Experimental High-Accuracy and Broadband Quantum Frequency Sensing via Geodesic Control

Abstract: Accurate frequency estimation of oscillating signals over a broad bandwidth is a central task in quantum sensing, yet it is often compromised by spurious responses to higher-order harmonics in realistic multi-frequency environments. Here we experimentally demonstrate a high-accuracy and broadband quantum frequency sensing protocol based on geodesic control, implemented using the electron spin of a single nitrogen-vacancy center in diamond. By engineering an intrinsically single-frequency response, geodesic control enables bias-free frequency estimation with strong suppression of harmonic-induced systematic errors across a wide spectral range spanning from the megahertz to the gigahertz regime. Furthermore, by incorporating synchronized readout, we achieve millihertz-level frequency resolution under noisy signal conditions. Our results provide systematic experimental benchmarking of geodesic control for quantum frequency sensing and establish it as a practical approach for high-accuracy metrology in realistic environments.

• The paper experimentally validates a geodesic control protocol using single NV centers in diamond that boosts accuracy in estimating AC signals by sufficiently suppressing higher-order harmonics, unlike conventional DD methods.
• The described approach achieves millihertz-level frequency resolution by applying synchronized readout techniques and demonstrates robust performance against high-frequency noises, enhancing its potential for sensitive quantum sensing and signal processing.
• Geodesic control remains compatible with existing quantum hardware and architectures and results in higher fidelity, eliminating concerns in bias-free frequency estimation even when spurious noise exceeds signal amplitude up to twice.

Context and motivation

Frequency estimation of AC signals is a core task in quantum sensing, with applications spanning nanoscale NMR, microwave photon detection, and wireless communications. The dominant approach uses dynamical decoupling (DD) sequences such as CPMG and XY families to impart frequency selectivity to the sensor while suppressing broadband noise. A well-known deficiency of these sequences is their intrinsic sensitivity to higher-order odd harmonics at $\omega_n = k\omega_s$ ($k = 3, 5, \ldots$), arising from the square-wave modulation function of the toggling frame. In realistic multi-frequency environments these spurious harmonic responses produce systematic biases in frequency estimation that cannot be removed by post-processing. Theoretical work had proposed geodesic control—shaping the control trajectory so the modulation function is intrinsically single-frequency—as a route to eliminating harmonic responses, but its experimental feasibility and compatibility with broadband, high-resolution techniques remained untested.

Geodesic sensing protocol

The experiment uses the electron spin of a single nitrogen–vacancy (NV) center in diamond, operated in the $\{\ket{m_s=0}, \ket{m_s=-1}\}$ subspace with splitting $\omega_0 \approx 2\pi\times1.47$ GHz under a static field $B_0 \approx 500$ Gs. For MHz-regime signals parallel to the NV axis, the GD$_\parallel$ sequence applies $N_s$ repetitions of a geodesic pulse block containing $N$ $\pi$ pulses whose rotation axes rotate continuously in the $x$–$k = 3, 5, \ldots$0 plane, with pulse phase $k = 3, 5, \ldots$1. In the interaction picture of this control, the effective sensing Hamiltonian is $k = 3, 5, \ldots$2, where for large $k = 3, 5, \ldots$3 the modulation function approaches the pure sinusoid $k = 3, 5, \ldots$4 rather than the square wave of conventional DD. The accumulated phase therefore builds constructively only when $k = 3, 5, \ldots$5, yielding an intrinsically single-frequency filter.

To access GHz-regime signals without requiring sub-50 ns pulses, the authors combine geodesic control with heterodyne detection: the transverse field component $k = 3, 5, \ldots$6 is down-converted to detunings $k = 3, 5, \ldots$7 in the rotating frame, and a GD$k = 3, 5, \ldots$8 sequence with axes rotating in the $k = 3, 5, \ldots$9–$\{\ket{m_s=0}, \ket{m_s=-1}\}$0 plane senses $\{\ket{m_s=0}, \ket{m_s=-1}\}$1. Both variants use identical hardware parameters: Rabi frequency $\{\ket{m_s=0}, \ket{m_s=-1}\}$2 MHz, pulse duration $\{\ket{m_s=0}, \ket{m_s=-1}\}$3 ns, $\{\ket{m_s=0}, \ket{m_s=-1}\}$4 (parallel) or $\{\ket{m_s=0}, \ket{m_s=-1}\}$5 (perpendicular), and $\{\ket{m_s=0}, \ket{m_s=-1}\}$6 pulses per block.

Spectral characterization and harmonic suppression

The modulation spectra were reconstructed by applying controlled random-phase test signals and averaging accumulated phases over uniformly distributed initial phases, exploiting the wide-sense-stationary equivalence to extract the filter-function magnitude $\{\ket{m_s=0}, \ket{m_s=-1}\}$7. With scan frequency fixed at $\{\ket{m_s=0}, \ket{m_s=-1}\}$8 MHz, the reconstructed spectrum of GD$\{\ket{m_s=0}, \ket{m_s=-1}\}$9 shows a dominant response at the fundamental while suppressing higher harmonics; the XY sequence at the same scan frequency exhibits clear residual responses at odd harmonics. The same contrast holds between GD$\omega_0 \approx 2\pi\times1.47$0 and CPMG.

Robustness was quantified directly by adding noise tones at the 3rd, 5th, and 7th harmonics with amplitudes up to $\omega_0 \approx 2\pi\times1.47$1 kHz. Under GD control the final-state fidelity remains above 0.9 across the full amplitude range, whereas XY and CPMG fidelities degrade rapidly; lower-order harmonics cause stronger degradation, consistent with the reconstructed filter functions. This establishes that geodesic control suppresses harmonic-induced errors through the geometry of the evolution itself, not through filtering or post-processing.

Bias-free frequency estimation

In two-tone experiments, GD$\omega_0 \approx 2\pi\times1.47$2 with a target at $\omega_0 \approx 2\pi\times1.47$3 MHz and noise components at $\omega_0 \approx 2\pi\times1.47$4 MHz and $\omega_0 \approx 2\pi\times1.47$5 MHz (noise amplitudes twice the signal amplitude) produces a single resonance dip at the target frequency, while XY yields three dips with biases up to $\omega_0 \approx 2\pi\times1.47$6 kHz, preventing unambiguous identification of $\omega_0 \approx 2\pi\times1.47$7. Similarly, GD$\omega_0 \approx 2\pi\times1.47$8 shows only the target dip where CPMG exhibits biases of $\omega_0 \approx 2\pi\times1.47$9 kHz and $B_0 \approx 500$0 kHz. These results demonstrate bias-free estimation even when spurious noise exceeds the target amplitude—a regime where conventional DD fails qualitatively, not merely quantitatively.

Broadband operation and millihertz resolution

Combining GD$B_0 \approx 500$1 with heterodyne detection extends sensing into the GHz regime, since only the low-frequency detuning must be resolved by the control sequence. Resolution was further improved via synchronized readout, in which repeated sensing procedures sample slices of a continuous signal at intervals $B_0 \approx 500$2 (71 μs for the parallel schemes over 23 minutes; 31 μs for the perpendicular schemes over 10 minutes), followed by Fourier transformation of the photon-count time trace. Frequency resolution improves from kHz-level (half-widths of $B_0 \approx 500$3 kHz and $B_0 \approx 500$4 kHz for GD$B_0 \approx 500$5 and GD$B_0 \approx 500$6) to 1 mHz and 2 mHz, respectively. Notably, synchronized readout applied to XY also sharpens peaks, but harmonic responses generate multiple spurious peaks whose amplitudes exceed the true signal peak, degrading the signal-to-noise ratio and complicating identification—so the accuracy advantage of geodesic control persists at high resolution.

Limitations and open questions

The paper concedes several constraints. The single-frequency response is exact only in the limit of large $B_0 \approx 500$7 and instantaneous pulses; finite pulse widths ($B_0 \approx 500$8 ns) leave residual spectral leakage visible in the reconstructed filter functions, and the available microwave power bounds how far direct scanning could extend without heterodyne assistance. The demonstration uses engineered, well-separated noise tones rather than dense stochastic noise backgrounds, so performance under realistic broadband noise spectra remains to be characterized. The millihertz resolution relies on long acquisition times (10–23 minutes) and assumes strict phase coherence of the target across the trace, which may not hold for free-running sources. Finally, the analysis neglects fast-oscillating counter-rotating terms under conditions such as $B_0 \approx 500$9 and $_\parallel$0, which bound the dynamic range of the protocol.

Conclusion

This work provides systematic experimental benchmarking of geodesic control for quantum frequency sensing on a single NV center. By engineering an intrinsically single-frequency modulation function, the protocol achieves bias-free frequency estimation with strong suppression of third-, fifth-, and seventh-harmonic noise, extends operation into the GHz regime through integration with heterodyne detection, and reaches millihertz-level resolution via synchronized readout—all using standard NV control hardware. The approach is compatible with existing multi-sensor and array-based architectures, though its behavior under dense stochastic noise and free-running signals remains an open question for future characterization.