Large-scale search for stable tin alloys with machine learning potentials

We have recently developed an automated framework for generating accurate machine learning potentials (MLPs) to accelerate ab initio structure prediction. Following our predictions of new thermodynamically stable Li-Sn compounds, we have expanded the MLP-guided evolutionary ground state searches to several M-Sn binary systems (M = Na, Mg, Ca, Cu, Pd, and Ag). The systematic exploration of the full binary composition ranges has uncovered a number of new crystal structure phases thermodynamically stable at different pressures and temperatures.