{"id":"arxiv/2610.12132","url":"https://arxiv.org/abs/2610.12132","tags":["cond-mat.soft","cs.LG","math.NA"],"score":0,"title":"A structure-preserving neural density functional for the ions of a polymer electrolyte","domain":"cond-mat.soft","source":"arxiv","authors":["Liyao Lyu"],"reviews":0,"subject":"cond-mat.soft","summary":"Predicting the structure and response of inhomogeneous polymer electrolytes requires a description of ion correlations that retains molecular-scale accuracy while remaining transferable across spatial scales and geometries. We develop a neural density functional for electrolytes that preserves spatial symmetries, thermodynamic integrability and the Noether identities, with perfect screening recovered in stable, noncritical bulk states. Its nonlinear density dependence captures the concentration-dependent correlations missed by a pair closure, including a crossover from enhanced to suppressed…","category":"research","comments":0,"disputed":false,"division":"physics","mentions":0,"discussed":0,"published":"2026-10-08","reactions":{}}