Predictive model for the discovery of sinter-resistant supports for metallic nanoparticle catalysts by interpretable machine learning

Neural Network Potential-driven Molecular Dynamics Database for Catalyst Support Materials This database contains Neural Network Potential driven Molecular Dynamics (NN-MD) simulation results for 3nm Pt nanoparticle supported on various metal oxides supports at 800 °C for 500 ps and high-throughput screening results for catalyst support materials, organized into three datasets for comprehensive analysis of sinter-resistant catalyst supports. Dataset Contents Dataset 1: NN-MD_Database_Custom_Selection.zip Source: Custom-selected material systems Content: Molecular dynamics simulation results for specifically chosen material surfaces Files: MD trajectories (PDB), analysis plots (SVG), and numerical data (CSV) Dataset 2: NN-MD_Database_From_OC22.zip Source: OC22 (Open Catalyst 2022) dataset Content: NN-MD simulation results for catalyst surfaces from OC22 database Files: MD trajectories (PDB), analysis plots (SVG), and numerical data (CSV) Dataset 3: Atomic_Structures_for_Promising_Sinter-Resistant_Supports.zip Source: High-throughput screening using interpretable generalized additive model (iGAM) Content: Atomic structures of promising sinter-resistant catalyst supports Files: VASP POSCAR format structure files (Extracted from the last frame of the local optimization trajectory) File Structure and Naming Folders: Named as {Material_name}_{MP-ID}_{Miller_index} (e.g., MoO3_mp-20589_001) or {Material_name}_{Miller_index}_{Trajectory._ID_in_OC22} (e.g., {FePtO4-rutile}_{001}_{FePtO4_FePtO4-rutile_clean_Tj36HCumzE} ) Key Files: NNMD_output.pdb: MD trajectory data MD_contact_angle_Normalized_MSI_descriptor.svg: Contact angle and MSI descriptor plots MD_Pt_Eadh_ChemicalPotential.svg: Platinum adsorption energy analysis MD_Pt_contact_angle_adhesion_energy.csv: Quantitative adhesion data POSCAR: VASP-format atomic structures extracted from the last frame of traj. file in OC22 database (Dataset 3 only) Technical Specifications Simulation Method: Neural network potential-based molecular dynamics Coordinate System: Cartesian coordinates Units: Distance (Å), Energy (eV), Time (ps) Structure Format: VASP POSCAR format for atomic structures Screening Method: iGAM (interpretable Generalized Additive Model) Applications This database supports research in: Catalyst support design and optimization Sinter-resistance analysis Platinum-support interaction studies High-throughput materials screening Neural network potential validation Usage Compatible with standard molecular dynamics analysis software (VMD, OVITO) and materials science visualization tools (VESTA, Materials Studio). CSV files contain ready-to-use numerical data for statistical analysis. Keywords molecular dynamics, neural network potential, catalyst support, sinter-resistance, high-throughput screening, OC22 dataset, materials informatics Related Resources Open Catalyst 2022 (OC22) Dataset Materials Project Database