Item

Abstract

Significant advances have been made in predicting new topological materials using high-throughput empirical descriptors or symmetry-based indicators. This line of research has produced extensive lists of candidate topological materials that still await experimental validation. To date, these approaches have been limited to materials already known in databases, leaving a much larger portion of the materials space unexplored. Here we uncover a novel two-dimensional descriptor for fast and reliable identification of the topological characters of complex alloyed systems. Using tetradymites with widely varying stoichiometric compositions as examples, we obtain this descriptor by applying a recently developed data-analytics approach named SISSO (Sure Independence Screening and Sparsifying Operator) to training data from high-level electronic structure calculations. By leveraging this descriptor that contains only two elemental properties (the atomic number and electronegativity) of the constituent species, we can readily scan over four million alloys in the tetradymite family. Strikingly, nearly two million new topological insulators are discovered, thus drastically expanding the territory of the topological materials world. The strong predictive power of the descriptor beyond the initial scope of the training data also testifies the increasing importance of such data-driven approaches in materials discovery.