Distilling Large-Scale Machine Learning Interatomic Potentials to Predict Long-Timescale Electrolyte Transport Properties

Ishan Amin; Yue Jian; Nitesh Kumar; Christine Zhang; Jianwei Lai; Casey S. Mezerkor; Aaron M. Melemed; Jonathan R. Thurston; Neil P. Dasgupta; J. David Bazak; Ethan J. Crumlin; Samuel Blau; Aditi S. Krishnapriyan

arXiv · 2610.39513

Distilling Large-Scale Machine Learning Interatomic Potentials to Predict Long-Timescale Electrolyte Transport Properties

Ishan Amin·Yue Jian·Nitesh Kumar·Christine Zhang·Jianwei Lai·Casey S. Mezerkor·Aaron M. Melemed·Jonathan R. Thurston·Neil P. Dasgupta·J. David Bazak·Ethan J. Crumlin·Samuel Blau·Aditi S. Krishnapriyan

View original source ↗Read source document ↗Export metadata ↓

Abstract

Accurately predicting the transport properties of liquid electrolytes is critical for next-generation battery design and requires force fields that are both quantum-mechanically accurate and fast enough for long-timescale simulation. Recent large-scale pretrained machine learning interatomic potentials (MLIPs) approach density-functional-theory accuracy, but remain too slow to access the necessary timescales to converge transport properties. We provide a knowledge distillation procedure to transfer the accuracy of a large-scale "teacher" MLIP into fast, system-specific "students", without any additional ab initio data. Our distilled models run ~40x faster, reproduce teacher-predicted structural observables, and produce stable 20 ns trajectories in days, rather than the months the teacher would require. Across a diverse set of sodium- and lithium-ion electrolytes, student-predicted self-diffusivities, ionic conductivities, and shear viscosities track experiment and outperform classical force-field predictions, establishing distillation as a practical route to accurate, long-timescale simulations necessary for high-throughput electrolyte screening.

Explore related subjects

physics.chem-ph cond-mat.mtrl-sci cond-mat.soft cs.LG physics.comp-ph

Keep this discovery

Explore connections, maps & timelines

BibTeX RIS

Ishan Amin, Yue Jian, Nitesh Kumar, Christine Zhang, Jianwei Lai, Casey S. Mezerkor, Aaron M. Melemed, Jonathan R. Thurston, Neil P. Dasgupta, J. David Bazak, Ethan J. Crumlin, Samuel Blau, Aditi S. Krishnapriyan. 2026-10-29. Distilling Large-Scale Machine Learning Interatomic Potentials to Predict Long-Timescale Electrolyte Transport Properties. https://arxiv.org/abs/2610.39513

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

Unravelling carbene moiety role in the N-heterocyclic carbene-phosphinidenes coordination chemistry: bonding and reactivity in gold chemistry

The coordination properties of the emerging class of N-heterocyclic carbene-phosphinidene (NHCP) ligands have been reported to be strongly affected by the carbene (NHC) moiety, but a clear understanding remains strikingly limited. In this work the bonding features and reactivity of 13 NHCP gold hydride complexes, [NHCPAuH], bearing different classes of NHCs have been systematically explored and compared to carbene analogues, using a consistent computational protocol. The analyses reveal that, although the $π$-acceptor ability trend of NHCPs qualitatively parallels that of NHCs, the $σ$-donor ability trend reverses, with more $σ$-donating NHC moieties generating less $σ$-donating NHCP ligands. Concurrently, the nature of the NHC moiety impacts the coordination geometry at the P center, with the CNHC-P-Au bond angle appearing as a structural descriptor to monitor and/or design NHCP ligands with desired $σ$-donor properties. The NHCPs stronger $σ$-donor and weaker $π$-acceptor abilities compared to NHCs and their unique structural flexibility and electronic adaptability have been showcased to be directly controlled by the carbene moiety via modulation of the HOMO lone pair energy and its atomic phosphorous 3p character. The reactivity of [NHCPAuH] complexes with CO2, taken as a probe for potential NHCP applications in small molecule activation processes, demonstrates qualitatively similar mechanisms, with remarkably different activation barriers, which directly correlate with the NHCPs $σ$-donor ability, thus reflecting the high tunability of their bonding properties. This work provides insights and perspectives on the design principle of NHCP ligands, with a spotlight on the pivotal role of NHC moiety in modulating their electronic and steric properties, offering opportunities for burgeoning applications across diverse fields.

physics.chem-ph↗

arXiv2026

Third-Order Algebraic Diagrammatic Construction Renormalized by Particle-Particle Ladders

The algebraic diagrammatic construction (ADC) of the electronic self-energy is based on Møller--Plesset perturbation theory. Its vertices contain bare energy denominators that do not capture renormalization of Hartree--Fock excitations through excited-state correlations. As we show here, replacing the Møller--Plesset amplitudes by amplitudes obtained from Epstein--Nesbet (EN) partitioning of the electronic Hamiltonian overcomes this deficiency at no extra cost by introducing dressed denominators that resum to infinite order the diagonal elements of ladder diagrams described by the pair propagator in the particle-particle $T$-matrix approximation. Third-order ADC with EN-dressed denominators leads to an almost fourfold increase in accuracy compared to standard ADC(3) on a challenging benchmark set for ionization potentials of closed-shell molecules, reaching an accuracy between that of equation-of-motion coupled cluster with singles and doubles (EOM-CCSD) and with singles, doubles, and triples. For open-shell systems, EN-ADC(3) halves the error of ADC(3) for weakly spin-contaminated molecules, outperforming EOM-CCSD, and still improves on ADC(3) for strongly spin-contaminated ones, for which EOM-CCSD remains more accurate.

physics.chem-ph↗

arXiv2026

A Geometry-Induced Finite Warburg element for Reversible Diffusion Impedance at Microdisk Electrodes

A compact analytical representation is developed for the diffusion impedance of a disk electrode under reversible conditions, obtained by transforming the chronoamperometric relaxation function into the frequency domain. A Padé approximant is constructed for the uniform-concentration boundary condition, preserving the finite low-frequency diffusion resistance of the steady-state disk current and the high-frequency Cottrell regime. It is extended to reactions in which both oxidized and reduced species contribute, with unequal diffusion coefficients and surface concentrations. Its accuracy is quantified against converged numerical solutions, including an exact solution of the coupled problem with finite-rate kinetics. The approximant reproduces the Saito resistance and the $45^{\circ}$ Warburg phase exactly; its modulus is within 0.6\% of the converged solution for $ωa^2/D\le12$ and 7\% below the Cottrell amplitude at high frequency. The uniform-concentration and uniform-flux boundary conditions both give a diffusion impedance of geometric origin, but differ by up to 11\% in modulus because each couples differently to the spreading diffusion field; a criterion based on the rate constant and the disk radius delineates where each applies. The exact solution shows that this crossover is broad and that the decomposition into a charge-transfer resistance and a diffusion impedance is accurate to about 5\% for $1\le2k^0a/D\le100$. Fitted to a published spectrum of a platinum microdisk, it reproduces the reported diffusion coefficient within 1\%, while the two limiting elements lie 4.8\% below and 4.1\% above it. The element is a practical tool for microdisk impedance analysis.

physics.chem-ph↗