Enabling full localization of qubits and gates with a multi-mode coupler

Tunable couplers are a key building block of superconducting quantum processors, enabling high on-off ratios for two-qubit entangling interactions. While qubit-qubit interaction can be turned off, residual wavefunctions delocalize single-qubit excitations over the device, yielding weak effective couplings that manifest as unintended crosstalk. Moreover, conventional single-mode couplers lack independent control over interactions in the one- and two-excitation manifolds, leading to unitary errors such as leakage during gate operations. Here, we propose a multi-mode tunable coupler that enforces complete localization, yielding near-perfect qubit isolation at the decoupled point. We further show that the additional degrees of freedom in the coupler enable independent and nonlinear control of effective interactions across distinct excitation manifolds, with large on-off ratios. This architecture provides a new route toward the next generation of couplers for scalable and high-fidelity gate operations in superconducting quantum processors.

💡 Research Summary

The paper addresses fundamental limitations of conventional single‑mode tunable couplers (SMCs) used in superconducting quantum processors. Although SMCs can turn off the effective two‑qubit interaction, they inevitably leave residual wavefunction delocalization: single‑excitation states spread across neighboring qubits, generating unwanted crosstalk and weak effective couplings. Moreover, because a single tunable mode mediates the interaction, the coupling strengths in the one‑excitation and two‑excitation manifolds are not independently controllable. This leads to leakage and unitary errors during common two‑qubit gates such as iSWAP (where unwanted |11⟩↔|02⟩ or |20⟩ transitions appear) and CPHASE (where spurious ZZ phase shifts arise).

To overcome these issues, the authors propose a multi‑mode tunable coupler (TMC) architecture that incorporates two (or more) adjustable resonant modes together with a tunable inter‑mode coupling term H_cc(λ). The key insight is that by diagonalizing the coupler Hamiltonian with a unitary transformation U_c(λ) that depends on the external control λ (e.g., flux bias), the full system Hamiltonian can be reorganized into N disjoint blocks, each containing a single qubit and its associated set of coupler modes. In this block‑diagonal form, each qubit evolves only within its own invariant subspace, achieving “localized decoupling”: the qubit wavefunction remains confined to its own block, eliminating inter‑qubit delocalization even when the nominal interaction is turned off.

Because the inter‑mode coupling λ is tunable, the effective frequencies and coupling strengths of each block can be independently adjusted. This enables separate control of the effective exchange coupling J_00 in the one‑excitation manifold and the conditional coupling J_11 (and related higher‑order terms) in the two‑excitation manifold. Consequently, an iSWAP gate can be realized by turning on only J_00 while keeping J_11 strictly zero, suppressing leakage from the computational subspace. Conversely, a CPHASE gate can be performed by activating only J_11, providing a clean ZZ interaction without contaminating the iSWAP channel. The architecture also supports simultaneous activation of both manifolds for continuous‑fSim‑type gates.

The authors develop a numerical “overlap method” to extract effective couplings directly from eigenvectors and wavefunction localization, avoiding frame‑dependent ambiguities of conventional Schrieffer‑Wolff‑type reductions. Simulations on realistic circuit parameters (transmon qubits at 5–7 GHz, SQUID‑based tunable modes, and capacitive inter‑mode links) demonstrate on‑off ratios exceeding two orders of magnitude and residual couplings below 10 kHz, well within fault‑tolerance thresholds. Time‑domain simulations of gate sequences show gate errors reduced to ≈0.05 % for both iSWAP and CPHASE, a substantial improvement over state‑of‑the‑art SMC implementations.

Practical circuit realizations are outlined: two flux‑tunable SQUID resonators provide the two modes, while a flux‑biased coupler loop implements the tunable H_cc(λ). The design is compatible with planar fabrication processes and can be scaled to multi‑qubit modules. By arranging each module as an isolated qubit‑coupler block and connecting modules via low‑crosstalk microwave buses, the authors propose a modular chip architecture that preserves localized decoupling while allowing selective long‑range interactions when needed. This modularity is especially advantageous for surface‑code error‑correction layouts, where high connectivity and low static crosstalk are simultaneously required.

In summary, the paper makes four major contributions: (1) a theoretical framework for achieving complete wavefunction localization through tunable multi‑mode couplers, (2) independent, nonlinear control of interaction strengths in distinct excitation manifolds, (3) demonstration of dramatically improved on‑off ratios and reduced leakage/ZZ errors, and (4) concrete circuit designs and a scalable modular architecture for large‑scale superconducting quantum processors. These advances pave the way toward meeting the stringent error budgets needed for fault‑tolerant quantum computation.