint . the constraint basically carves out a `` hole '' in the distribution at small @xmath55 , resulting in a very asymmetric distribution for small force . for @xmath110 the distribution function is practically unaffected by the presence of the wall and coincides with that shown in fig . [ distribut - nowall ] . for @xmath71 , measured via our monte - carlo procedure . left : @xmath98 , together with the marko - siggia formula ; right : @xmath111 , together with our formula , eq . ( [ zunif]).,title="fig:",width=307 ] for @xmath71 , measured via our monte - carlo procedure . left : @xmath98 , together with the marko - siggia formula ; right : @xmath111 , together with our formula , eq . ( [ zunif]).,title="fig:",width=307 ] , @xmath105 , with @xmath71 . left : @xmath98 ; right : @xmath111 . distributions at higher forces are essentially unaffected by the constraint.,title="fig:",width=307 ] , @xmath105 , with @xmath71 . left : @xmath98 ; right : @xmath111 . distributions at higher forces are essentially unaffected by the constraint.,title="fig:",width=307 ] in this work we used biased monte - carlo simulations , based on an extension of our frenet algorithm , to calculate the force - extension relations and the distribution functions of the extension for filaments of arbitrary persistence to contour length ratio . we showed that in the presence of external force , curvature fluctuations are no longer gaussian and that their long wavelength fourier modes are progressively suppressed by the force . this effect was taken into account in a self - consistent manner in the simulations by gradually increasing the force and adjusting the corresponding fourier amplitudes of curvature fluctuations . analytic results were obtained both for small fluctuations about the rigid rod limit , and in the limit of large force . in the former case , we calculated the force - induced renormalization of the length - scale - dependent bending moduli that govern the various fourier modes of curvature . as expected , the largest effect is on the zeroth ( constant curvature ) mode ; the corresponding modulus increases quadratically and linearly with the force in the weak and strong force limits , respectively . the underlying physical picture is reminiscent of the suppression of excluded volume by tensile force in the self - avoiding random walk model of polymers - while large scale fluctuations are strongly suppressed , small scale fluctuations ( within the so called pincus blobspincus ) are not affected by the force . for flexible filaments with @xmath95 our simulation reproduces the marko - siggia force - extension curves . however , our simulations and analytic results show that as the rigid rod limit @xmath91 is approached , the ms expression increasingly overestimates the response of the filament at finite values of the force , and approaches the correct result only asymptotically , in the limit of infinite force . the discrepancy can be attributed to the neglect of orientational effects ( the constant curvature mode is not taken into account ) in the ms derivation which is strictly valid only in the @xmath112 @xmath113 limit . these orientational effects are however seen in the short - chain limit of the gaussian model investigated recently by winklerwinkler . the large force limit of the gaussian model for short chains is however very different from the worm - like chain model studied herein . for long chains , the gaussian model can reproduce either the weak or the strong force limit , by appropriate choice of the persistence length , but not both this is an outcome of the length - scale dependence of the effective moduli discussed above and in appendix c. examination of the distribution functions for the extension shows that increasing rigidity leads to a more rapid shift of the peak of the distribution and narrowing of its width , with external force . the largest deviations from gaussian statistics of the distributions are observed for stiff filaments in the weak to intermediate force range ( @xmath104 ) , where fluctuations of initial orientation are not yet completely suppressed and where wall effects play an important role . while it may be difficult to approach the rigid rod limit in mechanical experiments on dsdna that has a persistence length of @xmath114 , it can be readily accessed by experiments on actin filaments@xcite for which the persistence length of order @xmath115 @xmath116 m . finally , we would like to comment on possible extensions of the present work . since we only considered here filaments with symmetric cross section and without spontaneous curvature , twist rigidity did not affect the distribution of extension and played no role in our considerations . however , since twist rigidity affects even the equilibrium distributions of ribbons with strongly asymmetric cross sections@xcite , it is expected that both twist rigidity and spontaneous twist will affect the deformation of such ribbons as well . perhaps even more interesting is the effect of spontaneous curvature . indeed , we have shown that a helical filament with sufficiently large twist to bending rigidity and radius to pitch ratios , undergoes a sequence of stretching instabilities under the action of tensile force , even in the absence of any thermal fluctuations@xcite . in the presence of thermal fluctuations or mechanical noise these instabilities are expected to lead to coexistence between different shapes of the deformed object@xcite . dak and yr acknowledge the support of the israel science foundation . we also would like to thank j. fineberg and g. cohen for stimulating discussions . since we are interested in three dimensions , we work in terms of the euler angles @xmath117 , @xmath118 and @xmath119 for convenience we will orient the force along the @xmath55-axis , i.e. along the direction @xmath120 @xmath121 . the energy due to the force is then @xmath122 the euler angles satisfy the differential equations@xcite : @xmath123 we need to expand to quadratic order about the @xmath124 values @xmath125 and @xmath126 , and to linear order about @xmath127 , ( since the energy does not involve @xmath128 directly ) giving @xmath129 where @xmath130 and @xmath131 are the rotated curvatures @xmath132 thus , to second order in the curvatures @xmath133 we will first compute the force - extension curve , via the partition function . we do this in two stages , depending on the value of the parameter @xmath134 . here we write @xmath138where @xmath139 is given by @xmath140 \right . \notag \\ & \ \hspace{0.5in}\left . -\cos \theta _ { 0}\cos \phi _ { 0}\int_{0}^{s}\,ds^{\prime } \lambda ( s^{\prime } ) \int_{0}^{s^{\prime } } \omega _ { 3}-\sin \phi _ { 0}\int_{0}^{s}\,ds^{\prime } \kappa ( s^{\prime } ) \int_{0}^{s^{\prime } } \omega _ { 3}\right ] \end{aligned}\]]we expand @xmath141 to quadratic order , yielding @xmath142 } \notag \\ & \ \times \left [ 1-f\delta + \frac{f^{2}}{2}\delta ^{2}\right ] \end{aligned}\]]where @xmath143 . the functional integrals are all gaussian and can be easily done by going to fourier space . for example , @xmath144so that @xmath145where the mixed terms will vanish latter upon functional integration and are omitted in the following . defining @xmath146and carrying out the functional integration by integrating over all the fourier components @xmath147we get @xmath148where @xmath149 is the ( normalization ) integral over the exponential factor only . similarly , @xmath150 ^{2}=\int \mathcal{d}_{0}\kappa \left [ \frac{\kappa _ { 0}l^{2}}{2}+\sum_{n=1}^{\infty } \frac{\kappa _ { n}^{s}l^{2}}{2\pi n}\right ] ^{2}=\mathcal{n}\frac{l^{3}}{3a}.\]]the functional integration with respect to @xmath151 is carried out in the same manner . putting this all together , we get @xmath152\]]the integrals over solid angle are as follows : @xmath153collecting all terms we finally get @xmath154 \right\ } \label{z - smallchi}\ ] ] from this , we can differentiate to get the average projection of the end - to - end vector on the @xmath55-axis , @xmath155 @xmath156 \label{frey}\]]at large @xmath157 ( but still small @xmath158 ) , this goes as @xmath159 . for small @xmath157 this projection goes linearly with @xmath85 as @xmath160 and we conclude that in the limit @xmath91 the effective `` spring constant '' is @xmath94 . we start by expanding @xmath125 , @xmath126 about the values @xmath161 , @xmath162 which minimize @xmath163 . to lowest order , @xmath164 so that the expansion reads @xmath165 + \frac{fl}{2}\left [ \left ( \theta_{0}-\theta_{0}^{\ast}\right ) ^{2}+\left ( \phi_{0}-\phi_{0}^{\ast } \right ) ^{2}\right ] \notag \\ & \ + \frac{f}{2}\int_{0}^{l}\,ds\left [ \left ( \int_{0}^{s}ds^{\prime}\kappa\right ) ^{2}+\left ( \int_{0}^{s}ds^{\prime}\lambda\right ) ^{2}\right]\end{aligned}\ ] ] let us now consider the functional integral over @xmath130 , the integral over @xmath131 being exactly the same . we have @xmath166 ^{2 } } \notag \\ & = \int\mathcal{d}\kappa\exp\left\ { -\frac{al}{2}\kappa_{0}^{2}-\frac{al}{4}\sum_{n=1}^{\infty}\left ( ( \kappa_{n}^{c})^{2}+(\kappa_{n}^{s})^{2}\right ) \right . \notag \\ & \ \left . { } -\frac{fl^{3}}{2}\left [ \frac{\kappa_{0}^{2}}{12}+\sum _ { n=1}^{\infty}\left ( \frac{(\kappa_{n}^{c})^{2}}{2(2\pi n)^{2}}+\frac{(\kappa_{n}^{s})^{2}}{2(2\pi n)^{2}}-\frac{2\kappa_{0}\kappa_{n}^{c}}{(2\pi n)^{2}}\right ) \right ] \right\}\end{aligned}\ ] ] we can diagonalize this expression by shifting @xmath167 , defining @xmath168 with @xmath169 yielding an effective modulus of the zero mode @xmath170 we can now calculate @xmath171 : @xmath172 ^{1/2}\end{aligned}\ ] ] where @xmath173 is the @xmath130 partition function at @xmath174 . the entire partition function is then @xmath175 for small @xmath176 , this yields @xmath177 which agrees precisely with the large @xmath85 expansion of eq . ( [ z - smallchi ] ) . the average extension in this second region ( @xmath178 ) is given by @xmath179 which for small @xmath176 gives @xmath180 in accord with the large @xmath157 limit of the first region ( @xmath181 ) . at large @xmath157 , it agrees with the marko siggia result , eq . ( [ mc ] ) . a uniform approximation for the force - extension relation is then @xmath182 this result is confirmed by simulations , as shown in data for @xmath89 , fig . [ extenfig ] . it even works well when the persistence length slightly shorter than the chain length , as can be seen in the data for @xmath111 in the right half of fig . [ force - ext - wall ] . since our analysis is limited to the small fluctuation ( stiff rod and/or large force ) regime , we expect it to break down for very short persistence lengths in the small force limit . this is clearly observed in the case of @xmath98 , shown in the left half of fig . [ extenfig ] . we now turn to the computation of the effective moduli of the various modes . we first treat the zero modes , starting as above with the small @xmath176 limit . we proceed by calculating a modified partition function , @xmath183 with a modified modulus of the zero modes , @xmath184 , replacing the bare modulus @xmath66 . then @xmath185 tracking the changes induced in the partition function calculation above , we find @xmath186 \left [ \frac{\cosh fl}{fl}-\frac{\sinh fl}{(fl)^{2}}\right ] \right\}\ ] ] giving @xmath187 so that @xmath188\ ] ] this implies an effective zero mode modulus of @xmath189 at small and large forces , this reads @xmath190 at large @xmath85 , the effective zero mode modulus was calculated above , eq . ( a0highf ) . the small @xmath176 limit of this agrees with the large force limit of eq . ( [ a0lims ] ) , as it must . a uniform approximation for @xmath191 reads @xmath192 this result is confirmed by the simulations as seen in fig . [ 0modeamp ] . now we turn to the finite wave - number modes . the interesting thing to note here is that the _ cos _ modes renormalize differently than the _ sin _ modes , due to the coupling of the zero mode to the _ cos _ modes . at large force , the effective modulus of the _ sin _ modes is the trivial @xmath193 so that the long wavelength modes are the most effected by the force . the effective modulus of the _ cos _ modes follows from @xmath194 with @xmath195 this expression is fairly messy , so let us examine its various limits . for large enough mode number , the effective modulus is @xmath196 as for the _ sin _ modes . for intermediate forces , all modes are large enough for this to be the case . for strong forces , however , the effective modulus is @xmath197 up to some mode number , at which point it crosses over to the previous result .	we develop a novel biased monte - carlo simulation technique to measure the force - extension curves and the distribution function of the extension of fluctuating filaments stretched by external force . the method is applicable for arbitrary ratio of the persistence length to the contour length and for arbitrary forces . the simulation results agree with analytic expressions for the force - extension curves and for the renormalized length - scale - dependent elastic moduli , derived in the rigid rod and in the strong force limits . we find that orientational fluctuations and wall effects produce non - gaussian distributions for nearly rigid filaments in the small to intermediate force regime . we compare our results to the predictions of previous investigators and propose new experiments on nearly rigid rods such as actin filaments .
searching for possible exotic states of nuclear matter is one of the currently most interesting and stimulating subjects in high energy nuclear physics ( see references [ 110 ] ) to create exotic matter such as the quark - gluon plasma ( qgp ) [ 14,5]strange matter @xcite,@xcite , multi@xmath11-matter , pion condensate , kaon condensate @xcite,@xcite it is essential to generate extreme states of nuclear matter regarding density , temperature or energy density @xcite . such extreme states are considered to be produced most probably in high energy heavy ion reactions[1014 ] or in the annihilation of energetic antiprotons(@xmath12)on a nucleus @xcite the possibility suggested by witten @xcite that strange quark - matter might be the lowest energy state of large baryon number , would have many consequences in physics and astrophysics and also cosmology relevance [ 1620 ] . small drops of strange matter(strangelets ) may be natural results of the evaporation of a qgp formed in heavy ion collisions . enhanced strangeness production @xcite in ultrarelativistic heavy ion collisions ( urhic ) is one of the most widely discussed signature for creating qgp @xcite,@xcite . several experiments have been performed to investigate strangeness production in reactions involving relativistic nuclei , particularly at super proton synchrotron ( sps ) at cern : na35 [ 2232 ] , na36 [ 3337],wa85,wa94[3845],[128130],na44 [ 132133 ] , hellios collab.[4650 ] , and at the alternating gradient synchrotron(ags ) at brook- + haven e802 [ 5155 ] , e810 @xcite,@xcite and e814 @xcite,[134 ] . these experiments have measured also the yield of several particle species containing one or two strange quarks ( @xmath13 , @xmath14,@xmath15,@xmath16 and their antiparticles ) or three strange quarks(@xmath17 ) . in all the cases their production relative to pions or negative hadrons and some other ratios are larger in nucleus - nucleus collisions compared to proton - proton . now experiments with truly heavy ion projectile have just started(with a gold beam at bnl)@xcite,@xcite or are under way ( @xmath18 at cern)@xcite . the soft physics data have stimulated a great deal of theoretical activity and in the present introduction it is impossible to give more than a superficial overview of the main types of the monte carlo models used.many models for urhic have been developed : dual partons models(dpm ) [ 5966],@xcite quark gluon string models(qgsm)[6769 ] , venus models for very energetic nuclear scattering [ 7075 ] fritiof model @xcite,@xcite , attila @xcite model , relativistic quantum molecular dynamics ( rqmd ) [ 7983][135138],parton string model(psm)[139 ] , hijet model@xcite . an excelent review and detailed comparision of the models was done by werner @xcite . on the other hand , no unconventional explanation has so far put forward for the observed increase of antyhyperons ( @xmath19,@xmath16 ) . actually , it is the old pomeron exchange picture that is revitalized in some models with a new ingredient that is the pomeron is made of two colored strings.thus the problem of interaction of strings becomes important.this interaction may give rice to the fusion of strings[8588][137,138 ] making a new string with more color charge.these new objects could explain the strangeness enhancement @xcite , @xcite,[137,138]especially @xmath19 enhancement . such models can be seen as an interplay between independent string fragmentation and qgp formation.one version of venus generator also include such interaction @xcite , @xcite . we note also that an interpretation based on a particular dependence of the cross sections for urhic was taken into account @xcite . in this report we have performed a systematic study of strange particle production in proton proton(@xmath20),proton nucleus(@xmath21 ) and nucleus nucleus ( @xmath22 ) interactions at sps cern energies using hijing(heavy ion jet interacting generator ) monte carlo model [ 9197 ] developed at berkeley . the formulation of hijing is guided by the lund fritiof ( @xcite@xcite ) and dpm @xcite models phenomenology for soft nucleus - nucleus reactions at intermediate energies @xmath23 and the successfull implementation of perturbative qcd ( pqcd ) processes in the pythia model @xcite@xcite for hadronic interactions . hijing incorporate the pqcd to multiple jet processes and the nuclear effects such as parton shadowing and jet quenching . hijing provide a link between the physics at intermediate cern sps energies and the highest collider energies at relativistic heavy ion collider(rhic)(@xmath24 and the cern large hadron collider ( lhc ) ( @xmath25 ) . using hijing monte carlo model @xcite,@xcite a detailed discussion and comparison with a wide variety of data in @xmath20,@xmath21 and @xmath22 collisions was reported.however no study on the strangeness production at sps - cern energies in @xmath20,@xmath21 and @xmath22 data exist within this approach . in this paper we present some theoretical calculations for mean multiplicities of strange particle production in the full rapidity range as well as their ratios to negative hadron predicted by hijing model(hij.01 version -see section 2 and 3 for explanation of this version)for @xmath20 , nucleon - nucleon @xmath26,minimum biascollisions @xmath27 and + central collisions@xmath28 at the energy of @xmath29 . also a qualitative prediction for normalised distributions for rapidity , transverse kinetic energy and transverse momenta are done at 200 gev / nucleon in @xmath30 and sau collisions and the results are compared with recent experimental data from na 35 collaboration[140].we note that theoretical calculations are done for full phase space . for giving an idea of abundances and spectra in experiments planned in the near future at cern , we give some predictions for sw and pbpb interactions . we note that the version hij.01 of the model includes neither the formation of qgp nor some collective dynamics on either the hadronic or string level.although the existing trends of hadron distributions are quite satisfactorily approximated by the hijing model @xcite,@xciteand also neutral strange production in @xmath20 and @xmath21 interactions are quite well described , this version is getting less reliable for strange particle produced in heavy colliding systems . the same result is stressed out and in other models qgsm@xcite,@xcite , rqmd@xcite , fritiof@xcite , psm@xcite . a brief description of the hijing monte carlo model and theoretical background are given in section 2.detailed numerical results of this version hij.01 for @xmath31 and @xmath32 reactions especially at cern - sps energies@xmath33 for strangeness production are compared to experimental data and other models predictions like qgsm @xcite,@xcite , dpm @xcite,[141],142],venus @xcite , fritiof @xcite , and rqmd @xcite,[140 ] in section 3 . section 4 concludes with a summary and discussion of future applications . a detailed discussion of the hijing monte carlo model was reported in references @xcite,@xcite,@xcite . the formulation of hijing was guided by the lund - fritiof and dual parton model phenomenology for soft nucleus - nucleus reactions at intermediate energies ( @xmath34)and implementation of perturbative qcd(pqcd ) processes in the phythia model for hadronic interactions.comparison between some monte carlo generators are given in reference @xcite all pqcd models employ initial and final state radiation in an iterative fashion : a parton with a squared mass @xmath35 radiates a parton and leaves a remainder parton with reduced @xmath35,the latter one radiating again and so on .the hadronization finally is done by ussing string model(pythia).a link to the soft semihard models is provided by pythia , where it was realized that the perturbatively calculated cross section is an inclusive one and may well exceed the total cross section meaning multiple scattering.therefore the cross sections was eikonalized : for a given impact parameter multiple scattering occurs acording to a poissonian distribution , with the average number of scatterings , being @xmath36dependent , proportional to some profile function @xmath37.hijing is based on a particular model of high energy pp inelastic collisions . we give in this section brief review of the aspect relevants to hadronic interaction : 1 . the models includes multiple minijet production with initial and final state radiation along the lines of the pythia model @xcite,@xcite and with cross sections calculated within the eikonal formalism . soft beam jets are modeled by quark - diquark strings with gluon kinks along the lines of the dpm and fritiof models.multiple low @xmath38 exchanges among the end point constituents are included . exact diffuse nuclear geometry is used to calculate the impact parameter dependence of the number of inelastic processes @xcite . an impact parameter dependent parton structure function is introduced to study the sensitivity of observable to nuclear shadowing , especially of the gluon structure functions . 5 . a model for jet quenching is included to enable the study of the a - dependence of moderate and high @xmath38 observable on an assumed energy loss of partons traversing the produced dense matter . hijing does not incorporate the mechanism for final state interactions among low @xmath38 produced particles . the rate of multiple minijet production in hijing is constrained by the cross sections in nucleon - nucleon collision .within an eikonal formalism @xcite the total elastic cross sections @xmath39,total inelastic cross sections @xmath40 and total cross sections @xmath41 can be expressed as : 0.3 cm @xmath42 0.3 cm @xmath43 0.3 cm @xmath44 0.3 cm strong interactions involved in hadronic collisions can be generally divided into two categories depending on the scale of momentum transfer @xmath45 of the processes . if @xmath46 the collisions are nonperturbative and are considered _ soft_and if @xmath47 the subprocesses on the parton level are considered _ hard _ and can be calculated via pqcd @xcite . in the limit that the real part of the scattering amplitude is small and the eikonal function @xmath48is real , the factor 0.3 cm @xmath49 0.3 cm in terms of semiclassical probabilistic model can be interpreted as _ the probability for an inelastic event of nucleon nucleon collisions at impact parameter @xmath36 , which may be caused by hard , semihard or soft parton interactions_. to calculate the probability of multiple minijet a main dynamical assumption is that they are independent.this holds as long as their average number is not too large.the independence should apply up to lhc energies @xcite when shadowing can be neglected , the probability of no jets and @xmath50 independent jet production in an inelastic event at impact parameter @xmath36,can be written as : 0.3 cm @xmath51 0.3 cm @xmath52^{j}}{j!}\cdot exp(-2\,\chi_{h}(b , s))\,\,\,\,\,\ , j \geq 1 \label{e6}\ ] ] 0.3 cm where @xmath53 is the eikonal function for soft interaction , @xmath54 is the average number of hard parton interactions at a given impact parameter , @xmath55 is the probability for no soft interaction . summing eqs.(5 ) and ( 6 ) over all values of @xmath50 leads to : 0.3 cm @xmath56 comparing with eq.([e4 ] ) one has : 0.3 cm @xmath57 0.3 cm if we consider that the parton distribution function is factorizable in longitudinal and transverse directions and the shadowing can be neglected the average number of hard interaction @xmath58 at the impact parameter @xmath36 is given by : 0.3 cm @xmath59 0.3 cm where @xmath60 is the effective partonic overlap function of the nucleons at impact parameter @xmath36 . 0.3 cm @xmath61 0.3 cm with normalization 0.3 cm @xmath62 0.3 cm and@xmath63is the is the pqcd cross section of parton interaction or jet production @xcite,@xcite if we note @xmath64,where @xmath65provide a measure of the geometrical size of the nucleon @xmath66 we get assuming the same geometrical distribution for both soft and hard overlap functions 0.3 cm @xmath67 @xmath68 @xmath69\chi_{0}(\xi ) \label{e14}\ ] ] 0.3 cm we note that@xmath70 is a function not only of @xmath71 but also of @xmath72because of the @xmath73 dependence on the jet cross section @xmath74.geometrical scaling implies on the other hand , that @xmath75 is only a function of@xmath76 . thefore geometrical scaling is broken at high energies by the introduction of the nonvanishing @xmath77 of jet production . now we can rewrite the cross sections of nucleon - nucleon collisions as : 0.3 cm @xmath78 @xmath79 @xmath80 0.3 cm the calculation of these cross sections requires specifying @xmath81 with a coresponding value of cut - off momenta @xmath82 @xcite,@xcite in the energy range @xmath83,where only soft parton interactions are important cross sections @xmath81 is fixed by the data on total cross sections @xmath84 directly.in and the above @xmath85 energy range @xmath86 we fix @xmath81 at a value of 57 mb with @xmath87,in order to fit the data of the cross sections.between the two regions @xmath88,we simply use a smooth extrapolation for @xmath81@xcite . in hijing a nucleus - nucleus collisions is decomposed into binary collisions involving in general excited or wounded nucleons . wounded nucleon are assumed to be @xmath89 string like configurations that decay on a slow time scale compared to the collision time of the nuclei.in the fritiof scheme wounded nucleon interactions follow the same excitation law as the original hadrons . in the dpm scheme subsequent collisions essentially differ from the first since they are assumed to involve sea partons instead of valence ones.the hijing model adopt a hybrid scheme , iterating string - string collisions as in fritiof but utilizing dpm like distributions . by incorporating the successful multistring phenomenology for low @xmath38 interactions at intermediate energies , hijing provide a link between the dominant nonperturbative fragmentation physics at intermediate cern - sps energies and perturbative qcd physics at the highest collider energies(rhic;lhc ) strange particle production provide some information about the question of whether or not a quark - gluon plasma occurs in heavy ion collisions.this is major motivation to first study the production of strange particles in pp scattering.in addition , strange particle give some insight into dynamics already for pp collisions . we run the program hijing , with default parameters , mainly ihpr2(11)=1 , which means choise of baryon production model with diquark - antidiquark pair production allowed , initial diquark treated as unit ; ihpr2(12)=1 , decay of particle such as @xmath90,@xmath91,@xmath15,@xmath92,@xmath93 , @xmath94 are allowed;ihpr2(17)=1 - gaussian distribution of transverse momentum of the sea quarks ; ihpr2(8)=0 - jet production turned off for theoretical predictions in hijing model- hij.01 , and ihpr2(8)=10-the maximum number of jet production per nucleon - nucleon interaction for for theoretical predictions @xmath95 . the energy dependence of average multiplicities for different particle species , gives a first impression about the predictive power of the model and provide an important check of model approach . in table 1 we give average multiplicities of particle at @xmath96 in proton - proton interaction , @xmath97 being the laboratory energy . the theoretical values @xmath98 are obtained for @xmath99 generated events and in a full phase space . the values @xmath95 are for very central @xmath20 collisions(@xmath100 ) and with minijet production . the experimental data are taken from reference @xcite . the large difference between pions and kaons is due to the suppressed strangeness production from string fragmentation.pions(@xmath101 ) and kaons(@xmath101 ) are more frequent than the negative ones due to charge conservation.we note that the multiplicities for neutral strange particle @xmath102 and for antiproton @xmath103are quite well described in the limits of three standard error deviations for @xmath20 interactions at @xmath104 .however the values for @xmath105 and @xmath106 are slightly overpredicted by the model . in order to give some comparision at ultrahigh energy and to test the predictive power of generator we perform calculations at centre of mass energy @xmath107(@xmath85-energies ) , for @xmath108 interactions . by using many pieces of data from the different collider experiments ( charge particle @xmath109 and @xmath110 from * ua5 * , @xmath111 from * ua2 * ) , the * ua5 * collaboration has attempt to piece together a picture of the composition of a typical soft event at the @xmath85 @xcite.the measurements were made in various different kinematic regions and have been extrapolated in the full transverse momenta(@xmath38 ) range and rapidity range for comparison as described in reference @xcite . the experimental data are compared with @xmath112 results in table 2 . the number of generated events was @xmath113 . it was stressed out @xcite that the data show a substantial excess of photons compared to the mean @xmath114.it was suggested like explanation the possible emission of gluon cerenkov radiation in hadronic collision @xcite.at this energy the model @xmath95 seems to work fairlly well and the data are well described . our calculations ruled out such hypothesis.taken into account direct gamma production the agreement with the data is in the limit of experimental errors . the experimental ratio @xmath115 are predicted also by@xmath116 model ( 0.99 ) . we note that a study of the ratios of invariant cross sections of kaons to that of pions as a function of transverse momenta in the central region was done by wang and gyulassy @xcite,@xcite . the behaviour of average multiplicities for different particles species as a function of the energy can be reproduced by hijing model@xcite and also by venus model(see reference @xcite ) . appropriate variable to describe single particle properties are the transverse momentum @xmath38 and the rapidity @xmath117 0.3 cm @xmath118 0.3 cm with @xmath119,and @xmath120 being energy , longitudinal momentum and transverse mass 0.3 cm @xmath121 0.3 cm with @xmath122 being particle mass . in figures 1a , b and 2a , b we show normalized rapidity ( 1a,2a)and transverse momentum distributions(1b,2b ) for @xmath123 s ( fig.1a , b ) and for @xmath124 s for@xmath125 scattering at@xmath126 . the theoretical values hij.01 are compared with experimental data taken from jaeger et al.@xcite hijing results and data @xcite are very similar , although @xmath127 does not show some structures present in data . we see the two distinct lambda peaks , being due ( in zeroth order)the fragmentation of the forward baryonic string and the backward baryonic string into leading baryons.therefore lambda distributions are similar to proton distributions , just reduced by some factor and with the diffractive peak missing(see also reference @xcite ) . the kaons are peaked at mid rapidity and are similar to the pions , though again less in magnitude.the distributions are narrower compared to proton s and pion s . in the following we investigate strange particle production in hijing model approach , for proton - nucleus and nucleus- nucleus collisions . we have studied the production of neutral strange particle @xmath128 in comparison with recent experimental data from na35 cern - experiments , for @xmath129 interactions . we investigate the average multiplicities for negative hadrons @xmath2,negative pions(@xmath130 and neutral strange particles ( @xmath131,@xmath132 , @xmath106 ) in @xmath133 minimum bias collisionsand @xmath134 central collisions at @xmath135 . the default parameters were used during the simulation and also option jet no(ihpr2(8)=0 ) . the number of monte carlo generated events was @xmath99 for @xmath136 and @xmath137interactions , @xmath138 for@xmath139 and @xmath140 for @xmath141 and @xmath142 collisions . the results are given in tabel 3 in comparison with experimental data @xcite.the corresponding data for @xmath20 and @xmath143 at 200 gev @xcite are also shown in table 3 . we remark that hijing model describe quite well neutral strange particle multiplicities for @xmath136 and @xmath144 interactions , but underestimate by a factor of two the values for nucleus - nucleus interactions for @xmath132,@xmath106,@xmath131 in symmetric collisions @xmath145(expected values ) and for @xmath132 , @xmath106 in assymmetric ones ( @xmath146 ) . we note also that our theoretical calculations are done for @xmath144 minimum bias collisions and the experimental data are for the events with charged particle multiplicity greater than five , which contain a significant fraction ( about 90 % ) of the minimum bias events @xcite . in order to get some idea about the discrepancies between theoretical calculations and experimental data we try to run the program in the limiting condition for very central events ( @xmath147 ) and jet allowed . the results are given in table 4 and are compared also with recent theoretical approach in rqmd @xcite , qgsm @xcite,@xcite and dpm models ( @xmath148 are from reference @xcite , variant of dpm which include additionally @xmath149 production from the sea into the chain formation process , @xmath150 are from reference @xcite , variant of dpm which include chain fusion , good candidate to explain the anomalous antihyperon production ) . in order to study the total production of strangeness in these collisions in a model independent way and to use all available experimental information na35 collaboration @xcite has studied the ratio defined as : 0.3 cm @xmath151 0.3 cm we estimate this ratio in hijing approach for the above interactions and give the numerical results in table 5 together with the mean multiplicities for negative pions . we can conclude from this analysis that the factor of two for @xmath152 which was reclaimed from multiplicities analisis for symmetric interactions are ruled out.experimental error are to higher and new data with higher statistics are needed for draw a definite conclusions . taken into account a factor of two in neutral strange particle production we get a value @xmath153 independent of the target nucleus mass number for @xmath22 interaction from @xmath154 to @xmath142 . without factor of two the theoretical @xmath155 values for @xmath144 and @xmath32 are aproximatively equal to the values for nucleon - nucleon interactions . however the experimental values of @xmath155 are two times higher then the corresponding @xmath156 and @xmath144 values , a fact established from recent na35 data . therefore new ideas are required to modify hadronic models in order to reproduce strange particle production at high energies . a different approach of finding signals for qgp is simply to compare nucleus - nucleus ( aa ) or hadron - nucleus(ha ) data with nucleon - nucleon(nn ) data . however , in particular this assumption is certainly not correct , since at high energies aa scattering is not simply a superposition of nn collisions.the time scale between two nn interactions is too small for hadronization and therefore some intermediate object is involved . simple theoretical estimates are very unrealiable .comparing aa with nn data or pa and nn data does not help , because new results do not necessarily mean qgp.calculation based on qcd are not feasible , because we are in domaine where pqcd does not apply . soft interactions involving small transverse momenta transfer should be taken into account . this ratios for multistrange particles were analised in terms of thermal models@xcite,@xcite , + @xcite , @xcite . a constraint between thermal fireball parameters arises from the requirement that the balance of strangeness in a fireball is nearly zero.the impact of this constraint on ( multi-)strange ( anti-)baryon multiplicities comparing hadron gas and qgp has been analised.the data are compatible with the qgp hypothesis and appear to be inconsistent with the picture of an equilibrated hadron gas fireball @xcite,@xcite . however the two scenarios ( slow and rapid)for the expansion of a qgp were recently studied in detail .production ratios @xmath157 and @xmath158 for hadron gas at @xmath159 and plasma break - up have been compared to experimental data at baryonic chemical potential @xmath160.present data on strange particle production can not provide a distinction between the two scenarios @xcite . we calculate the ratios of the mean strange particles multiplicities to the mean multiplicities of negative hadrons and we show in table 6 a comparison with experimental na35 data @xcite for the same interactions like in table 3.the agreement can not be obtained without factor of two in neutral strange multiplicities . we note that negative hadrons as well as negative pions are well described even for very complex @xmath32 interactions . in tabel 7 we give the multiplicity ratios for negative hadrons as well as for neutral strange particle for pa / pp , aa / pp and aa / pa resuts from hijing ( @xmath1 ) and @xmath161 models are compared to na5 @xcite and na35 @xcite experimental data.the calculations on the models @xmath1 and @xmath161 reproduce the relative increase of the @xmath91 ratios in central @xmath162 collisions , but can not describe the data concerning @xmath15 and @xmath19 generation . comparing experimental data with theoretical prediction of the hijing model we see that these ratios can not be described witouth factor of two for @xmath163 and @xmath164 . data for charge kaon production have already been published by helios collaboration for @xmath165 and @xmath166 at @xmath167 in the limited target rapidity region @xmath168 and by na35 collab . for @xmath169and @xmath170 collisions at @xmath167 in a wider rapidity interval . both experiments have shown an increase of the ratio @xmath171 as compared with the elementary interaction , while no significant increase was observed for @xmath172 ratio . the experimental values @xcite,@xcite are compared with hijing results in tabel 8a and tabel 8b . in table 9 the results on average multiplicities of @xmath13 and @xmath173 and multiplicities ratios in @xmath174 interactions extrapolated to the full phase space are compared with model predictions : hijing ( hij.01),venus and fritiof . the experimental data are taken from reference @xcite . all models underestimate the @xmath175 ratio , but reproduce well the @xmath176 ratio . the models seems to overpredict the average @xmath177 multiplicities in nucleus - nucleus collisions.the secondary collisions induced by the pion and kaon mesons in nuclei may influence the ratios @xmath178 @xcite . experimental values for multistrange baryon ratios in nucleus nucleus interactions were recently reported @xcite , @xcite , @xcite.we check if the model could describe these ratios , but our calculations are for full phase space and experimental data are reported in limited rapidity range . we see from table 10 that only ratios @xmath8 are well described by the model.the model overestimate the productions of @xmath179 and @xmath180 . the ratios are well described for @xmath181 but the ratios for multistrange particles and their antiparticle could not be predicted by the model.a detailed analysis in limited rapidity range is required in order to draw some definite conclusions ( calculations are now in progress ) . we remark that the ratios for multistrange particle production are not predicted by this version of hijing model , nor by more sophisticated models like qgsm , rqmd , dpm , spm . a qualitative prediction for rapidity and transverse kinetic energy are performed at @xmath182 in @xmath183 and @xmath184 collisions and are represented in comparison with experimental data from t.alner et al.,@xcite . in fig.3a , b , c , d we give normalized rapidity distributions for @xmath123 produced in interactions quoted above : fig.3a for @xmath185 ; fig.3b for @xmath186 , fig.3c for @xmath187;and fig.3d for @xmath184 . in fig.4a , b , c , d and in fig.5a , b , c , d we give the rapidity distributions for @xmath188 and @xmath124 respectively . in fig.6a , b , c , d ; fig.7a , b , c , d and in fig.8a , b , c , d normalized distributions of transverse kinetic energy defined as @xmath189,are represented for the same interactions . theoretical hijing values ( hij.01 ) are multiplied by a factor of two ( hij.01 * 2 ) shown in the figures.the total number of events generated in full phase space are the same as for the results from table 3(see section 3.1 ) . transverse momentum distributions for @xmath123 , @xmath188 and @xmath124 particles produced in @xmath190 interactions are given in figures 9a , b , c . the ratio of @xmath188 to @xmath123 rapidity distributions for @xmath190 collisions are done in fig.10 . analysing the figures ( fig.3a , b , c , d fig.5a , b , c , d ) we see that similarly to @xmath191 collisions , the rapidity spectra of @xmath91 and @xmath19 particles in @xmath192 interactions have a sharp peak centered at mid rapidity and their width are close to the ones for the proton induced case . for @xmath15 particles the distribution is naturally symmetrical and its width is larger due to possible production of strange baryons in the target(projectile ) fragmentation region . a detailed discussion of @xmath174 collision system was done also by werner @xcite and amelin et al . @xcite , @xcite.the projectile and target mass asymmetry is reflected in the @xmath123 rapidity distributions which for @xmath185 , @xmath187and @xmath184 interactions reach a maximum at about 1.0 which is compatible with those remarked in experiment ( about 1 to 1.5 ) . the @xmath188 and @xmath124 rapidity distributions are less sensitive to the initial asymmetry of the system . as follows from figures , the shape of rapidity for kaons , lambdas and antilambdas are reproduced fairly well by hijing model , but the model uderestimate the production of @xmath15 in the central region of rapidity spectra , even for proton - sulphur interaction . new experimental data and also data on proton production in hadron - nucleus and nucleus - nucleus interactions are required in order to draw a definite conclusions . for transverse kinetic energy distributions ( taken into account the factor of two ) we obtain a good agreement with the data except the low @xmath120 values and in some cases the values at the tails . the normalised transverse momentum distributions for @xmath193 interactions are well described except some discrepacies seen for @xmath15 ( see fig.9a , b , c ) . the ratio of the @xmath19 and @xmath15 yields is suggested to reflect the net baryon density of the particle source . in fig.10 the ratio of @xmath19 to @xmath15 rapidity distributions for @xmath191 is presented in comparison with experimental data .the theoretical values are higher than experimental ones . it should be stressed out that in the most string models the produced strings do not interact and decay independently . introduction of some collective effects like string fusion @xcite,@xcite or firecreacker model @xcite can change the predictions essentially , especially for rapidity plateau heights and high multiplicity distribution tails . some analysis was done recently by sorge @xcite , @xcitein the framework of the rqmd .a doubling of lambda production in the target fragmentation region for collision on heavy targets were found due to nuclear cascading of the produced mesons.also in rqmd approach the resonances play an important role in creating strange quark pairs.werner @xcite has shown that rescattering should be considered at cern - sps energies with the effect of increasing neutral strange particles production especially in the central region.hijing does not incorporate these ingredients . no model seems to get a real description of absolute values for neutral strange particle production without additional mechanisms . since recently @xcite the data on multiple strange particle production ( @xmath194 ) was reported , in figures 11a - b we give rapidity distribution for full phase space @xmath16 ( fig.11a ) and @xmath195 ( fig.11b)for @xmath174 interaction . we can see from these figures that the ratio of multiplicities @xmath196 has strogly dependence on rapidity bins . improvement of the statistics and considerations of limited rapidity experimental intervals , are needed in order to due some more toward comparison with experimental data . for giving an idea about of the abundances of particles produced in the near planed experiments at cern we have studied @xmath197 interaction at @xmath198 and in addition to multiplicities given in tabel 3 and tabel 5 , normalised rapidity distributions are given in figures 12a b for ( @xmath199 ) . the number of generated events was@xmath200 and the results are represented for full phase space . the dependences rapidity - transvers momentum , which define theoretical values for acceptances as well as their bidimensional projections are also given in figures 13 a - b for @xmath15 , in figures 14 a - b for @xmath19 and for all negatives charges in @xmath186,@xmath201 and @xmath202 interactions in figures 15a , b;16a , b and 17a , b(respectively ) and for all positives charges in @xmath203 interactions in figures 18a - b . we hope that from such kind of correlations we will get more inside the dynamics of the systems and we will find some limits of the model concerning neutral strange particles production . in this report , we give a systematic study of strange particle production in pp , pa and aa collisions mainly at sps cern - energies , using hijing monte carlo model developed for high energy collisions . we analyse data from cern experiments ( na35,na36 , + wa85 ) and we give a detailed comparison for mean multiplicities of neutral strange particle for pp , ps , pag , pw , pau , ss , sw and pbpb interactions . one should note that hijing model includes neither hipotheses of qgp . strange particle abundances are quite well described in pp and pa collisions and are underpredicted by a factor of two in nucleus - nucleus interactions , by this version ( theoretical values for hij.01 ) .the ratios for multistrange particles could not be predicted by the model hij.01 .the abundances of @xmath16 and @xmath17 are underpredicted and the abundances of @xmath179 and @xmath180 are overpredicted . more carreful analisis are necessary in limited rapidity interval . we present also some calculations for the ratios @xmath204 and @xmath205 given by hijing model compared with other models predictions like venus and fritiof.all models seems to overpredict @xmath205 ratios and underpredict @xmath204 ratios . rescattering should be important even at sps -energies . a qualitative prediction for rapidity distributions of neutral strange particles in @xmath206 , @xmath207,@xmath208,@xmath209 @xmath210per nucleon are done compared also with recent experimental data[140 ] . the models describes correctly the shapes of the rapidity in central @xmath211 collisions for @xmath212 and @xmath188 but absolute numbers for multiplicities are underestimated by a factor of two for @xmath213 in symmetric collisions and for @xmath214 in asymmetric ones . some discrepancies are seen for @xmath15 rapidity distribution and transverse momentum distributions even for @xmath190 interactions.the factor of two can not be explained taken into account minijets and very central events(@xmath215 ) . the theoretical model predictions hij.01 presented above for central @xmath201,@xmath184 and @xmath202 collisions give an idea for particle abundances in experiments which are under the way or are planned in the near future at cern . the hijing model simulations are however still somewhat too low compared to the data.also the ratios for multistrange baryons could not be described . new ideas are necessary in order to improve the comparison of theoretical calculations with experimental data.the main goal is to increase the rate of neutral strange particle production by a factor of two in nucleus - nucleus interaction at @xmath216 . we remark that other models ( qgsm , venus , dpm and psm ) include now some aditional hypothesis in order to improve the discrepancies ( colective string - string interactions , double or multiple colour exchanges , string fusion ) . one of the authors ( vtp ) would like to thank to professor miklos gyulassy and dr xin - nian wang for providing the monte carlo program hijing and many helpful communications . + _ nucleus collisions _ , gattinburg , tennesse , usa , november 1115(1991 ) , eds . t. c. awes , f. e. obenshaim , f . plasil , m . r. strayer and c. y. wong(qm91 ) + nucl . a544(1992 ) aussois , france , march 1620,1992,eds . m.bex , b.borderie,m.ploszajczak,j.richter,p.schuck and e.surand + nucl.phys.a545(1992 ) wiesbaden , germany + , july 26august 1,1992 , eds . r. bock , h . emling , e . grosse , u . grundinger + k. d. hildenbrand and j. knoll + nucl.phys.a553(1993 ) university of aarhus , denmark,2024 may 1991,eds.j.madsen and p.haensel + nucl.phys.b25(1992 ) eds .hans.h.gutbrod and johann rafelski , plenum press new york and london ( 1993 ) edited by e. stenlund , h . a. gustafsson , a . oskarsson and i. otterlund + nucl.phys.a566(1994 ) + perugia , italy 28 june2 july 1993,book of abstracts , courmayeur , aosta valley , italy 1992,eds.f.iazzi and a.zenoni + nucl.phys.a558(1993 ) s.a.bonometto,ornela pantano , phys.rep.228 * ( 1993)177 p.koch.b.muller and j.rafelski phys.rep.c142(1986)167 s.nagamiya in _ perspective of meson science _ eds.t.yamazaki,k.nakai,k.nagamine(north holland , amsterdam - london - new york -tokyo,1992),cp 24 _ strange meson production in heavy ion collisions_,p735 m.sano and m.wakai(see reference [ 11]),cp25 , _ strangeness production in high energy collisions_,p763 j.schukraft,nucl.phys.a553(1993)31c h. r. schmidt and j. schukraft , cern - ppe-/9242 ( 1992 ) j.of phys.g(1993)1705 s.ahmad et al.nucl.phys.a558(1993)393c e.witten,phys.rev.d30(1984)272 e.farhi and r.l.jaffe,phys.rev.d30(1984)2379 k.kajantie and l.mc.lerran,ann.rev.and part.sci . * 37(1987)293 h.stocker,c.greiner,j.schafner , nucl.phys.b24(1991)239c,246c h.bartz,nucl.phys.b24(1991)211c e.quercigh , see reference [ 5 ] p 499 j.bartke et al.,z.phys.c48(1990)191 r.stock et al.,nucl.phys.a525(1991)221c p.seyboth et al.,nucl.phys.a544(1992)293c m.kowalski et at.,nucl.phys.a544(1992)609 h.bialkowska et al.,z.phys.c55(1992)491 j.baechler et al.,z.phys.c58(1993)367 t.akeson et al.,nucl.phys.b353(1991)1 t.akeson et al.,z.phys.c46(1990)369 a.bamberger et al.,z.phys.c43(1989)25 i.derado et al.,z.phys.c50(1991)31 m.gazdzicki and o.hansen,nucl.phys . , a528(1991)754 d.e.greiner et al.,nucl.phys.a544(1992)309 e.anderson et al.,nucl.phys.a553(1993)817c e.andersen et al.,phys.lett.b294(1992)127 e.andersen et al.,nucl.instr.methods a301 * ( 1991)69 e.andersen et al.,phys.lett.b316(1993)603 s.abatzis et al.,phys.lett.b244(1990)130 s.abatzis et al.,phys.lett.b259(1991)508 s.abatzis et al.,phys.lett.b270(1991)123 s.abatzis et al.,nucl.phys.a525(1991)445c s.abatzis et al.,nucl . phys . * a566(1994)225c j.b.kinson et al.,nucl.phys.a544(1992)321c di bari , doctoral thesis ( january 1993 ) s.abatzis et al.,phys.lett.b316(1993)615 t.akesson et al.,preprint cern - ppe/87 - 176(1987 ) t.akesson et al.,preprint cern - ppe/89 - 97(1987 ) t.akesson et al.,preprint cern - ppe/87 - 170(1987 ) t.akesson et al.,preprint cern - ppe/89 - 111(1989 ) t.akesson et al.,preprint cern - ppe/92 - 130(1992 ) t.akesson et al.,z.phys.c53(1992)183 t.abbot et al.,phys.rev.lett.64(1990)847 t.abbot et al.,phys.rev.lett.66(1991)1567 t.abbot et al.phys.rev.d45(1992)3906 y.miake et al.,nucl.phys.a525(1991)231c t.abbot et al.,phys.rev.c47(1993)1351 k.j.foley et al.,nucl.phys.a544(1992)335c , w.a.love et al.,e810 collaboration , in _ proceedings of the workshop on heavy ion physics at ags , bnl_,(1990),edited by o.hansen r.a.salmeron,nucl.phys.b389(1993)301 a.capella et al.,phys.lett.b81(1979)68 a.capella and j.tran thanh , z.phys.c10 * ( 1981)249 a.capella,j.kwiecinski,j.tran thanh , phys.lett.b108(1982)347 a.capella et al.,z.phys.c33(1987)343 j.ranft and s.ritter,z.phys.c27 ( 1985)413 h.j.mohring,j.ranft,s.ritter , z.phys.c27(1985)419 h.j.mohring,j.ranft,z.phys.c52 ( 1991)643 h.j.mohring et al.,nucl.phys.a525 ( 1991)493c a.b.kaidalov,sov.j.nucl.phys.33(1981)733 a.b.kaidalov and o.i.piskunova,z.phys.c30 ( 1986)145 a.b.kaidalov , nucl.phys.a525(1991)39c k.werner,phys.lett.b197(1987)225 k.werner,z.phys.c38(1988)193 k.werner,phys.rev.d39(1989)780 k.werner,z.phys.c42(1989)85 k.werner and p.koch,z.phys.c47 ( 1990)215 k.werner,nucl.phys.a525(1991)501c b.andersson,g.gustafson and nilsson - almqvist nucl.phys.b281(1987)289 b.nilsson-almqvistand e.stenlund , comp.phys.commun.43(1987)387 m. gyulassy , cern report th-4794/87(1987 ) j. aichelin et al.,phys.rev.c37 ( 1988)2451 r. mattielo , h . sorge , h . stocker and w. greiner , phys.rev.lett.63(1989)1459 h. sorge , h . stocker and w. greiner , nucl.phys.a498(1989)567c a. von keitz , h . sorge , h.stocker and w.greiner,nucl.phys.a527(1991)601 h. sorge et al.,see ref.[5 ] p 335 k. werner , preprint hd - tvp-93 - 1(1993 ) phys.rep.232(1993)87 m. braun and c.pajares,nucl.phys.b390 ( 1993)542 m. braun and c.pajares,nucl.phys.b390 ( 1993)559 m. braun and c.pajares,phys.lett.b287 ( 1992)154 m. braun and c.pajares,phys.rev.d47 ( 1993)114 c. merino , c.pajares and j.ranft,phys.lett . * b276(1992)168 h. mohring , j.ranft , c.merino and c.pajares , phys.rev.d47(1993)4142 xin - nian wang , phys.rev.d43(1991)104 xin - nian wang and miklos gyulassy , phys.rev . d44(1991)3501 xin - nian wang and miclos gyulassy , phys.rev . d45(1992)844 xin - nian wang and miclos gyulassy , phys.rev.lett . 68(1992)1480 berndt muller and xin - nian wang , phys , rev.lett.68(1992)2437 xin - nian wang and miclos gyulassy , preprint duke - th-91 - 24(1991 ) xin - nian wang , preprint duke - th-92 - 33 ( 1993 ) j. ranft , phys.rev.d37(1988)1842 t. sjostrand , comput.phys.commun . 39(1986)347 t. sjostrand , comput.phys.commun . 43(1987)367 n. s. amelin et al.,phys.rev.c44 * ( 1991)1541 n. s. amelin et al.,phys.rev.c47 ( 1993)2299 r. c. hwa , phys.rev.d37(1988)1830 d. r. ward _ properties of soft proton antiproton collisions _ in advances series on direction in high energy physics , vol.4 , eds.g.altarelli and l.di lella,1989 , p.85 i. m. dremin , sov.j.nucl.phys . * 33(1981)726 g.j.alner et al . , phys.rep.154(1987)247 jaeger et al.,phys.rev.d11 * ( 1975)2405 liu bo and chiang huan ching , j. phys . g.nucl . part . phys.19(1993)161 h.strobele et al . nucl.phys.a525(1991)59c h.sorge,l . a. winckelmann , h. stocker , w . greiner , z.phys.c59(1993)85 j. letessier , a . tounsi , u . heinz , j. sollfrank and j. rafelski , _ strangeness conservation in hot fireballs_preprint par / lpthe/92 - 27(1992 ) j. cleymans , k . redlich , h . satz and e. suhonen , _ the hadronisation of a quark - gluon plasma _ , preprint cern - th 6759/92(1992 ) helmut satz , nucl.phys.a544(1992)371c e. quercigh , nucl . phys.a566(1994)321c r. s. longacre et al . , nucl . phys.a566(1994)167c johana stachel et . nucl.phys.a566(1994)183 g. e. diebold et . al . , phys . * c48(1993)2984 m. n. namboodiri et . al . , nucl . phys.a566(1994)443c r. folman , a . shor , nucl . phys.a566(1994)917 a.capella,u.sukhatme,c . i. tan and j. tran thanh van , phys . rep.236(1994)225 j. letessier , j . rafelski , a . tounsi , phys . b323(1994)393 j. sollfrank , m . gazdzichi , u . heinz , j . rafelski , z. phys . c61(1994)659 e.schnedermann,j . sollfrank and u.heinz , phys . c48(1993)2462 u.heinz , nucl . a566(1994)205 m. gazdzicki et al . , nucl . a566(1994)503c e. andersen et . a566(1994)217c e.andersen et . a566(1994)487c s.abatzis et . al . , nucl . phys . a566(1994)225c f. antinori et . al . , nucl . a566(1994)491c s.abatzis et . al . , nucl . a566(1994)499c j. p. blaizot , nucl . a566(1994)333c m. sarabura et al.,nucl . a544(1992)125c m. murray et al.,nucl . phys . a566(1994)515c j. barette et al.,z.phys.c59(1993)211 h. sorge , phys.rev.c49(1994)1253 h. sorge , m.berenguer,h.stocker,w.greiner , preprint la - ur-92 - 1078(1992 ) h. sorge , preprint la - ur-93 - 1103(1993 ) h. sorge , nucl.phys.a566(1994)633c n.s.amelin et al.,preprint us - ft/1 - 94 * ( santiago university)(1994 ) t.alber et.al , preprint ikf - henpg/1 - 94(university of frankfurt)(1994 ) h. j. mohring et al.,phys.rev . * d47(1993)4146 h. j. mohring et al.,phys.rev . d47(1993)4142 ben - hao sa and an tai , preprint lund university , may 1994(private communication ) table 3 . * + _ average multiplicities for negative hadrons , neutral strange particle in pp , pa and aa interactions predicted by hijing model are compared with experimental data.the values for @xmath218 are theoretical values for @xmath199 and@xmath219 multiplied by a factor two and the values for@xmath220 are theoretical values for @xmath199 multiplied by a factor two . _ * table 5 . * + _ the mean multiplicities of negative pions and @xmath221 ratios(see the text for definition ) for nuclear collisions at 200 gev per nucleon .the experimental data are from na 35 collaboration(reference[140 ] ) and the nn data are from reference[32 ] . _ * table 6 . * + _ the ratios of the mean strange particle multiplicity to the mean multiplicity of negative hadrons . the values for @xmath218 and @xmath220 have the same meaning as in table 3.the experimental data are from t.alner et al . [ 140 ] _ * table 7 * + _ multiplicity ratios for negative hadrons as well as for several identified strange particles and diferent reactions at @xmath222;results from the hij.01 model are compared to dpm model[65 ] , na5[31 ] and na35[22 ] data_. * table 9 * + _ average multiplicities of @xmath13 and @xmath173 and multiplicity ratios in @xmath174 interactions extrapolated to the full phase space . the experimental data are from baechler et al.[27 ] . the results for @xmath224 interaction [ 32 ] are also included for comparison . _ 0.3 cm + 0.3 cm * fig.1 + * 0.3 cm rapidity distributions ( fig.1a ) and transverse momentum distributions(fig.1b ) for @xmath15 particles produced in @xmath136 interactions at @xmath225 . experimental data are taken from jaeger et al.[107 ] . 0.3 cm * fig.2 * + 0.3 cm rapidity distributions ( fig.2a)and transverse momentum distributions(fig.2b ) for @xmath91 particles produced in @xmath136 interactions at @xmath225 . experimental data are taken from jaeger et al.[107 ] . 0.3 cm * fig.3 * + 0.3 cm rapidity distributions for @xmath123 particles produced in minimum bias @xmath185 ( a ) interactions and central @xmath186(b),@xmath187(c ) and @xmath184 ( d ) collisions at @xmath226 per nucleon.the experimental data ( full circles ) are taken from t.alner et al.,[140 ] . the dashed histograms are theoretical values@xmath227 and the solid histograms are theoretical predictions @xmath1 multiplied by a factor of two @xmath228 . the open circles show the distributions for @xmath186 collisions reflected at @xmath229 0.3 cm * fig.4 * + 0.3 cm rapidity distributions for @xmath188 particles produced in minimum bias @xmath185 ( a ) interactions and central @xmath186(b),@xmath187(c ) and @xmath184 ( d ) collisions at @xmath226 per nucleon . the experimental data ( full circles ) are taken from t.alner et al.,[140].the dashed histograms are theoretical values@xmath227 and the solid histograms are theoretical predictions @xmath1 multiplied by a factor of two @xmath228 . the open circles show the distributions for @xmath186 collisions reflected at @xmath229 0.3 cm * fig.5 * + 0.3 cm rapidity distributions for @xmath230 particles produced in minimum bias @xmath185 ( a ) interactions and central @xmath186(b),@xmath187(c ) and @xmath184 ( d ) collisions at @xmath226 per nucleon . the experimental data ( full circles ) are taken from t.alner et al.,[140].the dashed histograms are theoretical values@xmath227 and the solid histograms are theoretical predictions @xmath1 multiplied by a factor of two @xmath228 . the open circles show the distributions for @xmath186 collisions reflected at @xmath229 0.3 cm * fig.6 * + 0.3 cm transverse kinetic energy distributions for @xmath123 particles produced in minimum bias @xmath185 ( a ) interactions and central @xmath186(b ) @xmath187(c ) and @xmath184 ( d ) collisions at @xmath226 per nucleon . the experimental data ( full circles ) are taken from t.alner et al.,[140 ] . the solid histograms are theoretical hijing values ( @xmath1 ) or @xmath231 shown in the figures.the vertical scale is given in @xmath232 . 0.3 cm * fig.7 * + 0.3 cm transverse kinetic energy distributions for @xmath188 particles produced in minimum bias @xmath185 ( a ) interactions and central @xmath186(b ) @xmath187(c ) and @xmath184 ( d ) collisions at @xmath226 per nucleon . the experimental data ( full circles ) are taken from t.alner et al.,[140 ] . the solid histograms are theoretical hijing values ( @xmath1 ) or @xmath231 shown in the figures.the vertical scale is given in @xmath232 . 0.3 cm * fig.8 * + 0.3 cm transverse kinetic energy distributions for @xmath124 particles produced in minimum bias @xmath185 ( a ) interactions and central @xmath186(b ) @xmath187(c ) and @xmath184 ( d ) collisions at @xmath226 per nucleon . the experimental data ( full circles ) are taken from t.alner et al.,[140 ] . the solid histograms are theoretical hijing values ( @xmath1 ) or @xmath231 shown in the figures.the vertical scale is given in @xmath232 . 0.3 cm * fig.9 * + 0.3 cm transverse momentum distributions for @xmath123 ( a ) , @xmath188 ( b ) and @xmath124 ( c ) particles produced in minimum bias @xmath185 interactions at @xmath226 per nucleon.the solid histograms are theoretical hijing values ( @xmath1).the vertical scale is given in @xmath233 . 0.3 cm * fig.10 * + 0.3 cm the ratio of @xmath188 to @xmath123 rapidity distributions for @xmath193 interactions at @xmath226 per nucleon . the corresponding ratio calculated using hijing model is shown by solid histogram . the experimental data [ 140 ] are represented by full circles . 0.3 cm * fig.11 * + 0.3 cm rapidity distributions for @xmath234 ( a ) and for @xmath235 ( b ) particles produced in @xmath236 interactions at @xmath226 per nucleon . histograms are hijing results . 0.3 cm * fig.12 * + 0.3 cm predicted rapidity distributions for @xmath123 ( a ) and for @xmath188 ( b ) particle produced in central @xmath202 collisions at @xmath237 0.3 cm * fig.13 * + 0.3 cm theoretical values for acceptances are given ( a ) for @xmath123 particles produced in @xmath202 collisions at @xmath237 . a bidimensional plot ( rapidity y - transverse momentum @xmath38 ) ( b ) for the same interaction . 0.3 cm * fig.14 * + 0.3 cm theoretical values for acceptances are given ( a ) for @xmath188 particles produced @xmath202 collisions at @xmath237 . a bidimensional plot ( rapidity y - transverse momentum @xmath38 ) ( b ) for the same interaction . 0.3 cm * fig.15 * + 0.3 cm theoretical values for acceptances are given ( a ) for all negatives charges produced in @xmath186 central collisions at @xmath238 . a bidimensional plot ( rapidity y - transverse momentum @xmath38 ) ( b ) for the same interaction . 0.3 cm * fig.16 * + 0.3 cm theoretical values for acceptances are given ( a ) for all negatives charges produced in @xmath201 central collisions at @xmath238 . a bidimensional plot ( rapidity y - transverse momentum @xmath38 ) ( b ) for the same interaction . 0.3 cm * fig.17 * + 0.3 cm theoretical values for acceptances are given ( a ) for all negatives charges produced in @xmath202 central collisions at @xmath237 . a bidimensional plot ( rapidity y - transverse momentum @xmath38 ) ( b ) for the same interaction . 0.3 cm * fig.18 * + 0.3 cm theoretical values for acceptances are given ( a ) for all positives charges produced in @xmath202 central collisions at @xmath237 . a bidimensional plot ( rapidity y - transverse momentum @xmath38 ) ( b ) for the same interaction .	in this report we have made a systematic study of strangeness production in proton - proton(pp),proton - nucleus(pa ) and nucleus- nucleus(aa ) collisions at cern super proton synchroton energies , using@xmath0 + ( version @xmath1 ) . numerical results for mean multiplicities of neutral strange particles , as well as their ratios to negatives hadrons(@xmath2 ) for p - p , nucleon - nucleon(n - n),p - s , p - ag , p - au(min . bias)collisions and p - au , s - s , s - ag , s - au ( central)collisions are compared to experimental data available from cern experiments and also with recent theoretical estimations given by others models(@xmath3 + @xmath4 and @xmath5 ) . neutral strange particle abundances are quite well described for p - p , n - n and p - a interactions , but are underpredicted by a factor of two in a - a interactions for @xmath6 in symmetric collisions(s - s , pb - pb)and for @xmath7 in asymmetric ones(s - ag , s - au , s - w ) . the ratios antistrange / strange are well described for @xmath8 but the ratios for multistrange particles ( @xmath9 ) and their antiparticle , could not be predicted by the model.a detailed analysis in limited rapidity range is required in order to draw some definite conclusion . a qualitative prediction for rapidity , transverse kinetic energy and transverse momenta normalized distributions are performed at 200 gev / nucleon in p - s , s - s , s - ag and s - au collisions in comparison with recent experimental data . hijing model predictions for coming experiments at cern for s - au , s - w and pb - pb interactions are given . the theoretical calculations are estimated in a full phase space . the factor of two can not be explained by considering very central events ( @xmath10 ) and taking into account minijets . new ideas are necessary in order to improve the comparison with experiment for nucleus - nucleus interactions . = 16truecm = 24truecm 1 = = .5truecm = -2truecm 1truecm * dipartamento di fisica `` g.galilei '' + * * istituto di fisica nucleare + * * sezione di padova + * italy * * 2truecm * strangeness production in pp , pa , aa * + 0.3 cm * interactions * + 0.3 cm * at sps energies . hijing approach * + 1truecm * wa94 collaboration * + dipartimento di fisica``g.galilei'',infn sezione di padova , via marzolo 8 , 35131 padova,italy + , * romania + * and + dipartamento di fisica `` g.galilei'',infn sezione di padova , via marzolo 8 , 35131 padova,italy + * dfpd-94-np-42 8 june 1994 * _ work presented at the international conference `` physics of high energy heavy ion collisions '' , vuosaari(helsinki),finland,17 - 22 june 1994 _ 1truecm = -1truecm = -2truecm
a signal can be transformed from one domain into another in various ways . some signals that cover the whole considered interval in one domain ( where signals are dense in that domain ) could be located within much smaller regions in another domain . we say that signals are sparse in a transformation domain if the number of nonzero coefficients is much smaller that the total number of signal samples . for example , a sum of discrete - time complex sinusoidal signals , with a number of components being much lower than the number of signal samples in the time domain , is a sparse signal in the discrete fourier transform ( dft ) domain . sparse signals could be reconstructed from much fewer samples than the sampling theorem requires . compressive sensing is a field dealing with the problem of signal recovery from a reduced set of samples @xcite-@xcite . as a study case , in this paper we will consider signals that are sparse in the fourier transform domain . signal sparsity in the discrete fourier domain imposes some restrictions on the signal . reducing the number of samples in the analysis manifests as a noise , whose properties are studied in @xcite and used in @xcite to define a reconstruction algorithm . the input noise influence is also an important topic in this analysis since the reduced number of available samples could increase the sensitivity of the recovery results to this noise @xcite . additive noise will remain in the resulting transform . however , if a reconstruction algorithm for a sparse signal is used in the reconstruction of nonsparse signal then the noise , due to missing samples , will remain and behave as an additive input noise . a relation for the mean square error of this error is derived for the partial dft matrix case . it takes into account very important fact that if all samples are available then the error will be zero , for both sparse and nonsparse recovered signals . theory is illustrated and checked on statistical examples . the paper is organised as follows : after the introduction part in section 1 , the definition of sparsity is presented in section 2 . in section 3 , the reconstruction algorithm is presented for both one step reconstruction and the iterative way . also in section 3 , the analysis of the influence of additive noise will be expanded . the reconstruction of nonsparse signals with additive noise is shown in section 4 . in the appendix the conditions in which the reconstruction of sparse signals is possible in general are presented . consider a signal @xmath0 and its transformation domain coefficients @xmath1 , @xmath2 or @xmath3 where @xmath4 is the transformation matrix with elements @xmath5 , @xmath6 is the signal vector column , and @xmath7 is the transformation coefficients vector column . for the dft @xmath8 . a signal is sparse in the transformation domain if the number of nonzero transform coefficients @xmath9 is much lower than the number of the original signal samples @xmath10 , fig . [ posterslika ] , i.e. , if @xmath11 for@xmath12 the number of nonzero samples is@xmath13 where @xmath14 and @xmath15 is the notation for the number of nonzero transformation coefficients in @xmath7 . counting the nonzero coefficients in a signal representation can be achieved by using the so called @xmath16-norm denoted by @xmath17 . this form is referred to as the @xmath16-norm ( norm - zero ) although it does not satisfy norm properties . by definition @xmath18 for @xmath19 and @xmath20 @xmath21for @xmath22 . a signal @xmath0 , whose transformation coefficients are @xmath1 , is sparse in this transformation domain if@xmath23 for linear signal transforms the signal can be written as a linear combination of the sparse domain coefficients @xmath1 @xmath24 a signal sample can be considered as a linear combination ( measurement ) of values @xmath1 . assume that samples of @xmath0 are available only at a reduced set of random positions @xmath25 here @xmath26 is the set of all samples of a signal @xmath0 and @xmath27 is its random subset with @xmath28 elements , @xmath29 . the available signal values are denoted by vector @xmath30 , fig.[posterslika ] , @xmath31^{t}.\ ] ] [ ptb ] posterslika.eps the available samples ( measurements of a linear combination of @xmath1 ) defined by ( [ measdft ] ) , for @xmath32 , can be written as a system of @xmath28 equations @xmath33{c}x(n_{1})\\ x(n_{2})\\ ... \\ x(n_{m } ) \end{array } \right ] = \left [ \begin{array } [ c]{cccc}\psi_{0}(n_{1 } ) & \psi_{1}(n_{1 } ) & & \psi_{n-1}(n_{1})\\ \psi_{0}(n_{2 } ) & \psi_{1}(n_{2 } ) & & \psi_{n-1}(n_{2})\\ ... & ... & & ... \\ \psi_{0}(n_{m } ) & \psi_{1}(n_{m } ) & & \psi_{n-1}(n_{m } ) \end{array } \right ] \left [ \begin{array } [ c]{c}x(0)\\ x(0)\\ ... \\ x(n-1 ) \end{array } \right]\ ] ] or @xmath34 where @xmath35 is the @xmath36 matrix of measurements / observations / available signal samples . the fact that the signal is sparse with @xmath37 for @xmath38 is not included in the measurement matrix @xmath35 since the positions of the nonzero values are unknown . if the knowledge that @xmath37 for @xmath38 were included then a reduced observation matrix would be obtained as @xmath33{c}x(n_{1})\\ x(n_{2})\\ ... \\ x(n_{m } ) \end{array } \right ] = \left [ \begin{array } [ c]{cccc}\psi_{k_{1}}(n_{1 } ) & \psi_{k_{2}}(n_{1 } ) & & \psi_{k_{k}}(n_{1})\\ \psi_{k_{1}}(n_{2 } ) & \psi_{k_{2}}(n_{2 } ) & & \psi_{k_{k}}(n_{2})\\ ... & ... & & ... \\ \psi_{k_{1}}(n_{m } ) & \psi_{k_{2}}(n_{m } ) & & \psi_{k_{k}}(n_{m } ) \end{array } \right ] \left [ \begin{array } [ c]{c}x(k_{1})\\ x(k_{2})\\ ... \\ x(k_{k } ) \end{array } \right]\ ] ] or @xmath39 matrix @xmath40 would be formed if we knew the positions of nonzero samples @xmath41 . it would follow from the measurement matrix @xmath35 by omitting the columns corresponding to the zero - valued coefficients @xmath1 . assuming that there are @xmath9 nonzero coefficients @xmath1 , out of the total number of @xmath10 values , the total number of possible different matrices @xmath40 is equal to the number of combinations with @xmath9 out of @xmath10 . it is equal to @xmath42 . although the @xmath16-norm can not be used in the direct minimization , the algorithms based on the assumption that some coefficients @xmath1 are equal to zero , and the minimization of the number of remaining nonzero coefficients that can reconstruct sparse signal , may efficiently be used . the reconstruction process can be formulated as finding the positions and the values of @xmath9 nonzero coefficients @xmath1 of a sparse signal ( or all signal @xmath0 values ) using a reduced set of signal values @xmath43 , @xmath44 such that@xmath45 where @xmath46 . consider a discrete - time signal @xmath0 . signal is sparse in a transformation domain defined by the basis functions set @xmath5 , @xmath47 . the number of nonzero transform coefficients @xmath9 is much lower than the number of the original signal samples @xmath10 , i.e. , @xmath37 for @xmath12 @xmath48 . a signal@xmath49 of sparsity @xmath9 can be reconstructed from @xmath28 samples , where @xmath29 . in the case of signal @xmath0 which is sparse in the transformation domain there are @xmath9 nonzero unknown values @xmath50 , @xmath51, ... ,@xmath52 . other transform coefficients @xmath1 , for @xmath38 * * , * * are zero - valued . just for the beginning assume that the transformation coefficient positions @xmath53 , @xmath54 , ... , @xmath55 are known . then the minimal number of equations to find the unknown coefficients ( and to calculate signal @xmath0 for any @xmath56 ) is @xmath9 . the equations are written for at least @xmath9 time instants @xmath57 , @xmath58 , where the signal is available / measured , @xmath59 in a matrix form this system of equations is @xmath60 where @xmath61 is the vector of unknown nonzero coefficients values ( at the known positions ) and @xmath30 is the vector of available signal samples , @xmath62^{t}\\ \mathbf{y}=[x(n_{1})~~x(n_{2})~~ ... ~~x(n_{m})]^{t}\nonumber\\ \mathbf{a}_{k}=\left [ \begin{array } [ c]{cccc}\psi_{k_{1}}(n_{1 } ) & \psi_{k_{2}}(n_{1 } ) & ... & \psi_{k_{k}}(n_{1})\\ \psi_{k_{1}}(n_{2 } ) & \psi_{k_{2}}(n_{2 } ) & ... & \psi_{k_{k}}(n_{2})\\ ... & ... & ... & .... \\ \psi_{k_{1}}(n_{k } ) & \psi_{k_{2}}(n_{k } ) & ... & \psi_{k_{k}}(n_{k } ) \end{array } \right ] .\label{martr_sampl}\ ] ] matrix @xmath40 is the measurements matrix @xmath35 with the columns corresponding to the zero - valued transform coefficients @xmath63 , @xmath54 , ... , @xmath55 being excluded . for a given set @xmath64 the coefficients reconstruction condition can be easily formulated as the condition that system ( [ susmat ] ) has a ( unique ) solution , i.e. , that there are @xmath9 independent equations , @xmath65 note that this condition does not guarantee that another set @xmath64 can also have a ( unique ) solution , for the same set of available samples . it requires @xmath66 for any submatrix @xmath67 of the measurements matrix @xmath68 system ( [ sist_rj ] ) is used with @xmath69 . its solution , in the mean squared sense , follows from the minimization of the difference of the available signal values and the values produced by inverse transform of the reconstructed coefficients , @xmath70 where @xmath71 or @xmath72 where exponent @xmath73 denotes the hermitian conjugate . the derivative of @xmath74 over a specific coefficient @xmath75 , @xmath76 , is @xmath77 \psi_{p}^{\ast}(n).\ ] ] the minimum of quadratic form error is reached for @xmath78 when @xmath79 in matrix form this system of equations reads @xmath80 its solution is @xmath81 it can be obtained by a symbolic vector derivation of ( [ e_vec_h ] ) as @xmath82 if we do not know the positions of the nonzero values @xmath1 for @xmath83 , @xmath54 , ... , @xmath84 then all possible combinations of @xmath53 , @xmath54 , ... , @xmath85 should be tested . there are @xmath42 of them . it is not a computationally feasible problem . thus we must try to find a method to estimate @xmath53 , @xmath54 , ... , @xmath55 in order to recover values of @xmath1 . solution of the minimization problem , assuming that the positions of the nonzero signal coefficients in the sparse domain are known , is presented in the previous subsection . the next step is to estimate the coefficient positions , using the available samples . a simple way is to try to estimate the positions based on signal samples that are available , ignoring unavailable samples . this kind of transform estimate is @xmath86 where for the dft @xmath87 and @xmath88 . since @xmath89 this relation can be written as @xmath90 where @xmath91 is the measurement matrix . with @xmath92 the coefficients @xmath93 , calculated with @xmath28 samples , are random variables . note that using ( [ ms_summ ] ) in calculation is the same as assuming that the values of unavailable samples @xmath0 , @xmath94 , is zero . this kind of calculation corresponds to the result that would be achieved for the signal transform if @xmath95-norm is used in minimization . * algorithm * a simple and computationally efficient algorithm , for signal recovery , can now be implemented as follows : \(i ) calculate the initial transform estimate @xmath93 by using the available / remaining signal values @xmath96 \(ii ) set the transform values @xmath1 to zero at all positions @xmath97 except the highest ones . alternative : \(ii ) set the transform values @xmath1 to zero at all positions @xmath97 where this initial estimate @xmath93 is below a threshold @xmath98 , @xmath99 this criterion is not sensitive to @xmath98 as far as all nonzero positions of the original transform are detected ( @xmath93 is above the threshold ) and the total number @xmath100 of transform values in @xmath93 above the threshold is lower than the number of available samples , i.e. , @xmath101 . all @xmath102 transform values that are zero in the original signal will be found as zero - valued . \(iii ) the unknown nonzero ( including @xmath102 zero - valued ) transform coefficients could be then easily calculated by solving the set of @xmath28 equations for available instants @xmath103 , at the detected nonzero candidate positions @xmath104 , @xmath105,@xmath106 this system of the form @xmath107 is now reduced to the problem with known positions of nonzero coefficients ( considered in the previous subsection ) . it is solved in the least square sense as ( [ rjesenje ] ) @xmath108 the reconstructed coefficients @xmath109 , @xmath105 , ( denoted by vector @xmath61 ) are exact , for all frequencies . if some transform coefficients , whose true value should be zero , are included ( when @xmath110 ) the resulting system will produce their correct ( zero ) values . * comments : * in general , a simple strategy can be used by assuming that @xmath111 and by setting to zero value only the smallest @xmath112 transform coefficients in @xmath93 . system ( [ sist_rj ] ) is then a system of @xmath28 linear equations with @xmath111 unknown transform values @xmath109 . if the algorithm fails to detect a component the procedure can be repeated after the detected components are reconstructed and removed . this simple strategy is very efficient if there is no input noise . large @xmath100 , close or equal to @xmath28 , will increase the probability that full signal recovery is achieved in one step . in this paper , it will be shown that in the case of an additive ( even small ) input noise in all signal samples , a reduction of the number @xmath100 as close to the true signal sparsity @xmath9 as possible will improve the signal to noise ratio . _ example : _ consider a discrete signal @xmath113 for @xmath114 , sparse in the dft domain since only three dft values are different than zero . assume now that its samples @xmath115 , @xmath116 , @xmath117 , and @xmath118 are not available . we will show that , in this case , the exact dft reconstruction may be achieved by : \(i ) calculating the initial dft estimate by setting unavailable sample values to zero @xmath119 where @xmath120 \(ii ) detecting , for example @xmath121 positions of maximal dft values , @xmath122 , @xmath54 , and @xmath123 , and ( 3 ) calculating the reconstructed dft values at @xmath122 , @xmath54 , and @xmath123 from system@xmath124 where @xmath125 are the instants where the signal is available . the discrete - time signal @xmath0 , with @xmath114 is shown in fig . [ recon_expl_cs ] . the signal is sparse in the dft domain since only three dft values are different than zero ( fig . [ recon_expl_cs ] ( second row ) ) . the cs signal , with missing samples @xmath115 , @xmath116 , @xmath117 , and @xmath126 being set to @xmath127 for the initial dft estimation , is shown in fig . [ recon_expl_cs ] ( third row ) . the dft of the signal , with missing values being set to @xmath128 is calculated and presented in fig . [ recon_expl_cs ] ( fourth row ) . there are three dft values , at @xmath129 , @xmath130 , and @xmath131 @xmath132 above the assumed threshold , for example , at level of @xmath133 . the rest of the dft values is set to @xmath127 . this is justified by using the assumption that the signal is sparse . now , we form a set of equations , for these frequencies @xmath129 , @xmath130 , and @xmath131 as @xmath124 where @xmath125 are the instants where the signal is available . since there are more equations than unknowns , the system @xmath134 is solved using @xmath135 . the obtained reconstructed values are exact , for all frequencies @xmath97 , as in fig . [ recon_expl_cs ] ( second row ) . they are shown in fig . [ recon_expl_cs ] ( fifth row ) . if the threshold was lower , for example at @xmath136 , then six dft values at positions @xmath137 are above the assumed threshold . the system with six unknowns@xmath138 where @xmath125 will produce the same values for @xmath139 , @xmath140 , and @xmath141 while the values @xmath142 will be obtained . if the threshold is high to include the strongest signal component only , then the solution is obtained through an iterative procedure described in the next subsection . [ ptb ] recon_expl_cs.eps if components with very different amplitudes exist and the number of available samples is not large , then the iterative procedure should be used . this procedure could be implemented as follows . the largest component is detected and estimated first . it is subtracted from the signal . the next one is detected and the signal is estimated using the frequency from this and the previous step(s ) . the estimated two components are subtracted from the original signal . the frequency of next components is detected , and the process of estimations and subtractions is continued until the energy of the remaining signal is negligible or bellow an expected additive noise level . * algorithm * \(i ) calculate the initial transform estimate @xmath143 by using the available / remaining signal values @xmath144 @xmath145 set the transform values @xmath93 to zero at all positions @xmath97 except the highest one at @xmath146 , @xmath147 . set the counter to @xmath148 form the matrix @xmath149 using the available samples in time @xmath150 and detected index @xmath151 with one nonzero component . calculate the estimate of the transformation coefficient at @xmath146 @xmath152 calculate the signal estimation ( as the inverse dft)@xmath153 and check @xmath154 if , for example @xmath155 , stop the calculation and use @xmath156 . if not then go to the next step . \(ii ) set the counter to @xmath157 . form a signal @xmath158 at the available sample positions and calculate the transform @xmath159 set the transform values @xmath160 to zero at all positions @xmath97 except the highest one at @xmath161 . form the set of @xmath162 indices , using union of the previous maxima positions and the detected position , as @xmath163 form matrix @xmath164 using the available samples in time @xmath103 and detected @xmath165 indices @xmath166 . calculate the estimate of @xmath167 transformation coefficients @xmath168 calculate the signal@xmath169 and check @xmath170 if , for example @xmath155 , stop the calculation and use @xmath171 else repeat step ( ii ) . _ example : _ signal@xmath172 with @xmath173 is shown in fig.[iterativereport_knjiga ] . small number of samples is available @xmath174 with different signal amplitudes , making one - step recovery impossible . the available signal samples @xmath175 are shown in fig.[iterativereport_knjiga ] ( second row , left ) . the iterative procedure is used and , for the detected dft positions during the iterations , the recovered signal is calculated according to the presented algorithm . the recovered dft values in the @xmath162th iteration are denoted as @xmath176 and presented in fig.[iterativereport_knjiga ] . after first iteration the strongest component is detected and its amplitude is estimated . at this stage , other components behave as noise and make amplitude value inaccurate . accuracy improves as the number of detected components increases in next iterations . after five steps the agreement between the reconstructed signal and the available signal samples was complete . then the algorithm is stopped . the dft of the recovered signal is presented as @xmath177 in the last subplot of fig.[iterativereport_knjiga ] . its agreement with the dft of the original signal , fig.[iterativereport_knjiga ] ( first row , right ) is complete . [ ptb ] iterativereport_knjiga.eps the initial dft calculation in reconstruction algorithms is done assuming zero - valued missing samples . the initial calculation quality has a crucial importance to the successful signal recovery . with a large number of randomly positioned missing samples , the missing samples manifest as a noise in this initial transform . once the reconstruction conditions are met for a sparse signal and the exact reconstruction is achieved , the noise due to missing samples does not influence the results in a direct way . it influences the possibility to recover a signal . the accuracy of the recovery results is related to the additive input noise only . the input noise is transformed by the recovery algorithm into a new noise depending on the signal sparsity and the number of available samples . a simple analysis of this form of noise will be presented in the second part of this section . consider a sparse signal in the dft domain with nonzero coefficients @xmath1 at the positions @xmath178 @xmath179 where @xmath180 are the signal component amplitudes . the initial dft is calculated using @xmath88@xmath181 we can distinguish two cases : \(1 ) for @xmath182 then , with @xmath183 , @xmath184 the value of @xmath185 with random set @xmath186 , for @xmath187 , can be considered as a random variable . its mean over different realizations of available samples ( different realizations of sets @xmath188 ) is @xmath189.the mean value of @xmath109 is @xmath190 \(2 ) for @xmath191 the mean value of ( [ ms_summm ] ) is @xmath192 the mean value of ( [ ms_summm ] ) for any @xmath97 is of the form @xmath193 its variance is @xcite @xmath194 \text{. } \label{sign}\ ] ] the ratio of the signal amplitude @xmath50 and the standard deviation @xmath195 for @xmath196 is crucial parameter ( welsh bound for coherence index @xmath197 of measurement matrix @xmath35 ) for correct signal detection . its value is @xmath198 note that the variance in a multicomponent signal with @xmath199 is a sum of the variances of individual components at all frequencies @xmath97 except at @xmath200 when the values are lower for @xmath201 since all component values are added up in phase , without random variations . according to the central limit theorem , for @xmath187 the real and imaginary parts of the dft value for noise only positions @xmath202 can be described by gaussian distribution , @xmath203 with zero - mean and variance @xmath204 . real and imaginary parts of the dft value , at the signal component position @xmath200 , can be described by the gaussian distributions@xmath205 respectively , where @xmath206 , according to ( [ sign ] ) . _ example : _ for a discrete - time signal @xmath207 with @xmath173 , @xmath208 , @xmath209 , @xmath210 , the dft is calculated using a random set of @xmath174 samples . calculation is performed with @xmath211 random realizations with randomly positioned @xmath28 samples and random values of @xmath212 and @xmath123 . histogram of the dft values , at a noise only position @xmath213 and at the signal component @xmath146 position , is presented in fig.[hist_miss_samp ] ( left ) . histogram of the dft real part is shown , along with the corresponding gaussian functions @xmath214 and @xmath215 , shown by green dots . the same calculation is repeated with @xmath216 , fig.[hist_miss_samp ] ( right ) . we can see that the mean value of the gaussian variable @xmath1 can be used for the detection of the signal component position . also the variance is different for noise only and the signal component positions . it can also be used for the signal position detection . in the case with @xmath174 , the histograms are close to each other , meaning that there is a probability that a signal component is missdetected . histograms are well separated in the case when @xmath216 . it means that the signal component will be detected with an extremely high probability in this case . calculation of the detection probability is straightforward with the assumed probability density functions . [ ptb ] hist_miss_samp.eps the spark based relation can be obtained within the framework of the previous analysis if we assume that the noises ( [ noise_epsee ] ) due to missing samples coming from different components of the same ( unity ) amplitude @xmath217 are added up ( equal amplitudes are the worst case for this kind of analysis ) with the same phase to produce @xcite , @xmath218 at some frequency @xmath191 . random variable @xmath219 ( since @xmath103 is random ) should also assume its maximal possible value ( calculated over all possible @xmath220 and all possible positions @xmath97 , @xmath221 ) . the maximal possible value of this variable is related to the coherence index @xmath197 of the partial dft matrix defined as@xmath222 it means that maximal possible value of this variable is @xmath223 . it should also be assumed that @xmath224 remaining noise components ( due to missing samples ) at the component position @xmath225 assume the same maximal value @xmath223 and that all of them subtract in phase from the signal mean value @xmath28 at @xmath225 . condition for the correct detection of a component position at @xmath225 is then such that the minimal possible amplitude of the component @xmath226 is greater than the maximal possible noise @xmath227 at @xmath191 , i.e. , @xmath228 or @xmath229 where @xmath230 is the spark of the measurement matrix @xmath35 ( spark of matrix @xmath35 is defined as the smallest number of dependent columns or rows ) . according to many very unlikely assumptions that has been made , we can state that this is a very pessimistic bound for @xmath9 . therefore , for a high degree of randomness , a probabilistic approach may be more suitable for the analysis than the spark based relation . assume an additive noise @xmath231 in the input signal . in a matrix form this system of @xmath28 linear equations with @xmath9 unknowns reads @xmath232 the solution follows for @xmath233 where @xmath234 and @xmath235 are the reconstructed signal and noise components respectively . assume that the reconstruction conditions are met and the positions of nonzero coefficients @xmath236 , @xmath54 , ... , @xmath55 can be determined through a single step or iterative procedure @xcite . the equations to find the unknown coefficients are written for @xmath237 time instants @xmath57 , @xmath58 @xmath238 where @xmath239^{t}$ ] is the vector of unknown nonzero coefficients values ( at the determined positions ) and @xmath30 is the vector of the available signal samples @xmath240^{t}.$]the matrix @xmath40 is the measurements matrix @xmath35 with the columns corresponding to the zero - valued transform coefficients @xmath241 , @xmath54 , ... , @xmath55 being excluded . for a given set @xmath64 the coefficients reconstruction condition can be easily calculated as @xmath242 where @xmath243 and @xmath244 is the noise influence to the reconstructed signal coefficients . the input signal - to - noise ( snr ) ratio , if all signal samples were available , is @xmath245 assume the noise energy in @xmath28 available samples used in the reconstruction is@xmath246the correct amplitude in the signal transform at the frequency @xmath220 , in the case if all signal samples were used , would be @xmath247 . to compensate the resulting transform for the known bias in amplitude when only @xmath28 available samples are used we should multiply the coefficient by @xmath248 . it means that is a full recovery , a signal transform coefficient should correspond to the coefficient of the original signal with all signal samples being used . the noise in the transform coefficients will also be multiplied by the same factor . therefore , its energy would be increased to @xmath249 . the signal - to - noise ratio in the recovered signal would be @xmath250 if the distribution of noise in the samples used for reconstruction is the same as in other signal samples then @xmath251 and @xmath252 therefore , a signal reconstruction that would be based on the initial estimate ( [ ms_sum ] ) would worsen snr , since @xmath253 . since only @xmath9 out of @xmath10 dft coefficients are used in the reconstruction the energy of the reconstruction error is reduced for the factor of @xmath254 as well . therefore , the energy of noise in the reconstructed signal is @xmath255 the output signal to noise ratio in the reconstructed signal is @xcite@xmath256 it is related to the input signal to noise ration @xmath257 as @xmath258 _ example : _ theory is illustrated on a four component noisy signal@xmath259 as well , where @xmath260 @xmath261 , @xmath262 , @xmath263 , @xmath264 and @xmath265 . the signal is reconstructed using iterative calculation to find nonzero coefficients @xmath236 , @xmath54 , ... , @xmath55 and ( [ rec_rel ] ) to find the signal . the results are presented in the table [ tab1 ] . the agreement of the numerical statistical results @xmath266 with this simple theory in analysis of noise influence to the reconstruction of sparse signals @xmath267 is high for all considered @xmath257 . [ c]\|\|crrrr\|\|snr in [ db ] & @xmath268 & @xmath269 & @xmath270 & @xmath271 + @xmath257 & 3.5360 & 3.5326 & 3.5788 & 3.5385 + @xmath267 & 18.5953 & 19.5644 & 20.3562 & 21.0257 + @xmath266 & 18.7203 & 19.5139 & 20.2869 & 21.7302 + according to the results in previous section , the missing samples can be represented by a noise influence . assume that we use a reconstruction algorithm for a signal of sparsity @xmath9 on a signal whose dft coefficients @xmath7 are not sparse ( or not sufficiently sparse ) . denote by @xmath61 the sparse signal with @xmath9 nonzero coefficients equal to the largest @xmath9 coefficients of @xmath7 . suppose that the number of components @xmath9 and the measurements matrix satisfy the reconstruction conditions so that a reconstruction algorithm can detect ( one by one or at once ) largest @xmath9 components ( @xmath272 , @xmath273, ... @xmath274 ) and perform signal reconstruction to get @xmath275 . the remaining @xmath276 components ( @xmath277,@xmath278, ... ,@xmath279 ) will be treated as noise in these @xmath9 largest components . variance of a signal component is @xmath280 after reconstruction the variance is @xmath281 the total energy of noise in the reconstructed @xmath9 largest components @xmath275 will be@xmath282 denoting the energy of remaining signal , when the @xmath9 largest are removed from the original signal , by @xmath283 we get @xmath284 if the signal is sparse , i.e. , @xmath285 , then @xmath286 the same result follows if @xmath287@xmath286 that is , the error will be zero if a complete dft matrix is used in calculation . using schwartz inequality @xmath288 follows @xmath289 it means that if @xmath290 is minimized then the upper bound of the error @xmath291 is also minimized . based on the previous results we can easily get the following result . consider a signal @xmath292 , with transformation coefficients @xmath7 and unknown sparsity , including the case when the signal is not sparse . assume that @xmath293 time domain signal samples , corrupted with additive white noise with variance @xmath294 , are available . the signal is reconstructed assuming that its sparsity is @xmath9 . denote the reconstructed signal by @xmath295 , set of its nonzero positions by @xmath296 , and the corresponding original signal transform by @xmath297 where @xmath298 for @xmath299 and @xmath300 for @xmath301 . the total error in the reconstructed signal , with respect to the original signal at the same nonzero coefficient positions , is @xmath302 _ example : _ consider a nonsparse signal @xmath303 where @xmath304 @xmath305 are random frequency indices from @xmath127 to @xmath306 . using @xmath307 and @xmath308 the first @xmath309 components of signal are reconstructed . the remaining @xmath310 signal components are considered as disturbance . reconstruction of @xmath309 largest components is done in @xmath311 independent realizations with different frequencies and positions of available samples . the result for @xmath312 in the noise free case , obtained statistically and by using the theorem , is@xmath313 note that the calculation of @xmath314 is simple since we assumed that the amplitudes of disturbing components are coefficients of a geometric series . one realization with @xmath309 is presented in fig . [ input_noise_stat_snr_noise ] . the case when @xmath315 is presented in fig . [ input_noise_stat_snr_noise_k10 ] . red signal ( with dots ) represents the reconstructed signal with assumed sparsity and the signal with black crosses represents the original nonsparse signal . [ ptb ] input_noise_stat_snr_noise.eps [ ptb ] input_noise_stat_snr_noise_k10.eps in the case of additive complex - valued noise of variance @xmath316 the results are @xmath317 the decrease in the snr due to noise is @xmath318 the simulation is repeated with @xmath268 and the same noise . the snr values are @xmath319 and @xmath320 the goal of compressive sensing is to reconstruct a sparse signal using a reduced set of available samples . it can be done by minimizing the sparsity measure and available samples . a simple algorithm for signal reconstruction is presented . one step reconstruction and an iterative procedure of the reconstruction algorithm are given . noisy environment is taken into account as well . the input noise can degrade the reconstruction limit . however , as far as the reconstruction is possible , the noise caused by missing samples manifests its influence to the results accuracy in simple and direct way through the number of missing samples and signal sparsity . the accuracy of the final result is related to the input noise intensity , number of available samples and the signal sparsity . a theorem presenting error in the case when the reconstruction algorithm defined for reconstruction of sparse signals are used in for nonsparse signals reconstruction is defined as well . the theory is checked and illustrated on numerical examples . consider an @xmath10-dimensional vector @xmath7 whose sparsity is @xmath9 and its @xmath28 measurements @xmath321 , where the measurements matrix @xmath35 is an @xmath36 matrix , with @xmath322 . a reconstruction vector @xmath7 can be achieved from a reduced set of samples / measurements using the sparsity measures minimization . the @xmath323-norm based solution of constrained sparsity measure minimization @xmath324 is the same as the @xmath16-norm based solution of @xmath325 if the measurements matrix @xmath35 satisfies the restricted isometry property for a @xmath326 sparse vector@xmath327 with a sufficiently small @xmath328 . constant @xmath329 is the energy of columns of measurement matrix @xmath35 . for normalized energy @xmath330 , while for the measurement matrix obtained using @xmath28 rows of the standard dft matrix @xmath331 . if the signal @xmath7 is not sparse then the solution of minimization problem ( [ minl1 ] ) denoted by @xmath275 will satisfy @xmath332 where @xmath61 is @xmath9 sparse signal corresponding to @xmath9 largest values of @xmath7 . if the signal @xmath7 is of sparsity @xmath9 then @xmath333 and @xmath334 . in the case of noisy measurements when @xmath335 then @xcite @xmath336 where @xmath337 and @xmath338 are constants depending on @xmath328 . for example , with @xmath339 constants are @xmath340 and @xmath341 , @xcite . e. j. cands , j. romberg , and t. tao , robust uncertainty principles : exact signal reconstruction from highly incomplete frequency information , _ ieee transactions on information theory _ , vol . 52 , no . 2 , 489509 , 2006 . r. e. carrillo , k. e. barner , and t. c. aysal , robust sampling and reconstruction methods for sparse signals in the presence of impulsive noise , _ ieee journal of selected topics in signal processing _ , 2010 , 4(2 ) , pp . 392408 . m. a. figueiredo , r. d. nowak , and s. j. wright , gradient projection for sparse reconstruction : application to compressed sensing and other inverse problems,_ieee journal of selected topics in signal processing , _ , vol . 1 , no . 4 , pp . 586597 , 2007 . b. turlach , on algorithms for solving least squares problems under an l1 penalty or an l1 constraint , _ proc . of the american statistical association ; statistical computing section _ , pp . 25722577 , alexandria , va , 2005 . y. d. zhang and m. g. amin , compressive sensing in nonstationary array processing using bilinear transforms , " in proc . _ ieee sensor array and multichannel signal processing workshop _ , hoboken , nj , june 2012 . l. stankovi , s. stankovi , and m. g. amin , missing samples analysis in signals for applications to l - estimation and compressive sensing , _ signal processing _ , elsevier , volume 94 , jan . 2014 , pages 401408 . e. sejdi , a. cam , l. f. chaparro , c. m. steele , and t. chau , compressive sampling of swallowing accelerometry signals using tf dictionaries based on modulated discrete prolate spheroidal sequences , _ eurasip journal on advances in signal processing _ , 2012:101 doi:10.1186/168761802012101 . i. daubechies , m. defrise , and c. de mol , an iterative thresholding algorithm for linear inverse problems with a sparsity constraint , _ communications on pure and applied mathematics _ , vol . 57 , no . 11 , pp . 14131457 , 2004 . l. stankovi , m. dakovi , s. vujovi , adaptive variable step algorithm for missing samples recovery in sparse signals , _ iet signal processing _ , vol.8 , no.3 , 2014 , pp.246256 , doi : 10.1049/iet - spr.2013.0385 . s. stankovi , i. orovi , and l. stankovi , `` an automated signal reconstruction method based on analysis of compressive sensed signals in noisy environment '' , _ signal processing _ , elsevier , volume 94 , in print .	signals sparse in a transformation domain can be recovered from a reduced set of randomly positioned samples by using compressive sensing algorithms . simple reconstruction algorithms are presented in the first part of the paper . the missing samples manifest themselves as a noise in this reconstruction . once the reconstruction conditions for a sparse signal are met and the reconstruction is achieved , the noise due to missing samples does not influence the results in a direct way . it influences the possibility to recover a signal only . additive input noise will remain in the resulting reconstructed signal . the accuracy of the recovery results is related to the additive input noise . simple derivation of this relation is presented . if a reconstruction algorithm for a sparse signal is used in the reconstruction of a nonsparse signal then the noise due to missing samples will remain and behave as an additive input noise . an exact relation for the mean square error of this error is derived for the partial dft matrix case in this paper and presented in form of a theorem . it takes into account very important fact that if all samples are available then the error will be zero , for both sparse and nonsparse recovered signals . theory is illustrated and checked on statistical examples .
it was known to euclid that two plane polygons of the same area are related by scissors congruence : one can always cut one of them up into polygonal pieces that can be re - assembled to give the other . in the 19th century the analogous result was proved with euclidean geometry replaced by 2dimensional hyperbolic geometry or 2dimensional spherical geometry . the 3rd problem in hilbert s famous 1900 congress address @xcite posed the analogous question for 3dimensional euclidean geometry : are two euclidean polytopes of the same volume `` scissors congruent , '' that is , can one be cut into subpolytopes that can be re - assembled to give the other . hilbert made clear that he expected a negative answer . one reason for the nineteenth century interest in this question was the interest in a sound foundation for the concepts of area and volume . by `` equal area '' euclid _ meant _ scissors congruent , and the attempt in euclid s book xii to provide the same approach for 3dimensional euclidean volume involved what was called an `` exhaustion argument '' essentially a continuity assumption that mathematicians of the nineteenth century were uncomfortable with ( by hilbert s time mostly for aesthetic reasons ) . the negative answer that hilbert expected to his problem was provided the same year by max dehn @xcite . dehn s answer is delighfully simple in modern terms , so we describe it here in full . consider the free @xmath2module generated by the set of congruence classes of 3dimensional polytopes . scissors congruence group _ @xmath3 is the quotient of this module by the relations of scissors congruence . that is , if polytopes @xmath4 can be glued along faces to form a polytope @xmath5 then we set @xmath6=[p_1]+\dots+[p_n]\quad\text{in } { \operatorname{\mathcal p}}({{\mathbb e}}^3).\ ] ] ( a _ polytope _ is a compact domain in @xmath7 that is bounded by finitely many planar polygonal `` faces . '' ) volume defines a map @xmath8 and hilbert s problem asks and @xmath9 represent the same element of @xmath3 if and only if they are _ stably scissors congruent _ rather than scissors congruent , that is , there exists a polytope @xmath10 such that @xmath11 ( disjoint union ) is scissors congruent to @xmath12 . but , in fact , stable scissors congruence implies scissors congruence ( @xcite , see @xcite for an exposition ) . ] about injectivity of this map . dehn defined a new invariant of scissors congrence , now called the _ dehn invariant _ , which can be formulated as a map @xmath13 , where the tensor product is a tensor product of @xmath2modules ( in this case the same as tensor product as @xmath14vector spaces ) . if @xmath15 is an edge of a polytope @xmath5 we will denote by @xmath16 and @xmath17 the length of @xmath15 and dihedral angle ( in radians ) at @xmath15 . for a polytope @xmath5 we define the _ dehn invariant _ @xmath18 as @xmath19 we then extend this linearly to a homomorphism on @xmath3 . it is an easy but instructive exercise to verify that * @xmath20 is well - defined on @xmath3 , that is , it is compatible with scissors congruence ; * @xmath20 and @xmath21 are independent on @xmath3 in the sense that their kernels generate @xmath3 ( whence @xmath22 and @xmath23 ) ; * the image of @xmath20 is uncountable . in particular , @xmath24 is not just non - trivial , but even uncountable , giving a strong answer to hilbert s question . to give an explicit example , the regular simplex and cube of equal volume are not scissors congruent : a regular simplex has non - zero dehn invariant , and the dehn invariant of a cube is zero . of course , this answer to hilbert s problem is really just a start . it immediately raises other questions : * are volume and dehn invariant sufficient to classify polytopes up to scissors congruence ? * what about other dimensions ? * what about other geometries ? the answer to the first question is `` yes . '' sydler proved in 1965 that @xmath25 is injective . later jessen @xcite simplified his difficult argument somewhat and proved an analogous result for @xmath26 and the argument has been further simplified in @xcite . except for these results and the classical results for dimensions @xmath27 no complete answers are known . in particular , fundamental questions remain open about @xmath28 and @xmath29 . note that the definition of dehn invariant applies with no change to @xmath28 and @xmath29 . the dehn invariant should be thought of as an `` elementary '' invariant , since it is defined in terms of 1dimensional measure . for this reason ( and other reasons that will become clear later ) we are particularly interested in the kernel of dehn invariant , so we will abbreviate it : for @xmath30 @xmath31 in terms of this notation sydler s theorem that volume and dehn invariant classify scissors congruence for @xmath7 can be reformulated : @xmath32 it is believed that volume and dehn invariant classify scissors congruence also for hyperbolic and spherical geometry : [ sufficiency ] @xmath33 is injective and @xmath34 is injective . on the other hand @xmath35 is also surjective , but this results from the existence of similarity transformations in euclidean space , which do not exist in hyperbolic or spherical geometry . in fact , dupont @xcite proved : @xmath33 and @xmath34 have countable image . thus the dehn invariant sufficiency conjecture would imply : [ rigidity ] @xmath36 and @xmath37 are countable . the following collects results of bkstedt , brun , dupont , parry , sah and suslin ( @xcite , @xcite , @xcite , @xcite ) . [ divisibility ] @xmath28 and @xmath29 and their subspaces @xmath36 and @xmath37 are uniquely divisible groups , so they have the structure of @xmath14vector spaces . as @xmath14vector spaces they have infinite rank . the rigidity conjecture thus says @xmath36 and @xmath37 are @xmath14vector spaces of countably infinite rank . [ image of volume ] the subgroups @xmath38 and @xmath39 of @xmath40 are @xmath14vector subspaces of countable dimension . many generalizations of hilbert s problem have been considered , see eg @xcite for an overview . there are generalizations of dehn invariant to all dimensions and the analog of the dehn invariant sufficiency conjectures have often been made in greater generality , see eg @xcite , @xcite , @xcite . the particular dehn invariant that we are discussing here is a codimension 2 dehn invariant . conjecture [ sufficiency ] appears in various other guises in the literature . for example , as we shall see , the @xmath41 case is equivalent to a conjecture about rational relations among special values of the dilogarithm function which includes as a very special case a conjecture of milnor @xcite about rational linear relations among values of the dilogarithm at roots of unity . conventional wisdom is that even this very special case is a very difficult conjecture which is unlikely to be resolved in the forseeable future . in fact , dehn invariant sufficiency would imply the ranks of the vector spaces of volumes in corollary [ image of volume ] are infinite , but at present these ranks are not even proved to be greater than 1 . even worse : although it is believed that the volumes in question are always irrational , it is not known if a single one of them is ! as we describe later , work of bloch , dupont , parry , sah , wagoner , and suslin connects the dehn invariant kernels with algebraic @xmath0theory of @xmath1 , and the above conjectures are then equivalent to standard conjectures in algebraic @xmath0theory . in particular , the scissors congruence rigidity conjectures for @xmath42 and @xmath43 are together equivalent to the rigidity conjecture for @xmath44 , which can be formulated that @xmath45 ( indecomposable part of quillen s @xmath46 ) is countable . this conjecture is probably much easier than the dehn invariant sufficiency conjecture . the conjecture about rational relations among special values of the dilogarithm has been broadly generalized to polylogarithms of all degrees by zagier ( section 10 of @xcite ) . the connections between scissors congruence and algebraic @xmath0theory have been generalised to higher dimensions , in part conjecturally , by goncharov @xcite . we will return to some of these issues later . we also refer the reader to the very attractive exposition in @xcite of these connnections in dimension 3 . i would like to acknowledge the support of the australian research council for this research , as well as the the max institut fr mathematik in bonn , where much of this paper was written . thurston s geometrization conjecture , much of which is proven to be true , asserts that , up to a certain kind of canonical decomposition , 3manifolds have geometric structures . these geometric structures belong to eight different geometries , but seven of these lead to manifolds that are describable in terms of surface topology and are very easily classified . the eighth geometry is hyperbolic geometry @xmath41 . thus if one accepts the geometrization conjecture then the central issue in understanding 3manifolds is to understand hyperbolic 3manifolds . suppose therefore that @xmath47 is a hyperbolic 3manifold . we will always assume @xmath48 is oriented and for the moment we will also assume @xmath48 is compact , though we will be able to relax this assumption later . we can subdivide @xmath48 into small geodesic tetrahedra , and then the sum of these tetrahedra represents a class @xmath49 which is an invariant of @xmath48 . we call this the _ scissors congruence class of @xmath48_. note that when we apply the dehn invariant to @xmath50 the contributions coming from each edge @xmath15 of the triangulation sum to @xmath51 which is zero in @xmath52 . thus the scissors congruence class @xmath50 lies in @xmath36 . how useful is this invariant of @xmath48 ? we can immediately see that it is non - trivial , since at least it detects volume of @xmath48 : @xmath53 now it was suggested by thurston in @xcite that the volume of hyperbolic 3manifolds should have some close relationship with another geometric invariant , the chern simons invariant @xmath54 . a precise analytic relationship was then conjectured in @xcite and proved in @xcite ( a new proof follows from the work discussed here , see @xcite ) . we will not discuss the definition of this invariant here ( it is an invariant of compact riemmanian manifolds , see @xcite , which was extended also to non - compact finite volume hyperbolic 3manifolds by meyerhoff @xcite ) . it suffices for the present discussion to know that for a finite volume hyperbolic 3manifold @xmath48 the chern simons invariant lies in @xmath55 . moreover , the combination @xmath56 turns out to have good analytic properties and is therefore a natural `` complexification '' of volume for hyperbolic manifolds . given this intimate relationship between volume and chern simons invariant , it becomes natural to ask if @xmath57 is also detected by @xmath50 . the answer , unfortunately , is an easy `` no . '' the point is that @xmath57 is an orientation sensitive invariant : @xmath58 , where @xmath59 means @xmath48 with reversed orientation . but , as gerling pointed out in a letter to gauss on 15 april 1844 : scissors congruence can not see orientation because any polytope is scissors congruent to its mirror image tetrahedra occur in @xmath60 mirror image pairs . ] . thus @xmath61 and there is no hope of @xmath57 being computable from @xmath50 . this raises the question : [ question1 ] is there some way to repair the orientation insensitivity of scissors congruence and thus capture chern simons invariant ? the answer to this question is `` yes '' and lies in the so called `` bloch group , '' which was invented for entirely different purposes by bloch ( it was put in final form by wigner and suslin ) . to explain this we start with a result of dupont and sah @xcite about ideal polytopes hyperbolic polytopes whose vertices are at infinity ( such polytopes exist in hyperbolic geometry , and still have finite volume ) . [ ideal suffices ] ideal hyperbolic tetrahedra represent elements in @xmath28 and , moreover , @xmath28 is generated by ideal tetrahedra . to help understand this proposition observe that if @xmath62 is a non - ideal tetrahedron and @xmath15 is the ideal point at which the extension of edge @xmath63 meets infinity then @xmath62 can be represented as the difference of the two tetrahedra @xmath64 and @xmath65 , each of which have one ideal vertex . we have thus , in effect , `` pushed '' one vertex off to infinity . in the same way one can push a second and third vertex off to infinity , and the fourth , but this is rather harder . anyway , we will accept this proposition and discuss its consequence for scissors congruence . the first consequence is a great gain in convenience : a non - ideal tetrahedron needs six real parameters satisfying complicated inequalities to characterise it up to congruence while an ideal tetrahedron can be neatly characterised by a single complex parameter in the upper half plane . we shall denote the standard compactification of @xmath41 by @xmath66 . an ideal simplex @xmath67 with vertices @xmath68 is determined up to congruence by the cross - ratio @xmath69=\frac{(z_3-z_2)(z_4-z_1)}{(z_3-z_1)(z_4-z_2)}.\ ] ] permuting the vertices by an even ( ie orientation preserving ) permutation replaces @xmath70 by one of @xmath71 the parameter @xmath70 lies in the upper half plane of @xmath1 if the orientation induced by the given ordering of the vertices agrees with the orientation of @xmath41 . there is another way of describing the cross - ratio parameter @xmath72 $ ] of a simplex . the group of orientation preserving isometries of @xmath41 fixing the points @xmath73 and @xmath74 is isomorphic to the multiplicative group @xmath75 of nonzero complex numbers . the element of this @xmath75 that takes @xmath76 to @xmath77 is @xmath70 . thus the cross - ratio parameter @xmath70 is associated with the edge @xmath78 of the simplex . the parameter associated in this way with the other two edges @xmath79 and @xmath80 out of @xmath73 are @xmath81 and @xmath82 respectively , while the edges @xmath83 , @xmath84 , and @xmath85 have the same parameters @xmath70 , @xmath81 , and @xmath82 as their opposite edges . see figure 1 . this description makes clear that the dihedral angles at the edges of the simplex are @xmath86 , @xmath87 , @xmath88 respectively , with opposite edges having the same angle . .35 figure 1 now suppose we have five points @xmath89 . any four - tuple of these five points spans an ideal simplex , and the convex hull of these five points decomposes in two ways into such simplices , once into two of them and once into three of them . we thus get a scissors congruence relation equating the two simplices with the three simplices . it is often called the `` five - term relation . '' to express it in terms of the cross - ratio parameters it is convenient first to make an orientation convention . we allow simplices whose vertex ordering does not agree with the orientation of @xmath41 ( so the cross - ratio parameter is in the lower complex half - plane ) but make the convention that this represents the negative element in scissors congruence . an odd permutation of the vertices of a simplex replaces the cross - ratio parameter @xmath70 by @xmath90 so if we denote by @xmath91 $ ] the element in @xmath28 represented by an ideal simplex with parameter @xmath70 , then our orientation rules say : @xmath92=[1-\frac 1z]=[\frac 1{1-z}]=-[\frac1z]=-[\frac{z-1}z]=-[1-z ] . \label{invsim}\ ] ] these orientation rules make the five - term scissors congruence relation described above particularly easy to state : @xmath93=0.\ ] ] the cross - ratio parameters occuring in this formula can be expressed in terms of the first two as @xmath94 = : x\qquad[z_0{\hbox{$\,:\,$}}z_2{\hbox{$\,:\,$}}z_3{\hbox{$\,:\,$}}z_4]=:y\\ [ z_0{\hbox{$\,:\,$}}z_1{\hbox{$\,:\,$}}z_3{\hbox{$\,:\,$}}z_4]=\frac yx\quad [ z_0{\hbox{$\,:\,$}}z_1{\hbox{$\,:\,$}}z_2{\hbox{$\,:\,$}}z_4]=\frac { 1-x^{-1}}{1-y^{-1}}\quad [ z_0{\hbox{$\,:\,$}}z_1{\hbox{$\,:\,$}}z_2{\hbox{$\,:\,$}}z_3]=\frac { 1-x}{1-y}\end{gathered}\ ] ] so the five - term relation can also be written : @xmath95-[y]+[\frac yx]-[\frac{1-x^{-1}}{1-y^{-1}}]+[\frac{1-x}{1-y}]=0 . \label{5term}\ ] ] we lose nothing if we also allow degenerate ideal simplices whose vertices lie in one plane so the parameter @xmath70 is real ( we always require that the vertices are distinct , so the parameter is in @xmath96 ) , since the five - term relation can be used to express such a `` flat '' simplex in terms of non - flat ones , and one readily checks no additional relations result . thus we may take the parameter @xmath70 of an ideal simplex to lie in @xmath97 and every such @xmath70 corresponds to an ideal simplex . one can show that relations ( [ invsim ] ) follow from the five - term relation ( [ 5term ] ) , so we consider the quotient @xmath98 of the free @xmath2module on @xmath97 . proposition [ ideal suffices ] can be restated that there is a natural surjection @xmath99 . in fact dupont and sah ( loc . prove : [ ideal really suffices ] the scissors congruence group @xmath28 is the quotient of @xmath100 by the relations @xmath91=-[\overline z]$ ] which identify each ideal simplex with its mirror image . thus @xmath100 is a candidate for the orientation sensitive scissors congruence group that we were seeking . indeed , it turns out to do ( almost ) exactly what we want . the analog of the dehn invariant has a particularly elegant expression in these terms . first note that the above theorem expresses @xmath28 as the `` imaginary part '' @xmath101 ( negative co - eigenspace under conjugation turns out to be a @xmath14vector space and is therefore the sum of its @xmath102 eigenspaces , so `` co - eigenspace '' is the same as `` eigenspace . '' ] ) of @xmath100 . [ complex dehn ] the dehn invariant @xmath103 is twice the `` imaginary part '' of the map @xmath104\mapsto ( 1-z)\wedge z\ ] ] so we shall call this map the `` complex dehn invariant . '' we denote the kernel of complex dehn invariant @xmath105 and call it the `` bloch group of @xmath1 . '' ( we shall explain this proposition further in an appendix to this section . ) a hyperbolic 3manifold @xmath48 now has an `` orientation sensitive scissors congruence class '' which lies in this bloch group and captures both volume and chern simons invariant of @xmath48 . namely , there is a map @xmath106 introduced by bloch and wigner called the _ bloch regulator map _ , whose imaginary part is the volume map on @xmath107 , and one has : let @xmath48 be a complete oriented hyperbolic 3manifold of finite volume . then there is a natural class @xmath108 associated with @xmath48 and @xmath109 . this theorem answers question [ question1 ] . but there are still two aesthetic problems : * the bloch regulator @xmath110 plays the rle for orientation sensitive scissors congruence that volume plays for ordinary scissors congruence . but @xmath21 is defined on the whole scissors congruence group @xmath28 while @xmath110 is only defined on the kernel @xmath107 of complex dehn invariant . * the chern simons invariant @xmath54 is an invariant in @xmath55 but the invariant @xmath111 only computes it in @xmath112 . we resolve both these problems in section [ extended bloch ] . we describe the bloch regulator map @xmath110 later . it would be a little messy to describe at present , although its imaginary part ( volume ) has a very nice description in terms of ideal simplices . indeed , the volume of an ideal simplex with parameter @xmath70 is @xmath113 , where @xmath114 is the so called `` bloch wigner dilogarithm function '' given by : @xmath115 and @xmath116 is the classical dilogarithm function . it follows that @xmath113 satisfies a functional equation corresponding to the five - term relation ( see below ) . to worry about the second `` aesthetic problem '' above could be considered rather greedy . after all , @xmath57 takes values in @xmath55 which is the direct sum of @xmath117 and uncountably many copies of @xmath14 , and we have only lost part of the former summand . however , it is not even known if the chern simons invariant takes _ any _ non - zero values in @xmath112 . as we shall see , this would be implied by the sufficiency of dehn invariant for @xmath118 ( conjecture [ sufficiency ] ) . the analogous conjecture in our current situation is : [ csufficiency ] @xmath119 is injective . again , the following is known by work of bloch : @xmath119 has countable image . thus the complex dehn invariant sufficiency conjecture would imply : [ bloch rigidity ] @xmath107 is countable . [ cdivisibility ] @xmath100 and its subgroup @xmath107 are uniquely divisible groups , so they have the structure of @xmath14vector spaces . as @xmath14vector spaces they have infinite rank . note that the bloch group @xmath107 is defined purely algebraically in terms of @xmath1 , so we can define a bloch group @xmath120 analogously in the literature vary in ways that can mildly affect its torsion if @xmath121 is not algebraically closed . ] for any field @xmath121 . this group @xmath120 is uniquely divisible whenever @xmath121 contains an algebraically closed field . it is not hard to see that the rigidity conjecture [ bloch rigidity ] is equivalent to the conjecture that @xmath122 is an isomorphism ( here @xmath123 is the field of algebraic numbers ; it is known that @xmath122 is injective ) . suslin has conjectured more generally that @xmath124 is an isomorphism if @xmath121 is the algebraic closure of the prime field in @xmath0 . conjecture [ csufficiency ] has been made in greater generality by ramakrishnan @xcite in the context of algebraic @xmath0theory . conjectures [ csufficiency ] and [ bloch rigidity ] are in fact equivalent to the dehn invariant sufficiency and rigidity conjectures [ sufficiency ] and [ rigidity ] respectively for @xmath41 and @xmath118 together . this is because of the following theorem which connects the various dehn kernels . it also describes the connections with algebraic @xmath0theory and homology of the lie group @xmath125 considered as a discrete group . it collates results of of bloch , bkstedt , brun , dupont , parry and sah and wigner ( see @xcite and @xcite ) . there is a natural exact sequence @xmath126 moreover there are natural isomorphisms : @xmath127 , @xmath128 , @xmath129 and @xmath130 , where @xmath131 is generated by the class of the 3sphere and @xmath132 is the subgroup generated by suspensions of triangles in @xmath133 with rational angles . the cheeger simons map @xmath134 of @xcite induces on the one hand the bloch regulator map @xmath119 and on the other hand its real and imaginary parts correspond to the volume maps on @xmath135 and @xmath36 via the above isomorphisms . the isomorphisms of the theorem are proved via isomorphisms @xmath136 and @xmath137 . we have described the geometry of the isomorphism @xmath138 in theorem [ ideal really suffices ] . the geometry of the isomorphism @xmath130 remains rather mysterious . the exact sequence and first isomorphism in the above theorem are valid for any algebraically closed field of characteristic @xmath139 . thus conjecture [ bloch rigidity ] is also equivalent to each of the four : * is @xmath140 an isomorphism ? is @xmath45 countable ? * is @xmath141 an isomorphism ? is @xmath142 countable ? the fact that volume of an ideal simplex is given by the bloch wigner dilogarithm function @xmath113 clarifies why the @xmath41 dehn invariant sufficiency conjecture [ sufficiency ] is equivalent to a statement about rational relations among special values of the dilogarithm function . don zagier s conjecture about such rational relations , mentioned earlier , is that any rational linear relation among values of @xmath114 at algebraic arguments must be a consequence of the relations @xmath143 and the five - term functional relation for @xmath114 : @xmath144 differently expressed , he conjectures that the volume map is injective on @xmath145 . if one assumes the scissors congruence rigidity conjecture for @xmath41 ( that @xmath146 ) then the dehn invariant sufficiency conjecture for @xmath41 is just that @xmath114 is injective on the subgroup @xmath147 , so under this assumption zagier s conjecture is much stronger . milnor s conjecture , mentioned earlier , can be formulated that the values of @xmath148 , as @xmath149 runs through the primitive @xmath150-th roots of unity in the upper half plane , are rationally independent for any @xmath150 . this is equivalent to injectivity modulo torsion of the volume map @xmath114 on @xmath151 for the cyclotomic field @xmath152 . for this field @xmath153 modulo torsion . this is of finite rank but @xmath154 is of infinite rank , so even when restricted to @xmath155 zagier s conjecture is much stronger than milnor s . zagier himself has expressed doubt that milnor s conjecture can be resolved in the forseeable future . conjecture [ csufficiency ] can be similarly formulated as a statement about special values of a different dilogarithm function , the `` rogers dilogarithm , '' which we will define later . to define the dehn invariant of an ideal polytope we first cut off each ideal vertex by a horoball based at that vertex . we then have a polytope with some horospherical faces but with all edges finite . we now compute the dehn invariant using the geodesic edges of this truncated polytope ( that is , only the edges that come from the original polytope and not those that bound horospherical faces ) . this is well defined in that it does not depend on the sizes of the horoballs we used to truncate our polytope . ( to see this , note that dihedral angles of the edges incident on an ideal vertex sum to a multiple of @xmath156 , since they are the angles of the horospherical face created by truncation , which is an euclidean polygon . changing the size of the horoball used to truncate these edges thus changes the dehn invariant by a multiple of something of the form @xmath157 , which is zero in @xmath158 . ) now consider the ideal tetrahedron @xmath159 with parameter @xmath70 . we may position its vertices at @xmath160 . there is a klein 4group of symmetries of this tetrahedron and it is easily verified that it takes the following horoballs to each other : * at @xmath161 the horoball @xmath162 ; * at @xmath139 the horoball of euclidean diameter @xmath163 ; * at @xmath164 the horoball of euclidean diameter @xmath165 ; * at @xmath70 the horoball of euclidean diameter @xmath166 . after truncation , the vertical edges thus have lengths @xmath167 , @xmath168 , and @xmath169 respectively , and we have earlier said that their angles are @xmath170 respectively . thus , adding contributions , we find that these three edges contribute @xmath171 to the dehn invariant . by symmetry the other three edges contribute the same , so the dehn invariant is : @xmath172 to understand the `` imaginary part '' of @xmath173 we use the isomorphism @xmath174 to represent @xmath175 ( the equality on the third line is because tensoring over @xmath2 with a divisible group is effectively the same as tensoring over @xmath14 ) . under this isomorphism we have @xmath176 . the scissors congruence invariant @xmath177 turns out to be a very computable invariant . to explain this we must first describe the `` invariant trace field '' or `` field of definition '' of a hyperbolic 3manifold . suppose therefore that @xmath47 is a hyperbolic manifold , so @xmath178 is a discrete subgroup of the orientation preserving isometry group @xmath179 of @xmath41 . @xcitethe _ invariant trace field _ of @xmath48 is the subfield of @xmath1 generated over @xmath14 by the squares of traces of elements of @xmath178 . we will denote it @xmath180 or @xmath181 . this field @xmath180 is an algebraic number field ( finite extension of @xmath14 ) and is a commensurability invariant , that is , it is unchanged on passing to finite covers of @xmath48 ( finite index subgroups of @xmath178 ) . moreover , if @xmath48 is an arithmetic hyperbolic 3manifold ( that is , @xmath178 is an arithmetic group ) , then @xmath180 is the field of definition of this arithmetic group in the usual sense . see @xcite . now if @xmath121 is an algebraic number field then @xmath120 is isomorphic to @xmath182 , where @xmath183 is the number of conjugate pairs of complex embeddings @xmath184 of @xmath121 . indeed , if these complex embeddings are @xmath185 then a reinterpretation of a theorem of borel @xcite about @xmath44 says : [ borel regulator ] the `` borel regulator map '' @xmath186 induced on generators of @xmath187 by @xmath91\mapsto({\operatorname{vol}}[\sigma_1(z)],\dots,{\operatorname{vol}}[\sigma_{r_2}(z)])$ ] maps @xmath188 isomorphically onto a full lattice in @xmath189 . a corollary of this theorem is that an embedding @xmath190 induces an embedding @xmath191 . ( this is because the theorem implies that an element of @xmath120 is determined modulo torsion by the set of volumes of its galois conjugates , which are invariants defined on @xmath107 . ) moreover , since @xmath107 is a @xmath14vector space , @xmath192 . now if @xmath48 is a hyperbolic manifold then its invariant trace field @xmath180 comes embedded in @xmath1 so we get an explicit embedding @xmath193 whose image , which is isomorphic to @xmath194 , we denote by @xmath195 . the element @xmath177 lies in the subspace @xmath196 . in fact neumann and yang show that @xmath177 is well defined in @xmath197 for some explicit multi - quadratic field extension @xmath0 of @xmath180 , which implies that @xmath198 is actually well defined in @xmath199 for some @xmath200 . moreover , one can always take @xmath201 if @xmath48 is non - compact , but we do not know if one can for compact @xmath48 . in view of this theorem we see that the following data effectively determines @xmath177 modulo torsion : * the invariant trace field @xmath180 . * the image of @xmath177 in @xmath189 under the borel regulator map of theorem [ borel regulator ] . to compute @xmath177 we need a collection of ideal simplices that triangulates @xmath48 in some fashion . if @xmath48 is compact , this clearly can not be a triangulation in the usual sense . in @xcite it is shown that one can use any `` degree one ideal triangulation '' to compute @xmath177 . this means a finite complex formed of ideal hyperbolic simplices plus a map of it to @xmath48 that takes each ideal simplex locally isometrically to @xmath48 and is degree one almost everywhere . these always exist ( see @xcite for a discussion ) . special degree one ideal triangulations have been used extensively in practice , eg in jeff weeks program snappea @xcite for computing with hyperbolic 3manifolds . oliver goodman has written a program snap @xcite ( building on snappea ) which finds degree one ideal triangulations using exact arithmetic in number fields and computes the invariant trace field and high precision values for the borel regulator on @xmath177 . such calculations can provide numerical evidence for the complex dehn invariant sufficiency conjecture . here is a typical result of such calculations . to ensure that the bloch group has rank @xmath202 we want a field with at least two complex embeddings . one of the simplest is the ( unique ) quartic field over @xmath14 of discriminant @xmath203 . this is the field @xmath204 with @xmath205 . this polynomial is irreducible with roots @xmath206 and @xmath207 . the field @xmath121 thus has two complex embeddings @xmath208 up to complex conjugation , one with image @xmath209 and one with image @xmath210 . the bloch group @xmath120 is thus isomorphic to @xmath211 modulo torsion . this field occurs as the invariant trace field for two different hyperbolic knot complements in the standard knot tables up to @xmath212 crossings , the @xmath213crossing knot @xmath214 and the @xmath215crossing knot @xmath216 , but the embeddings in @xmath1 are different . for @xmath214 one gets @xmath217 and for @xmath216 one gets @xmath218 . the scissors congruence classes are @xmath219 + [ 1-\tau ] + [ \frac12(1-\tau^2+\tau^3)]\in { { { \mathcal b}}}(k)\\ \beta(7_7)&= : \beta_2=4[2-\tau-\tau^3]+4[\tau+\tau^2+\tau^3 ] \in { { { \mathcal b}}}(k ) \end{aligned}\ ] ] where @xmath220 is the class of @xmath221 in @xmath222 . these map under the borel regulator @xmath223 ( with respect to the embeddings @xmath208 ) to @xmath224 in particular , the volumes of these knot complements are @xmath225 and @xmath226 respectively snap has access to a large database of small volume compact manifolds . searching this database for manifolds whose volumes are small rational linear combinations of @xmath227 and @xmath228 yielded just eight examples , three with volume @xmath229 , four with volume @xmath230 and one with volume @xmath231 . the complex dehn invariant sufficiency conjecture predicts ( under the assumption that the rational dependencies found are exact ) that these should all have invariant trace field containing @xmath217 . checking with snap confirms that their invariant trace fields equal @xmath217 and their scissors congruence classes in @xmath232 ( computed numerically using the borel regulator ) are @xmath233 , @xmath234 , and @xmath235 respectively . in section [ hyperbolic 3-manifolds ] we saw that @xmath100 and @xmath236 play a role of `` orientation sensitive '' scissors congruence group and kernel of dehn invariant respectively , and that the analog of the volume map is then the borel regulator @xmath110 . but , as we described there , this analogy suffers because @xmath110 is defined on the dehn kernel @xmath107 rather than on the whole of @xmath100 and moreover , it takes values in @xmath237 , rather than in @xmath238 . the repair turns out to be to use , instead of @xmath97 , a certain disconnected @xmath239 cover of @xmath97 to define `` extended versions '' of the groups @xmath100 and @xmath107 . this idea developed out of a suggestion by jun yang . we shall denote the relevant cover of @xmath97 by @xmath240 . we start with two descriptions of it . the second will be a geometric interpretation in terms of ideal simplices . let @xmath5 be @xmath97 split along the rays @xmath241 and @xmath242 . thus each real number @xmath243 outside the interval @xmath244 $ ] occurs twice in @xmath5 , once in the upper half plane of @xmath1 and once in the lower half plane of @xmath1 . we denote these two occurences of @xmath243 by @xmath245 and @xmath246 . we construct @xmath240 as an identification space from @xmath247 by identifying @xmath248 we will denote the equivalence class of @xmath249 by @xmath250 . @xmath240 has four components : @xmath251 where @xmath252 is the set of @xmath253 with @xmath254 and @xmath255 ( mod @xmath256 ) . we may think of @xmath257 as the riemann surface for the function @xmath258 defined by @xmath259 . if for each @xmath260 we take the branch @xmath261 of this function on the portion @xmath262 of @xmath257 we get an analytic function from @xmath257 to @xmath263 . in the same way , we may think of @xmath240 as the riemann surface for the collection of all branches of the functions @xmath264 on @xmath97 . we can interpret @xmath240 in terms of ideal simplices . suppose we have an ideal simplex @xmath67 with parameter @xmath265 . recall that this parameter is associated to an edge of @xmath67 and that other edges of @xmath67 have parameters @xmath266 with opposite edges of @xmath67 having the same parameter ( see figure 1 ) . note that @xmath267 , so the sum @xmath268 is an odd multiple of @xmath269 , depending on the branches of @xmath270 used . in fact , if we use the standard branch of log then this sum is @xmath269 or @xmath271 depending on whether @xmath70 is in the upper or lower half plane . this reflects the fact that the three dihedral angles of an ideal simplex sum to @xmath156 . [ flattening ] we shall call any triple of the form @xmath272 with @xmath273 a _ combinatorial flattening _ for our simplex @xmath67 . thus a combinatorial flattening is an adjustment of each of the three dihedral angles of @xmath67 by a multiple of @xmath156 so that the resulting angle sum is zero . each edge @xmath15 of @xmath67 is assigned one of the components @xmath274 of @xmath275 , with opposite edges being assigned the same component . we call @xmath274 the _ log - parameter _ for the edge @xmath15 and denote it @xmath276 . for @xmath253 we define @xmath277 and @xmath278 is then a map of @xmath240 to the set of combinatorial flattenings of simplices . [ cover is flattenings ] this map @xmath278 is a bijection , so @xmath240 may be identified with the set of all combinatorial flattenings of ideal tetrahedra . we must show that @xmath279 determines @xmath250 . it clearly suffices to recover @xmath70 . but up to sign @xmath70 equals @xmath280 and @xmath281 equals @xmath282 , and the knowledge of both @xmath70 and @xmath281 up to sign determines @xmath70 . we shall define a group @xmath283 as @xmath284(relations ) , where the relations in question are a lift of the five - term relations ( [ 5term ] ) that define @xmath100 , plus an extra relation that just eliminates an element of order @xmath256 . we first recall the situation of the five - term relation ( [ 5term ] ) . if @xmath285 are five distinct points of @xmath286 , then each choice of four of five points @xmath287 gives an ideal simplex . we denote the simplex which omits vertex @xmath288 by @xmath289 . we denote the cross - ratio parameters of these simplices by @xmath290 $ ] . recall that @xmath291 can be written in terms of @xmath292 and @xmath293 as @xmath294 the five - term relation was @xmath295=0 $ ] , so the lifted five - term relation will have the form # 1(x_#1;p_#1,q_#1 ) @xmath296 with certain relations on the @xmath297 and @xmath298 . we need to describe these relations . using the map of lemma [ cover is flattenings ] , each summand in this relation ( [ 5-term ] ) represents a choice @xmath299 of combinatorial flattening for one of the five ideal simplices . for each edge @xmath15 connecting two of the points @xmath288 we get a corresponding linear combination @xmath300 of log - parameters ( definition [ flattening ] ) , where we put @xmath301 if the line @xmath15 is not an edge of @xmath289 . this linear combination has just three non - zero terms corresponding to the three simplices that meet at the edge @xmath15 . one easily checks that the real part is zero and the imaginary part can be interpreted ( with care about orientations ) as the sum of the `` adjusted angles '' of the three flattened simplices meeting at @xmath15 . [ geometric five term ] we say that the @xmath302 satisfy the _ flattening condition _ if each of the above linear combinations ( [ edge sums ] ) of log - parameters is equal to zero . that is , the adjusted angle sum of the three simplices meeting at each edge is zero . in this case relation ( [ 5-term ] ) is an instance of the _ lifted five - term relation_. there are ten edges in question , so the flattening conditions are ten linear relations on the ten integers @xmath303 . but these equations turn out to be linearly dependant , and the space of solutions is @xmath304dimensional . for example , if the five parameters @xmath305 are all in the upper half - plane ( one can check that this means @xmath306 is in the upper half - plane and @xmath221 is inside the triangle with vertices @xmath307 ) then the conditions are equivalent to : @xmath308 which express @xmath309 , @xmath310 , @xmath311 , @xmath312 , @xmath313 in terms of @xmath314 , @xmath315 , @xmath316 , @xmath317 , @xmath318 . thus , in this case the lifted five - term relation becomes : @xmath319 this situation corresponds to the configuration of figure 2 for the ideal vertices @xmath287 , .5 -7 mm figure 2 with @xmath73 and @xmath77 on opposite sides of the plane of the triangle @xmath320 and the line from @xmath73 to @xmath77 passing through the interior of this triangle . the extended pre - bloch group @xmath283 is the group @xmath321 @xmath322+[x;p',q']=[x;p , q']+[x;p',q].\ ] ] ( we call relation ( [ transfer ] ) the _ transfer relation _ ; one can show that if one omits it then @xmath283 is replaced by @xmath323 , where the @xmath324 is generated by the element @xmath325+[x,0,0]-[x,1,0]-[x,0,1]$ ] , which is independant of @xmath221 . ) the relations we are using are remarkably natural . to explain this we need a beautiful version of the dilogarithm function called the _ rogers dilogarithm _ : @xmath326 the extra @xmath327 is not always included in the definition but it improves the functional equation . @xmath328 is singular at @xmath139 and @xmath164 and is not well defined on @xmath97 , but it lifts to an analytic function @xmath329 we also consider the map @xmath330 relation ( [ transfer ] ) is clearly a functional equation for both @xmath331 and @xmath332 . it turns out that the same is true for the lifted five - term relation . in fact : if @xmath302 , @xmath333 satisfy the flattening condition , so @xmath334 is an instance of the lifted five - term relation , then @xmath335 and @xmath336 moreover , each of these equations also characterises the flattening condition . thus the flattening condition can be defined either geometrically , as we introduced it , or as the condition that makes the five - term functional equation for either @xmath331 or @xmath332 valid . in any case , we now have : @xmath331 and @xmath332 define maps @xmath337 we call @xmath332 the _ extended dehn invariant _ and call its kernel @xmath338 the _ extended bloch group_. the final step in our path from hilbert s 3rd problem to invariants of 3manifolds is given by the following theorem . [ final step ] a hyperbolic 3manifold @xmath48 has a natural class @xmath339 . moreover , @xmath340 . to define the class @xmath341 directly from an ideal triangulation one needs to use a more restrictive type of ideal triangulation than the degree one ideal triangulations that suffice for @xmath177 . for instance , the triangulations constructed by epstein and penner @xcite in the non - compact case and by thurston @xcite in the compact case are of the appropriate type . one then chooses flattenings of the ideal simplices of @xmath0 so that the whole complex @xmath0 satisfies certain `` flatness '' conditions . the sum of the flattened ideal simplices then represents @xmath341 up to a @xmath342 correction . the main part of the flatness conditions on @xmath0 are the conditions that adjusted angles around each edge of @xmath0 sum to zero together with similar conditions on homology classes at the cusps of @xmath48 . if one just requires these conditions one obtains @xmath341 up to @xmath60torsion . additional mod @xmath256 flatness conditions on homology classes determine @xmath341 modulo @xmath213torsion . the final @xmath342 correction is eliminated by appropriately ordering the vertices of the simplices of @xmath0 . it takes some work to see that all these conditions can be satisfied and that the resulting element of @xmath343 is well defined , see @xcite . there is a commutative diagram with exact rows and columns : @xmath345 here @xmath346 is the group of roots of unity and the labelled maps that have not yet been defined are : @xmath347-[z,0,0]\in { \widehat{{\mathcal p}}}({{\mathbb c}});\\ \xi[z]&=\log z \wedge \pi i;\\ \epsilon(w_1\wedge w_2)&= ( e^{w_1}\wedge e^{w_2 } ) . \end{aligned}\ ] ] the use of the disconnected cover @xmath240 of @xmath97 rather than the universal abelian cover ( the component @xmath257 of @xmath240 ) in defining the extended bloch group may seem unnatural . if one uses @xmath257 instead of @xmath240 one obtains extended bloch groups @xmath348 and @xmath349 which are non - trivial @xmath324 extensions of @xmath283 and @xmath343 . theorem [ theoremq ] then implies a natural _ isomorphism _ @xmath350 . the homomorphism of theorem [ theoremq ] is given explicitely by `` flattening '' homology classes in the way sketched after theorem [ final step ] , and the isomorphism @xmath350 presumably has a similar explicit description using `` @xmath257flattenings , '' but we have not yet proved that these always exist ( note that an @xmath257flattening of a simplex presupposes a choice of a pair of opposite edges of the simplex ; changing this choice turns it into a @xmath351 or @xmath352flattening ) . for the same reason , we do not yet have a simplicial description of the class @xmath353 for a closed hyperbolic manifold @xmath48 , although this class exists for homological reasons . it is essential here that @xmath48 be closed the class @xmath339 almost certainly has no natural lift to @xmath349 in the non - compact case . the formula of @xcite for @xmath57 used in the programs snappea and snap uses ideal triangulations that arise in dehn surgery . these triangulations are not of the type mentioned after theorem [ final step ] , but by modifying them one can put them in the desired form and use theorem [ final step ] to compute @xmath341 , reconfirming the formula of @xcite . the formula computes @xmath57 up to a constant for the infinite class of manifolds that arise by dehn surgery on a given manifold . it was conjectured in @xcite that this constant is always a multiple of @xmath356 , and this too is confirmed . the theorem also gives an independent proof of the relation of volume and chern simons invariant conjectured in @xcite and proved in @xcite , from which a formula for eta - invariant was also deduced in @xcite and @xcite . one way to prove the bloch group rigidity conjecture [ bloch rigidity ] would be to show that @xmath107 is generated by the classes @xmath177 of 3manifolds . this question is presumably much harder than the rigidity conjecture , although modifications of it have been used in attempts on it . more specifically , one can ask for totally real number fields ( ie @xmath358 ) the answer is trivially `` yes '' while for number fields with @xmath359 the existence of arithmetic manifolds again shows the answer is `` yes . '' but beyond this little is known . in fact it is not even known if for every non - real number field @xmath360 a 3manifold exists with invariant trace field @xmath121 . ( for a few cases , eg multi - quadratic extensions of @xmath14 , the author and a reid have unpublished constructions to show the answer is `` yes . '' ) jun yang has pointed out that `` gromov norm '' gives an obstruction to a class @xmath361 being realizable as @xmath177 ( essentially the same observation also occurs in @xcite ) . we define the _ gromov norm _ @xmath362 as @xmath363,\quad z_i\in { { \mathbb c}}\bigr\},\ ] ] and it is essentially a result of gromov , with proof given in @xcite , that : in particular , since @xmath362 is invariant under the action of galois , for @xmath366 one sees that the @xmath368 component of the borel regulator is the largest in absolute value and equals @xmath369 . this suggests the question : this question is rather naive , and at this point we have no evidence for or against . another naive question is the following . for @xmath372 , where @xmath121 is a number field , we can define a stricter version of gromov norm by @xmath373,\quad z_i\in k\bigr\}.\ ] ] if not , then @xmath375 gives a sharper obstruction to realizing @xmath371 as @xmath177 since it is easy to show that for @xmath366 one has @xmath376 for some at most quadratic extension @xmath0 of @xmath121 . , _ hyperbolic @xmath380manifolds with equal volumes but different chern simons invariants _ , from : `` low - dimensional topology and kleinian groups '' , dba epstein ( editor ) , london math . soc lecture notes 112 ( 1986 ) 209215 , _ combinatorics of triangulations and the chern simons invariant for hyperbolic @xmath380manifolds _ , from : `` topology 90 , proceedings of the research semester in low dimensional topology at ohio state '' , walter de gruyter verlag , berlin new york ( 1992 ) 243272 , * aw reid * , _ arithmetic of hyperbolic manifolds _ , from : `` topology 90 , proceedings of the research semester in low dimensional topology at ohio state '' walter de gruyter verlag , berlin new york ( 1992 ) 273310	this paper is an expansion of my lecture for david epstein s birthday , which traced a logical progression from ideas of euclid on subdividing polygons to some recent research on invariants of hyperbolic 3manifolds . this `` logical progression '' makes a good story but distorts history a bit : the ultimate aims of the characters in the story were often far from 3manifold theory . we start in section 1 with an exposition of the current state of hilbert s 3rd problem on scissors congruence for dimension 3 . in section 2 we explain the relevance to 3manifold theory and use this to motivate the bloch group via a refined `` orientation sensitive '' version of scissors congruence . this is not the historical motivation for it , which was to study algebraic @xmath0theory of @xmath1 . some analogies involved in this `` orientation sensitive '' scissors congruence are not perfect and motivate a further refinement in section [ extended bloch ] . section [ more ] ties together various threads and discusses some questions and conjectures .
we have known since 1937 that the gravitational potentials of galaxy clusters are too deep to be caused by the detected baryonic mass and a newtonian @xmath3 gravitational force law @xcite . proposed solutions either invoke dominant quantities of non - luminous `` dark matter '' @xcite or alterations to either the gravitational force law @xcite or the particles dynamical response to it @xcite . previous works aimed at distinguishing between the dark matter and alternative gravity hypotheses in galaxies @xcite or galaxy clusters @xcite have used objects in which the visible baryonic and hypothesized dark matter are spatially coincident , as in most of the universe . these works favor the dark matter hypothesis , but their conclusions were necessarily based on non - trivial assumptions such as symmetry , the location of the center of mass of the system , and/or hydrostatic equilibrium , which left room for counterarguments . the actual existence of dark matter can only be confirmed either by a laboratory detection or , in an astronomical context , by the discovery of a system in which the observed baryons and the inferred dark matter are spatially segregated . an ongoing galaxy cluster merger is such a system . given sufficient time , galaxies ( whose stellar component makes up @xmath4 of the mass @xcite under the assumption of newtonian gravity ) , plasma ( @xmath5% of the mass @xcite ) , and any dark matter in a typical cluster acquire similar , centrally - symmetric spatial distributions tracing the common gravitational potential . however , during a merger of two clusters , galaxies behave as collisionless particles , while the fluid - like x - ray emitting intracluster plasma experiences ram pressure . therefore , in the course of a cluster collision , galaxies spatially decouple from the plasma . we clearly see this effect in the unique cluster 1e0657@xmath0558 @xcite . the cluster has two primary galaxy concentrations separated by 0.72 mpc on the sky , a less massive ( @xmath6 kev ) western subcluster and a more massive ( @xmath7 kev ) eastern main cluster @xcite . both concentrations have associated x - ray emitting plasma offset from the galaxies toward the center of the system . the x - ray image also shows a prominent bow shock on the western side of the western plasma cloud , indicating that the subcluster is currently moving away from the main cluster at @xmath84700 km / s . as the line - of - sight velocity difference between the components is only @xmath9 km / s @xcite , the merger must be occurring nearly in the plane of the sky and the cores passed through each other @xmath10 myr ago . two galaxy concentrations that correspond to the main cluster and the smaller subcluster have moved ahead of their respective plasma clouds that have been slowed by ram pressure . this phenomenon provides an excellent setup for our simple test . in the absence of dark matter , the gravitational potential will trace the dominant visible matter component , which is the x - ray plasma . if , on the other hand , the mass is indeed dominated by collisionless dark matter , the potential will trace the distribution of that component , which is expected to be spatially coincident with the collisionless galaxies . thus , by deriving a map of the gravitational potential , one can discriminate between these possibilities . we published an initial attempt at this using an archival vlt image @xcite ; here we add three additional optical image sets which allows us to increase the significance of the weak lensing results by more than a factor of 3 . in this paper , we measure distances at the redshift of the cluster , @xmath1 , by assuming an @xmath11 cosmology which results in @xmath12 kpc/@xmath13 plate - scale . none of the results of this paper are dependent on this assumption ; changing the assumed cosmology will result in a change of the distances and absolute masses measured , but the relative masses of the various structures in each measurement remain unchanged . llcccccc 2.2 m eso / mpg & @xmath14 & @xmath15 & r & 14100 & @xmath16 & 15 & @xmath17 + wide field imager & @xmath14 & & b & 6580 & & & @xmath18 + & @xmath14 & & v & 5640 & & & @xmath19 + 6.5 m magellan & @xmath20 & @xmath21 radius & r & 10800 & @xmath22 & 35 & @xmath23 + imacs & @xmath20 & & b & 2700 & & & @xmath19 + & @xmath20 & & v & 2400 & & & @xmath17 + hst acs & @xmath24 & 35@xmath2535 & f814w & 4944 & @xmath26 & 87 & @xmath27 + subcluster & @xmath24 & & f435w & 2420 & & & @xmath27 + & @xmath24 & & f606w & 2336 & & & @xmath27 + main cluster & @xmath24 & 35@xmath2535 & f606w & 2336 & @xmath28 & 54 & @xmath27 + we construct a map of the gravitational potential using weak gravitational lensing @xcite , which measures the distortions of images of background galaxies caused by the gravitational deflection of light by the cluster s mass . this deflection stretches the image of the galaxy preferentially in the direction perpendicular to that of the cluster s center of mass . the imparted ellipticity is typically comparable to or smaller than that intrinsic to the galaxy , and thus the distortion is only measurable statistically with large numbers of background galaxies . to do this measurement , we detect faint galaxies on deep optical images and calculate an ellipticity from the second moment of their surface brightness distribution , correcting the ellipticity for smearing by the point spread function ( corrections for both anisotropies and smearing are obtained using an implementation of the ksb technique @xcite discussed in @xcite ) . the corrected ellipticities are a direct , but noisy , measurement of the reduced shear @xmath29 . the shear @xmath30 is the amount of anisotropic stretching of the galaxy image . the convergence @xmath31 is the shape - independent increase in the size of the galaxy image . in newtonian gravity , @xmath31 is equal to the surface mass density of the lens divided by a scaling constant . in non - standard gravity models , @xmath31 is no longer linearly related to the surface density but is instead a non - local function that scales as the mass raised to a power less than one for a planar lens , reaching the limit of one half for constant acceleration @xcite . while one can no longer directly obtain a map of the surface mass density using the distribution of @xmath31 in non - standard gravity models , the locations of the @xmath31 peaks , after adjusting for the extended wings , correspond to the locations of the surface mass density peaks . our goal is thus to obtain a map of @xmath31 . one can combine derivatives of @xmath32 to obtain @xcite @xmath33 which is integrated over the data field and converted into a two - dimensional map of @xmath31 . the observationally unconstrained constant of integration , typically referred to as the `` mass - sheet degeneracy , '' is effectively the true mean of @xmath34 at the edge of the reconstruction . this method does , however , systematically underestimate @xmath31 in the cores of massive clusters . this results in a slight increase to the centroiding errors of the peaks , and our measurements of @xmath31 in the peaks of the components are only lower bounds . for 1e0657@xmath0558 , we have accumulated an exceptionally rich optical dataset , which we will use here to measure @xmath32 . it consists of the four sets of optical images shown in table [ table1 ] and the vlt image set used in @xcite ; the additional images significantly increase the maximum resolution obtainable in the @xmath31 reconstructions due to the increased number of background galaxies , particularly in the area covered by the acs images , with which we measure the reduced shear . we reduce each image set independently and create galaxy catalogs with 3 passband photometry . the one exception is the single passband hst pointing of main cluster , for which we measure colors from the magellan images . because it is not feasible to measure redshifts for all galaxies in the field , we select likely background galaxies using magnitude and color cuts ( @xmath35 and not in the rhombus defined by @xmath36 , @xmath37 , and @xmath38 for the acs images ; similar for the other image sets ) that were calibrated with photometric redshifts from the hdf - s @xcite . each galaxy has a statistical weight based on its significance of detection in the image set @xcite , and the weights are normalized among catalogs by comparing the rms reduced shear measured in a region away from the cores of the cluster common to all five data sets . to combine the catalogs , we adopt a weighted average of the reduced shear measurements and appropriately increase the statistical weight of galaxies that occur in more than one catalog . we use the combined catalog to create a two - dimensional @xmath31 reconstruction , the central portion of which is shown in fig . two major peaks are clearly visible in the reconstruction , one centered @xmath39 east and @xmath40 north of the subcluster s brightest cluster galaxy ( bcg ) and detected at @xmath2 significance ( as compared to @xmath41 in @xcite ) , and one centered @xmath42 east and @xmath43 south of the northern bcg in the main cluster ( @xmath44 west and @xmath45 north of the southern bcg ) detected at @xmath46 . we estimate centroid uncertainties by repeating bootstrap samplings of the background galaxy catalog , performing a @xmath31 reconstruction with the resampled catalogs , and measuring the centroid of each peak . both peaks are offset from their respective bcg by @xmath47 , but are within @xmath48 of the luminosity centroid of the respective component s galaxies ( both bcgs are slightly offset from the center of galaxy concentrations ) . both peaks are also offset at @xmath49 from the center of mass of their respective plasma clouds . they are skewed toward the plasma clouds , which is expected because the plasma contributes about 1/10th of the total cluster mass @xcite ( and a higher fraction in non - standard gravity models without dark matter ) . the skew in each @xmath31 peak toward the x - ray plasma is significant even after correcting for the overlapping wings of the other peak , and the degree of skewness is consistent with the x - ray plasma contributing @xmath50 of the observed @xmath31 in the main cluster and @xmath51 in the subcluster ( see table [ table2 ] ) . because of the large size of the reconstruction ( @xmath52 or 9 mpc on a side ) , the change in @xmath31 due to the mass - sheet degeneracy should be less than @xmath53 and any systematic effects on the centroid and skewness of the peaks are much smaller than the measured error bars . lccccc main cluster bcg & @xmath54 & @xmath55 & @xmath56 & @xmath57 & @xmath58 + main cluster plasma & @xmath59 & @xmath60 & @xmath61 & @xmath62 & @xmath63 + subcluster bcg & @xmath64 & @xmath65 & @xmath66 & @xmath67 & @xmath68 + subcluster plasma & @xmath69 & @xmath70 & @xmath71 & @xmath72 & @xmath73 + the projected cluster galaxy stellar mass and plasma mass within 100 kpc apertures centered on the bcgs and x - ray plasma peaks are shown in table [ table2 ] . this aperture size was chosen as smaller apertures had significantly higher kappa measurement errors and larger apertures resulted in significant overlap of the apertures . plasma masses were computed from a multicomponent 3-dimensional cluster model fit to the chandra x - ray image ( details of this fit will be given elsewhere ) . the emission in the chandra energy band ( mostly optically - thin thermal bremsstrahlung ) is proportional to the square of the plasma density , with a small correction for the plasma temperature ( also measured from the x - ray spectra ) , which gives the plasma mass . because of the simplicity of this cluster s geometry , especially at the location of the subcluster , this mass estimate is quite robust ( to a 10% accuracy ) . stellar masses are calculated from the @xmath74-band luminosity of all galaxies equal in brightness or fainter than the component bcg . the luminosities were converted into mass assuming @xcite @xmath75 . the assumed mass - to - light ratio is highly uncertain ( can vary between 0.5 and 3 ) and depends on the history of recent star formation of the galaxies in the apertures ; however even in the case of an extreme deviation , the x - ray plasma is still the dominant baryonic component in all of the apertures . the quoted errors are only the errors on measuring the luminosity and do not include the uncertainty in the assumed mass - to - light ratio . because we did not apply a color selection to the galaxies , these measurements are an upper limit on the stellar mass as they include contributions from galaxies not affiliated with the cluster . the mean @xmath31 at each bcg was calculated by fitting a two peak model , each peak circularly symmetric , to the reconstruction and subtracting the contribution of the other peak at that distance . the mean @xmath31 for each plasma cloud is the excess @xmath31 after subtracting off the values for both peaks . the total of the two visible mass components of the subcluster is greater by a factor of 2 at the plasma peak than at the bcg ; however , the center of the lensing mass is located near the bcg . the difference of the baryonic mass between these two positions would be even greater if we excluded a contribution of the non - peaked plasma component between the shock front and the subcluster . for the main cluster , we see the same effect , although the baryonic mass difference is smaller . note that both the plasma mass and the stellar mass are determined directly from the x - ray and optical images , respectively , independently of any gravity or dark matter models . a key limitation of the gravitational lensing methodology is that it only produces a two - dimensional map of @xmath31 , and hence raises the possibility that structures seen in the map are caused by physically unrelated masses along the line - of - sight . because the background galaxies reside at a mean @xmath76 , structures capable of providing a significance amount of @xmath31 must lie at @xmath77 . by comparing the measured shear for galaxies divided into crude redshift bins using photometric redshifts @xcite , we further limit the redshift of the lensing objects to @xmath78 . this range is consistent with the cluster redshift , but corresponds to a large volume in which a structure unassociated with the cluster could exist and be projected onto the lensing map . however , the number density of structures with these lensing strengths in blank field surveys is @xmath79 arcmin@xmath80 for the subcluster @xcite , and an order of magnitude less for the main cluster , resulting in a @xmath81 probability of having two structures within a square arcminute of the observed cluster cores . further , all such lenses observed in these cosmic shear surveys are clusters with enough plasma and galaxies to be easily observable . there is no evidence , however , in our deep imaging for additional cluster sized concentrations of galaxies or of plasma hotter than @xmath82 kev ( the lower bound of the chandra energy band ) near the observed lensing structures . another alternate explanation of the lensing signal is related to the fact that clusters form at the intersections of matter filaments @xcite . in principle , one could imagine two line - of - sight filaments of intergalactic gas ( too cool to be visible with chandra and too diffuse to have cooled into stars ) extending from the cluster at the locations of the weak lensing peaks . to explain the measured surface mass density , such filaments would have to be several megaparsecs long , very narrow , and oriented exactly along the line of sight . the probability of such an orientation for two such filaments in the field is @xmath83 . further , because the two cluster components are moving at a relative transverse velocity of 4700 km / s compared to the typical peculiar velocities in the cmb frame of a few hundred km / s , the filaments could coincide so exactly with each of the bcgs only by chance . this is an additional factor of @xmath84 reduction in probability . while such projections become more important in non - standard gravity models because in such models the thin lens approximation breaks down @xcite and structures with a given surface density produce a greater amount of lensing the more they are extended along the line - of - sight , two such projections would still have a @xmath85 probability . finally , we mention that two other merging clusters , ms1054@xmath003 @xcite and a520 ( in preparation ) , exhibit similar offsets between the peaks of the lensing and baryonic mass , although based on lensing reconstructions with lower spatial resolution and less clear - cut cluster geometry . a final possibility is that some alternative gravity models may be able to suppress the lensing potential of the central peak in a multiple - peak system , as in @xcite . that work used a model of a gas disk located between two symmetric mass concentrations representing the galaxy subclusters . in their @xmath31 map , derived in the teves framework @xcite , the relative signal from this disk may be suppressed , but would still be easily visible with the noise levels of our reconstruction . our @xmath31 map , however , has no evidence of any mass concentration between the two galaxy subclusters other than the small perturbations consistent with the gas mass contribution in newtonian gravity . furthermore , such a suppression has also only been shown to work for symmetric systems which have the central peak directly between the two outer peaks . in 1e0657@xmath0558 , however , the x - ray plasma , which would provide the central peak , lies north of the line connecting the two @xmath31 peaks . further , the absolute @xmath31 levels of the peaks observed in 1e0657@xmath0558 are in good agreement with those in systems with similar velocity dispersions and x - ray temperatures ( e.g. * ? ? ? * ) which have the gas and the galaxies coincident . the @xmath31-to - light ratios are also consistent with those in normal clusters with coincident gas and galaxies . therefore one would need to not only suppress the inner peak in the @xmath31 map relative to the two outer peaks in this system , but also enhance the strength of the outer peaks to make up for the missing plasma mass . any non - standard gravitational force that scales with baryonic mass will fail to reproduce these observations . the lensing peaks require unseen matter concentrations that are more massive than and offset from the plasma . while the existence of dark matter removes the primary motivation for alternative gravity models , it does not preclude non - standard gravity . the scaling relation between @xmath31 and surface mass density , however , has important consequences for models that mix dark matter with non - newtonian gravity : to achieve the @xmath86 ratio in @xmath31 between the dark matter + galaxy component and the plasma component ( table [ table2 ] ) , the true ratio of mass would be even higher ( as high as 49:1 for a constant acceleration model ; although mond @xcite would not reach this ratio as the dark matter density would become high enough to shift the acceleration into the quasi - newtonian regime ) , making the need for dark matter even more acute . such high concentrations of dark matter , however , are extremely unlikely based on the measured x - ray plasma temperatures @xcite and cluster galaxy velocity dispersions @xcite . the spatial separation of the dominant baryonic component in a galaxy cluster from the hypothesized dark matter produced during a cluster merger has enabled us to directly compare the dark matter hypothesis to one with only visible matter but a modified law of gravity . the observed displacement between the bulk of the baryons and the gravitational potential proves the presence of dark matter for the most general assumptions regarding the behavior of gravity . we wish to thank bhuvnesh jain , hongsheng zhao , and peter schneider for useful discussions . this work was supported by nasa through grant number ltsa04 - 0000 - 0041 ( dc and dz ) . support for program 10200 was provided by nasa through grant go-10200.01 from the space telescope science institute , which is operated by the association of universities for research in astronomy , inc . , under nasa contract nas 5 - 26555 . mb acknowledges support from the nsf grant ast-0206286 . mm acknowledges support from chandra grant go4 - 5152x .	we present new weak lensing observations of 1e0657@xmath0558 ( @xmath1 ) , a unique cluster merger , that enable a direct detection of dark matter , independent of assumptions regarding the nature of the gravitational force law . due to the collision of two clusters , the dissipationless stellar component and the fluid - like x - ray emitting plasma are spatially segregated . by using both wide - field ground based images and hst / acs images of the cluster cores , we create gravitational lensing maps which show that the gravitational potential does not trace the plasma distribution , the dominant baryonic mass component , but rather approximately traces the distribution of galaxies . an @xmath2 significance spatial offset of the center of the total mass from the center of the baryonic mass peaks can not be explained with an alteration of the gravitational force law , and thus proves that the majority of the matter in the system is unseen .
currently a large variety of nucleon - nucleon ( nn ) potentials exist . despite the fact that all these potentials were designed to more or less reproduce the deuteron properties and the low - energy nn scattering data , they are quite different in details . in particular , the deuteron @xmath3-state percentage given by different potentials ranges from 4.4% for bonn a @xcite to 7.0% for hamada - johnson @xcite . exact ( or nearly exact ) calculations for the @xmath4=3 and @xmath4=4 systems @xcite show that there is a strong correlation between the tensor strength of the one - boson - exchange nn potential and the binding energy . the nn potential which has a weaker tensor strength tends to yield a larger binding energy . in ref.@xcite , however , it is found that when the nn potentials with different parametrizations are optimally fitted ( with @xmath5 per datum about 1 ) to the phase - shifts obtained in the `` new and comprehensive ( energy - dependent ) partial - wave analysis '' , they result in nearly identical triton properties . it is pointed out that the tensor force is tightly constrained by the mixing parameter @xmath6 , which can be determined very well in their multi - energy analysis . for example , three local nn potentials , nijmii , reid93 and av18 with parameters newly determined by their fit , have deuteron @xmath3-state percentages of 5.64% , 5.70% and 5.65% respectively . it is the main purpose of this work to find out whether or not the above observation made in the study of the @xmath4=3 system also holds for larger nuclei . to this end , we will perform nuclear shell - model ( sm ) calculations for @xmath0 and @xmath1 using effective interactions derived from the reid93 and nijmii potentials . to set a scale for the comparison , the same calculations will be performed using the hamada - johnson hard - core potential ( h - j ) @xcite and the original reid - soft - core potential ( rsc ) @xcite , both of which were widely used in nuclear - structure calculations in the past . we will adopt a no - core model space @xcite for which the effective interaction can reasonably be approximated by the brueckner g - matrix @xcite evaluated at appropriate starting energies . this avoids the difficulty encountered in doing the infinite summation over the core - polarization diagrams for the effective interaction in the presence of an inert core . of course , for each nucleus , the effective interactions will be calculated in exactly the same manner so the difference reflected in the nuclear sm results is solely due to the difference in the nn potentials . the brueckner g - matrix @xcite is calculated according to the following equation using the technique developed in ref.@xcite : @xmath7 where @xmath8 is the starting energy , @xmath9 is the nn potential , @xmath10 is the single - particle hamiltonian with the one - body mean field @xmath11 approximated by a shifted harmonic - oscillator ( ho ) potential @xmath12 : @xmath13 the parameter @xmath14 approximately represents the depth of the mean field @xmath11 in which the two nucleons in the ladder - scattering processes are moving . this parameter can be absorbed into the starting energy by writing the g - matrix equation ( [ g ] ) as @xmath15 where @xmath16 and @xmath17 . for the two - body g - matrix element @xmath18 , the shifted starting energy @xmath19 is taken as @xmath20 where @xmath21 are the ho single - particle energies : @xmath22 roughly speaking , the quantity @xmath23 represents the interaction energy of the two nucleons bound by the ho potential . it is negative and depends on the two - nucleon state . in this work , @xmath23 is regarded as an adjustable parameter whose value will be so chosen that we obtain a reasonable binding energy . it should be emphasized that @xmath23 will only be adjusted when we go from one nucleus to another ; its value is fixed when we consider the same nucleus but different nn potentials . we chose @xmath24 mev , @xmath25 mev for @xmath4=6 and @xmath26 mev , @xmath27 mev for @xmath4=14 . in the no - core model space , the starting - energy independent two - body effective interaction is simply the sum of ladder diagrams ( g - matrix ) plus the folded diagrams . when the starting energy for the g - matrix is properly chosen , the folded diagrams have a small contribution to the effective interaction , which can , therefore , be well approximated by the g - matrix alone . since we will be performing a @xmath2 calculation , the pauli operator @xmath28 is defined as ( @xmath29 and @xmath30 are the principal quantum numbers for the two single - particle orbitals of the intermediate states in the ladder diagrams ) : @xmath31 namely , in the ladder diagrams , the scattering into a two - particle state with @xmath32 is forbidden because these states will be included in the @xmath2 calculation . in addition , for @xmath0 , the scattering into a two - particle state with one particle in the @xmath33 orbit is also forbidden because the @xmath33 orbit is mostly occupied . for @xmath1 , the scattering into a two - particle state with one particle in the @xmath33 or @xmath34 orbits is not allowed for the same reason . this is also shown in fig.1 . notice that in fig.1 we have cut off the `` wings '' which extend out from the mostly occupied orbitals , at a large enough @xmath35 , so as to obtain a converged result . it should be pointed out that our treatment of the pauli operator @xmath28 is not exact ; a more accurate definition is possible . such a definition should be reasonable enough to allow us to do sensible calculations and meaningful comparisons . here a more important issue is that the effective interactions for different potentials are calculated in exactly the same way . once the g matrices are obtained , the hamiltonian for the sm calculation is written as : @xmath36 where @xmath37 is the center - of - mass ( cm ) kinetic energy and @xmath38 is the ho cm hamiltonian . the parameter @xmath39 is taken to be large enough ( @xmath40 ) to force the cm motion into its lowest configuration for all the low - lying states . our calculations are performed using the oxbash sm code @xcite . the results for the ground - state energy , the low - lying positive - parity energy spectrum , and selected electromagnetic ( em ) static and transitional properties in @xmath0 and @xmath1 are shown in table i for the four nn potentials used in this work . the deuteron @xmath3-state percentage @xmath41 for each potential is also listed . before we proceed to discuss the results , it should be emphasized that our calculations involve some parameters , such as @xmath23 and @xmath42 . the calculated ground - state energy , which is quite sensitive to the choice of these parameters , should not be taken as an accurate result . nevertheless , since we perform exactly the same calculation for different potentials , the results are valid for the purpose of comparisons . in ref.@xcite , the reid93 and nijmii potentials have been shown to give almost the same results for the triton by solving the faddeev equation @xcite . it is evident from table i that the effective interactions calculated from these potentials also produce remarkably similar results for the more complex @xmath4=6 and 14 systems . this is more obvious when we compare the results for these two potentials with those for the h - j and rsc potentials . note that the similarity between the reid93 and nijmii results lies not only in the bulk properties , such as the binding energy , but also in more subtle observables , like excitation energies and em transition rates . the latter reflects the equivalence in the wave functions for these two interactions . it is then safe to conclude that the reid93 and nijmii potentials are not only equivalent in describing the @xmath4=2 and 3 systems as shown in ref.@xcite , they are also equivalent for the larger systems @xmath0 and @xmath1 studied in this work . for the @xmath4=3 and @xmath4=4 systems , it has been found that there is a strong correlation between the nuclear binding energy and the strength of the tensor component in the nn potential , characterized by the deuteron @xmath3-state percentage @xmath41 @xcite apparently , the above correlation persists for the more complex nuclei @xmath0 and @xmath1 in a truncated - space sm calculation , as performed here . the calculated binding energies for the hj and rsc potentials , which have larger deuteron @xmath3-state percentages , are much smaller than those for the reid93 and nijmii potentials . note , however , that not all nuclear - structure properties are sensitive to the details of the nn potential . by examining table i , one can see that the _ excitation _ energies for the low - lying states and some em observables for different nn potentials are actually quite similar . this makes it difficult to test the nn potential in a nucleus by studying nuclear structure . on the other hand , besides the binding energy , there do exist a few observables which have a strong dependence on one or more particular components of the nn potential . two famous examples are the isospin splitting between the energies of the @xmath43 , @xmath44=0 and @xmath43 , @xmath44=1 states in @xmath45 and the gamow - teller strength @xmath46 between the ground state in @xmath47 and @xmath1 . the former is sensitive to the tensor component in the nn potential @xcite and the latter is sensitive to the spin - orbit and tensor components @xcite . these have been studied in great detail in @xcite . in table i , one does see a sizable change in the @xmath46 value from the ground state of @xmath47 , when going from one potential to the other . the @xmath48 value is closely related to the @xmath46 value and also shows some change . however , both the @xmath46 and @xmath48 values are not a monotonic function of @xmath41 , suggesting that besides the tensor force strength , other differences between the four nn potentials also play a role . from table i , we see that the calculated quadrupole moment for the ground state of @xmath0 also experiences quite significant changes when the nn potential is altered . in our truncated - space sm calculation with bare operators [ i.e. , @xmath49 and @xmath50 , none of the nn potentials yields the experimental result . the smallness of the experimental quadrupole moment for @xmath0 and of the @xmath46 value from @xmath47 seems to indicate that in a truncated sm calculation , one needs a weaker tensor force . this is consistent with the observations made in ref.@xcite , where some other quantities which are sensitive to the tensor force were considered . in this work , we perform a no - core space @xmath2 sm calculation for the @xmath4=6 and 14 systems using effective interactions calculated from four different nn potentials . the reid93 and nijmii potentials , which were found to be equivalent in the description of the @xmath4=2 and @xmath4=3 systems , are shown to yield almost identical results , not only for the binding energy , but also for the energy spectrum and the em static and transitional properties for the larger nuclei considered in this work . it is verified that the nn potential with a weaker tensor component ( i.e. , smaller deuteron @xmath3-state percentage ) tends to give a larger nuclear binding energy , consistent with the observations made in the study of the smaller @xmath4=3 and 4 systems . many nuclear - structure properties are found to be rather insensitive to the details of a nn potential and similar results are obtained for them , when different nn potentials are used . among the various observables considered , there are a few which are sensitive to the various components of a nn potential , in particular , the tensor force . in order to reproduce the experimental results for the gt transition strength from the ground state of @xmath47 and the quadrupole moment of @xmath0 in a truncated space sm calculation , a weaker tensor force is needed . we thank j. j. de swart and m. rentmeester for providing us the nn potentials reid93 and nijmii . this work was supported by the national science foundation , grant no . phy-9103011 . 99 r. machleidt , adv . in nucl . 19 , 189(1989 ) . t. hamada and i.d . johnson , nucl . phys . * 34 * , 382(1962 ) . j.l . friar , g.l . payne , v.g.j . stoks , and j.j . de swart , phys . lett . * b311 * , 4(1993 ) . w. glckle and h. kamada , phys . lett . * 71 * , 971 ( 1993 ) . reid , ann . of phys . * 50 * , 411 ( 1968 ) . j.p . vary , in _ theory and applications of moment methods in many - fermion systems _ , eds b.j . dalton , s.m . grimes , j.p . vary , and s.a . williams , plenum press ( new york ) , 1980 , p.423 . brueckner , phys . rev . * 97 * , 1353(1955 ) ; * 100 * , 36(1955 ) . b.r . barrett , r.g.l . hewitt , and r.j . mccarthy , phys . rev . * c 3 * , 1137(1971 ) . a. etchegoyen , w.d.m . rae , n.s . godwin , b.a . brown , w.e . ormand , and j.s . winfield , the oxford buenos aires msu shell model code ( oxbash ) ( unpublished ) . faddeev , zh . * 39 * , 1459(1960 ) . j. blomqvist and a. molinari , nucl . phys . * a106 * , 545(1968 ) ; d.j . millener and d. kurath , nucl . phys . * a255 * , 315(1975 ) ; b.r . barrett , phys . rev . * 159 * , 816(1967 ) . wong , nucl . a108 * , 481(1968 ) ; r.r . scheerbaum , phys . lett . * 63b * , 381(1976 ) . d.c . zheng and l. zamick , ann . ( ny ) * 206 * , 106(1991 ) ; l. zamick , d.c . zheng , and h. mther , phys . * c 45 * , 2763(1992 ) ; l. zamick and d.c . zheng , phys . , in press . * table i. * the ground - state energies and excitation energies of low - lying states , in units of mev , of @xmath0 and @xmath1 in a @xmath2 calculation . the g - matrices are calculated in exactly the same way from different nn potentials including hamada - johnson ( h - j ) , reid - soft - core ( rsc ) , new reid - soft - core ( reid93 ) , and new nijmegen ( nijmii ) . for @xmath0 ( @xmath1 ) , a harmonic - oscillator basis with @xmath24 mev ( @xmath26 mev ) is used . the gt strength @xmath46 and m1 strength @xmath48 ( in @xmath51 ) is calculated for the transition @xmath52 . the magnetic dipole moment @xmath53 ( in @xmath54 ) and electric quadrupole moment @xmath28 ( in @xmath55 ) are calculated for the @xmath56 state . bare operators [ @xmath57 , @xmath58 , @xmath59 , @xmath60 are used in these calculations . in the table , @xmath41 is the deuteron @xmath3-state percentage . [ cols="^,^,>,>,>,>,^",options="header " , ] * fig.1 * the pauli operator @xmath28 for @xmath4=6 and @xmath4=14 used in this work .	two `` phase - shift equivalent '' local nn potentials with different parametrizations , reid93 and nijmii , which were found to give nearly identical results for the triton by friar _ et al . _ , are shown to yield remarkably similar results for @xmath0 and @xmath1 in a @xmath2 no - core space shell - model calculation . the results are compared with those for the widely used hamada - johnson hard - core and the original reid soft - core potentials , which have larger deuteron @xmath3-state percentages . the strong correlation between the tensor strength and the nuclear binding energy is confirmed . however , many nuclear - structure properties seem to be rather insensitive to the details of the nn potential and , therefore , can not be used to test various nn potentials . d. c. zheng and b. r. barrett + _ department of physics , university of arizona , tucson , az 85721 _
we begin by defining the infinite finitary symmetric and alternating groups and their corresponding wreath products , and then we state our results regarding growth series identities . for an infinite set @xmath2 , the _ finitary symmetric group _ @xmath0 is the group of permutations of @xmath2 with finite support . the _ permutational wreath product _ of a group @xmath3 with @xmath0 is the group @xmath4 defined as follows : a. the group @xmath5 is the group of functions from @xmath2 to @xmath3 with finite support . b. the action of permutations @xmath6 on functions @xmath7 is defined by @xmath8 c. multiplication in the semi - direct product is defined for @xmath9 and @xmath10 by @xmath11 the _ finitary alternating group _ @xmath1 is the subgroup of @xmath0 of permutations with even signature , and the permutational wreath product @xmath12 is defined as above . in this paper , we only consider permutational wreath products with finite group @xmath3 . we now define some general terminology . for any group @xmath13 generated by a set @xmath14 , the _ word length _ @xmath15 of any element @xmath16 is the smallest nonnegative integer @xmath17 such that there exist @xmath18 with @xmath19 . the _ conjugacy length _ @xmath20 is the smallest word length appearing in the conjugacy class of @xmath21 . if @xmath17 is any natural number , we denote by @xmath22 the number of conjugacy classes in @xmath13 with smallest word length @xmath17 . if @xmath23 is finite for all @xmath17 , then we may define the conjugacy growth series of a group @xmath13 with generating set @xmath14 to be the following @xmath24-series : @xmath25\in\text{conj}(g)}q^{\kappa_{g , s}(g)}=\sum\limits_{n=0}^\infty\g_{g , s}(n)q^n,\ ] ] where the first sum is over representatives of conjugacy classes of @xmath13 . bacher and de la harpe @xcite prove conjugacy growth series identities for sufficiently large . ] generating sets @xmath14 of @xmath0 , @xmath26 of @xmath1 , and @xmath27 of @xmath28 relating the finitary permutation groups and their wreath products to the partition function . explicitly , we have the fascinating identities @xmath29 for the finitary symmetric group , @xmath30 for the finitary alternating group denotes the number of partitions of @xmath31 into an even number of parts . ] , and @xmath32 for wreath products @xmath28 , where @xmath33 is the number of conjugacy classes of @xmath34 . from now on , we denote @xmath35 for convenience . recall dedekind s eta function @xmath36 for @xmath37 , where @xmath38 denotes the upper half complex plane and @xmath39 . equation ( [ cgsalt ] ) can be written as the linear combination of eta - quotients @xmath40 which is essentially the sum of a modular form of weight @xmath41 and a modular form of weight @xmath42 , up to multiplication by @xmath43 . studying such linear combinations may shed light on properties of sums of mixed weight modular forms . it is natural to consider the number @xmath44 as a function of the number of conjugacy classes @xmath33 in order to study properties of the coefficients of the above @xmath24-series . here we obtain a universal recurrence for these numbers . this result requires the ordinary divisor function @xmath45 . we also must define , for @xmath46 , the polynomial @xmath47 the polynomials @xmath48 are fairly straightforward to compute using only the partitions of @xmath17 ; the first few are listed below . these polynomials have been used in earlier work @xcite on divisors of modular forms and the rogers - ramanujan identities . [ thm ] let @xmath50 be defined as above . let @xmath34 be a finite group with @xmath33 conjugacy classes , @xmath2 an infinite set , and @xmath51 a wreath product generated by a sufficiently large set @xmath27 . then we have @xmath52 where @xmath44 satisfies the recurrence relation @xmath53 in recent work , nekrasov and okounkov obtained a different formula for the infinite products in theorem [ thm ] in terms of hook lengths of partitions . let @xmath54 denote that @xmath55 is a partition of the number @xmath56 . the hook length of a partition @xmath57 is defined using the ferrers diagram of @xmath55 . for example , figure 1 below is a ferrers diagram of the partition @xmath58 , figure 2 represents a hook length of 4 , and figure 3 shows all hook lengths associated to @xmath55 . + + + + + + + + + + + + more generally , for each box @xmath59 in the ferrers diagram of a partition @xmath55 , its _ hook length _ @xmath60 is defined as the number of boxes @xmath61 such that a. @xmath62 , b. @xmath61 is in the same column as @xmath59 and below @xmath59 , or c. @xmath61 is in the same row as @xmath59 and to the right of @xmath59 . the _ hook length multi - set _ @xmath63 is the set of all hook lengths of @xmath55 . theorem [ thm ] implies the following formula for @xmath44 in terms of hook lengths . [ no ] we have that @xmath64 kostant observed @xcite that the coefficients of the nekrasov - okounkov hook length identity are polynomials in the variable @xmath65 , but he did not give an explicit formula for computing them . following bacher s and de la harpe s proofs of equations ( [ cgs ] ) , ( [ cgsalt ] ) , and ( [ cgswr ] ) , we prove ] the corresponding growth series identity for the wreath product @xmath66 with sufficiently large generating set @xmath67 , namely @xmath68 where @xmath69 is the number of conjugacy classes of @xmath70 . we provide the proof of equation ( [ cgswralt ] ) in section [ 2 ] . from now on , we denote @xmath71 for convenience . + in analogy with theorem [ thm ] , one may ask if the coefficients @xmath72 in the alternating case can be seen as a function of the number of conjugacy classes @xmath69 . we obtain a similar recurrence relation in this case . [ thm2 ] let @xmath50 be defined as above . let @xmath70 be a finite group with @xmath69 conjugacy classes , @xmath2 an infinite set , and @xmath73 a wreath product generated by a sufficiently large set @xmath67 . then we have @xmath74 where @xmath72 satisfies the recurrence relation @xmath75\bigg),\ ] ] and the @xmath76 are defined by their generating function @xmath77 it may be possible to interpret the coefficients @xmath72 in terms of hook lengths from formulas of han @xcite or others , as in the symmetric case . the author does not make this connection here . it is also natural to study the _ _ modified exponential rate of conjugacy growth _ _ to @xmath78 . with denominator @xmath17 , most of the growth series that we study have exponential rate of conjugacy growth zero . ] of a group @xmath13 generated by a set @xmath14 , namely @xmath79 it is useful to notice that @xmath80 is the radius of convergence of the conjugacy growth series @xmath81 . for permutational wreath products , we apply a theorem of cotron , dicks and fleming @xcite on the asymptotic behavior of the generalized partition function defined in section [ 2 ] ( see equations ( [ gpf ] ) and ( [ asymp ] ) ) . let @xmath28 be a wreath product where @xmath34 is a finite group , @xmath33 is the number of conjugacy classes of @xmath34 , and @xmath2 is an infinite set . it is easy to see from equation ( [ cgswr ] ) that the conjugacy growth series of such a wreath product is the generating function of the generalized partition function @xmath82 for the vector @xmath83 . this implies the following corollary . [ asymptotic ] let @xmath28 be a wreath product where @xmath34 is a finite group , @xmath33 is the number of conjugacy classes of @xmath34 , and @xmath2 is an infinite set . if @xmath27 is a sufficiently large generating set of @xmath84 , then we have @xmath85 we now give the modified exponential rate of conjugacy growth for wreath products in the symmetric case using this asymptotic formula . the modified exponential rate of conjugacy growth for the group @xmath28 defined above is @xmath86 we can also apply the theorem to wreath products in the alternating case using equation ( [ cgswralt ] ) . [ altasymptotic ] let @xmath66 be a wreath product where @xmath70 is a finite group , @xmath69 is the number of conjugacy classes of @xmath70 , and @xmath2 is an infinite set . if @xmath67 is a sufficiently large generating set of @xmath87 , then we have @xmath88 we also give the modified exponential rate of conjugacy growth in the alternating case using the above asymptotic formula . the modified exponential rate of conjugacy growth for the group @xmath66 defined above is @xmath89 we are interested in finding relationships between wreath products of @xmath0 and wreath products of @xmath1 . let @xmath28 and @xmath90 be two wreath products of @xmath0 , where @xmath91 are finite groups with @xmath92 conjugacy classes respectively . let @xmath66 and @xmath93 be two wreath products of @xmath1 , where @xmath94 are finite groups with @xmath95 conjugacy classes respectively . [ ratioqn ] what is the asymptotic behavior of the following ratios ? @xmath96 in particular , when do the ratios approach some nonzero finite number ? the asymptotic behavior of the ratios follows from corollaries [ symasymptotic ] and [ altasymptotic ] . [ ratio ] let @xmath97 and @xmath98 be groups as above . then as @xmath99 , we have @xmath100e^{\pi\sqrt{\frac{2n}{3}}\left(\sqrt{m_s}-\sqrt{m_s'}\right)}.\\\\ ( 2)\hspace{.3cm}\frac{\g_{w_s}(n)}{\g_{w_a}(n)}&\sim&\left(\frac{m_s^{\frac{1+m_s}{4}}}{m_a^{\frac{1 + 2m_a}{4}}}\right)\left[2^{\frac{8m_a-3m_s-1}{4}}(3n)^{\frac{2m_a - m_s}{4}}\right]e^{\pi\sqrt{\frac{2n}{3}}\left(\sqrt{m_s}-\sqrt{2m_a}\right)}.\\\\ ( 3)\hspace{.3cm}\frac{\g_{w_a}(n)}{\g_{w_s}(n)}&\sim&\left(\frac{m_a^{\frac{1 + 2m_a}{4}}}{m_s^{\frac{1+m_s}{4}}}\right)\left[2^{\frac{1 + 3m_s-8m_a}{4}}(3n)^{\frac{m_s-2m_a}{4}=}\right]e^{\pi\sqrt{\frac{2n}{3}}\left(\sqrt{2m_a}-\sqrt{m_s}\right)}.\\\\ ( 4)\hspace{.3cm}\frac{\g_{w_a}(n)}{\g_{w_a'}(n)}&\sim&\left(\frac{m_a^{\frac{1 + 2m_a}{4}}}{m_a'^{\frac{1 + 2m_a'}{4}}}\right)\left[4^{\left(m_a'-m_a\right)}(3n)^{\frac{m_a'-m_a}{2}}\right]e^{2\pi\sqrt{\frac{n}{3}}\left(\sqrt{m_a}-\sqrt{m_a'}\right)}.\end{aligned}\ ] ] we now observe for which pairs @xmath101 and @xmath102 these ratios asymptotically approach zero , infinity , or some nonzero finite number . corollary [ l ] follows from the asymptotic behavior of the exponential functions in corollary [ ratio ] . [ l ] let @xmath97 and @xmath98 be groups as above . then as @xmath99 , we have the following asymptotic behavior . 1 . if @xmath103 , then @xmath104 . if @xmath105 , then @xmath106 . + if @xmath107 , then @xmath108 . + + 2 . if @xmath109 , then @xmath110 . if @xmath111 , then @xmath112 . + if @xmath113 , then @xmath114 . + + 3 . if @xmath115 , then @xmath116 . if @xmath117 , then @xmath118 . + if @xmath119 , then @xmath120 . + + 4 . if @xmath121 , then @xmath122 . if @xmath123 , then @xmath124 . + if @xmath125 , then @xmath126 . moreover , the converses of all of the above statements hold as well . given any two wreath products of @xmath0 or @xmath1 , corollary [ l ] guarantees the asymptotic behavior of the ratios between the coefficients of their conjugacy growth series . in other words , for any two wreath products @xmath127 and @xmath128 , we know the expected relationship between the number of conjugacy classes of @xmath3 in @xmath127 and the number of conjugacy classes of @xmath129 in @xmath128 with minimal word length @xmath17 for any @xmath17 . although we know the asymptotic behavior of the above ratios , this does not mean that the ratios of the coefficients are always exactly equal to the above values . for example , consider the wreath products @xmath28 and @xmath130 , where @xmath131 are finite groups with @xmath132 conjugacy classes respectively . we expect the ratio of the coefficients of @xmath84 to the coefficients of @xmath87 to be asymptotic to @xmath133 . we compute the following coefficients with maple . + [ cols="^,^,^,^",options="header " , ] the author would like to thank ken ono for his invaluable advice and guidance on this project , and also pierre de la harpe and the referee for their helpful comments . we give the proofs of equation ( [ cgswralt ] ) and theorems [ thm ] and [ thm2 ] here . we also explain what it means for a generating set to be sufficiently large and give remarks on corollaries [ no ] and [ altasymptotic ] . + a set @xmath14 of transpositions of a set @xmath2 is called _ partition - complete _ ( _ pc _ ) @xcite if + for the corresponding property of _ partition - complete for wreath products _ ( _ pcwr _ ) @xcite , we must first establish more notation . let @xmath2 be an infinite set , @xmath3 a group , and @xmath138 . the group @xmath127 acts naturally on the set @xmath139 ; namely , for @xmath140 , the action is defined by @xmath141 for @xmath142 and @xmath143 , we let @xmath144 denote the permutation that maps @xmath145 to @xmath146 if @xmath147 , and to @xmath148 otherwise . then @xmath149 generates the group @xmath5 . now , let @xmath150 , and define the subsets @xmath151 let @xmath152 and @xmath153 be subsets , and let @xmath154 . such a set @xmath14 is said to be _ pcwr _ if + a. the transposition graph @xmath155 is connected , and b. for all @xmath156 and partitions @xmath135 , @xmath155 contains a forest of @xmath136 trees @xmath157 , with @xmath158 having @xmath159 vertices , including one vertex @xmath160 such that @xmath161 for all @xmath142 . + this proof follows from the proofs of equations ( [ cgsalt ] ) and ( [ cgswr ] ) in @xcite . for each @xmath162 , we can split @xmath163 into a product of an even number of cycles of even length , denoted @xmath164 , and a product of cycles of odd length , denoted @xmath165 , so that @xmath166 . let @xmath167 denote the set of conjugacy classes of @xmath70 ; we write @xmath168 for the class @xmath169 . to each conjugacy class in @xmath87 we associate an @xmath167-indexed family of partitions . using the same notation as in @xcite , we associate the conjugacy classes in @xmath70 to the family of partitions @xmath170 where @xmath171 and @xmath172 each have an even number of positive parts , in the following way . let @xmath173 be the finite subset of @xmath2 that is the union of the supports of @xmath174 and @xmath163 . let @xmath163 be the product of the disjoint cycles @xmath175 , where @xmath176 with @xmath177 and @xmath178 . we include cycles of length @xmath179 for each @xmath180 such that @xmath181 and @xmath182 , so that @xmath183 define @xmath184 to be the conjugacy class of the product @xmath185 . for @xmath186 and @xmath187 , let @xmath188 denote the number of cycles @xmath189 in @xmath190 such that @xmath191 and @xmath192 . let @xmath193 be the partition with @xmath194 parts equal to @xmath195 , for all @xmath187 . note that @xmath196 also observe that the partition @xmath197 does not contain parts of size @xmath179 , because if @xmath198 , then @xmath199 . using the same notation as above , let @xmath200 be the partition with @xmath201 parts equal to @xmath202 . we can write @xmath203 as above , so @xmath204 and @xmath205 each split into two partitions , one of which has an even number of parts . define the _ type _ of @xmath206 to be the family @xmath207 . then two elements in @xmath87 are conjugate if and only if they have the same type . thus , each @xmath167-indexed family of partitions @xmath207 is the type of one conjugacy class in @xmath87 . consider an @xmath167-indexed family of partitions @xmath208 and the corresponding conjugacy class in @xmath87 . let @xmath209 be the sums of the parts of @xmath210 respectively , and let @xmath211 be the number of parts of @xmath210 respectively . hence , the contribution to @xmath216 from @xmath208 is @xmath217 it follows that @xmath218\\ & & \times \prod_{\eta \in { ( h_a)}_{}\setminus1 } \left [ \left(\prod_{u_{\eta}=1}^\infty \frac{1}{1-q^{u_{\eta } } } \right ) \left(\frac{1}{2 } \prod_{v_{\eta}=1}^\infty \frac{1}{1-q^{v_{\eta } } } + \frac{1}{2 } \prod_{v_{\eta}=1}^\infty \frac{1}{1+q^{v_{\eta } } } \right ) \right]\\ & = & \left [ \left ( \frac{1}{2 } \prod_{n_{1}=1}^\infty \frac{1}{1-q^{2 n_{1 } } } + \frac{1}{2 } \prod_{n_{1}=1}^\infty \frac{1}{(1-q^{n_{1}})^{2 } } \right ) \right]\\ & & \times \prod_{\eta \in { ( h_a)}_{}\setminus1 } \left [ \left ( \frac{1}{2 } \prod_{n_{\eta}=1}^\infty \frac{1}{1-q^{2 n_{\eta } } } + \frac{1}{2 } \prod_{n_{\eta}=1}^\infty \frac{1}{(1-q^{n_{\eta}})^{2 } } \right ) \right]\\ & = & \left ( \frac{1}{2 } \prod_{k=1}^\infty \frac{1}{1-q^{2 k } } + \frac{1}{2 } \prod_{k=1}^\infty \frac{1}{(1-q^{k})^{2 } } \right)^{\left\|{(h_a)}_{*}\right\|}.\end{aligned}\ ] ] the equality between the first and second line is given in the appendix of @xcite . the _ generalized partition function _ @xmath82 is defined for the vector @xmath219 by its generating function @xmath220 the following theorem gives an asymptotic formula for the generalized partition function , which was obtained by using properties of modular forms . corollaries [ asymptotic ] , [ altasymptotic ] , [ ratio ] , and [ l ] all follow from the above theorem , since the conjugacy growth series for permutational wreath products correspond to the generalized partition function with vectors @xmath226 in the symmetric case and @xmath227 in the alternating case . by the binomial theorem applied to the conjugacy growth series in equation ( [ cgswralt ] ) , we find that @xmath228.\ ] ] but , intuitively , the summands corresponding to @xmath229 grow much more slowly than the summand corresponding to @xmath230 , since the instance of @xmath136 in the exponential function is negative . therefore , the above sum is asymptotic to the @xmath230 term , so we have @xmath231 we now introduce the proofs of theorems [ thm ] and [ thm2 ] . in a paper by bruinier , kohnen , and ono @xcite , the universal polynomial @xmath232 is defined as + @xmath233 and it is used to define a recursion relation for coefficients of meromorphic modular forms on @xmath234 . frechette and the author @xcite modify this polynomial to the above @xmath48 and use it to define a recursion relation for coefficients of quotients of rogers - ramanujan - type @xmath24-series . their proof surprisingly only requires properties of logarithmic derivatives applied to a @xmath24-series infinite product identity . the proof below is adapted from the proof in @xcite and can be applied to any @xmath24-series infinite product identity , including the famous identity of nekrasov and okounkov @xcite . define the @xmath24-series identity @xmath235 so that @xmath236 and @xmath237 . we take logarithms of both sides to obtain @xmath238 by the taylor expansion for @xmath239 . then we take the derivatives of both sides to obtain @xmath240 so we have @xmath241 for convenience , define @xmath242 . expanding the right hand side and equating coefficients , we now have @xmath243 the symmetric power functions @xmath244 and the elementary symmetric functions and @xmath245 are both functions of @xmath246 . ] @xmath247 exhibit a similar relationship ; namely , we have the identity @xmath248 evaluating equation ( [ eq ] ) at @xmath249 , where @xmath250 are the roots of the polynomial @xmath251 we have that @xmath252 for @xmath253 . then we have @xmath254 . using the fact that @xmath255 we see that @xmath256 is exactly @xmath257 plus the additional term with @xmath258 and @xmath259 . we arrive at the recursion @xmath260 thus , we have @xmath261 theorem [ thm ] gives a recurrence formula for the coefficients @xmath44 of the conjugacy growth series of a permutational wreath product in which the group @xmath34 has @xmath33 conjugacy classes . now , we consider the more general infinite product @xmath262 for any complex number @xmath263 , and we ignore its implications for finite groups . then the above proof also applies to the coefficients of the nekrasov - okounkov hook length formula @xcite @xmath264 if we change variables @xmath265 and @xmath266 for @xmath37 . the coefficients @xmath267 of the infinite product @xmath268 therefore satisfy the recurrence relation @xmath269 although for @xmath270 we can no longer observe the relationship between the number of conjugacy classes of @xmath34 and the coefficients of the conjugacy growth series of @xmath271 , we do obtain a simple recursion for the nekrasov - okounkov hook length formula which is independent of complex analysis and hook lengths . this proof closely follows the proof of theorem [ thm ] . define the @xmath24-series identity @xmath272 so that @xmath273 . then , by the binomial theorem , we have @xmath274 it suffices to find recurrence relations for each summand . define @xmath275 we take the logarithmic derivative of both sides as in the proof of theorem [ thm ] . first , we take logarithms of both sides to obtain @xmath276 by the taylor expansions for @xmath239 and @xmath277 . then we take the derivatives of both sides to obtain @xmath278q^{n-1},\end{aligned}\ ] ] so we have @xmath279q^n\right)\left(1+\sum\limits_{n=1}^\infty a_k(n)q^n\right).\ ] ] for convenience , define @xmath280 $ ] . expanding the right hand side and equating coefficients , we now have @xmath281 using the same identity between the symmetric power functions and the elementary symmetric functions as in the proof of theorem [ thm ] , we arrive at the recursion @xmath282\\ & = & \widehat{f}_n\big(a_k(1),\dots , a_k(n-1)\big)-\sum_{\d\mid n}\d\cdot\left[(-1)^\d(k - m_a)-(k+m_a)\right].\end{aligned}\ ] ] thus , we have @xmath283\right).\ ] ] this recurrence relation gives the coefficients @xmath72 in terms of the coefficients @xmath284 , @xmath285 , @xmath286 of each summand . since the linear combination of infinite products is raised to the @xmath287th power in the conjugacy growth series , presumably there is no simple way to obtain a recurrence relation for @xmath72 in terms of @xmath288 as in the symmetric case .	we examine the conjugacy growth series of all wreath products of the finitary permutation groups @xmath0 and @xmath1 for an infinite set @xmath2 . we determine their asymptotics , and we characterize the limiting behavior between the @xmath1 and @xmath0 wreath products . in particular , their ratios form a limit if and only if the dimension of the symmetric wreath product is twice the dimension of the alternating wreath product .
[ [ optical - setup ] ] * optical setup * + + + + + + + + + + + + + + + lattice beams with gaussian spatial profiles , lying and polarized in the horizontal plane , were centered onto the condensate . the large @xmath65 beam - waist diameters of the lattice beams ensured that the trapping frequencies were not strongly modified by the lattice potential . laser alignment and relative intensities were tuned to produce six - fold symmetric diffraction patterns of condensates released from lw- and sw - only lattices . in the combined lattice , the relative displacement of the lw and sw lattices was measured using two two - color mach - zehnder interferometers , one for beams 1 and 2 and the other for beams 1 and 3 , and stabilizing all four interference signals by feedback to separate piezo - actuated mirrors in the optical paths . a rotatable glass plate within each interferometer introduces a controlled relative shift between the two lattice colors that , following stabilization , is then imparted onto the lattice beams at the location of the optical lattice . for the kaptiza - dirac diffraction experiments , the active stabilization was engaged with the lattice light at negligible intensity before the brief high - intensity pulse of the lattice potential . for the bloch - state analysis experiments , the stabilization was involved at the beginning of the ramp - on of the lattice beams . to reduce deleterious effects of atomic interactions in the time - of - flight analysis , the vertical optical confinement was switched off about 0.5 ms before all other optical trapping beams , including the lattice beams , were extinguished . the optical frequencies of the sw and lw lattice beams were stable within a few tens of mhz , at the wavelengths of 532.15 nm and 1063.96 nm respectively . while the wavelengths are not perfectly commensurate , we note that the relative displacement of the sw and lw lattices varies by only ( 18,9 ) nm over the @xmath66 transverse diameter of the condensates in our experiments . [ [ bloch - state - analysis ] ] * bloch - state analysis * + + + + + + + + + + + + + + + + + + + + + + momentum populations at the reciprocal lattice wavevectors are measured by integrating under the narrow atomic distributions observed in time of flight images , thereby accounting for effects of finite temperature , atomic interactions , and finite size during the expansion of the gas . the wannier state fourier components that appear in eq . [ eq : firstorder ] are determined , assuming the wannier state to be gaussian , from the second - order diffraction populations as @xmath67 . the limiting values of the momentum population ratios enable us to distinguish different lattice geometries as summarized in table [ tab : bragg ratio main ] . .the momentum population ratios for various lattice geometries assuming the wannier wavefunctions at all sites to be @xmath56-functions in real space . the hallmark of the kagom lattice is the limiting value of 1/9 as observed in fig . [ fig : unitcellpop ] . although the 1d stripe and the decorated triangular lattices have the same momentum ratios in this table , they can be easily distinguished through their rotational symmetries , as can be seen in fig . [ fig : bragg].[tab : bragg ratio main ] [ cols="^,^",options="header " , ] we thank s. schreppler for experimental assistance . c. k. thomas acknowledges support by the department of energy office of science graduate fellowship program ( doe scgf ) , made possible in part by the american recovery and reinvestment act of 2009 , administered by orise - orau under contract de - ac05 - 06or2310 . p. h. was supported by nsf - dmr 0645691 . this work was supported by the nsf and by the army research office with funding from the darpa optical lattice emulator program .	geometrically frustrated systems with a large degeneracy of low energy states are of central interest in condensed - matter physics @xcite . the kagom net a pattern of corner - sharing triangular plaquettes presents a particularly high degree of frustration , reflected in the non - dispersive orbital bands . the ground state of the kagom quantum antiferromagnet , proposed to be a quantum spin liquid or valence bond solid @xcite , remains uncertain despite decades of work . solid - state kagom magnets @xcite suffer from significant magnetic disorder or anisotropy that complicates the interpretation of experimental results . here , we realize the kagom geometry in a two - dimensional optical superlattice for ultracold @xmath0rb atoms . we employ atom optics to characterize the lattice as it is tuned between various geometries , including kagom , one - dimensional stripe , and decorated triangular lattices , allowing for a sensitive control of frustration . the lattices implemented in this work offer a near - ideal realization of a paradigmatic model of many - body quantum physics . ultracold atoms trapped within optical lattices exhibit correlated phases , such as mott insulators of bosons @xcite and fermions @xcite and effective antiferromagnetic ordering in one - dimensional chains @xcite , which are analogous to those studied in condensed - matter physics . while most experiments have been performed in primitive bravais lattices , with a single lattice site per unit cell , recent works have explored non - standard optical lattices with a few - site basis such as the two - dimensional honeycomb @xcite and checkerboard @xcite geometries , and double - well superlattices @xcite . such lattices add a low - energy orbital degree of freedom , corresponding to the assignment of a particle to each site within the unit cell , that can yield non - trivial ordering and dynamics . the kagom lattice is obtained by eliminating every fourth site from a triangular lattice of spacing @xmath1 , with the eliminated sites forming a triangular lattice of spacing @xmath2 . the remaining sites generate three connected @xmath3-orbital bands within a bandwidth on the order of the intersite tunneling energy . intriguingly , the frustration of antiferromagnetic interactions in the kagom geometry implies that one of these bands be non - dispersing . such flat bands accentuate the role of interparticle interactions , leading possibly to crystalline ordering @xcite and supersolidity @xcite for scalar bosons , and ferromagnetism of itinerant fermions @xcite . here , we realize a two - dimensional kagom lattice for ultracold atoms by overlaying two commensurate triangular optical lattices generated by light at the wavelengths of 532 and 1064 nm ( fig . [ fig : scheme ] ) . stabilizing and tuning the relative position of the two lattices , we explore different lattice geometries including a kagom lattice , a one - dimensional ( 1d ) stripe lattice , and a decorated triangular lattice . we characterize these geometries using kapitza - dirac ( kd ) diffraction and by analyzing the bloch - state composition of a superfluid released suddenly from the lattice . the bloch - state analysis also allows us to determine the ground - state distribution within the superlattice unit cell as the potential is tuned between geometries . the atom - optical tools developed in this work are useful for the characterization of other superlattices and for superlattice - based atom interferometry . three plane waves of light of equal intensity @xmath4 , equal frequency , and wavevectors ( and linear polarizations ) lying in a plane and intersecting at equal angles , produce a triangular lattice of points with zero intensity , and a honeycomb lattice of points with maximum intensity @xmath5 separated by a triangular lattice of intensity saddle points with intensity @xmath6 . in our work , the short - wavelength ( sw ) lattice is formed by light that is blue - detuned from the principal atomic resonances of rubidium , so that atoms are attracted to the triangular lattice of zero - intensity sites with a lattice spacing of @xmath7 . the long - wavelength ( lw ) lattice is red detuned , so that its zero - intensity points are potential - energy maxima for rubidium atoms . a unit cell of the lw lattice contains four sites of the sw triangular lattice , labeled a , b , c and d in fig . [ fig : scheme ] . aligning the positions of the lw potential maxima to coincide with sites d lowers the potential energies @xmath8 at the indicated sites by equal amounts @xmath9 where @xmath10 is the maximum scalar potential depth of the lw lattice ( we ignore the @xmath11 vector shift in this lattice @xcite ) . as @xmath12 is increased , atoms are excluded from sites d , while the remaining sites form the kagom optical lattice . the kagom geometry persists until @xmath13 , at which point atoms become preferentially confined in the lw honeycomb lattice . compared with previous proposals @xcite , our simpler approach to creating a kagom lattice allows one to tune the lattice geometry , thereby controlling its degree of frustration . aligning the lw potential maxima with the sw lattice saddle points disfavors population in two sites of the four - site unit cell ( e.g. @xmath14 ) producing a 1d stripe lattice ( fig . [ fig : scheme]c ) . aligning the lw potential maxima with the sw potential maxima disfavors population in three sites of the four - site unit cell ( e.g. @xmath15 ) , producing a decorated triangular lattice with lowest - energy sites forming a triangular lattice while the remaining sites form a kagom lattice of local potential minima . experiments were conducted with scalar bose - einstein condensates ( becs ) of @xmath16 @xmath0rb atoms produced at temperatures of about 80 nk in a red - detuned crossed optical dipole trap with trap frequencies of @xmath17 , with @xmath18 applying vertically . the relative position of the lw and sw lattices was measured using two separate two - color interferometers and stabilized actively by displacing mirrors in the relevant optical paths . the relative position was tuned along two linearly independent directions by adding within each interferometer a variable relative phase shift between the two lattice colors ( see methods ) . the first atom - optical method we used to demonstrate the tunable superlattice is kapitza - dirac diffraction @xcite , for which the lattice potential is suddenly switched on and off , after which the condensate is released from the optical trap and imaged after a time of flight to evaluate its momentum - space distribution . the pulse is sufficiently short that one can neglect the kinetic energy of the diffracted momentum components over the pulse duration @xmath19 . the effect of the pulse is then to imprint a phase @xmath20 proportional to the lattice potential @xmath21 onto the condensate wavefunction ( which is initially nearly uniform ) . the corresponding momentum - space distribution is sensitive to the relative displacement of the lw and sw lattices . to exhibit this sensitivity we blocked one of the incident bi - chromatic lattice beams and examined the resulting one - dimensional superlattice formed by two bichromatic light beams intersecting at an angle of 120@xmath22 , with potential energy given as @xmath23 where @xmath24 is the 1d lw lattice spacing , and @xmath25 is the distance between the lw and sw intensity minima . considering diffraction up to second order in the phase modulation depth , the atomic populations at wavevectors @xmath26 are given as@xmath27 where @xmath28 is the @xmath29-order bessel function , evaluated at @xmath30 according to the superscript . the lack of inversion symmetry of the lattice produced by an incommensurate value of @xmath25 appears as a left / right momentum asymmetry in the diffracted matter wave ( fig . [ fig : kd ] ) . a second method to characterize the optical superlattice is the momentum - space analysis of a superfluid occupying the ground state of the lattice potential . for such analysis , the optical lattice potential depth was ramped up from zero over 90 ms , held at a constant level for 100 ms , and then suddenly switched off . imaging after a time of flight reveals coherent momentum peaks of the condensate wavefunction at the reciprocal lattice vectors @xmath31 of the lw lattice . @xmath32s pulse of the lattice potential followed by 26 ms time of flight , exhibit left / right momentum asymmetry ( defined as @xmath33 that varies as the displacement @xmath25 between the lw- and sw - lattice intensity minima is varied , reflecting broken inversion symmetry of the lattice , in close agreement with the predicted behavior ( solid line ) . ] identified by color according to the inset . the data ( averages over 4 - 5 repeated measurements ) agree with calculations of the single - particle ground state ( solid lines ) with @xmath34 and @xmath35 measured by kapitza - dirac scattering from lw - only or sw - only lattices . @xmath12 was higher than the chemical potential @xmath36 of the condensate in the sw - only lattice , so that modifications of the transverse condensate wavefunction due to interactions are negligible . ] . the d - site population is extinguished ( b ) as the energy offset @xmath12 exceeds the condensate chemical potential ( @xmath37 ) , and the momentum population ratios reach the asymptotic value of 1/9 expected for a kagom lattice ( a ) . data points represent averages of 7 10 measurements . dashed curves indicate expected population ratios @xmath38 for the ideal kagom configuration while the shaded region indicates the expected variation in @xmath38 given a shot - to - shot instability of @xmath3920 nm in the relative position of the lw and sw lattices . ] varying the relative position of the two lattices we identify the three high - symmetry lattice configurations ( fig . [ fig : bragg]a ) . given that the scalar condensate occupies the ground state of the lattice potential , its wavefunction can be taken as real and positive ; thus , its momentum distribution is symmetric under inversion . expansion from both the kagom and the decorated triangular lattices shows the three - fold rotational symmetry of the optical superlattice . in the 1d stripe geometry , one expects equally weak occupation of two sites ( e.g. , a and d ) , and equally strong occupation of the other two sites ( b and c ) of the superlattice unit cell . such a distribution is ( nearly ) invariant under displacements of @xmath1 along the a - d axis , and condensate diffraction along that axis should reflect the shorter periodicity of the sw lattice . the momentum distribution should also be symmetric under reflection about the a - d axis . both traits are observed experimentally . the bloch - state momentum distributions allow one to quantify the ground - state wavefunction within a unit cell of the superlattice , which we express as @xmath40 where @xmath41 is the normalized wannier state wavefunction , @xmath42 the position and @xmath43 the fractional atomic population of site @xmath44 of the unit cell . at low @xmath45 , we approximate @xmath46 as cylindrically symmetric and identical for all @xmath47 . from the momentum - space populations @xmath48 ( @xmath49 ) in the three first - order diffraction peaks of the lw lattice corresponding to the inner hexagon of peaks in time - of - flight images and that at zero wavevector @xmath50 , one determines the distinct quantities @xmath51 where @xmath52 and @xmath53 are now fourier components of the wannier function , and @xmath54 , @xmath55 , @xmath56 and @xmath57 label the four sites so that @xmath58 . together with the normalization @xmath59 these quantities determine the atomic distribution in the unit cell . we note that inverting eqs . [ eq : firstorder ] gives a discrete degeneracy of solutions that can be resolved by prior knowledge of the approximate populations . we measured the population ratios @xmath60 as the superlattice geometry was gradually tuned , using data from the second - order diffraction peaks to characterize the wannier state ( see methods ) . translating the relative position of the two lattices ( fig . [ fig : bragg]b ) , one advances from the kagom geometry , with equal population in the three ratios , to the 1d stripe geometry , with two identically small ratios , and then to another kagom - geometry lattice . our data agree with a calculation of the ground - state single - particle wavefunction for the known lattice depths . we focus finally on the kagom - geometry lattice alignment , and examine the transition between the triangular and kagom geometries ( fig . [ fig : unitcellpop ] ) . at zero lw lattice depth , the atoms are confined in a sw triangular lattice , and the first - order lw lattice diffraction orders are absent , indicating a unit - cell population of @xmath61 . as the lw lattice depth is increased , the population ratios @xmath38 increase and the kagom geometry is achieved by gradually expelling atoms from one site of the unit cell . the population ratios tend toward a limiting value of 1/9 that is a hallmark of diffraction from a kagom lattice wherein the atoms are distributed as @xmath62 ( see methods ) . here , the ground state of the kagom lattice does not suffer from frustration , allowing for the stability of a bose - einstein condensate in the lattice . in the future , effects of frustration may be explored by transferring bosons into the excited @xmath3-orbital flat band , or by changing the sign of the hopping energy @xcite so that the flat band becomes the lowest - energy band . we note that the present choice of wavelengths also yields kagom lattices for the fermionic isotopes @xmath63li and @xmath64k . introducing fermions into the lattice at the appropriate fillings would place the fermi energy within the flat band , allowing for studies of flat - band ferromagnetism due to repulsive interactions @xcite or of enhanced cooper pairing for attractive interactions @xcite . furthermore , remarkable versatility of the 2d superlattice opens new possibilities to emulate a distorted kagom lattice as well as a nearly ideal one enabling us to explore the possible various quantum ground states of the kagom quantum antiferromagnet .
m55 ( = ngc 6809 , = c1936 - 310 ) is a metal - poor halo globular cluster that is rich and at the same time relatively easy to study owing to its proximity and relative openness . its reddening and apparent distance moduli are estimated at @xmath11 and @xmath12 ( harris 1996 ) . the cluster was selected as one of the targets in an ongoing survey for eclipsing binaries in globular clusters ( eg . kaluzny , thompson & krzeminski 1997 ) . as a side result of that survey we obtained extensive time - series photometry for several rr lyr stars belonging to m55 . more than half of these variables are new discoveries . this contribution is devoted exclusively to the presentation and analysis of photometry of m55 rr lyr variables from the cluster field . results obtained for other variables will be published elsewhere ( thompson et al . , in preparation ) . time - series photometry of m55 was obtained during the interval 1997 may 09 september 17 with the 1.0-m swope telescope at las campanas observatory . the ccd camera used for the observations has a field of view of @xmath13 with a scale of 0.435 arcsec / pixel . more than 700 @xmath14-band frames and 65 @xmath15-band frames were obtained with exposure times ranging from 150 to 300 seconds for the @xmath14 filter and 200 to 360 seconds for the @xmath15 filter depending on the seeing . instrumental photometry was measured using dophot ( schechter , mateo & saha 1993 ) and the transformation to the standard @xmath16 system was based on observations of several landolt fields ( landolt 1992 ) . a more detailed description of the observations and reductions can be found in a complementary paper presenting results for eclipsing binaries and sx phe stars identified in m55 ( thompson et al . , in preparation ) . this search for variable stars in m55 has identified 15 rr lyrae variables . six of them ( stars v1v6 ) were previously known ( bailey 1902 , king 1951 , king and bruzual 1976 ) . of the remaining nine newly discovered variables , 5 are type rrc and 4 are type rrab . the periods , intensity averaged @xmath14 brightnesses , amplitudes , @xmath17 colors and equatorial coordinates of the 15 rr lyr variables are listed in table 1 . a color - magnitude diagram of m55 derived from our observations of m55 is presented in fig . open circles denote the rrc stars and filled circles correspond to the rrab variables belonging to m55 . the filled triangles correspond to two rrab stars that are more distant than m55 by about 3.5 magnitudes . these two variables are most likely members of the sagittarius dwarf galaxy ( ibata et al . the presence of a noticeable population of stars belonging to the sagittarius dwarf in the field of m55 has been noted by mateo & mirabel ( 1996 ) and fahlman et al . moreover the first authors reported identification of three rr lyr stars from the sagittarius dwarf in their survey for variables in m55 . it is likely that the two distant rrab stars discovered by us are among the variables already noted by mateo & mirabel ( 1996 ) . the periods of the cluster rrc stars are between 0.310 and 0.406 days with a mean period of 0.36 days . the periods of the cluster rrab variables are between 0.580 and 0.722 days with a mean value of 0.66 days . these properties clearly place m55 among the oosterhoff type ii clusters . we have fit our @xmath14-band light curves to fourier series with the form : @xmath18 where @xmath19 and @xmath20 is the pulsation period of the star . a method developed by schwarzenberg - czerny ( 1997 ) was used to determine values of @xmath21 , @xmath22 and @xmath23 . although the formal errors of our periods from a least squares fit are quite small , we estimate their actual value to be as large as 0.000010 . this is caused by correlations of residuals not accounted for in the least squares solution . our estimate for the errors in the derived periods corresponds to a phase uncertainty of @xmath24 over the entire length of the observations , a reasonable value given the quality of our light curves . its magnitude is in agreement with the scatter of periods obtained by fitting fourier series with different numbers of harmonics . in figs . 2 , 3 and 4 we show @xmath14 , @xmath15 and @xmath4 light curves of the m55 rrab stars , the m55 rrc stars , and the rrab stars from the sagittarius dwarf galaxy , respectively . the @xmath4 colors of each star were derived from equation ( 1 ) by calculating the @xmath14 brightness at the epoch of each of the @xmath15 observations . the values of the peak to peak amplitudes @xmath25 presented in table 1 are used to plot the period - amplitude diagram shown in fig . again open circles denote rrc stars , filled circles rrab stars , and filled triangles the two sagittarius dwarf rrab stars identified in the field of m55 . the solid line represents a linear fit to rrab variables in m3 ( kaluzny et al 1998 ) . the data presented in fig . 5 agree with the well established fact that metallicity of m55 is lower than the metallicity of m3 . at the same time the metallicities of the two rrab stars from the sagittarius dwarf are likely to be slightly higher than the metallicity of m3 . in a series of papers simon and teays ( 1982 ) , simon ( 1989 ) and simon and clement ( 1993 ) have presented a method of estimating the masses , luminosities , effective temperatures and the helium abundance of rrc stars based only on a fourier decomposition of the @xmath14-band light curves . the equations of simon and clement ( 1993 ) are : @xmath26 @xmath27 @xmath28 @xmath29 where @xmath30 is the mass of the star in solar units , @xmath20 is the first overtone pulsation period in days , @xmath31 is the luminosity in solar units , @xmath32 the effective temperature in kelvins and @xmath33 ( cf . equation 1 ) . using the above equations we computed the masses , luminosities and effective temperatures of the rrc stars in our sample . these are presented in table 2 together with the values of @xmath34 , @xmath35 and @xmath36 . the errors presented in table 2 are calculated from the error propagation law . due to the fact that simon and clement ( 1993 ) used a cosine fourier series the values of @xmath36 in table 2 differ from their typical values by @xmath37 . we exclude from our sample variables v9 , v10 and v12 due to their irregular light curves ( see subsection 3.2 for a detailed discussion ) . we also omit variable v11 due to the low accuracy of the estimation of @xmath36 . the mean values of the mass , luminosity , effective temperature , and helium abundance for the remaining 5 rrc variables are @xmath38 , @xmath39 , @xmath32=@xmath40 k , and @xmath3 , respectively . these values are broadly consistent with those for the rrc variables in the sequence of globular clusters discussed by kaluzny et al ( 1998 ) ( cf . their table 4 ) . in particular we conclude that m55 is similar to ngc 2298 . the mean @xmath41 and @xmath32 for this cluster are 1.75 and 7200 k , respectively , with a helium abundance of @xmath42 = 0.26 . previous determinations of the metallicity of m55 vary between @xmath43 ( bica and pastoriza 1983 , pilachowski 1984 , smith 1984 ) and @xmath44 ( zinn 1980 , richtler 1988 ) . the lower estimates agree with those of suntzeff et al ( 1991 ) who obtained @xmath45}=-1.81 $ ] and mcwilliam et al ( 1992 ) who obtained @xmath45}=-1.91 $ ] . considering that the values of @xmath41 , @xmath32 and @xmath42 obtained from the rrc stars are consistent with the sequence from kaluzny et al ( 1998 ) , we conclude that this analysis supports the more metal - poor determinations of [ fe / h ] for m55 . however , the fourier analysis suggests a mean mass for the m55 rrc stars that is discrepant with the kaluzny et al ( 1998 ) sequence . our mean value is @xmath38 . it places m55 between globular clusters such as ngc 6171 ( [ fe / h]=@xmath46 , @xmath47 and @xmath32=7447 k ) and m5 ( [ fe / h]=@xmath48 , @xmath49 and @xmath32=7388 k ) , these values disagree markedly with the values determined in this present work . we must take into account that the simon , teays and clement formulae for estimates of @xmath50 , @xmath41 , @xmath51 and @xmath42 are all linear functions of @xmath52 and @xmath36 . let us assume for simplicity that the distributions of @xmath52 and @xmath36 in the cluster are independent , with standard deviations of 0.04 and 0.6 , respectively . then by propagation of errors rule one can calculate the covariance matrix of @xmath50 , @xmath41 , @xmath51 and @xmath42 . the resulting correlation coefficients are all above 0.9 , except for @xmath53 for the correlation of @xmath50 and @xmath51 . this may suggest that the simon , teays and clement formulae do not effectively use all of the information contained in @xmath54 and @xmath36 . for this reason it may be preferable to consider physical similarities between different clusters directly on the @xmath54 and @xmath36 plane . for our sample of the rrc variables in m55 these values are @xmath55 and @xmath56 , respectively . in fig . 6 we show a plot of @xmath41 against mean @xmath14 . the solid lines have a slope of 0.4 and are separated by 0.04 in @xmath41 , which represents the uncertainty in the values of @xmath41 computed from @xmath36 and @xmath20 . it is clear that our points do not lie between these lines . but the accuracy in the estimates for the masses , the temperatures and the luminosities does not come from the experimental errors in the measurements of the periods and the values of @xmath36 , but probably in the systematic errors in the simon and clement ( 1993 ) analysis . three of the rrc variables presented in fig . 3 , namely v9 , v10 and v12 display a modulation of their light curves . on certain occasions the amplitudes of the light curves change on the time scale of days . this is shown in fig . 7 , where we present average light curves with selected light curves from individual nights . such behavior is not typical for stars exhibiting the blazhko effect ( blazhko 1907 ; see also the discussion in smith 1995 ) . in particular , in a sample of 46 stars for which blazhko periods have been determined , the shortest period is 10.9 days . we find no statistically significant periods for v9 , v10 , and v12 in the range 5 days to 60 days . this behavior is common in bimodal pulsators ( rr lyr type rrd ) . to clarify the nature of this modulation we performed a period analysis of the light curves . power spectra , clean ( roberts et al . 1987 ) and multi - harmonic periodograms of raw light curves reveal no radial periods other than those listed in table 1 and their harmonics and @xmath57 aliases . however , the periods of all 3 stars are very close to 1/3 of a day so that the same portions of the light curves are observed over several weeks . in order to judge the magnitude of any effects of aliasing , we simulated @xmath14-band observations of a bimodal pulsating star with @xmath58 from table 1 and overtone - to - fundamental period ratio @xmath59 = 0.745 , typical for rrd variables ( smith 1995 ) , using the formula : @xmath60 where @xmath61 denotes a gaussian noise component with standard deviation 0.025 mag . synthetic light curves were constructed for each of v9 , v10 and v12 , with the sampling corresponding to the times of real observations . the amplitudes were selected to closely mimic the light curves observed in fig . the fundamental period @xmath62 was not detected in power spectra of any of the simulated light curves . next we prewhitened our real and synthetic light curves by removing a sinusoid with the main period . in all of the power spectra of the prewhitened synthetic light curves , the fundamental period and its aliases were prominent . no periods close to @xmath62 were detected in any of the prewhitened observed light curves , excluding the possibility of double mode radial pulsations in these three variables . the upper panel in fig . 8 shows the power spectrum of the observations of variable v9 generated with clean software ( roberts et al . the most prominent peak is the radial frequency with period @xmath63 . the arrow marks the position of the fundamental period , assuming @xmath59 = 0.745 . a second , smaller peak is apparent in the vicinity of the larger peak . to check on the reality of this secondary peak we prewhitened our observed light curves removing a sinusoid with the main period and its two harmonics . the power spectrum of the prewhitened light curve of v9 is presented in the second panel of fig . the highest peak now corresponds to the frequency of the secondary peak in the first panel , with a period of @xmath64 . the prewhitened observed light curve , phased with this period , is presented in the lower panel of fig . we performed a similar analysis for variables v10 and v12 . in the power spectrum of v10 we again found two very close peaks ( upper panel of fig . after prewhitening of the observations , the highest peak in the power spectrum corresponds to a period of @xmath65 ( middle panel of fig . the lower panel of fig . 9 shows the observations phased with this period . variable v12 also shows multiperiodic behavior ( the upper panel of fig . after prewhitenning of the observations the most dominant period is @xmath66 ( the middle panel of fig . the lower panel of fig . 10 shows the observations phased with this period . non - radial oscillations with many frequencies are common in @xmath10 scuti variables ( main sequence stars in the instability strip ) but have not yet been observed in rr lyr stars . the theoretical calculations performed by van hoolst et al ( 1998 ) clearly show that low - degree non - radial modes can be excited in rr lyr stars . in their model , a large number of unstable low - degree ( @xmath67 ) modes in the vicinity of the radial modes are partially trapped and therefore have the largest growth rate and as a result are presumably most likely to be excited ( see their fig . 1 ) . we propose that this is exactly what we observe for variables v9 , v10 , and v12 in m55 . the @xmath10-scuti stars have periods about 10 times shorter than rr lyr stars . data sets consisting of observations over a few consecutive nights cover many tens of cycles of variability with good phase coverage , with the result that the complex period structure can be well determined . observations of rr lyr light curves are usually not so extensive , with a typical light curve containing 100 - 200 measurements . our data are significantly more detailed in both phase coverage and number of observations , with the light curves of our variables v9 , v10 and v12 containing over 700 points collected during four months . we conclude that other rr lyr variables may also pulsate with non - radial modes but that one needs excellent photometric coverage in order to detect these non - radial frequencies . recently jurcsik and kovacs ( 1996 ) , kovacs and jurcsik ( 1996 ) , kovacs and jurcsik ( 1997 ) , and jurcsik ( 1998 ) have extended the work of clement , simon and teays , developing methods for obtaining the metallicity , absolute magnitudes , intrinsic colors and temperatures of rrab stars basing on a fourier decomposition of @xmath14-band light curves . their formulae are : @xmath68 } = -5.038 - 5.394\log p + 1.345\phi_{31}\ ] ] @xmath69 @xmath70 @xmath71}\ ] ] where @xmath72 ( cf . equation 1 ) . the above equations are valid only for rrab stars with regular light curves , i.e. variables with a deviation parameter @xmath73 smaller than 3 ( see jurcsik and kovacs 1996 ) . in our sample , only variable v1 satisfies this condition . in order to increase the sample size , we include rrab variables with @xmath74 , a condition satisfied by all of the rrab stars in table 1 belonging to m55 . the results are listed in table 3 which contains values and errors of @xmath34 , @xmath35 , @xmath36 , @xmath75 , @xmath76 , [ fe / h ] , @xmath32 and @xmath73 . the errors of @xmath34 , @xmath35 , @xmath36 and @xmath75 come from the least squares fitting ( cf . equation 1 ) and the errors of @xmath76 and [ fe / h ] are computed from formulae given by jurcsik and kovacs ( 1996 ) and kovacs and jurcsik ( 1996 ) . in fig . 11 we plot @xmath77 - @xmath76 versus @xmath77 , where @xmath77 is the mean @xmath14 magnitude for the variables . the horizontal line is a fit to the data , giving an apparent distance modulus of @xmath8 , comparable to the value of 13.76 listed by harris ( 1996 ) . as discussed in section 3.1 , previous determinations of the metallicity of m55 vary between @xmath43 and @xmath44 . the fourier analysis of the rrab stars is consistent with this range , with only variable v8 indicating a larger value of [ fe / h ] . however , this star has the largest value of @xmath73 in the present sample of m55 rrab stars . our determinations of [ fe / h ] based on a fourier decomposition of the light curves of rrab variables are consistent with the period - amplitude diagram presented in fig . the reddening of rrab stars can be calculated from the metallicity , expressed in terms of @xmath78 ( preston 1959 ) . from blanco ( 1992 ) we have the relation @xmath79 where @xmath80 is the observed mean color in the 0.50.8 phase interval . based on the globular cluster metallicity scale adopted by zinn and west ( 1984 ) and zinn ( 1985 ) , suntzeff et al . ( 1991 ) derived the following @xmath78[fe / h ] relation : @xmath68 } = -0.408 - 0.158 \delta s\ ] ] using above formulae and adopting a metallicity for m55 of @xmath45}=-1.82 $ ] ( zinn and west 1984 ) we obtain @xmath81 . the average value of @xmath82 calculated in this way for our four m55 rrab stars is @xmath5 . the reddening in blanco s analysis is fairly independent of the metallicity over the quoted range . using a metallicity of @xmath45}=-1.54 $ ] ( smith 1984 ) we derive @xmath83 . a similar calculation for the reddening of the rrab star in sagittarius with a well defined light curve ( v14 with @xmath73 = 1.40 ) gives @xmath84 , accepting the value of @xmath45}$ ] determined from the fourier analysis for this star . these values are slightly larger than previous determinations of between 0.06 and 0.08 mag ( lee 1977 , reed et al . recently schlegel et al . ( 1998 ) have published a new all - sky reddening map based on the cobe / dirbe and iras / issa maps . their value of @xmath82 at the position of m55 @xmath82 = 0.135 mag , consistent with our estimate . adopting a value of @xmath5 we estimate the absolute distance modulus to be @xmath85 . the red and blue edges of the rr lyrae zone on the horizontal branch compiled by smith ( 1995 ) vary in @xmath86 color between 0.155 and 0.19 mag for the blue edge and between 0.38 and 0.44 mag for the red edge . this corresponds to ranges in effective temperature of 76007400 k and 62506100 k , respectively . in our case the instability strip lies between @xmath87 and @xmath88 mag . using a color excess of @xmath89 mag , we obtain @xmath90 and @xmath91 mag for the edges of the instability strip , consistent within the errors with estimates made for other globular clusters ( smith 1995 ) . we have identified nine new rr lyr variables in the field of the globular cluster m55 . two of these variables are probable members of the sagittarius dwarf galaxy . the number of known rr lyr variables in m55 is now thirteen , four rrab stars and nine rrc stars . the periods of the variables indicate that m55 is an oosterhoff type ii cluster . three of our nine rrc variables exhibit marked amplitude modulation on a time scale of less than a week . we excluded the hypothesis that such behavior is caused by the blazhko effect or double mode radial pulsations . a detailed analysis of the power spectra showed other clear frequencies in the vicinity of the main peak . these frequencies are too close to the main period to be radial pulsations . we conclude that we have detected non - radial pulsation in these rrc stars . we used fourier decomposition of the @xmath14-band light curves of the rr lyr variables to measure luminosities , effective temperatures , metallicities and masses of these variables . the mean mass of the rrc stars in m55 is equal to @xmath0 , the mean @xmath92 , and the mean effective temperature @xmath2 k. the helium abundance of m55 is @xmath3 . the blue edge of the instability strip lays at @xmath930.20 mag and the red edge at @xmath930.38 mag . the values of @xmath45}$ ] obtained from the fourier decomposition of the light curves of the rrc stars and the period - amplitude measurements presented in fig . 5 both indicate a metallicity of @xmath94 for m55 , consistent with the measurements of suntzeff et al ( 1991 ) , mcwilliam et al ( 1992 ) , among others . a similar analysis for the rrab stars suggests a metallicity of @xmath95 . this value is based only on one star with a deviation parameter @xmath73 smaller than 3.0 , and as a result is likely of low accuracy . using the blanco ( 1992 ) dependence of the color excess @xmath82 upon the mean color at minimum light of rrab stars , the pulsation period of the star , and spectroscopic parameter @xmath78 we obtained a color excess for m55 of @xmath5 for an adopted metallicity of @xmath96}=-1.8 $ ] . we also estimated the values of the visual apparent and dereddened distance moduli to be equal to @xmath8 and @xmath9 , respectively . we identified in the m55 field two rrab variables belonging most probably to the sagittarius dwarf galaxy . from the analysis of light curves we obtained for both of them @xmath45}=-1.13 $ ] . this result is consistent with earlier determinations of metallicity for rr lyr variables in the sagittarius dwarf ( sarajednini and layden 1995 , mateo et al . 1995 , marconi et al . 1998 ) . we would like to thank to prof . w. dziembowski , prof . b. paczyski and dr . p. moskalik for helpful hints and comments . we are also grateful to dr . g. pojmaski for his software help . ao , jk , wp and wk were supported by the polish committee of scientific research through grant 2p03d-011 - 12 and by nsf grant ast-9528096 to bohdan paczyski . asc acknowledges support by kbn grant 2p03c00112 . bailey , s.i . , 1902 , harv . annals , 38 , 1 bica , e.l . , pastoriza , m.g . , 1983 , ap&ss , 91 , 99 blanco , v. , 1992 , aj , 104 , 734 blazhko , s. , 1907 , astron . nachr . , 175 , 325 fahlman , g.g . , mandushev , g. , richer , h.b . , thompson , i.b . , and sivaramakrishnan , a. , 1996 , apjl , 459 , l65 . harris , w.e . 1996 , aj , 112 , 1487 ibata , r.a . , gilmore , g. , and irwin , m.j . , 1994 , nature 370 , 194 jurcsik j. , 1998 , a&a , 333 , 571 . jurcsik j. , kovacs g. , 1996 , a&a , 312 , 111 kaluzny , j. , thompson , i. , krzeminski , w. 1997 , aj , 113 , 2219 kaluzny , j. , hilditch , r.w . , clement , c. , and rucinski , s.m . , 1998 , mnras , 296 , 347 king , i. 1951 , harvard college obs . bull . , 920 , 16 king , i. , bruzual g.a . , 1976 , a&a , 50 , 459 kovacs g. , jurcsik j. , 1996 , apjl , 466 , l17 kovacs g. , jurcsik j. , 1997 , a&a , 322 , 218 landolt , a. 1992 , aj , 104 , 340 lee , y .- w . , 1977 , a&a , s 29 , 1 marconi , g. , buonauno , r. , castellani , m. , iannicola , g. , molaro , p. , pasquini , l. , and pulone , l. , 1998 , a&a , 330 , 453 . mateo , m. , udalski , a. , szymaski , m. , kaluzny , j. , kubiak , m. , and krzemiski w. , 1995 , aj , 109 , 588 . mateo , m. , and mirabel , n. , 1996 , in the origins , evolution and destinies of binary stars in clusters , eds . milone and j.c . mermilliod , vol . 90 , p. 87 . mcwilliam , a. , geisler , d. , and rich , r.m . , 1992 , pasp , 104 , 1193 pilachowski , c.a . , 1984 , apj , 281 , 614 preston , g.w . , 1959 , apj , 130 , 507 reed , b.c . , hesser , j.e . , and shawl , s.j . , 1988 , pasp , 100 , 545 richtler , t. 1988 , a&a , 204 , 101 roberts , d.h . , lehar , j. , dreher , j.w . , 1987 , aj , 93 , 968 . sarajednini , a. , layden , a. , 1995 , aj , 109 , 1089 . schechter , p.l . , mateo , m. , saha , a. 1993 , pasp , 105 , 1342 schlegel , d.j . , finkbeiner , d.p . , and davis , m. , 1998 , apj , in press ( astro - ph/9710327 ) schwarzenberg - czerny , a. , 1997 , apj , 489 , 941 simon n.r . , teays t.j . , 1982 , apj , 261 , 586 simon n.r . , 1989 , apjl , 343 , l17 simon n.r . , clement c.m . , 1993 , apj , 410 , 526 smith , h.a . , 1984 , apj , 281 , 148 smith , h.a . , 1995 , rr lyrae stars , cambridge university press , cambridge suntzeff , n.b . , kinman , t.d . , and kraft , r.p . , 1991 , apj , 367 , 528 van hoolst , t. , dziembowski , w.a . , and kawaler , s.d . , 1998 , mnras , 297 , 536 zinn , r. , 1980 , apjs , 42 , 19 zinn , r. , 1985 , apj , 293 , 424 zinn , r. , and west , m.j . , 1984 , apjs , 55 , 45	we present the results of a photometric study of rr lyrae variables in the field of the globular cluster m55 . we have discovered nine new rr lyrae stars , increasing the number of known variables in this cluster to 15 objects . five of the newly discovered variables belong to bailey type rrc and two to type rrab . two background rrab stars are probable members of the sagittarius dwarf galaxy . fourier decomposition of the light curves was used to derive basic properties of the present sample of rr lyrae variables . from an analysis of the rrc variables we obtain a mean mass of @xmath0 , luminosity @xmath1 , effective temperature @xmath2 k , and helium abundance @xmath3 . based on the @xmath4 colors , periods and metallicities of the rrab stars we estimate the value of the color excess for m55 to be equal to @xmath5 . using this value we derive the colors of the blue and red edges of the instability strip in m55 . the blue edge lies at @xmath6 mag and the red edge lies at @xmath7 mag . we estimate the values of the visual apparent and dereddened distance moduli to be @xmath8 and @xmath9 , respectively . the light curves of three of the rrc variables exhibit changes in amplitude of over 0.1 mag on the time scale of less than a week , rather short for the blazhko effect , but with no evidence for another radial pulsational frequency . however we do detect other periodicities which are clearly visible in the light curve after removing variations with the first overtone radial frequency . this is strong evidence for the presence of non - radial pulsations , a behavior common for @xmath10 scuti stars but not yet observed among rr lyr variables .
the main theoretical framework to understand the behavior of the density of states of a fermionic system has been , and still is , the independent particle model . in the degenerate gas approximation , when the excitation energy of the gas is much smaller than the fermi energy @xmath1 , the excitation spectrum relies on the properties of the single particle ( sp ) spectrum near @xmath1 . in most theoretical calculations only the average sp density of states , @xmath2 , is taken into account . for a system of @xmath3 noninteracting fermions moving in a mean field potential , the number of excited states of the many body ( mb ) system contained in a small energy window @xmath4 at energy @xmath0 , @xmath5 , is @xcite @xmath6 here @xmath0 is measured with respect to the ground state energy of the gas . for a given potential , @xmath2 is in general a function of @xmath3 . besides the condition @xmath7 , this expression assumes that the excitation energy is large compared to the sp mean level spacing @xmath8 . for simplicity , and because our aim here is not to compare with experimental data , we ignore angular momentum conservation , isospin , etc . after bethe s work , more accurate calculations of @xmath9 were made @xcite . schematic shell corrections related to a periodic fluctuation of the sp density were computed in @xcite ( see also @xcite ) , introducing the so called back - shifted bethe formula . several phenomenological modifications of eq.([bethe ] ) have been proposed to match the experimental results . these models take into account , for example , shell effects , pairing corrections and residual interactions @xcite , introducing a multitude of coexisting phenomenological parameterizations . the present status of the understanding do not allow to draw a clear theoretical picture of the functional dependence of @xmath9 with @xmath3 and @xmath0 . our purpose is , within a sp picture , to further develop the theoretical analysis to include several important effects that are missing in eq.([bethe ] ) . this presentation is based on the work reported in @xcite . there are different ways in which eq.([bethe ] ) can be improved . as mentioned before , it is the leading term of an expansion valid for a large number of particles @xmath3 and for high energies @xmath0 ( compared to the sp mean level spacing @xmath10 ) . for instance , assuming the gas is confined by a woods saxon like mean field potential , then @xmath11 where we used the approximation @xmath12 mev . this expression of @xmath2 is the leading order term of an expansion in decreasing powers of @xmath3 . corrections to it can be incorporated by considering lower order terms in the weyl series , that includes surface corrections , curvature corrections , etc @xcite . a similar expansion in decreasing powers of @xmath3 holds for the energy or mass of the nucleus ( the liquid drop formula ) . as is well known , the coefficient of each term in the expansion of the energy obtained from a fermi gas is not correct , since the renormalization due to interactions is out of the range of the model . similarly , the coefficient @xmath13 in eq.([afermi ] ) does not give a good description of the average trend of the experimental data obtained from neutron resonances , which is closer to @xmath14 ( cf fig.1 ) . although of great interest , we will not discuss these @xmath3 corrections and their renormalization due to interactions , but concentrate on two other corrections that are within the scope of a noninteracting model . the first one ( next section ) leads to smooth lower order terms in the excitation energy @xmath0 . the second type of corrections incorporates oscillatory terms in @xmath3 and @xmath0 . these oscillations , superimposed to the smooth @xmath15 trend mentioned above , are clearly visible in the experimental data shown in fig.1 . appearing in the fermi gas level density formula as a function of the mass number @xmath3 ( taken from ref.@xcite ) . ] the only property of the sp spectrum in eq.([bethe ] ) is the average density of levels @xmath2 . this approximation treats the sp spectrum as a continuum , and ignores the influence in the many body density of states of the discreteness of the spectrum and of the exact position of the sp levels . one step beyond , that incorporates the discreteness but ignores the fluctuations , is to consider a locally perfectly regular sp spectrum ( 1d harmonic oscillator ) , given by the sp energies @xmath16 , where @xmath17 is an arbitrary integer , and @xmath18 is the distance between neighboring levels . the fermi gas model consists of @xmath3 noninteracting fermions that occupy the equidistant levels , with occupation number 0 or 1 . in the ground state the particles fill the lowest @xmath3 sp states . we restrict the analysis to the approximation @xmath7 , in which the number of excited particles is small compared to the total number @xmath3 , and effects due to a finite number of particles can be ignored . in the excited states particles occupy sp levels above @xmath1 , thus creating holes in the fermi sea . the possible excited energies , measured with respect to the ground state energy , are @xmath19 , where @xmath20 is an arbitrary positive integer . the excitation energies are thus trivial . the nontrivial information comes from the degeneracy of each of theses energies , since there are many different many body configurations with the same excitation energy . the problem then is how to compute the degeneracy of each excited state of energy @xmath21 . ( next five columns ) of a gas of noninteracting fermions in a perfectly regular sp spectrum . each excited state is identify with a decomposition of the excitation energy 4 as a sum of integers ( bottom of each column ) . ] the solution is obtained by realizing that the problem we are facing is exactly equivalent to a well known problem in mathematics , i.e. in how many different ways an integer can be decomposed as a sum of integers ( partitions of an integer , @xmath22 ) . we will simply provide here a graphical illustration of this equivalence ( fig.2 ) . consider for example the case @xmath23 , the argument easily generalizes to any @xmath20 . @xmath23 can be decomposed as a sum of integers in 5 different ways ( and thus @xmath24 ) : @xmath25 , @xmath26 , @xmath27 , @xmath28 and @xmath29 . as fig.2 shows , each of these decompositions is in a one - to - one correspondence with one of the possible excited states of the fermi gas of total energy @xmath30 . the ground state is shown on the left column . in the second column from the left four particles have been pushed one level up , hence the equivalence with @xmath25 ( other mappings may be used as well ) . in the third column the upper particle was pushed up by two , the next two by one . etc . as has been emphasized many times , there is also a ( trivial ) one to one correspondence between the partitions of an integer and the excited states of a gas of noninteracting bosons in a one dimensional harmonic oscillator potential . in this case each integer in the decomposition of the excitation energy indicates the level occupied by one of the excited particles . the partition of an integer grows very fast with @xmath20 : it is 5 for @xmath23 , whereas @xmath31 ( this number was computed by hand by p. macmahon in 1918 ) . the generating function of @xmath22 was obtained in 1753 by l. euler . an explicit expression for @xmath22 came much later , under the form of an asymptotic ( exact ) formula ( initially obtained by hardy and ramanujan @xcite , improved and made rigorous later on by rademacher @xcite ) . to facilitate the comparison with more general approaches , we express the result in terms of @xmath32 , where here @xmath33 is the inverse of the exact spacing between neighboring levels . the first terms of the expansion are given by , @xmath34 we emphasize with the notation @xmath35 that this is the many body density of states for a fermi gas in a 1d harmonic oscillator spectrum . as expected , the first two terms of this equation reproduce bethe s formula . however , eq.([shr2 ] ) goes much further : it is a truncation of the _ exact _ asymptotic expansion of the density of states , whose precision can be improved by adding further terms . notice that all the terms in eq.([shr2 ] ) are smooth functions of the excitation energy @xmath0 . it is natural to ask about the relevance of eq.([shr2 ] ) in realistic systems , where the true sp spectrum consists of discrete energy levels arranged , in most cases , with no particular order . equation ( [ shr2 ] ) clearly goes beyond bethe s result , since it describes the exact behavior of the many body density of states of a gas of particles that occupy a discrete , perfectly regular arrangement of sp levels . it is therefore not unreasonable to expect that the result is valid for an arbitrary spectrum having the same average density @xmath2 , and that the effects of the fluctuations of the sp energy levels with respect to a perfectly regular arrangement come on top of it . although we do not have for the moment an explicit proof of this statement , numerical simulations to be presented below seem to confirm this hypothesis . the exact dependence of @xmath9 on @xmath3 and @xmath0 is sensitive to the detailed arrangement of the sp energy levels around the fermi energy . strong deviations with respect to a regular spacing , with possible degeneracies , may be produced by the presence of symmetries of the confining self consistent potential . these deviations induce , in turn , oscillations in the thermodynamic functions of the gas . shell effects are therefore due to deviations ( or bunching ) of the sp levels with respect to a perfectly regular spectrum . the degeneracies of the electronic levels of an atom produced by the rotational symmetry are an extreme manifestation of level bunching . in general , in systems that have other symmetries , or have no symmetries at all , there will still be level bunching , but its importance will typically be minor . therefore , depending on the presence or absence of symmetries , the shell effects may be more or less important . the level bunching , and more generally the fluctuations of the sp energy levels , are thus a very general phenomenon . the theories that describe those fluctuations make a neat distinction between systems with different underlying classical dynamics ( e.g. , regular or chaotic ) . our understanding of these fluctuations and of their connections with the regular or chaotic nature of the sp motion has greatly increased in the last decades . for a recent review see @xcite . oscillations in the many body density of states of nuclei are clearly visible in the experimental data available from neutron resonances as a function of the mass number @xmath3 . the parameter @xmath36 extracted from the data , plotted in fig.1 , shows oscillations around an average growth ( @xmath15 ) . the effective description of the oscillations through the parameter @xmath36 is an artifact of the analysis , since the fermi gas model relates @xmath36 to @xmath2 , a smooth function . as we will see below , the theoretical analysis leads to oscillatory corrections that enter as an additional term in the exponent of @xmath9 , and thus can not in general be interpreted as an effective @xmath36 . there are two distinct and important scales to describe the sp fluctuations . the first one , the smallest one , is the mean spacing between sp energy levels @xmath10 . the fluctuations of the sp energy levels on that scale have been shown to be , for a large class of systems , _ universal_. their statistical properties are described by uncorrelated sequences for integrable systems , and by random matrix theory in chaotic ones . the second relevant energy scale is @xmath37 , where @xmath38 is of the order of the time of flight across the system ( a precise definition will be given below ) . for @xmath3 large , the ratio @xmath39 is much larger than one ( for instance , @xmath40 in a three dimensional cavity ) . in contrast to the fluctuations on scales @xmath10 , the fluctuations on scales @xmath41 are long range modulations of the sp spectrum whose structure and amplitude are system specific . there is therefore no universality on this scale . for an arbitrary sp spectrum the computation of the density of excited states of a fermionic system is a difficult combinatorial problem for which no exact solution is known . what we are seeking here are the variations of the mb density of states due to fluctuations of the sp spectrum with respect to a perfectly ordered spectrum , described in the previous section . an explicit formula that includes these effects has been obtained recently @xcite . the result , obtained from a saddle point approximation of an inverse laplace transform of the mb density , takes the form @xcite @xmath42 \right\ } \ .\ ] ] the function @xmath43 is the fluctuating part of the energy of the gas at chemical potential @xmath44 and temperature @xmath45 , defined as follows . the energy is , as usual , given by @xmath46^{-1 } \ , \ ] ] where the function @xmath47 is the sp density of states defined by the sp energies @xmath48 @xmath49 the sp density of states is usually decomposed into a part @xmath2 that has a smooth dependence on the energy @xmath50 plus another contribution @xmath51 that describes the deviations with respect to the smooth behavior . the fluctuating part of the energy is defined as @xmath52^{-1 } \ .\ ] ] in eq . ( [ rhodef ] ) the arguments of @xmath53 are functions of @xmath3 and @xmath0 , @xmath54 and @xmath55 . these two functions are determined by inversion of the equations @xmath56 the function @xmath57^{-1}$ ] is the particle number . we are ignoring here ( small ) variations of the chemical potential with temperature . to lowest order , in which only the average behavior of the energy and of the particle number is taken into account , the second equation in ( [ spc ] ) leads to the usual relation between temperature and excitation energy @xmath58 and @xmath44 is simply a function of @xmath3 , @xmath59 . for instance , for a fermi gas in a 3d cavity of volume @xmath60 and particle mass @xmath20 , @xmath61 equation ( [ rhodef ] ) shows that , superimposed to the smooth growth of the density described by eq.([bethe ] ) , there are oscillatory corrections in the exponent of the density of states . these corrections are directly related to energy fluctuations of the system . the oscillatory nature of the new corrections is clearly displayed through a semiclassical theory , that expresses the quantum properties of the fermion gas in terms of classical solutions of the sp equations of motion . in this approach @xmath43 is written as a sum over the classical periodic orbits of the mean field potential @xcite @xmath62 each orbit is characterized by its action @xmath63 , stability amplitude @xmath64 , and maslov index @xmath65 . @xmath66 is the period of the periodic orbit , and @xmath67 is a temperature factor that introduces the time scale @xmath68 conjugate to the temperature . this expression describes the departures of @xmath69 with respect to its mean behavior due to the fluctuations of the sp spectrum . the behavior of @xmath53 strongly depends on whether @xmath44 or @xmath45 ( or @xmath3 or @xmath0 ) are varied . a temperature variation modifies the prefactors of the summands in @xmath53 ( through the function @xmath70 ) , and therefore produces gentle variations of the fluctuating part in the exponent of the density of states @xmath71 \ .\ ] ] in contrast , @xmath64 , @xmath66 and @xmath63 depend on @xmath44 ( and therefore on @xmath3 ) . for large values of @xmath44 , @xmath72 and the dominant variation with the particle number ( or any other parameter that modifies the actions ) comes from the argument of the cosine function in @xmath53 . rapid oscillations of @xmath73 are therefore generically expected when the number of particles is varied . noninteracting fermions in a rectangular billiard of sides @xmath74 and @xmath75 . the value of @xmath76 is 100 . dots : numerical computation at three different excitation energies ; solid curves : theoretical prediction ( [ hro ] ) ; dashed curves : smooth part eq.([shr2]).,width=321,height=264 ] for a given mean field potential there exists an infinite number of periodic orbits @xmath77 . the spectrum of periods @xmath66 has no upper bound . it has , however , a lower bound , given by the period @xmath38 of the shortest periodic orbit . since @xmath38 is usually the smallest characteristic time scale in the system , it determines the largest energy scale @xmath37 in which modulations ( bunching ) of the sp energy levels occur if the classical periodic orbits are known in a particular system , the corrections can be computed explicitly . the periodic orbit sum is dominated by the short orbits , and rapidly gives a good approximation of the result . to illustrate our findings we have tested some of the predictions by a direct numerical counting of the mb density of states in a particular system . figure 3 shows the results obtained for a gas of about 2000 fermions contained in a two dimensional rectangular cavity , an integrable system . for each number of particles we compute the mb density of states at three different temperatures , measured in units of @xmath78 the theoretical curve is computed according to the expression @xmath79 where @xmath73 is calculated from the periodic orbits of the rectangle . if the smooth corrections to the density of eq.([shr2 ] ) are not included , and only the result of bethe is used , a systematic deviation is observed in the average behavior of @xmath9 between theory and numerics . although the theoretical analysis of ref.@xcite does not include those corrections , the numerical results seem to indicate that their validity goes beyond the simple 1d harmonic oscillator spectrum . notice the extremely good accuracy of eq.([hro ] ) , either for the average value of the density as well as for the fluctuations . notice also that the typical size of the fluctuations can be quite important ( it is of the order of the average for the lowest excitation energy computed ) . equation ( [ hro ] ) decomposes the exponent of the density of states into a smooth part ( given by the hardy ramanujan result ) plus oscillatory contribution , @xmath80 , associated in a semiclassical theory to a sum over periodic orbits . @xmath80 relates the shell corrections in the mb density of states to the fluctuations of the energy of the gas . @xmath53 depends on temperature only through the function @xmath70 ( cf eq.([grandosc ] ) ) . the effect of this function is to exponentially suppress the contribution of periodic orbits whose period @xmath81 @xcite . since there is no suppression at @xmath82 ( because @xmath83 ) , only orbits whose period @xmath84 contribute to the difference @xmath85 . at temperatures such that @xmath86 , the term @xmath43 becomes exponentially small , and only @xmath87 remains . the shell correction @xmath80 therefore decays as @xmath88 at @xmath89 . this decay contrasts with the more common exponential damping observed in other thermodynamic quantities @xcite . a clearer picture of how the fluctuations behave may be obtained through a statistical analysis . the most simple property of the fluctuations is @xmath90 , where the brackets denote an average over a suitable chemical potential ( or particle number ) window . this result is valid only to first order in the expansion leading to eq.([rhodef ] ) @xcite ; it can be shown that higher order terms contribute to a non zero average . the next non trivial statistical property is the variance @xmath91 , that may be computed using eq.([grandosc ] ) . the result is @xmath92 ^ 2 /x^4 $ ] , where @xmath93 is the rescaled form factor of the sp spectrum ( cf eq.(36 ) in ref.@xcite ) . the latter function depends on the rescaled heisenberg time @xmath94 . it describes system dependent features for @xmath95 of the order of @xmath96 , while it is believed to be universal for @xmath97 . the universality class depends on the regular or chaotic nature of the dynamics , and on its symmetry properties . taking into account the basic properties of @xmath93 , in * chaotic systems * three different regimes for @xmath91 as a function of temperature are found @xcite ( remember the relation ( [ tq ] ) between @xmath45 and @xmath0 ) : * low temperatures @xmath98 . in this regime @xmath99 where @xmath100 is the value at @xmath101 of the constant @xmath102 ^ 2 \ .\ ] ] * intermediate temperatures @xmath103 . in this regime the size of the fluctuations saturates at a universal constant @xmath104 where @xmath105 and @xmath106 for systems with ( without ) time reversal invariance . * high temperatures @xmath89 . the size of the fluctuations decreases with excitation energy , @xmath107/ t^2 \ .\ ] ] after an exponential transient , a power law decay @xmath108 is thus obtained . the situation is different in * integrable systems * , where only two regimes are found . at low temperatures the result is identical to that of chaotic systems . the linear growth at low temperatures is thus totally universal and independent of the system . the difference is that in integrable systems the growth extends up to much higher temperatures , @xmath109 , without saturation . at that temperature the variance of the fluctuations is of order @xmath76 . in integrable systems , the maximum amplitude of the fluctuations is therefore reached at @xmath109 , and its typical size is much larger than in chaotic systems . at high temperatures @xmath89 the decay is almost identical to that of chaotic systems ( the coefficient 8 is replaced by a 12 ) . a schematic representation of the temperature dependence of the variance of the fluctuations for chaotic and integrable systems is given in fig.4 . . ] it has been shown @xcite that @xmath43 is dominated , at any @xmath45 , by the shortest classical periodic orbits . in contrast , the difference @xmath110 depends on orbits whose period @xmath84 . for temperatures @xmath111 the statistical properties of these orbits are universal , and correspondingly the probability distribution function of @xmath73 is expected to be universal , in the sense that at a given temperature it should only depend on the nature of the underlying classical dynamics ( regular or chaotic ) , and the symmetries of the system . this statement is supported by the fact that in the limit @xmath112 , @xmath113 . the probability distribution of the latter quantity was studied in ref.@xcite ; it was shown that it coincides at low temperatures with that obtained from a poisson spectrum for integrable systems and from a random matrix spectrum for chaotic ones . this confirms the universality of the distribution at low temperatures . as the temperature is raised , the universality of the probability distribution of @xmath73 will be lost for temperatures of the order or greater than @xmath41 , where system specific features ( i.e. , short periodic orbits ) are revealed . in this respect , the decay ( [ htr ] ) is only indicative . its exact form depends on the details of the short periodic orbits spectrum , treated here only through a rough approximation . two improvements with respect to bethe s formula of @xmath9 have been discussed . the first one incorporates the discreteness of the sp spectrum by considering a set of equidistant sp states ( 1d harmonic oscillator ) . the exact many body density of states is obtained by mapping the problem to the computation of the number of decompositions of an integer as a sum of integers . the exact formula adds finite@xmath0 corrections to the asymptotic result . although a proof is missing , numerical simulations suggest that the range of validity of this result extends beyond the 1d harmonic oscillator spectrum to any spectrum , in the sense that it describes with good accuracy the smooth part of the growth of @xmath9 in systems with an arbitrary sp spectrum ( cf fig.3 ) . beyond the smooth behavior , the second improvement concerns the fluctuations or shell effects in @xmath9 . in a semiclassical theory , each periodic orbit has been shown to contribute to @xmath114 with a fluctuating term as a function of the number of particles @xmath3 , of wavelength @xmath115 at low excitation energies long orbits contribute to the fluctuations while short ones are exponentially suppressed , leading to wild oscillations and universality of their statistical properties . as the temperature increases , the situation is reversed , long orbits are exponentially suppressed while short orbits come into play . at @xmath116 , oscillations of wavelength of order @xmath117 are predicted . these oscillations are clearly visible , with the correct wavelength , in fig.3 . with the raise of the short periodic orbits as @xmath45 increases , the universality of the statistical properties of the fluctuations disappears for @xmath45 of the order or higher than @xmath118 . concerning the typical size of the oscillations , a nontrivial dependence as a function of excitation energy ( or temperature ) was found . the results are summarized in fig.4 . the behavior is quite different in chaotic and integrable systems . at low temperatures the variance grows linearly with @xmath45 in all cases . however , a plateau is rapidly reached at @xmath119 in chaotic systems , whereas the amplitude of the fluctuations continue to grow in regular ones . at @xmath116 the differences reaches its maximum amplitude , where the variance in regular systems is @xmath76 times larger than in chaotic ones ( remember , moreover , that @xmath120 in a three dimensional cavity , so that the difference increases with an increasing number of particles ) . at this point the situation is similar to what was found for the ratio of the variance of the fluctuations of the nuclear mass due to regular and chaotic motion @xcite . at @xmath89 the variance decreases as @xmath121 for both types of dynamics . the validity of the results discussed here in the analysis and interpretation of the experimental data on the nuclear level density will be presented elsewhere . 9 h. a. bethe , _ phys . _ , * 50 * , 332 ( 1936 ) . c. bloch , _ phys _ , * 93 * , 1094 ( 1954 ) . n. rosenzweig , _ phys _ , * 108 * , 817 ( 1957 ) . a. bohr and b. r. mottelson , _ nuclear structure _ , vol.i , benjamin , reading , massachusetts , 1969 . a. gilbert and a. g. w. cameron , _ can . _ , * 43 * , 1446 ( 1965 ) . a. v. ignatyuk , g. n. smirenkin , and a. s. tishin , _ sov . _ , * 21 * , 255 ( 1975 ) . k. kataria , v. s. ramamurthy , and s. s. kapoor , _ phys . _ c , * 18 * , 549 ( 1978 ) . y. alhassid , g. f. bertsch , and l. fang , _ phys . _ c , * 68 * , 044322 ( 2003 ) . p. leboeuf , a. g. monastra and a. relao , _ phys . _ , * 94 * , 102502 ( 2005 ) . h. p. baltes and e. r. hilf : _ spectra of finite systems _ , bibliographisches institut - wissenschaftsverlag , mannheim , 1976 . g. h. hardy and s. ramanujan , _ proc . london math . * 17 * , 75 ( 1918 ) . h. rademacher , _ proc . london math . _ , * 43 * , 241 ( 1937 ) . p. leboeuf , _ regularity and chaos in the nuclear masses _ , lectures delivered at the viii hispalensis international summer school , sevilla , spain , june 2003 ( to appear in _ lecture notes in physics _ , springer verlag , eds . j. m. arias and m. lozano ) ; nucl - th/0406064 . v. m. strutinsky and a. g. magner , sov . * 7 * ( 1976 ) 138 . p. leboeuf and a. g. monastra , _ ann . _ , * 297 * , 127 ( 2002 ) . k. richter , d. ullmo and r. jalabert , _ phys . rep . _ , * 276 * , 1 ( 1996 ) . o. bohigas and p. leboeuf , _ phys . _ , * 88 * , 092502 ( 2002 ) ; * 88 * , 129903 ( 2002 ) .	it has been shown by h. bethe more than 70 years ago that the number of excited states of a fermi gas grows , at high excitation energies @xmath0 , like the exponential of the square root of @xmath0 . this result takes into account only the average density of single particle ( sp ) levels near the fermi energy . it ignores two important effects , namely the discreteness of the sp spectrum , and its fluctuations . we show that the discreteness of the sp spectrum gives rise to smooth finite@xmath0 corrections . mathematically , these corrections are associated to the problem of partitions of an integer . on top of the smooth growth of the many body density of states there are , generically , oscillations . an explicit expression of these oscillations is given . their properties strongly depend on the regular or chaotic nature of the sp motion . in particular , we analyze their typical size , temperature dependence and probability distribution , with emphasis on their universal aspects . address = laboratoire de physique thorique et modles statistiques , bt . 100 , + universit de paris - sud , 91405 orsay cedex , france
during the first decade of the third millennium we have begun to put observational constraints on the status quo of galaxy formation at roughly one billion years after the big bang ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? statistical samples of star - forming galaxies at @xmath9 either selected on the basis of their large ( _ i__z _ ) color due to the lyman break redshifted to @xmath0 ( _ i - dropouts _ ) , or on the basis of the large equivalent width of ly@xmath3 emission ( ly@xmath3 emitters ) suggest that they are analogous to the population of lyman break galaxies ( lbgs ) found at @xmath10 ( e.g. * ? ? ? a small subset of the @xmath2-dropouts has been found to be surprisingly massive or old @xcite . the slope of the uv luminosity function at @xmath9 is very steep and implies that low luminosity objects contributed significantly to reionizing the universe @xcite . cosmological hydrodynamic simulations are being used to reproduce the abundances as well as the spectral energy distributions of @xmath9 galaxies . exactly how these objects are connected to local galaxies remains a highly active area of research ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? the discovery of highly luminous quasi - stellar objects ( qsos ) at @xmath0 ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) is of equal importance in our understanding of the formation of the first massive black holes and galaxies . @xcite absorption troughs in their spectra demarcate the end of the epoch of reionization ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? assuming that high redshift qsos are radiating near the eddington limit , they contain supermassive black holes ( smbhs ) of mass @xmath11 @xmath12 ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? the spectral properties of most @xmath0 qsos in the rest - frame uv , optical , ir and x - ray are similar to those at low redshift , suggesting that massive , and highly chemically enriched galaxies were vigorously forming stars and smbhs less than one billion years after the big bang ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? hierarchical formation models and simulations can reproduce the existence of such massive objects at early times ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , provided however that they are situated in extremely massive halos . large - scale gravitational clustering is a powerful method for estimating halo masses of quasars at low redshifts , but can not be applied to @xmath0 qsos because there are too few systems known . their extremely low space density determined from the sloan digital sky survey ( sdss ) of @xmath131 gpc@xmath14 ( comoving ) implies a ( maximum ) halo mass of @xmath15@xmath16 @xmath12 @xcite . a similar halo mass is obtained when extrapolating from the ( @xmath8 ) relationship between black hole mass and bulge mass of @xcite , and using @xmath17 @xcite . because the descendants of the most massive halos at @xmath0 may evolve into halos of @xmath18 @xmath12 at @xmath8 in a @xmath19cdm model , ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * but see @xcite , @xcite and sect . [ sec : qso ] of this paper ) , it is believed that the qsos trace highly biased regions that may give birth to the most massive present - day galaxy clusters . if this is true , the small - scale environment of @xmath0 qsos may be expected to show a significant enhancement in the number of small , faint galaxies . these galaxies may either merge with the qso host galaxy , or may form the first stars and black holes of other ( proto-)cluster galaxies . observations carried out with the advanced camera for surveys ( acs ) on the _ hubble space telescope _ ( hst ) , allowed a rough measurement of the two - dimensional overdensities of faint @xmath2-dopouts detected towards the qsos j0836 + 0054 at @xmath20 @xcite and j1030 + 0524 at @xmath21 @xcite . recently @xcite presented results from a sample of 5 qso fields , finding some to be overdense and some to be underdense with respect to the hst / acs great observatories origins deep survey ( goods ) . @xcite find enhancements in the number counts of sub - mm galaxies . substantial overdensities of @xmath5-dropouts and ly@xmath3 emitters have also been found in non - qso fields ( e.g. * ? ? ? * ; * ? ? ? * ) , suggesting that massive structures do not always harbour a qso , which may be explained by invoking a qso duty - cycle . at @xmath22 , significant excesses of star - forming galaxies have been found near qsos ( e.g. * ? ? ? * ; * ? ? ? * ) , radio galaxies ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , and in random fields @xcite . although the physical interpretation of the measurements is uncertain , these structures are believed to be associated with the formation of clusters of galaxies . the idea of verifying the presence of massive structures at high redshift through the clustering of small galaxies around them has recently been explored by , e.g. , @xcite using the excursion set formalism of halo growth @xcite . however , the direct comparison between models or simulations and observations remains difficult , mainly because of complicated observational selection effects . this is especially true at high redshift . in order to investigate how a wide variety of galaxy overdensities found in surveys at @xmath23 are related to cluster formation , we have carried out an analysis of the progenitors of galaxy clusters in a set of cosmological @xmath24-body simulations . our results will be presented in a series of papers . in paper i , we use the milennium run simulations @xcite to simulate a large mock survey of galaxies at @xmath0 to derive predictions for the properties of the progenitors of massive galaxy clusters , paying particular attention to the details of observational selection effects . we will try to answer the following questions : + ( i ) where do we find the present - day descendants of the _ i_-dropouts ? \(ii ) what are the typical structures traced by _ i_-dropouts and ly@xmath3 emitters in current surveys , and how do they relate to protoclusters ? \(iii ) how do we unify the ( lack of excess ) number counts observed in qso fields with the notion that qsos are hosted by the most massive halos at @xmath0 ? + the structure of the present paper is as follows . we describe the simulations , and construction of our mock _ i - dropout _ survey in section 2 . using these simulations , we proceed to address the main questions outlined above in sections 35 . we conclude the paper with a discussion ( section 6 ) , an overview of recommendations for future observations ( section 7 ) , and a short summary ( section 8) of the main results . we use the semi - analytic galaxy catalogues that are based on the millennium run ( mr ) dark matter simulation of @xcite . detailed descriptions of the simulations and the semi - analytic modeling have been covered extensively elsewhere , and we kindly refer the reader to those works for more information ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * and references therein ) . the dark matter simulation was performed with the cosmological simulation code gadget2 @xcite , and consisted of @xmath25 particles of mass @xmath26 @xmath27 @xmath12 in a periodic box of 500 @xmath27 mpc on a side . the simulations followed the gravitational growth as traced by these particles from @xmath28 to @xmath8 in a @xmath19cdm cosmology ( @xmath29 , @xmath30 , @xmath31 , @xmath32 , @xmath33 ) consistent with the wmap year 1 data @xcite . the results were stored at 64 epochs ( `` snapshots '' ) , which were used to construct a detailed halo merger tree during postprocessing , by identifying all resolved dark matter halos and subhalos , and linking all progenitors and descendants of each halo . redshift versus the ( co - moving ) @xmath34-coordinate for all objects within a slice of width @xmath35 @xmath27 mpc along the @xmath36-axis . ] galaxies were modeled by applying semi - analytic prescriptions of galaxy formation to the stored halo merger trees . the techniques and recipes include gas cooling , star formation , reionizaton heating , supernova feedback , black hole growth , and `` radio - mode '' feedback from galaxies with a static hot gas atmosphere , and are described in @xcite . the photometric properties of galaxies are then modeled using stellar population synthesis models , including a simple dust model . here we use the updated models ` delucia2006a ' of @xcite that have been made publicly available through an advanced database structure on the mr website @xcite . the simulation box showing the positions in co - moving coordinates of all objects identified as @xmath2-dropout galaxies to @[email protected] mag . ] we used the discrete mr snapshots to create a large , continous mock survey of @xmath2-dropout galaxies at @xmath0 . the general principle of transforming a series of discrete snapshots into a mock pencil beam survey entails placing a virtual observer somewhere in the simulation box at @xmath8 and carving out all galaxies as they would be observed in a pencil beam survey along that observer s line of sight . this technique has been described in great detail in @xcite and @xcite . in general , one starts with the snapshot @xmath39 at @xmath8 and records the positions , velocities and physical properties of all galaxies within the cone out to a comoving distance corresponding to that of the next snapshot . for the next segment of the cone , one then use the properties as recorded in snapshot @xmath40 , and so on . the procedure relies on the reasonable assumption that the large - scale structure ( positions and velocities of galaxies ) evolves relatively slowly between snapshots . by replicating the simulation box along the lightcone and limiting the opening angle of the cone , one can in principle construct unique lightcones out to very high redshift without crossing any region in the simulation box more than once . the method is straightforward when done in comoving coordinates in a flat cosmology using simple euclidean geometry @xcite . because the comoving distances or redshifts of galaxies recorded at a particular snapshot do not correspond exactly to their effective position along the lightcone , we need to correct their magnitudes by interpolating over redshift as follows : @xmath41=m(z_i)+\frac{dm}{dz}[z(d)-z_i)],\ ] ] where @xmath42 $ ] is the observer - frame absolute magnitude at the observed redshift , @xmath43 ( including peculiar velocities along the line of sight ) , @xmath44 is the magnitude at redshift @xmath45 corresponding to the @xmath5th snapshot , and @xmath46 is the first order derivative of the observer - frame absolute magnitude . the latter quantity is calculated for each galaxy by placing it at neighbouring snapshots , and ensures that the @xmath47-correction is taken into account @xcite . finally , we apply the mean attenuation of the intergalactic medium using @xcite and calculate the observer - frame apparent magnitudes in each filter . in this paper , we use the fact that the selection of @xmath0 galaxies through the _ i - dropout _ technique is largely free of contamination from objects at lower ( and higher ) redshift @xcite provided that the observations are deep enough . because the transverse size of the mr simulation box ( 500 @xmath27 mpc ) corresponds to a comoving volume between @xmath48 and @xmath49 ( the typical redshift range of @xmath5-dropouts surveys ) we can use the three simulation snapshots centered at @xmath50 , @xmath51 and @xmath52 to create a mock survey spanning this volume , while safely neglecting objects at other redshifts . we extracted galaxies from the mr database by selecting the @xmath53axis of the simulation box to lie along the line - of - sight of our mock field . in order to compare with the deepest current surveys , we calculated the apparent magnitudes in the hst / acs @xmath54 , @xmath2 and @xmath37 filters and the 2mass @xmath55 , @xmath56 and @xmath57 filters . we derived observed redshifts from the comoving distance along the line of sight ( including the effect of peculiar velocities ) , applied the @xmath47-corrections and igm absorption , and calculated the apparent magnitudes in each band . [ fig : zx ] shows the spatial @xmath34-coordinate versus the redshift of objects in the simulated lightcone . [ fig : xy ] shows the entire simulated volume projected along the @xmath58- or redshift axis . these figures show that there exists significant filamentary and strongly clustered substructure at @xmath59 , both parallel and perpendicular to the line - of - sight . our final mock survey has a comoving volume of @xmath60 gpc@xmath61 , and spans an area of @xmath62 when projected onto the sky . it contains @xmath63 galaxies at @xmath64 with @xmath65 mag ( corresponding to an absolute magnitude ] of @xmath66 mag , about one mag below @xmath67 ) . for comparison and future reference , we list the main @xmath5-dropout surveys together with their areal coverage and detection limit in table [ tab : surveys ] . in the left panel of fig . [ fig : color ] we show the @xmath68 vs. @xmath69 colour - colour diagram for all objects satisfying @xmath70 . the @xmath5-dropouts populate a region in colour - colour space that is effectively isolated from lower redshift objects using a simple colour cut of @xmath71 . note that although our simulated survey only contains objects at @xmath72 , it has been shown @xcite that this colour cut is an efficient selection criterion for isolating starburst galaxies at @xmath0 with blue @xmath37@xmath73 colours ( see right panel of fig . [ fig : color ] ) . for reference , we have overplotted colour tracks for a 100 myr old , continuous star formation model as a function of redshift ; different colour curves show results for different amounts of reddening by dust . as can be seen , these simple models span the region of colour - colour space occupied by the mr galaxies . at @xmath74 , galaxies occupy a tight sequence in the plane . at @xmath75 , objects fan out because the @xmath68 colour changes strongly as a function of redshift , while the @xmath69 colour is more sensitive to both age and dust reddening . because of the possibility of intrinsically red interlopers at @xmath76 , the additional requirement of a non - detection in @xmath54 , or a very red @xmath77 colour , if available , is often included in the selection@xmath2@xmath37 filter system in order to compare with the deepest surveys available in literature . other works based on groundbased data commonly use the sdss - based @xmath78 filterset , but the differences in colours are minimal . ] . because the selection based on @xmath79 introduces a small bias against objects having strong ly@xmath3 emission at @xmath80 @xcite , we have statistically included the effect of ly@xmath3 on our sample selection by randomly assigning ly@xmath3 with a rest - frame equivalent width of 30 to 25% of the galaxies in our volume , and recalculating the @xmath69 colours . the inclusion of ly@xmath3 leads to a reduction in the number of objects selected of @xmath81% ( see also * ? ? ? * ) . in table [ tab : surfdens ] we list the surface densities of @xmath2-dropouts selected in the mr mock survey as a function of limiting @xmath37-magnitude and field size . for comparison , we calculated the surface densities for regions having areas comparable to some of the main @xmath2-dropout surveys : the sdf ( 876 arcmin@xmath82 ) , two goods fields ( 320 arcmin@xmath82 ) , a single goods field ( 160 arcmin@xmath82 ) , and a hudf - sized field ( 11.2 arcmin@xmath82 ) . the errors in table [ tab : surfdens ] indicate the @xmath83 deviation measured among a large number of similarly sized fields selected from the mock survey , and can be taken as an estimate of the influence of ( projected ) large - scale structure on the number counts ( usually referred to as `` cosmic variance '' or `` field - to - field variations '' ) . at faint magnitudes , the strongest observational constraints on the @xmath2-dropout density come from the hst surveys . our values for a goods - sized survey are 105 , 55 and 82% of the values given by the most recent estimates by b07 for limiting @xmath37 magnitudes of 27.5 , 26.5 and 26.0 mag , respectively , and consistent within the expected cosmic variance allowed by our mock survey . because the total area surveyed by b06 is about 200@xmath84 smaller than our mock survey , we also compare our results to the much larger sdf from @xcite . at @xmath85 mag the number densities of @xmath130.18 arcmin@xmath86 derived from both the real and mock surveys are perfectly consistent . last , we note that in order to achieve agreement between the observed and simulated number counts at @xmath0 , we did not require any tweaks to either the cosmology ( e.g. , see * ? ? ? * for the effect of different wmap cosmologies ) , or the dust model used ( see * ? ? ? * ; * ? ? ? * for alternative dust models better tuned to high redshift galaxies ) . this issue may be further investigated in a future paper . redshift histograms derived from the mr mock @xmath5-dropout survey at the depth of @xmath37=27.5 mag using the selection criteria @xmath2@xmath37@xmath87 ( thick solid line , error bars indicate the 1@xmath88 scatter expected among goods - sized fields ) , @xmath2@xmath37@xmath89 ( dashed line ) , @xmath2@xmath37@xmath90 ( blue line ) , and @xmath2@xmath37@xmath91 ( red line ) . the thin solid line indicates the model redshift distribution from b06 based on the hudf . ] in fig . [ fig : nz2 ] we show the redshift distribution of the full mock survey ( thick solid line ) , along with various subsamples selected according to different @xmath2@xmath37 colour cuts that we will refer to later on in this paper . the standard selection of @xmath2@[email protected] results in a distribution that peaks at @xmath93 . we have also indicated the expected scatter resulting from cosmic variance on the scale of goods - sized fields ( error bars are 1@xmath88 ) . some , mostly groundbased , studies make use of a more stringent cut of @xmath2@xmath37@xmath89 to reduce the chance of foreground interlopers ( dashed histogram ) . other works have used colour cuts of @xmath2@xmath37@xmath94 ( blue histogram ) and @xmath2@xmath37@xmath95 ( red histogram ) in order to try to extract subsamples at @xmath80 and @xmath96 , respectively . as can be seen in fig . [ fig : nz2 ] , such cuts are indeed successful at broadly separating sources from the two redshift ranges , although the separation is not perfectly clean due to the mixed effects of age , dust and redshift on the @xmath2@xmath37 colour . for reference , we have also indicated the model redshift distribution from b06 ( thin solid line ) . this redshift distribution was derived for a much fainter sample of @xmath37@xmath9729 mag , which explains in part the discrepancy in the counts at @xmath98 . evolution across the redshift range will furthermore skew the actual redshift distribution toward lower values ( see discussion in * ? ? ? * ) . this is not included in the b06 model , and its effect is only marginally taken into account in the mr mock survey due to the relatively sparse snapshot sampling across the redshift range . unfortunately , the exact shape of the redshift distribution is currently not very well constrained by spectroscopic samples @xcite . a more detailed analysis is beyond the scope of this paper , and we conclude by noting that the results presented below are largely independent of the exact shape of the distribution . physical properties of @xmath5-dropouts in the mr mock survey satisfying @xmath99 mag . we plot the cumulative fractions of galaxies with stellar masses , star formation rates , stellar ages and halo masses greater than a given value . _ top left : _ distribution of stellar masses . the median stellar mass is @xmath100 @xmath12 h@xmath101 ( dotted line ) . _ top right : _ distribution of sfrs . the median sfr is @xmath102 @xmath12 yr@xmath101 ( dotted line ) . _ bottom left : _ distribution of mass - weighted ages . the median age is @xmath13160 myr ( dotted line ) . _ bottom right : _ distribution of halo masses . the median halo mass is @xmath103 @xmath12 h@xmath101 ( dotted line ) . ] although a detailed study of the successes and failures in the semi - analytic modeling of galaxies at @xmath0 is not the purpose of our investigation , we believe it will be instructive for the reader if we at least summarize the main physical properties of the model galaxies in our mock survey . unless stated otherwise , throughout this paper we will limit our investigations to @xmath5-dropout samples having a limiting magnitude of @xmath37=26.5 mag magnitude of @xmath10426.5 mag for an unattenuated galaxy at @xmath105 would correspond to a sfr of @xmath1047 @xmath12 yr@xmath101 , under the widely used assumption of a 0.1125@xmath12 salpeter initial mass function and the conversion factor between sfr and the rest - frame 1500 uv luminosity of @xmath106 erg s@xmath101 hz@xmath101 / @xmath12 yr@xmath101 as given by @xcite . ] , comparable to @xmath107 at @xmath9 ( see * ? ? ? this magnitude typically corresponds to model galaxies situated in dark matter halos of at least 100 dark matter particles ( @xmath108 m@xmath109 h@xmath101 ) . this ensures that the evolution of those halos and their galaxies has been traced for some time prior to the snapshot from which the galaxy was selected . in this way , we ensure that the physical quantities derived from the semi - analytic model are relatively stable against snapshot - to - snapshot fluctuations . a magnitude limit of @xmath37=26.5 mag also conveniently corresponds to the typical depth that can be achieved in deep groundbased surveys or relatively shallow hst - based surveys . number density versus halo mass of the @xmath8 dark matter halos hosting descendants of @xmath5-dropouts at @xmath0 to a limiting depth of @[email protected] mag . the median @xmath5-dropout descendant halo mass is a few times @xmath111 @xmath12 ( dotted line ) . the halo mass function of all mr halos at @xmath8 is indicated by the dashed line . the mass range occupied by the halos associated with galaxy clusters is indicated by the hatched region . ] in fig . [ fig : z6props ] we plot the cumulative distributions of the stellar mass ( top left ) , sfr ( top right ) , and stellar age ( bottom left ) of the @xmath2-dropouts in the mock survey . the median stellar mass is @xmath112 @xmath12 h@xmath101 , and about 30% of galaxies have a stellar mass greater than @xmath113 @xmath12 . the median sfr and age are @xmath102 @xmath12 yr@xmath101 and @xmath13160 myr , respectively , with extrema of @xmath114 @xmath12 yr@xmath101 and @xmath13400 myr . these results are in general agreement with several studies based on modeling the stellar populations of limited samples of @xmath2-dropouts and ly@xmath3 emitters for which deep observations with _ hst _ and _ spitzer _ exist . @xcite have analyzed a statistically robust sample and find stellar masses ranging from @xmath115 @xmath12 for irac - undetected sources to @xmath116 @xmath12 for the brightest 3.6@xmath117 m sources , and ages ranging from @xmath11840 to 400 myr ( see also * ? ? ? * ; * ? ? ? * ; * ? ? ? * for additional comparison data ) . we also point out that the maximum stellar mass of @xmath116 @xmath12 found in our mock survey ( see top left panel ) is comparable to the most massive @xmath5-dropouts found , and that `` supermassive '' galaxies having masses in excess of @xmath119 @xmath12 are absent in both the simulations and observations @xcite . last , in the bottom right panel we show the distribution of the masses of the halos hosting the model @xmath5-dropouts . the median halo mass is @xmath120 @xmath12 . our results are in the range of values reported by @xcite and @xcite based on the angular correlation function of large @xmath2-dropout samples , but we note that halo masses are currently not very well constrained by the observations . number density versus stellar mass of the galaxies at @xmath8 that have at least one @xmath5-dropout progenitor at @xmath0 . the median descendant mass is @xmath108 @xmath12 ( dotted line ) . the distribution of stellar mass of all mr @xmath8 galaxies is indicated for comparison ( dashed line ) . ] in this section we study the relation between local overdensities in the @xmath5-dropout distribution at @xmath0 and the sites of cluster formation . throughout this paper , a galaxy cluster is defined as being all galaxies belonging to a bound dark matter halo having a dark matter mass@xmath121 , is the mass within the radius at which the halo has an overdensity corresponding to the value at virialisation in the top - hat collapse model ( see * ? ? ? * ) . ] of @xmath15@xmath122 h@xmath101 @xmath12 at @xmath8 . in the mr we find 2,832 unique halos , or galaxy clusters , fulfilling this condition , 21 of which can be considered supermassive ( @xmath15@xmath123 h@xmath101 @xmath12 ) . furthermore , a proto - cluster galaxy is defined as being a galaxy at @xmath124 that will end up in a galaxy cluster at @xmath8 . note that these are trivial definitions given the database structure of the millennium run simulations , in which galaxies and halos at any given redshift can be related to their progenitors and descendants at another redshift @xcite . in fig . [ fig : z0halos ] we plot the distribution of number densities of the central halos that host the @xmath8 descendants of the @xmath5-dropouts in our mock survey as a function of the halo mass . the median halo mass hosting the @xmath5-dropout descendants at @xmath8 is @xmath125 @xmath12 h@xmath101 ( dotted line ) . for comparison we indicate the mass distribution of all halos at @xmath8 ( dashed line ) . the plot shows that the fraction of halos that can be traced back to a halo hosting an @xmath5-dropout at @xmath0 is a strong function of the halo mass at @xmath8 . 45% of all cluster - sized halos at @xmath8 ( indicated by the hatched region ) are related to the descendants of halos hosting @xmath5-dropouts in our mock survey , and 77% of all clusters at @xmath8 having a mass of @xmath15@xmath126 @xmath12 h@xmath101 can be traced back to at least one progenitor halo at @xmath0 hosting an @xmath5-dropout . this implies that the first seeds of galaxy clusters are already present at @xmath0 . in addition , many @xmath5-dropout galaxies and their halos may merge and end up in the same descendant structures at @xmath8 , which was not accounted for in our calculation above where we only counted unique halos at @xmath8 . in fact , about @xmath1334% ( @xmath132% ) of all @xmath5-dropouts ( @[email protected] ) in the mock survey will end up in clusters of mass @xmath127 ( @xmath126 ) @xmath12 h@xmath101 at @xmath8 . this implies that roughly one third of all galaxies one observes in a typical @xmath5-dropout survey can be considered `` proto - cluster '' galaxies . the plot further shows that the majority of halos hosting @xmath5-dropouts at @xmath0 will evolve into halos that are more typical of the group environment . this is similar to the situation found for lyman break or dropout galaxies at lower redshifts @xcite . in fig . [ fig : z0galaxies ] we plot the stellar mass distribution of those @xmath8 galaxies that host the descendants of the @xmath5-dropouts . the present - day descendants are found in galaxies having a wide range of stellar masses ( @xmath15@xmath128 @xmath12 ) , but the distribution is skewed towards the most massive galaxies in the mr simulations . the median stellar mass of the descendants is @xmath108 @xmath12 ( dotted line in fig . [ fig : z0galaxies ] ) . we will now focus on to what extent local overdensities in the @xmath5-dropout distribution at @xmath59 may trace the progenitor seeds of the richest clusters of galaxies in the present - day universe . in fig . [ fig : field ] we plot the sky distribution of the @xmath5-dropouts in our @xmath62 mr mock survey ( large and small circles ) . large circles indicate those @xmath5-dropouts identified as proto - cluster galaxies . we have plotted contours of @xmath5-dropout surface density , @xmath129 , @xmath130 and @xmath131 being the local and mean surface density measured in circular cells of 5 radius . negative contours representing underdense regions are indicated by blue lines , while positive contours representing overdense regions are indicated by red lines . the green dashed lines indicate the mean density . the distribution of proto - cluster galaxies ( large circles ) correlates strongly with positive enhancements in the local @xmath5-dropout density distribution , indicating that these are the sites of formation of some of the first clusters . in fig . [ fig : npointing ] we plot the frequency distribution of the @xmath5-dropouts shown in fig . [ fig : field ] , based on a counts - in - cells analysis of 20,000 randomly placed acs - sized fields of @xmath132 ( solid histograms ) . on average , one expects to find about 2 @xmath5-dropouts in a random acs pointing down to a depth of @[email protected] , but the distribution is skewed with respect to a pure poissonian distribution as expected due to the effects of gravitational clustering . the poissonian expectation for a mean of 2 @xmath5-dropouts is indicated by a thin line for comparison . the panel on the right shows a zoomed - in view to give a better sense of the small fraction of pointings having large numbers of @xmath5-dropouts . also in fig . [ fig : npointing ] we have indicated the counts histogram derived from a similar analysis performed on @xmath5-dropouts extracted from the goods survey using the samples of b06 . the goods result is indicated by the dotted histogram , showing that it lies much closer to the poisson expectation than the mr mock survey . this is of course expected as our mock survey covers an area over 200@xmath84 larger than goods and includes a much wider range of environments . to illustrate that the ( small ) fraction of pointings with the largest number of objects is largely due to the presence of regions associated with proto - clusters , we effectively `` disrupt '' all protoclusters by randomizing the positions of all protocluster galaxies and repeat the counts - in - cells calculation . the result is shown by the dashed histograms in fig . [ fig : npointing ] . the excess counts have largely disappeared , indicating that they were indeed due to the proto - clusters . the counts still show a small excess over the poissonian distribution due to the overall angular clustering of the @xmath5-dropout population . counts - in - cells frequency distribution of the @xmath5-dropouts shown in fig . [ fig : field ] , based on 20,000 randomly placed acs - sized fields of @xmath132 . the panel on the right shows a zoomed - in view to give a better sense of the small fraction of pointings having large numbers of @xmath5-dropouts . in both panels , the thick solid line indicates the frequency distribution of the full mr mock survey . the dashed line indicates how the distribution changes if we `` disrupt '' all protocluster regions of fig . [ fig : field ] by randomizing the positions of the galaxies marked as proto - cluster galaxy . the dotted line indicates the frequency distribution of a large sample of @xmath5-dropouts selected from the goods survey by b06 using identical selection criteria . thin solid lines indicate the poisson distribution for a mean of 2 @xmath5-dropouts per pointing . ] panels show the cumulative probability distributions of finding regions having a surface overdensity @xmath133 of @xmath5-dropouts for the four samples extracted from the mr mock survey based on colour cuts of @xmath2@xmath37@xmath87 ( top - left ) , @xmath2@xmath37@xmath89 ( top - right ) , @xmath134@xmath2@xmath37@xmath90 ( bottom - left ) , and @xmath2@xmath37@xmath91 ( bottom - right ) . the inset plots show the full probability distributions . dashed , coloured lines indicate the joint probability of finding cells having an overdensity @xmath133 and those cells consisting of at least 25% ( blue ) , 50% ( green ) and 75% proto - cluster galaxies . ] we can use our counts - in - cells analysis to predict the cumulative probability , @xmath135 , of randomly finding an @xmath5-dropout overdensity equal or larger than @xmath136 . the resuls are shown in fig . [ fig : pcumul ] . the four panels correspond to the subsamples defined using the four different @xmath2@xmath37 colour cuts ( see 2.5 and fig . [ fig : nz2 ] ) . panel insets show the full probability range for reference . the figure shows that the probability of finding , for example , cells having a surface overdensity of @xmath5-dropouts of @xmath1373 is about half a percent for the @xmath2@[email protected] samples ( top left panel , solid line ) . the other panels show the dependence of @xmath135 on @xmath5-dropout samples selected using different colour cuts . as the relative contribution from for- and background galaxies changes , the density contrast between real , physical overdensities on small scales and the `` field '' is increased . the average fraction of @xmath5-dropouts marked as proto - cluster galaxies contained in acs - sized cells as a function of cell overdensity . error bars are @xmath138 . there is a clear trend showing that larger surface overdensities are associated with a larger contribution from galaxies in proto - clusters , albeit with significant scatter . ] the results presented in fig . [ fig : pcumul ] provide us with a powerful way to interpret many observational findings . specifically , overdensities of @xmath5-dropouts have been interpreted as evidence for large - scale structure associated with proto - clusters , at least qualitatively . although fig . [ fig : pcumul ] tells us the likelihood of finding a given overdensity , this is not sufficient by itself to answer the question whether that overdensity is related to a proto - cluster due to a combination of several effects . first , because we are mainly working with photometrically selected samples consisting of galaxies spanning about one unit redshift , projection effects are bound to give rise to a range of surface densities . second , the number counts may show significant variations as a function of position and environment resulting from the large - scale structure . the uncertainties in the cosmic variance can be reduced by observing fields that are larger than the typical scale length of the large - scale structures , but this is often not achieved in typical observations at @xmath0 . third , surface overdensities that are related to genuine overdensities in physical coordinates are not necessarily due to proto - clusters , as we have shown that the descendents of @xmath5-dropouts can be found in a wide range of environments at @xmath8 , galaxy groups being the most common ( see fig . [ fig : z0halos ] ) . we have separated the contribution of these effects to @xmath135 from that due to proto - clusters by calculating the fraction of actual proto - cluster @xmath5-dropouts in each cell of given overdensity @xmath139 . the results are also shown in fig . [ fig : pcumul ] , where dashed histograms indicate the combined probability of finding a cell of overdensity @xmath140 consisting of more than 25 ( blue lines ) , 50 ( green lines ) and 75% ( red lines ) protocluster galaxies . the results show that while , for example , the chance of @xmath141 is about 1% , the chance that at least 50% of the galaxies in such cells are proto - cluster galaxies is only half that , about 0.5% ( see top left panel in fig . [ fig : pcumul ] ) . the figure goes on to show that the fractions of protocluster galaxies increases significantly as the overdensity increases , indicating that the largest ( and rarest ) overdensities in the @xmath5-dropout distribution are related to the richest proto - cluster regions . this is further illustrated in fig . [ fig : deltaclusfrac ] in which we plot the average and scatter of the fraction of proto - cluster galaxies as a function of @xmath139 . although the fraction rises as the overdensity increases , there is a very large scatter . at @xmath142 the average fraction of protocluster galaxies is about 0.5 , but varies significantly between 0.25 and 0.75 ( 1@xmath88 ) . it will be virtually impossible to estimate an accurate cluster mass at @xmath8 from a measured surface overdensity at @xmath0 . although there is a correlation between cluster mass at @xmath8 and @xmath5-dropout overdensity at @xmath0 , the scatter is significant . many of the most massive ( @xmath15@xmath92@xmath143 @xmath12 ) clusters have very small associated overdensities , while the progenitors of fairly low mass clusters ( @xmath15@xmath13@xmath144 @xmath12 ) can be found associated with regions of relatively large overdensities . however , the largest overdensities are consistently associated with the progenitors of @xmath15@xmath145 @xmath12 clusters . lines show overdensity as a function of radius for each of the protocluster regions shown in fig . [ fig : topclusters_radec ] . ] although the above sections yield useful statistical results , it is interesting to look at the detailed angular and redshift distributions of the @xmath2-dropouts in a few of the overdense regions . in fig . [ fig : topclusters_radec ] we show 16 30@xmath8430regions having overdensities ranging from @xmath146 ( bottom left panel ) to @xmath81 ( top right panel ) . in each panel we indicate the relative size of an acs pointing ( red square ) , and the redshift , overdensity and present - day mass of the most massive protoclusters are given in the top left and right corners . field galaxies are drawn as open circles , while protocluster galaxies are drawn as filled circles . galaxies belonging to the same proto - cluster are drawn in the same colour . while some regions contain relatively compact protoclusters with numerous members inside the [email protected] acs field - of - view ( e.g. panels # 0 , 1 and 8) , other regions may contain very few or highly dispersed galaxies . also , many regions contain several overlapping protoclusters as the selection function is sensitive to structures from a relatively wide range in redshift inside the 30@xmath8430 regions plotted . although the angular separation between galaxies belonging to the same protocluster is typically smaller than @xmath1310 or 25 mpc ( comoving ) , fig . [ fig : topclusters_radius ] shows that the overdensities of regions centered on the protoclusters are significantly positive out to much larger radii of between 10 to 30 , indicating that the protoclusters form inside very large filaments of up to 100 mpc in size that contribute significantly to the overall ( field ) number counts in the protocluster regions . in fig . [ fig : topclusters_z ] we plot the redshift coordinate against one of the angular coordinates using the same regions and colour codings as in fig . [ fig : topclusters_radec ] . protoclusters are significantly more clumped in redshift space compared to field galaxies , due to flattening of the velocity field associated with the collapse of large structures . in each panel , a red dashed line marks @xmath147 , which roughly corresponds to , respectively , the upper and lower redshift of samples selected by placing a cut at @xmath2@xmath37@xmath972 and @xmath2@xmath37@xmath1372 ( see the redshift selection functions in fig . [ fig : nz2 ] ) . such colour cuts may help reduce the contribution from field galaxies by about 50% , depending on the redshift one is interested in . we also mark the typical redshift range of @xmath148 probed by narrowband filters centered on the redshift of each protocluster using blue dotted lines . as we will show in more detail in [ sec : ouchi ] below , such narrowband selections offer one of the most promising methods for finding and studying the earliest collapsing structures at high redshift , because of the significant increase in contrast between cluster and field regions . however , such surveys are time - consuming and only probe the part of the galaxy population that is bright in the ly@xmath3 line . our mock survey of @xmath5-dropouts constructed from the mr , due to its large effective volume , spans a wide range of environments and is therefore ideal for making detailed comparisons with observational studies of the large - scale structure at @xmath0 . in the following subsections , we will make such comparisons with two studies of candidate proto - clusters of @xmath5-dropouts and ly@xmath3 emitters found in the sdf and sxdf . when analysing the sky distribution of @xmath5-dropouts in the 876 arcmin@xmath82 subaru deep field , @xcite ( henceforward o08 ) discovered a large surface overdensity , presumed to be a proto - cluster at @xmath0 . the magnitude of the overdensity was quantified as the excess of @xmath5-dropouts in a circle of 20 mpc comoving radius . the region had @xmath149 with @xmath150 significance . furthermore , this region also contained the highest density contrast measured in a 8 mpc comoving radius @xmath151 ( @xmath152 ) compared to other regions of the sdf . by relating the total overdensity in dark matter to the measured overdensity in galaxies through an estimate of the galaxy bias parameter , the authors estimated a mass for the proto - cluster region of @xmath153 @xmath12 . we use our mock survey to select @xmath5-dropouts with @xmath2@xmath37@xmath89 and @xmath37@xmath154 , similar to o08 . the resulting surface density was 0.16 arcmin@xmath86 in very good agreement with the value of 0.18 arcmin@xmath86 found by o08 . in fig . [ fig : ota ] we plot the sky distribution of our sample , and connect regions of constant ( positive ) density @xmath155 . next we selected all regions that had @xmath156 . these regions are indicated by the large red circles in fig . [ fig : ota ] . we find @xmath1330 ( non - overlapping ) regions in our entire mock survey having @xmath157 at @xmath158 significance , relative to the mean dropout density of @xmath159 . analogous to fig . [ fig : field ] , we have marked all @xmath5-dropouts associated with proto - clusters with large symbols . it can be seen clearly that the proto - cluster galaxies are found almost exclusively inside the regions of enhanced local surface density indicated by the contour lines , while the large void regions are virtually depleted of proto - cluster galaxies . although the 30 regions of highest overdensity selected to be similar to the region found by o08 coincide with the highest peaks in the global density distribution across the field , it is interesting to point out that in some cases the regions contain very few actual proto - cluster galaxies , e.g. , the regions at ( ra , dec)=(10,150 ) and ( 80,220 ) in fig . [ fig : ota ] . we therefore introduce a proto - cluster `` purity '' parameter , @xmath160 , defined as the ratio of galaxies in a ( projected ) region that belong to protoclusters to the total number of galaxies in that region . we find @xmath1611650% . the purest or richest proto - clusters are found in regions having a wide range in overdensities , e.g. , the region at ( 175,225 ) with @xmath162 , @xmath16350% ) , and the region at ( 200,40 ) with @xmath164 , @xmath16340% . following o08 we also calculate the maximum overdensity in each region using cells of 8 mpc radius . we find @xmath165 with @xmath166 significance . these sub - regions are indicated in fig . [ fig : ota ] using smaller circles . interestingly , there is a very wide range in proto - cluster purity of @xmath167080% . the largest overdensity in fig . [ fig : ota ] at ( 175,225 ) corresponds to the region giving birth to the most massive cluster . by @xmath8 , this region has grown into a `` supercluster '' region containing numerous clusters , two of which have @xmath168 @xmath12 . we conclude that local overdensities in the distribution of @xmath5-dropouts on scales of @xmath1310 - 50 comoving mpc similar to the one found by o08 indeed trace the seeds of massive clusters . because our mock survey is about 80@xmath84 larger than the sdf , we expect that one would encounter such proto - cluster regions in about one in three ( 2.7 ) sdf - sized fields on average . however , the fraction of actual proto - cluster galaxies is in the range 1650% ( 080% for 8 mpc radius regions ) . this implies that while one can indeed find the overdense regions where clusters are likely to form , there is no way of verifying which galaxies are part of the proto - cluster and which are not , at least not when using photometrically selected samples . these results are consistent with our earlier finding that there is a large scatter in the relation between the measured surface overdensity and both cluster `` purity '' and the mass of its descendant cluster at @xmath8 ( sect . [ sec : detclus ] ) . the sky distribution of @xmath5-dropouts selected using criteria matched to those of @xcite . grey solid lines are surface density contours of @xmath169 and @xmath170 . large red dashed circles mark overdense regions of @xmath171 , corresponding to similar overdensities as that associated with the candidate @xmath0 proto - cluster region found by @xcite in the subaru deep field . small red circles inside each region mark a subregion having the largest overdensity @xmath172 measured in a 8 mpc co - moving radius ( projected ) cell ( see text for further details ) . ] mock ly@xmath3 survey at @xmath173 constructed from the mr mock @xmath2-dropout sample . grey solid lines are surface density contours of @xmath174 to 3.25 with a step increase of 0.5 as in fig . 2 of @xcite . the black dashed line marks the average field density . small circles indicate field galaxies . large circles indicate protocluster galaxies . ] the addition of velocity information gives studies of ly@xmath3 samples a powerful edge over purely photometrically selected @xmath2-dropout samples . as explained by ( * ? ? ? * and references therein ) , peculiar velocity fields are influenced by the large - scale structure : streaming motions can shift the overall distribution in redshift , while the dispersion can both increase and decrease as a result of velocity gradients . galaxies located in different structures that are not physically bound will have higher velocity dispersions , while galaxies that are in the process of falling together to form non - linear structures such as a filaments , sheets ( or `` pancakes '' ) and proto - clusters will have lower velocity dispersions . using deep narrow - band imaging observations of the sxdf , @xcite ( o05 ) were able to select ly@xmath3 candidate galaxies at @xmath175 . follow - up spectroscopy of the candidates in one region that was found to be significantly overdense ( @xmath176 ) on a scale of 8 mpc ( comoving ) radius resulted in the discovery of two groups ( a and b ) of ly@xmath3 emitting galaxies each having a very narrow velocity dispersion of @xmath177 km s@xmath101 . the three - dimensional density contrast is on the order of @xmath178 , comparable to that of present - day clusters , and the space density of such proto - cluster regions is roughly consistent with that of massive clusters ( see o05 ) . in order to study the velocity fields of collapsing structures and carry out a direct comparison with o05 , we construct a simple ly@xmath3 survey from our mock sample as follows . first , we construct a ( gaussian ) redshift selection function centred on @xmath20 with a standard deviation of 0.04 . as it is not known what causes some galaxies to be bright in ly@xmath3 and others not , our simulations do not include a physical prescription for ly@xmath3 as such . however , empirical results suggest that ly@xmath3 emitters are mostly young , relatively dust - free objects and a subset of the @xmath2-dropout population . the fraction of galaxies with high equivalent width ly@xmath3 is about 40% , and this fraction is found to be roughly constant as a function of the rest - frame uv continuum magnitude . therefore , we scale our selection function so that it has a peak selection efficiency of 40% . next , we apply this selection function to the @xmath2-dropouts from the mock survey to create a sample with a redshift distribution similar to that expected from a narrowband ly@xmath3 survey . finally , we tune the limiting @xmath37 magnitude until we find a number density that is similar to that reported by o05 . by setting @[email protected] mag we get the desired number density of @xmath130.1 arcmin@xmath86 . the mock ly@xmath3 field is shown in fig . [ fig : ouchi ] . the correlations between surface overdensity , cluster purity and velocity dispersion for ly@xmath3 galaxies selected from the mock ly@xmath3 survey shown in fig . [ fig : ouchi ] using randomly drawn cells of 8 mpc ( comoving ) radius . dashed lines indicate the median trends . red points highlight regions of purity @xmath1790.5 . blue points highlight regions of @xmath1790.5 and @xmath180 . shaded areas mark the values obtained by @xcite for a protocluster of ly@xmath3 galaxies in the sdf . see text for details . ] in the top left panel of fig . [ fig : ouchiresults ] we plot the overdensities measured in randomly drawn regions of 8 mpc ( comoving ) radius against the protocluster purity parameter , analogous to fig . [ fig : deltaclusfrac ] . although the median purity of a sample increases with overdensity ( dashed line ) , the scatter indicated by the points is very large even for overdensities as large as @xmath181 found by o05 ( marked by the shaded region in the top panel of fig . [ fig : ouchiresults ] ) . to guide the eye , we have plotted regions of purity @xmath920.5 as red points , and regions having purity @xmath920.5 and @xmath182 as blue points in all panels of fig . [ fig : ouchiresults ] . next , we calculate the velocity dispersion , @xmath183 , from the peculiar velocities of the galaxies in each region using the bi - weight estimator of @xcite that is robust for relatively small numbers of objects ( @xmath184 ) , and plot the result against @xmath139 and cluster purity in the top right and bottom left panels of fig . [ fig : ouchiresults ] , respectively . although gravitational clumping of galaxies in redshift space causes the velocity dispersions to be considerably lower than the velocity width allowed by the bandpass of the narrowband filter ( @xmath185 km @xmath186 compared to @xmath187 km @xmath186 for @xmath188 ) , the velocity dispersion is not a decreasing function of the overdensity ( at least not up to @xmath189 ) and the scatter is significant . this can be explained by the fact that proto - clusters regions are rare , and even regions that are relatively overdense in angular space still contain many galaxies that are not contained within a single bound structure . a much stronger correlation is found between dispersion and cluster purity ( see bottom left panel of fig . [ fig : ouchiresults ] ) . although the scatter in dispersion is large for regions with a purity of @xmath190 , the smallest dispersions are associated with some of the richest protocluster regions . this can be understood because the `` purest '' structures represent the bound inner cores of future clusters at @xmath8 . the velocity dispersions are low because these systems do not contain many field galaxies that act to inflate the velocity dispersion measurements . therefore , the velocity dispersion correlates much more strongly with the protocluster purity than with the surface overdensity . the overdensity parameter helps , however , in reducing some of the ambiguity in the cluster richness at small dispersions ( compare black and blue points at small @xmath183 in the bottom left panel ) . the shaded regions in fig . [ fig : ouchiresults ] indicate the range of measurements of o05 , implying that their structure has the characteristics of ly@xmath3 galaxies falling into a protocluster at @xmath0 . for reasons explained in the introduction , it is generally assumed that the luminous qsos at @xmath0 inhabit the most massive dark matter in the early universe . the hst / acs , with its deep and wide - field imaging capabilities in the @xmath2 and @xmath37 bands , has made it possible to test one possible implication of this by searching for small neighbouring galaxies tracing these massive halos . in this section , we will first investigate what new constraint we can put on the masses of the host halos based on the observed neighbour statistics . @xcite have addressed the same problem based on the excursion set formalism . our analysis is based on semi - analytic models incorporated in the mr simulation , which we believe is likely to provide a more realistic description of galaxy properties at @xmath0 . we will use the simulations to evaluate what we can say about the most likely environment of the qsos and whether they are associated with proto - clusters . we finish the section by presenting some clear examples from the simulations that would signal a massive overdensity in future observations . several searches for companion galaxies in the vicinity of @xmath0 qsos have been carried out to date . in table [ tab : surveys ] we list the main surveys , covering in total 6 qsos spanning the redshift range @xmath191 . we have used the results given in @xcite , @xcite and @xcite to calculate the surface overdensities associated with each of the qso fields listed in table [ tab : surveys ] . only two qsos were found to be associated with positive overdensities to a limiting magnitude of @[email protected] : j0836 + 0054 image is not available to reject lower redshift interlopers . ] ( @xmath192 ) and j1030 + 0524 ( @xmath21 ) both had @xmath193 , although evidence suggests that the overdensity could be as high as @xmath194 when taking into account subclustering within the acs field or sources selected using different @xmath195 or colour cuts ( see * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * for details ) . the remaining four qso fields ( j1306 + 0356 at @xmath196 , j1630 + 4012 at @xmath197 , j1048 + 4637 at @xmath198 , and j1148 + 5251 at @xmath199 ) were all consistent with having no excess counts with @xmath200 spanning the range from about @xmath2011 to @xmath2020.5 relatively independent of the method of selection @xcite . focusing on the two overdense qso fields , fig . [ fig : pcumul ] tells us that overdensities of @xmath203 are fairly common , occurring at a rate of about 17% in our @xmath204 simulation . the probability of finding a random field with @xmath205 is about 5 to 1% . it is evident that none of the six quasar fields have highly significant overdensities . the case for overdensities near the qsos would strengthen if all fields showed a systematically higher , even if moderate , surface density . however , when considering the sample as a whole the surface densities of @xmath5-dropouts near @xmath0 qsos are fairly average , given that four of the qso fields have lower or similar number counts compared to the field . with the exception perhaps of the field towards the highest redshift qso j1148 + 5251 , which lies at a redshift where the @xmath5-dropout selection is particularly inefficient ( see fig . [ fig : nz2 ] ) , the lack of evidence for substantial ( surface ) overdensities in the qso fields is puzzling . panels show the number of neighbours ( @xmath2-dropouts ) in cubic regions of @xmath206 mpc@xmath61 versus the mass of the most massive halo found in each of those regions . top and bottom panels are for snapshots at @xmath50 and @xmath51 , respectively . left and right panels are for neighbour counts down to limiting magnitudes of @xmath37=27.5 ( left ) and @xmath37@xmath207 mag , respectively . there is a wide dispersion in the number of neighbours , even for the most massive halos at @xmath0 . the highest numbers of neighbours are exclusively associated with the massive end of the halo mass function , allowing one to derive a lower limit for the mass of the most massive halo for a given number of neighbour counts . the scatter in the number of neighbours versus the mass of the most massive halo reduces signficantly when going to fainter magnitudes . the small squares in the panels on the right correspond to the three richest regions ( in terms of @[email protected] mag dropouts ) that are shown in close - up in fig . [ fig : z6panels ] . ] in fig . [ fig : boxtest ] we have plotted the number of @xmath2-dropouts encountered in cubic regions of @xmath208 @xmath27 mpc against the mass of the most massive dark matter halo found in each region . panels on the left and on the right are for limiting magnitudes of @xmath37=27.5 and 26.5 mag , respectively . because the most massive halos are so rare , here we have used the full mr snapshots at @xmath50 ( top panels ) and @xmath51 ( bottom panels ) rather than the lightcone in order to improve the statistics . there is a systematic increase in the number of neighbours with increasing maximum halo mass . however , the scatter is very large : for example , focusing on the neighbour count prediction for @xmath50 and @[email protected] ( top right panel ) we see that the number of neighbours of a halo of @xmath209 @xmath27 @xmath12 can be anywhere between 0 and 20 , and some of the most massive halos of @xmath111 @xmath27 @xmath12 have a relatively low number of counts compared to some of the halos of significant lower mass that are far more numerous . however , for a given @[email protected] neighbour count ( in a @xmath208 @xmath27 mpc region ) of @xmath210 , the halo mass is _ always _ above @xmath211 @xmath27 @xmath12 , and if one would observe @xmath212 @xmath2-dropout counts one could conclude that that field must contain a supermassive halo of @xmath213 @xmath27 @xmath12 . thus , in principle , one can only estimate a lower limit on the maximum halo mass as a function of the neighbour counts . the left panel shows that the scatter is much reduced if we are able to count galaxies to a limiting @xmath37-band magnitude of 27.5 instead of 26.5 , simply because the poisson noise is greatly reduced . we can therefore conclude that the relatively average number of counts observed in the qso fields is not inconsistent with the qsos being hosted by very massive halos . however , one could make an equally strong case that they are , in fact , not . if we translate our results of fig . [ fig : boxtest ] to the qso fields that cover a smaller projected area of @xmath214 mpc@xmath82 , and we add back in the average counts from the fore- and background provided by our lightcone data , we estimate that for qsos at @xmath215 we require an overdensity of @xmath216 in order to be able to put a lower limit on the qso host mass of @xmath217 @xmath12 , while a @xmath218 is consistent with @xmath108 @xmath12 . at @xmath51 , we would require @xmath219 for @xmath220 @xmath12 and @xmath221 for @xmath222 @xmath12 . comparison with the relatively low surface overdensities observed thus suggests that the halo mass is uncontrained by the current data . nonetheless , we can at least conclude quite firmly that the qsos are in far less rich environments ( in terms of galaxy neigbours ) compared to many rich regions found both in the simulations and in some of the deep field surveys described in the previous section . in order to get a feel for what the qso fields might have looked like if they were in highly overdense regions , we present some close - up views in fig . [ fig : z6panels ] of the three richest regions of @[email protected] mag @xmath2-dropouts as marked by the small squares in fig . [ fig : boxtest ] . the central position corresponding to that of the most massive halo in each region is indicated by the green square . large and small dots correspond to dropout galaxies having @[email protected] and @xmath11827.5 mag , respectively . for reference , we use blue circles to indicate galaxies that have been identified as part of a protocluster structure . the scale bar at the top left in each panel corresponds to the size of an acs / wfc pointing used to observe @xmath0 qso fields . we make a number of interesting observations . first , using the current observational limits on depth ( @xmath37=26.5 mag ) and field size ( 3.4 , see scale bar ) imposed by the acs observations of qsos , it would actually be quite easy to miss some of these structures as they typically span a larger region of 23 acs fields in diameter . going too much fainter magnitudes would help considerably , but this is at present unfeasible . note , also , that in three of the panels presented here , the galaxy associated with the massive central halo does not pass our magnitude limits . it is missed due to dust obscuration associated with very high star formation rates inside these halos , implying that they will be missed by large - area uv searches as well ( unless , of course , they also host a luminous , unobscured qso ) . finally , we investigate the level of mutual correspondence between the most massive halos selected at @xmath9 and @xmath8 . in fig . [ fig : z6panels ] we already saw that the richest regions are associated with a very large number of galaxies that will become part of a cluster when evolved to @xmath8 . in the top row of fig . [ fig : hh ] we show the mass of the most massive progenitors at @xmath50 ( left ) , @xmath51 ( middle ) ad @xmath52 ( right ) of halos selected at @xmath8 ( see also * ? ? ? the dotted line indicates the threshold corresponding to massive galaxy clusters at @xmath8 . although the progenitor mass increases systematically with increasing local mass , the dispersion in the mass of the most massive @xmath0 progenitors is about or over one order of magnitude , and this is true even for the most massive clusters . as explained in detail by @xcite this observation leads to an interesting complication when using the refinement technique often used to simulate the most massive regions in the early universe by resimulating at high redshift the most massive region identified at @xmath8 in a coarse grid simulation . in the bottom panels of fig . [ fig : hh ] we show the inverse relation between the most massive halos selected at @xmath0 and their most massive descendant at @xmath8 . from this it becomes clear that eventhough the most massive @xmath0 halos ( e.g. those associated with qsos ) are most likely to end up in present - day clusters , some evolve into only modest local structures more compatible with , e.g. , galaxy groups . this implies that the present - day descendants of some of the first , massive galaxies and supermassive black holes must be sought in sub - cluster environments . although our findings of the previous section show that the apparent lack of excess neighbour counts near @xmath0 qsos is not inconsistent with them being hosted by supermassive dark matter halos as suggested by their low co - moving abundance and large inferred black hole mass , it is interesting to note that none of the qso fields have densities that would place them amongst the richest structures in the @xmath0 universe . this leads to an intriguing question : where is the large - scale structure associated with qsos ? one possibility that has been discussed ( e.g. * ? ? ? * ) is that while the dark matter density near the qsos is significantly higher compared to other fields , the strong ionizing radiation from the qso may prohibit the condensation of gas thereby suppressing galaxy formation . although it is not clear how important such feedback processes are exactly , we have found that proto - clusters in the mr form inside density enhancements that can extend up to many tens of mpc in size . although we do not currently know whether the @xmath0 qsos might be associated with overdensities on scales larger than a few arcminutes as probed by the acs , it is unlikely that the qso ionization field will suppress the formation of galaxies on such large scales @xcite . an alternative , perhaps more likely , explanation for the deficit of @xmath2-dropouts near qsos , is that the dark matter halo mass of the qsos is being greatly overestimated . @xcite suggest that the combination of the steepness of the halo mass function for rare high redshift halos on one hand , combined with the sizeable intrinsic scatter in the correlation between black hole mass and stellar velocity dispersion or halo mass at low redshift on the other hand , makes it much more probable that a @xmath223 @xmath12 black hole is found in relatively low mass halos than in a very rare halo of extremely high mass . depending on the exact value of the scatter , the typical mass of a halo hosting a @xmath0 qso may be reduced by @xmath224 in log @xmath15@xmath225 without breaking the low redshift @xmath15@xmath226 relation . the net result is that qsos occur in some subset of halos found in substantially less dense environments , which may explain the observations . this notion seems to be confirmed by the low dynamical mass of @xmath227 @xmath12 estimated for the inner few kpc region of sdss j1148 + 5251 at @xmath199 based on the co line emission @xcite . this is in complete contradiction to the @xmath217 @xmath12 stellar mass bulges and @xmath228 @xmath12 mass halos derived based on other arguments . if true , models should then explain why the number density of such qsos is as observed . on the other hand , recent theoretical work by @xcite suggests that in order to facilitate the formation of a supermassive ( @xmath11 @xmath12 ) by @xmath0 in the first place , it may be required to have a rare pair of dark matter halos ( @xmath228 @xmath12 ) in which the intense uv radiation from one halo prevents fragmentation of the other so that the gas collapses directly into a supermassive blackhole . this would constrain the qsos to lie in even richer environments . the predicted large - scale distributions of @xmath2-dropouts and ly@xmath3 emitters as shown in , e.g. , figs . [ fig : field ] , [ fig : ota ] and [ fig : ouchi ] show evidence for variations in the large - scale structure on scales of up to @xmath131 - 2 , far larger than currently probed by deep hst or large - area ground - based surveys . a full appreciation of such structures could be important for a range of topics , including studies of the luminosity function at @xmath0 and studies of the comparison between @xmath19cdm predictions and gravitational clustering on very large scales . the total area probed by our simulation is a good match to a survey of @xmath229 degree@xmath82 targeting @xmath2-dropouts and ly@xmath3 emitters at @xmath0 planned with the forthcoming subaru / hypersurpimecam ( first light expected 2013 ; m. ouchi , private communications , 2008 ) . we found that the @xmath2-dropouts associated with proto - clusters are almost exclusively found in regions with positive density enhancements . a proper understanding of such dense regions may also be very important for studies of the epoch of reionization . simulations suggest that even though the total number of ionizing photons is much larger in very large proto - cluster regions covering several tens of comoving mpc as compared to the field ( e.g. * ? ? ? * but see @xcite ) , they may still be the last to fully re - inionize , because the recombination rates are also much higher . if regions associated with qsos or other structures were to contain significant patches of neutral hydrogen , this may affect both the observed number densities and clustering of lbgs or ly@xmath3 emitters relative to our assumed mean attenuation @xcite . since our work mostly focuses on @xmath59 when reionization is believed to be largely completed , this may not be such an issue compared to surveys that probe earlier times at @xmath230 ( e.g. * ? ? ? * ; * ? ? ? our evaluation of the possible structures associated with qsos leads to several suggestions for future observations . while it is unlikely that the wide field camera 3 ( wfc3 ) to be installed onboard hst in early 2009 will provide better contraints than hst / acs due to its relatively small field - of - view , we have shown that either by surveying a larger area of @xmath1310@xmath8410 or by going deeper by @xmath131 mag in @xmath37 , one significantly reduces the shot noise in the neighbour counts allowing more reliable overdensities and ( lower ) limits on the halo masses to be estimated . a single pointing with acs would require @xmath231 orbits in @xmath37 to reach a point source sensitivity of 5@xmath88 for a @xmath37=27.5 mag object at @xmath0 . given the typical structure sizes of the overdense regions shown in fig . [ fig : z6panels ] , a better approach would perhaps be to expand the area of the current qso fields by several more acs pointings at their present depth of @xmath37=26.5 mag for about an equal amount of time . however , this may be achieved from the ground as well using the much more efficient wide - field detector systems . although this has been attempted by @xcite targeting three of the qso fields , we note that their achieved depth of @xmath37=25.5 was probably much too shallow to find any overdensities even if they are there . we would like to stress that it is extremely important that foreground contamination is reduced as much as possible , for example by combining the observations with a deep exposure in the @xmath54 band . this is currently not available for the qso fields , making it very hard to calculate the exact magnitude of any excess counts present . while a depth of @xmath37=27.5 mag seems out of reach for a statistical sample with hst , narrow - band ly@xmath3 surveys targeting the typically uv - faint ly@xmath3emitters from the ground would be a very efficient alternative . although a significant fraction of sources lacking ly@xmath3may be lost compared to dropout surveys , they have the clear additional advantage of redshift information . most ly@xmath3 surveys are carried out in the atmospheric transmission windows that correspond to redshifted ly@xmath3 either at @xmath215 or @xmath232 for which efficient narrow - band filters exist . we therefore suggest that the experiment is most likely to succeed around qsos at @xmath215 rather than the qsos at @xmath233 looked at so far . it is , however , possible to use combinations of , e.g. , the @xmath215 narrow - band filter with medium or broad band filters at @xmath139000to place stronger constaints on the photometric redshifts of @xmath2-dropouts in qso fields ( e.g. , see * ? ? ? * ) . in the next decade , _ jwst _ will allow for some intriguing further possibilities that may provide some definite answers : using the target 0.70.9@xmath117 m sensitivity of the near infrared camera ( nircam ) on _ jwst _ we could reach point sources at 10@xmath88 as faint as @xmath37=28.5 mag in a 10,000 s exposure , or we could map a large @xmath1310@xmath8410 region around qsos to a depth of @xmath37=27.5 mag within a few hours . the near infrared spectrograph ( nirspec ) will allow @xmath92100 simultaneous spectra to confirm the redshifts of very faint line or continuum objects over a @xmath929 arcmin@xmath82 field of view . the main findings of our investigation can be summarized as follows . @xmath234 we have used the @xmath24-body plus semi - analytic modeling of @xcite to construct the largest ( 4@xmath235 ) mock galaxy redshift survey of star - forming galaxies at @xmath0 to date . we extracted large samples of @xmath2-dropouts and ly@xmath3 emitters from the simulated survey , and showed that the main observational ( colours , number densities , redshift distribution ) and physical properties ( @xmath236 , sfr , age , @xmath237 ) are in fair agreement with the data as far as they are constrained by various surveys . @xmath234 the present - day descendants of @xmath2-dropouts ( brighter than @xmath67 ) are typically found in group environments at @xmath8 ( halo masses of a few times @xmath111 @xmath12 ) . about one third of all @xmath2-dropouts end up in halos corresponding to clusters , implying that the contribution of `` proto - cluster galaxies '' in typical @xmath2-dropout surveys is significant . @xmath234 the projected sky distribution shows significant variations in the local surface density on scales of up to 1@xmath4 , indicating that the largest surveys to date do not yet probe the full range of scales predicted by our @xmath19cdm models . this may be important for studies of the luminosity function , galaxy clustering , and the epoch of reionization . @xmath234 we present counts - in - cells frequency distributions of the number of objects expected per [email protected] hst / acs field of view , finding good agreement with the goods field statistics . the largest positive deviations are due to structures associated with the seeds of massive clusters of galaxies ( `` protoclusters '' ) . to guide the interpretation of current and future hst / acs observations , we give the probabilities of randomly finding regions of a given surface overdensity depending on the presence or absence of a protocluster . @xmath234 we give detailed examples of the structure of proto - cluster regions . although the typical separation between protocluster galaxies does not reach beyond @xmath1310 ( 25 mpc comoving ) , they sit in overdensities that extend up to 30 radius , indicating that the proto - clusters predominantly form deep inside the largest filamentary structures . these regions are very similar to two proto - clusters of @xmath2-dropouts or ly@xmath3 emitters found in the sdf @xcite and sxdf @xcite fields . @xmath234 we have made a detailed comparison between the number counts predicted by our simulation and those measured in fields observed with hst / acs towards luminous @xmath0 qsos from sdss , concluding that the observed fields are not particularly overdense in neighbour counts . we demonstrate that this does not rule out that the qsos are in the most massive halos at @xmath0 , although we can also not confirm it . we discuss the possible reasons and implications of this intriguing result ( see the discussion in section 6 ) . @xmath234 we give detailed recommendations for follow - up observations using current and future instruments that can be used to better constrain the halo masses of @xmath0 qsos and the variations in the large - scale structure as probed by @xmath2-dropouts and ly@xmath3 emitters ( see section 7 ) . many colleagues have contributed to this work . we thank tom abel , jrmy blaizot , bernadetta ciardi , soyoung kim , sangeeta malhotra , james rhoads , massimo stiavelli , and bram venemans for their time and suggestions . we are grateful to masami ouchi for a careful reading of our manuscript and insightful comments . we owe great gratitude to volker springel and simon white and their colleagues at mpa responsible for the millennium run project . the millennium simulation databases used in this paper and the web application providing online access to them were constructed as part of the activities of the german astrophysical virtual observatory . rao acknowledges the support and hospitality of the aspen center for physics where part of this research was carried out . ajiki , m. , et al . 2006 , pasj , 58 , 113 barth , a. j. , martini , p. , nelson , c. h. , & ho , l. c. 2003 , apjl , 594 , l95 beers , t. c. , flynn , k. , & gebhardt , k. 1990 , aj , 100 , 32 bertoldi , f. , et al . 2003 , a&a , 409 , l47 begelman , m. c. , volonteri , m. , & rees , m. j. 2006 , mnras , 370 , 289 blaizot , j. , wadadekar , y. , guiderdoni , b. , colombi , s. t. , bertin , e. , bouchet , f. r. , devriendt , j. e. g. , & hatton , s. 2005 , mnras , 360 , 159 bouwens , r. j. , et al . 2003 , apj , 595 , 589 bouwens , r. j. , et al . 2004a , apjl , 606 , l25 bouwens , r. j. , illingworth , g. d. , blakeslee , j. p. , & franx , m. 2006 , apj , 653 , 53 bouwens , r. j. , illingworth , g. d. , franx , m. , & ford , h. 2007 , apj , 670 , 928 bouwens , r. j. , illingworth , g. d. , blakeslee , j. p. , broadhurst , t. j. , & franx , m. 2004b , apjl , 611 , l1 bruzual , g. , & charlot , s. 2003 , mnras , 344 , 1000 calzetti , d. 2001 , pasp , 113 , 1449 ciardi , b. , stoehr , f. , & white , s. d. m. 2003 , mnras , 343 , 1101 croton , d. j. , et al . 2006 , mnras , 365 , 11 dav , r. , finlator , k. , & oppenheimer , b. d. 2006 , mnras , 370 , 273 de lucia , g. , kauffmann , g. , & white , s. d. m. 2004 , mnras , 349 , 1101 de lucia , g. , & blaizot , j. 2007 , mnras , 375 , 2 , m. et al . 2004 , apjl , 600 , l99 djorgovski , s. g. , stern , d. , mahabal , a. a. , & brunner , r. 2003 , apj , 596 , 67 dow - hygelund , c. c. , et al . 2005 , apjl , 630 , l137 dijkstra , m. , et al . 2008 , mnras , submitted ( arxiv:0810.0014 ) eyles , l. p. , bunker , a. j. , ellis , r. s. , lacy , m. , stanway , e. r. , stark , d. p. , & chiu , k. 2007 , mnras , 374 , 910 fan , x. , et al . 2003 , aj , 125 , 1649 fan , x. , et al . 2006b , aj , 132 , 117 fan , x. , et al . 2006a , aj , 131 , 1203 fan , x. , et al . 2004 , aj , 128 , 515 fan , x. , et al . 2001 , aj , 122 , 2833 finlator , k. , dav , r. , & oppenheimer , b. d. 2007 , mnras , 376 , 1861 gayler harford , a. , & gnedin , n. y. 2006 , submitted to apj ( astro - ph/0610057 ) goto , t. 2006 , mnras , 371 , 769 gunn , j. e. , & peterson , b. a. 1965 , apj , 142 , 1633 guo , q. , & white , s. d. m. 2008 , arxiv:0809.4259 haiman , z. , & loeb , a. 2001 , apj , 552 , 459 iliev , i. t. , mellema , g. , pen , u .- l . , merz , h. , shapiro , p. r. , & alvarez , m. a. 2006 , mnras , 369 , 1625 jiang , l. , fan , x. , vestergaard , m. , kurk , j. d. , walter , f. , kelly , b. c. , & strauss , m. a. 2007 , aj , 134 , 1150 jiang , l. , et al . 2006 , aj , 132 , 2127 kauffmann , g. , colberg , j. m. , diaferio , a. , & white , s. d. m. 1999 , mnras , 303 , 188 kashikawa , n. , et al . 2004 , pasj , 56 , 1011 kashikawa , n. , et al . 2006 , apj , 648 , 7 kashikawa , n. , kitayama , t. , doi , m. , misawa , t. , komiyama , y. , & ota , k. 2007 , apj , 663 , 765 khochfar , s. , silk , j. , windhorst , r. a. , & ryan , r. e. , jr . 2007 , apjl , 668 , l115 kitzbichler , m. g. , & white , s. d. m. 2007 , mnras , 376 , 2 kim , s. , et al . 2008 , arxiv e - prints , 805 , arxiv:0805.1412 kurk , j. d. , et al . 2007 , arxiv e - prints , 707 , arxiv:0707.1662 lai , k. , huang , j .- s . , fazio , g. , cowie , l. l. , hu , e. m. , & kakazu , y. 2007 , apj , 655 , 704 lemson , g. , & virgo consortium , t. 2006 , e - print ( arxiv : astro - ph/0608019 ) lemson , g. , & springel , v. 2006 , astronomical data analysis software and systems xv , 351 , 212 li , y. , et al . 2007 , apj , 665 , 187 madau , p. 1995 , apj , 441 , 18 madau , p. , pozzetti , l. , & dickinson , m. 1998 , apj , 498 , 106 malhotra , s. , et al . 2005 , apj , 626 , 666 maiolino , r. , et al . 2005 , a&a , 440 , l51 magorrian , j. , et al . 1998 , aj , 115 , 2285 mclure , r. j. , cirasuolo , m. , dunlop , j. s. , foucaud , s. , & almaini , o. 2008 , arxiv e - prints , 805 , arxiv:0805.1335 mclure , r. j. , et al . 2006 , mnras , 372 , 357 mcquinn , m. , hernquist , l. , zaldarriaga , m. , & dutta , s. 2007 , mnras , 381 , 75 miley , g. k. , et al . 2004 , nature , 427 , 47 monaco , p. , mller , p. , fynbo , j. p. u. , weidinger , m. , ledoux , c. , & theuns , t. 2005 , a&a , 440 , 799 muoz , j. a. , & loeb , a. 2008a , mnras , 385 , 2175 muoz , j. a. , & loeb , a. 2008b , mnras , 386 , 2323 narayanan , d. , et al . 2007 , arxiv e - prints , 707 , arxiv:0707.3141 nagamine , k. , cen , r. , furlanetto , s. r. , hernquist , l. , night , c. , ostriker , j. p. , & ouchi , m. 2006 , new astronomy review , 50 , 29 nagamine , k. , ouchi , m. , springel , v. , & hernquist , l. 2008 , apj , submitted ( arxiv:0802.0228 ) night , c. , nagamine , k. , springel , v. , & hernquist , l. 2006 , mnras , 366 , 705 oesch , p. a. , et al . 2007 , apj , 671 , 1212 ota , k. , kashikawa , n. , malkan , m. a. , iye , m. , nakajima , t. , nagao , t. , shimasaku , k. , & gandhi , p. 2008 , arxiv e - prints , 804 , arxiv:0804.3448 ota , k. , kashikawa , n. , nakajima , t. , & iye , m. 2005 , journal of korean astronomical society , 38 , 179 ouchi , m. , et al . 2004 , apj , 611 , 685 ouchi , m. , et al . 2005 , apjl , 620 , l1 overzier , r. a. , bouwens , r. j. , illingworth , g. d. , & franx , m. 2006 , apjl , 648 , l5 overzier , r. a. , et al . 2008a , apj , 677 , 37 overzier , r. a. , et al . 2008b , apj , 673 , 143 priddey , r. s. , ivison , r. j. , & isaak , k. g. 2007 , arxiv e - prints , 709 , arxiv:0709.0610 robertson , b. , li , y. , cox , t. j. , hernquist , l. , & hopkins , p. f. 2007 , apj , 667 , 60 shimasaku , k. , et al . 2003 , apjl , 586 , l111 , k. , ouchi , m. , furusawa , h. , yoshida , m. , kashikawa , n. , & okamura , s. 2005 , pasj , 57 , 447 springel , v. 2005b , mnras , 364 , 1105 spergel , d. n. , et al . 2003 , apjs , 148 , 175 springel , v. , et al . 2005a , nature , 435 , 629 , e. r. , bunker , a. j. , & mcmahon , r. g. 2003 , mnras , 342 , 439 stiavelli , m. , et al . 2005 , apjl , 622 , l1 steidel , c. c. , adelberger , k. l. , shapley , a. e. , erb , d. k. , reddy , n. a. , & pettini , m. 2005 , apj , 626 , 44 steidel , c. c. , adelberger , k. l. , dickinson , m. , giavalisco , m. , pettini , m. , & kellogg , m. 1998 , apj , 492 , 428 stanway , e. r. , et al . 2007 , mnras , 376 , 727 suwa , t. , habe , a. , & yoshikawa , k. 2006 , apjl , 646 , l5 trenti , m. , santos , m. r. , & stiavelli , m. 2008 , arxiv e - prints , 807 , arxiv:0807.3352 venemans , b. p. , mcmahon , r. g. , warren , s. j. , gonzalez - solares , e. a. , hewett , p. c. , mortlock , d. j. , dye , s. , & sharp , r. g. 2007 , mnras , 376 , l76 vestergaard , m. 2004 , apj , 601 , 676 venemans , b. p. , et al . 2007 , a&a , 461 , 823 volonteri , m. , & rees , m. j. 2006 , apj , 650 , 669 walter , f. , carilli , c. , bertoldi , f. , menten , k. , cox , p. , lo , k. y. , fan , x. , & strauss , m. a. 2004 , apjl , 615 , l17 wang , r. , et al . 2007 , aj , 134 , 617 wang , j. , de lucia , g. , kitzbichler , m. g. , & white , s. d. m. 2008 , mnras , 384 , 1301 white , m. 2001 , a&a , 367 , 27 white , r. l. , becker , r. h. , fan , x. , & strauss , m. a. 2003 , aj , 126 , 1 willott , c. j. , percival , w. j. , mclure , r. j. , crampton , d. , hutchings , j. b. , jarvis , m. j. , sawicki , m. , & simard , l. 2005 , apj , 626 , 657 willott , c. j. , mclure , r. j. , & jarvis , m. j. 2003 , apjl , 587 , l15 wyithe , j. s. b. , loeb , a. , & carilli , c. 2005 , apj , 628 , 575 yan , h. , dickinson , m. , giavalisco , m. , stern , d. , eisenhardt , p. r. m. , & ferguson , h. c. 2006 , apj , 651 , 24 yan , h. , & windhorst , r. a. 2004a , apjl , 612 , l93 yan , h. , & windhorst , r. a. 2004b , apjl , 600 , l1 zentner , a. r. 2007 , international journal of modern physics d , 16 , 763 zheng , w. , et al . 2006 , apj , 640 , 574	motivated by recent observational studies of the environment of @xmath0 qsos , we have used the millennium run ( mr ) simulations to construct a very large ( @xmath1 ) mock redshift survey of star - forming galaxies at @xmath0 . we use this simulated survey to study the relation between density enhancements in the distribution of @xmath2-dropouts and ly@xmath3 emitters , and their relation to the most massive halos and protocluster regions at @xmath0 . our simulation predicts significant variations in surface density across the sky with some voids and filaments extending over scales of 1@xmath4 , much larger than probed by current surveys . approximately one third of all @xmath0 halos hosting @xmath5-dropouts brighter than @xmath6=26.5 mag ( @xmath7 ) become part of @xmath8 galaxy clusters . @xmath5-dropouts associated with protocluster regions are found in regions where the surface density is enhanced on scales ranging from a few to several tens of arcminutes on the sky . we analyze two structures of @xmath5-dropouts and ly@xmath3 emitters observed with the subaru telescope and show that these structures must be the seeds of massive clusters - in - formation . in striking contrast , six @xmath0 qso fields observed with hst show no significant enhancements in their @xmath2-dropout number counts . with the present data , we can not rule out the qsos being hosted by the most massive halos . however , neither can we confirm this widely used assumption . we conclude by giving detailed recommendations for the interpretation and planning of observations by current and future ground- and space based instruments that will shed new light on questions related to the large - scale structure at @xmath0 . [ firstpage ] cosmology : observations early universe large - scale structure of universe theory galaxies : high - redshift galaxies : clusters : general galaxies : starburst .
recently undoped vanadium oxides , such as lavo@xmath3 or yvo@xmath3 , have attracted much attention due to their puzzling magnetic properties @xcite . large on - site coulomb repulsion @xmath10 ev between @xmath11 electrons leads to suppression of charge fluctuations and to the effective superexchange ( se ) interaction @xmath12 in the low energy range , where @xmath13 ev is the hopping element . the cubic vanadates belong to the class of mott insulators with orbital degeneracy as two @xmath14 electrons , localized on the v@xmath15 ion , occupy two out of three ( almost ) degenerate @xmath8 orbitals . thus , the se model includes @xmath8 orbital degrees of freedom @xcite , and _ by itself _ explains the observed @xmath0-type antiferromagnetic ( af ) phase in lavo@xmath3 , with ferromagnetic ( fm ) chains staggered in two other directions . even more interesting and not yet understood are the properties of doped la@xmath1sr@xmath2vo@xmath3 , where the following phases were experimentally observed : @xmath0-af and insulating one up to @xmath16 hole concentration , @xmath0-af metallic one for @xmath17 , and , finally , paramagnetic and metallic one for @xmath18 @xcite . theoretical explanation of the existence of this variety of phases has to take into account that the motion of doped holes is strongly affected by : ( @xmath19 ) intersite se interaction @xmath20 , ( @xmath21 ) on - site hund s interaction @xmath22 ev @xcite . such a theoretical model was recently discussed by ishihara @xcite , but there the hund s term was introduced in a nonsystematic way via certain projection operators , with @xmath23 limit being assumed implicitly , and the model was solved in a way suitable for small doping only . here we start with an appropriate kondo - lattice model , systematically introduce the large @xmath6 limit in which the kinetic and the hund s parts of the hamiltonian reduce to the double exchange ( de ) model @xcite , and solve it for any realistic value of doping . since in this paper we are mainly concerned with the explanation of the existence of the metallic phase and the associated magnetic order , we will simplify the se part of the hamiltonian as much as possible and investigate only the competition of the de and the generic se interaction due to the occupied @xmath4 obitals ( @xmath24 ) , leaving the study of the full hamiltonian as a future problem . in particular we want to answer three questions : ( @xmath19 ) can @xmath0-af and metallic phase be predicted by de , ( @xmath21 ) if yes , then why such a phase contradicts the `` intuitive '' picture of the de mechanism which supports homogenous fm phase , and ( @xmath25 ) in what way the generic features of degenerate @xmath8 orbitals help to understand the answers obtained in points ( @xmath19 ) and ( @xmath21 ) , especially the differences to the case of the @xmath26 electrons , playing a role in the manganites @xcite . thereby , we shall use a similar approximation to that employed in manganites @xcite , assuming localized @xmath4 electrons and treating them as `` core spins '' . for the sake of clarity we start with a fm kondo - lattice hamiltonian but we already include important simplifications in the se part : @xmath27 where : @xmath28 are `` core spin '' @xmath5 operators of @xmath8 electrons in occupied @xmath4 orbitals , and @xmath29 is a vector of pauli matrices . the fermion operators @xmath30 create an electron with spin @xmath31 in @xmath32 and @xmath33 orbitals , labeled here as @xmath34 and @xmath35 as these orbitals have no amplitude along @xmath34 and @xmath35 axis @xcite , respectively . in the relevant regime of large coulomb interaction @xmath9 , the fermion operators act in the restricted hilbert space without double occupancies in the @xmath36 orbitals which is implemented by the projection operators @xmath37 . then : ( @xmath19 ) itinerant electrons in @xmath32 or @xmath33 degenerate orbitals at site @xmath38 can hop in the allowed [ @xmath39 or @xmath40 plane to the nearest neighbor ( nn ) site @xmath19 , but _ only _ if there are no other electrons at site @xmath19 in these orbitals ( first and second term ) , ( @xmath21 ) hund s exchange @xmath41 aligns spins of itinerant electrons with `` core spins '' ( third terms ) , and ( @xmath25 ) occupied @xmath4 orbitals are responsible for af se interactions between nn sites in the @xmath42 plane ( last term ) , while the se interactions due to itinerant electrons are not included . the above simplifications in the se part @xmath20 are motivated by the experimental observation that the structural transitions ( which favor the orbital order and occupied @xmath4 orbitals ) occur in la@xmath1sr@xmath2vo@xmath3 at lower temperature than the magnetic ones in the entire doping range @xcite . note that the structural transition does not occur in the metallic phase and then one does not expect any orbital order , similar to the fm metallic manganites @xcite . note that the se interactions due to the excitations of @xmath32 and @xmath33 electrons could in principle be included , and would : ( @xmath19 ) enhance the af interactions and induce their alternating orbital ( ao ) order in @xmath42 planes , and ( @xmath21 ) would give fm interactions along the @xmath43 direction , and thus further stabilize the @xmath0-af phase @xcite . as in ref . @xcite , we treat `` core spins '' @xmath5 classically and we make a local unitary transformation of the basis @xmath44 in the hilbert space . the transformation makes the quantization axes of the spin of an itinerant electron locally parallel to the one of the `` core spin '' , and thus it diagonalizes hund s term in the spin variables . assuming also infinite hund s exchange @xmath45 , we leave out the itinerant electrons with antiparallel spin and what is left is a renormalized ( _ a priori _ nonhomogenous ) hopping @xmath46 of spinless electrons . the absolute value of the complex parameter @xmath47 is @xmath48 , with @xmath49 being the relative angle between `` core spins '' at site @xmath19 and @xmath38 ( the phase of @xmath47 can be chosen arbitrarily at @xmath50 ) . furthermore , we will discuss only uniform magnetic structures and follow the so - called uniform hopping approach ( uha ) @xcite . one obtains then the following _ classical _ ( i.e. zener - type @xcite ) de hamiltonian : @xmath51 where : @xmath52 , @xmath53 , @xmath54 @xmath55 is the relative angle between spins in the @xmath56 @xmath57 direction and @xmath58 is the number of sites in the crystal . the restricted fermion creation operators are now @xmath59 , @xmath60 @xmath61 creates a spinless electron at site @xmath19 in @xmath62 orbital . the fm order along the @xmath43 direction is already assumed in eq . ( [ eq:2 ] ) as it minimalizes the total energy of the system by maximalizing the possibility of hopping in this direction . note that all but one ( spin 1/2 being classical ) of the assumptions necessary to derive eq . ( [ eq:2 ] ) from eq . ( [ eq:1 ] ) seem to be reasonable approximations for cubic vanadates _ and _ for investigation of the problems adressed in section [ sec:1 ] . we leave quantum treatment of the de for further studies but we believe that the main features of the de mechanism in cubic vanadates can be captured by this classical hamiltonian . obtained by solving eq . ( [ eq:2 ] ) within sbmf approximation . under increasing hole doping @xmath63 , the @xmath0-af order changes via the canted ca * * phase , with spins canted on two bonds , to a homogenous fm phase.,width=207,height=170 ] the ground state of eq . ( [ eq:2 ] ) was found by introducing kotliar - ruckenstein slave boson representation @xcite . following ref . @xcite , we adopted this method to the orbital case and made the mean - field approximation ( the unrestricted slave boson method , identical with the gutzwiller approximation and called sbmf ) . it leads to the effective hopping terms in @xmath64 : @xmath65 where : @xmath66 @xmath67 are pseudofermion creation operators for @xmath32 @xmath68 orbital , and @xmath69 @xmath70 are hole numbers in these orbitals with the constraint @xmath71 . now ( [ eq:2 ] ) is exactly solvable by minimizing the total energy with respect to the @xmath72 and @xmath73 variational parameters @xcite , which determine the magnetic structure . the resulting magnetic phases are shown in fig . [ fig:1 ] for different values of se @xmath74 and hole doping @xmath63 . we should stress that equal number of holes in both orbitals ( @xmath75 ) is always concomitant with the obtained magnetic phases , though we did not impose such a condition . this result follows almost entirely from the fermion properties of the system , namely the condition that the fermi energies of the systems with electrons filling partly @xmath32 orbitals and @xmath33 orbitals should be the same . note also that @xmath75 is a neccessary condtion for the existence of the ao order . thus , including real se interactions from section [ sec:1 ] : ( @xmath19 ) would not contradict our results in the orbital sector , ( @xmath21 ) would only further support the stability of the @xmath0-af phase in the spin sector , which is already stable for a broad range of parameters . let us further analyze these results . one finds that for the broad range of realistic values of parameters , the de mechanism does not win in @xmath76 planes , which instead have af order and are insulating , while the holes hop only along the fm @xmath43 direction . in this way the @xmath0-af order characterized by the _ one - dimensional metallic _ behavior is stabilized . remarkably , such a phase is stable for @xmath17 and @xmath77 , in agreement with the experimental observations @xcite . these results are striking since at the first sight one expects that the de should win here due to the small `` core spin '' value @xmath5 , much smaller than in the manganites ( @xmath78 ) . instead we get results which are qualitatively similar to those obtained by van den brink and khomskii in the de model for electron doped srmno@xmath3 @xcite . though , the reason is different this behavior here follows from very specific features of @xmath8 orbitals : ( @xmath19 ) the af se interactions are strictly two - dimensional ( 2d ) since for the occupied @xmath4 electrons virtual hopping processes ( leading to the se ) can occur only in the @xmath76 plane this enables metallic behavior without any loss of the magnetic energy ( hopping in the @xmath43 direction does not destroy the magnetic order ) ; ( @xmath21 ) in addition , the itinerant electrons hop between _ the other two _ @xmath8 orbitals , allowing for the coexistence of the magnetic order and the metallic behavior , and costing no extra magnetic energy . to understand better this behavior we show in fig . [ fig:2](a ) stable magnetic structures in the @xmath79 direction for the systems with 2d planar ( @xmath79 , @xmath80 ) af interactions , and holes being doped _ only _ into @xmath33 orbital , i.e. , assuming empty @xmath32 orbitals , @xmath81 , in eq . ( [ eq:2 ] ) . in addition , we also studied the above case but when the hopping between @xmath33 orbitals occurs only along the @xmath79 direction , which resembles holes doped into @xmath82 orbitals in the manganites [ stable phases are displayed in fig . [ fig:2](b ) ] . one finds that for this @xmath26-type orbital the canted ( ca ) structure is stabilized in a broad range of parameters , in contrast with @xmath8 case where af order in the plane can be conserved . this is one of the reasons why the af order along the @xmath43 direction can be so easily destabilized by doping in lamno@xmath3 , leading to the fm metallic phase @xcite . by looking at fig . [ fig:2](a ) , one can also realize better the role of orbital degeneracy : even if we had a system with planar af interactions and we doped it with holes into , e.g. , @xmath33 orbitals , we would not get the af plane stable for a _ realistic _ range of parameters ! instead , one af bond has to be broken . this tendency is suppressed if we have two degenerate orbitals as in eq . ( [ eq:2 ] ) . , with displayed magnetic order along the @xmath79 axis , as obtained by solving eq . ( [ eq:2 ] ) and assuming : ( a ) no electrons in @xmath32 orbital , and ( b ) _ in addition _ no hopping along the @xmath43 direction . the af order for @xmath83 [ marked in grey ( red online ) ] was found by finite size scaling of the results obtained for large clusters . note that the af ( fm ) order is stable in the entire range of @xmath63 along the @xmath80 ( @xmath43 ) axis [ though the type of order along @xmath43 direction does not matter in ( b ) ] . ] in summary , we have shown that the existence of the metallic @xmath0-af phase in doped la@xmath1sr@xmath2vo@xmath3 can be explained by the de model with extremely simplified se interactions . the stability of this phase is due to the specific features of the @xmath8 orbitals namely due to strictly 2d and strictly flavor conserving hopping between these degenerate orbitals @xcite . in our view these are very prospective results and , in connection with those of ref . @xcite , they suggest that including the real se interactions could indeed explain the observed insulating phase and the metal - insulator transition which occurs at finite doping .	we investigate the role of the double exchange mechanism for stability of the metallic @xmath0-type antiferromagnetic ( @xmath0-af ) phase , which was observed experimentally in hole doped la@xmath1sr@xmath2vo@xmath3 . the double exchange model treats localized electrons in @xmath4 orbitals as classical @xmath5 spins , which interact by hund s exchange @xmath6 with @xmath7 electrons in partly filled @xmath8 orbitals . including strong on - site coulomb repulsion @xmath9 between @xmath8 electrons , and using slave boson method we demonstrate that @xmath0-af and metallic phase can be stabilized due to the specific features of the hopping in degenerate and partly filled @xmath8 orbitals .
the luminosity function ( lf ) of galaxies is a fundamental statistical tool to study the populations of galaxies . its dependence on morphological type , on wavelength and on look - back time provides constraints on the evolution of the properties of the whole population of galaxies , of the populations of various morphological types and on their contribution to the luminosity density at different wavelength . the parameters derived by the best - fitting of the lf , the characteristic luminosity m@xmath13 , the slope @xmath9 and the normalization @xmath14 , provide strong constraints to the models of galaxy formation . the recent estimates of the lf based on local wide surveys in the optical , such as the two degree field ( 2df ; folkes et al . 1999 ; madgwick et al . 2002 ; norberg et al . 2002 ) and the sloan digital sky survey ( sdss ; blanton et al . 2003 ) and in the near - ir , such as the two micron all sky survey ( 2mass ; kochanek et al . 2001 ) , have provided a comprehensive view of the local lf for different morphological types and wavelengths . it is now well established that the lf depends on morphological type , that the faint - end is increasingly dominated by galaxies with late - type morphology and spectra and that their number density increases going to lower luminosities . on the contrary , the bright - end is dominated by early - type galaxies whose fraction increases with luminosities ( e.g. marzke et al . 1994 , 1998 ; folkes et al . 1999 ; kochanek et al . 2001 ; bell et al . 2003 ) . the studies of the lf based on optically selected redshift surveys at @xmath19 have provided the first evidences of a differential evolution of galaxies . lilly et al . ( 1995a ) , using the canada france redshift survey ( cfrs ; lilly et al . 1995b ) show that the rest - frame b - band lf of the blue population of galaxies brightens by about 1 mag to @xmath20 contrary to the red population which shows a little evolution over the redshift range probed . they used the observed i - band photometry to derive the rest - frame b - band luminosities . subsequent studies confirmed the different behaviors followed by the lf of the different populations of galaxies at various redshifts ( e.g. lin et al . 1997 ; liu et al . 1998 ) and the differential evolution they underwent ( e.g. bell et al . 2004 ; wolf et al . 2003 ) . at higher redshift ( @xmath21 ) the studies of the lf in the optical rest - frame confirm the presence of the bi - modality and provide evidences of luminosity and of density evolution down to @xmath22 ( e.g. poli et al . 2003 ; gabasch et al.2004 ; giallongo et al . 2005 ) . contrary to the light at uv and at optical wavelengths , the near - ir light is less affected by dust extinction and by ongoing star formation e.g. rix & riecke 1993 ; kauffmann & charlot 1998 ) . moreover , since the near - ir light of a galaxy is dominated by the evolved population of stars , the near - ir light is weakly dependent on galaxy type and it is more related to the stellar mass of the galaxy . therefore , the evolution of the lf of galaxies in the near - ir rest - frame can provide important clues on the history of the stellar mass assembly in the universe rather than on the evolution of the star formation . the studies of the k - band luminosity function conducted so far have shown no or little evolution of the population of galaxies at @xmath23 ( e.g. glazebrook et al . 1995 ; feulner et al . 2003 ; pozzetti et al . 2003 ) with respect to the local population ( kochaneck et al . 2001 ; cole et al . 2001 ; glazebrook et al . 1995 ; gardner et al . 1997 ) . on the contrary , evidences of evolution emerge at @xmath24 , even if some discrepancies are present among the results obtained by the various authors ( see e.g. cowie et al . 1996 ; pozzetti et al . 2003 ; drory et al . 2003 ; caputi et al . 2004 , 2005 ; dahlen et al . one of the possible reasons of these discrepancies is related to the rest - frame k - band luminosities , usually extrapolated from the observed k - band photometry making use of the best - fitting template which should reproduce the unknown sed of the galaxy . to minimize the uncertainties , photometry at wavelengths long - wards the rest - frame k - band would be required . given the obvious difficulties in getting deep mid - ir observations , some authors have used the observed k - band magnitudes to construct the rest - frame j - band lf ( e.g. pozzetti et al . 2003 ; bolzonella et al . 2002 ; dahlen et al . 2005 ) and to constrain the evolution of the lf in the near - ir . another reason could be that deep k - band observations allowing a good sampling of the lf at magnitudes fainter than m@xmath13 over a wide redshift range are not so common . consequently the lf faint - wards of the knee is progressively less constrained with increasing redshift affecting the lf estimate . in this work , we aim at measuring the rest - frame js - band and ks - band lf of galaxies and its evolution to @xmath25 through a sample of about 300 galaxies selected at ks@xmath0 on the hubble deep field - south ( hdf - s ) . the extremely deep near - ir observations collected on this field coupled with the optical hst data have allowed us to probe the lf over a wide redshift range and to sample the faint end of the lf down to unprecedented faint luminosities . we use photometric redshift spectroscopically calibrated with a sample of more than 230 spectroscopic redshift , @xmath26 of which in the hdf - s . we pay particular attention to the calculation of the rest - frame near - ir luminosities by comparing different methods and libraries of models . in 2 we describe the photometric catalogue and the criteria adopted to construct the final sample of galaxies used to derive the lf . in 3 we describe the procedure used to obtain calibrated photometric redshifts and we present the redshift and color distribution of galaxies . in 4 we discuss the derivation of the rest - frame js - band and ks - band absolute magnitudes and assess the influence of the different estimates on the lf . the resulting lf at various redshifts is derived in 5 . in 6 our lfs are compared with the local estimates and with those previously obtained by other authors at comparable redshift to depict the evolution of the lf of galaxies up to @xmath27 . in 7 we summarize and discuss the results . throughout this paper , magnitudes are in the vega system unless explicitly stated otherwise . we adopt an @xmath28 , @xmath29 cosmology with h@xmath30 km s@xmath31 mpc@xmath31 . and drop at fainter magnitudes in agreement with the 100% completenes estimated at ks@xmath32 . lower panel photometric errors as a function of ks - band magnitude for the whole sample of galaxies in the hdf - s . at ks@xmath33 the photometric error is still lower than 0.2 mag for most of the galaxies while is much larger at fainter magnitudes . the dotted line marks the limiting magnitude of the k23 selected sample.,width=302,height=302 ] the ks - band selected catalog we use has been obtained by combining the version 2 of the hubble space telescope ( hst ) images in the u@xmath34 , b@xmath35 , v@xmath36 and i@xmath37 bands ( casertano et al . 2000 ) with the deep ( @xmath38 hrs per filter ) vlt - isaac observations centered on the hdf - s carried out in the framework of the fires project ( franx et al . 2000 ; labb et al . 2003a ) in the js , h and ks bands . the formal limiting magnitudes ( 3@xmath39 within 0.2 arcsec@xmath40 ) reached are 26.7 , 29.1 29.5 , 28.6 , 26.2 , 24.5 and 24.4 in the u@xmath34 , b@xmath35 , v@xmath36 , i@xmath37 , js , h and ks bands respectively . the detection has been performed using sextractor ( bertin & arnouts 1996 ) on the ks - band image . the ks - band magnitude is the mag_best measured by sextractor while the colors have been measured within the ks - band detection isophote by using sextractor in double image mode . the near - ir data reduction , the detection and the magnitude estimate are described in detail in saracco et al . ( 2001 ; 2004 ) . to construct our catalog , we considered the area of the hdf - s fully covered by all the images in the various bands resulting in an area of @xmath41 arcmin@xmath40 . on this area we detected 610 sources at ks@xmath42 . in fig . 1 the observed number counts are shown . from this ks - band selected catalog we extracted a complete sample of galaxies to derive the lf . on the basis of the simulations described in saracco et al . ( 2001 ) , we estimated that our catalogue is 100% complete at ks@xmath32 . this is consistent with the number counts shown in fig . 1 which raise till ks@xmath43 . down to ks@xmath33 the near - ir photometric errors are lower than 0.2 magnitudes while they rapidly increase at fainter magnitudes as shown in the lower panel of fig . 1 making very uncertain the photometric redshift estimate . thus , according to these considerations , we constructed the sample by selecting all the sources ( 332 ) brighter than ks=23 ( k23 sample hereafter ) . we then identified and removed the stars from the k23 sample relying both on the sextractor star / galaxy classifier and on the colors of the sources . in fig . 2 the stellar index ( class_star ) computed by sextractor for the 332 sources of the k23 sample is shown as a function of the ks - band apparent magnitude . for magnitudes brighter than ks@xmath44 point like sources ( class_star@xmath45 ) and extended sources ( class_star@xmath46 ) are well segregated while they tend to mix at fainter magnitudes . we defined stellar candidates those sources with magnitudes ks@xmath47 having a class_star@xmath48 both in the ks - band image and in the i@xmath37-band image . we obtained a sample of 23 stellar candidates . among them , 7 bright ( ks@xmath49 ) sources display spikes in the hst - wfpc images implying that they are most likely stars . for the remaining 16 candidates we verified that their optical and near - ir colors were compatible with those of stars 3 the colors js - ks and v@xmath36-i@xmath37 of the k23 sample are plotted as a function of the ks magnitude and of the js - ks color respectively . the 16 star candidates and the 7 bright stars are marked by filled points and starred symbols respectively . as can be seen , all but 2 of the candidates occupy the stellar locus at j - k@xmath50 and are well segregated both in the color - magnitude and in the color - color plane . on the contrary , the two candidates lying far from the stellar locus of fig . 3 display colors not compatible with those of stars . in particular , while their j - k color ( j - k@xmath51 ) could be consistent with the near - ir color of an m6-m8 spectral type , the other optical colors differ even more than 2 magnitudes from those characterizing these stars . thus , these two sources are most probably misclassified compact galaxies . this has made us to classify 21 stars out of the 23 candidates originally selected on the basis of the sextractor stellar index at ks@xmath47 . at magnitudes fainter than ks=21 other sources ( 18 ) lie in the locus occupied by stars defined by a color j - k@xmath50 . by a visual inspection of the 18 sources on the hst images we clearly identified the extended profile for 7 of them . on the contrary , the remaining 11 sources have fwhm and surface brightness profile consistent with a point - like source . we probed the stellar nature of these 11 star candidates by comparing their observed spectral energy distribution ( sed ) defined by broad - band photometry with a set of spectral templates of stellar atmospheres from the kurucz s atlas ( kurucz 1993 ) . the observed seds of all the 11 star candidates are very well fitted by those of m and k stars . all but one are not targets of spectroscopic observations , as we verified through the eso archive data available . the spectrum available for the remaining one ( j32m58.07s-3258.9 ) confirms its stellar nature as m star . in the left panel of fig . 4 we show the reduced @xmath52 of the best fitting template to the sed of the 18 sources with j - k@xmath50 obtained with the stellar atmosphere templates of kurucz ( @xmath53 ) and with the stellar population templates of bc03 ( @xmath54 ) . it is interesting to note the effectiveness of this method in identifying stars and galaxies among faint k - selected objects . in the right panel of fig . 4 we show , as example , the seds of 5 out of the 11 stars together with the relevant best - fitting kurucz s templates . therefore , we classified stars also these 11 sources and we removed a total of 32 stars from the k23 sample which results in 300 galaxies . photometric redshifts have been derived by comparing the observed flux densities to a set of synthetic templates based on the latest version of the bruzual & charlot ( 2003 ) models . the @xmath52 minimization procedure of hyperz ( bolzonella et al . 2000 ) has been used to obtain the best fitting templates . the set of templates used has been selected among a large grid of models to provide us with the most accurate estimate of the redshift for the galaxies in our photometric sample . the grid from which we selected the final set of templates , includes declining sfrs with time scale @xmath55 in the range 0.1 - 15 gyr besides the simple stellar population ( ssp ) and the constant star formation ( cst ) models . the templates have been produced at solar metallicity with salpeter , scalo and miller - scalo initial mass functions ( imfs ) . various ranges of extinction a@xmath56 and two different extinction laws ( calzetti et al . 2000 and prevot et al . 1984 ) have been considered in the best - fitting procedure . the selection of the best set of templates has been performed by comparing the photometric redshifts @xmath57 provided by the various sets of templates obtainable by the whole grid of models , with the spectroscopic redshifts @xmath58 of a control sample of 232 galaxies . this sample includes 151 galaxies in the hdf - n ( cohen et al . 2000 ; dawson et al . 2001 ; fernndez - soto et al . 2002 ) and 81 in the hdf - s ( vanzella et al . 2002 ; sawicki & malln - ornelas 2003 ; labb et al . 2003b ; trujillo et al . 2004 ; rigopoulou et al . 2005 ) with known spectroscopic redshift . the best set of templates is the one minimizing the mean and the standard deviation of the residuals defined as @xmath59 and the number of outliers defined as those sources with @xmath60 . both mean and standard deviation are iteratively computed applying a @xmath61 clipping till the convergence . the final number of outliers includes all the galaxies removed by the clipping procedure . .parameters defining the best set of templates obtained by comparing the photometric redshift with the spectroscopic redshift of 232 galaxies , 151 in the hdf - n and 81 in the hdf - s . [ cols="<,^",options="header " , ] based on k - band imaging of an optically selected sample based on k - band data of an optically selected sample and on photometric redshift . based on photometric redshift . the number of galaxies ( 439 ) is the total number of galaxies in the sample . we have probed the evolution of the rest - frame js - band and ks - band lfs of field galaxies with a complete sample of about 300 galaxies selected in the hdf - s at ks@xmath0 ( vega ) . photometric redshifts have been obtained from template sed fitting to u@xmath34 , b@xmath35 , v@xmath36 , i@xmath37 , js , h and ks photometry after having calibrated and optimized the procedure and the set of templates with a control sample of 232 spectroscopic redshifts , 151 from the hdf - n and 81 from the hdf - s . the accuracy in the redshift estimate we obtained is 0.06 ( rms ) . we investigate the reliability of the rest - frame near - ir absolute magnitudes obtained using the conventional method based on the extrapolation of the observed photometry on the best - fitting template and using the photometry approaching the rest - frame near - ir . we find that the rest - frame js - band absolute magnitudes obtained through the photometry in the redder bands ( h and ks according to the redshift of the galaxy ) are consistent with those obtained from the js - band photometry at least down to @xmath1 . this shows that the lf is not dependent either on the extrapolation made on the best - fitting template or on the library of models used . we derived the js - band and the ks - band lf in the three redshift bins [ 0,0.8 ) , [ 0.8,1.9 ) and [ 1.9,4 ) centered at the median redshift @xmath8 , @xmath12 and @xmath15 . each bin contains 101 , 100 and 84 galaxies of which 52 ( 50% ) , 12 ( 12% ) and 10 ( 12% ) with spectroscopic redshift respectively . the analysis of the observed lf at different redshifts and the comparison with those previously found by other authors at comparable redshift and at @xmath10 provided the following results : \1 . we find hints of a raise of the faint end ( m@xmath6 and m@xmath7 ) of the near - ir lf at @xmath8 . the raise is defined by 9 galaxies at @xmath62 with irregular morphology as found for the local lf at optical wavelength ( e.g. marzke et al . 1994 ; marzke et al . 1998 ; zucca et al . 1997 ; folkes et al . 1999 ) . they account for a comoving number density @xmath63 mpc@xmath64 at @xmath65 , almost one order of magnitude higher than that of brighter galaxies . however , given the low statistics , such raise of the faint end needs a larger sample to be established ; \2 . we find no evidence of a steepening with redshift of the near - ir lfs of galaxies in the hdf - s . the value of @xmath9 we find is consistent with the local value . given the size of our sample , we estimate that if @xmath9 changes with redshift its evolution is constrained within @xmath66 from @xmath67 to @xmath27 . we do not find evidence of strong evolution of the lf up to @xmath3 , where 50% of our galaxies has spectroscopic redshift . the comparison with the local estimates of the near - ir lf ( cole et al . 2001 ; kochanek et al . 2001 ) shows that m@xmath13 has evolved not more than 0.2 - 0.3 magnitude in this redshift range , while the number density of galaxies ( @xmath14 ) has not evolved significantly in this redshift range , in agreement with the previous studies ( e.g. drory et al . 2003 ; pozzetti et al . 2003 ; feulner et al . 2003 ; glazebrook et al . 1995 ; caputi et al . 2004 , 2005 ) . we clearly detect an evolution of the lf at @xmath68 characterized by a brightening of m@xmath13 of about 0.6 and a decrease of @xmath14 by a factor 2 - 3 to @xmath12 both in the js - band and in the ks - band . by fixing @xmath69 , this evolution is characterized by @xmath70 ( at 1@xmath39 ) coupled with @xmath71 ( at @xmath72 ) . the brightening persists ( at 1@xmath39 ) up to @xmath15 together with the decline of @xmath14 ( at @xmath73 ) even if at a lower extent . our results agrees , at least qualitatively , with most of the analysis previously done ( drory et al . 2003 ; pozzetti et al . 2003 ; feulner et al . 2003 ; caputi et al . 2004 , 2005 ) while deviates from the recent estimate of dahlen et al . taking into account that the near - ir light is rather tightly connected to the stellar mass , our results can give insights on the evolution of the stellar mass of the galaxies besides their luminosity . up to @xmath74 , clearly little luminosity evolution and no density evolution is observed . this suggests that the population of local bright galaxies was already formed at @xmath75 and that their stellar mass growth was already completed at this redshift . this agrees with the results on the evolution of the stellar mass density obtained by many authors : fontana et al . ( 2003 ) and rudnick et al . ( 2003 ) in the hdf - s , fontana et al . ( 2004 ) on the k20 sample ( cimatti et al . 2002 ) , yamada et al . ( 2005 ) in the subaru deep survey field , bundy et al . ( 2005 ) and drory et al . ( 2005 ) in the goods and fdf fields and feulner et al . ( 2005a ) in the munics fields . indeed , all of them find very little or even no evolution in the stellar mass function of galaxies and in the specific star formation rate ( ssfr ) at @xmath19 for high - mass galaxies . at @xmath16 the evolution is stronger and the larger brightening observed is accompanied by a decrease of the number density of bright / massive galaxies . in the redshift bin [ 0.8,1.9 ) the number density of bright galaxies is @xmath76% of the local value and reaches @xmath77% at @xmath78 . this decline in the number density suggests that up to @xmath79% of the local massive / bright galaxies has grown at @xmath17 through star formation , merging or both . thus , we should expect to observe massive star - forming galaxies with disturbed / irregular morphology and merging systems in this high redshift range , as indeed found in same cases ( e.g. daddi et al . 2004 ) . on the other hand , the observed evolution implies also that at least 30% of the local bright / massive galaxies was already in place at @xmath80 as indeed found by some authors ( fontana et al . 2004 ; glazebrook et al . 2004 ; f@xmath81rster schreiber et al . pozzetti et al . ( 2003 ) and caputi et al . ( 2005a ) show that the bright end of the lf is dominated by early - type galaxies and that their contribution to the bright end does not decline with redshift . thus , that 30% should be dominated by early - types implying that we should observe massive evolved galaxies fully assembled at @xmath80 and even at @xmath82 . this is indeed observed by many authors ( cimatti et al . 2004 ; mccarthy et al . 2004 ; saracco et al . 2004 , 2005 ; longhetti et al . 2005 ; daddi et al . 2005 ; labb et al . it is worth noting that this picture could be accounted for , at least qualitatively , in the hierarchical picture of galaxy formation as suggested by the recent results obtained by nagamine et al . ( 2005 ) . from these considerations we can gather that the growth of massive galaxies does not follow a unique way but displays different behaviours . a significant fraction ( 50 - 70% ) of the brighetst / most massive galaxies increases their stellar mass over a large redshift range at @xmath21 . the remaining fraction reaches their final mass in a narrower redshift range at @xmath18 since it is already in place by this redshift . this suggests that , for the former , the stellar mass growth has been less efficient and has proceeded more slowly than for the latter . on the contrary , in the latter , the stellar mass has to be grown rapidly in a short interval , surely much shorter than 1 gyr given their redshift . this is supported by the recent results derived by mid - ir massive galaxies observed at high - z ( e.g. caputi et al . 2005b ) and suggests a high efficiency in the accretion of the stellar mass in massive haloes in the early universe , possibly through a very efficient star formation . the recent results obtained on the evolution of the ssfr of galaxies seems to point toward this direction . the strong increase in the ssfr of the most massive galaxies with redshift ( feulner et al . 2005b ) favors indeed an efficient sfr in the brightest galaxies at @xmath83 constraining their growth in a very short interval . this work is based on observations made with the eso - vlt telescopes at paranal observatory under the programs 164.o-0612 and 70.b-0144 and with the nasa / esa hubble space telescope . we are in debt with m. bolzonella for her helpfulnes . we thank the anonymous referee for the useful comments which improved the presentation of the results . balogh m. l. , christlein d. , zabludoff a. i. , zaritsky d. , 2001 , apj , 557 , 117 bell e. f. , mcintosh d. h. , katz n. , weinberg m. d. , 2003 , apjs , 149 , 289 bell e. f. , et al . 2004 , apj , 608 , 752 bertin e. , arnouts s. , 1996 , a&as , 117 , 393 blanton m. r. , et al . , 2003 , apj , 592 , 819 bolzonella m. , miralles j .- m . , pello ' r. , 2000 , a&a , 363 , 476 bolzonella m. , pell r. , maccagni d. 2002 , a&a , 395 , 443 bruzual a.,g . & charlot s. 2003 , mnras , 344 , 1000 bundy k. , ellis r. s. , conselice c. j. , 2005 , apj , 625 , 621 calzetti d. , armus l. , bohlin r. c. , kinney a. l. , koornneef j. , storchi - bergmann t. , 2000 , apj , 533 , 682 caputi k. i. , dunlop j. s. , mclure r. j. , roche n. d. , 2004 , mnras , 353 , 30 caputi k. i. , dunlop j. s. , mclure r. j. , roche n. d. , 2005a , mnras , 361 , 607 caputi k. i. , dole h. , lagache g. , et al . , 2005b , apj , in press , [ astro - ph/0510070 ] casertano s. , et al . , 2000 , aj , 120 , 2747 cimatti a. , mignoli m. , daddi e. , et al . 2002 , a&a , 392 , 395 cimatti a. , et al . , 2004 , nat . , 430 , 184 cohen j. g. , hogg d. w. , blandford r. , et al . 2000 , apj , 538 , 29 cole s. , norberg p. , baugh c. m. , et al . 2001 , mnras , 326 , 255 coleman g. d. , wu c .- c . , weedman d. w. 1980 , apjs , 43 , 393 cowie l. l. , songaila a. , hu e. m. , cohen j. g. , 1996 , aj , 112 , 839 daddi e. , cimatti a. , renzini a. , et al . , 2004 , apj , 600 , l127 daddi e. , et al . , 2005 , apj , 626 , 680 dahlen t. , mobasher b. , somerville r. s. , moustakas l. a. , dickinson m. , ferguson h. c. , giavalisco m. , 2005 , apj , 631 , 126 dawson s. , stern d. , bunker a. j. , spinrad h. , dey a. , 2001 , aj , 122 , 598 drory n. , bender r. , feulner g. , hopp u. , maraston c. , snigula j. , hill g. j. 2003 , apj , 595 , 698 drory n. , salvato m. , gabash a. , bender r. , hopp u. , feulner g. , pannella m. , 2005 , apj , 619 , l131 fernndez - soto a. , lanzetta k. m. , chen h .- w . , levine b. , yahata n. , 2002 , mnras , 330 , 889 feulner g. , bender r. , drory n. , hopp u. , snigula j. , hill g. j. , 2003 , mnras , 342 , 605 feulner g. , goranova y. , drory n. , hopp u. , bender r. , 2005a , mnras , 358 , l1 feulner g. , gabasch a. , salvato m. , drory n. , hopp u. , bender r. , 2005b , apj , 633 , l9 fioc m. , rocca - volmerange b. , 1997 , a&a , 326 , 950 folkes s. , ronen s. , price i. , et al . , 1999 , mnras , 308 , 459 fontana a. , donnarumma i. , vanzella e. , et al . 2003 , apj , 594 , l9 fontana a. , pozzetti l. , donnarumma i. , et al . 2004 , a&a 424 , 23 f@xmath81rster schreiber n. m. , et al . 2004 , 616 , 40 franx m. , et al . , 2000 , messenger , 99 , 20 gabasch a. , et al . , 2004 , a&a , 421 , 41 gardner j. p. , sharples r. m. , frenk c. s. , carrasco b. e. , 1997 , apj , 480 , l99 giallongo et al . 2005 , apj , 622 , 116 glazebrook k. , peacock j. a. , miller l. , collins c. a. , 1995 , mnras , 275 , 169 glazebrook k. , abraham r. g. , mccarthy p. j. , et al . , 2004 , nat . , 430 , 181 huang j .- s . , glazebrook k. , cowie l. l. , tinney c. , 2003 , apj , 584 , 203 ilbert o. , et al . 2005 , a&a , 439 , 863 kashikawa n. , et al . , 2003 , aj , 125 , 53 kochanek c. s. , pahre m. a. , falco e. e. , et al . 2001 , apj , 560 , 566 labb i. , et al . 2003a , aj , 125 , 1107 labb i. , et al . 2003b , apj , 591 , l95 labb i. , et al . 2005 , apj , 624 , l81 lilly s. j. , tresse l. , hammer f. , crampton d. , le fevre o. 1995 , apj , 455 , 108 lilly s. j. , le fevre o. , crampton d. , hammer f. , tresse l. , 1995b , apj , 455 , 50 lin h. , yee h. k. c. , carlberg r. g. , ellingson e. , 1997 , apj , 475 , 494 liu c. t. , green r. f. , hall p. b. , osmer p. s. , 1998 , aj , 116 , 1082 longhetti m. , saracco p. , severgnini p. , et al . , 2005 , mnras , 361 , 897 loveday j. , 2000 , mnras , 312 , 557 madgwick d. s. , et al . , 2002 , mnras , 333 , 133 mannucci f. , basile f. , poggianti b. m. , cimatti a. , daddi e. , pozzetti l. , vanzi l. , 2001 , mnras , 326 , 745 maraston c. , 2005 , mnras , 362 , 799 marzke r. o. , geller m. , j. , huchra j. , corwin h. , g. jr . , 1994 aj , 108 , 437 marzke r. o. , da costa l. n. , pellegrini p. s. , willmer c. n. a. , geller m. , j. , 1998 , apj , 503 , 617 mccarthy p. j. , le borgne d. , crampton d. , et al . 2004 , apj , 614 , l9 mobasher b. , sharples r. m. , ellis r. s. , 1993 , mnras , 263 , 560 nagamine k. , cen r. , hernquist l. , ostriker j. p. , springer f. , 2005 , apj , 627 , 608 norberg p. , et al . , 2002 , mnras , 336 , 907 poli f. , et al . , 2003 , apj , 593 , l1 pozzetti l. , cimatti a. , zamorani g. , et al . , 2003 , a&a , 402 , 837 prvot m. l. , lequeux j. , prvot l. , maurice e. , rocca - volmerange b. , 1984 , a&a , 132 , 389 rigopoulou d. , vacca w. d. , berta s. , franceschini a. , aussel h. , 2005 , a&a , 440 , 61 rix h. , riecke m. j. , 1993 , apj , 418 , 123 rudnick g. , rix h .- w . , franx m. , et al . 2003 , apj , 599 , 847 saracco p. , giallongo e. , cristiani s. , et al . , 2001 , a&a , 375 , 1 saracco p. , longhetti m. , giallongo e. , et al . 2004 , a&a , 420 , 125 saracco p. , longhetti m. , severgnini p. , et al . , 2005 , mnras , 357 , l40 sawicki m. , malln - ornelas g. , 2003 , aj , 126 , 1208 schmidt m. 1968 , apj , 151 , 193 szokoly g. p. , subbarao m. u. , connolly a. j. , mobasher b. , 1998 , apj , 492 , 452 takeuchi t. t. , yoshikawa k. , ishii t. t. 2000 , apjs , 129 , 1 trujillo i. , et al . , 2004 , apj , 604 , 521 vanzella e. , et al . , 2002 , a&a , 396 , 847 vanzella e. , et al . , 2001 , aj , 122 , 2190 wolf c. , meisenheimer k. , rix h. w. , borch a. , dye s. , kleinheinrich m. , 2003 , a&a , 401 , 73 yamada t. , et al . , 2005 , apj , 634 , 861 zucca e. , zamorani g. , vettolani g. , et al . , 1997 , a&a , 326 , 477	we present the rest - frame js - band and ks - band luminosity function ( lf ) of a sample of about 300 galaxies selected in the hdf - s at ks@xmath0 ( vega ) . we use calibrated photometric redshift together with spectroscopic redshift for 25% of the sample . the accuracy reached in the photometric redshift estimate is 0.06 ( rms ) and the fraction of outliers is 1% . we find that the rest - frame js - band luminosities obtained by extrapolating the observed js - band photometry are consistent with those obtained by extrapolating the photometry in the redder h and ks bands closer to the rest - frame js , at least up to @xmath1 . moreover , we find no significant differences among the luminosities obtained with different spectral libraries . thus , our lf estimate is not dependent either on the extrapolation made on the best - fitting template or on the library of models used to fit the photometry . the selected sample has allowed to probe the evolution of the lf in the three redshift bins [ 0;0.8 ) , [ 0.8;1.9 ) and [ 1.9;4 ) centered at the median redshift @xmath2 $ ] and to probe the lf at @xmath3 down to the unprecedented faint luminosities m@xmath4 and m@xmath5 . we find hints of a raise of the faint end ( m@xmath6 and m@xmath7 ) near - ir lf at @xmath8 , raise which can not be probed at higher redshift with our sample . the values of @xmath9 we estimate are consistent with the local value and do not show any trend with redshift . we do not see evidence of evolution from @xmath10 to @xmath8 suggesting that the population of local bright galaxies was already formed at @xmath11 . on the contrary , we clearly detect an evolution of the lf to @xmath12 characterized by a brightening of m@xmath13 and by a decline of @xmath14 . to @xmath12 m@xmath13 brightens by about 0.4 - 0.6 mag and @xmath14 decreases by a factor 2 - 3 . this trend persists , even if at a less extent , down to @xmath15 both in the js - band and in the ks - band lf . the decline of the number density of bright galaxies seen at @xmath16 suggests that a significant fraction of them increases their stellar mass at @xmath17 and that they underwent a strong evolution in this redshift range . on the other hand , this implies also that a significant fraction of local bright / massive galaxies was already in place at @xmath18 . thus , our results suggest that the assembly of massive galaxies is spread over a large redshift range and that the increase of their stellar mass has been very efficient also at very high redshift at least for a fraction of them . [ firstpage ] galaxies : evolution galaxies : luminosity function galaxies : formation .
fano resonance @xcite , a phenomenon that causes asymmetric conductance peaks due to interference between resonant and nonresonant states , is observed in several mesoscopic systems @xcite . recently , first observations of fano resonances are reported in crossed carbon nanotubes @xcite . al . _ claimed that fano resonances can not be observed in single - walled carbon nanotubes ( swnts ) and only appear when two nanotubes cross each other because the fano resonances are attributed to the resonant scattering by the crossed region . fano resonances are observed in multi - walled carbon nanotubes ( mwnts ) as well , and are considered to be due to impurities , defects or disorders @xcite . also theoretical prediction of fano resonances for a multiply - connected carbon nanotube has been reported where asymmetric resonance peaks are originated from the scattering by the localized states on the heptagonal defects @xcite . on the contrary , babi _ et . _ reported that , even in swnts without crossed contacts , fano resonances can be observed due to the interference between narrow and wide orbitals @xcite . in this report , we investigate the possibility of fano resonances in a single armchair nanotube with the aid of the c@xmath0-algebraic method as in our previous work on the ab - ring @xcite . the tight - binding hamiltonian is adopted for describing the electronic states of swnts and only the linear response regime is considered . we show that , when the swnt - electrode coupling is axially asymmetric , fano - like resonances exist for an armchair swnt . we consider a system consisting of the two - dimensional electrodes and a finite - size armchair swnt without caps ( see fig.[latticenanotube](a ) ) . in this section , the electronic structure of the finite - size swnt is reviewed @xcite . the nanotube can be regarded as a rolled graphene sheet and it is convenient to adopt the primitive lattice vectors of the graphene sheet for specifying the atomic positions : @xmath1 , @xmath2 , where @xmath3 and @xmath4 are the integers , @xmath5 , @xmath6 , @xmath7 and @xmath8 is the nearest - neighbor bond length . the armchair nanotube is constructed by rolling the graphene sheet along the @xmath9-direction in fig.[latticenanotube](b ) . the nearest - neighbor tight - binding hamiltonian is used to describe the nanotube : @xmath10 where @xmath11 is the annihilation operator of the @xmath12 electron with spin @xmath13 at site @xmath14 , @xmath15 is the nearest - neighbor hopping strength and the on - site energy @xmath16 linearly depends on the gate voltage applied to the nanotube . throughout this report , the repeated spin index means the summation over it . the coulomb interaction is neglected . in the case of the armchair nanotube , the site @xmath14 is identified with @xmath17 and , then , the condition : @xmath18 is imposed . in terms of new annihilation operators : + @xmath19 + where @xmath20 , @xmath21 and @xmath22 , the hamiltonian of an armchair swnt of length @xmath23 becomes : @xmath24 . \notag \end{aligned}\ ] ] we remark that @xmath25 satisfies : @xmath26 . the hamiltonian @xmath27 is , then , diagonalized by a unitary transformation : @xmath28 , @xmath29 . and one has @xmath30 where @xmath31 . note that , as @xmath32 for @xmath33 , the armchair swnt is metallic . . ( b ) unit cell ( dotted line ) for an armchair nanotube in 2d graphite sheet . ] now we consider the effect of the electrodes . the total hamiltonian is given by : @xmath34 where @xmath35 is the energy of the left or right electrode and @xmath36 is the single - electron energy : @xmath37 with @xmath38 the source - drain voltage . the second term @xmath27 stands for the nanotube hamiltonian ( [ hamiltonian ] ) and @xmath39 represents the tunneling between the nanotube and electrodes . we denote @xmath40 and @xmath41 are given by : @xmath42 @xmath43 where @xmath44 , @xmath45 , @xmath46 , + @xmath47 is the phase of @xmath48 and the product @xmath49 stands for the tunneling strength between each edge site of the nanotube and the electrode @xmath50 . in order to calculate the steady - state current from the left to right electrodes , we derive the asymptotic in - coming fields @xmath51 that are solutions of the heisenberg equation : @xmath52=-\omega _ { k\lambda}\beta _ { k\sigma \lambda}$ ] . they are given by : @xmath53 where @xmath54 , + @xmath55 , + @xmath56 , + @xmath57 , + @xmath58 , @xmath59 . in the above , @xmath60 . as in @xcite , if the system admits no bound state , a nonequilibrium steady state ( ness ) is rigorously constructed from the time - evolution . the ness is found to satisfy wick s theorem and is fully specified by the two - point function : @xmath61 where @xmath62 indicates the fermi - dirac distribution function of the electrode @xmath50 with temperature @xmath63 and chemical potential @xmath64 . + the current operator is the time - derivative of the left - electrode number operator : @xmath65 and its average is @xmath66 by expressing @xmath67 with the asymptotic field operators @xmath51 and by using ( [ eq : soukan ] ) , the steady - state current casts into the landauer formula @xcite : @xmath68 where @xmath69max@xmath70 . the transmission coefficient @xmath71 is @xmath72 ) . here , we are interested in the linear response regime at @xmath73 and the variables @xmath74 and @xmath75 are taken to be the same and @xmath76-independent , i.e. , @xmath77 . as the nanotube length is finite , the integrand of ( [ eq : denryu ] ) is a highly - oscillatory function of @xmath78 with period @xmath79 . so we consider the average conductance : @xmath80 where @xmath38 is the source - drain voltage . in fig . [ carbonkinji1 - 1 ] , the average conductance for @xmath81 is shown as a function of @xmath16 which corresponds to the gate voltage . the structural parameters of the nanotube are @xmath82 , which corresponds to the length of @xmath83 @xmath84 m and the diameter of @xmath85 nm . one can see that the envelope forms an asymmetric resonance peak around at @xmath86 . the case of @xmath87 corresponds to the half - filled swnt band and the flat region observed from @xmath88 to @xmath89 comes from the conduction band . the sharp peak below @xmath89 seems to arise from the van hove singularity of the valence band nearest to the conduction band . we remark that the conductance is suppressed just below the conductance maximum ( ( @xmath90 ) . the dip seems to be caused by the interference between the two bands , the conduction and valence bands . this means that the valence band nearest to the conduction band contributes to the interference . in the system of the wave - guide geometry , even without the confined states , similar suppression of the transmission due to the interference has been reported @xcite . we note that such asymmetric peaks are very sensitive to the coupling between the nanotube and electrodes . in the above calculation , four tunneling matrix elements @xmath91 involved in @xmath92 are taken as the same real numbers . this assumption implies that the nanotube contacts with both electrodes _ via _ single sites . this choice breaks the axial symmetry of the coupling , but the left and right contacts are equivalent . indeed , when nanotube - electrode coupling is axially symmetric , fano - like shapes disappear in the conductance . this could be understood as follows : suppose that an asymmetric peak arises from the interference between axially symmetric and asymmetric states . then , if the coupling is axially symmetric , non - symmetric nanotube states do not contribute to the transport and the asymmetric peak would disappear . + and ( n , m)=(7000,15 ) . ] we found that conductance shows fano - like line shapes even in a single swnt due to the interference between conduction and valence bands . these peaks are strongly affected by the contact condition . our analysis implies that , even for the same swnts , fano resonances may or may not appear depending on the contact condition with electrodes . therefore , we think that the observations by kim _ et . al . _ @xcite and by the babi @xcite are consistent . the exact form of the peaks depends on the nanotube structure and the contact conditions . therefore , in order to reproduce experimentally observed conductance , more detail information on the experimental setting is required . moreover , the neglected effects of the electron - electron interaction should be taken into account . in the present work , we have shown the possibility of fano like conductance peak and its dependence on the contact condition . we believe the essence of these features is not altered . the authors thank professor m. eto , professor s. w. kim , professor b. li and professor s. pascazio for fruitful discussions and valuable comments . this work is partially supported by a grant - in - aid for scientific research of priority areas `` control of molecules in intense laser fields '' and the 21st century coe program at waseda university `` holistic research and education center for physics of self - organization systems '' both from the ministry of education , culture , sports , science and technology of japan and by waseda university grant for special research projects ( 2004a-161 ) .	electrical transport through single - walled carbon nanotubes ( swnts ) is investigated by using the nearest - neighbor tight - binding model coupled with two electron reservoirs . when the swnt - electrode coupling is not axially symmetric , asymmetric resonance peaks are found in the conductance and are considered to be due to the interference between two transport channels . these fano resonances are sensitive to the coupling with electrodes . when the coupling is axially symmetric , no asymmetric resonance peaks are observed . and fano resonance , carbon nanotube , c@xmath0-algebra , nonequilibrium steady states , tight - binding model 72.10.-d , 73.63.-b , 73.63.fg
in the solar neighborhood reside stars of differing metallicity . stars with a metallicity of approximately the solar value belong overwhelmingly to the galactic disk . stars of lower metallicity are on orbits that identify them as members of the galactic halo . the metallicity [ fe / h ] @xmath9 may be taken as a rough boundary between disk and halo stars . one tool for unravelling the evolution of the galaxy is the measurement of the chemical compositions of stars as a function of metallicity . in this paper , we present the first extensive series of measurements of the rubidium abundance in disk and halo stars . rubidium is potentially a special diagnostic of the neutron capture @xmath5-process . rubidium is present in two isotopic forms : @xmath10rb which is stable and @xmath11rb which with a half - life of 5 @xmath12 yr may be deemed effectively stable from the astrophysical point of view . as we remark later , astronomical detection of rb must rely on the rbi lines that do not permit measurement of the relative isotopic rb abundances from stellar spectra . hence , we discuss the elemental rb abundance . analysis of the solar system abundances of rb and adjoining elements shows that the neutron capture @xmath5- and @xmath4-processes are about equally responsible for the synthesis of rb . scrutiny of the @xmath5-process abundances shows that the ` main ' @xmath5-process not the ` weak ' @xmath5-process is the principal source of rb s @xmath5-process component . the main @xmath5-process which manufactured elements heavier than about rb is identified with the he - burning shell of intermediate and low mass agb stars . the weak @xmath5-process is identified with he - core and possibly c - core burning of massive stars . evolution of the galaxy s @xmath5 and @xmath4-process products is rather directly observed from the stellar abundances of elements that are predominantly attributable to either the @xmath5 or to the @xmath4-process . traditional tracers include ba for the @xmath5-process and eu for the @xmath4-process . elucidation of the operation of the neutron capture processes requires observations of more than a single element per process . as an example , we note that the abundance ratio of a ` light ' to a ` heavy ' @xmath5-process elements , say zr to ba , provides information on the integrated exposure of material to neutrons . rubidium with a roughly equal mix of @xmath5 and @xmath4-process contributions , and an unfavorable electronic structure for ready detection in stellar spectra would seem to be an element of little interest . closer inspection of the working of the @xmath5-process shows , however , that rb offers a unique insight into the process : rb s role as a monitor of the neutron density at the @xmath5-process site . along the @xmath5-process path , rb is preceded by krypton with the path entering kr at @xmath13kr and exiting at either @xmath14kr or @xmath15kr . unstable @xmath10kr controls the exit . at low neutron density at the @xmath5-process site , stable @xmath14kr is converted by neutron capture to @xmath10kr that decays to stable @xmath10rb with the path continuing to @xmath15sr . at high neutron densities , @xmath10kr does not @xmath16-decay but experiences a neutron capture and is converted to stable @xmath15kr . subsequent neutron capture by @xmath15kr leads by @xmath16-decay of @xmath11kr to ( effectively ) stable @xmath11rb . when a steady flow along the @xmath5-process path is attained , the density of a nuclide is given approximately by the condition that @xmath17constant where @xmath18 is the neutron capture cross - section of nuclide @xmath19 and @xmath20 is the abundance of that nuclide . since @xmath21 for the rb isotopes , the switch of the @xmath10kr branch from its low neutron density routing through @xmath10rb to its high neutron density routing through @xmath11rb increases the total rb abundance by about an order of magnitude relative to the abundance of other elements in this section of the @xmath5-process path , such as sr , y , and zr . the isotopic mix of rb is obviously altered as a function of neutron density but this is not measureable for cool stars . ( krypton is undetectable spectroscopically in cool stars . ) operation of the @xmath10kr branch is more complicated than sketched , for example , neutron capture from @xmath14kr feeds not only the @xmath10kr ground state but a short - lived isomeric state that at all reasonable neutron densities provides some leakage to @xmath10rb . a thorough discussion of the @xmath10kr branch is provided by beer & macklin ( 1989 ) and its use in determining the effective neutron density of the @xmath5-process in stars is discussed by tomkin & lambert ( 1983 ) and lambert et al . ( 1995 ) . when detailed abundance measurements are available as in the case of the carbonaceous chondrites , several branches along the @xmath5-process path serve as neutron densitometers but rb is the only low neutron density branch available to stellar spectroscopists . ( at higher neutron densities , a branch controlled by @xmath22zr is exploitable in cool stars showing zro bands [ lambert et al . 1995 ] . ) a primary reason for the near neglect of rb in reports on quantitative spectroscopy of stars is that it is a trace alkali element . the rb atom s low ionization potential ( 4.177ev ) ensures that rb is primarily ionized but the rare - gas electronic structure of rb@xmath23 provides resonance lines in the far ultraviolet . detection of rb via the rb resonance lines at 7800 and 7947 at the expected low rb abundances is possible for cool dwarfs and giants , as our exploratory synthetic spectra indicated . stars for observation at high - spectral resolution were selected from schuster & nissen s ( 1988 ) catalog of dwarfs , and from pilachowski , sneden , & kraft s ( 1996 ) list of giants . emphasis was placed on metal - poor stars such that the metallicity range @xmath24 [ fe / h ] @xmath2 is well represented but metallicities [ fe / h ] @xmath25 are poorly represented . the following sections describe the observations , the method of analysis , the results , and present an interpretation of the rb abundances relative to the abundances of other elements ( fe , y , zr , and ba ) obtained here . the program stars are listed in table 1 . they comprise 32 g and k dwarfs and subgiants with metallicities of @xmath26 [ fe / h ] @xmath1 and 12 g and k giants with metallicities of @xmath0 [ fe / h ] @xmath27 . the observations were made at the mcdonald observatory with the 2.7-m telescope and 2dcoud echelle spectrometer ( tull et al . all the program stars were observed at the f3 focus at a resolving power of r = 60,000 and , in addition , eight of the brighter stars were also observed at the f1 focus at a resolving power of 200,000 . in order to minimise the influence of cosmic rays , two observations in succession , rather than one longer observation , were generally made of each star . four different detectors were used for the observations with resolving power 60,000 : a texas instruments ccd with 15@xmath28m@xmath29 pixels in a 800@xmath30800 format , a tektronix ccd with 27@xmath28m@xmath29 pixels in a 512@xmath30512 format , the goddard advanced imaging system ccd with 21@xmath28m@xmath29 pixels in a 2048@xmath302048 format , and a tektronix ccd with 24@xmath28m@xmath29 pixels in a 2048@xmath302048 format . the first two of these ccds provided partial coverage of the wavelength interval @xmath315500 @xmath318000with large gaps between the end of one spectral order and the beginning of the next . the last two ccds , which are much larger , provided nearly complete coverage from @xmath314000 @xmath319000 ; coverage was complete from the start of this interval to 5600 and substantial , but incomplete , from 5600 to the end of the interval . the tektronix ccd with 24@xmath28m@xmath29 pixels in a 2048@xmath302048 format was used for the 200,000 resolving - power observations , which provided partial coverage of the region from @xmath315500 @xmath318000 . the typical signal - to - noise ratio of the extracted one - dimensional spectra is between 100 and 300 at red and near - infrared wavelengths for the 60,000 resolving - power observations , while it is typically between 100 and 250 at the same wavelengths for the 200,000 resolving - power observations . the only accessible rb lines in stellar spectra are the two rb resonance lines at 7800.3 and 7947.6 . since these lines are typically weak we concentrated our attention on the stronger 7800.3 line and did not pursue the 7947.6 line , which is half as strong as the 7800.3 line . figure [ spec ] shows examples of the 7800.3rb line in the program stars ; as may be seen , the line is partially blended with a stronger si line at 7800.0 . the data were processed and wavelength calibrated in a conventional manner with the iraf package of programs on a sparc5 workstation . the 7800.3 rb line was analyzed by spectrum synthesis because of the presence of the si line . equivalent widths were measured for lines of other elements used in the investigation ; these were lines of fe and fe and the available lines ( y , zr , ba , and nd ) of other heavy elements besides rb . lines suitable for measurement were chosen for clean profiles , as judged by inspection of the solar spectrum at high resolution and signal - to - noise ratio ( kurucz et al . 1984 ) , that could be reliably measured in all , or most of , the program stars . moore , minnaert , & houtgast ( 1966 ) was our primary source of line identification . the equivalent width of each line was measured with the iraf measurement option most suited to the situation of the line ; usually this was the fitting of a single , or multiple , gaussian profile to the line profile . these equivalent widths were measured from the spectra with 60,000 resolving power ; the 200,000 resolving - power spectra were not used because their much more limited wavelength coverage excluded most lines of interest . table 2 gives basic information for the rb line and the lines of the other elements . the equivalent widths of the lines are available at jt s world wide web site ( http://anchor.as.utexas.edu/tomkin.html ) . the spectrum of an asteriod ( iris ) , observed with the same instrumental setup as that used for the 60,000 resolving - power observations of the program stars and reduced and measured in the same manner , provided solar equivalent widths for these lines . an lte model - atmosphere abundance analysis was made relative to the sun . here we briefly discuss the selection of atomic data for the lines , the model atmospheres , and the abundance determinations for rb and the other heavy elements . the lines used for the abundance determinations are given in table 2 . our first choice for @xmath32-values for the lines was modern laboratory @xmath32-values ; basic data for the lines , including the @xmath32-values and their sources , are given in table 2 . in particular we note that the @xmath32-value of the 7800.3 rb line , log@xmath32 = + [email protected] ( wiese & martin 1980 ) , is reliably determined . for some lines , for which reliable @xmath32-values are not available , we used solar @xmath32-values instead . these were calculated with the solar equivalent widths given in table 2 , the solar atmosphere of holweger & mller ( 1974 ) , a microturbulence of 1.15(tomkin et al . 1997 ) , and the abundances given in table 2 . the solar equivalent widths were measured from the spectrum of the asteriod iris . the adoption of the laboratory @xmath32-values for the fe lines follows the prescription of lambert et al . in particular , we make a small correction ( see table 2 ) to the @xmath32-values of may , richter , & wichelmann ( 1974 ) to normalize them to those of obrian et al . ( 1991 ) and bard , kock , & kock ( 1991 ) . no correction is needed to put the solar fe line @xmath32-values on the same scale as the laboratory @xmath32-values ; we find : log @xmath32(solar ) log @xmath32(lab ) = + [email protected] ( 4 lines , 6 @xmath32-values ) . accurate experimental @xmath32-values are not available for our fe lines , therefore we have used solar @xmath32-values for these lines . although the three zr lines all have modern laboratory @xmath32-values ( bimont et al . 1981 ) , we find their solar @xmath32-values are significantly larger than the laboratory ones : log @xmath32(solar ) log @xmath32(lab ) = + [email protected] . our adopted solar zr abundance log @xmath34(zr ) = 2.60 ( anders & grevesse 1989 ) is not a factor in this discrepancy because it is based on bimont et al.s @xmath32-values and is very similar to the value log @xmath34(zr ) = 2.56 bimont et al . derived in their own investigation of the solar zr abundance . some of the discrepancy is attributable to the use of different ionization potentials for zr ; we have used the accurate value of 6.634ev ( hackett et al . 1986 ) while bimont et al . must have used the significantly higher old value of 6.84 ( allen 1973 ) . this accounts for 0.20dex of the discrepancy . line - to - line variation of individual line abundances may also account for some of the discrepancy ; the individual line abundances bimont et al . derive for these three lines are on average 0.06dex larger than the average zr abundance they determine from the zr lines . although the remaining 0.15dex discrepancy is not readily accounted for , the manner in which these three lines strengthen in the cooler program stars leaves no doubt that they are low excitation lines of a neutral species and is thus consistent with their identification as zr lines . in order to minimize the influence of whatever is causing the residual discrepancy between the solar and laboratory @xmath32-values for these lines , we have chosen to adopt the solar @xmath32-values for them in our analysis . two y lines ( 5200.42 and 5402.78 ) were rejected when it became evident , during an initial abundance analysis of all the stars , that they gave significantly larger y abundances than the other y lines in many of the stars . atomic data and solar equivalent widths for the lines are given in table 2 . plane parallel , line blanketed , flux constant , lte marcs model atmospheres , which derive from a development of the programs of gustafsson et al . ( 1975 ) , were used for the abundance analysis . the determination of parameters ( effective temperature , surface gravity , metallicity , and microturbulence ) for the model atmospheres was done in two steps . first we chose preliminary parameters for each star which were used to calculate an initial set of model atmospheres . we then used the initial model atmospheres and the equivalent widths of the fe and fe lines to iteratively adjust the atmospheric parameters to determine an adopted set of parameters and atmospheres that are consistent with the fe and fe line data . we now briefly describe these two steps . strmgren photometry provided the primary means of determining preliminary parameters for the dwarfs and subgiants , which are mostly from the catalogue of schuster & nissen ( 1988 ) supplemented by three stars from carney et al . ( 1994 ) and four stars from the bright star catalogue ( hoffleit & jaschek 1982 ) . the stars are intrinsically faint and nearby ; an examination of their parallaxes ( see below ) shows they are all within 100 pc , except for two stars which are at 111@xmath3317 and 137@xmath3318pc . the interstellar reddening of the stars is therefore negligible ( schuster & nissen 1989a ) so we have not corrected their indices . our chief source of preliminary effective temperatures for the dwarfs and subgiants was the color index @xmath35 coupled with the vs. @xmath35 calibration of alonso , arribas , & martnez - roger ( 1996a , equation 9 ) , who used the infrared flux method to determine effective temperatures for metal - deficient stars . for the components of the visual binary hd23439 , which do not have their own @xmath35 , and for hd150281 , which also does not have @xmath35 , we took the effective temperatures from carney et al . ( 1994 ) ; for 61 cyg a and b , which are too cool for the applicable range of alonso et al.s calibration , we took the effective temperatures from alonso , arribas , & martnez - roger ( 1996b ) . initial surface gravities for the dwarfs and subgiants were determined by the relations @xmath36 and @xmath37 where @xmath38 is the stellar mass , @xmath39 the radius , and @xmath40 the luminosity , with the luminosities being set by the hipparcos parallaxes . we followed the prescriptions of nissen , hg , & schuster ( 1997 ) , who have successfully applied this method to determine surface gravities for 54 metal - poor stars . most of the stars have hipparcos parallaxes and all of these parallaxes are of sufficient accuracy ; the largest uncertainty ( @xmath41 ) is 0.15 with most uncertainties being much smaller than this . for the small number ( five ) of stars without hipparcos parallaxes we took trigonometric parallaxes from gliese & jahreiss ( 1979 ) or determined photometric parallaxes from @xmath42 and @xmath43 , with @xmath43 estimated from the @xmath44 photometry and the recipes of nissen & schuster ( 1991 ) . preliminary metallicities for the dwarfs and subgiants were estimated from the @xmath45 photometry and the calibration of schuster & nissen ( 1989b , equation 3 ) . for a small number ( five ) of stars the results of carney et al . ( 1994 ) or alonso et al . ( 1996b ) were used instead . an initial microturbulence of 1.0 , which is representative of dwarfs in this temperature range ( feltzing & gustafsson 1998 ) , was used for the dwarfs and subgiants . for the giants , which are taken from pilachowski et al.s ( 1996 ) medium - resolution spectroscopic investigation of na abundances in metal - poor giants , we adopted pilachowski et al.s atmospheric parameters as initial parameters . pilachowski et al.s effective temperatures and gravities are photometrically based as modified by their spectroscopic results , while their metallicities and microturbulences are from their spectroscopic analysis . model atmospheres were computed using the marcs code ( gustafsson et al . 1975 ) . those for the dwarfs and subgiants were calculated by interpolation in a grid of marcs models , which spanned the range of dwarf and subgiant parameters and was provided by b. edvardsson . the models of pilachowski et al . ( 1996 ) , who also used marcs models for their abundance analysis , were used as preliminary models for the giants and were provided by c. sneden ; iterations of the giant models with modified atmospheric parameters were calculated directly using the marcs code . models with the preliminary parameters and the line analysis code moog ( sneden 1973 ) were then applied to the equivalent width data for the fe and fe lines . trends of the fe abundances with line excitation potential were used to check the preliminary effective temperatures and trends of the fe abundances with equivalent width were used to check the preliminary microturbulences . where necessary revised parameters were determined , new models calculated , and a new round of abundance calculations done for the fe and fe lines . next the fe abundances from this round of calculations were used to calculate a new set of models and do another round of abundance calculations for the fe lines . in a final round of calculations the surface gravities were adjusted , and new models were calculated , so the fe and fe lines gave the same fe abundance . these final adopted parameters for the program stars are given in table 1 . the adopted parameters are generally only moderately different from the initial parameters . for the dwarfs and subgiants the analysis of the fe lines led to revised effective temperatures for 13 stars and an average temperature increase for these stars of 140@xmath3375k ( @xmath46 of the individual differences ) . the revised gravities of the dwarfs and subgiants tend to be lower than the preliminary gravities , but the inconsistency is small ; for the 28 dwarfs and subgiants with measurements of both fe and fe lines , for which spectroscopic gravities can thus be determined , the average downward revision of the log @xmath47 is [email protected] ( @xmath46 of the individual differences ) . this suggests that there is no serious inconsistency between the gravities of these stars and the four remaining dwarfs and subgiants for which we adopt preliminary gravities because they have no fe lines . it is of interest to see how the differences that we find between our spectroscopic gravities and the preliminary hipparcos - based gravities compare with what allende prieto et al . ( 1999 ) found in a thorough examination , of spectroscopic gravities , for nearby stars , taken from the literature versus hipparcos - based gravities . we confine the comparison of log @xmath48 log @xmath49 for our results and theirs to the temperature range ( 4900 5500k ) of our dwarfs and subgiants . ( this range excludes the four coolest dwarfs and subgiants because they do not have any measured fe lines . ) as mentioned earlier , we find an average difference log @xmath48 log @xmath49 = [email protected] ( 28 stars , @xmath46 of individual differences ) for our dwarfs and subgiants , while allende prieto et al . find an average difference of [email protected] ( 9 stars ) for their sample of stars . our results and allende prieto et al.s thus both show that the spectroscopic gravities tend to be smaller than the hipparcos gravities . also , the consistency between the spectroscopic and hipparcos gravities of our stars is somewhat better than it is for allende prieto et al.s sample of stars . revisions of effective temperature were required for five of the giants ; the average temperature increase for these stars was 100@xmath33140k . all the giants have measurements of both fe and fe lines thus allowing revision of their preliminary gravities for all 12 stars ; the average change of log @xmath47 was [email protected] . the adopted effective temperatures and gravities of the giants thus are in good agreement with the preliminary values ( pilachowski et al . our results , which use higher resolution spectra and more numerous fe lines than those of pilachowski et al . , thus confirm their results . our fe abundances are also in good agreement ; the average difference between our [ fe / h ] determinations and theirs is [email protected] . we conclude this section with a brief discussion of the potential influence of non - lte on the effective temperatures derived from the fe line excitation . as remarked earlier , the primary source of our initial effective temperatures ( for the dwarfs and subgiants ) was @xmath35 and the vs. @xmath35 calibration of alonso , arribas , & martnez - roger ( 1996a , ) , where the effective temperatures in their calibration were determined by the infrared flux method . the @xmath35 based effective temperatures are thus free of non - lte effects . one way to estimate the possible influence of non - lte on the effective temperatures derived from the excitation of fe lines , therefore , is to consider the difference between the excitation - based temperatures and the @xmath35-based temperatures . of course non - lte effects are not the only possible source of such a difference so this check is indicative , rather than conclusive . the average difference ( fe ) @xmath50 ( @xmath35 ) = + 45@xmath3368k ( @xmath46 of the individual differences , 20 stars ) , where the calculation includes not only stars whose initial effective temperatures were revised , but also stars for which no revision was necessary - as long as they had enough fe lines to define the line excitation . this difference is small and indicates that any non - lte influence on the determination of temperatures from the fe lines is minor . abundances were determined by matching the observed line strengths and theoretical line strengths calculated by moog with the adopted model atmospheres . as remarked earlier , the rb line was treated by means of spectrum synthesis , while the lines ( y , zr , ba , and nd ) of the other heavy elements were treated by means of equivalent widths . the spectrum synthesis of the 7800.29 rb line includes the hyperfine structure of the @xmath10rb and @xmath11rb isotopes , each of which is split into two components , and the blending si line at 7800.00 . the accurately known wavelengths and relative line strengths of the hyperfine structure components were taken from lambert & luck s ( 1976 ) analysis of rb in arcturus . we adopted a terrestrial abundance ratio ( @xmath10rb/@xmath11rb = 3 ) for the rb isotopes . although in principle it would be desirable to make direct measurements of the stellar isotopic rb abundances from the exact shape of the rbi line profile , in practice extreme departures from the terrestrial isotope ratio are required before there is appreciable distortion of the line profile . lambert & luck , for example , found that for their spectra , which had a resolving power of 195,000 that is similar to the high resolution spectra of the present investigation , the @xmath10rb/@xmath11rb ratio had to be as low as 1 or as high as 10 to cause even a small variation of the line profile ; they concluded that the isotope ratio in arcturus is terrestrial with a large uncertainty . direct determination of rb isotopic abundances , therefore , is not the thrust of our investigation , although we note that comparison of the observed and synthesised spectra of the rb line in our program stars does not show any variations attributable to non - terrestrial isotope ratios . we also note that the indeterminacy of the rb isotope ratios does not interfere with the measurement of the elemental rb abundances ; the rb line is weak in all the program stars so the rb abundances it provides are not affected by changes of the isotopic mixture . the synthesis of the rb line also includes the si line to the blue . no reliable experimental oscillator strength is available for the si line so a solar oscillator strength ( log @xmath32 = -0.65 ) was adopted . the instrumental and macroturbulent broadening , as well as thermal and microturbulent broadening , were included in the synthesis . the macroturbulent broadening was set by matching the profile of the clean nearby ni line at 7797.6 . spectrum synthesis of the solar rb line , using the solar model of holweger & mller ( 1974 ) and the kurucz et al . ( 1984 ) solar atlas , provides a solar rb abundance log @xmath34(rb ) = [email protected] , which is the same as the rb abundance in anders & grevesse s ( 1989 ) compilation of solar abundances . we note , however , the discrepancy between the photosperic abundance and the somewhat lower meteoritic abundance of [email protected] ( anders & grevesse 1989 ) . although it might be speculated that the relatively low melting and boiling points of rb ( 39 and 688c , respectively ) may make it behave like a volatile element and so explain the low meteoritic abundance , this does not appear to be the case . potassium , which is isoelectronic with rb and has only slightly higher melting and boiling points ( 63 and 759c , respectively ) , shows no discrepancy of its photospheric and meteoritic abundances , which are log @xmath34(k ) = [email protected] and [email protected] , respectively . although the discrepancy is a potential source of concern , we note that there is no evidence of the rb line being affected by an unknown blend ; in particular , in our program stars the strengthening of the line with decreasing effective temperature is consistent with the behavior of a resonance line of a heavy - element neutral species . the observed and synthesised rb line profiles for a sample of stars are shown in figure [ spec ] . table 3 gives the abundances of rb and the other heavy elements . the rb abundances derived from observations made at the f1 focus ( resolving power = 200,000 ) and the f3 focus ( resolving power = 60,000 ) are highly consistent ; for stars observed at both foci the average difference between the f1- and f3-based [ rb / h ] is [email protected] ( s.e . , five stars ) . for bright stars , such as arcturus and @xmath28 cas , the greater detail provided by the higher resolution f1 observations allows for more precise determination of the rb abundance . in fainter stars there is not much to choose between the f1 and f3 observations because the greater spectral detail of the f1 observations tends to be counterbalanced by their lower signal - to - noise ratio . the two main sources of errors in the abundances are measurement error and analysis error caused by errors in the adopted model atmosphere parameters . the scatter of the abundances provided by individual lines of the same species , which are caused by measurement errors of the equivalent widths and , to a lesser extent , by errors in the line oscillator strengths , is a good guide to measurement error . this scatter , as measured by the standard deviation of the individual line abundances , is given in table 3 . ( although the standard deviations of the abundances from individual lines are larger than the standard deviations of the mean abundances , the heavy element abundances are based on only a few lines for each element see table 2 so we prefer to consider the standard deviations of the individual line abundances . ) for rb , whose abundance is based on spectrum synthesis of the 7800 rb line , table 3 gives errors estimated from the fit of the observed and synthesised spectra . inspection of table 3 shows the measurement - related abundance errors range up to @xmath330.15dex with larger errors in a few cases ; a representative error in the [ x / h ] abundances is @xmath510.07 dex , while a representative error of the [ x / fe ] abundances is @xmath510.1 dex . estimated errors in the adopted effective temperatures are between @xmath3350k , for stars with a good selection of fe lines providing a well - determined excitation temperature , and @xmath33100k , for stars for which we adopted color - based effective temperatures . representative errors in the adopted log @xmath47 and metallicities are @xmath330.2 and @xmath330.1dex , respectively . a representative uncertainty in the microturbulence is @xmath330.5 , although we note that the rb abundances provided by the weak rb line have little , or no , microturbulence dependence and that because of the metal deficiency of most of the program stars the abundances of most other elements also have only a small , or negligible , microturbulence dependence . adopting a representative effective temperature error of @xmath33100k and the stated errors of the other parameters , we find that for a typical dwarf the combined effects of these errors change the fe abundance ( from fe lines ) by @xmath330.11dex . the corresponding figure for a typical giant is @xmath330.17dex . for the heavy elements the abundance of the element relative to fe , [ el / fe ] , holds the most interest . this ratio is less dependent on the atmospheric parameters than the absolute abundance . in the typical dwarf the combined effects of the errors in the atmospheric parameters change [ el / fe ] by @xmath330.04 ( rb ) , @xmath330.05 ( y ) , @xmath330.05 ( zr ) , @xmath330.11 ( ba ) , and @xmath330.06dex ( nd ) . the corresponding figures for a typical giant are : @xmath330.07 ( rb ) , @xmath330.06 ( y ) , @xmath330.00 ( zr ) , @xmath330.14 ( ba ) , and @xmath330.07dex ( nd ) . we estimate representative total errors , caused by measurement error and errors in the model atmosphere parameters together , to be 0.1 0.2 dex in [ fe / h ] and 0.1 0.2 dex for [ el / fe ] . we now briefly consider how the fe and rb abundances of the dwarfs and subgiants would change if the preliminary effective temperatures , which are mostly based on the infrared flux method ( alonso et al . 1996a ) , and preliminary gravities , which are mostly hipparcos - based , were used instead of the adopted effective temperatures and gravities . as discussed earlier , the preliminary effective temperatures were revised upward by an average of 140@xmath3375k ( @xmath46 of the individual differences ) for 13 of the dwarfs and subgiants , while no revisions of the preliminary effective temperatures were made for the other 19 dwarfs and subgiants . use of the lower preliminary effective temperatures for these 13 stars would decrease their fe abundances ( from fe lines ) by an average of 0.14dex , while their [ rb / fe ] , as set by rb and fe lines , would increase by an average of 0.04dex . the adopted gravities for 28 of the dwarfs and subgiants are spectroscopically determined , while those of the other four dwarfs and subgiants , for which we could not determine spectroscopic gravities , are the preliminary gravities . use of the preliminary gravities , instead of the adopted spectroscopic gravities , for these 28 stars would not change their fe or rb abundances significantly ; the adoption of the preliminary log @xmath47 , which are [email protected] ( @xmath46 of the individual differences ) higher on average than the spectroscopic log @xmath47 , in place of the spectroscopic log @xmath47 would change the [ fe / h ] and [ rb / fe ] from neutral lines by + 0.01 and 0.00dex , respectively , on average . before we consider the rb and other heavy - element abundances , we compare our results with those published in the literature . first we consider the results for [ fe / h ] . figure [ fecomp ] compares the [ fe / h ] determinations for stars in common to this study and earlier high signal - to - noise ratio , high resolution studies . the comparison is not exhaustive , but does include all recent studies ( since 1990 ) which have two , or more , stars in common with the present study . the agreement of the [ fe / h ] determinations is good over most of the metallicity range and , although the [ fe / h ] of this study tend to be slightly more negative than the literature [ fe / h ] in the most metal - deficient stars , the overall agreement is not unsatisfactory . previous studies which , to our knowledge , have determined rb abundances for stars in common with those of the present investigation are mckle et al.s ( 1975a ) study of arcturus and gratton & sneden s ( 1994 ) study of heavy - element abundances in metal - poor stars . in table 4 we compare our [ rb / fe ] with those of the two earlier investigations . because our rb abundances are based on the 7800 rb line , while mckle et al.s ( 1975a ) abundance is based on both the 7800 and 7947lines , we also include in the table their abundance for the 7800 rb line alone ( mckle et al . ( gratton & sneden s results are based only on the 7800 line . ) in order to make the rb abundances of our and the earlier studies directly comparable we have adjusted the rb abundances of the earlier studies to reflect the values they would have if the stellar parameters ( , log @xmath47 , [ m / h ] , and @xmath52 ) used in the earlier studies had been the same as those used here . for arcturus the difference between our [ rb / fe ] and mckle et al.s is only 0.03 dex pleasingly small and not unexpected for the case of such a bright star . for the three stars that we have in common with gratton & sneden we note that both investigations determined rb abundances for two of the stars ( hd64606 and 187111 ) , but were only able to determine upper limits for the third star ( hd122956 ) . the differences between our [ rb / fe ] and gratton & sneden s are 0.25 ( hd64606 ) , 0.09 ( hd122956 ) , and + 0.17 ( hd187111 ) . that these differences are much larger than in the case of arcturus can be ascribed to the fact that all three stars are quite metal - deficient and have only a weak or undetectable rb line . we estimate that for these three stars the uncertainty in the rb abundance associated with fitting our observed and synthetic spectra of the rb line is 0.1dex for hd64606 and 187111 and 0.2dex for hd122956 ; gratton & sneden s rb abundances must be subject to similar uncertainties also . we conclude , therefore , that our [ rb / fe ] and gratton & sneden s are probably the same to within the errors of measurement for these three stars . as is customary , the abundances in table 3 are plotted as [ el / fe ] against [ fe / h ] , in fig . [ rb.y ] , [ zr.ba ] , and [ nd ] to reveal trends in the relative abundance of element el and iron . two points are immediately apparent : ( i ) three stars are unusually rich in the heavy elements note especially the [ y / fe ] ratios of hd23439a and b , and bd+5@xmath33640 which we shall dub ch stars , and ( ii ) the distinctive behavior of [ rb / fe ] in the metal - poor stars [ rb / fe ] @xmath53 when the other heavy elements show [ el / fe ] @xmath54 . before commenting on these striking results , we compare our results for the y , zr , ba , and nd abundances with results in the literature . previous extensive abundance determinations of heavy ( and other ) elements in metal - poor stars have shown that the run of [ el / fe ] against [ fe / h ] is smooth down to about [ fe / h ] = 2 with a ` cosmic ' scatter less than the scatter that results from the errors of measurement . cosmic scatter is present for more metal - poor stars but our sample is devoid of such stars . therefore , our results are expected to agree well with previous studies despite the lack of a complete overlap in stellar samples . key papers reporting results on heavy elements are zhao & magain ( 1991 ) and gratton & sneden ( 1994 ) with reviews by wheeler , sneden , & truran ( 1989 ) , lambert ( 1989 ) , and mcwilliam ( 1997 ) amongst others . our results for [ y / fe ] , [ ba / fe ] , and [ nd / fe ] are in excellent agreement with previous results , for example , zhao & magain ( 1991 ) and gratton & sneden ( 1994 ) find [ y / fe ] @xmath55 at [ fe / h ] = 1 with the relative underabundance of y increasing to about 0.25 at [ fe / h ] = 2 which agree well with figure [ rb.y ] . a discrepancy appears when comparing results for [ zr / fe ] . zhao & magain ( 1991 ) and gratton & sneden ( 1994 ) report [ zr / fe ] @xmath56 for [ fe / h ] in the range of 1 to 2 but our results ( figure [ zr.ba ] ) show [ zr / fe ] to be consistently less than zero : a difference in [ zr / fe ] of about 0.3 to 0.4dex relative to the previous studies . this difference is most probably due to our exclusive use of zr lines . brown , tomkin , & lambert ( 1983 ) found that zr lines in mildly metal - poor giants gave a clear zr underabundance which was plausibly attributed to non - lte effects such as over - ionization of zr atoms to zr@xmath23 ions . for metal - poor stars , gratton & sneden remark that their selection of zr lines gives a systematically lower zr abundance than the zr lines : the difference of [email protected] would account in part for our largely negative values of [ zr / fe ] . our results clearly show a relative overabundance of rb in metal - poor stars : the mean value [ rb / fe ] = + [email protected] ( s.e . ) is found from nine stars with [ fe / h ] @xmath57 , excluding the three ch stars . this is consistent with the four measurements reported by gratton & sneden ( 1994 ) and with their six upper limits to [ rb / fe ] . non - lte effects such as overionization warrant consideration . as a guide to the non - lte effects on rb , we consider those calculated for lithium , another alkali . carlsson et al . ( 1994 ) predict that the lte abundances from the li 6707 resonance doublet require correction by not more than 0.03dex for non - lte effects , a negligible correction in the present and almost all contexts . the abundances of li and rb are similar : lithium has the abundance log @xmath34(li ) @xmath58 in the warmer dwarf stars comprising the spite plateau and rb declines from log @xmath34(rb ) = 2.6 at solar metallicity to 1.9 at [ fe / h ] = 1 and to 0.8 at [ fe / h ] = 2 . the key point is that optical depth effects in lines and continua are slight for both elements . rb is probably more affected by photoionization because the atom s ionization potential is 4.18ev versus 5.39ev for the lithium atom . photoionization of rb will be enhanced relative to the rate for li but collisional ionization rates will also be enhanced . the different wavelengths of the resonance ( and excited ) atomic lines for rb and li will introduce no more than slight differences in the non - lte corrections . if non - lte effects were large for rb , we would anticipate that dwarfs and giants of the same metallicity would yield systematically different rb abundances . this is not the case : four dwarfs with [ fe / h ] in the range 1 to 2 give a mean [ rb / fe ] = + [email protected] ( s.e . ) and five giants in the same [ fe / h ] range give the mean [ rb / fe ] = + [email protected] ( s.e . ) . ( the mean for the dwarfs excludes the three ch stars . ) although a non - lte analysis for rb would be of interest , we suggest that our results derived from lte analyses are not substantially different from non - lte results . scatter of the [ el / fe ] results at a given [ fe / h ] is not significantly different from that expected from the measurement errors . obviously , the three stars over - abundant in the heavy elements and dubbed ch stars are set aside as special cases . the scatter for [ rb / fe ] between [ fe / h ] of 0.5 and 1.0 is small and consistent with the measurement errors . there is an apparent moderate increase in scatter of [ rb / fe ] below [ fe / h ] = 1 but this is probably again due to the measurement errors because the rb line is very weak in these metal - poor stars . the results are roughly consistent with a constant [ rb / fe ] in stars with [ fe / h ] @xmath57 . these three stars , which are all dwarfs , are consistently overabundant in all of the five heavy elements investigated in this study . figure [ ch ] shows a zr line and a v line in hd23439a and b and hd103095 , a non - ch star with otherwise similar properties . the much greater strength of the zr line relative to the v line in hd23439a and b as compared with hd103095 is evident . as may be seen in fig . [ rb.y ] , [ zr.ba ] , and [ nd ] the heavy - element abundance enhancements are similar in the three stars . the average enhancements for the three stars are : [ rb / fe ] = + 0.41 , [ y / fe ] = + 0.34 , [ zr / fe ] = + 0.51 , [ ba / fe ] = + 0.27 , and [ nd / fe ] = + 0.32 ( hd23439a and bd+5@xmath33640 ) ; actual values of [ x / fe ] for the individual stars are given in table 5 . hd23439 is a nearby visual binary composed of a k1v primary and a k2v secondary ; the hipparcos catalogue gives 7.@xmath59307 and 40.83mas for the separation of its components and parallax , respectively . these numbers set a lower limit on the a b linear separation of 179au . to the best of our knowledge hd23439 is the first case of a binary in which both components have been found to be ch stars . we note that hd23439b is a single - lined spectroscopic binary with a period of 48.7d and a mass function of 0.0022 ( latham et al . could the unseen companion of hd23439b be the white dwarf descendant of the agb star responsible for the mass transfer that changed hd23439a and b into ch stars ? perhaps , but in this scheme it is hard to explain the very similar heavy - element enhancements of hd23439a and b ( see table 5 ) . just as today the unseen companion is much closer to component b than to component a so in the past the putative agb predecessor of the unseen companion must also have been much closer to b than a. how did the agb star manage to give the a and b components the same heavy - element enhancements ? this difficulty with the agb - star scenario suggests that the heavy - element enhancements must be primordial . heavy elements are synthesised by the neutron capture @xmath5- and @xmath4-processes . ( in considering elemental abundances , the small contribution from @xmath60-processes may be neglected . ) detailed dissection of the isotopic abundances measured for carbonaceous chondrites has provided an isotope by isotope resolution of the abundances into @xmath5- and @xmath4-process contributions ( cf . kppeler , beer , & wisshak 1989 ) . as is well known , ba and eu are primarily @xmath5- and @xmath4-process products respectively : cowan ( 1998 ) estimates that ba is 85% an @xmath5-process product , and eu is 97% a @xmath4-process product see gratton & sneden ( 1994 ) for similar estimates . rubidium is of mixed parentage ; cowan gives the @xmath5- and @xmath4-process fractions as 50% each gratton & sneden provide quite similar estimates ( 48% for @xmath5- and 52% for the @xmath4-process ) . as noted in the introduction , the @xmath5-process contribution may be broken into a ` weak ' and a ` main ' component . gratton & sneden divide the 48% @xmath5-process rb contribution into 5% from the weak and 43% from the main @xmath5-process . our goal is to use the ba and eu abundances as monitors of the @xmath5- and @xmath4-processes respectively to predict the rb abundances , and then to comment on the consistency between the predicted and observed rb abundances . the other heavy elements considered here are also a mix of @xmath5- and @xmath4-processes : cowan gives the following @xmath61 % : ( 72 , 28 ) for y , ( 81 , 19 ) for zr , and ( 47 , 53 ) for nd . gratton & sneden ( 1994 ) put the weak contribution to the total @xmath5-process as 16% for y , 10% for zr , and less than 1% for ba , nd , and eu . especially interesting is the roughly 50 - 50 split for nd that matches the split for rb . then , the simplest possible scenario of unvarying yields of @xmath5 and @xmath4-process products over the life of the galaxy would predict that [ rb / fe ] and [ nd / fe ] would vary identically with [ fe / h ] . inspection of fig . [ rb.y ] and [ nd ] shows that this is not the case . there are several factors pertinent to the understanding of the run of rb and other heavy elements with [ fe / h ] . * it is now well known that , the distribution of heavy elements at low [ fe / h ] resembles a @xmath4-process pattern and is not the mix of @xmath5- and @xmath4-processes that prevails at solar metallicities ( truran 1981 ; sneden & parthasarathy 1983 ; wheeler et al . 1989 ; lambert 1989 ) . there is evidence that the abundance distribution of the @xmath4-process was largely invariant from low metallicities to the present [ fe / h ] @xmath54 . then , it should suffice in modelling the [ el / fe ] vs [ fe / h ] relations to adopt the relative @xmath4-process abundances that are obtained from dissection of the measurements on carbonaceous chondrites . * europium is assigned to the @xmath4-process : cowan s ( 1998 ) resolution of the meteoritic abundances is 97% @xmath4-process and a mere 3% @xmath5-process . with declining metallicity the @xmath5/@xmath4 ratio declines . therefore , we may assume that eu is a @xmath4-process product throughout the evolution of the galaxy . the run of [ eu / fe ] against [ fe / h ] is taken from mcwilliam s ( 1997 ) review : [ eu / fe ] = 0 at [ fe / h ] = 0 with a smooth transition to [ eu / fe ] @xmath62 at [ fe / h ] = 1.0 and to the metallicity limit [ fe / h ] = 2.5 of interest to us . * red giants enriched in @xmath5-process heavy elements are likely the major donors of these elements to the galaxy s interstellar medium . analyses of such red giants of differing [ fe / h ] show that the pattern of @xmath5-process products has evolved with [ fe / h ] . smith ( 1997 ) , who has collated published results , defines y and zr as ` light ' @xmath5-process elements ( here , ls ) and ba , la , and ce as ` heavy ' @xmath5-process elements ( here , hs ) . he finds that [ hs / ls ] , which by definition is 0 at [ fe / h ] = 0 , increases to [ hs / ls ] @xmath63 0.6 at [ fe / h ] = 1.5 . this evolution of [ hs / ls ] is attributed to an increase in the average exposure to neutrons in the he - burning shell of the agb stars that are the site of the main @xmath5-process , as expected on theoretical grounds . * there is limited evidence also from abundance analyses of red giants that the rb abundance relative to other ls elements increases with decreasing [ fe / h ] ; smith s ( 1997 ) collection of results implies [ rb / zr ] increases by about 0.7 dex from [ fe / h ] = 0 to 1.5 . this increase is attributed to a higher mean neutron density in the he - burning shell of the metal - poor agb stars . as noted by smith , this increase is expected on theoretical grounds . * if the magnitude of the weak @xmath5-process contributions to the abundances in carbonaceous chondrites were representative of the @xmath5-process at all relevant [ fe / h ] , the weak @xmath5-process could be safely dropped from our search for an explanation of the run of [ el / fe ] . a thorough direct check on the weak @xmath5-process is not possible because the majority of the elements between rb and the fe - group are inaccessible spectroscopically . zinc , which offers at least a hint of the behavior of the weak @xmath5-process , is assigned 34% to the @xmath5-process and 66% to the @xmath4-process in cowan s breakdown of solar system abundances . the @xmath5-process component is essentially entirely due to the weak @xmath5-process . this breakdown neglects a possible contribution to zn from the sources that contribute the fe - group elements . in the metallicity range of interest , [ zn / fe ] = [email protected] ( sneden & crocker 1988 ) . this result , which is barely compatible with the increase of [ eu / fe ] with declining metallicity , implies a drop in [ @xmath5/fe ] with metallicity and justifies our neglect of the weak @xmath5-process contribution to rb and other elements . guided by these facts , it is possible to predict relative abundances of the heavy elements including rb . we begin by considering ba , nd , and eu . eu defines the evolution of the @xmath4-process products . the observed run of [ ba / fe ] against [ fe / h ] provides the evolution of the hs component of the @xmath5-process after a small correction for this element s @xmath4-process component based on the eu abundances and the meteoritic @xmath4-process ba / eu ratio . the adopted runs of [ eu / fe ] and [ ba / fe ] against [ fe / h ] are shown in fig . [ calc.abun]a . then , it is a simple matter to predict the run of [ nd / fe ] using the meteoritic 5050 split into @xmath5- and @xmath4-process contributions . this prediction which is shown too in fig . [ calc.abun]a is slightly inconsistent with the observations ( fig . [ nd ] ) that show [ nd / fe]@xmath63 0 at all metallicities rather than the predicted [ nd / fe ] = 0.14 at [ fe / h ] @xmath64 1 . earlier , we noted that our nd abundances are quite consistent with previously published results . this small inconsistency appears not to have been noted previously . it is likely that when the measurement errors are included the prediction and observations will overlap . note that the prediction uses the eu and ba abundances as well as the meteoritic @xmath5 to @xmath4 ratios for ba , nd , and eu . a change in the nd ratio from 50% @xmath5 and 50% @xmath4 to 75% @xmath5 and 25% @xmath4 reduces [ nd / fe ] to 0.0 for metal - poor stars . yttrium abundances may be predicted from the [ ba / fe ] observations and smith s estimates of [ hs / ls ] from heavy element enriched red giants such as s and barium stars . this prediction is shown in fig . [ calc.abun]b . the relative underabundance of y ( i.e. , [ y / fe ] @xmath65 ) results largely from the steep increase in [ hs / ls ] with decreasing metallicity that offsets the increase in the @xmath4-process contribution . this particular prediction assumes a meteoritic ratio of 72% @xmath5-process and 28% @xmath4-process and makes no attempt to separate main from weak @xmath5-process contributions . this prediction matches the observations quite well ( see fig . [ rb.y ] ) . the rb prediction corresponding to the y prediction which is also shown in figure [ calc.abun]b does not correspond to the observed [ rb / fe ] ratios in metal - poor stars . the limited evidence gathered by smith suggests that rb in heavy element enriched red giants is progressively overabundant ( relative to zr ) in metal - poor giants . this increase is attributable to a higher neutron density at the @xmath5-process site in red giants . if a smooth curve is drawn through smith s assembled data , the resulting run of [ rb / fe ] is shown in fig . [ calc.abun]b . this prediction is in good agreement with the observations . the principal novel result of our survey of rb abundances in stars is that rubidium relative to iron is systematically overabundant in metal - poor stars . this increase reflects partly the growth of the @xmath4-process abundances relative to iron in metal - poor stars ; we model this increase using observed eu abundances and the assumption that the pattern of @xmath4-process abundances is solar - like at all metallicities . a second and major factor accounting for the increase in the rb to fe ratio in metal - poor stars is that the @xmath5-process contribution to rb increases with decreasing metallicity . we model this increase using published abundances of rb and other heavy elements collated by smith ( 1997 ) for @xmath5-process enriched red giants that are presumed to be representative of the donors of @xmath5-processed material to the interstellar medium and so to control the chemical evolution of the galaxy as far as the @xmath5-process is concerned . two factors influence the rb abundance : ( i ) the total exposure to neutrons at the @xmath5-process site , the he - burning shell of an agb star , increases with decreasing metallicity of the red giant , and ( ii ) the neutron density at the @xmath5-process site increases with decreasing metallicity . that ( i ) is true follows from the observed increase of the relative abundance ratio of heavy to light @xmath5-process elements in the @xmath5-process enriched red giants . it is this effect that accounts , for example , for the drop in [ y / fe ] in metal - poor stars . that ( ii ) is true follows from the limited data on rb abundances in @xmath5-process enriched red giants . as explained above , neutron density influences the rb abundance through the branch in the @xmath5-process path at @xmath10kr . in this thoroughly empirical way we account for the relative enrichment of rb in metal - poor stars . in short , the observed [ rb / fe ] ratios of metal - poor stars are consistent with the expectation that agb stars control the input of main @xmath5-process products to the galaxy s interstellar medium . a serendipitous discovery is the finding that both members of a visual binary are mild ch stars , hd23439a and b , with @xmath5-process overabundances relative to other stars of the same metallicity . mass transfer across a binary system is now thought to account for ch stars and the barium stars , the higher metallicity counterparts of the ch stars . hd23439b is a single - lined spectroscopic binary and the visible star might have been transformed to a ch star by mass transfer from the companion , then an agb star and now a white dwarf . hd23439a appears to be a single star that can not have captured significant amounts of mass from a very distant agb star orbiting hd23439b . we suggested , therefore , that these ch stars testify that the halo s interstellar medium was not entirely chemically homogeneous . this is not too surprising given that @xmath5-process products are injected into the interstellar medium at low velocity by red giants whereas iron and other elements are injected at very high velocity by supernovae . if the timescale for star formation is shorter than the timescale for thorough mixing of supernovae and red giant ejecta , abundance anomalies will result . we thank bengt edwardsson for providing the grid of dwarf and subgiant marcs model atmospheres , chris sneden for providing pilachowski et al.s ( 1996 ) giant marcs model atmospheres , and m. busso and verne smith for helpful discussions . we also thank pilachowski et al . for a list of the stars on their observing program given to us in advance of publication . this research has made use of the simbad database , operated at cds , strasbourg , france . this work has been supported in part by nsf grant ast 9618414 and the robert a. welch foundation of houston , texas . smith , v. v. 1997 , in synthesis of the elements in stars : forty years of progress ; wallerstein , g. , iben , i. jr . , parker , p. , boesgaard , a. m. , hale , g. m. , champagne , a. e. , barnes , c. a. , kppeler , f. , smith , v. v. , hoffman , r. d. , timmes , f. x. , sneden , c. , boyd , r. n. , meyer , b. s. , and lambert , d. l. , rev . mod .	we report the first extensive study of stellar rb abundances . high - resolution spectra have been used to determine , or set upper limits on , the abundances of this heavy element and the associated elements y , zr , and ba in 44 dwarfs and giants with metallicities spanning the range @xmath0 [ fe / h ] @xmath1 . in metal - deficient stars rb is systematically overabundant relative to fe ; we find an average [ rb / fe ] of + 0.21 for the 32 stars with [ fe / h ] @xmath2 and measured rb . this behavior contrasts with that of y , zr , and ba , which , with the exception of three new ch stars ( hd23439a and b and bd+5@xmath33640 ) , are consistently slightly deficient relative to fe in the same stars ; excluding the three ch stars , we find the stars with [ fe / h ] @xmath2 have average [ y / fe ] , [ zr / fe ] , and [ ba / fe ] of 0.19 ( 24 stars ) , 0.12 ( 28 stars ) , and 0.06 ( 29 stars ) , respectively . the different behavior of rb on the one hand and y , zr , and ba on the other can be attributed in part to the fact that in the sun and in these stars rb has a large @xmath4-process component while y , zr , and ba are mostly @xmath5-process elements with only small @xmath4-process components . in addition , the rb @xmath5-process abundance is dependent on the neutron density at the @xmath5-processing site . published observations of rb in @xmath5-process enriched red giants indicate a higher neutron density in the metal - poor giants . these observations imply a higher @xmath5-process abundance for rb in metal - poor stars . the calculated combination of the rb @xmath4-process abundance , as estimated for the stellar eu abundances , and the @xmath5-process abundance as estimated for red giants accounts satisfactorily for the observed run of [ rb / fe ] with [ fe / h ] . 0 0 0 st_1@xmath6 _ nd_2@xmath7 _ rd_3@xmath8 _
by @xmath3 we denote the free group on @xmath4 . let @xmath5 be the normal closure of the defining relators @xmath6 in the free group @xmath3 and @xmath7 be the quotient group @xmath8 . for any word @xmath9 , its area is defined by the minimal number @xmath10 where @xmath11 @xmath12 . using van kampen s lemma @xcite , the area of @xmath13 is equal to the smallest number of cells in a van kampen diagram if its boundary label is @xmath13 . if two words @xmath14 are equal in @xmath7 then we denote by @xmath15 the number @xmath16 . the dehn function @xmath17 is the map from @xmath18 to itself defined as @xmath19 where @xmath20 is the length of the word @xmath13 . the r.thompsons group @xmath21 was discovered by richard thompson in the 1960s . the group @xmath21 is torsion - free , and it can be defined by the following presentation @xmath22 where @xmath23 . @xmath21 is finitely presented , it can be given by @xmath24 where @xmath25 for any @xmath26 . for richard thompson s group @xmath0 , it has the following finite presentation : @xmath27 where @xmath28 for any @xmath26 . let @xmath29 , @xmath30 be two non - decreasing functions from @xmath18 to itself . @xmath31 means that there exists a positive integer constant @xmath32 such that @xmath33 for all @xmath34 . @xmath35 if and only if @xmath31 and @xmath36 . for a finitely presented group there exists a unique dehn function , modulo the equivalence relation @xmath37 . the group has solvable word problem if and only if its dehn function has s recursive upper bound @xcite . a finitely presented group is word - hyperbolic group if and only if its dehn function @xmath38 @xcite . in 1997 , guba and sapir @xcite proved that the dehn function @xmath39 of group @xmath21 is strictly subexponential , namely @xmath40 . after one year , guba @xcite proved that the dehn function of @xmath21 satisfies @xmath41 . finally , guba @xcite proved that @xmath42 in 2006 . for r.thompsons group @xmath0 , guba @xcite proved that @xmath43 . we improve this result as the following : [ t ] _ the dehn function of r.thompsons group _ @xmath0 _ satisfies polynomial upper bound and _ @xmath1 . in this section , we improve some inequalities by using van kampen diagrams which shall be used to estimate the dehn function of @xmath0 in next section . [ 1 ] _ for any _ @xmath44 , \(1 ) @xmath45 \(2 ) @xmath46 \(3 ) @xmath47 for ( [ l11 ] ) , we just use the relation @xmath48 one time , @xmath49 now let us prove ( [ l12 ] ) . let @xmath50 . it is easy to see that @xmath51 , @xmath52 . at first , we prove the case @xmath53 , @xmath54 . consider the following van kampen diagram : applying ( [ l11 ] ) the above diagram shows that @xmath55 by the result of guba s which says the dehn function of r.thompson group @xmath21 is quadratic . we have @xmath56 @xmath57 @xmath58 so we get @xmath59 . for @xmath60 , we consider the following diagram which shows that @xmath61 by guba s result . so we proved ( [ l12 ] ) . for ( [ l13 ] ) , let @xmath62 . clearly , @xmath63 . now let @xmath26 , consider the following diagram shows that @xmath64 . so we have that @xmath65 , which obtains ( [ l13 ] ) . [ 2 ] _ let _ @xmath66 _ _ , _ _ @xmath67 , @xmath68 _ be integers . _ \(1 ) @xmath69 \(2 ) @xmath70 \(3 ) @xmath71 \(4 ) @xmath72 \(5 ) @xmath73 _ each of the equalities _ ( [ l21 ] ) ( [ l22 ] ) ( [ l23 ] ) ( [ l24 ] ) _ needs _ @xmath74 _ elementary steps and _ ( [ l25 ] ) _ needs _ @xmath75 _ steps_. for ( [ l21 ] ) , in case @xmath76 and consider the following diagram the word is @xmath77 labelled by the bottom path , we see that @xmath78 . in case @xmath79 . if @xmath80 , then @xmath81 . if @xmath82 , then the following diagram shows that @xmath83 . for ( [ l23 ] ) , lemma [ 1 ] and the following diagram shows that @xmath84 . now we return to prove ( [ l21 ] ) , in case @xmath85 . we consider the following diagram and apply ( [ l23 ] ) and ( [ l21 ] ) we have @xmath86 . for ( [ l22 ] ) and ( [ l24 ] ) , they can be obtained from ( [ l21 ] ) and ( [ l23 ] ) . in addition , they all need @xmath74 elementary steps . finally , let us prove ( [ l25 ] ) . we construct the following diagram so @xmath87 , which implies @xmath88 . the definition of the complexity @xmath89 which will be occured in the following lemma can be found in @xcite and we improve the corresponding lemma in it . [ 3 ] _ let _ @xmath90 _ be a positive word _ , @xmath91 . _ then there exist a positive word _ @xmath13 _ and an integer _ @xmath92 _ such that _ @xmath93 _ equals _ @xmath94 _ in _ @xmath0 _ and _ @xmath95 , _ where _ @xmath96 . we prove the lemma by induction on @xmath97 , which @xmath97 is the length of the word @xmath90 . it is obviously in the case of @xmath98 . now let @xmath99 . then we can write @xmath100 for some @xmath101 . by the induction assumption , set @xmath102 , we have a positive word @xmath103 , an integer @xmath104 , such that @xmath105 and @xmath106 . consider several cases . \(1 ) @xmath107 . applying lemma [ 2 ] we have @xmath108 we take @xmath109 . then we can obtain @xmath110 . we now compute the complexity of @xmath13 @xmath111 \(2 ) @xmath112 . in this case , @xmath113 . if @xmath114 , we take @xmath115 , @xmath116 . if @xmath117 , we take @xmath118 , @xmath119 . if @xmath120 , then we take @xmath121 , @xmath119 . it is easy to verify by using lemma [ 2 ] in all these cases that @xmath122 and @xmath95 . let @xmath13 be a word of length @xmath34 over the alphabet @xmath123 , then we claim that there exists positive words @xmath124 and an integer @xmath125 such that @xmath126 equals @xmath13 in @xmath0 , @xmath127 and @xmath128 . if @xmath129 , the claim is trivial . let @xmath130 . we write @xmath131 , where @xmath132 or @xmath133 . for @xmath103 by the inductive assumption , there exist positive words @xmath134 of complexity @xmath135 and an integer @xmath136 such that @xmath137 in @xmath0 . \(1 ) @xmath138 . let @xmath139 . after a simple calculation we have @xmath140 . if @xmath141 , then @xmath142 , @xmath143 . if @xmath144 , then by lemma [ 2 ] @xmath145 . in this case , we set @xmath146 , @xmath147 . @xmath148 . and we have the following inequality @xmath149 ( 2)@xmath150 . by @xciteand the lemma in @xcite , there exist a positive word @xmath151 , integers @xmath152 such that @xmath153 , @xmath154 , @xmath155 . if @xmath141 , we just take @xmath156 and @xmath142 . now let @xmath157 . if @xmath158 , then take @xmath147 , @xmath146 . so clearly by lemma [ 2 ] , @xmath159 . if @xmath160 . \(3 ) @xmath133 . in this case , it is similar to be obtained from @xcite , but we have the inequality more precisely @xmath168 thus we have proved the claim by induction . now it is not difficult to get @xmath169 , where @xmath13 is any word of length @xmath167 on the alphabet @xmath170 and @xmath171 in @xmath0 . baumslag , g.,miller , c.f . , short , h. , isoperimetric inequalities and the homology of groups . invent . math . * 113 * , 531 - 560(1993 ) . egbert r. van kampen . , on some lemmas in the theory of groups . _ amer.j.math._ , 55:268 - 273 , 1933 . gersten , s.,isoperimetric and isodiametric functions of finite presentations . in : geometric group theory , vol.1(sussex1991 ) . * 181,79 - 96(1993 ) . guba , v. s. and m.v . sapir , the dehn function and a regular set of normal forms for r. thompson s group @xmath21 , _ j.austral . a , 62(1997),315 - 328 . guba , v. s. , polynomial upper bounds for the dehn function fo r.thompsons group f. j.group theory 1,203 - 211(1998 ) . guba , v. s. , polynomial isoperimetric inequalities for the richard thompson s groups @xmath21 , @xmath0 , and @xmath172 . in : algorithmic problems in groups and semigroups , ed . by j .- c . birget et al . , pp . 91 - 120 . boston , berlin : birkhuser 2000 . guba , v. s. , the dehn function of richard thompson s group f is quadratic . 163 , 313 - 342(2006 ) . j.w.cannon , w.j.floyd , and w.r.parry , introductorary notes on richard thompson s groups , _ enseign ( 2 ) , 42(1996 ) , 215 - 256 . madlener , k. , otto , f. , pseudo - natural algorithms for the word problem for finitely presented monoids and groups . . comput . 1 * , 383 - 418(1985 ) . m.gromov , hyperbolic groups , in _ essays in group theory _ , s.gersten , ed . , msri publications 8 , springer - verlag , berlin , new york , heidelberg , 1987,75 - 263 . noel brady , tim riley and hamish short , the geometry of the word problem for finitely generated groups . advanced courses in mathematics crm barcelona , birkhuser , basel , 2007 . isbn 3 - 7643 - 7949 - 9 .	we improve guba s result about the dehn function of r.thompsons group @xmath0 and get that @xmath1 , where @xmath2 is the dehn function of group @xmath0 .
in this section we compare the expected oscillation periods of magnetic field sweeps for abelian and non - abelian candidate states at @xmath0 . as explained in the main text , the net phase accumulation in a quantum hall interferometer consists of two contributions : the ab phase due to the magnetic flux change by @xmath31 and the statistical phase due to introducing @xmath33 additional quasiparticles of charge @xmath22 to the interferometer area : @xmath78 if we assume that the active area of the interferometer remains constant throughout a magnetic field sweep , these two contributions are not independent ; @xmath32 and hence @xmath79 . \label{eq : net_phase_2}\ ] ] at @xmath0 , the smallest charge excitation is an @xmath4 quasihole / quasipartice irrespective of the exact nature of the state . since the smallest charge carriers are expected to dominate tunneling at the low temperature , weak tunneling limit @xcite ( or , in the case of the anti - pfaffian state , to vary with temperature in the same way as transport due to @xmath27 quasiparticles ) , we will first focus on these @xmath4 excitations . as shown in the main text , if the @xmath0 state is non - abelian ( whether it is moore read or anti - pfaffian ) , this should manifest itself via a @xmath80 oscillation period corresponding to the even - odd effect . an additional @xmath81 period of purely abelian nature is also expected , although it may be washed out by fermion parity fluctuations . if , on the other hand , the @xmath0 state is an abelian @xmath2 state @xcite , the @xmath80 oscillation period should not be observed there is no even - odd effect in this case . an abelian phase can be calculated using eq . ( [ eq : net_phase_2 ] ) with a slight caveat . the statistical angle @xmath82 can be either @xmath83 or @xmath84 , depending on whether the spin of a quasiparticle going around the interferometer and that of a quasiparticle inside the loop are the same or opposite . since the @xmath2 state is spin - unpolarized , excitations of both spins may carry charge around the interference loop . hence we can use the average value of the statistical angle , i.e. @xmath85 , per quasiparticle ( or , more precisely , @xmath86 per pair with opposite spins ) . this yields @xmath87 = - 2{\pi } \left({\phi}/{\phi_0}\right)$ ] . the resulting oscillation period is @xmath81 . so far in our discussion we have neglected other types of charged excitations , particularly the charge @xmath27 excitations that are also expected at @xmath0 . these excitations are always abelian and their statistical angle is @xmath88 , irrespective of the nature of the state . ( [ eq : net_phase_2 ] ) then yields @xmath89 implying a @xmath90 periodicity . while backscattering of these quasiparticles at quantum point contacts should be suppressed by comparison to @xmath4 quasiparticles ( except in the anti - pfaffian state ) , it is likely that their coherence length is much longer , which in turn enhances their contribution to the coherent interference signal @xcite . it therefore remains a puzzle that period @xmath90 oscillations are not convincingly seen in our data , particularly in view of the fact that @xmath27 oscillations were a prominent feature seen in the side - gate voltage sweeps @xcite . one possible explanation is the relatively narrow magnetic field window available for @xmath28-sweeps at @xmath0 , which in turn makes it difficult to see longer - period oscillations . this window is limited by the narrow width of the @xmath0 plateau immediately flanked by the reentrant compressible integer states @xcite see more on this in the next section . it is also worth mentioning that the data presented here was measured at @xmath91mk , while previously reported side - gate voltage sweeps were performed at @xmath92mk and higher ; lowering the temperature should lead to the suppression of the @xmath27 contribution to oscillations . testing this argument by measuring @xmath28-sweeps at higher temperatures and for different device sizes is an important future direction . additional sets of data showing fourier transforms of oscillations at integer , @xmath63 , and @xmath93 filling factors from different sample preparations are shown in fig . s[fig : s1 ] . note that the @xmath93 peak is split and is centered at @xmath94 for all the data sets in this figure . in roughly half the data sets examined , we were able to resolve the predicted splitting of the 5/2 fft peak at five times the fundamental integer oscillation frequency ( @xmath95 ) . this splitting at @xmath96 to @xmath97 , due to modulation of the non - abelian oscillation by the low frequency ab / anyonic oscillations at @xmath0 , can be seen only when we take the fourier transform the resistance over a substantial range in @xmath28-field : such data is displayed in fig . 3 of the main text , and in fig . s[fig : s1 ] . some @xmath28-field sweeps in this study ran over only a small range of @xmath28 near 5/2 filling to focus on measuring the small period ( @xmath6 ) oscillations attributed to non - abelian @xmath4 expression / suppression : these results are shown in fig . 4 of the main text , and here in fig . s[fig : s2 ] . the other measurements , covering a larger range in @xmath28 , noted above , facilitate resolution of the splitting at 5/2 . however , the range of @xmath28-field of either of these oscillations is limited around 5/2 by the presence of reentrant integer states @xcite adjacent to @xmath93 filling on both the high- and low - field sides . this is a more prominent effect in the larger area devices . when side gate sweeps are applied rather than @xmath28-field sweeps , the multiple devices of different areas examined here display the same ab properties at 5/2 as observed in previous studies @xcite , namely alternation of @xmath4 and @xmath982 period oscillations . see fig . s[fig : s3 ] . this alternation is attributed to the side gate excursion changing not only the enclosed magnetic flux number but also the localized non - abelian @xmath4 quasiparticle number . there are oscillations corresponding to ab interference of @xmath4 particles when an even number of @xmath4 quasiparticles are enclosed , and the @xmath27 oscillations are apparent when that number is odd . according to theoretical predictions @xcite , the @xmath27 oscillations are pervasive but are more easily observed when the larger amplitude @xmath4 oscillations are suppressed . the data of the figure show this alternation for all three rudimentary device sizes . further demonstration and details of these measurements can be found in reference @xcite . ( center ) showing small period oscillations consistent with expression / suppression of non - abelian @xmath4 interference , along with ffts of both spectra ( right ) for a series of interferometer devices of different areas . in our model , the magnetic field @xmath99 necessary to add two @xmath4 quasiparticles is determined from @xmath100 , or @xmath101 . the black line is an average of several ( typically , eight ) b - field sweeps , and the blue line shows the same data smoothed by local averaging . in the fft panel the two vertical blue lines differ by a factor of 5 in frequency . data are taken at @xmath102mk . ] . all devices used here are fabricated from the same heterostructure wafer , with representative bulk transport shown in the left panel . the central column shows representative transport through each device ; note prominence of the @xmath0 minimum and the presence of fqhe state at @xmath1 in these longitudinal resistance traces . the right hand column shows longitudinal resistance change with side gate ( 2 ) sweep at @xmath0 ; each device demonstrates oscillations consistent with the ab effect at periods corresponding to charge @xmath4 . the marked vertical lines of these periods are derived from similar measurements at integers and @xmath63 fillings , defining the period corresponding to @xmath4 charge . in each device the previously observed @xcite alternation of @xmath4 and @xmath27 periods is affirmed in large side gate voltage excursions . temperature in all data is @xmath102mk . ]	we show that the resistance of the @xmath0 quantum hall state , confined to an interferometer , oscillates with magnetic field consistent with an ising - type non - abelian state . in three quantum hall interferometers of different sizes , resistance oscillations at @xmath1 and integer filling factors have the magnetic field period expected if the number of quasiparticles contained within the interferometer changes so as to keep the area and the total charge within the interferometer constant . under these conditions , an abelian state such as the @xmath2 state would show oscillations with the same period as at an integer quantum hall state . however , in an ising - type non - abelian state there would be a rapid oscillation associated with the even - odd effect and a slower one associated with the accumulated abelian phase due to both the aharonov - bohm effect and the abelian part of the quasiparticle braiding statistics . our measurements at @xmath0 are consistent with the latter . [ [ introduction . ] ] introduction . + + + + + + + + + + + + + the origin of the fractional quantum hall effect @xcite at filling factor @xmath3 @xcite has been a long - standing open issue , which is important because it has been conjectured that this state of matter supports non - abelian anyons @xcite . two point - contact fabry prot interferometers have been proposed to observe the aharonov bohm ( ab ) effect and the anyonic braiding statistics of quasiparticles @xcite . in a non - abelian state , not only the phase but the also the amplitude of the observed oscillations is indicative of the braiding statistics @xcite . specifically , if the @xmath3 state is indeed non - abelian , the quasiparticle parity within the interferometer dictates whether the resistance of an interferometer oscillates with enclosed area ( controlled by a side gate ) with a period associated with charge @xmath4 quasiparticles . such oscillations should only be seen when the parity is even the even - odd effect @xcite . previous experiments @xcite are broadly consistent with these predictions @xcite , although some puzzles remain , as we discuss below . in this paper , we examine the magnetic field dependence of the resistance of a series of interferometers with a large range of active areas . two of them are shown in fig . [ fig : devices ] . we formulate a model based on the assumption that the total charge in the interferometer and the enclosed area both remain constant as the magnetic field is varied . we test it at @xmath1 and integer filling factors and show that it is consistent with the experimental data in the @xmath1 case , it predicts a resistance oscillation with the somewhat surprising flux period @xmath5 . we thereby determine the effective area of the interference loop in each device ( and each ` preparation ' of each device , which we describe in the next paragraph ) . the model also predicts that the resistance in the @xmath0 state will oscillate as the product of two oscillations , one with flux period @xmath6 and the other with flux period @xmath7 , as we explain and compare to our experimental data below . [ [ interferometers . ] ] interferometers . + + + + + + + + + + + + + + + + the interferometers used in this paper are fabricated from high - mobility ( @xmath8@xmath9/v@xmath10s ) , high - density ( @xmath11@xmath12 ) gaas / algaas quantum well heterostructures . a @xmath13 nm sin layer is applied to the heterostructure . the size and shape of the 2d electron channel , which is @xmath14 nm below the surface , is controlled by @xmath15 nm thick al top gates that are deposited on the sin layer , as shown in fig . [ fig : devices ] . prior to charging the top gates , the samples can be briefly illuminated to enhance mobility and to provide different sample preparations since the illumination changes the distribution of localized charges in the device , as do different cool - downs @xcite . further description of the device preparations and measurement details is presented in supplemental material . two interfering edge currents result from quasiparticle tunneling across constrictions defined by gate sets 1 and 3 . the longitudinal resistance @xmath16 is measured with contacts labeled * a * through * d * in the electron - micrograph in fig . [ fig : devices](a ) by the voltage drop from contact * a * to * d * , with current driven from * b * to * c * , using standard lock - in techniques . the two standard top gate designs shown in electron micrographs in fig . [ fig : devices](a ) are labeled with device dimension parameters _ x _ and _ y _ adjusted to produce three separate samples with ratios of areas of roughly 3:2:1 . in addition , the functional areas in these devices are defined by applying the gate voltages , resulting in a range of areas from 0.1 to 0.6m@xmath17 . the temperature of the measurements is 20mk in all data presented here . . ( bottom ) the longitudinal resistance @xmath16 for the two samples shows minima corresponding to fractional quantum hall states at @xmath187/3 , 5/2 , and 8/3 . ] [ [ previous - results . ] ] previous results . + + + + + + + + + + + + + + + + + oscillations in @xmath16 have previously been observed as a function of side gate voltage , @xmath19 , which changes the area of the interferometer @xcite . interpreted as due to the aharonov - bohm effect , the expected period of oscillation is @xmath20 ( @xmath21 is the fundamental flux quantum and @xmath22 is the charge of the interfering quasiparticles ) , from which the quasiparticle charge @xmath22 could be obtained if the proportionality constant between @xmath23 and @xmath24 were known . assuming its independence of magnetic field , this constant could be determined from the period of @xmath16 oscillations at integer filling factors , where @xmath25 , or at @xmath1 , where @xmath26 is expected . both filling fractions give similar proportionality constants between @xmath23 and @xmath24 , which supports the idea that this constant is approximately independent of magnetic field . at @xmath0 , @xmath16 oscillations in some intervals of @xmath19 appear consistent with ab oscillations corresponding to @xmath4 charges while in other intervals they seem consistent with @xmath27 charges @xcite . these results have been interpreted as a manifestation of the `` even - odd effect '' @xcite : charge @xmath4 oscillations should be observed only when there is an even number of charge @xmath4 quasiparticles in the interferometer ; charge @xmath27 oscillations should always be observed . when there is an odd number of charge @xmath4 quasiparticles in the interferometer , this should be the only type of oscillation visible @xcite . in refs . only e/2 oscillations are visible in certain side - gate voltage intervals ( when , according to this interpretation , an odd number of e/4 quasiparticles is in the interferometer ) , but it is not clear whether both e/4 and e/2 oscillations or only e/4 are present in the other intervals . by contrast , in this letter we focus on @xmath16 measurements during magnetic field sweeps . at integer filling , a @xmath28-field sweep produces ab oscillations of @xmath16 with period @xmath29gm@xmath17 , where @xmath30 is the current - encircled area of the interferometer ; we thereby determine the active area for each of the different devices and sample preparations . [ [ model . ] ] model . + + + + + + the key assumption in our interpretation of the experimental data is that the charge contained within the interference loop and the area of the loop remain constant as the magnetic field is varied . it is natural to assume that the charge contained within the loop remains constant if it is primarily determined by the local electrostatic potential or , in other words , if the coulomb energy dominates . in such a case , as the magnetic field is varied , one of two possibilities will occur . quasiparticles will be created in the bulk or else the quantum hall droplet will shrink or expand ; in the former case , the area of the interference loop will remain constant . we expect this scenario to hold if there are localized states in the bulk that have very low energy as a result of disorder so that it is energetically favorable to create quasiparticles there , rather than to change the charge density at the edge . when this scenario holds , increasing the flux through the interferometer by @xmath31 causes the number of charge @xmath22 quasiparticles to change by @xmath32 . meanwhile , changing the flux by @xmath31 and the number of charge @xmath22 quasiparticles by @xmath33 causes a change @xmath34 in the phase acquired by a quasiparticle taking one path around the interferometer relative to the phase acquired by a quasiparticle going around the other : @xmath35 \label{eq : ab_phase_general}\end{aligned}\ ] ] the first term on the right - hand - side is the ( ordinary electromagnetic ) ab phase seen by a charge @xmath36 quasiparticle encircling flux @xmath37 . the second term is the statistical phase seen by a charge @xmath36 quasiparticle when it encircles @xmath33 such quasiparticles ; the phase acquired when a single charge @xmath36 quasiparticle encircles another is @xmath38 , assuming that the particles are abelian . for non - abelian particles , more care is required , as we will see below . the relative minus sign can be understood using the argument in @xcite , where it is explained why the ab and statistical phases should cancel under certain conditions . in an integer quantum hall state , @xmath25 and @xmath39 , so @xmath40 and @xmath16 will oscillate with magnetic field period @xmath41gm@xmath17 . now consider the @xmath1 state . if it is in the same universality class as the @xmath42 laughlin state , then @xmath26 and @xmath43 . then @xmath44 . consequently , @xmath16 will show oscillations with period @xmath45gm@xmath17 i.e half that in an integer quantum hall state . now consider the case of @xmath0 . if the system is in an ising - type topological phase such as the moore read state @xcite or the anti - pfaffian state @xcite , then when there is an even number of charge @xmath4 quasiparticles in the interference loop , the ising topological charge will be @xmath46 or @xmath47 , but when there is an odd number in the interference loop , the ising topological charge will be @xmath48 . as a result , if one particular topological charge is energetically favorable for even quasiparticle number let us suppose , for the sake of concreteness , that it is @xmath46 then the non - abelian ising topological charge has a periodicity of two quasiparticles or , taking @xmath49 , a flux period @xmath50 . hence , @xmath16 oscillates with magnetic field period @xmath51gm@xmath17 . however , there is also an abelian phase ( [ eq : ab_phase_general ] ) which can have a different periodicity . the abelian phase acquired when a charge @xmath4 quasiparticle encircles an @xmath52 quasiparticles with ising charge @xmath46 ( or , equivalently , @xmath53 charge @xmath27 quasiparticles ) is @xmath54 and @xmath55 . hence , eq . ( [ eq : ab_phase_general ] ) now reads : @xmath56 . therefore , there is also a slower oscillation in @xmath16 with magnetic field period @xmath57gm@xmath17 . if , however , the ising charge is not fixed to @xmath46 for any even number of quasiparticles , but may be randomly either @xmath46 or @xmath47 , then the slower , period @xmath58 , oscillation will be afflicted by random @xmath59 phase shifts that could wash it out . if the system were in an abelian @xmath2 state , then similar considerations lead to a period @xmath58 oscillation but no rapid period @xmath6 oscillation . and also at @xmath1 ( insets ) , with ffts of these oscillations shown in the left hand panels . the ratio between these two oscillation periods is the same in eight device preparations of varying size . ] [ [ comparison - with - experiment . ] ] comparison with experiment . + + + + + + + + + + + + + + + + + + + + + + + + + + + the overall @xmath28-sweep trace between filling factors 2 to 3 of @xmath16 across two of the interferometers is shown in the bottom two panels of fig . [ fig : devices ] ; they clearly demonstrate fractional quantum hall states at @xmath187/3 , 8/3 , and 5/2 . this overall trace is averaged locally to define a background which we subtract from the raw @xmath16 measurement to make the oscillations clearer . measurement of these oscillation sets was repeated for multiple interferometric areas . the @xmath60 period should change with area according to our model . from the three devices used and the multiple preparations and gate values employed , the measured @xmath60 periods show active areas ranging from 0.1m@xmath17 to @xmath61m@xmath17 . to put our picture to test , we first consider @xmath62 and @xmath63 . oscillations of @xmath16 with @xmath28 are shown in the upper left of fig . [ fig1 ] . the period @xmath60 of oscillations is found to be similar near integer filling 2 , 3 and 4 for each device , consistent with this being an ab oscillation and not coulomb effects @xcite . from the periodicity @xmath60 , we determine the active area of this preparation . ( for instance , for the preparation displayed in fig . [ fig3 ] , @xmath64 g , from which we deduce @xmath65m@xmath17 . ) we now turn to @xmath1 . the results for @xmath28-sweeps are shown in fig . [ fig1 ] . oscillations at 7/3 and integer filling factors are shown in fig . [ fig1 ] insets with their corresponding fourier transforms , which show peaks . the 7/3 peak frequency is twice the @xmath62 peak frequency ( or half the period ) , consistent with the analysis above . the same @xmath16 measurements comparing @xmath1 and integer filling factors were carried out on the three different devices and different preparations of this study , as summarized in the right panel of fig . [ fig1 ] . the @xmath16 oscillations at @xmath1 consistently occur at twice the frequency of their respective integer filling factor oscillations over the full range of device areas studied . we conclude that the assumptions and analysis outlined above are valid . as a function of magnetic field and the associated fourier transforms at ( a ) @xmath62 , ( b ) @xmath1 , and ( c ) @xmath0 . the vertical blue lines mark 5x , 2x , and the integer frequency . the oscillations at @xmath1 are observed to have twice the frequency of those at @xmath62 , which is consistent with the theoretical model explained in the text . the oscillations at @xmath0 show beating between a fast oscillation with a period that is @xmath66 that at @xmath62 and a slow one with the same period as at @xmath62 ] . [ fig2 ] fig . [ fig2 ] presents the comparison between interference oscillations of @xmath68 at @xmath0 , 7/3 , and @xmath62 observed in the same sample / prepapration . ( the overall @xmath28-sweep trace of @xmath16 for this interferometer is shown in the bottom right panel of fig . [ fig : devices ] . ) sets of oscillations are shown with their respective fourier transforms . once again , the @xmath1 oscillations are observed at half the magnetic field period of those at @xmath69 . interestingly , the oscillations at 5/2 contain a higher frequency component , and the fft spectrum demonstrates the predominant frequency is 5 times that of the integer oscillation frequency . this value is consistent with the expected oscillation frequency for expression / suppression of non - abelian @xmath4 interference due to the changing number of quasiparticles with varying magnetic field . moreover , the peak centered around five times the integer frequency is _ split _ , with the splitting being roughly twice the frequency observed at integer plateau . this corresponds to beats which are further consistent with the above prediction for the interplay between the ab and statistical contributions for a _ non - abelian _ @xmath0 state . other such data sets are presented in supplemental materials . ( right inset ) , @xmath69 ( left inset)and at @xmath0 ( top panel ) . from the fourier transforms of @xmath70 vs. @xmath28 ( left panel ) , we see that the oscillation period at @xmath0 is @xmath66 as large as at @xmath71 . sometimes it shows beating with a more rapid oscillation with the same period as at @xmath72 . ] gm@xmath73 . equivalently ( inset ) the oscillation period in units of magnetic field is inversely proportional to the area . ] the observation of a small oscillation period at @xmath0 in a series of samples with different interferometer sizes and different sample preparations is further demonstrated in figs . [ fig3 ] , [ fig4 ] . fig . [ fig3 ] , top panel , shows the @xmath28-sweep results for one of the sample preparations in device area 2 , focusing on the small period resistive oscillations corresponding to multiple parity changes in the enclosed @xmath4 quasiparticle number near 5/2 . the set of oscillations is measured with no adjustments to the voltages of the quantum point contacts or central top gates . if our model is correct , the five periods of oscillation shown here represent ten parity changes . the @xmath28-field range near 5/2 where this data is taken is marked in the overall @xmath16 trace . the distinct resistive oscillations ( black trace ; the blue trace is a coarse smoothing of the data ) near 5/2 in this preparation have a period @xmath74 g , compared to a period at integer fillings of @xmath75 g . this is precisely the same fivefold ratio shown in fig . [ fig2 ] , once again in agreement with our model . note here that splitting in the 5/2 peak is not resolved , which may be an indication that the _ fermion _ parity in the interferometer is not constant , a possibility discussed above . or , alternatively , sweeps through a wider @xmath28-field interval may be necessary to observe this slow oscillation in some samples / preparations . wider sweeps may also reveal oscillations due to transport by charge-@xmath27 quasiparticles , which should have a period @xmath76 ( by essentially the same argument as for @xmath63 ) . they are not apparent in the magnetic field sweeps in figs . [ fig2 ] , [ fig3 ] , even though they are seen in side - gate voltage sweeps @xcite . fig . [ fig4 ] summarizes the principal result of this study . the oscillation period in units of flux ( i.e. , @xmath77 ) at @xmath0 measured for different samples / preparations is approximately independent of the device area derived from @xmath60 . the observed values of @xmath77 is shown to be in reasonable agreement with the expected value of 8gm@xmath17 which corresponds to the change in the parity of enclosed quasiparticles . to conclude , this experiment provides the necessary complement to prior measurements @xcite where ab oscillations were examined as a function of the active interferometer area @xmath30 controlled by the gate voltage . by sweeping the @xmath28 -field in these multiple area devices instead , the previous experimental limitation coming from slow gate charging has been avoided . the resistance oscillations observed near filling factor @xmath0 in multiple devises show a period consistent with the additional magnetic field needed to add one quasihole to their respective ( different ) active areas ( thereby changing the quasiparticle parity ) . we stress that the presence of such a period is indicative of a non - abelian nature of the @xmath0 state . while this interpretation is based on several assumptions discussed earlier , using the @xmath1 fqh state for control measurements significantly strengthens our case . rlw , cn and ks acknowledge the hospitality of kitp supported in part by the nsf under grant phy11 - 25915 . cn and ks are supported in part by the darpa - quest program . ks is supported in part by the nsf under grant dmr-0748925 . 24 natexlab#1#1bibnamefont # 1#1bibfnamefont # 1#1citenamefont # 1#1url # 1`#1`urlprefix[2]#2 [ 2][]#2 , , , * * , ( ) . , , , , , , * * , ( ) . , , , , , , , , , * * , ( ) , . , , , , * * , ( ) , . , * * , ( ) . , , , * * , ( ) . , * * , ( ) , . , * * , ( ) , . , , , * * , ( ) , . , , , , , * * , ( ) , . , , , , * * , ( ) , . , , , * * , ( ) , . , * * , ( ) , . , , , * * , ( ) , . , , , * * , ( ) . , , , * * , ( ) , . , , , , * * , ( ) , . , , , , , * * , ( ) , . , * * , ( ) , . , , , , * * , ( ) , . , , , , * * , ( ) , . , , , * * , ( ) , . , , , , p. ( ) , . , , , ( ) , . @#1@@@(s#1@italiccorr )
the bulk of the data on strong interaction effects in kaonic atoms is due to experiments of over three decades ago . with the exception of the very light atoms of @xmath2h and @xmath2@xmath4he the rest of the data could be described rather well with the help of @xmath2-nucleus optical potentials @xcite . recent experiments on @xmath2h and @xmath2@xmath4he with much reduced background removed the ` puzzles ' with these two atoms @xcite . however , the depth of the attractive @xmath2-nucleus real potential , which in phenomenological analyses came out in the range of 150 - 200 mev @xcite , presented a theoretical challenge in as much as in - medium chiral _ threshold _ @xmath0 scattering amplitude input led to a lower value of order 120 mev @xcite , or even to a considerably lower value of 40 - 50 mev @xcite . this outstanding discrepancy is of current interest , since it is relevant to the role of @xmath2 mesons in multistrange self - bound matter and in compact stars @xcite . the problem has been largely resolved very recently @xcite noting that in - medium chiral _ subthreshold _ @xmath0 scattering amplitudes provide the relevant input and thereby demonstrating the need to supplement the model by multi - nucleon terms , as discussed in the present work . this leads to deep real potentials in agreement with the purely phenomenological analyses . the present paper is an update of refs . @xcite , based on a recent in - medium coupled channel chirally motivated separable interaction model which produces good fits to all the low energy antikaon - nucleon data , including the latest @xmath2h atom results from the siddharta experiment @xcite . section [ sec : knampl ] outlines the self - consistent handling of subthreshold @xmath0 amplitudes while section [ sec : knucleus ] deals with the resulting @xmath2-nucleus amplitudes . section [ sec : potls ] reports on global optical model fits to kaonic atom data and the last section provides summary and conclusions . the potential experienced by a @xmath2 meson of energy @xmath5 interacting with a nucleus of density @xmath6 is given in the single - nucleon approximation by @xmath7 where @xmath8 is the in - medium @xmath0 scattering amplitude , reducing in the low - density limit @xmath9 to the free - space @xmath0 c.m . forward scattering amplitude @xmath10 , @xmath11 is the relative @xmath0 momentum , @xmath12 is the lorentz invariant mandelstam variable equal to the square of the total @xmath0 energy in the two - body c.m . frame , and the nucleon energy @xmath13 is approximated by its mass @xmath14 in the kinematical factor in front of @xmath15 . the in - medium amplitude @xmath8 in this work is a chirally motivated amplitude constructed within a full octet @xmath16 meson octet @xmath17 baryon coupled channel separable interaction model @xcite which in its latest next to leading order ( nlo ) version nlo30 @xcite incorporates the recent siddharta data for the atomic @xmath2h @xmath3 level shift and width @xcite . in this separable interaction model , the in - medium coupled channel scattering amplitudes assume the form @xmath18 with form factors @xmath19 . the momentum dependence introduced by the form factor @xmath20 in the separable interaction model of refs . @xcite is relatively weak for the applications discussed in the present work and is secondary to the strong energy dependence of the reduced amplitude @xmath21 generated by the @xmath22 subthreshold resonance . reduced amplitude ( [ eq : isoscalar ] ) in version nlo30 of the chiral model @xcite below and above @xmath23 mev . dashed curves : free - space amplitude ; dot - dashed curves : pauli blocked amplitude at @xmath24 ; solid curves : including meson and baryon self energies ( se ) , also at @xmath24.,scaledwidth=60.0% ] free - space and in - medium reduced amplitudes for half nuclear matter density @xmath25 are shown in fig . [ fig : nlo30 ] for the isospin - averaged combination @xmath26 corresponding to symmetric nuclear matter . similar results are obtained at full nuclear matter density @xmath27 . of the two in - medium amplitudes shown in the figure , the one marked without se " imposes pauli blocking on intermediate @xmath28 states for @xmath29 @xcite , whereas the one marked with se " adds self consistently hadron self energies in intermediate states @xcite , following a procedure suggested in ref . the real part of all three amplitudes exhibits strong energy dependence , switching from weak attraction above @xmath0 threshold to strong attraction below threshold . as a rule of thumb , @xmath30 fm translates into a sizable attraction @xmath31 mev . the imaginary part of these amplitudes exhibits a peak , related to the subthreshold @xmath22 resonance , with a steep decrease at lower energies , becoming vanishingly small near the @xmath32 threshold about 100 mev below the @xmath0 threshold . in between the two limits of the energy scale in fig . [ fig : nlo30 ] , @xmath33 mev , the three amplitudes differ appreciably from each other . at threshold , in particular , the real part of the with se " amplitude is about half of that without se " , corresponding to a depth @xmath34 mev , in agreement with ramos and oset @xcite . it was recognized in the early 1970s that the strong energy dependence of the two - body amplitude , particularly in the subthreshold region where the @xmath28 quasibound state @xmath22 dominates , provides the underlying structure for @xmath2 nuclear interactions at and near the @xmath2 nucleus threshold @xcite . this idea has been reformulated and applied recently in refs . @xcite to a comprehensive study of kaonic atoms . the essential idea is to replace the two - body variables @xmath11 and @xmath35 of the in - medium scattering amplitude by appropriate density dependent averages in the nuclear medium . this may be summarized by the following relationships : @xmath36 upon neglecting quadratic terms in the binding energies @xmath37 near threshold ( @xmath38 ) , and @xmath39 where @xmath40 in both of these substitutions . replacing in eqs . ( [ eq : sqrts ] ) and ( [ eq : p^2 ] ) the kinetic energy @xmath41 in the local density approximation by @xmath42 where @xmath43 , with @xmath44 the @xmath2 coulomb potential generated by the finite - size nuclear charge distribution , and approximating the nucleon kinetic energy @xmath45 in the fermi gas model by @xmath46 mev , eqs . ( [ eq : sqrts ] ) and ( [ eq : p^2 ] ) become @xmath47 where all the terms following @xmath48 on the r.h.s . are negative , thus implementing the anticipated downward energy shift into the @xmath0 subthreshold energy region , and @xmath49 , \;\;\ ; ( \rm { in~mev } ) \label{eq : p^2final}\ ] ] where both terms on the r.h.s . are positive for attractive potentials @xmath50 . the dominant contribution in eqs . ( [ eq : sfinal ] ) and ( [ eq : p^2final ] ) arises from @xmath51 , resulting in a downward energy shift of up to 60 mev as demonstrated in fig . [ fig : nlo30evsrho ] , and in values of @xmath52 as high as 275 mev / c for @xmath2 nuclear potential depths reaching 180 mev in phenomenological studies @xcite . these momenta are well within the nlo30 momentum dependence scale @xmath53 mev / c . having transformed the dependence of the in - medium scattering amplitudes @xmath54 ( @xmath55 ) on @xmath11 and @xmath35 into a density dependence , we denote the resultant in - medium scattering amplitudes by @xmath56 . in order to allow for different proton and neutron distributions in the actual calculations detailed below , the in - medium amplitude @xmath57 which should substitute for @xmath58 in the construction of the @xmath2-nucleus potential eq . ( [ eq : t ] ) is further replaced by an effective in - medium amplitude @xmath59 : @xmath60 with @xmath61 and @xmath62 normalized to @xmath63 and @xmath64 , respectively , and @xmath65 . the reduced amplitudes @xmath66 and @xmath67 are evaluated at @xmath35 given by eq . ( [ eq : sfinal ] ) , where the @xmath2 atomic binding energy @xmath68 is neglected with respect to the average nucleon binding energy @xmath69 mev . in our kaonic atom global fit hardly matters within reasonable limits . sensitivity to @xmath70 is expected in studies limited to light nuclei , as exhibited recently by analyzing finuda data @xcite of @xmath71 hypernuclear formation with stopped @xmath2 mesons on targets from lithium to oxygen @xcite . ] a similar approximation is made in eq . ( [ eq : p^2final ] ) for @xmath72 when using the form factors @xmath73 of eq . ( [ eq : sep ] ) . the @xmath2-nucleus potential @xmath74 is calculated by requiring self consistency in solving eq . ( [ eq : sfinal ] ) with respect to @xmath75 , i.e. , the value of @xmath76 in the expression for @xmath77 and in the form factors @xmath20 has to agree with the resulting @xmath76 . this is done at each radial point and for every target nucleus in the data base . the present model transforms the energy dependence of subthreshold effective amplitudes , eq . ( [ eq : effa ] ) , into density dependence . this transformation is hardly sensitive to the nucleus involved , as is seen in fig . [ fig : nlo30evsrho ] calculated in the without se " version of the nlo30 model . in the with se " version the energies for a given density differ from the plotted values by 2 - 3 mev . the energy shifts do not vanish for zero density because we have used a fixed average nucleon binding energy of 8.5 mev . replacing it by a position - dependent @xmath78 , in order to satisfy the low - density limit , causes the energy shift to vanish far outside the nucleus , with minor overall effects on the present results . it is seen from the figure that , e.g. , for a density of 50% of nuclear matter density , the downward energy shift is @xmath79 mev which , from fig . [ fig : nlo30 ] , implies a real amplitude at least twice larger than the threshold value in the with se " version . the density dependencies of the real and of the imaginary part of the effective amplitude @xmath80 are of particular interest because they are related to characteristic features of phenomenological optical potentials . it was shown already in 1993 @xcite that with empirical density - dependent potentials , where the effective @xmath0 scattering amplitude within a @xmath81 model depends on the density , improved fits to the data were obtained compared to fits using fixed amplitudes . it was observed that in addition to the increased depth of the best - fit real potentials , these were characterised by _ compression _ relative to the corresponding nuclear densities , with r.m.s . radii of the real potential smaller than the corresponding nuclear radii . reduced r.m.s . radii of optical potentials mean that the underlying in - medium @xmath0 interaction increases with density , a robust feature that is insensitive to details . figure [ fig : effa ] shows the nlo30 effective amplitudes for ni and pb as function of nuclear density , calculated in the without se " version . qualitatively similar results are obtained also in the with se " version , see ref . regions of low density , i.e. large radii , are the most effective in determining the r.m.s . radius of a distribution and the sharp rise of the real part in the extreme surface region can lead to compression of the real potential . the opposite dependence is observed for the imaginary part , thus implying _ inflation _ of the imaginary potential relative to the nuclear density . quantitatively , however , the change of r.m.s . radii relative to nuclear densities is found to disagree with the empirical trends of ref . @xcite , with too little compression for the real part and far too strong inflation for the imaginary part . inevitably this is reflected in the quality of agreement with the data , as demonstrated in the next section . strong - interaction effects in kaonic atoms are dominated by absorption , as is evident from level widths being significantly larger than the corresponding level shifts . moreover , the shifts are always repulsive although the real potential is attractive , again pointing to the dominance of absorption . therefore it is argued that the above marked decrease of the imaginary part of the effective scattering amplitude is the main deficiency of the present model , a decrease originating in the sharp decrease of the imaginary part of the free amplitude ( fig . [ fig : nlo30 ] ) towards the @xmath32 threshold , which is typical of the single - nucleon approach . consequently it is reasonable to expect that additional , multi - nucleon terms are required to obtain good fit to the data . strong interaction level shifts and widths in kaonic atoms have been calculated by solving a klein - gordon equation @xcite with the optical potential of eq . ( [ eq : t ] ) transformed to the @xmath2-nucleus c.m . system , and where the in - medium @xmath0 scattering amplitude is given by the effective amplitude eq . ( [ eq : effa ] ) : @xmath82 with @xmath83 the kaon - nucleus reduced mass and @xmath84 . two - parameter fermi distributions were used for both densities , with @xmath61 obtained from the known charge distribution by unfolding the finite size of the charge of the proton . for @xmath62 averages of the ` skin ' and ` halo ' forms of ref . @xcite were adopted with the difference between r.m.s . radii given by @xmath85 fm . nuclear potentials for @xmath2 atoms of ni . dashed curves : derived from in - medium nlo30 amplitudes ; solid curves : plus phenomenological terms from global fits.,scaledwidth=60.0% ] nuclear potentials for @xmath2 atoms of pb . dashed curves : derived from in - medium nlo30 amplitudes ; solid curves : plus phenomenological terms from global fits.,scaledwidth=60.0% ] figures [ fig : nipotl ] and [ fig : pbpotl ] show optical potentials for @xmath2-ni and for @xmath2-pb , respectively . the potentials marked nlo30 follow directly from the in - medium @xmath0 amplitudes in model nlo30 , without any adjustable parameters . the agreement with the full data set of 65 points , covering the whole periodic table , is poor , with @xmath86 per point in the range of @xmath87 . this is not surprising in view of the obvious deficiency of the single - nucleon approach where the imaginary part of the amplitude goes down rapidly towards far subthreshold energies . adding to the potential an empirical term linear in the nuclear density does not improve much the fit to kaonic atoms data and only by further addition of a @xmath88 or a @xmath89 term good fits to the data are possible . both imaginary and real parts of the additional phenomenological potential are then found to be dominated by @xmath90 or @xmath91 terms which are likely to represent multi - nucleon absorptive and dispersive contributions , respectively . figures [ fig : nipotl ] and [ fig : pbpotl ] also show potentials obtained when adjustable @xmath92 terms are added and the four parameters @xmath93 and @xmath94 are determined by requiring best fit to the data . the quality of the fits is then quite acceptable , with @xmath86 per point of 2 to 2.3 . qualitatively similar results to those displayed in figs . [ fig : nipotl ] and [ fig : pbpotl ] are obtained also in the with se " version , in agreement with the discussion for ni in ref . @xcite . considering values of the potentials at the nuclear center , the additional phenomenological part appears too large to be regarded as a correction term to the basic nlo30 amplitude . however , values of the potential at the center are rather meaningless in the context of kaonic atom observables . the sensitivity of calculated level shifts and widths to the @xmath2 nuclear potentials was found @xcite to be around the nuclear surface and certainly not at the center . with that in mind and focusing on the imaginary potential as noted above , it is remarkable that the additional term modifies the shape of the imaginary potentials in the surface region , bringing their r.m.s . radii closer to empirical values . changes of the imaginary potentials near the surface due to the phenomenological term are of the order of 30% of the nlo30 potentials , consistent with the fraction of multi - nucleon absorption estimated from experiments with emulsions and bubble chambers @xcite . the emerging phenomenology is similar to that for @xmath95 in pionic atom studies where theoretically motivated single - nucleon contributions are supplemented by phenomenological @xmath96 terms representing @xmath97 processes @xcite . ( see also ref . a simple ansatz for transforming the strong energy dependence of subthreshold @xmath0 scattering amplitudes in the nuclear medium to appropriate density dependent averages was presented and employed in global analyses of kaonic atom data , following refs . @xcite . with chirally motivated coupled channel separable interaction scattering amplitudes in model nlo30 @xcite that respect the low energy @xmath98 data , including the recent siddharta results for kaonic hydrogen @xcite , the connection between this model and deep real optical potentials was re - established . effective @xmath2-nucleus amplitudes were derived self - consistently and were critically reviewed with respect to empirical features of phenomenological optical potentials . we focused in the present update on the in - medium effects arising exclusively from the strong energy dependence of subthreshold @xmath0 amplitudes , using for this purpose the without se " version of the nlo30 in - medium model . the introduction of self - energy effects in the with se " version is necessarily model dependent to some extent . nevertheless , all of our findings and conclusions hold true in both without se " and with se " versions , with minor differences exhibited already within earlier versions @xcite . in the present update , as well as in the preceding studies @xcite , the steep decrease of the imaginary part of the amplitude as function of the nuclear density , due to the single - nucleon nature of the model , was identified as a major deficiency of the single - nucleon approach . this conclusion is valid also upon adding effective @xmath2-nucleus amplitudes generated by the @xmath99-wave @xmath100 subthreshold resonance , as discussed in ref . @xcite where @xmath99-wave effects were found secondary to @xmath22-dominated @xmath101-wave effects . good agreement with experiment was achieved by adding to the potential a phenomenological part which was found to be dominated by a @xmath90 or a @xmath91 term . including systematically multi - nucleon processes should be the next step in trying to obtain @xmath2-nucleus potentials from in - medium @xmath0 interaction input . we thank ale ciepl for communicating to us the results of the nlo30 chiral model , and him and our other collaborators daniel gazda and ji mare @xcite for fruitful discussions . acknowledges gratefully stimulating discussions with wolfram weise and the hospitality extended to him in fall 2011 by members of the t39 group at the technische universitt mnchen . this work was supported by the eu initiative fp7 , hadronphysics2 , under project no .	recent work on the connection between in - medium subthreshold @xmath0 amplitudes and kaonic atom potentials is updated by using a next to leading order chirally motivated coupled channel separable interaction model that reproduces @xmath1 observables at low energies , including the very recent siddharta results for the atomic @xmath2-hydrogen @xmath3 level shift and width . the corresponding @xmath2-nucleus potential is evaluated self - consistently within a single - nucleon approach and is critically reviewed with respect to empirical features of phenomenological optical potentials . the need to supplement the single - nucleon based approach with multi - nucleon interactions is demonstrated by showing that additional empirical absorptive and dispersive terms , beyond the reach of chirally motivated @xmath2-nucleus potentials , are required in order to achieve good agreement with the bulk of the data on kaonic atoms . in - medium subthreshold scattering amplitudes , coupled channel chiral models , kaonic atoms 13.75.jz , 21.65.jk , 36.10.gv
it is generally perceived that dwarf galaxies are primarily of low surface brightness , and that lower luminosity galaxies have fainter surface brightnesses . in particular , low surface brightness galaxies ( lsbgs ) seen towards a cluster are conventionally assumed to be members , while apparently faint , but high surface brightness galaxies ( hsbgs ) are presumed to be luminous objects in the background ( e.g. sandage , binggeli & tammann 1985 ) . indeed , it has frequently been asserted that dwarf galaxies ( in particular ) exhibit a strong surface brightness luminosity relation ( e.g. sandage , binggeli & tammann 1985 ; bingelli , sandage & tammann 1985 ; kormendy 1985 ; ferguson & sandage 1988 ) . binggeli ( 1994 ) has particularly persuasively argued this point with reference to the local group . nevertheless , other authors have noted that a good deal of the evidence for the correlation relies on the ` eyeball ' selection of dwarf galaxies ` similar ' to ones already known , that any correlation breaks down at the low luminosity end , and/or that there may be strong selection effects limiting the detected galaxies photometric parameters ( e.g. phillipps , davies & disney 1988 ; irwin et al . 1990 ) . the failure of the relation , i.e. the existence of large background lsbgs ( such as the serendipitously discovered malin 1 ; bothun et al . 1987 ) , or of a population of high surface brightness ( compact ) dwarfs in a cluster ( drinkwater & gregg 1998 ) , or indeed of a population of galaxies so compact as to masquerade as stars and , hence , be missed altogether from galaxy samples ( e.g. arp 1965 ; drinkwater et al . 2000b ; phillipps et al . 2001 ) could have a dramatic effect on our perception of the galaxy population as a whole ( see impey & bothun 1997 ) . the primary factor preventing a resolution of the controversy has been the non - existence of suitable samples of confirmed dwarfs of known distance ( and hence luminosity ) with well defined selection limits . generally , dwarf galaxy samples have been either subjectively ( for instance , morphologically ) selected sets of cluster galaxies ( e.g. ferguson 1989 ) or objectively chosen sets of galaxies with unconfirmed cluster membership / unknown distances ( e.g. irwin et al . 1990 ) . however , the development of a new generation of multi - object spectrographs , exemplified by the ` 2 degree field ' , or 2df , multi - fibre spectrograph on the anglo - australian telescope ( aat ) ( see e.g. taylor , canon & parker 1998 ) , has made it possible to conduct a truly complete spectroscopic survey of a given area on the sky , down to well determined , faint limits , irrespective of image morphology or any other pre - selection of target type , and thus obtain redshifts for a large and unbiased sample of targets ( phillipps 1997 ) . our fornax cluster spectroscopic survey ( fcss ) aims to exploit the huge multiplexing advantage of 2df in this way by surveying , in total , a region of 12 deg@xmath1 centred on the fornax cluster of galaxies . the prime reason for choosing this area of sky area was , of course , the presence of the fornax cluster itself . fornax , with virgo , is the nearest reasonably rich cluster ( approximately abell richness class 0 it is supplementary cluster s0373 in abell , corwin & olowin 1989 ) . the targets of the fcss encompass both cluster galaxies , of a wide range of types and magnitudes , and background galaxies ( over a similarly wide range of morphologies ) , as well as galactic stars and qsos ( drinkwater et al . our main motivation for an all - object survey was to determine cluster membership for as complete a sample of objects as possible across a wide range of surface brightnesses . we wished to test , in particular , whether the usual assignment of lsbgs to the cluster and high ( or even ` normal ' ) surface brightness faint galaxies to the background is justified . the situation regarding hsbgs is even less clear than for lsbgs the question of the very existence of normal or high surface brightness dwarf galaxies ( excepting possibly pathological cases like m32 ) has been highly controversial see , for example , the contrasting views expressed in ferguson & sandage ( 1988 ) and irwin et al . an up to date discussion is given in trentham & hodgkin ( 2002 ) . drinkwater et al . ( 2000a ) ( hereafter paper i ) outlined the aims and methods of the fcss and presented the results from the first 2 diameter field surveyed . drinkwater et al . ( 1999 ) ( paper ii ) described the nature of several compact galaxies found beyond the cluster . meyer et al . ( 2001 ) ( paper iv ) discussed the quasars found in the fcss data . drinkwater et al . ( 2000b ) ( paper iii ) and phillipps et al . ( 2001 ) ( paper v ) described the discovery of 5 extremely compact , high surface brightness dwarf galaxies within the cluster . these objects ( dubbed ` ultra - compact dwarfs ' -ucds ) have intrinsic sizes @xmath6100 pc . they are more compact and significantly less luminous than other known compact dwarf galaxies , yet are much brighter than globular clusters . the data from the first field ( see paper i ) revealed that the fornax cluster is clearly defined kinematically . a total of 24 galaxies from the data lie within the cluster ( this includes the 5 ucds described in papers iii and v ) . this is , to our knowledge , the first complete sample of dwarf galaxies of known distance ( and hence luminosity ) with well defined selection limits . in this paper we study the nature of the objects in this complete sample . section 2 briefly describes the methods and present status of the fcss . section 3 describes the contents of the complete fornax cluster sample and presents @xmath7 and @xmath8 photometry , radial intensity profiles and radial colour distributions for all 24 confirmed cluster members . using this data we are able to explore ( in section 4 ) the distribution in luminosity and surface brightness of the confirmed cluster members . section 5 presents our conclusions . the fcss survey - area will ultimately comprise four separate 2df fields , covering the majority of the fornax cluster area ( see fig . 1 and table 1 of paper i ) . the results presented in this paper are derived from the first of these four fields to be completed . field 1 is centred on the large galaxy ngc 1399 at the centre of the cluster ( r.a.=03@[email protected]@xmath11 , dec.=352701 , j2000 ) . in common with the main 2df galaxy redshift survey ( 2dfgrs ; see colless et al . 2001 and references therein ) , we have chosen to select our targets from catalogues based on ukst sky survey plates digitised by the automatic plate measuring ( apm ) facility at cambridge ( see irwin , maddox & mcmahon 1994 ) . however , unlike other galaxy surveys , which only select resolved images for spectroscopic measurement , we avoid any morphological pre - selection and include all objects , both resolved and unresolved ( i.e. ` stars ' and ` galaxies ' ) . this means that we can include galaxies with the greatest possible range of surface brightnesses : our only selection criterion is the ( blue ) magnitude limit . including objects normally classified as stars greatly increases the size of our sample but it is the only way to ensure completeness . in order both to cover a large number of targets and to go significantly deeper than previous spectroscopic cluster surveys we chose to limit our survey at a @xmath12 magnitude of 19.7 . this is marginally deeper than the main 2dfgrs , which is limited at @xmath13 . @xmath12 is the natural @xmath7 band photographic system of the ukst data . this is related to the standard cousins @xmath7 magnitude by @xmath14 ( blair & gilmore 1982 ) . targets are also chosen to be fainter than @xmath15 , since the photometry of brighter objects is problematic with the ukst / apm data ( particularly because of image saturation ) . note that the selection limits of the overall apm object catalogue ( images larger than a certain area at a given isophotal detection threshold ) do not impinge on our target selection , since any galaxies brighter than @xmath12 = 19.7 and with surface brightness high enough to be observable with 2df ( see below ) are well above the apm limit ( see also paper i ) . our input catalogue for the fcss is a standard apm ` catalogues ' file ( irwin et al . 1994 ) of field f358 from the ukst southern sky survey . the field is centred at r.a.=03@xmath937@xmath10559 , dec.=345014 ( j2000 ) and the region scanned for the catalogue file is [email protected] deg@xmath1 , approximately centered on the fornax cluster . the apm image catalogue lists image positions , magnitudes and morphological classifications ( as ` star ' , ` galaxy ' , ` noise ' , or ` merged ' ) measured from both the blue and red survey plates . the ` merged ' image classification indicates two overlapping images : at the magnitudes of interest for this project the merged objects nearly always consisted of a star overlapping a much fainter galaxy . all the positions are measured from the more recent red survey plate ( epoch 1991 september 13 compared to 1976 november 18 for the blue plate ) to minimise problems with proper motions . the apm catalogue magnitudes are calibrated for unresolved ( stellar ) objects only ( see bunclark & irwin 1984 ) , so we supplemented these with total magnitudes for the galaxies measured by direct analysis of the digitised plate data . these magnitudes were obtained by fitting exponentials to the apm galaxy surface brightness profiles ( see phillipps et al . 1987 and morshidi - esslinger , davies & smith 1999 ) across the surface brightness range 22.7 to 25.7 b mag arcsec@xmath17 ( i.e. between the possibly saturated inner parts of the images and the sky noise dominated outer parts ) . central surface brightnesses and exponential scale lengths could be determined in this way and combined to give pseudo - total magnitudes ( see paper i for further details ) ; we refine these measurements for the cluster objects in the present paper . our target selection then consisted simply of taking all objects from the apm catalogue in each of our four 2df fields with magnitudes in the range @xmath18 . we take the distance to the cluster to be 20 mpc ( a distance modulus of 31.5 mag ) as derived by drinkwater , gregg & colless ( 2001b ) . this leads to an absolute magnitude range for cluster galaxies 14.8 @xmath6 @xmath19 @xmath6 11.6 , so all the cluster galaxies from fcss are dwarfs in terms of their luminosity . we did not apply any morphological selection , although the apm image classifications from the blue survey plate were used to determine which photometry to use ( i.e. the catalogued apm magnitude in the case of unresolved images , or our fitted magnitudes for resolved images ) . although our targets cover a wide range of surface brightnesses ( see fig . 1 of paper i ) , our final spectroscopic sample inevitably suffers from incompleteness at the low surface brightness end . it is not possible to measure spectra for the faintest lsbgs catalogued ( at central surface brightness around 24.5 b mag arcsec@xmath17 ) in reasonable exposure times even though the multiplex advantage of 2df enables us to profitably expose for longer times and hence go fainter than most previous nearby cluster work . two hour exposures enabled us to obtain measurable spectra for galaxies with central surface brightnesses down to 23.2 b mag arcsec@xmath17 with a success rate of @xmath2080 per cent ( see paper i ) . in order to try to extend our surface brightness limit , we have also experimented with much longer exposures ( around 4 hours ) for the lowest surface brightness subset and have obtained successful spectra for some objects with surface brightnesses as faint as 23.7 b mag arcsec@xmath17 . this forms the practical lower limit to our cluster galaxy sample . we observed all our targets with an identical observing setup for 2df : the 300b grating and a central wavelength setting of 5800 giving a wavelength coverage of 36008010 at a resolution of 9 ( a dispersion of 4.3 per pixel ) . the 2df facility includes its own data reduction package , 2dfdr , which permits fast , semi - automatic reduction of data direct from the instrument . however , when we started the fcss project 2dfdr was still under development , so we chose instead to reduce the data with the dofibers package in iraf , as described in paper i. we have since reduced the data independently with 2dfdr , obtaining entirely comparable results ( deady 2002 ) . in particular , 98 per cent of spectra gave the same velocity shift ( see below ) to within the quoted errors ; the remaining 2 per cent turned out , on further inspection , to have spurious features which had not been removed by the cleaning of the original data . redshifts are determined in a standard fashion via cross - correlation of the spectra with templates , using rvsao in iraf ( kurtz & mink 1998 ) . however , as we required an automated system capable of coping with a large number of stars as well as galaxies , we have used stellar template spectra of various types , rather than absorption line galaxy spectra . these were chosen from the library of jacoby , hunter & christian ( 1984 ) . in addition we use emission line galaxy and quasar templates . for each spectrum we determined the best - matching template , the tonry - davies ( 1979 ) r coefficient , the redshift and its error . ( again see paper i for more details and illustrations of the template spectra used ) . the redshifts are measured as radial velocities in units of @xmath21 and are subsequently converted to heliocentric values . by choosing the template giving the best @xmath8 coefficient we can determine not only the redshift , but a first estimate of the object type . we only accept identifications with @xmath22 . objects with redshifts of @xmath23500 kms@xmath24 or less are galactic stars for which the best template indicates the stellar spectral type . at higher redshifts , external galaxies are separated into absorption - line types if they match one of the stellar spectra or emission - line types if they match the emission - line galaxy template best . typical velocity errors are around 64 km s@xmath24 , while comparison with 44 galaxies with redshifts given in the nasa / ipac extragalactic database ( ned ) reveals a negligible external mean error of @xmath25 km s@xmath24 and rms scatter of 110 km s@xmath24 , consistent with our internal errors given the heterogeneous data compiled in ned . [ cols= " < , > , > , > , < , < , > , > " , ] as noted earlier , initial photometry for all the catalogue galaxies was derived from the apm scan data in automated fashion , assuming pure exponential radial profiles and circular isophotes ( cf . morshidi - esslinger et al . 1999 ) . we have now improved on this photometry by using photographic data from a iiiaj @xmath7 plate and a tech pan @xmath8 film from the ukst , digitised by the supercosmos measuring machine at the royal observatory , edinburgh ( e.g. hambly , irwin & macgillivray 2001 ) . table 2 presents the results of this photometry . the data were calibrated ( deady 2002 ) onto the standard cousins @xmath7 and @xmath8 system via comparison of the surface brightness profiles of four suitable , lowish surface brightness dwarfs ( so as to avoid problems of saturation at moderate to high surface brightness on the plates ) with corresponding ccd data from davies , phillipps & disney ( 1990 ) . the methodology was as in phillipps & parker ( 1993 ) . note that none of the calibrating galaxies are actually in the present cluster galaxy sample , either being outside the 2df area or of too low a surface brightness for 2df spectroscopy . the davies et al . ( 1990 ) calibration is closely similar ( to 0.1 mag ) to both that of the original apm low surface brightness galaxy survey of davies et al . ( 1988 ) and the fornax cluster catalog of ferguson ( 1989 ) . after independently calibrated curtis schmidt data became available ( karick et al . , in prep . ) we were able to confirm that our photographic and ccd @xmath7 magnitude scales were in excellent agreement , too . to determine total magnitudes for our galaxies , we used the sextractor package ( bertin & arnouts 1996 ) to generate kron magnitudes ( kron 1978 ) . we selected an aperture size of 3.5 kron radii ( cf . metcalfe et al . 1995 ) , which contains 99.3 per cent of the light for an exponential profile . our magnitudes should therefore be total to within @xmath50.01 mag . a more significant source of error will be the determination of the background sky level . a realistic measure of the uncertainty in this was judged to be @xmath26 , where @xmath27 is the width ( standard deviation ) of the histogram of local sky values . repeating the kron radius and magnitude calculation with the background varied by this amount gives our estimate of the magnitude errors from this source . allowing also for possible systematic zero point errors of @xmath50.05 mag we find final photographic magnitude errors @xmath50.10.2 mag . the @xmath7 and @xmath8 magnitudes for all our cluster galaxies are summarized in columns 4 and 5 of table 2 . note , however , that the photographic data of the ucds ( denoted by an asterix in table 2 ) suffer from saturation , so the @xmath7 and @xmath8 values in table 2 only provide faint magnitude limits for these objects . column 6 presents @xmath19 values assuming d=20 mpc ( see section 2 ) . in this column , we have avoided the problems of the saturation of the photographic data for the ucds , by including , for these objects , the absolute magnitudes derived from curtis schmidt ccd data by phillipps et al . ( 2001 ) , and corrected from @xmath12 to @xmath7 by assuming @[email protected] . the @xmath29 colours are all rather similar ( nearly all between 1.2 and 1.5 ) , and are consistent with the spectral classifications as quiescent de or ds0 galaxies . aside from the ucds , the galaxies have a mean colour of @[email protected] , essentially identical to the @xmath30=1.38 reported by davies et al . ( 1990 ) for their ( presumed ) cluster lsbgs . although the numbers are small , there is a ( comforting ) slight trend for redder colours for the objects which match best to later type stellar spectra ( in particular the two m0 matches have an average @xmath29 of 1.54 ) . the starlink routine elprof was used to determine the radial intensity profile , as a function of semi - major axis length , for the cluster dwarfs ( as well as the calibrating galaxies ) from the photographic @xmath7 and @xmath8 data . the radial intensity profiles and radial colour distributions are illustrated in fig . most of the radial intensity profiles are well fitted by a simple exponential law ( straight line in the plot ) , typical of dwarf galaxies , though some do show some curvature . in some cases this is due merely to the saturation of the @xmath7 band measurements at @xmath522 b mag arcsec@xmath17 . this is less of a problem with the @xmath8 profiles as the tech pan films have somewhat better dynamic range . for completeness we have therefore also fitted the more general sersic ( 1968 ) law ( cf . davies et al . this fitting confirms that 13 of the 19 normal ( i.e. not ultra - compact ) dwarfs have best fitting indices @xmath32 ( in the @xmath7 band ) , where @xmath33 corresponds to a pure exponential . four of the remainder have @xmath34 between 0.4 and 0.8 ( a pure de vaucouleurs ( 1959 ) law , appropriate for giant ellipticals , would have @xmath35 ) , while two have @xmath36 . given that the differences in @xmath34 calculated for the @xmath7 and @xmath8 band profiles are typically around 0.2 ( even when there is no systematic colour gradient ) , that is probably a reasonable estimate of the level of uncertainty in @xmath34 . for subsequent fitting we thus assume that the standard exponential profile is a sufficiently good fit in all cases . columns 2 and 3 of table 2 therefore contain the scale sizes @xmath3 ( in arcsec ) and extrapolated central surface brightnesses @xmath37 of the exponential fit to the @xmath7 band radial intensity profiles . column 7 lists the scale size in kpc ( assuming d=20 mpc ) . fitting was carried out in the range between 22.5 and 26 b mag arcsec@xmath17 . for the compact objects , the tabulated values are merely limits on the intrinsic surface brightnesses and scale lengths due to the effects of the point - spread function ( as discussed in more detail in paper v ) . errors in the scale sizes from the fitting procedure are typically of order 10 per cent , and the differences between the @xmath7 and @xmath8 scale lengths are at the same level in almost all cases ( though a couple show systematic colour gradients and therefore differences up to @xmath3820 per cent ) . one galaxy , fcss j034100.8 - 354433 , is difficult to treat as simply as the others as it appears to have a two component structure . in the table we give the parameters for its ` outer disc ' ( which has a very low surface brightness around 24.5 b mag arcsec@xmath17 , making it an outlier in the plots below ) , but it also has a much more steeply rising ` inner bulge ' with actual central surface brightness more like 23 b mag arcsec@xmath17 ( hence the successful spectroscopy of this object ) , and a scale size of around 2 arcsec . given the small range of colours for most of the sample , unsurprisingly no correlation between colour and surface brightness or scale length is apparent . two necessary steps towards any full understanding of the processes of galaxy formation and evolution are ( i ) the compilation of a full and complete census of the local extragalactic population and ( ii ) an accurate and detailed quantification of this population . traditionally astronomers have sought to quantify the local galaxy population by constructing galaxy luminosity functions ( lfs ) . the lf of a sample of galaxies is the space density of the galaxies in that sample as a function of absolute luminosity . in theory , if one could select a complete sample of galaxies over a sufficiently large section of the local universe such that large - scale structure and clustering effects could be ignored then the lf which could be constructed would provide a definitive measure of the local luminosity density . in practice the many studies of the field population lf ( loveday et al . 1992 ; lin et al . 1997 ; marzke et al . 1997 ; bromley et al . 1998 ; muriel , valotto & lambas 1998 ; zucca et al . 1997 ; folkes et al . 1999 ) have produced significantly differing results and have generally not been complete to very faint limits . an alternative approach is to seek for clues about galaxy formation and evolution by constructing the lf for particular environments and then studying the difference in the derived lfs between environments . the many recent studies of the lfs of particular clusters ( e.g. driver et al . 1994 ; de propris et al . 1995 ; lobo et al . 1997 ; wilson et al . 1997 ; valotto et al . 1997 ; smith , driver & phillipps 1997 ; trentham 1997a , 1997b , 1998 ; driver , couch & phillipps 1998b ; garilli , maccagni & andreon 1999 ) fall into this category . however , it is now well established that the inclusion of a galaxy in an optically selected sample is critically dependent not only on its absolute luminosity but also on its surface brightness ( disney 1976 ; phillipps , davies & disney 1990 ; impey & bothun 1997 ) . it is also well established that lsbgs exist with a wide range of total luminosities . such galaxies will have been excluded from most determinations of the lf . a more thorough and physically meaningful way to quantify the extragalactic population is the construction of a bivariate brightness distribution ( bbd ) , i.e. the space density of galaxies as a function of their total luminosity and their surface brightness or , equivalently , as a function of their size and surface brightness ( boyce & phillipps 1995 ; dalcanton 1998 ; cross & driver 2002 ) . a consideration of any possible surface brightness luminosity relation is then encompassed as part of the wider determination of the bbd . past attempts at deriving a bbd ( e.g. choloniewski 1985 ; van der kruit 1987 ; sodre & lahav 1993 ) have generally suffered from poor statistics . the advent of wide - field imaging surveys and dedicated spectroscopic surveys has made the measurement of the bbd for statistically significant populations possible for the first time ( driver 1999 ; cross et al . 2001 ; de jong & lacey 2000 ; blanton et al . 2001 ) . these authors have all studied the ` field ' population to relatively bright limits . a thorough attempt , therefore , to catalogue and quantify the contents of a cluster , requires , as a first stage , the detection of a complete sample of cluster objects to well defined ( preferable very faint ) limits in absolute luminosity and surface brightness . secondly , the morphologies of the various objects found need to be determined . finally , the bbd for each morphological type can be constructed . even this could be further refined since , for example , one might expect the morphological composition of the cluster core to differ from its outer regions . hence , one may wish to study the bbd as a function of cluster radius . whilst we do not yet have a large enough sample of fornax cluster objects to undertake a full determination of the cluster bbd , we do have a complete sample close to the cluster centre with well defined luminosity and surface brightness limits and can , therefore , draw some general conclusions about the luminosity surface brightness or size surface brightness distribution for the first time . 3 shows the size surface brightness distribution of the fornax cluster sample using the b - band data from table 2 ( i.e. plotting @xmath37 against @xmath3 ) . the limits imposed by the apm magnitude selection limits ( @xmath39 ) are noted . these boundaries become slightly fuzzy in terms of the final photometry since the target selection was in terms of apm @xmath40 magnitudes . the actual apm catalogue limit ( on isophotal size at a threshold surface brightness ) hardly impinges on our sample , as noted earlier , since the low surface brightness galaxies which we are biased against including because of their small isophotal sizes ( see e.g. davies et al . 1988 ) have too low a surface brightness to be successfully observed by 2df anyway . also noted in fig . 3 is the position of our absolute surface brightness limit at 23.7 b mag arcsec@xmath17 . the surface brightness effective limit is a fuzzy boundary between 23.2 and 23.7 b mag arcsec@xmath17 where the 2df success rate falls from @xmath580 per cent to zero . since we are dealing with fibre spectra , it is only the central region of the object which is observed . this can have a positive effect on the inclusion of low surface brightness objects in the case where they are nucleated ( i.e. de , ns ) since the nucleus may be bright enough to obtain a successful spectrum even if it sits in a very low surface brightness galaxy . for example , we noted in section 3.2 that j034100.8 - 354433 has a prominent nuclear bulge and that the values of @xmath41 and @xmath3 included in table 2 refer only to the outer disk component of this galaxy . in fact , according to the fcc , 12 of our sample are nucleated ( see table 1 ) . however , in our imaging data only 3 ( j034023.5 - 351636 , j034100.8 - 354433 and j034217.3 - 353226 ) show an observed central surface brightness significantly brighter than the extrapolated value listed in table 2 . for these 3 objects , we show on fig . 3 the observed as well as the extrapolated central surface brightness . this makes it clear how j034100.8 - 354433 in particular has been included in the sample despite having an extrapolated central surface brightness of 24.6 b mags arcsec@xmath17 . from fig . 3 it can be clearly seen that the fornax sample falls into two distinct areas of the size surface brightness plane . firstly , the ucds can be seen around @xmath3@xmath51 arcsec ( @xmath5100 pc at the distance of fornax ) between @xmath42=20.522.3 b mag arcsec@xmath17 . secondly , the rest of the ( non - compact ) dwarfs have @xmath433 arcsec . above @xmath3=3 arcsec ( @xmath5300 pc ) , the cluster sample essentially fills all of the available ( @xmath3,@xmath42 ) parameter space and therefore shows no correlation between the parameters . this non - correlation was claimed previously on purely photometric grounds by davies et al . ( 1988 ) and irwin et al . however , the area of parameter space between @xmath3=13 arcsec contains no galaxies . there appears to be a minimum scale size which a fornax de can have . whilst fig . 3 re - emphasises that the ucds appear to form a distinct population from the larger ` normal ' dwarfs , one can not extrapolate too much about the relative properties of the ucds and the non - compact objects from this figure alone . as noted above , due to the unresolved nature of the ucds in the photographic data , the plotted values of @xmath42 and @xmath3 are merely limits : the ucds may be smaller and have higher surface brightness than we measure . in fact a preliminary analysis of hst imaging of these objects ( drinkwater et al , in prep . ) suggests that the ucds actually have scale sizes as small as 0.1 arcsec and central surface brightnesses up to 3 mag brighter than measured from our photographic data . the bright central surface brightnesses of the ucds will make them detectable to fainter magnitudes than the non - compact objects and hence the relative number densities of the two populations are not clearly illustrated by fig . the nature of the ucds is currently a subject of some discussion ( drinkwater et al . 2000b ; hilker et al . 1999 ; mieske , hilker & infante 2002 ) , the debate focussing on whether the ucds form a distinct population of galaxies not linked to globular clusters or whether there is a smooth transition between both populations . we defer further consideration of this issue until presentation of our hst results . from fig . 3 , we can say little about the area with surface brightness fainter than 23.7 b@xmath44 . however , it is noteworthy that we have detected 6 objects with surface brightness between 23.223.7 b mag arcsec@xmath17 , despite the completeness of the sample as whole falling from 80 to 0 per cent in this range . we might reasonably expect further dwarfs to lie within this region . 4 shows the relationship in its alternative form , showing @xmath7 central surface brightness against absolute @xmath7 magnitude . the upper and lower @xmath12 magnitude selection limits are once again shown along with the absolute surface brightness limit at 23.7 b mag arcsec@xmath17 . the outlier fcss j034100.8 - 354433 is again seen apparently defying these limits at @[email protected] b mag arcsec@xmath17 , @[email protected] . also shown in fig . 4 is the locus for a galaxy with a pure exponential profile with a constant scale size of 3 arcsec ( @xmath5300 pc at the distance of fornax ) , assuming a circular image . there is a large area of parameter space to the upper right of this line within which the fcss would have detected galaxies , but the only galaxies within it are the ucds . however , as noted above , the ucds may have central surface brightness up to 3 mag brighter than derived from our photometry . if so , then the plotted area of parameter space to the upper right of the @xmath3=3 arcsec locus would be completely empty . below the @xmath3=3 arcsec locus the sample fills the entirety of available parameter space . it is notable that this limit in scale - size imposes on the sample the upper - bound of an apparent surface brightness magnitude ` correlation ' . however , to our limits of @xmath19 and @xmath42 there is no evidence for a lower bound to any such correlation . if such a surface brightness magnitude correlation does exist for the fornax dwarfs it must be fairly broad . for example , at @xmath19=15.0 the surface brightness distribution is @xmath42.5 mag in width . in fig . 5 we again present the fornax dwarf sample on a plot of central surface brightness against absolute luminosity . we also show on this figure the positions of local group galaxies taken from mateo ( 1998 ) . one needs to be wary of reading too much into this diagram . whilst we might reasonably expect that local group galaxies form the most complete sample of galaxies at faint luminosities and low surface brightness available , the local group data does not form a sample to strictly defined completeness limits . we are also , of course , comparing galaxies from two different environments ( i.e. the centre of the fornax cluster and the much more diffuse local group ) . the presence of several dis in the local group , where none was found in field 1 of the fcss , is testimony to this . in common with the fornax des , all of the local group des also have @xmath3@xmath4300 pc . most of the local group dis are also constrained by this scale size limit , apart from two outliers , both of which lie in the region of ( @xmath19,@xmath42 ) space occupied by the ucds . however , as noted above , the ucds have central surface brightnesses up to 3 mag brighter than derived from our photometry , in which case the local group dis do not in reality lie in a similar region of ( @xmath19,@xmath42 ) space . the dsphs of the local group do appear to show a correlation between surface brightness and absolute magnitude , although these objects are generally considered to form a distinct population from the brighter des . many of the local group dsphs have scale - sizes less than 300 pc . our fcss data are not sensitive enough to detect any similar objects in fornax . we also note on fig . 5 the region at bright absolute luminosities ( @xmath4516.0 ) within which the galaxies from the faint end of cross et al.s ( 2001 ) 2dfgrs sample lie . this is one of the largest samples of bright galaxies yet assembled and formed the basis of cross et al.s determination of the bbd ( see also cross & driver 2002 ) . it should be noted , however , that cross et al.s sample is not differentiated by morphological type ( it includes ellipticals , bulge dominated spirals , disc dominated spirals and irregulars ) . in order to compare the cross et al . sample with our data , we have translated their effective surface brightnesses to central surface brightness by assuming a shift of @xmath51.8 mag appropriate for an exponential profile . within the limits of their sample , cross et al . found a correlation between surface brightness and absolute magnitude such that @xmath46 scales as 0.42@xmath19 . the relationship is around 2.5 mags wide in surface brightness at each absolute magnitude . it is clear from the fornax and local group data that this relationship does not extend to fainter magnitudes . as noted , the lower limit to the scale size of the des ( at @xmath3@xmath5300 pc ) creates an apparent upper - bound to a surface brightness - magnitude ` correlation ' . obviously the @xmath46 of this upper - bound scales as @xmath19 , i.e. much steeper than the cross et al . there is , though , no evidence that there is a low surface brightness bound to any such correlation for dwarfs . the correlation between surface brightness and absolute magnitude for local group dsphs is also much steeper than that from cross et al.s bright galaxy sample . a prime aim of the fcss is to test the reliability of using optical morphology as a basis for determining cluster membership . firstly , it is worth noting that only @xmath550 per cent of our complete cluster sample galaxies actually have @[email protected] b mag arcsec@xmath17 , the criterion often used to define a low surface brightness galaxy . in other words , though all our objects are dwarfs , many of them extreme dwarfs , in terms of luminosity , half of them are _ not _ low surface brightness galaxies . so , any selection based on the assumption that all cluster dwarfs are going to be lsbgs will be flawed . of course , many other lsb dwarfs which are really present will be too faint for us to obtain a spectroscopic redshift , so the true fraction of non - lsbg dwarfs will be considerably smaller than 50 per cent , but the important point as regards morphological ( pre)selection is that such dwarfs do exist . in addition we find 15 background objects with @[email protected] b mag arcsec@xmath17 which would morphologically be likely to be assigned to the cluster . our best point of comparison is with ferguson s ( 1989 ) fcc . ferguson catalogued 2678 objects from photographic plates of an area 40 deg@xmath1 centered on the fornax cluster . he gave each object one of five classifications : ` class 1 - definite cluster member ' ; ` class 2 - probable cluster member ' ; ` class 3 - possible cluster member ' ; ` class 4 - likely background object ' ; and ` class 5 - definite background object ' . of our complete sample of 24 fornax cluster dwarfs , a total of 17 ( 71 per cent of our cluster sample ) were ascribed by ferguson to the cluster : 12 as class 1 ; 4 as class 2 ; and 1 as class 3 . one of our cluster sample ( fcss j034159.4 - 352053 ) was described as class 5 ( ` definite background object ' ) . this object has the smallest scale size ( 3.23 arcsec ) of any of the non - ucds in our sample , although its surface brightness and absolute magnitude are not remarkable compared to the rest of the cluster sample . another of our sample ( fcss j033816.7 - 353027 ) was listed by ferguson as a ` class 4 ' ( ` possible background object ' ) . this does not have a remarkable surface brightness or absolute magnitude although it has one of the smaller scale lengths of the non - ucd objects ( 5.43 arcsec ) . ferguson did not include any of the ucds within the fcc either as background galaxies or cluster members . presumably these were so compact as to be considered foreground stars . in addition , within the fcss survey of field 1 we found that a further 5 objects which ferguson had ascribed to the cluster ( 1 as class 1 ; 4 as class 2 ) are actually background objects . so within our field and our selection limits , ferguson ascribed 22 objects to the cluster . fcss has found that 5 of these objects ( i.e. 23 per cent of ferguson s cluster objects ) are actually background objects . a further 9 objects which were listed by ferguson as class 3 ( ` possible cluster member ' ) but included in his table of background galaxies , are confirmed by fcss as background objects . to summarise : ferguson did not include any of the fcss s ucds within the fcc ( either as cluster or background objects ) , presumably assuming them to be stars . ferguson did include within the fcc all 19 of the non - compact cluster dwarfs from the fcss : however , he mis - classified 2 of these as ` background ' objects . hence , the fcc contains 89 per cent of the non - compact dwarfs within the fcss cluster sample ( but only 71 per cent if one considers all dwarfs in the fcss ) . however , the fcc contains a further 5 objects as ` cluster ' members which have been shown by the fcss to be background objects : i.e. within the fcss s field and selection limits , 23 per cent of those objects described as cluster members by the fcc are shown by the fcss to be background objects . this clearly illustrates that attempts to determine cluster membership solely on the basis of observed morphology can produce significant errors : firstly because many compact galaxies will be mis - classified as stars ; and secondly because many non - compact background objects will be mis - classified as cluster members and a smaller number of cluster members will be mis - classified as background objects . we have obtained spectra for a morphologically unbiased set of cluster dwarf galaxies as part of the fornax cluster spectroscopic survey . the present paper provides the photometric study of these confirmed cluster dwarfs . we find that , despite their low luminosities , not all the dwarfs also have low surface brightness . there is therefore a wide spread of parameters in the space of surface brightness versus luminosity or scale size . the only correlation evident in these plots is the one in the luminosity surface brightness plane caused by the absence of ` normal ' dwarfs with scale sizes below about 300 pc . even then , a separate class of ultra - compact dwarfs fills some of this otherwise empty area of parameter space . morphological selection of cluster dwarfs solely on the grounds of their low surface brightness may therefore overlook a significant number of dwarf galaxies which are virtually impossible to discriminate from background , larger galaxies on appearance alone . the data presented within this paper result from the completion of only one of the intended 4 fcss fields . we now have full data for the second field and are in the process of reducing these . the completion of all 4 fields will not only improve the statistics of our results but will also enable us to study the distribution of dwarf galaxies of different morphologies ( i.e. de , di , ucd ) as a function of position within the cluster . such studies will be assisted by the deep ccd g , r , i , z imaging of the cluster region which we have now obtained using the 4.0-m blanco telescope at the cerro tololo inter - american observatory . these data are currently being reduced . important scientific advances could also potentially be made by improving the sensitivity of our spectroscopy to fainter luminosities and/or surface brightnesses . this can be clearly seen by studying figs . 4 and 5 . using the ccd data we should be able to define a sample to a considerably fainter @xmath7 magnitude limit . the standard 2df set - up would be adequate to obtain spectra of the higher surface brightness objects in such a sample . this would include many potential ucds . however , even bigger advances will require us to obtain spectra for objects at fainter surface brightness limits than the present limit of 23.7 b mag arcsec@xmath17 . this possibility now exists with the commissioning of multi - object spectrographs on the new generation of 8-m telescopes . the number of cluster objects found in field 1 is dwarfed by the number of confirmed background objects ( 1175 ) . such a sample is potentially of great value to studies of the bbd of field galaxies . although the present magnitude limit is only slightly lower than that of the 2dfgrs used by cross et al . ( 2001 ) for their determination of the bbd , the fcss background sample has the advantage that the fcss sources were not pre - selected as galaxies . hence , our sample will contain a fairer representation of compact objects in the background than cross et al.s ( see paper ii ) . for the lower surface brightness objects in our sample we have also used longer exposure times than the 2dfgrs , so fcss extends to fainter surface brightness than 2dfgrs . however , if the fcss sample could be pushed to even fainter luminosity and surface brightness limits then , not only will this enable us to study the cluster bbd better , but the resulting sample will be ideal for studying the field bbd to fainter surface brightness limits than yet obtained . for example , cross et al.s correlation between luminosity and surface brightness has only been tested between [email protected] and [email protected] . at present the fscc is @xmath580 per cent complete to a surface brightness about 0.5 mag fainter than this . further improvements in the sensitivity of the spectroscopy will enable us to sample whole new regions of ( @xmath19,@xmath42 ) space . this project would not have been possible without the superb 2df facility provided by the aao and the generous allocations of observing time from patt and atac . we wish to thank the other members of the fcss team : jonathon davies , elaine sadler and quentin parker . jbj , pjb and jhd acknowledge support from the uk particle physics and astronomy research council , and mjd that from the australian research council . part of this work was done at the institute of geophysics and planetary physics , under the auspices of the u.s . department of energy by lawrence livermore national laboratory under contract no . w-7405-eng-48 . this research has made use of the nasa / ipac extragalactic database ( ned ) which is operated by the jet propulsion laboratory , caltech , under agreement with the national aeronautics and space administration . abell g.o . , corwin h.g . ( jr ) , olowin r.p . , 1989 , apjs , 70 , 1 arp , h. , 1965 , apj , 142 , 402 bertin e. , arnouts s. , 1996 , a&as , 117 , 393 binggeli b. , 1994 , in meylan g. , prigniel p. , eds , proc eso / ohp workshop , dwarf galaxies . eso , garching , p.13 binggeli b. , sandage a. , tammann g.a . , 1985 , aj , 90 , 1681 blanton m.r . et al . , 2001 , aj , 121 , 2358 blair m. , gilmore g. , 1982 , pasp , 94 , 742 bothun g.d . , impey c.d . , malin d.f . , mould j.r . , 1987 , aj , 94 , 23 boyce p.j . , phillipps s. , 1995 , a&a , 296 , 26 boyce p.j . , phillipps s. , jones j.b . , driver s.p . , smith r.m . , couch w.j . , 2001 , mnras , 328 , 277 bromley b.c . , press w.h . , lin h. , kirshner r.p . , 1998 , apj , 505 , 25 bunclark p , irwin m.j . , 1984 , in capaccioli , m. , ed . iau colloq . 78 , astronomy with schmidt - type telescopes . reidel , dordrecht , p. 147 caldwell , n. , 1987 , aj , 94 , 1116 choliniewski j. , 1985 , mnras , 214 , 197 colless m.m . et al . , 2001 , mnras , 328 , 1039 cross n.m . , driver s.p . , 2002 , mnras , 329 , 579 cross n.m . et al . , 2001 , mnras , 324 , 825 dalcanton j.j . , 1998 , apj , 495 , 251 davies j.i . , phillipps s. , cawson m.g.m . , disney m.j . , kibblewhite e.j . , 1988 , mnras , 232 , 239 davies j.i . , phillipps s. , disney m.j . , 1990 , mnras , 244 , 385 deady j.h . , 2002 , phd thesis , univ bristol de jong r. , lacey c. , 2000 , apj , 545 , 781 de propris r. , pritchet c.j . , harris w.e . , mcclure r.d . , 1995 , apj , 450 , 534 de vaucouleurs g. , 1959 , handbk . phys . , 53 , 120 disney m.j . , 1976 , nature , 263 , 573 drinkwater m.j . , gregg m.d . , 1998 mnras . , 296 , l15 drinkwater m.j . , phillipps s. , gregg m.d . , parker q.a . , smith r.m . , davies j.i . , jones j.b . , sadler e.m . , 1999 , apj , 511 , l97 ( paper ii ) drinkwater m.j . et al . , 2000a , a&a , 355 , 900 ( paper i ) drinkwater m.j . , jones j.b . , gregg m.d . , phillipps s. , 2000b , pub . australia , 17 , 227 ( paper iii ) drinkwater m.j . , gregg m.d . , holman b.a . , brown m. , 2001a , mnras , 326 , 1076 drinkwater m.j . , gregg m.d . , colless m.m . , 2001b , apj , 548 , 139 driver s.p . , 1999 , apj , 526 , l69 driver s.p . , phillipps s. , davies j.i . , morgan i. , disney m.j . , 1994 , mnras , 268 , 393 driver s.p . , couch w.j . , phillipps s. , smith r.m . , 1998a , mnras , 301 , 357 driver s.p , couch w.j . , phillipps s. , 1998b , mnras , 301 , 369 ferguson h.c . , 1989 , aj , 98 , 367 ( fcc ) ferguson h.c . , sandage a. , 1988 , aj , 96 , 1520 folkes s. et . al . , 1999 , mnras , 308 , 459 garilli b. , maccagni d. , andreon s. , 1999 , a&a , 342 , 408 hambly n.c . , irwin m.j . , macgillivray h.t . , 2001 , mnras , 326 , 1295 hilker m. , infante l. , vieira g. , kissler - patig m. , richtler t. , 1999 , a&as , 134 , 75 hodge p.w . , 1960 , pasa , 72 , 188 hodge p.w . , pyper d.m . , webb c.j . , 1965 , aj , 70 , 559 impey c.d . , bothun g.d . , 1997 , ara&a , 35 , 267 irwin m.j . , davies j.i . , disney m.j . , phillipps s. , 1990 , mnras , 245 , 289 irwin m.j . , maddox s. , mcmahon r. , 1994 , spectrum , 2 , 14 jacoby g.h . , hunter d.a . , christian c.a . , 1984 , apjs , 56 , 257 kambas a. , davies j.i . , smith r.m . , bianchi s. , haynes j.a . , 2000 , aj , 120 , 1316 kormendy j. , 1985 , apj , 295 , 73 kron r.g . , 1978 , phd thesis , univ . california , berkeley kurtz m.j . , mink d.j . , 1998 , pasp , 110 , 934 lin h. , yee h.k.c . , carlberg r.g . , ellingson e. , 1997 , apj , 475 , 494 lobo c. , biviano a. , durret f. , gerbal d. , le fevre o. , mazure , a. , slezek e. , 1997 , a&a , 317 , 385 loveday j. , peterson b.a . , efstathiou g. , maddox s.j . , 1992 , apj , 390 , 338 marzke r.o . , da costa l.n . , pellegrini p.s . , willmer c.n.a . , geller m.j . , 1997 , apj , 503 , 617 mayer l. , governato f. , colpi m. , moore b. , quinn t. , wadsley j. , stadel j. , lake g. , 2001 , apj , 559 , 754 mateo m. , 1998 , ar&a , 36 , 435 metcalfe n. , shanks t. , fong r. , roche n.,1995 , mnras , 273 , 257 meyer m.j . , drinkwater m.j . , phillipps s. , couch w.j . , 2001 , mnras , 324 , 343 ( paper iv ) miekse s. , hilker m. , infante l. , 2002 , a&a , 383 , 823 moore b. , lake g. , katz n. , 1998 , apj , 495 , 139 morshidi - esslinger z. , davies j. , smith r. , 1999 , mnras , 304 , 311 muriel h. , valotto c.a . , lambas d.g . , 1998 , apj , 506 , 540 phillipps s. , 1997 , in kontizas e. , konizas m. , morgan d.h . , vettolani g.p . , eds , proc working group iau commission 9 , wide - field imaging . kluwer , dordrecht , p. 281 phillipps s. , parker q.a . , 1993 , mnras , 265 , 385 phillipps s. , driver s.p . , 1995 , mnras , 274 , 832 phillipps s. , disney m.j . , kibblewhite e.j . , cawson m.g.m . , 1987 , mnras , 229 , 505 phillipps s. , davies j.i . , disney m.j . , 1988 , mnras , 233 , 485 phillipps s. , davies j.i . , disney m.j . , 1990 , mnras , 242 , 235 phillipps s. , drinkwater m.j . , gregg m.d . , jones j.b . , 2001 , apj , 560 , 201 ( paper v ) sandage a. , binggeli b. , tammann g. , 1985 , aj , 90 , 385 sersic j.l . , 1968 , atlas de galaxies australes , observatorio astronomico , cordoba sodre l. ( jr ) , lahav o. , 1993 , mnras , 260 , 285 smith r.m . , driver s.p . , phillipps s. , 1997 , mnras , 287 , 415 taylor k. , canon r.d . , parker q.a . , 1998 , in mclean b.j . , golombek d.a . , hayes j.j.e . , payne h.e . , eds , proc . iaus 179 , new horizons from multi - wavelength sky surveys . kluwer , dordrecht , p.135 tonry j.l . , davis m. , 1979 , aj , 84 , 1511 trentham n. , 1997a , mnras , 286 , 133 trentham n. , 1997b , mnras , 290 , 334 trentham n. , 1998 , mnras , 295 , 360 trentham n. , hodgkin s. , 2002 , mnras , 333 , 423 valotto c.a . , nicotra m.a . , muriel h. , lambas d.g . , 1997 , apj , 479 , 90 van der kruit p. , 1987 , a&a , 173 , 59 wilson g. , smail i. , ellis r.s . , couch w.j . , 1997 , mnras , 284 , 915 zucca e. et al . , 1997 , a&a , 326 , 477	the fornax cluster spectroscopic survey ( fcss ) project utilises the two - degree field ( 2df ) multi - object spectrograph on the anglo - australian telescope . its aim is to obtain spectra for a complete sample of all 14000 objects with @xmath0 _ irrespective of their morphology _ in a 12 deg@xmath1 area centred on the fornax cluster . a sample of 24 fornax cluster members has been identified from the first 2df field ( 3.1 deg@xmath1 in area ) to be completed . this is the first complete sample of cluster objects of known distance with well defined selection limits . 19 of the galaxies ( with [email protected] ) appear to be conventional dwarf elliptical ( de ) or dwarf s0 ( ds0 ) galaxies . the other 5 objects ( with [email protected] ) are those galaxies which we described in drinkwater et al . ( 2000b ) and labelled ` ultra - compact dwarfs ' ( ucds ) . a major result is that the conventional dwarfs all have scale - sizes @xmath3@xmath43 arcsec ( @xmath5300 pc ) . this apparent minimum scale size implies an equivalent minimum luminosity for a dwarf of a given surface brightness . this produces a limit on their distribution in the magnitude - surface brightness plane , such that we do not observe des with high surface - brightnesses but faint absolute magnitudes . above this observed minimum scale - size of 3 arcsec , the des and ds0s fill the whole area of the magnitude surface brightness plane sampled by our selection limits . the observed correlation between magnitude and surface brightness noted by several recent studies of brighter galaxies is not seen with our fainter cluster sample . a comparison of our results with the fornax cluster catalog ( fcc ) of ferguson illustrates that attempts to determine cluster membership solely on the basis of observed morphology can produce significant errors . the fcc identified 17 of the 24 fcss sample ( i.e 71 per cent ) as being ` cluster ' members , in particular missing all 5 of the ucds . the fcc also suffers from significant contamination : within the fcss s field and selection limits , 23 per cent of those objects described as cluster members by the fcc are shown by the fcss to be background objects . = = = = = = = = # 1 # 1 # 1 # 1 @mathgroup@group @mathgroup@normal@groupeurmn @mathgroup@bold@groupeurbn @mathgroup@group @mathgroup@normal@groupmsamn @mathgroup@bold@groupmsamn = `` 019 = ' ' 016 = `` 040 = ' ' 336 = " 33e = = = = = = = = # 1 # 1 # 1 # 1 = = = = = = = = [ firstpage ] galaxies : luminosity function galaxies : statistics galaxies : clusters : individual : fornax surveys techniques : spectroscopic
the anisotropy field in the 2-dimensional distribution of iras sources on angular scales of @xmath8 was recently used to derive a characteristic length scale of about @xmath9mpc ( fabbri & natale 1993 ; the hubble constant @xmath10is here @xmath11 km s@xmath12mpc@xmath12 ) , significantly larger than the dipole - dominating scale ( villumsen & strauss 1987 ; clowes et al . 1987 ) and comparable to the distance @xmath13 mpc at which the galaxy - galaxy correlation function @xmath5 becomes negative ( guzzo et al . 1991 ; see also davis & peebles 1983 ) . torres et al . ( 1994 ) have shown that the fourier transform of @xmath5 , suitably modeled to reproduce the data of guzzo et al . , should induce a peak in the spectral function @xmath14of the inhomogeneities at a wavelength @xmath15 ; further , they find that best fitting on cobe data , involving the angular correlation function of the cosmic background radiation , provides @xmath16 @xmath17 mpc . we are thereby lead to suspect that the characteristic length in iras may be close to a peak of @xmath3 , although not necessarily to its absolute maximum . this interpretation , even in its weakest form , seems to be not consistent with several works inferring @xmath3 from the 3-dimensional distributions of visible and ir galaxies and clusters ( vogeley et al . 1992 ; fisher et al . 1993 ; einasto et al . 1993 ; jing & valdarnini , 1993 ) . these works , based on fourier transforms of redshift - space distributions , provided spectra which , although not completely consistent with each other , steadily increase up to or beyond @xmath18 mpc : when some peak is detected , it is located at large wavelengths , in the range @xmath19 mpc . however , bahcall et al . ( 1993 ) argue that large - scale velocities induced by the inhomogeneity field itself , not properly taken into account in the above works , may lead to largely overestimate the large - scale portion of the spectrum in real ( rather than redshift ) space . corrections are strongly model dependent , and for some of the cdm models considered by bahcall et al . they shift the peak wavelengths in the iras and cfa spectra down to @xmath20 mpc . is is thereby important in our view to exploit alternative techniques that are not affected by the problem of large scale velocities . in this paper we reconsider the question of the significance and reliability of the preferred scale that we identified in version i of the iras point source catalog . we use a larger sample including 12511 sources derived from version ii of the catalog , and we perform three statistical tests based on finite - size bins . these include the shapes of the functions ( i ) @xmath21 , the rms fluctuation in the source number density vs. angular scale @xmath22 in the range @xmath23 , ( ii ) @xmath24 , the autocorrelation function of the number density at@xmath25 , and ( iii ) @xmath26 , the number of local maxima in the source distribution vs. gaussian beam dispersion @xmath27(in the range @xmath28 ) and peak - height threshold @xmath29 . the experimental data are tested against a large number ( about 2000 ) of theoretical models . first we model the inhomogeneity field by isotropically stochastic perturbations endowed with a single wavelength @xmath30 , and search for the best values of @xmath30 , the cosmological density parameter @xmath31 , and the product @xmath32 ( the biasing factor of iras sources times the local velocity ) . although the present sample allows us detecting some dependency of @xmath33 on the chosen angular range ( in particular using the statistics on @xmath34 ) , nevertheless we can identify a preferred range around @xmath35mpc , irrespective of @xmath31 which can not be determined by the present tests . we then examine continuous spectra modeling both hot and cold matter with the usual requirement that@xmath36 at large wavelengths . we find that the hdm model in the @xmath37 cosmology gives excellent fits for all of our tests , except for one data point of @xmath38 at the largest scale @xmath39 . the hdm peak wavelength @xmath40 is only slightly larger than @xmath41found for single - scale models . individual tests can give best values as high as @xmath42 mpc with large error bars , but two joint tests combining the peak number data with rms fluctuation and correlation function data , respectively , give @xmath43 mpc and @xmath44 mpc . although we can not derive stringent limits on @xmath32 from individual hdm tests , when we require to simultaneously satisfy all of our tests we get a preferred value of @xmath45 km / s ( in close agreement with results that we find from single - scale models for @xmath37 ) . basically the same picture arises for other sharply - peaked spectral forms , for which the only important parameter is the location @xmath46 of the peak of@xmath47 . thus similar results for @xmath48are found for simple two - power spectra with slopes around @xmath49 at large @xmath50 . the investigation of cdm models on the other hand provides a quite different picture . for individual tests we find best - fit values of the shape parameter @xmath51 in one of the ranges @xmath52 or @xmath53 . low - density models are favoured by several studies ( e.g. , efstathiou et al . now for @xmath54 the peak wavelength is @xmath55 mpc ; in fact , single - scale models are quite unable to represent so broad spectra as those implied by cdm models . however , the quality of cdm fits is less satisfactory than for hdm models , with the exception of the peak number statistics . in addition , no range of @xmath51 is completely satisfactory for all of our tests ; this latter fact recalls the results of scharf and lahav ( 1993 ) . it is well known that pure cdm models do not generally provide excellent fits to available data , a fact which motivates studies involving mixed matter ( davis et al . 1992 , taylor & rowan - roinson 1992 ) or a cosmological constant ( efstathiou et al . however , the comparison of hdm and cdm fits in our work ( sect . 4 ) is rather impressive . the spectral window functions that can be suitably defined for @xmath21 and @xmath24 show that such tests weight small scales less than other 2-dimensional tests from other authors ( in particular , scharf and lahav 1993 ) . this provides support to the idea that the preferred scale arising from sharply peaked spectra should correspond to something of physical . it is also clear that such a feature should be regarded as generated by the behaviour of @xmath5 around @xmath56 mpc measured by guzzo et al . ( 1991 ) . since we do not directly construct an explicit spectral form from data , we do not claim that our tests on the distribution of iras sources necessarily imply an absolute maximum of @xmath14at scale @xmath46 , nor can we claim that hdm models satisfy the bulk of available astronomical data . however , our results make us believe that the shape of @xmath3 below @xmath18 mpc should not be so simple and featureless as generally believed , and that a positive slope up to or beyond @xmath57 mpc is probably incorrect . according to recent measurements of the angular correlation function in the apm survey the spectrum rises above a power law around @xmath58 mpc and seems to be peaked at @xmath18 mpc ( baugh & efstathiou 1993 ) , a scale smaller than declared by most authors on the basis of redshift surveys . in our view the problem of the detailed shape of @xmath3 is not so clearly solved and deserves further investigations . our models assume that the large - scale inhomogeneities are described by a isotropic , gaussian field of perturbations . the theory of source - count anisotropies has been developed ( fabbri 1988 , 1992 ) for a friedmann background with an arbitrary density parameter @xmath31 , perturbed by a cosmic density - wave spectrum @xmath3 such that @xmath59 here the dimensionless wavenumber @xmath50 is connected to the physical wavelength by @xmath60 with @xmath61 for @xmath62 and @xmath63 . all the quantities that will be evaluated theoretically are derived from the squared coefficients of a harmonic expansion of the source - count anisotropy field @xmath64 . more precisely , we consider the expectation values @xmath65 where we sum up the cosmic - structure contribution@xmath66 given by @xmath67 and the noise term @xmath68 originated by discrete sources @xmath69 in eq . ( [ cosmical ] ) @xmath70 denotes the sign of space curvature and the functions @xmath71 are suitable coefficients introduced by fabbri ( 1988 ) . we could also define window functions @xmath72 as in scharf and lahav ( 1993 ) by setting @xmath73 , with @xmath74the biasing factor of iras sources . the analysis of @xmath71 ( fabbri 1992 ) shows that the source properties affect the calculation through the biasing factor , the luminosity function , the number evolution factor and , to a very small extent , the spectral index . we shall provide explicit results for the exponential - logarithmic luminosity function of saunders et al . ( 1990 ) , which best fits data with a number evolution factor @xmath75 . however , using the double - power functions of lawrence et al . ( 1986 ) or villumsen and strauss ( 1987 ) does not change the conclusions of sect . 4 ( adopting no source - number evolution of course , since they were obtained with this assumption ) . since in any case the assumed source distribution extends farther than probed in redshift surveys ( strauss et al . 1990 ) , we checked the influence of a cutoff placed at some @xmath76 . figure 1 displays some typical results for @xmath77 , showing that we get no significant differences provided @xmath78 . a sharp cutoff at redshifts @xmath79 would decrease the amplitude of high order harmonics relative to low order ones . thus we just assume that sources still exist at a distance ( along the light cone ) equal to roughly three times the median depth of psc . we do not regard this as an onerous requirement : in practice we only assume that the median depth of the catalog is determined by the source luminosity function and the lower flux limit . the global properties of the universe are involved through @xmath31 . further , it was shown that , because of the presence of the discrete source term in eq . ( [ al ] ) , the shape of the harmonic spectrum of anisotropies depends both on the shape of @xmath3 and on the product @xmath32 , with @xmath80the local velocity with respect to the cosmic background radiation . since @xmath80is generated by the perturbation field itself , it is also a measure of the absolute normalization of @xmath3 . ( this implies that the shape of @xmath77 vs. @xmath81should be sufficient to fix its own normalization . ) however , @xmath82 may be affected by nonlinear disturbances . on the other hand , the absolute normalization of @xmath77 is affected by high redshift sources more than the shape of the harmonic spectrum ( fabbri 1988 ) , and it might introduce a systematic error on the normalization of @xmath3 . we thereby prefer to fit @xmath83and the normalization factor of @xmath3 as two independent parameters . we then check a posteriori whether we obtain consistent results for @xmath84 , the biased fluctuation in @xmath85mpc spheres , computed from these two parameters when they are both available . for the peak number test , only the first method ( i.e. , derivation from @xmath32 ) can be used . as to the shape of @xmath3 , here we consider both single - scale models and continuous spectra . for the reasons outlined above , single - scale models only depend on three important parameters , i.e. @xmath30 , @xmath31 and @xmath32 . for a full investigation of their influence we generated a regularly spaced lattice of 900 models for the adopted optimal set of source parameters [ i.e. , using the luminosity function of saunders et al . ( 1990 ) and a mean spectral index of @xmath86 ; including also the luminosity functions of lawrence et al . ( 1986 ) and villumsen and strauss ( 1987 ) , we investigated about 1500 single - scale models . continuous spectra , which were investigated only in conjunction with the exponential - logarithmic luminosity function , were modeled by simple analytic forms for hot and cold matter . hdm models are described by @xmath87 , \ ] ] with @xmath88 the cutoff wavenumber . the peak wavelength is given by @xmath89 ^{-1}$ ] . the number of models computed with the spectrum in eq . ( [ hot ] ) was about 200 , and we limited ourselves to the critical density cosmology . we also considered about 200 models with power - law behaviours at large @xmath50 . we analyzed about 200 cdm models with @xmath90 ^{-2},\ ] ] where @xmath91 . in this case the relevant parameter is @xmath51 , that we let vary in the range @xmath92 . the corresponding peak wavelength is given by @xmath93 mpc . all of the above continuous spectra are so adjusted as to reproduce the standard harrison - zeldovich slope at large scales and exhibit a maximum . for sharply peaked spectra , which include hdm ( but not cdm ) models , only the location of this maximum is really important . thus the small - scale behaviour in eq . ( [ hot ] ) , theoretically predicted by simple modeling of dissipative processes in hot matter , can be easily changed without affecting our conclusions . ( this was tested investigating the sharply - peaked two - power spectra . ) although one can argue that eq . ( [ hot ] ) can not reproduce such astronomical data as the galaxy correlation function on scales of tens of megaparsec ( see e.g. guzzo et al . 1991 ) , it is the simplest one can imagine and turns out to pass our tests quite well . also , for the best fitted parameters arising from the tests it gives reasonable values of @xmath94 . for each model we calculated 300 harmonic amplitudes , in order to provide reliable results down to small angular scales . we then derived both the squared fluctuation in the source number per unit solid angle ( for two - beam experiments ) and the angular correlation function smeared over finite beamwidths , which are respectively given by the equations @xmath95\ ] ] and @xmath96 with @xmath97 the angular separation of beam centers ( beamthrow ) , @xmath98 the effective beamwidth , and @xmath99 the legendre polynomials . the coefficients @xmath100 are affected by the beam angular response and are well known for both gaussian and sharp - edge beams . in this paper any theoretical results based on eqs . ( [ rms ] ) and ( [ c ] ) refer to sharp - edge circular beams , for which @xmath27denotes the angular diameter . equation ( [ rms ] ) was used to evaluate @xmath101 with @xmath102 , while using ( [ c ] ) @xmath27is kept fixed . in both cases we checked that calculations are accurate down to @xmath103 , while at @xmath104 we would need more harmonic amplitudes . although the similarity of eqs . ( [ rms ] ) and ( [ c ] ) seems to imply that these two tests are quite equivalent , this is not the case . in the limiting case @xmath105 , the discrete - noise terms give a vanishing contribution to @xmath24 because @xmath106 for @xmath107 , and for the reasons explained in fabbri ( 1992 ) any dependence on the parameter @xmath32 disappears . for finite values of @xmath108 this is no more strictly true , but the test based on the shape of @xmath109 remains quite insensitive to @xmath32 . a question which is very important for the main issue of our paper is , how length scales are weighted in eqs . ( [ rms ] ) and ( [ c ] ) . since @xmath110 and @xmath24 depend linearly on @xmath77 , which in its turn is a linear function of @xmath3 , it is possible to define two window functions @xmath111 and @xmath112such that the cosmic - structure contributions to the above quantities are expressed as @xmath113 . ( 8) of scharf and lahav ( 1993 ) . ] from eqs . ( [ cosmical ] ) , ( [ rms ] ) and ( [ c ] ) we get @xmath114\end{aligned}\ ] ] and @xmath115 figure 2 reports some results for @xmath116 and @xmath117 in critical - density models which illustrate general features . ( for consistency with beamwidths used in our tests , we set @xmath102 and @xmath118 , respectively , for such calculations . ) it appears that small scales are weighted in two of our tests less than in the work of scharf and lahav ( 1993 ) using the 2-jy iras survey . this is due to the statistical bias of the iras _ redshift _ survey including 2650 objects . generally speaking , we expect a large , flux - limited sample to include more far - away sources than a redshift survey . both @xmath116 and @xmath119 increase monotonically with wavelength . the actual contribution of a length scale to the signal depends on the spectrum and can be estimated as @xmath120 . for instance , for rms fluctuations in the angular range @xmath121 the signals reach their maxima at @xmath122 mpc for a hdm spectrum with @xmath123 mpc , and at @xmath124 mpc for a cdm spectrum with @xmath125 . for the same models , the signals fall to 10% of their maxima at @xmath126 mpc , i.e. around the boundary of the linear region , and decrease very rapidly at smaller scales . thus nonlinear disturbances are small . we further consider the number of local maxima in the density of sources over the sky . we define @xmath26 as the number of peaks higher than a threshold @xmath29 , counted in the source distribution after smoothing by a gaussian beam with dispersion @xmath127 . following bond and efsthatiou ( 1987 ) we take the threshold height @xmath128 in units of @xmath129 . the peak statistics is described by @xmath130 @xmath131 \nonumber \\ & & \left [ \sum_\ell \left ( \ell -1\right ) \ell \left ( \ell + 1\right ) \left ( \ell + 2\right ) a_\ell ^2f_\ell ^2\right ] ^{-1},\end{aligned}\ ] ] and@xmath132 ^2\left [ \sum_\ell a_\ell ^2f_\ell ^2\right ] ^{-1 } \nonumber \\ & & \left [ \sum_\ell \left ( \ell -1\right ) \ell \left ( \ell + 1\right ) \left ( \ell + 2\right ) a_\ell ^2f_\ell ^2\right ] ^{-1 } , \label{gamma}\end{aligned}\ ] ] where the function@xmath133 was calculated by fabbri ( 1992 ) from the theory of bond & efsthatiou ( 1987 ) , and @xmath134now pertains to gaussian beams . inspection of eqs . ( [ peak]-[gamma ] ) shows a highly non - linear dependence on @xmath77 , so that it is not possible to define a window function . however , the strong weight given to higher order harmonics in eqs . ( [ teta ] ) and ( [ gamma ] ) proves that perturbations with small @xmath135must have a strong bearing on the statistics for small beamwidths , so that we expect to be able to provide a lower limit to a dominating scale by this test . taking advantage of 300 harmonic amplitudes allowed us to evaluate@xmath136 accurately down to @xmath137 . since now @xmath138denotes a beam dispersion rather than a diameter , the investigation of small scales is limited somewhat more strongly than for other quantities . we used a galaxy sample based on the iras point source catalog , version ii , including sources at 60 @xmath0 m in the flux interval 0.6 - 20 jy . the galaxies were selected basically using the same criteria as in strauss et al . ( 1990 ) , but we also set the lower flux limit to 0.6 jy , the completeness limit of the iras survey ( yahil et al . 1986 ) , and extended the region of avoidance around the galactic plane to @xmath139 . to reduce the residual contamination we also excluded other sky regions after source binning ; see below . this sample differs from the one of fabbri and natale ( 1993 ) , not only for the higher number of sources but also for a different calibration in version ii of the catalog , leading to smaller spectral indexes of the sources . the average spectral index calculated from fluxes at 60 and 25 @xmath140 m is @xmath141 , and the one from fluxes at 60 and 100 @xmath0 m is @xmath142 . for the theoretical modelling in sect . 2 we adopted the mean value @xmath143 , but the results are quite insensitive to this choice . sources were collected in bins spanning equal intervals @xmath144 ( i.e. , in galactic longitude ) and @xmath145 ( in the sine of galactic latitude ) . we considered three resolution levels leading to three maps with 128@xmath14664 , 256@xmath146128 and 512@xmath146256 bins , respectively , over the entire celestial sphere . an effective angular diameter @xmath147 such that @xmath148 @xmath149 can be defined for the bins . for the above resolution levels we have @xmath150 , @xmath151 and @xmath152 , respectively . for the purpose of our analysis clean bins with a vanishing number of sources are quite significant , but we must exclude regions contaminated by galactic emission or not sufficiently observed during the survey . hence to each source map we applied a mask excluding bins which are ( i ) not observed at least twice in the survey , or ( ii ) contained within @xmath139 around the galactic plane , or ( iii ) confused due to large number of sources ( such as the orion complex , taurus and ophiuchus star forming regions , the large and small magellanic clouds ) , or ( iv ) contaminated by cirrus emission . figure 3 gives the map of our full sample of 12511 sources , obtained by using the above mask . by means of this procedure , we constructed 3 matrices with indexes ( @xmath153 , @xmath154 ) spanning galactic coordinates , whose elements contained either the number of sources in the bins per unit solid angle @xmath155 or a masking flag . these matrices will be referred to as bin maps in the following . in the first part of our analysis we compute all the possible differences between the values of @xmath155 in adjacent windows and then the rms value . this amounts to calculating the quantity @xmath156 of eq . ( [ rms ] ) for sharp - edge beams , except for the fact that observation windows are not circular . in order to minimize errors introduced by non - circular shapes , we maintain beam self - similarity in the generation of larger beamwidths : starting from the above bin sizes we merge adjacent beams so that both @xmath144 and @xmath145 are doubled at each step . the effective beam diameter @xmath27covers the range from @xmath157 to @xmath39 at logarithmically spaced intervals . this procedure is similar to the one described in fabbri & natale ( 1990 ) . the present analysis only differs in the treatment of masked bins in the merging process : composed windows are now rejected when more than one bin is masked , otherwise the masked bin is filled with the average number of sources evaluated on all unmasked bins in the whole map . the results obtained for @xmath158 and @xmath159 are given in fig . 4 with error bars at 1 sigma . ideally the two data sets should coincide up to a multiplicative factor . the figure shows that this is verified to a good accuracy : quite small deviations arise from differences in the masks constructed at different resolutions . the data set with @xmath160 is not reported here , since it gives no new results and the first data point could not be compared to theory for the reasons explained in sect.2 . we also computed the correlation function at @xmath161 by means of the equation @xmath162 ^{-1 } \nonumber \\ & & \sum \left [ n(i_b , i_l)-\left\langle n\right\rangle \right ] \left [ n(j_b , j_l)-\left\langle n\right\rangle \right ] , \end{aligned}\ ] ] where the summation is performed over all pairs of beams such that the angular separation of center coordinates lies in the range @xmath163 . the number of such pairs is @xmath164 , and @xmath165 with @xmath166 an integer . masked bins are simply skipped in this computation . now we can not maintain self - similarity ; therefore we excluded two galactic polar cups at @xmath167 where the stretching of bin shapes is strong . figure 5 reports data in the range @xmath168 . in order to treat the peak number statistics , we should take into account that theoretical predictions are available for smooth distributions over the celestial sphere . for any distribution of discrete sources we encounter the following problem , that moving a sharp - edge beam across the sky the detected source number varies in a discontinuous manner . therefore we must limit ourselves to smooth beams , and in this paper we shall consider gaussian beams . in principle , for each beam position in the sky we should weight all sources in the sample by a smoothing gaussian factor @xmath169 $ ] , with @xmath170the source angular separation from the beam center . in practice it is more efficient to replace the actual content of each bin by an effective , smoothed number of sources according to the formula @xmath171 } { { \sum } \exp \left [ -\frac 12\left ( \vartheta _ { i_b i_l } /\sigma \right ) ^2\right ] } , \ ] ] where @xmath172 is the angular separation between cells @xmath173 and @xmath174 . masked bins obviously are disregarded . now a complication arises from the fact that the measured number of peaks explicitly depends on both @xmath27and @xmath157 , while theory assuming a continuous sampling of the source number - density field only predicts the @xmath127-dependence . therefore , to consistently compare data at varying values of @xmath27we had to require similar geometries , i.e. equal ratios @xmath175 . since we considered four logarithmically spaced values of @xmath27 in the range @xmath176 , from our bin maps we can build three sets of data with @xmath177 , 0.47 , and 0.93 respectively . each set includes a triplet of values of @xmath127 ; for instance , for @xmath178 we can collect data at @xmath179 , @xmath180 and @xmath181 taking advantage of all of our bin maps . in practice starting from 3 original bin maps we must build 9 smoothed maps , corresponding to specific values of @xmath27and @xmath147 , and in each of them we search for local maxima of the @xmath182 distribution . the search was performed comparing the source number of each bin with those of eight contiguous bins . we handled with the presence of holes in the map developing algorithms which extended the analysis to larger neighborhoods of potential maxima . the presence of very few uncertain cases does not affect our results . the identified peaks were then normalized to the rms source - number fluctuation calculated for each smoothed map , and the distribution @xmath26 was computed for 7 equally spaced values of @xmath29 between 0.1 and 1.6 . since the number of peaks should then be extrapolated to the whole sky , we have to evaluate the effective area associated to the distribution of detected peaks . we adopt two procedures : ( a ) we only include sky bins connected to the two big unmasked continents at @xmath183 and surrounded by no more than 5 masked bins , and ( b ) we accept all unmasked bins . figure [ ex6a ] gives @xmath26 , extrapolated to the whole sky from the entire set of unmasked bins using the first method above . since only data sets with fixed @xmath175 are homogeneous , the figure shows 3 distinct meshes . for the case @xmath184 only two values of @xmath108 are considered because data at @xmath185 could not be compared with theory . accepting all unmasked bins for the computation of areas slightly lowers the peak numbers , but this has no important effect as will be seen in sect . 4 . figure [ ex6b ] reports @xmath186 for the case @xmath187 , derived from the north and south halves of the sample . in this figure the peak number is extrapolated to one half of the sky . error bars at 1 sigma are evaluated as @xmath188 , with @xmath189 the unmasked sky fraction . such errors are reported in figs . [ ex7a ] and [ ex7b ] . for each of our tests we basically followed the same procedure which can be outlined as follows . after selecting values for the parameters describing the source properties , we are left with two free parameters related to cosmic inhomogeneities , i.e. the products @xmath190 ( or @xmath191 ) and @xmath32 . for single - scale models one more free parameter is the cosmic density parameter @xmath31 ; for hdm models we set @xmath37 , and for cdm models the spectral properties only depend on the shape parameter @xmath192 . for each parameter set we first minimize @xmath193 with respect to a normalization factor , which in the @xmath194and @xmath195 tests determines the normalization of @xmath3 . for the peak statistics this factor should not be fitted in the ideal case of a pointlike sampling of a continuous field , namely for @xmath196 , but in our case @xmath175 can be as large as 0.93 , and we have no warrant that theoretical predictions should reproduce finite - bin data exactly . however the normalization factor turned out to be usually close to unity ( within @xmath197@xmath198 and it was always so in the neighborhoods of the absolute minima . we then obtain 2- or 3-dimensional @xmath199grids , in which we search for @xmath200 . statistical errors on each parameter are calculated projecting out the extremal points of ( @xmath201 ) contours on the coordinate axes of two - parameter planes , after one more minimization with respect to the third , unplotted parameter when available . the errors quoted in the following tables are quadratic combinations of such 1-sigma errors and grid half spacing . for single - scale models , adopting the optimal source parameters we calculated a 3-dimensional lattice of equally spaced points in the ranges@xmath202 mpc , @xmath203 and @xmath204 km / s . table 1 lists fitted parameter values that we obtained from several tests , as well as @xmath205 compared to the number of degrees of freedom . since errors on @xmath31 often exceeded the range investigated , we enclosed the best values within parentheses to declare their loss of significance . lllll case@xmath206 & @xmath190 ( mpc ) & @xmath31 & @xmath32 ( km / s ) & @xmath207dof + f@xmath208 & 40@xmath209 & ( 1.0 ) & 500@xmath210 & 4.1/2 + f@xmath211 & 40@xmath212 & @xmath213 & 400@xmath214 & 15.2/3 + f@xmath215 & 40@xmath216 & @xmath217 & 400@xmath218 & 14.5/2 + c@xmath219 & 30@xmath220 & ( 0.9 ) & 700@xmath221 & 21.1/7 + p@xmath179(a ) & 20@xmath222 & ( 0.9 ) & 500@xmath223 & 14.2/17 + p@xmath179(a , n)@xmath224 & 30@xmath225 & ( 1.0 ) & 300@xmath226 & 15.6/17 + p@xmath179(a , s)@xmath224 & 20@xmath227 & ( 0.8 ) & @xmath228 & 20.8/17 + p@xmath179(b ) & 20@xmath229 & ( 0.9 ) & 500@xmath230 & 14.6/17 + p@xmath180(a ) & 50@xmath231 & ( 0.9 ) & 800@xmath232 & 24.0/17 + p@xmath180(b ) & 50@xmath216 & ( 0.9 ) & 800 @xmath232 & 24.5/17 + p@xmath28(a)@xmath233 & 30@xmath234 & ( 0.8 ) & 400@xmath235 & 30.0/22 + f+p@xmath236 & 40@xmath237 & @xmath238 & 500@xmath239 & 47.3/27 + c+p@xmath236 & 30@xmath240 & ( 0.9 ) & 400@xmath241 & 51.2/32 + @xmath242label f denotes rms fluctuation tests in the specified beamthrow range ; rounded - off bin size is within parentheses . c means correlation test in the specified beamthrow range ; bin size is @xmath243 . p marks peak number tests for triplets of @xmath108 around quoted values . for the meaning of ( a ) and ( b ) labels see sect . @xmath244n , s = northern and southern half of the sky . @xmath245full range of @xmath127 used . @xmath246joint tests combine p@xmath247(a ) data with f@xmath248 and c @xmath249 , respectively . results reported in the table for the rms fluctuation tests ( denoted by the letter f ) refer to two beamthrow ranges , 1.12@xmath250 and 2.24@xmath251 , and two values of @xmath147 . it was especially important to check a possible dependence of the best - fitted length scale on the beamthrow range . no difference appears in the quoted best values of @xmath190 , but @xmath252is high in the upper range for the difficulty of fitting the data point at @xmath39 . all of the rms fluctuation and correlation function data ( see the fourth row in the table ) are consistent with @xmath253 mpc and @xmath254 km / s . the parameter @xmath32 is correlated to @xmath31 ; if we assume @xmath37 ( which is not required by our data ) we can also derive a preferred range @xmath255 km / s . results from the peak statistics reported at rows 5 - 10 of the table ( denoted by the letter p ) are associated to a beamwidth value . each entry is in fact obtained using 3 beamwidths ( of which the label value is the geometric mean ) referring to data points with a single value of @xmath175 . thus for instance , the cases labelled by p@xmath179(a ) and p@xmath179(b ) use each 21 data points with @xmath256 , 2.4@xmath257 and 4.8@xmath257 and @xmath258 . the letters a and b refer to the methods for extrapolating the number of peaks to the whole sky described in sect . 3 . the test combining all of our @xmath259 data is denoted by the full range @xmath28 . entries in the 6th and 7th row show that consistent results were obtained for the north or south hemispheres ( which was verified for @xmath259 tests at all beamwidths ) . we do not report results for the self - similar case with @xmath184 because no significant result could be found after exclusion of data with @xmath260 . inspection of the table shows that in all cases we consistently find @xmath261 mpc and 50 mpc for smaller and larger beamwidths , respectively . the former can be regarded as a safe lower limit to a preferred scale . figure [ ex8 ] gives a few contours in the @xmath262 plane corresponding to @xmath263 , with @xmath264@xmath265 and 6.17 , which for a gaussian field should delimit the ellipses at confidence levels of 1 and 2 sigmas , respectively [ cf . press et al . ( 1988 ) , chapt . the figure shows that contours for the p tests do not overlap at 2 sigmas , since the preferred length scale show some dependence on angular scale . however , all of the other tests ( included those not reported in the figure ) provide intermediate results . finally , since as already remarked @xmath259 weights high - order harmonics much more strongly than @xmath266 and @xmath24 do , we treated @xmath259 data as quite independent from the other two sets of data . this allows us combining couples of tests , simply summing up the respective @xmath193 fields . entries labeled as f+p and c+p in the last two rows of table 1 respectively combine data from the p@xmath28(a ) case with data from f@xmath267 and c@xmath219 . the resulting @xmath205 are only slightly higher than the sums of minima in the component fields , which obviously refer to different values of the fitted parameters : this fact is due to the slow variations of @xmath193 around the minima . therefore the rather large values of @xmath205 in table 1 , which emphasize the limits of the single - scale approximation , mainly arise from some difficulties that the models encounter in each of our basic tests . in the case of continuous - spectrum models after minimization of @xmath193 with respect to the normalization factor we built 2-dimensional grids in the plane @xmath268 or @xmath269 for hdm and cdm respectively . llllll case & @xmath191 ( mpc ) & @xmath32 ( km / s ) & @xmath270 & @xmath271 & @xmath205/dof + f@xmath208 & @xmath272 & @xmath273 & 0.60 & 0.75 & 2.5/3 + c@xmath219 & @xmath274 & @xmath275 & 1.14 & 1.23 & 8.3/8 + p@xmath28(a ) & @xmath276 & @xmath277 & 0.89 & & 25.6/23 + f+p@xmath278 & @xmath279 & @xmath280 & 0.62 & & 30.3/27 + c+p@xmath278 & @xmath281 & @xmath282 & 0.89 & & 34.1/33 + @xmath242calculated from @xmath283 in the preceding column . @xmath245calculated from the normalization factor of @xmath3 . @xmath284joint tests combine p@xmath28(a ) data with f@xmath208 and c@xmath219 , respectively . lllll case & @xmath285 & @xmath286 ( km / s ) & @xmath287/dof & @xmath288/dof + f@xmath289 & @xmath290 & @xmath291 & @xmath292 & @xmath293 + c@xmath294 & @xmath295 & @xmath296 & @xmath297 & @xmath298 + p@xmath299 & @xmath300 & @xmath301 & @xmath302 & @xmath303 + f+p@xmath304 & @xmath305 & @xmath306 & @xmath307 & @xmath308 + c+p@xmath304 & @xmath309 & @xmath310 & @xmath311 & @xmath312 + @xmath313starred quantities ( last column ) refer to best fits limited to the range @xmath314 . @xmath315joint tests combine p@xmath316(a ) data with f@xmath317 and c@xmath318 , respectively . for hdm models the range limits were the same as for single - scale models , but the 2-dimensional @xmath193 grid was not equally spaced in the @xmath319 direction . ( the finest spacing , adopted around the @xmath193 minima , was 2.5 mpc . ) table 2 reports the results obtained in a subset of our best fitting procedures , which is sufficient to illustrate our findings . the table shows that the values of @xmath193 are very close to the numbers of degrees of freedom for all of the reported tests . the best fitted values of @xmath94 , derived both from @xmath283and from the normalization factor of @xmath3 , are often fairly close to the canonical value 0.7 ( lahav et al . 1990 , saunders et al . 1992 ) although there is some spreading in the results . one might have expected to find lower values of @xmath320for hdm models ; however , this is not the case because our spectra are peaked at low wavelengths . the only real limitation for so satisfactory results is the exclusion of the data point at @xmath39 in the case of the rms fluctuation test . the behaviour of the hdm model represented by the full line in fig . 4 shows how such a point spoils the quality of a fit which would otherwise be very good . in general , the quality of fits for any sharply - peaked ( including single - scale and double - power ) models is somewhat worse at larger angles . thus in fig . 5 models do not strictly follow the oscillations of @xmath321 around @xmath322 , and finally , all of our theoretical distributions of density peaks slightly deviate from experiment at our largest beamwidth , the measured @xmath323 being systematically lower than predicted for @xmath324 . we could not extend the analysis to angles larger than @xmath325 because the small number of expected peaks is completely masked by poisson noise . for the above reasons , we performed the global test f+p for hdm combining the p@xmath28(a ) and f@xmath208 data . contours at 1- and 2-sigma confidence levels for the f+p and c+p tests are reported in fig . these global tests appear to be somewhat complementary , setting more stringent limits on @xmath32 and @xmath191 respectively . we can satisfy both simultaneously with @xmath326 km / s and @xmath327 mpc . table 3 collects results for cdm tests corresponding to those of table 2 . in all of these tests @xmath51 is allowed to vary in the range ( 0.1 , 1.2 ) . however in the last column we also report @xmath328 , the minimum chi - square obtained in the smaller range ( 0.5 , 1.2 ) . the latter range is suitable to test critical density models , letting @xmath329vary ; in this case the constrained minima are always found at @xmath330 , but the fits are unsatisfactory with the exception of peak statistics . allowing for low - density models the situation improves substantially : for tests involving peak statistics only , @xmath205 is even slightly smaller than for hot matter . however , the chi - square minima identified by our tests are spread in the range @xmath331 and do not appear so mutually consistent as in the case of hot matter . the situation is made clear by inspection of fig . [ ex10 ] : for cold matter the c+p and f+p tests allow us identifying an overlap region ( @xmath332 km / s ) only at a formal confidence level of 2 sigmas . in order to decide whether our tests really argue against cdm models in low - density cosmologies , we should discuss the significance of error bars we attach to fitted parameters . the confidence regions derived from @xmath333 contours stem from purely statistical errors . ( error bars quoted in tables 1 - 3 also include grid spacing . ) as pointed out by white , krauss and silk ( 1993 ) cosmic variance can be more important than instrumental noise in anisotropy experiments such as those on the cosmic background radiation . in our case the instrumental noise ( i.e. , the error in counting extragalactic sources in sky bins in a definite flux interval ) is negligible in front of the noise term @xmath264@xmath334 , which is the source of our error bars . cosmic variance is just this noise term . in order to check that this interpretation is correct , we attributed the standard cosmic - variance uncertainty @xmath264@xmath335 to all of the models which best fitted any one of our tests , and derived the corresponding uncertainties @xmath336 on the theoretical quantities @xmath264@xmath337 , @xmath24 and @xmath259 . we found that our declared errors in figs . 4-[ex7b ] are typically @xmath338 times @xmath339 , and this ratio is larger as it must when theoretical predictions lie below experimental data points . the average ratio @xmath340 @xmath341is consistent with theoretical expectation taking into account the unmasked portion of the sky . as noted by scaramella and vittorio ( 1993 ) , taking cosmic variance into account in the computation of the @xmath193 field does not provide exactly the same results as simulations : models predicting anisotropies larger than experiment are affected by a larger cosmic variance , and can be rejected less easily than we could infer from inspection of the @xmath193 field . from these considerations and the inspection of table 3 and fig . [ ex10 ] , it follows that low - density cdm models are not clearly ruled out by our tests . however , very simple hdm models work remarkably better . the results expounded in sect . 4 show that a common picture arises from both single - scale and hdm models . the scale @xmath342 mpc suggested by the former is only slightly smaller than the peak wavelength @xmath46 arising from the latter ( in particular , the value @xmath2 mpc implied by the simultaneous consideration of the global tests ) . the linear component of our local velocity ( in practice , the velocity of the local group ) is estimated around 500 @xmath343 km / s , in agreement with current views on the large - scale velocity field , although not with inferences of about 1000 km / s ( cf . collins et al . our basic conclusions remain unchanged for different source parameters and sharply - peaked spectral shapes provided only that iras sources already existed at @xmath344 . the quality of hdm fits suggests that @xmath3 is characterized by some length scale smaller than appearing from the analyses of redshift surveys . possible objections to this result might involve the reliability of the sample , the ability of our tests to effectively probe the full perturbation spectrum and contamination from small , non - linear scales . as to the first point , we are confident that our sample is reliable . similar results ( using single - scale models and two of the present tests ) were obtained from another sample derived from iras psc , version i with more conservative selection criteria ( fabbri & natale 1990 , 1993 ) . thus there is no chance that our results might be affected by star contamination . no problem should arise from the finite depth of the sample . applying the peak statistics separately to the south and north halves of the sky we reduced the effective linear scale of the sample by a factor 2@xmath345 , but we observed no shift of the best - fitted wavelength or the confidence regions to lower scales . also , our models carefully take into account the flux interval of the sample , the source luminosity function and the relativistic expression of the luminosity distance ; they explicitly allow for the effective depth arising from the interplay of such factors . so there is no reason why such a length scale should strongly affect our tests , nor why a spectrum like ( [ hot ] ) should fit data better than equation ( [ cold ] ) . [ in fact , the luminosity function of saunders et al . ( 1990 ) does not even allow us identifying a length scale so sharply . ] the depth of the sample could drastically affect the fittings only if it arised from a strong cutoff , totally independent of the above factors and effective at redshifts @xmath346 0.1 . there is no evidence for such a cutoff . we previously studied the effect of varying the lower flux limit on version i of psc and found no effects of this kind ( fabbri & natale 1990 ) . we further notice that jing & valdarnini ( 1993 ) use the redshift field of 2-jy iras galaxies to infer a peak wavelength of @xmath347 mpc . the median depth of their sample is quite smaller than ours . thus the different results arise from the different techniques , not from peculiarities of our sample . it is therefore important to notice that the shape of the window functions which can be defined for our tests do not weight small length scales to a pathological extent ; actually they weight them less than a 2-dimensional analysis based on iras redshift survey ( scharf and lahav 1993 ) . all of the above considerations can not prove of course that our best value for @xmath46 should be the location of the absolute maximum of @xmath3 . we are aware that best - fit procedures are not equivalent to directly constructing the shape of @xmath3 from data . this consideration is especially important because we find that our tests , when applied to cdm models , point to larger length scales . this apparently strange result may also be linked to the fact that in 2-dimensional analyses length scales are smeared out ; under this respect 3-dimensional studies ( e.g. feldman et al . 1994 ) are superior . however , the quality of our hdm low - wavelength fits and the arguments discussed in the introduction makes us suspect that current spectra obtained from redshift catalogs may not represent the correct @xmath3 in real space very accurately . although there is a qualitative agreement between spectra inferred from visible and iras galaxies and clusters , a close inspection of the papers by fisher et al . ( 1993 ) , einasto et al . ( 1993 ) and jing & valdarnini ( 1993 ) shows significant differences between the spectra proposed by these authors . a conservative interpretation of our results is that at least a spectral feature should exist , not detected in the above works . strictly speaking , a local peak around @xmath348 mpc appears in fig . 5 of jing & valdarnini , but we believe its strength is too weak to justify our findings . a recent work by baugh & efstathiou ( 1993 ) studies the 2-dimensional correlation function in the apm survey and recovers @xmath3 by the method of lucy ( 1974 ) . their spectrum rises above a power law around 30@xmath349 mpc and appears to be peaked at 100@xmath349 mpc . although the authors find that their results may be compatible with previous work , still they declare that the discrepancy showed by their fig . 11 is disturbing . in our opinion the support for a steady increase of @xmath3 up to very large wavelengths is not overwhelming . [ difficulties may arise also in connection with data on the anisotropies of the cosmic background radiation ( fabbri & torres 1994 ) ] . the problems arising from the comparison of results found with different techniques may indicate that the cosmic spectrum is more complicated than the smooth theoretical shapes appearing in the current literature . we wish to thank m. t. dibari for useful discussions on iras psc . this work is partially supported by agenzia spaziale italiana under contract # 92rs64 , and by the italian ministry for the university and scientific and technological research ( progetti nazionali e di rilevante interesse per la scienza ) .	we investigate the existence and meaning of a preferred length in the large - scale distribution of 60-@xmath0 m iras sources applying three tests to the 2-dimensional distribution on the celestial sphere , involving respectively the rms fluctuation and the correlation function of the source number density , and the peak number statistics . for hdm models we find a best value @xmath1 @xmath2 mpc for the peak of the spectral function @xmath3 , close to the preferred wavelength @xmath4 mpc of nave single - scale perturbation models . a feature should indeed be generated in this range by the anticorrelation region of @xmath5 , although it is not detected by recent fourier transform analyses of redshift catalogs . our tests provide acceptable fits for cdm models with either @xmath6 or @xmath7 , in agreement with results in current literature . however , the quality of such fits is less satisfactory than for hdm models , with the exception of the peak number statistics , and we are not able to find a parameter interval satisfying all of our tests simultaneously . these results may indicate that the cosmic spectrum is more complicated than the smooth shapes predicted by current theoretical models .
let @xmath4 be a @xmath11 bounded domain of @xmath3 , @xmath12 , and let us consider the following finsler - laplacian of @xmath13 , namely the operator @xmath14 defined as @xmath15 where @xmath2 is a suitable smooth norm of @xmath3 ( see section [ 2.1 ] for the precise assumptions ) . the aim of the paper is to study the existence of solutions of the equation @xmath16 where @xmath6 , @xmath7 and @xmath8 is a suitable function in @xmath9 , bounded from below , with the boundary condition @xmath17 we will refer to the solutions of which satisfy as blow - up solutions . we are also interested in the asymptotic behavior of the solutions . moreover , we study the behavior of the blow - up solutions of when @xmath10 . problems which deal with finsler - laplacian type operators have been studied in several contexts ( see , for example , @xcite ) . when @xmath2 is the euclidean norm , namely @xmath18 , blow - up problems for equations depending on the gradient have been studied by many authors . we refer the reader , for example , to @xcite . in the euclidean setting , problem - reduces to @xmath19 the interest in problems modeled by has been grown since the seminal paper by lasry and lions @xcite . the equation in is a particular case of hamilton - jacobi - bellman equations , which are related to stochastic differential problems . indeed , in @xcite the authors enlightened the relation between problem and a model of stochastic control problem involving constraints on the state of the system by means of unbounded drifts . we briefly recall a few facts about this link . let us consider the stochastic differential equation @xmath20 where @xmath21 is a standard brownian motion . we assume that @xmath22 , where @xmath23 is the class of feedback controls such that the state process @xmath24 , solution to the above sde , remains in @xmath4 with probability @xmath25 , for all @xmath26 and for any @xmath27 . thanks to the dynamic programming principle due to bellman , the function @xmath28 , @xmath29 which solves can be represented as the value function @xmath30e^{-\lambda t}dt\ ] ] where @xmath31 is the expected value , @xmath6 , @xmath32 , @xmath33 , and @xmath34 is a discount factor . in @xcite there are several results regarding the existence , uniqueness and asymptotic behavior of the solutions of . when @xmath35 tends to zero , the limit of @xmath36 is known as ergodic limit . this kind of problems have been largely studied ( see , for example , @xcite ) . a typical result states that @xmath36 tends to a value @xmath37 and @xmath38 , for fixed @xmath39 , tends to a function @xmath40 which solves @xmath41 problem is seen as the ergodic limit , as @xmath42 , of the stochastic control problem just destribed . the scope of the present paper is to obtain existence , uniqueness and asymptotic behavior of the solutions to problem - , in the spirit of the work by lasry and lions @xcite , when @xmath2 is a general norm of @xmath43 . the interest in this kind of problems is twofold . first , in analogy with the relation between the quoted sde and the elliptic problem , we stress that the finsler laplacian @xmath5 can be interpreted as the generator of a `` h - finslerian diffusion '' , which generalizes the standard brownian motion in @xmath3 . stochastic processes of this type arise in some biology problems , as in the theory of evolution by endo - symbiosis in which modern cells of plants and animals arise from separately living bacterial species . we refer the reader to @xcite ( and to the bibliography cited therein ) for the stochastic interpretation of @xmath5 and for the quoted applications . second , apart from the stochastic motivation , the nonlinear elliptic problem we study is of interest in its own right . in our case , the operator in is , in general , anisotropic and quasilinear , with a strong nonlinearity in the gradient , and generalizes to this setting some extensively studied problems in the isotropic case . actually , this brings several difficulties and differencies with respect to the euclidean case . moreover , in @xcite the asymptotic behavior of the solutions of near to the boundary of @xmath4 is strongly related to a precise behavior of @xmath8 with respect to the distance to @xmath44 . in our case , the anisotropy of the operator leads to use an appropriate distance function to the boundary related to @xmath2 . on the other hand , unless @xmath45 , the function @xmath46 is always discontinuous in @xmath47 . hence , also giving smoothess assumptions on @xmath2 and on the data , it is not possible to apply classical caldern - zygmund type regularity results to get strong solutions in @xmath48 , @xmath29 . we deal , in fact , only with solutions in @xmath49 . furthermore , this lack of regularity does not permit to obtain , in general , the same gradient estimates for the solutions of - proved in the euclidean case , which play a central role in the study of the ergodic problem . actually , we are able to treat also the case @xmath10 , obtaining existence results for the limit problem @xmath50 and some properties of the ergodic constant @xmath51 . we refer the reader to section [ 2.3 ] for the complete scheme of the obtained results . the paper is organized as follows . in section [ 2 ] we give the precise assumptions on @xmath2 and recall some basic facts of convex analysis . moreover , we state our results . in section [ 3 ] we prove some a priori estimates for the gradient . finally , in section [ 4 ] we give the proof of the main results . throughout the paper we will consider a function @xmath52 convex , @xmath25-homogeneous , that is @xmath53 and such that @xmath54 for some constant @xmath55 . under this hypothesis it is easy to see that there exists @xmath56 such that @xmath57 moreover , we will assume that @xmath58 in all the paper we will denote with @xmath4 a set of @xmath3 , @xmath12 such that @xmath59 the hypothesis on @xmath2 assure that the operator @xmath5 is elliptic , hence there exists a positive constant @xmath60 such that @xmath61 for any @xmath62 and for any @xmath63 . we will consider as solutions of equation the strong solutions , namely functions @xmath64 such that the equality in @xmath65 holds almost everywhere in @xmath4 . in this context , an important role is played by the polar function of @xmath2 , namely the function @xmath66 defined as @xmath67 it is not difficult to verify that @xmath66 is a convex , @xmath25-homogeneous function that satisfies ( with different constants ) . moreover , @xmath68 the assumption on @xmath69 implies that @xmath70 is strongly convex , in the sense that it is a @xmath11 set and all the principal curvatures are strictly positive functions on @xmath71 . this ensures that @xmath72 ( see @xcite for the details ) . the following well - known properties hold true : @xmath73 analogous properties hold interchanging the roles of @xmath2 and @xmath66 . the open set @xmath74 is the so - called wulff shape centered at the origin . more generally , we denote @xmath75 and @xmath76 . due to the nature of the problem , it seems to be natural to consider a suitable notion of distance to the boundary . the anisotropic distance of @xmath77 to the boundary of @xmath44 is the function @xmath78 it is not difficult to prove that @xmath79 . moreover , the property gives that the @xmath80 satisfies @xmath81 furthermore , if @xmath44 is @xmath82 , then @xmath83 is @xmath82 in a suitable neighborhood of @xmath44 in @xmath84 ( see @xcite ) . since @xmath44 is @xmath82 , it is possible to extend @xmath83 outside @xmath85 to a function which is still @xmath82 in a suitable neighborhood of @xmath44 in @xmath86 . indeed , let @xmath87 and define the signed anisotropic distance function @xmath88 as @xmath89 -\widetilde d_h(x ) & \text { if } x\in { \mathbb r}^n\setminus\bar\omega . \end{array } \right.\ ] ] the following result is proved in @xcite . [ cmteo ] let @xmath4 be as in . then there exists @xmath90 such that @xmath88 is @xmath91 , with @xmath92 . the first result concerns the case when @xmath8 blows up at the boundary at most as @xmath93 , with @xmath94 . [ fpocodiv ] let @xmath95 bounded from below and such that @xmath96 then , there exists a unique solution @xmath97 of such that @xmath13 blows up at @xmath44 . moreover , any subsolution @xmath98 of is such that @xmath99 in @xmath4 . finally , if @xmath100 is the unique positive solution of @xmath101 if @xmath102 , @xmath103 if @xmath104 , then @xmath105 { c_{0}\log\frac{1}{d_{h}(x ) } } & \text{if } q=2 , \end{cases}\ ] ] as @xmath106 . the second main result we are able to prove is the case in which @xmath8 blows up very fast on @xmath44 . [ fmoltodiv ] let @xmath4 be a bounded domain of @xmath3 , and suppose that @xmath95 is bounded from below and satisfies @xmath107 then , any solution @xmath97 of bounded from below blows up at @xmath44 . moreover there exists a maximum solution of in @xmath108 and , among all the solutions bounded from below in @xmath4 , there exists a minimum one which is the increasing limit of sequences of subsolutions of . if in addition there exists @xmath109 such that @xmath110 then the blow up solution @xmath13 is unique and , as @xmath106 , @xmath111 with @xmath112 . finally , we prove what happens when @xmath113 . we will denote with @xmath114 a blow up solution of , and @xmath115 , where @xmath116 is any fixed point chosed in @xmath4 . [ teoerg ] let @xmath6 , and suppose that @xmath117 is bounded from below and such that , as @xmath106 , @xmath118 denote with @xmath114 the unique solution of in @xmath119 such that @xmath114 blows up at @xmath44 . then , @xmath120 and @xmath36 are bounded in @xmath121 and @xmath122 , @xmath123 where the convergence is uniformly on compact sets of @xmath4 . moreover , @xmath40 verifies and it is a solution of the ergodic equation @xmath124 in addition , if @xmath125 is such that the equation @xmath126 admits a blow - up solution in @xmath127 , then necessarily @xmath128 . we will refer to the unique constant @xmath51 such that admits a blow - up solution as the ergodic constant relative to . we observe that the ergodic constant @xmath51 , in the case @xmath104 , is related to an eigenvalue problem . indeed , if @xmath40 is a solution of the ergodic problem , performing the change of variable @xmath129 and using the properties of @xmath2 we have that @xmath130 satisfies @xmath131 this observation will be useful in the proof of the uniqueness , up to an additive constant , of the blow - up solutions of ( theorem [ q=2theo ] below ) . as a matter of fact , @xmath51 is the smallest eigenvalue of . we refer to the proof of theorem [ q=2theo ] for the details . when @xmath1321,2[$ ] , due to the nonlinearity of the principal part of the operator , and the fact that problem is non - variational , the uniqueness up to an additive constant of the solution of does not seem to be easy to prove . [ q=2theo ] if @xmath104 , under the hypotheses of theorem [ teoerg ] , and assuming also that @xmath117 satisfies @xmath133 for some @xmath134 , if @xmath40 and @xmath135 are blow - up solutions in @xmath136 of , then @xmath137 , for some constant @xmath138 . in this section we prove a local gradient bound for the solutions of @xmath139 such estimates are crucial in order to prove theorem [ teoerg ] on the ergodic problem . the method we will use relies in a local version , contained in @xcite ( see also@xcite ) , of the classical bernstein technique ( see @xcite ) . [ gradbound ] let @xmath4 be a bounded open set , and suppose that @xmath97 solves . for any @xmath140 , let us consider the set @xmath141 . if @xmath142 , for some @xmath1430,1[$ ] , then @xmath144 depends on @xmath145 , @xmath146 , @xmath147 and @xmath148 . actually , we will prove in section 4 that the estimate holds also under different assumptions on @xmath8 ( see remark [ remgradbound ] ) . the regularity assumptions on @xmath2 imply that @xmath149 ( see @xcite ) . for the sake of simplicity , we put @xmath150^{2 } \right\}_{\xi_{i}\xi_{j}}.\ ] ] hence the equation can be written as ( here and in the following the einstein summation convention is understood ) @xmath151^{q}+\lambda u = f.\ ] ] if @xmath152 , we can derive the equation with respect to @xmath153 , obtaining that @xmath154 let us consider @xmath155 such that @xmath156 in @xmath157 , @xmath158 on @xmath159 and @xmath160 for some @xmath161 0,1\right [ $ ] that will be determined later and some constant @xmath162 . multiplying by @xmath163 and summing we get @xmath164 denoting @xmath165 , equation can be rewritten as @xmath166 or @xmath167v - a^{ij } \varphi_{x_{i } x_{j } } v + 2\frac{v}{\varphi } ( a^{ij}\varphi_{x_{i}}\varphi_{x_{j}}).\end{gathered}\ ] ] let @xmath116 be a maximum point for @xmath168 in @xmath4 . obviously , @xmath169 , otherwise @xmath170 in @xmath4 . for the same reason , we can assume that @xmath171 . then by the maximum principle we get the following inequality in @xmath116 : @xmath172v + \\- a^{ij } \varphi_{x_{i } x_{j } } v + 2\frac{v}{\varphi } ( a^{ij}\varphi_{x_{i}}\varphi_{x_{j}}).\end{gathered}\ ] ] now , being @xmath173 @xmath25-homogeneous , and recalling that @xmath174 , it follows that @xmath175 are homogeneous of degree @xmath176 , and then @xmath177 hence , using the above inequality , the boundedness of @xmath178 , and the young inequality , we get @xmath179 on the other hand , @xmath180 } \ge c \left(h(\nabla u)^{q}+\lambda u -f\right)^{2}.\ ] ] hence , for @xmath181 sufficiently small , recalling and that @xmath182 is bounded from below , we have @xmath183^{2}\varphi \le \\ \le c\left\ { \frac{\|\nabla \varphi \|^{2}}{\varphi } \|\nabla u\|^{2}+2\varphi \,\left\vert \nabla f\right\vert \left\vert \nabla u\right\vert + q\left\vert h\left ( \nabla u\right ) \right\vert ^{q-1}\left\vert h_{\xi } \left ( \nabla u\right ) \right\vert \left\vert \nabla \varphi \right\vert v - a^{ij}\varphi_{x_{i}x_{j}}v+2\frac{v}{\varphi } ( a^{ij}\varphi _ { x_{i } } \varphi_{x_{j}}).\right\}\end{gathered}\ ] ] now using conditions , , the boundedness of @xmath178 and the @xmath184-homogeneity of @xmath185 , we get @xmath186^{2}\varphi \le c\left(\varphi v^{\frac{1}{2}}+\varphi ^{\theta } v^{\frac{q+1}{2}}+\varphi^{\theta } v\right)\ ] ] that means @xmath187 easy computation show that if @xmath188 , then @xmath189 being @xmath190 in @xmath191 , we get that @xmath192 and the proof is complete . actually , we can prove a more precise estimate of the gradient of the solutions when we precise the behavior of the datum @xmath8 near the boundary . [ gradbound2 ] let @xmath4 be a bounded open set , and suppose that @xmath97 solves . supposing that @xmath117 satisfies @xmath193 , @xmath134 , and @xmath194 for some @xmath195 . then @xmath196 where @xmath197 only depends on @xmath198 and the diameter of @xmath4 . let @xmath199 , define @xmath200 and consider @xmath201 , @xmath202 , for @xmath203 . the function @xmath204 solves @xmath205 the hypothesis on @xmath8 gives that @xmath206^{q'-\beta}=c_{4}\ ] ] where @xmath207 , and , similarly , @xmath208 now , using the estimate , we have @xmath209 where @xmath197 depends on @xmath210 . we split the proof considering first the case of @xmath8 bounded , then we consider the general case , with @xmath211 such that holds . * case 1 : @xmath212. we look for solutions which blow up approaching to the boundary . to this aim , we consider functions of the type @xmath213 , with @xmath214 and @xmath215 . recall that the anisotropic distance function is @xmath216 , where @xmath217 , with @xmath218 sufficiently small , is a tubular neighborhood of @xmath44 . if we substitute such functions in , by and property we get that @xmath219 moreover , if @xmath220 is the unique minimum point of , that is @xmath221 , then @xmath222 ( see ( ? ? ? 3.3 ) ) , and then by we have @xmath223 moreover , using , we finally have @xmath224 hence , computing the anisotropic laplacian and using and it follows that @xmath225= \\ = c_0 \alpha(\alpha+1 ) d_h(x)^{-\alpha-2 } \frac{h^o_\xi(x-\bar y_x)\cdot(x-\bar y_{x})}{h^{o}(x-\bar y_{x } ) } + \\- c_{0}\,\alpha \ , d_h(x)^{-\alpha-1 } \sum_{i , j=1}^{n } h_{\xi_i\xi_j}(\nabla d_h(x ) ) { \partial}_{x_i x_j}d_h(x)= c_0 \alpha(\alpha+1 ) d_h(x)^{-\alpha-2 } -k(x ) d_h(x)^{-\alpha-1},\end{gathered}\ ] ] where @xmath226 is bounded in @xmath227 , being @xmath228 bounded on @xmath71 , and @xmath229 . hence @xmath230 if @xmath8 is in @xmath231 , the most explosive term in is @xmath232 if @xmath102 , this leads to the choice of @xmath233 otherwise , for @xmath104 , @xmath234 , and it leads to the choice of @xmath235 . we construct , by means of the signed distance function @xmath236 , defined in , a suitable family of subsolutions and supersolutions of . to this aim , recalling that @xmath237 , where @xmath238 is given in theorem [ cmteo ] , it is possible to construct a function @xmath239 in @xmath240 such that @xmath241 d(x ) \ge \delta_{0 } & \text{if } x\in \omega,\text { and } d_{h}(x ) > \delta_{0 } , \\[.1 cm ] d(x)=-\widetilde d_{h}(x)&\text{if } x\not\in \bar \omega,\text { and } \widetilde d_{h}(x ) \le \delta_{0 } , \\[.1 cm ] d(x ) \le -\delta_{0 } & \text{if } x\not\in \omega,\text { and } \widetilde d_{h}(x ) > \delta_{0 } , \end{array } \right.\ ] ] where @xmath242 is a positive constant smaller than @xmath243 . hence , if @xmath102 , for @xmath244 and @xmath147 such that @xmath245 , we define @xmath246 { \overline w_{{\varepsilon},\delta}}(x ) = ( c_{0}+{\varepsilon})(d(x)-\delta)^{-\alpha } + c_{{\varepsilon } } \quad x \in \omega_{\delta } , \end{array}\ ] ] where @xmath247 is a constant which will be chosed later , and @xmath248 \omega_{\delta } : = \ { x \in \omega \colon d ( x ) > \delta \ } \subset \omega . \end{array}\ ] ] for suitable choices of @xmath247 , the functions in are a super and subsolution of in @xmath249 and @xmath252 , respectively ( we may assume @xmath253 in @xmath254 ) . indeed , for @xmath255 and @xmath100 as in , we get @xmath256 + \\ & { } \phantom{spaziospiospaziospaziospaziospaziospazio}+ \lambda c_{{\varepsilon}}-\bar c(1+(d-\delta)^{-\alpha-1})\ge \\ & \ge \nu { \varepsilon}(d-\delta)^{-\alpha-2}+\lambda c_{{\varepsilon}}-c(1+(d-\delta)^{-\alpha-1 } ) , \end{split}\ ] ] for some @xmath257 and @xmath258 . we stress that @xmath259 is bounded in @xmath260 being @xmath261 and @xmath262 . by choosing @xmath247 sufficiently large , last term in the above inequalities is nonnegative , and @xmath263 is a supersolution of in @xmath191 . the same argument shows that @xmath264 is a subsolution of in @xmath249 . now , fixed @xmath265 , let us consider the approximating problem @xmath266 u_{m}=\underline w_{{\varepsilon } , 1/ m } & \text{on } { \partial}\omega . \end{array } \right.\ ] ] observe that @xmath267 on @xmath44 . then , performing the change of unknown @xmath268 , problem can be rewritten as @xmath269 v_{m}=0 & \text{on } { \partial}\omega . \end{array } \right.\ ] ] problem admits a sub and supersolution in @xmath270 ( it is sufficient to take two suitable constants ) . under hypotheses and , by ( * theorem 2.1 ) , we get that problem admits a weak solution @xmath271 , namely @xmath272 satifies @xmath273\,dx= \int_{\omega } \left(f-\lambda c_{{\varepsilon},m}\right)\varphi\,dx,\\ \forall \varphi\in w_{0}^{1,2}(\omega)\cap l^{\infty}(\omega).\end{gathered}\ ] ] then also admits a weak solution @xmath274 . moreover , such solutions are in @xmath275 ( see @xcite , and the remarks contained in @xcite ) , and in @xmath276 ( see @xcite ) . now we apply the comparison principle contained in ( * ? ? ? * theorem 3.1 ) ( see also ( * ? ? ? * theorem 3.1 ) ) . we stress that the hypothesis ( 22 ) in @xcite holds , because the function @xmath277 is @xmath184-homogeneous , and then @xmath278 hence we have that , for @xmath279 and for any @xmath280 , @xmath281 last inequality in follows observing that @xmath282 near the boundary of @xmath4 , being @xmath283 is finite on @xmath44 , while @xmath284 , and then using the comparison principle . hence gives that the functions @xmath285 , @xmath265 are uniformly bounded in @xmath286 . this estimate , since @xmath211 , allows to apply ( * ? ? ? * theorem 1 ) in order to obtain a @xmath287 estimate . actually , in any compact set @xmath288 , by ( * ? ? ? * theorem 1 ) we have @xmath289 where @xmath290 is a constant which depends only on @xmath291 and on the @xmath292 bound of @xmath285 in @xmath293 . then by ascoli - arzel theorem @xmath285 , as @xmath294 , converges locally uniformly to a function @xmath295 . moreover , @xmath296 is a weak solution of and , recalling , @xmath297 next step consists in constructing a maximum blow - up solution of . to this end , we may argue as before to get the minimal solution @xmath301 of in @xmath191 which diverges on @xmath302 . we have that @xmath303 moreover , if @xmath299 is any blow up solution of , being @xmath40 bounded on @xmath304 , we have that @xmath305 passing to the limit as @xmath306 in , using and , reasoning as before we get a maximal blow - up solution @xmath307 of such that @xmath308 for any @xmath309 . as matter of fact , we claim that @xmath310 indeed , by it follows that @xmath311 hence , being @xmath312 and @xmath313 divergent near the boundary , we get that for any @xmath3140,1[$ ] there exists a neighbourhood of @xmath44 , dependent on @xmath315 , in which @xmath316 with @xmath317 . the function @xmath318 is a subsolution of , and by maximum principle @xmath319 in all @xmath4 . as @xmath320 , we get that @xmath321 in @xmath4 , and we get the claim . we further emphasize that inequality clearly holds also if @xmath322 is any subsolution of problem . passing to the limit , we get @xmath323 . + * case 2 : * @xmath8 unbounded . the proof runs analogously as in the previous case , except what concerns the existence of the minimum explosive solution . indeed , if @xmath324 or @xmath325 near @xmath44 , substituting @xmath326 in , with @xmath327 , we have that the most explosive term in , when @xmath328 approaches the boundary , is @xmath329 d_h^{-q'}(x).\ ] ] hence , as before we can construct a maximum explosive solution @xmath330 of such that @xmath331 where @xmath332 is the function defined in . as regards the existence of the minimum solution , differently from the bounded case we have that @xmath333 defined in is a subsolution of in @xmath249 , with @xmath8 replaced by @xmath334 c_{2}+c_{3}(d+\delta)^{-q ' } & \text { in } \omega^{\delta}\setminus \omega , \end{cases}\ ] ] with @xmath335 are positive constant such that @xmath336 , and @xmath337 in @xmath4 . now , @xmath338 is bounded in @xmath4 , and from the first case we get that there exists a unique explosive solution @xmath339 of with @xmath8 replaced by @xmath338 , and @xmath340 . hence , being @xmath341 , the comparison principle gives that @xmath342 . passing to the limit , we obtain a minimal solution @xmath343 of , with @xmath344 , that satisfies . again , the uniqueness and the comparison with subsolutions follows as before . [ remgradbound ] we observe that by taking a closer look to the proof of theorem [ fpocodiv ] , we are able to conclude that the thesis of the theorem [ gradbound ] holds also if @xmath345 and is satisfied . indeed , by using the approximating problems @xmath346 \tilde u_{m}=\underline w_{{\varepsilon } , 1/ m } & \text{on } { \partial}\omega , \end{array } \right.\ ] ] with @xmath347 sequence of smooth functions such that @xmath348 in @xmath287 , the solutions @xmath349 are uniformly bounded in @xmath286 and converge , up to a subsequence , to the unique blow - up solution @xmath13 of problem . then applying the bound in @xmath191 to @xmath349 and passing to the limit we get the same bound also for @xmath13 . once we prove the claim , the thesis of the theorem follows by adapting the proof contained in ( * ? ? * theorems iii.2 and iii.3 ) and the arguments used in theorem [ fpocodiv ] in order to construct a minimum and a maximum solution and , under the additional hypothesis , that such solutions coincide . in order to prove the claim , we may suppose , without loss of generality , that @xmath351 in @xmath4 and @xmath352 for some positive constant @xmath353 . let @xmath116 be a point in @xmath4 such that @xmath354 . hence @xmath355 , and from the equation we get that @xmath356 where @xmath357 . this means that @xmath13 is a supersolution of @xmath358 and , obviously , @xmath359 is a subsolution of . applying again ( * ? ? ? * theorem 2.1 ) and @xcite , there exists a strong solution @xmath360 of . hence , @xmath361 in @xmath362 . defining @xmath363 , for @xmath364 , with @xmath365 , it follows that @xmath366 solves @xmath367 for @xmath368 , multiplying the above equation by @xmath369 and integrating , we easily get , by , , and being @xmath370 , that @xmath371 and , for @xmath372 , @xmath373 where @xmath290 is a constant independent on @xmath374 . hence , the classical stampacchia lemma ( see @xcite ) assures that @xmath366 is uniformly bounded in @xmath375 . moreover , by @xcite @xmath366 is the unique bounded solution of , which is also radial with respect to @xmath66 , due to the symmetry of the data . that is , @xmath376 , @xmath377 . reasoning as in theorem [ fpocodiv ] we get that @xmath378 , where @xmath379 , @xmath377 solves @xmath380 as a matter of fact , @xmath381 solves the problem @xmath382,\\[.3 cm ] u_{0}(1)= 0,\ ; u_{0}'(0)=0 . \end{array } \right.\ ] ] hence , by the maximum principle @xmath383 . this implies that , for @xmath102 , @xmath384 diverges as @xmath350 . as regards the case @xmath104 , this method allows only to say that @xmath385 as matter of fact , arguing as in @xcite , we have that for any @xmath309 there exists @xmath386 such that for @xmath387 with @xmath388 , then @xmath389 . now , putting @xmath390 , and repeating exactly the above argument for @xmath40 ( at least for @xmath391 ) , we get that @xmath392 letting @xmath181 go to zero , and iterating the argument , we get that @xmath13 diverges as @xmath393 also if @xmath104 . the argument of the proof of theorem [ fpocodiv ] allows to obtain that the solution @xmath394 of problem satisfies , if @xmath395 , @xmath396 for all @xmath309 , @xmath3970,1]$ ] , and for some @xmath398 . in the case @xmath104 , the functions @xmath399 have to be replaced with @xmath400 . by , @xmath36 is uniformly bounded from below and in @xmath286 . moreover using theorem [ gradbound ] and remark [ remgradbound ] we get that also @xmath120 is uniformly bounded in @xmath286 . then , @xmath401 , for some fixed @xmath39 , is uniformly bounded with respect to @xmath4020,1]$ ] in @xmath403 . hence , for any @xmath404 , there exists a constant @xmath405 independent on @xmath35 such that @xmath406 passing to the limit we obtain , up to a subsequence , the convergence of @xmath407 to a constant @xmath51 and of @xmath408 to @xmath184 . we finally prove that @xmath409 converges to a blow - up solution of . first observe that @xmath410 satisfies the following equation in @xmath157 : @xmath411 hence , using again the arguments of the proof of the previous results , we can pass to the limit in , obtaining that @xmath409 converges to a solution @xmath322 of the problem . now we prove a lower bound for @xmath40 . let @xmath412 , with @xmath4130,c_{0}[$ ] fixed . then in a sufficiently small inner tubular neighbourhood of @xmath44 , namely @xmath414 , we have that @xmath415 on the other hand , @xmath409 is bounded from below in @xmath416 , namely there exists a constant @xmath417 such that @xmath418 adapting the methods used in theorem [ fpocodiv ] it is possible to obtain that @xmath419 passing to the limit , also @xmath40 satisfies . now we show that for any couple @xmath420 of problem , with @xmath135 such that blows up at the boundary , @xmath135 diverges as in . to this aim , it is possible to consider @xmath421 as in which is supersolution of the ergodic equation in @xmath422 , for some @xmath423 . hence , by the comparison principle , and letting @xmath147 go to zero , we can conclude that @xmath424 hence , @xmath135 is such that @xmath425 , with @xmath426 . the bounds in and the condition assure that @xmath427 and also satisfies . by theorem [ fpocodiv ] we get that @xmath135 satisfies . as regards the uniqueness of the ergodic constant @xmath51 , the proof runs similarly as in @xcite , supposing by contradiction that @xmath430 . let us choose @xmath309 , and @xmath431 . first , observe that obviously @xmath40 satisfies @xmath432 on the other hand , we have from the @xmath25-homogeneity of @xmath2 that @xmath433 moreover , since @xmath40 and @xmath135 diverge as @xmath399 near to the boundary of @xmath4 , then @xmath434 . hence , from the above inequality it follows that @xmath435 where last inequality holds for @xmath315 sufficiently near to @xmath25 and for @xmath436 sufficiently small . hence , @xmath437 is a subsolution of . by theorem [ fpocodiv ] , @xmath438 . as @xmath439 , @xmath440 . this is in contradiction with the fact that any function of the type @xmath441 , with @xmath442 solves the ergodic problem with the same constant @xmath125 . the hypothesis @xmath104 allows to perform a suitable change of variable . let @xmath322 be a solution of the ergodic equation with @xmath443 on @xmath44 . then the function @xmath444 belongs to @xmath445 . let us verify that @xmath446 . due the condition , we have that @xmath235 in , and then @xmath447 . moreover , using also we can apply theorem [ gradbound2 ] , obtaining that @xmath448 is bounded . hence @xmath449 now , observe that using the properties of @xmath2 it holds that the function @xmath130 is a @xmath450 solution of @xmath451 the ergodic constant @xmath51 is a critical point of the rayleigh quotient @xmath452=\dfrac{\displaystyle\int_{\omega } h(\nabla \psi)^{2}dx+\int_{\omega}f(x)\psi^{2}dx}{\displaystyle\int_{\omega}\psi^{2}dx}.\ ] ] as a matter of fact , we claim that @xmath51 is the minimum eigenvalue , namely @xmath453,\ ] ] and @xmath51 is the only eigenvalue associated to a positive eigenfunction . the claim follows observing , first of all , that being @xmath8 bounded from below , @xmath454 , the rayleigh quotient @xmath455 $ ] satisfies @xmath452 \ge -c.\ ] ] then the existence of the minimum value of @xmath455 $ ] easily follows by using standand arguments of calculus of variations . moreover the simplicity of @xmath51 and the fact that it is the unique eigenvalue associated to a positive eigenfunction follows by adapting the proof contained , for example , in @xcite and @xcite . hence problem admits , up to a multiplicative constant , a unique solution . this implies that if @xmath456 and @xmath457 solve , then @xmath456 and @xmath457 differ by a constant .	in this paper we prove existence results and asymptotic behavior for strong solutions @xmath0 of the nonlinear elliptic problem @xmath1 where @xmath2 is a suitable norm of @xmath3 , @xmath4 is a bounded domain , @xmath5 is the finsler laplacian , @xmath6 , @xmath7 and @xmath8 is a suitable function in @xmath9 . furthermore , we are interested in the behavior of the solutions when @xmath10 , studying the so - called ergodic problem associated to . a key role in order to study the ergodic problem will be played by local gradient estimates for . + : 35j60 ; 35j25 ; 35b44 + : anisotropic elliptic problems , finsler laplacian , blow - up solutions +
let @xmath4 be a positive integer , @xmath5 a real number in @xmath6 $ ] , and denote by @xmath7 the probability measure on @xmath8 which is the product of @xmath4 bernoulli measures with parameter @xmath5 . @xmath9 we write @xmath10 instead of @xmath7 when no confusion is possible . if @xmath2 is a subset of @xmath8 , we say that @xmath2 is _ monotone _ if and only if : @xmath11 where @xmath12 is the partial order on @xmath8 defined coordinate - wise . it follows from an elementary coupling device that for @xmath2 a monotone subset , the mapping @xmath13 is increasing . for many examples of interest ( see section [ examples ] for some examples ) , a threshold phenomenon occurs for property @xmath2 in the sense that the function @xmath14 `` jumps '' from near 0 to near 1 over a very short interval of values of @xmath5 . such threshold phenomena have been shown to occur in most discrete probabilistic models , such as random graphs ( see bollobs@xcite ) , percolation ( see grimmett @xcite ) , satisfiability in random constraint models ( see creignou and daud @xcite , friedgut @xcite , bolllobs et al . @xcite ) , local properties in random images ( see coupier et al . @xcite ) , reliability ( see paroissin and ycart @xcite ) and so on . to make the statement of a threshold phenomenon more precise , one need first to define the _ threshold width _ of a _ non trivial _ monotone subset @xmath2 . we say that @xmath2 is non trivial if it is non empty and different from @xmath8 itself . when @xmath2 is non trivial and monotone , the mapping @xmath13 is invertible . thus , for @xmath15 $ ] , let @xmath16 be the unique real in @xmath6 $ ] such that @xmath17 . the threshold width of a subset is the length of the `` transition interval '' , that is to say , the interval over which its probability raises from @xmath18 to @xmath19 . [ defwidth ] let @xmath2 be a non trivial monotone subset of @xmath8 . let @xmath20 . the _ threshold width of @xmath2 at level @xmath18 _ is : @xmath21 [ ] [ ] @xmath22 [ ] [ ] @xmath23 [ ] [ ] @xmath18 [ ] [ ] @xmath19 [ ] [ ] @xmath5 [ ] [ ] @xmath24 [ ] [ ] @xmath25 .,title="fig:",width=226 ] when one investigate the threshold of a monotone property , for example connectivity in the random graph , one has to do with a sequence of non trivial monotone subsets @xmath26 where @xmath27 is an increasing sequence of integers . in the sequel , we shall suppose that @xmath27 is only nondecreasing , for technical reasons . remark that , in order to get an intrinsec notion of width or localisation order , one has to keep in mind the size @xmath28 in which the subsets @xmath29 take place . therefore , if the threshold width of a subset @xmath29 of @xmath30 is of order @xmath31 , we should rather express it as @xmath32 , where @xmath33 is the pseudo - inverse of @xmath34 : @xmath35 in order to describe the asymptotic behaviour of a property , we shall therefore use the following definitions . let @xmath26 be a monotone property , @xmath31 and @xmath36 be two sequences of real numbers in @xmath6 $ ] , and @xmath37 $ ] . + the property @xmath2 is _ located at @xmath34 _ if : @xmath380,1[,\;p_{a_n,{\varepsilon}}\xrightarrow[n\rightarrow + \infty]{}\alpha\;.\ ] ] _ the location of @xmath2 is of order @xmath39 _ if : @xmath380,1[,\;p_{a_n,{\varepsilon}}=o\left(a(\alpha_n)\right)\;,\ ] ] as @xmath4 tends to infinity . + _ the threshold width of @xmath2 is of order @xmath40 _ if : @xmath380,1[,\;\tau(a_n,{\varepsilon})=o\left(b(\alpha_n)\right)\;,\ ] ] as @xmath4 tends to infinity . + the property @xmath2 _ has a sharp threshold _ if : @xmath380,1[,\;\frac{\tau(a_n,{\varepsilon})}{p_{a_n,1/2}(1-p_{a_n,1/2})}\xrightarrow[n\rightarrow + \infty]{}0\;.\ ] ] the property @xmath2 _ has a coarse threshold _ if it does not have a sharp threshold intuitively one would be tempted to say that a subset @xmath2 will have a narrow threshold unless a few coordinates have a strong influence on its definition ( as an example , think of @xmath41 ) . in many examples , this idea is captured by the notion of symmetry . [ defsymmetric ] the subset @xmath2 of @xmath8 is said to be symmetric if and only if there exists a subgroup @xmath42 of @xmath43 ( group of permutations ) acting transitively on @xmath44 , such that @xmath2 is invariant under the action of @xmath42 : @xmath45 for a symmetric subset , no coordinate has a stronger influence than any other . in friedgut and kalai @xcite , it is proven that the threshold width of any symmetric subset @xmath46 is at most of order @xmath1 . for properties whose threshold is located away from 0 and 1 , friedgut and kalai show that this upper bound is tight and that the threshold width is at least of order @xmath0 . in order to deepen the link between the invariance group of @xmath2 and the largest possible threshold width for @xmath2 , bourgain and kalai @xcite introduce , for any permutation group @xmath47 , @xmath48 they obtain nearly optimal asymptotics for @xmath49 when @xmath42 is a primitive permutation group . recall that a permutation group @xmath47 is primitive if its action on @xmath44 has no nontrivial group blocks , where a group block is a subset @xmath3 of @xmath44 such that for all @xmath50 , @xmath51 or @xmath52 . essentially , bourgain and kalai @xcite show that there are some gaps in the possible behaviours of @xmath49 for primitive groups @xmath42 . when @xmath42 is @xmath53 ( the alternating group ) or @xmath43 , @xmath49 is of order @xmath0 . for any other primitive group , @xmath49 is either of order @xmath54 , for @xmath55 belonging to arbitrarily small intervals around a value of the form @xmath56 , where @xmath57 is a positive integer depending only on @xmath42 , or of order @xmath58 , with @xmath59 which tends to one as @xmath4 tends to infinity . these results concern the worst threshold intervals for a given transitive group . in order to complete these results , it is natural to ask , given an increasing sequence of positive real numbers @xmath31 between @xmath60 and @xmath61 , whether there exists a symmetric property @xmath2 whose threshold width is @xmath62 . only few types of such asymptotics are known . the main result of this paper is theorem [ thmlargeursvariees ] , which gives a positive answer to this question under a mild hypothese of smoothness on the sequence @xmath31 . this result is achieved by using a combination of two properties @xmath2 and @xmath3 that we shall simply call the product of @xmath2 and @xmath3 . this paper is organized as follows . section [ examples ] is devoted to some examples of properties with explicit threshold widths and locations . some of them , which arise from reliability theory , will be used further as elementary building blocks to derive more general widths . in section [ product ] , we derive the basic properties of the product of @xmath2 and @xmath3 which turns out to have a simple interpretation in terms of failure sets . we prove that the product of @xmath2 and @xmath3 has a threshold width which is the product of that of @xmath2 and @xmath3 as soon as the threshold of @xmath3 is located away from 0 and 1 . this result allows us to obtain in section [ seuilsintermediaires ] some symmetric properties of @xmath8 with arbitrary threshold widths between @xmath1 and @xmath0 . for the sake of completeness , we also study the case where the threshold of @xmath3 tends to 0 or 1 . although we do not give an extensive understanding of what may happen , we show in section [ coarsesharp ] that if @xmath2 and @xmath3 have a threshold located respectively in 0 and 1 , then @xmath63 has a sharp threshold . in presenting the following examples of thresholds , our aim is twofold . first , we want to describe some of the few already known types of behaviour . second , we shall use some of these examples in section [ seuilsintermediaires ] , to derive more general widths thanks to the product of properties . one of the typical examples of threshold phenomena is that of the random graphs @xmath64 ( see erds and rnyi @xcite , bollobs @xcite , spencer @xcite ) . the graph @xmath65 has @xmath4 vertices , and each one of the @xmath66 possible edges is present with probability @xmath5 , independently from the others . once a labelling of the vertices is choosen , one can denote by @xmath67 the probability space of the random graph @xmath65 . small balanced subgraphs + let @xmath68 be a fixed connected graph with @xmath69 vertices and @xmath70 edges , and suppose that @xmath68 is balanced , that is to say none of its subgraphs has average degree strictly smaller than @xmath68 . denote by @xmath71 the property for a graph to contain at least one copy of @xmath68 . the threshold of @xmath71 is located at @xmath72 , i.e @xmath73 , and has width of the same order ( cf . spencer @xcite for instance ) . this implies that @xmath71 has a coarse threshold . connectivity + it is known ( see bollobs @xcite ) , that the probability for @xmath64 to be connected goes from @xmath74 to @xmath75 when @xmath76 , and @xmath55 goes from @xmath77 to @xmath78 . in this example , the threshold is located around @xmath79 i.e @xmath80 , and its width is of order @xmath81 , i.e @xmath82 . thus , this threshold is sharp . let us turn to examples occuring in reliability theory . in this framework , at instant @xmath83 , two characteristic quantities of the system are especially important : the reliability , that is the probability that there never occured any breakdown before @xmath83 , and the non - availability , which is the probability that the system is down at instant @xmath83 ( see for instance barlow and proschan @xcite ) . of course , these quantities differ if the system is repairable . the analysis of the reliability of large systems , for instance its asymptotic behaviour , is generally much more difficult than the analysis of the non - availablity . we shall only focus on the latter one , but want to stress the fact that when one deals with a large system composed of repairable markovian components , it is natural to expect strong similarities between the asymptotics of the two quantities ( see for example paroissin and ycart @xcite ) . when @xmath84 denotes a system composed of @xmath4 binary components , one can describe the states of these components as a state in @xmath8 , 1 standing for a failed component , and 0 for a working component . one can therefore associate to @xmath84 its failure subset , which is the subset @xmath2 of @xmath8 containing all the configurations of the @xmath4 components such that the system @xmath84 fails . if we assume that a component is failed independently from the others with probability @xmath5 , @xmath7 is the distribution of the state of @xmath84 in @xmath8 , and @xmath85 is the non - availability of @xmath84 . it is very natural to assume that the subset @xmath86 is monotone ( if the system is down , and a component fails , then the system remains down ) . the question of how quickly @xmath85 `` jumps from 0 to 1 '' is of great importance ( see paroissin and ycart @xcite for an application of the works of friedgut and kalai @xcite and bourgain et al . @xcite in this context ) . the main result of this article , theorem [ thmlargeursvariees ] , relies on examples [ exksurn ] and [ exparser ] . [ exksurn]@xmath57-out - of-@xmath4 system + the system is failed when the total number of failed components is greater than a certain threshold @xmath87 . the failure subset is therefore : @xmath88 note that the particular cases of @xmath89 and @xmath90 correspond respectively to parallel and series system . obviously , @xmath91 is monotone and invariant under every permutation of the coordinates . it is therefore a monotone symmetric subset of @xmath8 . since the sum @xmath92 has mean @xmath93 and variance @xmath94 when @xmath95 is distributed according to @xmath10 , one can guess , intuitively , that @xmath91 has a threshold located at @xmath96 , and of order @xmath97 . we shall precise this intuition when @xmath98 in lemma [ lemseuilsqrtn ] . [ exparser]parallel - series system + a parallel - series system contains @xmath99 components which are assembled into @xmath100 blocks containing @xmath101 composants . the system is failed as soon as a block is failed , and a block fails if all its components are failed . of course , the non - availability of such a system is very easy to derive . let @xmath86 denote its failure subset : @xmath102 for example , when @xmath103 , @xmath104 and @xmath105 , the threshold of @xmath86 is located at @xmath106 with a width of order @xmath1 ( see lemma [ lemseuillogn ] below ) . remark that @xmath86 is monotone and symmetric ( under permutation of the components inside a block and permutation of the blocks ) . such systems , with multi - states components instead of binary ones , have been studied by kolowrocki @xcite , and a concrete application is presented in @xcite . one can also define the dual system called series - parallel system , in which components are assembled into @xmath100 blocks containing @xmath101 composants , the system is failed when all blocks are failed , and a block is failed as soon as one component is failed . [ exkconn]consecutive @xmath57-out - of-@xmath4 system + components are arranged around a circle . the system is failed as soon as there are at least @xmath87 consecutive components down . this model has an asymptotic behaviour similar to the parallel - series system with @xmath107 blocks of @xmath57 components . for example , when @xmath108 , the threshold of the failure subset is located at @xmath106 , with a width of order @xmath1 ( for a similar result , see paroissin and ycart @xcite ) . this model was introduced by kontoleon @xcite to model some problems arising in engineering science , such as oil transportation using pipelines , telecommunication system by spacecraft relay station or transmission of data in a ring of computer ring networks , etc . as far as we know , whereas the influence of simple operations between properties has been extensively studied whithin the so - called 0 - 1 laws which occur in logic ( see compton @xcite ) , no such work has been undertaken regarding the threshold phenomena . the first combinations of properties that come to mind , union and intersection , behave quite in an unpleasant way with respect to the threshold width ( see @xcite , chapter 3 ) . in this section , we will show the nice behaviour of another combination which we simply call the product . even though linearity does not play any role in this setting , it is worth noting the similarity between this product and the kronecker product of matrices . given two properties @xmath2 and @xmath3 , on two distinct spaces , their product is a property combining the belongings to @xmath2 and @xmath3 in the following way . [ deftens ] let @xmath2 be a subset of @xmath109 and @xmath3 a subset of @xmath110 . the _ product of @xmath2 and @xmath3 _ , denoted by @xmath63 is the subset of @xmath111 defined by : @xmath112 where @xmath113 in order to visualize the precise meaning of this definition , it is convenient to consider this product via the language of reliability theory . let @xmath2 denote the failure set of a system @xmath84 composed of @xmath100 components , and @xmath3 be the failure set of another system @xmath114 , with @xmath101 components . then @xmath63 is the failure subset of the system obtained by replacing the components in @xmath114 by @xmath101 independent copies of @xmath84 . for example , one can obtain the so - called parallel - series and series - parallel systems from some elementary building blocs : the series and parallel systems ( see figure [ fig : tenseurparser ] ) . this building set can be continued , embedding systems one in another ( see figure [ fig : tenseurfiab ] ) . now , let us describe the basic properties of this product . a very nice feature is the link between the probability of @xmath63 and those of @xmath2 and @xmath3 . it is also easy to get some invariance and monotonicity properties for @xmath63 providing some similar hypotheses for @xmath2 and @xmath3 . in the sequel , if @xmath115 belongs to @xmath111 , with @xmath116 for every @xmath117 , we will denote by @xmath118 the @xmath119-th coordinate of @xmath120 , which is therefore 0 or 1 . in this way , we identify @xmath111 and @xmath121 . [ propprodtensgen ] let @xmath122 and @xmath123 . 1 . for every @xmath5 in @xmath6 $ ] , @xmath124 2 . if @xmath2 and @xmath3 are monotone , then @xmath63 is monotone . if @xmath2 is invariant under the action of a subgroup @xmath42 of @xmath125 and @xmath3 is invariant under the action of a subgroup @xmath68 of @xmath126 , then @xmath63 is invariant under the action of the subgroup @xmath127 of the permutations of @xmath128 defined by : @xmath129 if @xmath130 are independant and distributed according to the law @xmath131 , then @xmath132 has law @xmath133 . this proves the first assertion + let us prove now the second assertion . let @xmath134 and @xmath135 belong to @xmath111 . suppose that @xmath136 , i.e @xmath137 since @xmath2 is monotone , @xmath138 suppose now that @xmath139 . @xmath140 since @xmath3 is monotone , il follows from ( [ eqcroisstens1 ] ) and ( [ eqcroisstens2 ] ) that @xmath141 , which proves the monotonicity of @xmath63 . + let us prove now the last point of proposition [ propprodtensgen ] . let @xmath142 , @xmath143 and let us denote @xmath144 . @xmath145 which can be restated as : @xmath146 on the other hand , @xmath147 with @xmath148 . and also : @xmath149 therefore , @xmath150 @xmath151 which means : @xmath152 thus @xmath153 , and the proof is complete . @xmath154 [ ] [ ] @xmath84 : [ ] [ ] @xmath114 : [ ] [ ] set of failure states @xmath2 , + parallel system . [ ] [ ] set of failure states @xmath3 , + series system . [ ] [ ] set of failure states @xmath63 , + parallel - series system . [ ] [ ] set of failure states @xmath155 , + series - parallel system . [ ] [ ] @xmath84 : [ ] [ ] @xmath114 : [ ] [ ] set of failure states @xmath2 . [ ] [ ] set of failure states @xmath3 . [ ] [ ] set of failure states @xmath63 . intuitively , the first assertion in proposition [ propprodtensgen ] suggests that if the threshold of @xmath3 is located away from 0 and 1 , the threshold effects of @xmath2 and @xmath3 will conjugate and give birth to a threshold width the order of which will be the product of the widths of @xmath2 and @xmath3 . this is indeed the case , and this is roughly the statement of proposition [ propprodtensseuil ] . actually , this result is valid as long as the threshold of @xmath3 is located away from zero and one , and some additional hypotheses of homogeneity hold for the threshold widths of @xmath2 and @xmath3 . when a threshold phenomenon occurs for a property @xmath2 , it is usually true that the threshold width is homogeneous , in the sense that all the transition intervals shrink at the same speed . this allows to consider the exact order of the threshold width , since this one does not depend on the level @xmath18 . we will use the following definitions of homogeneity and strong homogeneity . [ defseuilhomogene ] let @xmath156 be a non trivial monotone property , and @xmath157 be a sequence of positive real numbers . + the threshold width of the property @xmath2 is _ homogeneous of order @xmath158 _ if : @xmath1590,1[,\mbox { s.t . } \beta<\gamma,\;p_{a,\gamma}-p_{a,\beta}=\theta(a_n)\;.\ ] ] the threshold width of the property @xmath2 is _ strongly homogeneous of order @xmath158 _ if in addition , for all sequences of real numbers @xmath160 and @xmath161 such that @xmath1620,1[,\;\forall n\in\mathbb{n},\;{\varepsilon}<\beta_n<\gamma_n < 1-{\varepsilon}\;,\ ] ] we have @xmath163 we are now able to state the main result about the width of a product . [ propprodtensseuil]let @xmath164 and @xmath165 be two nondecreasing sequences of integers , @xmath166 and @xmath167 be two monotone properties . suppose that the threshold width of @xmath2 is strongly homogeneous of order @xmath158 , and the threshold width of @xmath3 is homogeneous of order @xmath168 . suppose in addition that the threshold of @xmath3 is located away from 0 and 1 : @xmath380,1[,\;\exists\delta\in]0,1[,\;\forall n\in \nn^*,\;\delta < p_{b,{\varepsilon}}<1-\delta\;.\ ] ] then , the threshold of @xmath169 has a homogeneous width of order @xmath170 . moreover , if the threshold of @xmath2 is located at @xmath37 $ ] , so does the threshold of @xmath63 . let @xmath18 be a real number in @xmath1710,1/2[$ ] . according to proposition [ propprodtensgen ] , @xmath172 therefore , @xmath173 then , @xmath174 @xmath175 since the threshold width of @xmath3 is of order @xmath168 , @xmath176 recall that , by hypothese , @xmath1770,1[,\;\forall n\in \nn,\;\delta < p_{b,{\varepsilon}}<p_{b,1-{\varepsilon}}<1-\delta\;.\ ] ] thus , the fact that @xmath2 has a strongly homogeneous threshold width of order @xmath158 ( see definition [ defseuilhomogene ] ) implies that : @xmath178 therefore , the threshold of @xmath169 has a homogeneous width of order @xmath170 . + now , suppose that @xmath2 is located at @xmath37 $ ] . let @xmath18 be a real number in @xmath1710,1[$ ] . recall that @xmath1770,1[,\;\forall n\in \nn,\;\delta < p_{b,{\varepsilon}}<1-\delta\;.\ ] ] since @xmath179 , @xmath180 thus @xmath181 tends to @xmath34 as @xmath4 tends to infinity . this completes the proof . in this section , we show how to derive from proposition [ propprodtensseuil ] a large variety of threshold widths , ranging from @xmath1 to @xmath0 . to this end , we need some elementary building blocks the threshold of which are easy to study , and which we shall eventually combine in order to obtain the desired threshold width . these blocks will be taken from the reliability examples of section [ examples ] . + recall that for any @xmath182 , we denote by @xmath91 the following subset of configurations in @xmath8 ( see example [ exksurn ] ) : @xmath183 in the sequel , we shall use @xmath184 and @xmath185 for different values of @xmath4 , @xmath100 and @xmath101 . [ lemseuilsqrtn ] let @xmath186 . for every @xmath187 , @xmath188 moreover , @xmath2 has a strongly homogeneous threshold , located at @xmath106 , with a width of order @xmath0 . the simplest way to show that @xmath2 has a trheshold located at @xmath106 with a width of order @xmath0 is perhaps to use the concentration property of the binomial law . indeed , hoeffding s inequality @xcite ensures that : @xmath189 and @xmath190 let @xmath18 belong to @xmath1710,1[$ ] , and let @xmath191 , so that @xmath192 . if @xmath22 is such that @xmath193 , then @xmath194 can not be too far away from @xmath195 . inequality ( [ ineggdbingauche ] ) and ( [ ineggdbindroite ] ) imply that : @xmath196 therefore , the threshold of @xmath197 is located at @xmath106 : @xmath380,1[,\;p({\varepsilon})\xrightarrow[]{}\frac{1}{2}\;,\ ] ] and its threshold width is at most of order @xmath0 : @xmath1980,1/2[,\;\tau(a_{\lfloor n/2\rfloor , n},{\varepsilon})\leq 2\sqrt{\frac{\log \frac{1}{{\varepsilon}}}{2n}}\;.\ ] ] to see that this is the right order , one can express the derivative @xmath199 as follows : @xmath200 then , by cauchy - schwarz inequality , @xmath201 one can easily integrate this differential inequation . let us define : @xmath202 we have : @xmath2030,1[,\;j'(x)=\frac{1}{\sqrt{x(1-x)}}\;.\ ] ] therefore , integrating [ ineqdiff ] between @xmath22 and @xmath23 gives : @xmath204 when @xmath4 tends to infinity , @xmath22 and @xmath23 tend to @xmath106 . therefore : @xmath205 therefore , @xmath206 . finally , to prove the strong homogeneity of the width , we need a sharp minoration of @xmath199 . a smooth way to do this is to use one of the discrete isoperimetric inequalities la margulis - talagrand . the work of margulis @xcite has impulsed a number of more and more accurate discrete isoperimetric inequalities ( see @xcite ) . for example let @xmath207 denote the gaussian density , i.e @xmath208 , and @xmath209 the gaussian cumulative distribution , i.e @xmath210 . one can deduce from the main result in tillich and zmor @xcite that : @xmath211 where @xmath212 stands for @xmath213 . therefore , @xmath214 let @xmath160 and @xmath161 be two sequences of real numbers in @xmath1710,1[$ ] . integrating this inequality between @xmath215 and @xmath216 leads to : @xmath217 now , suppose that there exists @xmath2180,1[$ ] such that @xmath219 since @xmath212 is continuous and strictly positive on @xmath1710,1[$ ] , @xmath220 finally , @xmath221 and thus , the strong homogeneity of the threshold width of @xmath2 holds . @xmath154 now , consider a parallel - series system composed of @xmath100 blocks containing @xmath101 components ( see example [ exparser ] ) . the system is failed as soon as a block is failed , and a block fails if all its components are failed . the failure subset of such a system is @xmath222 . it is symmetric and monotone . one can easily derive explicitly its probability : @xmath223 thus , for any @xmath2240,1[$ ] , @xmath225 [ lemseuillogn ] for every @xmath226 , let @xmath227 be the following parallel - series failure subset : @xmath228 where @xmath229 stands for the logarithm to base 2 , and @xmath230 . the property @xmath227 has a sharp threshold located at @xmath106 , with a homogeneous width of order @xmath231 . more precisely , the threshold width of @xmath227 satisfies the following asymptotic expansion : @xmath232 let @xmath34 belong to @xmath1710,1[$ ] . we have : @xmath233 therefore , the threshold of @xmath227 is located at @xmath106 . moreover , for any @xmath2340,1/2[$ ] , @xmath235 since @xmath236 is equivalent to @xmath57 as @xmath57 tends to infinity , @xmath237 thus , @xmath227 has a homogeneous width of order @xmath231 . @xmath154 we are now able to prove our main result . [ thmlargeursvariees ] let @xmath59 be a nondecreasing sequence of integers such that : @xmath238 and suppose that : @xmath239 then , there exists an increasing sequence of integers @xmath240 and a monotone symmetric property @xmath241 whose threshold is located at @xmath106 , with a width of order @xmath242 . the condition @xmath243 implies that @xmath55 increases rather smoothly . because of the way we choosed to build the property mentioned in theorem [ thmlargeursvariees ] , we can not avoid this condition . of course , any condition of the type @xmath244 , with @xmath100 an integer strictly greater than 1 would be sufficient , since @xmath55 is nondecreasing . remark that in most natural choices of a nondecreasing function , this condition is fulfilled . nevertheless , one can build some `` unnatural '' examples where this condition fails . indeed , let @xmath245 be the increasing sequence of integers defined recursively by : @xmath246 remark that @xmath247 for every @xmath248 . let @xmath59 be the nondecreasing sequence of integers defined by : @xmath249,\;c(n)=\log(n)\;,\\ \forall n\in [ 2a_j , a_{j+1}-1],\;c(n)=\sqrt{2a_j}\;.\end{aligned}\ ] ] on one hand , @xmath250 and on the other hand , @xmath251 therefore , @xmath252 is not upperbounded . _ proof of theorem [ thmlargeursvariees ] _ : for any integer @xmath105 , let @xmath227 be the failure subset of a parallel - series system composed of @xmath253 blocks ocontaining @xmath254 components . suppose that @xmath255 . let us define the following integer : @xmath256 and the following monotone symmetric subset of @xmath257 : @xmath258 according to lemma [ lemseuilsqrtn ] , @xmath259 has a threshold located at @xmath106 with a strongly homogeneous width of order @xmath260 . from lemma [ lemseuillogn ] , @xmath261 has a threshold located at @xmath106 with a homogeneous width of order @xmath262 . therefore , one can deduce from proposition [ propprodtensseuil ] that @xmath263 has a threshold located at @xmath106 , with a homogeneous width of order @xmath264 . now , one can see that this device allows us to get nearly any order of threshold width between @xmath265 and @xmath266 . indeed , let @xmath267 be the following function : @xmath268&\rightarrow & \rr \\ x&\mapsto&x\left(\log\frac{n}{x}\right)^2\end{array}\right.\ ] ] the derivative of @xmath267 is easy to compute : @xmath269,\;\phi_n'(x)=\log\frac{n}{x}\times\log \frac{ne^{-2}}{x}\;.\ ] ] therefore , @xmath267 is a bijection from @xmath270 $ ] to @xmath271 $ ] . let @xmath59 be a sequence of integers such that : @xmath272 define @xmath273 . thus , @xmath274 let @xmath275 $ ] . the subset @xmath263 has a threshold width of order @xmath276 . it is clear that @xmath277 and @xmath59 are equivalent as @xmath4 tends to infinity . therefore , @xmath263 has a threshold width of order @xmath278 . furthermore , suppose now that @xmath55 is nondecreasing . since @xmath279 , we have : @xmath280 now , we suppose that @xmath281 . then @xmath282 . finally , @xmath263 has a threshold width of order @xmath242 . whence the result . when the localisation of the threshold of @xmath3 is not bounded away from 0 and 1 , proposition [ propprodtensseuil ] is useless in describing the threshold width of @xmath283 . moreover , one can say that its conclusion is not valid any longer . indeed , consider the failure subset @xmath227 defined in lemma [ lemseuillogn ] : @xmath284 one can easily compute the probability of @xmath285 : @xmath286 therefore , @xmath287 thus , when @xmath4 tends to infinity , @xmath22 is equivalent to @xmath288 , and @xmath289 is equivalent to @xmath290 . this is the typical example of a coarse threshold . similarly , we get : @xmath291 therefore , @xmath292 thus , when @xmath4 tends to infinity , @xmath22 is equivalent to @xmath293 , and @xmath294 is again equivalent to @xmath290 . + consequently , @xmath295 and @xmath296 have threshold widths of order , respectively @xmath297 and @xmath298 . according to lemma [ lemseuillogn ] , their product is a subset of @xmath299 , where @xmath300 , and has a threshold width of order @xmath231 . this is much bigger than the order @xmath301 which one would get if the conclusion of proposition [ propprodtensseuil ] remained valid . + nevertheless , this example witnesses an interesting phenomenon . the two subsets @xmath295 and @xmath296 of @xmath299 clearly have coarse thresholds , but their product has a sharp one . we shall prove this to be a very general behaviour : as soon as @xmath2 and @xmath3 have thresholds located respectively at 0 and 1 , even if these are coarse , their product @xmath63 has a sharp threshold . + to this end , we shall use a well known tensorisation property of the entropy . the major role of this property in concentration and threshold topics has been pointed out many times ( see for example @xcite ) . first , let us recall the definition of the entropy of a non negative function @xmath302 on a probability space @xmath303 : @xmath304 entropy satisfies the following tensorisation inequality ( see for instance ledoux @xcite , proposition 5.6 p. 98 ) : for every non negative function @xmath302 on @xmath8 , @xmath305 where @xmath306 means that only the @xmath119-th coordinate is concerned with the integration . the following lemma is the key towards the main result of this section , proposition [ propcoarsesharp ] . the following formula is easy to obtain , by considering the derivative of @xmath309 with respect to @xmath5 . for any real function @xmath302 on @xmath8 , @xmath310 where @xmath311 on the other hand , if @xmath2 is a monotone subset , @xmath312 therefore , @xmath313 remark that for any subset @xmath2 , @xmath314 thus , when applied to @xmath315 , equation ( [ eqtens ] ) gives @xmath316 now , remark that @xmath317 since @xmath318 when applied to @xmath319 , equation [ eqtens ] gives @xmath320 @xmath154 [ propcoarsesharp ] let @xmath164 and @xmath165 be two increasing sequences of integers , @xmath166 and @xmath167 be two non trivial monotone properties . suppose that the threshold of @xmath2 is located at 1 , and the one of @xmath3 is located at 0 : @xmath3210,1[,\;p_{a,{\varepsilon}}\xrightarrow[n\rightarrow + \infty]{}1\;,\ ] ] and @xmath3210,1[,\;p_{b,{\varepsilon}}\xrightarrow[n\rightarrow + \infty]{}0\;,\ ] ] then , the thresholds of @xmath63 and @xmath155 are sharp . first , let us remark that @xmath63 has a coarse threshold if and only if , for every @xmath2180,1[$ ] ( see friedgut and kalai @xcite or rossignol @xcite , p.99 ) , @xmath322{}+\infty\;.\ ] ] now , for all monotone subsets @xmath2 and @xmath3 , and every @xmath5 in @xmath1710,1[$ ] , we get from proposition [ propprodtensgen ] : @xmath323 let us denote by @xmath324 the quantity @xmath325 . we have therefore : @xmath326 according to lemma [ lementropie ] , @xmath327 thus , @xmath328 we shall focus on @xmath63 , the threshold of @xmath155 can be treated in the same way , by switching the roles of @xmath2 and @xmath3 . now , suppose that the threshold of @xmath3 is located at 0 . let @xmath18 belong to @xmath1710,1[$ ] . from equation ( [ eqprobatens ] ) , one get @xmath174 let us note @xmath329 . we get from inequality [ eqtensorparser ] : @xmath330 since @xmath331 tends to zero as @xmath4 tends to infinity , @xmath332 since @xmath333 is equivalent to @xmath334 as @xmath95 tends to 1 , if @xmath335 is bounded away from zero , inequality ( [ eqcoarsesharp1 ] ) implies that @xmath63 has a sharp threshold . if @xmath335 is not bounded away from zero , one need to show that the location of @xmath2 at 1 implies that @xmath336 is asymptotically negligible compared to @xmath337 . this follows from inequality ( [ eqtensp ] ) . indeed , integrating this inequality between @xmath338 and @xmath339 leads to : @xmath340 since @xmath339 tends to 1 as @xmath4 tends to infinity , @xmath341 is asymptotically negligible compared to @xmath337 . @xmath154	we investigate the threshold widths of some symmetric properties which range asymptotically between @xmath0 and @xmath1 . these properties are built using a combination of failure sets arising from reliability theory . this combination of sets is simply called a product . some general results on the threshold width of the product of two sets @xmath2 and @xmath3 in terms of the threshold locations and widths of @xmath2 and @xmath3 are provided .
the hubbard model is one of the basic models in the theory of strongly correlated systems . during its forty years of lifetime numerous approaches have been proposed for the investigation of the possible states of the system , the spectrum of its quasi - particles and the collective modes , the transport properties , the different types of ordered states and the phase transitions among them . such long period of development of a model which could look simple at a first glance since it contains only two parameters , the bare bandwidth @xmath6 and the on - site coulomb repulsion @xmath1 is determined by the circumstance that the case @xmath9 is of main physical significance . but just in this case the theory does not contain a small parameter . already the first researchers tried to avoid perturbative theories and used different non - perturbative approaches . starting from the pioneering works of hubbard @xcite @xcite @xcite , the method of decoupling of the double - time green s functions ( gf ) was treated successfully . the works based on projecting the equations of motion for the basic operators come here @xcite @xcite @xcite . the most productive application of this approach has been done with the method of composite operators @xcite @xcite @xcite @xcite used widely not only for the hubbard model but also for many other models @xcite of strongly correlated electronic systems . the method of the spectral density moments uses in essence the cut short of the equations of motions for the basic operators as well @xcite @xcite . also the variational method of gutzwiller belongs to the non - perturbative approaches @xcite , and made it possible to investigate qualitatively the behavior of a vast class of strongly correlated systems during the last four decades . the method of slave particles ( slave bosons ) represents an important direction of investigation also @xcite@xcite@xcite@xcite . the basic operators are expressed through a product of conventional fermi and bose operators with subsequent exclusion of unphysical states . the suitable choice of a slave particle representation makes it possible to catch the physics of low energy states in the scope of the mean field approximation . unfortunately there is no standard recipe for constructing such representations , and it is not always clear which one among the possible representations is the most adequate . during the last decade the method of the dynamical mean field theory ( dmft ) has become quite popular @xcite@xcite . by means of this method it has been possible to investigate the behavior of almost all the models in the theory of strongly correlated systems in the region of strong and intermediate interactions . apparently dmft is the most efficient method of investigation of these systems , although not exempt from some defects : it demands a huge amount of computations and has problems with the description of collective modes ( see the review @xcite ) . we do not mention here the numerical methods like quantum monte carlo and small cluster diagonalization , because we concentrate our efforts on the analytical approaches . we want to pay attention to one of the analytical approaches where there is a possibility to derive a consistent perturbative theory with respect to the parameter @xmath5 . definitely , such an approach corresponds to the perturbative theory near the atomic limit . the approach is based on the introduction of a generating functional @xmath10 $ ] , describing the interaction of the system with fluctuating fields depending on space and time . this functional corresponds to the generalization of the partition function of the system for the case of interactions with external fluctuating fields . for a proper choice of the @xmath11 operator the different gfs of the system are expressed through variational derivatives with respect to fluctuating fields . at the beginning this method was developed for a weak interaction by kadanoff and baym @xcite@xcite forty years ago . it could be generalized to a strongly correlated system when we express the hamiltonian through some basic operators taking into account the correlations ( for instance the hubbard @xmath0 operators)[3 ] instead of the conventional ones . the first time such an approach has been applied to the hubbard model was in the limit @xmath12 ( with an additional small parameter @xmath13 , where @xmath14 is the degeneracy of the electronic states ) @xcite . afterwards this approach has been developed farther in the works @xcite@xcite@xcite@xcite . recently we have provided a general framework for the generating functional approach ( gfa ) and we have applied it to a set of basic models of spin and strongly correlated electronic systems : heisenberg model , hubbard model for @xmath15 , @xmath16-model , @xmath17-model , double exchange model @xcite@xcite . the results of these investigations have been generalized in the monograph @xcite , published in russian , and in the course of lectures delivered in an international school @xcite . in this paper we apply the gfa to the hubbard model with a finite coulomb interaction @xmath1 . supposing that @xmath1 is large but of the order of @xmath6 we express the hamiltonian of the model in terms of the @xmath0 operators and calculate the electronic and bosonic gfs . the latter describes the two types of collective modes : magnons and doublons . the electronic gf is a matrix with respect to the spin index @xmath18 , the index @xmath19 , indicating the hubbard subbands , and the index @xmath20 corresponding to the particle - hole representation . we have derived the equation in the variational derivatives with respect to fluctuating fields for it . because the basic operators do not commute on @xmath21-values , the electronic green s function is characterized by two functions of four - momenta : the self - energy @xmath3 and the terminal part @xmath4 . for @xmath3 and @xmath4 the equations with the variational derivatives have been derived too , whereas it is possible to make iterations with respect to the parameter @xmath5 . just these iterative series represent the perturbative theory near the atomic limit @xcite . we have limited ourselves to the first and second order corrections for @xmath3 and extracted from them a mean field type @xmath22 part , which includes contributions depending only on the wave vector @xmath23 , but not on the frequency . @xmath22 consists of a term giving a shift to the hubbard subbands and renormalizing its width . the last term was extracted from the second order correction @xmath24 , which is an uncutable " term ( with respect to the hopping matrix element ) , while a cutable " term @xmath25 describes the dynamical interaction with boson - type excitations . a procedure of extraction of the static part from @xmath24 was borrowed from the composite operator method ( com ) @xcite@xcite@xcite@xcite . the main idea of this approach is that bosonic correlators , describing for example static fluctuations of charge , spin and pair , should not be calculated by some uncontrollable approximation ( like decoupling or use of the equation of motion ) , but must be determined by means of general properties of the electronic gf [ 10 ] . the gfa , restricted to the mean field approximation , and the com , restricted to a two - pole approximation , have a different structure for the electronic gf . in spite of this , the results obtained by these two methods for different properties of the hubbard model turned out to be in very good agreement . in particular , such mean field gfs give two quasiparticle subbands with a gap between them , which vanishes for half - filling at some critical value @xmath26 , and an insulator - metal phase transition occurs . detailed comparison of the mean field approximation in gfa and com will be discussed below . using the electron gfs we found , we can calculate bose - like gfs for plasmons , magnons and doublons . in this paper we study only doublons collective modes , describing motion of double occupied states of the lattice sites . the equation for the doublon gf has been derived . this equation contains variational derivatives of the electronic gf with respect to the corresponding fluctuating fields , coupled with charge densities . in the mean field approximation for the electronic @xmath3 we have obtained the closed equation for the doublon gf . for the paramagnetic state at half filling ( @xmath7 ) the doublon gf has a soft mode at momentum @xmath27 . it indicates a possible instability of the uniform state against a charge density wave formation . when the filling deviates from unity ( @xmath28 ) , the pole of the doublon gf has a gap @xmath29 , thus having the activative character . the content of the paper is the following . in part 2 , based on the @xmath0 operators formalism , the gfa is constructed . in part 3 it is derived the equation of motion for the electronic gf in the form of equation with variational derivatives . this equation is decoupled into two : one for the self energy and one for the terminal part . in part 4 the iterations of these equations with respect to the parameter @xmath5 are implemented and the gf in the hartree - fock approximation " is calculated . in part 5 we formulate a mean field approximation and compare gfa and com approaches . a bose - like gf for doublons is calculated in part 6 with the electronic gf taken in the mean field approximation . in part 7 we calculate the doublon susceptibility in the hydrodynamical regime . finally in part 8 we discuss the obtained results and propose suggestions for further study of the hubbard model . let us consider the conventional hubbard model for nondegenerate states . in terms of the fermi operators the model hamiltonian is @xmath30 where @xmath31 is the operator of annihilation ( creation ) of an electron on the site @xmath32 with spin @xmath18 , @xmath33 is the electron number on the same site with the given spin . under the condition of a strong on - site coulomb repulsion @xmath34 ( where @xmath35 is the hopping matrix element for the nearest neighbors and @xmath36 is the coordination number ) it is useful to express the hamiltonian ( [ eq:2.1 ] ) in terms of the @xmath0 operators . the operator @xmath37 for the site @xmath32 describes the transitions between the four possible states @xmath38 without any electron , with one electron possessing the spin projection @xmath18 or @xmath39 and a pair of electrons , respectively . the @xmath0 operators could be represented through the conventional fermi operators by means of the relations @xmath40 the operators @xmath41 and @xmath42 describe the correlated creation of an electron and are fermi - like @xmath43-operators ; @xmath44 and @xmath45 describe the flip of a spin on a site and the creation of a pair ; they are bose - like @xmath46-operators , respectively . the remaining @xmath0 s are called diagonal . we note that there are the hermitian - conjugate operators @xmath47 . the sixteen @xmath0 operators comprise thus the whole set , forming the algebra with the corresponding property of the product @xmath48 and the permutation relations of the anticommuting type for the @xmath43-operators while commutating for the @xmath46-operators . we note that the conventional fermi operators are expressed through the linear combinations of the @xmath0 operators of the @xmath43-type @xmath49 these relations express the motion of the correlated electrons in the two hubbard subbands . it is convenient to introduce the two - component spinors for the the @xmath43-operators : @xmath50 then the hamiltonian ( [ eq:2.1 ] ) is represented as @xmath51 , where @xmath52 @xmath53 here we added to hamiltonian ( 2.1 ) the term @xmath54 , where @xmath55 is the chemical potential and @xmath56 is the external magnetic field , that is why new notation appears : @xmath57 . in the quadratic form ( [ eq:2.7 ] ) @xmath58 represents the component of the spinor @xmath59 ; in addition we have introduced the matrix @xmath60 note that the index @xmath19 numerates the hubbard subbands . with the help of the rule of multiplication ( [ eq:2.3 ] ) for @xmath0 operators , one can write the permutation relations of the spinor @xmath43-operators : @xmath61_{+}= \delta_{ij}f_{i}^{\sigma}\\ & \left [ \varpsi(i\sigma)\otimes\varpsi^{\dag}(j\bar{\sigma } ) \right]_{+}= \delta_{ij}x_{i}^{\bar{\sigma}\sigma}\tau^{z}\\ & \left [ \varpsi(i\sigma)\otimes\varpsi(j\bar{\sigma } ) \right]_{+}= \delta_{ij}\sigma x_{i}^{02}(i \tau^{y } ) \end{split } \right\ } , \label{eq:2.9}\ ] ] where @xmath62 are the pauli matrices , and @xmath63 is a @xmath64 matrix , composed of @xmath0 operators : @xmath65 the permutation relations between @xmath43- and @xmath46-operators have a commutator character : @xmath66_{-}= \delta_{ij}\delta_{\sigma_{1}\sigma_{2 } } \varpsi(i\bar{\sigma_{1}})\\ & \left [ \varpsi(i\sigma_{1}),x_{j}^{20 } \right]_{-}= \delta_{ij}\bar{\sigma}_{1 } \varpsi^{\dag}(i\bar{\sigma}_{1})\tau^{x}\\ \end{split } \right\}. \label{eq:2.11}\ ] ] in other cases of permutations , relations of type ( [ eq:2.9 ] ) and ( [ eq:2.11 ] ) give zero . thus , an anticommutator of two @xmath67-operators is expressed either through a diagonal or a @xmath46-operator , but the commutator of @xmath67- and @xmath46-operators is naturally a @xmath67-operator . note the two relations @xmath68 @xmath69 which complete the algebra of the @xmath0 operators . let us write the equation of motion for the @xmath43-operator . for the thermodynamical time @xmath70 we start from the heisenberg equation @xmath71\,,\ ] ] which could be written in the case of the hamiltonian ( [ eq:2.6 ] ) ( [ eq:2.7 ] ) in the following form @xmath72 here a double - row matrix with respect to the spinor index was introduced @xmath73 here and in the following the numerical indexes indicate the four - dimensional coordinates including the site and the time @xmath74 , i.e. @xmath75 ; a summation over the primed indexes is understood ( it is a summation over the sites @xmath32 and an integration over the time @xmath74 ) . and finally the value @xmath76 has been introduced , representing the matrix over the spinor indexes ( the last circumstance has been specified by the symbol @xmath77 ) . thus the operator @xmath78 represents the linear combination of the @xmath43-operators , with the bosonic @xmath46-operators as the coefficients , and the matrixes @xmath79 and @xmath80 too . following the method we have applied many times to different quantum models @xcite@xcite@xcite@xcite , we introduce the generating functional @xmath81=\text{tr}\left(\mathrm{e}^{-\beta \mathcal{h}}t\mathrm{e}^{-v}\right ) \equiv\mathrm{e}^{\phi } , \label{eq:2.13}\ ] ] where @xmath82 is the symbol of the chronological product and the trace is taken over the whole set of variables of the system . for the hamiltonian ( [ eq:2.6 ] ) ( [ eq:2.7 ] ) it is convenient to choose the operator @xmath11 in the form @xmath83 it represents the linear combination of the whole diagonal and @xmath46-operators with the single point fields @xmath84 . thus , differentiating the equation @xmath10 $ ] with respect to the different @xmath84 s , we can express the different gfs through the variational derivatives with respect to the corresponding fields . for instance , for the single particle bose - like gfs of the plasmons , magnons and the doublons we have the expressions : @xmath85}\frac{\displaystyle\delta^{2}z[v ] } { \displaystyle\delta v_{1}^{\sigma_{1}\sigma_{1 } } \delta v_{2}^{\sigma_{2}\sigma_{2 } } } , \label{eq:2.20}\ ] ] @xmath86 @xmath87 here and further symbol @xmath88 , where @xmath89 means averading over gibbs ansamble with hamiltonian @xmath90 . having been introduced in such a way , the gfs are functionals of the fluctuating fields . directing these fields to zero after taking the variational derivatives , we shall obtain the actual gfs , describing our system . the fermionic gf can not be obtained by differentiation of @xmath10$](@xmath91 ) with respect to the single - point fields and it is necessary to determine the equation of motion for them . we make use of the general equation of motion ( see appendix ) and write it for the expression @xmath92 , determining the electronic gf : @xmath93 let us calculate now the anticommutator and the commutator of the @xmath67-operators in ( [ eq:3.1 ] ) . according to relations ( [ eq:2.9 ] ) and ( [ eq:2.11 ] ) , we have : @xmath94 @xmath95 here @xmath6 is the double - row matrix composed with the fluctuating fields : @xmath96 after the substitution of expression ( [ eq:2.14 ] ) and the commutators in equation ( [ eq:3.1 ] ) , the latter could be represented in the form : @xmath97 \nonumber\\ & + \bar{\sigma}_{1}v_{1}^{02}\tau^{x } ( \!(t\varpsi^{\dag}(\scriptstyle 1 \displaystyle\bar{\sigma}_{1 } ) \varpsi^{\dag}(\scriptstyle 2 \displaystyle\sigma_{2})e^{-v})\ ! ) \label{eq:3.5 } \\ & + ( \!(tf_{1}^{\sigma_{1 } } \hat{t}(\scriptstyle 11^{\prime } \displaystyle ) \varpsi(\scriptstyle 1^{\prime } \displaystyle\sigma_{1 } ) \varpsi^{\dag}(\scriptstyle 2 \displaystyle\sigma_{2})e^{-v})\ ! ) \nonumber\\ & + \tau^{z } \hat{t}(\scriptstyle 11^{\prime } \displaystyle ) ( \!(tx_{1}^{\bar{\sigma}_{1}\sigma_{1 } } \varpsi(\scriptstyle 1^{\prime } \displaystyle\bar{\sigma}_{1 } ) \varpsi^{\dag}(\scriptstyle 2 \displaystyle\sigma_{2})e^{-v})\ ! ) \nonumber\\ & + \sigma_{1}(\!(tx_{1}^{02 } \varpsi^{\dag}(\scriptstyle 1^{\prime } \displaystyle\bar{\sigma}_{1 } ) \varpsi^{\dag}(\scriptstyle 2 \displaystyle\sigma_{2})e^{-v})\ ! ) \hat{t}(\scriptstyle 1^{\prime}1 \displaystyle)i\tau^{y}. \nonumber\end{aligned}\ ] ] here the quantity @xmath98 has been introduced , which defines the zeroth - order approximation propagator of the electrons in the fluctuating single - point fields . this quantity is the @xmath64 matrix with respect to spinor indexes . expressing the mixed gfs through the variational derivatives of @xmath10 $ ] , we can represent the obtained equation as @xmath99 + \hat{a}_{1}(\sigma_{1}\sigma_{1}^{\prime } ) \hat{t}(\scriptstyle 11^{\prime}\displaystyle ) ( \!(t\varpsi(\scriptstyle 1^{\prime}\displaystyle\sigma_{1}^{\prime } ) \varpsi^{\dag}(\scriptstyle 2 \displaystyle\sigma_{2})e^{-v})\ ! ) \label{eq:3.7}\\ & -\sigma_{1}\left[v_{1}^{02}\tau^{x}\delta_{11^{\prime}}- i\tau^{y}\,\hat{\tilde{t } } ( \scriptstyle 11^{\prime } \displaystyle ) \frac{\displaystyle\delta}{\displaystyle\delta v_{1}^{02}}\right ] ( \!(t\varpsi^{\dag}(\scriptstyle 1^{\prime } \displaystyle\bar{\sigma}_{1 } ) \varpsi^{\dag}(\scriptstyle 2 \displaystyle\sigma_{2})e^{-v})\ ! ) . \nonumber\end{aligned}\ ] ] here the double - row matrix is the differential operator with respect to the single point fluctuating fields : @xmath100 where @xmath101also , let us note that @xmath102 is the transposed matrix of @xmath80 . as usual , we pass from the functional @xmath10 $ ] to the functional @xmath103 $ ] using the substitution : @xmath81=e^{\phi[v]}. \label{eq:3.10}\ ] ] then , the equation ( [ eq:3.5 ] ) results in a direct equation for the electronic gf : @xmath104 \langle t \varpsi(\scriptstyle 1^{\prime } \displaystyle\sigma_{1}^{\prime } ) \varpsi^{\dag}(\scriptstyle 2 \displaystyle\sigma_{2})e^{-v } \rangle \nonumber\\ = & -\delta_{12 } [ \hat{a}_{1}(\sigma_{1}\sigma_{2})\phi]- \sigma_{1}\delta_{11^{\prime}}v_{1}^{02 } \langle t \varpsi^{\dag}(\scriptstyle 1^{\prime } \displaystyle\bar{\sigma}_{1})\tau^{x } \varpsi^{\dag}(\scriptstyle 2 \displaystyle\sigma_{2})e^{-v } \rangle \label{eq:3.11}\\ & -\sigma_{1 } \left ( \frac{\delta\phi}{\delta v_{1}^{02}}+ \frac{\delta}{\delta v_{1}^{02 } } \right ) \langle t \varpsi^{\dag}(\scriptstyle 1^{\prime } \displaystyle\bar{\sigma_{1 } } ) \hat{t}(\scriptstyle 1^{\prime}1 \displaystyle ) i\tau^{y } \varpsi^{\dag}(\scriptstyle 2 \displaystyle\sigma_{2})e^{-v } \rangle . \nonumber\end{aligned}\ ] ] we see that the equation for the gf @xmath105 contains the anomalous gf @xmath106 . then , it is necessary to write the equation for it , too . let us introduce the matrix gf : @xmath107 the underlined numerical index @xmath108 in the left part represents the cumulative index , containing the space - time point @xmath109 , the spin @xmath110 , the spinor index @xmath111 and one more index @xmath112 , accepting two values , specifying the matrix elements ( [ eq:3.12 ] ) , so that @xmath113 the matrix @xmath114 is an @xmath115 matrix with respect to the collection of the discrete indexes . a matrix of such a rank appears automatically in the hubbard model . its arising is described `` normal '' states ( without the cooper s pairs ) but also with broken symmetries as well . the set of four equations for the gfs in ( [ eq:3.12 ] ) could be written as a single matrix equation : @xmath116 \mathcal{l } ( \scriptstyle\underline{1}\,^{\prime}\underline{2}\,\,\displaystyle)= ( \hat{a}\phi ) ( \scriptstyle\underline{1}\,\underline{2}\,\displaystyle ) . \label{eq:3.14}\ ] ] here we introduced the operator matrix @xmath117 where each element represents the @xmath64 matrix with respect to the spinor indexes , hidden in the pauli matrices and the matrix @xmath118 , having the variational derivatives with respect to the fluctuating fields as its elements . besides , the equation ( [ eq:3.14 ] ) contains the matrix @xmath119 the value @xmath120 represents the double - row matrix @xmath121 where @xmath122 is given by the expression ( [ eq:3.6 ] ) , and @xmath123 by its transposition : @xmath124 if replace back in equation ( [ eq:3.14 ] ) term with by the mixture gfs one can see that the matrix equation ( [ eq:3.14 ] ) is equivalent that derived by plakida @xcite . the equation ( [ eq:3.14 ] ) is of the same type of the equation for the single particle gf , that we derived for the hubbard model in the limit @xmath125 @xcite and for the heisenberg model as well . in the above models the matrix @xmath126 degenerated into a scalar , but now it is a matrix with respect to the discrete indexes @xmath19 and @xmath20 , likewise the other values in ( [ eq:3.15 ] ) . by virtue of the noted similarity of the equation ( [ eq:3.14 ] ) with the respective equations of the models considered before we could expect the same structure in the solutions of these equations , in particular the multiplicative character of the electronic gfs . let us represent them as a product of the propagator @xmath127 and the terminal @xmath128 parts , respectively , namely : @xmath129 the propagator part satisfies the dyson equation @xmath130 let us represent the equation for the self - energy part like the sum of the two terms : @xmath131 which took place for the models considered before . then , inserting ( eq:3.19 ) and ( [ eq:3.20 ] ) in ( [ eq:3.18 ] ) and comparing with the initial equation ( [ eq:3.14 ] ) , we can obtain the two equations for @xmath132 and @xmath133 : @xmath134 @xmath135 in obtaining these equation we have taken into account the identity @xmath136 l(\scriptstyle\underline{3}\,^{\prime}\underline{2}\,\displaystyle ) , \label{eq:3.23}\ ] ] which is the generalization of the well known identity expressing the differentiation of a gf through the differentiation of its inverse : @xmath137 the equations ( [ eq:3.21 ] ) and ( [ eq:3.22 ] ) for the terminal and self - energy parts of the gf have a structure analogous to the respective equations of the other models . these are the equations for the variational derivatives for @xmath128 and @xmath133 . the contribution @xmath138 in the self - energy part @xmath3 is not cutable through the line of the interaction " , representing the value @xmath139 . the cutable part @xmath140 has been already extracted in the equation ( [ eq:3.20 ] ) like the second contribution . from the set of the equations ( [ eq:3.18 ] ) ( [ eq:3.20 ] ) it follows an important consequence , which could be represented in the form of the following equation for the gf @xmath141 : @xmath142 here @xmath143 is determined by the two relations : @xmath144 the solution of the equation ( [ eq:3.24 ] ) could be written as : @xmath145 ^{-1 } ( \scriptstyle\underline{1}\,\underline{2}\,\displaystyle ) , \label{eq:3.25}\ ] ] where @xmath146 as it follows from the definition , the value @xmath147 is not cutable through the line @xmath139 , therefore the equation ( [ eq:3.24 ] ) for the gf is the larkin s equation , expressing a gf through an irreducible part ( with respect to a line of interaction " ) . from this equation it follows the locator representation ( [ eq:3.25 ] ) for the electronic gf , also . so , this issue is a diagrammatic justification of the multiplicative representation ( [ eq:3.18 ] ) for one - particle electron gf . similar representations for one - particle gfs in other models of strongly correlated electron and spin systems was discussed in details in a review @xcite . the equations ( [ eq:3.21 ] ) and ( [ eq:3.22 ] ) could be solved by iterations . at the first orders with respect to @xmath139 we obtain : @xmath148 @xmath149 \label{eq:3.28}\ ] ] @xmath150 l(\scriptstyle\underline{2}\,^{\prime}\underline{5}\,^{\prime}\displaystyle ) \left [ \hat{a } ( \scriptstyle\underline{3}\,^{\prime}\underline{6}\,^{\prime}\displaystyle ) l_{0v}^{-1 } ( \scriptstyle\underline{5}\,^{\prime } \underline{2}\,\displaystyle ) \right]+ ... \ , , \nonumber\ ] ] in ( [ eq:3.27 ] ) the operator @xmath126 , acting on @xmath91 , brings the mean value of the diagonal and @xmath46-operators ; a repeated action of the operator @xmath126 will produce bosonic gfs of the different types . an action of the operator on @xmath151 will result in expressions composed of different @xmath152-symbols . the problem is contained in the multiplication of the matrices in the equations ( [ eq:3.27 ] ) and ( [ eq:3.28 ] ) , taking into account that the matrix @xmath153 contains derivatives , which should act on the corresponding values . to fulfil the matrix multiplication accounting for the operator character of the several factors , we rewrite the expressions ( [ eq:3.27 ] ) and ( [ eq:3.28 ] ) in another form : @xmath154 @xmath155 @xmath156 in these expressions all the factors are arranged in the order of the matrix multiplication , but we should not forget which factors the derivatives of the matrix @xmath126 act on . according to definition ( [ eq:3.12 ] ) , the electronic gf @xmath141 takes into account the possibility of states with coupled electrons . in this paper we shall consider the normal system , described completely by the matrix element of the electronic gf @xmath157 , namely @xmath158 the normal gf @xmath2 can be looked in the standard multiplicative form @xmath159 with @xmath160 obeying the dyson equation @xmath161 and the self - energy part being a sum of two terms , uncutable @xmath133 and cutable @xmath162 : @xmath163 in equations ( [ eq:4.2 ] ) ( [ eq:4.4 ] ) all quantities are @xmath64 matrices with respect to spinor indexes , with arguments of the type @xmath164 . iterations in general equations ( [ eq:3.21 ] ) and ( [ eq:3.22 ] ) allow to get series for @xmath133 and @xmath4 , determined by equations ( [ eq:4.2 ] ) ( [ eq:4.4 ] ) . calculations of these series are done in appendix b , and here we present the results within the limit of the first two orders . we have , for the zeroth order of @xmath4 : @xmath165 where @xmath166 is the average number of electrons on a site with spin @xmath18 . the first order correction for @xmath4 is the following @xmath167 where @xmath168 , \nonumber\end{aligned}\ ] ] @xmath169 . \nonumber\end{aligned}\ ] ] the quantities @xmath170 , @xmath171 and @xmath172 are the fourier transforms of the bosonic gfs , determined by relations ( [ eq:2.20 ] ) ( [ eq:2.22 ] ) with 4-momentum @xmath173 . here @xmath174 is the fourier transform of @xmath175 , which is actually the bare electron energy in the lattice . the contribution of the first order in @xmath133 is given by : @xmath176 where @xmath177 . \label{eq:4.11}\ ] ] the second order correction is equal to : @xmath178 where @xmath179 , \nonumber\end{aligned}\ ] ] and the quantity @xmath180 is given by a change of spinor indexes @xmath181 in ( [ eq:4.13 ] ) . here @xmath182 is a linear combination of the matrix elements of the electronic gf : @xmath183 finally we write down the second order contribution in the cutable part of @xmath3 , that is , the expression for @xmath184 . because in momentum representation @xmath185 is equal to @xmath186 , with @xmath187 being the @xmath64 matrix determined in ( [ eq:2.8 ] ) , we find , according to ( [ eq:4.8 ] ) : @xmath188 where @xmath189 , \nonumber\end{aligned}\ ] ] @xmath190 and @xmath191 we see that the correction @xmath192 depends neither on momentum nor on frequency and determines only a shift of electron spectrum , but it depends on spin . the second order corrections @xmath193 and @xmath194 depend both on momentum and frequency . the contribution @xmath195 is determined by the interaction of electrons with bosonic excitations , while @xmath196 is determined by electronic gfs only . the simplest approximation of a mean field type is the hubbard - i , which takes into account a term in @xmath3 equal to @xmath197 . to it , one can add a first order term @xmath198 , not depending on frequency . the second order correction @xmath24 depends on the frequency , however we shall try to extract from it a static part by the following ansatz . let us consider that in @xmath24 both expressions for @xmath199 , @xmath180 include a factor @xmath200 in the summation over @xmath201 . so it can be factorized in the nearest neighbor approximation , and a term proportional to @xmath174 can be taken out from the static part of @xmath24 for the cubic lattice . thus in the static approximation @xmath24 can be approximated by the expression @xmath202 here @xmath203 and @xmath204 are some spin dependent constants . their expressions can be explicitly written out , but we will not do it , because we shall try to calculate them from some general conditions for electronic gfs , which should be satisfied . such conditions were formulated in works by mancini and coworkers ( see general discussion in paper @xcite and refs . therein ) , where it is developed a method using linearized equation of motion for composite operators . the condition is demanding that the electronic gf @xmath205 is equal to zero when arguments coincide . below we will use this idea for the determination of unknown parameters @xmath203 and @xmath204 . first we write down the self - energy part in an approximation which includes the hubbard - i term , the first order correction @xmath133 ( [ eq:4.10 ] ) and @xmath24 in the form ( [ eq:5.1 ] ) . all these three contributions give @xmath22 , corresponding to a mean field approximation . so we have : @xmath206 it is clear that the first term is responsible for a shift of the hubbard subbands , and the second one for a renormalization of their widths . the propagator part of the gf in the mean field approximation is determined by a matrix equation : @xmath207^{-1}= \left[g_{0}^{\sigma}(k)\right]^{-1}- \sigma_{mf}^{\sigma}(k).\ ] ] we look for a solution of the form @xmath208 the poles @xmath209 and their residues @xmath210 are written in the form : @xmath211 \\ & ( \mathcal{a}_{\scriptscriptstyle 1,2 \displaystyle } ^{\sigma})_{22}(\bm{k})= \frac{\displaystyle 1}{\displaystyle 2 } \left [ 1 \mp \frac{\displaystyle r^{\sigma}(\bm{k } ) } { \displaystyle 2q^{\sigma}(\bm{k } ) } \right ] \\ & ( \mathcal{a}_{\scriptscriptstyle 1,2 \displaystyle } ^{\sigma})_{12}(\bm{k})= \mp \frac{\displaystyle \eta^{\sigma } + ( 1-\langle n^{\bar{\sigma}}\rangle+p_{2}^{\sigma } ) \varepsilon(\bm{k } ) } { 2q^{\sigma}(\bm{k } ) } \\ & ( \mathcal{a}_{\scriptscriptstyle 1,2 \displaystyle } ^{\sigma})_{21}(\bm{k})= \mp \frac{\displaystyle \eta^{\sigma } + ( \langle n^{\bar{\sigma}}\rangle - p_{1}^{\sigma } ) \varepsilon(\bm{k } ) } { 2q^{\sigma}(\bm{k } ) } \end{split } \right\ } , \label{eq:5.4}\ ] ] @xmath212 here @xmath213 \varepsilon(\bm{k}),\ ] ] whilst expressions for @xmath214 and @xmath215 will be written later . the electronic gf @xmath216 in the mean field approximation is found with the help of the general relation ( [ eq:4.2 ] ) @xmath217 where @xmath218 is given by the matrix ( [ eq:4.5 ] ) . the electronic gf depends on parameters @xmath55 , @xmath219 , @xmath220 , @xmath203 and @xmath204 , which must be determined in a self consistent way from the equations @xmath221 and also from equation ( [ eq:4.11 ] ) , determining the parameter @xmath219 . parameters @xmath203 and @xmath204 will be determined [ 10 ] from conditions which follow from the properties of @xmath0 operators , namely : @xmath222 thus a complete system of equations for all five parameters can be written in the form : @xmath223 @xmath224 , \label{eq:5.8}\ ] ] @xmath177 , \label{eq:5.9}\ ] ] @xmath225 @xmath226 ( we assume homogeneous states , so all averages do not depend on site index ) . from the comparison of the last two equations we find a relation between parameters @xmath203 and @xmath204 : @xmath227 therefore , the parameter @xmath204 can be replaced in all the expressions above . thus the equations ( [ eq:5.4 ] ) for the residues of gf are : @xmath228 \\ & ( \mathcal{a}_{\scriptscriptstyle 1,2 \displaystyle } ^{\sigma})_{22}(\bm{k})= \frac{\displaystyle 1}{\displaystyle 2 } \left [ 1 \mp \frac{\displaystyle u } { \displaystyle 2q^{\sigma}(\bm{k } ) } \right ] \\ ( \mathcal{a}_{\scriptscriptstyle 1,2 \displaystyle } ^{\sigma})_{12}&(\bm{k})= ( \mathcal{a}_{\scriptscriptstyle 1,2 \displaystyle } ^{\sigma})_{21}(\bm{k})= \mp \frac{\displaystyle \eta^{\sigma } + ( \langle n^{\bar{\sigma}}\rangle - p_{1}^{\sigma } ) \varepsilon(\bm{k } ) } { 2q^{\sigma}(\bm{k } ) } \end{split } \right\}. \label{eq:5.13}\ ] ] expressions for @xmath214 and @xmath215 , determining poles , are now equal to : @xmath229^{2 } } \end{split } \right\}. \label{eq:5.14}\ ] ] after the replacement of parameter @xmath204 , the two equations ( [ eq:5.10 ] ) and ( [ eq:5.11 ] ) reduce to only one , which allows to find the unknown parameter @xmath203 . taking the summation over frequencies in all equations ( [ eq:5.8 ] ) ( [ eq:5.11 ] ) , we write our system in the form : @xmath230 @xmath231 @xmath232 where we use the definitions of the paper [ 8 ] : @xmath233 , \label{eq:5.18}\ ] ] @xmath234 . \label{eq:5.19}\ ] ] we have to add to them equations ( [ eq:5.15 ] ) ( [ eq:5.17 ] ) and equation ( [ eq:5.7 ] ) for chemical potential . the energy of the system can be found by averaging the hamiltonian ( [ eq:2.6 ] ) ( [ eq:2.7 ] ) over a gibbs ensemble . it is quite easy to express it by means of electronic gfs : @xmath235 where @xmath236 after substituting here the expressions for the matrix elements @xmath237 , we find the expressions for the energy @xmath238 and the double occupation parameter @xmath239 : @xmath240,\nonumber\end{aligned}\ ] ] @xmath241 in fig . 1 the parameters @xmath242 and @xmath243 are plotted as functions of electron concentration at different @xmath1 . such results are typical for other fixed parameters of the system . for all different @xmath244 and @xmath1 the parameter @xmath242 is positive and @xmath243 is negative . a negative solution for parameter @xmath242 was not found . the behavior of @xmath242 is rather similar to the com1 solution for the parameter @xmath245 in works @xcite@xcite ( com1 is a name authors @xcite@xcite gave for the solution with @xmath246 ) . in fig.2 the concentration dependence of chemical potential is given for two values of @xmath1 . in the same figure a com1 solution , that we found from equations of paper @xcite , is presented for two variants of density of states for the bare electron band : a two - dimensional square lattice with nearest - neighbor hopping and a model density of states of this type : @xmath247 we see that com and our gfa give similar results . the com1 solution for the 2d - system and for the model density of states are quantitatively very close , and because of this we shall use hereafter for simplicity the model density of states ( [ eq:5.24 ] ) . slightly worse is the comparison of results for @xmath248 ( fig.3 ) , however there is a qualitative coincidence of com1 and gfa calculations . the parameter of double occupation @xmath239 gives again a satisfactory coincidence of the two approaches ( fig.4 ) . it is useful to show the dependence of @xmath248 on @xmath244 in a whole electron concentration interval at different values of @xmath1 ( fig.5 ) . when decreasing @xmath1 a part of the curve denoted by dash lines approaches to the abscess line , and when @xmath249 one can see that @xmath250 , as it should be in the case of noninteracting electrons . calculations show that when @xmath1 decreases , the value of the jump of the chemical potential at @xmath7 decreases too , and at the same critical value @xmath251 becomes equal to zero . this corresponds to the closing up the two hubbard subbands , and the insulator - metal phase transition occurs . the evolution of the density of states of the quasiparticle spectrum when @xmath1 changes is shown for two different bare density of states : the model one ( [ eq:5.24 ] ) , fig.6 , and the semielliptic one @xmath252 fig.7 . in com1 the critical value is @xmath253 @xcite , which is close to our value @xmath251 , obtained for the density of states ( [ eq:5.24 ] ) . at half - filling it is easy to get an expression for the gap between the two hubbard subbands with energies @xmath254 and @xmath255 : @xmath256 from here follows the critical value @xmath257 , when @xmath258 . it is equal to @xmath259 so that when @xmath260 the system is an insulator , and when @xmath261 a metal . compare now the two approaches for the hubbard model : gfa and com . the mean field approximations in the framework of these approaches are close to each other both as what regards the gfs structure and physical properties of the model calculated with their help . in both cases the electronic gf has a two - poles structure . the com approach includes only the parameter @xmath245 , which has to be found from the equation @xmath262 . in the gfa two parameters , ( @xmath242 and @xmath243 ) , appear , determined through two equations : @xmath262 and @xmath263 . due to this pair of equations one of these parameters can be eliminated , and as a result we have only one parameter , @xmath242 . the physical meanings of the parameters @xmath245 and @xmath242 are close . in the com approach the parameter @xmath245 describes the static fluctuation of charge , spin and pair . in gfa the parameter @xmath242 includes traces of static charge and spin fluctuations as well . corrections for the self - energy due to dynamical interaction of electrons with bosons in both approaches practically coincide and correspond to scba . the equations for the determination of parameters @xmath55 , @xmath264 , @xmath248 , @xmath242 in gfa and @xmath55 , @xmath264 , @xmath265 , @xmath245 in com are rather similar , but have different solutions . in com at fixed external parameters ( @xmath244 , @xmath1 , @xmath6 ) one has two solutions : with @xmath246 and @xmath266 , while in gfa there is only one solution with @xmath267 ( the second parameter @xmath243 is always negative , but it does not enter in the electronic gf explicitly ; but it only guarantees the satisfaction of the two conditions @xmath262 and @xmath263 , simultaneously ) . a remarkable conclusion follows from numerical calculations with different sets of external parameters . in spite of some difference in gfa and com , the calculated quantities of the model are rather close to each other , if in com only com1 solutions with @xmath246 are taken into account . the two - poles gf of this approximation can be used farther for the calculation of corrections to the self - energy @xmath268 from the dynamical fluctuations @xcite and for bosonic gfs ( magnon , plasmon , doublon ) , describing these fluctuations @xcite . the complete system of 16 @xmath0 operators contains two bose - like operators @xmath269 and @xmath270 ( and their conjugates @xmath271 and @xmath272 ) , which determine the two bose - like gfs ( 2.15 ) and ( 2.16 ) . they describe propagation of a spin flip ( magnon ) and a dyad ( doublon ) , representing the two types of the bose - like collective modes . these gfs could be represented as the variational derivatives of @xmath10 $ ] with respect to the fluctuating fields : @xmath273 @xmath274 to write the equations of motion for the gfs @xmath275 and @xmath276 we need the equations of motion for the bose - like operators : @xmath277 @xmath278 we see that in the right hand sides of these relations @xmath67-operators occur ; therefore in the corresponding equations of motions for the magnon and the doublon gfs the @xmath82-mixed product of @xmath43- and @xmath46-operators will appear . they could be represented as the the variational derivative of the electronic gf with respect to the fluctuating field @xmath279 in the first case and @xmath280 in the second . one of the important feature of the doublon gf is that it includes the anomalous " electronic gf , composed of the two operators @xmath281 and @xmath282 , while the equation for the magnon gf should include the normal electronic gf , composed of the operators @xmath283 and @xmath284 . by itself these anomalous gfs are equal to zero when the fields are absent , however their derivatives with respect to the fields @xmath279 and @xmath280 are not equal to zero and determine the contribution in the equation of motion , caused by the interactions of the electronic and bosonic degrees of freedom . now our task is to determine the equations of motion for the magnon and doublon gfs and to obtain their approximate solution . this will let us determine the spectrum of the corresponding collective modes . in this paper we study only the doublon gf . let us derive the equation of motion for the doublon gf ( [ eq:2.22 ] ) ; to this purpose we write the equation of motion for the mean value of the operator @xmath270 : @xmath285 we substitute in it the expression ( [ eq:6.4 ] ) for @xmath286 , and also the relation @xmath287 then our initial equation could be rewritten in the form : @xmath288 where @xmath289 taking into account the definition of the electronic gf we write its last term in the form @xmath290,\ ] ] where @xmath291 is the anomalous component of the electronic gf . the mean values @xmath292 of @xmath0 operators are expressed through the variational derivative of the functional @xmath103 $ ] , and we come to the final form of the equation for the generating functional : @xmath293 . \label{eq:6.8}\ ] ] in the same way it is possible to write the equation for @xmath294 and reduce it to the form @xmath295 . \label{eq:6.9}\ ] ] differentiating now the equation ( [ eq:6.8 ] ) with respect to @xmath296 , and the equation ( [ eq:6.9 ] ) with respect to @xmath297 , we come to the pair of conjugate equations for the doublon gf : @xmath298 . \label{eq:6.10}\ ] ] @xmath299 . \label{eq:6.11}\ ] ] here we introduced the number of electrons on the site , @xmath300 , where @xmath301 . we see that the exact equations for the doublon gf contain terms with variational derivatives of anomalous electronic gf with respect to the fields @xmath302 and @xmath303 . to obtain a close equation for doublon gf we have to calculate these terms by the same approximate way . let us calculate the derivative of off - diagonal ( with respect to the upper spinor indexes ) electronic gfs @xmath304 and @xmath305 . we use the multiplicative representation ( [ eq:3.18 ] ) . in the normal state we could use the expression for the variational derivative . @xmath306^{12 } } { \delta v_{2}^{20 } } \mathcal{l}^{22 } ( \scriptstyle 4^{\prime } 4\displaystyle ) + l^{11}(\scriptstyle 33^{\prime}\displaystyle ) \frac{\delta\pi^{12 } ( \scriptstyle 3^{\prime } 4\displaystyle ) } { \delta v_{2}^{20 } } \label{eq:6.12}\ ] ] we take the inverse propagator gf @xmath307 in the approximation , when in the general expression ( [ eq:3.20 ] ) the term @xmath308 is neglected , and @xmath128 is taken in the zeroth order approximation . it is easy to obtain the relations : @xmath309 ^{12 } } { \delta v_{2}^{20}}= \sigma _ { 3}\delta _ { \bar{\sigma}_{3}\sigma _ { 4 } } \delta_{23}\delta _ { 34}\tau ^{x},\ ] ] @xmath310 then , within the first order approximation with respect to @xmath35 the eq . ( [ eq:6.10 ] ) is determined by the expression @xmath311 \mathcal{d}^{02 } ( \scriptstyle 3^{\prime } 2 \displaystyle ) \biggr\}. \nonumber\end{aligned}\ ] ] after substituting this relation into the equation ( [ eq:6.10 ] ) , we represent equation for the doublon gf in the form @xmath312 \mathcal{d}^{02 } ( \scriptstyle 1^{\prime}2 \displaystyle)= ( 1-n_{1})\delta _ { 12}+\mathcal{p}_{l}^{02 } ( \scriptstyle 12 \displaystyle ) , \label{eq:6.14}\ ] ] @xmath313 , \label{eq:6.15}\ ] ] @xmath314 . \label{eq:6.16}\ ] ] the index @xmath315 of the terminal and the self - energy part indicates the left " form of the equations for @xmath276 . in the same way starting from the equation ( [ eq:6.11 ] ) , it is possible to come to the right " form of the equation for @xmath276 : @xmath316 = ( 1-n_{1})\delta _ { 12 } + \mathcal{p}_{r}^{02 } ( \scriptstyle 12 \displaystyle ) , \label{eq:6.17}\ ] ] @xmath317 , \label{eq:6.18}\ ] ] @xmath318 . \label{eq:6.19}\ ] ] to recover the symmetry of the doublon gf let us symmetrize the equations ( [ eq:6.14 ] ) and ( [ eq:6.17 ] ) making their sum . then , the doublon gf is equal to @xmath319 where @xmath320 , \qquad \mathcal{p}^{02}(q)= \frac{\displaystyle1}{\displaystyle2}\ , [ \mathcal{p}_{l}^{02}(q)+\mathcal{p}_{r}^{02}(q)].\ ] ] the self - energy and the terminal part are equal to @xmath321- \mathrm{tr}[\im\,\widetilde{g}^{\sigma}(k - q)\,i\tau^{y } ] \biggr\ } \nonumber \\ + \frac{\displaystyle 1}{\displaystyle 2 } \sum\limits_{k\sigma } & \varepsilon(\bm{k } ) \biggl\{\varepsilon(\bm{k - q } ) \mathrm{tr}[\im\,g^{\sigma}(k)\,i\tau^{y}\ , \im\,\widetilde{\mathcal{g}}^{\bar{\sigma}}(k - q ) ] \label{eq:6.21 } \\ + & \varepsilon(\bm{k } ) \mathrm{tr}[i\tau^{y}\,\im\,\mathcal{g}^{\sigma}(k)\ , \im\,\widetilde{g}^{\bar{\sigma}}(k - q ) ] \biggr\ } , \nonumber\end{aligned}\ ] ] @xmath322 + \mathrm{tr}[\im\,\widetilde{g}^{\sigma}(k - q)\,\tau^{x}\ , \mathcal{g}^{\bar{\sigma}}(k ) ] \biggr\}. \label{eq:6.22}\ ] ] in the same way we can calculate the doublon gf @xmath323 . it is possible to represent the result of the computation in the form @xmath324 where the values @xmath325 and @xmath326 are expressed through @xmath327 and @xmath328 : @xmath329 thus we see that the condition of symmetry is fulfilled @xmath330 or @xmath331 in the coordinate space , which follows directly from the definition ( [ eq:6.2 ] ) for the doublon gf . after the computation of the trace in the expressions ( [ eq:6.21 ] ) and ( [ eq:6.22 ] ) , we can represent them in the form : @xmath332 \label{eq:6.26 } \\ + \frac{\displaystyle 1}{\displaystyle 2 } \sum\limits_{k\sigma } \varepsilon(\bm{k } ) & \biggl [ \varepsilon(\bm{k - q})g^{\sigma}(k ) \sum\limits_{\alpha\beta } \widetilde{\mathcal{g}}_{\alpha\beta}^{\bar{\sigma}}(k - q)+ \varepsilon(\bm{k})\widetilde{g}^{\sigma}(k - q ) \sum\limits_{\alpha\beta } \mathcal{g}_{\alpha\beta}^{\bar{\sigma}}(k)\biggr ] . \nonumber\end{aligned}\ ] ] @xmath333\ , [ \mathcal{g}_{12}^{\bar{\sigma}}(-k+q)+ \mathcal{g}_{22}^{\bar{\sigma}}(-k+q ) ] \label{eq:6.27 } \\ + & [ g_{12}^{\sigma}(k)+g_{22}^{\sigma}(k)]\ , [ \mathcal{g}_{11}^{\bar{\sigma}}(-k+q)+ \mathcal{g}_{21}^{\bar{\sigma}}(-k+q ) ] \nonumber\\ + & [ g_{11}^{\sigma}(-k+q)+g_{12}^{\sigma}(-k+q)]\ , [ \mathcal{g}_{21}^{\bar{\sigma}}(k)+ \mathcal{g}_{22}^{\bar{\sigma}}(k ) ] \nonumber\\ + & [ g_{21}^{\sigma}(-k+q)+g_{22}^{\sigma}(-k+q)]\ , [ \mathcal{g}_{11}^{\bar{\sigma}}(k)+ \mathcal{g}_{12}^{\bar{\sigma}}(k ) ] \biggr\}. \nonumber\end{aligned}\ ] ] now we calculate the expression ( [ eq:6.26 ] ) in the mean field approximation for the electronic gf . substituting here the formula ( [ eq:5.3 ] ) and summing over frequencies , we write the result as a sum of two contributions of the first and second order with respect to @xmath35 : @xmath334 where @xmath335- f\left[e_{\scriptscriptstyle 2\displaystyle}^{\sigma}(\bm{k } ) \right ] \biggr ] , \label{eq:6.28}\ ] ] @xmath336- f\left[e_{\scriptscriptstyle m \displaystyle}^{\bar{\sigma}}(\bm{k - q } ) \right]}{i\omega_{\nu}- e_{\scriptscriptstyle n \displaystyle}^{\sigma}(\bm{k})- e_{\scriptscriptstyle m \displaystyle}^{\bar{\sigma}}(\bm{k - q})}. \label{eq:6.29}\end{aligned}\ ] ] where @xmath337 @xmath338(\bm{k } ) \biggl\{\bigl[(\mathcal{a}_{m}^{\bar{\sigma}})_{11}+ ( \mathcal{a}_{m}^{\bar{\sigma}})_{21}\bigl](\bm{k } ) \bigl(\lambda_{11}^{\bar{\sigma}}+ \lambda_{12}^{\bar{\sigma}}\bigl ) + \nonumber \\ + & \bigl[(\mathcal{a}_{m}^{\bar{\sigma}})_{12}+ ( \mathcal{a}_{m}^{\bar{\sigma}})_{22}\bigl](\bm{k } ) \bigl(\lambda_{21}^{\bar{\sigma}}+ \lambda_{22}^{\bar{\sigma}}\bigl ) \biggl\ } \label{eq:6.30}\end{aligned}\ ] ] and here @xmath339 , @xmath340 are determined by formulas ( [ eq:5.13 ] ) and ( [ eq:4.5 ] ) remarkable is the fact that expressions @xmath341 and @xmath342 vanish at wave vector @xmath8 . because @xmath328 is nothing but the self - energy of a doublon , we see from eq . ( [ eq:6.23 ] ) , that at half - filling , when @xmath343 , a doublon is a soft mode in the vicinity of the point @xmath344 . this observation pushes to study its dispersion law and attenuation . we postpone the study of doublons at arbitrary electron concentration and fix ourselves on the case @xmath7 . we are limited now to the hydrodynamical regime . it is well known that collective modes in a disordered ( symmetrical ) phase in the hydrodynamical regime are ruled by the conservation laws @xcite . thus the spin gf @xmath275 should be determined by the total spin conservation law , while the pseudospin gf @xmath276 is determined by the pseudospin conservation law @xcite@xcite@xcite@xcite . the three pseudospin components @xmath345 with @xmath346 obey permutation relations @xmath347=(2\mu - u)p^{+},\quad [ p^{-},\mathcal{h}]=-(2\mu - u)p^{-},\quad [ p^{z},\mathcal{h}]=0 , \label{eq:7.2}\ ] ] from which it is clear that at half filling @xmath348 all pseudospin components are conserved . this leads to the diffusion form of the pseudospin ( doublon ) susceptibility , which is the retarned doublon gf @xmath349 . according to kubo - mori theory , this susceptibility is expressed through the memory function @xmath350 by the relation @xmath351 where we introduce a notation for the static susceptibility , @xmath352 . on the other hand the memory function is expressed through the irreducible retarded gf of pseudospin currents ( see @xcite ) : @xmath353 here @xmath354 means the time derivative of operator @xmath355 : @xmath356= ( u-2\mu)x_{i}^{02}-\sigma^{\prime}(\tau^{x}\varpsi)_{\alpha^{\prime } } ( i\bar{\sigma}^{\prime})(\hat{t}\varpsi)_{\alpha^{\prime}}(i\sigma^{\prime } ) . \label{eq:7.5}\ ] ] further to this , we consider the half - filling case , when @xmath343 . then @xmath357 now we use the approximation of interacting modes , well known in the relaxation theory , by mori @xcite : the two - particle electron correlations in expression ( [ eq:7.6 ] ) are decomposed into pair correlators and then expressed through the imaginary parts of the retarded electron gfs . as a result we come to the following expression , determining the memory function : @xmath358^{2 } \int d\omega^{\prime}\int d\omega_{1 } \bigl [ f(\omega_{1}-\omega')-f(\omega_{1 } ) \bigr ] \label{eq:7.7 } \\ & \times \frac{\displaystyle tr\bigl\{[im\mathcal{g}^{\bar{\sigma}^{\prime } } ( \bm{q}-\bm{k},\omega^{\prime}-\omega_{1 } ) ] [ im(\im\mathcal{g}^{t\sigma^{\prime}}(\bm{k},\omega_{1})\im ) ] \bigr\ } } { \omega^{\prime}(\omega-\omega^{\prime}+i\delta)}\nonumber.\end{aligned}\ ] ] here @xmath359 is the transposed matrix @xmath216 . the quantity @xmath360 is the retarded electron gf . it can be obtained from our matsubara gfs by analytical continuation from discrete imaginary frequencies into real ones : @xmath361 . expression ( [ eq:7.7 ] ) is similar to those obtained in the interacting modes approximation for other dynamical susceptibilities . for example , the spin susceptibility is : @xmath362 by similar decoupling of the irreducible gfs of the currents we obtain : @xmath363^{2 } \int d\omega^{\prime}\int d\omega_{1 } \bigl[f(\omega_{1}-\omega^{\prime})-f(\omega_{1})\bigr ] \label{eq:7.9 } \\ & \times \frac{\displaystyle tr\bigl\{[im\mathcal{g}^{\sigma}(\bm{k}-\bm{q},\omega_{1}-\omega^{\prime } ) ] [ im(\im\mathcal{g}^{\bar{\sigma}}\im)(\bm{k},\omega ] \bigr\ } } { \displaystyle\omega^{\prime}(\omega-\omega^{\prime}+i\delta)}.\nonumber\end{aligned}\ ] ] it is remarkable that the memory gf for the spin susceptibility vanishes at @xmath364 , while for the doublon susceptibility it vanishes at @xmath365 . this difference originates from the total spin conservation law ( fourier component of the spin density at @xmath364 ) , while the component of the pseudospin density is conserved at @xmath365 and only for half - filling . there is another important difference in expressions ( [ eq:7.7 ] ) and ( [ eq:7.9 ] ) . arguments of the electron gfs appear in a different way in these expressions . this reflects the fact that the spin collective mode is formed through excitations of a particle and a hole , while the pseudospin collective mode ( doublone ) is formed through excitations of two particles ( or two holes ) . consider now the hydrodynamical limit corresponding to small frequencies @xmath366 and small wave point @xmath201 ( for expression ( [ eq:7.7 ] ) ) . in the hydrodynamical limit @xmath367 , where @xmath84 is a characteristic electron velocity on the fermi surface . under these conditions from eqs . ( [ eq:7.7 ] ) and ( [ eq:7.9 ] ) the asymptotic expressions follow : @xmath368 @xmath369 where the coefficients of spin and pseudospin stiffness are equal @xmath370 \biggr\ } , \label{eq:7.12}\ ] ] @xmath371 \biggr\}. \label{eq:7.13}\ ] ] here @xmath372 is the unit wave vector , and @xmath373 is the derivative of the fermi function . expression ( [ eq:7.13 ] ) is valid at arbitrary @xmath1 ; in the case of @xmath374 it is consistent with the result of @xcite for the @xmath16-model . notice that if we use the electron gf in the mean field approximation ( without attenuation of quasiparticles ) both expressions ( [ eq:7.12 ] ) and ( [ eq:7.13 ] ) vanish . it is easy to show that if the attenuation of quasiparticles @xmath375 obeys the condition @xmath376 , both expressions become finite . in the general case expressions ( [ eq:7.7 ] ) and ( [ eq:7.9 ] ) for the memory function give in the hydrodynamical limit correct asymptotic values , therefore the susceptibilities have the diffusion form , which is @xmath377 where @xmath378 . we have applied the gfa to investigate the hubbard model in the @xmath0 operator representation . this means that we discussed the case of sufficiently strong electronic correlations @xmath379 . we have derived the exact equation for the electronic gf in terms of the variational derivatives with respect to the fluctuating fields @xmath380 , @xmath279 , @xmath381 coupled with the spin and charge densities . the electronic gfs represent generally an 8x8 matrix with respect to the three discrete indexes @xmath18 , @xmath19 , @xmath20 . in the matrix representation the equation has the same structure with the gf for the hubbard model in the limit @xmath15 , for the @xmath16- and @xmath17-models and for the gfs of the transverse spin components in the heisenberg model as well . the electronic gf @xmath2 has a multiplicative character in the sense that it is expressed by a product of two quantities , @xmath382 , where @xmath160 is the propagator satisfying the dyson equation with the self energy @xmath3 , and @xmath4 is the terminal part . from the equation for @xmath2 , a pair of equations with variational derivatives for @xmath3 and @xmath4 follow . their iteration generates a power series in the parameter @xmath5 . this corresponds to the perturbation theory close to the atomic limit . the iteration corrections of the first two orders allow to formulate a mean field approximation essentially equivalent to that of com . taking the electronic gf in the mean field approximation we derived an equation for the doublon gf . the properties of the poles of the doublon gf depends substantially on the electronic concentration @xmath244 . for @xmath28 there is a pole which has a real part @xmath383 , corresponding to the activated mode with the quadratic dispersion law . for @xmath384 , @xmath385 . the investigation of the special case @xmath7 reveals that a soft mode with @xmath8 may exist . however at @xmath8 the paramagnetic phase of the hubbard model has an instability to antiferromagnetic ordering . it means that two possible instabilities doublon and magnon ones should compete , and a final result concurring a type of ordering at half filling demands a farther investigations . it will be a subject of next study . the other direction is to investigate magnetically ordered states . we should go out of the scope of the mean field approximation and take into account the second order correction @xmath25 , including the interaction of electrons with magnons . the preliminary analysis reveals that it contains a singular kondo - like term @xmath386 , which , as it has been pointed out in the works @xcite@xcite , leads to a stable ferromagnetism . after the extraction of the relevant term in the second order correction we could write a more exact equation for the magnon gf and calculate the spin - wave spectrum . all of this will form the subject of a next paper . one of the authors ( yu.a . i. ) would like to thank prof . f. mancini for hospitality during his stay in the dipartimento di fisica e.r . caianiello " , universit degli studi di salerno , where this paper has been started . izuymov , n.i . chaschin and d. s. alexeev thank the russian foundation of support of scientific school , grant ns747.2003.2 . consider an average of an arbitrary @xmath82-product of the operators @xmath387 , @xmath388 , @xmath389 , @xmath390 , ... ( it could be @xmath0 operators , spin operators or other ) , taken in the heiseberg representation : @xmath391 and so on . then the following identity is valid : @xmath392 here the average over the gibbs statistical ensemble is denoted by @xmath393 the relation ( [ eq : a.2 ] ) represents the result of the differentiation of the initial average with respect to the time @xmath394 ascribed to the operator @xmath387 . in the right hand side of the relation ( [ eq : a.1 ] ) @xmath395 represents the time derivative @xmath396 . \tag{a.4 } \label{eq : a.4}\ ] ] the curl brackets of the kind @xmath397 mean @xmath398_{\pm } ( \tau_{1 } ) , \tag{a.5 } \label{eq : a.5}\ ] ] where @xmath399_{\pm}$ ] is an anticommutator or a commutator depending on the kind of the @xmath400 and @xmath401 operators . the signs of the terms in the second line of eq.([eq : a.2 ] ) are determined by the signs of the transpositions of @xmath43-operators in the @xmath82-product from their original place to the second one in the energy term . in the last term of eq.([eq : a.2 ] ) the expression @xmath402 is a commutator because the operator @xmath11 is implied to be boson - like . formally this term and the first one could be merged and @xmath403 may be considered in a sense as the hamiltonian of a system immersed in fluctuating fields . the identity ( [ eq : a.2 ] ) could be proved by differentiating the @xmath82-product expressed through the @xmath404 functions and the expansion of the exponent @xmath405 in a series , with subsequent recollection of all the terms back to the exponent . this identity serves as a basis for the derivation of the gfs in the fluctuating fields , defined as @xmath406 where @xmath407 $ ] is the generating functional . the normal gf @xmath408 can be represented in a multiplicative form similar to that of ( [ eq:3.18 ] ) for the whole gf , namely @xmath409 where bold indexes mean @xmath410 . thus quantities @xmath2 , @xmath160 , @xmath4 and @xmath133 ( determined by dyson equation ( [ eq:4.3 ] ) ) are @xmath64 matrices with respect to the spinor index @xmath19 . we find the equations for @xmath4 and @xmath3 from the general matrix equations ( [ eq:3.21 ] ) and ( [ eq:3.22 ] ) by writing down the equation for the matrix element @xmath411 : @xmath412\centerdot \left [ \left(l_{0v}^{-1}\right)^{11 } ( \scriptstyle\textbf{3}^{\prime}\textbf{2}\displaystyle)- \sigma^{\prime 11 } ( \scriptstyle\textbf{3}^{\prime}\textbf{2}\displaystyle ) \right ] \nonumber\\ & -\left [ ( \hat{t}l^{12 } ) ( \scriptstyle\textbf{4}^{\prime}\textbf{3}^{\prime}\displaystyle ) \hat{a}^{11 } ( \scriptstyle\textbf{1}\textbf{4}^{\prime}\displaystyle)- ( \hat{\tilde{t}}l^{22 } ) ( \scriptstyle\textbf{4}^{\prime}\textbf{3}^{\prime}\displaystyle ) \hat{a}^{12 } ( \scriptstyle\textbf{1}\textbf{4}^{\prime}\displaystyle ) \right]\centerdot \left [ \left(l_{0v}^{-1}\right)^{21 } ( \scriptstyle\textbf{3}^{\prime}\textbf{2}\displaystyle)- \sigma^{\prime 21 } ( \scriptstyle\textbf{3}^{\prime}\textbf{2}\displaystyle ) \right ] \nonumber\end{aligned}\ ] ] @xmath413\centerdot \left [ \left ( l_{0v}^{-1 } \right)^{11 } ( \scriptstyle\textbf{3}^{\prime}\textbf{2}\displaystyle)- \sigma^{\prime 11 } ( \scriptstyle\textbf{3}^{\prime}\textbf{2}\displaystyle ) \right ] \nonumber\\ & -\left [ ( \hat{t}l^{12 } ) ( \scriptstyle\textbf{4}^{\prime}\textbf{3}^{\prime}\displaystyle ) \hat{a}^{21 } ( \scriptstyle\textbf{1}\textbf{4}^{\prime}\displaystyle)- ( \hat{\tilde{t}}l^{22 } ) ( \scriptstyle\textbf{4}^{\prime}\textbf{3}^{\prime}\displaystyle ) \hat{a}^{22 } ( \scriptstyle\textbf{1}\textbf{4}^{\prime}\displaystyle ) \right]\centerdot \left [ \left ( l_{0v}^{-1 } \right)^{21 } ( \scriptstyle\textbf{3}^{\prime}\textbf{2}\displaystyle)- \sigma^{\prime 21 } ( \scriptstyle\textbf{3}^{\prime}\textbf{2}\displaystyle ) \right ] . \nonumber\end{aligned}\ ] ] for the normal phase , matrix elements @xmath414 , however the derivatives of them with respect to the fields @xmath280 and @xmath381 should not vanish , therefore in equations ( [ eq : b.2 ] ) and ( [ eq : b.3 ] ) we must keep such derivatives . in the first order in @xmath80 equations ( [ eq : b.2 ] ) and ( [ eq : b.3 ] ) lead respectively to an equation for the self - energy of the normal gf ( @xmath415 ) . matrix operators @xmath416 and @xmath417 include the derivatives with respect to fluctuating fields which act on @xmath418 and @xmath419 , and therefore we come to the expression : @xmath420 \nonumber \\ & = -\delta_{12 } \biggl [ ( tg_{11}^{\bar{\sigma } } ) ( \scriptstyle 11 \displaystyle)- ( tg_{22}^{\bar{\sigma } } ) ( \scriptstyle 11 \displaystyle ) \biggr ] \begin{pmatrix } 1 & -1\\ -1 & 1 \end{pmatrix}. \nonumber\end{aligned}\ ] ] here in the last line we have used a more concise definition of @xmath421 , and also made use of the expression for the transposed matrix @xmath422 the calculation of second order contribution in the uncutable part @xmath423 demands much more efforts , but it involves nothing else then standard and straightforward iterations of equations ( [ eq : b.2 ] ) and ( [ eq : b.3 ] ) . we present the final result : @xmath424 where @xmath425 in a similar way we calculate the terminal part @xmath4 of the electronic gf . firstly we write down the expression for a matrix element @xmath426 from equation ( [ eq:3.29 ] ) . this element is coupled with the off - diagonal element @xmath427 . we have the following pair of coupled equations @xmath428\lambda ( \scriptstyle\textbf{3}^{\prime}\textbf{2}\displaystyle ) \nonumber\\ & + \left [ ( \hat{t}l^{12 } ) ( \scriptstyle\textbf{4}^{\prime}\textbf{3}^{\prime}\displaystyle ) \hat{a}^{11 } ( \scriptstyle\textbf{1}\textbf{4}^{\prime}\displaystyle)- ( \hat{\tilde{t}}l^{22 } ) ( \scriptstyle\textbf{4}^{\prime}\textbf{3}^{\prime}\displaystyle ) \hat{a}^{12 } ( \scriptstyle\textbf{1}\textbf{4}^{\prime}\displaystyle ) \right]\pi^{21 } ( \scriptstyle\textbf{3}^{\prime}\textbf{2}\displaystyle ) , \nonumber\end{aligned}\ ] ] @xmath429\lambda ( \scriptstyle\textbf{3}^{\prime}\textbf{2}\displaystyle ) \nonumber\\ & + \left [ ( \hat{t}l^{12 } ) ( \scriptstyle\textbf{4}^{\prime}\textbf{3}^{\prime}\displaystyle ) \hat{a}^{21 } ( \scriptstyle\textbf{1}\textbf{4}^{\prime}\displaystyle)- ( \hat{\tilde{t}}l^{22 } ) ( \scriptstyle\textbf{4}^{\prime}\textbf{3}^{\prime}\displaystyle ) \hat{a}^{22 } ( \scriptstyle\textbf{1}\textbf{4}^{\prime}\displaystyle ) \right]\pi^{21 } ( \scriptstyle\textbf{3}^{\prime}\textbf{2}\displaystyle ) . \nonumber\end{aligned}\ ] ] from here we find the zeroth order expressions for @xmath430 and @xmath427 @xmath431 @xmath432 substituting these expressions in equations ( [ eq : b.8 ] ) and ( [ eq : b.9 ] ) leads to the first order correction for the terminal part : @xmath433 which includes bosonic gfs determined by definitions ( [ eq:2.20 ] ) ( [ eq:2.22 ] ) . fourier transformations of expressions ( [ eq : b.4 ] ) , ( [ eq : b.6 ] ) , ( [ eq : b.10 ] ) and ( [ eq : b.12 ] ) give the final results for the self - energy and the terminal part of the electronic gf . they are given by formulas ( [ eq:4.10]),([eq:4.12 ] ) , ( [ eq:4.5 ] ) and ( [ eq:4.7 ] ) , respectively . s. ishihara , h. matsumoto , s. odashima , m. tachiki , f. mancini , phys . b * 49 * , 1350 ( 1994 ) ; h. matsumoto , t. saikawa , f. mancini , phys . * * 54*5 , 14445 ( 1996 ) ; d. villani , e. lange , a. avella , g. kotliar , phys . rev . lett . 85 * , 804 ( 2000 ) ; v. fiorentino , f. mancini , e. zasinas , a. barabanov ; phys . b * 64 * , 214515 ( 2001 ) ; f. mancini , n. perkins , n. plakida , phys a * 284 * , 286 ( 2001 ) ; a. avella , f. mancini , d. villani , h. matsumoto , eur . j. b * 20 * , 303 ( 2001 ) ; m. bak , f. mancini , physica b * 312 * , 732 ( 2002 ) ; a. avella , f. mancini , r. hayn , eur . j. b * 37 * , 465 ( 2004 ) yu . a. izyumov , in lectures on the physics of highly correlated electron systems vii , seventh training course in the physics of correlated electron systems and high - tc superconductors , ed . a. avella , f. mancini , aip conference proceedings 678 , 2003	the method of generating functional , suggested for conventional systems by kadanoff and baym , is generalized to the case of strongly correlated systems , described by the hubbard @xmath0 operators . the method has been applied to the hubbard model with arbitrary value @xmath1 of the coulomb on - site interaction . for the electronic green s function @xmath2 constructed for fermi - like @xmath0 operators , an equation using variational derivatives with respect to the fluctuating fields has been derived and its multiplicative form has been determined . the green s function is characterized by two quantities : the self energy @xmath3 and the terminal part @xmath4 . for them we have derived the equation using variational derivatives , whose iterations generate the perturbation theory near the atomic limit . corrections for the electronic self - energy @xmath3 are calculated up to the second order with respect to the parameter @xmath5 ( @xmath6 width of the band ) , and a mean field type approximation was formulated , including both charge and spin static fluctuations . this approximation is actually equivalent to the one used in the method of composite operators , and it describes an insulator - metal phase transition at half filling reasonably well . the equations for the bose - like green s functions have been derived , describing the collective modes : the magnons and doublons . the main term in this equation represents variational derivatives of the electronic green s function with respect to the corresponding fluctuating fields . the properties of the poles of the doublon green s functions depend on electronic filling . the investigation of the special case @xmath7 demonstrates that the doublon green s function has a soft mode at the wave vector @xmath8 , indicating possible instability of the uniform paramagnetic phase relatively to the two sublattices charge ordering . however this instability should compete with an instability to antiferromagnetic ordering . the generating functional method with the @xmath0 operators could be extended to the other models of strongly correlated systems . + + yu.a . izyumov , n.i . chaschin , d.s . alexeev + _ institute for metal physics , ural division of the russian academy of sciences , + 620219 , ekaterinburg , s.kovalevskaya str.18 , russia _ f. mancini + _ dipartimento di fisica e.r . caianiello " , universit degli studi di salerno , + via s. allende , i-84081 baronissi ( sa ) , italy _ +
a classical theorem of vizing @xcite asserts that the _ chromatic index _ of every cubic bridgeless graph , i.e. , the smallest number of colors needed to properly edge - color such a graph , is @xmath2 or @xmath3 . cubic cyclically @xmath3-edge - connected graphs with chromatic index @xmath3 are known as _ snarks _ and it is known that smallest counter - examples , if any , to several deep open conjectures in graph theory , such as the cycle double cover conjecture , must be snarks . our research is motivated by edge - colorings of cubic bridgeless graphs with no short cycles . the girth conjecture of jaeger and swart @xcite asserted that there are no snarks with large girth . this conjecture was refuted by kochol @xcite who constructed snarks of arbitrary large girth . hence , it is natural to ask whether it can be said that snarks of large girth are close to being @xmath2-edge - colorable in some sense . one of the relaxations of ordinary colorings are circular colorings , introduced by vince @xcite . a _ @xmath4-coloring _ of a graph @xmath5 is a coloring of vertices with colors from the set @xmath6 such that any two adjacent vertices receive colors @xmath7 and @xmath8 with @xmath9 . circular colorings naturally appear in different settings , which is witnessed by several equivalent definitions of this notion as exposed in the surveys by zhu @xcite . the infimum of the ratios @xmath10 such that @xmath5 has a @xmath4-coloring is the _ circular chromatic number _ of @xmath5 . it is known that the infimum is the minimum for all finite graphs and the ceiling of the circular chromatic number of a graph is equal to its chromatic number . thus , the circular chromatic number is a fractional relaxation of the chromatic number . the _ circular chromatic index _ of a graph is the circular chromatic number of its line - graph . zhu @xcite asked whether there exist snarks with circular chromatic index close or equal to @xmath3 , and as there are snarks with arbitrary large girth , it is also interesting to know whether there exist such snarks of arbitrary large girth . afshani et al . @xcite showed that the circular chromatic index of every cubic bridgeless graph is at most @xmath11 and kaiser et al . @xcite showed that for every @xmath12 , there exists @xmath13 such that every cubic bridgeless graph with girth at least @xmath13 has circular chromatic index at most @xmath14 . this latter result was generalized to graphs with bounded maximum degree @xcite . moreover , the circular chromatic indices of several well - known classes of snarks have been determined @xcite . the petersen graph is the only cubic bridgeless graph that is known to have the circular chromatic index equal to @xmath11 . in fact , it is the only example of a cubic bridgeless graph with circular chromatic index greater than @xmath1 . this leads to the following ( wild ) conjecture . [ conj ] every cubic bridgeless graph different from the petersen graph has circular chromatic index strictly less than @xmath11 , maybe , at most @xmath1 . in their paper , kaiser et al . @xcite , formulated a problem to determine the smallest girth @xmath15 such that every cubic bridgeless graph with girth at least @xmath15 has circular chromatic index at most @xmath1 , and they showed that @xmath16 . note that @xmath17 because of the petersen graph . in this paper , we prove that every cubic bridgeless graph having a @xmath18-factor composed of cycles of length different from @xmath2 and @xmath19 has circular chromatic index at most @xmath1 . this implies that @xmath20 , i.e. , the circular chromatic index of every cubic bridgeless graph with girth at least @xmath21 is at most @xmath1 . our result also applies to subcubic graphs with odd - girth at least @xmath0 . the core of our argument is formed by decomposing the graph obtained by contracting a @xmath18-factor of a cubic bridgeless graph into trails . we first introduce notation related to such decompositions and then prove their existence . an _ _ is a graph with multiple edges , loops and half - edges allowed with a fixed cyclic ordering of the ends around each vertex . formally , an @xmath22 is comprised of a vertex - set @xmath23 and an edge - set @xmath24 . each edge has two ends : one of them is incident with a vertex , and the other may , but need not , be incident with a vertex . an edge that has exactly one end incident with a vertex is a _ half - edge_. an edge with both ends incident with the same vertex is a _ loop_. the _ degree _ @xmath25 of a vertex @xmath26 is the number of ends incident with @xmath27 ( in particular , loops are counted twice ) . moreover , for each vertex , there is a cyclic ordering of the ends incident with it . these orderings are represented by a surjective mapping @xmath28 such that @xmath29 is an edge incident with @xmath27 and @xmath30 for every @xmath31 . an edge has two pre - images in @xmath32 if and only if it is a loop incident with @xmath27 . an naturally yields an embedding of the corresponding multi - graph on a surface . _ deleting a vertex @xmath26 _ from an @xmath33 yields an @xmath34 with @xmath35 , @xmath36 and @xmath37 being the restriction of @xmath38 to @xmath39 . in other words , the vertex @xmath27 and half - edges and loops incident with @xmath27 are removed . the other edges incident with @xmath27 become half - edges . a _ trail _ in an is a sequence of mutually distinct edges @xmath40 such that * @xmath41 and @xmath42 have a common end - vertex @xmath43 , for @xmath44 ; and * @xmath45 unless @xmath46 is a loop , for @xmath47 . if @xmath48 is not a half - edge , then the trail _ starts at a vertex_. similarly , if @xmath49 is not a half - edge , the trail _ ends at a vertex_. furthermore , a trail that consists of a single half - edge either starts or ends at a vertex . for a given trail , a linear ordering on its edges is naturally defined . let @xmath50 and @xmath51 be two edge - disjoint trails and @xmath27 a vertex . if @xmath50 ends at @xmath27 and @xmath51 starts at @xmath27 , then we can _ link _ the two trails by identifying the end of @xmath50 with the beginning of @xmath51 : thereby , we obtain a new trail that first follows the edges of @xmath50 and then those of @xmath51 . we also define the linking of a trail @xmath52 that starts and ends at the same vertex with itself : in this case , we obtain the same trail @xmath52 except that the ordering of the edges becomes cyclic . a trail obtained in this way closed_. trails that are not closed are _ open_. note that a trail that starts and ends at the same vertex can be either open or closed . trails @xmath53 form a _ compatible decomposition _ of an @xmath5 if all of them are open , every edge is contained in exactly one of the trails and for every vertex @xmath27 of odd degree of @xmath5 , there is an index @xmath54 such that the following pairs of edges are consecutive ( regardless of their order ) in some of the trails ( and thus are not the same edge ) : * @xmath55 and @xmath56 ; * @xmath57 and @xmath58 ; and * @xmath59 and @xmath60 . note that if an @xmath5 has a compatible decomposition , then it has no vertices of degree @xmath61 or @xmath19 . we now prove that every with no vertices of degree @xmath61 , @xmath2 or @xmath19 has a compatible decomposition . in the next two lemmas , which form the base of our inductive argument , with a single vertex are analyzed . [ lm-7 ] every @xmath5 that has a single vertex and the degree of this vertex is @xmath0 has a compatible decomposition . let @xmath27 be the vertex and let @xmath46 be the image of @xmath62 for each @xmath63 . further , let @xmath64 be the set of trails obtained by the following process : initially , each edge of @xmath5 forms a single trail . we then link the two trails starting or ending with @xmath46 and @xmath65 , next we link the two trails starting or ending with @xmath42 and @xmath66 , and last the two trails starting or ending with @xmath67 and @xmath68 . this operation does not always yield a compatible decomposition , since it may create closed trails . for instance , @xmath64 contains a closed trail if @xmath69 , or if @xmath70 and @xmath71 ( in this last case , after having linked @xmath42 with @xmath66 , thereby obtaining a trail @xmath52 , the last linking amounts to linking @xmath52 with itself ) . on the other hand , if all the obtained trails are open , then @xmath64 is a compatible decomposition , with @xmath72 being the index @xmath54 of the definition . our goal is to show that at least one of the sets @xmath64 is composed only of open trails . since the degree of @xmath27 is odd , the vertex @xmath27 is incident with at least one half - edge . by symmetry , we assume that this half - edge is @xmath73 . if @xmath74 contains a closed trail , then one of the following four cases applies ( recall that @xmath73 is a half - edge ) : * @xmath75 , * @xmath76 , * @xmath77 and @xmath78 , or * @xmath79 and @xmath80 . similarly , if @xmath81 contains a closed trail , then one of the following four cases applies : * @xmath82 , * @xmath76 , * @xmath83 and @xmath84 , or * @xmath85 and @xmath86 . comparing the two sets of four possible cases , we conclude that if none of @xmath74 and @xmath81 is compatible , then at least one of the following six cases applies ( these cases are referred to as _ main _ cases in figures [ fig-7-a ] and [ fig-7-b ] ) : * @xmath75 and @xmath82 , * @xmath76 , * @xmath77 , @xmath78 and @xmath82 , * @xmath79 , @xmath80 and @xmath82 , * @xmath75 , @xmath83 and @xmath84 , or * @xmath75 , @xmath85 and @xmath86 . in the first case , the set @xmath87 contains only open trails . in the last four cases , the sets @xmath88 , @xmath87 , @xmath89 and @xmath90 , respectively , are composed of open trails ( see figure [ fig-7-a ] for an illustration ) . we now focus on the second case . if @xmath87 contains a closed trail ( and @xmath76 ) , it holds that @xmath91 . however , in this case , the set @xmath88 contains no closed trails ( see figure [ fig-7-b ] ) . this finishes the proof of the lemma . [ lm - single ] every @xmath5 that has a single vertex and the degree of this vertex is different from @xmath61 , @xmath2 and @xmath19 has a compatible decomposition . let @xmath27 be the only vertex of the graph @xmath5 . if the degree of @xmath27 is even , then there is nothing to prove as every decomposition into open trails is compatible . if the degree of @xmath27 is @xmath0 , then the statement follows from lemma [ lm-7 ] . hence , we assume that the degree of @xmath27 is odd and it is at least @xmath92 . let @xmath46 be the image of @xmath62 for @xmath93 . as the degree of @xmath27 is odd , @xmath27 is incident with at least one half - edge . without loss of generality , we can assume that @xmath94 is the half - edge . let @xmath64 be the set of trails defined as in the proof of lemma [ lm-7 ] . assume that both the sets @xmath81 and @xmath87 contain a closed trail ; if any of them were composed of open trails only , then it would form a compatible decomposition . as in the proof of lemma [ lm-7 ] , we infer from the facts that both @xmath81 and @xmath87 contain a closed trail that one of the following six cases applies ( replace @xmath73 with @xmath94 and @xmath64 with @xmath95 for @xmath96 in the analysis done in the proof of lemma [ lm-7 ] ) : * @xmath82 and @xmath97 , * @xmath98 , * @xmath99 , @xmath100 and @xmath97 , * @xmath101 , @xmath102 and @xmath97 , * @xmath82 , @xmath103 and @xmath104 , or * @xmath82 , @xmath80 and @xmath105 . in the last four cases , the sets @xmath88 , @xmath88 , @xmath74 and @xmath74 , respectively , contain no closed trails ( see figure [ fig - single - a ] for an illustration ) . let us focus on the first two cases , now . suppose that @xmath82 and @xmath97 . if @xmath74 contains a closed trail , then @xmath75 . further , if @xmath88 contains a closed trail , then @xmath106 . thus , the set @xmath107 contains no closed trails ( see figure [ fig - single - b ] ) , and hence is a compatible decomposition . + = 6.5 cm = 6.5 cm assume now that @xmath98 . if @xmath74 contains a closed trail , then @xmath108 . further , if @xmath88 contains a closed trail , then @xmath109 . again , the set @xmath107 is then a compatible decomposition ( see figure [ fig - single - c ] ) . to summarize , we have shown that at least one of the sets @xmath64 , @xmath110 , is composed of open trails only , and thus it forms a compatible decomposition . using lemma [ lm - single ] , we show that every with no vertices of degree @xmath61 , @xmath2 or @xmath19 has a compatible decomposition . [ lm - mgraph ] every @xmath5 without vertices of degree @xmath61 , @xmath2 or @xmath19 has a compatible decomposition . the proof proceeds by induction on the number of vertices of @xmath5 . if @xmath5 has one vertex , then the statement follows from lemma [ lm - single ] . otherwise , let @xmath27 be an arbitrary vertex of @xmath5 and let @xmath111 be an obtained from @xmath5 by removing @xmath27 . by the induction hypothesis , @xmath111 has a compatible decomposition @xmath112 into trails . let @xmath113 be the set of trails obtained from @xmath112 by adding the set of loops and half - edges incident with @xmath27 in @xmath5 . observe that @xmath113 is a set of trails of @xmath5 in which there is no trail `` traversing '' @xmath27 . if the degree of @xmath27 is even , the set @xmath113 is a compatible decomposition of @xmath5 as there is no restriction on how the trails pass through the vertex @xmath27 . assume that the degree @xmath25 of @xmath27 is odd . we now define an auxiliary @xmath114 . the @xmath114 contains a single vertex @xmath115 of degree @xmath25 , and for @xmath116 , we have @xmath117 if and only if @xmath113 contains a trail starting with the edge @xmath62 and ending with @xmath118 . in other words , trails of @xmath112 that both start and finish with an edge incident with @xmath27 correspond in @xmath114 to loops incident with @xmath115 , the loops incident with @xmath27 are preserved and the trails starting or finishing at @xmath27 ( but not both ) correspond to half - edges . trails containing no edge incident with @xmath27 have no counterparts among the edges of @xmath114 . by lemma [ lm - single ] , the @xmath114 has a compatible decomposition @xmath119 . we can now obtain a compatible decomposition @xmath120 of @xmath5 as follows : all the trails of @xmath113 neither starting nor ending at @xmath27 are added to @xmath120 . each trail @xmath121 of @xmath119 has a corresponding trail in @xmath120 that is obtained by replacing every edge of @xmath121 with the corresponding trail of @xmath113 and linking these trails . it is straightforward to verify that @xmath120 is a set of open trails of @xmath5 . by induction , the trails pass through vertices of @xmath5 different from @xmath27 in the way required by the definition of a compatible decomposition . the trails also pass through @xmath27 in the required way because @xmath119 is a compatible decomposition of @xmath114 . hence , @xmath120 is a compatible decomposition of @xmath5 . we are now ready to prove our main theorem . [ thm - main ] the circular chromatic index of every cubic graph with a @xmath18-factor composed of cycles of lengths different from @xmath2 and @xmath19 is at most @xmath1 . let @xmath5 be a cubic graph and @xmath122 a @xmath18-factor of @xmath5 composed of cycles of lengths different from @xmath2 and @xmath19 , and let @xmath123 be the perfect matching complementary to @xmath122 . the multi - graph obtained by contracting @xmath122 can be viewed as an @xmath114 : the vertices of @xmath114 correspond to the cycles of the @xmath18-factor @xmath122 , and the order in which the edges of @xmath123 are incident with cycles of @xmath122 naturally defines the function @xmath38 . note that @xmath114 has no half - edges and its loops correspond to chords of cycles of @xmath122 . since no cycle of @xmath122 has length @xmath2 or @xmath19 , no vertex of @xmath114 has degree @xmath124 or @xmath19 . thus , by lemma [ lm - mgraph ] , the @xmath114 has a compatible decomposition @xmath113 . color the edges of every trail of @xmath113 with @xmath125 and @xmath61 in an alternating way . since the edges of @xmath114 correspond one - to - one to the edges of @xmath123 , we have obtained a coloring of the edges of @xmath123 with @xmath125 and @xmath61 . we now construct a @xmath126-edge - coloring of the edges of @xmath5 . let @xmath127 be a cycle of @xmath122 and @xmath128 the color of the edge of @xmath123 incident with the vertex @xmath43 , for @xmath129 . if the length @xmath130 of @xmath131 is even , we color the edges of @xmath131 with @xmath2 and @xmath19 in an alternating way . let us consider the case where @xmath130 is odd . since @xmath113 is a compatible decomposition , there exists an index @xmath132 such that @xmath133 , @xmath134 and @xmath135 ( indices are taken modulo @xmath130 ) as the colors of the edges of trails of @xmath113 alternate . we now show that there exists an index @xmath136 such that @xmath137 . if @xmath138 , then set @xmath139 . otherwise , @xmath140 . since either @xmath141 or @xmath142 is equal to @xmath143 , the index @xmath136 can be set to @xmath132 or @xmath144 . by symmetry , we can assume in the remainder that @xmath145 , @xmath146 and @xmath147 . color the edge @xmath148 with @xmath18 , the edge @xmath149 with @xmath3 and the edge @xmath150 with @xmath21 . the remaining edges are colored with @xmath2 and @xmath19 in the alternating way ( see figure [ fig - main ] ) . we have obtained a proper coloring of @xmath131 . as we can extend the coloring of the edges of @xmath123 to all cycles of @xmath122 , the resulting @xmath126-edge - coloring witnesses that the circular chromatic index of @xmath5 does not exceed @xmath1 . petersen s theorem @xcite asserts that every cubic bridgeless graph has a perfect matching ; this yields the next corollary of theorem [ thm - main ] . [ cor - bridgeless ] the circular chromatic index of every cubic bridgeless graph with girth @xmath21 or more is at most @xmath1 . finally , we show that the assumption that the given graph is cubic can be relaxed in corollary [ cor - bridgeless ] . [ cor - main ] the circular chromatic index of every subcubic graph with odd - girth @xmath0 or more is at most @xmath1 . let @xmath5 be a subcubic graph with odd - girth @xmath0 or more that has a circular chromatic index greater than @xmath1 and that has the smallest number of vertices among all such graphs . consequently , the minimum degree of @xmath5 is at least @xmath18 . similarly , @xmath5 is connected . the graph @xmath5 is also bridgeless : otherwise , each of the two graphs obtained from @xmath5 by splitting along the bridge has a @xmath126-edge - coloring and these edge - colorings ( after rotating the colors if necessary ) combine to a @xmath126-edge - coloring of @xmath5 . if @xmath5 has no vertices of degree @xmath18 , then @xmath5 is a cubic bridgeless graph . theorem @xcite ensures that @xmath5 has a a @xmath18-factor @xmath122 . by our assumption , no cycle of @xmath122 has length @xmath2 or @xmath19 , and therefore @xmath5 can not be a counter - example by theorem [ thm - main ] . so , assume that @xmath5 has at least one vertex of degree @xmath18 . let @xmath114 be a @xmath2-edge - connected cubic graph of odd - girth at least @xmath0 from which we remove an edge . we construct the graph @xmath111 as follows . we take two disjoint copies of @xmath5 . for each pair @xmath151 of corresponding vertices of degree @xmath18 ( one in each copy of @xmath5 ) , we add a copy of @xmath114 , join @xmath152 to a vertex of degree @xmath18 of @xmath114 , and @xmath27 to the other vertex of degree @xmath18 in @xmath114 . the resulting graph @xmath111 is cubic and has odd - girth at least @xmath0 . moreover , since @xmath5 is bridgeless , @xmath111 has at most two bridges . more precisely , @xmath111 has two bridges if and only if @xmath5 has exactly one vertex of degree @xmath18 , and @xmath111 is bridgeless otherwise . therefore , @xmath111 has a perfect matching by tutte s theorem @xcite . consequently , since @xmath111 has odd - girth at least @xmath0 , theorem [ thm - main ] implies the existence of a @xmath126-edge - coloring of @xmath111 . this edge - coloring restricted to @xmath5 yields a @xmath126-edge - coloring of @xmath5 , a contradiction . the first and the last authors would like to thank mohammad ghebleh and luke postle for discussions and insights on circular edge - colorings of cubic bridgeless graphs . 99 p. afshani , m. ghandehari , m. ghandehari , h. hatami , r. tusserkani , x. zhu : circular chromatic index of graphs of maximum degree 3 , j. graph theory 49 ( 2005 ) , 325335 . m. ghebleh , d. krl , s. norine , r. thomas : the circular chromatic index of flower snarks , electron . j. combin . 13 ( 2006 ) , # n20 , 7pp . m. ghebleh : the circular chromatic index of goldberg snarks , discrete math . 307 ( 2007 ) , 3220 - 3225 . m. ghebleh : the circular chromatic index of generalized blanua snarks , electron . j. combin . 15 ( 2008 ) , # r44 , 9pp . f. jaeger , t. swart : conjecture 1 , in : combinatorics 79 ( m. deza , i. g. rosenberg , eds . ) , ann . discrete math . vol . 9 , north - holland , amsterdam , 1980 , p. 305 . t. kaiser , d. krl , r. krekovski : a revival of the girth conjecture , j. combin . theory ser . b 92 ( 2004 ) , 4153 . t. kaiser , d. krl , r. krekovski , x. zhu : the circular chromatic index of graphs of high girth , j. combin . theory ser . b 97 ( 2007 ) , 113 . m. kochol : snarks without small cycles , j. combin . theory ser . b 67 ( 1996 ) , 3447 . l. lovsz , m. d. plummer : _ matching theory _ , volume 121 of north - holland mathematics studies . north - holland publishing co. , amsterdam , 1986 . annals of discrete mathematics , 29 . j. mazk : circular chromatic index of type 1 blanua snarks , j. graph theory 59 ( 2008 ) , 8996 . j. petersen : die theorie der regulren graphs , acta math . 15 ( 1891 ) , 193220 . w. t. tutte : the factorization of linear graphs , j. london math . soc . 22 ( 1947 ) , 107111 . a. vince : star chromatic number , j. graph theory 12 ( 1988 ) , 551559 . v. g. vizing : on an estimate of the chromatic class of a @xmath153-graph ( in russian ) , diskret . analiz . 3 ( 1964 ) , 2430 . x. zhu : circular chromatic number : a survey , discrete math . 229 ( 2001 ) , 371410 . x. zhu : recent developments in circular colorings of graphs , in m. klazar , j. kratochvl , j. matouek , r. thomas , p. valtr ( eds . ) : topics in discrete mathematics , springer , 2006 , 497550 .	we show that the circular chromatic index of a ( sub)cubic graph with odd - girth at least @xmath0 is at most @xmath1 .
this paper deals with the one parameter simultaneous approximation problem @xmath3 where @xmath0 are real numbers which we will assume to be linearly independent together with @xmath4 and @xmath5 are integers to be determined in dependence of the parameter @xmath6 in order to minimize @xmath7 . to be more precise , we define the function @xmath8 for @xmath9 by setting @xmath8 the minimum over all @xmath10 such that there are @xmath11 linearly independent vectors @xmath12 that satisfy the system ( [ eq : rof ] ) . in the sequel we will restrict to approximation vectors with @xmath13 , which clearly is no loss of generality as @xmath14 does not affect approximation constants . another equivalent way to view the functions @xmath15 is to consider the lattice @xmath16 and the convex body ( in fact the parallelepiped ) @xmath17 defined as the set of points @xmath18 with @xmath19 and to define @xmath20 as the @xmath11-th _ successive minimum _ of @xmath21 with respect to @xmath17 . this @xmath11-th minimum is defined as the infimum over all @xmath22 for which the @xmath23-span of @xmath24 has dimension at least @xmath11 , or equivalently @xmath25 contains @xmath11 linearly independent points of @xmath21 . + with respect to these successive minima @xmath26 , the functions @xmath8 can also be determined by @xmath27 one has the inequalities @xmath28 as we will show later , and dirichlet s theorem states @xmath29 minkowski s second convex body theorem yields for any convex body @xmath30 with volume @xmath31 and any lattice @xmath21 @xmath32 see @xcite , so that in our special case , as @xmath33 for every @xmath34 , we have @xmath35 uniformly in the parameter @xmath34 . with @xmath36 and taking logarithms , this yields @xmath37 with some constant @xmath38 not depending on @xmath34 . + another important property of the joint behaviour of the functions @xmath15 is that for any given @xmath39 there are arbitrarily large values @xmath40 such that @xmath41 provided that @xmath42 are linearly independent over @xmath43 , see theorem 1.1 in @xcite . to quantify the behaviour of @xmath8 summerer and schmidt introduced the quantities @xmath44 and gave the estimates @xmath45 where ( [ eq:1001alk ] ) requires @xmath42 to be linearly independent over @xmath43 again . each of these bounds will be shown to be best possible in corollary [ korol1 ] . moreover , ( [ eq : asdf ] ) implies @xmath46 in order to study the dynamical behaviour of the functions @xmath8 it will be convenient to work with functions @xmath47 as these functions are piecewise linear with slopes among @xmath48 . therefore we have ( [ eq : mbni ] ) . defintion ( [ eq : zokz ] ) is equivalent to @xmath49 we also introduce the classical approximation constants @xmath50 defined by jarnik , bugeaud in addition to @xmath51 . for fixed @xmath0 and for every @xmath52 define the functions @xmath53 as the supremum over all real numbers @xmath54 ( in fact the maximum ) such that the system @xmath55 has @xmath11 linearly independent solutions @xmath56 . the approximation constants @xmath50 are now defined as @xmath57 we will put @xmath58 and denote by @xmath59 the vector of classical approximation constants ( relative to @xmath0 ) . very similar to the proof of theorem 1.4 in @xcite , which treats the special case @xmath60 , one obtains @xmath61 one just needs to replace a solution by linearly independent solutions at any place it occurs in the proof . combining ( [ eq:2 ] ) with ( [ eq:1000alk]),([eq:1001alk ] ) for @xmath42 linearly independent over @xmath43 we obtain the bounds @xmath62 for the constants @xmath63 as well as @xmath64 for the constants @xmath65 . each considered individually , these bounds again are best possible . + in the present paper , we will put our focus on simultaneous approximation of numbers that allow good individual as well as simultaneous approximation . liouville numbers , that is real numbers @xmath66 for which the inequality @xmath67 has infinitely many rational solutions @xmath68 for arbitrarily large @xmath69 , will be suitable examples since they all satisfy @xmath70 , where @xmath71 is defined by ( [ eq:1 ] ) in the onedimensional case . + in section 2 , propositions [ propo ] , [ prop2 ] , we establish a connection between the @xmath1-adic expansions ( @xmath2 ) of the components @xmath72 of @xmath73 and the approximation constants @xmath74 . these results are then applied to the case where all @xmath72 admit good approximations in one fixed base @xmath1 independent of @xmath11 . after these considerations for suitable arbitrary @xmath73 we put our focus on liouville numbers , using heavily the fact that @xmath70 in this case . theorem [ satz1 ] will allow to compute all classical approximation constants @xmath50 for a special type of liouville numbers and the resulting corollary [ koroo ] will lead us to the construction of vectors @xmath73 with prescribed approximation constants @xmath50 that are subject to certain restrictions . as consequences of these results we will be able to give an explicit example of a vector @xmath0 that shows a conjecture by wolfgang schmidt concerning successive minima of a lattice to be true . a non - constructive proof was given by moshchevitin in a nonconstructive way . moreover we will construct cases where all functions @xmath15 simultaneously take all possible values of their spectrum for arbitrarily large @xmath34 . + inspired by methods used to deal with liouville numbers , we then gegeneralize theorem [ satz1 ] to a wider class of vectors @xmath73 for which @xmath75 . this will be the subject of theorems [ satz2],[satz6 ] and lead to many more explicit constructions of special cases of the schmidt conjecture . + in the last section we will first discuss the special case where @xmath76 for @xmath77 and give a constructive existence proof for the degenerate case @xmath78 in arbitrary dimension . troughout the paper we will illustrate the derived results by matlab plots of the functions @xmath79 for the special cases we consider to visualize derived results . these plots shall also lead to some insight into the dynamical behaviour of these functions in general . one should mention at this point that the plots often seem curved although the functions are piecewise linear , which is due to the non sufficient digital reslution , i.e. by zooming in one can see that they are indeed piecewise linear . in the sequel let @xmath2 be an integer and @xmath80 for @xmath81 . for each @xmath81 the non vanishing digits of the @xmath1-adic expansions of such @xmath82 and @xmath83 define two sequences @xmath84 and @xmath85 by @xmath86 we call the sequence @xmath87 the _ dual expansion _ of @xmath82 in base @xmath1 . set @xmath88 the monotonically ordered sequence of all @xmath89 and similarly @xmath90 the monotonically ordered sequence of all @xmath85 . the following theorem expresses the simultaneous approximation constant @xmath91 of @xmath0 in terms of the @xmath1-adic presentations of @xmath82 @xmath92 by using these two orderd sequences . the proof is introductory to the rest of the work and for this purpose quite detailed . [ propo ] we have @xmath93 where @xmath94 ( or any other norm since they are all equivalent in @xmath95 ) . we first prove ( [ eq:3 ] ) . by definition of @xmath88 as the mixed ordered sequence of the sequences @xmath96 , all numbers @xmath0 will have zeros at the positions @xmath97 behind the comma in base @xmath1 for any @xmath2 . since multiplication of @xmath72 by @xmath98 only shifts the comma @xmath99 positions to the right , this means , for any @xmath77 all @xmath100 start with @xmath101 zeros in base @xmath1 behind the comma . for this reason any pair @xmath102 satisfies @xmath103 for any @xmath77 , where @xmath104 denotes the smallest distance of a real number to an integer . analoguously , for all @xmath105 and all pairs @xmath102 we have @xmath106 we conclude that for any pair @xmath102 @xmath107 surely , @xmath108 or @xmath109 is equivalent to @xmath110 , and we claim that ( [ eq:3 ] ) follows directly from the definition of the approximation constant @xmath91 . to see this we take a sequence of pairs @xmath111 with @xmath112 , for which @xmath113 or @xmath114 tend to the @xmath115-values on the right hand side of ( [ eq:3 ] ) . putting @xmath116 or @xmath117 , where @xmath118 is an arbitrary bijection @xmath119 , we obtain a sequence of @xmath120-values and @xmath121-values that leads via ( [ eq : rmna ] ) to an approximation constant @xmath91 in ( [ eq:1 ] ) at least as large as both @xmath115-values . + to prove ( [ eq : qpqp ] ) , we first show the following assertion : it suffices to prove , that for any sufficiently large real parameter @xmath120 there is a @xmath122 , such that @xmath123 where @xmath124 is the largest exponent for which @xmath125 holds for all @xmath126 . + for any sequence @xmath127 let @xmath128 be the largest exponent , for which ( [ eq : merz ] ) holds with @xmath129 in place of @xmath54 for all @xmath130 . the existence of @xmath122 with ( [ eq : anelim ] ) for any @xmath120 implies the existence of a sequence @xmath131 with @xmath132 for all @xmath133 with @xmath134 of the shape @xmath135 hence of the shape of the expressions involved in ( [ eq : qpqp ] ) in the case @xmath136 . by definition of @xmath91 we may choose the sequence @xmath127 such that @xmath137 . furthermore we can assume without loss of generality that @xmath127 satisfies @xmath138 for any @xmath133 , as the exponent @xmath54 in the definition of @xmath91 in ( [ eq:1 ] ) for a fixed @xmath121 decreases with growing @xmath120 . combining all these observations we get @xmath139 where @xmath134 fits in the @xmath115 term of ( [ eq : qpqp ] ) if we set @xmath140 , where @xmath141 plays the role of @xmath142 above , for @xmath143 . + it remains to prove that for such sequences we have @xmath144 . this , however , is easy to see . as @xmath138 the definition of @xmath141 guarantees that the number @xmath145 minimizes @xmath146 among all @xmath147 . on the other hand clearly @xmath148 for any @xmath43-linearly independent @xmath149 and so by definition of @xmath150 we also have @xmath151 . consequently the sequence @xmath150 can not be bounded as only finitely many ( strictly positive ) values @xmath152 would appear , which proves @xmath153 . + to complete the proof we have to find a value @xmath122 for which ( [ eq : anelim ] ) holds . note first , that for sufficiently large @xmath120 und @xmath154 we have @xmath155 . indeed for @xmath156 and @xmath157 the index , for which the minimum is attained , we have @xmath158\cup[\frac{s-1}{s},1)\}}$ ] , so the first digit after the coma in base @xmath1 is neither @xmath159 nor @xmath160 . so we can assume @xmath120 to be large enough to ensure @xmath161 for all @xmath77 and hence @xmath162 as well . it is now easy to see that putting @xmath163 is an appropriate choice , since ( [ eq : merz ] ) says that all @xmath164 respectively @xmath165 start with @xmath166 digits zero in base @xmath1 behind the coma . this yields @xmath167 for all @xmath77 , therefore @xmath168 respectively the same facts for @xmath169 . we easily deduce the following corollary : [ korolar1 ] we have @xmath170 similarly , we can give a lower bound for @xmath171 with respect to the @xmath1-adic representation of a real number . [ prop2 ] for any @xmath172 we have @xmath173 by definition of the supremum it is suffient to prove @xmath174 for any base @xmath1 separately . so let @xmath1 be fixed and put @xmath175 . by definition of @xmath171 for arbitrary @xmath176 and sufficiently large @xmath177 we have to find an approximation vector @xmath56 with @xmath178 and @xmath179 for @xmath176 and large @xmath120 let @xmath180 be defined by @xmath181 respectively @xmath182 . put @xmath183 respectively @xmath184 where @xmath11 is the index , such that the inner maximum from the definition of @xmath185 is attained for the given @xmath180 . by definition of @xmath186 as the mixed sequence the first @xmath187 positions behind the comma of each @xmath188 , respectively @xmath189 , are zeros in base @xmath1 . we infer that putting @xmath190 for all @xmath191 respectively @xmath192 for all @xmath191 we have @xmath193 for the left hand side inequalties compare the proof of proposition [ prop1 ] , the right hand side inequalities follow from @xmath194 and @xmath195 respectively . as ( [ eq : rumenige]),([eq : rummenige ] ) holds for every large @xmath120 , we may let @xmath196 tend to @xmath197 to conclude that ( [ eq : mileniza ] ) has a solution for all sufficiently large @xmath120 . hence the exponent of @xmath120 in ( [ eq : rumenige ] ) and ( [ eq : rummenige ] ) respectively is larger than @xmath198 . now we turn to simultaneous approximation of vectors @xmath73 with good approximation in one fixed simultaneous base @xmath2 , and we skip the dual expansion . we want to use corollary [ korolar1 ] and proposition [ prop2 ] to give estimates for the simultaneous approximation constants @xmath74 . with respect to the notation above , meaning that the @xmath1-adic digits of @xmath82 are given by @xmath199 as in ( [ eq : ttt ] ) and the ordered mixed sequence by @xmath88 , we get [ lemma1 ] for any @xmath2 we have @xmath200 the right hand inequality is trivial . for the left hand inequality keep @xmath1 fixed and put @xmath201 and choose @xmath180 large enough , such that for all @xmath133 and all @xmath202 we have @xmath203 ( @xmath1 has been dropped in the notation ) . for arbitrary @xmath204 , there exist @xmath205 with @xmath206 by definition of @xmath186 . the interval @xmath207 $ ] contains at most @xmath208 numbers @xmath209 , since it contains at most one element of every sequence @xmath210 for @xmath81 . by the pigeon hole principle there are two numbers @xmath211 in the interval @xmath212 $ ] whose quotient @xmath213 is at least @xmath214 . the lemma follows with @xmath215 . in combination with corollary [ korolar1 ] we observe @xmath216 getting lower bounds for @xmath171 by just considering the sequences @xmath217 is more complicated and to some extent impossilbe as we will see in corollary [ korol1 ] . in fact even if @xmath218 we can have @xmath219 , which is the weakest lower bound for @xmath171 by ( [ eq:2001neu ] ) . we only mention that if we construct sequences @xmath217 for which @xmath220 , proposition [ prop2 ] yields @xmath221 and consequently @xmath222 in view of ( [ eq:2001neu ] ) . in the following theorem , we compute the classical approximation constants @xmath50 for a special type of liouville numbers @xmath0 , whose best approximation vectors @xmath223 to @xmath73 are easy to guess . the main arguments of the compilation will be carried out in the proofs of the following theorems . [ satz1 ] let @xmath208 be a positive integer and for @xmath77 let @xmath224 , where @xmath225 are natural numbers , such that @xmath226 and such that @xmath227 where @xmath228 satisfy @xmath229 then the classical approximation constants relative to the vector @xmath230 are given by @xmath231 and @xmath232 we start with the constants @xmath63 and intend to prove the inequalities @xmath233 and @xmath234 seperately for @xmath9 . + @xmath235 + let @xmath236 be the @xmath237-th quotient of the maximum labeled @xmath238 . we give a detailed proof of @xmath239 and then mention how to generalize the proof to derive all the other inequalities @xmath240 for @xmath241 . + to prove @xmath242 , we will construct @xmath11 sequences of approximation vectors @xmath243 which are linearly independent for each fixed @xmath244 and such that @xmath245 follows for @xmath246 . indeed for @xmath247 in a @xmath11-element subset of @xmath248 and any @xmath176 we claim for @xmath133 sufficiently large @xmath249 in analogy to the definition of @xmath250 in subsection 2.1 let @xmath186 be the combined sequence of the logarithms of the integers @xmath251 in increasing order , which means for any nonnegative integer @xmath252 and @xmath253 we have @xmath254 . by ( [ eq : nurrr ] ) we have @xmath255 and thus by putting the first approximation vector @xmath256 with arbitrary @xmath133 we may let @xmath196 tend to infinity , to obtain @xmath70 : indeed applying ( [ eq : erwin ] ) we derive that all the remainder terms @xmath257 are small due to ( [ eq : nurrrrr])-([eq : nurrr ] ) . in order to estimate @xmath63 for @xmath258 we construct a sequence of parameters @xmath120 and approximation vectors @xmath259 with @xmath178 explicitely . for the fixed choice @xmath260 we will get @xmath242 . to see this let @xmath261 and @xmath262 for @xmath191 . define the second approximation vector by taking @xmath263 and again @xmath264 . by means of ( [ eq : nurrrrr]),([eq : nurrrr]),([eq : nurrr ] ) and the definition of @xmath265 we claim that for each @xmath266 @xmath267 holds for @xmath268 large enough . this follows from @xmath269 in view of the definition of @xmath270 and our assumption ( [ eq : erwin ] ) . in every case all the values of @xmath271 for @xmath191 are bounded by @xmath272 . using ( [ eq : nurrrrr]),([eq : nurrrr]),([eq : nurrr ] ) this leads to ( [ eq : mota ] ) . + similarly , defining the @xmath11-th approximation vector for @xmath273 by @xmath274 and for @xmath275 by @xmath276 and then putting @xmath277 yields the corresponding inequalities . + we now check that these vectors are linearly independent as required . to do this we prove that all the matrices @xmath278 obtained by writing the @xmath279-th approximation vector @xmath280 in the @xmath279-th row , i.e. @xmath281 are nonsingular . observe that if we subtract @xmath282 times the @xmath283-th row from the @xmath279-th row of the matrix @xmath284 , all entries in the new @xmath279-th line will be zero apart from a one in position @xmath285 . starting with this process at @xmath286 and repreating it until @xmath287 we end up with the matrix @xmath288 which is easily seen to have absolute value of the determinant equal to @xmath289 . therefore also @xmath290 , as required . + to obtain all the other inequalities @xmath291 for @xmath292 , where the upper bound of @xmath133 depends on @xmath11 , we proceed analoguously . in the definition of @xmath293 we replace @xmath294 by @xmath295 and again for @xmath191 we define @xmath296 for the first approximation vector . we define all the others by taking @xmath297 and again @xmath298 for @xmath191 and @xmath299 . this construction yields the desired lower bounds ( or @xmath159 which is omitted in @xmath300 ) as above again by ( [ eq : nurrrrr]),([eq : nurrrr]),([eq : nurrr ] ) . + @xmath301 + we have to show that for @xmath9 the approximation vectors @xmath302 constructed in the first step of the proof are somehow best possible . we split the proof of this assertion in 3 steps . to simplify notation let @xmath303 be the ordered mixed sequence @xmath304 . + first step : for an arbitrary approximation vector @xmath305 let @xmath279 be the index determined by @xmath306 and let @xmath307 be the largest integer such that the index @xmath308 satisfies @xmath309 . since @xmath310 and consequently @xmath311 we clearly have @xmath312 . when @xmath313 so does @xmath279 and we claim that for @xmath314 @xmath315 furthermore in the case @xmath316 ( i.e. the assumption of ( [ eq : memina ] ) ) , the inequality @xmath317 contradicts that @xmath318 holds for @xmath314 . + second step : let @xmath120 be a real parameter from the definition of the approximation constants @xmath63 and @xmath319 be the index such that @xmath320 . then for @xmath9 a set of @xmath11 vectors @xmath321 , satisfying the inequalities @xmath322 can only be linearly independent if at least one @xmath323 is not divisable by @xmath324 . + third step : we intend to show by combining the first two steps and using ( [ eq : nurrrrr])-([eq : nurrr ] ) , that for arbitrary @xmath120 the choice of approximation vectors in the proof of @xmath233 is somehow optimal , i.e. the approximation constants of this case can not be improved . + proof of first step : we first make the assumption @xmath325 , and will explain at the end how to extend this easily to the case where @xmath326 belongs to another sequence . this assumption is equivalent to @xmath327 for some @xmath328 and it follows that @xmath329 . by ( [ eq : erwin ] ) we have @xmath330 for all @xmath331 and @xmath332 for all @xmath333 , in particular @xmath334 . recall @xmath312 . to prove the assertions we now consider the corresponding cases seperately : + case 1 : @xmath335 . note that since @xmath336 this is equivalent to @xmath337 or @xmath338 . we can write @xmath339 with @xmath340 and @xmath341 , since by our definition of @xmath307 we have @xmath342 . denote by @xmath343 the congruence class of @xmath307 in the residue system @xmath248 mod @xmath208 . note , that @xmath344 is the smallest value @xmath345 with @xmath307 in the residue class @xmath343 , or equivalently @xmath346 , which we will make use of . we claim that @xmath347 where @xmath348 denotes the closest distance to an integer an @xmath349 the fractional part of a real number . inequality ( [ eq : trunschuh ] ) relies on the fact that @xmath350 and @xmath351 , which is seen to be true because @xmath352 and @xmath341 by definition , so putting these together we get @xmath353 , but @xmath354 . combination of these two facts and recalling that exactly @xmath355 are the indices greater than @xmath307 belonging to the residue class @xmath343 shows that @xmath356 is of the form @xmath357 with @xmath358 ( note @xmath359 ) . the assertion now follows by a combination of @xmath360 , @xmath361 and @xmath362 , which is true by the assumption of case 1 . + inequality ( [ eq : truhnschuh ] ) follows from the fact that for any @xmath363 by virtue of ( [ eq : erwin ] ) we have @xmath364 which holds in particular for those @xmath365 with @xmath1 in the residue class @xmath343 . so all quantities @xmath366 with @xmath363 are integers . thus the sum of quantities of order smaller than @xmath367 in @xmath368 , i.e. @xmath369 , has the same fractional part as the entire sum . now on the one hand we have @xmath370 , and on the other hand @xmath371 by the assumption of case 1 . together with ( [ eq : pfred ] ) these assertions yield ( [ eq : truhnschuh ] ) . + summing ( [ eq : trunschuh ] ) and ( [ eq : truhnschuh ] ) and noting that by ( [ eq : nurrr ] ) we have @xmath372 , so that we can use the triangle inequality on the fractional parts , we further have @xmath373 by ( [ eq:18neu]),([eq:17neu ] ) and ( [ eq : nurrr ] ) the expression @xmath374 is montonically decreasing in @xmath375 and the error term @xmath376 is obviously @xmath377 by ( [ eq : nurrr ] ) . hence for @xmath314 we obtain @xmath378 this establishes ( [ eq : mimena ] ) in this case as @xmath379 by ( [ eq:18neu ] ) and ( [ eq : nurrrrr])-([eq : nurrr ] ) . + if @xmath326 belongs to another sequence @xmath380 , which means @xmath381 with @xmath382 , we look at the case @xmath383 and again obtain @xmath384 as in the proof of the special case ( without using @xmath385 as above from which we derived the weaker but sufficient conditions ( [ eq : trunschuh]),([eq : truhnschuh ] ) ) . however , by ( [ eq : nurrrrr])-([eq : nurrr ] ) we again have @xmath386 as @xmath387 ( or equivalently @xmath388 as @xmath337 ) and the rest of the argumentation is almost as above . thus ( [ eq : mimena ] ) holds in any case . + case 2 : @xmath389 . in this case it is more convenient to work directly with the values @xmath390 instead of @xmath391 . as in case 1 let @xmath339 with @xmath392 and @xmath393 . again @xmath394 and the definition of @xmath307 ensures @xmath342 . analoguously to the proof of ( [ eq : trunschuh]),([eq : truhnschuh ] ) in case 1 we deduce @xmath395 using again the triangle inequality and ( [ eq:17neu ] ) , we again deduce @xmath396 but by ( [ eq : nurrrrr])-([eq : nurrr ] ) again @xmath397 for @xmath314 so that finally @xmath398 but @xmath399 , so we have ( [ eq : memina ] ) in this case . if @xmath326 belongs to another sequence @xmath380 , @xmath400 , we can apply very similar estimates with respect to @xmath401 instead of @xmath402 . so our assumption is no loss of generality in this case either . thus ( [ eq : memina ] ) holds in any case . + we still have to prove that @xmath403 contradicts ( [ eq : wilena ] ) . for simplicity we again discuss the case @xmath325 first . write @xmath339 with @xmath404 and @xmath405 . note that again we have @xmath342 by the assumption @xmath316 , so @xmath406 , and the definition of @xmath307 . assume we have @xmath407 . the fractional part of @xmath408 is @xmath409 as higher order summands are integers by definition of @xmath410 . we split this expression in @xmath411 and using @xmath412 we infer @xmath413 which is obviously @xmath414 as @xmath314 by ( [ eq : nurrr ] ) . + on the other hand , by definition of @xmath415 and @xmath416 as @xmath316 by assumption and a very similar argument as in case 1 we have @xmath417 on the other hand by @xmath418 the sum of the remainder terms of @xmath419 , i.e. @xmath420 , is bounded above by @xmath421 with very similar estimates as in ( [ eq : pfred ] ) . so @xmath422 by a very similar argument as in case 1 . as @xmath423 , we have @xmath424 ( note that @xmath314 if @xmath423 ) and thus by ( [ eq : dranele ] ) @xmath425 using triangular inequality on ( [ eq : elenard]),([eq : drenale ] ) thus gives @xmath426 our last assertion is proved in this case and the assumption @xmath427 can obviously be dropped again . + proof of second step : without loss of generality assume that @xmath428 , the proof for the other cases is essentially the same . this means @xmath429 for some @xmath430 and consequently @xmath431 . suppose @xmath432 divides @xmath323 for all @xmath433 . on the one hand , by our assumption the @xmath434 numbers @xmath435 belong to the sequences @xmath436 . on the other hand , @xmath437 for all @xmath438 combined with @xmath439 for all @xmath433 implies that for all @xmath440 the number @xmath365 divides @xmath323 . from these two facts we conclude that for @xmath441 , i.e. @xmath442 , the partial sum @xmath443 of @xmath444 is an integer , since every summand @xmath445 is . as terms of order lower than @xmath446 in @xmath444 for @xmath447 obviously add up to a quantity smaller than @xmath448 , for @xmath433 and @xmath447 we have @xmath449 and combined with ( [ eq:29neu ] ) eventually @xmath450 in view of our assumption @xmath451 the results ( [ eq : haarz]),([eq : zraah ] ) are also valid for @xmath452 . to sum up , for all @xmath453 we have ( [ eq : zraah ] ) , which obviously yields @xmath454 thus in the matrix , whose @xmath133-th row is the @xmath133-th approximation vector @xmath455 ( @xmath433 ) , the first @xmath456 columns together have rank @xmath4 . the rank of the whole matrix therefore can not exceed @xmath457=j-1<j$ ] . this means the @xmath11 rows are linearly dependent , a contradiction . so @xmath432 can not divide all the numbers @xmath323 , as stated . + proof of third step : we will prove for arbitrary @xmath11 , that @xmath53 is for @xmath313 asymptotically bounded above by one of the fractions ( depending on @xmath458 ) involved in the definition of @xmath238 , by which we mean that for any @xmath176 and @xmath177 large enough we have @xmath459 . since @xmath460 , @xmath461 shows the required result . + so let @xmath120 be arbitrary but fixed and let @xmath279 be the index determined by @xmath462 . we first prove that without loss of generality we may restrict to the case where @xmath120 lies an interval of the shape @xmath463 . + this is the case because the logarithm to the base @xmath464 of @xmath465 for vectors @xmath466 with @xmath467 is asymptotically the same as to the base @xmath326 . indeed we have @xmath468 and hence @xmath469 on the other hand , since we can restrict to @xmath466 belonging to some @xmath63 with @xmath258 , all expressions @xmath470 are bounded above by @xmath471 ( see ( [ eq:2222neu ] ) ) for @xmath279 sufficiently large . together with ( [ eq : zarteawa ] ) and since this holds for every vector @xmath466 for which @xmath178 , the definition of the quantities @xmath63 immediately implies that they remain unaffected by this change of base . + so let @xmath472 be the index determined by @xmath473 . by the second step of the proof ( putting @xmath474 ) at least one of the @xmath11 linearly independent approximation vectors @xmath475 has to satisfy the condition @xmath476 . consider one of the @xmath11 approximation vectors with this property . this means if we let @xmath308 be the largest index with @xmath309 as in step 1 , we have @xmath477 , i.e. @xmath478 . further let @xmath133 be the index , for which @xmath479 belongs to the sequence @xmath380 . at this point one should mention that we will repeatedly use step 1 in the following , neglecting the @xmath480-terms in the estimates ( [ eq : mimena]),([eq : memina]),([eq : wilena ] ) as they do not affect the asymptotic behaviour we aim to prove . + first we treat the case @xmath481 ( case 1 step 1 ) . note , that @xmath482 is monotonically decreasing as @xmath375 increases by ( [ eq : nurrrrr])-([eq : nurrr ] ) and ( [ eq:18neu ] ) , which we already used before . thus by @xmath478 and ( [ eq : mimena ] ) we have @xmath483 so @xmath484 implies @xmath485 it is now easy to see by ( [ eq : nurrrrr])-([eq : nurrr ] ) that for @xmath279 in a fixed residue class @xmath486 of the residue system @xmath248 mod @xmath208 , the right hand side tends to one of the fractions ( depending on @xmath486 ) in the definition of @xmath487 or to zero as @xmath314 or equivalently @xmath313 . clearly , each expression in @xmath238 is induced by some @xmath486 in that way as well . this shows , that indeed we have @xmath488 for any @xmath176 and @xmath177 large enough . + in the remaining case @xmath489 ( case 2 step 1 ) by ( [ eq : memina ] ) and as @xmath326 in ( [ eq : memina ] ) corresponds to @xmath490 , @xmath491 is essentially bounded below by @xmath492 ( omitting the lower order terms and @xmath133 in the residue system @xmath248 mod @xmath208 ) . we distinguish three cases now . + if we have @xmath493 , approximation relative to base @xmath120 is bad , as in this case we have @xmath494 as a consequence of ( [ eq : nurrr ] ) , so again by @xmath495 , the expression @xmath496 tends to @xmath159 as @xmath313 , and we are done again . + in the case @xmath497 , or equivalently @xmath498 , due to @xmath499 we know again by ( [ eq : memina ] ) that @xmath500 and so @xmath501 implies @xmath502 hence by ( [ eq : nurrr ] ) , for @xmath314 we have the asyptotic @xmath503 the right hand side , however , converges to @xmath504 for @xmath505 by ( [ eq : nurrrrr ] ) . this shows that @xmath506 for any @xmath176 and @xmath177 large enough and together with @xmath507 for @xmath9 we get @xmath234 as desired . + we devide the remaining case @xmath508 , which means @xmath509 , again into two cases . if @xmath510 , it follows that @xmath491 is essentially bounded below by @xmath511 in view of ( [ eq : wilena ] ) . this gives the estimate @xmath512 which tends to @xmath159 for @xmath313 by ( [ eq : nurrr ] ) . + otherwise @xmath513 , which clearly implies @xmath514 by ( [ eq : nurrr ] ) . in particular for every @xmath176 we have @xmath515 for @xmath516 sufficiently large . note @xmath517 and that @xmath313 is equivalent to @xmath505 . combination of these facts together with the fact that @xmath491 is essentially bounded below by @xmath518 by ( [ eq : memina ] ) yields the inequality @xmath519 for any @xmath176 and @xmath516 sufficiently large . however , the right hand side is of the form @xmath520 as @xmath505 and @xmath461 , which tends to @xmath504 for @xmath505 as in the case @xmath497 above and so it is no improvement either . this shows step three . + now it only remains to determine the approximation constants @xmath65 . however , for @xmath258 they are easily seen to be zero as a consequence of @xmath70 . indeed in this case we have @xmath78 by ( [ eq:2 ] ) and if for some @xmath176 we had @xmath521 , we would obtain @xmath522 for a sequence of arbitrary large values @xmath523 , a contradiction to ( [ eq : zokz ] ) . so ( [ eq:2 ] ) again yields @xmath524 for @xmath525 . + it remains to determine @xmath171 . let @xmath120 be a real number of the form @xmath526 , so that in particular we have @xmath527 . + putting @xmath60 in ( [ eq : limeta ] ) and noting @xmath528 in the case @xmath337 we obtain @xmath529 if we now fix a residue class @xmath486 for the values of @xmath279 in the residue system @xmath248 mod @xmath208 , the right hand side of ( [ eq : harzharzl ] ) tends to one of the fractions in the definition of @xmath530 as @xmath314 . in fact , as @xmath486 runs through the residue system @xmath248 mod @xmath208 this induces a bijection between the residue classes @xmath486 of the residue system @xmath248 mod @xmath208 and the expressions of @xmath171 . + in the case @xmath316 as @xmath491 is essentially bounded below by @xmath492 again , we have the upper esitmate @xmath531 the right hand side , however , is smaller than the corresponding value in ( [ eq : harzharzl ] ) for every @xmath279 , so the case @xmath316 does never give any improvement . thus , by its definition , the quantity @xmath171 can be estimated above by the minimum of the expressions of @xmath530 , which simply is @xmath530 . + on the other hand , fixing a residue classes @xmath486 for @xmath279 in the residue system @xmath248 mod @xmath208 again and putting @xmath532 and @xmath533 for @xmath191 , we obtain a bijection between the resulting values @xmath534 as @xmath486 runs through @xmath535 and the expressions involved in the definition of @xmath536 . hence @xmath171 is at least as large as the minimum of these expressions , which again is @xmath171 . note , that in the special case @xmath537 we have @xmath538 , and the approximation constants of @xmath539 in theorem [ satz1 ] become @xmath540 particularly , we see that @xmath541 , which is easily seen by straight forward computation with repeated use of ( [ eq:2 ] ) to be equivalent to jarnik s identity in the form @xmath542 , see theorem 1.5 in @xcite . + so far we have not asked for the numbers @xmath0 to be @xmath43-linearly independent together with @xmath4 , which is the usual assumption . for this purpose , we apply theorem [ satz1 ] in the special case @xmath543 with suitable sequences @xmath544 for @xmath77 . [ koroo ] for @xmath77 let @xmath544 be sequences with the properties @xmath545 and for @xmath546 as in theorem [ satz1 ] put @xmath547 and @xmath548 for @xmath77 . then the corresponding approximation constants are given as in theorem [ satz1 ] . clearly , if we put @xmath549 , all conditions of theorem [ satz1 ] are satisfied . now one can easily prove that there are uncountably many vectors @xmath550 , such that additionally @xmath42 are linearly independent over @xmath43 . the arguments of the proof of the following proposition [ prop1 ] are suitable to prove the existence of vectors @xmath73 for which @xmath42 are linearly independent subject to certain approximation properties if this existence was established without the linear independence condition . [ prop1 ] one can choose sequences @xmath544 in corollary [ koroo ] such that @xmath551 is linearly independent over @xmath43 . note that in the case @xmath552 ( [ eq : nurrrrr])-([eq : nurrr ] ) simply yield @xmath553 and it follows from liouville s theorem that the corresponding number of the form @xmath554 is transcendental , in particular @xmath555 is linearly independent over @xmath43 . in the case @xmath556 consider the numbers @xmath557 for @xmath77 and define the sequences @xmath544 by the recurrence relations @xmath558 one checks that ( [ eq:23neu])-([eq:26neu ] ) imply ( [ eq : nurrrrr])-([eq : nurrr ] ) . now we prove that we can change ( [ eq:23neu])-([eq:26neu ] ) slightly such that @xmath551 is linearly independent over @xmath43 . + let @xmath186 be the ordered combined set of all @xmath559 defined as above . note that we can obviously disturb the system ( [ eq:23neu])-([eq:26neu ] ) a little by adding one to the elements of the form @xmath560 where @xmath561 with @xmath562 an arbitrary subset of @xmath563 , without violating ( [ eq:23neu])-([eq:26neu ] ) . noting that ( [ eq:23neu])-([eq:26neu ] ) imply @xmath564 , by the considerations of the case @xmath552 we know @xmath565 is a @xmath43-linearly independent set , where @xmath402 is generated by the sequence @xmath566 defined above , ie @xmath567 . if we now consider the set of all sequences @xmath568 which arise from @xmath569 as above by adding @xmath4 to @xmath560 with @xmath570 for an arbitrary subset @xmath571 of @xmath563 , we see that @xmath572 is uncountable . so there must be a number @xmath573 generated by an @xmath574 with the property that @xmath575 are linearly independent over @xmath43 , since the @xmath576-span of @xmath565 is only countable . now we proceed analoguously with sets @xmath577 for @xmath578 and finally get a @xmath43-linearly independent set @xmath579 . as mentioned above , the set @xmath580 fulfills all the requirements . note : since algebraic numbers have countable cardinality one can readily generalize the proof above to show that we can even ask @xmath539 to be algebraically independent . + we now give some applications of the above theorem . note that ( [ eq : mbni]),([eq:11 ] ) and ( [ eq:1000alk ] ) imply that for all @xmath176 and sufficiently large @xmath581 we have the bounds @xmath582 in the first corollary we construct @xmath0 for which each @xmath8 takes each of the values inside of the corresponding intervals @xmath583 for arbitrarily large @xmath34 simultaneously for all @xmath9 . so roughly speaking in this case all @xmath15 take their possible range of values for arbitrarily large @xmath584 . in particular the bounds in ( [ eq:1000alk ] ) are best possible . [ korol1 ] there exist @xmath0 for which the set @xmath585 is @xmath43 linearly independent and such that @xmath586 note that by means of proposition [ prop1 ] for every @xmath587 subject to the restrictions of theorem [ satz1 ] we can construct @xmath230 together with @xmath4 linearly independent over @xmath43 such that theorem [ satz1 ] holds . putting @xmath588 in theorem [ satz1 ] immediately gives all the stated equalities for this @xmath539 . now we want to give our first explicit construction of special cases of schmidt s conjecture , which was proved by moshchevitin in a nonconstructive way in @xcite . the conjecture states , that for each integer pair @xmath589 with @xmath590 there exists a vector @xmath550 with @xmath591 linearly independent over @xmath43 such that @xmath592 and @xmath593 . note that we can not have @xmath594 for any @xmath133 because of the assumption of linear independence because of ( [ eq : asdf ] ) , see also the introduction in @xcite . we now give a generalisation of this fact in the special case @xmath595 for arbitrary @xmath556 . let @xmath556 and @xmath596 be integers . then there exists @xmath550 with @xmath591 linearly independent over @xmath43 such that @xmath597 the case @xmath598 clearly implies schmidt s conjecture for @xmath599 . we may assume @xmath600 , because for @xmath537 the corollary only states that @xmath601 and @xmath602 is possible , which only requires @xmath601 by ( [ eq : zokz ] ) and for any choice of @xmath603 the construction of theorem [ satz1 ] gives an example . we apply theorem [ satz1 ] with @xmath604 defined by @xmath605 the parameter @xmath606 will be chosen later in dependence of @xmath607 . first note that by ( [ eq:2 ] ) our inequalities translate to @xmath608 now note , that the left hand side of ( [ eq : a2 ] ) trivially holds by theorem [ satz1 ] . with ( [ eq : b1]),([eq : b2 ] ) the quantity @xmath530 of theorem [ satz1 ] becomes @xmath609 moreover the assumption @xmath610 implies that ( [ eq : a1 ] ) holds . to obtain the remaining inequalities it is obviously sufficient to prove @xmath611 for some @xmath612 , which by ( [ eq : b1]),([eq : b2 ] ) and theorem [ satz1 ] is equivalent to @xmath613 since the last term in @xmath238 is easily seen to be the largest in @xmath614 in our special case of constant ratios . for @xmath598 this reduces to @xmath615 , which is obviously true if we choose any @xmath616 , so we can assume @xmath617 . defining the functions @xmath618 shows that ( [ eq:300 ] ) in the cases left to consider is equivalent to @xmath619 for @xmath620 . it is easy to check that all these @xmath621 are continuous , @xmath622 and @xmath623 . further more from @xmath624 we deduce that they are monotonically increasing in @xmath612 . combination of these properties implies that for fixed @xmath625 there exists some @xmath626 such that @xmath627 by intermediate value theorem . it further follows from these considerations on the one hand @xmath628 for @xmath629 , and on the other hand the existence of an interval @xmath630 with @xmath631 such that @xmath632 . thus , for all @xmath633 we have @xmath619 . we now aim to give similar results for vectors @xmath230 whose components @xmath72 have one - dimensional approximation constant @xmath75 . hence in particular the simultaneous aproximation constant @xmath91 is finite too , as by definition it can not exceed the minimum of the one - dimensional constants . as in theorem [ satz1 ] each @xmath72 will be the sum of the re ciprocals of integers @xmath251 that satisfy ( [ eq : qq]),([eq : erwin ] ) , but the conditions ( [ eq : nurrrrr])-([eq:17neu ] ) will be altered in ways to obtain a symmetric situation in all @xmath72 , which will be more convenient for the purposes of chapter 2 . we start with proving the following [ satz2 ] for @xmath77 let @xmath634 where @xmath635 are sequences of integers for which ( [ eq : qq]),([eq : erwin ] ) are satisfied and that for @xmath636 the inequality @xmath637 is satisfied . then the first @xmath638 approximation constants relative to @xmath0 are given by @xmath639 further more we have the inequlities @xmath640 denote the right hand side expressions by @xmath614 respectively @xmath641 . we prove the inequalities @xmath642 for @xmath9 and then @xmath643 for @xmath644 . this obviously yields the assertions of the theorem . + throughout the proof , for arbitrary @xmath120 let the integer @xmath279 be determined as the index for which @xmath527 . obviously @xmath313 is equivalent to @xmath314 , which will often be used implicitly . we first prove @xmath645 for @xmath9 . + assume @xmath11 arbitrary but fixed . observe that for every @xmath279 if we put @xmath646 , @xmath647 for @xmath191 and @xmath433 , we obtain @xmath648 with @xmath649 , as explained in the proof of theorem [ satz1 ] . choose an integer sequence @xmath650 of values for @xmath279 in ( [ eq : tiryl ] ) such that with @xmath651 we have @xmath652 which is possible by definition of the @xmath115 . leaving out the lineary independence condition , the so constructed vectors @xmath321 , lead to the values @xmath614 as lower bounds for @xmath63 in view of ( [ eq : tiryl ] ) . however , the missing lineary independence condition is now obtained exactly as in theorem [ satz1 ] . + to prove @xmath653 , consider any sequence @xmath650 and the same approximation vectors as in the proof of @xmath645 but take logarithms to base @xmath654 instead of base @xmath655 . this yields a lower estimate for @xmath53 for @xmath656 . as this is valid for any sequence @xmath650 we obtain the lower bounds @xmath657 for @xmath65 by definition of @xmath658 , as we claimed . so far we have established the lower bounds @xmath614 ( respectively @xmath641 ) for @xmath63 ( respectively @xmath65 ) for @xmath9 . + for the upper bounds note first that with basically the same arguments as in the proof of step 3 in theorem [ satz1 ] we can restrict to the case @xmath659 . further step 1 and step 2 of @xmath660 in the proof of theorem [ satz1 ] remain valid in the present situation . indeed , the estimates ( [ eq : mimena]),([eq : memina ] ) are already valid under the assumtion @xmath661 , which is weaker than ( [ eq : nurrr ] ) used in theorem [ satz1 ] . the proof of step 2 is analoguous . + now for every fixed @xmath120 we divide all approximation vectors @xmath305 with @xmath178 into two categories . let @xmath307 be the largest integer such that @xmath309 for an approximation vector @xmath305 as in step 1 of theorem [ satz1 ] . the distinction of vectors with @xmath662 , which we will call vectors of category 1 , and @xmath316 , which we will call vectors of category 2 , now leads to 2 cases . + case 1 : if for fixed @xmath120 with @xmath663 we have @xmath664 for an approximation vector ( i.e. it belongs to category 1 ) , ( [ eq : mimena ] ) implies that @xmath665 so we have that for every @xmath120 the quantity @xmath614 is an upper bound for @xmath666 , by which we mean the supremum over all real numbers @xmath54 such that ( [ eq:1 ] ) has @xmath11 linearly independent vector solutions all of which are of the first category . hence if we define @xmath667 , we get @xmath668 in order to give a connection between the approximation constants @xmath65 and approximation vectors of category 1 , we define @xmath669 . we start with an arbitrary sequence @xmath670 with corresponding subsequence @xmath671 of @xmath672 and define a sequence @xmath673 by putting @xmath674 . in view of ( [ eq : mimena ] ) and observing that the fractions @xmath675 are monotonically decreasing by our assumption @xmath661 , we get the upper estimate for the approximation constants @xmath676 @xmath677 for @xmath9 . if we specify a sequence @xmath670 for which the corresponding sequence @xmath650 has the property @xmath678 which is possible again by definition of @xmath658 , we put @xmath679 in the definition of @xmath641 so that the right hand side of ( [ eq : tomoy ] ) tends to @xmath641 as @xmath680 . thus @xmath681 exists and is bounded above by @xmath641 . in particular @xmath682 case 2 : in the other case @xmath316 ( i.e. the vector belongs to category 2 ) , consider first an arbitrary sequence @xmath670 that tends monotonically to infinity and the corresponding subsequence @xmath671 of @xmath672 determined by @xmath683 . recall that without loss of generality we can assume @xmath684 , as this does not affect the approximation constants . by ( [ eq : memina ] ) we have @xmath685 furthermore define @xmath686 as the supremum of all @xmath54 , such that ( [ eq:1 ] ) has @xmath11 linearly independent vector solutions with at least one vector of category 2 , for every fixed @xmath52 . taking logarithms to the bases @xmath687 for @xmath680 , ( [ eq : huhnfisch ] ) implies that for @xmath9 the expression @xmath688 is bounded above by @xmath689 note that the bound in ( [ eq : qply ] ) is valid for any sequence @xmath670 and the corresponding sequence @xmath671 . hence on the one hand we have @xmath690 simply by the definition of @xmath115 . observe @xmath691 ( where @xmath196 in the definition of @xmath692 corresponds to @xmath279 in the definition of @xmath693 ) and hence @xmath694 on the other hand , for any sequence @xmath670 by ( [ eq : huhnfisch ] ) and our assumption @xmath695 , for @xmath696 sufficiently large and @xmath644 we have @xmath697 where the right inequality is a consequence of the construcions @xmath642 in the case @xmath698 ( see ( [ eq : tiryl]),([eq : mtiryl ] ) ) . since this holds for any sequence @xmath670 , we have @xmath699 we now combine the results above for the quantities @xmath700 , @xmath701 to derive the required upper bounds . as every approximation vector is either of category 1 or category 2 for fixed @xmath52 , the defintions of @xmath702 , imply @xmath703 for every @xmath52 . observe that for any functions @xmath704 we have @xmath705 applying this on @xmath706 implies that @xmath63 , which is by definition @xmath707 , equals the maximum of @xmath667 and @xmath708 . by ( [ eq : yaqmko]),([eq : yaqomk ] ) for @xmath644 this maximum is @xmath709 , which proves the upper bounds for @xmath63 , @xmath644 . + it remains to check the upper estimates for the constants @xmath65 . in view of @xmath703 , ( [ eq : yaqokm ] ) and ( [ eq : schmetzeling ] ) , we obtain @xmath710 for @xmath644 . remark : it is rather clear from the proof that theorem [ satz2 ] remains valid for @xmath711 too . we will need this later in theorem [ theo9 ] . [ korolar ] let the assumptions of theorem [ satz2 ] be satisfied and assume further the existence of the limit of the quotients @xmath712 , i.e. @xmath713 then the first @xmath638 approximation constants are given by @xmath714 for the remaining approximation constants we have the inequlities @xmath715 for every @xmath716 we have @xmath717 and hence theorem [ satz2 ] yields the claimed result . remark : the bounds for @xmath718 could be improved further to @xmath719 by a rather long and technical proof that we will not present here . in particular in the case @xmath720 , where @xmath721 is the largest real root of @xmath722 , we have @xmath723 let us call the assumptions of theorem [ satz2 ] without the growth condition of @xmath712 the _ basic assumptions _ of theorem [ satz2 ] in the sequel . we can generelize the idea of the proof of theorem [ satz2 ] to get [ satz6 ] given the basic assumptions of theorem [ satz2 ] , we consider some fixed @xmath724 and define @xmath725 to be the largest real root of @xmath726 . then if 1 . @xmath727 , + 2 . the sequence @xmath728 is monotonically increasing + are satisfied , the first @xmath729 approximation constants are given by @xmath730 furthermore we have the inequalities @xmath731 we proceed as in the proof of theorem [ satz2 ] using the fact that @xmath732 is increasing in place of the equivalent fact that @xmath675 is monotonically decreasing , which we deduced from the stronger assumptions in theorem [ satz2 ] ( where it was infered from the stronger assumptions in theorem [ satz2 ] ) up to equation ( [ eq : qply ] ) . instead of ( [ eq : qply ] ) , by our weaker assumption @xmath733 instead of @xmath734 , we obtain the weaker upper bound @xmath735 which yields @xmath736 as an upper bound for @xmath63 instead of @xmath692 . we proceed analogously again till ( [ eq : ronzor ] ) , instead of which we obtain @xmath737 this gives @xmath738 as an upper bound for @xmath65 instead of @xmath739 . the remainder of the proof is essentially the same as in theorem [ satz2 ] . remark : we clearly have @xmath740 . on the other hand @xmath741 which can esily be derived using the well known monotonic convergence of @xmath742 to the euler number @xmath743 . + again , we easily deduce the following [ kor2 ] let the basic assumptions of theorem [ satz2 ] and condition 1 ) from theorem [ satz6 ] be satisfied . let us further assume @xmath744 then the first @xmath729 approximation constants are given by @xmath745 further more we have the inequalities @xmath746 let us illustrate the results of corollary [ korolar ] and corollary [ kor2 ] in the case @xmath747 . in the first plot we put @xmath748 which ( for numerical purposes ) are the initial terms of @xmath749 which clearly satisfy the conditions of corollary [ korolar ] with @xmath750 . notice the special behaviour of @xmath751 in comparison to the first @xmath752 functions which behave as predicted in corollary [ korolar ] . + the assumptions of corollary [ kor2 ] are weaker in the sense that either @xmath753 or the quotients @xmath712 converge to @xmath750 without being strictly larger than @xmath754 for every sufficiently large @xmath196 . to illustrate this latter case we may put @xmath755 which are the initial terms of @xmath756 with @xmath757 and @xmath758 for all @xmath759 , which fufills the conditions of corollary [ kor2 ] with @xmath760 . indeed , we will see a different behaviour of @xmath761 compared to the previous picture . only @xmath762 has the predicted shape . we can apply corollary [ kor2 ] to construct many more cases of schmidt s conjecture explicitely . for simplicity of the proof we first deduce another easy corollary from corollary [ kor2 ] . [ coroll ] for @xmath556 let @xmath763 be an arbitrary integer . for any @xmath764 there exists a sequence of positive integers @xmath186 such that @xmath765 for all @xmath196 and @xmath766 . if @xmath544 for @xmath77 are @xmath208 sequences satisfying ( [ eq : frure ] ) such that @xmath186 is their ordered mixed sequence , then for @xmath543 the result of corollary [ kor2 ] is valid . + furthermore we can choose the sequence @xmath186 such that @xmath767 are @xmath43-linearly independent . the sequence @xmath186 defined by @xmath768 and @xmath769 with @xmath770 sufficiently large that @xmath771 clearly satisfies the stated properties . putting @xmath549 we see that the assumptions of corollary [ kor2 ] are satisfied , since it clearly makes no difference if we take the logarithm to base @xmath754 instead of @xmath772 as the quotients @xmath712 do nt change . by a variation of @xmath186 as in the proof of corollary [ prop1 ] we can guarantee the linear independence . with the notation of corollary [ coroll ] there exists a constant @xmath773 such that the following holds : * @xmath774 for @xmath208 sufficiently large . + * for fixed @xmath775 there is some @xmath776 , such that there exists a sequence @xmath186 of positive integers satisfying @xmath777 such that the corresponding vector @xmath778 constructed via corollary [ coroll ] with @xmath779 has approximation constants that satisfy + @xmath780 + a possible choice of @xmath773 is @xmath781 . this provides _ explicit examples _ for schmidt s conjecture . let @xmath208 be an arbitrary but fixed integer . in view of ( [ eq:2 ] ) for a given @xmath775 we need to find @xmath776 such that a vector @xmath73 that arises from corollary [ coroll ] with @xmath779 satisfies @xmath782 to obtain ( [ eq : frrranr ] ) . we will implicitly identify such a vector @xmath73 with the resulting value @xmath783 from the limit of the quotients @xmath712 in corollary [ coroll ] . this is well defined as the approximation constants we consider do nt depend on the choice of the exact vector but depend only on @xmath783 by corollary [ coroll ] . as was shown there the set of such vectors @xmath73 is nonempty for every @xmath784 and we can assume @xmath778 together with @xmath4 to be @xmath43 linearly independent . + for any positive integer @xmath625 define the function @xmath785 . each function @xmath786 is easily seen to be continuous and @xmath787 increases on @xmath788 $ ] and decreases on @xmath789 with limit @xmath159 as @xmath790 . + we use the notation of theorem [ satz6 ] , in particular @xmath725 is the largest real root of @xmath791 . we first prove that we can choose @xmath776 such that ( [ eq : frrranr ] ) holds for a given @xmath792 that has the property @xmath793 with the constructions of corollary [ coroll ] and the particular choice @xmath779 . + put @xmath794 . if ( [ eq : turmm ] ) is valid , the facts about the functions @xmath787 show that there is @xmath795 such that @xmath796 and @xmath797 already decreases at @xmath121 . furthermore there is an interval @xmath798 such that @xmath799 . since @xmath800 we can apply corollary [ coroll ] with @xmath801 and obtain @xmath802 , as intended + now assume @xmath208 is fixed , @xmath803 and that for @xmath804 @xmath805 holds . by definition of @xmath806 we have @xmath807 and ( [ eq : skkiforn ] ) further implies @xmath808 for such @xmath792 , i.e. ( [ eq : turmm ] ) . combining what we have shown so far , ( [ eq : frrranr ] ) can be obtained for @xmath792 and @xmath809 , provided ( [ eq : skkiforn ] ) holds . we now show that ( [ eq : skkiforn ] ) is true for @xmath775 . + in view of @xmath810 , inequality ( [ eq : skkiforn ] ) is equivalent to @xmath811 . since @xmath812 , whose largest root is @xmath813 , increases on the interval @xmath814 , this is equivalent to @xmath815 , i.e. @xmath816 basic rearrangements show this is equivalent to @xmath817 to finish up the proof we are left to show that @xmath818 holds for @xmath208 sufficiently large . to see this we claim that for sufficiently large @xmath208 we have @xmath819 > \frac{k}{2}\left[1+\frac{\log(k^{\frac{1}{k}}-1)}{\log(k)}\right]>\frac{k}{\log(k)}.\ ] ] for @xmath820 only the right most inequality is non trivial . it is equivalent to @xmath821 or @xmath822 . however , this is easily seen to be true by @xmath823 for @xmath824 for arbitrary @xmath252 , which follows from the monotonic convergence of @xmath825 for every @xmath252 . putting @xmath826 arbitrary proves the claim . remark : we have @xmath827 , but the convergence is very slow , in particular we have seen it is slower than @xmath828 . in this chaper we want to treat the case of @xmath0 that define functions @xmath15 with the property we will exclude the generic case where all approximation constants @xmath830 are zero , and call the cases where @xmath78 the _ degenerate cases_. note , that the numbers corollary [ korolar ] deals with lead to functions @xmath15 that satisfy the equalities of ( [ eq : rindvieg ] ) for @xmath831 , but not for @xmath832 in the case @xmath833 ( see the bounds given in the remark following corollary [ korolar ] ) . a _ special case _ of ( [ eq : rindvieg ] ) , for which an idealized picture is shown below , is of particular interest . if @xmath834 denotes the sequence of the first coordinates of approximation vectors @xmath305 , then we consider the special case where there is a sequence of @xmath834 with ( ideally ) constant value of @xmath835 such that let the equistence of a sequence @xmath839 such that ( [ eq : pattricia]),([eq : trapizia ] ) holds and additionally for @xmath840 every @xmath841 consecutive approximation vectors @xmath842 belonging to @xmath843 ( ie @xmath844 ) are linearly independent be our definition of the _ special case _ mentioned above . + roy shows , that numbers he defines as _ extremal numbers _ @xmath66 in the introduction of @xcite satisfy the property of the special case of ( [ eq : rindvieg ] ) for @xmath537 and @xmath845 and yield @xmath846 for @xmath847 and @xmath848 with @xmath849 . we are interested in other particular cases of the special case of ( [ eq : rindvieg ] ) . + it follows in general by ( [ eq : pattricia]),([eq : trapizia ] ) that all the values @xmath850 are determined by the value @xmath851 ( or equivalently @xmath91 ) . this holds for the degenerate case in particular . however , we will show in theorem [ theo9 ] that this phenomenon holds for _ all _ @xmath73 in the degenerate case of ( [ eq : rindvieg ] ) . by virtue of corollary [ korolar ] we can easily provide concrete examples for the degenerate case . before we do so , for the sake of completeness we give a general result about the degenerate case of ( [ eq : rindvieg ] ) . first note that in general if @xmath78 we have @xmath854 for @xmath525 by means of ( [ eq : zokz ] ) , see the proof of theorem [ satz1 ] . consequently by ( [ eq : rindvieg ] ) we have ( [ eq : unddasauch ] ) and ( [ eq : unddies ] ) . + for ( [ eq : dasauch ] ) note first that @xmath855 is always a lower bound as established in ( [ eq:1000alk]),([eq:1001alk ] ) . so by ( [ eq : rmuo ] ) it suffices to prove @xmath856 . + suppose we had @xmath857 . this means for some sequence @xmath858 tending to infinity we have @xmath859 for some @xmath860 . putting @xmath861 and using @xmath862 we have [ theo9 ] for any @xmath556 there exist real numbers @xmath149 as in corollary [ korolar ] with @xmath711 together with @xmath4 linearly independent over @xmath43 that satisfy the degenerate case of ( [ eq : rindvieg ] ) and hence ( [ eq : obdasgeht])-([eq : unddies ] ) by proposition [ kokorp ] . @xmath866 which gives ( [ eq:90alk])-([eq:93alk ] ) . however , in the case @xmath537 and @xmath711 we can also apply corollary [ korolar ] with a slightly more sophisticated argumentation . of course we directly infer ( [ eq:90alk ] ) and the left equation in ( [ eq:91alk ] ) follows as for @xmath600 . from ( [ eq:90alk ] ) we can immediately deduce @xmath867 for @xmath868 as in the proof of theorem [ satz1 ] , which is a rephrasing of the left hand side of ( [ eq:92alk ] ) and ( [ eq:93alk ] ) . by the left equation in ( [ eq:91alk ] ) and jarnik s identity @xmath869 ( see comments between the end of proof of theorem [ satz1 ] and corollary [ koroo ] ) we get @xmath870 , i.e. the right hand side of ( [ eq:92alk ] ) . for the missing right hand equation of ( [ eq:91alk ] ) note that @xmath871 is always true by ( [ eq : rmuo ] ) and on the other hand @xmath872 by ( [ eq:2222neu ] ) , so by the left hand equality in ( [ eq:91alk ] ) we infer the right hand equality of ( [ eq:91alk ] ) . this allows to show that the bounds in ( [ eq:2000neu])-([eq:2004neu ] ) are best possible if considered independently by using only three types of vectors @xmath73 depending on the dimension @xmath208 . these types are : * a set of together with @xmath4 @xmath43-linearly independent _ algebraic _ numbers @xmath0 ( leading to the generic case ) + * @xmath0 as in corollary [ korolar ] with @xmath711 , for example @xmath873 for @xmath77 + * @xmath0 as in corollary [ korol1 ] in corollary [ korol1 ] we have seen , that the upper bounds in ( [ eq:2000neu]),([eq:2222neu ] ) , ( [ eq:2002neu ] ) as well as the lower bounds in ( [ eq:2001neu]),([eq:2003neu]),([eq:2004neu ] ) can not be improved . + in theorem [ theo9 ] we ve just seen , that the left hand side of ( [ eq:2002neu ] ) and the right hand side of ( [ eq:2001neu ] ) are optimal . + however , all the other bounds are @xmath874 and it is well known that all constants @xmath50 are equal to @xmath874 in the generic case . to give concrete examples , an implication of schmidt s subspace theorem says , that for all @xmath43-linearly independent algebraic numbers all approximation constants take the value @xmath874 ( which follows already from @xmath875 by ( [ eq:2 ] ) and ( [ eq : zokz ] ) ) . so the lower bounds of ( [ eq:2000neu]),([eq:2222neu ] ) such as the upper bounds of ( [ eq:2003neu ] ) , ( [ eq:2004neu ] ) can not be improved either , and the list is complete . let @xmath0 be real numbers that lead to a special case of ( [ eq : rindvieg ] ) , i.e. ( [ eq : pattricia]),([eq : trapizia ] ) hold . it follows directly from ( [ eq : pattricia]),([eq : trapizia ] ) that all the constants @xmath50 only depend on @xmath91 . it is easy to check that more precisely we have @xmath882 differentiating shows that @xmath883 decreases on @xmath884 and increases on @xmath885 , so its global mimimum on @xmath886 is at @xmath887 . combining this with @xmath888 , ( [ eq : hunggar ] ) and ( [ eq : roott ] ) we obtain remarks : 1 ) one can proof that for @xmath556 we have @xmath894 . + observe , that in corollary [ kor2 ] with arbitrary @xmath783 we always have @xmath895 . proposition [ wasqob ] shows that given @xmath91 the resulting special case of ( [ eq : rindvieg ] ) leads to a larger value of @xmath171 . it may be conjectured that among all @xmath0 linearly independent together with @xmath4 with prescribed @xmath896 , the quantity @xmath171 is maximised for the special case of ( [ eq : rindvieg ] ) with the value @xmath896 . + 2 ) observe that the inequality @xmath897 always holds as established by jarnik , see theorem 1 page 331 in @xcite . so together with proposition [ wasqob ] in the special case of ( [ eq : rindvieg ] ) we have the author thanks l. summerer for help in translation and w. schmidt for an idea to generalize my initial versions of theorems [ satz1 ] , [ satz2 ] , [ satz6 ] which simplified the proofs and made them more concise . 99 p.m.gruber , c.g . lekkerkerker : _ geometry of numbers _ , north - holland verlag ( 1987 ) v. jarnik : _ contribution to the theory of linear homogeneous diophantine approximation _ , czechoslovak math . j. @xmath899 ( 1954 ) , no . 79 , 330 - 353 n.g . moshchevitin : _ proof of w.m . schmidt s conjecture concerning successive minima of a lattice _ , j. london math . ( 2012 ) doi : 10.1112/jlms / jdr076 , 12 mar 2012 d. roy : _ diophantine approximation in small degree _ , number theory , 269 - 285 , crm proc . lecture notes , 36 , amer . soc . , providence , ri , 2004 w.m . schmidt , l. summerer : _ parametric geometry of numbers and applications _ , acta arithm . 140.1 ( 2009 ) w.m . schmidt , l. summerer : _ diophantine approximation and parametric geometry of numbers _ , to appear in monatshefte fr mathematik m. waldschmidt , _ report on some recent advances in diophantine approximation _ ( 2009 )	this paper introduces some methods to determine the simultaneous approximation constants of a class of well approximable numbers @xmath0 . the approach relies on results on the connection between the set of all @xmath1-adic expansions ( @xmath2 ) of @xmath0 and their associated approximation constants . as an application , explicit construction of real numbers @xmath0 with prescribed approximation properties are deduced and illustrated by matlab plots .
it is a widely held belief that new physics must exist beyond that predicted by the standard model(sm ) if for no other reason than that it leaves us with too many unanswered questions and too many free parameters . just when , where and how new physics will make its first appearance has been and continues to be a matter of some speculation . in the past few years we have seen a number of potential new physics signatures vanish as either more statistics was accumulated or the data and analysis thereof improved . these signatures have each in their turn induced some excitement in the community with the hope that a new window beyond the sm was finally opening . though later vanishing , these affects have each taught us something new about what kinds of models can be constructed , even if they were not necessarily realized in nature , given the ever - tightening constraints provided by experiment . still , new physics is out there somewhere waiting to be discovered ; it is only a question of looking . we should learn to expect the unexpected . searching for new physics is a multi - pronged attack on the unknown . while the production of a non - sm particle at a collider would be the most obvious and undeniable signature that one can imagine , the first sign of something new maybe more subtle . for example , one can imagine a significant deviation from sm expectations in a precision measurement , , the @xmath3 mass or polarized forward - backward asymmetry for @xmath4quarks , @xmath5 . instead , one might imagine the observation of a process forbidden by the sm , such as @xmath6 . however , as is well known , new physics rarely contributes to only one of these scenarios . for example , susy leads to new particle production at colliders , a potentially observable shift in the @xmath3 mass and/or @xmath7 and can enhance many rare decay processes beyond their sm expectations . the same is of course true with other forms of new physics . in fact , it likely that once new physics is found all three types of experiments will be necessary to unravel its detailed nature . attempts to understand the excess of events at high@xmath0 recently observed at hera in both the neutral current(nc ) and charged current(cc ) channels@xcite provides just such an example of the strong interplay between the various new physics search scenarios and the constraints that arise from the three classes of experiments . if the hera excesses are real and non - resonant then possible explanations include , , the presence of higher dimensional operators@xcite signalling compositeness or exotic modifications in the parton densities at large @xmath8@xcite ; both these proposed scenarios face some very serious difficulties . instead , if the excess is resonant a popular@xcite explanation is the @xmath9channel production of a @xmath10 gev scalar ( , spin-0 ) leptoquark(lq ) with fermion number(@xmath11 ) equal to zero the subject of the present work . ( given the most recent results@xcite it appears that the h1 excess in the nc channel is apparently clustered at @xmath12 gev while that for zeus is at @xmath13 gev . whether this is reconcilable with a single resonance is still unknown . ) how did we so quickly deduce the lq spin and @xmath11 quantum number from the data ? any discussion of lq models has been historically based on the classic work by buchm " uller , r " uckl and wyler(brw)@xcite . those authors provided not only a set of assumptions under which consistent lq models can be constructed but then classified them according to their possible spins and fermion number thus leading to the 10 states displayed in table 1 . these assumptions may be stated as follows : \(a ) lq couplings must be invariant with respect to the sm gauge interactions \(b ) lq interactions must be renormalizable \(c ) lqs couple to only a single generation of sm fermions \(d ) lq couplings to fermions are chiral \(e ) lq couplings separately conserve baryon and lepton numbers \(f ) lqs only couple to the sm fermions and gauge bosons l\|ccrccc & leptoquark ' '' '' & su(5 ) rep & & @xmath14 & coupling & @xmath15 + scalars & & & & & & + @xmath16 ' '' '' & @xmath17 & * 5 * & & @xmath18 & @xmath19 , @xmath20 & @xmath21 + & @xmath22 & * 5 * & & @xmath18 & @xmath23 & 1 + & @xmath24 & * 45 * & & @xmath25 & @xmath26 & 1 + & & & & @xmath25 & @xmath27 & 1 + & @xmath28 & * 45 * & @xmath29 & @xmath18 & @xmath30 , @xmath31 & @xmath21 + & & & & @xmath32 & @xmath33 & 0 + [ 0pt]@xmath34 & [ 0pt]@xmath35 & [ 0pt]45 & [ 0pt]@xmath36 & @xmath37 & @xmath38 & 1 + & & & & @xmath39 & @xmath40 & 0 + & [ 0pt]@xmath41 & [ 0pt]45 & [ 0pt]@xmath36 & @xmath37 & @xmath42 & 1 + & & & & @xmath39 & @xmath43 & 1 + & [ 0pt]@xmath44 & [ 0pt]10/15 & [ 0pt]@xmath36 & @xmath39 & @xmath45 & 1 + & & & & @xmath46 & @xmath47 & 0 + vectors & & & & & + ' '' '' [ 0pt]@xmath16 & [ 0pt]@xmath48 & [ 0pt]24 & [ 0pt]@xmath36 & @xmath25 & @xmath49 & 1 + & & & & @xmath18 & @xmath20 & 0 + & [ 0pt]@xmath50 & [ 0pt]24 & [ 0pt]@xmath36 & @xmath25 & @xmath26 & 1 + & & & & @xmath18 & @xmath23 & 1 + & [ 0pt]@xmath51 & [ 0pt]10/15 & [ 0pt]@xmath36 & @xmath18 & @xmath19 & 1 + & & & & @xmath32 & @xmath52 & 0 + @xmath34 & @xmath53 & * 10 * & & @xmath39 & @xmath45 , @xmath40 & @xmath21 + & @xmath54 & * 10 * & & @xmath39 & @xmath26 & 1 + & @xmath55 & * 75 * & & @xmath37 & @xmath42 & 1 + & & & & @xmath37 & @xmath56 & 1 + & @xmath57 & * 40 * & @xmath29 & @xmath39 & @xmath58 , @xmath40 & @xmath21 + & & & & @xmath46 & @xmath59 & 0 + if we strictly adhere to these rules then the requirements of gauge invariance and renormalizability fix all of the spin-1 lq couplings and thus its production cross section at the tevatron@xcite is simply a function of its mass . the possibility that such particles can exist in the mass range below approximately 350 gev can then be excluded based on the direct searches by both cdf and d0@xcite . ( as noted by blmlein@xcite , the introduction of non - renormalizable anomalous couplings for the lq may allow us to somewhat soften this conclusion but spin-1 lqs are still found to be excluded in the range of interest for the hera excess . ) if the lq is a scalar but is of the @xmath60 type then we would have expected to see an event excess show up in the @xmath61 nc channel and not the @xmath62 channel as is the case . this is demonstrated in table 2 which shows the event rate for each of the brw scalar lqs assuming a yukawa coupling strength of @xmath63 and a mass of 200 gev normalized to a luminosity of 100 @xmath64 . even with the great disparity in integrated luminosity collected by the experiments in both channels and allowing for the free adjustment of the strength of the yukawa coupling we must conclude that the lq is of the @xmath34 type . we note from table 1 that all @xmath34 scalar lqs must have @xmath65 and lie in @xmath66 doublets , implying that more than one type of lq must exist . given the recent strengthening of the tevatron search reach and the possible cc excess at hera , this poses a serious challenge to the scalar lq interpretation , although not as serious as was found in the case of vector lqs . using the next - to - leading order cross section formulae of krmer , the @xmath67 cl lower limit on the mass of a @xmath65 scalar lq is found by d0 to be 225 gev . d0 has also performed a combined search for first generation leptoquarks by using the @xmath68 , @xmath69 and @xmath70 channels . for fixed values of the leptoquark mass below 225 gev , these search constraints can be used to place an upper limit on @xmath15 . for @xmath71=200(210,220 ) gev , d0 obtains the constraints @xmath72 at @xmath67 cl . of course if cdf and d0 combine their searches in the future , then the 225 gev bound may rise to @xmath73 gev , in which case even stronger upper bounds on @xmath15 will be obtained . allowing the lq to have decays into the @xmath74 final state with a reasonable branching fraction would solve this problem and would yield the desired cc signal at hera . however the models in table 1 do not allow for this possibility . .expected number of events per @xmath75 for each electron charge and state of polarization for a 200 gev scalar leptoquark at hera assuming @xmath76 , @xmath77 , and an electron - jet invariant mass @xmath78 gev . these results have been smeared with a detector resolution of @xmath79 in @xmath80 . [ cols="<,^,^,^,^",options="header " , ] to directly test the proposed models we must look for new physics signatures _ beyond _ those suggested by the lq interactions described eqns . 1 and 2 . one possibility is to probe for the additional interactions between the vlfs and the lq ; this is obviously difficult since the vlfs are so massive . a second possibility is to directly look for the influences of the vlfs themselves . for the time being this must be an indirect search since the lhc will be required to directly produce the vlfs in the range of interest to us here . once the lhc is available , however , a 1 tev color triplet will produce 1000 events / yr even with a luminosity of 10 @xmath81 . in the first category a potential new process of interest is the pair production of like sign lqs at the tevatron through @xmath82 and @xmath83channel @xmath84 exchange ( in those models where it is present ) when @xmath84 is a majorana field : @xmath85 . this rate for this process goes as the fourth power of the @xmath86 or @xmath87 yukawa coupling but these are assumed to be of order unity so that potentially large cross sections are obtainable . one finds the subprocess cross section to be @xmath88 ^ 2\,,\ ] ] where @xmath89 is the mass of the @xmath84 . note that as @xmath90 the rate vanishes as one might expect for a majorana fermion induced process . the cross section for this reaction at the tevatron main injector for @xmath91 and @xmath92 gev is approximately 50 fb for @xmath89=1 tev and falls off quickly with increasing as @xmath89 increases . since the signature for this process is 2 jets plus like - sign leptons there is little sm backgrounds and so it may be observable during run ii . the best indirect tests for the presence of vlfs which mix with the conventional fermions are searches for new physics associated with deviations in couplings from sm expectations@xcite . two of the best tests here are our old friends quark - lepton universality in the guise of @xmath93 and the leptonic decays of the @xmath94 discussed above . to clarify this point we note that in models where , , the @xmath95 and @xmath96 mix with isodoublet vlfs the @xmath3 couples as @xmath97 whereas if the vlfs are isosinglets the corresponding coupling is @xmath98 . ( the corresponding couplings can be also written down for the case of leptonic mixing . ) we note also the general feature we find is that in models where the vlfs are in isodoublets @xmath99 and differences in @xmath100 , whereas the converse is true when the relevant vlfs are isosinglets . leptonic mixing will never show up as a shift in the value of @xmath101 since the modification in the amplitude occurs not only in the process @xmath102 but also in @xmath103 decay so that it is absorbed into the definition of @xmath104 . on the otherhand if quark mixing occurs @xmath101 will experience an apparent small shift(to leading order in the mixing angles ) @xmath105 in models with isodoublet vlfs and @xmath106 in the isosinglet vlf case . for the ratio @xmath107 in @xmath94 decay any modification of the hadronic matrix element will factor out so that there is no sensitivity to quark mixing . however , leptonic mixing no longer factorizes and we find a shift in @xmath107 by an amount @xmath108 in the case of isodoublet mixing and @xmath109 for isosinglets . recall that isodoublet leptonic mixing is only viable in models where the neutrino is dirac or the right - handed neutrino appears as an ordinary neutrino . in these same isodoublet models , the presence of both left- and right - handed cc couplings and heavy vlfs can lead to a contribution to the @xmath110 of the electron and @xmath111 which are typically both or order @xmath112 , neither of which are far from the present level of sensitivity . the mixing of the sm fermions with the vlf modify their couplings to the @xmath113 . these are difficult to observe particularly in the case of quarks due to the small size of the effect and qcd correction uncertainties . in the case of leptonic mixing there is not only the contribution due to mixing but there is an overall normalization change in the couplings due to our redefinition of @xmath104 . this mixing can lead to a shift in @xmath114 width by @xmath115 mev and an apparent shift in @xmath116 from the asymmetries of @xmath117 . again , shifts of this size are near the present limit of experimental sensitivity . in a similar manner sm expectations for @xmath118 in apv measurements may also be modified if sm mixing with the vlf occurs . however , in this case it is easy to show that the fractional change in @xmath118 due to these effects is only at the level of @xmath119 . if lqs are indeed real and we also believe that there is experimental evidence for coupling constant unification then we must begin to examine schemes which contain both ingredients as pointed out in@xcite . in the scenarios presented here the sm quantum numbers of the lq are fixed but new vlfs have now been introduced as well , all of which will alter the usual rge analysis of the running couplings . before discussing susy models we note with some curiosity that coupling unification _ can _ occur in lq models containing exotic fermions even if susy is not introduced as was shown many years ago in @xcite . of course in the work of murayama and yanagida @xcite , the lq was an isodoublet and one of the particular models on the brw list , now excluded by the combined hera and tevatron data . in the scenarios presented above the lq is now a @xmath120 isosinglet so that the murayama and yanagida analysis does immediately apply . fortunately , we see from the results of ref . @xcite that a second possibility does exist for just this case : one adds to the sm spectrum the lq and its conjugate as well as a vector - like pair of color - triplet , isodoublets together with the field @xmath121 . this is the just particle content of the model a. to verify and update this analysis , let us assume for simplicity that all the new matter fields are introduced at the weak scale and take @xmath122 as input to a two - loop rge analysis . we then obtain the predictions that coupling unification occurs at @xmath123 gev and @xmath124 is predicted to be 0.118 . if unification does indeed occur we can estimate the proton lifetime@xcite to be @xmath125 years , safely above current constraints@xcite . we find this situation to be rather intriguing and we leave it to the reader to further ponder . of course there are other reasons to introduce susy beyond that of coupling constant unification . this subject has been discussed at some length in @xcite from which we extract several important observations : ( @xmath126 ) to _ trivially _ preserve the successful unification of the susy - sm , only complete @xmath127 representations can be added to the mssm spectrum . as is well - known , the addition of extra matter superfields in complete @xmath127 representations delays unification and brings the gut scale closer to the string scale . of course , there still remains the rather unnatural possibility of adding incomplete , but ` wisely chosen ' , split representations . employing split representations certainly allows for more flexibility at the price of naturalness but still requires us to choose sets of @xmath128 representations which will maintain asymptotic freedom and perturbative unification . an example of this rather bizarre scenario is the possibility of adding a @xmath129 from a * 15 * and a @xmath130 from a * 10 * to the low energy spectrum@xcite . here the notation refers to the @xmath131 quantum numbers of the representation . we remind the reader that the lq itself transforms as @xmath132 ; the smallest standard @xmath127 representation into which the @xmath133 can be embedded is a * 10@xmath134 while in flipped-@xmath135@xcite , it can be placed in a 5@xmath136 . ( @xmath137 ) since we are using vlfs in these models , it is clear that only pairs of representations , r@xmath138 , can be added to the mssm spectrum in order to maintain anomaly cancelation . of course this is also true for the lq superfield in that both lq and @xmath139 fields must now be present as discussed above . ( @xmath140 ) to preserve perturbation theory and asymptotic freedom up to the gut scale when adding complete representations , at most one 10@xmath141 or three 5@xmath142 can be appended to the low energy spectrum of the mssm apart from sm singlets . the reason for this is the general observation that if one adds more than three , vector - like , color triplet superfields to the mssm particle content then the one - loop qcd beta function changes sign . recall that the lq itself _ already _ accounts for one of these color triplets . this same consideration also excludes the introduction of light exotic fields in higher dimensional @xmath143 representations . complete @xmath127 representations larger than 10@xmath141 are found to contribute more than this critical amount to the running of the qcd coupling . these are highly restrictive constraints on the construction of a successful gut scenario containing both vlfs and lqs and we see than none of the models discussed above can immediately satisfy them _ unless _ the lq and vlf superfields can be placed into a single @xmath127 representation . in the standard @xmath127 picture , we can then place @xmath144 , an isosinglet @xmath145 and @xmath139 into a single 10 * with the corresponding conjugate fields in the @xmath146 . this would form a hybrid of model a with the ` skeleton ' model d , which we ve denoted by ad above . of course we pay no penalty for also including ` skeleton ' model c here as well , which then yields model acd . instead , when we consider the flipped-@xmath135 case , it would appear that we can place @xmath147 and @xmath139 into a @xmath148 with the conjugate fields in the * 5 * ; this is exactly model b. it would also seem that no penalty is paid as far as unification is concerned for including the ` skeleton ' model c here as well _ except _ that this would violate assumption ( @xmath140 ) about the chirality of lq couplings to fermions . however , this model is no longer truly unified since the hypercharge generator is not fully contained within the @xmath127 group itself and lies partly in the additional @xmath149 . while the @xmath143 and @xmath66 couplings will unify , @xmath150 will not join them even when arbitrary additional vector - like singlet fields are added . thus unification no longer occurs in this scenario so that this possibility is now excluded . the lq embedding situation becomes more perplexing if the lq and vlfs can not occupy the same gut multiplet . in this case unification and asymptotic freedom constraints become particularly tight and we are forced to consider the split multiplet approach mentioned above . this means that we add the fields @xmath151 and their conjugates at low energies but constrain them to be from different @xmath127 representations . in this case the combination @xmath152 corresponds to the isosinglet lq and its conjugate so what remains can only be the vlf fields . note that we have again arrived back at models ad and acd . are these the only solutions ? we have performed a systematic scan over a very large set of vlfs with various electroweak quantum numbers under the assumption that they are either color singlets or triplets , demanding only that ( @xmath126 ) qcd remains asymptotically free and ( @xmath137 ) the model passes the so - called `` b - test''@xcite which is highly non - trivial to arrange . essentially the b - test takes advantage of the observation that if we know the couplings at the weak scale and we demand that unification takes place _ somewhere _ then the values of the one - loop beta functions must be related . note that it is a necessary but not sufficient test on our choice of models but is very useful at chopping away a large region of parameter space . using the latest experimental data@xcite , we find that @xmath153 where the @xmath154 is an estimate of the corrections due to higher order as well as threshold effects and the @xmath155 are the one - loop beta functions of the three sm gauge groups . note that @xmath156 clearly satisfies the test . if we require that ( @xmath126 ) and ( @xmath137 ) be satisfied and also require that the unification scale not be too low then only the solutions described above survive after examining @xmath157 combinations of matter representations . while not completely exhaustive this search indicates the solutions above are fairly unique . it is interesting to observe that models constructed around model a produce successful grand unification both with and without susy . in this talk we have seen how a wealth of data from low and medium energy experiments as well as high energy colliders can be combined to point us in a fixed direction for lq model building . i have also discussed a general framework for the construction of new @xmath34 scalar lq models which go beyond the original classification by buchmller , rckl and wyler . this approach is based on the observation that in any realistic extension of the sm containing lqs it is expected that the lqs themselves will not be the only new ingredient . this construction technique is , of course , far more general than that required to address the specific issue of the hera excess . while the assumptions of gauge invariance and renormalizability are unquestionable requirements of model building , it is possible that the other conditions one usually imposes are much too strong unless they are clearly demanded by data . this observation implies that for lqs to be experimentally accessible their couplings to sm fermions must be essentially chiral and separately conserve both baryon and lepton numbers . the assumption that lqs couple to only a single sm generation is surely a convenient way of avoiding numerous low energy flavor changing neutral current constraints but is far from natural in the mass eigenstate basis . what is required to obtain a new class of lq models is that the lqs themselves must be free to couple to more than just the sm fermions and gauge fields . given the fixed gauge structure of the sm the most likely new interactions that lqs may possess are with the higgs field(s ) responsible for spontaneous symmetry breaking and with new vlfs that are a common feature in many extensions of the sm . in the discussion above it has been shown how two new forms of the effective interactions of lqs with the sm fermions , consistent with tevatron searches , the hera excess in both the nc and cc channels and low - energy data , can arise through the action of vlfs and ordinary symmetry breaking . the typical vlf mass was found to lie in the low tev region and they could thus be directly produced at future colliders with known rates . we saw that we could construct ten new models which fell into two broad classes according to the chirality of the resulting lq couplings to the sm fermions . the vlfs themselves were shown to lead to a number of model - dependent effects which are close to the boundary of present experimental sensitivity . lqs within the framework of models containing vlfs were also shown to be consistent with grand unification in both a supersymmetric _ and _ non - supersymmetric context . the common feature of both schemes is the structure associated with model a , , the vlfs are color triplet , weak isodoublets in a @xmath129 representation and both @xmath158 and @xmath121 higgs fields are required to be present as is @xmath139 field . in both scenarios the gut scale is raised appreciably from the corresponding model wherein lqs and vector - like fermions are absent . in the susy case a @xmath159 field is also required with the optional addition of a sm singlet , corresponding to models ad and acd . in some sense , acd is the `` anti-@xmath160 '' model in that the color triplet vlfs are in isodoublets while the color singlet fields are all isosinglets . interestingly , in this scenario there is a vector - like fermion corresponding to every type of sm fermion . realistic lq models provide a rich source of new physics beyond the standard model . the author would like to thanks j.l . hewett for collaboration on the work on which this talk is based . the author also appreciates comments , discussions and input from s. eno(d0 ) , g. landsberg(d0 ) , j. conway(cdf ) and h. frisch(cdf ) regarding the current tevatron constraints . in addition , he thanks y. sirois(h1 ) and d. krakauer(zeus ) for discussions of the hera data . # 1 # 2 # 3 mod . # 1 * , # 2 ( # 3 ) # 1 # 2 # 3 nucl . phys . * # 1 * , # 2 ( # 3 ) # 1 # 2 # 3 phys . lett . * # 1 * , # 2 ( # 3 ) # 1 # 2 # 3 phys . rep . * # 1 * , # 2 ( # 3 ) # 1 # 2 # 3 phys . rev . * # 1 * , # 2 ( # 3 ) # 1 # 2 # 3 phys . # 1 * , # 2 ( # 3 ) # 1 # 2 # 3 rev . mod . phys . * # 1 * , # 2 ( # 3 ) # 1 # 2 # 3 z. phys . * # 1 * , # 2 ( # 3 ) # 1 # 2 # 3 int . j. mod . phys . * # 1 * , # 2 ( # 3 ) c. adloff , h1 collaboration , c74 191 1997 ; j. breitweg , zeus collaboration , c74 207 1997 . for a comprehensive analysis of this possibility and original references , see v. barger , hep - ph/9707412 . the analysis by these authors apparently excludes this possibility as the source of the hera excess in the nc channel . s. kuhlman , h.l . lai and w.k . tung , hep - ph/9704338 ; k.s . babu , c. kolda and j. march - russell , hep - ph/9705399 ; j.f . gunion and r. vogt , hep - ph/9706252 ; s. rock and p. bosted , hep - ph/9706436 ; w. melnitchouk and a.w . thomas , hep - ph/9707387 . g. altarelli , hep - ph/9703279 ; j. blmlein , hep - ph/9703287 ; j. kalinowski , hep - ph/9703288 ; k.s . babu , hep - ph/9703299 ; m. suzuki , hep - ph/9703316 ; g.k . leontaris and j.d . vergados , hep - ph/9703338 ; i. montvay , hep - ph/9704280 ; s.f . king and g.k . leontaris , hep - ph/9704336 ; j. elwood and a. faraggi , hep - ph/9704363 ; b. dutta , r.n . mohapatra and s. nandi , hep - ph/9704428 ; m. heyssler and w.j . stirling , hep - ph/9705229 ; j. ellis , s. lola and k. sridhar , hep - ph/9705416 ; s. jadach , b.f.l . ward and z. was , hep - ph/9705429 ; a. blumhofer and b. lampe , hep - ph/9706454 ; e. keith and e. ma , hep - ph/9707214 ; n.g . deshpande and b. dutta , hep - ph/9707274 . j.l hewett and t.g . rizzo , slac - pub-7430 , hep - ph/9703337 , to appear in physical review * d * and slac - pub-7549 , hep - ph/9708419 . b. straub , zeus collaboration , talk given at the _ xviii international symposium on lepton and photon interactions _ , july 28-august 1 , 1997 , hamburg , germany . for the most recent results , see e. elsen , h1 collaboration , talk given at the _ international europhysics conference on high energy physics _ , jerusalem , august 19 - 26 , 1997 . w. buchm " uller , r. r " uckl , and d. wyler , b191 442 1987 . hewett , t.g . rizzo , s. pakvasa , h.e . haber , and a. pomarol , in _ proceedings of the workshop on physics at current accelerators and the supercollider _ , june 1993 , ed . hewett , a. white , and d. zeppenfeld . f. abe , cdf collaboration , hep - ex/9708017 , submitted to phys . lett . ; b. abbott , d0 collaboration , hep - ex/9707033 , submitted to phys . .for the most recent results , see b. klima , d0 collaboration , talk given at the _ international europhysics conference on high energy physics _ , jerusalem , august 19 - 26 , 1997 . m. krmer , t. plehn , m. spira and p. zerwas , hep - ph/9704322 . s. davidson , d. bailey , and b.a . campbell , c61 613 1994 ; m. leurer , d50 536 1994 , and * d49 * , 333 ( 1994 ) . wood , science * 275 * 1759 ( 1997 ) . for a recent review and updated , see j.l . rosner , hep - ph/9704331 . barnett , ( particle data group ) , d54 1 1996 . z. kunszt and w.j . stirling , hep - ph/9703427 ; t. plehn , h. spiesberger , m. spira and p.m. zerwas , hep - ph/9703433 . for an updated discussion on the effects of qcd and initial and final state radiation , see c. freiberg , e. norrbin and t. sjstrand , hep - ph/9704214 . j.l . hewett and t.g . rizzo , 183 193 1989 . hewett and t.g . rizzo , d36 3367 1987 ; j. bl " umlein and r. r " uckl , b304 337 1993 ; j. blmlein and e. boos , _ nucl . phys._(proc . suppl . ) _ 37b _ , 181 ( 1994 ) ; j. bl " umlein , e. boos , and a. kryukov , b392 150 1997 ; d. choudhury , b346 291 1995 ; j.e . cieza - montalvo and o.j . eboli , d47 837 1993 ; t.g . rizzo , d44 186 1991 ; h. dreiner , 3a 443 1988 ; g. bhattacharyya , d. choudhury , and k. sridhar , b349 118 1995 ; m.s . berger , hep - ph/9609517 . k. ackerstaff , opal collaboration , hep - ex/9708024 . r. rckl and p. zerwas , in _ proceedings of the lathuile workshop on physics at future accelerators _ , lathuile , italy , 1987 , ed . mulvay , cern report 87 - 07 ( 1987 ) . a. buras , talk given at the _ seventh international symposium on heavy flavor physics _ , university of california , santa barbara , july 7 - 11 , 1997 . this analysis relies heavily on the work presented in e. hagberg , nucl - ex/9609002 . m. peskin and t. takeuchi , 65 964 1990 , and d46 381 1992 ; w. marciano and j. rosner , 65 2963 1990 ; g. altarelli and r. barbieri , b253 161 1990 ; d. kennedy and p. langacker , 65 2967 1990 , and d44 1591 1991 ; i. maksymyk , c.p . burgess , and d. london , d50 529 1994 ; c.p . burgess , b326 276 1994 . o. eboli , hep - ph/9508342 ; o.j . eboli , m.c . gonzalez - garcia , and j.k . mizukoshi , hep - ph/9612254 , and b443 20 1995 ; g. bhattacharyya , j. ellis , and k. sridhar , b336 100 1994 , and erratum * b338 * , 522 ( 1994 ) . see , a. djouadi , j. ng , and t.g . rizzo , in _ electroweak symmetry breaking and beyond the standard model _ t. barklow , ( world scientific , singapore ) , hep - ph/9504210 , and references therein . h. murayama and t. yanagida , a7 147 1992 . rizzo , d45 3903 1992 . marciano , talk given at the _ 1983 international symposium on lepton and photon interactions at high energies _ , ithaca , ny , august 1983 . i. antoniadis , j. ellis , j.s . hagelin and d.v . nanopolous , b194 321 1987 . m. peskin , slac - pub-7479 , 1997 . a. bohm , l3 collaboration , talk presented at the _ @xmath161 recontres de moriond : electroweak interactions and unified theories _ , le arces , france , 15 - 22 march , 1997 .	attempts to understand the recent observation of an excess of events in the neutral and charged current channels at high@xmath0 at hera has provided an excellent example of how experiments at both low and high energies can be used to simultaneously constrain scenarios which predict new physics beyond the standard model . in this talk i will discuss this subject from the point of view of the construction of new models of leptoquarks . psfig.sty -0.5 in 6.5 in 8.5 in 4.0=0.5 in -0.5 in # 1_.25ex # 1_-.1em#1 # 1_-.2em#1 # 1_-.1em.25ex # 1_.25ex # 1_.4ex # 1^#1 # 1^#1 / plus0pt minus0pt # 1#20.7ex # 1@xmath1 # 1@xmath2 # 1 # 1 # 1[#1 ] * leptoquarks : pride and prejudice * thomas g. rizzo + + 0.95 in talk given at the _ workshop on physics beyond the standard model : beyond the desert : accelerator and nonaccelerator approaches _ , ringberg castle , germany , june 8 - 14 , 1997 .
@xmath3 formation in the ism ( interstellar medium ) is a fundamental process in astrophysics . the radiative association of two hydrogen atoms is a process too rare to be efficient because it involves forbidden roto - vibrational transitions , and gas - phase three - body reactions are rare in the diffuse ism to explain @xmath3 abundance . it has long been recognized that hydrogen recombination occurs on surfaces of dust grains , where the grains act as the third body in the @xmath4 reaction . ( @xcite ) . the critical physical steps in which interstellar grains act as catalysts for the @xmath3 synthesis are : sticking and accommodation of gas - phase atoms on the grain surface ; mobility of h adatoms to meet each other within their residence time on the cold grain ; and ejection of the newly formed molecule into the gas phase . the efficiency of these processes depends on the structure and the chemical activity of the grain surface , whether it is silicon - bearing ( silicates can be not very stable in uv radiation fields ) @xcite , carbonaceous ( carbon can possibly be distributed on silicon - bearing grain surfaces ) or icy ( @xcite ) . in the case of more chemically active surfaces , like carbonaceous ones , the possible presence of unsaturated bonds will increase the percentage of chemisorption events . the morphology of the surface should influence the sticking and tunneling processes ; it is reasonable to assume that sticking will be higher and diffusion length lower on amorphous than on crystalline surfaces . amorphous is the most probable structure for grains in space , as shown by observations ( @xcite ) . observed @xmath3 abundance in diffuse clouds can be explained if the formation rate @xmath5 on grains balances its destruction rate ; the canonical value of @xmath5 is @xmath6 ( @xcite ) ; on grain surfaces , the recombination rate can be described as @xmath7 where , @xmath8 and @xmath9 are the number densities of hydrogen atoms and dust grains , respectively , and @xmath10 is their relative velocity . @xmath0 ( sticking coefficient ) is the probability that an atom hitting a dust grain remains on the surface , and @xmath1 is the probability that an atom , once on the surface , makes an encounter with another h atom and recombines with it . for a typical diffuse cloud environment , ( @xcite ) it is assumed that , on average , @xmath11 . from a theoretical standpoint , the complexity of the problem lies in the fact that , due to the quantum nature of the h atom , a quantum mechanical calculation has to be done on a realistic , heterogeneous surface resembling dust grains in ism . hollenbach and salpeter ( hs ) ( @xcite ) calculated @xmath0 using simple semiclassical method for the atom - surface interaction , and found @xmath0 to be between @xmath12 and 1 in most cases . leitch - devlin and williams ( @xcite ) used a quantum mechanical approach and assumed a perfect single crystal with energy loss due to single phonon excitations . their sticking coefficient increases with gas temperature , reaches a maximum at @xmath13 comparable with phonon energies and then decreases again . buch and zhang ( @xcite ) numerically evaluated the sticking of hydrogen atoms on a cluster ( amorphous in structure ) of water molecules . they obtained @xmath14 , where @xmath15 is a parameter(@xmath16 for @xmath17 and @xmath18 for @xmath19 ) in hs s calculation ( @xcite ) , the tunneling efficiency of adsorbed hydrogen was high enough to ensure a very fast scanning of virtually all possible adsorption sites on a grain surface in a fraction of the residence time of the adsorbed species . in their model , the grain surface has two adsorption sites for h , one weak ( @xmath20 ) and the other strong ( @xmath21 ) . if only one weak adsorption site is present , h atoms would nt spend enough time on the surface to meet other h atoms and they would evaporate in a time @xmath22 , where @xmath23 is a typical time of a vibration of a h atom in an adsorption site . smoluchowski ( @xcite ) did a quantum mechanical calculation and found that @xmath17 would get trapped in the deepest sites of an amorphous surface after a few hops . his efficiency of @xmath3 production is several orders of magnitude smaller than hs s . pirronello and averna ( @xcite ) investigated the possibility that @xmath3 could be produced in dense clouds by cosmic rays bombardment of grain mantles . within the smoluchowsky s model , they found that the cosmic rays mechanism would be dominant over @xmath3 recombination on grains at 10 k. although sticking has been extensively studied experimentally ( @xcite ) , it has been done in conditions and on surfaces of almost no astrophysical interest , except for the following studies . brackmann and fite ( @xcite ) measured a sticking probability of 0.2 on a @xmath3-free surface and 0.5 on top of a @xmath3 layer in the temperature range of 2.5 to 8 k. due to poor vacuum conditions of their apparatus , it is assumed that their surfaces were covered with thick cryodeposits of background gas . there have been even fewer determinations in the laboratory of the hydrogen recombination coefficient on surfaces and up to now in conditions of no astrophysical relevancy . schutte et al . ( @xcite ) measured the h recombination on a surface of a bolometer at 3 k , a temperature significantly smaller than the one of interstellar grains ( 10 - 15 k ) . they found @xmath24 on a hypothetically @xmath3-free surface . other experiments ( @xcite ) measured recombination rates in situations where either the surface temperature or the kinetic energy of the h atoms were high . on the other hand , @xmath17 beam scattering experiments have shown that the interaction of _ low energy _ @xmath17 beams with single crystal graphite surfaces is rather weak ; the ground state is @xmath25 @xmath26 deep . ( @xcite ) experiments of scattering of @xmath17 from graphite single crystals above @xmath27 k gave a sticking coefficient of less than 0.1 and negligibly small above 21 k. ( @xcite ) the experimental studies mentioned above , although gave interesting information about physical / chemical process at surfaces , did nt really address the question of measuring the hydrogen recombination rate on surfaces in situations which could be related to astronomical observations . our study attempts to address this need . in this report , we present the results of a set of experiments to measure the recombination coefficient on the surface of an olivine slab ( a silicate rich in fe , mg , si and oxygen ) , an astrophysically relevant material , even if polycrystalline , as natural stones on earth are , and not amorphous , as required by observations ( @xcite ) . low kinetic energy of h beams , low fluxes , and low surface temperature have been used . the experiments were performed in an ultra - high vacuum apparatus consisting of a scattering chamber and of two triply differentially pumped beam lines ( see fig . each atomic beam is produced by rf dissociation of molecular hydrogen ( or deuterium ) in a water - cooled pyrex tube placed in a rf cavity . an eni power supply provides 100 - 150 watt of power to the cavities via a power splitter and impedance matching networks . by using @xmath17 and @xmath19 beams , the recombination product ( @xmath28 ) can be formed only on the surface , and not in the beam source due to imperfect dissociation , as in the case when using only one beam . the dissociation rate is measured by the quadrupole mass spectrometer in the main chamber . the maximum dissociation rate is over 90 @xmath29 , but in most cases was lower , around 70 - 85 @xmath29 , and it remained constant during experimental runs . each beam is made by the expansion of a low pressure gas ( 0.1 - 0.2 torr ) into vacuum through a short aluminum capillary of 1 mm in diameter connected to a @xmath30 reservoir . the beams enter the scattering chamber through 3 mm collimators . the estimated solid angle is @xmath31 and the flux for h is @xmath32 atoms/@xmath33/sec . to obtain a lower flux , a mechanical chopper with a duty cycle of @xmath34 was used . the @xmath17 beam hits the surface of the sample perpendicularly , while the @xmath19 beam hits it at 38 deg . from the normal . the sample is mounted on a copper sample holder that is in good thermal contact with a helitrans cold finger and its temperature is measured via two calibrated chromel / iron - gold thermocouples pressed against the top and bottom surfaces of the sample . the lowest temperature reached with the cold finger is @xmath35 ( top thermocouple ) . the sample can be heated by radiation and electron - beam bombardment with a heater placed in a small cavity of the sample holder just behind the sample . the incoming and reflected beams are detected by a differentially - pumped quadrupole mass spectrometer . most of the results reported here are for a sample of olivine donated by dr . p. plescia of the cnr institute for the treatments of minerals ( rome ) . it consists of a mixture of @xmath36 and @xmath37 . prior insertion into the apparatus , the sample was cleaned with mild solvents ( acetone , methanol , freon ) in an ultrasonic bath . the sample was then placed on the sample holder in the uhv chamber and the apparatus was baked to 150 @xmath38c for a couple of days . the base pressure was in the mid @xmath39 torr range . in a typical experiment , the sample is first heated to 200 @xmath38c for cleaning . after cooling , it is exposed for a given amount of time to the @xmath17 and @xmath19 beams . during the adsorption time , the detector is placed in front of the sample and the mass is tuned to 3 ( i.e. , mass of @xmath28 . there is no other background gas contributing to the spectrometer signal at mass `` 3 '' ) . at the end of the exposure time , the sample temperature is quickly ( @xmath40 k / sec ) raised to over 30 k by shutting the he flow to the tip of the cryostat ; the amount of @xmath28 desorbed as a function of time is recorded into a multichannel scaler . a set of representative traces is shown in fig . an analysis of desorption kinetics is given elsewhere . ( @xcite ) the background pressure of @xmath28 due to finite pumping capacity and/or recombination of @xmath17 and @xmath19 on the walls of the chamber is measured prior to and after each adsorption / desorption experiment , in front and back of the sample , and is subtracted from the recorded signal . we checked that after the blow - off of the @xmath28 layer there is no re - adsorption of @xmath28 from the gas phase . the sample holder is hugged by a copper radiation shield ; no adsorption / desorption of @xmath28 has been detected on / from the shield whose temperature is considerably higher than the sample . exposure times have been changed over several decades , see fig . 3 , where the desorption yield ( which is the integral of curve in fig.2 ) is plotted vs. exposure to @xmath17 and @xmath19 . the data are fitted well by a langmuir adsorption kinetics . the saturation of the signal is interpreted as due to the completion of the first layer . the time it takes to form a @xmath28 layer is consistent with the calculated flux and a low ( @xmath41 ) sticking coefficient ( @xcite ) . 4 shows the amount of @xmath28 desorbed following adsorption of @xmath17 and @xmath19 at various sample temperatures . the amount of @xmath28 produced and released _ during _ the adsorption process ( as measured by the mass spectrometer ) is typically much smaller than the amount released in the thermal desorption run . this means that at low sample temperatures ( 5 - 7 k ) , @xmath17 sticks ( although with a sticking probability well below 1 ) and readily recombines ; because of the low temperature , just @xmath28 that has just formed remains on the surface . at higher sample temperatures ( 10 - 15 k ) , @xmath17 does nt stay on the surface long enough to recombine , and the @xmath28 yield is much smaller . likely , only those @xmath17 atoms which became trapped in strong binding sites are retained at these temperatures ; @xmath28 is then produced at reduced rates because of slow diffusion of @xmath17 atoms out of these deeper energy sites . we could nt measure the @xmath17 signal during adsorption / desorption in a quantitative way because of the presence of background of @xmath17 in a stainless steel system ( even at 5 @xmath39 torr total base pressure ) . an analysis of desorption kinetics after adsorption of @xmath17 at high sample temperature could yield some valuable clues ; at present , however , the desorption signals are too low for a meaningful quantitative analysis . the recombination efficiency of @xmath28 , here defined as the fraction of @xmath17 atoms which stick and recombine to from @xmath28 - or @xmath42 , is calculated as follows ( note that this is different from the recombination rate @xmath5 defined earlier ) : @xmath43 , where @xmath44 is the @xmath28 signal during adsorption , @xmath45 is the total amount of @xmath46 desorbed after the adsorption process , and @xmath47 the the amount of h sent into the chamber . the signals are corrected for velocity - dependent detection efficiency and for the measured solid angle of the detector.@xmath48 is also corrected for the probability of forming @xmath28 into other reaction channels , and for the different measured intensities of the @xmath17 and @xmath19 beams . as mentioned before , only a small fraction of the amount of @xmath28 which desorbs in the thermal programmed desorption leaves the sample during the _ adsorption process_. consequently , the recombination efficiency drops with increasing surface temperature similarly to the fall off of the desorption yield as a function of temperature , see fig . the recombination efficiency was measured for different exposures to @xmath17 and @xmath19 , as seen in fig . other candidate materials for grains in the ism are : carbonaceous solids , both amorphous and crystalline , and icy mantles ( @xcite ) . we plan to investigate the formation of molecular hydrogen also on these other materials and we have already obtained some preliminar results ( to be published elsewhere ) for @xmath49 recombination on hopg ( highly oriented pyrolitic graphite ) . our estimate of @xmath50 at 5 - 6 k is not far from schutte et al. s value obtained on a uncharacterized surface of a semiconductor bolometer but at considerably lower temperature ( 3k ) . at the astrophysically relevant grain temperature range of 10 - 15 k , our result , @xmath51 is noticeably lower than the value of hs s model ( @xcite ) , @xmath52 at 10 - 15 k. we might try to speculate on the reason of this difference . calculations are for very idealized surfaces . there are a few assumptions that can influence the outcome of the calculation , such as the semiclassical description , the energetics of the binding sites , and the tunneling time between sites . the sample used by us presents a far more heterogeneous environment , both energetically and morphologically , than it has been considered in theoretical models . in a real surface , tunneling might proceed much more slowly , and the sticking coefficient for h on surfaces of olivine and graphite might be much different than previously calculated on model surfaces . considering that many models ( @xcite ) assume @xmath53 , and that a determination of @xmath0 for @xmath17 on graphite at @xmath54 gives @xmath55 ( @xcite ) , it is conceivable that @xmath0 is only partially responsible for the fact that our @xmath56 is a factor 10 lower than hollenbach s , and that @xmath1 might be appreciably less than @xmath57 , contrary to what is most often assumed in models . what are the consequences of this determination of @xmath58 on the recombination rate @xmath5 giving the constraints posed by astronomical observations ? in order to reconcile our value for @xmath59 with the accepted value ( @xcite ) for @xmath5 ( @xmath40 to @xmath60 ) , we propose that the surface area of grains ( which , in a simple model , enters the expression for @xmath61 as a multiplicative factor ) might be larger ( by a factor 5 to 10 or even more for amorphous surfaces ) . this could be indeed the case if grains have a larger surface area ( i.e , they are fluffier ) ( @xcite ) , than previously considered . this is not an unlikely scenario , considering the type of processing that grains are subjected to in the ism , and more theoretical and experimental effort should be devoted to verify it . duley , w.w . and williams , d.a . , interstellar chemistry ( 1984 ) hong , s.s . and greenberg , j.m . 1980 , astron.astrophys . greenberg , j.m . and li , a. 1996 , astron.astrophys . 309 , 258 , and references cited therein . mathis , j.m . 1993 , rep.prog.phys.56 , 605 . leger , a. , klein , j. , de cheveigne , s. , guinet , c. , defourneau , d. , belin , m. , astron . 1979 , 79 , 256 . jura , m. , astrophys.j . 1975 , 197 , 575 . hollenbach , d. and salpeter , e.e . , 1970 , j. chem phys . 53 , 79 ; 1971 , astrophys . j. 163 , 155 . leitch - devlin , m.a . and williams , d.a . 1984 , m.n.r.a.s . 210 , 577 ; 1985 , ibidem 213 , 295 . buch , v. and zhang , q. 1991 , astrophys . j. 379 , 647 . smoluchowski , r. 1981 , astrophys . space sci . 75 , 353 ; 1983 , j. phys . 87 , 4229 . pirronello , v. , and averna , d. 1988 , astron . 196 , 201 ; averna , d. and pirronello , v. 1991 , astron . 245 , 239 . rendulic , k.d . 1992 , surf . science 272 , 34 . brackmann , r.t . , and fite , w.l . 1961 , j.chem.phys . 34 , 1572 . schutte , a. , bassi , d. , tommasini , f. , turelli , f. , scoles , g. , herman , l.j.f . 1976 , j.chem.phys . 64 , 4135 . schermann , c. , gough , s.f . , pichou , f. , landau , m. , hall , r.i . , curie , m. 1993 , in molecules and grains in space , ed . by i.nenner , ( aip confer.proc . 312 ) , p.801 king , a.b . , and wise , h. 1963 , j.phys.chem . 67 , 1163 . ghio , e. , mattera , l. , salvo , c. , tommasini , f. , and valbusa , u. 1980 , j.chem.phys . lin , j. , and g.vidali , g. 1996 , in the cosmic dust connection by m.greenberg and v.pirronello ( kluwer , dodrect ) in press . draine , b.t . , and lee , h.m . 1984 , astophys . j. 285 , 89 . vidali , g. , liu , c. , and shen , l. 1996 , to be submitted . mathis , j.s . , and whiffen , g. 1989 , astrophys.j . 341 , 808 . support from nasa - astrophysics division is gratefully acknowledged . we thank dr . plescia of the cnr - rome for providing the olivine sample and eni of rochester , n.y . for donating the rf splitter . 0.3 true in @xmath62 to whom correspondence should be addressed ; e - mail : [email protected]	we report on the first results of experiments to measure the recombination rate of hydrogen on surfaces of astrophysical interest . our measurements give lower values for the recombination efficiency ( sticking probability @xmath0 x probability of recombination upon h - h encounter @xmath1 ) than model - based estimates . we propose that our results can be reconciled with _ average _ estimates of the recombination rate ( @xmath2 ) from astronomical observations , if the actual surface of an average grain is rougher , and its area bigger , than the one considered in models .
let @xmath0 be a graph ( we allow parallel edges but not loops ) . a _ walk _ in @xmath1 is a sequence w & = & ( v_0,e_1,v_1,e_2,v_2 , , e_m , v_m),m 0 , where @xmath2 for all @xmath3 , @xmath4 for all @xmath5 , and @xmath6 has endpoints @xmath7 and @xmath8 for all @xmath5 . we say that @xmath9 are the _ internal vertices _ of the walk @xmath10 . note that since we are allowing repetitions , the vertices @xmath11 could also be internal vertices . the walk @xmath10 is said to be _ closed _ when @xmath12 and is said to be a _ trail _ when the edges @xmath13 are distinct . now assume that the edges of @xmath1 are colored with a set @xmath14 of colors , where @xmath15 , the coloring being given by @xmath16 . we say that @xmath17 is an _ edge - colored graph_. the walk @xmath10 above is _ internally alternating _ when @xmath18 for each @xmath19 , and is _ alternating _ when in addition if @xmath10 is closed then @xmath20 ( note that a walk can be closed and internally alternating without being alternating , but if @xmath21 , there is no distinction between internally alternating and alternating walks and we use the word alternating in this case ) . a closed alternating walk ( respectively , trail ) is abbreviated as _ caw _ ( respectively , _ cat _ ) . given an edge - colored graph @xmath0 and a set @xmath22 of vertices called _ terminals _ , the _ alternating reachability problem _ is to either find an alternating trail connecting distinct terminals or show that none exists . our motivation for considering this problem arises from the following two special cases with two colors , say red and blue . * ( a ) * let @xmath0 be a simple graph and let @xmath23 be a matching . color the edges in @xmath24 red and the edges in @xmath25 blue , and let @xmath26 be the set of exposed vertices of @xmath24 . the alternating reachability problem for these data is equivalent to finding an augmenting path with respect to @xmath24 , if one exists . * ( b ) * let @xmath0 be a graph whose edges have been colored red and blue . the _ alternating cone _ is defined as the set of assignments of nonnegative weights to the edges such that at each vertex , the total red weight equals the total blue weight ( we define the alternating cone more formally below ) . in a companion paper @xcite we showed how the search for an integral weight vector within a given box in the alternating cone can be reduced to the alternating reachability problem in a 2-colored graph . we now outline the main results of this paper . in section [ sec4 ] we show , generalizing the approach of edmonds @xcite as explained in lovsz and plummer @xcite , that given an edge - colored graph @xmath0 and a set of terminals , either there is an alternating trail connecting distinct terminals , or else there is a subset of nonterminals , called a _ tutte set _ , that acts as an obstruction to such alternating trails ( for a precise definition of a tutte set , see section [ sec4 ] ) . moreover , we present a polynomial - time algorithm that finds one or the other . the alternating reachability problem ( in a slightly different version ) was first considered by tutte , in a nonalgorithmic form , in his book on graph theory @xcite and in the paper @xcite . tutte @xcite called the obstructions to alternating trails connecting distinct terminals @xmath27-barriers . every @xmath27-barrier is a tutte set but not conversely . it turns out that there is a very minor error in tutte s work : tutte s theory actually produces tutte sets in our sense and not @xmath27-barriers in his sense . also , it is easy to give an example where there exists a unique tutte set , which is not an @xmath27-barrier . in section [ sec5 ] we use tutte sets to prove a combinatorial result on cat s in edge - colored bridgeless graphs . to motivate this result consider the following statements : let @xmath28 be a directed graph in which every vertex has positive indegree and outdegree . then @xmath28 has a directed circuit ( this is easily proved ) . let @xmath17 be an edge - colored graph such that for every vertex @xmath29 of @xmath1 , we can find edges of two different colors incident with @xmath29 . then @xmath17 has a caw ( this can be proved by a simple alternating walk argument similar to the proof of theorem 2.2 of @xcite ) . theorem [ catbg ] strengthens the hypothesis of statement ( ii ) above by assuming in addition that @xmath1 is bridgeless . it concludes , using tutte sets , that @xmath17 has a cat . we also deduce a result ( theorem [ seymour - giles ] ) due to giles and seymour @xcite on cycles in bridgeless graphs from a special case of theorem [ catbg ] , where there are two colors , red and blue , and the red edges form a matching . in section [ sec6 ] we use theorem [ catbg ] to prove an intersection theorem for certain polyhedral cones associated to _ 2-colored graphs _ ( i.e. , edge - colored graphs where the number of colors is two ) that were defined in @xcite . we now recall some definitions from @xcite . let @xmath0 be a graph . assume that the edges of @xmath1 are colored red or blue , the coloring being given by @xmath30 . consider the real vector space @xmath31 , with coordinates indexed by the set of edges of @xmath1 . we write an element @xmath32 as @xmath33 . for an edge @xmath34 , the characteristic vector @xmath35 is defined by @xmath36 the _ cone of closed alternating walks _ or simply the _ alternating cone _ @xmath37 of a 2-colored graph @xmath17 ( when the coloring @xmath38 is understood , we suppress it from the notation and write @xmath39 ) , is defined to be the set of all vectors @xmath40 in @xmath31 satisfying the following system of homogeneous linear inequalities : [ 3 ] _ e e_r(v ) x(e ) - _ e e_b(v ) x(e ) & = & 0 , v v , + [ 4 ] x(e ) & & 0,e e . we refer to ( [ 3 ] ) as the _ balance condition _ at vertex @xmath29 . figure [ fig1_1 ] illustrates a 2-colored graph together with an integral vector in its alternating cone . in @xcite we determined the extreme rays and dimension of the alternating cone and showed that searching for an integral vector within a given box in the alternating cone can be reduced to searching for an alternating trail connecting two given vertices in a residual 2-colored graph . the characteristic vector of a walk @xmath41 is defined to be @xmath42 . a caw @xmath10 in a 2-colored graph is said to be _ irreducible _ if @xmath43 can not be written as @xmath44 for any caw s @xmath45 and @xmath46 . similarly , a a cat @xmath47 is said to be _ irreducible _ if @xmath48 can not be written as @xmath49 for any cat s @xmath50 and @xmath51 . figure [ fig1_5 ] depicts an irreducible caw ( with direction of walk indicated by an arrow ) and figure [ fig1_6 ] depicts an irreducible cat . irreducibility is easily seen . a simple alternating walk argument ( see @xcite ) shows that every integral vector in the alternating cone is a nonnegative integral combination of characteristic vectors of irreducible caw s . circulations in directed graphs can be thought of in terms of flows along the arcs obeying the conservation constraint at every vertex . for example , the characteristic vector of a directed circuit corresponds to a unit of flow along the circuit . such an interpretation is not available in the case of vectors in the alternating cone . the irreducible caw of figure [ fig1_5 ] does not correspond to a flow in an intuitive sense . on the other hand , the characteristic vector of an irreducible cat _ can _ be thought of as a unit of flow around the trail . for a 2-colored graph @xmath52 , it is thus natural to consider the convex polyhedral cone @xmath53 generated by the characteristic vectors of the cat s in @xmath17 . we call @xmath54 the _ cone of closed alternating trails _ , or simply the _ trail cone _ , of @xmath17 . for example , it is easily seen that the integral vector in the alternating cone from figure [ fig1_7 ] is not in the trail cone . on the other hand , the integral vector in the alternating cone from figure [ fig1_1 ] can be written as a sum of characteristic vectors of three cat s , as shown in figure [ fig1_8 ] . consider a cat in a 2-colored graph . its characteristic vector satisfies the balance condition at every vertex . if we ignore the colors , the edge set of the cat is a disjoint union of the edge sets of some cycles in the underlying graph . this shows that a nonnegative integral combination ( that is to say , a linear combination with nonnegative integral _ coefficients _ ) of characteristic vectors of cat s satisfies the balance condition at every vertex and can be written as a nonnegative integral combination of characteristic vectors of cycles in the underlying graph . we conjecture that the converse of this observation is also true : [ conj ] let @xmath52 be a 2-colored graph and let @xmath55 . then @xmath56 is a nonnegative integral combination of characteristic vectors of cat s in @xmath17 if and only if @xmath56 satisfies the balance condition at every vertex , i.e. , @xmath57 ; @xmath56 can be written as a nonnegative integral combination of characteristic vectors of cycles in @xmath1 . let @xmath58 denote the cone in @xmath31 generated by the characteristic vectors of the cycles in @xmath1 . seymour @xcite found the linear inequalities determining @xmath59 ( theorem [ sey ] ) . the observation in the paragraph preceding conjecture [ conj ] shows that @xmath60 . from conjecture [ conj ] it is easy to show that @xmath61 : take a rational vector @xmath62 ; for a suitably large positive integer @xmath63 , @xmath64 satisfies conditions ( i ) and ( ii ) of the conjecture ; by the conjecture @xmath64 ( respectively , @xmath56 ) is a nonnegative integral ( respectively , nonnegative rational ) combination of characteristic vectors of cat s . in section [ sec6 ] we prove that indeed @xmath65 . our proof is an adaptation of seymour s argument for @xmath58 . seymour s inductive proof is based on the giles - seymour lemma ( theorem [ seymour - giles ] ) and likewise , our inductive proof uses theorem [ catbg ] . we remark that in this paper we focus on graph - theoretical aspects of the alternating cone and not on algorithmic efficiency . we do consider algorithms , but always with a view to obtaining graph - theoretical results . we now collect in one place certain commonly used definitions in the rest of the paper . let @xmath0 be a graph and consider a walk [ w ] w & = & ( v_0,e_1,v_1,e_2,v_2 , , e_m , v_m),m 0 , in @xmath1 . we say that @xmath10 is a @xmath66-@xmath67 walk of length @xmath68 . we call @xmath69 the _ first _ edge of @xmath10 and @xmath70 the _ last _ edge of @xmath10 . the walk @xmath71 is the @xmath67-@xmath66 walk obtained by reversing the sequence ( [ w ] ) . the walk @xmath10 is said to be when the edges @xmath13 are distinct and the vertices @xmath72 are distinct ; when @xmath10 is closed , the edges @xmath13 are distinct , and the vertices @xmath73 are distinct . we have defined paths and cycles as special classes of walks . however , sometimes it is more convenient to think of paths and cycles as subgraphs , as is done usually . this will be clear from the context . if @xmath45 is a @xmath74-@xmath29 walk and @xmath46 is a @xmath29-@xmath75 walk , then the _ concatenation _ of @xmath45 and @xmath46 , denoted @xmath76 , is the @xmath74-@xmath75 walk obtained by walking from @xmath74 to @xmath29 along @xmath45 and continuing by walking from @xmath29 to @xmath75 along @xmath46 . note that if @xmath45 and @xmath46 are trails , then @xmath76 is a trail whenever @xmath45 and @xmath46 have no edges in common . now let @xmath17 be a 2-colored graph . the walk @xmath10 in ( [ w ] ) is said to be when @xmath10 is a cycle of even length and @xmath10 is alternating ; an even alternating cycle will also be called simply an _ alternating cycle _ ; when @xmath10 is a @xmath66-@xmath66 cycle of odd length and @xmath10 is internally alternating . let @xmath0 be a graph and let @xmath77 be an edge coloring . in this section we consider the _ alternating reachability problem _ : given a set @xmath26 of vertices called _ terminals _ , either find an alternating trail connecting distinct terminals or show that no such trail exists . for the rest of this section we consider @xmath1 , @xmath78 and @xmath26 as fixed . the problem of finding a cat through a given edge @xmath79 in an edge - colored graph can be easily reduced to the alternating reachability problem : let @xmath80 , @xmath81 . remove @xmath79 from the graph , add two new vertices @xmath82 and @xmath83 , add a new edge with the color @xmath84 between @xmath82 and @xmath85 and one between @xmath83 and @xmath86 , and let @xmath87 . clearly the alternating reachability problem in the new graph is equivalent to the original problem . as mentioned in the introduction , the problem of finding an augmenting path with respect to a given matching is also reducible to the alternating reachability problem . the alternating reachability problem was first considered by tutte @xcite . we discuss tutte s work at the end of this section . our solution to the alternating reachability problem is along the lines of the blossom forest algorithm of edmonds @xcite , as explained in section 9.1 of lovsz and plummer s book @xcite . the solution is in terms of tutte sets , defined below . [ dfs ] a subset @xmath88 is a _ tutte set _ when each component of @xmath89 has at most one terminal ; @xmath90 can be written as a disjoint union ( denoted @xmath91 , empty blocks allowed ) @xmath92 such that conditions ( a ) , ( b ) , and ( c ) below hold . a vertex @xmath93 is said to have _ color _ @xmath94 if @xmath95 ( there is a unique such @xmath94 ) . an edge @xmath34 is said to be _ mismatched _ if @xmath79 connects a vertex @xmath93 with a vertex @xmath96 and @xmath97 is different from the color of @xmath74 , or @xmath79 connects two vertices @xmath98 and @xmath97 is different from both the colors of @xmath74 and of @xmath29 . conditions ( a ) , ( b ) , and ( c ) are as follows : if @xmath99 is a component of @xmath100 containing a terminal , then there is no mismatched edge with an endpoint in @xmath99 ; if @xmath99 is a component of @xmath100 containing no terminals , then there is at most one mismatched edge with an endpoint in @xmath99 ; there are no mismatched edges with both endpoints in @xmath90 . the next theorem shows that a tutte set is an obstruction to the existence of an alternating trail connecting distinct terminals . [ ts ] suppose a tutte set @xmath90 exists . let @xmath85 be a terminal , let @xmath99 be a component of @xmath100 containing the vertex @xmath86 but not @xmath85 , and assume that there is no mismatched edge with an endpoint in @xmath99 . then there is no alternating @xmath85-@xmath86 trail . in particular , there is no alternating trail connecting distinct terminals . we assert that if an alternating trail @xmath47 enters @xmath90 from @xmath101 via an edge that is not mismatched , then the next time @xmath47 leaves @xmath90 , it can only be via a mismatched edge . specifically , suppose that @xmath102 is an alternating trail with @xmath103 , @xmath104 , and @xmath105 . assume that there exists some @xmath106 with @xmath107 , and let @xmath3 be the least such @xmath5 . proving the assertion amounts to showing @xmath108 . we show by induction that all @xmath109 satisfy @xmath110 . the base case @xmath111 follows from hypothesis . now assume that the statement holds for @xmath112 , where @xmath113 . thus @xmath114 , where @xmath94 is the color of @xmath115 . since @xmath47 is alternating , @xmath116 . since @xmath117 , it follows from condition ( ii)(c ) in definition [ dfs ] that the color of @xmath118 is @xmath119 . we have shown that @xmath120 . since @xmath47 is alternating , @xmath121 , and thus @xmath108 , which proves the assertion . now suppose that @xmath47 is an alternating @xmath85-@xmath86 trail . since @xmath85 and @xmath86 are in different components of @xmath100 , @xmath47 must enter @xmath90 . by ( ii)(a ) in definition [ dfs ] , the first time @xmath47 enters @xmath90 , it must be via an edge that is not mismatched . since @xmath122 , @xmath47 must leave @xmath90 . by the assertion , the first time @xmath47 leaves @xmath90 , it is via a mismatched edge . by hypothesis , the component @xmath123 of @xmath100 that @xmath47 enters upon leaving @xmath90 ( for the first time ) can not be the destination component @xmath99 containing @xmath86 , and thus @xmath47 must leave @xmath123 and enter @xmath90 again . by ( ii)(b ) in definition [ dfs ] , this entry must be via an edge that is not mismatched ( since the only mismatched edge has already been used for entering @xmath123 ) . therefore , by the assertion , when @xmath47 leaves @xmath90 for the next time , it must be via a mismatched edge . continuing this argument we see that every time @xmath47 leaves @xmath90 , it must be via a mismatched edge . thus @xmath47 can never reach the destination component @xmath99 containing @xmath86 , a contradiction . for later use we record the following lemma . [ os ] suppose a tutte set @xmath90 exists . if @xmath85 is a terminal and @xmath124 , then the last edge of each alternating @xmath85-@xmath74 trail has color @xmath94 . let @xmath102 be an alternating trail with @xmath125 and @xmath126 . by ( ii)(a ) in definition [ dfs ] , we see that when @xmath47 enters @xmath90 for the first time , it must be via an edge that is not mismatched . just as in lemma [ ts ] , we can now show that every time @xmath47 leaves @xmath90 , it must be via a mismatched edge , and every time @xmath47 enters @xmath90 , it must be via a edge that is not mismatched . since @xmath93 , @xmath47 must enter @xmath90 for the last time , say via the edge @xmath127 . since @xmath127 is not mismatched , @xmath128 . the induction argument used in proving the assertion in the proof of theorem [ ts ] now shows that @xmath129 . we now want to show that the converse of theorem [ ts ] holds , i.e. , if there is no alternating trail connecting distinct terminals , then a tutte set exists . we shall present an algorithm that finds such a trail or a tutte set . first , we need to make a few definitions . for a non root vertex @xmath74 in a rooted forest , the _ predecessor edge _ of @xmath74 is the first edge in the unique path from @xmath74 to the root of the component containing @xmath74 . given a partition @xmath130 of the vertex set of @xmath1 , by the _ shrunken graph _ @xmath131 we mean the graph obtained from @xmath1 by shrinking each block of @xmath130 into a vertex and discarding loops . in other words , the vertex set of @xmath132 is the set of blocks of @xmath130 , the edge set of @xmath132 is the set of edges of @xmath1 whose endpoints lie in different blocks of @xmath130 , and the endpoints of an edge @xmath79 ( in @xmath132 ) are the blocks of @xmath130 in which the endpoints of @xmath79 ( in @xmath1 ) lie . for @xmath133 , the block of @xmath130 containing @xmath29 is denoted by @xmath134 $ ] . for a subset @xmath135 , the subgraph of @xmath1 induced on @xmath136 is denoted by @xmath137 $ ] . [ db ] a subgraph @xmath138 of @xmath1 is said to be a _ blossom with base @xmath74 _ when @xmath139 ; for each @xmath140 such that @xmath141 , @xmath142 has two @xmath29-@xmath74 alternating trails whose first edges have different colors . condition ( ii ) allows us to extend any alternating trail reaching @xmath142 up to @xmath74 . note that a subgraph consisting of a single vertex @xmath74 is a blossom with base @xmath74 . [ dcb ] let @xmath143 . a subgraph @xmath138 of @xmath1 is said to be a _ @xmath94-blossom with base @xmath74 _ when @xmath139 ; for each @xmath140 such that @xmath141 , @xmath142 has two @xmath29-@xmath74 alternating trails whose first edges have different colors and whose last edges have colors different from @xmath94 ; @xmath142 has a @xmath74-@xmath74 internally alternating trail of positive length whose first and last edges have colors different from @xmath94 . conditions ( ii ) and ( iii ) allow us to extend any alternating trail reaching @xmath142 ( even a trail reaching @xmath74 via an edge of color @xmath94 ) up to @xmath74 and then be ready to continue with color @xmath94 . note that for all @xmath94 , a subgraph consisting of a single vertex @xmath74 is _ not _ a @xmath94-blossom . the following are examples of blossoms . if there are only two colors @xmath144 and @xmath145 , an odd internally alternating @xmath74-@xmath74 cycle with two @xmath144 edges incident at @xmath74 is a @xmath145-blossom ( see figure [ fig_15 ] ) . assuming there are three colors @xmath144 , @xmath145 and @xmath1 , figure [ fig_16 ] depicts two @xmath145-blossoms with base @xmath74 . figure [ fig_17 ] depicts two more @xmath145-blossoms with base @xmath74 . at this point this can be verified directly from definition [ dcb ] . later we shall see that the first of these @xmath145-blossoms arises by shrinking and the second by fusion . [ dcbf ] by a _ colored blossom forest _ we mean a triple @xmath146 , where @xmath147 , @xmath130 is a partition of @xmath148 , and @xmath149 is a rooted forest in @xmath150\times \pi$ ] , satisfying the following conditions : @xmath149 has @xmath151 components and the roots of @xmath149 are @xmath152 $ ] , @xmath153 ; for each terminal @xmath154 , the induced subgraph @xmath155 $ ] is a blossom with base @xmath85 ; let @xmath134 $ ] be an non root vertex of @xmath149 satisfying @xmath156=1 $ ] , let @xmath79 be the predecessor edge of @xmath134 $ ] in @xmath149 , and let @xmath157 be the edges between @xmath134 $ ] and its children in @xmath149 ; then @xmath158 all have colors different from @xmath97 ; let @xmath134 $ ] be an non root vertex of @xmath149 satisfying @xmath156 \geq 2 $ ] , let the predecessor edge of @xmath134 $ ] in @xmath149 have color @xmath94 , and let its endpoint ( in @xmath1 ) that is contained in @xmath134 $ ] be @xmath159 $ ] ; then the induced subgraph @xmath160 $ ] is a @xmath94-blossom with base @xmath74 . figure [ fig_18 ] depicts a colored blossom forest , where the blocks of @xmath130 are the blossoms , @xmath94-blossoms , and singleton inner vertices indicated in the figure . a colored blossom forest always exists : set @xmath161 , take @xmath130 to be the trivial partition of @xmath26 whose blocks are the singletons , and take @xmath149 to be the rooted forest with no edges and roots @xmath152 $ ] , @xmath154 ; then @xmath162 is a colored blossom forest . given a colored blossom forest @xmath163 , we classify the vertices of @xmath1 as follows . vertices in @xmath164 are called _ out - of - forest vertices_. for @xmath165 , if @xmath156 = 1 $ ] and @xmath134 $ ] is a non root vertex of @xmath149 , then we call @xmath29 an _ inner vertex_. all other vertices in @xmath148 are called _ blossom vertices_. if @xmath29 is an inner or blossom vertex in @xmath166 , we also call @xmath134 $ ] an inner or blossom vertex in @xmath149 , respectively . in item ( ii ) of definition [ dcbf ] , we say that @xmath85 is the _ base _ of the blossom vertex @xmath152 $ ] of @xmath149 , and in item ( iv ) of definition [ dcbf ] , we say that @xmath74 is the _ base _ of the blossom vertex @xmath134 $ ] of @xmath149 . we now define certain positive - length internally alternating trails in @xmath1 w.r.t . a colored blossom forest @xmath146 . let @xmath74 be a blossom vertex such that @xmath167 $ ] is not a root of @xmath149 . let the predecessor edge of @xmath167 $ ] in @xmath149 have color @xmath94 and assume that @xmath74 is the base of the blossom vertex @xmath167 $ ] of @xmath149 . pick an internally alternating @xmath74-@xmath74 trail of positive length in @xmath168 $ ] whose first and last edges have colors different from @xmath94 ( such a trail is guaranteed by the definition of a colored blossom forest ) , and call it @xmath169 . let @xmath74 be a blossom vertex such that @xmath167 $ ] is not a root of @xmath149 . let the predecessor edge of @xmath167 $ ] in @xmath149 have color @xmath94 and now assume that @xmath29 , @xmath141 , is the base of the blossom vertex @xmath167 $ ] of @xmath149 . pick two alternating @xmath74-@xmath29 trails in @xmath168 $ ] whose first edges have different colors and whose last edges have colors different from @xmath94 ( such trails are guaranteed by the definition of a colored blossom forest ) , and call them @xmath170 and @xmath171 . let @xmath172 be a blossom vertex such that @xmath167 $ ] is a root of @xmath149 . let @xmath154 be the base of the blossom vertex @xmath167 $ ] of @xmath149 . pick two alternating @xmath173 trails contained in @xmath155 $ ] , whose first edges have different colors ( such trails are guaranteed by the definition of a colored blossom forest ) , and call them @xmath170 and @xmath171 . the next lemma defines certain internally alternating trails in @xmath1 , and states their properties . some of these trails may be of length zero and some may be equal . these conventions prevent some case distinctions later on . [ pl ] let @xmath146 be a colored blossom forest , and let @xmath174 be vertices of @xmath1 satisfying the following conditions : * @xmath74 and @xmath29 both belong to @xmath148 ; * @xmath167 $ ] is a descendent of @xmath134 $ ] in @xmath149 ; * @xmath29 is either an inner vertex or the base of the blossom vertex @xmath134 $ ] of @xmath149 . then @xmath1 has internally alternating trails @xmath175 , @xmath176 satisfying the following properties . if @xmath74 is an inner vertex , then @xmath177 . the trail @xmath175 is of length zero precisely when @xmath178 . the trail @xmath176 is of length zero precisely when @xmath179 is inner or @xmath180 for some @xmath154 . the edges of the trails @xmath175 and @xmath176 include all the edges in the @xmath167$]-@xmath134 $ ] path in @xmath149 . if an edge @xmath79 in @xmath175 or @xmath176 is not in @xmath149 , then both its endpoints in @xmath1 are contained in a blossom vertex of @xmath149 lying on the @xmath167$]-@xmath134 $ ] path in @xmath149 . if @xmath74 is an inner vertex and @xmath141 , then the first edge of @xmath175 is the predecessor edge of @xmath167 $ ] in @xmath149 . if @xmath29 is an inner vertex and @xmath181 , then the last edges of @xmath175 and @xmath176 have colors different from the color of the predecessor edge of @xmath134 $ ] in @xmath149 . if @xmath74 is a blossom vertex and @xmath182 , then the first edges of @xmath183 and @xmath176 have different colors . if @xmath184 is the base of a blossom vertex , then the first and last edges of @xmath185 have colors different from the color of the predecessor edge of @xmath134 $ ] in @xmath149 . if @xmath184 is the base of a blossom vertex and @xmath181 , then the last edges of @xmath175 and @xmath176 have colors different from the color of the predecessor edge of @xmath134 $ ] in @xmath149 . the proof is by induction on the distance @xmath186,[v])$ ] in @xmath149 between the vertices @xmath167 $ ] and @xmath134 $ ] . first assume that @xmath186,[v ] ) = 0 $ ] , i.e. , @xmath74 and @xmath29 belong to the same block of @xmath130 . the following cases arise . * case ( a ) : * @xmath74 and @xmath29 are both inner . define @xmath187 , the zero - length trail starting and ending at @xmath74 . * case ( b ) : * @xmath74 and @xmath29 are both blossom vertices and @xmath188 . define @xmath189 * case ( c ) : * @xmath74 and @xmath29 are both blossom vertices and @xmath190 . define @xmath191 * case ( d ) : * @xmath74 and @xmath29 are both blossom vertices and @xmath179 , @xmath192 . define @xmath193 it is easily seen that @xmath175 and @xmath176 are internally alternating and conditions ( i)(ix ) in the statement of the lemma are satisfied ( when restricted to @xmath174 satisfying @xmath186,[v])=0 $ ] ) . now assume that @xmath186,[v ] ) > 0 $ ] . then @xmath181 , and let @xmath79 be the predecessor edge of @xmath167 $ ] in @xmath149 . let the endpoints of @xmath79 in @xmath1 be @xmath194 and @xmath56 , where @xmath195 $ ] . then @xmath196,[v])=d([u],[v])-1 $ ] and @xmath197 and @xmath198 will have been defined . the following cases arise . * case ( a ) : * @xmath74 is inner . we have @xmath199 . by induction and ( vii ) and ( iii ) of definition [ dcbf ] , one of @xmath200 is alternating ; define @xmath177 to be that trail ( breaking ties arbitrarily ) . * case ( b ) : * @xmath74 is a blossom vertex and @xmath201 . by the definition of @xmath170 , induction and ( vii ) , one of @xmath202 is alternating ; define @xmath175 to be that trail ( breaking ties arbitrarily ) . similarly , one of @xmath203 is alternating ; define @xmath176 to be that trail ( breaking ties arbitrarily ) . * case ( c ) : * @xmath74 is a blossom vertex and @xmath199 , @xmath172 . by ( iii ) of definition [ dcbf ] , induction and ( vii ) , one of @xmath204 is alternating ; define @xmath175 to be that trail ( breaking ties arbitrarily ) . similarly and by the definition of @xmath169 , one of @xmath205 is alternating ; define @xmath176 to be that trail ( breaking ties arbitrarily ) . * case ( d ) : * @xmath74 is a blossom vertex and @xmath199 , @xmath192 . this implies @xmath179 and thus @xmath186,[v])=0 $ ] . so this case can not occur . it is easily checked that @xmath175 and @xmath176 are alternating and conditions ( i)(ix ) in the statement of the lemma are satisfied . let @xmath146 be a colored blossom forest and let @xmath79 be an edge in @xmath1 , with endpoints @xmath74 and @xmath29 belonging to @xmath148 , that is not an edge of @xmath149 . assume that * @xmath167 $ ] and @xmath134 $ ] are not in the same component of @xmath149 . * if @xmath167 $ ] is an inner vertex of @xmath149 , then @xmath97 is different from the color of the predecessor edge of @xmath167 $ ] in @xmath149 . * if @xmath134 $ ] is an inner vertex of @xmath149 , then @xmath97 is different from the color of the predecessor edge of @xmath134 $ ] in @xmath149 . in this situation we say that we have a _ breakthrough_. [ btl ] assume we have a breakthrough , with the notation as in the preceding paragraph . let @xmath152,[t]$ ] , @xmath206 , @xmath81 be the roots of the components of @xmath149 containing @xmath167 $ ] and @xmath134 $ ] , respectively . then @xmath1 has an alternating @xmath85-@xmath86 trail . by ( vii ) of lemma [ pl ] and the definition of breakthrough , for some @xmath207 , @xmath208 is an alternating @xmath85-@xmath86 trail in @xmath1 . we now discuss three operations on a colored blossom forest @xmath146 : growing , shrinking and fusing . the operation of fusing does not occur in the classical case of searching for augmenting paths in a nonbipartite graph . let @xmath79 be an edge of @xmath1 between vertices @xmath74 and @xmath29 that is not an edge of @xmath149 . assume that * @xmath209 , @xmath210 . * if @xmath167 $ ] is an inner vertex of @xmath149 , then @xmath97 is different from the color of the predecessor edge of @xmath167 $ ] in @xmath149 . add the singleton block @xmath211 to @xmath130 to get a partition @xmath212 of @xmath213 . let @xmath214 denote the rooted forest in @xmath215 obtained by adding the inner vertex @xmath211 to the vertices of @xmath149 and adding the edge @xmath79 to the set of edges of @xmath149 . it is easily seen that @xmath216 is a colored blossom forest . we say that @xmath216 is obtained from @xmath146 by _ growing_. we now define the operation of shrinking . let @xmath79 be an edge of @xmath1 between vertices @xmath74 and @xmath29 that is not an edge of @xmath149 . assume that * @xmath217 . * @xmath167 $ ] and @xmath134 $ ] are in the same component of @xmath149 and @xmath167\neq [ v]$ ] . * if @xmath167 $ ] is an inner vertex of @xmath149 , then @xmath97 is different from the color of the predecessor edge of @xmath167 $ ] in @xmath149 . * if @xmath134 $ ] is an inner vertex of @xmath149 , then @xmath97 is different from the color of the predecessor edge of @xmath134 $ ] in @xmath149 . adding the edge @xmath79 to @xmath149 creates a unique cycle @xmath123 . let @xmath218=[v_0]$ ] denote the unique common ancestor in @xmath149 of @xmath167 $ ] and @xmath134 $ ] that belongs to @xmath123 . denote the vertices on the @xmath218$]-@xmath167 $ ] path in @xmath149 by @xmath218,[u_1],\ldots , [ u_{k-1}],[u_k]=[u]$ ] , and denote the edge of @xmath149 between @xmath219 $ ] and @xmath220 $ ] by @xmath221 . similarly , denote the vertices on the @xmath222$]-@xmath134 $ ] path in @xmath149 by @xmath222,[v_1],\ldots , [ v_{l-1}],[v_l]=[v]$ ] , and denote the edge of @xmath149 between @xmath223 $ ] and @xmath224 $ ] by @xmath225 . see figure [ fig_19 ] . note that @xmath63 or @xmath226 may be zero but @xmath227 . replace the blocks @xmath218,\ldots,[u_k],[v_0],\ldots,[v_l]$ ] of @xmath130 by their union to obtain a partition @xmath212 of @xmath148 with @xmath228 . define a rooted forest @xmath214 in @xmath229 by throwing away the edges @xmath230 from @xmath149 ( the remaining edges in @xmath149 have obvious endpoints in @xmath229 ) . for convenience , we denote the block of @xmath130 containing @xmath231 by @xmath232 $ ] and the block of @xmath212 containing @xmath75 by @xmath233 . we assert that @xmath234 is a colored blossom forest . clearly , conditions ( i ) and ( iii ) in definition [ dcbf ] are satisfied by @xmath234 , and we need to check condition ( ii ) or ( iv ) only for the vertex @xmath235 ( according as @xmath218 $ ] is a root of @xmath149 or not ) . we assume that @xmath218 $ ] is not a root of @xmath149 and check condition ( iv ) ; the case when @xmath218 $ ] is a root of @xmath149 and we need to check condition ( ii ) is similar , and we omit it . without loss of generality we may assume that if @xmath218 $ ] is a blossom vertex of @xmath149 , then its base is @xmath236 . let @xmath94 be the color of the predecessor edge of @xmath218 $ ] in @xmath149 . we show that @xmath237 $ ] is a @xmath94-blossom with base @xmath236 . first we check condition ( ii ) of definition [ dcb ] . let @xmath238 , @xmath239 . without loss of generality we may assume that @xmath240 $ ] for some @xmath241 . the following two cases arise . * case ( a ) : * @xmath220 $ ] is a blossom vertex of @xmath149 . by ( vii ) and ( ix ) of lemma [ pl ] , we have that @xmath242 and @xmath243 have first edges of different colors and have last edges with colors different from @xmath94 . * case ( b ) : * @xmath220 $ ] is an inner vertex of @xmath149 . in this case @xmath244 . from the assumption on @xmath167 $ ] and @xmath134 $ ] and ( vii ) of lemma [ pl ] , it follows that for some @xmath207 , the @xmath75-@xmath236 trail @xmath245 is alternating . since @xmath75 is an inner vertex , it now follows from ( v ) and ( vi ) of lemma [ pl ] that @xmath242 and @xmath246 have first edges of different colors , and by ( ix ) of lemma [ pl ] their last edges have colors different from @xmath94 . we have verified condition ( ii ) in definition [ dcb ] . now consider condition ( iii ) . as before , for some @xmath207 , the @xmath236-@xmath236 trail @xmath247 of positive length is internally alternating with first and last edges of colors different from @xmath94 . this completes the proof of the assertion that @xmath234 is a colored blossom forest . we say that @xmath234 is obtained from @xmath146 by _ shrinking_. now we define the operation of fusion . let @xmath79 be an edge of @xmath1 between vertices @xmath74 and @xmath29 that is an edge of @xmath149 . assume that * @xmath167 $ ] and @xmath134 $ ] are ( necessarily distinct ) blossom vertices of @xmath149 . then @xmath79 must be the predecessor edge of one of @xmath167 $ ] or @xmath134 $ ] . replace the blocks @xmath167 $ ] and @xmath134 $ ] of @xmath130 by their union to obtain a partition @xmath212 of @xmath148 . define a rooted forest @xmath214 in @xmath248 by throwing away the edge @xmath79 from @xmath149 ( the remaining edges in @xmath149 have obvious endpoints in @xmath229 ) . again we denote the block of @xmath212 containing a vertex @xmath231 by @xmath233 . we assert that @xmath234 is a colored blossom forest . without loss of generality , assume that @xmath79 is the predecessor edge of @xmath167 $ ] . clearly , conditions ( i ) and ( iii ) in definition [ dcbf ] are satisfied by @xmath234 , and we need to check condition ( ii ) or ( iv ) only for the vertex @xmath249 ( according as @xmath134 $ ] is a root of @xmath149 or not ) . we assume that @xmath134 $ ] is not a root of @xmath149 and check condition ( iv ) ; the case when @xmath134 $ ] is a root of @xmath149 and we need to check condition ( ii ) is similar , and we omit it . let @xmath194 be the base of the blossom vertex @xmath134 $ ] of @xmath149 , and let @xmath94 be the color of the predecessor edge of @xmath134 $ ] in @xmath149 . we show that @xmath250 $ ] is a @xmath94-blossom with base @xmath194 . condition ( iii ) in definition [ dcb ] clearly holds , since @xmath251 $ ] is contained in @xmath250 $ ] . condition ( ii ) of definition [ dcb ] follows from lemma [ pl ] : given that @xmath252 , @xmath253 , the two alternating @xmath75-@xmath194 trails @xmath254 and @xmath255 are contained in @xmath250 $ ] , have first edges of different colors , and have last edges of colors different from @xmath94 . this proves the assertion that @xmath234 is a colored blossom forest . we say that @xmath216 is obtained from @xmath146 by _ we now put a partial order on colored blossom forests . given colored blossom forests @xmath256 and @xmath257 , we say that @xmath258 if @xmath148 is a proper subset of @xmath259 , or @xmath260 and @xmath261 ( as partitions , i.e. , every block of @xmath130 is contained in a block of @xmath212 and @xmath262 ) . the next theorem is the promised converse of theorem [ ts ] . [ cts ] suppose that @xmath1 has no alternating trail connecting distinct terminals . let @xmath163 be a maximal colored blossom forest . let @xmath90 be the set of inner vertices of @xmath1 with respect to @xmath263 . for each @xmath143 , define @xmath264 \text { in } f \text { has color } c\}.\ ] ] then @xmath90 is a tutte set with coloring given by @xmath265 . before proving theorem [ cts ] , we prove the following properties of the maximal colored blossom forest @xmath163 . [ cfs ] under the assumptions of theorem [ cts ] : no edge in @xmath1 connects a blossom vertex and an out - of - forest vertex . if an edge @xmath79 in @xmath1 connects an inner vertex and an out - of - forest vertex , then @xmath97 agrees with the color of the inner vertex . no edge in @xmath1 connects blossom vertices contained in two different vertices of @xmath149 . if an edge @xmath79 in @xmath1 connects two inner vertices , then @xmath97 agrees with the color of one of them . if an edge @xmath79 in @xmath1 connects a blossom vertex @xmath29 and an inner vertex @xmath74 , then @xmath97 agrees with the color of @xmath74 , except when @xmath29 is the base of a blossom and @xmath79 is the predecessor edge of @xmath134 $ ] in @xmath149 , in which case the colors are different . if condition ( i ) or ( ii ) does not hold , we can grow the colored blossom forest @xmath146 , contradicting its maximality . we now consider condition ( iii ) . let @xmath79 be an edge in @xmath1 between blossom vertices @xmath74 and @xmath29 . if @xmath167 $ ] and @xmath134 $ ] are in different components of @xmath149 , then we have a breakthrough which , by lemma [ btl ] , contradicts our assumption that there are no alternating trails connecting distinct terminals . if @xmath167 $ ] and @xmath134 $ ] are in the same component of @xmath149 and @xmath79 is an edge of @xmath149 , then we can fuse , contradicting the maximality of @xmath146 . if @xmath167 $ ] and @xmath134 $ ] are in the same component of @xmath149 and @xmath79 is not an edge of @xmath149 , then we can shrink , again contradicting the maximality of @xmath146 . now we verify condition ( iv ) . let @xmath79 be an edge in @xmath1 between the inner vertices @xmath74 and @xmath29 . if @xmath79 is an edge of @xmath149 , then it is the predecessor edge of one of @xmath167 $ ] and @xmath134 $ ] , and thus @xmath97 agrees with the color of @xmath74 or @xmath29 . if @xmath79 is not an edge of @xmath149 and @xmath167 $ ] and @xmath134 $ ] are in different components of @xmath149 and @xmath97 is different from the colors of @xmath74 and @xmath29 , then we have a breakthrough , a contradiction . if @xmath79 is not an edge of @xmath149 and @xmath167 $ ] and @xmath134 $ ] are in the same component of @xmath149 and @xmath97 is different from the colors of @xmath74 and @xmath29 , then we can shrink , a contradiction . finally , consider condition ( v ) . let @xmath79 be an edge in @xmath1 between a blossom vertex @xmath29 and an inner vertex @xmath74 . if @xmath79 is an edge of @xmath149 , then either @xmath79 is the predecessor edge of @xmath167 $ ] , in which case @xmath97 agrees with the color of @xmath74 , or @xmath29 is the base of the blossom vertex @xmath134 $ ] of @xmath149 and @xmath79 is the predecessor edge of @xmath134 $ ] , in which case @xmath97 is different from the color of @xmath74 . if @xmath79 is not an edge of @xmath149 , then @xmath97 agrees with the color of @xmath74 , because otherwise we have a breakthrough ( if @xmath134 $ ] and @xmath167 $ ] are in different components of @xmath149 ) or we can shrink ( if @xmath134 $ ] and @xmath167 $ ] are in the same component of @xmath149 ) . let @xmath266 be the components of the subgraph of @xmath1 induced by the out - of - forest vertices . write @xmath267 and list the non inner vertices of @xmath149 as @xmath268,[s_2],\ldots , [ s_k],[s_{k+1}],\ldots , [ s_l]\ ] ] ( so that the set of blossom vertices in @xmath166 is precisely @xmath269 \cup \cdots \cup [ s_l]$ ] ) . statements ( i ) and ( iii ) in lemma [ cfs ] imply that the components of @xmath100 are precisely @xmath270,\ldots , g[[s_l]],k_1,\ldots , k_p.\ ] ] condition ( i ) in definition [ dfs ] now follows . statement ( iv ) in lemma [ cfs ] proves condition ( ii)(c ) in definition [ dfs ] . we verify condition ( ii)(a ) in definition [ dfs ] using statement ( v ) in lemma [ cfs ] , by noting that @xmath269,\ldots,[s_k]$ ] are roots of @xmath149 and therefore have no predecessor edge . condition ( ii)(b ) in definition [ dfs ] follows from statements ( ii ) and ( v ) in lemma [ cfs ] : from ( ii ) we see that there are no mismatched edges with an endpoint in @xmath271 , and from ( v ) we see that there is exactly one mismatched edge with an endpoint in each of @xmath272,\ldots , g[[s_l]]$ ] . in the spirit of the gallai - edmonds decomposition , the out - of - forest , inner , and blossom vertices w.r.t . a maximal colored blossom forest can be characterized as follows . define @xmath273 to be the set of all vertices @xmath274 such that for all @xmath154 , there is no alternating @xmath85-@xmath86 trail in @xmath1 . define @xmath275 to be the set of all vertices @xmath274 such that for some @xmath154 , there are two alternating @xmath85-@xmath86 trails whose last edges have different colors . for a color @xmath143 , define @xmath276 to be the set of all vertices @xmath274 satisfying the following property : there are alternating trails starting from @xmath26 and ending in @xmath86 , and the last edges of all such alternating trails have the color @xmath94 . set @xmath277 . lemma [ os ] implies that if @xmath90 is a tutte set , then @xmath278 . theorem [ oib ] below and theorems [ ts ] and [ cts ] show that if a tutte set exists , then @xmath279 is one ( indeed , if a tutte set exists , theorem [ ts ] shows that @xmath1 has no alternating trail connecting distinct terminals ; then theorem [ cts ] shows that the inner vertices of a maximal colored blossom forest form a tutte set ; and theorem [ oib ] shows that @xmath279 are the inner vertices ) . [ oib ] suppose that @xmath1 has no alternating trail connecting distinct terminals . let @xmath163 be a maximal colored blossom forest . then @xmath280 let @xmath281 be the set of out - of - forest vertices , and let @xmath282 be the set of blossom vertices . define @xmath90 and @xmath283 as in theorem [ cts ] . we first show that @xmath284 , @xmath285 , and @xmath286 . write the components of @xmath100 ( in the notation of the proof of theorem [ cts ] ) as @xmath270,\ldots , g[[s_l]],k_1,\ldots , k_p.\ ] ] by lemma [ cfs](ii ) , there is no mismatched edge with an endpoint in @xmath287 . it now follows from theorem [ ts ] that there is no alternating trail from any vertex in @xmath26 to any out - of - forest vertex , i.e. , @xmath284 . for each @xmath95 , it follow by lemmas [ pl](v ) and [ os ] that @xmath288 . thus @xmath285 . lemma [ pl](vii ) implies that @xmath289 . since @xmath281 , @xmath290 , @xmath282 partition @xmath166 and @xmath273 , @xmath279 , @xmath291 are disjoint , it follows that @xmath292 , @xmath293 , @xmath294 . the theory presented in this section produces in polynomial time either an alternating trail connecting distinct terminals or a tutte set . indeed , start with the trivial colored blossom forest @xmath130 defined after definition [ dcbf ] , and perform growing , shrinking and fusing operations in any order until a breakthrough or a maximal colored blossom forest is achieved . discovering that one of these operations is possible and performing it or discovering a breakthrough takes polynomial time . as for the number of operations , initially @xmath130 has @xmath151 blocks . shrinking and fusing decrease the number of blocks of @xmath130 and keep the number of out - of - forest vertices constant . growing increases the first by one and decreases the second by one , so at most @xmath295 growing steps can occur in total . it follows that termination must occur within @xmath296 operations . we now comment on tutte s work on the alternating reachability problem . tutte gives a nonalgorithmic solution to a slightly different version of the alternating reachability problem . tutte calls the obstructions to the existence of alternating trails @xmath27-barriers @xcite . there is a small but important difference between our definition of tutte set and the definition of @xmath27-barrier . if we were to apply tutte s definition of @xmath27-barrier to the version of alternating reachability considered in this paper , then condition ( c ) in definition [ ts](ii ) would read : \(c ) if an edge @xmath79 connects two vertices of @xmath90 , then these vertices have different colors and one of them has the color @xmath84 , instead of our condition paraphrased : \(c ) if an edge @xmath79 connects two vertices of @xmath90 , then one of them has the color @xmath84 . thus every @xmath27-barrier is a tutte set but not conversely . it is easy to find instances of the alternating reachability problem where there are no alternating trails connecting distinct terminals , but all obstructions are tutte sets and not @xmath27-barriers . we give such an example for tutte s original version of the alternating reachability problem . we follow the definitions and notation of tutte s paper @xcite without reproducing them here . consider the graph in figure [ ct_ex_1 ] ( 1 and 2 are colors , as in tutte s paper ) . in this graph there are no bicursal edges and the only acursal edge is between @xmath74 and @xmath29 , all vertices are unicursal , and @xmath297 , @xmath298 . this is a counter - example to tutte s theorem 5.1 , since @xmath299 is not an @xmath27-barrier : the edge between @xmath74 and @xmath29 violates condition ( ii ) in the definition of an @xmath27-barrier . now consider the graph in figure [ fig_counter_example_tutte ] . it is easy to convince oneself of the following : there is no `` coloured path '' in @xmath300 from @xmath27 to @xmath301 . there is no @xmath27-barrier such that @xmath301 is a vertex of an inaccessible outer component . in other words , with tutte s definition of an @xmath27-barrier , his main result theorem 5.3 is false . the solution is to replace condition ( ii ) in tutte s definition of an @xmath27-barrier by the following : \(ii ) if both ends of an edge @xmath90 of @xmath1 are inner vertices of @xmath302 , then one of them has the color of @xmath90 . this is precisely our notion of a tutte set . we believe that the only ( very minor ) error in * @xcite * occurs in the proof of theorem 5.1 , and that the theory presented in that paper actually yields a tutte set rather than an @xmath27-barrier . tutte has several applications of this theorem in his book on graph theory @xcite . it is unlikely that these applications need the stronger notion of barriers ; in most applications , tutte sets would do . edges like the one between @xmath74 and @xmath29 in the first example and between @xmath194 and @xmath56 in the second are acursal in tutte s terminology , and therefore will not be used by @xmath300 . in this section we give an application of tutte sets to closed alternating trails in edge - colored bridgeless graphs . this result will be used in the next section where we determine the inequalities defining the trail cone . [ catbg ] let @xmath0 be a bridgeless graph with an edge - coloring @xmath303 . assume that for each @xmath133 , there are two edges with different colors incident at @xmath29 . then @xmath304 has a cat . suppose that @xmath304 has no cat . let @xmath79 be an edge in @xmath1 between @xmath85 and @xmath86 . form a new edge - colored graph @xmath142 from @xmath1 by removing @xmath79 , and adding two new vertices @xmath82 and @xmath83 and two new edges @xmath305 between @xmath82 and @xmath85 and @xmath306 between @xmath83 and @xmath86 , both with the color @xmath97 . since @xmath1 has no cat , it follows that @xmath142 has no @xmath82-@xmath83 alternating trail . consider the alternating reachability problem for @xmath142 with @xmath307 and consider the sets @xmath276 , @xmath143 defined in the previous section . let @xmath90 denote the set of inner vertices of a maximal colored blossom forest . it follows from theorem [ oib ] that @xmath308 is a tutte set with @xmath82 and @xmath83 in different components of @xmath309 that are incident with no mismatched edges . since @xmath79 is not a bridge in @xmath1 , @xmath82 and @xmath83 are in the same component of @xmath142 , and hence @xmath310 . we first make the following two observations . there are edges of two different colors incident at every vertex of @xmath142 other than @xmath82 and @xmath83 . if @xmath311 , then @xmath85 must be in the component of @xmath312 containing @xmath82 ; and similarly if @xmath122 , then @xmath86 must be in the component of @xmath309 containing @xmath83 . by definition of @xmath279 , there exist alternating trails in @xmath142 starting at @xmath82 or @xmath83 and ending at a vertex in @xmath90 . let @xmath47 denote the longest such alternating trail , starting at @xmath82 without loss of generality and ending at @xmath313 . let @xmath119 be the last edge of @xmath47 . by lemma [ os ] , @xmath314 . by observation ( i ) above we can find an edge @xmath315 of @xmath142 incident with @xmath301 satisfying @xmath316 . let @xmath75 be the other endpoint of @xmath315 . we shall now derive a contradiction . the following three cases arise . * case ( a ) : * @xmath47 contains @xmath315 . since @xmath317 , it follows from @xmath314 that @xmath47 must have traversed @xmath315 in the direction from @xmath301 to @xmath75 . thus , the portion of @xmath47 starting with @xmath315 and ending with @xmath119 is a cat in @xmath142 . since @xmath82 and @xmath83 have degree 1 in @xmath142 , this cat can not contain the edges @xmath305 or @xmath306 , so is a cat in @xmath1 , a contradiction . * case ( b ) : * @xmath318 and @xmath47 does not contain @xmath315 . in this case , @xmath319 is an alternating trail that is longer than @xmath47 and ends at a vertex in @xmath90 , a contradiction . * case ( c ) : * @xmath320 and @xmath47 does not contain @xmath315 . let @xmath99 be the component of @xmath309 containing @xmath75 . since @xmath316 , it follows that @xmath315 is mismatched and @xmath99 does not contain @xmath82 or @xmath83 . hence by observation ( ii ) above , @xmath321 or @xmath322 , and similarly for @xmath86 , so @xmath79 does not connect @xmath90 and @xmath99 . by the proof of lemma [ os ] , @xmath47 can enter a component of @xmath309 only via a mismatched edge . it follows ( since @xmath315 is the only mismatched edge incident with @xmath99 ) that @xmath47 never enters @xmath99 , and so @xmath47 has no vertices in @xmath99 . since @xmath301 is an inner vertex and @xmath75 is a blossom vertex , it follows from lemma [ cfs](v ) that @xmath75 is the base of a @xmath323-blossom , which coincides with @xmath99 as in the proof of theorem [ cts ] . if @xmath315 were the only edge of @xmath142 between @xmath90 and @xmath99 , then it would follow that @xmath315 is a bridge not only in @xmath142 but also in @xmath1 , a contradiction . thus @xmath1 has some edge @xmath324 , @xmath325 between @xmath326 and @xmath327 . since @xmath47 has no vertices in @xmath99 , it contains neither @xmath315 nor @xmath324 nor any edge of @xmath99 . by lemma [ pl](ii ) , ( vii ) , ( ix ) we see that one of @xmath328 and @xmath329 is an alternating trail from @xmath82 to @xmath90 that is longer than @xmath47 , a contradiction . the algorithm of section [ sec4 ] gives a polynomial - time method of finding a cat in an edge - colored graph satisfying the hypothesis of theorem [ catbg ] : for each edge @xmath79 of @xmath1 , construct the corresponding graph @xmath142 and look for an alternating @xmath82-@xmath83 trail . for some edge @xmath79 we will find an alternating @xmath82-@xmath83 trail in @xmath142 , which easily yields a cat in @xmath1 . actually , the proof of theorem [ catbg ] gives a method of finding a cat in @xmath1 by solving the alternating reachability problem ( with @xmath307 ) for a single @xmath142 ( for an arbitrary edge @xmath79 in @xmath1 ) . we do not give the details . conjecture [ conj ] can be seen as strengthening both the hypothesis and conclusion of theorem [ catbg ] ( in the case of 2 colors ) . the nontrivial part of the conjecture states that if a sum of cycles is balanced , then it is a sum of cat s . we may delete bridges and then isolated vertices from @xmath0 , since they are irrelevant to the conjecture . then some positive vector @xmath55 is a sum of cycles . theorem [ catbg ] assumes that @xmath330 ( the support of @xmath56 ) has edges of both colors at every vertex ; conjecture [ conj ] assumes more , namely that @xmath56 is balanced . theorem [ catbg ] concludes that @xmath1 has a cat , whereas conjecture [ conj ] concludes more , namely that @xmath56 is a sum of cat s . we conclude this section by deducing the lemma of giles and seymour on cycles in bridgeless graphs ( see @xcite ) from theorem [ catbg ] . [ seymour - giles ] let @xmath0 be a bridgeless graph and let @xmath331 map each vertex @xmath29 to an edge incident at @xmath29 . then @xmath1 has a cycle @xmath14 such that for each vertex @xmath75 of @xmath14 , @xmath332 is an edge of @xmath14 . partition @xmath330 as @xmath333 , where for @xmath334 , @xmath335 consists of those edges @xmath79 such that @xmath336 for exactly @xmath3 endpoints @xmath29 of @xmath79 . define a 2-colored graph as follows . first subdivide each edge in @xmath337 by introducing a new vertex , i.e. , for each edge @xmath338 between @xmath74 and @xmath75 , introduce a new vertex @xmath339 and replace @xmath79 with the two edges between @xmath74 to @xmath339 and between @xmath75 to @xmath339 . there is an obvious map @xmath340 from the edges of the resulting graph @xmath142 onto @xmath330 ( @xmath340 is the identity on @xmath341 and @xmath342 and takes each subdivided edge onto its parent edge in @xmath337 ) . we color the edges in @xmath341 blue and the edges in @xmath342 red . now consider an edge @xmath338 between @xmath74 and @xmath75 , say with @xmath343 and @xmath344 . then we color the edge between @xmath74 and @xmath339 red and the edge between @xmath75 and @xmath339 blue . it is easily seen that @xmath142 is bridgeless . every vertex of @xmath142 has positive red and blue degrees . the red edges form a matching in @xmath142 . from ( i ) and ( ii ) , it follows by theorem [ catbg ] that @xmath142 has a cat @xmath47 . from ( iii ) , @xmath47 must be an even alternating cycle . then @xmath345 is a cycle in @xmath1 with the required properties . let @xmath346 be a 2-colored graph . in this section we show , using theorem [ catbg ] , that @xmath347 . seymour @xcite found the linear inequalities determining @xmath348 and our proof is modeled after his . given a nonempty proper subset @xmath281 of @xmath166 , the subset @xmath349 of edges between @xmath281 and @xmath350 will be called a _ cut_. we say that @xmath281 and @xmath350 are the two _ sides _ of the cut , and their _ sizes _ are @xmath351 and @xmath352 . let @xmath28 be a cut , @xmath353 , and @xmath14 a cycle in @xmath1 . if @xmath14 contains @xmath79 , then @xmath14 must also contain an edge in @xmath354 . thus the characteristic vector @xmath355 of @xmath14 satisfies the following inequality [ cutcondition ] x(e ) _ f d - e x(f ) , where we write @xmath356 for @xmath354 . we abbreviate the right - hand side of ( [ cutcondition ] ) by @xmath357 . we call ( [ cutcondition ] ) the _ cut condition _ for the pair @xmath358 . if it holds with equality , the pair @xmath358 is said to be _ tight for @xmath194_. given a graph @xmath0 , we let @xmath362 denote the set of all cycles in @xmath1 . if @xmath363 is a 2-coloring of @xmath1 , then we denote the set of all cat s in @xmath304 by @xmath364 . [ il1 ] let @xmath0 be a graph and let @xmath365 . let @xmath28 be a cut in @xmath1 , and let @xmath353 be such that @xmath358 is tight for @xmath366 . suppose @xmath367 , which means @xmath366 can be be expressed as @xmath368 let @xmath369 with @xmath370 . then @xmath371 is either empty or equal to @xmath372 for some @xmath373 ( we think of @xmath14 as a set of edges ) . suppose that @xmath374 is a 2-coloring and @xmath375 , which means @xmath366 can be be expressed as @xmath376 let @xmath377 with @xmath378 . then @xmath379 is either empty or equal to @xmath372 for some @xmath373 . we prove ( i ) ; the proof of ( ii ) is similar . we have _ c # ( c\{e } ) ( c ) & = & _ c , e c ( c ) + & = & p(e ) + & = & p(d - e ) + & = & _ h d - e _ ( c ) + & = & _ c # ( c ( d - e ) ) ( c ) . since each @xmath380 satisfies @xmath381 , it follows that each @xmath380 with @xmath382 satisfies @xmath383 . since @xmath384 , the result follows . [ il2 ] let @xmath385 be a 2-colored graph , let @xmath28 be a cut in @xmath1 with sides @xmath281 and @xmath350 , and let @xmath353 be an edge with an endpoint @xmath386 . let the pair @xmath358 be tight for the weight function @xmath387 . suppose that @xmath388 and that @xmath366 satisfies the balance condition at the unique vertex of @xmath281 . then each edge of @xmath389 has color opposite @xmath97 . it follows that for each @xmath390 , the intersection @xmath391 is either empty or equal to @xmath372 for some @xmath373 . suppose that @xmath392 and that @xmath366 satisfies the balance condition at both vertices of @xmath281 . then each edge in @xmath389 with color opposite @xmath97 has @xmath393 as an endpoint , and each edge in @xmath389 with color @xmath97 does not have @xmath393 as an endpoint . it follows that for each @xmath390 , the intersection @xmath391 is either empty or equal to @xmath372 for some @xmath373 . ( i ) this is clear . \(ii ) let @xmath394 . without loss of generality we may assume that @xmath395 . set x_1 & = & p(e ) , + x_2 & = & _ d p(d ) , , + x_3 & = & _ d p(d ) , , + x_4 & = & _ d p(d ) , , + x_5 & = & _ d p(d ) , , + x_6 & = & _ d p(d ) , , + x_7 & = & _ d p(d ) , . since the pair @xmath358 is tight we have [ bal1 ] x_1 = x_2 + x_3 + x_4 + x_5 . the balance condition at @xmath393 gives [ bal2 ] x_1+x_2+x_7=x_3+x_6 , and the balance condition at @xmath396 gives [ bal3 ] x_5+x_6=x_4+x_7 . adding ( [ bal2 ] ) and ( [ bal3 ] ) gives @xmath397 comparing this equation with ( [ bal1 ] ) we get @xmath398 , and hence @xmath399 by nonnegativity . this implies that there are no red edges in @xmath389 incident with @xmath393 and no blue edges in @xmath28 incident with @xmath396 . for a graph @xmath0 and a nonempty proper subset @xmath281 of @xmath166 , we denote by @xmath400 the graph obtained by shrinking @xmath281 to a single vertex ( and deleting the resulting loops ) . [ il3 ] let @xmath0 be a graph and let @xmath401 . let @xmath281 be a nonempty proper subset of @xmath166 and let @xmath402 denote @xmath366 restricted to the edges of @xmath400 . if @xmath366 is cut - admissible for @xmath1 , then @xmath402 is cut - admissible for @xmath400 . this follows since each cut in @xmath400 is also a cut in @xmath1 . let @xmath0 be a graph with a 2-coloring @xmath403 . for @xmath404 , we denote by @xmath405 the set of all edges of @xmath1 with one endpoint in @xmath281 and the other endpoint in @xmath282 . let @xmath28 be a cut in @xmath1 with sides @xmath281 and @xmath350 of sizes at least 3 , and let @xmath353 be an edge between @xmath386 and @xmath406 . let @xmath407 be a weight function . * delete all edges in @xmath410 ( respectively , @xmath411 ) . * replace @xmath281 ( respectively , @xmath350 ) with @xmath412 ( respectively , @xmath413 ) , where @xmath414 are two new vertices . * the endpoints of each edge in @xmath415 ( respectively , @xmath416 ) remain the same . * the endpoint of each edge @xmath417 in @xmath350 ( respectively , @xmath281 ) is the same as before , and the endpoint in @xmath281 ( respectively , @xmath350 ) is @xmath393 ( respectively , @xmath418 ) if @xmath419 and is @xmath420 ( respectively , @xmath421 ) if @xmath422 . * add a new edge @xmath423 ( respectively , @xmath424 ) between @xmath393 and @xmath420 ( respectively , @xmath418 and @xmath421 ) . * the color of @xmath423 ( respectively , @xmath424 ) is opposite @xmath97 . all other edges retain their original color . * define a weight function @xmath425 on the edges of @xmath408 by setting @xmath426 , where the sum is over all @xmath373 with @xmath427 , and @xmath428 for all other edges @xmath315 of @xmath408 . similarly , define a weight function @xmath429 on the edges of @xmath409 by setting @xmath430 , where the sum is over all @xmath373 with @xmath427 , and @xmath431 for all other edges @xmath315 of @xmath409 . our restriction on the sizes of @xmath281 and @xmath350 ensures that @xmath408 and @xmath409 have fewer vertices than @xmath1 . [ il4 ] let @xmath432 be a 2-colored graph , let @xmath28 be a cut in @xmath1 with sides @xmath281 and @xmath350 of sizes at least 3 , and let @xmath353 be an edge between @xmath386 and @xmath406 . let the pair @xmath358 be tight for the weight function @xmath433 . suppose that @xmath366 satisfies the balance condition at each vertex of @xmath1 and is cut - admissible for @xmath1 . then @xmath425 ( respectively , @xmath429 ) satisfies the balance condition at each vertex of @xmath408 ( respectively , @xmath409 ) and is cut - admissible for @xmath408 ( respectively , @xmath409 ) . furthermore , @xmath358 is tight for @xmath425 and @xmath429 . suppose that @xmath425 is a nonnegative integral combination of characteristic vectors of cat s in @xmath408 , and @xmath429 is a nonnegative integral combination of characteristic vectors of cat s in @xmath409 . then @xmath366 is a nonnegative integral combination of characteristic vectors of cat s in @xmath1 . \(i ) from the definition of @xmath434 and @xmath435 and the hypothesis that @xmath358 is tight for @xmath366 and that @xmath366 satisfies the balance condition at each vertex of @xmath1 , it is clear that @xmath425 and @xmath429 satisfy the balance condition at each vertex of @xmath408 and @xmath409 , respectively , and that @xmath358 is tight for @xmath425 and @xmath429 . we shall now verify that @xmath425 is cut - admissible for @xmath408 ; the proof for @xmath429 is the same . consider the graph @xmath400 . we retain the name @xmath393 for the vertex obtained by shrinking @xmath281 . let @xmath402 be @xmath366 restricted to the edges of @xmath400 . note that @xmath28 is a cut in @xmath400 and @xmath358 is tight for @xmath402 . by lemma [ il3 ] , @xmath402 is cut - admissible for @xmath400 and thus , by theorem [ sey ] , we can write @xmath436 consider a cycle @xmath437 in @xmath438 with @xmath439 . we obtain a cycle @xmath440 in @xmath408 from @xmath437 as follows : * if @xmath437 does not intersect @xmath28 , then @xmath437 is a cycle in @xmath408 and we set @xmath441 . * if @xmath437 intersects @xmath28 then , by lemma [ il1](i ) , this intersection must be @xmath372 for some @xmath373 . if @xmath393 is an endpoint of @xmath315 in @xmath408 ( i.e. , if @xmath442 ) then @xmath437 is also a cycle in @xmath408 and we set @xmath441 . if @xmath393 is not an endpoint of @xmath315 in @xmath408 ( i.e. , if @xmath443 ) we define @xmath440 to be the cycle in @xmath408 obtained from @xmath437 by inserting @xmath423 between @xmath79 and @xmath315 . to prove the assertion , we examine separately the two kinds of edges of @xmath408 : edges of @xmath444 and @xmath423 if @xmath305 is an edge of @xmath444 , then each @xmath445 with @xmath439 satisfies @xmath446 , hence now sum ( [ hfixed ] ) over all @xmath453 such that @xmath454 . since for each @xmath445 such that @xmath439 and @xmath455 there is exactly one edge @xmath453 such that @xmath450 and @xmath456 , the left - hand side of ( [ hfixed ] ) sums to @xmath457 . the right - hand side of ( [ hfixed ] ) sums to \(ii ) the hypothesis on @xmath425 ( respectively , @xmath429 ) implies that there is a multiset @xmath459 ( respectively , @xmath460 ) of cat s in @xmath408 ( respectively , @xmath409 ) such that every edge @xmath315 in @xmath408 ( respectively , @xmath409 ) appears @xmath461 ( respectively , @xmath462 ) times in the various cat s contained in @xmath459 ( respectively , @xmath460 ) . we now build a multiset @xmath463 of cat s in @xmath1 such that every edge @xmath315 in @xmath1 appears @xmath464 times in the cat s contained in @xmath463 . this will prove the result . we begin with some notation . let @xmath47 be a cat in @xmath1 whose intersection with @xmath28 has exactly 2 edges @xmath79 and @xmath315 for some @xmath373 . let @xmath465 and @xmath466 be the endpoints of @xmath315 in @xmath281 and @xmath350 , respectively . then an appropriate cyclic shift of @xmath47 must have the form @xmath467 where @xmath468 is a @xmath393-@xmath465 alternating trail whose vertices are in @xmath281 , and @xmath469 is a @xmath466-@xmath418 alternating trail whose vertices are in @xmath350 . consider a cat in @xmath459 or @xmath460 that intersects @xmath28 . by lemma [ il1](ii ) , the intersection of each such cat with @xmath28 must be @xmath372 for some @xmath373 . for @xmath373 , let @xmath470 ( respectively , @xmath471 ) consist of the cat s in @xmath459 ( respectively , @xmath460 ) whose intersection with @xmath28 is @xmath372 . by the definition of @xmath459 and @xmath460 , @xmath425 and @xmath429 , we have @xmath472 for each @xmath473 . for each @xmath373 , fix a bijection @xmath474 . * add to @xmath463 all cat s in @xmath459 whose vertices are contained in @xmath350 ( each such cat is added the same number of times as it appears in @xmath459 ) . * add to @xmath463 all cat s in @xmath460 whose vertices are contained in @xmath281 . * for every @xmath373 and every @xmath475 add the cat @xmath476 to @xmath463 . it is easily checked that each edge @xmath315 in @xmath1 appears @xmath464 times in the cat s contained in @xmath463 ( in particular , this holds for @xmath477 because @xmath358 is tight for @xmath366 ) . let @xmath432 be a 2-colored graph . then @xmath478 we have already seen in the introduction that @xmath479 . consider a nonnegative rational vector @xmath480 that satisfies the balance condition at every vertex of @xmath1 and that is cut - admissible for @xmath1 . we will show that @xmath481 , which will prove the result . without loss of generality we may assume that @xmath482 for all @xmath34 ( we may drop edges @xmath79 with @xmath483 and maintain the balance condition and cut - admissability ) . the proof is by induction on the pairs @xmath484 ordered lexicographically . for a suitably large positive integer @xmath486 , the vector @xmath487 is integral . thus , by lemma [ il4](i ) , @xmath425 @xmath488respectively , @xmath429@xmath489 satisfies the balance condition at every vertex of @xmath408 @xmath488respectively , @xmath409@xmath489 and is cut - admissible for @xmath408 @xmath488respectively , @xmath409@xmath489 . since @xmath408 and @xmath409 have fewer vertices than @xmath1 , we see by induction and lemma [ il4](ii ) that for a suitably large positive integer @xmath24 , the vector @xmath490 is a nonnegative integral combination of characteristic vectors of cat s in @xmath1 . it follows that @xmath481 , as required . since @xmath485 is positive on every edge and cut - admissible for @xmath1 , it follows that @xmath1 is bridgeless . since @xmath485 satisfies the balance condition at every vertex , it follows from theorem [ catbg ] that @xmath17 has a cat @xmath47 . consider the vector @xmath492 . clearly , @xmath493 is balanced for all @xmath494 and nonnegative for all sufficiently small @xmath495 . we assert that there is a positive rational @xmath496 such that @xmath497 is cut - admissible for @xmath1 . indeed , let @xmath28 be a cut in @xmath1 with sides @xmath281 and @xmath350 , and let @xmath353 . we have the following two cases . * case ( b ) : * @xmath500 . by assumption , one of @xmath281 and @xmath350 , say @xmath281 , has size at most 2 . by lemma [ il2 ] we see that either @xmath47 contains no edge of @xmath28 or it contains precisely two edges of @xmath28 , @xmath79 and @xmath315 , for some @xmath373 . it follows that @xmath501 for all @xmath494 . * subcase ( ii.1 ) : * @xmath504 for some @xmath505 . by dropping @xmath305 we obtain a graph with the same number of vertices as @xmath1 but with fewer edges , while maintaining balance and cut - admissability . thus by induction @xmath375 , and hence @xmath506 , as required . * subcase ( ii.2 ) : * @xmath507 for all @xmath505 . from case ( b ) above we see that @xmath508 implies @xmath509 . since @xmath366 is positive on every edge , it must be that the cutoff determining @xmath496 occurs by case ( a ) above and not by case ( b ) or by the requirement that @xmath510 . therefore there is a cut @xmath511 and an edge @xmath512 such that @xmath513 and @xmath514 . thus @xmath366 is a positive rational vector , balanced and cut - admissible for @xmath1 , and more pairs @xmath358 are tight for @xmath366 than for @xmath485 . we may now repeat the whole argument with @xmath366 in place of @xmath485 . since the total number of pairs @xmath358 where @xmath28 a cut in @xmath1 and @xmath353 is finite , eventually we will reach case ( i ) or subcase ( ii.1 ) . finally , we would like to state the following problems . in @xcite we saw that the problem of finding an integral vector in the intersection of the alternating cone with a box leads to the alternating reachability problem . we can ask a similar question for the trail cone , but with the integrality restriction dropped . given a 2-colored graph with nonnegative rational upper and lower bounds on the edges , is there an augmenting - path - type algorithm for either finding a rational vector in the trail cone satisfying these bounds , or showing that no such vector exists ? in * @xcite * seymour makes the following conjecture for a graph @xmath0 : if @xmath515 , then @xmath56 is a sum of cycles , i.e , @xmath56 can be written as a nonnegative integer linear combination of characteristic vectors of cycles in @xmath1 . a vector in @xmath58 is a nonnegative rational combination of characteristic vectors of cycles , i.e. , is a fractional sum of cycles . so seymour s conjecture can be stated as follows : a fractional sum of cycles that is an even integer on every edge is a sum of cycles . conjecture [ conj ] states that a balanced sum of cycles is a sum of cat s . is there any relation between these conjectures ?	we consider a graph with colored edges . a trail ( vertices may repeat but not edges ) is called _ alternating _ when successive edges have different colors . given a set of vertices called _ terminals _ , the _ alternating reachability _ problem is to find an alternating trail connecting distinct terminals , if one exists . a special case with two colors is searching for an augmenting path with respect to a given matching . in another special case with two colors red and blue , the _ alternating cone _ is defined as the set of assignments of nonnegative weights to the edges such that at each vertex , the total red weight equals the total blue weight ; in a companion paper we showed how the search for an integral weight vector within a given box in the alternating cone can be reduced to the alternating reachability problem in a 2-colored graph . we define an obstacle , called a _ tutte set _ , to the existence of an alternating trail connecting distinct terminals in a colored graph , and give a polynomial - time algorithm , generalizing the blossom algorithm of edmonds , that finds either an alternating trail connecting distinct terminals or a tutte set . we use tutte sets to show that an an edge - colored bridgeless graph where each vertex has incident edges of at least two different colors has a closed alternating trail . a special case with two colors one of which forms a matching yields a combinatorial result of giles and seymour . we show that in a 2-colored graph , the cone generated by the characteristic vectors of closed alternating trails is the intersection of the alternating cone with the cone generated by the characteristic vectors of cycles in the underlying graph .
our starting point is kannan s result in metric fixed point theory involving contractive type mappings which are not necessarily continuous @xcite . it has been shown in @xcite that kannan s theorem is independent of the famous banach contraction principle ( see , e.g. @xcite ) , and that it also characterizes the metric completeness concept @xcite . let @xmath0 be a metric space . let us call @xmath1 a _ kannan map _ if there exists some @xmath2 such that @xmath3 for all @xmath4 . for complete metric spaces , kannan proved the following : ( @xcite)[thm : if is ] if @xmath0 is a complete metric space , and if @xmath5 is a kannan map on @xmath6 , then there exists a unique @xmath7 such that @xmath8 . and subrahmanyam ( in @xcite ) has proved the counterpart by showing that if all the kannan maps on a metric space have fixed points then that space must necessarily be complete . as in @xcite , from this point onwards let @xmath0 stand for a complete metric space . let us select arbitrarily a point @xmath9 , and call it the `` zero '' of @xmath6 . we denote @xmath10 let @xmath11 $ ] be fixed constants , and let @xmath12\rightarrow[0,\,\infty)$ ] denote a preassigned increasing function that vanishes ( with continuity ) at zero . then , for a map @xmath1 , pata shows that the following theorem holds . ( @xcite)[thm : pata ] if the inequality @xmath13^{\beta}\label{eq : w}\ ] ] is satisfied for every @xmath14 $ ] and every @xmath15 then @xmath5 possesses a unique fixed point @xmath16 ( @xmath17 ) . motivated by this generalization of the banach fixed point theorem , we can come up with an analogous generalized form of theorem [ thm : if is ] . with all the other conditions and notations remaining the same except for a more general @xmath18 , our goal is to prove the following : [ thm : if - the - inequality]if the inequality @xmath19^{\beta}\label{eq : b}\end{gathered}\ ] ] is satisfied @xmath20 $ ] and @xmath21 , then @xmath5 possesses a unique fixed point @xmath22 since we can always redefine @xmath23 to keep ( [ eq : b ] ) valid no matter what initial @xmath9 we choose , we are in no way restricting ourselves by taking that zero instead of a generic @xmath7 @xcite . we claim first that such an @xmath24 , if it exists , is unique . to see that this is the case , let , if possible , @xmath25 such that @xmath26 then ( [ eq : b ] ) implies , @xmath20,$ ] @xmath27^{\beta}.\ ] ] in particular , @xmath28 gives us @xmath29 which is a contradiction . we now bring into play the two sequences @xmath30 but before we proceed any further , we will need the following : [ lem :- is - bounded.]@xmath31 is bounded . from ( [ eq : b ] ) , considering again the case of @xmath32 we see that for @xmath33 , @xmath34 now , @xmath35 , @xmath36}\\ & \leq d(tx_{n},\ , tx_{0})+2c_{1}\\ & \leq\frac{1}{2}\left\ { d(x_{n+1},\ , x_{n})+d(x_{1},\ , x_{0})\right\ } + 2c_{1 } & \textrm{[using ( \ref{eq : b } ) with } \varepsilon=0\,\textrm{]}\\ & \leq\frac{1}{2}(c_{1}+c_{1})+2c_{1 } & \textrm{[(\ref{eq : c})]}\\ & = 3c_{1}. \end{aligned}\ ] ] and hence the lemma is proved . next we strive to show that : [ lem :- is - cauchy.]@xmath37 is cauchy . in light of ( [ eq : b ] ) , for @xmath33 , @xmath38^{\beta}\\ & \leq\frac{1-\varepsilon}{2}\left\ { d(x_{n+1},\ , x_{n})+d(x_{n},\ , x_{n-1})\right\ } + c\varepsilon^{\alpha}\psi(\varepsilon)\end{aligned}\ ] ] for @xmath39 ( on account of lemma [ lem :- is - bounded . ] ) . but then , @xmath40,$ ] @xmath41\nonumber \\ & \leq k^{n}d(x_{1},\ , x_{0})\nonumber \\ & + \frac{2c\varepsilon^{\alpha}}{1+\varepsilon}\psi(\varepsilon)(1+k+\cdots+k^{n-1 } ) & \textrm{[letting } k=\frac{1-\varepsilon}{1+\varepsilon}\geq0\textrm{]}\nonumber \\ & \le k^{n}d(x_{1},\ , x_{0})\nonumber \\ & + \frac{2c\varepsilon^{\alpha}}{1+\varepsilon}\psi(\varepsilon)(1+k+\cdots+k^{n-1}+\cdots ) & \textrm{[because } k\geq0 ] \nonumber \\ & \le k^{n}d(x_{1},\ , x_{0})\nonumber \\ & + \frac{2c\varepsilon^{\alpha}}{1+\varepsilon}\psi(\varepsilon)\frac{1}{1-k}\nonumber \\ & = k^{n}d(x_{1},\ , x_{0})+c\varepsilon^{\alpha-1}\psi(\varepsilon ) & \textrm{[putting } k=\frac{1-\varepsilon}{1+\varepsilon}]\label{eq : z}\end{aligned}\ ] ] for all @xmath42 . at this point we note that if @xmath43,$ ] then @xmath44 therefore , taking progressively lower values of @xmath45 that approach zero but never quite reach it , the r.h.s . of ( [ eq : z ] ) can be made as small as one wishes it to be as @xmath46 indeed , since @xmath47 as @xmath48 for an arbitrary @xmath49 @xmath50 such that @xmath51 again , for this @xmath45 ( @xmath52 ) , @xmath53 such that @xmath54 @xmath55 because @xmath56 as @xmath46 together that gives us @xmath57 in other words , @xmath58 hence , from ( [ eq : b ] ) , using the same @xmath59 , and letting @xmath60 , @xmath48 @xmath61 } \end{aligned}\ ] ] uniformly over @xmath62 which basically assures us that @xmath37 is cauchy . equipped with ( [ lem :- is - cauchy . ] ) and taking into note the completeness of @xmath6 , we can now safely guarantee the existence of some @xmath63 to which @xmath37 converges . finally , all that remains to show is that : for this we observe that , @xmath65 @xmath66}\nonumber \\ & \implies\frac{1}{2}d(tx^{},\ , x^{})\leq\frac{1}{2}d(x_{n},\ , x_{n+1})+d(x_{n+1},\ , x^{})\label{eq : f}\end{aligned}\ ] ] as @xmath60 ( and @xmath67 ) , we know that : @xmath68;}\\ & d(x_{n+1},\ , x^{})\rightarrow0 & \textrm{[since } x_{n}\rightarrow x^{*}].\end{aligned}\ ] ] so ( [ eq : f ] ) actually gives us that @xmath69 which is the required result . the requirements of theorem [ thm : if - the - inequality ] are indeed weaker than those of kannan s theorem . to see that , let us start from ( [ eq : a ] ) with @xmath70 ( barring the trivial case where @xmath71 ) . we have , @xmath20,$ ] @xmath72}\nonumber \\ & \leq\frac{1-\varepsilon}{2}\{d(x,\ , tx)+d(y,\ , ty)\}\nonumber \\ & + \frac{\lambda}{2}\varepsilon^{\frac{1}{\lambda}}\left[1+\left\ { d(x,\ , x_{0})+d(x_{0},\ , tx)\right\ } + \left\ { d(y,\ , x_{0})+d(x_{0},\ , ty)\right\ } \right]\nonumber \\ & = \frac{1-\varepsilon}{2}\{d(x,\ , tx)+d(y,\ , ty)\}+\frac{\lambda}{2}\varepsilon^{1+\gamma}[1+\left\vert x\right\vert + \left\vert tx\right\vert + \left\vert y\right\vert + \left\vert ty\right\vert ] \nonumber \\ & \textrm{[taking } 1<\frac{1}{\lambda}=1+\gamma\,\textrm { for some } \gamma>0]\nonumber \\ & \leq\frac{1-\varepsilon}{2}\{d(x,\ , tx)+d(y,\ , ty)\}+\frac{\lambda}{2}\varepsilon\varepsilon^{\gamma}[1+\left\vert x\right\vert + \left\vert tx\right\vert + \left\vert y\right\vert + \left\vert ty\right\vert ] .\label{eq : g}\end{aligned}\ ] ] then a quick comparison between ( [ eq : b ] ) and ( [ eq : g ] ) with @xmath73 ( @xmath74 ) provides us with what we need .	we prove a generalization of kannan s fixed point theorem , based on a recent result of vittorino pata .
in algebraic geometry , the linear systems of divisors on curves are well studied ( cf . other authors studied linear systems of divisors on metric graphs , for example @xcite . baker and norine @xcite proved a graph - theoretic analogue of the riemann - roch theorem , and it was generalized to tropical curves ( which may contain unbounded rays ) independently by gathmann and kerber @xcite and by mikhalkin and zharkov @xcite . the theory of linear systems on metric graphs is applied to algebraic geometry , for example in @xcite . haase , musiker and yu @xcite studied the cell complex structure of @xmath0 and the tropical semi - module structure of @xmath2 , where @xmath1 is a divisor on a metric graph . this work is an extension of @xcite . we focus on the computation of the cell complex @xmath0 , namely given a metric graph @xmath3 and a divisor @xmath1 on it , how to find the cells in @xmath0 and the @xmath4-vector of @xmath0 . since @xmath0 may contain a large number of cells and some of these are complicated , the complexity of computation could be high . we introduce the _ anchor cell _ that serves as the landmarks to find other cells in @xmath0 . as a byproduct we can compute the extremal generators of @xmath2 . we implemented the algorithms and computed some examples - namely the canonical linear systems on some trivalent graphs . a _ metric graph _ @xmath3 is a connected undirected graph whose edges have positive lengths . a _ divisor _ @xmath1 on @xmath3 is a formal finite @xmath5-linear combination @xmath6 of points @xmath7 in the edges of @xmath3 . the divisor is _ effective _ if @xmath8 for all @xmath9 . the _ degree _ of a divisor @xmath1 is @xmath10 . the _ support _ of a divisor @xmath1 on @xmath3 is the set @xmath11 , denoted as @xmath12 . a _ ( tropical ) rational function _ @xmath4 on @xmath3 is a continuous function @xmath13 that is piecewise - linear on each edge of @xmath3 with finitely many pieces and integer slopes . the _ order _ @xmath14 of @xmath4 at a point @xmath9 is the sum of the outgoing slopes at @xmath15 along all directions . note that if @xmath15 is an interior point of a linear piece of @xmath4 , then there are two directions at @xmath15 , and @xmath15 has two opposite outgoing slopes , so @xmath16 . the _ principal divisor _ associated to @xmath4 is @xmath17 so the support of @xmath18 is always finite . two divisors @xmath1 and @xmath19 are _ linearly equivalent _ if @xmath20 for some rational function @xmath4 , denoted as @xmath21 . for any divisor @xmath1 on @xmath3 , let @xmath2 be the set of all rational functions @xmath4 on @xmath3 such that the divisor @xmath22 is effective , and @xmath23 , the _ linear system _ of @xmath1 . [ exa : c4 ] let @xmath3 be a metric graph with graph - theoretic type @xmath24 and equal edge lengths . below are examples of @xmath1 and @xmath25 . ( 2,2 ) (0,2 ) ( 0,0 ) ; ( 0,0 ) circle ( 1pt ) ; ( 2,0 ) circle ( 1pt ) ; ( 2,2 ) circle ( 1pt ) ; ( 0,2 ) circle ( 1pt ) ; at ( 0,0 ) q ; at ( 2,0 ) r ; at ( 0,2 ) p ; at ( 2,2 ) s ; ( 0,0 ) ( 2,0 ) ( 2,2 ) (0,2 ) ( 0,0 ) ; ( 0,0 ) circle ( 1pt ) ; ( 2,0 ) circle ( 1pt ) ; ( 2,2 ) circle ( 1pt ) ; ( 0,2 ) circle ( 1pt ) ; at ( 0,0 ) 2 ; at ( 2,0 ) 2 ; at ( 0,2 ) 2 ; at ( 2,2 ) 2 ; ] ( 0,0 ) ( 2,0 ) ( 2,2 ) (0,2 ) ( 0,0 ) ; ( 2,2 ) circle ( 1pt ) ; ( 0.5,2 ) circle ( 1pt ) ; ( 1.5,2 ) circle ( 1pt ) ; ( 1.33,0 ) circle ( 1pt ) ; ( 0,1 ) circle ( 1pt ) ; at ( 2,2 ) 1 ; at ( 0.5,2 ) 1 ; at ( 1.5,2 ) 1 ; at ( 0,1 ) 2 ; at ( 1.33,0 ) 3 ; at ( 0,0 ) q ; at ( 2,0 ) r ; at ( 0,2 ) p ; at ( 2,2 ) s ; the metric graph @xmath3 is determined by its graph - theoretic type and the lengths of its edges . the graph - theoretic type of @xmath3 is called the _ skeleton _ of @xmath3 , and the lengths of edges of @xmath3 are called the _ metric _ of @xmath3 and denoted by @xmath26 . for an edge @xmath27 of @xmath3 we denote by @xmath28 the length of @xmath27 in @xmath26 . for any metric graph @xmath3 , the _ canonical divisor _ @xmath29 of @xmath3 is the divisor on @xmath3 with @xmath30 when @xmath15 is a vertex of @xmath3 and @xmath31 otherwise . when we fix the skeleton of @xmath3 , for any metric @xmath26 and divisor @xmath1 on @xmath3 we denote by @xmath32 the set @xmath2 and by @xmath33 the linear system @xmath0 . given a metric graph @xmath34 and a divisor @xmath1 on @xmath3 . @xmath1 is _ vertex - supported _ if @xmath35 . note that @xmath12 is always a finite set . if @xmath1 is not vertex - supported , then we can refine @xmath3 to get a new metric graph whose set of vertices is @xmath12 . we shall assume from now on that @xmath1 is vertex - supported . also we may assume that @xmath0 is not empty , so if @xmath1 is not effective , we can consider an effective divisor @xmath36 . it is obvious that @xmath37 . we shall assume from now that @xmath1 is vertex - supported . in section [ sec:\|d\| ] , we present the cell complex structure of @xmath0 and introduce the _ anchor cells_. we use them as landmarks to find all cells in the complex and prove a combinatorial formula ( corollary [ cor : fvec ] ) for the @xmath4-vector of @xmath0 . we introduce an algorithm to compute the cells of @xmath0 given @xmath3 and @xmath1 . in section [ sec : r(d ) ] , we regard @xmath2 as a tropical semi - module ( convex set ) and introduce an algorithm to find the extremal generators of @xmath2 , using a result in @xcite based on the chip - firing technique . in section [ sec : exp ] we apply our algorithms to examples of trivalent graphs . finally we raise some open problems in section [ sec : q ] . in this section we present the cell complex structure of @xmath0 . note that if @xmath38 is a constant rational function on @xmath3 , then for any rational function @xmath4 on @xmath3 , the divisor @xmath39 is equal to @xmath22 . let @xmath40 be the set of constant functions on @xmath3 . the set @xmath41 can be identified with the linear system @xmath0 by the map @xmath42 . we adapt the formulation of @xmath0 as a cell complex in @xcite , which originates from @xcite and @xcite . [ def : cell ] we identify each open edge @xmath43 with the interval @xmath44 ( this implicitly gives an orientation of @xmath27 , while it is independent to the cell complex ) . then each cell of @xmath0 is characterized by the following data : * a nonnegative integer @xmath45 for each @xmath46 ; * an ordered partition @xmath47 of positive integers for some edges @xmath43 ; * an integer @xmath48 for each @xmath43 . a divisor @xmath49 belongs to this cell if and only if the following statements hold : * @xmath50 for each @xmath46 ; * for those @xmath43 with the partition above , the divisor @xmath49 on @xmath27 is expressed as @xmath51 , where @xmath52 ; for other edges @xmath27 , @xmath53 for all points @xmath15 in the interior of @xmath27 ; * for any @xmath25 such that @xmath54 , the outgoing slope of @xmath4 at the point @xmath55 is @xmath48 for each @xmath43 . the slopes @xmath48 are also required because two distinct rational functions in @xmath2 may lead to different effective divisors in @xmath0 with exactly the same values of the @xmath45 and the ordered partitions of @xmath56 . [ ext : cou ] let @xmath57 be a loop @xmath27 with one vertex @xmath58 and the length of @xmath27 equal to @xmath59 . let @xmath60 . then the two effective divisors in figure [ fig : cir ] belong to @xmath0 . for both , all @xmath45 are @xmath55 and all @xmath56 are @xmath59 with one part in the partition , while the slopes @xmath48 are different : one is @xmath61 and the other is @xmath62 . ( 2,2 ) circle [ radius=2 ] ; ( 2,0 ) circle ( 1pt ) ; ( 0.27,3 ) circle ( 3pt ) ; at ( 2,0 ) v ; at ( 0.27,3 ) 3 ; at ( 1,0 ) -2 ; at ( 3,0 ) -1 ; ( 2,2 ) circle [ radius=2 ] ; ( 2,0 ) circle ( 1pt ) ; ( 3.73,3 ) circle ( 3pt ) ; at ( 2,0 ) v ; at ( 3.73,3 ) 3 ; at ( 1,0 ) -1 ; at ( 3,0 ) -2 ; a cell is represented by any divisor @xmath22 in it , because once we have the rational function @xmath25 , we can compute all data in definition [ def : cell ] . there is a natural question : how can we tell the dimension of a cell from a representative ? the following proposition is a good answer . ( ( * ? ? ? * proposition 13))[prop : dim ] let @xmath1 be a vertex - supported effective divisor on a metric graph @xmath3 and @xmath63 be the set of vertices in @xmath3 . let @xmath64 be a cell in @xmath0 and @xmath49 a representative of @xmath64 . let @xmath65 . then @xmath66 is one less than the number of connected components in the graph @xmath67 . [ cor : dim ] let @xmath1 be a vertex - supported effective divisor on a metric graph @xmath3 and @xmath68 . then the dimension of the cell complex @xmath0 is at most @xmath69 . in addition , if @xmath3 is @xmath70-connected , then the dimension of the cell complex @xmath0 is at most @xmath71 . let @xmath64 be a cell in @xmath0 and @xmath49 a representative of @xmath64 . then @xmath72 . so the support of @xmath49 contains at most @xmath69 points . by definition , @xmath73 , so @xmath74 contains at most @xmath69 points as well . now we consider the connected components in the graph @xmath67 . first @xmath3 is connected . and each time we delete a point in @xmath74 , the number of connected components can increase by at most @xmath75 , because the deleted point is always interior to an edge of the current graph . therefore there are at most @xmath76 connected components in the graph @xmath67 . hence , by proposition [ prop : dim ] , we conclude that @xmath77 . if @xmath3 is @xmath70-connected , then when we delete one point in @xmath74 from @xmath3 , the remaining graph is still connected , so there are at most @xmath69 connected components in the graph @xmath67 . hence @xmath78 . in order to find the cells in @xmath0 , we need to know whether there are finitely many of them . fortunately we have the following theorem . * theorem 14))[thm : fin ] the cell complex @xmath0 has finitely many vertices . since each cell is uniquely determined by its vertices , we have the following corollary . [ cor : fin ] the cell complex @xmath0 has finitely many cells . in order to find the cells given @xmath3 and @xmath1 , one approach is that we first find the vertices of @xmath0 , then find other cells of @xmath0 . as for the vertices , it is straightforward to implement the method in the proof of ( * ? ? ? * theorem 6 ) . however , there might be many general cells in @xmath0 with rather complicated structures , because there might be many parts in the partition on one edge in the data in definition [ def : cell ] . to grasp the general cells in @xmath0 , we introduce an important type of divisors and cells in @xmath0 that serve as landmarks . [ def : anc ] a divisor @xmath49 on @xmath3 is an _ anchor divisor _ if for each edge of @xmath3 there is at most one its interior point @xmath15 with @xmath79 . a cell @xmath64 in @xmath0 is an _ anchor cell _ if all representatives @xmath49 of @xmath64 are anchor divisors . the number of anchor divisors in @xmath0 could be infinite . for example if an anchor cell has dimension @xmath75 , then it contains infinitely many divisors and each such divisor is an anchor divisor . nonetheless , they all represent the same anchor cell because they share the same data in definition [ def : cell ] . one property of anchor cells is that they contain all vertices of @xmath0 . [ lem : ver ] each vertex of @xmath0 is an anchor cell . suppose that @xmath80 is not an anchor divisor . there exists an edge @xmath27 of @xmath3 and two distinct interior points @xmath81 of @xmath27 such that @xmath82 , then @xmath83 . after deleting @xmath84 and @xmath85 the graph is no longer connected . so the number of connected components in the graph @xmath67 is at least @xmath70 . hence , by proposition [ prop : dim ] , @xmath49 is the representative of a cell with dimension at least @xmath75 . therefore , if a divisor represents a vertex of @xmath0 , it must be an anchor divisor . so each vertex is an anchor cell . [ def : sa ] let @xmath86 be an anchor cell of @xmath0 . for each edge of @xmath3 we consider whether there is an ordered partition in the data of @xmath86 . if @xmath87 are the edges of @xmath3 such that in the data of @xmath86 there is an ordered partition @xmath88 on @xmath89 , then we define @xmath90 ; otherwise we define @xmath91 . the motivation for anchor cells is the following theorem . ( association to anchor cells)[thm : anc ] let @xmath1 be a vertex - supported effective divisor on a metric graph @xmath3 . ( 1 ) suppose @xmath92 is the set of all cells in @xmath0 and @xmath93 is the set of all anchor cells in @xmath0 . then there is a function @xmath94 such that for @xmath95 , let @xmath96 be a representative of @xmath64 . then @xmath97 is represented by an anchor divisor @xmath98 such that a. @xmath99 for all vertices @xmath58 of @xmath3 ; b. the corresponding rational functions of @xmath98 and @xmath49 have the same slopes at the endpoints of every edge of @xmath3 . \(2 ) let @xmath100 be the same map as in ( 1 ) . then for any anchor cell @xmath101 , we have @xmath102 furthermore , for @xmath103 , there are @xmath104 cells of dimension @xmath105 in @xmath106 . \(1 ) given @xmath1 and @xmath49 , we construct @xmath98 as follows . let @xmath25 such that @xmath54 . then we construct another rational function @xmath107 . first we let @xmath108 for vertices @xmath58 of @xmath3 ; next we let @xmath109 for all points @xmath110 , where @xmath27 is an edge of @xmath3 such that @xmath111 ; for other edges @xmath27 of @xmath3 , we regard the open edge @xmath27 as the interval @xmath112 . suppose @xmath4 has @xmath113 linear pieces @xmath114 $ ] on @xmath27 , with slope @xmath115 , where @xmath116 . then for @xmath117 , we have @xmath118 . since the adjacent linear pieces have different slopes , @xmath119 . while @xmath49 is effective , so @xmath120 . now we let @xmath121 have two linear pieces on @xmath27 , the one containing @xmath122 with slope @xmath123 , the one containing @xmath124 with slope @xmath125 . note that if @xmath126 then @xmath121 coincides with @xmath4 on @xmath27 . if @xmath127 , since @xmath128 , the graph of the two pieces of @xmath121 will intersect within the interval , so @xmath121 is well - defined on @xmath27 . in addition let @xmath129 be the intersection point of these two pieces , then @xmath130 . we let @xmath131 . by definition of @xmath121 , @xmath98 is effective and anchor , and @xmath98 satisfies both ( i ) and ( ii ) . \(2 ) note that for @xmath95 , the two cells @xmath64 and @xmath97 share the same first and third parts of their data , and given any edge @xmath27 of @xmath3 , either both do not have a partition , or they have the partitions of the same positive integer . the only difference is that the partition of @xmath97 is always trivial , while the partition of @xmath64 could be arbitrary . given a positive integer @xmath132 , there are @xmath133 ordered partitions of @xmath132 objects in a row , because each partition corresponds to a @xmath134-tuple of @xmath135 , indicating whether or not to break the @xmath134 pairs of adjacent objects . in addition , for @xmath136 , there are @xmath137 ordered partitions with @xmath138 parts . then if @xmath139 is an interior point on edge @xmath89 with @xmath140 , the partition on @xmath89 of a preimage of @xmath86 has @xmath141 choices . once all partitions are determined , so is the preimage . then @xmath142 furthermore , note that if the number of parts in the partition on one edge increases by @xmath75 , the number of connected components in proposition [ prop : dim ] also increases by @xmath75 , which means the dimension of the new cell is one more . then for @xmath103 , the number of cells with dimension @xmath105 in @xmath106 equals to the number of the @xmath113-tuples of partitions of @xmath143 with @xmath144 parts in total . this number is @xmath104 by an easy argument of generating functions . the following corollary provides a combinatorial formula to compute the @xmath4-vector of @xmath0 given all of its anchor cells . [ cor : fvec ] let @xmath1 be a vertex - supported effective divisor on a metric graph @xmath3 . for an anchor cell @xmath86 in @xmath0 , define @xmath145 as in theorem [ thm : anc ] . if @xmath146 are all anchor cells in @xmath0 , then for each @xmath147 the number of @xmath69-dimensional cells in the cell complex @xmath0 is the coefficient of @xmath148 in the generating function @xmath149 by theorem [ thm : anc](1 ) , each cell in @xmath0 is associated to a unique anchor cell of @xmath0 , i.e. @xmath150 . by theorem [ thm : anc](2 ) , the generating function for @xmath151 is @xmath152 . so the total generating function is just their sum . let @xmath3 have the skeleton of the complete bipartite graph @xmath153 and let @xmath29 be the canonical divisor on @xmath3 . if @xmath49 is the left divisor in figure [ fig : div ] , then @xmath49 has two chips on two edges respectively . then @xmath154 cells of @xmath155 are associated to @xmath49 . by theorem [ prop : dim ] , their dimensions are @xmath156 respectively . with the anchor cells , we have the following approach for finding the cells of @xmath0 . a. computing the anchor cells of @xmath0 ; [ enu : anc ] b. given the anchor cells of @xmath0 , computing the other cells of @xmath0 . [ enu : asso ] step ( [ enu : asso ] ) is done by corollary [ cor : fvec ] . we then explain implementations of step ( [ enu : anc ] ) . we first introduce some properties of anchor cells . [ lem:2piece ] let @xmath1 be a vertex - supported effective divisor on a metric graph @xmath3 . if @xmath25 and @xmath22 is an anchor divisor , then @xmath4 has at most two linear pieces on each edge of @xmath3 . for any edge @xmath27 of @xmath3 , suppose @xmath15 is an interior point of @xmath27 and it is the intersection of two linear pieces of @xmath4 with different slopes . then , by definition , we have @xmath157 . since @xmath1 is vertex - supported , @xmath158 . then @xmath159 . since @xmath25 , the divisor @xmath22 is effective , so @xmath160 . however @xmath22 is an anchor divisor , meaning that there is at most one such point @xmath15 , so @xmath4 has at most two linear pieces on @xmath27 . therefore if the outgoing slopes at the two endpoints of @xmath27 sum to @xmath55 , then @xmath4 is linear on @xmath27 ; otherwise @xmath4 has two linear pieces on @xmath27 and there is one interior point @xmath15 of @xmath27 such that @xmath160 . [ cor : anc ] let @xmath1 be a vertex - supported divisor on a metric graph @xmath3 . if @xmath64 is an anchor cell and it is represented by a divisor @xmath22 , then @xmath64 is uniquely determined by the outgoing slopes of @xmath4 at all vertices of @xmath3 . it suffices to show that given all those slopes , the data of @xmath64 are uniquely determined . firstly , the @xmath48 are determined given those slopes . secondly we show that the @xmath56 are also determined . if @xmath64 is an anchor cell and @xmath22 is a representative of @xmath64 , then @xmath22 is an anchor divisor . by lemma [ lem:2piece ] , @xmath4 has at most two linear pieces on each edge @xmath27 of @xmath3 . if @xmath4 is linear on @xmath27 , then the integer @xmath56 is zero for @xmath22 ; otherwise @xmath4 has two linear pieces on @xmath27 . suppose they have a common point @xmath58 . then @xmath161 is minus the sum of the two outgoing slopes at the endpoints of @xmath27 , so @xmath56 is still determined by those slopes . finally for each vertex @xmath58 of @xmath3 , @xmath162 . @xmath163 is known , and each @xmath164 , which is the sum of the outgoing slopes at v , is also determined , so @xmath45 is determined too . ( ( * ? ? ? * lemma 7))[lem : bd ] let @xmath1 be an effective divisor on a metric graph @xmath3 and @xmath25 . then the slopes of all linear pieces of @xmath4 are between @xmath165 and @xmath166 . by corollary [ cor : anc ] and lemma [ lem : bd ] , we can find all anchor cells by considering the @xmath167-tuple of outgoing slopes of all @xmath25 with at most two linear piece(s ) on each edge of @xmath3 . in particular we have a proof of the finiteness of anchor cells . next we implement this approach using linear programming algorithms . suppose @xmath57 , where @xmath168 is the set of vertices and @xmath169 is the set of edges . for @xmath170 , the edge @xmath171 has endpoints @xmath172 and @xmath173 , where @xmath174 , and the length of @xmath89 is @xmath175 . for @xmath176 let @xmath177 and @xmath178 . [ lem : ancfun ] let @xmath4 be a rational function defined on @xmath57 such that there are at most two linear pieces on each edge of @xmath3 . denote by @xmath179 for @xmath176 and by @xmath180 the outgoing slope at @xmath172 and @xmath173 of @xmath171 for @xmath170 . then @xmath25 if and only if * for each vertex @xmath181 we have the equations @xmath182 * for @xmath170 , either @xmath183 or @xmath184 suppose @xmath25 . note that for each @xmath176 we have @xmath185 , then @xmath186 for each @xmath170 , @xmath4 has either one or two linear piece(s ) on the edge @xmath171 . if there is one linear piece , then @xmath187 , and by definition of slope we have @xmath188 ; if there are two linear pieces , then there exists an interior point @xmath129 of @xmath171 such that @xmath4 is linear on both line segments @xmath189 and @xmath190 , and @xmath191 . note that @xmath192 , so @xmath193 . let @xmath15 be the length of the segment @xmath189 , then the length of the segment @xmath190 is @xmath194 . and by definition of slope we have @xmath195 since @xmath196 can be any real number , we eliminate it and get @xmath197 then @xmath198 since @xmath129 is an anterior point of @xmath171 , we have @xmath199 . then @xmath200 and @xmath201 . conversely , suppose for each @xmath176 ( [ eq : ver ] ) holds and for each @xmath170 , either ( [ eq:1pc ] ) or ( [ eq:2pc ] ) holds . for each @xmath170 , if @xmath187 then by ( [ eq:1pc ] ) @xmath4 is well - defined on @xmath171 and @xmath202 for all points @xmath203 in the interior of @xmath171 ; otherwise ( [ eq:2pc ] ) holds , by the chain inequalities @xmath4 is well - defined on @xmath171 with two linear pieces . since @xmath193 , we have @xmath202 for all points @xmath203 in the interior of @xmath171 . finally for each @xmath176 , ( [ eq : ver ] ) holds . then @xmath204 . so @xmath4 is a well - defined rational function on @xmath3 with at most two linear pieces on each edge of @xmath3 and @xmath25 . note that if @xmath22 is an anchor divisor in @xmath0 , then it also has degree @xmath69 and its support has at most one intersection point with each open @xmath89 . thus we obtain a partition of @xmath69 into @xmath205 nonnegative integers : @xmath206 , where @xmath207 and @xmath208 . we call them as _ configurations _ of anchor divisors . note that there are @xmath209 configurations in total . [ cor : cfg ] suppose @xmath25 with the same notations as in lemma [ lem : ancfun ] . let @xmath206 be the configuration of the divisor @xmath22 , then for @xmath170 , we have @xmath210 and for @xmath176 we have @xmath211 now for each configuration @xmath206 , we consider the following system of linear constraints in ( [ eq:1pc ] ) , ( [ eq:2pc ] ) , ( [ eq : cfge ] ) , ( [ eq : cfgv ] ) . here @xmath212 are parameters and @xmath213 are variables . then among the solutions of this system , each @xmath214-tuple of integers @xmath180 gives an anchor cell of @xmath0 . for each @xmath215 , whether to apply ( [ eq:1pc ] ) or ( [ eq:2pc ] ) depends on the value of @xmath216 . lemma [ lem : ancfun ] and corollary [ cor : cfg ] gives rise to the following algorithm . the input is the metric graph @xmath57 , the edge lengths @xmath217 and the divisor @xmath1 . the output @xmath49 is a list of the anchor cells . @xmath218 @xmath219 @xmath220 @xmath221 @xmath222 @xmath223 the system of linear constraints in ( [ eq:1pc ] ) , ( [ eq:2pc ] ) , ( [ eq : cfge ] ) , ( [ eq : cfgv ] ) ( * ) next @xmath38 @xmath224 array of @xmath225 entries @xmath226 array of @xmath225 entries @xmath227 \gets [ minimize(s_{j,1},s),maximize(s_{j,1},s ) ] \cap \mathbb{z}$ ] @xmath228 \gets [ minimize(s_{j,2},s),maximize(s_{j,2},s ) ] \cap \mathbb{z}$ ] @xmath229 the cartesian product of @xmath230,\ldots , a[m],b[1],\ldots , b[m]$ ] @xmath231\}$ ] * return * @xmath49 \(1 ) we use integer programming method for step ( * ) . in maple 2015 there is a command lpsolve that is able to do it . + ( 2 ) in this algorithm we do not require the metric graph to be simple . we allow both loops and parallel edges . we can also deal with more general input . suppose the skeleton of @xmath3 is fixed , but the metric may vary . in this case the input is the metric @xmath232 , and the desired output is the set of anchor cells in @xmath0 . apparently we can apply the above approach once the @xmath217 are given , but since algorithm [ alg : anc ] needs to run the linear - programming subroutine @xmath209 times , it is not very efficient if we would like to compute for many different metrics . instead , we can also view the @xmath217 s as variables . however @xmath233 appears in ( [ eq:2pc ] ) . in order to make the system of constraints linear , we have to let @xmath180 become parameters . so we take the approach in the proof of proposition [ thm : fin ] . we still have the equations and inequalities ( [ eq:1pc ] ) , ( [ eq:2pc ] ) , ( [ eq : cfge ] ) , ( [ eq : cfgv ] ) , but @xmath180 are parameters instead . now the parameters are @xmath234 and the variables are @xmath235 . then we also need @xmath236 for @xmath22 being effective and @xmath237 for the edge lengths in @xmath26 are all positive . by lemma [ lem : bd ] , we have @xmath238 . so there are finitely many possible values of them . now we have an empty list first and for each choice of @xmath239 , we check the feasibility of the system of linear constraints formed by ( [ eq:1pc ] ) , ( [ eq:2pc ] ) , ( [ eq : cfge ] ) , ( [ eq : cfgv ] ) , ( [ ein : eff ] ) , ( [ ein : pos ] ) . if it is feasible , then we find one anchor cell and save it to our list . the output is the set of anchor cells represented by divisors @xmath49 such that there exists some metric @xmath240 with @xmath241 . the advantage of this approach is that we can compute this list of all possible anchor cells in @xmath0 beforehand . then given a specific metric , we just plug in the values of @xmath217 and check the feasibility of each anchor cell in the list . in this section we present the properties of @xmath2 . if @xmath4 and @xmath121 are rational functions defined on @xmath3 , then @xmath242 is the rational function on @xmath3 with @xmath243 and @xmath244 is the rational function on @xmath3 with @xmath245 . in other words , @xmath246 and @xmath247 are the _ tropical _ operations in the set of rational functions on @xmath3 . here we choose the max - plus algebra . ( ( * ? ? ? * lemma 4 ) ) let @xmath1 be any divisor on a metric graph @xmath3 . the space @xmath2 is a tropical semi - module , i.e. it is closed under tropical addition and tropical scalar multiplication . we would like to find a minimal set of generators of the tropical semi - module @xmath2 . we use the notion of _ chip - firing _ @xcite . for an effective divisor @xmath1 on @xmath3 we regard it as a distribution of @xmath166 chips on @xmath3 : there are @xmath248 chips at each point @xmath9 . suppose that @xmath1 is effective and @xmath25 . for each linear piece @xmath249 of @xmath3 , if @xmath4 has slope @xmath250 from @xmath84 to @xmath85 , then we say that @xmath84 fires @xmath251 chips to @xmath85 when adding @xmath18 to @xmath1 . for a metric graph @xmath3 , a _ subgraph _ is a compact subset with a finite number of components . fix an effective divisor @xmath1 on @xmath3 . we say that a subgraph @xmath252 of @xmath3 can _ fire _ for @xmath1 if , for each boundary point @xmath15 of @xmath253 , the number of edges pointing out of @xmath252 is no greater than @xmath248 . a function @xmath25 is called _ extremal _ if for any @xmath254 , the decomposition @xmath255 implies that @xmath256 . any generating set of @xmath2 must contain all extremal generators up to tropical scalar multiplication . ( ( * ? ? ? * theorem 14(a)))[lem : ext ] if @xmath1 is a vertex - supported divisor on a metric graph @xmath3 and @xmath25 is extremal , then @xmath22 is a vertex of the cell complex @xmath0 . by lemma [ lem : ext ] in order to find the extremal generators of @xmath2 , it suffices to search among the vertices of @xmath0 . the next lemma is an important criterion for the extremal generators of @xmath2 . ( ( * ? ? ? * lemma 5))[lem : fire ] let @xmath1 be any divisor on a metric graph @xmath3 . then @xmath25 is extremal if and only if there do not exist two proper subgraphs @xmath257 and @xmath258 of @xmath3 such that they cover @xmath3 and both can fire on @xmath22 . for a non - extremal @xmath25 , the proper subgraphs @xmath257 and @xmath258 may not be obvious . see the example in figure [ fig : div ] , where @xmath3 is the metric graph with skeleton @xmath153 and all - equal metric , @xmath259 is the canonical divisor on @xmath3 . ( 0,0 ) ( 0,6 ) ( 3,0 ) (6,6 ) ( 0,0 ) (6,6 ) ; ( 3,6 ) ( 3,0 ) ; ( 0,6 ) ( 6,0 ) ; ( 0,3 ) circle ( 3pt ) ; ( 3,0 ) circle ( 3pt ) ; ( 3,6 ) circle ( 3pt ) ; ( 6,3 ) circle ( 3pt ) ; at ( 0,3 ) 2 ; at ( 6,3 ) 2 ; at ( 3,0 ) 1 ; at ( 3,6 ) 1 ; ( 0,3 ) ( 0,6 ) ( 3,0 ) (3,6 ) ( 6,3 ) ; ( 0,6 ) ( 6,0 ) ; ( 0,3 ) ( 0,1.8 ) ; ( 6,3 ) ( 6,4.2 ) ; ( 3,0 ) ( 3.5,1 ) ; ( 3,6 ) ( 2.5,5 ) ; ( 0,1.8 ) ( 2.5,5 ) ; ( 6,4.2 ) ( 3.5,1 ) ; ( 0,0 ) ( 6,6 ) ; ( 0,3 ) circle ( 3pt ) ; ( 3,0 ) circle ( 3pt ) ; ( 3,6 ) circle ( 3pt ) ; ( 6,3 ) circle ( 3pt ) ; at ( 0,3 ) 2 ; at ( 6,3 ) 2 ; at ( 3,0 ) 1 ; at ( 3,6 ) 1 ; ( 0,3 ) ( 0,0 ) ( 3,6 ) (3,0 ) ( 6,3 ) ; ( 0,0 ) ( 6,6 ) ; ( 0,3 ) ( 0,4.2 ) ; ( 6,3 ) ( 6,1.8 ) ; ( 3,0 ) ( 2.5,1 ) ; ( 3,6 ) ( 3.5,5 ) ; ( 0,4.2 ) ( 2.5,1 ) ; ( 6,1.8 ) ( 6,0 ) ( 3.5,5 ) ; ( 0,6 ) ( 6,0 ) ; ( 0,3 ) circle ( 3pt ) ; ( 3,0 ) circle ( 3pt ) ; ( 3,6 ) circle ( 3pt ) ; ( 6,3 ) circle ( 3pt ) ; at ( 0,3 ) 2 ; at ( 6,3 ) 2 ; at ( 3,0 ) 1 ; at ( 3,6 ) 1 ; our discussion suggests the following approach for computing all extremal generators of @xmath2 : a. find all anchor divisors in @xmath2 . b. filter them by theorem [ prop : dim ] to get all vertices of @xmath0 . c. filter the vertices by lemma [ lem : fire ] to get all extremal generators of @xmath2.[enu : ext ] we introduce a method for step ( [ enu : ext ] ) . given an effective divisor @xmath54 on @xmath3 , we apply lemma [ lem : fire ] to check whether @xmath4 is extremal . note that if a proper subgraph can fire for @xmath49 , then its boundary is contained in @xmath260 . so we partition @xmath3 into connected subgraphs whose boundaries are contained in @xmath260 and we call them components . since @xmath260 is finite , so is the number of such components . we conclude that any subgraph that can fire for @xmath49 is a union of these components . for each such union , we ignore it if it is non - proper or can not fire for @xmath49 , then we have a finite list of all proper subgraphs of @xmath3 that can fire for @xmath49 . finally we check whether there is a pair in the list that covers @xmath3 . in this section we apply our methods to some nontrivial examples . a metric graph is _ trivalent _ if the degree of every vertex is @xmath59 . trivalent graphs appear in the berkovich skeleton of many curves ( * ? ? ? * example 5.29 ) . this fact motivates us to compute the examples of trivalent metric graphs and the canonical divisor @xmath29 . for the computations we performed below , our hardware is a laptop with intel core i5 - 6200u processor ( 2.3 ghz ) and 8 gb ram . the software is maple , version 2015 . all computations were single - threaded . the complete graph @xmath262 has @xmath263 vertices and its genus is @xmath59 . table [ tab : k4 ] shows the structure of @xmath264 and @xmath265 on @xmath261 given a particular metric @xmath26 . for @xmath266 let @xmath267 be the edge length between vertex @xmath138 and @xmath215 . a metric @xmath26 is given by @xmath268 . [ cols="^ " , ] + all - equal & 370 & 33 & @xmath269 & @xmath270 + & 460 & 63 & @xmath271 & @xmath272 + & 730 & 84 & @xmath273 & @xmath274 + [ tab : k33 ] [ rem : char ] by ( * ? ? ? * corollary 31 ) , the cell complex @xmath0 is contractible as a topological space . thus the euler characteristic of @xmath0 is always @xmath75 , which coincides with all the @xmath4-vectors we computed above . here are some open problems for further research on this topic . * fixing the skeleton of @xmath3 and @xmath1 , find the polyhedral cone decomposition of @xmath275 based on the cell structure of @xmath0 . in particular , determine the metrics @xmath26 such that @xmath33 has dimension @xmath276 . * given all the cells in @xmath0 , find the face lattice of @xmath0 . in general , if we apply chip - firing to a representative of a cell , we may obtain a representative of a cell on its boundary , but the rigorous argument is yet to be established . * given the skeleton of @xmath3 and @xmath1 , find non - trivial upper and lower bounds of the number of anchor cells ( or cells , vertices ) in @xmath0 and of the number of extremal generators in @xmath2 . 10 m. baker . specialization of linear systems from curves to graphs . , 2(6):613653 , 2008 . m. baker and x. faber . metrized graphs , laplacian operators , and electrical networks . , 415:1534 , 2006 . m. baker and s. norine . riemann - roch and abel - jacobi theory on a finite graph . , 215(2):766788 , 2007 . m. baker , s. payne , and j. rabinoff . non - archimedean geometry , tropicalization , and metrics on curves . , 3(1):63105 , 2016 . a. bjrner , l. lovsz , and p. w. shor . chip - firing games on graphs . , 12(4):283291 , 1991 . s. brodsky , m. joswig , r. morrison , and b. sturmfels . moduli of tropical plane curves . , 2(1):1 , 2015 . a. gathmann and m. kerber . a riemann roch theorem in tropical geometry . , 259(1):217230 , 2008 . p. griffiths and j. harris . . john wiley & sons , 2014 . c. haase , g. musiker , and j. yu . linear systems on tropical curves . , 270(3 - 4):11111140 , 2012 . d. maclagan and b. sturmfels . , volume 161 . american mathematical soc . , g. mikhalkin and i. zharkov . tropical curves , their jacobians and theta functions . , 465:203230 , 2008 . the author is grateful to bernd sturmfels for his guidance and encouragement throughout this project . the author thanks madhusudan manjunath , ralph morrison and kristian ranestad for helpful discussions and comments . the author thanks yang an , matt baker and josephine yu for their precious suggestions . the author also thanks two anonymous referees for their critical comments on an earlier version of this work .	the linear system @xmath0 of a divisor @xmath1 on a metric graph has the structure of a cell complex . we introduce the anchor divisors and anchor cells in it - they serve as the landmarks for us to compute the f - vector of the complex and find all cells in the complex . a linear system can also be identified as a tropical convex hull of rational functions . we compute its extremal generators using the landmarks . we apply these methods to some examples - namely the canonical linear systems of some small trivalent graphs .
energy efficiency is of paramount importance for future communication networks and is a main design target for all 5 g radio access solutions . it refers to an efficient utilization of the available energy and consequently extends the network lifetime and/or reduces the operation cost . specifically , conventional battery - powered communication systems suffer from short lifetime and require periodic replacement or recharging in order to maintain network connectivity . on the other hand , communication systems that are supported by a continuous power supply such as cellular networks require a power grid infrastructure and may result in large energy consumption that will further increase due to the increasing growth of data traffic . the investigation of energy - aware architectures as well as transmission techniques / protocols that prolong the lifetime of the networks or provide significant energy savings has been a hot research area over several years , often under the umbrella of the green radio / communications @xcite . due to the limited supply of non - renewable energy resources , recently , there is a lot of interest to integrate the energy harvesting ( eh ) technology to power communication networks @xcite . energy harvesting is a new paradigm and allows nodes to harvest energy from natural resources ( i.e. , solar energy , wind , mechanical vibrations etc . ) in order to maintain their operation . related literature concerns the optimization of different network utility functions under various assumptions on the knowledge of the energy profiles . the works in @xcite assume that the eh profile is perfectly known at the transmitters and investigate optimal resource allocation techniques for different objective functions and network configurations . on the other hand , the works in @xcite adopt a more networking point of view and maximize the performance in terms of stability region by assuming only statistical knowledge of the eh profile . although energy harvesting from natural resources is a promising technology towards fully autonomous and self - sustainable communication networks , it is mainly unstable ( i.e. , weather - dependent ) and thus less efficient for applications with critical quality - of - service ( qos ) requirements . an interesting solution that overcomes the above limitation is to harvest energy from man - made electromagnetic radiation . despite the pioneering work of tesla , who experimentally demonstrated wireless energy transfer ( wet ) in late 19th century , modern wireless communication systems mainly focus on the information content of the radio - frequency ( rf ) radiation , neglecting the energy transported by the signals . recently , there is a lot of interest to exploit rf radiation from energy harvesting perspective and use wireless energy transfer in order to power communication devices . the fundamental block for the implementation of this technology is the rectifying - antenna ( rectenna ) which is a diode - based circuit that converts the rf signals to dc voltage . several rectenna architectures and designs have been proposed in the literature for different systems and frequency bands @xcite . an interesting rectenna architecture is reported in @xcite , where the authors study a rectenna array in order to further boost the harvesting efficiency . although information theoretic studies ideally assume that a receiver is able to decode information and harvest energy independently from the same signal @xcite , this approach is not feasible due to practical limitations . in the seminal work in @xcite , the authors introduce two practical rf energy harvesting mechanisms for `` simultaneous '' information and energy transfer : a ) time switching ( ts ) where dedicated time slots are used either for information transfer or energy harvesting , b ) power splitting ( ps ) where one part of the received signal is used for information decoding , while the other part is used for rf energy harvesting . the employment of the above two practical approaches in different fundamental network structures , is a hot research topic and several recent works appear in the literature . in @xcite , the authors study the problem of beamforming design for a point - to - point multiple - input multiple - output ( mimo ) channel and characterize the rate - energy region for both ts and ps techniques . this work is extended in @xcite for the case of an imperfect channel information at the transmitter by using robust optimization tools . the work in @xcite investigates the optimal ps rule for a single - input single - output ( siso ) channel in order to achieve different trade - offs between ergodic capacity and average harvested energy . an interesting problem is discussed in @xcite , where the downlink of an access point broadcasts energy to several users , which then use the harvested energy for time - division multiple access ( tdma ) uplink transmissions . in @xcite , the authors study a fundamental multi - user multiple - input single - output ( miso ) channel where the single - antenna receivers are characterized by both qos and ps - eh constraints . on the other hand , cooperative / relay networks is a promising application area for rf energy harvesting , since relay nodes have mainly limited energy reserves and may require external energy assistance . the works in @xcite deal with the integration of both ts and ps techniques in various cooperative topologies with / without batteries for energy storage . the simultaneous information / energy transfer for a mimo relay channel with a separated energy harvesting receiver is discussed in @xcite . although several studies deal with the analysis of communication networks with rf energy harvesting capabilities , most of existing work refers to specific ( fixed ) single / multiple user network configurations . since harvesting efficiency is associated with the interference and thus the geometric distance between nodes , a fundamental question is to study rf energy harvesting for large - scale networks by taking into account random node locations . stochastic - geometry is a useful theoretical tool in order to model the geometric characteristics of a large - scale network and derive its statistical properties @xcite . several works in the literature adopt stochastic - geometry in order to analyze the outage probability performance or the transmission capacity for different conventional ( without harvesting capabilities ) networks e.g. , @xcite . large - scale networks with energy harvesting capabilities are studied in @xcite for different network topologies and performance metrics . these works model the energy harvesting operation as a stochastic process and mainly refer to energy harvesting from natural resources e.g. , solar , wind , etc . however , few studies analyze the behavior of a rf energy harvesting network from a stochastic - geometry standpoint . in @xcite , the authors study the interaction between primary and cognitive radio networks , where cognitive radio nodes can harvest energy from the primary transmissions , by modeling node locations as poisson point processes ( ppps ) . a cooperative network with multiple transmitter - receiver pairs and a single energy harvesting relay is studied in @xcite by taking the spatial randomness of user locations into consideration . the analysis of large - scale rf energy harvesting networks with practical ts / ps techniques , is an open question in the literature . in this paper , we study the performance of a large - scale network with multiple transmitter - receiver pairs , where transmitters are connected to the power grid , while receivers employ the ps technique for rf energy harvesting . by using stochastic - geometry , we model the randomness of node locations and we analyze the fundamental trade - off between outage probability performance and average harvested energy . specifically , we study two main protocols : a ) a non - cooperative protocol , and b ) a cooperative protocol with orthogonal relay assistance . in the non - cooperative protocol , each transmitter simultaneously transfers information and energy at the associated receiver via the direct link . the outage probability of the system as well as the average harvested energy are derived in closed form in function of the power splitting ratio . in addition , an optimization problem which minimizes the transmitted power under some well - defined outage probability and average harvesting constraints , is discussed and closed form solutions are provided . the cooperative protocol is introduced to show that relaying can significantly improve the performance of the system and achieve a better trade - off between outage probability performance and energy harvesting transfer . relaying cooperation is integrated in several systems and standards in order to provide different levels of assistance ( i.e. , cooperative diversity , energy savings , secrecy etc ) ; in this work , relays are used in order to facilitate the information / energy transfer . for the cooperative protocol , we introduce a set of potential dynamic - and - forward ( df ) relays , which are randomly distributed in the network according to a ppp ; these relays have similar characteristics with the transmitters and are also connected to the power grid . in this case , information and energy can be received at each destination via two independent paths ( in case of cooperation ) . for the relay selection , we adopt a random selection policy based on a sectorized selection area with central angle at the direction of each receiver . the outage performance of the system for a selection combining ( sc ) scheme as well as the average harvested energy are analyzed in closed form and validate the cooperative diversity benefits . numerical results for different parameter set - up reveal some important observations about the impact of the central angle and relay density on the trade - off between information and energy transfer . it is the first time , to the best of the authors knowledge , that stochastic - geometry is used in order to analyze a ps energy harvesting network with / without relaying . the remainder of this paper is organized as follows . section [ system ] describes the system model and introduces the considered performance / harvesting metrics . section [ direct_link ] presents the non - cooperative protocol and analyzes its performance in terms of outage probability and average harvested energy . section [ system2 ] introduces the cooperative protocol and analyzes both performance metrics considered . simulation results are presented in section [ num ] , followed by our conclusions in section [ conc ] . @xmath0 denotes the @xmath1-dimensional euclidean space , @xmath2 denotes the indicator function , @xmath3 is the lebesgue measure , @xmath4 denotes a two dimensional disk of radius @xmath5 centered at @xmath6 , @xmath7 denotes the euclidean norm of @xmath8 , @xmath9 denotes the probability of the event @xmath10 and @xmath11 represents the expectation operator . in addition , the number of points in @xmath12 is denoted by @xmath13 . we consider a 2-d large - scale wireless network consisting of a random number of transmitter - receiver pairs . the transmitters form an independent homogeneous ppp @xmath14 with @xmath15 of intensity @xmath16 on the plane @xmath17 , where @xmath18 denotes the coordinates of the node @xmath19 . each transmitter @xmath18 has a unique receiver @xmath20 ( not a part of @xmath21 ) at an euclidean distance @xmath22 in some random direction @xcite . all nodes are equipped with single antennas and have equivalent characteristics and computation capabilities . the time is considered to be slotted and in each time slot all the sources are active without any coordination or scheduling process . in the considered topology , we add a transmitter @xmath23 at the origin @xmath24 $ ] and its associated receiver @xmath25 at the location @xmath26 $ ] without loss of generality ; in this paper , we analyze the performance of this typical communication link but our results hold for any node in the process @xmath27 according to slivnyak s theorem @xcite . we assume a _ partial fading _ channel model , where desired direct links are subject to both small - scale fading and large - scale path loss , while interference links are dominated by the path - loss effects . according to the literature @xcite , this channel model is denoted as `` 1/0 fading '' and serves as a useful guideline for more practical configurations e.g. , all links are subject to fading @xcite . more specifically , the fading between @xmath18 and @xmath20 is rayleigh distributed so the power of the channel fading is an exponential random variable with unit variance . the path - loss model assumes that the received power is proportional to @xmath28 where @xmath29 is the euclidean distance between the transmitter @xmath30 and the receiver @xmath31 , @xmath32 denotes the path - loss exponent and we define @xmath33 . the euclidean distance between two nodes is defined as @xmath34 where the parameter @xmath35 refers to the minimum possible path - loss degradation and ensures the accuracy of our path - loss model for short distances @xcite . the instantaneous fading channels are known only at the receivers in order to perform coherent detection . in addition , all wireless links exhibit additive white gaussian noise ( awgn ) with variance @xmath36 . the transmitters are continuously connected to a power supply ( e.g. , battery or power grid ) and transmit with the same power @xmath37 . on the other hand , each receiver has rf energy harvesting capabilities and can harvest energy from the received electromagnetic radiation . the rf energy harvesting process is based on the ps technique and therefore each receiver splits its received signal in two parts a ) one part is converted to a baseband signal for further signal processing and data detection ( information decoding ) and b ) the other part is driven to the rectenna for conversion to dc voltage and energy storage . let @xmath38 denote the power splitting parameter for each receiver ; this means that @xmath39 of the received power is used for data detection while the remaining amount is the input to the rf - eh circuitry . we assume an ideal power splitter at each receiver without power loss or noise degradation , and that the receivers can perfectly synchronize their operations with the transmitters based on a given power splitting ratio @xmath40 @xcite . during the baseband conversion phase , additional circuit noise , @xmath41 , is present due to phase - offsets and circuits non - linearities and which is modeled as awgn with zero mean and variance @xmath42 @xcite . based on the ps technique considered , the signal - to - interference - plus - noise ratio ( sinr ) at the typical receiver can be written as @xmath43 where @xmath44 denotes the total ( normalized ) interference at the typical receiver with @xmath45 and @xmath46 denotes the channel power gain for the link @xmath47 . a successful decoding requires that the received sinr is at least equal to a detection threshold @xmath48 . on the other hand , rf energy harvesting is a long term operation and is expressed in terms of average harvested energy @xcite . since @xmath49 of the received energy is used for rectification , the average energy harvesting at the typical receiver is expressed as @xmath50 \bigg ) , \label{mean}\end{aligned}\ ] ] where @xmath51 $ ] denotes the conversion efficiency from rf signal to dc voltage ; for the convenience of analysis , it is assumed that @xmath52 . it is worth noting that the rf energy harvesting from the awgn noise is considered to be negligible . the first investigated scheme does not enable any cooperation between the nodes and thus communication is performed in a single time slot ; all the sources simultaneously transmit towards their associated receivers . [ model1 ] schematically presents the network topology for the non - cooperative case . the information decoding process is mainly characterized by the outage probability which denotes the probability that the instantaneous sinr is lower than the predefined threshold @xmath48 . by characterizing the outage probability for the typical transmitter - receiver link @xmath53 , we also characterize the outage probability for whole network ( @xmath54 ) . is the euclidean distance between a transmitter and its associated receiver and @xmath55 is the minimum path - loss distance . ] [ prop1 ] the outage probability for the non - cooperative protocol is given by @xmath56 where @xmath57 \bigg ) \nonumber \\ & \times\exp \left(-2 \pi \omega x y^{\alpha } z^{2-\alpha } \right).\end{aligned}\ ] ] see appendix [ app1 ] . for high transmitted powers i.e. , @xmath58 and @xmath59 the system becomes interference limited and the outage probability converges to a constant error floor given by @xmath60 as for the average harvested energy , by expanding we have the following proposition : [ prop2 ] the average harvested energy for the non - cooperative protocol is given by @xmath61 , \label{prop2e}\end{aligned}\ ] ] where @xmath62 from , we have : @xmath63 \bigg ) \nonumber \\ & = ( 1-\nu_d)p_t \bigg ( d_0^{-\alpha } \mathbb{e}(h_0)+\mathbb{e}(i_0)\bigg ) \nonumber \\ & = ( 1-\nu_d)p_t \big [ d_0^{-\alpha}+\psi(\lambda ) \big],\end{aligned}\ ] ] where @xmath64 for all @xmath19 , and the proof of @xmath65 can be found in appendix [ app2 ] . an interesting optimization problem is formulated when energy becomes a critical issue for the network and each receiver is characterized by both qos and rf energy harvesting constraints . due to the symmetry of the nodes , the minimization of the transmitted power for the typical transmitter , it also minimizes the total energy consumption for whole network . the optimization problem considered can be written as @xmath66 where the qos constraint ensures an outage probability lower than a threshold @xmath67 , while the rf energy harvesting constraint requires an average harvested energy at least equal to @xmath68 ( i.e. , it represents the minimum required energy to maintains operability at each device ) . for the case where the power splitting ratio is constant i.e. , @xmath69 , the solution to the optimization problem in is simplified as follows : @xmath70 \;\text{if}\ ; \pi_{\text{nc}}^{\infty}\leq c_i \\ \text{infeasible},\;\;\text{elsewhere } \end{array } \right.\end{aligned}\ ] ] where @xmath71 , @xmath72 $ ] , @xmath73 and @xmath74 . the asymptotic expression in is involved in the optimization problem and determines its feasibility . more specifically , if the outage probability floor in is higher than the outage probability constraint @xmath67 , there is not any transmitted power that can satisfy @xmath67 and therefore the optimization problem becomes infeasible . the constant power splitting case corresponds to a low implementation complexity and is appropriate for ( legacy ) systems where the rectenna s design is predefined and the power splitting parameter is not adaptable . for the general case , where both @xmath37 and @xmath40 are adjustable , it can be easily seen that the two main constraints are binding at the solution can further be reduced . by examining the cases where one constraint is binding and the other holds with inequality , we show that at the optimal solution the inequality constraint holds with equality . ] . in this case , the optimization problem is transformed to the solution of a standard quadratic equation and for @xmath75 the solution is given by @xmath76 we note that the optimization problem is infeasible for @xmath77 . the general optimization problem requires adaptive and dynamic rf power splitting and therefore refers to a higher implementation complexity . the implementation problem can be solved either by a central controller or in a distributed fashion . in the first case , a central unit that controls the network , solves the problem and broadcasts the common solution ( transmitted power , power splitting ratio ) to all nodes ; the transmitters and the receivers adjust their transmitted power and the power splitting ratio , respectively . in the distributed implementation , each node can locally solve the optimization problem without requiring external signaling ( but with the cost of a higher computational complexity ) . the optimization problem involves only deterministic and average system parameters such as geometric distances , network density , path - loss exponent and channel statistics ; these parameters are estimated at the beginning of the communication and remain constant for a long operation time . is the radius of the selection sector . ] the cooperative scheme exploits the relaying / cooperative concept in order to combat fading and path - loss degradation effects . the network topology considered is modified by adding a group of single - antenna df relays , which have not their own traffic and are dedicated to assist the transmitters . [ model2 ] schematically presents the network topology for the cooperative protocol . the location of all relay nodes are modeled as a homogeneous ppp denoted by @xmath78 with density @xmath79 ; this assumption refers to mobile relays where their position as well as their `` availability '' changes with the time . the relay nodes are also continuously connected to a power supply ( e.g. , battery ) and have equivalent computation / energy capabilities . we adopt an orthogonal relaying protocol where cooperation is performed in two orthogonal time slots @xcite . it is worth noting that although several cooperative schemes have been proposed in the literature e.g. , @xcite , the orthogonal relaying protocol has a low complexity and is sufficient for the purposes of this work . the cooperative protocol operates as follows : 1 . the first phase of the protocol is similar to the non - cooperative scheme and thus all transmitters simultaneously broadcast their signals towards the associated receivers . each transmitter @xmath18 defines a 2-d relaying area @xmath80 around its location and each relay node located inside this area is dedicated to assist this transmitter ; this means that all relays @xmath81 consider the signal generated by @xmath18 as a useful information and all the other signals as interference . in accordance to the general system model , we assume that direct links suffer from both small - scale fading and path - loss , while interference links are dominated by the path - loss attenuation ( 1/0 partial fading @xcite ) . by focusing our study on the typical transmitter @xmath23 , we define as @xmath82 the power fading gain for the link @xmath83 with @xmath84 . in this case , the direct link @xmath53 is characterized by , , while the sinr at the relay @xmath85 is written as + @xmath86 + where @xmath87 denotes the total ( normalized ) interference received at @xmath85 . if the relay node @xmath85 can decode the transmitted signal , which means that @xmath88 , it becomes a member of the transmitter s potential relay . it is worth noting that the relay nodes use all the received signal for information decoding , since they have not energy harvesting requirements . 2 . in the second phase of the protocol , one relay node ( if any ) that successfully decoded the transmitted signal accesses the channel and retransmits the source s signal . we assume a _ random selection _ process which selects a single relay out of all potential relays with equal probability . the random relay selection does not require any instantaneous channel feedback or any instantaneous knowledge of the geometry and is appropriate for low complexity implementations with strict energy constraints @xcite . more sophisticated relay selection policies , which take into account the instantaneous channel conditions @xcite , can also be considered in order to further improve the cooperative benefits . we define as @xmath89 the selected relay for the typical transmitter @xmath23 , @xmath90 is the channel power gain for the link @xmath91 and @xmath92 is the homogeneous ppp that contains all the selected relays for whole network . if the potential relay set is empty for a specific transmitter ( no relay was able to decode the source s transmitted signal ) , its message is not transmitted during the second phase of the protocol and therefore does not enjoy cooperative diversity benefits . the relaying link for the ( typical ) receiver is characterized by the following equations + @xmath93 \bigg ) , \label{mean2}\end{aligned}\ ] ] + where @xmath94 denotes the total ( normalized ) interference at the receiver @xmath20 , @xmath95 is the power splitting ratio used in the second phase of the protocol , @xmath96 is the transmitted power for each active relay and the binary variable @xmath97 is equal to one in case of a relaying transmission , while it takes the value zero when the relay set is empty . a perfect synchronization between the selected relay and the associated receiver is assumed for a given power splitting ratio @xmath98 . as for the decoding process at the receivers , we assume that the two copies of the transmitted signal are combined with a simple sc technique ; this means that information decoding is based on the best path between direct / relaying links @xcite . it is well known that sc only requires relative sinr measurements and thus it is simpler than maximum ratio combiner , which requires exact knowledge of the channel state information for each diversity branch . in addition , sc significantly reduces power consumption because continuous estimates of the channel state information are not necessary ; this is beneficial for the considered rf energy harvesting system , where energy saving is a critical requirement e.g. , @xcite . regarding the potential use of the non - selected branch for rf energy harvesting , here , we assume a simple / conventional implementation and the received energy allocated to the sc can not be used for rf energy harvesting purposes . on the other hand , the energy harvesting process exploits both transmission phases and the total average energy harvested becomes equal to @xmath99 the considered random relay selection process does not require any instantaneous channel feedback and is appropriate for scenarios with critical energy / computation constraints . however , the definition of the selection area @xmath80 has a significant impact on the system performance . in this work , we assume that @xmath80 is a circular sector and @xmath100 , circular discs around the different transmitters do not overlap at most of the time . in our case , we have circular sectors and therefore the probability of overlapping becomes much lower . ] with center @xmath18 , radius @xmath101 and central angle with orientation at the direction of the receiver @xmath20 . by appropriately adjusting the central angle of the sector , we can ensure that the relaying paths are shorter than the direct distance @xmath22 ; this parameterization avoids scenarios where the selected relay experiences more serious path - loss effects than the direct link . it is proven in appendix [ angle_sec ] that a selection area @xmath102,\ ; \theta \in [ -\theta_0\;\theta_0]\big\}$ ] with @xmath103 satisfies this requirement . for the cooperative protocol , an outage event occurs when ( a ) the direct link is in outage and no relay is able to decode the source s message or ( b ) a relay node is able to decode the transmitted signal but both direct and relaying link are in outage . based on these two cases , we have the following proposition : the outage probability for the cooperative protocol is given by @xmath104 where @xmath105 \right ) , \nonumber \\ & \rightarrow \underbrace{\exp\left(-\lambda_r\left[\int_{-\theta_0}^{+\theta_0}\int_{r_0}^{\eta } \xi(\lambda , r , r_0)r dr d\theta+\theta_0 r_0 ^ 2 \xi(\lambda , r_0,r_0 ) \right ] \right)}_{\pi_c^{\infty},\;\text{for}\ ; p_t\rightarrow \infty } , \label{prob_fst } \\ & \pi_r(\nu_r , p_r)=1-\frac{1}{\theta_0(\eta^2-r_0 ^ 2)}\int_{-\theta_0}^{\theta_0}\int_{r_0}^{\eta}\exp\left(-\frac{\sigma^2 \omega c^{\alpha}}{p_r } \right ) \nonumber \\ & \;\;\;\;\;\;\;\times \exp\left(-\frac{\sigma_c^2 \omega c^{\alpha}}{\nu_r p_r } \right ) \xi(\lambda [ 1-\pi_c ] , c , r_0)r dr d\theta , \nonumber \\ & \;\;\;\rightarrow \underbrace{1-\frac{1}{\theta_0(\eta^2-r_0 ^ 2)}\int_{-\theta_0}^{\theta_0}\int_{r_0}^{\eta}\xi(\lambda [ 1-\pi_c ] , c , r_0)r dr d\theta}_{\pi_r^{\infty},\;\text{for}\ ; p_r\rightarrow \infty},\end{aligned}\ ] ] with @xmath106 . see appendix [ app_c1 ] for the outage probability of the first hop ( @xmath107 ) and appendix [ app_c2 ] for the outage probability of the second ( relaying ) hop ( @xmath108 ) . we note that for the case where @xmath109 or @xmath110 with @xmath111 , the system becomes interference limited and the outage performance converges to a constant outage probability floor given by @xmath112 on the other hand , each receiver harvests energy from both phases of the cooperative protocol . in contrast to the information decoding , which highly depends on the relaying transmission and thus becomes inactive in case of an empty relay set , the rf harvesting process is active in all cases . more specifically , in case where the relay set is empty ( no relay reforwards the source s message ) , the corresponding receiver does not employ a ps technique and uses all the received energy ( e.g. , interference ) for rf energy harvesting . based on this fundamental remark , we have the following proposition the average harvested energy for the cooperative protocol is given by @xmath113+\pi_c \cdot p_r \psi \big(\lambda[1-\pi_c ] \big ) \nonumber \\ & \;\;\;\;\;+(1-\pi_c ) \cdot(1-\nu_r)p_r \bigg[\mathcal{z}+\psi(\lambda [ 1-\pi_c ] ) \bigg],\end{aligned}\ ] ] where @xmath114 denotes the average attenuation for the relaying link . from , we have @xmath115+\pi_c\cdot p_r \psi \big(\lambda [ 1-\pi_c ] \big ) \nonumber \\ & \;\;\;\;\;+(1-\pi_c)\cdot ( 1-\nu_r)p_r \bigg[\mathcal{z}+\psi(\lambda [ 1-\pi_c ] ) \bigg],\end{aligned}\ ] ] where the proof of @xmath116 is reported in appendix [ app_aver ] ; it is worth noting that appendix [ app_aver ] provides both the exact value of the average attenuation as well as the above simplified approximation . it is worth noting that a similar optimization problem with can be formulated for the relaying case ; the objective function could be the minimization of the total transmitted power i.e. , @xmath117 . however , as it can be seen from proposition 3 , the expressions of the outage probability are complicated in the relaying case and do not allow elegant closed form solutions for the optimization problem . computer simulations are carried out in order to evaluate the performance of the proposed schemes . the simulation environment follows the description in sections [ system ] , [ system2 ] with parameters @xmath118 distance units ( we will use meters ( m ) for the sake of presentation ) , @xmath119 m , @xmath120 , @xmath121 , @xmath122 db , @xmath123 and @xmath52 ; the average harvested energy is measured in watts . for the sake of simplicity , we assume that the relay nodes have similar computational / complexity characteristics with the transmitters ( e.g. , they could be inactive transmitters of the network ) and therefore transmit with a power @xmath124 . for the cooperative protocol we assume @xmath125 m and thus remark 1 corresponds to @xmath126 . the presented results concern the typical link @xmath53 but refer to any link of the network @xmath127 ( according to the slivnyak s theorem ( * ? ? ? 8.5 ) ) . fig.s [ fig1a ] , [ fig1b ] deal with the performance of the non - cooperative scheme for different network densities e.g. , @xmath128 . specifically , fig . [ fig1a ] plots the outage probability of the system versus the transmitted power @xmath37 . the first main observation is that the outage performance converges to a constant floor for high transmitted powers @xmath58 . this behavior is due to the fact that all nodes transmit with the same power without any coordination ( scheduling ) and therefore the system becomes interference limited as @xmath37 increases . as for the impact of the network density on the outage performance , it can be seen that as the density increases , the outage probability of the system increases ; for @xmath129 the outage probability converges to @xmath130 , while for @xmath131 it converges to @xmath132 . this observation shows that the network density and the related multi - user interference significantly affects the decoding ability of the receivers . in the same figure , we plot the analytical results given by , which match with the simulation results and validate our analysis . on the other hand , fig . [ fig1b ] plots the rf average harvested energy versus the transmitted power @xmath37 . it can be seen that the rf average harvested energy is a linear function of the transmitted power ( as it can be observed by ) and increases as the transmitted power increases . by comparing the different curves , we can see that as the network density increases , the average harvested energy increases ; e.g. , if @xmath133 db , we have @xmath134 watt for @xmath135 and @xmath136 watt for @xmath131 . this remark shows that a dense network facilitates the rf energy harvesting process and thus interference is beneficial from an energy harvesting standpoint . in fig . [ fig1b ] , we also show the percentage of the harvested energy which is from the direct link . as it can be seen , for small network densities , the direct component significantly contributes to the total average harvested energy while becomes less important as the network density increases . for high network densities i.e. , @xmath131 , interference dominates the rf energy harvesting process and the percentage of the direct link is almost negligible i.e. , @xmath137 % . the theoretical curves perfectly match with the simulation results and validate our analysis in proposition [ prop2 ] . it is worth noting that these two figures demonstrate the fundamental trade - off between information decoding and rf energy harvesting ; interference significantly degrades the achieved outage performance , while it becomes helpful for the rf energy harvesting process . [ figure2 ] [ cols="^,^,^,^,^",options="header " , ] fig.s [ fig2a ] , [ fig2b ] show the impact of the power splitting ratio @xmath40 on the outage performance and the rf average harvested energy , respectively . we assume @xmath133 db and the other parameters are similar to the previous simulation example . a higher @xmath37 affects ( decreases ) the outage probability in accordance with fig . [ fig1a ] but does not change the observed behavior of the outage probability versus @xmath40 curves ; therefore further simulation results with other @xmath37 parameters do not add value to the main remarks of the paper . it can be seen that the power splitting ratio significantly affects the performance of the system and defines the balance between the two conflicting objectives i.e. , outage probability vs energy harvesting . specifically , as @xmath40 increases the outage performance is improved while the harvesting process becomes less efficient , since most of the received energy is used for information decoding ; when @xmath40 decreases we have the inverse behavior , since most of the received energy is used for rf - to - dc rectification . regarding the network s density @xmath16 , our observations confirm the previous main remarks . in table [ tab ] , we deal with the optimization problem discussed in section [ optimization_sec ] . specifically , we present the optimal solution for the case of a constant splitting power ratio @xmath138 ( _ optimization i _ ) as well as for the case where both parameters @xmath139 can be adjusted ( _ optimization ii _ ) . the first observation is that _ optimization ii _ gives a solution @xmath140 which is much lower than this one of _ optimization i _ , since the power spitting parameter is optimized accordingly . the solution of _ optimization ii _ satisfies both constraints with equality , as it has been discussed in section [ optimization_sec ] . on the other hand , for the first set - up @xmath141 , the solution of the _ optimization i _ satisfies the harvesting constraint with equality , since @xmath142 is the dominant constraint . for the setup @xmath143 , the optimal solution satisfies the outage constraint with equality , since @xmath144 becomes the dominant constraint for this case . [ figure3 ] although our analysis is general and concerns any @xmath16 , @xmath79 , in the simulations results , we assume that @xmath79 is much higher than the network density @xmath16 in order to demonstrate the potential gains from relaying at the outage / harvesting performance . more specifically , with a small @xmath79 ( e.g. , @xmath145 or @xmath146 ) the probability of decoding at the relays ( according to ) becomes almost zero and therefore we can not show the impact of cooperation . on the other hand , as @xmath16 increases , the multi - user interference significantly increases and the achieved outage probability of the system has not practical interest . therefore , in order to reveal the potential benefits of cooperation , we assume a small @xmath16 which ensures a low ( non - cooperative ) probability outage floor as well as a higher @xmath79 which provides a non - empty relay set and therefore cooperative diversity . a further optimization of the network densities is an interesting problem that could be considered for future work @xcite ; here , we assume that network densities are fixed and can not be taken into account in the design . this network density setup with @xmath147 could refer to a bursty network with a small transmission probability , where relays are part of the same network and correspond to the inactive nodes @xcite . fig.s [ fig3a ] , [ fig3b ] show the performance of the cooperative protocol in terms of outage probability and average harvested energy , for a simulation setup with @xmath124 , @xmath120 , @xmath148 , @xmath125 m and @xmath149 with @xmath150 ( remark 1 ) ; the other parameters are defined as before . specifically , fig . [ fig3a ] plots the outage probability versus the transmitted power @xmath37 . as it can be seen the main observations are similar to the non - cooperative protocol and thus the outage probability converges to a constant outage floor for high @xmath37 , since there is not any coordination / scheduling in both phases of the protocol . as for the network density @xmath16 , we can see that it significantly affects the outage performance of the system ; the associated multi - user interference degrades the decoder s performance at both the receivers and the relays in the first phase of the protocol . in the same figure , we plot the theoretical expressions given by ; we can see that the theoretical results provide a near - perfect match to the simulations results and validate our analysis for the cooperative case . a direct comparison between fig.s [ fig1a ] , [ fig3a ] for high @xmath37 ( i.e. , @xmath58 ) , the comparison between non - cooperative and cooperative protocol is fair and the their performance gap is due to the cooperative diversity associated with the cooperative scheme . ] , shows that the cooperative protocol improves the outage probability of the system and achieves a lower outage probability floor e.g. , for @xmath129 , the outage probability converges to an outage probability equal to @xmath130 and @xmath151 , for the non - cooperative protocol and the cooperative protocol , respectively , ( a more significant gain can be observed for another simulation setup as it is reported in the following discussion ) . in case of a successful decoding at the relay nodes ( the relay set is not empty ) , the cooperative protocol provides a retransmission of the source s message from a shorter distance than the direct link as well as via an independent fading channel and therefore improves the achieved outage probability due to the cooperative diversity . in fig . [ fig3b ] , we plot the rf average harvested energy versus the transmitted power @xmath37 . as it can be seen the relaying operation further improves the average harvested energy since the receivers can harvest energy from the relaying links . for @xmath131 and @xmath152 db , the average harvested energy increases from @xmath153 watt to @xmath154 watt due to cooperation . it is worth noting that when a receiver has not relay assistance , it uses all the received power during the second phase of the cooperative protocol for energy harvesting . in fig . [ fig4 ] , we show the impact of the central angle @xmath155 and the relay density @xmath79 on the outage performance of the cooperative protocol ; we assume @xmath156 , @xmath157 , while the other simulation parameters follows the previous example . it can be seen that the combination @xmath158 achieves the best outage performance for high @xmath37 ( it converges to the lowest outage floor ) . this result reveals a very interesting relation between these two parameters as well as a multidimensional trade - off . specifically , a high @xmath79 ensures a non - empty relay set during the first phase of the cooperative protocol and provides cooperative diversity benefits . in this case , a smaller angle i.e. , @xmath159 ( remark 1 ) can further protect the system from large - distance relays and therefore achieves a better outage probability performance than @xmath160 . on the other hand , as @xmath79 decreases , the probability to have a non - empty relay set increases and a larger angle is required in order to still have a potential relay at the area of the transmitter ; the condition in remark 1 becomes less important , since successful relay decoding is the priority for the system . the combination @xmath158 seems to provide the best balance between successful relay decoding and protection from large - distance relays . in comparison to the non - cooperative protocol , the considered setting reveals a significant gain of the cooperative protocol against the non - cooperative scheme e.g. , for @xmath161 and @xmath159 , the outage probability converges to @xmath162 in comparison to @xmath130 reported in fig . [ fig1a ] . finally , fig . [ fig5 ] depicts the average harvested energy versus @xmath37 . we observe that a small angle is beneficial for the energy harvesting operation since an empty relay set allows the receiver to use all the received signal for energy harvesting in the second phase of the protocol . in addition , as @xmath79 increases , the probability of relaying increases which is beneficial for the energy harvesting process . for different @xmath79 and @xmath155 ; @xmath124 , @xmath129 , @xmath156 , @xmath121 , @xmath157 , @xmath122 db , @xmath119 m , @xmath125 m , @xmath118 m , @xmath120 and @xmath123 . ] for different @xmath79 and @xmath155 ; @xmath124 , @xmath129 , @xmath156 , @xmath121 , @xmath157 , @xmath122 db , @xmath119 m , @xmath125 m , @xmath118 m , @xmath120 and @xmath123 . ] this paper has dealt with the ps harvesting technique in large - scale networks with multiple transmitter - receiver pairs , where receivers are characterized by both qos and rf energy harvesting requirements . a non - cooperative scheme where all transmitters simultaneously communicate with their associated receivers without any coordination , is analyzed in terms of outage performance and average harvested energy by using stochastic - geometry . we show that network density and power splitting ratio significantly affects the fundamental trade - off between outage performance and energy harvesting . for this case , an optimization problem that minimizes the transmitted power under outage probability and harvesting constraints is formulated and solved in closed form . in addition , a cooperative scheme where sources transmissions are assisted by a random set of orthogonal relays is analyzed . a random relay selection policy is considered with a sectrorized selection area at the direction of the receivers . analytical and simulation results reveal the impact of relay density and selection area on the achieved outage - probability / average harvested performance . an extension of this work is to integrate a coordination ( scheduling ) between the different transmissions and study the trade - off between energy harvesting and potential diversity gains . in addition , more sophisticated cooperative protocols and diversity combining schemes can also be considered in order to further boost the simultaneous information / energy transfer . in order to calculate the outage probability for the non - cooperative protocol , we need to calculate the laplace transform of the normalized interference term @xmath163 . we have @xmath164^{\mathbb{e}\big(n(b(0,r_0 ) ) \big ) } \nonumber \\ & = \mathcal{l}_{i_0}'(s ) \exp \left(-s \pi \lambda r_0^{2-\alpha } \right),\label{teliko}\end{aligned}\ ] ] where @xmath165 denotes the average number of points @xmath166 falling in a disk of radius @xmath55 @xcite . for the computation of @xmath167 , we have @xmath168 \bigg ) , \label{e4}\end{aligned}\ ] ] where @xmath169 \}$ ] denotes the integration area , follows from the probability generating functional of a ppp ( * ? ? ? 4.6 ) , by using the transformation @xmath170 , by using the transformation @xmath171 , and from integration by parts ; @xmath172 is the lower incomplete gamma function @xcite . the outage probability for the typical transmitter - receiver link @xmath53 can be written as @xmath173 where follows from the cumulative distribution function of an exponential random variable with unit variance @xmath174 and the laplace transform in is given by . it is worth noting that although the above analytical method is similar to several stochastic geometry works e.g. , @xcite , our analysis / result concerns a different problem and is based on different system assumptions . let @xmath21 be a ppp with density @xmath16 and let @xmath163 ; by using campbell s theorem for the expectation of a sum over a point process @xcite , we have : @xmath176 where @xmath177 . we define as @xmath178 the angle which is formed by the relay node @xmath179 , the transmitter @xmath18 and the receiver @xmath20 , @xmath180 and @xmath181 , as depicted in fig . [ model2 ] . by using the cosine rule , the requirement that the relay - receiver distance should be shorter than @xmath22 gives : @xmath182.\end{aligned}\ ] ] in the case where the selection area is a sector with a constant central angle , by applying the above condition to the border of the sector ( i.e. , for a distance @xmath100 ) , we have @xmath183.\end{aligned}\ ] ] it is worth noting that the above condition gives the maximum range of the angle ; any angle defined in this range , it also supports the distance requirement . let @xmath5 be the distance between transmitter and relay . the relay nodes that are able to successfully decode the source s signal form the point process @xmath185 , which is generated by the homogeneous ppp process @xmath186 by applying a thinning procedure @xcite ; therefore @xmath185 is a ppp with intensity @xmath187 where for the above expression we have used the expression in proposition [ prop1 ] for a direct distance equal to @xmath5 and @xmath188 . if we focus on the typical transmitter , the mean of @xmath185 inside the area @xmath189 is equal to @xmath190\lambda_r \exp\left(-\frac{\sigma^2 \omega r_0^{\alpha}}{p_t } \right)\xi(\lambda , r_0,r_0 ) \nonumber \\ & = \int_{-\theta_0}^{+\theta_0}\!\!\int_{r_0}^{\eta } \lambda_r \exp\left(-\frac{\sigma^2 \omega r^{\alpha}}{p_t } \right)\xi(\lambda , r , r_0)r dr d\theta \nonumber \\ & \;\;\;\;\;+\lambda_r \theta_0 r_0 ^ 2 \exp\left(-\frac{\sigma^2 \omega r_0^{\alpha}}{p_t } \right)\xi(\lambda , r_0,r_0)\end{aligned}\ ] ] by using fundamental properties of a ppp process @xcite , the probability to have an empty relaying set is equal to @xmath191 by using the cosine rule , the distance relay - receiver can be expressed as @xmath193 , where @xmath5 denotes the distance transmitter - relay ( see also fig . [ model2 ] ) . in the case of a relaying transmission , the interference at each receiver is generated by all selected relays which form a ppp @xmath194 with density @xmath195 ( i.e. , one relay is selected for each transmitter with probability @xmath196 ) . for the outage probability of the relaying link , we can apply the derived expressions for the direct link as follows @xmath197 where @xmath198 gives the area of @xmath80 . we note that the above expression takes into account that the smallest distance between a communication pair is @xmath55 ; the points of @xmath80 with @xmath199 are considered to have a distance @xmath55 according to the considered radio propagation model in . let @xmath200 the distance relay - receiver and @xmath201 the distance transmitter - relay ; by using the cosine rule ( see also fig . [ model2 ] ) , we have @xmath202 . for a selection area @xmath203,\;\theta\in [ -\theta_0,\;\theta_0 ] \}$ ] , the average relay - receiver attenuation can be expressed as @xmath204 o. ozel , k. tutuncuoglu , j. yang , s. ulukus and a. yener , `` transmission with energy harvesting nodes in fading wireless channels : optimal policies , '' _ ieee j. select . areas commun . 17321743 , sept . o. orhan , d. gunduz , and e. erkip , `` throughput maximization for an energy harvesting communication system with processing cost , '' in _ proc . . theory work . _ , lausane , switzerland , sept . 2012 , pp . 8488 . j. jeon and a. ephremides , `` the stability region of random multiple access under stochastic energy harvesting , '' in _ proc . inf . theory _ , saint petersburg , russia , july 2011 , pp . 17961800 . a. a. nasir , x. zhou , s. durrani , and r. a. kennedy , `` relaying protocols for wireless energy harvesting and information processing , '' _ ieee trans . wireless commun . 36223636 , july 2013 . z. ding , s. m. perlaza , i. esnaola , and h. v. poor , `` power allocation strategies in energy harvesting wireless cooperative networks , '' _ ieee trans . wireless commun . _ , accepted for publication , july 2013 . [ available online:]http://arxiv.org / pdf/1307.1630v2.pdf d. s. michalopoulos , h. a. suraweera , and r. schober , `` relay selection for simultaneous information and wireless energy transfer : a tradeoff perspective , '' _ ieee trans . wireless commun . _ , submitted for publication , march 2013 . [ online available ] : http://arxiv.org/pdf/1303.1647.pdf b. k. chalise , w -k . ma , y. d. zhang , h. a. suraweera , and m. g. amin , `` optimum performance boundaries of ostbc based af - mimo relay system with energy harvesting receiver , '' _ ieee trans . 61 , pp . 41994213 , sept . 2013 . m. haenggi , j. g. andrews , f. baccelli , o. dousse , and m. franceschetti , `` stochastic geometry and random graphs for the analysis and design of wireless networks , '' _ ieee j. selec . areas commun . 10291046 , sept . 2009 . z. shens , d. l. goeckel . k. k. leung , and z. ding , `` a stochastic geometry approach to transmission capacity in wireless cooperative networks , '' in _ proc . ind . mob . _ , tokyo , japan , sept . 2009 , pp . 622626 . r. vaze , `` transmission capacity of wireless ad hoc networks with energy harvesting nodes , '' _ ieee trans . inf . theory _ , submitted for publication , may 2013 . [ available online : ] http://arxiv.org/pdf/1205.5649v1.pdf h. s. dhillon , li , p. nuggehalli , z. pi , and j. g. andrews , `` fundamental of heterogeneous cellular networks with energy harvesting , '' _ ieee trans . wireless commun . _ , submitted for publication , july 2013 . [ available online : ] http://arxiv.org/pdf/1307.1524v1.pdf k. song , j. lee , s. park , and d. hong , `` spectrum - efficient operating policy for energy - harvesting clustered wireless networks , '' in _ proc . ieee pers . ind . mob . radio commun . _ , london , uk , sept . 2013 , pp . 23932397 . h. wang , s. ma , t. -s . ng , and h. v. poor , `` a general analytical approach for opportunistic cooperative systems with spatially random relays '' , _ ieee trans . wireless commun . 41224129 , dec . 2011 . y. a. chau and k. y. huang , `` channel statistics and performance of cooperative selection diversity with dual - hop amplify - and - forward relay over rayleigh fading channels , '' _ ieee trans . wireless commun . _ , vol . 7 , pp . 17791785 , may 2008 . ioannis krikidis ( s03-m07-sm12 ) received the diploma in computer engineering from the computer engineering and informatics department ( ceid ) of the university of patras , greece , in 2000 , and the m.sc and ph.d degrees from ecole nationale suprieure des tlcommunications ( enst ) , paris , france , in 2001 and 2005 , respectively , all in electrical engineering . from 2006 to 2007 he worked , as a post - doctoral researcher , with enst , paris , france , and from 2007 to 2010 he was a research fellow in the school of engineering and electronics at the university of edinburgh , edinburgh , uk . he has held also research positions at the department of electrical engineering , university of notre dame ; the department of electrical and computer engineering , university of maryland ; the interdisciplinary centre for security , reliability and trust , university of luxembourg ; and the department of electrical and electronic engineering , niigata university , japan . he is currently an assistant professor at the department of electrical and computer engineering , university of cyprus , nicosia , cyprus . his current research interests include information theory , wireless communications , cooperative communications , cognitive radio and secrecy communications . krikidis serves as an associate editor for the ieee wireless communications letters , ieee transactions on vehicular technology and elsevier transactions on emerging telecommunications technologies . he was the technical program co - chair for the ieee international symposium on signal processing and information technology 2013 . he received an ieee communications letters and an ieee wireless communications letters exemplary reviewer certificate in 2012 . he was the recipient of the _ research award young researcher _ from the research promotion foundation , cyprus , in 2013 .	energy harvesting ( eh ) from ambient radio - frequency ( rf ) electromagnetic waves is an efficient solution for fully autonomous and sustainable communication networks . most of the related works presented in the literature are based on specific ( and small - scale ) network structures , which although give useful insights on the potential benefits of the rf - eh technology , can not characterize the performance of general networks . in this paper , we adopt a large - scale approach of the rf - eh technology and we characterize the performance of a network with random number of transmitter - receiver pairs by using stochastic - geometry tools . specifically , we analyze the outage probability performance and the average harvested energy , when receivers employ power splitting ( ps ) technique for `` simultaneous '' information and energy transfer . a non - cooperative scheme , where information / energy are conveyed only via direct links , is firstly considered and the outage performance of the system as well as the average harvested energy are derived in closed form in function of the power splitting . for this protocol , an interesting optimization problem which minimizes the transmitted power under outage probability and harvesting constraints , is formulated and solved in closed form . in addition , we study a cooperative protocol where sources transmissions are supported by a random number of potential relays that are randomly distributed into the network . in this case , information / energy can be received at each destination via two independent and orthogonal paths ( in case of relaying ) . we characterize both performance metrics , when a selection combining scheme is applied at the receivers and a single relay is randomly selected for cooperative diversity . rf energy harvesting , stochastic geometry , poisson point process , relay channel , power consumption , outage probability .
superradiance is a well - known cooperative phenomenon that was first predicted in theory by dicke in 1954 @xcite . an intriguing conclusion is that when @xmath0 identical atoms in the limit of a small dimension are uniformly excited by a single photon , the decay rate of the system is @xmath0 times that of the isolated atom . the cooperative effect is due to the exchange of real and virtual photons between atoms by the interaction field , and it is obvious that the distances between atoms play a key role in the cooperative spontaneous emission . the simplest theoretical model is the two - atom system @xcite , where the dependence of collective decay rate and lamb shift on the distance is clearly demonstrated . for the @xmath0-atom system , several models , such as atomic clouds with spherical , cubic and slab - shaped geometry , or atoms arranged as a straight line , have been studied with different approaches @xcite . the manipulations of cooperative spontaneous emission can realize many interesting behaviors and potential applications @xcite . summary review on the cooperative spontaneous emission under certain approximations can be found in refs . @xcite . recently , cooperative spontaneous emission received renewed interest when the emission from an @xmath0-atom system excited by a single photon was considered @xcite . with the development of experimental technology , it becomes possible to detect the single - photon superradiance @xcite , which might yield new tools for storing quantum information and deepening our understanding on the physics of virtual processes @xcite . the cooperative decay was observed more than 40 years ago @xcite , but the direct experimental observation of the cooperative lamb shift has been achieved just recently @xcite . this breakthrough is stimulating the study on the effect of the virtual processes in the cooperative spontaneous emission . for a special initial state , i.e. the timed dicke state , a directed emission with large atomic samples is predicted @xcite . the dynamic problem involving the counter - rotating terms for a single excitation in a large atomic sample was studied in ref . @xcite , where , under several approximations , the author obtained the analytic conclusions indicating that the evolution of the timed dicke state obeys the simple exponential decay . in most of the previous papers on the cooperative spontaneous emission , the details of atomic positions are ignored and an average interatomic distance @xcite or the continuum approximation @xcite ( sums over atoms replaced by integrals , actually , which is equivalent to the distance average ) is taken . furthermore , in most of recent papers about single - excitation cooperative spontaneous emission , the scalar photon theory which ignores the polarization and vector character of the field has been applied @xcite . in our previous works @xcite , we studied the same problem in optical vector theory but did not reckon in the electrostatic dipole - dipole interaction which dominates when the distance between atoms is much smaller than the wavelength . in another paper @xcite , we have investigated the influence of the atomic distribution on the cooperative spontaneous emission for a simple model of three atoms with the electrostatic dipole - dipole interaction in optical vector theory . we found that the atomic distribution significantly influences the cooperative effects of the system , which leads us to suspect the universal applicability of the continuum approximation for multi - atom systems . in this paper , we extend the study about the influence of the atomic distribution on the cooperative spontaneous emission from the three - atom system @xcite to the @xmath0-atom one . the purpose of this paper is to illustrate that the spontaneous emission in random atomic distribution is very different from that in uniform distribution and the conclusions under the continuum approximation may be questionable for real experiments . the strong superradiance which approaches to the dicke limit is hard to obtain in real experimental conditions where atoms are always distributed randomly . on the other hand , with the development of semiconductor quantum dots , the highly ordered arrays can be achieved in this artificial system @xcite and its superradiance has been observed in experiment @xcite . in the controllable artificial system , the cooperative effects can be studied in a regime which was difficult to achieve with real atoms @xcite , and our results may be valuable to its experimental research . we consider a system consisting of @xmath0 identical multilevel atoms located at positions @xmath1 , @xmath2 the total hamiltonian of the atoms and electromagnetic ( em ) fields can be written as ( @xmath3 ) @xcite @xmath4 where @xmath5 @xmath6 @xmath7 @xmath8 is the unperturbed hamiltonian of the atoms and fields , @xmath9 is the interaction hamiltonian between the atoms and the transverse fields , and @xmath10 is the electrostatic dipole - dipole interaction ( also called instantaneous coulomb interaction ) between the atoms @xcite . here @xmath11 is the energy of the level @xmath12 , @xmath13 ( @xmath14 ) is the creation ( annihilation ) operator of the @xmath15th em mode with frequency @xmath16 , and @xmath17 is the coupling strength between the @xmath15th em mode with unit polarization vector @xmath18 and the atomic transition between levels @xmath12 and @xmath19 with transition dipole moment @xmath20 , of which @xmath21 ( assumed to be real ) and @xmath22 are the magnitude and unit vector , respectively . the displacement between the @xmath23th and @xmath24th atoms is @xmath25 . in eq . ( [ eq4 ] ) , @xmath26 is the dipole moment operator of the @xmath24th atom . here we have assumed that all the atoms are identical and similarly oriented @xcite . in order to take into account the counter - rotating terms and simplify the calculation , we introduce a unitary transformation @xmath27 @xcite with @xmath28 where @xmath29 and @xmath30 . in addition , we subtract the divergent free - electron self - energy @xmath31 from the hamiltonian . the effective hamiltonian after the transformation can be written as @xcite @xmath32 where @xmath33 @xmath34 @xmath35 @xmath36 contains the non - dynamic lamb shift for single atoms , i.e. , the second term in eq . ( [ eqa2 ] ) , which is due to the counter - rotating terms . @xmath37 is the transformed interaction hamiltonian describing the light - atom couplings , contains only the rotating wave terms ( i.e. , the terms associated with @xmath38 and @xmath39 where @xmath40 is below @xmath19 ) , and @xmath41 is the transformed coupling strength . in contrast , the interaction hamiltonian ( [ eq3 ] ) before the transformation describing the interaction between light and atoms contains both counter - rotating and rotating wave terms . the emerging term @xmath42 describes the interatomic interaction due to the exchange of virtual photons ( the counter - rotating terms ) . @xmath10 does not change its form after the unitary transformation because it commutes with @xmath43 . @xmath44 contains terms of order @xmath45 and higher , and it will be neglected . here we consider the weak excitation case in which only one of the atoms is in the first excited state and all others are in the ground state . since the transformed interaction hamiltonian @xmath37 only contains the rotating wave terms and the initial excitation is in the first excited state , the populations in higher atomic levels due to the counter - rotating terms can be neglected formally , and the multilevel atoms are reduced to effective two - level ( @xmath46 and @xmath47 ) ones . in the interaction picture with respect to @xmath36 , the wave function at time t can be written as @xmath48 where @xmath49 , and @xmath50 with @xmath51 standing for the vacuum and @xmath52 for one photon in the @xmath15th mode of the em field . substituting eq . ( [ eq7 ] ) into the schrdinger equation yields the differential equations for @xmath53 and @xmath54 . formally integrating the differential equation for @xmath54 with the initial value @xmath55 , and then substituting into the differential equation for @xmath53 , under the markov approximation , we find @xmath56 where for @xmath57 , @xmath58 with @xmath59 the single - atom decay rate from @xmath46 to @xmath47 , @xmath60 the wave number of resonant light , and @xmath61 the dynamic energy shift of single atoms . for @xmath62 , @xmath63 with @xmath64 } \right\},\end{aligned}\ ] ] @xmath65 } \right\},\label{eq11}\end{aligned}\ ] ] where @xmath66 is the angle between the dipole moment @xmath67 and the displacement @xmath68 . the quantity @xmath69 accounts for the collective spontaneous emission effect and the quantity @xmath70 describes the vacuum - induced dipole - dipole interaction containing the electrostatic dipole - dipole interaction between atoms . the detailed calculations of above results can be found in ref . note that the two quantities @xmath69 and @xmath70 had emerged in the master equation in refs . @xcite where the markov approximation had also been applied . for the markov approximation to be valid , the atomic sample must not be too large such that the propagation effects , e.g. retardation , are negligible @xcite . the coefficients @xmath71 in the differential equation ( [ eq8 ] ) constitute an @xmath72 matrix . the problem reduces to the determination of complex eigenvalues @xmath73 and eigenstates @xmath74 of the matrix @xmath75 @xcite . the eigenstates decay exponentially in the long - time limit . the decay rates of the eigenstates are given by @xmath76 , while @xmath77 are the corresponding lamb shifts . for any initial state @xmath78 its time evolution is @xmath79 here , we select two kinds of initial states of the system , the symmetric dicke state @xcite @xmath80 and the timed dicke state @xcite @xmath81 where @xmath82 is the wave vector of the incident photon which prepares the single - excitation state . note that the different initial states will result in different cooperative spontaneous emission , which also depends on the atomic distribution and the sample size . next , by diagonalizing the matrix @xmath83 , we numerically analyze the influence of atomic distribution on the cooperative spontaneous emission for small and large samples , respectively . first , we focus on a small sample , where the strong superradiance appears in the original theory of dicke . the dimension of the atomic sample is smaller than the resonant wavelength , so that the additional position - dependent phase in eq . ( [ eq15 ] ) is negligible and the timed dicke state @xmath84 reduces to the single - excitation dicke state @xmath85 . in the following analysis , we will take the dicke state as the initial state . we select two different geometry cases , a square and a cube . for each case , we made the atomic distribution from totally random to slight disturbance and finally , completely regular . to produce the random or slightly disturbed positions of atoms , we first fix atoms regularly in their own positions [ see figs . [ fig1](a1 ) and [ fig2](a1 ) ] , where the lattice constant is denoted by @xmath86 , then we add a random displacement @xmath87 $ ] on each atom [ see figs . [ fig1](b1 ) , [ fig1](c1 ) , [ fig2](b1 ) and [ fig2](c1 ) ] , where @xmath88 and @xmath89 is a random number between 0 and 1 . so that we can adjust the randomness of atomic distribution by controlling the parameter @xmath90 . note that the dipole - dipole interaction is divergent as @xmath91 when the atomic distance approaches to zero [ see eq.([eq11 ] ) ] . actually , this is not physical and we can make our calculation valid by forbidding the small atomic distance which has the same order with the bohr radius . in figs . [ fig1 ] and [ fig2 ] , the side lengths of the square and the cube are both @xmath92 with @xmath93 and @xmath94 , respectively , so the lattice constants are both @xmath95 . we estimate that the transition wavelength @xmath96 and the bohr radius @xmath97 , so that if we control @xmath98 , the nearest atomic separation after the random displacing @xmath87 $ ] will be larger than @xmath99 . in the second row of figs . [ fig1 ] and [ fig2 ] , we plot the structures of the radiative eigenstates for the corresponding atomic distributions shown in the first row . the horizontal axis represents the decay rates @xmath100 of the eigenstates @xmath101 and the vertical axis is their weights @xmath102 with the initial dicke state @xmath103 . we can see that in figs . [ fig1](a2 ) and [ fig2](a2 ) , there is only one dot appearing with large @xmath102 and with its decay rate close to the superradiant limit @xmath104 , while the dots belonging to all other modes have negligible contribution for the initial dicke state . this means , the dicke state is a approximate radiative eigenstate with the large decay rate close to superradiant limit @xmath105 for the small atomic sample under the regular distribution . however , if the atoms are distributed randomly , the dicke state is composed of many different eigenstates and their decay rates are all much smaller than @xmath104 , see figs . [ fig1](b2 ) , [ fig1](c2 ) , [ fig2](b2 ) and [ fig2](c2 ) . to characterize the time evolution of the system more clearly , from eq.([e13 ] ) we calculate the time - dependent population in all atomic excited state @xmath106 for the small atomic sample given in fig . [ fig2 ] with different atom - distribution random parameter @xmath107 ( dash dot ) , @xmath108 ( dot ) and @xmath109 ( dash ) and the single - atom case ( solid ) . the insert is the amplification of the little range . here @xmath110 and the initial state is the symmetric dicke state @xmath85.,width=264 ] the populations corresponding to different atomic distributions are plotted in fig . [ fig3 ] . for comparison , we also plot the single - atom decay ( see the black solid line ) . from this figure , we can see the emission of small atomic sample for the initial dicke state is superradiant whatever the distribution is . but , when the atoms are distributed regularly ( see the red dashed dotted line ) , the decay of the system will be much faster than that in the case of random atomic distribution ( see the blue dashed line and the green dotted line ) . this result coincides with the structural analysis of the radiative eigenstates shown in fig . [ fig2 ] . here we try to give a physical interpretation of the above numerical results . the cooperative effect of spontaneous emission is influenced by the dipole - dipole interaction between atoms which has the @xmath91 behavior in the small atomic distance . if the atoms are distributed randomly , some part of them which are close to each other may converge together as a small subsystem . so the atomic ensemble is split up into multiple subsystems whose weak cooperative effects are restricted in their own interior . however , when the atoms are distributed uniformly , the photons can be uniformly exchanged among them and as a whole they can display the strong collective spontaneous emission . in refs . @xcite , the electrostatic dipole - dipole interaction is neglected unreasonably , then the interactions between atoms are less sensitive to the distances and the strong superradiance is still obtained under the random distribution . note that the effect of regular atomic distribution is universal for different geometries ( see figs . [ fig1 ] and [ fig2 ] ) , therefore , we only choose the cubic geometry in the below analysis . for large multi - atom samples , the timed dicke state is superradiant , while the symmetric dicke state is subradiant @xcite . this previous conclusion is also based on the continuum approximation of atomic distribution . here , we discuss the influence of atomic distribution ( random or regular distribution ) on the cooperative spontaneous emission of large atomic samples . in fig . [ fig4 ] , we plot the regular and random atomic distributions and their corresponding structures of the radiative eigenstates with the initial dicke and timed dicke states . in the cases of random large samples , both the dicke and timed dicke states consist of many components of different eigenstates , as shown in figs . [ fig4](b2 ) and [ fig4](b3 ) . instead , for regular large samples , the dicke state largely projects to a few eigenstates with one of them holding nearly @xmath111 probability , and the timed dicke state also mostly projects to a few eigenstates whose corresponding @xmath102 are biggish and similar , see figs . [ fig4](a2 ) and [ fig4](a3 ) , respectively . we also plot the probability that atoms are still in the excited states for the two cases of random ( blue dashed lines ) and regular ( red dashed dotted lines ) distributions for the dicke and timed dicke states in fig . [ fig5 ] , wherein the solid lines mean the decay of a single - atom for comparison . from fig . [ fig5](a ) we can see that the dicke state is surely a subradiant state and its decay is slower than single - atom one whatever the atomic distribution is . this result is similar to that in refs . @xcite , where the dicke state is not a good trapped state and its excitation will be slowly emitted due to the virtual processes . however , if the atoms are distributed regularly , the decay of the excitation will be slower than that in the case of general random distribution obviously . of course , it is not realistic to control the atoms in regular distribution for the sample of low - pressure gas in most experiments . however , the semi - conduct quantum dot is the burgeoning and hopeful system to control the distribution and compass the purpose of storing the photon by subradiance with the dicke state in large sample . here , our results demonstrate that the regular distribution is beneficial to realize the photon storage . in fig . [ fig5](b ) , we plot the upper - state population for the initial timed dicke state . here , we want to recall fig . [ fig4](a3 ) first , where there are four eigenstates that dominate the evolution of the timed dicke state . it clearly demonstrates that the timed dicke state is not an approximate eigenstate for large atomic sample . however , due to the real parts of the four eigenvalues are approximately equal , the timed dicke state can be approximately written in an exponentially decaying form even if it is not an eigenstate . this result is coincident with the conclusion in refs . @xcite . from fig . [ fig5](b ) , we can see the approximately exponentially decaying behavior indeed . in addition , the difference of the evolved upper - state populations between the regular and random distributions for the timed dicke state is small , and the decays under the two distributions are both superradiant . in our previous works @xcite , we studied the total spectrum of cooperative spontaneous emission which is the average of the spectra detected in each direction . here we investigate the directed properties of the spontaneous emission for the initial timed dicke state @xcite and the influence of atomic distribution on the emission spectrum . the spectrum detected by the detector at position @xmath112 is given by refs . @xcite @xmath113 where @xmath114 , @xmath115 is the coefficient in eq . ( [ eq12 ] ) , and @xmath116 . note that the detector is far away from the atoms in experiment ( @xmath117 ) . the optical mode whose wavevector is not parallel to @xmath118 leads to negligible contributions to the detector , so @xmath119 is replaced by @xmath118 . meanwhile , as the detuning @xmath120 is far less than the resonant frequency @xmath121 , we can replace @xmath122 by @xmath123 in the exponential term . based on these approximations , we have @xmath124 . next , we select the direction of the atomic dipole moment @xmath125 in the @xmath126 direction , and assume the wave vector of the incident light which prepares the timed dicke state along the @xmath127 direction . the position of the detector in the coordinate system is shown in fig . [ fig6 ] , where @xmath128 is the angle between the detector and the @xmath126 direction , @xmath129 is the azimuthal angle of the vector . in the selected coordinate system , the expression of spectrum becomes @xmath130 it is well known that the classical dipole radiation is maximal in the direction perpendicular to the dipole . here , we have the same law for cooperative spontaneous emission which can be shown by the factor of @xmath131 in eq . ( [ eq18 ] ) . what we are concerned about is not this kind of angular distribution described by @xmath132 , but the directionality of spontaneous emission associated with the direction of the incident light @xmath133 . so we set the detector in the x - y plane ( @xmath134 ) and investigate the spectra in different directions by changing the azimuthal angle @xmath129 . ( in units of @xmath135 ) for the ( a ) regular and ( b ) random atomic distributions as shown in figs . [ fig8](a1 ) and [ fig8](b1 ) , respectively . the initial state is the timed dicke state.,title="fig:",width=151 ] ( in units of @xmath135 ) for the ( a ) regular and ( b ) random atomic distributions as shown in figs . [ fig8](a1 ) and [ fig8](b1 ) , respectively . the initial state is the timed dicke state.,title="fig:",width=151 ] we plot the spectra at different angles @xmath136 for small and large samples in figs . [ fig7 ] and [ fig8 ] , respectively . in fig . [ fig7 ] , we can see that the spectra in the case of random distribution for small samples do not exhibit a distinguishable peak . actually , it is composed of a large number of peaks corresponding to their multiple eigenstates . this can be explained by fig . [ fig2](c2 ) . as mentioned above , under the random atomic distribution , the dicke state is not an approximate eigenstate but consists of many contributions associated with different eigenstates . therefore , we obtain the spectrum shown in figs . [ fig7](b2 ) , ( b3 ) , and ( b4 ) . however , if the atoms are distributed regularly ( uniformly ) , the ( timed ) dicke state is an approximate eigenstate of the small - sample system [ see fig . [ fig2](a2 ) ] , so , there is a significant peak with half width about @xmath104 associated to the superradiant state , see figs . [ fig7](a2 ) , ( a3 ) and ( a4 ) . those additional narrow peaks come from the little components of other eigenstates . note that there is little difference between the spectra in different directions . that means there is no directed spontaneous emission for small sample . in fig . [ fig8 ] , we can clearly see the spectra in the direction of @xmath137 and @xmath138 are negligible compared with that in the direction of @xmath139 . the emission is almost along the direction of the incident light , @xmath82 . in addition , for the main direction of emission @xmath140 , the spectra of regular distribution has two peaks because there are two pairs of dominant eigenstates with their own similar lamb shifts [ see fig . [ fig9](a ) ] , while the random distribution only has one peak because the fusion of many eigenstates with quasi - continuous lamb shifts [ see fig . [ fig9](b ) ] . in this paper , we investigated the cooperative spontaneous emission of the system consisting of @xmath0 identical multilevel atoms in vacuum field by optical vector method . we focused on the influence of the details of atomic distribution , which had been ignored by most of previous papers . we find that the dicke s original conclusion that the superradiance limit is @xmath104 in small samples , is only established when the atoms are distributed uniformly . actually , in real experiments , the atoms are usually randomly distributed , and the superradiance limit @xmath104 will never be approached even if the dimension of the atomic sample is smaller enough than the resonance wavelength . for the large sample , the conclusion by scully @xcite that the timed dicke state will approximately decay exponentially is correct , and the decay of symmetric dicke state under regular distribution is slower than that in the random distribution . we also investigated the spectra , and demonstrated the directional emission for large atomic sample with the initial timed dicke state . this work was supported by the national natural science foundation of china ( under grants no . 11174026 and no . 11174027 ) and the national basic research program of china ( grants no . 2011cb922203 and no . 2012cb922104 ) . g. s. agarwal , _ quantum statistical theories of spontaneous emission and their relation to other approaches _ , edited by g. hhler , springer tracts in modern physics vol . 70 ( springer - verlag , berlin , 1974 ) .	we study cooperative single - photon spontaneous emission from @xmath0 multilevel atoms for different atomic distributions in optical vector theory . instead of the average approximation for interatomic distance or the continuum approximation ( sums over atoms replaced by integrals ) for atomic distribution , the positions of every atom are taken into account by numerical calculation . it is shown that the regularity of atomic distribution has considerable influence on cooperative spontaneous emission . for a small atomic sample ( compared with radiation wavelength ) , to obtain strong superradiance not only needs the uniform excitation ( the dicke state ) but also requires the uniform atomic distribution . for a large sample , the uniform atomic distribution is beneficial to subradiance of the dicke state , while the influence of atomic distribution on the timed dicke state is weak and its time evolution obeys exponential decay approximately . in addition , we also investigate the corresponding emission spectrum and verify the directed emission for the timed dicke state for a large atomic sample .
neutrinoless double - beta decay ( @xmath2 ) is nowadays a topical issue that has attracted a lot of interest from both theoretical and experimental sides @xcite . this decay mode is still unobserved and would violate lepton number conservation . its existence would demonstrate that the neutrino is a massive majorana particle and would provide a measurement of the absolute mass scale of the neutrino . obviously , the nuclear matrix element ( nme ) involved in such a process must be determined accurately to extract a reliable estimate of the neutrino mass . on the other hand , the two - neutrino double - beta decay mode ( @xmath3 ) is perfectly allowed by the standard model . it is a rare second - order weak - interaction process that has been observed experimentally in several nuclei ( see ref . @xcite for a review ) . thus , to test the reliability of the nuclear structure calculations involved in the @xmath2 process , one checks first the ability of the nuclear models to reproduce the experimental information available . certainly , there are differences between the nmes involved in both processes , but there are also clear similarities . among the differences , one should notice that the @xmath3-decay nmes are dominated by gamow - teller ( gt ) transitions connecting the initial and final @xmath4 states with all @xmath5 states in the intermediate nucleus . on the other hand , in the @xmath2 process the intermediate states run over all @xmath6 values . another difference with consequences in the evaluation of the nmes is that in @xmath2 decay the average virtual neutrino momentum is much larger than the typical scale of nuclear excitations and then closure approximation can be safely used . in the case of @xmath3 decay the neutrino momenta are comparable with the nuclear excitation energies , preventing the use of closure . among the similarities , the two processes connect the same initial and final nuclear ground states to be described within a given nuclear model and they share the calculation for the intermediate @xmath5 states . therefore , reproducing the @xmath3 nmes is a requirement for any nuclear structure model aiming to describe the neutrinoless mode . one can go even further and compare the calculations , not only with the @xmath3 nmes extracted from the measured half - lives , which is nothing but a number , but also with the gt@xmath7 ( gt@xmath8 ) strength distribution of the single branch connecting the initial ( final ) ground state with all the @xmath5 states in the intermediate nucleus . this comparison provides a more detailed information on which the calculation can be tested . gt strength distributions have been measured in recent years from high resolution charge - exchange reactions ( cer ) , within a large experimental program aimed to explore the gt properties at low excitation energies of double @xmath9-decay partners @xcite . the purpose of this work is to investigate the possibility to describe the rich information available at present on the gt nuclear response within a formalism based on a deformed proton - neutron quasiparticle random - phase approximation ( qrpa ) . this information includes global properties such as the location and strength of the gt resonance , more detailed description in the low - lying excitations , and last but not least , the @xmath3-decay nmes . this study was started for the double @xmath9-decay partners with @xmath10 @xcite . here , we shall focus on the @xmath0cd @xmath11sn and @xmath1nd @xmath12sm decays , motivated by the recent experiments on cer performed for @xmath0cd@xmath13 in and @xmath0sn@xmath14 in @xcite , as well as for @xmath1nd@xmath15he,@xmath16pm and @xmath1sm@xmath17he)@xmath1pm @xcite . it is also worth mentioning the recent measurement of the electron capture in @xmath0 in @xcite that can be used as a benchmark for nuclear structure calculations . in addition , the double @xmath9 decay of @xmath1nd has received increasing attention in the last years and it is currently considered as one of the best candidates to search for the @xmath2 decay in the planned experiments sno+ , supernemo , and dcba . the reason for the interest in @xmath1nd is that it has a large phase - space factor and therefore a relatively short half - life . it also has a large @xmath18 energy that will reduce the background contamination . however , both @xmath1nd and @xmath1sm are deformed nuclei that require a deformed formalism to deal with them properly . the paper is organized as follows : in sec . [ sec : theory ] , we present a brief summary of the theoretical approach used to describe the gt strength distributions , as well as the @xmath3-decay basic expressions . section [ sec : results ] contains the results obtained from our approach , which are compared to experimental data and other available calculations . section [ sec : conclusions ] contains a summary and the main conclusions . the description of the deformed qrpa approach used in this work is given elsewhere @xcite . here we give only a summary of the method . we start from a selfconsistent deformed hartree - fock ( hf ) calculation with effective nucleon - nucleon skyrme interactions , assuming axial deformation and time reversal symmetry @xcite . most of the results in this work are performed with the skyrme force sly4 @xcite , which is one of the most widely used and successful interactions . in addition , we also present results obtained with other skyrme forces to investigate their suitability for the description of the spin - isospin properties of nuclei . in particular , besides sly4 , we shall show results from the simpler force sk3 @xcite and from the force sgii @xcite that was fitted taking into account nuclear spin - isospin properties . in our approach , the single - particle wave functions are expanded in terms of the eigenstates of an axially symmetric harmonic oscillator in cylindrical coordinates using twelve major shells . pairing correlations between like nucleons are included in bcs approximation taking fixed pairing gap parameters for protons and neutrons , which are phenomenologically determined . besides the selfconsistent hf solution , we also explore the potential energy curves , that is , the hf energy as a function of the quadrupole deformation @xmath9 , which are obtained from constrained hf+bcs calculations . these results can be seen in fig . [ fig1 ] for @xmath0cd and @xmath0sn , and in fig . [ fig2 ] for @xmath1nd and @xmath1sm for the three skyrme forces considered . we get very soft profiles for @xmath0cd and @xmath0sn , whereas for @xmath1nd and @xmath1sm we obtain two energy minima , oblate and prolate , but with clear prolate ground states in both cases . we obtain similar results with the three skyrme forces with the only exception of the force sk3 in @xmath0cd , which seems to miss a spherical solution . cd ( a ) and @xmath0sn ( b ) obtained from hf+bcs calculations with various skyrme forces.,width=321 ] , but for @xmath1nd ( a ) and @xmath1sm ( b).,width=321 ] in @xmath0cd as a function of the excitation energy in the daughter nucleus ( a ) and accumulated strengths ( b ) . sly4-qrpa calculations for various deformations are compared to data ( sasano 2012 : @xcite ) from @xmath19 reactions.,width=321 ] in @xmath0sn as a function of the excitation energy in the daughter nucleus ( a ) and accumulated strengths ( b ) . sly4-qrpa calculations for various deformations are compared to data ( sasano 2012 : @xcite ) from @xmath20 reactions , as well as from @xmath21he ) ( rakers 2005 : @xcite).,width=321 ] , but for @xmath22 in @xmath1nd . data ( guess 2011 : @xcite ) are from @xmath15he , t ) cers.,width=321 ] , but for @xmath23 in @xmath1sm . data ( guess 2011 : @xcite ) are from ( t,@xmath24he ) cers.,width=321 ] cd ( a ) , @xmath0sn ( b ) , @xmath1nd ( c ) , and @xmath1sm ( d ) . the results correspond to the prolate shapes with the skyrme forces sly4 , sgii , and sk3 . data are as in the previous figures.,width=321 ] , but in the low energy range . also shown is the gt strength extracted from electron capture of @xmath0 in ( wrede 2013 : @xcite).,width=321 ] , but in the low energy range . also shown ( open circle ) is the gt strength extracted from @xmath25 decay of @xmath0 in to which the data from @xcite are normalized.,width=321 ] , but in the low energy range.,width=321 ] , but in the low energy range.,width=321 ] decay of @xmath1nd as a function of the coupling strengths @xmath26 ( a ) and @xmath27 ( b ) for prolate shapes in both @xmath1nd and @xmath1sm . the shaded area indicates the experimental range extracted from the measured half - life using bare @xmath28 ( lower line ) and quenched @xmath29 ( upper line ) factors.,width=321 ] nme in @xmath0cd as a function of the intermediate excitation energy in @xmath0 in for various calculations using different quadrupole deformations for parent and daughter nuclei . thin lines correspond to calculations with the energy denominator @xmath30 ( [ den1 ] ) while thick lines correspond to @xmath31 ( [ den2 ] ) . ] , but for the double-@xmath9 decay of @xmath1nd . ] nme in @xmath0cd as a function of the intermediate excitation energy in @xmath0 in for the decay between prolate shapes . the results are obtained with the energy denominator @xmath31 for three different skyrme forces . ] , but for the decay of @xmath1nd . ] after the hf+bcs calculation , the qrpa equations are solved on the deformed ground - state basis to get the gt strength distributions and to compute the @xmath3-decay nme . to describe gt excitations in qrpa we add to the quasiparticle mean field a separable spin - isospin residual interaction in the particle - hole ( @xmath32 ) and particle - particle ( @xmath33 ) channels . the @xmath32 part is responsible for the position and structure of the gt resonance and its coupling constant @xmath26 is usually taken to reproduce them @xcite . the @xmath33 part consists of a proton - neutron pairing force and it is also introduced as a separable force @xcite . the coupling constant @xmath27 is usually fitted to the half - lives phenomenology @xcite . in this work we have chosen for the @xmath32 and @xmath33 coupling constants the law dependence on the mass number @xmath34 given by @xmath35 in ref . however , the different mean fields used , nilsson @xcite versus skyrme - hf here , make the constants somewhat different . specifically we use @xmath36 and @xmath37 mev . although slightly different values could be used for the different skyrme forces , we use in this work the same values for them , so that changes among calculations from various forces are caused by the underlying quasiparticle structure provided by the effective skyrme force . at this point it is worth mentioning the studies on @xmath1nd and @xmath1sm carried out in refs . @xcite , where deformed qrpa calculations were performed using realistic nucleon - nucleon residual interactions based on the brueckner @xmath38 matrix for the cd - bonn force on top of a phenomenological deformed woods - saxon potential . comparison with the results obtained from schematic separable forces @xcite shows that the latter reproduce fairly well the main characteristic of the gt strength distributions profiles and @xmath3-decay nmes . the technical details to solve the qrpa equations have been described in refs . @xcite . for each value of the excitation energy , the gt transition amplitudes in the intrinsic frame connecting the ground state @xmath39 to one - phonon states in the daughter nucleus @xmath40 , are found to be @xmath41 where @xmath42 in terms of the occupation amplitudes for neutrons and protons @xmath43 ( @xmath44 ) and the forward and backward amplitudes of the qrpa phonon operator , @xmath45 and @xmath46 , respectively . once the intrinsic amplitudes in eq . ( [ amplgt ] ) are calculated , the gt strength @xmath47(gt ) in the laboratory frame for a transition @xmath48 can be obtained as @xmath49 \ , . \label{bgt}\end{aligned}\ ] ] to obtain this expression we have used the initial and final states in the laboratory frame expressed in terms of the intrinsic states using the bohr and mottelson factorization . finally , a quenching factor @xmath50 is included in the calculations to take into account in an effective way all the correlations that are not properly considered in the present approach . we shall discuss later in subsec . [ dbd ] various attempts aiming to fit simultaneously single-@xmath9 and double-@xmath9 decay observables by adjusting the quenching factor . concerning the @xmath3-decay nmes , the basic expressions for this process , within the deformed qrpa formalism used in this work , can be found in refs . deformation effects on the @xmath3 nmes have been also studied within the projected hartree - fock - bogoliubov model @xcite . attempts to describe deformation effects on the @xmath2 decay within qrpa models can also be found in refs . the half - life of the @xmath3 decay can be written as @xmath51 ^{-1}= ( g_a)^4\ g^{2\nu\beta\beta}\ \left\| m_{gt}^{2\nu\beta\beta}\right\| ^2 \ , , \ ] ] in terms of the phase - space integral @xmath52 and the nuclear matrix element @xmath53 that contains all the information of the nuclear structure involved in the process , @xmath54 in this equation @xmath55 are the qrpa intermediate @xmath56 states reached from the initial ( final ) nucleus . the indices @xmath57 , @xmath58 label the @xmath56 states of the intermediate nucleus . the gt matrix elements are those in eq . ( [ amplgt ] ) . the overlaps are needed to take into account the non - orthogonality of the intermediate states reached from different initial @xmath59 and final @xmath60 ground states . their expressions can be found in ref . the energy denominator @xmath61 involves the energy of the emitted leptons , which is given on average by @xmath62 , as well as the excitation energies of the intermediate nucleus . in terms of the qrpa excitation energies the denominator can be written as @xmath63 where @xmath64 is the qrpa excitation energy relative to the initial ( final ) nucleus . it turns out that the nmes are quite sensitive to the values of the denominator , especially for low - lying states when the denominator reaches smaller values . thus , it is a common practice to use some experimental normalization of this denominator trying to improve the accuracy of the nmes . in this work we shall also consider the denominator @xmath31 , which is corrected with the experimental energy @xmath65 of the first @xmath56 state in the intermediate nucleus relative to the mean ground - state energy of the initial and final nuclei , in such a way that the first calculated @xmath56 state appears at the exact experimental energy , @xmath66 + \bar{\omega}_{1_1^+ } \ , .\ ] ] the running @xmath3 sums calculated later , will be shown for the two choices of the denominator @xmath30 and @xmath31 . in the case of the @xmath67 isobars , the ground state in the intermediate nucleus @xmath0 in is a @xmath56 state . then , the energy @xmath65 is given by @xmath68 written in terms of the experimental energies @xmath69 mev and @xmath70 mev of the decays of @xmath0 in into @xmath0cd and @xmath0sn , respectively . in the case of @xmath71 , although the ground state in @xmath1pm is not a @xmath56 state , the first @xmath56 state excited in cers appears at a very low excitation energy , e=0.11 mev @xcite . hence , we also determine @xmath65 from the experimental values of @xmath72 mev and @xmath73 mev for the decays of @xmath1pm into @xmath1nd and @xmath1sm , respectively . the various measurements reported for the @xmath3-decay half - lives ( @xmath74 ) have been analyzed in ref . @xcite , where recommended values of @xmath75cd)=@xmath76 yr and @xmath77nd)=@xmath78 yr were adopted . using the phase - space factors from the recent evaluation @xcite that involves exact dirac wave functions with finite nuclear size and electron screening , @xmath79cd)=@xmath80 yr@xmath81 , and @xmath82nd)@xmath83 yr@xmath81 , one gets the experimental nuclear matrix elements @xmath84cd)=0.138 mev@xmath81 ( 0.223 mev@xmath81 ) when the bare @xmath85 ( quenched factor @xmath29 ) is used , and @xmath86nd)=0.070 mev@xmath81 ( 0.114 mev@xmath81 ) when @xmath85 ( @xmath29 ) is used . next we discuss the results obtained for the energy distributions of the gt strength . in the upper panels ( a ) of figs . [ fig3 ] and [ fig4 ] we can see the gt strength distributions for @xmath0cd ( @xmath22 ) and @xmath0sn ( @xmath23 ) , respectively . the lower panels ( b ) show the respective accumulated gt strengths . we compare our qrpa results from sly4 obtained for prolate , oblate , and spherical shapes in the initial ( @xmath0cd ) and final ( @xmath0sn ) nuclei with the experimental strengths extracted from @xmath0sn@xmath21he)@xmath0 in cers @xcite and from @xmath0cd@xmath13 in and @xmath0sn@xmath14 in cers @xcite . one should notice that the measured strength extracted from the cross sections contains two types of contributions , namely gt ( @xmath87 operator ) and isovector spin monopole ( ivsm ) ( @xmath88 operator ) . both are associated to @xmath89 and @xmath90 transitions and therefore , they have similar angular distributions that can not be disentangled in the experiment . thus , the measured strength corresponds actually to b(gt+ivsm ) . different theoretical calculations evaluating the contributions from both gt and ivsm modes are available in the literature @xcite . in ref . @xcite a self - consistent skyrme hartree - fock plus tamm - dancoff approximation was used to separate the gt from the ivsm strength . well separated centroids of both resonances were found , with the ivsm resonance at much higher energy , as it corresponds to a @xmath91 mode . subsequent calculations including @xmath67 isobars were performed in refs . @xcite with different degrees of complexity , from a schematic single - particle model space up to realistic calculations within a qrpa approach in a woods - saxon basis and two - body interactions constructed with the bonn - a potential . in ref . @xcite the contribution from the ivsm component was also evaluated by employing a microscopic method based on qrpa with residual interactions obtained from the brueckner g - matrix for cd - bonn nucleon - nucleon force . in general , the conclusions of those theoretical calculations converge in an overall picture . in the @xmath19 direction the strength distribution is dominated by the gt component up to about 20 mev , although non - negligible contributions from ivsm components are found between 10 and 20 mev , overlapping with the gt strength of the gt giant resonance . above this energy , there is no significant amount of gt strength in the calculations . the interference effects are constructive below 20 mev and destructive above this energy , but they are not very important . in the @xmath20 direction the gt strength is expected to be strongly pauli blocked in nuclei with more neutrons than protons and therefore , the measured strength is mostly due to the ivsm resonance over the whole excitation - energy range . nevertheless , the strength found in low - lying isolated peaks is associated with gt transitions because the continuous tail of the ivsm resonance is very small at these energies and is not expected to exhibit any peak . in summary , the measured strength in the ( @xmath92 ) direction can be safely assigned to be gt in the low energy range below 10 mev and with some caution between 10 and 20 mev . beyond 20 mev the strength would be practically due to ivsm . on the other hand the measured strength in the ( @xmath93 ) direction would be due to ivsm transitions , except in the low - lying excitation energy below several mev , where the isolated peaks observed can be attributed to gt strength . this is the reason why we plot experimental data in fig . [ fig4 ] only up to 5 mev . calculations for the gt strength distributions in both directions b(gt@xmath94 ) for the @xmath67 isobars are also available from spherical qrpa calculations within a woods - saxon basis and two - body bonn - a realistic interactions @xcite . the results from this reference are quite similar to those in figs . [ fig3 ] and [ fig4 ] , although a direct comparison is not easy as they are obtained for particular values of both the strength of the @xmath33 residual interaction and the effective @xmath95 value . the b(gt@xmath7 ) strength distribution calculated in ref . @xcite is characterized by some marginal strength up to an excitation energy of 5 mev , then a small bump develops between 7 and 10 mev , and finally the gt resonance appears between 11 and 15 mev . both the energy location of the bumps and the strength contained agree with the features observed in fig . similarly , the b(gt@xmath8 ) distributions in ref . @xcite and in the corresponding fig . [ fig4 ] in this work , contain a similar amount of strength concentrated in the very low - lying excitation energy . [ fig5 ] and [ fig6 ] show the results for @xmath1nd and @xmath1sm , respectively . we compare the qrpa results from sly4 for prolate and oblate shapes with the data from @xmath15he,@xmath96 on @xmath1nd and from @xmath17he ) on @xmath1sm in ref . @xcite . in general terms , we reproduce fairly well the global properties of the gt strength distributions , including the location of the gt resonance and the total strength measured , although the experimental gt resonances appear more fragmented than the calculated ones . similar comments to the previous case @xmath67 relative to the contributions from gt and ivsm components apply now to the case @xmath71 . microscopic qrpa calculations with residual interactions obtained from the cd - bonn potential were performed in ref . @xcite to estimate the gt and ivsm components of the strength b(gt+ivsm ) measured . from this analysis it was also concluded that the strength b(gt@xmath7+ivsm@xmath7 ) below 20 mev correspond to a large extent to gt strength , whereas the measured strength b(gt@xmath8+ivsm@xmath8 ) below 3 mev could be assigned to the gt component . in fig . [ fig7 ] we show the results for the gt@xmath97 strength distributions obtained with three different skyrme forces to see the sensitivity of these distributions to the nucleon - nucleon effective interaction used in the calculations . we can see that the profiles obtained with the three forces are quite similar . the gt@xmath7 resonances appear at close energies and contain about the same strength for the three interactions . the gt@xmath8 strength is very small compared to the gt@xmath7 and is concentrated for the three forces in the excitation energy region below 5 mev , with a bump in @xmath1sm centered at about 15 mev . figures [ fig8 ] , [ fig9 ] , [ fig10 ] , and [ fig11 ] show in more detail the gt strength distributions with sly4 in the low excitation - energy range . in the case of @xmath0cd ( fig . [ fig8 ] ) we have included , in addition , the experimental gt strength measured by ground - state to ground - state electron capture on @xmath0 in @xcite . this value agrees well with the strength obtained with the prolate shape and to a less extent with the oblate one , but it is too low compared to the spherical calculation . similarly , in the case of @xmath0sn ( fig . [ fig9 ] ) we show with an open circle the experimental gt strength measured by the ground - state to ground - state @xmath25 decay ( @xmath0 in @xmath98sn ) . one should notice that the b(gt ) strength from ref . @xcite is calibrated to this value . although a detailed spectroscopy is beyond the capabilities of our model and the isolated transitions are not well reproduced by our calculations , the overall agreement with the total strength contained in this reduced energy interval , as well as with the profiles of the accumulated strength distributions is reasonable , especially for the prolate shapes in both @xmath67 partners in figs . [ fig8 ] and [ fig9 ] , as well as with the prolate profiles in the @xmath71 partners displayed in figs . [ fig10 ] and [ fig11 ] . it is well known that the @xmath3 nmes are very sensitive to the residual interactions , as well as to differences in deformation between initial and final nuclei @xcite . to illustrate the dependence of the nmes to the residual forces , we show in fig . [ fig12 ] the sly4-qrpa results with prolate shapes for the @xmath1nd @xmath12sm for various strengths of the coupling constants @xmath26 and @xmath27 . the shaded region corresponds to the experimental nmes extracted from the measured @xmath3 half - lives , using bare ( @xmath28 ) or quenched ( @xmath29 ) values . we can see that the experimental nmes contained in the shaded region are reproduced in windows of the parameters that include the values used in this work ( @xmath99 mev and @xmath100 mev in @xmath1nd ) . in the next figures we show the running sums for the @xmath3 nmes . these are the partial contributions to the nmes of all the @xmath56 states in the intermediate nucleus up to a given energy . we see in fig . [ fig13 ] the running sums for the @xmath3 decay in @xmath0cd for three combinations of the initial and final nuclear shapes . the nmes of crossing deformations produce much lower values than the experimental window and are not shown . similarly , we show in fig . [ fig14 ] the running sums for the @xmath1nd @xmath12sm decay using prolate shapes for both nuclei . results obtained with the energy denominator @xmath30 ( [ den1 ] ) are displayed with thin lines , whereas results obtained with @xmath31 ( [ den2 ] ) are shown with thick lines . we can see that @xmath31 denominators produce larger nmes than @xmath30 . the main difference is originated at low excitation energies , where the relative effect of using shifted energies is enhanced . the effect at larger energies is negligible and we get a constant difference between @xmath30 and @xmath31 , which is the difference accumulated in the first few mevs . we also observe negative contributions coming from the region of the gt resonances at about 15 mev . this feature was studied in ref . @xcite and was related to the strength of the @xmath33 residual interaction . the small contribution to the @xmath3 nmes from the region of the gt resonance is due to the joint effects of large energy denominators in eq . ( [ mgt ] ) and the mismatch between the branches of the gt@xmath7 and gt@xmath8 excitations in the energy region of the gt resonance . in figs . [ fig15 ] and [ fig16 ] we have the @xmath3 nmes calculated with the denominator @xmath31 for three skyrme forces in @xmath0cd and @xmath1nd , respectively . the results correspond to the prolate shapes in both initial and final nuclei . the sensitivity of the nmes to the skyrme interaction is manifest in the figures , but it is not as significant as the sensitivity found to other aspects such as the residual interactions , the deformation , or even the energy denominators . another issue worth to comment is the treatment of the quenching factor of the axial - vector coupling constant @xmath95 . the physical reasons for this quenching have been studied elsewhere @xcite and are related to the role of non - nucleonic degrees of freedom , absent in the usual theoretical models , and to the limitations of model space , many - nucleon configurations , and deep correlations missing in these calculations . the implications of this quenching on the description of single-@xmath9 and double-@xmath9 decay observables have been considered recently in several works @xcite . in those works both the effective value of @xmath95 and the coupling strength of the residual interaction in the particle - particle channel are considered free parameters of the calculation . the striking result found is that very strong quenching values are needed to reproduce simultaneously the observations corresponding to the @xmath3 half - lives and to the single-@xmath9 decay branches , namely @xmath101 values for the ground state to ground state decay of the intermediate nucleus into the initial one , and the @xmath102 values for the ground state to ground state decay of the intermediate nucleus into the final one . different procedures for adjusting those parameters have been explained in refs . @xcite . the most recent analysis performed in ref . @xcite within a qrpa formalism comes to the conclusion that considerable quenching is needed , with an average value @xmath103 . the need of a strong quenching is also required in other theoretical approaches , such as different qrpa calculations ( @xmath104 ) @xcite , ibm-2 models @xcite , where the quenching obtained is @xmath105 for @xmath0cd and @xmath106 for @xmath1nd , or shell - model calculations that found quenching factors in the interval @xmath107 @xcite for double-@xmath9 emitters from @xmath108ca up to @xmath1nd . one should note that the qrpa calculations leading to the strong quenching that fits the @xmath3 nmes have been performed within a spherical formalism neglecting possible effects from deformation degrees of freedom . because the main effect of deformation is a reduction of the nmes , deformed qrpa calculations shall demand less quenching to fit the experiment . although the goal of this paper is not to extract consequences on the quenching needed to fit the experiment , it will be very interesting to explore in the future the implications of deformation . this study will be part of a more ambitious project aiming to extract an effective @xmath95 value that fits the decay observables taking into account the uncertainties related to the theoretical description and including all possible @xmath3-decay candidates . in summary , using a theoretical approach based on a deformed hf+bcs+qrpa calculations with effective skyrme interactions , pairing correlations , and spin - isospin residual separable forces in the @xmath32 and @xmath33 channels , we have studied simultaneously the gt strength distributions of the double-@xmath9 decay partners ( @xmath0cd , @xmath0sn ) and ( @xmath1nd , @xmath1sm ) , reaching the intermediate nuclei @xmath0 in and @xmath1pm , respectively , as well as their @xmath3 nmes . our results for the energy distributions of the gt strength have been compared to recent data from charge - exchange reactions , whereas the calculated @xmath3 nmes have been compared with the experimental values extracted from the measured half - lives for these processes . we have discussed the sensitivity of our results to the various ingredients in the theoretical formalism . namely , to the skyrme force , to the residual interactions , to deformation , to effective @xmath95 values , and to the treatment of the energy denominators of the nmes . we found different sensitivities to them that have been analyzed and discussed . all in all , the method used in this work has demonstrated to be well suited to account for the rich variety of experimental information available on the nuclear gt response . the global properties of the energy distributions of the gt strength and the @xmath3 nmes are reasonably well reproduced , the exception is the detailed description of the low - lying gt strength distributions that could clearly be improved . it will be interesting in the future to extend these calculations by including all the double-@xmath9 decay candidates and to explore systematically the potential of this method . we are grateful to e. moya de guerra for useful discussions . this work was supported in part by mineco ( spain ) under research grant no . fis2011 - 23565 and by consolider - ingenio 2010 programs cpan csd2007 - 00042 . 99 j. suhonen and o. civitarese , phys . rep . * 300 * , 123 ( 1998 ) ; a. faessler and f. imkovic , j. phys . * 24 * , 2139 ( 1998 ) ; h. v. klapdor - kleingrothaus , _ seventy years of double beta decay _ , world scientific , singapore ( 2010 ) . a. s. barabash , phys . c * 81 * , 035501 ( 2010 ) . y. fujita , b. rubio , and w. gelletly , prog . phys . * 66 * , 549 ( 2011 ) . d. frekers , p. puppe , j. h. thies , and h. ejiri , nucl . phys . a * 916 * , 219 ( 2013 ) . s. rakers _ et al . c * 71 * , 054313 ( 2005 ) . m. sasano _ et al . _ , c * 85 * , 061301(r ) ( 2012 ) . c. j. guess _ et al . c * 83 * , 064318 ( 2011 ) . p. sarriguren , phys . c * 86 * , 034335 ( 2012 ) . p. sarriguren , o. moreno , and e. moya de guerra , rom . journ . phys . * 58 * , 1242 ( 2013 ) . c. wrede _ et al . c * 87 * , 031303(r ) ( 2013 ) . p. sarriguren , e. moya de guerra , a. escuderos , and a. c. carrizo , nucl . a * 635 * , 55 ( 1998 ) ; p. sarriguren , e. moya de guerra , and a. escuderos , nucl . a * 658 * , 13 ( 1999 ) ; nucl . a * 691 * , 631 ( 2001 ) . p. sarriguren , e. moya de guerra , and a. escuderos , phys . c * 64 * , 064306 ( 2001 ) . r. lvarez - rodrguez , p. sarriguren , e. moya de guerra , l. pacearescu , a. faessler , and f. imkovic , phys . c * 70 * , 064309 ( 2004 ) . d. vautherin and d. m. brink , phys . c * 5 * , 626 ( 1972 ) ; d. vautherin , phys . c * 7 * , 296 ( 1973 ) . a. chabanat , p. bonche , p. haensel , j. meyer , and r. schaeffer , nucl . a * 635 * , 231 ( 1998 ) . m. beiner , h. flocard , n. van giai , and p. quentin , nucl . a * 238 * , 29 ( 1975 ) . n. van giai and h. sagawa , phys . b * 106 * , 379 ( 1981 ) . p. moeller and j. randrup , nucl . a * 514 * , 1 ( 1990 ) . k. muto , e. bender , and h. v. klapdor , z. phys . a * 333 * , 125 ( 1989 ) ; k. muto , e. bender , t. oda , and h. v. klapdor - kleingrothaus , z. phys . a * 341 * , 407 ( 1992 ) . h. homma , e. bender , m. hirsch , k. muto , h. v. klapdor - kleingrothaus , and t. oda , phys . c * 54 * , 2972 ( 1996 ) . m. s. yousef , v. rodin , a. faessler , and f. imkovic , phys . c * 79 * , 014314 ( 2009 ) . d. fang , a. faessler , v. rodin , m. s. yousef , and f. imkovic , phys . c * 81 * , 037303 ( 2010 ) . f. imkovic , l. pacearescu , and a. faessler , nucl . a * 733 * , 321 ( 2004 ) . o. moreno , r. lvarez - rodrguez , p. sarriguren , e. moya de guerra , f. imkovic , and a. faessler , j. phys . g : nucl . part . phys . * 36 * , 015106 ( 2009 ) . s. singh , r. chandra , p. k. rath , p. k. raina , and j. g. hirsch , eur . j. a * 33 * , 375 ( 2007 ) . d. l. fang , a. faessler , v. rodin , and f. imkovic , phys . c * 82 * , 051301(r ) ( 2010 ) . m. t. mustonen and j. engel , phys . c * 87 * , 064302 ( 2013 ) . j. kotila and f. iachello , phys . c * 85 * , 034316 ( 2012 ) . i. hamamoto and h. sagawa , phys . c * 62 * , 024319 ( 2000 ) . d. r. bes , o. civitarese , and j. suhonen , phys . c * 86 * , 024314 ( 2012 ) . o. civitarese and j. suhonen , phys . c * 89 * , 044319 ( 2014 ) . j. suhonen and o. civitarese , nucl . phys . * a 924 * , 1 ( 2014 ) . f. osterfeld , rev . mod . phys . * 64 * , 491 ( 1992 ) . g. f. bertsch and i. hamamoto , phys . c * 26 * , 1323 ( 1982 ) . e. caurier , g. martinez - pinedo , f. nowacki , a. poves , and a. p. zuker , rev . * 77 * , 427 ( 2005 ) . j. suhonen and o. civitarese , phys . b * 725 * , 153 ( 2013 ) . a. faessler , g. l. fogli , e. lisi , v. rodin , a. m. rotunno , and f. imkovic , j. phys . * 35 * , 075104 ( 2008 ) . j. barea , j. kotila , and f. iachello , phys . c * 87 * , 014315 ( 2013 ) . e. caurier , f. nowacki , and a. poves , phys . b * 711 * , 62 ( 2012 ) .	the two gamow - teller ( gt ) branches connecting the double - beta decay partners ( @xmath0cd , @xmath0sn ) and ( @xmath1nd , @xmath1sm ) with the intermediate nuclei @xmath0 in and @xmath1pm are studied within a microscopic approach based on a deformed proton - neutron quasiparticle random - phase approximation built on a skyrme selfconsistent mean field with pairing correlations and spin - isospin residual forces . the results are compared with the experimental gt strength distributions extracted from charge - exchange reactions . combining the two branches , the nuclear matrix elements for the two - neutrino double - beta decay are evaluated and compared to experimental values derived from the measured half - lives .
the presence of molecules in interstellar space was predicted as early as 1930 s ( e.g. , @xcite ) , quickly followed by identification of simple diatomic species , ch , cn , ch@xmath0 ( e.g. , @xcite ) , in optical absorption spectra . these early observations lead to important insights into the physics and chemistry of the interstellar medium ( e.g. , @xcite ) . however , the optical and uv observations are limited to diffuse or translucent clouds and applications of molecular spectroscopy to studies of the dense interstellar medium had to await the development of microwave detection techniques in the 1960s . since the first microwave detections of molecules in space @xcite , molecular spectroscopy has been a premier tool for studying interstellar gas cloud physics , from the milky way @xcite to the high - redshift universe @xcite . in order to use molecules as quantitative tracers of star - forming clouds , it is necessary to understand the radiative and collisional processes that govern the excitation of their internal quantum states @xcite . for decades , it has been theorized that , under interstellar conditions , some molecules form via highly - exothermic reactions between ions and neutral molecules @xcite . however , quantitative models of chemistry and of excitation have conventionally been constructed independently , on the grounds that chemical time - scales are long compared with excitation time - scales . the most abundant interstellar molecule , molecular hydrogen , is a symmetric rotor without a permanent dipole moment , which only has weak quadrupole rovibrational transitions in its ground electronic state ( e.g. , @xcite ) . its uv electronic transitions can only be studied from space and the observations are limited to the relatively nearby diffuse and translucent clouds . the infrared rovibrational lines are accessible from the ground , but are only excited in warm , active environments , such as shocks or photon dominated regions . consequently , trace molecules , with rich rotational spectra in the millimeter to far - infrared wavelength range have been used as proxies of h@xmath1 , to understand the physical conditions in the star - forming gas reservoir of the milky way and external galaxies . of these , oblate symmetric top molecules , in particular ammonia @xcite , for which the @xmath2 level is the lowest energy level in each @xmath3 ladder , are of great utility for the temperature determination . the population of these `` metastable '' levels is assumed to be thermalized at the kinetic temperature of the medium , due to the absence of allowed radiative transitions out of these levels . the hydronium ion , , is another oblate symmetric rotor , isoelectronic with ammonia , which like ammonia has the characteristic inversion splitting of its rotational levels . in the case of , however , the splitting is very large @xcite and the inversion transitions occur at far - infrared wavelengths , as opposed to centimeter wavelengths for ammonia . the far - infrared inversion transitions , up to ( 11,11 ) , have only recently been detected for the first time by _ herschel _ , in absorption toward the central region of the milky way @xcite . earlier hifi observations of the hydronium ion toward sagittarius b2 indicated a high rotational temperature , 500 k , at velocities corresponding to the cloud envelope . @xcite suggested that , in regions exposed to ionizing radiation , such as those at the center of the milky way , highly - excited states of the hydronium ion and ammonia are populated mainly by exoergic chemical formation processes and temperatures derived from their spectra do not represent the physical temperature of the medium . the detection of highly - excited states of the hydronium ion in nearby active galaxies , with rotational temperatures similar to that observed in sagittarius b2 @xcite , demonstrated the universality of the shape of the observed population diagrams in _ active _ environments . here we present new sensitive _ herschel _ observations of the hydronium ion on the lines of sight toward sagittarius b2(n ) and w31c ( g10.6 - 0.4 ) , which clearly demonstrate that this effect is not limited to the environments of galactic nuclei , characterized by high x - ray and cosmic ray fluxes , but is a _ fundamental _ property , pervasive throughout the interstellar medium . the inversion lines of were first discovered in absorption in the full spectral scans of sagittarius b2(m ) and ( n ) carried out between 2010 march and 2011 april using the heterodyne instrument for the far - infrared ( hifi ; @xcite ) on the herschel space observatory @xcite , within the framework of the hexos guaranteed time key program @xcite . preliminary analysis of the sagittarius b2(n ) data was presented by @xcite . the data presented here have been re - reduced using an improved version of the hifi pipeline , which results in significantly lower noise levels in the high - frequency heb mixer bands . the source coordinates are : sagittarius b2(m ) @xmath4 ; sagittarius b2(n ) @xmath5 . the line survey data have been calibrated with hipe version 10.0 @xcite and the resulting double - sideband ( dsb ) spectra were subsequently reduced using the gildas class software package . basic data reduction steps included removal of spurious features or otherwise unusable portions of the spectra . the continuum emission was then subtracted from the dsb scans by fitting a low - order ( typically 01 , in a few cases 2 ) polynomial . the continuum - subtracted dsb data were deconvolved ( sideband separation through pure @xmath6 minimization ; @xcite ) to provide a single - sideband ( ssb ) spectrum for each hifi band . the variation of the continuum intensity as a function of frequency was determined from the least - squares fit to the emission - free spectral channels across all hifi bands . the fitted continuum was subsequently folded back into the deconvolved , baseline subtracted spectra . the data reduction procedure ensures that saturated absorption features reach the zero continuum level and there are no discontinuities in the overlap regions between the various hifi bands . in addition to the full hifi spectral scans , we obtained deep follow - up integrations at the frequencies of the ( 6,6 ) and ( 9,9 ) inversion transitions of the hydronium ion toward sagittarius b2(n ) and w31c . these observations were carried out in 2012 october 417 in the framework of the special , final call for _ herschel _ observing proposals , referred to by esa as `` must - do '' programs . the source coordinates for w31c are @xmath7 and for sagittarius b2 are the same as the earlier hexos observations . the data were taken in the double beam switching ( dbs ) mode , with the reference beams 180 away from the source , using the hifi wideband spectrometer ( wbs ) , which provides a 1.1 . mhz frequency resolution over the 2.4 ghz bandwidth of the band 7 heb receivers . three shifted lo settings were observed , which were subsequently averaged together after inspecting the spectra for possible line contamination from image sideband ( no contamination was found ) . the total observing time for each of the two sources was 53 and 56 min for the ( 6,6 ) and ( 9,9 ) lines , respectively . spectra from the hifi heb mixer bands are known to be adversely affected by standing waves originating in the signal amplification chain . this problem is particularly apparent in long integrations , such as the deep sagittarius b2(n ) and w31c observations presented here , where the detector power levels can drift significantly , resulting in a disparity between the `` on '' and `` off '' phases of the observation . since the standing waves occur in the electrical chain , where the reflecting components have complex frequency - dependent impedance , the resulting baseline ripple has a non - sinusoidal behavior and it is difficult to remove using standard standing wave removal techniques @xcite . fortunately the standing wave phase and amplitude are stable for a given mismatch between the `` on '' and `` off '' phases . therefore , the spectra can be corrected by subtracting a standing wave profile from similarly mismatched template spectra . this approach has been applied here to the individual integrations , before they were averaged together to produce the final spectra . lcccccccc ( 1,1 ) & 1655.834 & @xmath8 & 6 & 6 & 0 & & & + @xmath9 & 984.712 & @xmath10 & 4 & 12 & 7.3 & @xmath11 & @xmath12 & @xmath13 + ( 2,2 ) & 1657.248 & @xmath14 & 10 & 10 & 42.7 & @xmath15 & @xmath16 & @xmath17 + ( 3,3 ) & 1663.587 & @xmath18 & 28 & 28 & 103.1 & @xmath19 & @xmath20 & @xmath21 + ( 4,4 ) & 1674.866 & @xmath22 & 18 & 18 & 181.2 & @xmath23 & @xmath24 & @xmath25 + ( 5,5 ) & 1691.134 & @xmath26 & 22 & 22 & 276.9 & @xmath27 & @xmath28 & @xmath29 + ( 6,6 ) & 1712.461 & @xmath30 & 52 & 52 & 390.2 & @xmath31 & @xmath32 & @xmath33 + ( 7,7 ) & 1738.936 & @xmath34 & 30 & 30 & 521.1 & @xmath35 & @xmath36 & @xmath37 + ( 8,8 ) & 1770.679 & @xmath38 & 34 & 34 & 669.5 & @xmath39 & @xmath40 & @xmath41 + ( 9,9 ) & 1807.826 & @xmath42 & 76 & 76 & 835.3 & @xmath31 & @xmath43 & @xmath44 + ( 10,10 ) & 1850.536 & @xmath45 & 42 & 42 & 1019 . & @xmath46 & @xmath47 & @xmath48 + ( 11,11 ) & 1898.992 & @xmath49 & 46 & 46 & 1219 . & @xmath50 & @xmath51 & @xmath52 lccccccc @xmath9 & @xmath53 & @xmath54 & @xmath55 & @xmath56 & @xmath57 & @xmath58 & @xmath59 + ( 2,2 ) & @xmath60 & @xmath61 & @xmath62 & @xmath63 & @xmath64 & @xmath65 & @xmath66 + ( 3,3 ) & @xmath67 & @xmath68 & @xmath69 & @xmath70 & @xmath71 & @xmath72 & @xmath73 + ( 4,4 ) & @xmath74 & @xmath75 & @xmath76 & @xmath77 & @xmath78 & @xmath79 & @xmath80 + ( 5,5 ) & @xmath81 & @xmath82 & @xmath83 & @xmath84 & @xmath85 & @xmath86 & @xmath87 + ( 6,6 ) & @xmath88 & @xmath89 & @xmath90 & @xmath91 & @xmath92 & @xmath93 & @xmath94 + ( 7,7 ) & @xmath95 & @xmath96 & @xmath97 & @xmath98 & @xmath99 & @xmath100 & @xmath101 + ( 8,8 ) & @xmath102 & @xmath103 & @xmath104 & @xmath105 & @xmath106 & @xmath107 & @xmath108 + ( 9,9 ) & @xmath109 & @xmath110 & @xmath111 & @xmath112 & @xmath113 & @xmath114 & @xmath115 + ( 10,10 ) & @xmath116 & @xmath117 & @xmath118 & @xmath119 & @xmath120 & @xmath121 & @xmath122 + ( 11,11 ) & @xmath123 & @xmath124 & @xmath125 & @xmath126 & @xmath127 & @xmath128 & @xmath129 continuum - divided spectra of the inversion lines toward sagittarius b2(m ) and ( n ) from the hexos spectral scans are shown in figure [ fig : spectra ] , upper and lower panels , respectively . the lower panel is a revised version of figure 3 of @xcite , including latest improvements in the data processing . the ( 1,1 ) line is not included in the analysis , as it is blended with the strong @xmath130 line of h@xmath131o . instead , the @xmath132 ground state ortho- spectrum is included ( see energy level diagram in figure 2 of @xcite ) . gaussian fit results for the two sources are listed in tables [ tab : fitm ] and [ tab : fitn ] . we use the latest line frequencies of @xcite , which in some cases differ by up to 25 mhz from those used in @xcite . the column densities in the metastable levels can be derived directly from the velocity integrated optical depths of the corresponding inversion transitions using the formula ( see , e.g. , eq . 3 of @xcite ) : @xmath133 where @xmath134 is the spontaneous emission coefficient , @xmath135 and @xmath136 the upper and lower level degeneracies , and @xmath137 the wavelength . spectroscopic constants for the observed transitions are included for completeness in table [ tab : fitm ] and the resulting rotation diagrams ( plots of the logarithm of the integrated line strength as a function of the lower level energy ; e.g. , @xcite ) for the two sources are also shown in figure [ fig : spectra ] . toward sagittarius b2(m ) , the spectrum is dominated by a single component centered around 64 , with a line width of 15 , associated with the cloud envelope . three velocity components , centered around 5 , 65 , and 80 , with line widths of 25 , 19 , and 11 respectively , can be identified in the sagittarius b2(n ) spectra . the strongest component at 65 is associated with the cloud envelope and contains 60% of the total column density . the spectra are not sensitive enough to produce independent rotational diagrams for the individual velocity components . therefore the three components have been added together to produce a single rotational diagram for this source . ) . ] the rotational diagrams for the two sources naturally separate into `` warm '' and `` hot '' components ( blue and red points in figure [ fig : spectra ] ; lower level energies below and above 200 k , respectively ) . rotational temperatures , determined separately for the warm and hot components from the least square fits to the data , are shown in figure [ fig : spectra ] with the corresponding uncertainties , which should be viewed as _ uncertainties due to systematics , such as baseline removal . we conservatively assume a 25% maximum uncertainty for the individual measurements . the warm component toward the two sources is well described by a rotational temperature of @xmath138 k , while the temperatures of the hot component toward the two sources are consistent with @xmath139 k ( maximum uncertainties ) . a two - component fit is only an approximation to the observed rotation diagram and a continuous range of temperatures between these limiting values is possible . however , one clear conclusion is that the population of the highest and lowest - energy metastable levels is described by _ different _ rotational temperatures , 175 and 485 k , respectively . total column densities toward sagittarius b2(n ) and ( m ) , obtained by adding up the population in the observed rotational levels , are @xmath140 and @xmath141 , respectively , with 40% of the total column density in the hot component . the extended envelope of sagittarius b2 , with a velocity of 65 has a column density of @xmath142 on the line of sight toward sagittarius b2(n ) . this is 50% higher than the corresponding value toward sagittarius b2(m ) . this , together with the fact that the 80 component that is quite prominent in the sagittarius b2(n ) spectra is not seen toward sagittarius b2(m ) , suggests that significant variations in the column density are present on linear scales of order 1.7 pc . if the absorption originated in a foreground shocked screen between the sun and sagittarius b2 ( e.g. , @xcite ) , one would not expect such large column density variations . the new sagittarius b2(n ) deep integrations are much more sensitive than the hexos spectral scans . the continuum - divided spectra of the ( 6,6 ) and ( 9,9 ) transitions are shown in figure [ fig : sgrb2 ] ( blue and red lines , respectively ; average of the h and v polarizations , equally weighted ; left vertical scale ) . absorption is clearly detected over a wide range of lsr velocities between about 120 and + 90 . the ( 9,9 ) absorption at the most negative velocities is contaminated by the strong @xmath143 line of ammonia in the same sideband , which may affect the derived line ratio and the corresponding rotational temperature at these velocities . it is not possible to remove the contamination , because the only other para- line with comparable line strength and lower level energy is itself blended and the intrinsic line profile can not be determined . lccc galactic center & 92 to 69 & @xmath144 & @xmath145 + norma arm & 47 to 13 & @xmath146 & @xmath147 + galactic center & 9 to 8 & @xmath148 & @xmath149 + scutum arm & 12 to 22 & @xmath150 & @xmath151 + sagittarius b2 & 47 to 89 & @xmath152 & @xmath153 black points in figure [ fig : sgrb2 ] show the ( 9,9)/(6,6 ) optical depth ratio in 1 channels ( left vertical scale ) . only points corresponding to velocities at which absorption in both transitions is detected above @xmath154 level in individual channels are plotted . the corresponding @xmath155 uncertainties are computed from the difference between the h and v polarization spectra , which are assumed to be independent measurements . the rotational temperature describing the population of these high - energy ( @xmath156400 k above the ground rotational state ) metastable levels is uniformly high , in the range 400700 k ( figure [ fig : sgrb2 ] ; right vertical scale ) . the differential rotation of the milky way allows , in principle , separating in velocity space spectral features from gas clouds at different galactocentric radii ( see , e.g. , @xcite , however , the distance determination and hence the assignment to specific spiral arms is complicated due to the streaming motions in the arms , @xcite ) . all velocity components that can be clearly identified in the sagittarius b2(n ) and ( m ) spectral scan data correspond to the galactic center gas . there is evidence for broad absorption in the ( 3,3 ) and ( 6,6 ) lines toward sagittarius b2(n ) , but the signal - to - noise ratio is limited . however , absorption is clearly detected over a broad range of velocities in the deep integrations toward this source ( figure [ fig : sgrb2 ] ) . velocities of the various foreground components on this line of sight are marked with horizontal color bars above the abscissa : green galactic center gas , magenta norma arm , cyan scutum arm , and yellow sagittarius arm . absorption at velocities of the norma and scutum arms ( magenta and cyan , respectively ) is clearly detected , in addition to the galactic center gas . averaged optical depth ratios and rotational temperatures for five velocity ranges corresponding to different kinematic components are listed in table [ tab : trot ] . determining the uncertainty of the velocity - averaged rotational temperature values is not trivial , as the channels are not independent due to the presence of baseline ripples in the spectra . this aspect has been discussed in detail in the context of the measurements of the water ortho / para ratio by @xcite . based on our previous experience with hifi heb data , we allow a factor of 2 correction for the uncertainty of the average line ratio to account for the correlation between the spectral channels . this leads to conservative maximum uncertainties for the line ratio and the rotational temperature listed in table [ tab : trot ] . we conclude that , within the observational uncertainties , the rotational temperature of the hot component is the same , 500 k , in regions at different galactocentric distances . as a comparison , the kinetic temperature in the foreground clouds , determined from the ( 1,1 ) and ( 2,2 ) observations , is in the range 1060 k @xcite . these low - evergy transitions , however , are not sensitive to the presence of high - temperature gas . lc ( 1,1 ) & @xmath157 + @xmath9 & @xmath157 + ( 2,2 ) & @xmath158 + ( 3,3 ) & @xmath159 + ( 6,6 ) & @xmath160 + ( 9,9 ) & @xmath161 + gas motions in the galactic center region are quite complex , since the velocity field is dominated by the non - axisymmetric bar potential . in particular the gas in the so - called `` x2 '' non - circular orbits may overlap in velocity with the foreground disk clouds . one thus can not exclude that the absorption at _ all _ velocities toward sagittarius b2 is predominantly associated with the gas in the central molecular zone . the observations of w31c , a line of sight that does not intersect the galactic center region , provides independent constraints . the ( 6,6 ) and ( 9,9 ) excited inversion lines of the hydronium ion are also detected in the deep hifi observations on this line of sight ( figure [ fig : w31c ] ) , which samples the diffuse interstellar medium along a 4 kpc line of sight through the galactic disk . the foreground absorption covers velocities between 10 and 50 @xcite , with a strong feature at 40 clearly seen in the spectra . the observed absorption line intensities are listed in table [ tab : fitw ] . the resulting rotational diagram ( figure [ fig : pop ] ) can again be described by two components with temperatures of 65 k ( + 21,13 ) k and 380 k ( + 210,100 ) k. the uncertainties are maximum uncertainties , assuming a 35% maximum uncertainty for the individual measurements ( a conservative estimate of the combined statistical and systematic uncertainties ) . although the uncertainty in the rotational temperature of the hot component is large , the lower bound is well constrained , demonstrating that the population of the excited metastable levels is described by a temperature that is significantly higher than that describing the low - energy metastable rotational levels , in agreement with the sagittarius b2 results . the 65 k temperature describing the population of the low - energy metastable levels is consistent with the kinetic temperature of the gas in the diffuse interstellar medium in the galactic disk @xcite . o and this transition can be used in the analysis . note that the vertical scale is different for the ( 6,6 ) and ( 9,9 ) lines . ] from the two - component fit to the rotation diagram , we estimate a total column density of @xmath162 toward w31c , with 20% in the hot component , as compared to 40% toward sagittarius b2(n ) and ( m ) . earlier ground - based @xcite and space - borne @xcite observations of ammonia absorption toward sagittarius b2 have been interpreted in terms of a physically hot component of molecular gas in the central region of the galaxy , in which the observed high rotational excitation temperature traces a correspondingly high kinetic temperature . at first glance this seems plausible , because shock waves in this active high - mass star - forming region should be able to maintain some molecular gas at temperatures of 500 to 700 k , or because the region contains luminous heating sources of x - rays and cosmic rays . moreover , strong x - ray emission in iron fluorescence at 6.4 kev has been taken as evidence that the envelope of sagittarius b2 was recently illuminated by an x - ray flash , attributable to a flare from the region around the central black hole that is associated with the sagittarius a@xmath163 radio continuum source @xcite . the sagittarius b2 x - ray flux is now fading @xcite , with a characteristic decay time of 8 years , comparable to the light - travel time across the central part of the molecular cloud . while the competing mechanisms put forward to explain the hot gas component on the line of sight toward sagittarius b2 are all plausible , a generally accepted explanation has yet to be offered . interestingly , recent observations of the ( 8,8 ) to ( 15,15 ) ammonia inversion lines @xcite indicate rotational temperatures of 350450 k in several largely _ quiescent _ galactic center giant molecular clouds . this indicates that the effect is not limited to sagittarius b2 , but widespread in the galactic center environment . the hot gas component traced by the high - energy ammonia inversion lines would thus have to fill a few hundred parsec size region . the energy input required to heat such a vast volume of gas would then have to be carefully evaluated . moreover , the diffuse , widespread , low - density gas component in the central molecular zone can be independently probed through the infrared absorption spectroscopy of the h@xmath164 rovibrational transition @xcite , which indicate temperatures of `` only '' 250 k. inversion lines of ammonia , up to ( 10,10 ) , have also been detected in absorption toward pks 1830 - 211 , indicating rotational temperatures 600 k for the highest - energy lines @xcite . the absorption is attributed to molecular gas in spiral arms of an ordinary galaxy at a redshift of 0.89 . explanations similar to those previously invoked to explain the high - rotational temperatures of ammonia can not easily be applied to h@xmath165o@xmath0 : if the density were high enough to thermalize the populations of highly excited rotational states at a kinetic temperature of 400 to 600 k , then molecular collisions would also raise the excitation temperatures of the high-@xmath166 inversion doublets to values higher than the brightness temperature of the continuum radiation at @xmath167 . in that case , the inversion lines could not appear in absorption as is observed . a fundamentally different interpretation is required to explain simultaneously the high rotational excitation of h@xmath165o@xmath0 , its short chemical lifetime , and its relatively large total column density , @xmath168 @xmath169 , close to an intense far - infrared continuum source . the characteristic two - component rotational diagram for metastable states of h@xmath165o@xmath0 naturally results from its formation by exoergic ion - neutral reactions . the excess enthalpy ( heat of formation ) of the main source reaction ( e.g. , @xcite ) @xmath170 goes partly into internal rotational excitation of the product ion , the highly excited non - metastable states relax rapidly by spontaneous radiative transitions to the metastable states ( i.e. the lower @xmath2 inversion levels ) , whose populations are limited by the rates of excited - molecule formation and chemical destruction . populations are partly re - distributed by inelastic collisions and by absorption of the background far - infrared continuum radiation . the populations of the lowest metastable states may approach a collision - dominated thermal distribution . the turnover to a `` hot '' formation - dominated distribution is detectable when the formation rate is a non - negligible fraction of the inelastic collision rates . the relation between molecular excitation and formation can be outlined quantitatively as follows . the destruction rate per h@xmath165o@xmath0 molecule by dissociative recombination with electrons can be written @xmath171 s@xmath172 , where the rate coefficient for the process @xmath173 @xmath174 s@xmath172 at kinetic temperature @xmath175 @xcite and @xmath176 is the number density of free electrons . at @xmath177 to @xmath178 @xmath179 , and @xmath180 to @xmath181 k , @xmath182 to @xmath183 s@xmath172 . this is fast enough to be the dominant destruction mechanism . if we furter assume steady state between destruction and formation , then the formation rate per unit voulme can be written as @xmath184 @xmath179s@xmath172 . with no further knowledge of the chemistry , we can assert that the source of h@xmath165o@xmath0 in the envelope of sagittarius b2 must be of the order of @xmath185 \bigl[{{n(e)}\over{0.1\;{\rm cm}^{-3}}}\bigr ] \;{\rm cm}^{-3}{\rm s}^{-1},\ ] ] where @xmath186 is the characteristic length scale of the absorbing region . the interstellar chemistry of oxygen - containing ions is thought to be straightforward @xcite , with cosmic - ray or x - ray ionizations of hydrogen leading to h@xmath0 and h@xmath164 , which transfer charge to o@xmath0 or oh@xmath0 . the oxygen ions form h@xmath1o@xmath0 and finally h@xmath165o@xmath0 via reactions with h@xmath1 . in regions where the atomic fraction is small , h / h@xmath187 , the chemical source rate of the terminal ion h@xmath165o@xmath0 can be a significant fraction of the hydrogen ionization rate . thus we can express @xmath188 , where @xmath189 is the ionization rate , @xmath190 is the density of hydrogen , and @xmath191 is a chemical loss factor that accounts for minor channels in the ion chemistry that prevent the hydrogen ions from processing solely the oxygen . such an analysis has been applied to herschel / hifi observations of oh@xmath0 and h@xmath1o@xmath0 in diffuse molecular gas by @xcite , who concluded that those highly reactive ions arise mostly in components of gas with a high atomic content , unlike , which traces predominantly molecular gas . following these arguments , we see that the observed column density and estimated destruction rate of h@xmath165o@xmath0 toward sagittarius b2 require @xmath192 for @xmath193 pc and @xmath194 cm@xmath172 in the envelope of sagittarius b2 . this could be achieved with a cosmic - ray ionization rate @xmath196 s@xmath172 at @xmath197 @xmath179 , or with less extreme values if the electron density is lower . cosmic ray ionization rates in excess of @xmath198 s@xmath172 have been deduced from infrared absorption observations of h@xmath164 on several lines of sight through the central molecular zone of the galaxy ( @xcite , and references therein ) . the line of sight to w31c presents a simpler test case because it lies entirely outside the central molecular zone , where conditions are extreme . for illustration , consider the low - density limit , in which collisional excitation is neglected and the rotational populations are governed entirely by the formation process and radiative transitions . let the state - specific formation rate of h@xmath165o@xmath0 be described by @xmath199 for each state of rotational quantum numbers @xmath166 , @xmath3 , and parity @xmath200 , where @xmath201 is the statistical weight , @xmath202 is the energy , @xmath203 is the partition function , and @xmath204 is a parameter the formation temperature . the populations of the vibration - rotation levels are then computed by solving the rate equations with formation , destruction , and all radiative transitions included ( equations 1215 of @xcite ) , but with all the collisional terms set to zero . the results of one example are compared with the observed population diagram in figure 4 . in this case , @xmath205 s@xmath172 ( corresponding to @xmath206 @xmath179 at @xmath207 k ) and the total column density is @xmath208 @xmath169 . the formation temperature is @xmath209 k and the molecules are exposed to an average galactic background continuum similar to that described by @xcite . at the adopted destruction rate , @xmath210 @xmath179 s@xmath172 is required to attain a good match in column density over a total path length @xmath211 pc . this rate , in turn , would imply a rate of cosmic - ray ionization of h@xmath1 , @xmath212 s@xmath172 for a density @xmath213 @xmath169 and @xmath214 . this ionization rate would be consistent with that derived by @xcite toward w51 irs2 if the h@xmath165o@xmath0 is found preferentially in regions of higher molecular fraction and higher @xmath191 than the observed oh@xmath0 . these conditions are also in harmony with photochemical models of diffuse molecular gas as discussed by @xcite . a combined excitation - chemistry model , which includes explicitly the exothermicity of the chemical reactions leading to the formation of the hydronium ion and ammonia is beyond the scope of the present paper and will be presented separately ( j. black , in prep . ) such calculations are clearly needed to demonstrate that chemistry and excitation are closely coupled in some interstellar environments and should not be treated separately and to use as a quantitative tracer of the cosmic ray ionization rate . this aspect is of particular interest for understanding the energetics of the nuclear regions of ultraluminous , star - forming galaxies . highly - excited inversion lines of the hydronium ion have now been detected in the ultraluminous infrared galaxy arp 220 , as well as in ngc 4418 , which harbors a deeply buried , compact agn @xcite . in these environments , the hydronium ion absorption traces a relatively low - density molecular gas with a large filling factor , which is directly influenced by the feedback process of the central engine . the traditional explanation for these observations invokes a new phase of the interstellar medium , which would have to be heated by some combinations of x - rays , cosmic rays , shocks , or turbulence to 500 k. if the chemical formation pumping explanation is correct , this additional energy input is not required the observations can simply be explained by chemical formation pumping in a much lower temperature medium . this relaxes the heating requirements and simplifies the physical picture of the interstellar medium in these active environments . with the improved understanding of the excitation , the far - infrared inversion lines of the hydronium ion also provide an accurate measurement of the ionization rate in the mostly molecular gas component of high - redshift galaxies , using alma , highly complementary to the oh@xmath0 and h@xmath1o@xmath0 observations that primarily trace atomic gas , with low molecular fraction @xcite . symmetric top molecules have been proven to be some of the best tracers of the gas kinetic temperature in the interstellar medium @xcite . however , the results presented here suggest that they may sometimes fail as an interstellar thermometer , under specific conditions . this typically involves low - density , diffuse gas , as collisions and strong radiation field tend to drive the rotational level population to equilibrium conditions . consequently , care has to be taken when interpreting the derived rotational temperatures as physical temperatures of the medium . hifi has been designed and built by a consortium of institutes and university departments from across europe , canada and the united states ( nasa ) under the leadership of sron , netherlands institute for space research , groningen , the netherlands , and with major contributions from germany , france and the us . support for this work was provided by nasa ( _ herschel _ ot funding ) through an award issued by jpl / caltech . the research of ps , cc , and rh is supported by the collaborative research center 956 funded by the deutsche forschungsgemeinschaft ( dfg ) .	we present new _ _ herschel _ _ observations of the ( 6,6 ) , but we follow here the notation commonly used in the ammonia literature . ] and ( 9,9 ) inversion transitions of the hydronium ion toward sagittarius b2(n ) and w31c . sensitive observations toward sagittarius b2(n ) show that the high , 500 k , rotational temperatures characterizing the population of the highly - excited metastable rotational levels are present over a wide range of velocities corresponding to the sagittarius b2 envelope , as well as the foreground gas clouds between the sun and the source . observations of the same lines toward w31c , a line of sight that does not intersect the central molecular zone , but instead traces quiescent gas in the galactic disk , also imply a high rotational temperature of 380 k , well in excess of the kinetic temperature of the diffuse galactic interstellar medium . while it is plausible that some fraction of the molecular gas may be heated to such high temperatures in the active environment of the galactic center , characterized by high x - ray and cosmic ray fluxes , shocks and high degree of turbulence , this is unlikely in the largely quiescent environment of the galactic disk clouds . we suggest instead that the highly - excited states of the hydronium ion are populated mainly by exoergic chemical formation processes and temperature describing the rotational level population does not represent the physical temperature of the medium . the same arguments may be applicable to other symmetric top rotors , such as ammonia . this offers a simple explanation to the long - standing puzzle of the presence of a pervasive , hot molecular gas component in the central region of the milky way . moreover , our observations suggest that this is a universal process , not limited to the active environments associated with galactic nuclei .
the nature and detection of dark matter is one of the most pressing unsolved problems in modern physics . a myriad of cosmological observations indicate that @xmath030% of the energy density of the universe must be comprised of a cold and non - baryonic component , yet the particle content of this dark matter remains unknown @xcite . the most promising method of detecting dark matter in the laboratory is the search for their kev - scale nuclear recoils produced in elastic scattering events between dark matter particles in the milky way halo and target nuclei @xcite . this method of detection is possible if dark matter is in the form of a weakly interacting massive particle ( wimp ) . these particles are a popular and well motivated candidate which appear in extensions to the standard model such as supersymmetry , and freeze out in the early universe with an abundance that matches cosmological observations ( for reviews see e.g. , refs . @xcite ) . neutrinos are also weakly interacting particles . it is known that they must , too , elastically scatter off the same target nuclei of dark matter detectors . as a result of neutrinos being impossible to shield against , they are regarded as the ultimate background to experimental searches for wimps @xcite . current experiments such as xenon100 @xcite , lux @xcite , and cdms @xcite , which can probe to spin - independent ( si ) wimp - nucleon cross sections of the order @xmath1 , are not yet sensitive to the expected neutrino background ( for a recent review of direct detection experiments see e.g. , ref . however as the sensitivity of experiments increases with the next generation of ton - scale ( and beyond ) detectors such as xenon1 t @xcite and lz @xcite , the neutrino background will begin to become important @xcite . the limiting wimp parameters at which this occurs is known as the neutrino floor and is caused by a close similarity between the recoil energies and rates of wimps of certain masses and cross sections @xcite . for example in a xenon detector the recoil energy spectrum of a wimp with mass @xmath2 and si cross section @xmath3 very closely matches that of @xmath4 solar neutrinos @xcite . the neutrino floor is not however the true final limit to direct detection . as initially shown by ruppin @xcite , the differences in the tails of the recoil energy distributions of wimps and neutrinos allow the `` floor '' to be overcome albeit with very large numbers of events ( @xmath5 ) . furthermore other studies have shown that the neutrino background can be mitigated with the use of annual modulation effects @xcite , direction dependence @xcite or complementarity between multiple target nuclei @xcite . a recent work by dent @xcite made use of the non - relativistic effective field theory formalism which posits additional operators to describe the nuclear response to a wimp interaction . they found that for many of these additional operators , which induce significantly different nuclear recoil energy spectra , the neutrino bound is much weaker or not present . however the limits calculated by these studies all depend on the choice of astrophysical input , hence if an accurate prediction is to be made about when future experiments will be affected by the neutrino background at a statistically significant level , we must first establish the extent to which the uncertainty in the astrophysical input plays a role . the phase space structure of the milky way halo in the region of the solar system is an uncertain and much debated topic @xcite . most direct detection analyses use an isotropic and isothermal spherically symmetric assumption for the galactic halo known as the standard halo model ( shm ) . this model gives rise to an isotropic maxwellian velocity distribution for which there is an analytic form for the wimp event rate . however substantial evidence from n - body and hydrodynamic simulations indicate that the velocity structure is likely to contain significant deviations from this simple maxwellian form @xcite . in extreme cases there is some evidence that the velocity distribution may contain highly anisotropic features such as tidal streams @xcite , a dark disk @xcite and debris flows @xcite which have been shown to have a noticeable effect on direct detection signals @xcite . to add further complication , the astrophysical parameters not related to the velocity distribution directly such as the laboratory velocity @xmath6 and local density @xmath7 are also subject to a high degree of uncertainty @xcite . particularly in the case of the local density this is a source of concern as it is degenerate with the wimp - nucleon cross section . attempts have been made however to account for astrophysical uncertainties in direct detection analysis by calculating halo - independent discovery limits @xcite . in this work we will incorporate neutrino backgrounds into an analysis which will take into account these astrophysical uncertainties . the outline of this paper is as follows , in sec . [ sec : theory ] we present the nuclear recoil event rates in the case of wimp and neutrino elastic scattering . then in sec . [ sec : nufloor ] we describe the mock detector and outline the profile likelihood ratio test employed in order to calculate the neutrino floor . in sec . [ sec : astro ] we describe the various sources of astrophysical uncertainty and how estimates on each individual parameter and different speed distributions change the neutrino floor . section [ sec : parcon ] deals with the effect of the same uncertainties on the parameter constraints possible in the event that a wimp signal is detected . in sec . [ sec : future ] we discuss the potential for the upcoming generation of direct detection experiments to detect coherent neutrino - nucleus scattering , and in sec . [ sec : time ] we show how it is possible to go beyond the putative neutrino floor by including the time dependence of the wimp and solar neutrino event rates ; in both sections we emphasise the importance of understanding astrophysical uncertainties . finally in sec . [ sec : conc ] we summarise our results and conclude . the elastic scattering event rate as a function of recoil energy and time assuming spin - independent interactions with identical couplings to protons and neutrons , is given by @xcite , @xmath8 where @xmath9 is the wimp mass , @xmath10 is the wimp - nucleon reduced mass , @xmath7 is the local dark matter density , @xmath11 is the mass number of the target and @xmath12 is the wimp - nucleon cross section . the function @xmath13 is the form factor of the nucleus which only has a noticeable effect at large wimp masses . in this work we will primarily be concerned with light wimps where understanding the neutrino background is most important , hence we will assume the standard helm form factor for simplicity . finally , @xmath14 is the mean inverse speed which is calculated by integrating the velocity distribution @xmath15 from @xmath16 - the minimum wimp speed required to produce a nuclear recoil of energy @xmath17 , @xmath18 the lab velocity , @xmath19 , is the velocity of the observer relative to the rest frame of the galaxy and if taken to be time dependent is responsible for the annual modulation of the event rate @xcite . for the kev - scale nuclear recoils relevent for wimp detection the most important background to consider from neutrinos is due to coherent neutrino - nucleus scattering ( cns ) . we will ignore neutrino - electron elastic scattering , which for direct detection experiments has a very small event rate ( from @xmath20 neutrinos ) and only slightly adjusts the neutrino floor for wimp masses larger than @xmath0100 gev @xcite . despite the fact that cns is yet to be observed , the rate is fully predicted by the standard model @xcite . the differential cross section as a function of nuclear recoil energy ( @xmath17 ) and neutrino energy ( @xmath21 ) is , @xmath22 where @xmath23 is the weak nuclear hypercharge of a nucleus with @xmath24 neutrons and @xmath25 protons , @xmath26 is the fermi coupling constant , @xmath27 is the weak mixing angle and @xmath28 is the nucleus mass . the event rate per unit mass , as a function of the recoil energy is found by integrating the differential cross section with the neutrino flux from @xmath29 , which is the minimum neutrino energy required to generate a nuclear recoil with energy @xmath17 , @xmath30 the neutrino flux @xmath31 is the sum of multiple different components each with different individual energies and uncertainties . the relevent contributions to the neutrino background for wimp searches are displayed in the left hand of fig . [ fig : neutrino_flux ] with uncertainties listed in table [ tab : neutrino ] . in fact with advances in technology currently underway @xcite it will be possible for direct detection experiments to make competitive measurements of these neutrino fluxes @xcite and even constrain new physics such as the existence of sterile neutrinos @xcite or new interactions between neutrinos and nuclei or electrons @xcite . neutrinos from various fusion reactions in the interior of the sun dominate the low energy - high flux regime and are the dominant background for direct detection with a total flux at earth of around @xmath32 @xmath33 s@xmath34 @xcite . neutrinos from the initial @xmath20 reaction make up 86% of all solar neutrinos and have been detected most recently by the borexino experiment , determining the flux with an uncertainty of @xmath35% @xcite . however for the remaining solar neutrinos the theoretical uncertainties in the fluxes are as large as or larger than the measurement uncertainty and rely on an assumption of a solar model for their calculation . in this work we assume the high metallicity standard solar model ( ssm ) @xcite which is the model most consistent with existing solar data . due to their relatively low energies , solar neutrinos will influence the detection of wimps with masses less than 10 gev . for wimp masses between 10 and 30 gev , the neutrino floor is caused by the sub - dominant diffuse supernova neutrino background ( dsnb ) , the sum total of all neutrinos emitted from supernovae over the history of the universe . the background flux is calculated by performing a line of sight integral of the spectrum of neutrinos from a single supernova with the rate density of core - collapse supernovae as a function of redshift . see ref . @xcite for the full calculation of the predicted dsnb . the total flux of the dsnb is considerably smaller than for solar neutrinos , around 86 @xmath33 s@xmath34 , however it is an important background to consider as it extends to a higher energy range not occupied by solar neutrinos . the calculated spectra have a fermi - dirac form with temperatures in the range 3 to 8 mev . in this study we use a dsnb flux which is the sum of 3 temperatures : 3 and 5 mev for electron and anti - electron neutrinos respectively , and 8 mev for the sum of the remaining neutrino flavours . there are considerable theoretical uncertainties in this calculation , hence we will take a large systematic uncertainty of 50% on the total flux of dsnb neutrinos @xcite . the final type of neutrino remaining to be discussed are those from the atmosphere which provide the main neutrino background for wimp masses above 100 gev . these neutrinos occupy the high energy and low flux regime and will limit the sensitivity of experiments to spin - independent cross sections below around @xmath36@xmath37 @xcite . the flux of atmospheric neutrinos with energies less than 100 mev is difficulat to measure as well as predict theoretically @xcite although the expected flux is around 11 @xmath33 s@xmath34 . in this work we use a calculation that is a sum of the contributions from electron , anti - electron , muon and anti - muon neutrinos and place a @xmath38 uncertainty on the total flux @xcite . in the right hand panel of fig . [ fig : neutrino_flux ] we show the recoil energy spectrum for each neutrino type scattering with a xenon target . in addition we show the recoil energy spectra for two example wimps with masses of 6 gev and 100 gev . this is to demonstrate the similarity that certain wimp masses have with individual neutrino sources . this overlapping between wimp and neutrino event rates is the reason why neutrinos limit wimp discovery . for cross sections below the neutrino floor , an experiment which observes the excess in the number of observed events over the expected background can not determine whether these events were due to a wimp interaction or a slightly larger value of neutrino flux due to the systematic uncertainty . hence the neutrino floor limit divides the wimp parameter space into cross sections which induce enough events to be significant over the systematic uncertainty on the neutrino background and those which do not . .total neutrino fluxes with corresponding uncertainties . the maximum neutrino energy , @xmath39 , and maximum recoil energy of a xenon target , @xmath40 , are also shown.[tab : neutrino ] [ cols="^,^,^,^ " , ] before continuing with the complete wimp+neutrino analysis we can gain insight into the influence of each source of neutrino on the experiment s overall sensitivity to the astrophysical parameters by performing a wimp - only analysis on datasets comprised of a single neutrino source . the wimp - only hypothesis consists of a likelihood of the form , @xmath41 where @xmath42 is the expected number of wimp events , defined in eq . ( [ eq : nchi ] ) . in fig . [ fig : wimponly ] we show the results of a wimp - only analysis on each neutrino component individually . the exposure in each case was chosen so that the experiment observes 500 neutrino events above a threshold of 1 ev . in the left hand panel of fig . [ fig : wimponly ] one can observe the wimp masses that are fitted in the astrophysics fixed case ( filled contours ) match the positions of each peak in the neutrino floor seen in fig . [ fig : el_vs_mass_allneutrinos ] . with astrophysics parameters , @xmath7 , @xmath43 , @xmath44 unfixed , the interpreted wimp parameters are quite different , as seen in the unfilled contours . the neutrino contributions when fitted under the this model that are least sensitive to changes in the astrophysical parameters appear to be @xmath45 , @xmath46b , @xmath47n , @xmath48o , @xmath49f and atmospheric neutrinos because the contours in the @xmath50 plane do not change location when the astrophysical parameters are unfixed , and the contours in the @xmath51 plane are centered on the fiducial values . the @xmath52be , @xmath20 and @xmath53 neutrinos seem to prefer slightly larger values of @xmath44 and the dsnb seems to fit to a slightly smaller value of @xmath44 . these contours tell us how well each neutrino component mimics a wimp signal . we can also observe that once astrophysical uncertainties are accounted for , the allowed range of cross section and wimp mass becomes much larger , which is consistent with the result of fig . [ fig : dl_uncertainties ] . one feature that we will note now as it will become important in subsequent results is the correlation between reconstructed wimp mass and cross section for large values of @xmath54 ( @xmath55100 gev ) . this can be seen in the contours with atmospheric neutrino data . in this regime the energy dependence of wimp event rate has a much weaker sensitivity to changes in the wimp mass . at large wimp mass the event rate is both proportional to cross section and inversely proportional to wimp mass ( due to the constant @xmath7 , larger @xmath54 implies lower wimp number density ) . these dependencies give rise to the positive correlation between wimp mass and cross section in the contours at 100 gev . this phenomenon has been explored in more detail in previous studies on the reconstruction of wimp properties e.g. , refs . @xcite . following the wimp only analysis we now return to the full wimp+neutrino likelihood and discuss how the parameters of this model can be reconstructed using a bayesian approach . the 2d marginalised posterior distributions in the wimp mass - cross section plane for a range of input values are shown in fig . [ fig : parameter_reconstruction ] ( a full set of 2d marginalised distributions for all parameters are given in the appendix [ appendix1 ] ) . here we choose a range of input wimp masses and show the constraints recovered when the input cross section is large enough to produce 2000 wimp events ( green ) or lies just above the discovery limit derived in sec . [ sec : nufloor ] ( blue ) . in both cases we use an exposure of 1 ton - year . when the wimp mass lies just above the discovery limit the recovered constraints are very wide . particularly for larger wimp masses ( above 200 gev ) when only 3 - 4 wimp events are observed the constraints span over an order of magnitude in @xmath54 and @xmath12 . a feature that can be seen particularly in the constraints at 0.4 , 10 and 200 gev , is the 68% and 95% regions follow the shape of the discovery limit . this is because the discovery limit defines the region of the parameter space above which the wimp events can not be interpreted as fluctuations around the systematic uncertainty in the various neutrino fluxes . hence when fitting the data to wimp parameters the allowed regions should roughly follow the values permitted by the discovery limit . this effect can be seen in a slightly different way in the green contours which correspond to wimp masses and cross sections that give 2000 wimp events . here the contours get noticably wider at low wimp masses when the neutrino events mimic the wimp signal . comparing the left and right hand panels of fig . [ fig : parameter_reconstruction ] we observe the effect of unfixing the values of the astrophysical parameters . we see a large increase in the size of the contours in both wimp mass and cross section , particularly for light wimps ( @xmath56 10 gev ) . as in the previous section we interpret this as being a result of the fact that when the value of @xmath44 is allowed to vary it allows a greater range of wimp masses to explain recoil energies over a given energy range as well as a greater range of cross sections to explain the size of the event rate . the contours are also rounder and do not follow the shape of the discovery limit as closely , this is thanks to the freedom offered by a wider range of allowed parameter values to explain the data . in fig . [ fig : error_recon ] we show the 68% error in the reconstructed values of wimp mass and cross section calculated using the profile likelihood . in this case we show the dependence of this error over a range of wimp masses from 0.1 to 1000 gev where in each case the input cross section is chosen to produce a fixed number of wimp events ( 100 , 300 or 500 ) . here we see a similar effect to the upper contours in fig . [ fig : parameter_reconstruction ] but in greater detail . we can see now that at wimp masses which match the recoil energy range of the underlying background , for instance 6 gev and @xmath46b neutrinos , the increase in uncertainty in cross section is coincident with a decrease in the uncertainty in wimp mass . when @xmath44 , @xmath7 and @xmath43 are unfixed however both wimp mass and cross section suffer an increase in error of up to a factor of 10 over the full range of input values . when these parameters are allowed to vary , less of the parameter space is mimicked by the background and hence there is more freedom in the allowed values , and correspondingly a larger uncertainty in the reconstructed @xmath54 and @xmath12 . finally in fig . [ fig : error_recon_maxwell_errs ] we show the same 68% profile likelihood errors in the remaining wimp parameters : @xmath7 , @xmath44 and @xmath43 . for @xmath7 and @xmath43 the error in their reconstructed values seems to only very marginally improve with increasing wimp event number , @xmath57 , as many more events than 600 are needed to significantly improve constraints on these parameters . although around 0.2 gev there is a rise in the error on @xmath7 as well as a slight decrease in the error on @xmath43 . the former can be understood to be a result of the sharp increase in the input cross section at very low wimp masses , and the latter effect being a result of the recoil energies detectable at these masses being much more sensitive to the escape velocity . for the error on the reconstructed value of @xmath44 however we can observe an interesting dependency on the input value of @xmath54 : rising around 100 gev and reaching a minimum around 6 gev . in this section we have seen that the neutrino background has a detrimental effect on the potential wimp parameter constraints that might be achievable with future detectors . at wimp masses which are mimicked by individual neutrino backgrounds there is a large error in the reconstructed cross section induced by the systematic uncertainty in the neutrino flux . this problem remains when an estimation of the ingredient astrophysical parameters is attempted , and this is coupled with a large increase in the error in the reconstructed wimp parameters even at masses which are not mimicked by neutrino backgrounds . future direct detection experiments such as supercdms @xcite , xenon1 t @xcite and lz @xcite , are poised to make the first detection of coherent neutrino - nucleus scattering . in this work so far we have used a very small threshold of 3 ev . this allowed us to map out the neutrino floor at low wimp masses whilst keeping other experimental parameters such as the target nucleus and exposures consistent . one might argue however that this is an unrealistic expectation for future experiments , including those using a dual - phase liquid xenon detector . the next generation of experiments are predicted to be sensitive to neutrino backgrounds with thresholds larger than those used in this work - once the energy resolution of the detector is taken into account . for example with a xenon target , the maximum recoil energy from @xmath46b neutrinos is 4.5 kev . however with an energy resolution of say 0.5 kev it is expected that some @xmath46b xenon recoils will leak into the energy sensitive window even with @xmath58 kev , provided @xmath59 is nt very much larger than 4.5 kev . the energy resolution is taken into account by convolving the event rate with a gaussian resolution function defined by an energy dependent resolution @xmath60 , @xmath61 the results we will show in this section are for a 2 kev threshold xenon detector with a 10 ton target mass over 1000 days running time , which is a reasonable approximation to what is expected in the near future beyond experiments such as lz and xenon1 t . the energy resolution we take to be constant 80% at 1@xmath62 over the full energy range , i.e. @xmath63 . figure [ fig : dl_resolution ] shows the discovery limit for this detector when neutrinos and astrophysical uncertainties are included . we show results for three sets of values of the uncertainties and the limits when energy resolution is both included and ignored . as previously mentioned , the finite energy resolution at low wimp masses causes some recoils to leak into the energy sensitive window and hence the discovery limits below 10 gev are lower than in the case of perfect resolution . this is interesting when neutrinos are included as it means that more @xmath46b events are observed and the neutrino floor emerges around 6 gev . as found in the results of fig . [ fig : dl_uncertainties ] when the uncertainty on @xmath44 is larger , the floor appears at larger wimp masses for the same reasons as discussed in sec . [ sec : astro ] . in this case we see that for the largest uncertainties the discovery limits lie extremely close to the currently excluded region by lux . the opposing effect of energy resolution however is that for larger wimp masses some events are pushed _ below _ the energy threshold meaning that the discovery limit is raised to larger cross sections in this range , this can be seen for wimp masses greater than 10 gev . to reiterate the conclusion of sec . [ sec : astrounc ] we state that the neutrino floor may be encountered by future direct detection experiments much sooner than previously thought . even without the ultra - low thresholds used in previous sections , the neutrino floor still limits wimp discovery when a finite energy resolution and modest threshold is considered . it was shown in ref . @xcite that the time dependence of the wimp and solar neutrino event rates provides a distinguishing feature between the two signals which can help circumvent the neutrino floor . the wimp signal is time dependent because of a well known annual modulation effect due to the motion of the earth with respect to the sun @xcite . we can insert this time dependence into the wimp calculation by simply including the additonal velocity @xmath64 in the lab velocity @xmath6 . details on the time dependence of this velocity component can be found in ref . @xcite . the solar neutrino flux also exhibits an annual modulation in the event rate due to the eccentricity of the earth s orbit which causes a change in the earth - sun distance over the course of a year . the time dependence can be written , @xmath65 \ , , \label{eq : solarneutrinoflux}\ ] ] where @xmath66 is the time from january 1st , @xmath67 is the eccentricity of the earth s orbit , @xmath68 days is the time at which the earth - sun distance is shortest ( and hence the solar neutrino flux is largest ) and @xmath69 year . both the solar neutrino and wimp event rates have a @xmath70% annual modulation but they peak at times separated by about 5 months . we now perform the energy+time analysis by extending the likelihood , @xmath71 where now we bin in both energy and time with @xmath72 and @xmath73 bins respectively . the number of wimp events in bin @xmath74 is , @xmath75 and @xmath76 is the number of expected neutrino events from the @xmath77th neutrino species , @xmath78 where @xmath79 is the mass of the detector . in fig . [ fig : dl_time ] we show the the improvement on the discovery limits obtained for wimp masses which are mimicked best by solar neutrino backgrounds . the improvement is most noticeable between 0.4 - 1 gev when the @xmath52be , @xmath53 , @xmath47n , @xmath48o and @xmath49f neutrinos play the biggest role , as well as at 6 gev when the floor is induced by @xmath46b neutrinos . outside of these specific mass ranges when the event rate energy dependencies do not overlap the improvement offered by time information is very small . above 10 gev for example ( for exposures with a very small atmospheric and supernovae background rates ) the discovery limit cross section is simply set by poisson statistics at a value which produces a sufficient number of wimp events to be significant over the systematic uncertainty on the total background flux . moreover , because the annual modulation amplitudes are small , to observe the benefit of time information one needs to go to large exposures which see in excess of @xmath80 neutrino events . incorporating uncertainties on the shm parameters into the energy+time analysis we obtain limits shown in fig . [ fig : dl_time_astro ] . the results we obtain here are analogous to those of the fig . [ fig : dl_uncertainties ] only now that we are working with a larger exposure which sees a number of events sufficient to discriminate neutrinos and wimps using their respective time dependencies . again the discovery limit under the assumption of the largest values of uncertainty is up to an order of magnitude higher than the astrophysics fixed case . however it still remains below the energy only limit around the peaks due to the solar neutrino contributions . as we have performed this analysis at a very large detector mass of @xmath81 tons , we do not see the same proximity to the lux region as in figs . [ fig : dl_uncertainties ] and [ fig : dl_time_astro ] . for smaller exposures , closer to those used in previous results , we do not expect any extra discrimination power with the addition of time information . in this work we have demonstrated the impact of astrophysics uncertainties in the analysis of wimp and neutrino data . we have used two different statistical approaches to both evaluate the discoverability of a wimp signal by a given experiment as well as demonstrating the ability of an experiment to measure the underlying parameters of the wimp and neutrino signals . relaxing the assumption of perfectly known astrophysics parameters such as the solar velocity , escape speed and local wimp density results in a shift in the shape of the neutrino floor as a function of wimp mass and cross section . we find that if there are reasonably large uncertainties in the various astrophysics parameters ( close to those currently known ) then neutrinos will become important to future direct detection experiments sooner than previously realised . when attempting to reconstruct the input wimp and neutrino parameters we find that unfixing the astrophysics parameters induces a significant increase in the uncertainty of the reconstruction . when the astrophysical parameters are fixed and only wimp mass and cross section are reconstructed we see that for input wimp masses which suffer the most overlap in recoil energies with different neutrino contributions there is trade off between a small error in mass with a large error in reconstructed cross section . this error in cross section is almost entirely controlled by the systematic uncertainty in the relevant neutrino flux , whereas the error in wimp mass is small because of the similarity between the recoil energy ranges of wimp and neutrino induced recoils . for this reason then we see that including uncertainty in the value of the astrophysical parameters , which in turn induces an uncertainty in the allowed range of recoil energies , there is a large increase in the error of the reconstructed wimp mass . for the remaining parameters we see that the error on the reconstructed values has little dependence on the input wimp parameters , with only @xmath44 having a noticeable relationship by being the parameter that most affects the shape of the wimp recoil spectrum . the first detection of coherent neutrino - nucleus scattering is expected to be made with the forthcoming generation of ton - scale direct detection experiments @xcite . when this occurs it will be crucial to begin to implement strategies for dealing with neutrino backgrounds . this can be achieved in a number ways . as can be seen in this work , as well as that of ref . @xcite the number of events observed at these detector masses are not yet enough to utilise the time dependence of the wimp and neutrino signals to discriminate the two . for spin - dependent interactions as well as non - relativistic effective field theory operators , complementarity between target materials will be a powerful and relatively easy method for discriminating neutrinos @xcite . independent of the wimp - nucleus interaction however , directional detection , if experimentally feasible , will prove the most powerful scheme for distinguishing wimps and neutrinos @xcite . the angular signature of wimp and neutrino recoils are entirely distinct and this is true for any relationship set of astrophysical inputs or wimp - nucleus interactions . however for the upcoming generation of direct detection experiments which will lack sensitivity to either direction or time dependence , we have shown that a better understanding of the uncertainty in the astrophysical dependence of a predicted wimp signal will be vital to understand in order to deal with the neutrino background . the author thanks anne m. green for helpful discussion and comments . the author is supported by the science & technology facilities council ( stfc ) . in this appendix we show the full 2d marginalised constraints on each parameter in both the initial case with the astrophysical parameters fixed ( top of fig . [ fig : parameter_reconstruction_triangle ] ) , and then in the case where we allow the parameters to vary under a gaussian prior ( bottom of fig . [ fig : parameter_reconstruction_triangle ] ) . the prior ranges are shown in table [ tab : priors ] . in both cases we use asimov data as a representative to remove the dependence on the stochastic nature of an individual dataset , however we do not expect there to much dependence on this choice due to the large number of events observed in this example . for the wimp mass we pick 6 gev with a cross section that lies just above the discovery limit so there is maximum overlap between the wimp signal and the @xmath46b signal . in both cases the majority of the parameters are recovered accurately and with little to no correlation between them . there is one notable exception in the astrophysics fixed case between cross section and the @xmath46b flux , as is to be expected when the signal and background event rates are extremely similar . in the second case however we note two more additional correlations , a degeneracy between @xmath7 and @xmath12 because they both appear as coefficients in the wimp event rate and a negative correlation between @xmath44 and @xmath54 because wimp event number is both proportional to @xmath44 and inversely proportional to @xmath54 at low wimp masses ( i.e. , with recoil energies close to the threshold ) . e. aprile _ et al . _ [ xenon100 collaboration ] , phys . * 109 * ( 2012 ) 181301 [ arxiv:1207.5988 [ astro-ph.co ] ] . d. s. akerib _ et al . _ [ lux collaboration ] , phys . * 112 * ( 2014 ) 091303 [ arxiv:1310.8214 [ astro-ph.co ] ] . r. agnese _ et al . _ [ supercdms collaboration ] , phys . * 112 * ( 2014 ) no.24 , 241302 [ arxiv:1402.7137 [ hep - ex ] ] . t. marrodn undagoitia and l. rauch , j. phys . g * 43 * ( 2016 ) no.1 , 013001 [ arxiv:1509.08767 [ physics.ins-det ] ] . e. aprile _ et al . _ [ xenon collaboration ] , [ arxiv:1512.07501 [ physics.ins-det ] ] . d. s. akerib _ et al . _ [ lz collaboration ] , arxiv:1509.02910 [ physics.ins-det ] . b. cabrera , l. m. krauss and f. wilczek , phys . * 55 * ( 1985 ) 25 . a. gutlein _ et al . _ , astropart . * 34 * ( 2010 ) 90 [ arxiv:1003.5530 [ hep - ph ] ] . j. billard , l. strigari and e. figueroa - feliciano , phys . rev . d * 89 * ( 2014 ) no.2 , 023524 [ arxiv:1307.5458 [ hep - ph ] ] . j. h. davis , jcap * 1503 * ( 2015 ) 012 [ arxiv:1412.1475 [ hep - ph ] ] . grothaus , m. fairbairn and j. monroe , phys . d * 90 * ( 2014 ) no.5 , 055018 [ arxiv:1406.5047 [ hep - ph ] ] . c. a. j. ohare , a. m. green , j. billard , e. figueroa - feliciano and l. e. strigari , phys . d * 92 * ( 2015 ) 6 , 063518 [ arxiv:1505.08061 [ astro-ph.co ] ] . f. ruppin , j. billard , e. figueroa - feliciano and l. strigari , phys . d * 90 * ( 2014 ) no.8 , 083510 doi:10.1103/physrevd.90.083510 [ arxiv:1408.3581 [ hep - ph ] ] . j. b. dent , b. dutta , j. l. newstead and l. e. strigari , arxiv:1602.05300 [ hep - ph ] . m. vogelsberger _ et al . _ , mon . not . * 395 * ( 2009 ) 797 [ arxiv:0812.0362 [ astro - ph ] ] . m. kuhlen , n. weiner , j. diemand , p. madau , b. moore , d. potter , j. stadel and m. zemp , jcap * 1002 * ( 2010 ) 030 [ arxiv:0912.2358 [ astro-ph.ga ] ] . y. y. mao , l. e. strigari , r. h. wechsler , h. y. wu and o. hahn , astrophys . j. * 764 * ( 2013 ) 35 [ arxiv:1210.2721 [ astro-ph.co ] ] . k. freese , p. gondolo and h. j. newberg , phys . d * 71 * ( 2005 ) 043516 doi:10.1103/physrevd.71.043516 [ astro - ph/0309279 ] . c. w. purcell , a. r. zentner and m. y. wang , jcap * 1208 * ( 2012 ) 027 [ arxiv:1203.6617 [ astro-ph.ga ] ] . f. s. ling , phys . d * 82 * ( 2010 ) 023534 [ arxiv:0911.2321 [ astro-ph.ga ] ] . j. i. read , l. mayer , a. m. brooks , f. governato and g. lake , mon . not . * 397 * ( 2009 ) 44 [ arxiv:0902.0009 [ astro-ph.ga ] ] . m. lisanti and d. n. spergel , phys . dark univ . * 1 * ( 2012 ) 155 doi:10.1016/j.dark.2012.10.007 [ arxiv:1105.4166 [ astro-ph.co ] ] . m. kuhlen , m. lisanti and d. n. spergel , phys . rev . d * 86 * ( 2012 ) 063505 [ arxiv:1202.0007 [ astro-ph.ga ] ] . c. savage , k. freese and p. gondolo , phys . d * 74 * ( 2006 ) 043531 doi:10.1103/physrevd.74.043531 [ astro - ph/0607121 ] . c. a. j. ohare and a. m. green , phys . d * 90 * ( 2014 ) no.12 , 123511 [ arxiv:1410.2749 [ astro-ph.co ] ] . c. mccabe , phys . d * 82 * ( 2010 ) 023530 [ arxiv:1005.0579 [ hep - ph ] ] . m. t. frandsen , f. kahlhoefer , c. mccabe , s. sarkar and k. schmidt - hoberg , jcap * 1201 * ( 2012 ) 024 [ arxiv:1111.0292 [ hep - ph ] ] . d. z. freedman , phys . d * 9 * ( 1974 ) 1389 . n. mirabolfathi , h. r. harris , r. mahapatra , k. sundqvist , a. jastram , b. serfass , d. faiez and b. sadoulet , arxiv:1510.00999 [ physics.ins-det ] . l. e. strigari , arxiv:1604.00729 [ astro-ph.co ] . w. c. haxton , r. g. hamish robertson and a. m. serenelli , ann . astrophys . * 51 * ( 2013 ) 21 [ arxiv:1208.5723 [ astro-ph.sr ] ] . v. antonelli , l. miramonti , c. pena garay and a. serenelli , adv . high energy phys . * 2013 * ( 2013 ) 351926 [ arxiv:1208.1356 [ hep - ex ] ] . g. bellini _ et al . _ [ borexino collaboration ] , nature * 512 * ( 2014 ) no.7515 , 383 . j. f. beacom , ann . nucl . part . * 60 * ( 2010 ) 439 [ arxiv:1004.3311 [ astro-ph.he ] ] . g. battistoni , a. ferrari , t. montaruli and p. r. sala , astropart . * 23 * ( 2005 ) 526 . doi:10.1016/j.astropartphys.2005.03.006 m. honda , t. kajita , k. kasahara and s. midorikawa , phys . d * 83 * ( 2011 ) 123001 [ arxiv:1102.2688 [ astro-ph.he ] ] . g. battistoni , a. ferrari , t. montaruli and p. r. sala , astropart . * 19 * ( 2003 ) 269 [ hep - ph/0207035 ] . m. c. smith _ et al . _ , mon . not . * 379 * ( 2007 ) 755 [ astro - ph/0611671 ] . t. piffl _ et al . _ , astron . astrophys . * 562 * ( 2014 ) a91 [ arxiv:1309.4293 [ astro-ph.ga ] ] . n. bozorgnia , g. b. gelmini and p. gondolo , phys . d * 84 * ( 2011 ) 023516 [ arxiv:1101.2876 [ astro-ph.co ] ] . m. j. reid _ et al . _ , astrophys . j. * 783 * ( 2014 ) 130 [ arxiv:1401.5377 [ astro-ph.ga ] ] . o. bienaym _ et al . _ , astron . astrophys . * 571 * ( 2014 ) a92 [ arxiv:1406.6896 [ astro-ph.ga ] ] . m. lisanti , l. e. strigari , j. g. wacker and r. h. wechsler , phys . d * 83 * ( 2011 ) 023519 [ arxiv:1010.4300 [ astro-ph.co ] ] . f. s. ling , e. nezri , e. athanassoula and r. teyssier , jcap * 1002 * ( 2010 ) 012 [ arxiv:0909.2028 [ astro-ph.ga ] ] . n. bozorgnia _ et al . _ , arxiv:1601.04707 [ astro-ph.co ] . b. j. kavanagh , phys . d * 89 * ( 2014 ) 8 , 085026 [ arxiv:1312.1852 [ astro-ph.co ] ] . s. k. lee , jcap * 1403 * ( 2014 ) 047 doi:10.1088/1475 - 7516/2014/03/047 [ arxiv:1401.6179 [ astro-ph.co ] ] . j. stadel , d. potter , b. moore , j. diemand , p. madau , m. zemp , m. kuhlen and v. quilis , mon . not . * 398 * ( 2009 ) l21 [ arxiv:0808.2981 [ astro - ph ] ] . m. t. busha , a. e. evrard , f. c. adams and r. h. wechsler , mon . not . * 363 * ( 2005 ) l11 [ astro - ph/0412161 ] . g. angloher _ et al . _ [ cresst collaboration ] , eur . j. c * 76 * ( 2016 ) no.1 , 25 arxiv:1509.01515 r. agnese _ et al . _ [ supercdms collaboration ] , phys . rev . * 116 * ( 2016 ) no.7 , 071301 arxiv:1509.02448 . l. e. strigari and r. trotta , jcap * 0911 * ( 2009 ) 019 [ arxiv:0906.5361 [ astro-ph.he ] ] . a. h. g. peter , phys . d * 83 * ( 2011 ) 125029 [ arxiv:1103.5145 [ astro-ph.co ] ] . a. h. g. peter , v. gluscevic , a. m. green , b. j. kavanagh and s. k. lee , phys . dark univ . * 5 - 6 * ( 2014 ) 45 [ arxiv:1310.7039 [ astro-ph.co ] ] . c. strege , r. trotta , g. bertone , a. h. g. peter and p. scott , phys . d * 86 * ( 2012 ) 023507 doi:10.1103/physrevd.86.023507 [ arxiv:1201.3631 [ hep - ph ] ] . f. feroz and m. p. hobson , mon . not . * 384 * ( 2008 ) 449 [ arxiv:0704.3704 [ astro - ph ] ] . f. feroz , m. p. hobson and m. bridges , mon . not . * 398 * ( 2009 ) 1601 [ arxiv:0809.3437 [ astro - ph ] ] . f. feroz , m. p. hobson , e. cameron and a. n. pettitt , arxiv:1306.2144 [ astro-ph.im ] .	the search for weakly interacting massive particles ( wimps ) by direct detection faces an encroaching background due to coherent neutrino - nucleus scattering . for a given wimp mass the cross section at which neutrinos constitute a dominant background is dependent on the uncertainty on the flux of each neutrino source from either the sun , supernovae or atmospheric cosmic ray collisions . however there are also considerable uncertainties with regard to the astrophysical ingredients to the predicted wimp signal . uncertainties in the velocity of the sun with respect to the milky way dark matter halo , the local density of wimps , and the shape of the local wimp speed distribution all have an effect on the expected event rate in direct detection experiments and hence will change the region of the wimp parameter space for which neutrinos are a significant background . in this work we extend the wimp+neutrino analysis to account for the uncertainty in the astrophysics dependence of the wimp signal . we show the effect of uncertainties on projected discovery limits with an emphasis on low wimp masses ( less than 10 gev ) when solar neutrino backgrounds are most important . we find that accounting for astrophysical uncertainties changes the shape of the neutrino floor as a function of wimp mass but also causes it to appear at cross sections up to an order of magnitude larger , extremely close to existing experimental limits - indicating that neutrino backgrounds will become an issue sooner than previously thought . we also explore how neutrinos hinder the estimation of wimp parameters and how astrophysical uncertainties impact the discrimination of wimps and neutrinos with the use of their respective time dependencies .
the influence of environmental degrees of freedom ( dof ) on quantum phenomena , like e.g. tunneling , has been of great interest during the last decades @xcite . a significant progress came from the investigation of a more or less phenomenological model where a particle in a one - dimensional potential @xmath5 or a free particle is coupled to a bath of harmonic oscillators with the spectral density @xmath6 . the quantum dissipation generated by the bath depends qualitatively on the low frequency behavior of @xmath6 @xcite . a particular interesting case is ohmic damping , when @xmath7 for low enough frequencies . in that case and for a symmetric double well potential , the particle at zero temperature undergoes a transition from a delocalized state to a localized one if the coupling constant between the particle and bath exceeds a critical value @xcite . an interesting observation has been made by caldeira and leggett @xcite . the exponent of the exponential factor for the tunneling probability is multiplied by @xmath8 , the phenomenological friction coefficient of the corresponding _ classical _ dynamics . this relationship between classical and quantum dissipation has been deepened and generalized by leggett @xcite for an arbitrary linear coupling between the particle and bath coordinates . let @xmath9 be the fourier transform of the classical particle trajectory @xmath10 and @xmath11 the transformed classical equation of motion , where @xmath12 contains the dissipative influence of the bath . then the reduced euclidean particle propagator @xmath13 ( where the harmonic dof have been eliminated ) can be represented by a path integral in the imaginary time @xmath14 @xcite @xmath15e^{-\frac{1}{\hbar } s[q(\tau ) ] } \label{eq2}\qquad.\ ] ] the action @xmath16 = s_0[q(\tau ) ] + s_{nonlocal}[q(\tau)]\tag{3a}\label{eq3a}\ ] ] contains the local @xmath17 = \int\limits_0^td\tau\ , \left[\frac{m_p}{2}\dot{q}(\tau)^2+v(q(\tau ) ) + \frac{\mu}{2}q(\tau)^2\right ] \tag{3b } \label{eq3b}\ ] ] and the nonlocal part @xmath18 = - \int\limits_0^t \,d\tau\int\limits_0^{\tau}\,d\tau'\ , k(\tau-\tau')q(\tau)q(\tau ' ) = -\frac{1}{2\pi}\frac{1}{2}\int d\omega \,\widetilde{k}(\omega)\| \tilde{q}(\omega)\|^2 . \tag{3c}\label{eq3c}\ ] ] the second equality holds for @xmath19 . @xmath20 is the fourier transform of the integral kernel @xmath21 and it is related to @xmath12 by @xmath22 . if the euclidean lagrangian of the particle - bath system is @xmath23 with @xmath24 and @xmath25\tag{4c } \label{eq4c}\ ] ] then@xcite @xmath26 and @xmath27 with the spectral density @xmath28 for a finite @xmath2 the kernel @xmath21 and the paths @xmath29 can be periodically continued . then the fourier series for @xmath21 is given by the fourier coefficients @xcite @xmath30 with @xmath31 . here @xmath32 and @xmath33 are the masses of the particle and harmonic oscillators , respectively . @xmath34 are the oscillator frequencies and @xmath35 are the coupling constants between q and the coordinates of the oscillators @xmath36 . note that @xmath36 are not necessarily positions , but can represent normal mode coordinates of vibrations , etc . in the following we will restrict ourselves to a system of @xmath37 particles whose potential energy @xmath38 includes harmonic and anharmonic interactions . without an external field , @xmath39 must be translationally invariant . however , since the coordinates of langrangian ( 4 ) are not specified , it is not necessarily invariant under translations . this has motivated chudnovsky @xcite to apply the caldeira - leggett approach to a system of two particles ( @xmath40 ) with positions @xmath41 and masses @xmath42 , coupled to oscillators with coordinates @xmath43 , frequencies @xmath44 and masses @xmath45 . the corresponding euclidean lagrangian is of the form @xmath46 \label{eq6}\ ] ] with @xmath47 being the interaction energy between both particles . the coupling of particle @xmath48 to the oscillators is assumed to be zero . it is obvious that @xmath49 is translationally invariant under @xmath50 provided that @xmath51 are real space coordinates . surprisingly the elimination of harmonic dof does not lead for zero temperature to a nonlocal action of caldeira - leggett type ( cf . ( [ eq3c ] ) , ( [ eq5b]),([eq``1 '' ] ) ) . the fourier coefficients of chudnovsky s kernel have the form @xmath52 which differs qualitatively from the caldeira - leggett type result ( [ eq``1 '' ] ) . however , for @xmath53 one gets @xmath54 where the last term , up to the additional factor @xmath55 , corresponds to the second order time derivative of the kernel , identical to @xmath56 from eq . ( [ eq``1 '' ] ) , if one chooses @xmath57@xcite . then the question arises : does a translationally invariant model in general lead to a nonlocal action , which is not of caldeira - leggett type ? this is one of the main points we want to investigate among others in our paper . it will be done for a microscopic(within a born - oppenheimer approximation ) , explicitly translationally invariant lattice model with a defect which can not diffuse . we will show how the normal mode coordinates for the harmonic dof can exactly be separated from the anharmonic ones . this leads to a lagrangian of the form of eq . ( 4 ) with coupling constants @xmath58 , frequencies @xmath59 and a spectral density determined by the microscopic model parameters . such a microscopic justification of lagrangian ( 4 ) has been presented for quantum _ diffusion _ @xcite . there a particle diffusing through an elastic lattice is considered . if , however , that particle can not diffuse and is an integral part of the lattice , e.g. an impurity which can tunnel only between two positions , one has to separate the center of mass ( com ) and relative coordinates of _ all _ particles . for such a situation , sethna @xcite has estimated the coupling constants @xmath58 by comparing the strain field of an elastic monopole of an impurity with the displacement for a longitudinal mode . but , as far as we know , there is no microscopic derivation available for the quantities @xmath58 , @xmath59 and @xmath6 for a non - diffusing impurity in a lattice . besides such a microscopic derivation we will show that the quantum behavior of the defect is sensitive to its location . the outline of our paper is as follows : section [ model ] presents our model and outlines the main steps leading to lagrangian ( 4 ) . the implications for the quantum behavior will be discussed in the third section with a special emphasis on the role of defect location . a summary and conclusions are contained in the final section [ summary ] . appendices a , b and c contain details on the separation of the harmonic and anharmonic dof . we consider an open chain of @xmath37 particles with masses @xmath60 and harmonic nearest neighbor interactions and one anharmonic bond representing a defect . this model could describe a linear macromolecule with an impurity . the classical hamiltonian reads @xmath61 with the potential energy @xmath62 in which @xmath63 is the position of @xmath64-th particle , @xmath65 is the elastic constant of the harmonic nearest neighbor interaction , @xmath66 is the equilibrium length of the harmonic bonds and @xmath67 is the energy of the anharmonic bond defect located between the sites @xmath1 and @xmath68 . the potential energy @xmath39 is explicitly translationnally invariant . several ways exist to separate the harmonic and anharmonic dof , which finally lead to the langrangian ( 4 ) ( see appendices [ b ] and [ c ] ) . one ( see appendix [ b ] ) , which is also applicable in higher dimensions , is to introduce the com @xmath69 and relative coordinates of the _ bond defect _ @xmath70 then , @xmath71 are harmonic coordinates . their kinetic and potential energy can be diagonalized by introducing normal coordinates . @xmath72 is linearly coupled to these coordinates . as a result , one obtains the lagrangian ( 4 ) . here we will choose a different approach applicable to 1d systems , which , however , leads to the same lagrangian . let @xmath73 be the com of _ all _ the particles and @xmath74 @xmath75 be the relative coordinates , respectively . using notation @xmath76 , eq . ( [ eq12a]),([eq12b ] ) has the form @xmath77 with @xmath78 let @xmath79 be the canonical conjugate momenta of @xmath80 . it is easy to prove that eq . ( [ eq13a ] ) implies @xmath81 substituting @xmath82 into eq . ( [ eq13a ] ) yields @xmath83 here the first term is the kinetic energy of com , which will be dropped from now on . note that the use of relative coordinates introduces a coupling between the momenta . in the next step , we perform the canonical transformation @xmath84 @xmath85 @xmath86 @xmath87 @xmath88 this leads to @xmath89 where @xmath90 is the defect hamiltonian , @xmath91 is the harmonic part of the hamiltonian and @xmath92 is the coupling between the two . the matrix @xmath93 in ( [ eq16c ] ) depends on the masses @xmath94 ( see appendix [ a ] ) . let @xmath95 and @xmath96 be the normalized eigenvectors and eigenvalues of * t * , respectively , and @xmath97 be the orthogonal matrix , which diagonalizes * t * , i.e. @xmath98 then we can introduce the normal coordinates @xmath99 for @xmath100 such that @xmath101 becomes diagonal @xmath102\qquad .\tag{19a}\label{eq19a}\ ] ] the interaction term takes the form @xmath103 the @xmath1 dependent coupling constants are given by @xmath104\qquad .\label{eq17}\ ] ] the final step is the legendre transformation of @xmath105 , which leads to the euclidean lagrangian @xmath106 where @xmath107 and @xmath108 \qquad .\tag{21c}\label{eq21c}\ ] ] here we have used the equality @xmath109 ( cf . ( [ eq15b ] ) ) and @xmath110 which follows from the completeness of the set @xmath111 of eigenvectors and @xmath112 ( see below ) . ( [ eq19 ] ) allows us to include the counterterm @xmath113 in eq . ( [ eq16b ] ) into @xmath114 . this counterterm , the role of which has been discussed by caldeira and leggett @xcite , results from the canonical transformation , eq . ( [ eq15a ] ) , ( [ eq15b ] ) . this transformation eliminates the coupling between the momenta of the harmonic dof and that of the defect and generates coupling between the normal mode coordinates @xmath51 and the corresponding defect variable @xmath72 ( cf . ( [ eq16d ] ) ) . due to the disappearance of the counterterm , there is no frequency renormalization for the bond defect@xcite . the lagrangian ( 21 ) is identical to that of eq . the masses @xmath115 and frequencies @xmath59 follow from @xmath116 and the reduced defect mass @xmath117 is given by @xmath118 note that these results are exact for one dimensional systems . it can be shown that for an _ arbitrary _ impurity in a two- or three - dimensional system the langrangian ( 21 ) can be derived within a kind of harmonic approximation @xcite . in this section we will investigate the zero temperature quantum behavior of the anharmonic bond defect embedded in the harmonic chain as shown in fig . [ fig ] . particles . the masses @xmath119 , @xmath120 are chosen to be equal . @xmath66 is the equilibrium length of the harmonic bonds and @xmath121 the two degenerate equilibrium lengths of the anharmonic bond . ] we will assume that @xmath122 is a symmetric double - well potential with _ degenerate _ minima at @xmath123 and @xmath124 . then the classical ground state of @xmath125 ( cf . . [ eq10b ] ) is twofold degenerate ( see fig.1 ) . therefore the low - lying eigenstates form doublets . neglecting the excited doublets at zero temperature is justified if the bare tunneling splitting of the ground state doublet is much less than the frequency of the upper phonon band edge @xmath126 ( see eq . ( [ eq25a]))@xcite . if the total number @xmath37 of particles in the chain is macroscopically large and the particle number @xmath1 is in the bulk of the chain , i.e. @xmath127 , one might have expected a suppression of tunneling since the change from , e.g. @xmath128 to @xmath129 , would require a translation of the macroscopic mass of the left and right harmonic parts of the chain . we will see that this naive expectation is not always correct . on the other hand , if the defect is close to one of the free boundaries , i.e. either @xmath130 or @xmath131 , only a finite mass @xmath132 has to be translated . consequently tunneling can not be suppressed . this qualitative @xmath1-dependence should follow from that of the kernel @xmath21 , which itself results from the strong @xmath1-sensitivity of the spectral density @xmath6 . subsection [ subsec_a ] discusses this phenomenon . the limit @xmath133 ( or @xmath134 ) , motivated by the conclusions drawn in ref . @xcite , will be discussed in subsection [ subsec_b ] . as we will see in subsection [ subsec_b ] the form of the nonlocal action eq . ( [ eq3c ] ) does not depend qualitatively on the masses @xmath119 , even if one of the defect masses @xmath135 or @xmath136 becomes infinitely large . therefore we will choose for convenience @xmath137 . let us start with the situation where the bond defect is located within the bulk , i.e. @xmath127 , so that @xmath138 one can prove that the tunneling phenomena do not depend on @xmath139 if it is different from 0 and 1 . therefore we choose @xmath140 and without loss of generality @xmath37 to be even . this choice and the assumption @xmath137 allows us to determine the eigenfrequencies @xmath59 and the eigenvectors @xmath141 exactly for finite @xmath37 . a calculation , whose technical details are presented in appendix [ a ] , results in @xmath142 @xmath143 @xmath144 where @xmath145 for @xmath146 . it is easy to see that @xmath147 are the symmetric eigenvalues with respect to @xmath148 and @xmath149 are the antisymmetric ones . from this , @xmath150 and eq . ( [ eq17 ] ) it is obvious that the bond defect does not couple to the antisymmetric vibrational modes , as may be expected from the symmetry of the problem . with these results we can calculate the spectral density . making use of eqs . ( [ eq17 ] ) , ( [ eq23a ] ) , ( [ eq25a ] ) , ( [ eq25b ] ) and ( [ eq25c ] ) we get from eq . ( [ eq5c ] ) in the thermodynamic limit @xmath151 that @xmath152 @xmath153 in the limit @xmath154 we obviously have @xmath155 which corresponds to the situation of ohmic damping . the ohmic damping results from two facts : first , the density of states @xmath156 of the vibrational modes in a one - dimensional lattice is constant for @xmath154 and second the squared coupling constant involves the factor @xmath157 which , for @xmath158 , @xmath159 with @xmath160 , oscillates faster and faster so that it can be replaced by 1/2 . it is emphasized that the condition @xmath161 , i.e. @xmath162 or 1 , will change the shape of @xmath6 qualitatively . now we can calculate the kernel @xmath21 . for @xmath163 and @xmath164 the fraction in eq . ( [ eq5b ] ) can be well approximated by @xmath165 . the influence of the oscillators on tunneling of @xmath72 is determined by the large @xmath166 behavior , i.e. by the low frequency modes . therefore , substituting @xmath6 from eq . ( [ eq26b ] ) into eq . ( [ eq5b ] ) we find , of course , the well - known result for ohmic damping @xcite @xmath167 as a consequence , there exists a critical elastic constant @xmath0 so that the anharmonic bond can tunnel for @xmath168 , despite macroscopic masses have to be moved ( see figure 1 ) . for @xmath169 symmetry is broken . if the anharmonic bond is prepared in its ground state , e.g. @xmath170 , it will remain there on average . if , however , the bond defect is located close to one of the boundaries , so that either @xmath171 or @xmath172 ( @xmath173 0 or 1 in ( [ eq21 ] ) ) , the situation changes . note that we perform first the thermodynamic limit @xmath174 . then @xmath130 means that @xmath1 may equal @xmath175 or even a larger , but still finite number . for @xmath137 one can easily show that @xmath176 where @xmath177 is the normalization constant of @xmath178 ( see appendix [ a ] ) . replacing @xmath177 by its low frequency behavior @xmath179 and taking the limits @xmath180 and @xmath181 for the kernel . in order to be consistent we also replaced @xmath182 by its low frequency dispersion @xmath183 . the integrand of eq . ( [ eq26 ] ) involves two @xmath184-scales @xmath185 equating @xmath186 defines the timescale @xmath187 the physical meaning of @xmath188 is as follows : the path integral formalism @xcite allows one to investigate quantum tunneling by determining the instanton solutions , i.e. the solutions of the classical equation of motion for a double well potential in imaginary time . the width @xmath189 of a single instanton is @xmath190 if we assume that @xmath191 , the elastic constant of the harmonic bonds , then @xmath192 so that @xmath193 the @xmath194-dependence of @xmath195 is sensitive to whether @xmath196 or vice versa . let us start with the long time limit * @xmath197 . + then it follows from eqs . ( [ eq27 ] ) and ( [ eq28 ] ) that @xmath198 . the major contribution to the integral in eq . ( [ eq26 ] ) comes from @xmath199 . therefore , we are allowed to replace @xmath200 by @xmath201 , which leads to the spectral density @xmath202 at low frequencies corresponding to _ superohmic _ damping . it implies that @xmath203 . the precise result is @xmath204 + -dependence of @xmath195 for @xmath205 and @xmath206 ( from bottom to top ) on a log - log representation . the dotted and dashed line , corresponds to @xmath207 and @xmath208 , respectively . the crossover at @xmath209 from the @xmath207 behavior to that of @xmath208 can clearly be seen . the parameters have been chosen as follows : @xmath210 . ] * a different situation takes place for @xmath211 . + then @xmath212 and the integral in eq . ( [ eq26 ] ) must be decomposed in two contributions @xmath213 . the first integral yields a constant in the leading order in @xmath214 . for the second one we are allowed to replace @xmath215 by @xmath216 since the main contribution comes from @xmath217 so that the function @xmath200 is oscillating fast between zero and one , whereas @xmath218 varies slowly . accordingly @xmath219 corresponds to an effective spectral density @xmath220 , i.e. to ohmic damping . it is straightforward to estimate both integrals . as the final result we obtain @xmath221\label{eq31}\ ] ] + for @xmath222 . taking the limit @xmath223 in eq . ( [ eq31 ] ) restores the result ( [ eq24 ] ) for the kernel @xmath21 . for @xmath130 or @xmath224 , and @xmath174 we obtain for @xmath195 a crossover at @xmath188 from the power law @xmath207 for @xmath219 to @xmath225 for @xmath226 . figure 2 illustrates this behavior for @xmath195 , calculated numerically . on the log - log plot of figure 2 the crossover between both power laws can easily be observed . this crossover is related to the @xmath1-dependence of the spectral density because the defect - phonon coupling constants @xmath227 are @xmath1-dependent . if the observation time @xmath2 ( which is the time @xmath2 in @xmath228 from eq . ( [ eq2 ] ) ) is smaller than @xmath188 then @xmath229 will be smaller than @xmath188 , as well . consequently the kernel @xmath230 entering the nonlocal action ( eq . ( [ eq3c ] ) ) decays as @xmath231 . if , however , @xmath2 is larger than @xmath188 then it is possible that @xmath229 becomes larger than @xmath188 , as well . for those values of @xmath229 the kernel @xmath232 decays as @xmath233 . this discussion reveals that the choice of the observation time @xmath2 allows to fix the large-@xmath194 behavior of @xmath195 , where , of course , @xmath234 . if @xmath1 is far away from the chain end the crossover time @xmath188 is correspondingly large . increasing @xmath1 even more @xmath188 increases , too . nevertheless , the observation time dependence still exists . it disappears for @xmath235 , only . we remind the reader that the limit @xmath174 has to be taken first . now making use of the analogy @xcite between the calculation of the action of a multi - instanton configuration interacting via @xmath236 and a one - dimensional ising model with the coupling constants @xmath237 we can conclude the following : if the observation time @xmath2 is smaller than @xmath238 then we have @xmath239 , i.e. _ ohmic _ damping . in that case the bond defect may tunnel for @xmath240 , whereas symmetry becomes broken for @xmath241 . note , that this is not a sharp transition at @xmath242 since finite @xmath2 corresponds to a finite ising chain which does not exhibit a sharp phase transition . what really happens when increasing the coupling constant @xmath65 is an increase of the correlation length @xmath243 . as soon as @xmath243 equals the size @xmath2 of the ising - chain a long range order occurs . however , if @xmath2 is much larger than @xmath244 we have @xmath245 and tunneling is never suppressed @xcite . one might be puzzled by these conclusions since the transition for ohmic damping to decoherence for @xmath246 occurs for large @xmath2 , or to be more precise it becomes a sharp transition for @xmath247 , only . as we already stressed above , the transition for @xmath248 is _ not _ sharp . the relevant phonons contributing to @xmath195 have wavenumbers @xmath249 . this makes the effective spectral density ohmic . mapping the situation for @xmath250 again onto the ising chain of length @xmath49 results in the coupling constants @xmath251 decaying like @xmath252 for @xmath253 and like @xmath254 for @xmath255 . it is clear that there is _ no _ sharp phase transition for finite @xmath49 . but if @xmath256 , the coupling constants decay as @xmath257 . for a fixed temperature @xmath2 ( not to be confused with the observation time @xmath2 ) there will be no magnetic order ( @xmath258 coherent tunneling ) for @xmath259 . for @xmath260 a crossover to the long range order ( @xmath258 decoherent tunneling ) takes place . accordingly the quantum tunneling phenomenon is richer for @xmath250 than for @xmath261 . actually we may think also in the real time @xmath262 terms that as long as only the phonons with relatively high frequencies @xmath263 and hence short wave length such that @xmath264 participate in the interaction with the anharmonic defect , the latter does not feel that the chain is finite and behaves as in the ohmic case . in the course of time the lower frequency phonons with higher wave length reach the end of the chain and a crossover to a superohmic behavior takes place . in this subsection we will assume that @xmath265 for @xmath127 we may choose without loss of generality @xmath266 and @xmath37 being even . it is easy to prove that @xmath267 from eq . ( [ eq25b ] ) remain eigenvectors of * t * with @xmath268 , @xmath269 given by eq . ( [ eq25c ] ) . the remaining @xmath270 eigenvectors are of the form @xmath271 @xmath272 is a solution of transcendental equation . let us introduce the quantities @xmath273 , @xmath40 and @xmath274 . the limit @xmath275 ( or @xmath276 implies @xmath277 . since the results in @xcite motivate us to study , e.g. @xmath278 , we find @xmath279 in the limit @xmath280 . substituting eqs . ( [ eq36a ] ) , ( [ eq36b ] ) into eq . ( [ eq17 ] ) yields @xmath281 in the leading order in @xmath282 . as a result @xmath6 and therefore @xmath21 , too , are doubled as compared to the case of @xmath283 , i.e. we get @xmath284 for @xmath285 . the only essential result of changing @xmath286 from @xmath287 to infinity is that the critical elastic constant increases by a factor of two . for a translationally invariant chain with one anharmonic bond and otherwise harmonic nearest neighbor interactions , we have shown exactly how the anharmonic degree of freedom can be separated from the harmonic ones in their normal mode representation . as a result , we have obtained lagrangian ( 21 ) , which is of the form of lagrangian ( 4 ) . note , that this result can also be obtained for a three - dimensional system within the born - oppenheimer approximation starting with an arbitrary translationally invariant potential @xmath288 for a @xmath37-particle system @xcite . since the caldeira - leggett type nonlocal action @xcite is based on the form ( 4 ) ( or ( 21 ) ) of the lagrangian it is not the translation invariance and therefore not the conservation of momentum , which can lead to a different type of nonlocal action . the discrepancy between our results and those of ref . @xcite may have the following origin . since the harmonic part of lagrangian ( [ eq6 ] ) is diagonal in @xmath43 , the harmonic variables @xmath43 are already normal mode coordinates . in that case a translation of the full system only changes the goldstone mode amplitude , let us say @xmath289 , but leaves all the other normal mode coordinates unchanged , i.e. @xmath290 , @xmath291 for any translation . if @xmath292 and @xmath293 in eq . ( [ eq6 ] ) are real space coordinates then the coupling term @xmath294 is not translationally invariant for @xmath295 . our model has allowed us to calculate explicitly , e.g. for @xmath296 and @xmath297 the coupling constants @xmath58 , the eigenfrequencies @xmath59 and the spectral density @xmath6 . the @xmath298-dependence of @xmath299 is determined by the density of states @xmath156 and the coupling constants @xmath58 . although @xmath156 for @xmath300 is independent of the number @xmath1 , the frequency dependence of @xmath58 exhibits a sensitivity to the location @xmath1 , which makes the affect of the harmonic bath on quantum tunneling @xmath1 sensitive . as a consequence , the damping is ohmic if the bond defect is within the bulk of the chain and superohmic if it is close to the boundaries . for the former case there is a transition from a delocalized state ( due to tunneling ) to a localized one if the elastic constant exceeds a critical value @xmath0 , whereas tunneling is never suppressed in the latter case , provided the observation time @xmath2 is large enough compared to @xmath188 which is roughly @xmath1 times the instanton kink width . for @xmath248 ( since the thermodynamic limit @xmath174 had already been performed , @xmath1 must be finite , but can be arbitrary large ) the dissipation is effectively ohmic leading to a similar behavior when the bond defect is within the bulk . if @xmath127 ( e.g. @xmath297 ) and if one of the masses of the bond defect tends to infinity , no significant changes occur except for doubling of the critical constant @xmath0 . this is obvious since , e.g. @xmath301 , makes the part of the chain to the right of the defect inactive , i.e. the phonons to the right do not act as a reservoir for the bond defect . accordingly , only half of the harmonic chain is generating dissipation , which results in doubling of @xmath0 . although we are not aware of a concrete experimental system , these results could be relevant for linear macromolecules , which may be described by the model hamiltonian eq . ( 10 ) . if many of such molecules with a single defect are produced . the position of which can be controlled experimentally , one might observe , e.g. the location - dependent tunneling by spectroscopic methods . this work has been completed when two of us ( v.f . and r.s . ) were a member of the advanced study group 2007 at the mpipks dresden . v.f . and r.s . gratefully acknowledge the mpipks for its hospitality and financial support . v.f . acknowledges a support of israeli science foundation , grant n 944/05 . one of us ( r.s . ) thanks eugene chudnovsky for helpful discussions . the separation of the harmonic and anharmonic dof by using the center of mass and relative coordinates of the _ total _ chain has been described in section [ model ] . the transformation to normal coordinates requires the diagonalization of the matrix * t. here the most important steps of the diagonalization procedure are outlined . making use of eqs . ( 12)-(15 ) one obtains the matrix elements of the symmetric matrix t * in the form @xmath302 @xmath303 and @xmath304 the diagonalization of * t * can not be done analytically for arbitrary masses @xmath94 . therefore we take the simplest case of equal masses , @xmath305 . then it is straightforward to prove that the eigenvalue equation @xmath306 is solved by @xmath307 @xmath308 where @xmath309 is the normalization constant and @xmath310 is a coefficient depending on the location @xmath1 of the bond defect . the wave numbers @xmath311 are solutions of the transcendental equation @xmath312 since the l.h.s . of eq . ( [ eqa4 ] ) diverges at @xmath313 it is easy to see that its solutions are of the form @xmath314 with @xmath315 . there are @xmath316 solutions corresponding the @xmath316 harmonic dof . the remaining two dof are the com and the bond defect coordinate @xmath317 and @xmath72 , respectively . in the thermodynamic limit @xmath174 , the discrete set of @xmath311 wave vectors becomes a continuous variable @xmath184 within @xmath318 $ ] with a constant density which together with eq . ( [ eqa3 ] ) implies a constant low energy density of states . the normalization constant @xmath309 and the coefficient @xmath310 are functions of @xmath319 and @xmath37 . their explicit expressions are not given here . in this appendix we will describe the separation of harmonic and anharmonic dof using an approach alternative to that used in section [ model ] . it has the advantage that it can be straightforwardly applied to higher dimensional systems . the starting point is the introduction of com and relative coordinate @xmath320 and @xmath72 , respectively , of the _ bond defect _ ( see eqs . ( [ eq11a ] ) and ( [ eq11b ] ) ) . the corresponding canonical momenta @xmath321 @xmath322 substituting @xmath323 from eq . ( 11 ) and @xmath324 from ( b1 ) into eq . ( 10 ) yields @xmath325 where @xmath326 @xmath327 @xmath328 @xmath329q_m\quad .\tag{b2d}\label{eqb2d}\ ] ] here @xmath330 is the reduced mass of bond defect . note that @xmath331 from eq . ( [ eqb2b ] ) is identical to @xmath331 from eq . ( [ eq16b ] ) after replacing @xmath332 by @xmath333 . the transformation of @xmath101 ( eq . ( [ eqb2c ] ) ) to the normal coordinates may be more conveniently carried out using the notations @xmath334 @xmath335 and @xmath336 next we expand the potential part @xmath337 of @xmath101 around its equilibrium configuration , @xmath338 up to the second order terms in @xmath339 . note that this is not an approximation since @xmath340 is a harmonic potential . this leads to @xmath341 introducing the mass - weighted coordinates @xmath342 @xmath343 eq . ( [ eqb5 ] ) yields @xmath344 where the only nonzero matrix elements of the symmetric matrix @xmath345 are @xmath346 and @xmath347 let @xmath348 and @xmath349 be respectively the eigenvectors and eigenvalues of @xmath350 . the canonical transformation @xmath351 leads to the normal mode representation @xmath352.\tag{b9 } \label{eqb9}\ ] ] since @xmath101 in eq . ( [ eqb5 ] ) is still translation invariant there is a zero frequency mode which we choose for @xmath353 . with @xmath354 we get @xmath355\quad .\tag{b10}\label{eqb10}\ ] ] the first term in the r.h.s . of eq . ( [ eqb10 ] ) is the kinetic energy of the com of _ total _ chain . the second term corresponds to @xmath101 from eq . ( [ eq19a ] ) . using eqs . ( b3 ) , ( [ eqb4 ] ) , ( b6 ) and ( [ eqb8 ] ) brings the interaction part ( eq . ( [ eqb2d ] ) ) into the form @xmath356 with @xmath357\tag{b12 } \label{eqb12}\ ] ] again , the analytical diagonalization of @xmath350 can not be performed for arbitrary masses . accordingly , we choose @xmath358 as in appendix [ a ] . ( [ eqb7b ] ) and ( [ eqb7c ] ) result in @xmath359 @xmath360 @xmath361 all the other matrix elements vanish . again it is straightforward to prove that the eigenvalue equation @xmath362 for @xmath363 is solved by @xmath364 @xmath365 with @xmath366 being the normalization constant and @xmath367 a @xmath1-dependent coefficient . the remaining equations for @xmath348 with @xmath368 and @xmath369 yield a nontrivial solution if a corresponding determinant vanishes . this condition leads to the transcendental equation @xmath370 for the wave numbers @xmath371 . although eq . ( [ eqb17 ] ) looks quite different from the transcendental equation ( [ eqa4 ] ) it can be shown by use of identities for trigonometric functions that ( [ eqb17 ] ) and ( [ eqa4 ] ) are equivalent , i.e. the set of solutions @xmath372 of eq . ( [ eqb17 ] ) and @xmath373 of eq . ( [ eqa4 ] ) are identical . we have already stressed that @xmath331 from eq . ( [ eqb2b ] ) and that from eq . ( [ eq16b ] ) are identical , as well . straightforward but tedious calculations show that the complete lagrangian corresponding to the classical hamiltonian eq . ( b2 ) is identical to the lagrangian ( 21 ) . particularly , it can be proven that @xmath374 from eq . ( [ eqb12 ] ) is identical to @xmath227 from eq . ( [ eq17 ] ) . in this appendix we will show that separation of the harmonic and anharmonic dof can be done by taking the left and right harmonic parts separately . similarly to the approach used in section [ model ] our first step is to separate the com of the _ total _ chain from the relative coordinates . this leads to the hamiltonian from eq . ( [ eq11 ] ) . neglecting the kinetic energy of com ( [ eq11 ] ) can be rewritten as @xmath375 where @xmath376 @xmath377 @xmath378 @xmath379 and the nonzero matrix elements are @xmath380 with @xmath381 for @xmath382 and @xmath383 for @xmath384 . let @xmath385 and @xmath386 be the eigenvectors and eigenvalues of @xmath387 . then we use the notations @xmath388 @xmath389 in order to get @xmath390,\tag{c3a}\label{eqc3a}\ ] ] @xmath391\tag{c3b}\label{eqc3b}\ ] ] for the harmonic part and @xmath392 for the interaction with @xmath393 for the coupling constants . this type of approach describes the chain as a bond defect coupled to _ two _ baths of harmonic oscillators , the left and right part of the chain . for the path integral formalism we need the lagrangian . from ( [ eqc1a ] ) , ( [ eqc1b ] ) , ( c3 ) and ( [ eqc4a ] ) we can determine the velocities @xmath394 and @xmath395 as function of the momenta . solving for the momenta as a function of the velocities is straightforward but tedious . we report the final result @xmath396\tag{c5a}\label{eqc5a}\ ] ] @xmath397\tag{c5b}\label{eqc5b}\ ] ] @xmath398\tag{c5c}\label{eqc5c}\ ] ] with @xmath399^{-1}.\tag{c6}\label{eqc6}\ ] ] making use of ( c5 ) for the calculation of the legendre transform of @xmath105 from eq . ( c1 ) leads to the euclidean lagrangian @xmath400 where @xmath401 in which @xmath402 + \frac{1}{2 } \sum_{\mu=1}^{n - m-1 } \left [ \frac{1}{\lambda_{\mu}^r}(\dot{x}_{\mu}^r)^2 + c(x_{\mu}^r)^2\right],\ ] ] @xmath403 ^ 2 , \tag{c7c}\label{eqc7c}\ ] ] @xmath404.\tag{c7d}\label{eqc7d}\ ] ] this form of @xmath49 differs completely from that of eq . particularly , @xmath405 from eq . ( [ eqc7d ] ) is a coupling of the velocities and not of the bond defect coordinate @xmath72 with the normal mode coordinates @xmath406 as for @xmath405 from eqs . ( [ eq21a ] ) , ( [ eq21c ] ) . in addition , the harmonic part eq . ( [ eqc7c ] ) is not diagonal , i.e. due to the third term on the r.h.s . of eq . ( [ eqc7c ] ) there is an intra- and an inter - coupling between the phonons ( normal modes ) of the left and right harmonic part of the chain . in order to eliminate the harmonic degrees of freedom in the path integral representation of the propagator one has to diagonalize @xmath407 from eq . ( [ eqc7c ] ) . this can be done by a point transformation @xmath408 and @xmath409 . this transformation follows directly from eqs . ( [ eq15b ] ) , ( 18 ) and ( c2 ) : @xmath410 @xmath411 taking the time derivative of eq . ( c8 ) yields the transformation of the velocities . substituting this and the transformation ( c8 ) into eq . ( c7 ) diagonalizes @xmath407 and replaces the velocity coupling by a coupling of @xmath72 and @xmath51 . after a lengthy calculation one arrives at the lagrangian from eq . ( 21 ) , which , of course , is not a surprise .	we investigate quantum tunneling in a translation invariant chain of particles . the particles interact harmonically with their nearest neighbors , except for one bond , which is anharmonic . it is described by a symmetric double well potential . in the first step , we show how the anharmonic coordinate can be separated from the normal modes . this yields a lagrangian which has been used to study quantum dissipation . elimination of the normal modes leads to a nonlocal action of caldeira - leggett type . if the anharmonic bond defect is in the bulk , one arrives at ohmic damping , i.e. there is a transition of a delocalized bond state to a localized one if the elastic constant exceeds a critical value @xmath0 . the latter depends on the masses of the bond defect . superohmic damping occurs if the bond defect is in the site @xmath1 at a finite distance from one of the chain ends . if the observation time @xmath2 is smaller than a characteristic time @xmath3 , depending on the location @xmath1 of the defect , the behavior is similar to the bulk situation . however , for @xmath4 tunneling is never suppressed .
the unification of gauge couplings in the supersymmetric ( susy ) models allows one to embed the gauge group of the standard model ( sm ) into grand unified theories ( guts ) based on simple gauge groups such as @xmath3 , @xmath4 or @xmath0 . on the other hand gauge coupling unification permits one to incorporate susy models into superstring theories that make possible partial unification of gauge interactions with gravity . at high energies the @xmath0 symmetry in the superstring inspired models can be broken to rank5 subgroup @xmath5 where @xmath6 . two anomaly - free @xmath7 and @xmath8 symmetries originate from the breakdown of @xmath0 and @xmath4 respectively , i.e. @xmath9 . here we concentrate on a particular @xmath0 inspired supersymmetric model with an extra @xmath1 gauge symmetry that corresponds @xmath10 . only in this exceptional supersymmetric standard model ( e@xmath2ssm ) @xcite the right handed neutrinos do not participate in gauge interactions . therefore they may be superheavy shedding light on the origin of lepton mass hierarchy . the extra @xmath1 gauge symmetry survives to low energies and forbids a bilinear term @xmath11 in the superpotential of the considered model but allows the interaction @xmath12 . at the electroweak ( ew ) scale the scalar component of the sm singlet superfield @xmath13 acquires a non - zero vacuum expectation value ( vev ) , @xmath14 , breaking @xmath15 and an effective @xmath16 term is automatically generated . thus the @xmath17 problem in the e@xmath2ssm is solved in a similar way to the next - to - minimal supersymmetric standard model ( nmssm ) , but without the accompanying problems of singlet tadpoles or domain walls . the e@xmath2ssm is based on the @xmath18 gauge group which is a subgroup of @xmath0 . to ensure anomaly cancellation the particle content of the e@xmath2ssm is extended to include three complete * @xmath19 * representations of @xmath0 @xcite . each * @xmath20 * multiplet contains a sm family of quarks and leptons , right handed neutrino @xmath21 , sm singlet field @xmath22 which carries a non zero @xmath1 charge , a pair of @xmath23doublets @xmath24 and @xmath25 with the quantum numbers of higgs doublets and a pair of colour triplets of exotic quarks @xmath26 and @xmath27 which can be either diquarks ( model i ) or leptoquarks ( model ii ) . @xmath24 and @xmath25 form either higgs or inert higgs multiplets . we also require a further pair @xmath28 and @xmath29 from incomplete extra @xmath30 and @xmath31 representations to survive to low energies to ensure gauge coupling unification . our analysis reveals that the unification of the gauge couplings in the e@xmath2ssm can be achieved for any phenomenologically acceptable value of @xmath32 , consistent with the central measured low energy value , unlike in the mssm which requires significantly higher values of @xmath32 , well above central measured value @xcite . since right handed neutrinos have zero charges in the considered model they can acquire very heavy majorana masses . the heavy majorana right - handed neutrinos may decay into final states with lepton number @xmath33 , thereby creating a lepton asymmetry in the early universe . because the yukawa couplings of exotic particles are not constrained by the neutrino oscillation data the substantial values of the cp asymmetries in the considered model can be induced even for a relatively small mass of the lightest right handed neutrino ( @xmath34 ) so that the successful thermal leptogenesis may be achieved without encountering gravitino problem @xcite . the superpotential of the e@xmath2ssm involves a lot of new yukawa couplings in comparison to the sm . in general these new interactions violate baryon number conservation and induce non - diagonal flavour transitions . to suppress baryon number violating and flavour changing processes one can postulate a @xmath35 symmetry under which all superfields except one pair of @xmath24 and @xmath25 ( say @xmath36 and @xmath37 ) and one sm - type singlet field ( @xmath38 ) are odd . the @xmath35 symmetry reduces the structure of the yukawa interactions to : @xmath39 + h^{e}_{4j}(h_d h')e^c_j + \mu'(h'\overline{h}')+w_{\rm{mssm}}(\mu=0 ) , \end{array}\ ] ] where @xmath40 and @xmath41 . here we assume that all right handed neutrinos are relatively heavy so that they can be integrated out . the @xmath23 doublets @xmath42 and @xmath43 , that are even under @xmath35 symmetry , play the role of higgs fields generating the masses of quarks and leptons after the ew symmetry breaking ( ewsb ) . the singlet field @xmath13 must also acquire large vev to induce sufficiently large masses for the exotic charged fermions and @xmath44 boson to avoid conflict with direct particle searches at present and former accelerators . this implies that the yukawa couplings @xmath45 and @xmath46 should be large enough so that the evolution of the soft scalar mass @xmath47 of the singlet field @xmath13 results in negative values of @xmath47 at low energies , triggering the breakdown of @xmath1 symmetry . to guarantee that only @xmath42 , @xmath43 and @xmath13 acquire vevs in the e@xmath2ssm a certain hierarchy between the yukawa couplings is imposed , i.e. @xmath48 . after the breakdown of the gauge symmetry @xmath42 , @xmath43 and @xmath13 form three cp - even , one cp - odd and two charged states in the higgs spectrum . the mass of the one cp even higgs particle is always very close to the @xmath44 boson mass @xmath49 . the masses of another cp even , cp odd and charged higgs states are almost degenerate . as in the mssm and nmssm one of the cp even higgs bosons is always light irrespective of the susy breaking scale . however in contrast with the simplest susy models the lightest higgs boson in the e@xmath2ssm can be heavier than @xmath50 even at the tree level . in the two loop approximation the lightest higgs boson mass does not exceed @xmath51 @xcite . thus the sm like higgs boson in the e@xmath2ssm can be considerably heavier than in the mssm and nmssm . .particle spectrum in the constrained e@xmath2ssm for @xmath52 ( all mass parameters are given in gev ) . [ cols="^,^,^",options="header " , ] [ tabular ] the e@xmath2ssm contains many new parameters . even if we neglect @xmath53 , @xmath54 and @xmath55 the simplified superpotential of the e@xmath2ssm involves seven extra couplings ( @xmath56 , @xmath46 and @xmath45 ) as compared with the mssm with @xmath57 . the soft breakdown of supersymmetry gives rise to many new parameters . the number of fundamental parameters can be reduced drastically within the constrained version of the e@xmath2ssm ( cessm ) . constrained susy models imply that all soft scalar masses are set to be equal to @xmath58 at the scale @xmath59 , all gaugino masses @xmath60 are equal to @xmath61 and trilinear scalar couplings @xmath62 . thus cessm is characterised by the set of yukawa couplings , which are allowed to be of the order of unity , and universal soft susy breaking terms , i.e. @xmath63 where @xmath64 , @xmath65 and @xmath66 are the @xmath67quark , @xmath68quark and @xmath69lepton yukawa couplings . to simplify our analysis we assume that all parameters ( [ 3 ] ) are real and @xmath61 is positive . in order to guarantee the correct ewsb @xmath58 has to be positive . the set of the cessm parameters ( [ 3 ] ) should be in principle supplemented by @xmath56 and the associated bilinear scalar coupling @xmath70 . however since @xmath56 is not constrained by the ewsb and the term @xmath71 in the superpotential ( [ 2 ] ) is not suppressed by the @xmath0 the parameter @xmath56 is expected to be @xmath72 so that @xmath28 and @xmath29 decouple from the rest of the particle spectrum . as a consequence parameters @xmath70 and @xmath56 are irrelevant for our analysis . in order to calculate the particle spectrum within the cessm we evolve all mass parameters from the grand unification scale to the susy breaking scale for each set of gauge and yukawa couplings at the scale @xmath59 . in our analysis we use two loop renormalisation group ( rg ) equations for the gauge and yukawa couplings together with two loop rg equations for @xmath73 and @xmath74 as well as one loop rg equations for @xmath75 . at the next stage of our analysis we fix @xmath76 and @xmath77 , where @xmath78 and @xmath79 are the vevs of the higgs fields @xmath42 and @xmath43 , and choose @xmath80 , @xmath81 and @xmath61 so that the ew symmetry breaking constraints are satisfied , i.e. @xmath82 where @xmath83 is a higgs effective potential . although the correct ewsb is not guaranteed in the considered model , remarkably there is always a solution for sufficiently large values of @xmath46 which drive @xmath47 negative . finally at the last stage of our analysis we vary yukawa couplings , @xmath84 and @xmath76 to establish the allowed range of the parameters and qualitative pattern of the particle spectrum in the e@xmath2ssm . the results of our study of the particle spectrum in the e@xmath2ssm are summarised in the table 1 . we find that a set of the lightest susy particles in the ce@xmath2ssm always includes the gluino @xmath85 , the two lightest neutralino ( @xmath86 and @xmath87 ) and the lightest chargino @xmath88 . all other susy particles can be substantially heavier . nevertheless there exists a part of the e@xmath2ssm parameter space where exotic quarks , exotic squarks , inert higgs bosons and inert higgsinos can be also relatively light ( see scenario ii ) . if neutralino , chargino and gluino are the only lightest susy particles then one can expect to observe @xmath89 , @xmath90 , @xmath91 and @xmath92 pair production at the lhc . since each gluino will decay further into quark , antiquark and lightest neutralino ( or chargino ) the gluino pair production will result in an appreciable enhancement of the cross section of @xmath93 . if all squarks are much heavier than the gluino , i.e. particle spectrum has a very hierarchical structure , then gluinos can be relatively long lived particles because @xmath94 in the considered case . in particular their lifetime can be comparable with the lifetime of @xmath95 and @xmath96 bosons . at the same time the pair production of the second lightest neutralino also gives rise to a remarkable signature . when @xmath86 , @xmath87 and @xmath88 are the lightest susy particles the second lightest neutralino will decay predominantly into lepton , antilepton and @xmath86 . thus @xmath92 pair production would generate an excess in the cross section of @xmath97 that should be possible to observe at the lhc . other possible manifestations of the e@xmath2ssm at the lhc are related with the presence of a @xmath44 and of exotic multiplets of matter . for instance , @xmath44 boson can be discovered at the lhc if its mass is less than @xmath98 . when the yukawa couplings of the exotic particles have a hierarchical structure some of the exotic fermions can be relatively light so that their production cross section at the lhc can be comparable with the cross section of @xmath99 production . assuming that @xmath27 and @xmath26 couple most strongly with the third family quarks and leptons the lightest exotic quarks decay into either two heavy quarks @xmath100 or a heavy quark and @xmath69lepton @xmath101 , where @xmath102 is either a @xmath68- or @xmath67-quark . this can lead to the substantial enhancement of the cross section of either @xmath103 or @xmath104 . the discovery of the @xmath44 and exotic quarks predicted by the e@xmath2ssm would represent a possible indirect signature of an underlying @xmath0 gauge structure at high energies and provide a window into string theory . 9 s.f . king , s. moretti , r. nevzorov , `` theory and phenomenology of an exceptional supersymmetric standard model '' , phys . d * 73 * ( 2006 ) 035009 ; s.f . king , s. moretti , r. nevzorov , `` exceptional supersymmetric standard model '' , phys . b * 634 * ( 2006 ) 278 .	we present a self consistent @xmath0 inspired supersymmetric model with an extra @xmath1 gauge symmetry under which right - handed neutrinos have zero charge . we explore the particle spectrum within the constrained version of this exceptional supersymmetric standard model ( e@xmath2ssm ) and discuss its possible collider signatures .
in chirally stacked few layer graphene systems @xcite two sublattice sites , one in the top layer and one in the bottom layer , are disconnected from the near - neighbor interlayer hopping network and partially isolated . this geometric feature leads to a gapless semiconductor with weakly dispersive conduction and valence band edges , and therefore to an enhanced role for electron - electron interactions @xcite . the low - energy sublattice degree - of - freedom in chirally stacked @xmath1 layer graphene can be viewed as a @xmath2 pseudospin , with @xmath3-pseudospins corresponding to top layer states and @xmath4-pseudospins to bottom layer states . in this language the conduction and valence band eigenstates at wavevector @xmath5 have @xmath6 plane pseudospin orientations , _ i.e. _ they have equal weight in top and bottom layers , and azimuthal angles @xmath7 and @xmath8 respectively . a number of years ago we predicted@xcite on the basis of an unrestricted hartree - fock mean - field - theory that these circumstances would lead to ground states with pseudospin order , and in particular to a state in which the pseudospins rotate out of the @xmath6 plane by a @xmath9-dependent polar angle toward either the north or the south pole , breaking layer - inversion symmetry @xcite and inducing a gap @xcite in the charged excitation spectrum . recent bilayer experimental work @xcite has yielded clear evidence for broken symmetry states , although a consensus on the character of their order has not yet been achieved . at the same time a substantial body of theoretical work @xcite has used a variety of different approaches to study bilayer graphene . the general consensus of this work is that the ground state is a pseudospin ferromagnet . it has become clear however that because of quantum fluctuations that are not captured by mean - field theory , the gapped state competes closely with a gapless anisotropic state in which the net pseudospin alignment is in the @xmath6 plane , not the @xmath10 direction . this paper addresses that competition . as we will explain in more detail later , some features of the bilayer graphene system present awkward obstacles to theory . first among these is the property that the conduction and valence bands are weakly dispersive only over the small part of the brillouin zone where inter - layer tunneling plays an essential role . elsewhere in the brillouin zone interaction effects are expected to be similar to those in single - layer graphene , producing strong quasiparticle velocity enhancements,@xcite and spectral weight rearrangements,@xcite but not order . the small number ( @xmath11 ) of @xmath12-electrons per carbon atom in the strongly correlated region of momentum space is an obstacle for theories of the ordered state that are based primarily on a lattice model of graphene . the narrow momentum - space distribution of the most important virtual states also creates difficulties for continuum models because it makes the physics sensitive to the long - range of the coulomb interaction . in this paper we report on calculations in which lattice effects are completely ignored and the long - range of the coulomb interaction is not treated explicitly . in our view the continuum approach we use is strongly motivated by the low - density of strongly correlated electrons . our use of short - range interactions can be crudely justified by appealing to screening considerations , and arguing that the momentum - independent interaction parameters we use represent an average over the relevant portion of momentum space . although not fully rigorous in this sense , we believe that the conclusions we reach by following this approach systematically are nevertheless valuable and consistent with recent experimental work.@xcite this paper elaborates on ideas presented in a series of closely related earlier papers @xcite in which the ground state is estimated on the basis of perturbative renormalization group ( prg ) calculations . these calculations differ among themselves most essentially in the approximate ways they account for lattice scale physics and long - range interactions . in section ii we briefly describe the model we study and restate some identities and observations that are useful in developing its perturbation theory . section iii presents new results for the prg interaction flows of this model , emphasizing the role of exchange interactions between electrons with the same spin and valley states and commenting on the flavor number ( @xmath1 ) dependence of the interaction parameter flows . in section iv we explain how the prg pseudospin susceptibilities , whose divergences suggest the character of the low - temperature order , depend on the renormalized interaction strengths . finally in section v we present our conclusions and discuss the relationship of our work to experiments and to other theoretical work on this system . bilayer graphene ( blg ) is described approximately at low - energies by the @xmath13 version of the chiral band hamiltonian:@xcite @xmath14c_{{\bm q}\beta i}.\ ] ] in eq . ( [ eq : band ] ) @xmath15 are pauli matrices which act on the ( greek ) layer labels and @xmath16 is a spin - valley flavor label . we have used the notation @xmath17 and @xmath18 . here @xmath19 labels @xmath20 and @xmath21 valleys located at inequivalent brillouin zone ( bz ) corners and @xmath22 is wavevector measured from the bz corner . it is convenient to perform a rotation in blg pseudospin space@xcite in the @xmath23 valley to eliminate the valley dependence of the band hamiltonian , in order to make our discussion more concise in the following sections . the blg chiral band hamiltonian applies @xcite at energies smaller than the interlayer hopping scale @xmath24 mev and larger than the trigonal warping scale @xmath25 mev . we discuss the important role of trigonal warping in the band - structure later . the physics discussed in this paper applies to bilayers when @xmath13 and partly generalizes to @xmath26-layer chirally ( abc ) stacked graphene multilayers,@xcite although the chiral band hamiltonian @xmath27 applies over narrower ranges of energy for larger @xmath26 . we will focus our attention on the @xmath13 case hereafter . @xmath28 can be viewed as specifying a momentum @xmath29-dependent effective magnetic field that acts on the bilayer layer - pseudospin degree - of - freedom : @xmath30 where @xmath31 with @xmath32 , and the orientation angle of @xmath33 is @xmath34 . the corresponding matsubara green s function is @xmath35^{-1 } = { \mathcal g}_{{\bm q}\rm s}(i\omega_{\rm n})+{\mathcal g}_{{\bm q}\rm t}(i\omega_{\rm n } ) \ ; { \bm n } \cdot { \bm \sigma}\,\ ] ] where @xmath36 and @xmath37 . note that @xmath38 whereas @xmath39 . when expressed explicitly as a @xmath40 matrix , @xmath41 the off - diagonal _ triplet _ component captures processes in which electrons propagate between layers ; its momentum - orientation dependent phase factor plays an essential role in the one - loop prg calculations described in the next section . the frequency sums which appear in loop diagrams are readily evaluated : @xmath42 these two results express the property that fluctuations involve interband transitions . in the prg each loop diagram is multiplied by appropriate interaction constants ( discussed below ) and then integrated over momentum labels near the model s flowing cutoff @xmath43 : @xmath44 where @xmath45 is the blg density - of - states per flavor . the second frequency sum in eq.([eq : freq_sum_gf ] ) vanishes in the long - wavelength static limit because it has contributions from intra - band particle - hole excitations only . because @xmath46 rather than @xmath47 , this integral diverges logarithmically when the high - energy cut - off is scaled down by a factor of @xmath48 . the prg calculation is therefore very similar to the corresponding calculation for a 1des . this rather surprising property of blg is directly related to its unusual band structure which yields fermi points rather than fermi lines in neutral systems , and quadratic rather than linear dispersion.@xcite in systems with short - range interactions , finite momentum ( gradient ) corrections are irrelevant in the rg flows . we therefore ignore the wave - vector transfer dependence of scattering processes starting already at the @xmath13 chiral model s ultraviolet cut - off energy @xmath49 . in the low - energy continuum model of blg electrons carry spin , and both layer and valley pseudospin labels . in a scattering event , both the two incoming and two outgoing particles can therefore have one of eight labels and the general scattering function therefore has @xmath50 distinct values , even when gradient corrections are ignored . the number of distinct coupling constants in the rg flow equations is much smaller , however , because many values are zero and others are related to each other by symmetry . one simplification is that interactions conserve spin and , in the continuum model , both layer and valley pseudospin at each vertex . the role of spin in our model is therefore equivalent to the role of valley . we do however choose to allow the interaction strength between electrons in different ( @xmath51 ) layers to be weaker than that between electrons in the same ( @xmath52 ) layer so that the interactions are layer - pseudospin dependent . the internal loops in the prg calculation contain two fermion propagator ( green s function ) lines . because we seek a momentum - independent effective interaction we simplify the calculation by evaluating the coarse - graining correction to the effective interaction for the case where all incoming and outgoing electrons are at the dirac point . ( as we comment in the discussion section , this approximation is made less innocent by the layer pseudospin structure of the band eigenstates . ) these propagators conserve both spin and valley pseudospin , but as we have seen above , not the layer pseudospin . from eq.([eq : phasefactor ] ) we see that a phase factor @xcite @xmath53 appears when a propagator transfers electrons between layers . unless these layer transfers enter an equal number of times in each direction , the integrand in a feynman diagram will contain a net phase factor related to chirality and vanish upon angular integration of the virtual momentum . the total layer number ( the @xmath10 component of layer pseudospin ) is therefore also conserved.@xcite this property combined with flavor invariance limits the number of independent interaction parameters to three , two for interaction processes involving electrons in the same layer or different layers that conserve layer index , and one for interaction processes between electrons in different layers accompanied which interchange layer indices . we refer to the three distinct processes , illustrated in fig . [ fig : gamma ] , as @xmath54 , @xmath55 and @xmath56 respectively . if flavor invariance is broken , either due to microscopic lattice effects that violate @xmath57 invariance or due to broken symmetries , the number of interaction parameters increases . for example , if the bare interactions are spin and valley dependent in an arbitrary way , the number of independent _ antisymmetrized _ interaction parameters increases from three to ten:@xcite @xmath58,@xmath59 , @xmath60 , @xmath61,@xmath62 , @xmath63 , @xmath64 , @xmath65 , @xmath66 and @xmath67 . in this notation the first label follows the convention explained above for the layer pseudospin dependence of the interaction , whereas the second and third labels refer to valley and spin respectively , and distinguish the interaction between particles with the same ( @xmath52 ) or different ( @xmath51 ) spin or valley . the number of entries in this list of interactions has been reduced by two by appealing to fermi statistics for the case of @xmath68 interactions , those between particles with the same spin and valley labels . in that case interchanging the labels of outgoing electrons yields an indistinguishable and canceling amplitude for particles in the same layer . although we can choose to retain @xmath69 in the rg flows , it can not contribute to any physical observable . similarly , since @xmath61 and @xmath70 are interchanged by reversing outgoing labels they differ only by a sign and it is sufficient to retain one parameter . the number of independent interacting parameters in the analysis by vafek @xcite and by lemonik _ et al._,@xcite who impose the symmetry of the underlying blg crystal but do not explicitly account for antisymmetrization is nine . for @xmath57 invariant models the number of parameters is three , but antisymmetrization allows us to retain only one coupling constant in the single flavor ( @xmath71 ) case , and the physical implications of the flowing interactions are more apparent if we do so.@xcite a convenient choice @xcite is to retain @xmath61 . we take the view that the underlying lattice @xcite is unlikely to play an important role in determining whether the broken symmetry state is gapped @xcite or gapless,@xcite though it is key to selecting @xcite between distinct gapped states @xcite that are equivalent in an @xmath57 invariant model . therefore , the number of interactions parameters that appear in the calculation described below is three in the general case with unbroken flavor invariance and one for the special case of @xmath71 . in this subsection we explore the similarities and differences between blg and 1des s by temporarily neglecting the spin and valley degrees - of - freedom . we choose the single effective interaction parameter to be @xmath72 . a prg analysis determines how the bare interaction at the @xmath73 scale , @xmath74 , is renormalized by integrating out the high energy fermion degrees of freedom . at one loop level the coarse - graining contributions to the effective interaction are described @xcite by the three higher order diagrams labeled zs , zs , and bcs . the internal loops in these diagrams are summed over the high - energy labels . the main merit of the prg is that it treats all virtual processes on an equal footing , it is well known that in a 1des the zs loop vanishes while the zs and bcs diagrams cancel , implying that the interaction strength does not flow to large values and hence that neither the cdw repulsive interaction nor the bcs attractive interaction instabilities predicted by mean - field theories survive the quantum fluctuations they neglect . the key features of blg order physics can be understood in terms of two properties of these one - loop diagrams ; ( i ) the particle - particle ( bcs ) and particle - hole ( zs , zs ) loops have the same logarithmic divergences as in the 1des case in spite of the larger space dimension and ( ii ) the zs loop , which vanishes in the 1des case , is finite in the blg case and the bcs loop vanishes instead . both of these changes are due to the layer pseudospin triplet contribution to the single - particle green s function as we explain below . the net result is that interactions flow to strong coupling even more strongly than in the mean - field approximation . the key step in one - loop prg calculations is identifying the coupling factors attached to the loop in each diagram . since only opposite layer interactions need to be retained for @xmath71 models , all scattering functions have two incoming particles with opposite layer labels and two outgoing particles with opposite layer labels , much like the 1des case.@xcite the external legs in the scattering function feynman diagrams are labeled by the layer index ( @xmath75 top layer and @xmath76 bottom layer ) . the corresponding labels for the 1des case @xcite are the chirality , @xmath77 for right - going and @xmath78 for left - going . we refer to these as the single - particle labels below when a comment refers to both 1des and blg cases . as shown in fig . [ fig : zszs1](a ) , at the upper vertex of the zs diagram the incoming and the outgoing @xmath79 particles induce a @xmath80 particle - hole pair in the loop , while the incoming and outgoing @xmath80 particles at the lower vertex induce a @xmath79 particle - hole pair . the zs contribution is absent in the 1des case @xcite because propagation is always diagonal in interaction labels . however , this correction survives for blg because the single - particle green s function has a triplet contribution ( see eq . ( [ eq : freq_sum_gf ] ) ) which is off - diagonal in layer index . here we find @xmath81 the zs loop shown in fig . [ fig : zszs1](b ) corresponds to repeated interaction between a @xmath79 particle and a @xmath80 hole . this is the channel responsible for the 1des mean - field cdw instability @xcite in which coherence is established between @xmath77 and @xmath82 particles.@xcite in both 1des and blg cases it has the effect of enhancing repulsive interactions . its evaluations for the two cases correspond quite closely , because this loop diagram involves only particle - propagation that is diagonal in interaction label . we find that @xmath83 the bcs channel corresponds to repeated interaction between the two incoming particles . in the 1des case the contribution from this loop ( see fig . [ fig : bcs1 ] ) cancels the zs contribution@xcite , leading to marginal interactions and luttinger liquid behavior . this same kind of bcs correction for blg reads @xmath84 in the blg case there is an additional bcs loop contribution ( see fig . [ fig : bcs2 ] ) in which the incoming @xmath79 and @xmath80 particles both change layer labels before the second interaction . this contribution is possible because of the triplet pseudospin propagation and , in light of eq . ( [ eq : freq_sum_gf ] ) , gives a bcs contribution that has the opposite sign compared to the normal contribution : @xmath85 it follows that the bcs loop contribution is absent in the blg case because @xmath86 therefore , at one - loop level , the renormalization of interlayer interaction is @xmath87 these results and the comparison with 1des s are summarized in table [ table : one ] and imply the following rg flow equation for blg : @xmath88 if we follow the flows from the model s microscopic cutoff we obtain @xmath89 where @xmath74 is the bare interaction . @xmath72 diverges when @xmath90 . this instability criterion is similar to the stoner criterion for ferromagnetism . c c c diagrams & 1deg ' '' '' & blg + & @xmath91 ' '' '' & @xmath92 + _ zs _ & @xmath93 ' '' '' & @xmath92 + _ bcs _ & @xmath94 ' '' '' & @xmath91 + mean field & @xmath95 & @xmath92 + quantum fluctuations & @xmath96 & @xmath92 + full one - loop & @xmath97 & @xmath98 + [ table : one ] the interaction correction to the layer pseudospin response function @xmath99 is obtained by closing the scattering function with a @xmath100 vertex at top and bottom . the @xmath100 operator measures the charge difference between @xmath79 and @xmath80 layers . because it is an effective single - particle theory , fermion mean - field theory corresponds to response function diagrams with at most a single particle - hole pair . it follows that mean - field theory in blg is equivalent to a single - loop prg calculation in which the bcs and zs channels are neglected and only the zs channel is retained . in mean - field theory , ideal blg has an instability to a state in which charge is spontaneously transferred between the layers which is signaled by the divergence of @xmath99 . ( @xmath101 also diverges in mean - field theory but less strongly . ) the prg analysis demonstrates that the mean - field theory instability is enhanced by the reinforcing influence of the zs channel contribution . we discuss pseudospin response functions which indicate the character of the broken symmetry state further for the physically relevant @xmath102 case in section iv . assumming @xmath57 invariance , the one - loop flow equations for @xmath102 are derived in the same way as for the @xmath71 case , except that we need to keep track of three interaction parameters , @xmath103 ( @xmath104 ) , @xmath72 ( @xmath105 ) and @xmath106 ( @xmath107 ) . a tedious but elementary book - keeping exercise yields the following results : @xmath108 this equation is the @xmath57 invariant version of the more general ten parameter flow equations derived in ref . . the rg flow equations derived previously in ref . differ only in notation . the @xmath71 case is recovered by noting that @xmath103 has no physical effect for interactions among particles of the same flavor and @xmath72 and @xmath106 always enter in the combination @xmath109 . defining @xmath110 , it follows from eqs.([eq : full_rg_flow ] ) that @xmath111 yielding a one - parameter flow equation for the @xmath71 case . note that this @xmath112 also represents @xmath61 in the @xmath113 case.@xcite the only fixed point of these rg flow equations is the non - interacting one . the character of the broken symmetry state can be estimated by following the rg flows to strong interactions , or more reliably by following the flow equations for appropriate susceptibilities as discussed in the next section . in order to represent the property that bare same layer interactions should be slightly stronger than the bare different layer interactions and that the bare layer interchange interaction vanishes , we start the rg flow integrations from @xmath114 when the flows start from an interaction with these properties , the coupling constants diverge simultaneously at a finite value of @xmath115 , and for @xmath116 satisfy the following properties at the divergence point : @xmath117 the value of the first ratio in eq.([eq : rg_results ] ) follows from the rg flow equation for @xmath118 which satisfies @xmath119 the rate of growth of @xmath118 is proportional to @xmath120 , whereas @xmath103 and @xmath120 grow like @xmath121 . equivalent results in a different notation , using @xmath122 and @xmath123 , have been derived previously by vafek and yang.@xcite these results do not depend on the bare coupling constants as long as the physical conditions specified by eq.([eq : bare ] ) are satisfied . note that the rg flows for @xmath113 are still qualitatively different from those of the @xmath124 limit . typical rg flows , obtained by integrating eq.([eq : full_rg_flow ] ) numerically , are plotted in fig . [ fig : rgflow ] . we note that , when combined with the mean - field - theory of the gapped state,@xcite the value @xmath125 yields values of the spontaneous gap , the critical temperature , and the carrier density at which order is absent that are consistent with some recent experimental observations@xcite . we therefore choose to use the bare interaction parameters @xmath125 and @xmath126 to illustrate typical rg flows . we find that the interaction parameters flow away from the non - interacting fixed point and diverge at a finite value of @xmath48 . the instability criterion implied by the one - loop rg calculation is @xmath127 for @xmath113 , i .. e the instability tendency is enhanced compared to the @xmath71 case . ( although the @xmath103 , @xmath72 and @xmath106 interaction parameters do diverge at a similar value of @xmath115 in the @xmath71 case , they have no physical effect because they enter observables in the combination @xmath128 . ) because of the quadratic band dispersion , the ratio of the interlayer coupling scale to the gap ( @xmath129 ) can be associated @xcite with the square of the momentum scaling factor @xmath48 at the phase transition . this yields a value of @xmath130 on the order of @xmath131 mev , close to the scale of the gap values seen in some experiments.@xcite more insight into the likely nature of the broken symmetry state which occurs in blg can be obtained by using the prg calculations to estimate the long wavelength static limit of the pseudospin susceptibilities : @xmath132 where @xmath133 label layer - pseudospin components ( instead of top and bottom layers ) and @xmath134 label either a unit matrix ( @xmath135 ) or a generator of rotations ( @xmath136 ) in the @xmath137-component valley - spin flavor space : @xmath138 a physically transparent choice@xcite for the @xmath57 generators is to set @xmath139 to a pauli matrix which acts on the spin degree - of - freedom for @xmath140 , to a pauli matrix which acts on the valley degree - of - freedom for @xmath141 , and to a product of spin and valley pauli matrices for @xmath142 . it follows from @xmath57 invariance@xcite that the normal state susceptibilities in this case are @xmath143 and equal for all values of @xmath144 . below we refer to the @xmath135 susceptibility ( @xmath145 ) as the flavor charge susceptibility and to the @xmath146 susceptibility ( @xmath147 ) as the flavor spin susceptibility . the conservation of particle - number in each layer at long wavelengths , explained previously , implies that @xmath148 and that @xmath149 . divergences in @xmath150 signal ordered states in which the charge density is transferred between layers in the same sense for all flavors , while divergences in @xmath151 signal flavor dependent layer - inversion symmetry breaking with no overall charge transfer . divergences in @xmath152 for @xmath153 , on the other hand imply broken symmetry states in which the phase relationship between layers is altered and rotational symmetry within the layers is broken . broken layer - inversion symmetry states have an energy gap for charged excitations and large momentum space berry curvatures that lead to large anomalous hall effect contributions from individual flavors.@xcite these states will be referred to below as spontaneous quantum hall states.@xcite the broken rotational symmetry state will be referred to below as the nematic state . provided that the system has a single continuous phase transition , the ordered state character can in principle be determined by identifying the pseudospin susceptibility which diverges first as temperature is lowered . although we can not make that determination definitively from the rg interaction strength flows , we can obtain some guidance by examining how the interaction contributions to the susceptibilities grow under the rg flows . the pseudospin response functions of interest can be related to the renormalized interactions as summarized in the feynman diagrams of fig . [ fig : susc ] . fig . [ fig : susc](a ) represents the non - interacting pseudospin susceptibility @xmath154 and fig . [ fig : susc](b ) represents the interaction correction @xmath155 . in the long wavelength static limit the explicit expressions for these susceptibilities are : @xmath156 note that in these equations all propagators are diagonal in the implicit spin - valley flavor labels . in the noninteracting case , the inter - flavor ( d ) susceptibilities vanish because the single - particle hamiltonian is diagonal in flavor and the intra - flavor ( s ) susceptibilities are positive because of the band state pseudospin structure : @xmath157 these pseudospin susceptibilities capture the contribution of momentum @xmath158 vertical interband quantum fluctuations involving states with energies measured from the dirac point between @xmath73 and @xmath159 . the factor of two difference between @xmath160 and @xmath161 demonstrates that the band state is more susceptible to a gap opening perturbation than to a nematic perturbation . this property can be understood @xcite in terms of the pseudospin orientations of the band states ; in particular a @xmath10 pseudospin effective field is perpendicular to the valence band pseudospin for all momentum orientations , so that @xmath10 direction fields always yield a strong response . on the other hand @xmath6-plane pseudospin fields are not in general perpendicular to the valence band pseudospin and consequently produce a weaker response averaged over momentum orientations . the larger response to a @xmath10 pseudospin effective field is related to the well - known property of blg that a potential difference between the top and bottom layers lead to an energy gap at the dirac point . the corresponding charge susceptibility ( density - density channel ) , obtained by replacing the pseudospin - field pauli matrix by an identify matrix , vanishes because conduction and valence band states are orthogonal . when the interactions flow to strong values , the susceptibilities are dominated by their interaction contributions . when these are included we find that @xmath162 note that the large susceptibility in @xmath163 is attribute to the layer interchange processes @xmath106 while the strong divergences in @xmath164 are due to intra - flavor exchange interactions , recalling @xmath165 in eq.([eq : intra ] ) . for the @xmath71 case there is no different - flavor response and we therefore find that near the instability @xmath166 suggesting that a gapped spontaneous quantum hall state is most likely . for the physical @xmath113 case we can relate the flavor charge and flavor spin response functions to @xmath167 and @xmath168 by examining the particular case of responses to perturbations that are diagonal in flavor index , _ i.e. _ perturbations proportional to @xmath169 , @xmath170 or @xmath171 . we find that near the divergence point @xmath172 @xmath173 have the same relations to @xmath174 and @xmath175 . combining eq.([eq : inter ] ) and eq.([eq : intra ] ) with the results from eq.([eq : rg_results ] ) for @xmath113 we find that @xmath176 the property that all susceptibilities diverge together is an artifact of the single - loop rg calculation which becomes questionable when the scaled interactions are strong . still , the fact that the @xmath177 diverges most strongly suggests that the spin - channel gapped state is more likely than the charge channel nematic state . ( negative coefficients in eq.([eq : n4n ] ) imply that one - loop interaction corrections tend to reduce the susceptibilities @xmath178 and @xmath179 rather than to increase them and therefore do not suggest instabilities . ) in the @xmath180 case , which accounts for only spin or valley but not both , @xmath181 the @xmath137 dimensional @xmath182-matrices are replaced by @xmath183-dimensional pauli matrices . we find that @xmath184 which equally applies to the @xmath101 channel . combining the @xmath180 results in eq.([eq : rg_results ] ) with eq.([eq : inter ] ) and eq.([eq : intra ] ) we obtain @xmath185 in both @xmath116 cases interactions favor breaking layer - inversion symmetry in the flavor spin channel and orientational symmetry in the flavor charge channel . the strongest divergence occurs for the flavor spin channel broken layer - inversion symmetry state.@xcite the @xmath71 susceptibility and the flavor charge and spin channel susceptibilities for @xmath116 are illustrated in fig.[fig : susflow ] and agree with the above analytic analysis . because @xmath186 is always the most _ positvely _ divergent channel we conclude that the dominant many - body effect in blg is spin - channel spontaneous layer - inversion symmetry breaking yielding blg states with spontaneous charge gaps . these results verify our original theoretical predictions @xcite and are consistent with recent experimental observations in dual - gated ultra - clean suspended blg devices.@xcite finally , we point out that an interlayer potential breaks @xmath57 invariance in the @xmath113 case , leading to a range of stability for a state in which three flavors are polarized toward one layer and one flavor toward the other layer when an ordered state appears . for instance , a possible mass term in a mean - field theory with broken layer - inversion symmetry can be proportional to @xmath187 . this gapped state breaks parity inversion and time reversal symmetries , favored by small interlayer electric and magnetic fields @xcite that break @xmath57 symmetry . the difference between this state and other spontaneously broken symmetry states can be further appreciated as follows in our full model with _ approximated _ @xmath57 symmetry . this state reduces the full symmetry to @xmath188 with three goldstone modes , in contrast to the three possible states in flavor spin channel that break the symmetry down to @xmath189 with four goldstone modes , and there is no soft mode in the flavor charge channel . in this paper we have attempted to shed light on the pseudospin ferromagnet ground state of blg by carrying out a conventional perturbative renormalization group calculation for a model with @xmath57 invariance in spin - valley space , and examining interaction corrections to spin and charge pseudospin susceptibilities . the prg flows lead to strong layer interchange scattering processes which favor gapless nematic ( @xmath6-plane pseudospin ferromagnet ) ground states , competing with the gapped ( @xmath10-direction pseudospin ferromagnet ) states favored by intra - flavor exchange interactions and predicted @xcite by mean field theory . the role of exchange is especially dominant in the single - flavor @xmath71 case in which the prg flow equations lead to divergent interactions which vanish when projected onto the many - fermion hilbert space . for @xmath71 the interactions enter observable only in the combination @xmath128 that is uniquely allowed by the pauli exclusion principle . we conclude that the most likely ground states are gapped spin - channel ising pseudospin ferromagnets , in agreement with some @xcite recent experiments , although the competition becomes closer when the number of flavors is larger , and quite close for the physically relevant @xmath113 case . according to our estimates , broken symmetry states appear at energy scales slightly larger than those at which trigonal warping effects in the single - particle hamiltonian @xcite start to have a large influence on the band structure , splitting the vorticity @xmath13 vortex of the chiral model into four @xmath190 dirac points . the strength of our conclusions is limited by our neglect of the long - range character of coulomb interactions and by the fact that our one - loop prg calculations become unreliable once the effective interactions flow to strong values , _ i.e. _ once @xmath191 . the details of our calculations are conventional and in particular evaluate the loop corrections to scattering amplitudes for the case in which both incoming and out - going momenta are at the dirac point . this simplification is normally justified by the irrelevance of gradient corrections to the effective interactions . as we now explain , in the case of blg , this approximation likely overestimates the strength of the @xmath56 layer interchange interactions . consider for example the zs process which is non - zero because of the triplet layer off - diagonal contribution to the green s functions , as we have approximated in eq.([eq : zs ] ) . when the incoming electrons have finite momenta @xmath192 and @xmath193 , the momentum arguments of the virtual states in the loop diagram are @xmath194 and @xmath195 . the triplet propogator phase factors which are integrated over @xmath196 in the loop correction to @xmath56 therefore becomes @xmath197 where @xmath198 . notice that this phase factor has a singular dependence on @xmath192 and @xmath193 at small @xmath196 and that the approximation we have used previously , in which we replace @xmath199 by zero , is valid only for @xmath200 . when the phase factors are taken into account the @xmath56 scattering amplitude has an additional phase factor which will average to a value smaller than one in physical observables like the pseudospin susceptibilities . it therefore appears to us that our calculation likely overestimates the strength of the interactions which favor the nematic state . the gapped states have the interesting property that they support large momentum space berry curvatures,@xcite which can lead to interaction - induced quantum hall effects . the regions of momentum space near both k and k valleys flavor contribute @xmath201 to the hall conductivity for each spin . when all flavors make a contribution of the same sign , the total hall conductivity is exactly @xmath202 . this gapped bilayer state is therefore a quantized anomalous hall ( qah ) insulator @xcite with chiral edge states . because of its nontrivial overall orbital moment@xcite the qah state is favored by a perpendicular magnetic field . the state in which the sense of layer polarization is opposite for opposite spins and for opposite valleys does not break time - reversal symmetry and has a mean - field interaction terms with an effective spin - orbit coupling @xcite that leads to a quantum spin hall ( qsh ) state with zero charge hall conductivity . the qsh state has helical edge states protected by time - reversal symmetry . its @xmath203 classification in an interacting @xmath1-layer chiral graphene is given by@xcite @xmath204 . when lattice effects @xcite are taken into account , inter - valley exchange weakly favors the layer - antiferromagnet ( laf ) @xcite state . ( that observation does not definitively make the case for this particular state , since inter - valley exchange could be less important than correlation effects . ) for a laf state , @xmath205 even though time - reversal symmetry is broken by opposite spin polarizations on the top and bottom layers . in a continuum model that neglects lattice effects and has @xmath57 symmetry , the three states are all members of the continuous family of degenerate broken symmetry states that is signaled by divergence of @xmath186 . in addition to their difference in edge state properties , the qah , qsh and laf states also exhibit different responses @xcite to zeeman fields which couple to spin and can be realized in the absence of orbital coupling by applying magnetic fields parallel to graphene layers . in the qah case , the ground state is unchanged but the quasiparticle gap is reduced , vanishing when the zeeman - coupling strength is equal to the ground state gap via a mechanism reminiscent of the clogston limit in superconductors . the qsh and laf states respond to zeeman fields by establishing noncollinear spin states within each valley and evolving toward an unusual kind of exciton condensate in the strong zeeman - coupling limit . the quasiparticle gap of qsh and laf states is however independent of zeeman - coupling strength drawing a sharp distinction with the qah case . it is therefore possible to use zeeman responses and edge state signatures to identify the character of the bilayer ground state experimentally.@xcite on the other hand , the three states respond very similarly to an electric field between the layers , which can induce first order transitions at which the total layer polarization jumps.@xcite this state favored by an interlayer electric field is a quantum valley hall state with zero hall conductivity but , in the absence of scattering which mixes valleys counterpropagating valley - resolved edge states . we note that a small electric field between the layers can also possibly stabilize a state in which one flavor is polarized in a sense opposite to the other three and charge , valley , and spin hall conductivities are all nonzero.@xcite one consequence of the above mentioned spontaneous quantum hall effects @xcite in each valley is that the gapped states have a very simple adiabatic evolution with magnetic field @xmath80 in which the gap remains open , but the carrier density at which the gap appears varies with field . the total filling factor inside the gap at finite @xmath80 is @xmath206 for a state with a hall conductance equal to @xmath207 at zero field . the topological character @xcite of the gapped states in blg is induced @xcite by weak bare electron - electron interactions and is enhanced by the nearly flat bands with chiral pseudospin textures , and is distinct from the @xmath203 topological insulators @xcite that require strong intrinsic spin - orbital coupling . although interaction - induced topological states apparently do not occur in single - layer graphene , as proposed @xcite on the basis of extended hubbard models with at least finite interaction strength , it appears quite possible that they do occur in bilayers via a weak interaction instability.@xcite there are three previous articles @xcite which discuss similar rg calculations . reports equivalent rg flow equations for the @xmath57 invariant model in terms of interaction parameters that are related to ours by @xmath208 and @xmath123 . several references have started their rg flows from effective interactions which include symmetry - allowed corrections due to lattice effects . ref . attempts to account for the long - range of the coulomb interaction by explicitly accounting for screening and then replacing the coulomb amplitude by an average value , and estimates the ground state by applying mean - field theory to interactions from prg flows that have been truncated at an intermediate interaction strength . all calculations recognize the competition between nematic and gapped states , but small differences in approximations or in estimates of bare coupling parameters have lead to different final conclusions . we note that the recent elaborations@xcite of the calculations which motivated nematic state proposals@xcite recognize that gapped states@xcite are also a possibility . although these calculations , like ours , shed light on the physics at play in determining the ground state , it is difficult to make definitive conclusions based on theory alone and we must ultimately appeal to experiment . here the situation is also confusing . some experimental studies @xcite of high - quality suspended bilayers appear to reveal gaps whose size , carrier - density - dependence , and temperature - dependence , is roughly consistent with mean - field - theory predictions.@xcite these observations are consistent with the conclusions of this paper . on the other hand this behavior is not observed in all samples , even in samples which appear to be similar to the gapped samples . one experimental study@xcite focuses on the magnetic - field @xmath80 dependence of the gap at filling factor @xmath209 in a sample which does not show a gap at @xmath210 . the anomalous weak @xmath80 persistence of the @xmath209 gap in this experiment appears to be consistent with earlier studies by yacoby and collaborators.@xcite although interpreted as evidence for the nematic study , the persistence of the @xmath209 gap is , it appears to us , equally consistent with a field - driven crossover or transition to a gapped @xmath211 state . although this gap is likely not the ground state at @xmath210 it is favored at filling factor @xmath209 , because its gap is then pinned to the fermi level . in this interpretation the increased conductance observed at weak fields in ref . would be associated with a first order transition @xcite between @xmath211 and @xmath212 states which induces a network of current - supporting @xcite domain walls.@xcite similar spontaneously broken symmetry physics@xcite is also likely in thicker high quality graphene films with chiral ( abc ) stacking order.@xcite in an @xmath1-layer chiral stack,@xcite the low - energy sublattice sites are localized in the outermost layers , @xmath213 and @xmath214 . hopping occurs between these two sites via an @xmath1-step virtual process which leads to @xmath215 energy dispersions and an @xmath216 berry phase . for larger @xmath1 , the low - energy bands are increasingly flat and the pseudospin chirality larger , at least when weak remote hopping processes are neglected , leading to larger opening for many - body interaction effects . in the simplified chiral model , the density of states @xmath217 diverges as @xmath218 approaches zero for @xmath219 whereas it remains finite for @xmath180 . in the prg language , this difference corresponds to the fact that the short - range interactions at tree level are marginal for @xmath180 but relevant for @xmath219 . since @xmath220 at tree level , as discussed in section iii - c , it seems that gapped spontaneous quantum hall states @xcite which break layer - inversion symmetry and produce large momentum space berry curvatures can occur in multilayers as well . future experimental and theoretical work will be necessary to sort out the competition between interactions and remote hopping terms in the hamiltonian . fz has been supported by darpa under grant spawar n66001 - 11 - 1 - 4110 . ahm acknowledges valuable discussions with joel moore on the charge and spin susceptibilities of @xmath57 invariant systems . ahm was supported by welch foundation grant tbf1473 , nri - swan , and doe division of materials sciences and engineering grant de - fg03 - 02er45958 . we refer here to the valley - projected layer - inversion symmetry . this symmetry is broken in the sense that terms proportional to @xmath100 are generated by electron - electron interactions in the mean - field continuum theory . note that this layer - inversion symmetry is different than the parity inversion symmetry with the operator @xmath221 . the band hamiltoian of chiral graphene has parity inversion symmetry , as @xmath222 . all the spontaneously gapped states classified in ref . break the valley projected layer - inversion symmetry . however , the quantum anomalous hall and quantum spin hall states preserve the parity inversion symmetry . f. zhang , h. min , m. polini and a. h. macdonald , phys . rev . b * 81 * , 041402(r ) ( 2010 ) . also see the supplementary attached to arxiv:0907.2448 for the complete set of prg flow equations in the spinful and valleyful case . specifically we rotate by @xmath223 around the @xmath224-axis so that @xmath225 and @xmath226 . this choice does not change the interaction term in the hamiltonian , even in models that account for the difference between intralayer and interlayer interactions . it has the advantage of implying the conservation of the @xmath10 component of layer pseudospin that is used to simplify our diagrammatic calculations . however , we do have to remember to invert the transformation when we interpret the layer interchange interaction @xmath106 which appears in our calculations . the inter - valley @xmath106 scattering process is actually one in which two electrons in the same layer are both scattered to the opposite layer . it is of course possible to perform our calculations without making this transformation simply by replacing @xmath227 by @xmath228 in eq.([eq : phasefactor ] ) for valley k.	we use a perturbative renormalization group approach with short - range continuum model interactions to analyze the competition between isotropic gapped and anisotropic gapless ordered states in bilayer graphene , commenting specifically on the role of exchange and on the importance of spin and valley flavor degeneracy . by comparing the divergences of the corresponding susceptibilities , we conclude that this approach predicts gapped states for flavor numbers @xmath0 . we also comment briefly on the related gapped states expected in chiral ( abc ) trilayer graphene .
interactions of acceptor - bound holes with magnetic fields and strain are known for acceptor dopants in bulk silicon@xcite . in the @xmath12 subspace , interactions with magnetic fields @xmath189 are represented by the hamiltonian @xmath190 here , @xmath191 are @xmath188 matrices , c.p . refers to cyclic permutations , @xmath192 and @xmath193 are the linear and cubic land g - factors , and @xmath40 is the bohr magneton . interactions with strain @xmath194 are represented by the hamiltonian @xmath195 + b'((j_x^2-\tfrac{5}{4}i)\epsilon_{xx}+\textrm{c.p.})\nonumber\\+&(2d'/\sqrt{3})(\{j_x , j_y\}\epsilon_{xy}+\textrm{c.p.}),\end{aligned}\ ] ] where @xmath196 , @xmath197 , @xmath198 and @xmath199 are bir - pikus deformation potentials@xcite . in the spherical spin-3/2 basis @xmath200 , where @xmath201 is an effective total angular momentum@xcite , acceptor eigenstates take the form @xmath202 where @xmath203 is an effective total spin , and the spin-3/2 spin - orbit interaction has coupled states with @xmath204 . the @xmath205 and @xmath206 are radial envelope wavefunctions for envelope function spherical harmonics with @xmath207 and @xmath208 , respectively . the @xmath200 are found using the clebsch - gordan coefficients . states outside the @xmath11 subspace , given in ref . @xcite , contribute to envelope function asymmetry to realize the @xmath23 and @xmath24 terms in @xmath209 from the main text . the schrieffer - wolff calculation mixing in these higher excited states will be presented in a future paper . the low - energy holes are described in the @xmath12 basis by @xmath210 for @xmath211 , where @xmath39 , and @xmath41 for b : si@xcite . the unitary transform @xmath212 that diagonalizes @xmath213 is @xmath214 in the basis @xmath215 . here , @xmath61 and @xmath62 . applying this transformation we obtain an extended qubit hamiltonian for the spin qubit states and nearest excited spin states @xmath63 in the basis @xmath216 . here , the effective zeeman interactions @xmath64 appear on the off - diagonal . the effective zeeman interactions depend explicitly on the gate electric field due to the @xmath44-induced mixing of @xmath217 and @xmath47 . these off - diagonal zeeman interactions are @xmath218 the upper and lower @xmath219 blocks of @xmath220 are diagonalized by @xmath221 in the basis @xmath222 where , @xmath223 applying this transformation we obtain the following extended hamiltonian @xmath224 in the basis @xmath225 , where , @xmath226 our approximate qubit model from the main text takes @xmath74 to zeroth order to give qubit states @xmath227 and @xmath228 . in this approximation we obtain a qubit frequency @xmath229 a sweet spot occurs when the qubit frequency is insensitive to small fluctuations in electric fields . for an acceptor experiencing no static external applied field along @xmath230 and @xmath231 directions , this occurs for the roots of @xmath232 one root ( sweet spot ) is @xmath233 . substituting @xmath234 and @xmath235 , we obtain an equivalent condition @xmath236 for this sweet spot . another root ( sweet spot ) occurs at the roots of @xmath237 the sweet spot associated with the above root occurs at @xmath80 . the lowest order correction to @xmath31 due to coupling to levels @xmath238 is easily obtained from 2@xmath239-order perturbation theory , @xmath240 the exact solution in ( main text , fig . 2c , d ) shows that all higher order corrections ( including @xmath241 ) to the approximate solution presented in the main text do not qualitatively modify the qubit frequency . the total interaction with in - plane electric fields @xmath242 and @xmath243 described by @xmath30 and @xmath244 is @xmath245 in the basis @xmath215 , where @xmath246 and @xmath247 . the first matrix is the coupling due to the broken inversion symmetry of the interface and gate field , along the @xmath248 direction , while the second matrix describes the interaction due to the @xmath18 symmetry of the local field of the ion . rotated into the qubit basis using @xmath249 , we obtain @xmath250 in the basis @xmath225 , where , @xmath96 , @xmath97 , @xmath98 , and @xmath251 , and @xmath100 . edsr coupling @xmath105 in the qubit subspace arises from oscillating in - plane electric fields @xmath108 . we obtain the effective edsr interactions @xmath252 in the qubit basis using a schrieffer - wolff transformation . expanding in @xmath253 and @xmath254 and grouping terms according to powers in the electric fields @xmath255 and @xmath256 , we obtain @xmath257 the term proportional to @xmath258 is linear in the electric field and defines the electric dipole spin resonance term . substituting @xmath259 , @xmath260 , @xmath255 and @xmath256 we obtain the edsr matrix element @xmath261 because of the spin - orbit interaction , the electric dipole moment in each acceptor couples to spin . as a result , two qubits interacting only via mutual coulomb repulsion experience a spin - dependent interaction resembling a magnetic dipole - dipole interaction . here we determine this interaction in the coupled - qubit subspace @xmath262 . for the total hamiltonian we have @xmath263 , where @xmath264 is the single - acceptor hamiltonian for qubit @xmath265 , and @xmath266 is the electrostatic interaction between the qubits . we work in the tensor @xmath267 tensor product subspace of the two qubits @xmath268 where @xmath269 and @xmath270 , explicitly ignoring anti - symmetrization , _ i.e. _ , assuming spatial overlaps are negligible . without the coulomb interaction , the hamiltonian is @xmath271 where @xmath264 is given by equation ( [ eq : thop ] ) . meanwhile , in the direct product subspace the two - qubit coulomb interaction is @xmath272 when the separation @xmath273 between the acceptors is large compared to dipole moments @xmath274 of the system , we may use the multi - pole expansion of the coulomb interaction in equation ( [ eq : v12h ] ) . the lowest - order non - zero term is @xmath275\label{eq : d12d}\end{aligned}\ ] ] where @xmath276 @xmath277 is the position of qubit ion @xmath132 , and @xmath278 . the coulomb interaction is now a product of single - hole dipole matrix elements known in the basis @xmath279 from equation ( [ eq : hesup ] ) . the total hamiltonian is @xmath280 the combined effect of the coulomb and zeeman interactions can be projected into the coupled - qubit subspace @xmath262 using a schrieffer - wolff transformation . working out the effective interaction , to second order in off - diagonal terms and zeroth order in @xmath281 and @xmath282 , gives a spin - independent shift to all levels which does not influence the qubit physics . a transformation to third order in off - diagonal terms in the full @xmath267 space finds the spin - spin interaction of interest , @xmath283 which is an ising type spin - spin interaction . for convenience this can be re - written as @xmath284 where @xmath285 is the raising / lowering operator for qubit @xmath286 . for @xmath287 , we have contributions @xmath288 and @xmath289 from the @xmath230 and @xmath231 oriented dipoles , respectively such that the total coupling is @xmath290 , where @xmath291 @xmath292 , @xmath293 , @xmath294 , and @xmath295 . substituting @xmath296 , @xmath297 , we obtain @xmath298 such that @xmath299 the relaxation from @xmath300 to @xmath301 via emission of a phonon with energy @xmath302 can be determined from fermi s golden rule , @xmath303 where @xmath304 are the phonon polarizations , @xmath305 is the phonon wavevector , @xmath306 is the electron - phonon interaction , and @xmath307 is the phonon population . the deformation potential matrices @xmath308 are determined from the bir - pikus hamiltonian @xmath309\nonumber\\+&(2d'/\sqrt{3})[\{j_x , j_y\}\epsilon_{xys}+\textrm{c.p.}]\end{aligned}\ ] ] where @xmath197 , @xmath198 , and @xmath199 are bir - pikus deformation potentials@xcite and the strain @xmath310 of the phonon polarization @xmath311 is determined by the displacement@xcite @xmath312 where @xmath313 is the normalized phonon polarization vector@xcite , @xmath314 is the number of unit cells , @xmath315 is the unit cell volume , @xmath316 is the crystal volume , @xmath317 is the mass density , and @xmath318 ( @xmath319 ) creates ( destroys ) a phonon of wavevector @xmath305 and polarization @xmath311 . for @xmath320 we use the dipole approximation @xmath321 , which is appropriate since @xmath322 where @xmath323 is the the phonon wavevector and @xmath324 nm is the bohr radius . at low temperatures @xmath325 k we obtain @xmath326\end{aligned}\ ] ] where @xmath327 and @xmath328 are the longitudinal and transverse sound velocities . for our qubit , the relaxation from @xmath228 to @xmath227 can be evaluated using a schrieffer - wolff transformation of the non - diagonal elements of @xmath329 , where @xmath330 . we determine the coupling to lowest order in @xmath331 and @xmath332 while treating @xmath244 , @xmath333 and @xmath30 exactly within the @xmath11 subspace . we obtain @xmath334\end{aligned}\ ] ] at the sweet spot . for comparison , we also evaluated the phonon - mediated transition rate from the first excited level @xmath335 to the ground state @xmath336 in bulk unstrained silicon , which is allowed even in zero magnetic field , since @xmath335 and @xmath336 are not time - reversal symmetric . for @xmath337 along [ 001 ] directions we obtain @xmath338 in agreement with the low - temperature result in ref . . dephasing of a spin - orbit qubit occurs due to random fluctuations @xmath164 in the qubit level energy splitting . in this section we outline expected dephasing rates associated with realistic parameters for the qubit . first , we estimate the dephasing due to a single fluctuating charge trap ( random telegraph signal ) , assuming that a charge trap can only fluctuate when it is in tunneling proximity to carrier reservoir or gate . the dephasing rate is given by @xmath165 , where @xmath166 is the qubit energy shift when the trap is charged , and @xmath167 is the switching time@xcite . together with its image in the gate , a dipole @xmath339 is created , resulting in a dipole potential @xmath340 when the defect is a distance @xmath341 away from the acceptor qubit . the electric field for the dipole potential is @xmath342 . the change @xmath166 of the larmor frequency is readily calculated from the extended qubit hamiltonian ( eq . [ eq : hesup ] ) , where @xmath79 couples off - diagonally through the @xmath343 terms , and @xmath344 on - diagonally via the explicit dependence of @xmath235 and @xmath345 on @xmath21 . equivalently , the @xmath219 qubit model can be used , and it is easy to show that @xmath346 for the estimate of dephasing in the main text we assumed @xmath347 nm , and @xmath348 nm giving @xmath349 v / m and @xmath350 v / m . gate electric field noise is modeled as a white johnson voltage noise @xmath351 process applied across the gates producing a randomly fluctuating field of order @xmath352 where @xmath353 nm is the shortest envisioned distance between gates . the dephasing rate is given by @xmath354 where @xmath355 is the power spectral density of the random process @xmath356 . further approximation to @xmath357 from above gives @xmath358 we find that the first term is much larger for our qubit and dominates the power spectral density of @xmath359 , @xmath360 the quantity @xmath361 is determined from our analytic model . substituting @xmath362 as the white noise power spectrum , @xmath363 k and @xmath364 ohms , we find that the dephasing rate @xmath365 due to intrinsic johnson voltage noise on the gate is @xmath366 times smaller than the dephasing rate from charge noise . all theory preditions in the main text were compared with a numerical solution of the acceptor hamiltonian including the full spatial dependence of @xmath367 . the hamiltonian @xmath368 , where @xmath369 , was computed in the @xmath370 representation of valence band bloch states @xmath371@xcite , @xmath372 , where @xmath373 are envelope functions . the first step is to numerically diagonalize @xmath374 using @xmath375 @xmath376 in the @xmath377 , @xmath378 , @xmath379 , @xmath380 , @xmath381 , @xmath382 , basis , with @xmath383 @xmath384 , @xmath385 , @xmath386 with @xmath387 ( using a dirichlet boundary condition ) , and @xmath388 and @xmath389 from the main text . here , @xmath390 , @xmath391 , @xmath392 , and @xmath194 are the strains . the dielectric constant @xmath393 , lk parameters @xmath394 , @xmath395 and @xmath396 , and @xmath397 mev , bir - pikus deformation potentials @xmath398 ev and @xmath399 ev , and lande g - factors @xmath41 and @xmath400 were obtained from the literature@xcite . the hydrostatic deformation potential @xmath197 is not needed since it acts on the valence bands with an identity operator . the hamiltonian @xmath374 was diagonalized by finite differences in cartesian coordinates for different fields @xmath21 and acceptor depths @xmath24 , to obtain the eigen - energies ( eigen - states ) @xmath401 ( @xmath402 ) of @xmath374 . the inversion asymmetry - induced mixing of states outside the @xmath403 subspace is explicitly included due to the explicit representation of electric potentials of the ion , interface , and applied gate field , in the hamiltonian . the effect of the @xmath18 symmetry terms is absent in @xmath402 because the local field of the ion is only present in the unit cell containing the ion , which lies outside the validity of the lk approach@xcite . the @xmath18 symmetry term was included in the spectrum and eigenstates by diagonalizing @xmath404 in the basis @xmath402 . the @xmath113 values in @xmath244 were taken to reproduce lh - hh couplings known from experiments@xcite for calculations on bulk acceptors . we reduced @xmath113 by a multiplicative prefactor due to the interface and gate field , which was determined from the reduction of the charge in density at the ion in @xmath402 . we diagonalized @xmath404 using the first 8 states from @xmath374 . we note that states 5 and 6 contribute very little ; even though these predominantly split - off hole states have a non - zero density at the ion ( non - zero @xmath18 symmetry correction ) , they are already 20 mev away ( compared to the @xmath405 mev hh - lh level splitting ) . the even higher energy states 7 and higher may be neglected _ a priori _ at low fields , because they have even higher energies , but more importantly , since their principal quantum number is 2 or higher . consequently , the density at the ion , and hence the @xmath18 symmetry coupling , is even smaller . the states @xmath402 experience inversion asymmetry due to explicit presence of the interface and gate field . we have verified that inversion asymmetry - induced mixing of excited states into the @xmath11 subspace does not modify the symmetry of @xmath407 from the form in the main text , for different acceptor depths , by evaluating @xmath408 using eigenstates @xmath402 calculated non - perturbatively , that is , with applied potentials including the ion , interface , and gate electric field . this was checked for all static fields @xmath21 in the main text , and for fields @xmath409 for all three orthogonal directions . values for @xmath410 were obtained from eigen - energies , while values of @xmath411 were obtained from the off - diagonal elements coupling the lh and hh blocks . the lh - hh mixing caused by @xmath113 in @xmath244 changes the larmor frequency @xmath31 of our spin qubit @xmath412 dramatically compared with @xmath413 . numerical calculations of @xmath31 ( fig . 2c , d , blue squares ) are in agreement with the approximate ( fig . 2c , d , black lines ) and analytic solutions ( fig . 2c , d , green lines ) to the @xmath11 model . in particular , our numerical calculations show that mixing of states outside the @xmath11 subspace , and the inclusion of cubic lande g - factor @xmath193 with values obtained in experiments , does not qualitatively change the behaviour of the qubit splitting . therefore , the main electric - field dependence on electric fields is dominated by the lh - hh mixing inside the @xmath11 subspace . the numerical results for edsr ( fig . 2e , f , orange squares ) were obtained by transforming our numerical representations of @xmath30 and @xmath244 into the basis of @xmath414 , using @xmath415 , where @xmath416 is the unitary matrix that diagonalizes @xmath404 in the basis @xmath402 . 81ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ] + 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop ( ) , @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop , , vol . ( , , ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop ( , , ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( )	high fidelity entanglement of an on - chip array of spin qubits poses many challenges . spin - orbit coupling ( soc ) can ease some of these challenges by enabling long - ranged entanglement via electric dipole - dipole interactions , microwave photons , or phonons . however , soc exposes conventional spin qubits to decoherence from electrical noise . here we propose an acceptor - based spin - orbit qubit in silicon offering long - range entanglement at a sweet spot where the qubit is protected from electrical noise . the qubit relies on quadrupolar soc with the interface and gate potentials . as required for surface codes , @xmath0 electrically mediated single - qubit and @xmath1 dipole - dipole mediated two - qubit gates are possible in the predicted spin lifetime . moreover , circuit quantum electrodynamics with single spins is feasible , including dispersive readout , cavity - mediated entanglement , and spin - photon entanglement . an industrially relevant silicon - based platform is employed . in recent years , the coherence and control fidelity of solid - state qubits has dramatically improved@xcite and spin qubits@xcite with highly desirable properties have been demonstrated.@xcite however , many obstacles remain to efficiently entangle a large array of spin qubits on a chip . for example , exchange is inherently vulnerable to decoherence from electrical fluctuations@xcite , coupling spin to charge noise . minimizing decoherence and improving control in the face of noise is the key issue for large - scale quantum computing , because it ultimately determines if the error - correction resources can be managed for a large qubit array.@xcite moreover , exchange - based entanglement is inherently short - ranged , making fabrication challenging for gates in quantum dot arrays@xcite , and placing strict demands on si : p donor placement.@xcite here we propose a single - acceptor spin - orbit qubit where the unique properties of hole spins give a host of desirable attributes . first , spin - orbit coupling ( soc ) enables long - ranged entanglement via microwave photons or electric dipole - dipole interactions@xcite , of interest for hybrid quantum systems@xcite , improving error correction@xcite , and reducing fabrication demands compared with exchange coupled schemes . second , and most remarkably , we find a sweet spot where coherence is insensitive to electrical noise and electric dipole spin resonance@xcite ( edsr ) is maximized . consequently , coherence and gate timings are protected from electrical noise at the hamiltonian level , and one- and two - qubit gate times are optimized . in comparison , electric field noise dephases conventional spin - orbit qubits@xcite and acceptor charge qubits.@xcite the coherence of our spin - orbit qubit benefits from reduced hyperfine coupling of holes@xcite and @xmath2si enrichment@xcite , and has much longer phonon relaxation times than acceptor charge qubits.@xcite finally , the acceptors naturally confine single holes that can be manipulated in silicon nanoelectronic devices@xcite . the exceptional properties of the qubit derive from the quadrupolar soc@xcite contained in the spin-3/2 luttinger hamiltonian@xcite and in the interaction with the inversion asymmetric interface potential , not studied previously for acceptors . this soc is unusually strong for acceptors because it acts directly on the low - energy spin manifold , contrasting its indirect role in hole quantum dots.@xcite the soc must be considered non - perturbatively to obtain the sweet spot , and the interface strongly enhances edsr relative to a bulk acceptor . we find 0.2 ns one - qubit gate times , charge - noise immunity , and long phonon relaxation times at the sweet spot , allowing for @xmath3 operations in the coherence time . two - qubit entanglement based on spin - dependent electric dipole - dipole interactions@xcite is feasible with @xmath4 times of @xmath5 ns , and @xmath1 operations in the coherence time . edsr also enables circuit quantum electrodynamics@xcite ( cqed ) with single - spin dispersive readout , and long distance spin - spin entanglement with @xmath4 times of @xmath6 ns . resonant spin - photon coupling with @xmath7 mhz is also feasible . _ qubit concept . _ the qubit is a hole spin bound to a single si : b dopant@xcite , implanted@xcite or placed by scanning tunneling microscopy@xcite near an interface , in a strained silicon - on - insulator ( soi ) substrate ( fig . [ fdevice]a ) . the key quadrupolar interactions , associated with interface inversion asymmetry and products @xmath8 of spin-@xmath9 matrices where @xmath10 , originate from strong soc in the valence band , and have no analog in the conduction band.@xcite this soc acts on the @xmath11 ground state manifold @xmath12 , _ i.e. _ , the @xmath13 and @xmath14 kramers doublets composed mostly of @xmath15 bloch states.@xcite for si : b they are well isolated by @xmath16 mev from orbital excited states and 46 mev from the valence band edge@xcite ( fig . [ fdevice]b ) . the key quadrupolar interactions include the acceptor hole spin - mixing that is linear in electric fields , @xmath17 , associated with @xmath18 symmetry in the central cell @xcite . here , @xmath19 d is known for si : b@xcite ( 1 d = 0.021 e@xmath20 nm ) . an electric field @xmath21 further breaks the envelope function parity by mixing excited states outside the @xmath12 manifold.@xcite projected into the @xmath12 subspace , this interaction is governed by @xmath22 , where @xmath23 and @xmath24 split and mix the doublets , respectively . we verified that this holds for triangular interface wells , using ( i ) a schrieffer - wolff transformation@xcite with higher excited states in the spherical spin-3/2 basis@xcite , and ( ii ) numerical , non - perturbative luttinger - kohn ( lk ) calculations with explicit ion and interface well potentials@xcite . we find a splitting @xmath25 ( fig . [ fdevice]b ) , where @xmath26 from the interface is larger for shallower acceptors ( in agreement with experiments@xcite ) , and @xmath27 increases with increasing field . moreover , quadrupolar soc combining inversion asymmetry and in - plane electric fields is governed by terms @xmath28 that replace @xmath29 in @xmath30 . _ operating point and sweet spot . _ here we show that the qubit splitting @xmath31 ( between @xmath32 and @xmath33 , fig . 1b ) in an in - plane magnetic field @xmath34 depends on the electric field @xmath21 applied by the gates ( fig . 1a ) , and at the sweet spot , @xmath31 is insensitive to electric - field noise @xmath35 in all directions . including magnetic fields , strain @xmath36 ( supplemental material@xcite ) and the interface well , but not in - plane electric fields , we find an operating point hamiltonian , @xmath37 in the basis @xmath38 , where @xmath39 , @xmath40 is the bohr magneton , @xmath41 is the land g - factor for si : b.@xcite , and @xmath42 is the splitting between the light and heavy holes . the cubic g - factor@xcite @xmath43 is temporarily neglected . is determined by the strain , interface , and gate field @xmath21 . shown : @xmath44-induced mixing of states in the @xmath11 manifold due to the @xmath18 symmetry in the unit cell of the ion . not shown : lh - hh coupling from in - plane drive fields . ] inspecting @xmath45 , @xmath21 mixes @xmath46 and @xmath47 and these states have an avoided crossing when the interface well splitting compensates strain , _ i.e. _ , @xmath48 . in fig . 1a we show that for appropriate strains @xmath49 , the anti - crossing can be obtained at @xmath50 mv / m for @xmath51 nm acceptor depths . the field @xmath52 at such an anti - crossing is large enough that the level - repulsion gap @xmath53 exceeds the zeeman interactions , _ i.e. _ , @xmath54 . this unusual aspect of our hole spin - orbit qubit _ c.f . _ other proposals@xcite follows from the tunability of the spin-3/2 levels with strain and confinement , giving rise to the anti - crossing , and the strength of quadrupolar soc@xcite relative to typical spin qubit larmor frequencies . we treat the quadrupolar soc term @xmath44 by a rotation that maps @xmath44 exactly to the diagonal , to a basis @xmath55 leaving zeeman terms @xmath56 off - diagonal . we obtain @xmath57 , a low - energy kramers pair ( energy @xmath58 ) , and an excited kramers pair @xmath59 ( energy @xmath60 ) . here , @xmath61 and @xmath62 . in the basis @xmath55 eq . 1 becomes @xmath63 here , the zeeman terms @xmath64 depend explicitly on @xmath21 due to the gate - induced mixing of @xmath65 and @xmath66 . we find @xmath67 , @xmath68 and @xmath69 . we perform a final rotation that exactly maps @xmath70 and @xmath71 to the diagonal , leaving @xmath72 off - diagonal , defining a basis @xmath73 ( see supplemental material@xcite ) . to zeroth order in @xmath74 , the splitting of the kramers pair qubit states @xmath32 and @xmath33 is @xmath75 . when mixed by the gate electric field , the spin 1/2 and spin 3/2 states with different zeeman terms define a qubit @xmath76 where @xmath31 is maximized ( independent of electric fluctuations ) @xmath77 to first order when @xmath78 ( see supplemental material@xcite ) . as we will subsequently show , the qubit is also insensitive to in - plane electric noise @xmath79 , while a similar analysis yields another sweet spot at @xmath80 . energy levels @xmath81 for the qubit are shown alongside excited levels @xmath82 for @xmath83 nm ( @xmath84 nm ) in fig . [ fop]a ( fig . [ fop]b ) . here , blue ( red ) hue denotes the amplitude of @xmath85 ( @xmath86 ) . the qubit frequency is shown in fig . [ fop]c and fig . [ fop]d for approximate ( black ) and exact ( green ) solutions to @xmath45 , alongside the numerics ( squares ) . the maxima in @xmath31 in fig . [ fop]c ( fig . [ fop]d ) defines the sweet spot at @xmath87 mv / m ( @xmath88 mv / m ) , for @xmath89 , as expected . we note that the approximate solution ( fig . [ fop]c , d , black lines ) captures the essential behaviour of the analytic model ( fig . [ fop]c , d , green lines ) . corrections to zeeman interactions from interface inversion asymmetry and cubic land g - factor , although included in the numerics ( squares ) , have been neglected in the analytic model ( green ) . note that the interface prevents ionization ; although @xmath90 mv / m is much smaller than silicon s breakdown field , it well exceeds the ionization field of si : b.@xcite _ in - plane electric fields : edsr and noise immunity._we express interactions with in - plane electric fields in the basis @xmath91 , yielding @xmath92here , @xmath32 and @xmath33 are our kramers pair qubit states , @xmath93 and @xmath94 are zeeman terms , and @xmath95 are interaction terms with in - plane electric fields , where @xmath96 , @xmath97 , @xmath98 , @xmath99 , and @xmath100 . the qubit hamiltonian @xmath101 , where @xmath31 is the qubit frequency ( fig . 2c , d ) and @xmath102 is the edsr matrix element ( fig . 2e , f ) , is obtained by projecting the off - diagonal elements of @xmath103 to first order in @xmath104 using a schrieffer - wolff transformation.@xcite notably , qubit coherence is protected from in - plane electric noise since @xmath31 contains no terms to first order in @xmath104 . edsr drive comes from the transverse coupling @xmath105 in @xmath106 . we obtain @xmath107 , where @xmath108 . interestingly , the small splitting @xmath109 essential for spin mixing at the sweet spot also causes strong edsr , since @xmath110 . note that the edsr term is dominated by the inversion asymmetry quadrupolar soc parameter @xmath111 d ( fig . [ fop]f ) , since it is @xmath112 larger than the bare @xmath18 soc parameter @xmath113 . importantly , @xmath102 can be maximized at the sweet spot by choosing the angle @xmath114 relative to @xmath115 ( see fig . 2e , f ) . this yields fast gate times , but it also makes @xmath102 , and therefore all timings based on edsr , insensitive to fluctuations in electric field , protecting gate fidelity from noise at the hamiltonian level . since @xmath116 and @xmath117 at the sweet spot , @xmath102 is maximized with respect to @xmath118 at @xmath119 . as shown for @xmath83 nm ( @xmath84 nm ) in fig . [ fop]e ( fig . [ fop]f ) @xmath102 is maximized with respect to @xmath21 for the same choice @xmath118 . this result can be easily obtained analytically , and holds for the analytic ( green ) and numerical ( blue squares ) solutions . _ qubit operation_. the one - qubit and two - qubit gates employ edsr - mediated interactions at the sweet spot , where coherence is protected from noise , and their times @xmath120 are minimized and also insensitive to electrical noise . edsr driven @xmath121 rotations require @xmath122 ns ( @xmath123 ns ) for the @xmath83 nm ( 6.9 nm ) deep acceptor , assuming a modest in - plane microwave field @xmath124 v / cm . a @xmath125 ( 0 ) phase shift realizes a @xmath126 ( @xmath127 ) gate , and @xmath128 gates can be decomposed into a sequence of @xmath127 and @xmath126 gates . readout can be accomplished by energy - dependent@xcite or spin - dependent@xcite tunneling , or dispersive readout in cqed.@xcite initialization can be achieved by projective readout followed by spin rotation . two - qubit entanglement can be achieved via long - ranged coulomb interactions , owing to spin - dependent electric dipole - dipole interactions.@xcite their strength is given by @xmath129 , where @xmath130 is the inter - qubit displacement and @xmath131 is a spin - dependent charge dipole of qubit @xmath132 , which has the same magnitude as the edsr matrix element . for a @xmath133 nm distance with negligible tunnel coupling , we obtain a @xmath4 time of @xmath134 ns with @xmath135 . the @xmath136 times enhancement of edsr from the interface reduces @xmath137 by @xmath1 relative to acceptors in bulk silicon , and @xmath0 relative to bare magnetic dipole - dipole coupling . entanglement by heisenberg exchange is also possible and exchange is hydrogenic when @xmath138 exceeds @xmath139.@xcite we note that the advantage that holes do not have valley degrees of freedom@xcite which may complicate heisenberg exchange for electrons in si.@xcite _ circuit qed_. coplanar superconducting microwave cavities could be used to implement cqed including two - qubit gates , dispersive single - spin readout , and strong jaynes - cummings coupling on resonance with the cavity.@xcite we assume a coplanar waveguide resonator operating at @xmath140 t ( @xmath141 ghz ) and a vacuum electric field @xmath142 v / m . this can be obtained using a tapered resonator gap , or a superconducting nanowire resonator.@xcite at the sweet spot for @xmath83 nm ( @xmath84 nm ) , the vacuum rabi coupling is @xmath143 nev ( @xmath144 nev ) . for cavity mediated non - demolition readout and qubit coupling , we detune the qubit from the cavity by @xmath145.@xcite here , the spin state shifts the cavity resonance by @xmath146 mhz ( 1.25 mhz ) for @xmath83 nm ( @xmath84 nm ) . the two - qubit @xmath4 time is @xmath147 ns for @xmath148 nm , determined by the effective spin - spin interaction@xcite @xmath149 mhz . operating at zero detuning , spin / photon rabi oscillations require @xmath150 @xmath151s ( @xmath6 ns ) . assuming @xmath152 at @xmath153 t in state - of - the - art superconducting cavities@xcite @xmath154 ( 33 ) rabi cycles can be obtained for @xmath83 nm ( @xmath84 nm ) , where @xmath155 is the cavity loss rate . _ relaxation and dephasing_. we consider spin - lattice ( phonon ) relaxation and dephasing from a host of electrical noise sources , and compare them to gate times . since silicon is not piezoelectric , spin - lattice relaxation occurs only via the deformation potential@xcite . for temperatures @xmath156 , the spin relaxation time derived in the supplemental material@xcite follows @xmath157 , where @xmath158 at the sweet spot , and @xmath159 ( ev)@xmath160(s / m)@xmath161 . we obtain @xmath162 @xmath151s ( 5 @xmath151s ) for @xmath83 nm ( 6.9 nm ) at @xmath153 t that are @xmath163 times longer _ c.f . _ bulk unstrained silicon at @xmath140 t.@xcite random fluctuations in qubit splitting @xmath164 dephase the qubit . the dephasing rate from random telegraph signal ( rts ) in charge trap occupation is @xmath165 , where @xmath166 is qubit frequency shift , and @xmath167 is the average switching time.@xcite we take @xmath168 as the worst case , since slower fluctuations can be suppressed by dynamical decoupling . assuming a trap @xmath169 nm away , we find @xmath170 v / m and a large window of @xmath171 v / m ( @xmath172 v / m ) of gate space where @xmath173 at the sweet spot , for @xmath83 nm ( @xmath84 nm ) . in comparison , the same analysis gives @xmath174 ns for acceptor - based charge qubits with similar gate times . it is remarkable that in comparison , electrical noise has virtually no effect on coherence in our spin - orbit qubit , illustrating the advantages of inversion asymmetry and our spin - orbit qubit s sweet spot . we also find that dephasing from johnson - limited gate voltage noise , and from two - level ( tunneling ) systems ( tls ) , are @xmath175 and @xmath176 times weaker , respectively , compared with rts.@xcite there are only a few spin resonance experiments on acceptors@xcite , none of which feature strain and an interface.@xcite we expect hyperfine - induced decoherence in @xmath177si to be weak since it has only 4.7 % of spin - bearing isotopes and hyperfine interactions are weaker for holes than electrons.@xcite meanwhile , @xmath2si enrichment could be used to virtually eliminate the nuclear bath.@xcite the insensitivity to johnson noise and tunneling tls means spin - lattice @xmath178 limits coherence for few ( or slow enough ) traps at si / sio@xmath179 interfaces . for @xmath140 t , @xmath180 single qubit gates , @xmath181 dipole - dipole two - qubit gates , and @xmath182 cavity - mediated two - qubit gates can be achieved in a @xmath178 limited coherence time . therefore while @xmath178 is short compared to donors , many gate operations can performed . since @xmath183 , choosing @xmath184 t increases all ratios favourably to @xmath185 , @xmath186 , and @xmath187 . since @xmath178 is much longer at the @xmath80 sweet spot , adiabatically sweeping to @xmath80 opens a pathway for a long - lived quantum memory . _ conclusions_. the proposed single - acceptor spin - orbit qubit exploits the tunability of the @xmath188 manifold of acceptors and the associated quadrupolar soc arising from the ion and interface potential , providing for ( i ) fast one - qubit and long - ranged two - qubit gates ( ii ) at a sweet spot where the qubit phase and all gate timings are insensitive to electrical fluctuations , ( iii ) avoiding entirely the need for exchange interactions , ( iv ) in an industrially relevant silicon platform . @xmath0 single - qubit and @xmath1 two - qubit gates could be possible in the qubit coherence time . using cqed , dispersive single - spin readout , cavity - mediated spin - spin entanglement , and jaynes - cummings spin - photon entanglement are possible . we thank r. winkler , u. zuelicke , m. tong , and t. kobayashi for helpful discussions . js , jam and sr acknowledge funding from the arc centre of excellence for quantum computation and communication technology ( ce110001027 ) , and in part by the us army research office ( w911nf-08 - 1 - 0527 ) . dc acknowledges funding through the arc discovery project scheme .

Subsets and Splits

No saved queries yet

Save your SQL queries to embed, download, and access them later. Queries will appear here once saved.