Key points are not available for this paper at this time.
Logical qubits encoded into a quantum code exhibit improved error rates when the physical error rates are sufficiently low, below the pseudothreshold. Logical error rates and pseudothresholds can be estimated for specific circuits and noise models, and these estimates provide approximate goals for qubit performance. However, estimates often assume uniform error rates, while real devices have static and/or dynamic distributions of non-identical error rates and may exhibit outliers. These distributions make it more challenging to evaluate, compare, and rank the expected performance of quantum processors. We numerically investigate how the logical error rate depends on parameters of the noise distribution for the subsystem surface code and the compass code on a subdivided hexagonal lattice. Three notable observations are found: (1) the average logical error rate depends on the average of the physical qubit infidelity distribution without sensitivity to higher moments (e.g., variance or outliers) for a wide parameter range; (2) the logical error rate saturates as errors increase at one or a few "bad" locations; and (3) a decoder that is aware of location specific error rates modestly improves the logical error rate. We discuss the implications of these results in the context of several different practical sources of outliers and non-uniform qubit error rates.
Carroll et al. (Mon,) studied this question.