Inverse Design of Quantum Materials by Monte Carlo Tree Search and First Principles Calculations

Quantum materials that exhibit novel magnetism and unconventional superconductivity hold promise for future quantum sensing and quantum computing applications. However, designing novel materials with desired properties is challenging because of the vast geometrical and compositional degrees of freedom in the design space. We highlight our recent efforts in designing novel quantum materials by combining first principles calculations with a machine learning technique, Monte Carlo Tree Search. To predict the stability of compounds, we employ Density Functional Theory and MACE, a "foundational model" trained to the Materials Project, to calculate essential physical properties such as the formation energy and energy above hull. As a proof of concept, we inverse design and predict a few stable materials by performing an iterative, guided search for the best structure and composition using the Monte Carlo Tree Search algorithm.