Skip to content

Readout of a solid state spin ensemble at the projection noise limit

FieldValue
JournalNature Communications
Published2026-05-04
AuthorsRouven Maier, Cheng-I Ho, Andrej Denisenko, Marina Davydova, Peter Knittel
DOI10.1038/s41467-026-72721-0
Citations1

A repetitive nuclear-assisted readout moved a mesoscopic ensemble of ¹⁴N nitrogen-vacancy (NV) centers out of the photon shot-noise-dominated regime. At high photon counts, photon shot noise was 3.76(1) dB below the observed projection noise. The fitted model included spin relaxation during repeated measurements and gave a 3.8 dB reduction below the thermal projection-noise level. The readout exposed the ensemble’s spin distribution width as well as its mean state.

The experiment used a home-built confocal microscope at room temperature. The sample was a (111)-cut diamond chip with as-grown, delta-doped, preferentially aligned thin ensembles of ¹⁴N NV centers. A 2.7 T bias field was aligned with the principal NV axis. Electron spins were driven at 73 GHz by a lambda/2 bow-tie resonator. Nuclear spins were controlled with radio frequency delivered by a 50 µm copper wire. An avalanche photodiode detected the emitted photons.

The effective longitudinal relaxation time of the nuclear spins increases quadratically with magnetic field. At 2.7 T, the experiment could perform more than about 4,000 repetitive readouts, enough to cross into the projection-noise regime. The effective T1 still set the maximum repetition count.

Each readout mapped the nitrogen nuclear hyperfine levels onto electron spin eigenstates with a spin-selective narrowband microwave π pulse. A laser pulse then read out and reinitialized the electron spin. Counts a and b from the two projected states were subtracted, producing a Skellam photon distribution. Two reference readouts were inserted between spin measurements to estimate the baseline and remove slow laser-power drift.

The observed variance was modeled as

[ \sigma^2 = \sigma_n^2 + (2nc\tilde{\sigma}_{J_z})^2. ]

Here, σn is photon shot noise, c is optical contrast, n is the average detected photon number, and σ̃Jz is spin projection noise. The nitrogen state remained almost undisturbed during optical readout, so repeated projections sampled the same nuclear state without directly optically resetting it.

Because |0⟩ and |−1⟩ map to the same electron state, the three nuclear levels were treated as a pseudo spin-1/2 system. The analysis also included residual nuclear relaxation during the extended readout.

The detected photon number was increased by raising the repetition count from 1,250 to 25,000. At low n, the normalized standard deviation followed photon shot-noise scaling proportional to (1/\sqrt{n}). At high n, the curve levelled as spin projection noise became dominant. Photon shot noise at the crossover was 3.76(1) dB below the observed projection noise.

The 3.76(1) dB value compares two contributions at the same high photon count. The separate 3.8 dB figure compares the relaxed projection noise with the thermal level.

Both photon shot noise and thermal spin projection noise decrease as (1/\sqrt{N_{NV}}). The authors therefore place the crossover at the same normalized photon count for different ensemble sizes. Fitting yielded (N_{NV}=31(3)) active NV centers in the confocal spot. Three density spots showed a linear correlation with observed fluorescence.

After initialization into |+1⟩, resonant radio frequency produced long-lived Rabi oscillations in the nuclear ensemble. The mean state remained below 0.5 because initialization left population in |0⟩ and |−1⟩. Limited electron initialization fidelity also prevented complete inversion because the radio-frequency transition required correctly prepared NV- centers. During the oscillations, projection-noise readout tracked the distribution width as well as the mean. The width followed the expected dependence on eigenstate readout probability and peaked when the ensemble crossed the equator.

The readout then distinguished spatially correlated noise from uncorrelated phonon-induced relaxation. In T1 relaxometry, both sources caused similar decay of the mean spin, but their projection widths differed. The correlated source combined ten fully randomized frequencies within a 3 MHz band around the electron resonance. The thermal, spatially uncorrelated distribution was significantly narrower.

A second experiment used noise spectroscopy to detect a stochastic-phase field at 250 kHz. Its amplitude at the NV center was 1.84(7) µT. An XY8 dynamical decoupling sequence and different final π/2 pulses reconstructed marginal distributions along the X, Y, and Z axes. The result showed delocalization in the equatorial plane. This width measurement exposed common environmental noise that mean spin decay alone did not resolve.

The sensing benefit was calculated from experimental parameters, not demonstrated in a complete protocol. The model compared conventional single electron-spin readout with the repetitive nuclear-assisted readout. Its figure of merit, η, included effective contrast, reduced noise, and initialization and readout overhead.

The calculation predicted improvement for sensing times above 15 µs, which the authors describe as typical for T2 protocols. The plotted example showed an 8.9 dB sensitivity improvement at its optimum. The optimum was about 2,700 readouts for sensing times below 10,000 µs. With more readouts, spin relaxation limited noise reduction while added overhead reduced sensitivity.

Under the assumed quadratic field dependence, about 2,700 readouts could fit within the effective T1 for fields above 0.5 T. The model predicted several orders of magnitude of improvement for very long measurement sequences.

The calculation also considered spin squeezing up to 4 dB. A previously demonstrated solid-state ensemble value of η²=0.5 dB led to only 0.13 dB improvement in this model. The 4 dB case was therefore a modeled opportunity, not a squeezing result from this experiment.

Experimental limits and device implications

Section titled “Experimental limits and device implications”

The main experimental limit was the effective nuclear T1. It capped the repetition count and tied the usable readout sequence to magnetic field. The authors call the nuclear readout quantum non-demolition because the nitrogen state remained almost undisturbed. Repeated projections still produced residual polarization decay, which reduced the measured projection variance relative to an unperturbed thermal ensemble. Their variance model included this relaxation.

At high photon counts, the avalanche photodiode left its linear regime and artificially narrowed the detected distribution. The model used a brightness-dependent correction factor k of about 0.89 to 0.99. Initialization infidelity prevented full population inversion. Faster decay of |0⟩ than |+1⟩ during readout caused an incomplete dip at inversion.

The proposed widefield correlation map and higher-order cumulants for many-body states were not demonstrated. The reported experiment used local confocal readout on one diamond chip. For device design, field, T1, and readout overhead must be selected together. The model places the useful optimum near 2,700 readouts for sensing times below 10,000 µs. The demonstrated hardware required a 2.7 T field, 73 GHz microwave control, radio-frequency nuclear control, optical fluorescence collection, and photodiode detection.

For this quantum-sensing application, the material requirement is a (111)-cut diamond chip containing as-grown delta-doped, preferentially aligned ¹⁴N NV ensembles. 6CCVD’s diamond materials catalog lists MPCVD-grown single-crystal and polycrystalline plates and wafers up to 125 mm, with thickness from 0.1 to 500 micrometers for both material types and substrates up to 10 mm, plus global shipping. The paper’s orientation and NV ensemble requirements still need to be reconciled with those catalog options through the linked engineering support.

View Original Abstract

Abstract Spin ensembles are central to quantum science, from fundamental physics searches to magnetic resonance spectroscopy and quantum sensing. The standard quantum limit for their performance is ultimately posed by spin projection noise, yet solid-state implementations have so far been limited by significantly larger photon shot noise. Here, we demonstrate a direct quantum non-demolition readout of a mesoscopic ensemble of nitrogen-vacancy (NV) centers in diamond that surpasses the photon shot-noise limit and approaches the intrinsic spin projection noise. By stabilizing the intrinsic 14 N nuclear spin bath at high magnetic fields and employing a repetitive nuclear-assisted spin readout, we achieve a noise reduction of 3.8 dB below the thermal projection noise level. This enables direct access to the intrinsic fluctuations of the spin ensemble, allowing us to directly observe the signatures of correlated spin states. Our results establish projection noise-limited readout as a practical tool for solid-state quantum sensors, opening pathways to quantum-enhanced metrology, direct detection of many-body correlations, and the implementation of spin squeezing in mesoscopic solid-state ensembles.