Machine learning prediction of magnetic property of lanthanum substituted bismuth ferrite multiferroic with experimental validation

Abstract

Finding new Rare Earth Elements (REE) substituted Bi_1-xLa_xFeO₃ [x = 0.0, 0.03, 0.06] multiferroics with improved performance remains a significant challenge due to the time-consuming nature of conventional trial-and-error approaches and the vast unexplored composition space. Our study introduces a machine learning (ML) approach, explicitly employing a Random Forest (RF) regressor to accurately predict the magnetization (Ms) of REE-substituted multiferroics. Three compositions of Bi_1-xLa_xFeO₃ (x = 0.0, 0.03, 0.06) were synthesized and experimentally characterized to validate the model’s predictions. The RF model was applied to predict the magnetization of Bi_1-xLa_xFeO₃[x = 0.0, 0.03, 0.06], a material system relevant to multiferroic research. In addition, the trained model was tested on literature-reported compositions with other REE substitutions to assess its generalization capability. The model’s predictions showed strong agreement with experimental data, demonstrating strong alignment and confirming its predictive reliability. The model achieved a high coefficient of determination (R²) of 0.99 and the average error of 0.20 emu/g, indicating excellent predictive performance. This work provides a robust method for predicting Ms and paves the way for more efficient design and discovery of REE-substituted multiferroics through data-driven approaches.