Cavity-assisted dynamic nuclear polarisation

AI summary

Overview

Research area: Quantum physics / magnetic resonance — dynamic nuclear polarisation (DNP) of electron–nuclear spin ensembles coupled to superconducting microwave cavities, with relevance to spin-ensemble quantum memories and NMR sensitivity.

Technical level: Advanced. The paper assumes familiarity with spin Hamiltonians, hyperfine structure, cavity quantum electrodynamics (Purcell effect, cooperativity), and pulsed electron spin resonance techniques.

Scope (one sentence): The paper introduces and experimentally demonstrates a cavity-based DNP protocol that uses Purcell-enhanced radiative relaxation of bismuth donor spins in isotopically purified silicon to concentrate the ground-manifold population into a single state, reaching a cumulative enhancement of 4.6 (±0.2) against a theoretical limit of 9.

What This Paper Is About

Nuclear spins at millikelvin temperatures are very weakly polarised because their transition energies are far below the thermal energy, which limits both NMR detection sensitivity and the performance of spin-ensemble quantum memories. Conventional DNP transfers polarisation from electron spins using a dissipative channel, but at millikelvin temperatures phonon-mediated relaxation freezes out, making those routes ineffective. This paper shows that a cold electromagnetic bath — a tunable superconducting resonator — can supply that dissipation through the Purcell effect, and uses it to pump the populations of a bismuth donor ensemble's nine ground states into one.

Key Contributions

1. A new DNP mechanism based on cavity-induced dissipation. The authors introduce a "Purcell-DNP" protocol in which the dissipative channel is not a material property but an externally controlled parameter: the radiative decay of an electronic transition into a superconducting resonator. The protocol combines adiabatic microwave pulses, radio-frequency (hyperfine) pulses, and Purcell relaxation.
2. Experimental demonstration on Bi:²⁸Si. The protocol is implemented on an ensemble of bismuth donor spins (I = 9/2, S = 1/2) in isotopically purified ²⁸Si coupled to a frequency-tunable superconducting resonator, with a single cycle increasing the target ground state population by 1.79 (±0.04) and eight cycles reaching 4.6 (±0.2).
3. Quantitative identification of the limiting mechanisms. The shortfall from the ideal ninefold limit is explained by three effects: competition between Purcell and intrinsic relaxation, finite inversion fidelity, and a collective cavity-mediated spin flip-flop effect that slows radiative decay as nuclear polarisation builds up. A third-order cumulant-expansion model reproduces all build-up curves without adjustable parameters.
4. A roadmap for practical use in quantum memories. The authors show that Purcell-DNP increases the measured ensemble cooperativity by a factor 2.8 (±0.1) and argue that a low-impedance resonator with a 300 nm inter-chip distance would reach unit cooperativity in about 10 minutes even for an ensemble 20 times more dilute.

Main Findings

• Single-cycle enhancement: After an optimal recovery delay τ_W = 280 s, the echo amplitude at ω₆₋₁₅ increases by a factor of 1.79 (±0.04), against an ideal factor of 2 for one cycle. The echo stays at this level for about 700 s before decaying back to its unpumped level.
• Cumulative eight-cycle enhancement: Running n = 1, …, 8 cycles reaches 4.6 (±0.2) after eight cycles (total duration 45 min), then saturates for additional cycles. The theoretical limit set by the degeneracy of the ground manifold is 9.
• Spectroscopic confirmation: CPMG spectroscopy after one cycle shows the |6⟩↔|15⟩ peak area increasing by 1.7 (±0.2), with no change to the |5⟩↔|14⟩ peak. Continuous-wave spectroscopy gives C₈/C₀ = 2.8 (±0.1) after eight cycles; the lower value is attributed to weakly coupled, off-resonant spins that CPMG spectroscopy does not see and that are less efficiently repumped.
• Purcell dominance: On resonance, the |6⟩↔|15⟩ lifetime is T₁ = 32 (±2) s, versus an intrinsic lifetime of T₁^int = 24 (±2) min when the resonator is detuned — Purcell decay dominates by nearly two orders of magnitude. Modulating the resonator by ±150 kHz during relaxation broadens the reset bandwidth at the cost of T₁^eff = 41 (±1) s.
• Extracted coupling: From the on-resonance T₁ and the single-spin Purcell rate formula, the mean coupling for spins contributing to the echo is ḡ/(2π) = 16.3 (±0.5) Hz.
• Spin ensemble parameters: The |6⟩↔|15⟩ and |5⟩↔|14⟩ transitions are at 7.416635 GHz, with linewidths Γ₆₋₁₅/(2π) = 288 (±24) kHz and Γ₅₋₁₄/(2π) = 182 (±27) kHz, split by 940 (±11) kHz. Bare resonator linewidth κ/2π ∼ 210 kHz; cooperativities C⁶⁻¹⁵ = 0.1 (±0.01) and C⁵⁻¹⁴ = 0.06 (±0.01) — the low-cooperativity regime where the Purcell effect can occur.
• Pulse fidelities: Electronic WURST inversion efficiency η_e = 0.9996 (uncertainty given as ±0.0001 in the main text and ±0.0005 in the Fig. 3 caption), with 90 (±3)% of spins inverted across the full line and at most 11% change on the spectator |5⟩↔|14⟩ transition. Nuclear inversion probability η_n = 0.991 (±0.001).
• Slowdown of radiative relaxation: Fitting each cycle's build-up curve with independently measured lifetimes and pulse efficiencies fixed leaves the Purcell rate Γ_P,n decreasing systematically with cycle number (Fig. 5e). This is attributed to resonator-mediated spin–spin flip-flops that grow as population accumulates in |6⟩, pushing the system toward strong coupling and broadening the resonator to κ̃ = κ + κ_s.
• Model-identified loss budget: Suppressing cavity-mediated flip-flops would raise the enhancement to 4.9; perfect WURST pulses would give 6.7; additionally removing hyperfine diffusion would give 8.1; the residual gap to 9 is due to weakly coupled spins that are hard to reset radiatively.
• Polarisation lifetime: Hyperfine diffusion relaxes the accumulated population with a decay time T_hf = 166 (±5) min; repeated CPMG interrogation every 4 min produces a similar decay on a slightly shorter timescale.
• DNP enhancement parameter: Estimated at about 20, lower than the 100–500 reported in solid-state experiments, which the authors attribute to the already sizeable equilibrium nuclear polarisation (p_th ≈ 0.19 at T = 25 mK).

Methodology in Plain English

The experiment uses bismuth atoms implanted in a layer at the surface of an isotopically purified silicon chip, placed 1.1 µm above a superconducting niobium-titanium-nitride resonator. At low magnetic field, each bismuth donor has nine ground states; at 25 mK they are almost equally populated, so each nuclear transition has negligible polarisation.

The pumping cycle has three steps. First, a chirped (WURST) microwave pulse moves population from state |6⟩ up to the excited state |15⟩. Second, a radio-frequency pulse swaps the population of a neighbouring ground state (|7⟩) into the now-empty |6⟩. Third, the resonator is tuned onto the |6⟩↔|15⟩ transition so that the excitation in |15⟩ decays radiatively back into |6⟩ — much faster than it would on its own. Repeating this with successive nuclear swaps funnels the whole ground manifold into |6⟩.

Measuring the result is itself delicate, because the signal is tiny. The team uses Carr–Purcell–Meiboom–Gill (CPMG) sequences — one π/2 pulse, a 250 µs wait, then a train of 100 π pulses producing 100 echoes — read out by heterodyne detection, exploiting a spin coherence time T₂ = 3.4 ms. NMR-like ENDOR spectroscopy locates the nuclear transition frequencies, and continuous-wave spectroscopy of the resonator linewidth provides an independent, ensemble-averaged measure of polarisation through the cooperativity. Rate-equation and cumulant-expansion models, with parameters fixed by independent measurements, reproduce the full set of build-up curves.

Why This Matters

Impact on research. The work establishes radiative relaxation as a usable DNP mechanism in the millikelvin regime where phonon-mediated relaxation is frozen — a regime that matters for circuit QED experiments. It also reframes a loss channel (Purcell decay) as a resource, and gives a quantitatively validated account of why dense, strongly coupled ensembles fall short of ideal polarisation, namely cavity-mediated spin flip-flops.

Potential applications:

• Spin-ensemble quantum memories: higher cooperativity allows coherent storage and retrieval of single microwave photons from a superconducting processor while preserving the ensemble's coherence time; the demonstrated 2.8 (±0.1) cooperativity gain is described as a first step.
• NMR sensitivity in nanoscale and volume-limited samples, where the effective number of contributing spins is set by polarisation.
• Solid-state spin platforms beyond bismuth: rare-earth ions, NV centres, and molecular spins with spectrally resolved electron–nuclear transitions and sufficient resonator coupling.
• Suppressing spectral diffusion and lengthening coherence times by polarising nearby nuclear spin baths with weaker electron coupling, potentially without an isotopically purified host.

Industry relevance. The results bear on superconducting quantum processor architectures that need microwave quantum memories, on instrumentation for hyperpolarised magnetic resonance, and on the design of low-impedance resonators and flip-chip packaging for spin–cavity experiments.

Future Directions

• Engineering stronger Purcell rates. The authors propose a low-impedance resonator and a 300 nm inter-chip distance to reach unit cooperativity in roughly 10 minutes even for a spin ensemble 20 times more dilute — the regime relevant for quantum memories.
• Extending to weakly coupled nuclear baths. Applying the technique to host-material nuclear spins or radicals, which couple more weakly to the electron spin, remains open; success would offer a route to NMR sensitivity in nanoscale samples and could remove the need for isotopically purified silicon.
• Suppressing cavity-mediated flip-flops. Since removing these interactions alone would raise the enhancement to 4.9, finding ways to mitigate them (or to operate away from the onset of strong coupling) is a direct lever on efficiency.
• Improving pulse and diffusion limits. Perfect inversion pulses would give 6.7 and eliminating hyperfine diffusion 8.1, so better control over pulse bandwidth and population leakage is a clear target on the path to the ninefold limit.

Target Audience

Specialists and graduate students in quantum information and magnetic resonance: researchers working on spin-ensemble quantum memories coupled to superconducting circuits, DNP methodologists, and cavity-QED experimentalists. Readers interested in hyperpolarisation for NMR will find the mechanism and its limits clearly quantified, but the paper assumes comfort with spin Hamiltonians, hyperfine structure, and cavity cooperativity.

Authors’ abstract

Dynamic nuclear polarisation transfers polarisation from electron to nuclear spins by relying on a dissipative channel. Here we introduce a cavity-based DNP mechanism which uses the radiative relaxation of the electronic spin through the Purcell effect to bring about the nuclear spin polarisation. We implement this protocol on an ensemble of bismuth donor spins (I=9/2, S=1/2) in isotopically purified silicon coupled to a superconducting resonator. A single cycle, combining adiabatic microwave and radio-frequency pulses with Purcell relaxation, increases a target ground state's population by a factor of $1.79\pm0.04$; a cumulative enhancement of $4.6\pm0.2$ is reached with additional cycles, to be compared with the theoretical limit of ninefold set by the degeneracy of the ground manifold. We explain this difference by the competition between Purcell and intrinsic relaxation on the one hand and a collective effect that slows down the electronic spin radiative decay as the nuclear spin population grows on the other hand. Our results place radiative DNP as an available mechanism at millikelvin temperatures where phonon-mediated relaxation is frozen.