Bosonic Error Correction with Fluxonium

AI summary

Overview

Research area: Quantum physics / superconducting quantum computing, specifically bosonic quantum error correction (QEC) in microwave circuit quantum electrodynamics.

Technical level: Advanced. The paper assumes familiarity with GKP codes, dispersive qubit–resonator coupling, and measurement-free error correction. The headline result, however, is understandable with intermediate background knowledge.

Scope (one sentence): This paper demonstrates a standalone bosonic error-correcting code — finite-energy Gottesman–Kitaev–Preskill (GKP) states stabilized by measurement-free error correction — in a fully planar superconducting circuit, using a long-lived heavy fluxonium qubit as the control element and an on-chip Archimedean spiral resonator (ASR) as the storage mode.

What This Paper Is About

Bosonic codes store a logical qubit in the large Hilbert space of a single harmonic oscillator, which is far more hardware-efficient than building codes from many physical qubits. Previous superconducting bosonic-code experiments used centimeter-scale 3D microwave cavities controlled by fixed-frequency transmons, and their logical lifetimes were capped by the transmon's bit-flip errors (T₁ᵠ ≈ 200 µs). The goal here is to replace both the 3D cavity and the transmon with an on-chip resonator and a long-lived fluxonium, and then to show that a genuine GKP error-correction protocol works in that fully planar architecture.

Key Contributions

1. First standalone bosonic QEC code in a fully planar superconducting circuit. The authors replace the 3D storage cavity with an on-chip Archimedean spiral resonator and demonstrate an error-corrected GKP qubit in that monolithic, planar device.
2. First resonator control using a weakly coupled fluxonium. With a dispersive coupling of only χ/2π = 23 kHz, they still achieve fast echoed conditional displacement (ECD) control by using large resonator displacements to enhance the conditional-displacement interaction, which scales as χ|α|.
3. First experimental preparation of GKP logical states in a fully planar device. Finite-energy codewords |+z_L⟩, |−z_L⟩, and |+y_L⟩ were prepared at Δ = 0.438 via numerically optimized circuits of single-qubit rotations and ECD gates.
4. Measurement-free stabilization of GKP states with rapid fluxonium reset. Using the small-big-small (sBs) protocol plus a 236 ns pulsed fluxonium reset, they extend the logical lifetime by a Pauli-averaged factor of 1.59 ± 0.05.

Main Findings

• Long-lived control qubit: The heavy fluxonium, biased at its half-flux sweet spot, has a time-averaged bit-flip lifetime of T̄₁ᵠ = 451 ± 70 µs, exceeding any control qubit reported in prior experimental bosonic QEC demonstrations. Its echo coherence is T̄₂Eᵠ = 91 ± 10 µs, limited by shot-noise dephasing to its readout resonator.
• High-quality planar storage mode: The ASR has frequency 4.138 GHz, single-photon lifetime T₁ᵃ = 103 ± 2 µs, and overall quality factor Q = 2.65 × 10⁶.
• Weak coupling suppresses inherited nonlinearity: The 23 kHz dispersive coupling yields resonator self-Kerr nonlinearities of only K_g/2π = 2.4 Hz and K_e/2π = −0.007 Hz — hertz level, versus kilohertz-level resonator nonlinearity in prior strongly coupled work.
• Large displacements are supported: Out-and-back measurements were modeled up to mean photon number n̄ = 40,000, and the resonator reached more than 2000 photons before drive-induced unwanted state transitions (DUST) appeared in the fluxonium.
• Error correction works on all three Pauli axes: Corrected logical lifetimes T_X^C, T_Y^C, and T_Z^C all exceed their uncorrected counterparts T_X^U, T_Y^U, and T_Z^U, giving T_avg^C / T_avg^U = 1.59 ± 0.05. The individual fitted values and ratios are listed in Supplemental Table S4 and are not given in the main text excerpt.
• Code-space preservation is visible: Characteristic-function measurements after 5 uncorrected rounds show decay toward vacuum, whereas after 5 and 29 corrected rounds the state remains in the code space even as logical information is gradually lost.
• Break-even is not reached: The logical qubit does not yet outperform its constituents — both the Fock qubit (measured in Sec. S4.4) and the fluxonium have longer Pauli-averaged lifetimes.
• Dominant error identified: The error budget agrees closely with the measured logical error rate and identifies the resonator's intrinsic dephasing, T_φᵃ = 1.08 ± 0.16 ms, as the leading channel, likely from dispersive dephasing through coupled two-level systems such as surface defects.

Methodology in Plain English

The team built a single chip containing two coupled elements: a storage resonator (a spiral-shaped on-chip microwave cavity) and a control qubit (a fluxonium, chosen because it can be engineered to hold its state for a long time). Because the two are only weakly coupled, the natural interaction is too slow for fast control — so the team first pushes the resonator into a large-amplitude state and lets that large displacement amplify the interaction, enabling an echoed conditional displacement gate that displaces the resonator one way or the other depending on the qubit's state. This gate, plus ordinary qubit rotations, is enough to build any desired state.

To make GKP states, they used numerical optimization (run on GPUs via the JAX-based open-source JAXQuantum package) to search for short circuits — single-qubit rotations and ECD gates — that produce the target codewords, then executed those circuits on the hardware. To keep the states alive, they applied an error-correction cycle that repeatedly maps small errors from the resonator onto the fluxonium and then empties the fluxonium, so entropy is carried away. No mid-circuit measurement or real-time feedback is needed: the fluxonium is simply reset with a fast 236 ns pulsed cooling sequence, which is only 5.6% of the 4232 ns error-correction round. Logical information was read out through logical Pauli expectation values and characteristic functions, and lifetimes were extracted from zero-offset exponential fits to the distinguishability between logical states.

Why This Matters

Bosonic codes promise error correction with far fewer hardware components than qubit-register codes, but their practical value depends on being buildable in a scalable, manufacturable format. This work shows that the whole stack — storage mode, control qubit, tunable coupling, and error-correction cycle — can be built in a single planar chip, and it shifts the dominant error source away from the control qubit toward the resonator's own materials-limited dephasing. That reframing tells the field where to focus next.

Potential applications:

• Hardware-efficient quantum processors: Compact on-chip GKP qubits as the inner layer of concatenated codes, protected at the outer level by a surface code against residual logical errors.
• Dense integration: Planar resonators with mode volumes several orders of magnitude smaller than 3D cavities and footprints comparable to physical qubits in register architectures.
• Real-time tunable quantum hardware: On-chip broadband flux lines make real-time flux tunability and parametrically driven interactions easy to add, without introducing photon-loss channels into the storage mode.
• Improved decoding via soft information: The fluxonium state measured before each reset could feed the outer decoder, analogous to how erasure information improves surface-code decoding.

Industry relevance: The work sits squarely in the superconducting quantum-computing roadmap pursued by academic labs and industrial players (the paper was supported in part by the AWS Center for Quantum Computing and involved MIT Lincoln Laboratory fabrication). A planar, lithographically defined bosonic qubit is compatible with standard foundry-style fabrication, which is the practical prerequisite for scaling. Fluxonium-based control may also appeal to groups already pursuing flux-tunable architectures.

Future Directions

1. Reduce resonator intrinsic dephasing. T_φᵃ = 1.08 ± 0.16 ms is the dominant error channel; surface treatments, resonator geometry optimization, and materials engineering such as tantalum metallization are proposed, with planar internal quality factors potentially approaching the bulk silicon substrate limit of Q_i ∼ 10⁸.
2. Exploit flux tunability to suppress backaction. Threading the ASR frequency between the fluxonium plasmon transitions produces a zero crossing of the dispersive interaction near half flux; this could decouple the qubit from the storage mode during operations such as fluxonium reset.
3. Push past break-even. Since the Fock qubit and fluxonium both currently outlive the corrected GKP qubit, improving the storage mode is the route to a logical qubit that genuinely beats its parts.
4. Build the concatenated layer. Combine the on-chip GKP qubit with an outer code such as the surface code, and use pre-reset fluxonium measurement outcomes as soft information for the outer decoder.

Target Audience

Quantum-hardware researchers working on superconducting circuits and bosonic codes; graduate students and postdocs in circuit QED who want a concrete blueprint for planar bosonic QEC; quantum-algorithm and architecture researchers evaluating GKP-based concatenated schemes; and industry R&D teams assessing the manufacturability and scaling path of bosonic error correction. Readers without a background in GKP formalism or dispersive readout will find the experimental narrative accessible, but the protocol details require prior exposure to the field.

Authors’ abstract

Bosonic quantum error correction (QEC) offers a hardware-efficient route to fault-tolerant quantum computing. To date, however, superconducting circuit implementations of bosonic codes have utilized centimeter-scale 3D microwave cavities controlled by fixed-frequency transmon qubits, with logical lifetimes limited by transmon bit-flip errors. Here, we realize bosonic QEC in a fully planar architecture by pairing a heavy fluxonium, whose $451 \pm 70~μ\mathrm{s}$ bit-flip lifetime exceeds that of any control qubit in previous demonstrations, with an on-chip Archimedean spiral resonator several orders of magnitude smaller in mode volume than prior 3D cavities. We prepare finite-energy Gottesman-Kitaev-Preskill (GKP) states and stabilize them using measurement-free error correction with rapid fluxonium reset, extending the logical lifetime by a factor of $1.59 \pm 0.05$. These results provide the first demonstration of resonator control using a weakly coupled fluxonium and, with it, the first realization of standalone bosonic QEC in a fully planar superconducting circuit architecture.