Toward a temperature-insensitive composite diamond clock
Paper details
Section titled “Paper details”| Field | Value |
|---|---|
| Journal | Physical Review Applied |
| Published | 2026-04-21 |
| Authors | Sean Lourette, Andrey Jarmola, Jabir Chathanathil, Victor M. Acosta, A. Glen Birdwell |
| DOI | 10.1103/z2sl-6gcc |
Interleaving the electronic and nuclear spin frequencies of NV centers produced a composite diamond clock signal with fractional instability below 5 × 10⁻⁹ at 200 s and below 1 × 10⁻⁸ at 2 × 10⁵ s. The result came from a 10-day room-temperature test compared with an Rb vapor-cell clock. At 200 s and 2 × 10⁵ s, the composite signal improved on a clock based only on D by factors of 4 and 200, respectively.
Two transitions cancel first-order temperature drift
Section titled “Two transitions cancel first-order temperature drift”The NV electronic ground state provides an electronic zero-field splitting D of about 2.87 GHz. Its intrinsic 14N nuclear spin provides a quadrupole splitting Q of about minus 4.94 MHz. Both frequencies shift with temperature, but their fractional coefficients differ.
| Reference | Value at 297 K | Temperature derivative | Fractional sensitivity at 297 K |
|---|---|---|---|
| D | 2870.3 MHz | −72 kHz/K | −25.3 ppb/mK |
| Q | −4945.9 kHz | 35 Hz/K | −7.17 ppb/mK |
The authors form a weighted sum of the two detunings with α ≈ 1.4. This removes the first-order temperature term while retaining sensitivity to local oscillator drift. The same frequency pair can also provide a temperature estimate through a different weighting.
The measured half-sums are 0.5(f+ + f−) for D and 0.5(f1 + f2) for −Q. A 475 G bias field along the NV axis permits optical polarization and direct fluorescence readout of the 14N nuclear spin. This arrangement lets one diamond defect supply both the frequency reference and a temperature measurement.
TTZFS-8 measures both transition splittings
Section titled “TTZFS-8 measures both transition splittings”Each D and Q measurement uses a two-tone zero-field splitting sequence with a π/2, 2π, π/2 pulse pattern and two free-evolution periods. The sequence converts changes in the frequency half-sum into changes in fluorescence. It is insensitive to changes in the difference between the two addressed transition frequencies.
Optical pumping prepares the electronic state for the D measurement. For Q, optical pumping prepares the nuclear state, and an f1 π pulse transfers it to mI = 0. The authors interleave the two measurements at fixed free-evolution times.
Single-sequence measurements contain unwanted spectral components caused by pulse imperfections across the ensemble. TTZFS-8 combines eight pulse-phase configurations. It suppresses these components by about 60 dB relative to the desired peak.
Fitting the damped oscillations gives T2,D = 1.68(1) μs and T2,Q = 0.881(9) ms. The D sequence uses 50 ns, 200 ns, and 50 ns pulses. The Q sequence uses 15 μs, 60 μs, and 15 μs pulses.
The main sample is a [110]-cut plate from 99.99% carbon-12 diamond with about 4 ppm NV centers. Its reported ensemble coherence time was about 0.8 μs. A 532 nm laser excites a roughly 50 μm spot with about 30 mW of green light. A Halbach array supplies the 475 G field, and two layers of mu-metal shielding suppress external magnetic noise.
RF and microwave signals reach the ensemble through a 160 μm copper wire near the optical focus. A 10 MHz Rb standard stabilizes the RF and microwave synthesizers during the measurements.
Ten-day stability at room temperature
Section titled “Ten-day stability at room temperature”For averaging times below 100 s, the fractional instability scales as τ^-1/2. The authors attribute this behavior mainly to photon shot noise in the individual transition measurements.
At longer times, D and Q follow the common temperature drift, with instability scaling approximately as τ. The composite signal has better stability than either single transition after 200 s. Its Allan deviation reaches a minimum near 200 s and rises slowly at longer times.
The composite measurement suppresses temperature fluctuations by at least one to two orders of magnitude in the long-term comparison. It improves the fractional instability by a factor of 4 at 200 s and a factor of 200 at 2 × 10⁵ s relative to a clock based on D alone.
State-of-the-art microwave atomic clocks achieve fractional stabilities better than 10⁻¹² at 1 s. The composite diamond clock remains roughly two orders of magnitude above its spin-projection-noise limit.
Residual systematics set the next limits
Section titled “Residual systematics set the next limits”Each environmental parameter was varied independently, and its Allan deviation was measured over 200 s. The projected clock contribution combines the parameter’s measured fluctuation with the sensitivity of the composite signal.
| Parameter | 200 s fluctuation | D contribution | Q contribution | Composite contribution |
|---|---|---|---|---|
| Temperature at 294 K | 10 mK | 250 × 10⁻⁹ | 72 × 10⁻⁹ | <5 × 10⁻⁹ |
| Longitudinal field at 475 G | 50 μG | 0.08 × 10⁻⁹ | 0.03 × 10⁻⁹ | 0.06 × 10⁻⁹ |
| RF power at 0 dB | 0.001 dB | 0.4 × 10⁻⁹ | 1.4 × 10⁻⁹ | 2.0 × 10⁻⁹ |
| Laser power at 50 mW | 20 μW | 2.2 × 10⁻⁹ | 0.44 × 10⁻⁹ | 0.44 × 10⁻⁹ |
After temperature cancellation, RF power becomes the largest projected source. The authors do not establish its mechanism. They suspect a distribution of Rabi frequencies across the sensing volume and imperfect pulse durations. More uniform RF delivery and adiabatic pulses could reduce this sensitivity.
The measured transverse field variation was below the 50 μG experimental uncertainty. Transverse shifts can be quadratic in phase, so the authors suggest modulation and feedback. RF and laser power can also produce quadratic amplitude shifts that might be controlled in the same way.
Photon shot noise remains the measured limit below 100 s. Better photon collection, readout contrast, and coherence times could reduce it. Once stability passes 1 × 10⁻⁹, microwave phase noise sampled by the measurement cycle is expected to cause the Dick effect.
Comparison with cryogenic and stabilized operation
Section titled “Comparison with cryogenic and stabilized operation”The authors also measured D alone in a separate temperature-controlled setup. This experiment used a [111]-cut natural-abundance plate with 50 ppb NV centers. A field of 10 G to 20 G, near 15 G, was applied along one NV axis. The π/2 and 2π pulses lasted 70 ns and 280 ns.
They tested external temperature compensation, liquid-nitrogen cooling, and active stabilization to 300 K. All three improved stability, and active stabilization gave the best result.
At 77 K, the temperature sensitivity of D falls by about a factor of 15 to roughly minus 5 kHz/K. Room-temperature and cryogenic data have comparable stability over averaging times of about 10² s to 10³ s.
The authors suggest that a changing temperature difference between the diamond and the cryostat sensor may explain the difference between compensation and active stabilization. The composite clock has better long-term stability because the NV transitions sense the diamond temperature internally. Combining D with Q removes the need to predict the temperature at the sensing sites.
From a measured signal to a clock output
Section titled “From a measured signal to a clock output”The proposed clock output uses a voltage-controlled quartz oscillator at 10 MHz to stabilize the transition synthesizers. Balanced photodetector voltages feed an FPGA, which calculates the composite correction and applies it to the oscillator. This closes a feedback loop around the diamond signal.
The reported experiment instead used RF and microwave synthesizers already stabilized by the Rb standard. It measured the composite signal and its stability rather than demonstrating the complete autonomous oscillator loop.
First-order temperature cancellation does not remove every thermal effect. The paper describes second-order correction terms for use when the temperature drifts by more than 100 mK. A third transition could track A||, whose fractional temperature sensitivity is minus 91 ppb/mK, if additional terms require cancellation.
Absolute accuracy still requires characterization of laser-beam inhomogeneity, strain, charge distribution, and sample-to-sample variation. The authors also call for tests of day-to-day, week-to-week, and month-to-month reproducibility.
The result supports a compact solid-state clock in which the same diamond ensemble can provide timing and temperature metrology. It does not remove the need for magnetic-field control, stable optical and RF delivery, or long-term systematic characterization.
Materials and support
Section titled “Materials and support”The clock’s sensing element is an NV-active [110]-cut diamond plate with about 4 ppm NV centers. 6CCVD’s diamond materials catalog lists MPCVD-grown single-crystal and polycrystalline diamond plates up to 125 mm. The linked engineering support can be used to confirm the cut, carbon-12 content, defect density, and surface finish required before custom fabrication.
View Original Abstract
Frequency references based on solid-state spins promise simplicity, compactness, robustness, multifunctionality, ease of integration, and high densities of emitters. Nitrogen-vacancy (NV) centers in diamond are a natural candidate, but the electronic zero-field splitting ( <a:math xmlns:a=“http://www.w3.org/1998/Math/MathML” display=“inline”> <a:mi>D</a:mi> <a:mo>≈</a:mo> <a:mn>2.87</a:mn> <a:mspace width=“0.2em”/> <a:mi>GHz</a:mi> </a:math> ) exhibits a large fractional temperature dependence (around <d:math xmlns:d=“http://www.w3.org/1998/Math/MathML” display=“inline”> <d:mn>25</d:mn> <d:mspace width=“0.2em”/> <d:mi>ppb</d:mi> <d:mo>/</d:mo> <d:mi>mK</d:mi> </d:math> ), which has precluded its use as a stable clock transition. Here we show that this limitation can be overcome by forming a composite frequency reference that combines measurements of the electronic splitting <g:math xmlns:g=“http://www.w3.org/1998/Math/MathML” display=“inline”> <g:mi>D</g:mi> </g:math> with the nuclear quadrupole splitting ( <i:math xmlns:i=“http://www.w3.org/1998/Math/MathML” display=“inline”> <i:mi>Q</i:mi> <i:mo>≈</i:mo> <i:mn>4.94</i:mn> <i:mspace width=“0.2em”/> <i:mi>MHz</i:mi> </i:math> ) of the <l:math xmlns:l=“http://www.w3.org/1998/Math/MathML” display=“inline”> <l:msup> <l:mi/> <l:mn>14</l:mn> </l:msup> <l:mrow> <l:mrow> <l:mi mathvariant=“normal”>N</l:mi> </l:mrow> </l:mrow> </l:math> nuclear spin intrinsic to the NV center. We further benchmark this composite approach against alternative strategies for mitigating temperature sensitivity, including cryogenic operation, temperature compensation, and active stabilization. By implementing a specially designed pulse sequence with an eight-phase control scheme that suppresses pulse imperfections, we interleave measurements of <o:math xmlns:o=“http://www.w3.org/1998/Math/MathML” display=“inline”> <o:mi>D</o:mi> </o:math> and <q:math xmlns:q=“http://www.w3.org/1998/Math/MathML” display=“inline”> <q:mi>Q</q:mi> </q:math> in a high-density NV ensemble and demonstrate a temperature-compensated composite frequency reference. The stability of this composite diamond clock is characterized over a 10-day period at room temperature through a comparison to a <s:math xmlns:s=“http://www.w3.org/1998/Math/MathML” display=“inline”> <s:mi>Rb</s:mi> </s:math> vapor-cell clock, yielding a fractional instability below <u:math xmlns:u=“http://www.w3.org/1998/Math/MathML” display=“inline”> <u:mn>5</u:mn> <u:mo>×</u:mo> <u:msup> <u:mn>10</u:mn> <u:mrow> <u:mo>−</u:mo> <u:mn>9</u:mn> </u:mrow> </u:msup> </u:math> for an averaging time of <w:math xmlns:w=“http://www.w3.org/1998/Math/MathML” display=“inline”> <w:mi>τ</w:mi> <w:mo>=</w:mo> <w:mn>200</w:mn> <w:mspace width=“0.2em”/> <w:mrow> <w:mi mathvariant=“normal”>s</w:mi> </w:mrow> </w:math> and below <ab:math xmlns:ab=“http://www.w3.org/1998/Math/MathML” display=“inline”> <ab:mn>1</ab:mn> <ab:mo>×</ab:mo> <ab:msup> <ab:mn>10</ab:mn> <ab:mrow> <ab:mo>−</ab:mo> <ab:mn>8</ab:mn> </ab:mrow> </ab:msup> </ab:math> at <cb:math xmlns:cb=“http://www.w3.org/1998/Math/MathML” display=“inline”> <cb:mi>τ</cb:mi> <cb:mo>=</cb:mo> <cb:mn>2</cb:mn> <cb:mo>×</cb:mo> <cb:msup> <cb:mn>10</cb:mn> <cb:mn>5</cb:mn> </cb:msup> <cb:mspace width=“0.2em”/> <cb:mrow> <cb:mi mathvariant=“normal”>s</cb:mi> </cb:mrow> </cb:math> , corresponding to measured improvements by a factor of approximately <gb:math xmlns:gb=“http://www.w3.org/1998/Math/MathML” display=“inline”> <gb:mn>4</gb:mn> </gb:math> and approximately <ib:math xmlns:ib=“http://www.w3.org/1998/Math/MathML” display=“inline”> <ib:mn>200</ib:mn> </ib:math> , respectively, over a clock based purely on the single frequency <kb:math xmlns:kb=“http://www.w3.org/1998/Math/MathML” display=“inline”> <kb:mi>D</kb:mi> </kb:math> for the same periods. By characterizing the residual sensitivity to magnetic fields, optical power, and radio-frequency drive amplitudes, we find that temperature is no longer the dominant source of instability. These results establish complementary electron- and nuclear-spin transitions in diamond as a viable route to thermally robust frequency metrology, providing a pathway toward compact, multifunctional solid-state clocks and quantum sensors.
Single Crystal Cvd Diamond Plate
View Specs →
Boron Doped Single Crystal Diamond Plate
View Specs →
Customized Single Crystal Cvd Diamond Plate
View Specs →References we cover
Section titled “References we cover”- Nuclear Spin Gyroscope based on the Nitrogen Vacancy Center in Diamond
- Robust optical readout and characterization of nuclear spin transitions in nitrogen-vacancy ensembles in diamond
- Sensitivity optimization for NV-diamond magnetometry
Original source
Section titled “Original source”References
Section titled “References”- **** - 2018 European Frequency and Time Forum (EFTF) [Crossref]