A five-qubit fluxonium processor has paired near-coherence-limited single-qubit control with entanglement extending across all five qubits, according to an arXiv preprint. The reported median single-qubit gate fidelity was 99.957(5)%, close to the stated coherence limit of 99.96(1)%. The processor also produced a five-qubit GHZ state, a joint state used here to test multiqubit control, with an error-corrected tomographic fidelity of 90.0(2)%.
The hardware uses a fluxonium-resonator-fluxonium, or FRF, design with five fluxonium qubits in a connectivity-four unit cell. That unit cell is the central device reported in the study.
The supplied document is identified as arXiv:2608.25503v1 and dated 26 August 2026.
Single-qubit control came close to the stated limit
The study measured one-qubit control with simultaneous randomized benchmarking, using 48 ns gates on all qubits. In practical terms, the benchmark tested the qubits together, so the reported result reflects simultaneous operation across the processor.
The median fidelity of 99.957(5)% was close to the reported coherence limit of 99.96(1)%. Because the figure is a median, it summarizes the middle of the reported results rather than indicating identical performance for every gate.
The harder test was controlling two-qubit interactions
The researchers also looked for residual ZZ interaction, an unwanted coupling between connected qubits. For each connected pair, they ran 50 JAZZ sequences. The median residual rate was 30(215) Hz, while residual static ZZ was reported below 1 kHz across all coupled pairs.
Regular RIP-based CZ gates used a detuned, cosine-shaped 128 ns pulse on the coupler. In 30 interleaved randomized-benchmarking sequences, with a maximum length of 200 Cliffords, the pair labeled r15 reached 99.68(5)% fidelity when q2 was in its zero state, but 96.1(6)% when q2 was in its one state. Averaged across the two spectator-state cases, fidelity was 97.9(3)%.
The comparison showed that the measured two-qubit result could change with the state of q2. In this test, q2 acted as a spectator, meaning its state was varied while the benchmark assessed the gate on the target pair.
The refocused RIP protocol used two 96 ns coupler pulses. Averaged across the two spectator-state IRB sequences, it reached 99.20(5)% fidelity, with no measured dependence beyond statistical uncertainty.
Across the next-neighbor couplers, the median refocused CZ fidelity was 98.8(12)%, while the pair labeled r35 reached 96.2(2)%. The figures show that performance still varied from pair to pair even with the refocused sequence.
Entanglement reached five qubits, with readout still a constraint
The final test moved from individual gates to shared states. Three-, four- and five-qubit GHZ states were prepared and benchmarked with quantum state tomography, which reconstructs a state's density matrix from measurement results. Before that reconstruction, the analysis applied classical measurement-error correction to the recorded data.
The corrected tomographic fidelities were 96.9(2)% for three qubits, 95.4(2)% for four, and 90.0(2)% for five. The decline as more qubits were included gives a tougher test than the single-qubit median, while still reporting a five-qubit entangled state.
Readout remains an important caveat. The tomography appendix reported a median readout fidelity of 94%, with an asymmetric reported uncertainty of plus 1 and minus 3 percentage points. The smallest measured fidelity was 85.5%. The GHZ figures were obtained after classical measurement-error correction, so they are corrected tomographic results rather than raw readout results.
Taken together, the measurements describe the reported five-qubit unit cell as having strong one-qubit control, low residual static ZZ, a refocused two-qubit result without measured spectator-state dependence beyond statistical uncertainty, and entanglement measured through five qubits. The evidence concerns this unit cell, not a demonstrated larger array.
The acknowledgements report external financial support.
Paper data and sources
Original title: High-Fidelity Entangled States in a Connectivity-Four Fluxonium Quantum Processor
Authors: J. Schirk, N. Bruckmoser, S. M. Taubenberger et al.
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-26
DOI: Not available
Original paper · Full text