Performance of universal machine learning potentials in global optimization

Rapid development of universal machine learning potentials (uMLPs) and expansion of training data sets are reshaping the state of the art in atomistic simulation, highlighting the need for concurrent systematic benchmarking of their capabilities. Global optimization is among the most demanding uMLP applications because unconstrained exploration includes probing motifs not present in reference sets. We examine the latest generation of uMLPs in unconstrained evolutionary searches to assess whether these models can consistently predict complex grounds states across diverse inorganic systems. The observed performance of the considered uMLPs establishes considerable variance, from near ab initio to essentially non-predictive, in their ability to rank low-energy crystal structures. Additional tests for select ground states reveal that some uMLPs capture fine energy differences arising from subtle electronic structure features.