Bloch oscillations: Inverse problem

TL;DR

This paper employs neural networks to solve the inverse problem of Bloch oscillations, accurately determining physical parameters like field strength or lattice spacing from oscillation signals.

Contribution

It introduces a neural network method to accurately infer physical parameters from Bloch oscillation signals, addressing the inverse problem in condensed matter physics.

Findings

Over 80% accuracy in classifying physical parameters

Effective neural network approach for inverse problem solving

Potential for experimental application in material characterization

Abstract

We use a neural network approach to explore the inverse problem of Bloch oscillations in a monoatomic linear chain: given a signal describing the path of oscillations of electrons as a function of time, we determine the strength of the applied field along the direction of motion or, equivalently, the lattice spacing. We find that the proposed approach has more than 80% of accuracy classifying the studied physical parameters.