Scalable Fluxonium Quantum Processors via Tunable-Coupler Architecture

Abstract: Superconducting quantum processors have largely converged on transmon-based architectures, while alternative qubit modalities with intrinsic error protection have lacked a demonstrated path to scalable system integration. In particular, although tunable-coupler-mediated interactions have been validated for small fluxonium systems, it remains unclear whether such designs can be scaled to a multi-qubit lattice. Here, we establish a scalable fluxonium processor architecture based on a modular qubit-coupler unit cell engineered to suppress residual interactions and spectator errors in a many-qubit lattice. The system enables parallel single-qubit gate fidelities approaching 99.99% and two-qubit CZ gate fidelities around 99%. With an optimized gate duration of 32 ns, the best CZ gate fidelity reaches 99.9%. We further validate this architecture in a 22-qubit processor based on the same configuration, where parallel operations enable the deterministic generation of Greenberger-Horne-Zeilinger states involving up to 10 qubits. Together, these results demonstrate that the fluxonium-tunable-coupler unit cell composes without emergent interaction pathologies and establish fluxonium as a scalable superconducting qubit platform.

• The paper establishes a scalable fluxonium-based processor by integrating 22 qubits with modular tunable couplers to overcome spectral crowding and achieve 99.9% CZ gate fidelity.
• High-fidelity two-qubit gates are realized through the fluxonium’s multilevel structure, enabling operations as fast as 32 ns with state-dependent hybridization and detuning below 200 kHz.
• System-level validation via multi-qubit GHZ state preparation confirms robust suppression of residual interactions (ZZ <1 kHz, XX <10 kHz) and reliable scalability in parallel operations.

Scalable Fluxonium Quantum Processors via Tunable-Coupler Architecture: An Analysis

Architectural Overview and Motivation

The work establishes a scalable superconducting quantum processor based on fluxonium qubits coupled through modular, tunable transmon couplers. The architecture addresses scalability limitations inherent to transmon-based systems—most critically, spectral crowding and residual coupling—by leveraging the fluxonium modality’s large anharmonicity and low-frequency computational space.

The device layout comprises a chain of 22 fluxonium qubits ($Q_1$–$Q_{22}$) with 21 capacitively coupled transmon tunable couplers. Each modular “FTF” (Fluxonium-Transmon-Fluxonium) unit supports individualized frequency tuning and microwave drive, enabling strong, selective coupling during entangling gate operation and efficient decoupling otherwise. Capacitive qubit-qubit crosstalk is intrinsically suppressed by physical and spectral separation, distinguishing this approach from transmon architectures that require interference-based cancellation or complicated frequency allocation procedures.

Figure 1: Processor architecture for the 22-qubit device, featuring cascaded FTF units and discrete microwave/flux bias infrastructure.

High-Fidelity Parallel Two-Qubit Gates

Gate operations exploit the fluxonium’s multilevel structure: computational transitions ($|g\rangle \leftrightarrow |e\rangle$) reside in the 100–500 MHz band, while higher excited-state transitions ($|e\rangle \leftrightarrow |f\rangle$) sit in 4–5 GHz. Two-qubit controlled phase (CZ) gates are activated by flux-pulsing the intermediate transmon into resonance with the fluxonium’s plasmon mode, enabling microwave-driven, state-dependent hybridization with high spectral selectivity.

In the four-qubit demonstration subsystem, interleaved randomized benchmarking yields a best CZ gate fidelity of 99.9% at 32 ns, with a typical range near 99%. Parallel operations using random circuit sampling (RCS) extract error-per-cycle fidelities of 98–99%—establishing high stability under simultaneous activation and minimal crosstalk between neighboring modules.

Figure 2: Parallel two-qubit gate operation, including level hybridization diagrams, benchmarking data, and circuit sampling structure.

Suppression of Residual Interaction and Crosstalk

An essential criterion for scalability is the suppression of residual ZZ and XX interactions and spectator-induced errors, particularly during parallel operation. Experimental Ramsey-type and conditional $T_1$ measurements show that all residual ZZ couplings in 5-qubit chains are consistently below 1 kHz in both coupling-OFF and coupling-ON conditions. Residual XX interactions are under 10 kHz. Single-qubit gate fidelities remain in excess of 99.9% throughout large-scale arrays, underscoring the negligible effect of spurious coupling.

Spectator-induced errors (cross-talk during two-qubit gate execution) are quantitatively analyzed by measuring the target gate’s detuning and conditional phase error given different spectator qubit states. In the coupling-OFF regime, induced frequency shifts are under 200 kHz with corresponding phase errors $<0.005$ radians, indicating robust parameter coherence across spectator configurations.

Figure 3: Characterization of residual ZZ interactions and spectator errors in a multi-qubit chain. Both metrics confirm scalability without emergent crosstalk pathologies.

Multi-Qubit Entanglement and System-Level Validation

System-level controllability is benchmarked via the deterministic generation of multi-qubit GHZ states. Using deep circuit calibration, coordinated initialization, and parallel CNOTs (compiled from native CZs), the system reliably prepares GHZ states up to 10 qubits. For $N\leq5$, quantum state tomography reconstructs density matrices with high-fidelity off-diagonal coherence. Parity oscillation protocols for $N=4,7,10$ show oscillation order and amplitude matching theoretical predictions, confirming the state’s global coherence.

Fidelity decays with growing $N$, consistent with cumulation of two-qubit errors; however, fidelity remains above the multipartite entanglement threshold ($>0.5$) for $Q_{22}$0. The experimental decay tracks an independent error model based on characterized single- and two-qubit gates without significant additional loss from parallel operation or residual interaction, validating the architectural composition’s integrity.

Figure 4: Results for multi-qubit GHZ state preparation, including density matrices (tomography), parity oscillations, and fidelity/harmonic order scaling with qubit number.

Extended Device Characterization

Comprehensive device benchmarking maps the gate fidelities and coherence metrics across all 22 qubits. Single-qubit gate fidelities approach 99.99% on best qubits and exceed 99.9% in parallel operation. Two-qubit gate fidelities display broader variation, limited by residual inhomogeneities in device fabrication (with a maximum of 99.26% and average of 95.65%). These spread metrics are not architectural limitations—fidelity can reach the short-chain subsystem’s optimum where local coherence permits.

Figure 5: Parallel gate performance map across the 22-qubit chain, annotating single- and two-qubit fidelities on the system layout.

Theoretical and Practical Implications

The presented architecture demonstrates that a modular fluxonium–tunable coupler unit cell is physically composable to at least 22 qubits, enabling large-scale parallel quantum logic without unanticipated interaction or calibration complexity. The detuning-based coupling suppression is robust, even as the lattice grows, eliminating the fine-tuned cancellation engineering necessitated by certain transmon designs.

From a practical perspective, the high coherence, strong gate fidelity, and crosstalk resilience position the architecture as a candidate for surface code blocks and error-corrected logical qubits, provided two-dimensional arrangements are realized. The approach is extensible to planar processors by leveraging double-transmon coupler schemes, which can accommodate increased control wiring density.

Future Outlook

Improvements in fabrication uniformity and materials are required to close the gap between the best and worst performing units in long chains. The demonstrated modularity and minimal residual interaction pave the way for surface code layouts and logical qubit encoding. Efforts toward double-transmon coupler designs, as referenced, will be pivotal in enabling scalable, neighbor-rich topologies with manageable cross-module wiring.

In principle, the architecture supports scaling to hundreds of qubits without encountering emergent spectral crowding or crosstalk-induced fidelity collapse, provided device uniformity is maintained. The demonstrated system serves as a viable alternative to transmon-centric approaches, with fluxonium’s large anharmonicity and unique energy scaling likely to ease system integration at larger scales.

Conclusion

The fluxonium-tunable coupler processor provides a validated, scalable route to large-scale superconducting quantum computation. It achieves high-fidelity, parallelizable gates with effective isolation, supporting multi-qubit entanglement across extended systems without architectural pathologies. These findings establish fluxonium as a practical superconducting qubit platform for next-generation, error-corrected quantum computing deployments and indicate a clear path towards two-dimensional, high-connectivity layouts (2604.13363).