Data-Driven Accelerating the Discovery of Hole-Doping Induced 2d Magnets

Abstract

Hole-doping induced 2D magnets (termed as HDIM), with hole-mediated magnetic state, not only fill in the scarce situation of 2D intrinsic magnets but also bring the great promise for next-generation spintronic nanodevices. However, current frameworks based on these first-principles calculations approaches for excavating HDIM show great blindness and laboriousness. Herein, a data-driven high-throughput screening framework was proposed to overcome this challenge. Utilizing and learning the data of Computational 2D Materials Database (C2DB), we identified the hole effective mass (HEM) as a physical descriptor for efficiently discovering HDIM, and developed an effective HEM machine learning model. Via the HEM-driven high-throughput screening framework, we screened the 2DMatPedia database and obtained a set of 477 high-stability HDIM candidates. Combination with high-throughput calculations assess that up to 35% exhibit significant HDIM-associated properties. For example, finding a novel HDIM of ZrMo2O8, with a fantastic honeycomb-checkerboard 2D architecture, demonstrates hole-doped induced rarely half-metallic ferromagnetism and high Curie temperature. This proposed data-driven framework offers a high-efficiency approach toward accelerating the discovery of 2D tunable magnets.

Keywords: high-throughput screening, machine learning, hole-doping induced magnetism, 2D magnetic materials