Slow drives could connect distant superconducting qubits through a shared cavity

How can a shared connection remain selective?

A quantum processor needs operations that entangle pairs of qubits. If only neighbouring qubits can interact, connecting distant partners requires extra operations, adding time and opportunities for errors. A common resonator offers another possibility: many qubits could share one electromagnetic connection. The challenge is making only the intended partners interact.

In a preprint submitted to arXiv on November 9, 2026, Elena Lupo and colleagues propose a solution for superconducting flux qubits. Their work combines analytical calculations with numerical simulations, not experiments on a working processor.

The proposed architecture couples several flux qubits to one superconducting cavity, with each qubit controlled through its own local magnetic-flux line. Strong, relatively slow signals would activate an entangling interaction between a selected pair without requiring a separate tunable coupler for every connection.

Tilting the qubits slowly

The qubits have a static magnetic-flux setting called a sweet spot, where their transition frequencies are insensitive to small flux changes to first order. That static setting stays fixed during the proposed gate, although an alternating control signal is applied around it.

The signals oscillate much more slowly than the qubits' own transitions: typically hundreds of megahertz rather than several gigahertz. Under suitable conditions, the qubit states approximately follow the changing instantaneous energy states. This is the meaning of an adiabatic drive.

In the authors' mathematical description, the drive tilts each qubit's state axis. That changes how the qubit couples to the cavity, producing an effective interaction called ZZ. This interaction accumulates different phases for different joint qubit states and can therefore entangle the pair.

The interaction strength depends on the drive amplitudes and their relative phase, rather than on matching a drive to a qubit or cavity transition. However, frequency choice still matters. Intended partners receive matching drive frequencies; idle qubits remain undriven. Separate pairs can use different frequencies, provided unwanted harmonic coincidences and other resonances are avoided.

What the simulations show

The researchers numerically evolved a model treating the qubits as two-level systems coupled to a single cavity mode. For a two-qubit gate lasting 55 nanoseconds, they found broad drive-frequency windows with coherent errors around or below 0.0001. In one parameter regime, errors reached approximately one part in a million at the lowest drive frequencies considered.

Those figures require important qualifications. The fidelity calculation projects the evolution onto the cavity's zero-photon subspace, includes correcting single-qubit rotations and optimizes control parameters. It is not a measurement of total error in hardware.

A three-qubit simulation tested whether an undriven spectator spoiled a gate between the other two qubits. Broad frequency regions retained errors below 0.0001, but sharp error peaks appeared near unwanted resonances and near frequencies where qubit states strongly mix.

For four qubits, the team simulated two simultaneous gates on separate pairs. Away from problematic frequency combinations, optimized errors approached 0.0001. Driving both pairs at the same frequency instead generated unwanted interactions between them. Parallel operation therefore becomes a frequency-allocation problem, not an unrestricted ability to run every connection simultaneously.

Noise protection is not a complete error budget

The analysis predicts reduced correlated two-qubit errors from slow fluctuations in qubit frequency. Because the desired ZZ interaction commutes with single-qubit phase evolution, such fluctuations do not disrupt it in the same way as certain resonance-based gates. Single-qubit phase errors can nevertheless remain.

The authors also argue that complete drive cycles cancel the first-order response to effectively static flux noise. This is a theoretical noise-filtering argument, not a demonstrated noise-free gate.

Crucially, relaxation is separate from the quoted coherent errors. Using a representative qubit lifetime of 50 microseconds, the paper estimates a relaxation-related error of roughly 0.001 for a 55-nanosecond gate—larger than the best coherent simulation results.

A proposal awaiting hardware

The paper estimates small residual static crosstalk for an illustrative 20-qubit register and sketches connected cavity modules for larger processors. These are calculations and architectural possibilities, not demonstrations at those scales.

The central result is a proposed mechanism for selective connectivity through a shared cavity, supported by small-system simulations. Whether it delivers comparable performance in hardware remains an experimental question.