In silico design and prediction of metastable quaternary phases in Cu-Ni-Si-Cr alloys

TL;DR

This study combines quantum simulations, machine learning, and classical methods to predict and analyze novel metastable quaternary phases in Cu-Ni-Si-Cr alloys, advancing alloy design and understanding.

Contribution

It introduces a novel integrated computational approach for predicting stable quaternary phases and their properties in copper alloys using active learning and atomistic simulations.

Findings

Predicted two new quaternary phases in Cu-Ni-Si-Cr alloys.

Developed machine-learned potentials with quantum accuracy.

Elucidated structural and elastic properties of the phases.

Abstract

Quaternary phases formed in copper alloys are investigated through a combination of quantum-mechanical and classical computer simulations and active machine learning. Focus is given on nickel, silicon, and chromium impurities in a copper matrix. The analysis of the formation enthalpies of candidate quaternary structures leads to the prediction of two novel quaternary phases and the assessment of their stability. For the predicted two phases, machine learned atomistic potentials are developed using active learning with a quantum-mechanical accuracy. Use of these potentials in atomistic simulations further elucidates the structure, temperature-dependent dynamics, and elastic behavior of the predicted quaternary phases in copper alloys. The combined in silico approach is thus proven highly efficient in both designing materials and elucidating their properties and potential combining different spatiotemporal scales. In the case of alloys, this computational scheme significantly reduces the effort in searching the huge chemical space of possible phases, enhancing the efficiency in synthesizing novel alloys with pre-defined properties.