arXiv subjects

Garvit Agarwal

Publications and source records attributed to Garvit Agarwal.

5 recordsLinked to original sources

arXiv2026

Adsorption energies and decomposition barrier heights for ethylene carbonate on the surface of lithium from cluster-based quantum chemistry

For ethylene carbonate on the (100) surface of lithium, we calculate the adsorption energy in two binding motifs as well as the barrier height for a ring-opening decomposition reaction. We validate a scheme for producing results in the thermodynamic limit by correcting results obtained on finite lithium clusters containing only 40-100 atoms, which enables the use of hybrid density functionals, the random-phase approximation, and correlated wavefunction theories such as coupled-cluster theory and auxiliary-field quantum Monte Carlo. We find that the high-level theories agree to within 2-5 kcal/mol and can therefore serve as benchmarks for more affordable methods. Using our reference data, we demonstrate that generalized gradient approximation functionals, such as PBE, are not sufficiently accurate for reaction barrier heights, and we identify $\omega$B97X-V as an especially promising functional for the interfacial chemistry of electrolyte solvents at lithium metal anodes.

cond-mat.mtrl-sci↗

arXiv2025

Reaction dynamics of lithium-mediated electrolyte decomposition using machine learning potentials

We study the ring-opening decomposition of ethylene carbonate in the presence of a single lithium atom and on the surface of lithium metal. Combining accurate electronic structure theory, enhanced sampling, and machine learning, we fine-tune the MACE-MP0 foundation model and apply the resulting machine learning potentials to obtain statistically converged free energy profiles and reaction rates. We confirm that the level of electronic structure theory is important, and inaccurate density functionals can overestimate the reaction rate by up to nine orders of magnitude. We also find that harmonic transition state theory underestimates reaction rates by about one order of magnitude. For the surface reaction, we find and characterize a new, ultrafast decomposition pathway wherein the carbonyl is deeply inserted into the lithium surface and bent by about 70$^\circ$. This reaction, which occurs in a few tens of picoseconds, generates a ring-opened intermediate that is a precursor for CO or CO$_2$ formation; by contrast, an alternative pathway that yields CO$_3^{2-}$ and ethylene is found to be non-competitive, occurring on a timescale of tens of nanoseconds.

physics.chem-ph↗

arXiv2025

Efficient Long-Range Machine Learning Force Fields for Liquid and Materials Properties

Machine learning force fields (MLFFs) have emerged as a sophisticated tool for cost-efficient atomistic simulations approaching DFT accuracy, with recent message passing MLFFs able to cover the entire periodic table. We present an invariant message passing MLFF architecture (MPNICE) which iteratively predicts atomic partial charges, including long-range interactions, enabling the prediction of charge-dependent properties while achieving 5-20x faster inference versus models with comparable accuracy. We train direct and delta-learned MPNICE models for organic systems, and benchmark against experimental properties of liquid and solid systems. We also benchmark the energetics of finite systems, contributing a new set of torsion scans with charged species and a new set of DLPNO-CCSD(T) references for the TorsionNet500 benchmark. We additionally train and benchmark MPNICE models for bulk inorganic crystals, focusing on structural ranking and mechanical properties. Finally, we explore multi-task models for both inorganic and organic systems, which exhibit slightly decreased performance on domain-specific tasks but surprising generalization, stably predicting the gas phase structure of $\simeq500$ Pt/Ir organometallic complexes despite never training to organometallic complexes of any kind.

physics.chem-ph↗

arXiv2024

Robust and effective ab initio molecular dynamics simulations on the GPU cloud infrastructure using the Schr\"odinger Materials Science Suite

Ab initio Born-Oppenheimer molecular dynamics (AIMD) is a valuable method for simulating physico-chemical processes of complex systems, including reactive systems, and for training machine learning models and force fields. Speed and stability issues on traditional hardware preclude routine AIMD simulations for larger systems and longer timescales. We postulate that any practically useful AIMD simulation must generate a trajectory of a minimum 1000 MD steps a day on a moderate cloud resource. In this work, we implement a computing workflow that enables routine calculations at this throughput and demonstrate results for several non-trivial atomistic dynamical systems. In particular, we have employed the GPU implementation of the Quantum ESPRESSO code which we will show increases AIMD productivity compared to the CPU version. In order to take advantage of transient servers (which are more cost and energy effective compared to the stable servers), we have implemented automatic restart/continuation of the AIMD runs within the Schr\"odinger Materials Science Suite. Finally, to reduce simulation size and thus reduce compute time when modeling surfaces, we have implemented a wall potential constraint. Our benchmarks using several reactive systems (lithium anode surface/solvent interface, hydrogen diffusion in an iron grain boundary) show a significant speed up when running on a GPU-enabled transient server using our updated implementation.

cond-mat.mtrl-sci↗

arXiv2024

On the Impact and Utility of Single-Exomoon Modeling for Multi-Moon Systems

The search for exomoons in time-domain photometric data has to-date generally consisted of fitting transit models that are comprised of a planet hosting a single moon. This simple model has its advantages, but it may not be particularly representative, as most of the major moons in our Solar System are found in multi-moon satellite systems. It is critical that we investigate, then, the impact of applying a single-moon model to systems containing multiple moons, as there is the possibility that utilizing an inaccurate or incomplete model could lead to erroneous conclusions about the system. To that end, in this work we produce a variety of realistic multi-moon light curves, perform standard single-moon model selection, and analyze the impacts that this model choice may have on the search for exomoons. We find that the number of moons in a system fit with a single-moon model generally has little impact on whether we find evidence for a moon in that system, and other system attributes are individually not especially predictive. However, the model parameter solutions for the moon frequently do not match any real moon in the system, instead painting a picture of a ``phantom'' moon. We find no evidence that multi-moon systems yield corresponding multi-modal posteriors. We also find a systematic tendency to overestimate planetary impact parameter and eccentricity, to derive unphysical moon densities, and to infer potentially unphysical limb darkening coefficients. These results will be important to keep in mind in future exomoon search programs.

astro-ph.EP↗