Detecting Drift, Change, and Context Dependence in Qubit Experiments

Quantum information processors are sensitive and precisely controlled devices. As a result, their behavior — especially the operations implemented by “logic gate” pulses — is very easily influenced by external influences. This context dependence includes familiar failure modes such as drift (where the context is time) and crosstalk (where the context is the state or situation of a neighbor qubit). Detecting and characterizing context dependence is important for at least two reasons: (1) it can constitute a significant source of error on its own; and (2) it can corrupt protocols designed to probe other errors, like randomized benchmarking and tomography. We present two techniques that we've been using to detect and characterize context dependence in experimental systems. The first is simple, and generic for discrete contexts (e.g., whether a neighboring qubit is driven or not); the second is specifically adapted to tracking drift over time. Both protocols are specifically designed to analyze count data, as produced by (e.g.) randomized benchmarking or tomography, rather than requiring specialized experiments.