AI and the World's Hard Problems · Chemistry & materials · arXiv · 2026-10-23
Concentration-Dependent Speciation in ZnCl2 Electrolyte Solutions: Transferability of Machine-Learning Interatomic Potentials and Speciation Sensitivity of EXAFS
Predicting speciation in concentrated electrolytes requires accurate ion interactions and sampling of slowly interconverting coordination environments. Zn K-edge EXAFS measurements of aqueous ZnCl2 were combined with long molecular dynamics runs using several machine-learning interatomic potentials, including DeepMD, DPMD-SCAN and MACE-MATPES-r2SCAN. Increasing concentration reduced octahedral coordination, replaced water with chloride, and promoted chloride-bridged association. DPMD-SCAN was the most transferable across concentrations, and MACE-MATPES-r2SCAN gave comparable spectra more efficiently. Distance disorder and scattering-path interference can produce similar EXAFS signatures for different coordination ensembles, so ensemble-averaged spectra are needed.