Fast characterization of optically detected magnetic resonance spectra via data clustering

TL;DR

This paper introduces a data clustering algorithm that significantly improves the speed and accuracy of analyzing ODMR spectra, enabling faster quantum sensing with less data and noise.

Contribution

The paper presents a novel clustering-based algorithm for analyzing ODMR spectra, outperforming traditional methods in accuracy, resolution, and data efficiency.

Findings

Improves resonance frequency determination accuracy by ~1.3x

Enhances spectral resolution by ~4.7x

Reduces required data points by ~5x

Abstract

Optically detected magnetic resonance (ODMR) has become a well-established and powerful technique for measuring the spin state of solid-state quantum emitters, at room temperature. Relying on spin-dependent recombination processes involving the emitters ground, excited and metastable states, ODMR is enabling spin-based quantum sensing of nanoscale electric and magnetic fields, temperature, strain and pressure, as well as imaging of individual electron and nuclear spins. Central to many of these sensing applications is the ability to reliably analyze ODMR data, as the resonance frequencies in these spectra map directly onto target physical quantities acting on the spin sensor. However, this can be onerous, as relatively long integration times -- from milliseconds up to tens of seconds -- are often needed to reach a signal-to-noise level suitable to determine said resonances using traditional fitting methods. Here, we present an algorithm based on data clustering that overcome this limitation and allows determining the resonance frequencies of ODMR spectra with better accuracy (~1.3x factor), higher resolution (~4.7x factor) and/or overall fewer data points (~5x factor) than standard approaches based on statistical inference. The proposed clustering algorithm (CA) is thus a powerful tool for many ODMR-based quantum sensing applications, especially when dealing with noisy and scarce data sets.