Skip to content

Quantum gates via dynamical decoupling of a central qubit on IBMQ and N 15 -vacancy centers in diamond

FieldValue
JournalPhysical Review A
Published2026-03-09
AuthorsLucas Tsunaki, Michael Dotan, Kseniia Volkova, Boris Naydenov
DOI10.1103/6st7-qh12

A 24-pulse XYN sequence produced a simulated $^{15}$N population inversion in 9.04 µs, with pseudo-fidelity 0.980. The electron-spin experiment used a 0.364 µs pulse spacing, but did not measure nuclear polarization. Under equal microwave and RF field strengths, the simulated gate was almost 10 times faster than direct nuclear excitation.

How dynamical decoupling drives the target

Section titled “How dynamical decoupling drives the target”

Dynamical decoupling uses a fast central qubit to control a weakly coupled target without resonant pulses on the target. CPMG applies $N$ $\pi_x$ pulses separated by $\tau$, with $\pi/2$ pulses bracketing the sequence. XYN alternates the $\pi$-pulse axes between $x$ and $y$. In the minimal model, a target resonance appears when $\tau$ equals $1/(2\omega_0)$, with additional resonances at odd multiples. The target precession then adds constructively in the central qubit’s reference frame. Other frequencies are filtered from the central response.

The state transfer is mediated by entanglement, not a simple Bloch-sphere rotation. The target’s reduced state moves inside the Bloch sphere because population transfers into two-qubit off-diagonal terms. The authors use a static laboratory-frame model without rotating frames or the rotating wave approximation. This model accepts generalized couplings and includes finite pulse durations, frequency errors, and duration errors.

For the NV implementation, slight field misalignment creates perpendicular Zeeman terms that expose nuclear precession to the electron spin. The resonance condition therefore differs from the minimal model’s direct Larmor-frequency rule.

IBM Quantum Torino served as a digital simulator of the central-target model. Qiskit 2.1.0 implemented the circuits, while Qiskit Aer 0.17.1 supplied local noisy simulations. Pauli evolution gates used one-repetition Lie Trotterization. Measurements used 100 shots, except tomography used 500 shots. The minimal model omitted collapse operators, while Aer supplied the IBMQ noise model.

The one-target model used central and target Larmor frequencies of 50 and 1 frequency units. The drive frequency was 5 units, the coupling was 0.2 units, and the time step was 0.001. At $\tau=0.5$, a CPMG-12 sequence inverted the populations in six time units.

CaseOperating conditionResultEvidence
One targetCPMG-12, $\tau=0.5$Full two-qubit fidelity 0.945; target-only fidelity 0.995; pseudo-fidelity 0.963IBMQ and noisy Qiskit
Intermediate entanglementCPMG-6, $\tau=0.5$Concurrence 0.899IBMQ tomography
Two targetsCPMG-16 at $\tau=0.5$, then CPMG-11 at $\tau=1.0$Individual fidelities 0.702 and 0.695; combined fidelity 0.659IBMQ and noisy Qiskit

The one-target central and target observables oscillated at the same frequency and amplitude with opposite phase. The two-target signals decayed in both hardware and simulation. The authors attribute this result to digital emulation of a time-dependent Hamiltonian, Trotterization, and compilation. They do not identify the low two-target fidelities as a failure of the DD protocol itself.

A coupling sweep found that $T_\pi^{-1}$ was linear in $A_{zx}$, with $T_\pi\approx0.212/A_{zx}$. The perturbative relation $T_\pi=1/(2A_{zx})$ underestimated the simulated duration by over a factor of two. Stronger coupling shortened the gate, but reduced pseudo-fidelity because discrete pulse counts could miss the exact inversion time.

The tested NV center was fabricated in an electronic-grade diamond plate. The authors implanted $^{15}$N at energies between 20 and 1530 keV, with a dose of $10^9$ N/cm². The sample was annealed at 1000 °C under ultra-high vacuum. A 518 nm diode laser initialized and measured the electron spin. Microwave $\pi$ pulses lasted 12.85 ns.

The NV measurement used a 32 mT field with a 2.9 degree misalignment from the NV axis. The model included a 2.87 GHz zero-field splitting. The hyperfine tensor had $A_{xx}=A_{yy}=3.65$ MHz and $A_{zz}=3.03$ MHz. The authors selected $^{15}$N because its $I=1/2$ nuclear spin has no quadrupole interaction that suppresses precession.

The measured electron-spin XYN spectrum showed the expected nuclear resonance. QuaCCAToo reproduced the short-sequence response and predicted opposite electron and nuclear oscillations. The nuclear resonance did not follow $1/(2|\gamma_n|B_0|)$ because hyperfine coupling dominated the nuclear Zeeman term. Electron-state-dependent nuclear Larmor frequencies also contributed.

A direct RF $\pi$ pulse would last 83.4 µs if its field intensity matched the microwave field. The authors also describe the gate as close to 100 times faster than the reported $^{13}$C DD method.

Simulated polarization of the $^{15}$N nuclear spin

Section titled “Simulated polarization of the $^{15}$N nuclear spin”

The simulated CPMG sequence transferred polarization from an initialized electron spin to a mixed $^{15}$N nucleus. The simulation started with $|0\rangle\langle0|$ for the electron and identity divided by 2 for the nucleus. At $\tau=0.3534$ µs and $N=27$, $T_{\mathrm{pol}}$ was 9.54 µs. The nuclear expectation reached 0.475, giving an efficiency of 0.950 against the maximum of 0.5. Tomography gave a fidelity of 0.979 for the polarized nuclear state. The electron observable ended near 0.5, consistent with loss of its initial polarization.

In simulation, XYN left the nuclear expectation nearly constant for the mixed initial state. CPMG produced an antisymmetric response around the resonance because its filter function differed from XYN. The joint evolution remained unitary because entanglement changed the reduced nuclear state’s mixing.

The authors report that the 9.54 µs sequence was over three times faster than a NOVEL sequence. It was nearly five times faster than PulsedPol. These results remain predictions from the NV simulation.

Longer experimental XYN sequences departed from the simulation as decoherence and pulse errors accumulated. Simulations showed XYN pseudo-fidelity above 0.7 across pulse-length ratios of 0.7 to 1.3. It remained above 0.7 across frequency ratios of 0.97 to 1.03. CPMG fidelity fell sharply under small deviations from ideal pulses. The authors suggest antisymmetrization, random phases, uneven pulse spacing, and composite pulses for further improvement.

The model neglects interactions between target qubits. It also examines only two target initial states, $|0\rangle$ and identity divided by 2. The authors have not characterized the complete quantum channel produced by the DD gate. They also leave deterministic entangled-state generation as future work.

Only one NV center was measured, and no diamond grade or crystal-orientation comparison was performed. The results therefore validate the control mechanism for the studied center, not material performance across an ensemble.

The $^{15}$N nucleus offers a uniform hyperfine interaction across NV centers, unlike the random lattice positions of $^{13}$C nuclei. This uniformity supports the authors’ argument for scalable diamond quantum-token registers. The electron spin serves as the fast actuator, while the nucleus serves as the longer-lived target. The demonstrated nuclear gate uses only microwave pulses on the electron spin.

Device calibration must treat pulse separation, pulse count, field orientation, and coupling together. Material fabrication also affects NV coherence, which the authors list as a route to extending experimental performance.

For the $^{15}$N NV quantum-control application, 6CCVD’s matching format is MPCVD-grown single-crystal diamond plate up to 125 mm in the diamond materials catalog. The paper specifies an electronic-grade plate for $^{15}$N implantation and a 1000 °C anneal, plus optical initialization and microwave delivery. The engineering support page can confirm the substrate specification, while defect processing remains a separate requirement.

View Original Abstract

We demonstrate a hardware-agnostic protocol for realizing fast, high-fidelity gates through dynamical decoupling (DD) pulse sequences applied to a central qubit coupled to target qubits. The target qubits are controlled by leveraging their intrinsic interaction with the central qubit, eliminating the need for slow, error-prone direct control. We develop and implement the DD-gate protocol within two distinct frameworks: a general model with minimal assumptions, benchmarked on a gate-based digital quantum simulator given by the IBM quantum (IBMQ) platforms; and an experimentally realistic case with a <a:math xmlns:a=“http://www.w3.org/1998/Math/MathML”> <a:mmultiscripts> <a:mi mathvariant=“normal”>N</a:mi> <a:mprescripts/> <a:none/> <a:mn>15</a:mn> </a:mmultiscripts> </a:math> -vacancy center ( <c:math xmlns:c=“http://www.w3.org/1998/Math/MathML”> <c:mmultiscripts> <c:mi>NV</c:mi> <c:mprescripts/> <c:none/> <c:mn>15</c:mn> </c:mmultiscripts> </c:math> ) in diamond. Using IBMQ, we are able to elucidate the underlying quantum dynamics of the DD gates and test them, independently of experimental constraints. For <d:math xmlns:d=“http://www.w3.org/1998/Math/MathML”> <d:mmultiscripts> <d:mi>NV</d:mi> <d:mprescripts/> <d:none/> <d:mn>15</d:mn> </d:mmultiscripts> </d:math> , we realize the protocol considering system-specific properties, which could represent a significant reduction in gate duration and improved technological scalability compared with current dynamical-decoupling-based control. We also propose a simple application for high-efficiency polarization of the <e:math xmlns:e=“http://www.w3.org/1998/Math/MathML”> <e:mmultiscripts> <e:mi mathvariant=“normal”>N</e:mi> <e:mprescripts/> <e:none/> <e:mn>15</e:mn> </e:mmultiscripts> </e:math> nuclear spin that could potentially be less technically demanding than current methods. Altogether, this work provides a robust strategy for quantum control that can be implemented in arbitrary systems fitting the central-target qubit architecture. Beyond these results, our open-source simulations and implementations for both platforms provide a practical framework for simulating time-dependent qubit dynamics on noisy intermediate-scale quantum era gate-based quantum processors.

  1. 2020 - Diamond for Quantum Applications Part 1