Machine Learning in Electronic Quantum Matter Visualization Experiments

27 June 2019

A collaboration of experimental physicists led by Prof. JC Séamus Davis (University of Oxford), theoretical physicists led by Prof. Eun-Ah Kim (Cornell University), and computer scientists led by Prof. E. Kathami (San Jose State University), developed and trained a new Machine Learning (ML) protocol, based on a suite of artificial neural networks (ANN), that is designed to recognize different types of electronic ordered states which are hidden within electronic quantum matter image-arrays.
Electronic quantum matter studies using automated scientific instrumentation and large-scale data acquisition are now generating data sets of such volume and complexity as to defy human analysis. For example modern scanning tunneling microscopy (STM) visualization of electronic quantum matter (EQM) yields dense arrays of atomic-scale, electronic-structure images, that are often astonishingly complex.
The ANN suite analyzed one of the Davis Group experimental EQM image archives, spanning a wide range of electron densities and energies, in carrier-doped cuprate Mott insulators. The ANN suite discovered, throughout all the noisy and complex data, the features of a very specific ordered state of EQM: a Vestigial Nematic State.
This is a milestone for general scientific technique in that ANN’s can process and identify specific broken symmetries of highly complex image-arrays from non-synthetic experimental EQM data. It opens the immediate prospect of additional ML-driven scientific discovery in EQM studies.

Nature 570, 484 (2019).