Hardware-efficient random circuits to classify noise in a multiqubit system

Abstract

In this work, we extend a multiqubit benchmarking technique known as the binned output generation (BOG) in order to discriminate between coherent and incoherent noise sources in the multiqubit regime. While methods exist to discriminate coherent from incoherent noise at the single and few-qubit level, these methods scale poorly beyond a few qubits or must make assumptions about the form of the noise. On the other end of the spectrum, system-level benchmarking techniques exist, but they fail to discriminate between coherent and incoherent noise sources. We experimentally verify the BOG against randomized benchmarking (RB) (the industry standard benchmarking technique) in the two-qubit regime, then apply this technique to a six-qubit linear chain, a regime currently inaccessible to RB. In this experiment, we inject an instantaneous coherent Z-type noise on each qubit and demonstrate that the measured coherent noise scales correctly with the magnitude of the injected noise, while the measured incoherent noise remains unchanged as expected. This demonstrates a robust technique to measure coherent errors in a variety of hardware.