Benchmarking Universal Machine Learning Interatomic Potentials for Elastic Property Prediction

Abstract: Universal machine learning interatomic potentials have emerged as efficient tool for material simulation fields,yet their reliability for elastic property prediction remains unclear. Here we present a systematic benchmark of four uMLIPs-MatterSim, MACE, SevenNet, and CHGNet-against theoretical data for nearly 11,000 elastically stable materials from the Materials Project database. The results show SevenNet achieves the highest accuracy,MACE and MatterSim balance accuracy with efficiency, while CHGNet performs less effectively overall. This benchmark establishes a framework for guiding model selection and advancing uMLIPs in mechanical property applications.