Skip to content

Robust Quantum-Network Memory Using Decoherence-Protected Subspaces of Nuclear Spins

FieldValue
JournalPhysical Review X
Published2016-06-22
AuthorsAndreas Reiserer, Norbert Kalb, Machiel Blok, Koen J. M. van Bemmelen, T. H. Taminiau
DOI10.1103/physrevx.6.021040
Citations155

Quantum coherence survived more than 1,000 repetitions of a probabilistic remote-entangling protocol when the qubit was encoded in a two-nuclear-spin decoherence-protected subspace. The protected state used nuclear spins 2 and 3. Its Bloch vector length reached 1/e only after more than 1,000 repetitions, and the remaining decay matched intrinsic nuclear-spin dephasing. For single-spin memory, the 1/e point occurred after more than 100 repetitions for the strongest coupled spin and about 500 repetitions for the weakest.

The experiment used a diamond device with natural-abundance 13C nuclear spins, each with I = 1/2. A nitrogen-vacancy (NV) center served as the optical interface and ancilla for local control. The device was cooled to 4 K in a helium bath cryostat, and a 40 mT field was applied along the NV symmetry axis. Before each run, the authors prepared the NV center in its negative charge state and checked its resonance with the lasers.

Spin-selective optical transitions initialized the electronic spin with fidelity above 0.99, while single-shot readout had an average fidelity of about 0.94. Tailored microwave pulse sequences gave individual control and readout of five 13C nuclear spins.

The measured parameters were:

13C spinΔω/(2π) (kHz)A∥ (kHz)T₂* (ms)F_i,r
1-15.4-11.06(1)0.89(2)
218.421.213(1)0.96(2)
323.724.719(2)0.97(2)
4-37.0-36.010(1)0.92(2)
5-48.6-48.74(1)0.90(2)

Here Δω is the nuclear precession-frequency difference, A∥ is the parallel hyperfine coupling, and F_i,r is the combined initialization and readout fidelity. The hyperfine couplings were measured by dynamical decoupling spectroscopy, with uncertainty in the last digit.

The Barrett-Kok inter-node sequence initialized the electronic spin optically, applied a microwave π/2 pulse, and generated spin-photon entanglement through two optical excitations separated by a microwave π pulse. Interfering photons at a beamsplitter could then project the electronic spins into a maximally entangled state. Photon loss made success probabilistic.

After a failed attempt, the electronic spin had an unknown state and required reset. Its always-on hyperfine coupling changed the nuclear precession axis and frequency. With the electron in |0⟩, the nucleus precessed at the Larmor frequency. With the electron in |-1⟩, the axis tilted and the frequency shifted. In the strong-field limit used here, the tilt was suppressed quadratically, leaving linear frequency shifts from A∥ as the main dephasing source.

The sequence would cancel these shifts if reset were instantaneous and precisely timed, because the electron spent equal time in |0⟩ and |-1⟩. Stochastic optical reset left irreversible phase uncertainty. The authors set t = 2π/ω0 ≃ 2.3 µs and found a broad coherence optimum at τ ≃ 0.44 µs.

The authors prepared the electronic spin in |-1⟩, applied a repump laser for a variable duration, waited 2.5 µs, and measured the |0⟩ population. The decay was double-exponential, with a fast-to-slow relative amplitude of about 3:1.

At high laser power, driving |A₁,₂⟩ gave reset timescales of 29(1) ns and 463(14) ns. Driving |E₁,₂⟩ gave 48(1) ns and 432(26) ns. The authors attributed the slow component to metastable singlet states. Their rate-equation model used a 440 ns singlet lifetime and a 2:1:1 branching ratio toward |0⟩, |+1⟩, and |-1⟩.

The ground-state Zeeman splitting of about 2 GHz required a second laser beam. At 2000 nW, the initialization curve saturated below 0.01 after 2 µs for |E₁,₂⟩ and 1.5 µs for |A₁,₂⟩.

With optimized timing, all five single-spin coherences decayed exponentially with the number of repetitions. Even the spin with the largest A∥ magnitude, 48.7 kHz, remained above the 1/e point after more than 100 repetitions. The weakest spin, with A∥ of -11.0 kHz, reached 1/e near 500 repetitions.

Earlier NV-node experiments had used this optical sequence for remote entanglement. The new measurement added a simultaneously encoded local memory to repeated sequence attempts.

The authors used a BB1 sequence of five microwave pulses, each shorter than 0.2 µs, with a broad Hermite envelope. Replacing BB1 with single pulses did not change the DPS decay constant. Lower A∥ can extend memory, but weakly coupled spins generally require longer local control operations.

The protected basis used |↓i↑j⟩ and |↑i↓j⟩ for spins i and j with comparable A∥. To first order, effective dephasing followed the difference in parallel couplings rather than either magnitude. The |↑i↑j⟩ and |↓i↓j⟩ basis produced increased dephasing.

For spins 2 and 3, the parallel couplings were 21.2 kHz and 24.7 kHz. The protected pair had a measured effective frequency difference of 5.1 kHz. Its decoherence-protected subspace maintained coherence for more than 1,000 repetitions. After correction for intrinsic T₂* dephasing, the decay constant was fully explained by natural nuclear-spin dephasing. Reset-induced dephasing had become negligible.

Encoding and readout required extra nuclear control operations, so the initial fidelity was slightly reduced.

The authors measured the 1/e repetition count for all five single spins and all 20 possible two-spin subspaces under five repump conditions. These were 2000 nW for both transition configurations, and 190, 100, and 50 nW for |A₁,₂⟩.

The model reproduced exponential decay and faster loss at larger coupling. It did not agree quantitatively when τ was fixed at 0.44 µs. The model assumed an exponential repump-time distribution and an NV that stayed in |-1⟩ until reset. It neglected the measured double-exponential reset and time spent in |+1⟩, an excited state, or the metastable singlet.

Introducing a coupling offset C ≈ 2π × 15 kHz and leaving τ free gave reasonable agreement. The fitted values were τ = 0.43(3) µs for |E₁,₂⟩ and τ = 0.46(1) µs for |A₁,₂⟩. Both agreed with the slow reset scale and the measured optimum.

The protected pair remained limited by intrinsic T₂* because no nuclear-spin echo was applied. Population decay constants ranged from 1,000 to 10,000 repetitions depending on the spin. This decay did not limit the tested DPS, but the authors expect it to become relevant for the smallest effective couplings once an echo cancels intrinsic dephasing.

At 2 µW with the |A₁,₂⟩ configuration, the probability that the NV remained negative decayed with N_d = 2.82(8) × 10³ resets when every trial reset the electron. In the entanglement sequence, reset was required in 50% of trials on average, so the authors expected decay twice as slow. The scaling analysis excluded deionized runs, while a charge-state check identified the loss.

The authors propose high-quality optical cavities to increase direct spin flips through the Purcell effect without populating singlet states. Non-destructive electronic-spin readout or measurement within a decoupling sequence might remove the need for probabilistic repumping. Absorption-based network protocols could reduce the number of resets before a successful event is heralded. These remain proposed directions.

The study compares nuclear-spin sites within one diamond rather than diamond grades. For device engineering, A∥, T₂*, reset duration, and charge stability must be considered together. Closely matched A∥ values enable a protected subspace, while faster reset must be balanced against deionization at higher laser power. When natural dephasing dominates, a nuclear echo can extend coherence.

An NV-center quantum-network node requires a diamond material platform. 6CCVD’s diamond materials catalog lists single-crystal and polycrystalline diamond grown by MPCVD, which are the catalog families to compare with the paper’s device.

The paper’s implementation requires an optically active NV center with natural-abundance 13C nuclear spins at 4 K and a 40 mT magnetic field.

Engineering support can be used to check the material and fabrication requirements against those conditions.

View Original Abstract

The realization of a network of quantum registers is an outstanding challenge in quantum science and technology.We experimentally investigate a network node that consists of a single nitrogen-vacancy center electronic spin hyperfine coupled to nearby nuclear spins.We demonstrate individual control and readout of five nuclear spin qubits within one node.We then characterize the storage of quantum superpositions in individual nuclear spins under repeated application of a probabilistic optical internode entangling protocol.We find that the storage fidelity is limited by dephasing during the electronic spin reset after failed attempts.By encoding quantum states into a decoherence-protected subspace of two nuclear spins, we show that quantum coherence can be maintained for over 1000 repetitions of the remote entangling protocol.These results and insights pave the way towards remote entanglement purification and the realization of a quantum repeater using nitrogen-vacancy center quantum-network nodes.