Hydrolysis-argyrodite

Deciphering Surface Hydrolysis Mechanism in Argyrodite via Large-Scale Machine Learning Potential SimulationsThis dataset contains all simulation and validation files associated with the paper <i>“Deciphering Surface Hydrolysis Mechanism in Argyrodite via Large-Scale Machine Learning Potential Simulations.”</i> The repository is organized into two main directories: <b>MD_results</b> and <b>validation</b>.The <b>MD_results</b> directory includes the molecular dynamics (MD) simulation data for three different environments:LPSC–H2OLPSC–H3O+OH+H2OLPSnSC–H3O+OH+H2OEach subdirectory contains the final MD structures and the fine-tuned universal machine learning potential (uMLP) files used for the simulations. The uMLP files for the LPSC–H2O and LPSC–H3O+OH+H2O systems are identical.The <b>validation</b> directory includes data and scripts used to verify the accuracy of the fine-tuned machine learning potentials. Two systems are included:LPSC_H–OLPSnSC_H–OEach validation directory contains:<br>• checkpoint_best.pth: fine-tuned uMLP checkpoint file<br>• combined.xyz: structures used to compute single-point DFT energies<br>• correlation.py: Python script for generating validation plots<br>• fine-tuned.png: validation plot image<br>• MD_fidelity_DFTvsMLP.csv: comparison of MLP vs. DFT energies for MD-sampled structures, used to assess MD fidelity