SciPost Submission Page

Detecting Nematic Order in STM/STS Data with Artificial Intelligence

by Jeremy B. Goetz, Yi Zhang, Michael J. Lawler

Submission summary

As Contributors: Michael Lawler
Arxiv Link: https://arxiv.org/abs/1901.11042v1
Date submitted: 2019-02-04
Submitted by: Lawler, Michael
Submitted to: SciPost Physics
Domain(s): Theoretical
Subject area: Condensed Matter Physics - Computational

Abstract

Detecting the subtle yet phase defining features in Scanning Tunneling Microscopy and Spectroscopy data remains an important challenge in quantum materials. We meet the challenge of detecting nematic order from local density of states data with supervised machine learning and artificial neural networks for the difficult scenario without sharp features such as visible lattice Bragg peaks or Friedel oscillation signatures in the Fourier transform spectrum. We train the artificial neural networks to classify simulated data of isotropic and anisotropic two-dimensional metals in the presence of disorder. The supervised machine learning succeeds only with at least one hidden layer in the ANN architecture, demonstrating it is a higher level of complexity than nematic order detected from Bragg peaks which requires just two neurons. We apply the finalized ANN to experimental STM data on CaFe2As2, and it predicts nematic symmetry breaking with 99% confidence (probability 0.99), in agreement with previous analysis. Our results suggest ANNs could be a useful tool for the detection of nematic order in STM data and a variety of other forms of symmetry breaking.

Current status:
Editor-in-charge assigned

Submission & Refereeing History

Submission 1901.11042v1 on 4 February 2019

Login to report


Comments

Anonymous on 2019-04-08

The manuscript by Goetz et al. reports on detecting subtle nematic order from local density of states (LDOS) data measured by scanning tunneling microscopy (STM) using supervised machine learning. They simulate LDOS data using various tight binding methods to train and test their artificial neural network (ANN). They obtain 95% accuracy score with an ANN with a single hidden layer. They test the ANN on a real STM data that is observed to be anisotropic in previous studies. The ANN is able to identify nematic order with 99% confidence.

The manuscript is well written, the subject is novel and the results are compelling in my opinion.
However, the authors test the ANN only on a single anisotropic STM image. I would like to see its performance on few other isotropic as well as anisotropic STM images. The fact that the ANN performs poorly on simulated data (only %65 Fig. 2) which it is trained on compared to the real STM data (%99) is a bit controversial.

Typo:
For example, inhomogeneous behavior found in strongly correlated materials [can lead] to ...