figshare
Browse

<b>Deciphering Li₂S-like SEI Formation Mechanism at Li/Argyrodite Anodic Interfaces via Large-Scale Machine-Learned Reactive Dynamics</b>

dataset
posted on 2025-10-03, 13:54 authored by Ji Seon KimJi Seon Kim, Sang Uck Lee
<p dir="ltr"><b>Files are composed of </b><b>the following</b><b> directories.</b></p><ul><li>Structure</li><li>Validation</li></ul><h2><b>Structure</b></h2><p dir="ltr">This directory contains the structural data used for molecular dynamics simulations and the construction of training sets.</p><ul><li><b>MD_structure/</b></li><li><ul><li><b>initial_structure/</b>: CIF files of the starting models used as the initial configurations for MD simulations.</li><li><b>MD_structure/</b>: Representative CIF snapshots obtained from MD trajectories at selected time steps. These structures illustrate the evolution of the Li|argyrodite interface during the simulations.<br></li></ul></li><li><b>Trainingset/</b><br>This folder contains the <b>initial structures that were used before running ab initio molecular dynamics (AIMD)</b> to build the MTP training set. These structures serve as the starting point for generating AIMD data.</li></ul><p><br></p><ul><li><ul><li><b>bulk/</b>: Pristine bulk phases (Li, Li₆PS₅Cl).</li><li><b>slab/</b>: Surface slab models constructed from bulk structures.</li><li><b>interface/</b>: Li/argyrodite interface models that were used as the initial configurations for the active learning procedure.<br><br>In other words, the CIF files here are not trajectories or relaxed structures but the reference initial configurations from which AIMD simulations and MTP training data were generated.</li></ul></li></ul><h2><b>Validation</b></h2><p dir="ltr">This directory contains validation data for the trained machine-learning interatomic potentials (MTP).</p><ul><li><b>Bulk&slab/</b>: <br>Contains validation <b>cfg files</b> for Li metal and Li₆PS₅Cl (argyrodite) bulk and slab systems.</li></ul><p><br></p><ul><li><ul><li><b>Li.cfg, Arg.cfg</b>: Reference cfg files from <b>DFT calculations</b>, including atomic positions, energies, and forces.</li><li><b>Li_mtp_cfg, Arg_mtp.cfg</b>: Corresponding cfg files generated using the <b>trained MTP potential</b>, directly comparable to the DFT reference data.</li></ul></li></ul><p><br></p><ul><li><b>Interface/</b>: Contains validation <b>cfg files</b> for Li | Li₆PS₅Cl interface models.</li><li><ul><li><b>mtp/</b>: Contains cfg files generated by running the trained MTP on snapshot structures extracted from MD trajectories. Each cfg file includes atomic positions, energies, and forces, with a CSV file summarizing the predicted energies.</li><li><b>snapshot_cifs/</b>: Contains 50 CIF files sampled every 20 ps from a 1 ns MD trajectory of the Li | Li₆PS₅Cl interface. DFT single-point calculations were performed for these snapshots, and the resulting energies are provided in a CSV file.</li></ul></li></ul><ul><li><b>pot.mtp</b></li><li><ul><li>The <b>trained MTP potential file</b> required to run MLIP-based MD simulations.</li><li>This file was used in all validation tasks as well as in the production MD runs described in the manuscript.</li></ul></li></ul><p><br></p><p><br></p><p><b>CIF files</b> can be directly used as input for VASP or LAMMPS simulations.</p><p dir="ltr"><b>MTP files</b> and <b>cfg datasets</b> allow re-training or reproduction of the MLIP calculations.</p><p dir="ltr">This dataset provides the key inputs and reference results necessary to reproduce the structural models, MD simulations, and validation tests described in the manuscript.</p>

History

Usage metrics

    Licence

    Exports

    RefWorks
    BibTeX
    Ref. manager
    Endnote
    DataCite
    NLM
    DC