Detection of topological materials with machine learning

Abstract

Databases compiled using ab–initio and symmetry-based calculations now contain tens of thou- sands of topological insulators and topological semimetals. This makes the application of modern machine learning methods to topological materials possible. Using gradient boosted trees, we show how to construct a machine learning model which can predict the topology of a given existent mate- rial with an accuracy of 90%. Such predictions are orders of magnitude faster than actual ab initio calculations. We use machine learning models to probe how different material properties affect topo- logical features. Notably, we observe that topology is mostly determined by the “coarse–grained” chemical composition and crystal symmetry and depends little on the particular positions of atoms in the crystal lattice. We identify the sources of our model’s errors and we discuss approaches to overcome them.