Material discovery using quantum-inspired technologies: validating a band gap engineering approach for inorganic materials

Abstract There is growing interest in applying materials informatics to inorganic materials, such as power semiconductors and high-entropy alloys. Developing material exploration methods that can overcome the combinatorial explosion, particularly in optimizing compositions containing multiple elements, is an urgent issue. This research successfully developed a method for optimizing the element arrangement of diamond-structured C, Si, and Ge to achieve a desired band gap. We combined mathematical optimization techniques utilizing Fujitsu Quantum-inspired Computing Digital Annealer (DA), a quantum-inspired technology, with density functional theory (DFT) calculations using Quantum Espresso (QE). We optimized the element arrangement using FM-DA&GA, a combination of the factorization machines (FM), a machine learning technique, with DA and genetic algorithms (GA). This approach was confirmed to be superior to random search, conventional GA methods, and FM-DA alone in terms of the number of calculations and computation time. Furthermore, in multiobjective optimization incorporating formation energy in addition to band gap, FM-DA&GA successfully identified the real material 3C-SiC in a 2-element (C and Si). FM-DA&GA combines the fast model exploration of FM-DA with the global search capabilities of GA. While the method of combining FM-DA&GA with DFT calculations has previously been applied to organic compounds, this study represents its world's first application to inorganic materials. This achievement is expected to significantly expand the possibilities for inorganic material exploration.