Concentration-Dependent Speciation in ZnCl2 Electrolyte Solutions: Transferability of Machine-Learning Interatomic Potentials and Speciation Sensitivity of EXAFS
Abstract
Predicting speciation in concentrated electrolytes requires accurate ion interactions and adequate sampling of slowly interconverting coordination environments. We combine Zn K-edge extended X-ray absorption fine structure (EXAFS) measurements with long-timescale molecular dynamics (MD) using machine-learning interatomic potentials to study aqueous ZnCl2 from dilute conditions to 30 m. We train revPBE-based DeepMD models with D3 corrections included or excluded for Zn-containing pairs and compare them with DPMD-SCAN, classical MD (CMD), and MACE-MATPES-r2SCAN. EXAFS calculations are benchmarked against hydrated and chloride-coordinated Zn reference complexes, then applied to MD ensembles alongside Zn-Cl coordination free-energy analysis. Increasing concentration reduces octahedral coordination, replaces water with chloride, and promotes chloride-bridged association. All three DeepMD models yield similar EXAFS at 3 m, but DPMD-SCAN shows greater transferability across concentrations, reproducing measured spectral trends, key features of mixed ZnCl2/LiCl spectra, and octahedral fractions consistent with other studies. DFT treatment strongly affects coordination free energies and aggregation; excluding Zn-related D3 corrections does not systematically improve EXAFS agreement. CMD shows larger spectral discrepancies and less consistent coordination trends. MACE-MATPES-r2SCAN improves upon CMD and yields spectra comparable to DPMD-SCAN, which provides more consistent agreement and greater computational efficiency. Neighbor-resolved analysis shows that distance disorder and scattering-path interference can produce similar EXAFS signatures for distinct coordination ensembles, highlighting the need for rigorous sampling and ensemble-averaged spectra to distinguish isolated from chloride-bridged Zn species.
Related
- Distilling Large-Scale Machine Learning Interatomic Potentials to Predict Long-Timescale Electrolyte Transport Properties
- Reconstruction of Oxygen Evolution Precatalysts: Effects of Leaching, Electrolyte Concentration and Iron Incorporation
- Cross-exposure transferability and failure boundaries of self-supervised DNN denoising for AR-HAXPES depth profiling
- A predictive transferability index for crystal plasticity framework selection: Variational derivation, analytical bounds, and FFT-based plausibility assessment
- Sn Speciation and Spectroscopic Signatures of Sn(II) Sites in Alumina-Supported PtSn Propane Dehydrogenation Catalysts
BibTeX
@misc{chen2026concentrationdependent,
title = {{Concentration-Dependent Speciation in ZnCl2 Electrolyte Solutions: Transferability of Machine-Learning Interatomic Potentials and Speciation Sensitivity of EXAFS}},
author = {Junhan Chen and Luke D. Gibson and Nadia N. Intan and Darren M Driscoll and Mahalingam Balasubramanian and John L. Fulton and Gregory K. Schenter and Christopher J. Mundy and Mohammadhasan Dinpajooh},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.30513},
eprint = {2610.30513},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.30513},
note = {CC BY-NC-ND 4.0},
}
RIS
TY - UNPB
TI - Concentration-Dependent Speciation in ZnCl2 Electrolyte Solutions: Transferability of Machine-Learning Interatomic Potentials and Speciation Sensitivity of EXAFS
AU - Junhan Chen
AU - Luke D. Gibson
AU - Nadia N. Intan
AU - Darren M Driscoll
AU - Mahalingam Balasubramanian
AU - John L. Fulton
AU - Gregory K. Schenter
AU - Christopher J. Mundy
AU - Mohammadhasan Dinpajooh
PY - 2026
DA - 2026/10/21
PB - arXiv
DO - 10.48550/arxiv.2610.30513
UR - https://arxiv.org/abs/2610.30513
AB - Predicting speciation in concentrated electrolytes requires accurate ion interactions and adequate sampling of slowly interconverting coordination environments. We combine Zn K-edge extended X-ray absorption fine structure (EXAFS) measurements with long-timescale molecular dynamics (MD) using machine-learning interatomic potentials to study aqueous ZnCl2 from dilute conditions to 30 m. We train revPBE-based DeepMD models with D3 corrections included or excluded for Zn-containing pairs and compare them with DPMD-SCAN, classical MD (CMD), and MACE-MATPES-r2SCAN. EXAFS calculations are benchmarked against hydrated and chloride-coordinated Zn reference complexes, then applied to MD ensembles alongside Zn-Cl coordination free-energy analysis. Increasing concentration reduces octahedral coordination, replaces water with chloride, and promotes chloride-bridged association. All three DeepMD models yield similar EXAFS at 3 m, but DPMD-SCAN shows greater transferability across concentrations, reproducing measured spectral trends, key features of mixed ZnCl2/LiCl spectra, and octahedral fractions consistent with other studies. DFT treatment strongly affects coordination free energies and aggregation; excluding Zn-related D3 corrections does not systematically improve EXAFS agreement. CMD shows larger spectral discrepancies and less consistent coordination trends. MACE-MATPES-r2SCAN improves upon CMD and yields spectra comparable to DPMD-SCAN, which provides more consistent agreement and greater computational efficiency. Neighbor-resolved analysis shows that distance disorder and scattering-path interference can produce similar EXAFS signatures for distinct coordination ensembles, highlighting the need for rigorous sampling and ensemble-averaged spectra to distinguish isolated from chloride-bridged Zn species.
ER -