Machine Learning Assisted Vector Atomic Magnetometry

TL;DR

This paper introduces a machine learning-based vector magnetometry technique that encodes three-dimensional magnetic field information into optical signals and uses a neural network for measurement, simplifying the device design.

Contribution

It presents a novel machine learning approach for vector magnetometry that simplifies the architecture by using a single laser beam and neural network decoding.

Findings

Achieved magnetic field sensitivities of about 100 fT/√Hz.

Demonstrated angular sensitivities of about 100 μrad/√Hz.

Implemented a single-shot all optical vector magnetometer.

Abstract

We propose a novel paradigm to vector magnetometry based on machine learning. Unlike conventional schemes where one measured signal explicitly connects to one parameter, here we encode the three-dimensional magnetic-field information in the set of four simultaneously acquired signals, i.e., the oscillating optical rotation signal's harmonics of a frequency modulated laser beam traversing the atomic sample. The map between the recorded signals and the vectorial field information is established through a pre-trained deep neural network. We demonstrate experimentally a single-shot all optical vector atomic magnetometer, with a simple scalar-magnetometer design employing only one elliptically-polarized laser beam and no additional coils. Magnetic field amplitude sensitivities of about 100 $\textrm{fT}/\sqrt{\textrm{Hz}}$ and angular sensitivities of about 100 $\mu rad/\sqrt{\textrm{Hz}}$ (for a magnetic field of about 140 nT) are derived from the neural network. Our approach can reduce the complexity of the architecture of vector magnetometers, and may shed light on the general design of multiparameter sensing.