Fig. 3: Walking implementation.
a, Detecting region slices for a Z-type bulk detector in the walking circuit. b, Measured logical error probability for a distance-5 and embedded distance-3 surface codes implemented using the walking circuit and decoder H*. c, Comparison of different decoders. d, Measured average leakage on the A and B data qubit sublattices after each walking circuit gate layer, with leakage data per qubit shown in the upper-right corner. To mitigate an outlier, the highlighted qubit used a single-level reset instead of a multi-level reset. Average leakage for both sublattices over 12 cycles is shown in the lower right. The steady state is reached rapidly, and no rise in leakage over time is visible. e, Comparison of the average detector autocorrelation for three circuits: the standard surface code without DQLR (orange), with DQLR (blue) and the walking circuit without DQLR (green). Long-time correlations in the standard code without DQLR are removed using walking or DQLR. f, Detector error budget from the component benchmarks.
