Skip to content

Real-time Amplitude and Phase Estimation of AC Fields with Diamond Spins

FieldValue
JournalOpen MIND
Published2026-02-27
AuthorsC. T.-K. Lew, Samuel A. Wilkinson, Nicholas Gillespie, Brant C. Gibson, David A. Broadway
DOI10.48550/arxiv.2602.23712

An NV-center ensemble in diamond recovered a 4 MHz magnetic field’s amplitude and phase from one pair of consecutive measurements. The per-shot sensitivities were 78 nT for amplitude and 63 mrad for phase at a temporal resolution of 320 µs.

Each estimate came from two sequential readouts that formed orthogonal in-phase and quadrature measurements of a coherent field. The first readout measured $I(t_i)$, and the delayed readout measured $Q(t_i)$. Amplitude followed $R(t_i)=\sqrt{I(t_i)^2+Q(t_i)^2}$, while phase followed $\phi(t_i)=\operatorname{atan}(Q(t_i)/I(t_i))$. The separation was $\Delta t=(n+1/4)T_{\mathrm{AC}}$, where $n$ accounts for optical initialization and readout overhead. This orthogonality separates the protocol from CASR, which advances the signal phase through a small frequency detuning. Each pair returns amplitude and phase without averaging many readouts over seconds.

After a $\pi/2$ pulse along x prepared an equatorial superposition, the qubit accumulated phase during a free-evolution interval. The final $\pi/2$ pulse along y was orthogonal to the preparation pulse, placing the transition at maximum slope for phase-sensitive readout. CPDD built on XY8 by applying phase changes at defined intervals and repeating the sequence $N$ times. It suppresses amplitude and environmental frequency noise errors and tolerates the microwave-field inhomogeneity found in ensemble measurements. The sequence resonates when $\omega_L=\pi/T_{\mathrm{CPDD}}$ and $\omega_L=\Omega_{\mathrm{CPDD}}$. Arbitrary controls can tailor the frequency response during the sensing interval.

The ensemble sensor used a 100 µm thick NV-doped diamond with top and bottom surfaces oriented along the <111> direction. A bias field of approximately 6.1 mT along <111> split the $m_s=\pm1$ states, and the protocol used the 0 to -1 transition. Laser pulses of about 20 µs polarized the NV spins and provided optical readout. A coplanar waveguide delivered resonant spin control. An 8-turn planar coil terminated in 50 Ω delivered the test field. The default signal was 4 MHz with a 2 µT amplitude unless otherwise specified. The final $\pi/2$ y readout pulse was phase cycled between 0 and 180 degrees to cancel low-frequency charge-state and laser-intensity noise. Green side illumination produced photoluminescence collected from the top with a photodetector.

Calibration used CPDD-XY8 with $N=14$, corresponding to $T_{\mathrm{evo}}=8T_{\mathrm{CPDD}}N=14\ \mu\mathrm{s}$. The coil conversion was experimentally calibrated as $\alpha\approx670\ \mu\mathrm{T}/\sqrt{\mathrm{W}}$. The fitted weak-field photoluminescence slope was 4.85 mV/µT and served as the calibration for later work. A phase sweep from 0 to $2\pi$ gave near one-to-one agreement with small residuals.

The linear approximation applies only when $\xi\cos\phi$ and $\xi\cos(\phi+\pi/2)$ are both much less than one, with $\kappa=1/2$ for CPDD. Here, $\xi=\kappa\gamma_{\mathrm{NV}}R_{\mathrm{test}}T_{\mathrm{evo}}$. The changing-field test used amplitudes of 0.9, 1.6, and 2.1 µT. Its phases were $\pi/4$, $5\pi/8$, and $37\pi/24$ rad, with each setpoint held for 320 ms. Each setpoint included 1000 measurements at 320 µs intervals, and most IQ points lay near their expected values.

Amplitude and phase Allan deviations were calculated from a 10 s trace. The Allan deviation followed $\tau^{-1/2}$ behavior to about 100 ms, then departed at longer times. The authors associated this departure with colored noise, possibly from temperature and background-field drift.

MeasureResultCondition
Temporal resolution320 µs4 MHz test signal
Per-shot amplitude sensitivity78 nTAllan deviation at 320 µs
Per-shot phase sensitivity63 mradAllan deviation at 320 µs
Minimum amplitude Allan deviation4.2 nT$\tau=160$ ms
Minimum phase Allan deviation1.9 mrad$\tau=390$ ms
Normalized amplitude sensitivity1.7 nT/√HzExtracted from the 10 s trace
Normalized phase sensitivity1.2 mrad/√HzExtracted from the 10 s trace

The normalized values are comparable to a Qdyne demonstration at GHz frequencies. The authors describe them as one to three orders of magnitude less sensitive than the cited state-of-the-art MHz implementation using magnetic flux concentrators.

With CPDD fixed at $f_{\mathrm{probe}}=4$ MHz, the test frequency was swept in 10 kHz steps. A separate inset scan used 1 kHz steps. The amplitude peak occurred at $f_{\mathrm{test}}=f_{\mathrm{probe}}$, and its FWHM narrowed as $2f_{\mathrm{probe}}/(8N)$. Beyond the filter bandwidth, spurious harmonics appeared at multiples of $f_{\mathrm{probe}}$. The authors identify randomized global phase changes for each repetition and microwave pulse shaping as possible suppression methods.

Phase sweeps at detunings of 0, 10, 20, 28, and 100 kHz produced elliptical IQ traces, phase rotation, and reversal in the rotation direction. Ellipticity was defined as $\epsilon=(a-b)/a$, where $a$ and $b$ are the major and semiminor axes. The value $1-\epsilon$ equaled one for a circle and zero for an ellipse. The idealized model reproduced most features. The authors say the remaining discrepancies may reflect experimental imperfections and decoherence processes.

Fields above 5 µT were treated as the strong-field regime. Phase sweeps at 7, 12, 17, and 21 µT showed significant distortion and departures from circular IQ traces. The distortion follows phase wrapping because spin-state population must remain between zero and one. The idealized model reproduced this nonlinear behavior.

Adaptive sensing or alternative protocols may partially alleviate the limited dynamic range of slope detection. One proposed source of IQ asymmetry is Rabi detuning caused by hyperfine coupling to the nitrogen nuclear spin. Higher fields are proposed to suppress that effect by polarizing the nitrogen nuclear spin. Other broadening and experimental imperfections remain candidate causes.

The dynamic test cycled through three frequency, amplitude, and phase setpoints, holding each for 1.28 s.

Set pointTest frequencyTest amplitudeTest phase
17.8125 MHz1 µT$23\pi/36$ rad
24 MHz6.25 µT$11\pi/72$ rad
36.25 MHz5.45 µT$49\pi/36$ rad

Each measurement took 320 µs, giving 4000 points during every 1.28 s frequency change. With $f_{\mathrm{probe}}$ fixed at 4 MHz, the measured $I$ and $Q$ components approached zero as $f_{\mathrm{test}}$ moved away. Dynamic tracking changed $\Delta t$ and the CPDD parameters to match $f_{\mathrm{test}}$. Measured amplitude and phase agreed with the setpoints except near each frequency change, attributed to imperfect synchronization. The CPDD repetition count changed among 27, 14, and 22 to hold the filter linewidth near 71 kHz. The authors state that this linewidth can be adjusted to trade temporal resolution against sensitivity.

The 320 µs interval was dominated by laser-related measurement overhead. The same CPDD-XY8 sequence with $N=14$ could reduce the interval by at least a factor of four. With sufficiently high laser intensity, a pulse lasting about 2 µs would include the required post-laser wait. Simultaneously addressing multiple spin transitions with orthogonal control sequences could reduce overhead further.

Increasing $N$ can trade temporal resolution for sensitivity until the NV coherence time is reached. The authors identify longer NV coherence, higher photon collection efficiency, and tailored microwave filter functions as routes to further gains. Device selection must account for NV coherence, microwave-field uniformity, and optical collection efficiency. The CPDD choice addresses microwave-field inhomogeneity, while the readout still depends on photon collection. Probe detuning must also be controlled because it compresses the IQ response and can reduce recovered components toward zero. In the reported implementation, each estimate used a 320 µs measurement cadence and a probe frequency matched to the test field.

View Original Abstract

Nitrogen-vacancy centers in diamond have been shown to be capable of detecting AC magnetic fields with high sensitivity, spectral resolution, and spatial resolution. However, most studies so far have focused on the regime of time-averaged or time-correlated measurements, while little attention has been paid to the single-shot regime. Here we show that the amplitude and phase of an AC field can be retrieved from a single pair of two consecutive measurements. We demonstrate this concept by measuring a 4 MHz AC field with a per-shot amplitude and phase sensitivity of 78 nT and 63 mrad, respectively, at a temporal resolution of 320 us. We also investigate the effects and quantify the errors resulting from probe frequency detunings, as well as operating in the strong field regime. Moreover, we showcase the ability of the measurement protocol to dynamically change the probe frequency in real-time. This work advances the use of NV centers for real-time measurements of AC magnetic fields.