Machine-learning Spectral Indicators of Topology

Topological materials are promising for next-generation energy and information applications. However, the experimental determination of topology can be painstaking, with a few limitations such as limited sample types, high technical barriers, and limited sample environment.

Topological materials are promising for next-generation energy  and  information  applications.  However,  the experimental   determination   of   topology   can   be painstaking, with a few limitations such as limited sample types,  high  technical  barriers,  and  limited  sample environment.
In  this  work,  by  designing  a  machine  learning architecture  to  analyze  simple  X-ray  absorption  spectra, materials’ topological class can be determined with over 90% accuracy. This enables the topological determination of new topological materials with much reduced technical barriers  in  broader  types  of  material  candidates,  and facilitates  study  of  other  phenomena  like  topological phase transition and amorphous topological materials.

Designing Materials to Revolutionize and Engineer our Future (DMREF)