Crosstalk

In an ideal quantum processor, applying a gate to one qubit leaves all other qubits completely undisturbed. In practice, the physical couplings that enable two-qubit gates also introduce residual interactions that are always present, even when a qubit pair is supposed to be idle. In superconducting processors, the dominant source of this problem is the ZZ coupling: a dispersive interaction between adjacent transmon qubits that shifts the frequency of qubit B depending on the state of qubit A. Because transmons are coupled through shared resonators or direct capacitive links, this ZZ term is always on and accumulates phase errors on idle qubits whenever a neighboring qubit is being driven.

Spectator errors and frequency crowding

A spectator qubit is one that is not the intended target of a gate but whose state is nonetheless perturbed during the gate’s execution. Spectator errors arise from two mechanisms. The first is coherent: the always-on ZZ coupling imprints a conditional phase on the spectator that depends on the control qubit’s trajectory during the gate pulse. The second is incoherent: the drive field applied to a target qubit leaks power into nearby qubits through shared electromagnetic modes, rotating them slightly off their intended states. As processor qubit counts grow toward hundreds or thousands, frequency crowding compounds both mechanisms. Qubits must be assigned distinct transition frequencies to be individually addressable, but the available frequency band is finite. When frequencies are packed tightly, the leakage pathways multiply and spectator errors accumulate faster than can be compensated by simple calibration.

Impact on two-qubit gate fidelity

Crosstalk is one of the primary reasons that two-qubit gate fidelities measured in isolation (using randomized benchmarking on a pair of qubits while all others are idle) differ from fidelities observed during full-circuit execution (when many gates run simultaneously). Simultaneous gate benchmarking protocols deliberately run gates on all qubit pairs at once to expose crosstalk-induced degradation that single-pair benchmarks miss. On current superconducting hardware, crosstalk can degrade simultaneous two-qubit gate fidelity by one to several percent relative to isolated benchmarks, which is a meaningful penalty for deep circuits where thousands of gates must execute reliably.

Mitigation techniques

Several strategies reduce crosstalk without eliminating the physical couplings that enable computation. Dynamical decoupling inserts sequences of Pauli pulses on idle qubits timed to refocus the ZZ phase accumulation. Simultaneous gate calibration optimizes pulse amplitudes and phases for each qubit pair assuming all neighbors are also active, rather than calibrating each pair in isolation. Frequency allocation algorithms treat qubit frequency assignment as a graph coloring problem, seeking assignments that minimize ZZ coupling between connected pairs while keeping all qubits within their coherent operating band. Tunable coupling elements, used in some superconducting architectures, physically switch the qubit-qubit coupling toward zero during idle periods, suppressing the always-on ZZ interaction at the cost of additional control hardware and potential flux noise.

See also

• Superconducting Qubit
• Gate Fidelity
• Decoherence
• Qubit Connectivity