mavrl/matpes

MatPES（材料势能表面）数据集是由Materials Virtual Lab和Materials Project共同发起的，旨在解决现有材料势能表面数据集中的关键缺陷。该数据集使用静态DFT计算，并采用严格的收敛标准，以保证准确性。MatPES结构采用两阶段的维度降低编码集群与分层DIRECT采样方法，从大量扩展的MD结构中选取，确保了数据的全面性。数据集包括来自PBE功能以及高保真r2SCAN元GGA功能计算的数据，以改进不同键合和化学环境的描述。初始版本v2025.1包含约40万个来自30万MD模拟的结构，尽管规模小于其他文献中的PES数据集，但它在训练FPs时展现出可比或更高的性能和可靠性。

The MatPES (Materials Potential Energy Surface) dataset is an initiative by the Materials Virtual Lab and the Materials Project to address critical deficiencies in existing PES datasets for materials. It is computed using static DFT calculations with stringent convergence criteria to ensure accuracy. MatPES structures are sampled using a 2-stage version of Dimensionality-Reduced Encoded Clusters with sTratified DIRECT sampling from a greatly expanded configuration of MD structures, ensuring comprehensiveness. The dataset includes computed data from the PBE functional and the high-fidelity r2SCAN meta-GGA functional with improved descriptions across diverse bonding and chemistries. The initial v2025.1 release comprises ~400,000 structures from 300K MD simulations, and it achieves comparable or improved performance and reliability on trained FPs despite being smaller in size than other PES datasets in the literature.