Nanosecond MD of battery cathode materials with electron density description

<p>Data Description for the Paper "Nanosecond Molecular Dynamics of Battery Cathode Materials with Electron Density Description"</p><p>The data associated with this paper is provided in a compressed .zip folder identified as Data_MD_trajectories.zip. Within this archive, each trajectory is labeled with temperature annotations: T=600, T=500, and T=400. Additionally, each trajectory includes an index that corresponds to various initial structures and Na-ion concentrations.</p><p>The Python scripts for the PaINN model are included in the "graphNNP" zip.</p><p>The Python scripts for the charge prediction model are included in the "Charge_Density_Model_scripts.zip"&nbsp;</p><p>Four trained force and energy prediction models are available, identified as follows:</p><p>1. best_model_Node148Int4Force99.pth</p><p>2. best_model_Node148Int3Force99.pth</p><p>3. best_model_Node128Int4Force99.pth</p><p>4. best_model_Node128Int3Force99.pth</p><p>One trained charge model is available and identified as follows:</p><ol><li><a href="https://zenodo.org/api/records/10051133/draft/files/best_model_Charge_predict.pth/content">best_model_Charge_predict.pth</a></li></ol><p>The training datasets utilized for the development of the force and energy prediction models are denoted as follows:</p><p>1. MDDatabase_1.db: The original training dataset derived from Density Functional Theory (DFT) data, used to initially train the 4 energy and force prediction models</p><p>2. MDDatabase_2: Dataset obtained through the active learning framework detailed in the paper on a small unit size cell (-40 atoms)</p><p>3. MDDatabase_3: Another dataset created using the same active learning framework outlined in the paper, on a larger unit cell (-300 atoms)</p>