Nanophotonic magnetometry in a spin-dense diamond cavity

Abstract

Quantum sensors based on the nitrogen-vacancy (NV) center in diamond are leading platforms for high-sensitivity magnetometry with nanometer-scale resolution. State-of-the-art implementations, however, typically rely on bulky free-space optics or sacrifice spatial resolution to achieve high sensitivities. Here, we realize an integrated platform that overcomes this trade-off by fabricating monolithic whispering-gallery-mode cavities from a diamond chip containing a high density of NV centers and by evanescently coupling excitation to and photoluminescence from the cavity using a tapered optical fiber. Employing a lock-in-amplified Ramsey magnetometry scheme, we achieve a photon-shot-noise-limited DC sensitivity of $58\,\text{nT}/\sqrt{\text{Hz}}$ -- the best sensitivity reported to date for a nanofabricated cavity-based magnetometer. The microscopic cavity size enables sub-micrometer-scale spatial resolution and low-power operation, while fiber-coupling provides a path to scalable on-chip integration. Arrays of such sensors could enable NV-NMR spectroscopy of sub-nanoliter samples, new magnetic-gradient imaging architectures, and compact biosensing platforms.

Figures (16)

• Figure 1: Magnetometry using NV$^{-}$ centers in a diamond DNV cavity.a, Different geometries can be used for sensing with varying NV center quantity. A single NV achieves optimal spatial resolution, $\delta x$, whereas ensembles achieve better sensitivity. A cavity-integrated ensemble further enhances sensitivity by making the pump field multi-pass and by improving resonant PL collection (see text for discussion). b, In this work, a fiber-taper-coupled diamond cavity is used to perform magnetometry. Initialization and readout is achieved using evanescent coupling of light to and from the fiber-taper waveguide. Here, we sketch the fiber-taper atop a scanning electron micrograph of the diamond cavity.
• Figure 2: Fiber-coupled cavity ODMR.a, NV PL collected from a fiber-coupled cavity compared to that collected from the same diamond bulk material using a confocal microscope. The coupled cavity acts as a spectral filter, collecting only photons emitted into resonant modes that couple to the fiber-taper. This is used to enhance the SNR and suppress background PL. b, Optically detected magnetic resonance of NV centers in the cavity with and without a bias magnetic field, $B_{\text{bias}}$. We resolve a hyperfine structure with a linewidth of 773 (8) kHz and achieve Rabi frequencies up to 6.5 MHz (inset).
• Figure 3: Characterizing the spin-coherence time of the diamond microcavity.a, The inhomogeneous dephasing time, $T_2^*$, is extracted using Ramsey ODMR. The grey background shows the fitted decay envelope function ($\exp[-\tau/T_2^*]$) of the Ramsey signal (blue points). The blue lineplot is a least squares fit to the data (see \ref{['app:sensitivityOptimization']}). b, The spin-coherence time of the cavity is extended using $n$-CPMG pulse sequences, which dynamically decouple the NV centers from magnetic noise. c, The measured $T_2$ values for each $n$ are compared to $T_2^*$ and $T_1$, which is extracted from the measurement shown in d.
• Figure 4: Magnetometry using the microcavity sensor.a, Lock-in techniques suppress electronic and laser noise. Frequency-modulated microwaves (red) cause the NV PL signal (grey) to fluctuate at the modulation frequency, which can be extracted to suppress noise. b, Measurement of NV ensemble spin transitions split by a bias magnetic field using lock-in detection. The dashed black line indicates the operating point for maximum signal response. c, ODMR spectra comparing single-tone (grey) and three-tone (blue) microwave driving. Simultaneous driving of all three $^{15}$N nuclear hyperfine transitions increases signal contrast and results in five spectral features. Pulsed ODMR and Ramsey measurements are plotted as a function of detuning, $\Delta_{\text{MW}}/2\pi$. Lock-in measurements done for both the Pulsed and Ramsey ODMR protocols produce similar contrast to their non-lock-in counterparts; however, they suppress low-frequency noise and result in higher sensitivity. d, e, Magnetic field measurements and Allan deviation for different sensing schemes. The inset in d shows 200 ms of Ramsey-LI data. Lock-in pulsed ODMR and Ramsey measurements are shot-noise-limited, as demonstrated by the bar chart in d, which shows the predicted shot-noise-limited sensitivities ($\eta_{\text{sn}}$). The sensitivities measured in the data presented in d and e are in good agreement. The similarities between the Allan deviation and the magnetic noise signify the sensor's stability sensor -- it is not limited by drift.
• Figure 5: Comparison of the figures of merit of different diamond magnetometers. Comparison of the sensitivity (a) and optical control power (b) as a function of volumetric spatial resolution for different magnetometers comprised of single and ensembles of NV centers: 1 Taylor2008; 2 Fang2013; 3 Clevenson2015; 4 Barry2016; 5 Pelliccione2016; 6 Patel2020; 7 Strner2021; 8 Zhang2021; 9 Shim2022; 10 Graham2023; 11 Barry2024; 12 Sekiguchi2024; 13 Katsumi2025. In a, the black arrow and outlined star highlight the sensitivity improvement estimated given the modifications outlined in the main text. The $T_2^*$-limited spin projection limit is calculated as $\eta_{\text{sp}}= (\gamma \sqrt{NT_2^*})^{-1}$, where $N$ is the sample volume multiplied by the NV density, $\text{[NV]}=4.5\,$ppm.