Scalable Suppression of XY Crosstalk by Pulse-Level Control in Superconducting Quantum Processors

TL;DR

The paper tackles XY crosstalk in densely connected superconducting qubits by introducing a pulse-level framework that combines continuous frequency modulation (FM) and dynamical decoupling (DD) to suppress residual exchange interactions. By leveraging matched gate times and phase-averaging, the authors show that both idle and single-qubit gate infidelities can be reduced by multiple orders of magnitude, with scalable performance demonstrated in a five-qubit layout where a central qubit couples to four neighbors. The approach remains agnostic to coupling strengths and is compatible with existing hardware, offering a practical route to high-fidelity operation in dense architectures. Limitations and future directions include flux-noise concerns for FM, finite-width pulses for DD, and extending these strategies to two-qubit gates and larger, more complex arrays.

Abstract

As superconducting quantum processors continue to scale, high-performance quantum control becomes increasingly critical. In densely integrated architectures, unwanted interactions between nearby qubits give rise to crosstalk errors that limit operational performance. In particular, direct exchange-type (XY) interactions are typically minimized by designing large frequency detunings between neighboring qubits at the hardware level. However, frequency crowding in large-scale systems ultimately restricts the achievable frequency separation. While such XY coupling facilitates entangling gate operations, its residual presence poses a key challenge during single-qubit controls. Here, we propose a scalable pulse-level control framework, incorporating frequency modulation (FM) and dynamical decoupling (DD), to suppress XY crosstalk errors. This framework operates independently of coupling strengths, reducing calibration overhead and naturally supporting multi-qubit connectivity. Numerical simulations show orders-of-magnitude reductions in infidelity for both idle and single-qubit gates in a two-qubit system. We further validate scalability in a five-qubit layout, where crosstalk between a central qubit and four neighbors is simultaneously suppressed. Our crosstalk suppression framework provides a practical route toward high-fidelity operation in dense superconducting architectures.

Figures (19)

• Figure 1: Schematic of a nearest-neighbor qubit array with XY interactions, indicated by gray dashed lines. The pair $Q_1$ and $Q_2$ within the dashed orange outline defines a minimal two-qubit subsystem used to analyze XY-induced crosstalk under single-qubit control. The XY coupling between $Q_1$ and $Q_2$ is highlighted by red dashed lines.
• Figure 2: Idle gate infidelity, $1 - F^{\rm CD}_{I_1 I_2}$, as a function of time under crosstalk dynamics. Parameters are chosen as $\Delta/2\pi = 50~\mathrm{MHz}$ and $J/2\pi = 5~\mathrm{MHz}$.
• Figure 3: Suppression of idle gate infidelity under frequency modulation. (a) Waveform of a sinusoidal $Z_2$ drive with $N=4$ modulation cycles applied to $Q_2$. (b) Idle gate infidelity of one idle operation as a function of the coupling strength $J$ for different numbers of modulation cycles $N$, evaluated at the matched gate time $T_M = 20$ ns. (c) Idle gate infidelity of consecutive idle operations as a function of time for different numbers of modulation cycles $N$, with $T_M = 20$ ns and $J/2\pi = 5$ MHz. Results under crosstalk dynamics (CD) are shown for comparison.
• Figure 4: Suppression of $X_{1}$ gate infidelity under frequency modulation. (a) Waveforms of the $X_1$ control pulses applied to $Q_1$ and the sinusoidal $Z_2$ drive with $N=4$ modulation cycles applied to $Q_2$. (b) $X_{1}$ gate infidelity of one $X_1$ operation as a function of the coupling strength $J$ for different numbers of modulation cycles $N$. (c) $X_{1}$ gate infidelity of consecutive $X_1$ operations as a function of time for different numbers of modulation cycles $N$, with a fixed $J/2\pi = 5$ MHz. Odd numbers of consecutive $X_{1}$ gates are applied to perform a net $X_1$ operation.
• Figure 5: Suppression of idle gate infidelity under dynamical decoupling. (a) Waveform of the DD Z-4 sequence containing four $Z_2$ gates applied to $Q_2$. (b) Idle gate infidelity of one idle operation as a function of the coupling strength $J$ under DD Z-4. (c) Idle gate infidelity of consecutive idle operations as a function of time under DD Z-4, with $J/2\pi = 5$ MHz.