Topological materials: from a catalogue to machine learning
The advent of Topological Quantum Chemistry (TQC) and Magnetic Topological Quantum Chemistry (MTQC) provides a straightforward approach to look at the topological properties of a material through ab-initio calculations. Combined with the non-magnetic material database from the Inorganic Crystal Structure Database and the Bilbao Magnetic Material Database, this allows for a systematic search of topological materials. We present a complete catalogue of topology in all the bands of the more than 75000 stoichiometric materials existing in the world, with and without spin-orbit coupling. This systematic survey brings several new exciting material classes, including fragile phases and new, ”supertopological” compounds where all the bands (not only at the Fermi level) have topological features.
This trove of data makes possible the application of modern machine learning methods to topological materials. Using gradient boosted trees, we show how to construct a machine learning model which can predict the topology of a given existent material with an accuracy of 90%. Such predictions are orders of magnitude faster than actual ab-initio calculations. Through extensive testing of different models we determine which properties help detect topological materials.