id
stringlengths 24
24
| idx
int64 0
402
| paragraph
stringlengths 106
17.2k
|
|---|---|---|
626bdff9ebac3ac754e735e4 | 4 | Here, enzymes streamline synthesis by uniting small to medium-sized fragments of a larger target compound in an efficient manner -generally with excellent selectivity and without the need for any prefunctionalization or directing groups. As a testament to this prowess, the Narayan group recently employed P450 monooxygenases for crosscoupling reactions with coumarin-type phenols. In contrast to known organic or organometallic procedures, these P450 enzymes smoothly united different phenols, exerted surprising levels of atroposelectivity and proved highly amendable by protein engineering to improve their selectivity and substrate scope. In addition to such heterodimerization, the Sherman group reported the use of P450s for the asymmetric dimerization of diketopiperazines in the synthesis of naseseazine alkaloids. 27 Also employing radical transformations, the Sattley group used laccases to obtain pinoresinol analogues from coniferyl-type alcohols, which may serve as precursors to analogues of the topoisomerase inhibitor etoposide. In this example, the use of a dirigent protein provided nearly perfect stereocontrol in the dimerization of two planar units into a product featuring four vicinal stereocenters. Further complementing the work on Pictet-Spenglerases discussed above, the Hailes group recently explored their use in the convergent synthesis of isoquinoline scaffolds which could be elaborated by regioselective enzymatic methylation and chemical Pictet-Spengler reactions to obtain protoberberine alkaloids. Finally, the Merck team recently disclosed their catalytic route to the highly potent STING agonist MK-1454, a cyclic dinucleotide derived from guanosine and thymidine building blocks. The key convergent step unites two unusual, fluorinated thiophosphate nucleotides in a cascade comprising kinase-mediated phosphorylation and diastereoselective cyclization by a cyclic GMP-AMP synthase. |
626bdff9ebac3ac754e735e4 | 5 | Lastly, the combination of multiple enzymatic steps in onepot cascades 31 , sometimes combined with chemical reactions steps, has also found several recent applications in natural product synthesis, particularly for the rapid construction of molecular complexity. The main appeal of this approach is the smooth combination of multiple reaction steps in one pot, including those which may be incompatible by purely chemical means. Typically, cascade strategies grant significant savings in time and resources by circumventing the need for intermediate work-up steps, in addition to the possibility to drive (potentially unfavorable) equilibrium reactions to completion. For instance, following the impressive example of the development of a three-step biocatalytic cascade for the synthesis of the nucleoside analogue Islatravir, Merck recently developed a chemoenzymatic cascade towards the nucleoside analogue Molnupiravir, an investigational antiviral agent for the treatment of COVID-19. Following 5-acylation of ribose using an immobilized lipase, a five-enzyme cascade for 1phosphorylation and nucleobase installation with an engineered MTR kinase and an engineered uridine phosphorylase in combination with auxiliary enzymes for phosphate donor recycling yielded 5′-acyl uridine selectively. Chemical oxime formation completed this chemoenzymatic synthesis of Molnupiravir, which is not only 70% shorter compared to the initial chemical route, but also 7-fold higher yielding. In an alternative approach towards the same nucleoside analogue, the groups of Lovelock, Turner and Green employed an engineered cytidine deaminase for the installation of the N-hydroxyl group in cytidine by relying on the promiscuous activity of the enzyme with hydroxylamine. In addition to these examples, our group recently demonstrated the diversification of nucleoside analogues bearing alkylated sugar moieties through the use of hyperthermostable nucleoside phosphorylases. The inherent thermodynamic constrains in such phosphorolysis systems could either be manipulated by reaction engineering or selective product removal through strategies such as in situ formation of borate esters. In a remarkable example from the Allemann group, the chemo-enzymatic synthesis of natural as well as non-natural linear terpene precursors and terpene analogs has been demonstrated. Starting from prenol or prenol-based precursors, the corresponding diphosphates were accessed through sequential enzymatic phosphorylation in a one-pot cascade using two kinases together with an ATP recycling system relying on phosphoenolpyruvate. Extending this cascade with different prenyl transferases allowed easy access to methylated as well as hydroxylated farnesyldiphosphate derivatives. Upon further extension with terpene synthases, several sesquiterpenes and non-natural analogues thereof were obtained with impressive selectivity. Similarly, Tang and coworkers developed a cell-free biosynthetic cascade comprising a total of nine enzymes for the efficient production of the plant monoterpene nepetalactol from geraniol at a product titer of up to 1 g/L. 40 Key to the successful combination of the different enzymes in one pot was the selection of enzymes with orthogonal cofactor specificities and suitable, highly-selective cofactor regeneration systems. Efficient cofactor supply was also crucial in the cascade synthesis of methylated natural and non-natural tetrahydroisoquinoline alkaloids as reported by the Hailes group. Starting from L-tyrosine, (S)-norcoclaurine was generated in situ using four different enzymes, and further methylated in one pot by SAMdependent O-and N-methyltransferases exhibiting exceptional regioselectivities. Here, the necessary (but instable) SAM cofactor was formed in situ from L-methionine and ATP using a methionine adenosyltransferase, while the resulting S-adenosylhomocysteine by-product arising from methyl transfer had to be removed through enzymecatalyzed degradation to circumvent inhibition of the methyltransferases. |
626bdff9ebac3ac754e735e4 | 6 | Finally, the combination of different biosynthetic enzymes in in vitro cascades also proved highly fruitful in the gramscale synthesis of different non-standard sugar nucleotides as demonstrated by Wen and coworkers. Starting from the feedstock chemical N-acetyl glucosamine, the application of efficient cofactor regeneration systems in the reduction, oxidation, amination and/or acetylation steps was crucial to drive the concurrent cascades toward target sugar nucleotide formation. |
626bdff9ebac3ac754e735e4 | 7 | As apparent from the examples outlined in this review, enzymes are exquisite catalysts for the selective introduction of chirality and complexity in natural product synthesis. Inspired by nature, naturally occurring biosynthetic enzymes are currently not only used at single strategic steps of total syntheses but are also combined in complex cascades to mimic biosynthetic pathways in vitro. Moreover, chemoenzymatic cascades have already made an entrance into the pharmaceutical industry for the synthesis of active pharmaceutical ingredients from commodity chemicals, as their application can shorten development routes significantly and/or avoid unwanted side reactions and toxic wastes. In this respect, protein engineering of naturally occurring enzymes is a powerful tool to tailor their reactivities and selectivities to match the desired application. We are convinced that in the future we will see more and more applications of such specifically engineered enzymes in natural product synthesis, as this will allow access to a wider range of natural product derivatives as well as the design of "new-to-nature" biosynthetic pathways. |
66dca3da12ff75c3a1b864bc | 0 | Electrochemical water splitting is a potentially carbon-neutral method of splitting water into gaseous H2 fuel and O2 via electrical energy from renewable sources. H2O is oxidized to O2 in the Oxygen Evolution Reaction (OER) at the anode and reduced to H2 in the Hydrogen Evolution Reaction (HER) at the cathode. Electrochemical H2 production is generally performed entirely in alkaline media but HER is much more favorable in acidic media. Alkaline conditions are preferred at the anode as OER is generally more favorable in basic media. To achieve the greatest activity, a Proton-Exchange Membrane (PEM) must separate the two electrolytes and allows protons to flow from the anode to the cathode. The economics associated with PEMs, precious metal catalysts, and catalyst stability have led to widespread use of alkaline single electrolyte devices without expensive membranes or noble metal catalysts. The search space for such catalysts is much greater due to the greater stability of alkaline conditions. Various types of catalysts have been investigated from transition metal alloys, compounds (carbides, phosphides, chalcogenides) and their heterostructures. Particular attention has been paid to inexpensive Ni-based catalysts, especially binary compounds of Ni. In addition, the understanding of the origin of such success has been limited. Theoretical analyses of HER catalysis focuses on the free energy of hydrogen adsorption, ΔGH* determining reactivity. By the Sabatier principle, the reactive intermediate H* has an optimum binding energy that is neither too strong nor too weak. This is the basis of the so-called "Volcano relation" of The correlation between activity and ΔGH* has been used extensively in acid, as it captures the rate-determining step (RDS) of H adsorption/reduction. In alkaline media, catalysts suffer a ~2-3 order of magnitude decrease in activity which is not explained by shifts in ΔGH*. Proposed origins for lower alkaline activity are of two types: increased difficulty of water dissociation and increased energy associated with interfacial reorganization. The former originates from the greater O-H bond energy of H2O compared to that of H3O + . The sluggish H2O dissociation kinetics are thought to limit the Volmer step and make it ratelimiting. In 2016, Rossmeisl et al proposed that decreased HER activity in alkaline media was not a surface interaction, but due to reorganization of the double layer upon proton transfer to the surface from bulk. They further explained that this barrier originates from the entropy loss of a proton entering the well-structured Outer Helmholtz Layer (OHL), which has a configuration dependent on the surface itself. This idea was developed further by Koper et al by relating entropy change to the strength of the interfacial electric field and ultimately the Potential of Zero Charge (PZC). A recent models Koper have integrated the OH binding energy into previous ΔGH* correlations to explain alkaline HER on metal surfaces. The structure of the electrochemical interface and its influence on reaction barriers has been left relatively unexplored, especially in the case of composite materials. Previous studies that explicitly include the solvent do so using static calculations with the system following an energy minimum along the reaction path. Such investigations entirely neglect the role of interface dynamics on the reaction and instead use the solvent as a Helmholtz layer and source of hydrogen bonds, while assuming perfect solvation. In recent years, significant effort has been dedicated to understanding the dynamics and influence of the metal-water interface. The group of Groß et al. has performed much of the foundational work on the structure of the metal-water interface at finite temperature. They first established that the extent of order in the electrical double layer varies with metal on close-packed (111) surfaces using Ab Initio Molecular Dynamics (AIMD) simulations. Later studies found differences in hydrogen bonding above the noble metal (111) surfaces. Groß and Sakong recently outlined the structure of the electrochemical interface on Pt (111), addressing the effect of pH, charging, and HER adsorbates. Despite this diagnosis of the interface, studies have yet to attempt directly ascertaining the influence of interface dynamics on the alkaline Volmer reaction. With recent advances in computing, advanced sampling techniques are just beginning to allow for a quantification of the entropic contributions to barrier energies that cannot be measured in the single-point calculations. In this work, we revisit the Volmer step under alkaline conditions using both static and AIMD simulations. We begin with a look at what a Ni surface might look like under active conditions. Revisiting a previous publication, we show that Ni/Ni3S2 composite catalysts may result in a S-modified Ni interphase. We further use GCDFT calculations to show that unmodified Ni will be covered by a small amount of *OH, in contrast to Pt. However, the energetics on these surfaces with light coverage do not explain the disparities in reactivity. Further investigation of the Volmer barrier at potential shows that these modifications do not significantly change charging behavior of the catalyst. Having dismissed the energetic explanation, we use Car-Parinello molecular dynamics combined with metadynamics sampling to estimate the Volmer step barrier on pristine and modified metal surfaces. These barriers are contrasted with those obtained using static (NEB) calculations to quantify the entropic contributions to the barrier. Finally, these entropic "penalties" are explained as function of the interface structure through analysis of unbiased AIMD simulations. We conclude that the consistency of interfacial solvation governs the Volmer barrier under dynamics. |
66dca3da12ff75c3a1b864bc | 1 | Static calculations were performed using the Vienna Ab Initio Simulation Package (VASP). Valence electrons were described using Projector-Augmented Wave (PAW) pseudopotentials. The exchange-correlation terms of the Hamiltonian were computed using the Generalized Gradient Approximation (GGA) functional of Perdew, Burke, and Ernzerhof. The solvation of the electrode surface was considered using the VASPsol package. The first Brillouin zone was sampled using a 4x4x1 Gamma-centered k-point mesh. The basis-set cutoff was set to 400 eV to ensure proper treatment of O-metal bonds. |
66dca3da12ff75c3a1b864bc | 2 | Transition state energies were found using the nudged elastic band (NEB) method. Grand free energies as a function of potential were found using Grand Canonical Density Functional Theory. The CHE framework was used to account for the chemical potential of a proton/electron pair. All calculations were referenced to alkaline conditions (pH = 14). Transition state grand free energies were found in the same way as those of energy minima with individual points obtained by performing NEB calculations at certain net charges. All structures were optimized until the maximum force on an atom was less than 0.05 eV/Å. Molecular dynamics used the same cutoff and a timestep of 1 fs. The Brillouin zone was sampled only at the gamma-point to decrease computation time. Constant temperature was maintained using a Nose-Hoover thermostat. Temperature ramping and lowering were treated using an Andersen thermostat with no heat-bath coupling and velocities reassigned every 10 fs. All AIMD simulations were run for at least 10 ps. Van Der Waals interactions were considered during AIMD using the DFT-D2 method of Grimme. Trajectories were analyzed using the MDAnalysis python library. The last 5 ps of each trajectory were used to generate plots. The radial distribution functions (RDF) take the form of: |
66dca3da12ff75c3a1b864bc | 3 | Trouillier-Martins type in the form of Kleinman and Bylander 49 were used to describe the valence-core electron interaction in CPMD. A large cutoff of 80 Ry was used. Such a high cutoff is recommended for metal atoms using Trouillier-Martins pseudopotentials. Our testing found that the metal-water interface was unstable at lower cutoffs. Spin polarization was considered using the Local Spin Density (LSD) approximation. Our tests found that spin polarization was also essential to describe the interface. Molecular dynamics of the Car-Parrinello type were performed using a timestep of 0.12 fs and a fictitious electron mass of 600 amu. Van Der Waals forces were again considered using the DFT-D2 corrections of Grimme. . A chain of Nose-Hoover thermostats was employed to manage electronic and ionic dynamics. The traditional metadynamics algorithm was employed with hill dimensions of 0.1 Å and 0.023 eV. Sampling was performed until at least 5000 hills and 10 crossings were observed. Hills were deposited at a rate of 10 -2 fs -1 . |
66dca3da12ff75c3a1b864bc | 4 | Ni/Ni3S2 catalyst for alkaline HER. TEM images showed a highly disorganized structure where the lattice constant could not be identified. Our calculations demonstrated poor results for the pure close-packed phases of both Ni and Ni3S2 with superior reactivity at the interface. (These results were later supplemented with GCDFT calculations on both pure surfaces, shown in Figure .) We initially constructed an interface of 2 pristine phases to model the interface, but this may not reflect the complexity of the real surface. For this reason, we performed further simulated annealing on our original interface structure from Li. To mimic the conditions of synthesis, we first increased the temperature of the systems by 1K/fs for 1 ps using a micro-canonical ensemble. The structure was then kept at 1000K for 10 ps under an NVT ensemble. Finally, a micro-canonical ensemble brought the temperature down by 1K/fs for another 1 ps. |
66dca3da12ff75c3a1b864bc | 5 | We observed a significant reorganization in the structure during the constant temperature portion. Several S atoms began to move from the Ni3S2 section to pure Ni. These atoms were mobile on the surface of Ni but did not move back to Ni3S2 phase. In addition, some Inspection of the surface, reveals this structure to be poorly suited for alkaline HER just as the pristine Ni3S2 surface is. As in our previous publication, the formation of an unfavorable S-H bond is still required to break an O-H bond. In addition, the H adsorption energies on this amorphous Ni3S2-like phase show it to have a poor performance. Some representative hydrogen adsorption energies are listed in Table . |
66dca3da12ff75c3a1b864bc | 6 | Step sites for HER have already been thoroughly treated in literature and will not be discussed here. In addition, these sites are not necessarily inherent Finally, we present GCDFT energies of reactions relating to the stability of *S/*OH on Ni (111) in Figure . The adsorption of S was computed using H2S (g) and H2 (g) as references and without using an electron transfer term. S adsorption is very favorable and not a function of potential biasing. In contrast, OH adsorption is a function of potential, with significant grand canonical effects originating from a shift in PZC upon adsorption. |
66dca3da12ff75c3a1b864bc | 7 | This PZC shift is not seen in the *S case. The hydrogenation of adsorbed *S to an adsorbed thiol is very unfavorable. Even at extreme potentials of -1 V vs SHE, the reaction is still endergonic. The hydrogenation of adsorbed *OH to H2O (aq) is less unfavorable,as in the case of adsorption, grand canonical effects significantly reduce the energy of *OH hydrogenation at negative potentials. |
66dca3da12ff75c3a1b864bc | 8 | The next section discusses the coverage studies to quantify the amount of adsorbed *OH on Ni at varying potential. We discuss the surfaces for Pt and Pd, notably more active catalysts, for comparison. Coverage diagrams were generated for each surface by varying the H/OH coverage as well as applied potential. These diagrams are shown in Figure . |
66dca3da12ff75c3a1b864bc | 9 | Ni/Pd surfaces. The chain mechanism is significantly more favorable though the difference declines with decreasing potential. The 3H2O barrier could not be obtained on Pt due to its very large *OH binding energy. OH -(in solution) is more likely to be produced than *OH on a Pt surface, but the treatment of solvated OH -would require many more explicit water molecules. Explicit treatment of the electrochemical interface is relegated to our metadynamics studies of the Volmer reaction in the next section. |
66dca3da12ff75c3a1b864bc | 10 | Figure plots the barrier for the water chain mechanism on surfaces with 1 S atom and 1 OH group adsorbed. Both act as spectators and do not participate in the reaction, though *OH does alter the PZC of the surface. There is no significant effect of adding both adsorbates. This further emphasizes that the slope is inherent to reaction geometry of the reaction. |
66dca3da12ff75c3a1b864bc | 11 | Now we move to calculating the Volmer barrier under dynamic conditions. The collective variable (CV) selected was a standard distance difference commonly used for bond-breaking reactions. This CV configuration is illustrated in Figure . To account for the homogeneity of the metal surface, only the z-component of Ni-H distance was used, and that distance was assigned to the entire surface. Attention was paid to picking H2O molecules that form chains like those modeled in the previous section, although it was observed during sampling that such features are formed upon reaction regardless. |
66dca3da12ff75c3a1b864bc | 12 | The black line drawn from *H to *OH indicates the path that the *H atom effectively followed in a Grotthus-like mechanism. Sampling of the region from the initial state (H2O) to the transition state and early final state was found to be sufficient to estimate the forward Volmer barrier. We calculate this value by first calculating the barrier energy for each timestep from the difference between the initial state basin minimum and the transition state region maximum. The maxima of this function correspond to the points just before the system crosses from the initial state to the final state, where the bias potential is sufficient for the system to overcome the barrier. Figure shows the evolution of the CVs during metadynamics sampling and Figure shows the corresponding barrier energy evolution. The black lines on the CV plot and the points on the energy plot correspond to these crossing points. Figure The barrier energies in Figure were measured to be 1.36 eV at 300K and 1.19 eV at 100K. The 0.17 eV difference here corresponds to an increase in entropic contributions with temperature. It appears that a consistent hydrogen bonding environment comparatively stabilizes the transition state. The system is in the initial state for a much longer period than the transition state, so such differences in stabilization are inherently broken, while the liquid-like interface will have some number of crossings without hydrogen bonding. |
66dca3da12ff75c3a1b864bc | 13 | We also applied this methodology to the Pt surface for comparison and obtained the free energy barriers arranged in As with bare metals, the hydrogen bond lifetimes over these surfaces were computed from unbiased AIMD calculations. These results are shown in Figure and Table . This difference in hydrogen bonding between the OH-modified and S-modified surfaces provides an interesting explanation for superior activity of the Ni/Ni3S2 system. |
66dca3da12ff75c3a1b864bc | 14 | As shown in the first section, S can diffuse from the Ni3S2 phase to pure Ni, creating a surface modified Ni-S phase. In contrast, pure Ni catalysts will be covered by some amount of OH in alkaline conditions. S and OH occupy the same sites on the metal surface and so these surface modified states are mutually exclusive. The lesser hydrogen bonding consistency on the OH surface is a likely origin for the inferior activity of pure Ni. The adsorbed *OH disrupts the hydrogen bond network in the interface in a way that adsorbed *S does not. *OH partakes in hydrogen-bonding and *S simply blocks the surface. |
66dca3da12ff75c3a1b864bc | 15 | Our studies with other d 8 metals, Pt and Pd, found that Ni is unique in its OH binding energy of Ni which is much more favorable than that of Pd and especially Pt. From coverage calculations, we predict that significant OH coverage exists on Ni at alkaline HER conditions. These studies also revealed a significant difference in H2O dissociation. Ni was calculated to have a comparable or lower Volmer barrier despite its poor activity. We Ni was shown to have a more consistent solvation environment and a smaller free energy barrier when compared to OH-modified Ni. Our calculations offer a novel explanation of why such a metal/metal-compound heterostructure possesses a superior activity compared to that of its pure metal counterpart. |
65bd797a66c1381729d75b36 | 0 | Main Text: Organic compounds containing P-stereogenic centers constitute important classes of pharmaceuticals, agrochemicals, and ligands for transition metal complexes (Fig. ) . The absolute stereochemistry at phosphorus often underpins the function of these compounds, ranging from their biological activities to their roles in asymmetric synthesis . Given the absence of a P-stereogenic chiral pool available from Nature (2), chemists' access to P-stereogenic compounds in enantioenriched form necessarily relies on the development of asymmetric synthetic methods . Well-established asymmetric syntheses of P-stereogenic compounds rely on covalently attached C-stereogenic chiral auxiliaries to achieve diastereocontrol over transformations at the phosphorus center (Fig. , left) . Asymmetric catalytic approaches have been identified that rely on coupling at the phosphorus center, substitution at phosphorus, or desymmetrization of substituents (1, 7) (Fig. , right). In particular, recent organocatalytic approaches have established the viability of substitution at phosphorus as a successful principle for catalytic We noted, however, that no approach to date has drawn from the diverse manifold of powerful transformations involving phosphonium dealkylation events such as the Michaelis-Arbuzov, Appel, Staudinger, and Pudovik reactions , reactions for which there is over 125 years of demonstrated synthetic utility. Inspired by the capacity of chiral dual-hydrogen-bonddonors (HBDs) to bind anions to generate chiral ion-pair intermediates , we hypothesized that a suitable HBD catalyst might prove effective in binding a phosphonium ion pair and reorganizing it into a configuration that is favorable for enantioselective dealkylation (Fig. ). Because catalytic enantiocontrol over phosphonium ions has, to our knowledge, yet to be discovered in synthetic or biological systems, we posited that developing this approach might ultimately unlock catalytic asymmetric variants of the manifold of important phosphoniummediated reactions (Fig. ). |
65bd797a66c1381729d75b36 | 1 | To explore this hypothesis, we selected the Michaelis-Arbuzov reaction as a prototypical transformation proceeding through a phosphonium intermediate. Specifically, we directed our attention towards the protio-Michaelis-Arbuzov variant that yields enantioenriched Hphosphinates, P-stereogenic building blocks with broad synthetic utility but that are currently only accessible by stoichiometric approaches (Fig. , right) . |
65bd797a66c1381729d75b36 | 2 | After surveying various alkyl phenylphosphonites, we selected as a model reaction the dealkylation of dibenzyl phenylphosphonite 2a with HCl (Fig. ). Through a systematic evaluation of HBDs featuring a tert-leucine-bis(trifluoromethyl)aniline scaffold (30), we found that a thiourea (1c) promoted the reaction with higher enantioselectivities than its squaramide (1a) or urea (1b) counterparts (Fig. ). Furthermore, the size of the (poly)aromatic substituent as well as α-quarternary substitution at the pyrrolidine correlated positively with enantioselectivity. Ultimately, we found that thiourea 1g bearing a 2-phenanthryl substituent promoted dealkylation in 90% ee and quantitative yield. Although catalyst optimization studies were performed using two equivalents of HCl in methyl tert-butyl ether (MTBE), measurable enhancements in enantioselectivity were achieved by changing the reaction solvent to toluene (Fig. , entry 1), employing a single equivalent of HCl (entry 2), and diluting the reaction mixture to 50 mM (entry 3) to furnish 3a in 95% ee. |
65bd797a66c1381729d75b36 | 3 | A wide array of dibenzyl phosphonites are amenable as substrates in this enantioselective dealkylation, furnishing air-and moisture-stable chiral H-phosphinate products that can be purified by silica gel column chromatography (Fig. ). Various para-substituted phenylphosphonites underwent dealkylation with good yield and enantioselectivity (3a-3g), with lower enantioselectivity observed for highly electron-withdrawing substituents (3h-3j). Metasubstituted phenylphosphonites were dealkylated with comparable enantioselectivities to their para-substituted regioisomers (3k-3m). In contrast, steric volume at the ortho position of the phenyl group was deleterious for enantioselectivity-o-fluoro substitution decreased enantioselectivity from 90% (3f) to 73% ee (3n), whereas o-phenyl substitution ablated enantioselectivity (3o). Substrates bearing ortho-fused polyaromatic substituents, however, could be dealkylated with high enantioselectivity (3p and 3q). The method is also compatible with a variety of hetero-and poly-aromatic substituents (3r-3t and 3v), including acid-labile functional groups (N-Boc-protected indole, 3u). Phosphonites with non-aryl substituents such as isopropenyl (3w) and adamantyloxy (3x) underwent dealkylation with moderate levels of enantioselectivity while alkyl substituents such as cyclopropyl and methyl afforded low levels of enantioselectivity at 54% and 30% ee respectively (see Supplementary Materials, Section 3.1). The enantioselective dealkylation of 2a was performed successfully on a gram scale employing only 3 mol % 1g with high enantioselectivity (93% ee), yield (98%), and efficient catalyst recovery (95%). The substituents on the H-phosphinate products obtained through this enantioselective dealkylation are known to exhibit complementary reactivity as the proton and alkoxide groups on phosphorus are prone to substitution by electrophiles and nucleophiles, respectively . To highlight the synthetic utility of enantioenriched H-phosphinates, we explored the reactivity of (R)-benzyl phenylphosphinate 3a as an orthogonally bifunctionalizable P-stereogenic building block. We found it amenable to an array of stereospecific synthetic elaborations of the P-H moiety, followed by secondary derivatization of the P-OBn moiety (Fig. ). Building on established phospha-Mannich reactivity of phosphinates , we found that deprotonation with lithium phenoxide enabled 3a to participate in the Pudovik addition to Eschenmoser's salt , affording 20 α-amino phosphinate 5a. The benzyloxy group of 5a could subsequently be substituted in the presence of methyllithium, preserving enantioenrichment at phosphorus . We also explored a polarity reversal strategy in the context of the Atherton-Todd reaction , in which the nucleophilic P-H moiety was converted to electrophilic P-Cl then trapped with heteroatom-based nucleophiles. In this context, nucleophiles such as benzylamine and tyrosine could be employed to access phosphonate esters and phosphonamidates 7a/b with high yields and enantiospecificities. To investigate phosphoryl-radical-mediated reactivity, we adapted the lithium phenoxide conditions to a previously reported 1,6-coupling between H-phosphinates and benzoquinones , finding that O-phosphoryl hydroquinone derivative 8a was obtained with excellent stereospecificity. This result indicates that the 3a-derived phosphoryl radical possesses sufficient configurational stability in the absence of chiral control elements to engage in stereospecific reactions . Finally, 3a was subjected to a sulfuration-methylation sequence to afford phosphonothioate 9a with excellent enantiospecificity . |
65bd797a66c1381729d75b36 | 4 | Recognizing that gleaning insight into the present system might enable extension of this catalytic asymmetric strategy to other reactions involving the intermediacy of phosphonium species, we embarked on efforts to elucidate the catalytic mechanism as well as the origins of catalyst enantioinduction and rate acceleration. The canonical Michaelis-Arbuzov reaction proceeds through formation of an intermediate alkylphosphonium ion -in this case a protonated dibenzyl phosphonite -that subsequently undergoes dealkylation through nucleophilic substitution (SN2) by the counterion . Evidence for the intermediacy of a protiophosphonium ion in the catalytic system was obtained by monitoring the reaction of 2a with HCl catalyzed by thiourea 1g by 31 P NMR (Fig. ). At the earliest timepoints the signal due to 2a at δ 159 ppm is not observed, replaced instead by a doublet at δ 52 ppm that diminishes gradually with concomitant appearance of product 3a at δ 24 ppm (see 20 Supplementary Materials, Fig. ). The transient signal at δ 52 is assigned to the phosphonium chloride intermediate 10a based on the 710 Hz (J 1 P-H) coupling constant and the spectrum of an independently prepared phosphonium tetrafluoroborate. These observations not only support the intermediacy of phosphonium chloride 10a, but also indicate that it is the resting state of the substrate under the catalytic conditions. |
65bd797a66c1381729d75b36 | 5 | The resting-state form of catalyst 1g was investigated by employing diffusion-ordered NMR spectroscopy (DOSY) to measure its diffusion constant under catalytic reaction conditions as a means of estimating its molecular weight (see Supplementary Materials, Section 8.2). The trifluoromethyl groups on 1g provided a sensitive handle for 19 F NMR measurements at concentrations of 1g approximating the catalytic conditions (10 mM, 5% Et2O/toluene-d8, -70 °C). |
65bd797a66c1381729d75b36 | 6 | With the molecular composition of the resting-state complex established, we endeavored to determine the kinetic dependence on [1g]T and [10a] to elucidate the stoichiometry of the ratedetermining transition-state complex (Fig. ). The distinct infrared absorbance of the P=O bond in 3a (1240 cm -1 ) and the symmetric P-O stretch in 10a (1039 cm -1 ) provided excellent handles to monitor reaction progress by in situ IR spectroscopy. By systematically varying the concentration of 1g and applying Bures' normalized time scale treatment to the reaction profiles , we obtained excellent graphical overlay only when dividing the reaction profiles by [1g] n for n = 1, indicating that the reaction rate exhibits first-order dependence on catalyst [1g]T . Furthermore, the consumption of phosphonium species 10a and the formation of product 3a both follow a zerothorder kinetic rate behavior for the first ~80% of the reaction (Fig. ), consistent with a turnoverlimiting and enantiodetermining dealkylation transition state proceeding from a 1:1 1gꞏ10 restingstate complex. |
65bd797a66c1381729d75b36 | 7 | A catalytic cycle consistent with all available mechanistic data is depicted in Fig. . This cycle features (i) a protonation equilibrium that favors the phosphonium chloride 10a, (ii) binding of 10a to monomeric catalyst 1g forming the resting state, and (iii) turnover-limiting and enantiodetermining dealkylation to form product 3a and benzyl chloride, which dissociate from the catalyst to turn over the catalytic cycle. |
65bd797a66c1381729d75b36 | 8 | Having determined the molecular composition of the key dealkylation transition state, we turned to computational modeling to elucidate the nature of catalyst-substrate interactions leading to rate acceleration and enantioinduction. A systematic search of transition-state conformers led to the identification of low-energy diastereomeric structures leading to the major (R) and minor (S) enantiomers of 3a, with relative energies in excellent agreement with the experimentally observed enantioselectivities. In both transition structures, the catalyst engages in a network of stabilizing noncovalent interactions with both cationic and anionic components of the phosphonium ion pair (Fig. ), in agreement with previous studies on thiourea-arylpyrrolidine scaffolds . These interactions appear to stabilize both charged components of the ion pair, consistent with our original design hypothesis. Close analysis of the computational models reveals specific stabilizing interactions that might be responsible for enantioinduction. The diastereomeric transition state structures leading to the minor (S) and major (R) product enantiomers display almost identical catalyst geometries but are related by a 120-degree rotation of the phosphonium ions within the catalyst active site. TS cat,S positions the phosphonium P-Ph group below the 2-phenanthryl group of the catalyst, while splaying its benzyl groups (highlighted green). The lower energy TS cat,R , in contrast, positions the phosphonium P-Ph group into solvent and stacks the two benzyloxy groups, with one residing below the 2-phenanthryl group of the catalyst. While TS cat,S and TS cat,R both possess several noncovalent attractive interactions in common, the following interactions are likely enantiodifferentiating: TS cat,S possesses one additional H-bonding interaction between a benzylic C-H and the amide oxygen (Fig. . orange box) while TS cat,R incorporates two benzylic C-H-π interactions in (Fig. , blue box). The net energetic benefit of losing one benzylic C-H hydrogen bonding interaction and gaining two benzylic C-H-π interactions may be sufficient to account for the sense and magnitude of enantioinduction . High enantioselectivity in Arbuzov-type reactions promoted by 1g demands that the catalyzed pathway is faster than the uncatalyzed racemic background reaction, a phenomenon that was confirmed in kinetic experiments (Fig. ). Given that spectroscopic and kinetic studies establish the resting state in the catalytic reaction as 1g⋅10a (Fig. ), the phosphonium chloride ion pair must be more reactive when bound to HBD 1g than in the free state. This experimental observation would seem to contradict the established understanding that SN2 reactions involving ionic nucleophile-electrophile pairs are strongly inhibited by hydrogen bonding interactions such as those that arise in protic media . In such systems, hydrogen bonding is understood to give rise to rate deceleration by several orders of magnitude due to preferential stabilization of the fully charged resting state over the partially charged transition state (Fig. ). In attempting to understand how HBD catalyst 1g might achieve rate acceleration in the phosphonium dealkylation step, we noted that the computationally modeled transition-state structure (Fig. ) predicts that 1g binds simultaneously to both cationic and anionic components of the phosphonium ion pair. We hypothesized that the catalytic activity of 1g might arise from this synergistic, geometrically well-defined ion-pair binding mode that would not be possible in simple hydrogen bonding environments such as protic solvent. |
65bd797a66c1381729d75b36 | 9 | To investigate this possibility experimentally, we prepared structural variants of 1g designed to isolate its anion-and cation-binding domains (Fig. ). The simple anion-binding variant 11 was constructed by replacing the t-Leu-arylpyrrolidine moiety with a n-octyl group, and was found to be a strong inhibitor of dealkylation relative to the racemic background reaction (Fig. , left inset). The cation-binding variant 12 was synthesized by Smethylation of the thiourea, thereby removing the dual-HBD properties of the catalyst. Compound 12 also induced no rate acceleration compared to the background reaction. Combining 10 mol % each of anion-binding variant 11 and cation-binding variant 12 also did not result in rate acceleration compared to background. Taken together, these observations provide compelling evidence that the presence of both cation-and anion-binding domains as well as their precise relative spatial orientation as in 1g are necessary for catalysis. |
65bd797a66c1381729d75b36 | 10 | Computational modeling of the catalyzed and uncatalyzed dealkylation reactions of phosphonium chloride 10 was performed with continuum solvation in a low dielectric (PCM, toluene, 𝜖=2.38) to mimic the experimental reaction conditions (Fig. ). In the absence of catalyst, the phosphonium chloride 10 is found to rest as a tight ion pair with a cage-like structure in which the chloride anion engages in multiple stabilizing interactions with the cation. All of the stabilizing H-bonding interactions and a significant portion of the Coulombic attraction present in the ground state must be sacrificed to attain the linear geometry mandated by the SN2 mechanism as in TS uncat . In considering this geometric reorganization, the concerted pathway for dealkylation can be partitioned conceptually into an ion-pair reorganization phase followed by an ion-pair collapse phase. These two phases can be demarcated by a non-stationary state 10' located on the computed intrinsic reaction coordinate, wherein the chloride ion is positioned along the SN2 trajectory but formation of the C-O bond has yet to commence. By this analysis, >75% of the overall electronic activation barrier results from reorganization of the chloride anion in the first phase (10a 10'), whereas the ion-pair collapse phase corresponding to the covalent bond-breaking and forming events contributes <25% (ca. 4 kcal/mol) to the overall barrier. |
65bd797a66c1381729d75b36 | 11 | In the computed catalyzed pathway, the nucleophilic substitution is effectively broken into two discrete steps (Fig. ). Complexation of 1g to ion pair 10 results in chloride binding to the thiourea and association of the phosphonium ion to the cation-binding domain. Remarkably, the geometrical features of the phosphonium chloride in 1gꞏ10 are very similar to those in 10', i.e. the chloride is positioned in a nearly optimal pre-transition-state geometry (Fig. ). Thus, catalyst 1g can be seen as effectively turning the high-energy non-stationary point 10' on the uncatalyzed path into a relatively stable ground-state complex 1gꞏ10. |
65bd797a66c1381729d75b36 | 12 | The computational analysis enables a quantitative assessment of the effect of catalyst 1g on not only the ion-pair reorganization, but also the ion-pair collapse stage. Catalyst association leads to a higher barrier to ion-pair collapse relative to the uncatalyzed pathway (7.5 vs. 4.0 kcal/mol, Fig. ), consistent with the expected attenuating effect of H-bonding on the nucleophilicity of chloride. Yet inhibition of ion-pair collapse is more than compensated for by stabilization of phosphonium chloride ion pair in the reorganization phase. This results in overall rate acceleration by the HBD catalyst. |
65bd797a66c1381729d75b36 | 13 | Precise preorganization of reactants in favorable pre-transition-state geometries is well appreciated as a fundamental principle underlying rate acceleration in enzymatic catalysis, but is relatively underexplored in small molecule catalysis . The mechanistic insights gleaned from the catalytic Arbuzov reaction described here may hold broad implications for catalysis of ionic pathways where ion-pair reorganization represents a significant component of the turnoverlimiting reaction barrier. |
673cd39b7be152b1d0afdc25 | 0 | Molecular junctions generally consist of three main components: a molecular bridge , metal electrodes , and terminal anchor groups that electronically couple the bridge to the electrodes . Terminal anchors play a key role in controlling the electronic properties of single molecule junctions via molecular binding, junction stability, and electronic coupling. One class of robust terminal anchors is characterized by dative bonding interactions with metal electrodes, including amine , thiol , pyridine , and methyl thiol groups. However, molecular junctions generally require two terminal anchor groups to form a closed circuit, which poses synthetic challenges that tend to restrict the chemical space for molecular junction design. From this view, new binding modalities that avoid the requirement of two terminal anchors could expand the scope of the chemical toolbox available for molecular electronics. |
673cd39b7be152b1d0afdc25 | 1 | In addition to dative anchor groups, covalent anchors have been reported to yield stable and highly conductive molecular junctions . However, direct formation of Au-C covalent bonds in molecular junctions is challenging and typically requires cleavage or reaction of functional groups such as iodine , alkyne , or diazonium to generate Au-C bonds through in situ reactions. Recently, oriented external electric fields have gained increasing attention for their ability to reorganize the electron distribution of molecules, stabilize charge-separated resonant forms, and promote non-redox reactions .This prior work suggests that it may be possible to form dynamic anchors in nanoscale junctions under external electric fields without the need for cleavage of functional groups, which could provide a new molecular junction binding mechanism for single molecule electronics. However, the structural and electronic properties that promote the formation of Au-C contacts under external electric fields are not yet fully understood. |
673cd39b7be152b1d0afdc25 | 2 | Singly anchored molecules lack a second terminal anchor group to complete a closed electronic circuit. Prior work has reported that electron transport in singly anchored molecules occurs by non-covalent interactions such as π-π stacking and Au-π interactions . However, the unanchored terminus is exposed to a strong electric field due to the close proximity to the metal electrode, which could induce a dynamic second anchor via a redox reaction and subsequent formation of an Au-C bond . In addition, understanding the electron transport behavior in π-conjugated molecules with one terminal anchor group could provide additional insights into non-covalent intermolecular interactions. Intermolecular charge transport in π-stacked aromatic groups is critical to organic electronics . The efficiency of electron transport in π-stacked systems depends on the electronic coupling and the distance and orientation between neighboring πstacked molecules . Although recent work has examined electron transport in singly anchored organic molecules , we lack a complete understanding of the potential for dynamic anchor formation for singly anchored molecules under external electric fields. |
673cd39b7be152b1d0afdc25 | 3 | In this work, we investigate the electron transport behavior of singly anchored p-terphenyl derivatives using a combination of automated chemical synthesis, single molecule electronics experiments, molecular dynamics (MD) simulations, and non-equilibrium Green's function-density functional theory (NEGF-DFT) calculations. Synthesis of pterphenyl derivatives was performed using a rapid Suzuki-Miyaura cross coupling (SMCC) method with reduced reaction time and temperature . Following synthesis, the electronic properties of these molecules were characterized using the scanning tunneling microscope-break junction (STM-BJ) technique . Our results show that 4-amino-pterphenyl (PPP) has a surprisingly well-defined high conductance feature despite the presence of only one terminal anchor, but this high conductance feature is greatly diminished or absent in all other terphenyl derivatives studied in this work. Our results further reveal a low conductance feature for all singly anchored amino-p-terphenyl derivatives studied in this work, which arises due to non-covalent intermolecular interactions. Flicker noise analysis and machine learning methods such as correlation 2D analysis and Gaussian mixture modeling (GMM) are used to understand the conductance behavior. In addition, cyclic voltammetry (CV), bulk-scale electrolysis, and electron spin resonance (ESR) were used to understand the origin of the high conductance state, which arises due to Au-C bond formation due to a single electron oxidation event for molecular junctions . |
673cd39b7be152b1d0afdc25 | 4 | Synthesis of organic molecules for electronics experiments was carried out using Suzuki-Miyaura cross coupling (SMCC) by leveraging recent advances in iterative automated synthesis . Prior SMCC conditions for automated synthesis require more than twelve hours to prepare one molecule . Manual synthesis requires even longer time scales to synthesize a library of small molecules . In this work, we used automated iterative coupling based on a new rapid SMCC method (Supporting Information Section S1-S3) that significantly decreases the reaction time to ten minutes with high yield . |
673cd39b7be152b1d0afdc25 | 5 | In this work, the chemical space (Figure ) focuses on p-terphenyl molecules for molecular electronics due to their rigidity and extended π-conjugation, which results in low tunneling barriers . We designed a series of singly anchored p-terphenyl molecules based on the parent molecule PPP by including additional substituents on the three benzene rings. PPM, PMP, and MPP each contain a methyl group on one of the aromatic rings, which changes preferred resonance structures, alters ring torsional angles, or introduces steric hindrance during the formation of molecular junctions. Molecule PPF was incorporated to investigate the effect of electron withdrawing groups in contrast to the electron donating nature of the methyl substituted terphenyls. Three additional control molecules, PPN, PPS, and PPT, are characterized to understand the role of chemical anchor groups. |
673cd39b7be152b1d0afdc25 | 6 | The electronic properties of p-terphenyl derivatives were characterized at the single molecule level using the scanning tunneling microscope-break junction (STM-BJ) technique . The STM-BJ setup consists of a gold tip electrode that is repeatedly moved into and out of contact with a gold substrate electrode in a solution containing molecules, resulting in the continual formation and breakage of single molecule junctions (Supporting Information Section S1). The STM-BJ instrument is automated, and experiments are repeated over an ensemble of >5000 molecules for each experiment. Single molecule conductance data are then analyzed using one-and two-dimensional (1D and 2D) conductance histograms without data selection. The timescale of a single STM-BJ pulling trajectory is in the order of milliseconds , which allows for sampling a range of molecular conformations during a conductance measurement. Prior to understanding the electron transport behavior of singly anchored terphenyl derivatives, we characterized the molecular conductance behavior of the molecular analog with two terminal anchor groups (4,4'-diamino-p-terphenyl), and our results reveal a bimodal conductance distribution, with a prominent high conductance feature and weak low conductance feature, consistent with prior literature 6 (Supplementary Figure ). |
673cd39b7be152b1d0afdc25 | 7 | Single molecule conductance traces for PPP unexpectedly show two distinct populations, as demonstrated by characteristic single molecule conductance traces for high and low conductance features (Figure ). Singly anchored molecules are generally thought to form active junctions through intermolecular stacking interactions , where two different molecules, each anchored to a different electrode, form non-covalent dimeric interactions to complete the circuit. We hypothesized that the low-conductance feature in singly anchored p-terphenyl derivatives arises due to intermolecular stacked junctions (Figure ). However, in the STM-BJ setup, the applied bias between the two electrodes results in relatively high electric field gradients (~0.1-0.5 V/nm) in the nanoscale junction. We posited that the electric field could promote the formation of a dynamic terminal anchor in molecular junctions without requiring the cleavage or chemical reaction of additional functional groups (Figure ). We aimed to understand if such a mechanism could give rise to dynamic anchors resulting in well-defined conductance pathways in singly anchored organic molecules. |
673cd39b7be152b1d0afdc25 | 8 | We began by characterizing the electron transport properties of amino-p-terphenyl derivatives in non-polar solvents (1,2,4-trichlorobenzene, TCB) using the STM-BJ method. Our results show that amino-p-terphenyl derivatives exhibit a characteristic low conductance feature (Figure and Supplementary Figure ). Surprisingly, PPP shows an additional high conductance feature occurring in a significant molecular subpopulation over a large ensemble of molecules (Figure ). Unsupervised machine learning (Supporting Information Section S1), 2D correlation analysis, and Gaussian mixture modeling (GMM), were used to interpret the two-state conductance behavior observed for PPP. 2D correlation analysis and GMM indicates that the two conductance states are negatively correlated and therefore occur independently in distinct trajectories ). These results suggest that the two conductance populations arise from two different junction configurations. |
673cd39b7be152b1d0afdc25 | 9 | Flicker noise analysis was performed to differentiate between through-bond and through-space electron transport for the high and the low conductance states (Supporting Information Section S1). Prior work has shown that the conductance fluctuations (conductance noise power) exhibit a power law dependence on the mean conductance G values depending on through-space and through-bond transport characteristics . Conductance noise is quantified by numerically integrating the conductance noise power spectral density (PSD) between frequencies of 100 and 1000 Hz . The correlation is quantified by the scaling exponent n of the normalized noise power (noise power/G n ) versus the average normalized conductance G/G0, where G0 is the conductance quantum. A scaling exponent n ≈ 2 suggests through-space transmission whereas an exponent n ≈ 1 corresponds to through-bond transport. These results show that the low conductance state in PPP occurs by through-space electron transport, whereas the high conductance state shows dominant through-bond electron transport characteristics (Figure ). Flicker noise analysis for MPP (Supplementary Figure ) is also consistent with through-space electron transport for the low conductance state. Based on these results, we posit that the low conductance state observed for all aminop-terphenyl derivatives in this work arises due to non-covalent dimeric interactions. We posit that the high conductance state in PPP arises due to the formation of a dynamic covalent anchor binding to the metal electrode, which is consistent with through-bond transport for the high conductance state. Based on the structure of PPP and results from bulk electrochemical characterization (vide infra), we hypothesize that the second anchor involves the formation of an Au-C bond. To explore this hypothesis further, we pursued a series of additional experiments and simulations. |
673cd39b7be152b1d0afdc25 | 10 | 2D conductance histograms for amino-p-terphenyl derivatives indicate stark differences in electron transport pathways for different molecular composition. Notably, the high conductance state observed in PPP is significantly diminished or absent in all other singly anchored p-terphenyl molecules studied in this work (Figure and Supplementary Figure ), suggesting that molecular substituents disrupt the high conductance pathway in PPP. Although PPM and PMP show a weak high conductance feature, it is completely absent in MPP, implying that the presence of a methyl group at the para position on the aromatic ring abolishes the high conductance pathway. Oriented external electric fields are known to stabilize charge resonance structures and catalyze chemical reactions . It is possible that the quinoidal resonant form of the axisymmetric molecule PPP is relevant for electron transport, with electron density redistributing towards the para position on the phenyl ring farthest from the amine anchor. |
673cd39b7be152b1d0afdc25 | 11 | We hypothesize that the carbon atom at the terminal para position in amino-p-terphenyl derivatives PPP, PPM, and PMP forms a direct Au-C covalent linkage due to an oxidation event at the metal electrode, forming a terminal anchor that completes the circuit and leads to the high conductance pathway (Figure ). The absence of the high conductance feature in MPP is consistent with the hypothesis that the methyl group at the terminal para position inhibits the interaction between the molecule and the gold electrode, preventing the formation of the Au-C linkage. However, the para position remains available for binding in PPM and PMP, but the substituent groups on these molecules disrupt their symmetry, which disfavors the formation of the quinoidal resonance structure. PMP contains a methyl substituent on the central ring, which alters the torsional angles and hinders the formation of the planar quinoidal resonance form, resulting in a weak high conductance feature compared to PPP. |
673cd39b7be152b1d0afdc25 | 12 | We further characterized the electronic properties of PPF, which contains an electron withdrawing group as opposed to the electron donating group in the methyl substituted terphenyl molecules. PPF exhibits similar electron transport characteristics to PPM (Supplementary Figure ), suggesting that the nature of substituent does not significantly affect the high conductance state. These results indicate that disruption of symmetry in the amino-p-terphenyl derivatives can inhibit the high conductance state. In addition, we also characterized the electronic properties of PPN, PPT, and PPS, which contain thiol, pyridine, and N-methylamino as terminal anchor groups, respectively. The electron transport behavior of PPN, PPT, and PPS (Supplementary Figure ) indicates the absence of a high conductance state in these molecules. PPN exhibits a low conductance state, which is absent or significantly diminished in PPT and PPS, and likely arises due to non-covalent dimeric interactions. |
673cd39b7be152b1d0afdc25 | 13 | We conducted a series of additional single molecule experiments by varying temperature, concentration, and applied bias to further understand the high-and low-conductance populations observed for PPP. Temperature-dependent STM-BJ measurements were carried out at three different temperatures: 20 ºC, 30 ºC and 40 ºC. Increased temperature leads to enhanced molecular vibrations, reducing the likelihood of forming intermolecular junctions during the STM-BJ experiments . Our results indicate that as the temperature is increased, the low conductance state is significantly diminished (Figure ). These results are consistent with our hypothesis that the low conductance state arises due to non-covalent dimeric interactions. On the other hand, the high conductance state is largely unaffected by the 20ºC temperature increase, suggesting the high conductance state arises due to a stronger binding mechanism. Concentration-dependent STM-BJ experiments were also performed for PPP in the range on 0.1-10 mM (Figure ). These results indicate that the low conductance state of PPP exhibits strong concentration dependence, whereas the high conductance state is concentration independent. These results are consistent with our hypothesis that the low conductance state arises due to non-covalent dimeric interactions whereas the high conductance state arises due to through-bond transport. |
673cd39b7be152b1d0afdc25 | 14 | Bias dependent STM-BJ experiments (Figure ) reveal that at low applied bias (50-150 mV), only the high conductance state is observed. As the applied bias is increased, the low conductance state emerges, consistent with prior work reporting that increased bias regulates dimeric interactions in molecular junctions . The two-state conductance behavior of PPP is also observed in a polar solvent such as propylene carbonate (PC), but with lower counts for each molecular conductance sub-population as the higher dielectric strength of polar solvents likely reduces the effect of the electric field (Figure ). We also characterized the molecular conductance of PPP in PC in the presence of a reducing agent (sodium borohydride, NaBH4), which results in the disappearance of the high conductance state (Supplementary Figure ). Based on these results, we posit that an oxidation event leads to Au-C bond formation and the high conductance state observed for PPP. Bulk electrochemistry, electrolysis, and electron spin resonance were pursued to further understand the origin of the high conductance state. |
673cd39b7be152b1d0afdc25 | 15 | Integrating single molecule measurements with bulk experiments offers a powerful approach to understand electron transport . We performed a series of bulk electrochemical and spectroscopy experiments to understand the electronic properties of PPP, focusing on the origin of the high conductance state. Results from cyclic voltammetry (CV) experiments show a distinct oxidation wave on the forward scan at approximately 0.65 V vs. Ag/AgCl, with only a small reduction peak at 0.15 V on the return scan (Figure ). The absence of a prominent return peak in the CV indicates that a distinct chemical process occurs after oxidative electron transfer. A second distinct electron transfer event was observed when the positive limit of the potential window was expanded from 0.8 V to 1.0 V, however this second electron transfer event leads to the formation of a surfacebound species (Supplementary Figure ). Bulk electrolysis was performed to identify the nature of the products generated upon electrochemical oxidation of PPP. The potential was maintained at 0.8 V during electrolysis to isolate the first electron transfer event and avoid possible film formation on the electrode surface. The solution in the working electrode compartment exhibited dramatic changes in color, changing from colorless to purple upon oxidation (Figure ). All electrochemical experiments were performed in triplicates, and the integrated charge was used to determine the total number of electrons transferred during the electrochemical oxidation. Our results (Figure ) indicate that one electron was transferred, which suggests that the electrochemical oxidation of PPP forms a radical cation species. |
673cd39b7be152b1d0afdc25 | 16 | Electron spin resonance (ESR) spectroscopy experiments were performed on PPP before electrolysis, after electrolysis, and after preconcentration of the electrolysis product, as shown in Figure . The signals at ~3480 G and ~3650 G indicate the presence of a radical species in the electrolyzed solutions that is not present in the pre-electrolyzed sample. Based on these results, we posit that the high conductance state of PPP arises due to a radical species (Supplementary Figure ), which leads to the formation of a dynamic anchor based on a covalent Au-C linkage (Figure ). Computational modeling including MD simulations and NEGF-DFT calculations was further pursued to complement experimental results. |
673cd39b7be152b1d0afdc25 | 17 | A series of MD simulations was performed to elucidate the dimeric interactions between methyl substituted amino-p-terphenyl derivatives. For each terphenyl derivative, a pair of molecules was simulated in TCB and PC solvents, and the separation distance between their center-of-mass (COM) positions was systematically varied to provide molecular insights from both energetic and conformational perspectives (Supporting Information Section S1). The potential mean force (PMF) as a function of the separation distance between the COM of terphenyl pairs in solution was assessed using umbrella sampling and the weighted histogram analysis method (WHAM) . The PMFs of PPP, MPP, and PPM in TCB solution exhibit similar profiles, with the global minima located at a separation distance of 5.1 Å (Figure ). The corresponding binding energies (Supplementary Table ) between two terphenyl molecules is 0.63 ± 0.04 kBT for PPP, 0.68 ± 0.05 kBT for MPP, and 0.79 ± 0.04 kBT for PPM, where T is the absolute temperature at 300 K. On the other hand, the global minimum of PMP is situated at a separation distance of 5.9 Å with a binding energy of 0.19 ± 0.05 kBT due to the non- planar intramolecular conformation and steric hinderance caused by its methyl group at the middle phenyl ring (Supplementary Figure ). The observed binding energies for the terphenyl derivatives are below 1 kBT, which is consistent with results from the temperature-dependent STM-BJ experiments. |
673cd39b7be152b1d0afdc25 | 18 | We further characterized the structural features of the terphenyl molecules based on the MD trajectories of the PMF global minimum. Figure shows the angle distribution between the long axes of terphenyl molecules in TCB solvent, which quantifies the intermolecular structure between terphenyl dimers. An angle close to 180 0 indicates that the terphenyl molecules are aligned in the same parallel direction, increasing the likelihood of π-π stacking and providing an efficient conductance pathway. For MPP, a bimodal distribution is observed which could arise due to the methyl group leading to an offset in stacking for the dimeric structure. The symmetric geometry of PPP in TCB promotes a more parallel pair structure, ensuring an effective stacked dimeric structure. A similar trend in PMFs and conformational features was observed for peptides in PC solution (Supplementary Figure ). Molecular conformations generated by MD can be used in computationally efficient quantum mechanics (QM) calculations to aid in comparison between theory and experimental results. |
673cd39b7be152b1d0afdc25 | 19 | Electron transport calculations provide a powerful tool to validate experimentally observed conductance behavior . NEGF-DFT simulations were performed for methyl substituted amino-p-terphenyl derivatives in a stacked dimer geometry, corresponding to the low conductance state characterized by non-covalent intermolecular interactions (Supporting Information Section S1). For PPP, NEGF-DFT calculations were further carried out for the proposed high conductance state involving Au-C covalent linkages. Transmission plots for the stacked dimeric structures for PPP, PPM, and PMP show qualitatively similar behavior (Supplementary Figure ). Comparing the transmission values at the Fermi energy level (Figure ) shows that MPP exhibits lower conductance compared to the other derivatives, in both experiments and NEGF-DFT calculations. These results suggest that the presence of a methyl group at the terminal para position on the phenyl ring opposite to the amine can cause an offset in stacked molecular conformation. These results also support experimental results suggesting that the low conductance state arises due to non-covalent dimeric interactions. |
673cd39b7be152b1d0afdc25 | 20 | We also performed NEGF-DFT calculations for the high conductance state of PPP, which involves the formation of an Au-C covalent bond. Our results indicate that there is a tenfold difference in computed transmission values between the high and low conductance states of PPP (Figure ), in accordance with our experimentally observed STM-BJ results. The differences between experimental and computed transmission values could arise due to the semi-local exchange-correlation functional (PBE) used in our calculations, which tends to underestimate the HOMO-LUMO gap between the electrodes and molecule. In addition, the electron transmission calculations are carried out at zero applied bias. The experimental STM-BJ results indicate an increase in conductance with an increase in bias. Overall, NEGF-DFT calculations combined with MD simulations support the hypothesis that the low conductance state arises due to non-covalent intermolecular interactions and the high conductance state involves the formation of a robust Au-C covalent linkage. |
673cd39b7be152b1d0afdc25 | 21 | In this work, the electronic properties of singly anchored amino-p-terphenyl derivatives are characterized using automated chemical synthesis, single molecule electronics experiments, bulk scale electrochemistry, MD simulations, and NEGF-DFT calculations. Single molecule experiments reveal two well-defined conductance pathways in some amino-p-terphenyl derivatives due to a single electron oxidation event under an electric field. Bulk electrochemistry and spectroscopy experiments show that a radical cation state occurs when PPP is exposed to an electric field, which facilitates the formation of robust Au-C covalent linkages due to an oxidation event. Our results show that the formation of Au-C linkages is favored when the charge separated resonance state experiences minimal disruption. The introduction of substituents on the three benzene rings within the terphenyl system disrupts this state, hindering the high-conductance electron transport pathway. In addition, our work highlights the importance of non-covalent dimeric interactions in molecular electronics that can be leveraged for the design of materials for bulk scale measurements. MD simulations are used to understand the stacking conformations for various amino-p-terphenyl derivatives, and NEGF-DFT calculations are carried out to understand the electron transport behavior observed in single molecule experiments. Overall, our work highlights the formation of dynamic anchor groups in molecular junctions without requiring the cleavage or conversion of functional groups, which enhances the chemical toolbox available for constructing molecular electronics. |
638d8b950fd992e348387ddb | 0 | N EWTONIAN mechanics describe the force between two bodies as an action at a distance, and consequently, does not explain how force transfers through space between bodies. The concept of force cannot be understood mechanistically without introducing a property to the space more than just being an empty vacuum for the formation of force or action at a distance. Even though the conjecture about the existence of a medium in space has been introduced before even classical mechanics was invented, there have been controversial debates on its reality. Maxwellian electromagnetism specifies a property to the space through which electromagnetic waves as a field flow continuously through it in time. Later on, Einstein provided a geometrical explanation on basis of the Pythagorean theorem to calculate the distance between two points on a curved spacetime using the Riemann curvature tensor. Over a decade ago, Eric Verlinde, inspired by holographic principle , and its realization of ADS/CFT correspondence , described Gravity as an emergent phenomenon on the boundary of spacetime containing a large number of degrees of freedom. He found out that Newton's second law and gravitational force recover from the second law of thermodynamics. |
638d8b950fd992e348387ddb | 1 | The holographic principle developed by Hooft and Susskind states that the bulk properties of a system can be understood from the properties on the boundary of that system. A special realization of the Holographic principle discovered by Maldecena in so-called ADS/CFT correspondence shows that the gravity of a system in anti-de Sitter spacetime can be calculated from conformal field theory of entangled particles on the boundary of that system in one lower dimension. Understanding the bulk properties of a system from the boundary of that system, for the first time was hypothesized by Bekenstein in a thought experiment. He conjectured that M. Farshad is with the Department of Chemistry, Temple University, Philadelphia, PA, 19122 USA e-mail: [email protected]. the information of a particle that is absorbed by a black hole will lead to an increase in the area by expanding its boundary. Later on, Hawking postulated an entropy relation for its area using quantum field theory in the vicinity of the black hole, which elegantly supported Bekenstein's conjecture on information conservation. Insights from this principle and Jacobson's derivation of Einstein's gravity from the first law of thermodynamics and using the concept of entanglement entropy led Eric Verlinde to his coarse-grained hypothesis that it is the collective entropic force of microscopic degrees of freedom on the boundary of spacetime that is the cause of gravity. It is the emergent entropic force resulting from random Brownian motions of degrees of freedom that is governing the underlying physics behind the gravitational force. , , Recently we derived the displacement relation from the Schrodinger equation and Einstein's theory of diffusion. It is the complex version of the displacement relation that Verlinde postulated considering Bekestein's thought experiment on the black hole area and its entropy. We recover the gravitational entropic force when two bodies are at distance from each other using the first law of thermodynamics. Following this, we derive the attractive entropic force between two bodies by incorporating the radius of both bodies into the entropic force. In the end, we intuitively and visually describe that the repulsive entropic force can emerge from the rotation of degrees of freedom around spinning particles with opposite spin signs. |
638d8b950fd992e348387ddb | 2 | For the sake of simplicity, we assume our system is in equilibrium with its surrounding, and therefore the thermodynamics process is isothermal and reversible where ∆E = 0. Since the absorbed heat is proportional to the change of entropy of surounding or system (∆S sys = ∆S sur ) at a given temperature, the work in the form of expansion is equal to: |
638d8b950fd992e348387ddb | 3 | also, if M is the mass of N degrees of freedom within the emerging space, based on Einstein's theory of special relativity, the energy of such space is E = M c 2 . Using this equation and Eq. 13 from the equipartition theorem, we derive the Newtonian gravity from entropic gravity as follows: |
638d8b950fd992e348387ddb | 4 | We showed the derivation of Verlinde's entropic gravity from Unruh's formula. Also, we showed its relation to Newton's second law using the first law of thermodynamics and incorporation of the displacement and entropy relations into it. We further showed that this entropic force is equal to Newton's gravitational force using the equipartition theorem and the number of degrees of freedom relations. The number of degrees of freedom with the total mass of M in emerging space forms the entropic force through Brownian motions. Even though this intuitive incorporation of degrees of freedom explains the underlying physics behind the force, it does not assign any direction to these degrees of freedom. Moreover, the entropic force relation does not incorporate the radius of the bodies and the distance between them. To implement the direction of the force produced by these degrees of freedom as well as the radius of the bodies and the distance between them, first, we show a schematic representation of a two-dimensional system filled with these degrees of freedom around two large and small bodies. The left image in Figure shows two bodies with large blue and small black colors surrounded by homogeneously distributed degrees of freedom throughout the space. It is shown that a small round black particle is being attracted by gravitational force toward a black screen represented by a black line as the diameter of the large blue body. |
638d8b950fd992e348387ddb | 5 | Eq. 18 is the entropic force with the incorporation of the radius of both objects and their distance from each other. When the radius of the large pulling body goes to infinity or when the distance between two bodies goes to zero, we achieve Eq. 9. On the other hand, F = 0 when the radius of the large pulling body goes to zero or the distance between two bodies goes to infinity. |
638d8b950fd992e348387ddb | 6 | We applied this modification to the entropic force using a routine modification of the classical gravitational force. However, if we want to be mechanistic, we need to know how gravitational force on the body from the other body is created by entropic forces of the degrees of freedom continuously through space. Here, we suggest the ratio of degrees of freedom that apply force on the particle from up to down would be proportional to the entropic force. This is achieved by dividing the volume above the particle by the volume below the particle (Figure ). This gives: |
638d8b950fd992e348387ddb | 7 | We derived an entropic force for two bodies with the radii of R and R ′ at the center-to-center distance of r = R+h from each other, where h is the sum of the distance between two surfaces and R ′ . Without the extension of the entropic force formula to incorporate the distance between two bodies, it is assumed that the bodies are at contact and R → ∞. Under these circumstances, we return: |
638d8b950fd992e348387ddb | 8 | As is seen in Eq. 25, this is the same as the force between bodies in contact with each other. This resulted from the ratio of degrees of freedom that attract the body toward the other body (above) to the degrees of freedom that repel the body from the other body (below). Part of the degrees of freedom is blocked by the large body from repelling the other body. This disproportion in degrees of freedom results in a net gravitational entropic force on the body. |
638d8b950fd992e348387ddb | 9 | In the above derivation, we assumed only one body is applying force on the other body, this approximately makes sense when R >> R ′ . Otherwise, we should sum the two forces that two bodies apply to each other. In this regard, we derive the following equation for degrees of freedom relation: |
638d8b950fd992e348387ddb | 10 | We derived the attractive force between two particles at a distance resulting from free degrees of freedom in space pulling two bodies toward each other. However, we did not consider any rotating effects for these hard spheres. We believe the repulsion force will manifest from the rotation of two rotating bodies. We hypothesize that the degrees of freedom rotating around the bodies can be the root of repulsive force. Here, we present our conjecture with a visualized model in Figure . Figure shows that if the spinning particle, resulting from complex Browning motion of degrees of freedom, in turn, causes a rotation in the degrees of freedom. Two rotating bodies with the same spin signs would repel each other and with opposite spin signs attract each other. Depending on the spin sign of particles, degrees of freedom rotate clockwise or counterclockwise. The direction of rotation of degrees of freedom around bodies with the same spin opposes each other which causes a repulsive force between them. On the other hand, the rotation of degrees of freedom around bodies with the opposite sign is in the same direction. In our hypothetical model, interior degrees of freedom have higher force owing to their distance from the spinning particle. The rotation of degrees of freedom slows down with the increase in the distance and eventually goes to zero. |
638d8b950fd992e348387ddb | 11 | Despite being a simple concept of force as an emergent phenomenon through mechanical formation, it is as profound as a force resulting from the counterintuitive random Brownian motions of degrees of freedom. This means Brownian motions of matter in a medium are justified by allocating space filled with degrees of freedom. Our question is, what is the resulting emergent force that causes the Brownian motion of those degrees of freedom? We have two answers: 1) It is either formed by some other unknown degrees of freedom, presumably, even smaller than degrees of freedom that contribute predominantly to interactive forces of matter. 2) The universe is working like a mechanical cycle similar to a machine such as a mechanical watch. The second hypothesis is relatively intuitive compared with the first hypothetical solution as we can stop at a minimum size of a degree of freedom as inferred from quantum theory. Consequently, one does not need to go to infinitesimal sizes to fully understand the mechanism of force formation. However, the imagination of such a system to systematically explain the mechanical force among all degrees of freedom is extremely difficult. |
638d8b950fd992e348387ddb | 12 | We were able to derive a formula for emergent entropic force by considering that the space is filled by small interactive Brownian degrees of freedom. We mechanistically showed that if we find the ratio of attracting to repelling degrees of freedom, the gravitational force on the smaller body toward a much larger body will recover when they are in contact. We further modified the degrees of freedom ratio to incorporate both bodies' radii which is necessary for the calculation of other forces where the size of particles are comparable. After presenting the attractive force of still bodies, we hypothesize that the spin of bodies is responsible for the formation of repulsive force. |
62ab7ce404a3a956f9495494 | 0 | The large number of combinations of the twenty natural amino acids that form proteins impart to them high structural flexibility and functional diversity. Quantum mechanical (QM), non-covalent interactions play a critical role in these diverse structures and functions. For example, non-covalent interactions between amino acids range from stronger charge-assisted or low-barrier hydrogen bonds and salt bridges to weaker hydrogen bonds and dispersive interactions. Computational, atomistic modeling provides essential insight into the dynamics and non-covalent interactions of proteins. Nevertheless, the size of proteins along with the timescales over which they dynamically rearrange means that classical molecular modeling is nearly exclusively used to understand protein structure-function relationships . This is at odds with the fact that proteins are known to carry out a number of inherently quantum mechanical functions, including shaping the electric field of the active site to influence chemical bond formation and altering the strength of noncovalent interactions critical for catalytic action. |
62ab7ce404a3a956f9495494 | 1 | When large QM regions have been used in multi-scale quantum-mechanical-molecularmechanical (QM/MM) simulations of enzyme catalysis, distinct nuclear and charge dynamics have been observed in comparison to those using small QM regions. This is consistent with observations that much of the protein must be studied quantum mechanically to obtain asymptotic convergence of QM properties such as the favorability of proton or charge transfer , electric fields , excitation energies , bond critical points and partial charges . Nevertheless, conclusions from QM/MM studies of charge fluctuations are potentially limited by the significant effects of the boundary treatment and embedding method on the charge distribution in the QM region. Although some QM-derived effects can be captured using polarizable force field modeling , charge transfer and dynamical formation of charge-assisted hydrogen bonds remain outside the scope of those methods. |
62ab7ce404a3a956f9495494 | 2 | Recent advances in hardware and algorithms have made it possible for increasingly large proteins to be studied with a full QM treatment. These simulations have revealed the importance of first principles to accurately describing unexpected structures and to explaining charge transfer and polarization in water . Proteins are not just flexible but undergo concerted changes in shape, meaning that the motions of residues (e.g., changes in positions of Ca atoms or dihedral angles) are coupled. Analysis of geometric coupling has been extensively applied to understand this conformational allostery in proteins.[106-108] However, similar analysis has only recently begun to be applied to the quantum mechanical properties of proteins in order to understand the extent to which the QM charge distribution among protein residues varies dynamically. |
62ab7ce404a3a956f9495494 | 3 | We recently carried out a study of three diverse peptides (i.e., a lasso peptide, Trpcage, and a small helical protein, mini-CD4) that found charge distributions sampled from ab initio molecular dynamics to be broad. We showed that breadth of charge distributions was associated with significant pairwise coupling of the charges between residues, including those that are distant in both space and sequence. These observations added to a growing body of research that has demonstrated the value of studying these couplings to understand dynamic events in materials , interpret QM/MM simulations , and to guide QM method selection . In this recent work , we quantified both the mutual information and linear cross-correlation of charge distributions and determined that they bore little similarity to the more commonly studied geometric coupling values in these proteins. Nevertheless, one limitation of that study was that the diverse nature of the three proteins did not allow us to determine if small changes in protein sequence could be detected through charge coupling analysis. |
62ab7ce404a3a956f9495494 | 4 | In this work, we use the GFN2-xTB semi-empirical tight-binding method to simultaneously and rapidly capture the dynamics of both nuclei and their electrons in small model peptides that differ by only a few residues in their primary sequence. We specifically investigate the extent to which charge coupling can explain differences in stability for variants of Trp-cage that differ by a small number of N-terminal substitutions[117, 118] (Figure ). TC10b was designed to have greater helical propensity than TC5b, increasing the stability of the Nterminal a-helix . We focus on Trp-cage because its folding and unfolding has been widely studied . Furthermore Trp-cage's ca. 300 atom size makes it possible to carry out relatively long-time dynamics (i.e., 800 ps in total). Our analysis reveals a shift in the network of charge-coupled residues when an N-terminal Asn is replaced with a nearly isostructural Asp residue. This observation suggests that charge coupling analysis provides complementary insight into protein structure-function relationships alongside more established geometric coupling measures. We confirm that trends observed in the electronic structure of the semi-empirical models hold when snapshots are carried out using full DFT modeling with a range-separated hybrid functional, potentially guiding lower-cost strategies for obtaining charge coupling data. |
62ab7ce404a3a956f9495494 | 5 | The protonation states and hydrogen positions obtained from these NMR ensemble structures were used as a starting point for all simulations (Supporting Information Table ). Three residues differ between Trp-cage TC5b[117] and its TC10b mutant[118]: Asn1, Leu2, and Ile4 in the former are replaced by Asp1, Ala2, and Ala4 in the latter (Figure ). As a result, the net charge of TC5b is +1, whereas TC10b has a net charge of 0 (Supporting Information Table ). |
62ab7ce404a3a956f9495494 | 6 | The first 20 snapshots from the NMR ensemble were used to initiate molecular dynamics (MD) simulations (Figure ). Semi-empirical quantum mechanical (SQM) MD was carried out using GFN2-xTB[116] with D3 dispersion[126], as implemented in a developer version of TeraChem v1.9 . The conductor-like polarizable continuum model (C-PCM) was employed with a dielectric of 78.39 for water and a cavity generated from 1.2x Bondi's van der Waals radii[130] for all atoms, as implemented in TeraChem . These simulations were carried out in the NVT ensemble with an initial temperature of 300 K and a timestep of 0.5 fs. The Langevin thermostat was used with a damping frequency of 1 ps -1 . |
62ab7ce404a3a956f9495494 | 7 | Mulliken partial charges from GFN2-xTB were collected for 20 ps (i.e., 40,000 steps) for each of the 20 starting structures for both mutants. These charges were summed over each residue to mitigate shortcomings of the Mulliken partial charge scheme as in prior work . Both charges and structures were recorded at every timestep. A subset of 400 evenly-spaced structures (i.e., every 0.2 ps) from the SQM MD runs were used to collect density functional theory (DFT) partial charges with the long-range corrected, range-separated hybrid wPBEh functional (w = 0.2 bohr -1 )[134] and the cc-pVDZ[135] basis set using implicit C-PCM solvation. Charge covariance and mutual information were computed using sklearn and numpy with an in-house script following previous work . Geometric coupling and hydrogen bonding analysis were computed using the cpptraj utility[136] in AMBER18 . |
62ab7ce404a3a956f9495494 | 8 | Comparisons were also carried out for implicitly and explicitly solvated classical MD for 20 structures from each protein's NMR ensemble. These classical MD simulations were carried out using the AMBER ff14SB force field implemented in the GPU-accelerated version of the AMBER18 code . Explicit water simulations used the TIP3P[141] force field, and at least 20 Å solvent was prepared around the protein using tleap. The Trp-cage protein was neutralized using Cl -, which was modeled with the TIP3P-compatible Cl -parameters implemented in AMBER. In all simulations, both water and protein bonds with hydrogen atoms were fixed using the SHAKE algorithm to afford a timestep of 2 fs. All explicitly solvated classical MD simulations followed the same protocol: i) 1000 steps of restrained minimization, ii) 2000 steps of unrestrained minimization, iii) 10 ps NVT heating to 300 K, and iv) 1 ns of NpT equilibration. A Langevin thermostat with random seed was used for all constant-temperature dynamics with a collision frequency of 5 ps -1 and random seed. Constant-pressure dynamics employed a Berendsen barostat with a rescaling frequency of 2.0 ps -1 . A real-space cutoff of 10 Å was used for particle mesh Ewald summation. Following equilibration, NpT production dynamics were carried out for 250 ns. |
62ab7ce404a3a956f9495494 | 9 | Implicit solvation simulations employed the generalized Born model[139, 142] and equilibration only involved steps ii and iii of the explicitly solvated equilibration protocol with no cutoff employed. This equilibration was followed by 250 ns of NVT production dynamics with a Langevin thermostat. DSSP secondary structure analysis was carried out on both implicit and explicit solvent MD runs, as implemented in cpptraj. We made comparisons of the first 20 ps of explicit MD with that from implicit MD to confirm the suitability of implicit MD in the semiempirical simulations. This analysis suggested that while unfolding occurred in some cases at longer times, no unfolding occurred in the early stages considered with the semi-empirical molecular dynamics (Supporting Information Figures ). |
62ab7ce404a3a956f9495494 | 10 | We focus first on the effect of changing the N-terminal residue from Asn in TC5b (PDB ID: 1L2Y) to Asp in TC10b (PDB ID: 2JOF). The change in steric bulk between the two residues is limited, but the change in charge of the residue can be expected to alter which non-covalent interactions are preferred (Figure ). To carry out this analysis, we obtained geometric and charge correlations using DFT from 400 equally spaced snapshots obtained with SQM on 20 different trajectories each initialized from the NMR ensemble of the two structures (see Methods). Trends are qualitatively unchanged for both Ca coupling and for charges when considering the full SQM trajectory (Supporting Information Figure ). Most trends in geometric correlations between the first residue and the rest of the protein are unchanged with the Asn to Asp mutation from TC5b to TC10b (Figure ). In both proteins, the strongest positive geometric correlations of the first residue are to its nearest-neighbor residue (i.e., Leu2 in TC5b and Ala2 in TC10b, Figure ). Other strong positive correlations are with the Gly11, Gly12, Pro13, and Ser14 sequence at the opposite end of the a-helix to the N-terminal residue (Figure ). Considering instead charge coupling, individual trajectories have more distinct behavior between the TC5b and TC10b variants (Figure ). Here, a negative correlation indicates significant charge transfer between two residues. We note that very strong negative correlations are observed between Asp1 and Ala4, Lys8, or Ser20 in TC10b that are absent from interactions between Asn1 and the same residues in TC5b (Figure ). The observation of strong coupling between Asp1 and Ala4 is consistent with the design intent and with classical MD simulations[119] that indicated increased helical propensity was the source of higher stability for TC10b. This difference in helical propensity is also apparent in the SQM MD simulations, with the first four residues of TC10b having a higher fraction of snapshots assigned as helical in comparison to those of TC5b. (Supporting Information Figure ). These trends observed on the QM snapshots are consistent with those observed from SQM charges on the full 20 ps trajectories (Supporting Information Figure ). If we make comparisons of changes in coupling of the other two differing residues (i.e., Leu2 to Ala2 or Ile4 to Ala4 in Tc5b and Tc10b, respectively), both charge and geometric coupling appear more comparable between the two proteins (Supporting Information Figures ). The largest enhancements in coupling for Ala2 over Leu2 are with its nearest neighbor Asp1, although the geometric coupling for this same pair of residues is actually reduced in TC10b (Supporting Information Figure ). For Ala4 mutation from Ile4, enhanced charge coupling in TC10b is observed in individual trajectories with Asp1 or Lys8 (Supporting Information Figure ). Each of these residues has its strongest geometric couplings with its nearest neighbors in both protein variants, with only modest long-range coupling to more distant residues (Supporting Information Figures ). Observations are similar whether this analysis is carried out using SQM or QM charges (Supporting Information Figures ). Thus, from a geometric or charge coupling perspective, the main impact of mutations between TC5b and TC10b appears to be primarily due to changes in the electronic structure of the N-terminal Asn to Asp. |
62ab7ce404a3a956f9495494 | 11 | We next aimed to quantify the geometric and electronic structure characteristics of the three key trajectories for which Asp1 in TC10b has a large charge coupling with the previously identified residues (i.e., Asp4, Lys8, and Ser20). To confirm these charge couplings were not a consequence of the method of cross-correlation analysis , we also computed the mutual information between the charge distributions (Figure ). Indeed, we observe high mutual information values for these three trajectories between Asp1 and the respective residues (Figure ). This observation holds despite the fact that geometric couplings are modest or weakly anticorrelated for the same residues over each of these trajectories (Figure ). Thus, it appears that charge coupling captures interactions between Asp1 and proximal residues (i.e., Ala4) as well as more distant residues (i.e., Ser20 or Lys8). The same observations can be made from the full SQM trajectory (Supporting Information Figure ). To elucidate what structural orientations give rise to this coupling, we next computed the distributions of charges and identified extrema for these three pairs of residues (i.e., Asp1 with Ala4, Lys8, or Ser20). A wide distribution of both QM residue charges is observed, with variations on the order of 0.4 a.u., regardless of whether the residue is neutral (i.e., Ala4) or charged (i.e., Lys8 or Ser20, Figure ). These charge distributions are comparable to those we had previously observed in small proteins for charged residues but larger than the typical charge distributions we had noted for most neutral residues (i.e., non-polar or polar). The greater statistics in SQM trajectories lead to an apparently narrower distribution of charges but are otherwise consistent with our QM observations (Supporting Information Figure ). Focusing on the most extreme cases in the charge distributions reveals which structures favor a more positive charge on Asp1 versus a more negative charge (Figure ). In the trajectories for which Lys8 or Ala4 coupling are strongest, Asp1 in the most extreme cases takes on a weakly positive charge (Figure ). In the Asp1-Lys8 structure, a salt bridge is observed between the sidechain of the residues (Figure ). For all three trajectories, the most negative charge on Asp1 occurs when it is distant from the paired residue, forming no interactions and thus having a net charge as low as -0.4 a.u.. In the case of Asp1-Ser20, this orientation appears stabilized by a salt bridge between the N-terminal and C-terminal backbone atoms (Figure ). Thus, very strong couplings between sequence-distant residues can arise due to dynamic hydrogen-bonding interactions. |
62ab7ce404a3a956f9495494 | 12 | Although enhanced a-helical propensity in classical MD simulations of TC10b suggests that fixed-charge force fields are able to capture some degree of enhanced stabilization for the mutant[119], the variations in partial charges we observe on the residues as hydrogen bonds dynamically form and break cannot be captured by fixed-charge force fields. We next aimed to gain a global view of charge and structural differences between TC5b and TC10b by computing the linear correlations among QM charges and Ca motions for all residues in the two proteins over all 20 trajectories. Although it is known that TC10b has enhanced stability with respect to the TC5b protein, few variations in geometric coupling are apparent between the two proteins (Figure ). Furthermore, the areas of greatest change in structural correlations do not involve the N-terminus and instead correspond to differences in the anti-correlated motion of Ser13-Arg16 versus Asn5-Asp9 (Figure ). Furthermore, the strongest couplings are consistently nearest-neighbor interactions, with relatively few positively correlated interactions between Asp1/Asn1 and neighboring residues (Figure ). Variations in charge coupling are much more apparent between the two proteins (Figure ). Here, the more stable TC10b has enhanced negative correlations (i.e., charge-transfer-mediated couplings) between Asp1-Ala4 and neighboring residues 1-8 (Figure ). This enhanced charge coupling is likely a signature of stronger interactions in the a helix. The salt-bridge between Asp9 and Arg16 also has stronger coupling in the TC10b variant (Figure ). Thus, overall analysis of charge coupling suggests strengthened interactions in comparison to the TC5b variant for the first four residues of the protein both with each other and with more distant residues. While most charge correlations between Asp1 and surrounding residues are stronger in TC10b than those for Asn1 with surrounding TC5b residues, we note one exception. The Gln5 residue in TC5b appears to have a stronger correlation with Asn1 than Gln5 with Asp1 in TC10b (Figure ). To identify the structural significance of this variation, we examined the trajectories over which the Asp1-Gln5 or Asn1-Gln5 couplings were strongest (Figure ). We noted that in both of these cases, the coupling between the Asp1 terminal NH3 + and Ser20 terminal carboxylate is reduced compared to the average due to competing interactions, but the Asn1-Ser20 charge correlation is essentially absent in the case of TC5b (Figure ). In both cases, the N-terminus backbone forms a hydrogen bond with Gln5's carbonyl sidechain to orient Asp1 or Asn1 away from any other sidechains (Figure ). When Asp1 is more highly charged, this hydrogen bonding interaction is disrupted, in part because the Asp1 sidechain forms an intramolecular hydrogen bond with the terminal NH3 + (Figure ). This analysis suggests that the different hydrogen bonding capabilities of the Asn1 sidechain cause the residue to preferentially interact with Gln5 instead of with the C-terminus, potentially contributing to the lower stability of the TC5b variant. The cartoon structures are colored the same as the symbols in the graph that show the extreme values of charge (i.e., blue for positive Gln5 charge and orange for relatively negative Gln5 charge) for Asn1 or Asp1 with Gln5, and other atoms are colored as follows: nitrogen in dark blue, oxygen in red, and hydrogen in white. |
62ab7ce404a3a956f9495494 | 13 | While nearest-neighbor interactions dominate the coupling matrices for both geometries and charges, the character of the strongest non-nearest-neighbor interactions differs. For geometric coupling, these strongest interactions are still close in sequence, with the strongest couplings generally occurring between second-nearest neighbors (i.e., Ser14-Arg16 or Trp6-Lys8). These are particularly strong for residues at the end of the a-helix or in the turn of the Trp-cage. The strongest couplings are also essentially identical between TC5b and TC10b variants, with none of them involving the mutated residues (Figure ). Thus, from a geometric perspective, the Ca couplings do not distinguish the two peptides. Turning instead to the charge coupling, we observe the strongest non-nearest-neighbor couplings to form an interaction map in TC10b that is distinct from TC5b (Figure ). Namely, both peptides have strong interactions in the a-helix for residues Asn1-Gln5 (TC5b) or Asp1-Ala4 (TC10b) and Gln5-Lys8 as well as in the salt bridge between Asp 9 and Arg16 (Figure ). However, the more structurally stable TC10b forms a circle of tight interactions with strong couplings for Arg16-Ser20 and Asp1-Ser20 which are not analogously present in TC5b (Figure ). Thus, the charge coupling analysis suggests that a potential source of stability for TC10b is a network of residue couplings that form throughout the protein. The observations on Trp-cage suggest that charge coupling can provide insights into residue-residue interactions not otherwise evident from geometric coupling. However, it remains computationally demanding to obtain full electronic structure descriptions of proteins. To emphasize the value of such analysis, we computed the relationship between charge coupling and geometric coupling over all residue pairs in the two proteins (Figure ). For the residues most strongly geometrically coupled, charge couplings tend to be negative (i.e., charge mediated, Figure ). However, very weak geometric coupling occurs alongside both positive and negative charge coupling (e.g., 1-16 vs 1-20 in TC10b, Figure ). A potential path forward is highlighted by the fact that we consistently obtained good agreement between charges on 20 ps of SQM dynamics with those from a much smaller subset of DFT QM snapshots. Thus, a relatively small number of single points sampled from classical molecular dynamics may reveal charge-mediated interactions as long as the trajectories are somewhat consistent with full ab initio MD. It is also expected as computing power and algorithms advance, fully ab initio, fragment-based, or dynamics using neural network potentials could make these calculations even more tractable. |
62ab7ce404a3a956f9495494 | 14 | We have carried out extensive semi-empirical QM dynamics of two variants of the miniprotein Trp-cage, TC5b and a more stable mutant TC10b. Small differences (e.g., Asp1 vs Asn1) in the sequence of the two peptides lead to differences in their thermal stability. From extensive SQM molecular dynamics of these proteins, we have found limited evidence that geometric coupling alone can distinguish the two peptides over the timescale studied. We have shown that differences are more apparent in the coupling of by-residue-summed charge distributions either from the full SQM MD run or from snapshots obtained with DFT. In particular, Asp1 in TC10b shows significantly enhanced coupling to both sequence-adjacent (i.e., Ala2) and more sequence-distant residues (i.e., Ala4, Lys8, and Ser20). These couplings are evident both from linear correlations as well as mutual information analysis. In all cases, Asp1 in TC10b samples a wide distribution of charges (ca. 0.4 a.u.) from weakly positive to significantly negative, highlighting potential limitations in the use of fixed-charge force fields. This greater flexibility of the charge distribution in Asp1 in comparison to Asn1 also leads to stronger overall coupling of the charge distributions of residues near the N-terminus. In comparison, geometric coupling appears comparable between TC10b and TC5b. |
62ab7ce404a3a956f9495494 | 15 | Specifically, residues roughly one a-helical turn apart couple most strongly from a geometric perspective and are similarly coupled in TC5b and TC10b. In comparison, non-nearest-neighbor charge couplings are stronger in TC10b than TC5b and indicate a network of residues with significant couplings up to 8 residues apart in primary sequence. Although it is not apparent from geometric analysis, this network of interactions can help explain differences in stability of the two peptides. Thus, our study highlights the complementary benefit of charge coupling analysis to interpret protein structure-function relationships. Although fully first-principles molecular dynamics remains cost prohibitive for proteins of typical size, advances in modeling such as artificial neural network potentials, polarizable force fields, or fragment-based methods could be paired with judicious choice of snapshots with DFT to capture important aspects of charge coupling. This analysis is also expected to be useful in the systematic design of QM regions for multi-scale modeling of catalysis where strong QM coupling of residues to an enzyme active site motivate their treatment with QM. ASSOCIATED CONTENT Supporting Information. Protonation states of TC5b and TC10b residues; RMSD of implicit and explicitly solvated TC5b; RMSD of implicit and explicitly solvated TC10b; SQM charge and geometric coupling for residue 1 in TC5b vs TC10b; Secondary structure analysis of SQM MD trajectories; SQM charge and geometric coupling for residue 2 in TC5b vs TC10b; SQM charge and geometric coupling for residue 4 in TC5b vs TC10b; QM charge coupling for residue 2 in TC5b vs TC10b; QM charge coupling for residue 4 in TC5b vs TC10b; Trajectory-specific SQM charge and geometric coupling for res. 1 in TC10b; 2D KDEs of charges between correlated pairs from full SQM of TC10b (PDF) from the Burroughs Wellcome Fund, an AAAS Marion Milligan Mason Award, and an Alfred P. Sloan award in Chemistry, which supported this work. |
64fb0789b6ab98a41c139327 | 0 | ABSTRACT. We report the synthesis and full characterization of a family of phosphorus containing polymethine cyanines (phospha-cyanines). The compounds are easily prepared in two steps starting from readily available phosphanes. The impact of the P-substituents and the counterions on the structural and optical properties is investigated through a joint experimental/theoretical approach. Based on the study of the single crystal X-ray diffraction structures, all phospha-cyanines present a bond length alternation (BLA) close to zero, independently of the substituent and the counterions, which indicates an ideal polymethine state. |
64fb0789b6ab98a41c139327 | 1 | All these compounds display the typical cyanine-like UV-vis absorption with an intense and sharp transition with a vibronic shoulder. TD-DFT calculations allowed to fully rationalize the optical properties (absorption/emission wavelengths, luminescence quantum yields). Interestingly, due to the tetrahedral shape of the P-atom, the optical properties are independent of the counterion, which is in marked contrast with N-analogs, which enables predictive engineering of the phosphacyanines regardless of the medium in which they are used. |
64fb0789b6ab98a41c139327 | 2 | Polymethine cyanines are organic compounds in which a charge (either positive such A, Fig 1, or negative) is fully delocalized between two heteroatoms (or heterocycles) over an odd number of sp 2 carbon atoms. This particular structure, called the ideal polymethine state (IPS) leads to a series of hallmark properties: (i) a bond length alternation (BLA) equal to zero as all the C-C bonds display a "one and a half" bond length, (ii) a sharp and particularly intense absorption band with a vibronic shoulder at high energy, and (iii) luminescence with a very small Stokes shift. The molecular engineering of the structure (polymethine chain length, nature of the heterocycle, counter-ion, central substituent) allowed to tune the absorption/emission of these dyes from the visible to the near infrared. Polymethine cyanines thus represent a fascinating platform to study the effect of the molecular structures on electronic delocalization as their exceptional characteristics can vanish when the "cyanine limit" is crossed. Thanks to these hallmark properties, cyanine dyes have been used in many applications in chemical detection, bio-imaging 4 and material sciences 5 . So far, the vast majority of cyanine dyes display capping groups encompassing nitrogen (A, Fig. ), even though it was shown that substituting it by oxygen-or sulfur-heterocycles significantly redshifts the absorption/emission. During the last decades, systems in which the nitrogen is replaced by phosphorus, its higher group 15 analogue, became very attractive compounds. Such chromophores benefit from the particular structural and electronic features of the P atom, e.g., a reactive lone pair, a pyramidal geometry preventing staking, and the possibility to display - conjugation. 7 Surprisingly, while phospha-cyanines have sporadically appeared in the literature since the 1950s, no systematic structure-properties relationship has been established on an entire family (B, Fig. ). 9 Interestingly whilst several articles devoted to the reactivity of organophosphorus derivatives report P-compounds fulfilling the structural requirements of phospha-cyanines (cationic compounds in which a charge is delocalized between two P over an odd number of sp 2 carbon atoms), this aspect is not commented in the articles nor the optical properties reported. In this context, we wish to highlight Yamada's approach, which studied the reaction of perfluorocyclopentene with various nucleophiles, including a phosphonium ylide, and reported a compound that can be structurally characterized as a phospha-pentamethine cyanine (Fig. , R = Ph). In the present contribution, we adapt this synthetic method to prepare a full family of dyes and systematically study their structures, and optical properties using a joint experimental and theoretical approach. From this study, we can conclude that these "phospha-cyanines" possess all the structural and spectroscopic properties of cyanines in their IPS. . Here, a library of methyl-phosphonium 1-6[X] are prepared from commercially available phosphanes and then converted to the corresponding phospha-cyanines 7- highlighting the stability of the platform. Nevertheless, slow oxidation to the corresponding phosphine oxide was observed when the compound was stored several days in solution under ambient conditions (Fig. ). Scheme 1: Synthesis of 7-12[X] 7-11[X] are characterized by X-ray diffraction performed on single crystals (see Fig. , Fig. and Table and). In addition, to confirm the molecular structure of the synthesized compounds, the crystallographic structure provides the opportunity to determine the BLA and therefore the "cyanine" character of the dyes. |
64fb0789b6ab98a41c139327 | 3 | At the molecular level, the cation always displays the cis geometry of the C1-C2 and C4-C5 (see and Fig. ). This result in a BLA in the crystalline state close to 0 (BLA < 0.019 Å, see Table ). Importantly, this absence of BLA is independent of the structure of the cation or of the counterion. DFT calculations performed in DCM provide bond distances that are close to the experimental values, the difference between measured and computed bond lengths being of the order of the impact of the counterion on the experimental data. Similarly, the P-C bond lengths are almost identical (1.733 Å <dP-C< 1.753 Å) and characteristic of a one and a half R3P-C bond, with again similar DFT values. The lack of BLA change when modifying the counterion is in marked contrast to the observations made on the polymethine cyanines. This can be attributed to the bulkiness of tetravalent P atoms which prevents significant interactions between the ions. Given the impact of the anion on heptamethine dyes optical properties, the fact that phospha-cyanine conserves its IPS irrespective of the counterion is of prime importance for further applications. |
64fb0789b6ab98a41c139327 | 4 | [cm 7 The optical properties of 7-12[X] were investigated in diluted CH2Cl2 solutions (c = 5.10 -6 mol.L - 1 , Fig. , Fig. -16 and Table ). All derivatives display similar absorption spectra with a sharp absorption band between 419 nm and 468 nm with high extinction coefficients (70-135 000 L.mol - 1 .cm -1 ) and a clear shoulder. Such band topology is typical of cyanines in their IPS, in agreement with the structural observations (see above). As evidenced in Fig. , no aggregation occurred in diluted DCM, as expected due to the presence of the bulky P-atom. In addition, the absorption spectra are independent of the solvent (Fig. ). The absorption maxima are in the typical range of dimethylamino-substituted pentamethine (Cyanine A (Fig. , n =2): labs = 416 nm), showing that the presence of P does not affect the conjugation along the polymethine bridge. Surprisingly, the effect of substituents (either electron-donating -OMe or electron-withdrawing -F) on the phenyl |
64fb0789b6ab98a41c139327 | 5 | ring attached to the P is trifling (labs(7[OTf]) = 448 nm, labs(7-9-10-11[OTf]) = 3 nm). However, when the nature of the substituent is changed (Cy vs Ph vs Napht), the impact on the absorption wavelengths becomes more significant as the maximal absorption wavelengths shift from labs (8[OTf]) = 419 nm to labs (12[OTf]) = 468 nm. We have performed theoretical calculations to probe the nature of the excited states of these systems. The protocol is detailed in the SI, and goes beyond standard TD-DFT using wavefunction theories for transition energies, as cyanines are known to be challenging for this level of theory. The vertical absorption wavelength computed is 408 nm for 7 + logically blue shifted as compared to experiment (see below). More interesting is that very similar wavelengths are determined for 9 + (410 nm), 10 + (415) nm, and 11 + (404 nm), whereas significantly smaller and larger absorption wavelengths are respectively predicted for 8 + (387 nm) and 12 + (439 nm), in perfect agreement with the experiments. Fig. displays the frontier orbitals and electron density difference (EDD) plot for 7 + . Consistent with a cyanine nature, the EDD plot displays an alternation on odd/even carbon atoms that undergo large decrease/increase of electronic density upon absorption. The changes of density on the atoms not involved in the cyanine moiety are essentially negligible but for the fluorine atom attached to the central carbon atoms that acts as a secondary donor group. The LUMO is nevertheless slightly delocalized on the side phenyl rings. In Fig. , the same representation is given for 8 + and 12 + and one notes similar topologies for the EDD and MOs, which respectively slightly less and more delocalized LUMO as compared to 7 + . |
64fb0789b6ab98a41c139327 | 6 | More quantitative theory experiment can be reached by considering 0-0 energies (for emissive dyes) and band shapes. For the latter, one can find in Fig. the theoretical band shape of 7 + , and it clearly fits the experimental one with a marked shoulder and a molar extinction coefficient fitting the experimental value as well. As 0-0 energies, theory returns 2.69 eV, 3.17 eV, 3.00 eV, 2.88 eV, 3.06 eV, and 2.99 eV in the 7 + to 12 + series. For the three cases in which it can be compared to experiment, namely 9 + , 10 + , and 12 + , the discrepancies between theory and experiment are of the order of 0.2-0.3 eV, which is usual for such level of theory. It has been shown on various cyanines dyes that the nature of the counterion strongly impacts the absorption in apolar solvents and in solid state. Here, the band shape is not affected by the nature of the counterion (Fig. ). This absence of interaction is due to the bulkiness of the tetravalent P-atom which prevents the ions to strongly interact, in contrast to the "classical" cyanines, in full agreement with the structural study based on X-ray diffraction (vide supra). This feature is a clear advantage for considering further applications in the solid state, as the absorption properties will be similar to those in diluted solution. No complex anion engineering will thus be necessary to maintain the cyanine behaviour. |
64fb0789b6ab98a41c139327 | 7 | Finally, emission could be recorded for some compounds (Table and Fig. , Fig. ). If the associated luminescence quantum yield remains poor ( < 0.5%), the measured Stokes shift is small, again fitting the cyanine structure. The trend observed in luminescence wavelengths follows the one observed in absorption. We have used theory to obtain some first hints on this absence of emission, or very low . First, optimization of the lowest excited-state of 7 + yields a bright state with an oscillator strength larger than 0.5 with both TD-DFT and CC2, indicating that there is no dark state quenching appearing in the excited state, consistent with the experimentally measured small Stokes shifts (for the emissive dyes). Second, given the presence of "heavy" phosphorous atoms, we envisaged the possibility of intersystem crossing (ISC). Theory returns only one triplet below the lowest singlet (on the S1 optimal geometry), but this triplet is both strongly energetically far (ST = 0.77 eV) and weakly coupled (SOC of 0.22 cm -1 ) with the corresponding singlet, which phosphanes. The impact of the P-substituents and the counterions on the structural and optical properties is investigated using a joint experimental/theoretical approach. All the phosphacyanines present a BLA close to zero according to the crystallographic structures, confirming their Ideal Polymethine State (IPS). All these compounds display a sharp UV-vis absorption band with a vibronic shoulder, characteristic of a cyanine in its IPS. The low luminescence quantum yield was rationalized on the basis of TD-DFT calculations. Interestingly, the optical properties of these phospha-cyanines are independent of both the solvent and the counterion. This is contrast in to its N-analogs and enables predictive engineering of the phospha-cyanines regardless of the medium in which they are used (biological media or materials) which is a clear advantage for potential applications. Finally, this study also paves the way toward the preparation of unprecedented cyanines dyes featuring heavier main-group elements, as those heteroatoms were recently described to induce unusual structural/electronic properties to chromophore/fluorophores. ASSOCIATED CONTENT Supporting Information. Synthetic procedure, complete characterizations, X-ray crystallographic data and CIF files, computational details and Cartesian coordinates. The following files are available free of charge. |
6407337ccc600523a3cd2921 | 0 | Moore's Law , which states that the number of transistors in a typical integrated circuit doubles approximately every two years, has been a driver for semiconductor device miniaturisation. This has required aggressive decreases in dimensions to pack more and more transistors into a single device and ensure that Moore's Law continues. To keep these miniaturised devices connected, the size of the interconnect must correspondingly decrease. However, interconnect scaling is now reaching its limits with current materials and processing technology. The rate of increase in the number of transistors on a chip is slowing compared to Moore's Law and makes interconnect scaling a serious bottleneck and highly limiting factor in device miniaturisation. Even with a move away from Moore's Law, device interconnects will continue to provide stern challenges. |
6407337ccc600523a3cd2921 | 1 | Several layers of interconnect lines are needed to connect all the transistors in a device. The closer the interconnect is to the transistors the smaller it must be, resulting in extremely high aspect ratio structures at the lowest device levels. Currently, interconnects are made with Cu metal. To create a functioning interconnect, a diffusion barrier and a seed layer or liner (also referred to as adhesion promoter) need to be deposited in the interconnect via. |
6407337ccc600523a3cd2921 | 2 | As the name suggests, the role of the diffusion barrier is to prevent the diffusion of Cu atoms into the surrounding dielectric, while the liner material promotes deposition of smooth, conductive Cu films. Current diffusion barriers, such as TaN, are unable to promote the deposition of conductive Cu without the addition of the liner material. |
6407337ccc600523a3cd2921 | 3 | Common liner materials include Ta and Ru. The barrier and liner materials take up the limited available volume in the interconnect via, to the point that even with modern deposition methods such as Atomic Layer Deposition (ALD) they cannot be thinned further without compromising their efficiency and this means that insufficient volume remains in the via to deposit the amount of Cu needed for a functioning interconnect. New developments in interconnects are clearly needed in order to drive further advances in electronic devices. |
6407337ccc600523a3cd2921 | 4 | • Wettability and adhesion of Cu From an experimental perspective, a large variety of analysis methods are available to determine barrier and liner performance. However, it is usually not feasible to focus on more than one specific aspect of the system, such as the failure mechanism of the diffusion barrier. Additionally, there is overwhelmingly more literature on the topic of diffusion barriers than there is for liner materials, and there are few studies that examine a barrier and a liner material together, or test a material for both barrier and liner properties. Some of the materials studied as combined barrier/liner materials include NiP alloy , Cr , RuMo alloy , CoWB , Ru , MoC doped Ru , Ru(P) , RuCo , Co , Co(W) , MnN and Ti . There is also some precedence for building upon existing knowledge by combining known diffusion barrier and liner materials. Chakraborty et al. studied mixed-phase RuTaN. Eisenbraun used plasma enhanced ALD (PEALD) to create mixed phase barrier/liner materials by layering known barrier (TaN, WCN) and liner materials (Ru, Co), starting with the barrier and ending with a liner layer. These materials showed comparable barrier properties to TaN. Additionally, the ratios of metals in the material can be used to further tune the material to have low resistivity and good wettability while maintaining strong barrier properties. This type of work also includes our work on Ru-doped TaN which demonstrated that Cu wetting can be promoted on TaN surfaces by incorporating a known liner material and provides the foundation for the work presented in this paper. While the work on combined barrier/liner materials is relatively limited, there is even less work on using theoretical methods to screen for potential barrier, liner or barrier/liner materials. Just as a wide variety of experimental methods is needed to study all aspects of a potential barrier and liner material, theory also requires a range of methods to attempt to study all of the above properties. This is usually not feasible and therefore it is reasonable to select the primary properties of interest and focus on modelling these. Adhesion is the property that is most easily studied using theoretical methods. Activation energy for atom diffusion can be determined using nudged elastic band calculations and can give insight into surface mobility of atoms as well as the diffusion barrier properties. Ab initio molecular dynamics calculations can be used to gain some insight into thermal stability as well as wettability and growth processes. Additionally, electronic properties such as density of states (DOS), atomic charges and electron density can be studied, which give useful insight into conductivity and band gap of the material and can also reveal trends that drive the formation of favourable structures. These are useful in determining the basic material properties such as conductivity and band gap. Models have also been developed to screen for materials based on their electronic properties. These are particularly useful when combined with experimental methods. Additionally, machine learning can now be applied to modelling the overall performance of interconnects which can save valuable time and resources during the development process. Overall, theoretical studies make up only a small fraction of the research on advancing interconnect manufacturing leaving plenty of room to gain further insight into growth mechanisms and material interactions relevant for application in interconnect technology. |
6407337ccc600523a3cd2921 | 5 | In our previous work we showed that by doping Ru atoms into the top layer of a TaN surface, a single material with both barrier and liner properties can be created. Such a material can be deposited using ALD. Use of a single barrier/liner material frees up volume for Cu in the interconnect and reduces the number of process steps. As part of these studies, we also developed a computationally efficient method for predicting the thin-film morphology of Cu on potential single barrier/liner materials. This is based mainly on the understanding of the forces that drive the growth mechanism of a material. |
6407337ccc600523a3cd2921 | 6 | Materials tend to follow a classical homoepitaxial growth mechanism when there is a strong interaction between the deposited material and the substrate. An example of this would be aluminium grown on aluminium substrate. For strongly interacting substrates, the interaction with the substrate is stronger than the interaction between atoms of the deposited material. A higher temperature is associated with 2D growth, making annealing a reasonable approach to create a conformal film. In contrast, the opposite becomes true for growth on a weakly interacting substrate. Here, the interactions between the atoms in the deposited material are stronger than the interaction with the substrate. Consequently, increased temperature tends to promote 3D growth. This leads to faceted structures and rough surfaces which are not desirable for interconnects. Additionally, while migration of atoms across the surface can occur, Gervilla et al. described that upward migration of atoms to form new layers is not possible on strongly interacting substrates. This is a crucial difference between the two mechanisms. By enhancing the adhesion between the adatoms and the substrate a homoepitaxial growth mechanism, of for example copper on a nitride, can be promoted, which in turn promotes 2D growth of the thin film. Computing the activation energies for atoms to migrate between layers lets us determine which mechanism is dominant based on the magnitude of the activation energy. The different migration pathways for homoepitaxial growth and weakly interacting substrates are shown in Figure ). |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.