Bayesian optimization of 7-component (AlVCrFeCoNiMo) single crystal alloy’s compositional space to optimize elasto-plastic properties from molecular dynamics simulations

Abstract

Exploring the vast compositional space of high-entropy alloys (HEAs) promises materials with superior mechanical properties much needed in industrial applications. We demonstrate on the 7-component alloy AlVCrFeCoNiMo system with randomly ordered atoms that this exploration of the compositional space can be accelerated by combining molecular dynamics simulations with Bayesian optimization. Our algorithm is tested on maximizing the shear modulus, resulting in pure Mo, an unsurprising result based on Mo’s large density. Maximizing the yield stress results in Co-, Cr- and Ni-based alloys with the optimal composition varying depending on the presence of defects within the crystal. Finally, we optimize the plastic behaviour by aiming for high stresses while minimizing the deformation fluctuations, and find that a predominantly NiMo alloy’s high lattice distortions ensure a smooth stress response. The results suggest that mechanical properties of 2- to 4-component alloys with optimized composition may be superior to those of equiatomic HEAs without short-range order.