Unsupervised machine learning of topological phase transitions from experimental data

TL;DR

This paper demonstrates the use of unsupervised machine learning techniques to identify topological phase transitions directly from experimental data in quantum many-body systems, including noisy and finite-temperature data.

Contribution

It introduces the application of anomaly detection and influence functions to experimental ultracold atom data to map topological phases without prior knowledge of order parameters.

Findings

Successfully mapped the topological phase diagram of the Haldane model.

Applied methods to finite-temperature and Floquet systems.

Provided a benchmark for detecting exotic phases in complex systems.

Abstract

Identifying phase transitions is one of the key challenges in quantum many-body physics. Recently, machine learning methods have been shown to be an alternative way of localising phase boundaries also from noisy and imperfect data and without the knowledge of the order parameter. Here we apply different unsupervised machine learning techniques including anomaly detection and influence functions to experimental data from ultracold atoms. In this way we obtain the topological phase diagram of the Haldane model in a completely unbiased fashion. We show that the methods can successfully be applied to experimental data at finite temperature and to data of Floquet systems, when postprocessing the data to a single micromotion phase. Our work provides a benchmark for unsupervised detection of new exotic phases in complex many-body systems.