Prediction magnetocrystalline anisotripy Fe-Rh thin films via machine leaning

TL;DR

This paper applies Lasso regression to predict magnetocrystalline anisotropy energy in Fe-Rh thin films, revealing a linear relationship with orbital moment anisotropy based on first-principles data.

Contribution

It introduces a machine learning approach using Lasso to analyze magnetocrystalline anisotropy in Fe-Rh thin films, a novel application in this context.

Findings

Lasso regression effectively models MCA energy in Fe-Rh films.

A linear correlation between MCA energy and orbital moment anisotropy was identified.

First-principles data supported the regression analysis.

Abstract

Least absolute shrinkage and selection operator (Lasso) was originally formulated for least squares models and this simple case reveals a substantial amount about the behavior of the estimator. It also shows that the coefficient estimates need not be unique if covariates are collinear. Using this Lasso technique, we analyze a magnetocrystalline anisotropy energy which is a long-standing issue in transition-metal thin films, expectially for Fe-Rh thin film systems on a MgO substrate. Our LASSO regression took advantage of the data obtained from first principles calculations for single slabs with seven atomic-layers of binary Fe-Rh films on MgO(001). In the case of Fe-Rh thin films, we have successfully found a linear behavior between the MCA energy and the anisotropy of orbital moments.