Grover Search with Semiconductor Spin Qubits at Ambient Conditions

AI summary

Overview

Research area: Experimental quantum computing hardware and algorithm benchmarking — specifically nuclear spin qubits hosted by nitrogen-vacancy (NV) centers in diamond, operated at room temperature.

Technical level: Intermediate. The paper assumes familiarity with Grover's algorithm, randomized benchmarking, and spin-qubit control, but the experimental narrative is accessible to readers with a graduate-level quantum information background.

Scope: The paper reports a three-qubit Grover search executed on a commercially built, ambient-condition diamond quantum computer, together with the full calibration and benchmarking of its native universal gate set.

What This Paper Is About

Most quantum processors that run meaningful algorithms need either ultralow temperatures in vacuum (superconducting circuits, silicon spin qubits) or ultra-high vacuum with laser cooling (trapped ions, neutral atoms). This paper asks whether a solid-state spin system — the electronic and nuclear spins of an NV center in diamond — can run a genuine multi-qubit quantum algorithm with competitive success probability while sitting on a bench in air at roughly 296 K. The authors implement Grover's search over a search space of N = 8 using three nuclear spins as the computational qubits, and they benchmark every native gate so that the algorithmic result can be traced back to component-level performance.

Key Contributions

1. A three-qubit Grover search on a commercial room-temperature diamond quantum computer. The device is installed on the premises of Fraunhofer IWU. The authors describe it as the first demonstration of a search of this size with NV center technology, and state it is the highest quantum search performance achieved at ambient conditions for any quantum device to date.

2. An all-nuclear-qubit architecture. Unlike earlier NV work, the electronic spin is not used as a computational qubit. It serves only as an optical readout interface, as the mediator for the nuclear entangling operation, and as the intermediary for nuclear spin polarization. The computational register is the intrinsic ¹⁴N spin (I = 1, with only a two-level subspace used) plus two strongly coupled ¹³C spins (I = 1/2).

3. A calibrated universal native gate set. Single-qubit gates are radiofrequency pulses combined with virtual z-rotations; the entangling resource is a multi-qubit phase gate Phase₀₀₀ = 𝟙 − 2|000⟩⟨000|, locally equivalent to a Toffoli gate, realized through a geometric phase from selective microwave pulses on the electronic spin.

4. A cross-modality benchmark table. The paper compiles published Grover results for neutral atoms, liquid-state NMR, superconducting devices, trapped ions, photonic platforms, silicon spin qubits, and NV centers, listing search-space size, Grover repetitions, success probability, temperature, and cryogenic/vacuum requirements.

Main Findings

• Single marked state: With N = 8 and r = 2 Grover steps, the average algorithm success probability (ASP) across all eight possible target states is (77.3 ± 3.4) %, against an ideal maximum detection probability of 94.53 %. The average squared statistical overlap (SSO) with the ideal distribution is (90.6 ± 2.6) %.

• Two marked states: With r = 1 step, the average ASP is (87.0 ± 4.2) % and the mean SSO is (86.9 ± 4.2) %.

• Classical comparison: The maximum classical success probability is 37.5 % for two oracle queries and 46.4 % for finding one of two marked states with a single query — both far below the measured quantum values.

• Single-qubit gate fidelities: Randomized benchmarking (20 random circuits per Clifford-sequence length) gives average Clifford fidelity 99.90 % and average Sₓ fidelity 99.95 %. Per qubit, Clifford fidelities are (99.82 ± 0.07) % for ¹⁴N, (99.91 ± 0.08) % for ¹³C₁, and (99.96 ± 0.03) % for ¹³C₂; Sₓ fidelities are (99.91 ± 0.04) %, (99.95 ± 0.04) %, and (99.98 ± 0.02) % respectively — the 99.98 % figure being the highest fidelity quoted in the abstract.

• Entangling gate fidelities: Using a modified interleaved randomized benchmarking protocol with single-qubit Clifford twirls, two-qubit-subspace fidelities for the Phase₀₀₀ gate are (96.0 ± 0.3) % for q0,q1, (96.5 ± 0.3) % for q0,q2, and (94.8 ± 0.3) % for q1,q2 — 95.7 % on average. The gate time is 810 ns.

• Coherence: Ramsey coherence times of the nuclear spins are approximately 4.3 ms (¹⁴N), 1.5 ms (¹³C₁), and 1.4 ms (¹³C₂) — the paper notes nuclear T₂* of roughly 1–10 ms is about three orders of magnitude larger than typical electronic spin coherence times.

• Operating conditions: Average sample temperature T ≈ 296.3 K with thermal excursions of ΔT ≈ 0.8 K during measurements; no cryogenics and no vacuum equipment. Total operational power consumption is roughly 600 W.

• Comparison to the best cryogenic result: The authors explicitly state their result is "not entirely within striking distance" of the best average ASP at this size, 89.4 %, obtained with silicon spin qubits at 15 mK and with readout post-selection. Their own Table 1 also lists a superconducting result of 78 % at 20 mK (only one target state) and a trapped-ion result of 69 % at 6 K.

• Device and materials details: A type IIa, (100)-oriented single-crystal diamond substrate (ElementSix) with natural 1.1 % ¹³C abundance and less than 1 ppb of single nitrogen; sulphur implantation and activation were used to increase NV yield and stability. A bias field of approximately 51.6 mT along the NV symmetry axis enables selective addressing and nuclear polarization. A green laser (OBIS 514 nm LX 30, Coherent) in a confocal setup with NA = 0.95 in air drives initialization.

• Calibration precision: Nuclear frequencies are 2.940059(30) MHz, 5.990373(69) MHz, and 13.314576(63) MHz, with Rabi frequencies of 30.96(2) kHz, 52.33(8) kHz, and 108.61(6) kHz. Deterministic-benchmarking calibration pins pulse frequencies to about 60 Hz, on the order of thermal fluctuation of strongly coupled nuclear resonance frequencies.

• Not reported: The paper does not present full process tomography (it states this explicitly and chooses randomized benchmarking instead), does not demonstrate error correction, and does not extend beyond N = 8.

Methodology in Plain English

The qubits are atomic nuclei trapped inside a diamond crystal. The nitrogen atom at the center of the defect and two nearby carbon-13 atoms each carry a nuclear spin, and each can be flipped with radiofrequency pulses. Because nuclei couple only weakly to their environment, they hold their quantum state for milliseconds — long enough to run a short algorithm.

The electron spin of the NV center does the support work. Green laser light pushes it into a well-defined state, and a carefully tuned magnetic field of about 51.6 mT lets the researchers address each nuclear spin's transition individually and transfer polarization from the electron to the nuclei, raising the readout contrast by about 10 % to a total of 21.4 %. Microwave pulses on the electron create the conditional phase that entangles the nuclear qubits; the team shaped that microwave pulse using numerical simulations with hyperfine tensors, bringing the gate down to 810 ns. Residual unwanted phases are removed with virtual z-rotations.

Single-qubit gates come from radiofrequency pulses; the entangling Phase₀₀₀ gate, combined with those, makes the gate set universal. Each gate's average error rate was measured with randomized benchmarking, which cancels out state-preparation and measurement errors. For the three-qubit gate, the team used a modified interleaved protocol with single-qubit Clifford twirls because standard benchmarking becomes resource-prohibitive at larger gate dimensions.

Grover's algorithm itself starts from an equal superposition of all eight three-bit strings via three Hadamard gates. The oracle marking the target state is built from one Phase₀₀₀ gate framed by local Pauli operators; the amplification step is the reflection R = 𝟙 − 2|ψ⟩⟨ψ|, implemented as H^⊗3 Phase₀₀₀ H^⊗3. A single-target search needs two Grover iterations, hence four Phase₀₀₀ gates, while a two-target search succeeds with one iteration. Readout interrogates all 2³ electron transitions and converts the measured fluorescence into probabilities using a pre-measured matrix and a non-negative least-squares fit. To combat drift in an uncontrolled room, automatic correction procedures compensate for externally induced system shifts.

Why This Matters

Impact on research. The work pushes NV-based quantum information processing from two-qubit algorithmic demonstrations to a three-qubit search that beats classical bounds by a wide margin, and it frames the result against a literature table spanning seven hardware modalities. It also argues that purely nuclear registers — exploiting the long coherence of hyperfine-coupled spins rather than the optically active electron — are a viable computational architecture, and it provides a reusable calibration chain (ENDOR, Rabi, deterministic benchmarking, simulation-informed pulse shaping) for others building similar devices.

Real-world applications (implications, not demonstrations in this paper):

• Deployable quantum computing hardware for industrial or factory-floor settings where cryogenics and vacuum infrastructure are impractical.
• Compact, low-power quantum processors — roughly 600 W total consumption — for edge or embedded deployment.
• Quantum sensing and metrology, building on the same NV platform's ambient-condition optical readout.
• Materials and semiconductor process development, including ion implantation and doping recipes used here to raise NV yield and stability.

Industry relevance. The device is commercial (SAXON Q GmbH) and installed at a Fraunhofer institute, and the work is co-financed by the Saxon State Parliament through the SAX-QT project. That framing — an on-premises, room-temperature machine with published gate fidelities — is directly relevant to vendors and integrators arguing that solid-state spin hardware can leave the dilution refrigerator.

Future Directions

• Better characterization of the Phase₀₀₀ gate. The authors point to newer benchmarking approaches that address the full dimensionality of the gate, beyond the two-qubit-subspace IRB used here.
• Higher-fidelity control through improved pulse generation. Pulse quality is currently limited by the finite amplitude resolution of the arbitrary waveform generator; simulations suggest fidelities of at least 99 % with more precise pulse generation.
• Scaling the register. The paper presents N = 8 as a first demonstration of this size for NV technology, leaving open how the approach extends to more nuclear spins and longer algorithms.
• Closing the gap to cryogenic platforms. The explicitly acknowledged 89.4 % cryogenic result with readout post-selection sets a target for ambient-condition devices, alongside the open question of whether post-selection or error mitigation can be applied here.

Target Audience

Experimental quantum computing researchers working on spin qubits, NV centers, or solid-state quantum hardware; engineers building commercial or field-deployable quantum devices; benchmarking and characterization specialists interested in randomized and interleaved benchmarking protocols; and industrial or program-level decision makers evaluating room-temperature quantum hardware against cryogenic alternatives. Readers wanting a rigorous, numbers-first account of what an ambient-condition quantum processor can currently do will get the most from it.

Authors’ abstract

Grover's algorithm is executed on a commercial quantum computer based on nitrogen-vacancy centers in diamond operating under ambient conditions, achieving fidelities up to $99.98\,\%$. Within a $N=8$ search space of three solid-state nuclear spin qubits, the measured success probabilities of finding a single or two marked states are $(77.3 \pm 3.4)\,\%$ and $(87.0 \pm 4.2)\,\%$, respectively. These values surpass published results for superconducting qubits or any quantum computer operating at room temperature. In addition, the paper also details the calibrated fidelities of the implemented universal gate set.