Accelerating the prediction of stacking fault energy by combining ab initio calculations and machine learning

TL;DR

This paper presents a machine learning approach combined with ab initio calculations to rapidly predict stacking fault energies in metals and alloys, significantly speeding up the alloy design process.

Contribution

It introduces a physics-informed ML model that accelerates SFE predictions by approximately 80 times compared to traditional DFT calculations.

Findings

ML predictions are consistent with experimental data.

The approach links d-electron physics to deformation behavior.

Significantly reduces computational time for SFE estimation.

Abstract

Stacking fault energies (SFEs) are vital parameters for understanding the deformation mechanisms in metals and alloys, with prior knowledge of SFEs from ab initio calculations being crucial for alloy design. Machine learning (ML) algorithms employed in the present work demonstrate approximately 80 times acceleration in predicting generalized stacking fault energy (GSFE), which is otherwise computationally expensive to obtain directly from density functional theory (DFT) calculations, particularly for alloys. The features used to train the ML algorithms stem from the physics-based Friedel model, revealing a connection between the physics of d-electrons and the deformation behavior of transition metals and alloys. Predictions based on the ML model are consistent with experimental data. This model could aid in accelerating alloy design by offering a rapid method for screening materials based on stacking fault energies.