Data-Driven Studies of van der Waals Magnetic Heterostructures

Magnetic van der Waals (vdW) materials have the potential to revolutionize the semiconductor industry due to their various exotic physical properties. Herein, we use a data-driven framework to investigate magnetic vdW heterostructures of the form A^<i>i</i>A^<i>ii</i>B^<i>i</i>₄X^<i>i</i>₈/B^<i>ii</i>₄X^<i>ii</i>₆ based on a well-known heterostructure MnBi₂Te₄/Sb₂Te₃. Our study shows that combining a nonmagnetic B^<i>ii</i>₄X^<i>ii</i>₆ monolayer with a magnetic A^<i>i</i>A^<i>ii</i>B^<i>i</i>₄X^<i>i</i>₈ monolayer can alter the magnetic properties as well as the band gap. By training various machine learning (ML) models on the density functional theory (DFT)-generated data set, we rapidly predict the properties of 16,431,660 A^<i>i</i>A^<i>ii</i>B^<i>i</i>₄X^<i>i</i>₈/B^<i>ii</i>₄X^<i>ii</i>₆ heterostructures. The results from the ML predictions are used to identify promising candidate heterostructures. Our study aims to accelerate the design of vdW heterostructures that have applications in spintronics, optoelectronics, and topological quantum computing.