Physics-Based Machine Learning for the Electronic Structure of Phase Change Materials

Computing the electronic properties of amorphous and disordered materials is challenging but of great importance to accelerate the discovery and design of phase change materials for high-efficient in-memory computing. While first-principles calculations based on Density Functional Theory (DFT) have demonstrated the numerical accuracy required for most nanoelectronics applications, amorphous and polycrystalline systems represent a particular challenge due to their heterogeneity and the associated size of their structural approximants. Tight-binding methods offer the scalability required, but rely on their prior parametrization, a complex tax for multivalent and phase changing materials. In this work, we use machine learning to parameterize a tight-binding ansatz for the electronic structure of complex phase change materials. Using the DFT results for small unit cells of single and multivalent GexSbyTez (0< x,y,z<1) alloys, we map the complex atomic configurations to a tight-binding Hamiltonian using an atom-centered basis set and the local overlap matrix. This atomic-configuration-dependent tight-binding Hamiltonian allows us to evaluate the electronic structure of large-scale phase change materials in both crystalline, amorphous, and mixed phases.