Analyzing Nonequilibrium Quantum States through Snapshots with Artificial Neural Networks

Area of Science:

• Quantum simulation and condensed matter physics.
• Application of machine learning to quantum systems.

Background:

• Quantum simulation experiments are advancing into unexplored regimes of many-body dynamics.
• Identifying suitable observables for studying these dynamics is a key challenge.
• Understanding thermalization in quantum systems is fundamental.

Purpose of the Study:

• To investigate the dynamics and thermalization behavior of interacting quantum systems.
• To explore the transition from ergodic to many-body localized phases.
• To determine effective machine learning techniques for analyzing quantum many-body systems.

Main Methods:

• Utilized supervised and unsupervised machine learning (ML) techniques.
• Employed ML network performance as an indicator of thermalization.
• Tested ML methods on experimental data from ultracold atoms using a quantum gas microscope.

Main Results:

• Successfully distinguished between nonequilibrium and equilibrium data using ML.
• Demonstrated ML's capability to probe thermalization behavior in quantum systems.
• Validated the approach with real experimental snapshots of ultracold atoms.

Conclusions:

• Machine learning offers a powerful pathway for analyzing highly entangled quantum states.
• This method is effective for large system sizes where traditional numerical methods are intractable.
• The study provides a novel approach to understanding complex quantum dynamics and phase transitions.