⚛️ quantum physics

All-to-All Connectivity in Superconducting Qubits via Metamaterial-Mediated Interactions

This paper proposes a metamaterial-mediated architecture using dual-rail or spin-locked transmons to achieve scalable all-to-all qubit connectivity for quantum LDPC codes, effectively suppressing spurious spectator entanglement at the hardware level while demonstrating high-fidelity entangling gates and joint parity measurements limited primarily by photon loss and mode crowding.

Original author: Or Golan, Alexey V. Gorshkov, Andreas Butler, Gihwan Kim, Fernando G. S. L. Brandão, Oskar Painter, Alex Retzker

Published 2026-11-03

📖 5 min read🧠 Deep dive

CC BY 4.0

Original author: Or Golan, Alexey V. Gorshkov, Andreas Butler, Gihwan Kim, Fernando G. S. L. Brandão, Oskar Painter, Alex Retzker

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the author. For technical accuracy, refer to the original paper. Read full disclaimer

Building a computer that can solve problems beyond the reach of any machine today requires a delicate balancing act. The most promising candidates for these machines are quantum computers, which use the strange rules of physics to process information in ways classical bits cannot. However, these machines are incredibly fragile; the slightest disturbance from the environment can scramble their calculations. To fix this, scientists use error correction, a method where many physical components work together to protect a single piece of information. The most common approach, known as the surface code, works well but is inefficient, requiring thousands of physical parts to create just one reliable unit of information. A more efficient path exists using a different type of code, but it demands a level of connectivity that current machines cannot easily provide: the ability for any part of the computer to talk directly to any other part, regardless of distance.

Superconducting quantum processors, which use tiny circuits cooled to near absolute zero, are a leading platform for this technology. They naturally connect their neighbors well, but linking distant parts usually involves complex wiring or moving information through a shared channel. This shared channel, often a resonator, creates a problem: when one part of the system is active, it unintentionally disturbs all the other quiet parts nearby, creating errors that are hard to fix. Researchers at the AWS Center for Quantum Computing and the Hebrew University of Jerusalem have proposed a new architecture that solves this specific problem. By rethinking how the basic units of information are built, they have shown a way to connect many qubits to a single shared medium without the unwanted side effects that have previously limited scalability.

The researchers focused on a structure called a metamaterial waveguide, which is essentially a long chain of coupled resonators. This chain acts like a highway for quantum information, allowing signals to travel between any two points. In previous attempts to use such a highway, the act of driving a signal between two specific qubits would inadvertently shift the frequency of every other idle qubit connected to the same chain. This created a web of unwanted entanglement, where the "spectator" qubits became confused and entangled with the active gate, ruining the calculation. To fix this, earlier methods required complex, active pulse sequences to cancel out the interference, a process that became increasingly difficult as the number of qubits grew.

The new work introduces a different strategy: instead of trying to cancel the interference after it happens, the researchers prevent it from happening in the first place by changing the nature of the qubits themselves. They encode each logical qubit as a pair of superconducting circuits working together, a configuration known as a dual-rail transmon. This pairing creates a protective gap that naturally suppresses the unwanted interaction with the shared waveguide. In simulations, this hardware-level protection reduced the disruptive coupling by a significant margin, pushing the error rate down to a level where active cancellation is no longer strictly necessary. The result is a system where the unwanted noise is suppressed at the source, rather than fought with software.

With this protection in place, the researchers demonstrated how to perform two-qubit operations and joint measurements across the entire chain. By applying a specific two-tone drive to the waveguide, they could activate an interaction between any pair of encoded qubits. This interaction allows them to perform a gate that entangles the two qubits, a fundamental step for computation. Crucially, because the system is built on a metamaterial with many distinct modes, different pairs of qubits can be addressed simultaneously through different frequencies. This means the computer can run many operations in parallel, a feature that is essential for the efficiency of the advanced error-correcting codes the researchers aim to support.

The team simulated a system where two dual-rail qubits shared a chain of three resonators with a specific lifetime of 30 microseconds. In these simulations, the gate operation reached an infidelity of 1.4 × 10⁻³, meaning the error rate was very low. The primary source of this remaining error was the loss of photons from the resonator, a physical limitation of the hardware. The researchers also found that the size of the array is limited by a trade-off: as the chain gets longer, the frequency spacing between the available modes shrinks, making it harder to target a single pair without disturbing its neighbors. However, by increasing the lifetime of the resonators and the coupling strength between them, the system can support more qubits. In a scenario with a resonator lifetime of 150 microseconds, the simulations suggest that a single metamaterial could host 13 qubits on 25 resonators while staying within a strict error budget.

This work does not claim to have built a fully functional quantum computer yet, but it provides a clear theoretical and simulated path forward. The findings suggest that by encoding qubits in a dual-rail fashion, the fundamental barrier of unwanted spectator entanglement can be overcome. This opens the door to using superconducting circuits for the highly connected architectures required by the next generation of error-correcting codes. The study highlights that the key to scaling these machines may not just be in building better components, but in designing the components to be inherently compatible with the complex web of connections required for large-scale quantum computation.

Technical Summary: All-to-All Connectivity in Superconducting Qubits via Metamaterial-Mediated Interactions

Problem Statement
Superconducting qubits are a leading platform for fault-tolerant quantum computing, particularly for the surface code, which relies on native nearest-neighbor couplings. However, scaling to quantum low-density parity-check (qLDPC) codes, which offer constant encoding rates and growing distances, requires non-local connectivity that planar nearest-neighbor layouts cannot provide. Existing routes to non-local connectivity—direct links, one-to-many buses, and shuttling—suffer from limitations in reach, selectivity, or parallelism.

A specific challenge arises when using a shared bus, such as a metamaterial waveguide (a chain of coupled resonators), to mediate interactions. In a "bare-transmon" architecture, the dispersive coupling between a populated bus mode and every idle "spectator" qubit entangles the spectators with the gate operation. This spurious entanglement burdens the system, requiring active dynamical decoupling sequences (e.g., the "flowers" sequence) to cancel the error. As the number of qubits and driven modes increases, the total dispersive error grows linearly, making the approach difficult to scale.

Methodology
The authors propose suppressing this dispersive coupling at the hardware level by encoding physical qubits as either dual-rail transmons (DRTs) or spin-locked transmons (SLTs).

Encoding: In a DRT, two resonant transmons are strongly coupled to form a logical qubit. The logical states are symmetric and anti-symmetric superpositions of the two rails. This encoding protects against dephasing and, crucially, suppresses the dispersive coupling to the bus modes.
Mechanism: The dispersive coupling, which scales as $1/\Delta^2$ for bare transmons (where $\Delta$ is the qubit-mode detuning), is suppressed to fourth order ( $1/\Delta^4$ ) in the encoded architectures. This suppression is passive and built into the hardware, removing the need for extensive active decoupling sequences.
Interaction: A two-tone cross-resonance drive is applied to the encoded qubits. The drive frequencies are centered on a shared metamaterial mode, with one tone blue-detuned and one red-detuned by the dual-rail gap (or spin-locking gap). This drive activates a Mølmer-Sørensen interaction between the encoded qubits and the mode, enabling both entangling gates and joint parity measurements.
Architecture: The system utilizes a metamaterial waveguide consisting of $N$ identical resonators with nearest-neighbor coupling $J$ . Qubits are attached to the chain such that one DRT is placed per two resonators to minimize mediated cross-Kerr interactions.

Key Contributions

Hardware-Level Suppression of Dispersive Errors: The paper demonstrates that encoding qubits as DRTs or SLTs suppresses the spectator-qubit dispersive coupling from second to fourth order in the inverse detuning. This eliminates the primary scalability bottleneck of shared-mode architectures, where error correction sequences become increasingly burdensome with system size.
Selective and Parallelizable Operations: The two-tone drive allows for selective addressing of specific qubit pairs through different modes of the metamaterial. Unlike previous bus schemes that entangle all qubits simultaneously or require serial swapping, this approach enables parallel entangling gates and joint parity measurements across different pairs.
Joint Parity Measurement: The same interaction used for gates allows for the direct measurement of the joint logical operator $\tilde{X}\tilde{X}$ . By reading out the photon number of the shared mode, the system can perform a non-destructive parity check, a fundamental operation for Floquet codes and other qLDPC implementations.
Scalability Analysis: The authors derive a "mode-count ceiling" for the array size, bounded by the competition between resonator photon loss (which scales with gate duration) and spectral crowding (which requires longer pulses to resolve modes as the array grows). They show that the maximum number of qubits scales linearly with the mode lifetime ( $T_1$ ) and the inter-resonator coupling ( $J$ ).

Results

Gate Infidelity: Simulations of two dual-rail transmons sharing three resonators with a mode lifetime of $30\,\mu\text{s}$ yield a two-qubit gate infidelity of $1.4 \times 10^{-3}$ at a gate duration of $152\,\text{ns}$ . The error is dominated by photon loss (81%), with the remainder attributed to coherent leakage. This is lower than the infidelity of bare-transmon gates, even when those gates employ dynamical decoupling.
Array Capacity: The analysis indicates that the number of qubits a single metamaterial can host is limited by the error budget. For a target error budget of $10^{-3}$ , a system with a mode lifetime of $150\,\mu\text{s}$ and inter-resonator coupling of $80\,\text{MHz}$ can support 13 DRTs on 25 resonators. In this configuration, spectator erasure accounts for approximately one-seventh of the error budget.
Parallel Readout: The paper outlines a protocol for parallel parity measurements using auxiliary transmons and frequency combs, estimating measurement times of 3 to 5 $\mu\text{s}$ per pair for chains of 5 to 20 resonators.

Significance
The paper claims to present a viable architecture for achieving all-to-all connectivity in superconducting processors, a prerequisite for implementing qLDPC codes with constant encoding rates. By suppressing dispersive coupling at the hardware level, the approach removes the scalability barrier associated with active dynamical decoupling in shared-mode systems. The resulting architecture supports selective, parallelizable entangling gates and joint parity measurements, offering a path toward the non-local connectivity required for next-generation quantum error correction codes. The work suggests that with improved mode lifetimes and coupling strengths, superconducting processors could host arrays large enough to outperform surface codes in terms of logical qubit density.

More like this