Multi-objective optimization of High-entropy alloy properties

Body-centered cubic (bcc) high-entropy alloys possess high strength retention at high temperatures, while suffering

from limited ductility and embrittlement at low temperatures. Although both the strength and the ductility

can be characterized by numerical modeling from first principles, the search for a composition with an optimal trade-off is hampered by the complex alloy structure

and vast composition space. In this work, we present a methodology for exploring Pareto-optimal compositions by combining ab initio

based computational workflow with a Bayesian multi-objective optimization framework. One part of the computational workflow involves

evaluation of solid solution strengthening within a model relying on parameters derived from efficient ab initio calculations using

coherent-potential approximation and taking into account finite-temperature and exchange-correlation corrections for the equilibrium volume.

Another part of the workflow is based on a ductility model whose parameters are obtained

using machined-learned interatomic potentials parameterized on the fly by means of

active learning from ab initio calculations.

We demonstrate how our approach can be used to address the strength-ductility trade-off in multi-component alloys

based on refractory metals.