Prediction of Magnetic Janus Materials Based on Machine Learning and First-Principles Calculations

Abstract

Discovering the compact、 stable and easily controllable nanoscale non-trivial topological magnetic structures—magnetic skyrmions,is the key to develop next-generation high-density, high-speed,and lowenergy non-volatile information storage devices.Based on the topological generation mechanism,magnetic skyrmions could be generated through the Dzyaloshinskii–Moriya Interaction (DMI) induced by spacereversal symmetry broken.Two dimensional (2D) non-centrosymmetric Janus could generate vertical builtin electric fields to break spatial inversion symmetry. Therefore, seeking 2D Janus with intrinsic magnetism is fundamental to develop the novel chiral magnetic storage technologies.In this work, we combined detailed machine learning techniques and first-principles calculations to discover the magnetism of the unexplored 2D janus. we first collected 1179 2D hexagonal ABC-type Janus based on the Materials Project database, and used elemental composition as feature descriptors to construct four machine learning models: Random Forest(RF), Gradient Boosting Decision Trees (GBDT), Extreme Gradient Boosting (XGB), and Extra Trees(ET). These algorithms and models were constructed to predict lattice constants, formation energies, and magnetic moment, via hyperparameter optimization and ten-fold cross-validation. GBDT exhibits the highest accuracy and best prediction performance for magnetic moment classification. Subsequently, the collected data of 82,018 yet-undiscovered 2D Janus,were input into the trained models to generate 4,024 high magnetic moment 2D Janus with thermal stability. First-principles calculations were employed to validate random sample of 13 Janus with high magnetic moment. This study provides an effective machine learning framework for magnetic moment classification and high-throughput screening of 2D Janus, accelerating the exploration of magnetic properties in 2D Janus structures.