SEARCH · arXiv Science

Results for “physics.chem-ph”

Search indexed arXiv papers on artificial intelligence, large language models, computer vision and robotics. Read source abstracts and follow links to arXiv.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 415 records · Page 23Linked to original sources

arXiv2018

Carbonate-Bridged Dinuclear and Trinuclear Dysprosium(III) Single-Molecule Magnets

In 2016, we reported a single-ion magnet [Dy(bbpen)Br] with an energy barrier over 1000 K. Here a dimeric [Dy2(mu-CO3)(bbpen)2(H2O)].H2O.CH3OH (1) and a trimeric [Dy3(mu3-CO3)(bppen)3](CF3SO3).H2O (2) single-molecule magnets (SMMs) were obtained through replacing the Br- anion with the carbonate bridge. Their effective relaxation barriers at zero dc field are decrease to 51 K and 422 K, respectively, which are consist with their structural modifications.

cond-mat.mtrl-sci↗

arXiv2018

Machine Learning of Energetic Material Properties

In this work, we discuss use of machine learning techniques for rapid prediction of detonation properties including explosive energy, detonation velocity, and detonation pressure. Further, analysis is applied to individual molecules in order to explore the contribution of bonding motifs to these properties. Feature descriptors evaluated include Morgan fingerprints, E-state vectors, a custom "sum over bonds" descriptor, and coulomb matrices. Algorithms discussed include kernel ridge regression, least absolute shrinkage and selection operator ("LASSO") regression, Gaussian process regression, and the multi-layer perceptron (a neural network). Effects of regularization, kernel selection, network parameters, and dimensionality reduction are discussed. We determine that even when using a small training set, non-linear regression methods may create models within a useful error tolerance for screening of materials.

cond-mat.mtrl-sci↗

arXiv2018

High-resolution spectroscopy on cold electrically trapped formaldehyde

We present precision spectroscopy on electrically trapped formaldehyde (H$_2$CO), demonstrating key attributes which will enable molecular spectroscopy with unprecedented precision. Our method makes use of a microstructured electric trap with homogeneous fields in the trap center and rotational transitions with minimal Stark broadening at a 'magic' offset electric field. Using molecules cooled to the low millikelvin temperature regime via optoelectrical Sisyphus cooling, we reduce Stark broadening on the $J=5\leftarrow4$ ($K=3$) transition at 364 GHz to well below 1 kHz, observe Doppler-limited linewidths down to 3.8 kHz, and determine the line position with sub-kHz uncertainty. Our results and clear prospects for even narrower spectra pave the way towards in-trap precision spectroscopy on diverse molecule species.

physics.atom-ph↗

arXiv2018

Bond type restricted radial distribution functions for accurate machine learning prediction of atomization energies

Understanding the performance of machine learning algorithms is essential for designing more accurate and efficient statistical models. It is not always possible to unravel the reasoning of neural networks. Here we propose a method for calculating machine learning kernels in closed and analytic form by combining atomic property weighted radial distribution function (AP-RDF) descriptor with a Gaussian kernel. This allowed us to analyse and improve the performance of the Bag-of-Bonds descriptor, when the bond type restriction is included in AP-RDF. The improvement is achieved for the prediction of molecular atomization energies and is due to the incorporation of a tensor product into the kernel which captures the multidimensional representation of the AP-RDF. On the other hand, the numerical version of the AP-RDF is a constant size descriptor, and it is more computationally efficient than Bag-of-Bonds. We have also discussed a connection between molecular quantum similarity and machine learning kernels with first principles kind of descriptors.

physics.chem-ph↗

arXiv2021

Fast Fisher-Lee approach for conductance calculations on BTB-based molecular junctions: effects of isomerization and electrode coupling

In this work, we have implemented the Fisher-Lee formalism to couple non-equilibrium Green functions with tight-binding Density Functional to tackle large molecular systems. This method is used to determine the decay constant of a set of oligomers based on seven different monomers taken from the literature in the non-resonant tunneling regime. Results show good agreement with experimental measurements. The approach is then applied to explore the conformational pattern effect as well as the asymmetry and the strength of coupling with the electrode of bisthienylbenzene oligomers sandwiched between gold (Au),titanium (Ti)and graphene (G)) electrodes The results indicate that conformational patterns have low impact on the conductance, since the delocalization of π-electrons exhibits similar behavior for all the conformations explored. The calculated attenuation factor is found to be comparable with the strength of contact coupling (Ti > Au = G). Electronic analysis of metal-molecule interactions reveals the ionic nature of Ti-C bonds, through the emergence of a local dipole contributing to the work function variation of 0.35 eV. In addition, the Ti d-orbitals are found to be strongly coupled with the lowest unoccupied orbital (LUMO) of BTB, thus facilitating charge transfer from Ti to the molecule, at the origin of this strong interfacial dipole. However, the Au-C bond is found to be similar to the C-C bond, with pure covalent character. The results confirm the hole transport mechanism observed experimentally in the cases of Au-(BTB)n-Au and Au-(BTB)n-Ti, and predict possible combined mechanism of both hole and electron transport in the case of Ti(BTB)n-Ti.

cond-mat.mes-hall↗

arXiv2021

Better force fields start with better data -- A data set of cation dipeptide interactions

We present a data set from a first-principles study of amino-methylated and acetylated (capped) dipeptides of the 20 proteinogenic amino acids - including alternative possible side chain protonation states and their interactions with selected divalent cations (Ca$^{2+}$, Mg$^{2+}$ and Ba$^{2+}$). The data covers 21,909 stationary points on the respective potential-energy surfaces in a wide relative energy range of up to 4 eV (390 kJ/mol). Relevant properties of interest, like partial charges, were derived for the conformers. The motivation was to provide a solid data basis for force field parameterization and further applications like machine learning or benchmarking. In particular the process of creating all this data on the same first-principles footing, i.e. density-functional theory calculations employing the generalized gradient approximation with a van der Waals correction, makes this data suitable for data-driven force field development. To make the data accessible across domain borders and to machines, we formalized the metadata in an ontology.

q-bio.BM↗

arXiv2021

Sample-Efficient Generation of Novel Photo-acid Generator Molecules using a Deep Generative Model

Photo-acid generators (PAGs) are compounds that release acids ($H^+$ ions) when exposed to light. These compounds are critical components of the photolithography processes that are used in the manufacture of semiconductor logic and memory chips. The exponential increase in the demand for semiconductors has highlighted the need for discovering novel photo-acid generators. While de novo molecule design using deep generative models has been widely employed for drug discovery and material design, its application to the creation of novel photo-acid generators poses several unique challenges, such as lack of property labels. In this paper, we highlight these challenges and propose a generative modeling approach that utilizes conditional generation from a pre-trained deep autoencoder and expert-in-the-loop techniques. The validity of the proposed approach was evaluated with the help of subject matter experts, indicating the promise of such an approach for applications beyond the creation of novel photo-acid generators.

cs.LG↗

arXiv2022

Learning Small Molecule Energies and Interatomic Forces with an Equivariant Transformer on the ANI-1x Dataset

Accurate predictions of interatomic energies and forces are essential for high quality molecular dynamic simulations (MD). Machine learning algorithms can be used to overcome limitations of classical MD by predicting ab initio quality energies and forces. SE(3)-equivariant neural network allow reasoning over spatial relationships and exploiting the rotational and translational symmetries. One such algorithm is the SE(3)-Transformer, which we adapt for the ANI-1x dataset. Our early experimental results indicate through ablation studies that deeper networks - with additional SE(3)-Transformer layers - could reach necessary accuracies to allow effective integration with MD. However, faster implementations of the SE(3)-Transformer will be required, such as the recently published accelerated version by Milesi.

physics.chem-ph↗

arXiv2022

Direct numerical simulations of the modified Poisson-Nernst-Planck equations for the charging dynamics of cylindrical electrolyte-filled pores

Understanding how electrolyte-filled porous electrodes respond to an applied potential is important to many electrochemical technologies. Here, we consider a model supercapacitor of two blocking cylindrical pores on either side of a cylindrical electrolyte reservoir. A stepwise potential difference $2Φ$ between the pores drives ionic fluxes in the setup, which we study through the modified Poisson-Nernst-Planck equations, solved with finite elements. We focus our discussion on the dominant timescales with which the pores charge and how these timescales depend on three dimensionless numbers. Next to the dimensionless applied potential $Φ$, we consider the ratio $R/R_b$ of the pore's resistance $R$ to the bulk reservoir resistance $R_b$ and the ratio $r_{p}/λ$ of the pore radius $r_p$ to the Debye length $λ$. We compare our data to theoretical predictions by Aslyamov and Janssen ($Φ$), Posey and Morozumi ($R/R_b$), and Henrique, Zuk, and Gupta ($r_{p}/λ$). Through our numerical approach, we delineate the validity of these theories and the assumptions on which they were based.

cond-mat.soft↗

arXiv2022

Ground-state properties of metallic solids from ab initio coupled-cluster theory

Metallic solids are a challenging target for wavefunction-based electronic structure theories and have not been studied in great detail by such methods. Here, we use coupled-cluster theory with single and double excitations (CCSD) to study the structure of solid lithium and aluminum using optimized Gaussian basis sets. We calculate the equilibrium lattice constant, bulk modulus, and cohesive energy and compare them to experimental values, finding accuracy comparable to common density functionals. Because the quantum chemical "gold standard" CCSD(T) (CCSD with perturbative triple excitations) is inapplicable to metals in the thermodynamic limit, we test two approximate improvements to CCSD, which are found to improve the predicted cohesive energies.

cond-mat.mtrl-sci↗

arXiv2022

Learning Local Equivariant Representations for Large-Scale Atomistic Dynamics

A simultaneously accurate and computationally efficient parametrization of the energy and atomic forces of molecules and materials is a long-standing goal in the natural sciences. In pursuit of this goal, neural message passing has lead to a paradigm shift by describing many-body correlations of atoms through iteratively passing messages along an atomistic graph. This propagation of information, however, makes parallel computation difficult and limits the length scales that can be studied. Strictly local descriptor-based methods, on the other hand, can scale to large systems but do not currently match the high accuracy observed with message passing approaches. This work introduces Allegro, a strictly local equivariant deep learning interatomic potential that simultaneously exhibits excellent accuracy and scalability of parallel computation. Allegro learns many-body functions of atomic coordinates using a series of tensor products of learned equivariant representations, but without relying on message passing. Allegro obtains improvements over state-of-the-art methods on the QM9 and revised MD-17 data sets. A single tensor product layer is shown to outperform existing deep message passing neural networks and transformers on the QM9 benchmark. Furthermore, Allegro displays remarkable generalization to out-of-distribution data. Molecular dynamics simulations based on Allegro recover structural and kinetic properties of an amorphous phosphate electrolyte in excellent agreement with first principles calculations. Finally, we demonstrate the parallel scaling of Allegro with a dynamics simulation of 100 million atoms.

physics.comp-ph↗

arXiv2022

Ab Initio Prediction of Excited State and Polaron Effects in Transient XUV Measurements of $α$-Fe$_2$O$_3$

Transient X-ray and extreme ultraviolet (XUV) spectroscopies have become invaluable tools for studying photoexcited dynamics due to their sensitivity to carrier occupations and local chemical or structural changes. One of the most studied mate-rials using transient XUV spectroscopy is $α$-Fe$_2$O$_3$ because of its rich photoexcited dynamics, including small polaron formation. The interpretation of carrier and polaron effects in $α$-Fe$_2$O$_3$ is currently done using a semi-empirical method that is not transferrable to most materials. Here, an ab initio, Bethe-Salpeter equation (BSE) approach is developed that can incorporate photoexcited state effects for arbitrary materials systems. The accuracy of this approach is proven by calculating the XUV absorption spectra for the ground, photoexcited, and polaron states of $α$-Fe$_2$O$_3$. Furthermore, the theoretical approach allows for the projection of the core-valence excitons and different components of the X-ray transition Hamiltonian onto the band structure, providing new insights into old measurements. From this information, a physical intuition about the origins and nature of the transient XUV spectra can be built. A route to extracting electron and hole energies is even shown possible for highly angular momentum split XUV peaks. This method is easily generalized to K, L, M, and N edges to provide a general approach for analyzing transient X-ray absorption or reflection data.

cond-mat.mtrl-sci↗

arXiv2024

Enhanced Shadow Tomography of Molecular Excited States from Enforcing $N$-representability Conditions by Semidefinite Programming

Excited-state properties of highly correlated systems are key to understanding photosynthesis, luminescence, and the development of novel optical materials, but accurately capturing their interactions is computationally costly. We present an algorithm that combines classical shadow tomography with physical constraints on the two-electron reduced density matrix (2-RDM) to treat excited states. The method reduces the number of measurements of the many-electron 2-RDM on quantum computers by (i) approximating the quantum state through a random sampling technique called shadow tomography and (ii) ensuring that the 2-RDM represents an $N$-electron system through imposing $N$-representability constraints by semidefinite programming. This generalizes recent work on the $N$-representability-enhanced shadow tomography of ground-state 2-RDMs. We compute excited-state energies and 2-RDMs of the H$_4$ chain and analyze the critical points along the photoexcited reaction pathway from gauche-1,3-butadiene to bicyclobutane via a conical intersection. The results show that the generalized shadow tomography retains critical multireference correlation effects while significantly reducing the number of required measurements, offering a promising avenue for the efficient treatment of electronically excited states on quantum devices.

quant-ph↗

arXiv2024

Shadow Ansatz for the Many-Fermion Wave Function in Scalable Molecular Simulations on Quantum Computing Devices

Here we show that shadow tomography can generate an efficient and exact ansatz for the many-fermion wave function on quantum devices. We derive the shadow ansatz -- a product of transformations applied to the mean-field wave function -- by exploiting a critical link between measurement and preparation. Each transformation is obtained by measuring a classical shadow of the residual of the contracted Schrödinger equation (CSE), the many-electron Schrödinger equation (SE) projected onto the space of two electrons. We show that the classical shadows of the CSE vanish if and only if the wave function satisfies the SE and, hence, that randomly sampling only the two-electron space yields an exact ansatz regardless of the total number of electrons. We demonstrate the ansatz's advantages for scalable simulations -- fewer measurements and shallower circuits -- by computing H$_{3}$ on simulators and a quantum device.

quant-ph↗

arXiv2025

Chiral Vibrational Modes in Small Molecules

The development of quantitative methods for characterizing molecular chirality can provide an important tool for studying chirality induced phenomena in molecular systems. Significant progress has been made in recent years toward understanding the chirality of molecular normal vibrational modes, mostly focusing on vibrations of helical molecular structures. In the present study, we examine the applicability two methodologies previously used for helical structures for the quantification of the chirality of molecular normal modes across a range of small, not necessarily helical, molecules. The first approach involves the application of the Continuous Chirality Measure (CCM) to each normal mode by associating the mode with a structure formed by imposing the corresponding motion about a common origin. The second approach assigns to each normal mode a pseudoscalar defined as the product of atomic linear and angular momentum summed over all atoms. In particular, using the CCM also as a measure of the chirality of the underlying molecular structure, we establish the existence of correlation between the chirality of molecular normal modes and that of the underlying molecular structure. Furthermore, we find that normal modes associated with different frequency ranges of the molecular vibrational spectrum exhibit distinct handedness behavior.

physics.chem-ph↗

arXiv2026

Benchmarking Proton Tunneling Splittings with a Wavefunction-Based Double-Well Model: Application to the Formic Acid Dimer

Proton tunneling across hydrogen bonds is a fundamental quantum effect with implications for spectroscopy, catalysis, and biomolecular stability. While state-of-the-art instanton and path-integral methods provide accurate multidimensional tunneling splittings, simplified one-dimensional models remain valuable as conceptual and benchmarking tools. Here we develop a wavefunction-based framework for tunneling splittings using a Cornell-type double-well potential and apply it as a benchmark for hydrogen-bond tunneling. Analytical WKB estimates and numerical finite-difference solutions are compared across a range of barrier parameters, showing consistent agreement. As a test case, we map the formic acid dimer (FAD) barrier onto a quartic double-well model parameterized to reproduce the reported barrier height of $V_b \\approx 2848~\\text{cm}^{-1}$. The resulting tunneling splitting of about $0.037~\\text{cm}^{-1}$ matches the reduced-dimensional calculations of Qu and Bowman. The close agreement between numerical and semiclassical results highlights the pedagogical and diagnostic value of one-dimensional models, while comparison with molecular benchmarks clarifies their limitations relative to full multidimensional quantum treatments.

physics.chem-ph↗

arXiv2026

A Physics-Regularized Neural Network and Kirchhoff Markov Random Field Framework for Inferring Internal Electrochemical States from Operando Spectromicroscopy

Quantitative understanding of coupled reaction-transport dynamics in lithium-ion battery (LIB) composite electrodes is limited by the inaccessibility of key internal electrochemical states. Here we present a physics-integrated, data-driven framework to reconstruct latent states from operando microscopic X-ray absorption fine structure ($μ$ -XAFS) hyperspectral data of LIB cathodes. Normalized-local-lithium-stoichiometry (NLLS) maps derived from Co K-edge spectra are refined using a physics-regularized neural network that enforces spatial continuity and current conservation to resolve ambiguities in the weak-sensitivity region. The reconstructed NLLS dynamics are embedded in a Kirchhoff-based Markov random field incorporating Kirchhoff's laws, Ohm's law, and Butler-Volmer kinetics to infer interfacial current density, ionic current, electrolyte potential, and effective ionic conductivity. Application to electrodes with differential initial electrolyte concentrations reveals distinct reaction-propagation modes that are consistent with a mechanism involving the non-monotonic concentration dependence of electrolyte conductivity. The inferred electrolyte concentration profiles qualitatively resemble the spatially extended concentration changes observed by independent operando X-ray transmission imaging using a 1 M LiAsF$_6$ electrolyte.

physics.chem-ph↗

arXiv2026

SPIBER: Reconstructing Free Energy Landscapes from Short, Unconverged Trajectories with Generative Flow Networks

Molecular systems have many degrees of freedom, but their metastable behavior can often be described by a few collective variables. Identifying these variables and estimating free energies along them from limited simulation data remains a challenging, important problem. Separate short trajectories may sample different metastable states without capturing transitions or establishing their relative equilibrium populations. For unbiased trajectories generated with the same Hamiltonian at a single temperature, alternate methods based on histogram reweighting cannot correct this imbalance. Here we present SPIBER, which combines the State Predictive Information Bottleneck (SPIB) with Generative Flow Networks (GFlowNets). SPIB uses deep learning to approximate slow degrees of freedom through a past-future information bottleneck, retaining information needed to predict future metastable states. We show that this compression limits conditional entropy variations in populated regions, allowing conditional mean potential energies, which are much easier to calculate, to be used to approximate free energy differences. Given sufficient local sampling to estimate these energies, they define the target distribution for GFlowNets, energy-based generative samplers that sample according to estimated thermodynamic stability rather than observed populations. For a particle in a radial double-well potential, for alanine dipeptide, and for the nine-residue peptide AIB9, SPIBER recovers free energy differences between sampled metastable states to within one thermal energy unit of reference values. The method combines collective-variable learning and free energy estimation in up to four latent dimensions, without requiring converged state populations or additional molecular dynamics simulations.

physics.chem-ph↗