Searcharxiv⌕ Search

arXiv subjects

Kyle Reeves

Publications and source records attributed to Kyle Reeves.

3 recordsLinked to original sources

What Do LLMs Use to Judge Synthesis Recipes? An Experimental Analysis on 240 Solid-State Syntheses

Inorganic materials discovery is increasingly limited by synthesis rather than by material design. Large language models (LLMs) are being adopted to propose synthesis conditions with little experimental evidence of their capabilities and biases. Here, we carry out an experimental synthesis benchmark of 240 experiments across eight known and unknown inorganic oxides and subsequently evaluate an LLM as a zero-shot ranker over fixed banks of solid-state synthesis recipes. For each pair of recipes for the same oxide target, `GPT-5.6-Sol` selects the recipe expected to yield higher target-phase purity. We aggregate these pairwise preferences using a regularized Bradley-Terry model to obtain a ranked list of synthesis recipes for a given target material. By comparison with the actual purity, obtained through X-ray diffraction, we measure `GPT-5.6-Sol`'s performance as a zero-shot ranker (0.72 pairwise accuracy). By selective removal of the information exposed to the model during this ranking, we conclude that ranking is driven predominantly by thermal features, with reaction chemistry modulating the effect. Ranking accuracy tracks the purity-temperature correlation by target (Spearman ρ= 0.91\), \(p = 0.002\), and on the six thermally-driven targets withholding all chemical identity hardly impacts performance (0.68) compared to withholding temperature (0.56). Ranking performance was higher on known than unknown targets, but the group-level difference was not statistically conclusive. The methodological choices highlight the need for more experimentally validated synthesis benchmarks and position LLMs as a thermally informed prior.

cs.CE↗

Multi-scale simulation of the adsorption of lithium ion on graphite surface: from Quantum Monte Carlo to Molecular Density Functional Theory

The structure of the double-layer formed at the surface of carbon electrodes is governed by the interactions between the electrode and the electrolyte species. However, carbon is notoriously difficult to simulate accurately, even with well-established methods such as electronic Density Functional Theory and Molecular Dynamics. Here we focus on the important case of a lithium ion in contact with the surface of graphite, and we perform a series of reference Quantum Monte Carlo calculations that allow us to benchmark various electronic Density Functional Theory functionals. We then fit an accurate carbon--lithium pair potential, which is used in molecular Density Functional Theory calculations to determine the free energy of the adsorption of the ion on the surface in the presence of water. The adsorption profile in solution differs markedly from the gas phase results, which emphasize the role of the solvent on the properties of the double-layer.

cond-mat.mtrl-sci↗

A semiclassical Thomas-Fermi model to tune the metallicity of electrodes in molecular simulations

Spurred by the increasing needs in electrochemical energy storage devices, the electrode/electrolyte interface has received a lot of interest in recent years. Molecular dynamics simulations play a proeminent role in this field since they provide a microscopic picture of the mechanisms involved. The current state-of-the-art consists in treating the electrode as a perfect conductor, precluding the possibility to analyze the effect of its metallicity on the interfacial properties. Here we show that the Thomas-Fermi model provides a very convenient framework to account for the screening of the electric field at the interface and differenciating good metals such as gold from imperfect conductors such as graphite. All the interfacial properties are modified by screening within the metal: the capacitance decreases significantly and both the structure and dynamics of the adsorbed electrolyte are affected. The proposed model opens the door for quantitative predictions of the capacitive properties of materials for energy storage.

cond-mat.mtrl-sci↗