Investigating Crosstalk and Correlated Errors with Randomized Benchmarking on Multiple Qubits*

Quantum computing with superconducting circuits has recently shown progress in implementing multi-qubit quantum processors with promising performance. To ensure scalability of these systems for quantum applications, it is critical that any errors remain small, local, and uncorrelated when increasing the number of qubits. In particular, the threshold theorem for quantum error correction is typically derived under the assumption of uncorrelated errors and constant error rates for each qubit. In this talk, we will present characterizations of cross-talk during gate operations using an extended simultaneous randomized benchmarking (RB) protocol. We apply sequences of randomly chosen elements from the single qubit Clifford group simultaneously on up to four qubits and use single-shot readout to measure individual qubits and all σ_z-correlators between the qubits. From these measurements, we estimate the multi-qubit error as the average n-qubit infidelity per application of n single-qubit Clifford gates. We find the correlated errors to be composed mainly by two-qubit correlations. We also extend this method to analyse the crosstalk and correlated errors induced during two-qubit RB.