High-Fidelity Entangled States in a Connectivity-Four Fluxonium Quantum Processor

AI summary

Overview

Research area: Superconducting quantum computing hardware — specifically fluxonium qubits, multi-qubit coupling architectures, microwave gate implementation, and multi-qubit entanglement generation.

Technical level: Intermediate. The device physics and Hamiltonian modeling are dense, but the performance results (fidelities, interaction rates, entanglement benchmarks) are presented in a way that readers familiar with basic quantum-computing concepts can follow.

Scope: The paper reports a five-qubit fluxonium processor built from lumped-element resonator couplers that achieves a connectivity-four unit cell, benchmarking single-qubit gates, idling ZZ suppression, two-qubit (RIP) gates with a spectator-error refocusing protocol, and three- to five-qubit GHZ states.

What This Paper Is About

Fluxonium qubits have excellent coherence, but nobody had shown how to wire them into a two-dimensional lattice where every qubit touches four neighbors — the geometry the surface code needs — without introducing unwanted residual interactions or errors caused by the presence of extra coupled qubits. The authors build a five-qubit processor in a fluxonium-resonator-fluxonium (FRF) design, where each pair of neighboring qubits shares a grounded lumped-element resonator that both mediates two-qubit gates and suppresses static ZZ crosstalk. The goal is to demonstrate that this architecture supports high-fidelity single- and two-qubit gates and genuine multi-qubit entanglement in a connectivity-four geometry.

Key Contributions

1. First connectivity-four unit cell in fluxonium technology. The processor uses lumped-element resonator couplers with a grounded geometry (2 µm niobium wire as the inductive element, characteristic impedance of approximately 150 Ω) to realize a unit cell where each qubit couples to four neighbors with suppressed parasitic interactions, which the authors state is a first for this qubit type.

2. Simultaneous high-fidelity single-qubit operation with strongly suppressed idling crosstalk. Parallel single-qubit randomized benchmarking on all qubits yields a median fidelity of 99.957(5) %, and residual static ZZ interactions are held below 1 kHz across all coupled pairs, with a measured median of 30(215) Hz.

3. Resonator-induced phase (RIP) CZ gates with a refocusing scheme that cancels spectator errors. A standard RIP gate loses fidelity when a neighboring qubit sits in |1⟩; splitting the gate into two 96 ns pulses separated by a π-pulse on the spectator restores fidelity and removes the state dependence, raising the average across both IRB sequences from 97.9(3) % to 99.20(5) %.

4. Multi-qubit entanglement within the unit cell. Three-, four-, and five-qubit GHZ states are prepared and verified by quantum state tomography with fidelities of 96.9(2) %, 95.4(2) %, and 90.0(2) %, which the authors describe as the first multi-qubit entangled states in fluxonium qubits at connectivity four and the highest GHZ fidelities for fluxonium at each respective qubit count.

Main Findings

• Single-qubit gates near the coherence limit: Median simultaneous-RB fidelity is 99.957(5) % using 48 ns gates (32 ns Gaussian pulse plus a 16 ns delay for flux-line reflections), compared with a device coherence limit of 99.96(1) %. Individual qubit fidelities range from 99.953(1) % (q5) to 99.968(1) % (q1).
• ZZ suppression of roughly a few tens of hertz: Using a JAZZ sequence with 50 repetitions per pair, the median static ZZ rate is 30(215) Hz, far below the qubit decoherence rate of Γ₂ ∼ 10 kHz. Per-coupler values range from −6(111) Hz (r25) to 131(145) Hz (r45).
• Spectator error is state-dependent and ZZ-like: On coupler r15, the bare RIP gate reaches 99.68(5) % when the spectator q2 is in |0⟩ but drops to 96.1(6) % when q2 is in |1⟩. The echo-type refocusing removes this dependence, which the authors take as strong evidence that the error is ZZ-type.
• Refocused gate improves two-qubit fidelity across the device: Applying the refocused sequence to all next-neighbor qubits gives a median CZ fidelity of 98.8(12) %, with per-coupler values of 99.20(5) % (r15), 98.4(1) % (r25), 96.2(2) % (r35), and 99.44(3) % (r45).
• One pair limits the median: Coupler r35's 96.2(2) % is attributed to increased photon loss, because its small dispersive shift of χ = 0.81 MHz forces a smaller drive detuning to keep the same pulse time, increasing decoherence from the resonator linewidth.
• GHZ fidelities degrade gracefully with qubit count: Tomographic fidelities are 96.9(2) % (three qubits, q1–q2–q5), 95.4(2) % (four qubits, q1–q2–q4–q5), and 90.0(2) % (five qubits), each reconstructed from 30720 single shots with classical detector error correction applied. Uncorrected state fidelities remain above 60 %, which still exceeds the 50 % threshold for genuine multipartite entanglement.
• No residual coherent spectator signatures: Reconstructed GHZ density matrices show no pairwise off-diagonal elements of opposing sign or significant imaginary components, consistent with effective spectator suppression in multi-qubit circuits.
• Readout is the weak link: Median single-shot assignment fidelity is 94₋₃⁺¹ %, which the authors identify as insufficient for high-fidelity syndrome extraction.

Methodology in Plain English

Each qubit is a fluxonium — a superconducting circuit with a large inductance that makes it insensitive to charge noise — flux-biased at half a flux quantum and designed to the same target parameters (E_C/h = 1.5 GHz, E_J/h = 6.5 GHz, E_L/h = 0.9 GHz, qubit transition around 350 MHz). Neighboring qubits share a small lumped-element LC resonator made from a narrow niobium wire. Because the resonator is grounded while the qubits are floating, the direct qubit-qubit coupling (designed at 35–43 MHz) cancels the interaction the resonator would otherwise mediate, driving the static ZZ rate toward zero.

To perform a two-qubit gate, the resonator is driven with a microwave tone detuned from its frequency. The resonator's frequency shifts depending on which computational states the two qubits occupy, because the qubits are deliberately placed so the resonator sits between the |1⟩→|2⟩ and |0⟩→|3⟩ transitions. A shaped pulse therefore imprints a state-dependent phase — a controlled-Z gate — without needing any tunable Josephson element in the coupler. Because other qubits are permanently attached, their presence perturbs the phase accumulation; the fix is to split the pulse in half and flip the neighboring qubit(s) in between, so their unwanted phase contributions cancel while the intended gate phase survives.

Characterization proceeds layer by layer: standard decay and Ramsey/echo experiments for coherence, simultaneous randomized benchmarking for single-qubit gates, the JAZZ echo sequence for ZZ rates (with a virtual frequency injected to make the very small oscillation measurable in a fit), interleaved randomized benchmarking for two-qubit gates, and finally full state tomography of GHZ states with a measured detector-error matrix inverted to remove readout bias.

Why This Matters

The paper shows that a resonator-based coupling scheme — passive, junction-free, and easy to frequency-target — can support the four-neighbor connectivity that error-correcting codes demand, at fidelities that are at least in the right neighborhood. It also demonstrates a pulse-level fix for the spectator problem that appears the moment you add more neighbors, which is precisely the problem that blocks scaling from two-qubit demos to lattices. Because the coupler is a linear element and the gate is microwave-only, the authors avoid the flux-pulse crosstalk calibration overhead typical of flux-activated schemes.

Potential applications:

• Surface-code quantum error correction: The stated motivation; connectivity four with sub-kHz ZZ is a prerequisite for stabilizer-based logical qubits.
• Codes needing higher connectivity: The authors point to qLDPC codes as a natural fit for resonator-mediated long-range coupling.
• Quantum simulation: Densely connected qubit lattices suit lattice models and materials-inspired simulation.
• Gate-model algorithm hardware: GHZ states and high-fidelity CZ gates are building blocks for entanglement-heavy circuits used in cryptography and optimization workloads.

Industry relevance: The results bear directly on superconducting processor design choices — coupler material and geometry, impedance targets, capacitance budgeting under four couplers per qubit, and microwave control-line engineering. The honest accounting of the readout limit (94₋₃⁺¹ %) and the coupler-to-coupler residual coupling (up to 26 MHz) gives hardware teams concrete fabrication and layout targets, and the architecture avoids junctions in the coupler, which matters for yield and parameter reproducibility in foundry-style fabrication.

Future Directions

• Uniformity of couplers and dispersive shifts: The spread in dispersive shifts (from 0.81 MHz to 53.9 MHz across pairs) and coupler frequencies tracks the spread in gate fidelity. Tightening fabrication tolerances is named as the primary target, and would let the intrinsic fidelity ceiling of the RIP interaction be measured.
• Faster gates: The 128 ns RIP gate (2 × 96 ns when refocused) is slow compared with the fastest fluxonium two-qubit gates using direct or tunable coupling. Pulse shaping is the proposed route to shorten it without reintroducing leakage or spectator errors.
• Restoring full parallelism: Refocusing requires the spectator qubit to idle and accept an X_π pulse, so neighboring pairs cannot run two-qubit gates simultaneously — in surface-code terms, stabilizer measurements stretch over two cycles. Increasing coupler-coupler detuning and reducing inter-coupler capacitance are proposed to fix this at the hardware level. The authors speculate the observed spectator errors stem from residual neighbor-coupler coupling (up to 26 MHz) combined with small mutual detuning.
• Better readout: Median single-shot assignment fidelity of 94₋₃⁺¹ % is not sufficient for syndrome extraction. Purcell-filtered readout resonators, optimized dispersive-shift/linewidth combinations, and dynamic readout protocols are named as planned integrations.

Target Audience

Superconducting-qubit experimentalists and hardware architects, particularly groups working on fluxonium, coupler design, or scaling toward surface-code lattices. It is also useful for quantum-error-correction theorists who need realistic device-level numbers for connectivity, ZZ crosstalk, and gate fidelity to inform code design, and for graduate students in quantum device physics looking for a worked example of how an architecture is characterized from coherence times up through multi-qubit entanglement. Readers without a superconducting-circuit background will find the device sections demanding but the benchmarking narrative and conclusions accessible.

Authors’ abstract

A central challenge in fluxonium-based quantum processors is the extension of the qubit connectivity to two-dimensional lattices compatible with quantum code-error correction. Here, we present a fluxonium quantum processor that employs lumped-element resonator couplers which realizes, for the first time, a connectivity-four unit cell with suppressed parasitic interactions. We achieve parallel single-qubit gate fidelities exceeding 99.9 % in simultaneous randomized benchmarking experiments, while maintaining residual static ZZ interactions below 1 kHz across all coupled qubit pairs. We implement resonator-induced phase (RIP) gates and benchmark two-qubit gate fidelities exceeding 99 % using interleaved randomized benchmarking. To cancel spectator errors observed in two-qubit operations, we implement a refocused RIP gate, recovering coherent control in the presence of multi-qubit connectivity. Furthermore, we prepare Greenberger-Horne-Zeilinger states of up to five qubits with a tomographic fidelity of 90 %, verifying multi-qubit entanglement within the unit cell. These results establish the fluxonium-resonator-fluxonium architecture as a viable approach to realizing densely connected fluxonium processors and provide a scalable path toward quantum error-correction-compatible processor architectures.