Design of Hybrid Improper Ferroelectric Double Perovskite Oxides with Atomistic Simulations and Machine Learning

Abstract

The integration of machine learning (ML) in multiferroic materials research has newlineopened new avenues for accelerating the discovery and optimization of functional properties. newlineIn particular, understanding the basic mechanisms like Hybrid Improper Ferroelectricity newline(HIF) in Double perovskite oxides (DPOs) have emerged as promising candidates newlinefor spintronic and non-volatile memory applications due to their coupled polarization newlineand magnetization switching mechanisms. Understanding and predicting these newlineswitching barriers and dynamical behaviors are critical for the efficient design of nextgeneration newlinespintronic devices. However, conventional first-principles approaches are newlinecomputationally demanding when exploring vast compositional and structural spaces. newlineMachine learning provides an efficient alternative by leveraging data-driven models newlinetrained on density functional theory (DFT) and molecular dynamics (MD) simulations newlineto accurately predict polarization and magnetization switching barriers, phase transitions, newlineand stability trends. Advanced ML techniques, including deep learning and newlinephysics-informed neural networks, have demonstrated success in capturing structureproperty newlinerelationships, enabling inverse design strategies for tailored multiferroic functionalities. newlineIn double perovskites, ML models facilitate the prediction of metal-insulator transitions newlineand magnetoelectric coupling under varying strain conditions. Meanwhile, for newlinehybrid improper ferroelectrics, ML aids in quantifying octahedral rotation-induced polarization newlineswitching and energy barriers, providing insights into their switching dynamics. newlineThe synergy between ML and multiferroic research is poised to revolutionize the newlinefield by enabling rapid screening and optimization of materials for high-performance newlinespintronic applications, bridging the gap between theoretical predictions and experimental newlinerealizations newline