Approaching the Limit in Multiparameter AC Magnetometry with Quantum Control
Paper details
Section titled “Paper details”| Field | Value |
|---|---|
| Journal | Open MIND |
| Published | 2026-02-19 |
| Authors | Takuya Isogawa, Zhiyao Hu, Ayumi Kanamoto, Nutdech Phadetsuwannukun, Shilin Wang |
| DOI | 10.48550/arxiv.2602.17648 |
Quantum control removed a fundamental obstruction to simultaneous AC magnetometry and brought both parameter estimates close to their separate quantum limits. A nitrogen-vacancy center in diamond measured a microwave field’s amplitude and frequency at room temperature. The theoretical standard deviations were (4/\pi \approx 1.27) times the single-parameter benchmarks. The experiment also recovered (\delta B \propto N^{-1}) and (\delta\omega \propto N^{-2}). These results show that the quantum Fisher information matrix need not remain singular when parallel generators can be reshaped.
Why parallel generators prevent joint estimation
Section titled “Why parallel generators prevent joint estimation”Linearly polarized AC sensing is encoded by (H_0 = \gamma B \cos(\omega t)\sigma_x). The amplitude derivative is (\gamma \cos(\omega t)\sigma_x). The frequency derivative is (-\gamma Bt \sin(\omega t)\sigma_x). Both derivatives remain parallel to (\sigma_x). Their Heisenberg-picture generators therefore commute. The QFIM is singular for any probe state, including entangled states. Its inverse does not exist, so the joint quantum Cramér-Rao bound cannot be formed. This differs from ordinary multiparameter incompatibility, where the bound exists but may be unattainable. Measuring amplitude and frequency sequentially avoids the singular joint matrix, but it divides the measurement budget. Earlier control work mainly maximized generator variance while preserving its direction. The new strategy changes the direction of time evolution so the two parameters occupy different information directions.
Control turns a singular matrix into a diagonal one
Section titled “Control turns a singular matrix into a diagonal one”The protocol adds a control field (H_c = \gamma B_c \cos(\omega_c t)\sigma_x). The control parameters start from prior information and can be refined adaptively. In the ideal local-estimation case, (B_c = B) and (\omega_c = \omega). The resulting evolution is (U = \exp[-i\omega\sigma_z t/2]). When (\omega T \gg 2\pi), the amplitude and frequency generators become orthogonal. Their commutator and generator overlap both vanish. A maximally entangled Bell probe then produces a diagonal QFIM. The amplitude QFI scales as (T^2) and the frequency QFI as (T^4). The corresponding standard deviations scale as (T^{-1}) and (T^{-2}). Deviations from the long-time limit decay as ((\omega T)^{-1}). The control therefore changes generator geometry while retaining the optimal time scalings.
A nitrogen-vacancy center implements the protocol
Section titled “A nitrogen-vacancy center implements the protocol”The sensor used the electronic-nuclear spin system of a (^{14})NV center. The electron spin senses the microwave field, while the nitrogen nuclear spin serves as an ancilla. The sequence has three stages: Bell-state preparation, sensing, and Bell-state measurement. A nuclear RF (\pi/2) pulse and a selective microwave (\pi) pulse prepare (|\Phi^+\rangle) from (|0,+1\rangle). Reversing the entangling operations maps the Bell populations to electron-spin fluorescence. Averaged Bell-basis readout does not require single-shot detection.
Room temperature and a 357 G static bias along the NV axis defined the operating conditions. The control frequency followed (D - \gamma_e B_{z0} - A/2). The stated constants were (D = (2\pi)\times 2.87) GHz, (A = -(2\pi)\times 2.16) MHz, and (\gamma_e = (2\pi)\times 2.8) MHz/G. The control field was set to (B_c = 5.65) G and (\omega_c = (2\pi)\times 1870) MHz.
An always-on hyperfine interaction mixes the sensor and ancilla. The sequence alternates target and control intervals and applies electron (\pi) pulses for dynamical decoupling. One block lasts (\tau), and (N) blocks give the nominal interrogation time (T = N\tau). The practical field-switching arrangement doubles the nominal sequence duration to (2T). A measurement-basis rotation gives equal Bell outcomes of (1/4) and avoids a flat optimum. This choice increases the signal derivatives used for parameter estimation. Photon shot noise limits the readout, which uses averaged fluorescence rather than single-shot detection.
Both estimates improve at their expected rates
Section titled “Both estimates improve at their expected rates”Signal calibration varied target amplitude or frequency while holding the other parameter at the control value. The authors compared (N = 1) with (N = 8). Each plotted point used (n = 3\times 10^6) averages of complete sequences. At (N = 8), both signal gradients became steeper near the control setting. The steeper response produced smaller joint parameter uncertainties.
Linear fits to the signal slopes supplied the Jacobian and parameter covariance matrix. Simulation and experiment showed (\delta B \propto N^{-1}) and (\delta\omega \propto N^{-2}). The fitted simulation exponents were 1.00 and 2.00. Because (T = N\tau), these results match the predicted (T^{-1}) amplitude scaling and (T^{-2}) frequency scaling. The experiment therefore recovered both optimal trends at the same time. Although three Bell populations were available, the analysis used two signals because simulated sensitivities were nearly identical.
Comparison with single-parameter and sequential bounds
Section titled “Comparison with single-parameter and sequential bounds”The supplemental benchmark compares simultaneous estimation with separate optimal amplitude and frequency measurements in the long-time regime. The reported ratio is a constant-factor cost from sharing one probe between both parameters. It does not change the asymptotic time-scaling orders. The simultaneous protocol therefore recovers the same precision class while producing both estimates from every run.
In a sequential baseline, half of (M) repetitions estimate (B) and half estimate (\omega). The simultaneous protocol uses all (M) repetitions for both parameters. The paper reports a modest advantage at fixed total interrogation-time resources, despite lower per-shot QFI than the separate optima. This is a theoretical comparison, not a direct sequential experiment.
Practical limits remain in pulse control and finite time
Section titled “Practical limits remain in pulse control and finite time”The orthogonality result applies when (\omega T \gg 2\pi). At shorter interrogation times, off-diagonal QFIM elements remain and the joint estimate is not at its asymptotic optimum. The ideal control also assumes prior estimates close to the target values. The reported experiment used fixed control settings near the target. Adaptive refinement is proposed as a way to approach that ideal condition.
Practical microwave (\pi) pulses are imperfect because of hyperfine coupling. Their errors accumulate with pulse count and cause oscillations in sensitivity versus (N). Composite pulses could mitigate these effects and produce near-ideal (\pi) pulses. Averaged fluorescence readout also excludes single-shot operation, although it permits Bell-basis population estimation. The demonstration covers a linearly polarized AC field with one NV center at room temperature. Future work targets more general multiparameter sensing and methods for addressing QFIM singularities.
Engineering consequences for NV magnetometry
Section titled “Engineering consequences for NV magnetometry”For a sensor designer, the immediate requirement is a diamond host containing a usable NV center. The paper does not specify growth method, defect concentration, dimensions, or metallization, so those choices cannot be inferred from the experiment. The demonstrated operating point is room temperature, a 357 G axial bias, and a microwave control near 1870 MHz.
Implementation also requires calibrated control parameters, alternating field intervals, and controlled electron (\pi)-pulse errors as (N) grows. The method is most useful when prior estimates locate the target near the control point and (\omega T \gg 2\pi) is attainable. It replaces separate amplitude and frequency acquisition with one joint estimate at the same asymptotic scaling orders.
Materials and support
Section titled “Materials and support”The experiment needs a diamond host with an NV center for room-temperature microwave magnetometry. 6CCVD’s diamond materials catalog lists MPCVD-grown single-crystal diamond, which matches the host material family but does not specify NV creation. The engineering support page provides custom fabrication support for discussing the diamond format and any required Au, Pt, Pd, Ti, W, or Cu metallization.
View Original Abstract
Simultaneously estimating multiple parameters at the ultimate limit is a central challenge in quantum metrology, often hindered by inherent incompatibilities in optimal estimation strategies. At its most extreme, this incompatibility culminates in a fundamental impossibility when the quantum Fisher information matrix (QFIM) becomes singular, rendering joint estimation unattainable. This is the case for a canonical problem: estimating the amplitude and frequency of an AC magnetic field, where the generators are parallel to each other. Here, we introduce a quantum control protocol that resolves this singularity. Our control protocol strategically engineers the sensor’s time evolution so the generators for the two parameters become orthogonal. It not only removes the singularity but also restores the optimal scaling of precision with interrogation time for both parameters simultaneously. We experimentally validate this protocol using a nitrogen-vacancy center in diamond at room temperature, demonstrating the concurrent achievement of the optimal scaling for both parameters under realistic conditions.
High Concentration Quantum Grade Single Crystal Diamond
View Specs →Products
View Specs →