High-Speed NV Ensemble Magnetic Field Imaging via Laser Raster Scanning

TL;DR

The paper introduces a fast, high-sensitivity NV ensemble magnetic imaging platform that raster-scans a laser across a diamond with continuous microwaves, employing quasi-continuous-wave ODMR to read out spin states. By leveraging an acousto-optic deflector for rapid beam steering and a single photodetector with software-based common-mode noise suppression, the method achieves sub-millisecond temporal resolution across a ~170 μm field of view, with per-pixel sensitivity around 10 nT/√Hz in the 100–1000 Hz band. The authors demonstrate both single-pixel qCW-ODMR physics and wide-field imaging of a current-carrying microwire, highlighting a tunable trade-off between frame rate, spatial resolution, and sensitivity. They also compare favorably to camera-based NV imaging at high frame rates and discuss future enhancements like lock-in detection and double-quantum readout for further performance gains, making the approach promising for biomagnetic and other dynamic sensing applications.

Abstract

We present a technique that uses an ensemble of nitrogen-vacancy (NV) centers in diamond to image magnetic fields with high spatio-temporal resolution and sensitivity. A focused laser beam is raster-scanned using an acousto-optic deflector (AOD) and NV center fluorescence is read out with a single photodetector, enabling low-noise detection with high dynamic range. The method operates in a previously unexplored regime, quasi-continuous-wave optically detected magnetic resonance (qCW-ODMR). In this regime, NV centers experience short optical pump pulses for spin readout and repolarization -- analogous to pulsed ODMR -- while the microwave field continuously drives the spin transitions. We systematically characterize this regime and show that the spin response is governed by a tunable interplay between coherent evolution and relaxation, determined by the temporal spacing between pump laser pulses. Notably, the technique does not require precise microwave pulse control, thus simplifying experimental implementation. To demonstrate its capabilities, we image time-varying magnetic fields from a microwire with sub-millisecond temporal resolution. This approach enables flexible spatial sampling and, with our diamond, achieves $\text{nT}/\sqrt{\text{Hz}}$-level per-pixel sensitivity, making it well suited for detecting weak, dynamic magnetic fields in biological and other complex systems.

Figures (5)

• Figure 1: (a) Schematic of the scanning-based NV magnetometry setup. The beam of a 532 nm laser is raster-scanned across the layer of NV centers using a 2D AOD, with fluorescence detected by a PIN photodetector. A second photodetector provides a reference for common-mode rejection and technical noise suppression. (b) qCW-ODMR diagram: laser scanning (green), continuous microwave excitation (blue), and photodetector readout (red) across successive pixels and frames. (c) Fluorescence images of the diamond captured with a camera from above for different scan patterns as indicated in the images.
• Figure 2: Quasi-continuous-wave (qCW) ODMR characterization of a single pixel with the magnetic field aligned along the [110] crystal direction. (a) Normalized contrast and (b) linewidth, shown as a function of laser pulse spacing for two Rabi frequencies $\Omega_R$. Pulse spacing defines the time between successive 0.4µs laser pulses, while the microwave field is applied continuously. In (a), contrast is normalized to the CW-ODMR contrast at the corresponding $\Omega_R$. Vertical dashed lines indicate the expected $\pi$-pulse condition. Shaded regions represent the approximate regimes of coherent Rabi oscillations (gray), dephasing-dominated evolution (red), and relaxation-limited decay ($T_1$ regime, blue). In (b), solid horizontal lines indicate the corresponding linewidths obtained in CW for each Rabi frequency. (c) ODMR spectra of the $m_s = 0$ to $-1$ transition acquired under a Rabi drive of $\Omega_R = 70~\text{kHz}$ for a single pixel. The CW (purple) and qCW (red) spectra are shown; the qCW ODMR is measured with a pulse spacing of 100µs, revealing enhanced contrast compared to CW.
• Figure 3: Top: Noise floor levels (in $\mu\text{V}/\sqrt{\text{Hz}}$) for a single pixel as a function of pulse spacing, with total measurement time fixed at 1 s. As pulse spacing increases, fewer photons are collected, leading to a noise scaling consistent with $\sqrt{t}$. The dashed line shows the CW noise floor. Bottom: Sensitivity versus pulse spacing. The dashed purple line indicates the CW sensitivity baseline. The solid purple line shows the expected CW $\sqrt{t}$ scaling assuming constant contrast and linewidth. The red curve represents the actual sensitivity in the qCW regime, which initially improves beyond the expected CW scaling before degrading at longer spacings due to increasing noise and reduced contrast.
• Figure 4: Wide-field NV magnetometry of a current-carrying microwire. a) Time-traces from all 441 pixels in the $21\times 21$ scan (gray). The orange curve represents the spatial average of all traces, and the black dashed curve shows the recorded biphasic current stimulus. b) Microscope image showing the microwire relative to the NV sensing region. The oblique dark line indicates the wire position; scale bar: 170µm. c) Magnetic field maps at two selected times (26.6 ms and 53.1 ms), corresponding to the positive and negative phases of the applied stimulus. Pixels A and B (white X marks) lie on a horizontal cut (dashed line) used for subsequent analysis. d) Time-traces of pixels A (gold) and B (purple), overlaid with the applied stimulus (black). The two traces show opposite magnetic polarity due to their positions on opposite sides of the wire. e) Horizontal line profiles across the sensing plane at the same two timestamps as in (c), revealing the characteristic dipole-like magnetic field generated by the microwire. f) Noise performance of the sensor array. The standard deviation of successive averages follows the expected $1/\sqrt{t}$ scaling (red dashed line), demonstrating stable temporal averaging over three decades in time.
• Figure 5: (a) Magnetic sensitivity versus frame rate for conventional camera-based systems and our AOD scanning approach. Black diamonds represent state-of-the-art scientific and high-speed cameras, with sensitivity normalized to the effective area of a single AOD pixel (30µm), treating camera macropixels as matched sensing volumes. Labels correspond to camera models listed in the inset. The red star marks the performance of our AOD method, which achieves superior sensitivity at higher frame rates. (b) Projected root-mean-square (RMS) magnetic noise versus averaging time for our AOD system and a representative high-end camera (heliCam C4). The green region marks averaging times up to 24h, corresponding to the practical upper limit for continuous measurement in biological samples. The red region indicates integration times beyond this limit. The dashed line marks a 1nT detection threshold at signal-to-noise ratio (SNR) = 3. The AOD approach achieves lower noise and faster averaging compared to camera-based systems.