Fig. 7: Restoring code performance by post-selecting on leakage detection.
Improvement in the logical error rate εL via post-selecting on the detection of leakage for a MWPM decoder (green) and the decoder upper bound (red). The post-selection is based on the probabilities predicted by the HMMs (solid) or on those extracted from the density-matrix simulation (dashed), for 2 × 104 runs of 20 QEC cycles each. The physical error rate of a single transmon qubit under decoherence is also given (solid black). Detection of leakage allows for the restoration of the performance of the MWPM decoder, reaching the memory break-even point by discarding about ≈28% of the data. The logical error rates obtained from simulations without leakage (and without post-selection) are indicated by diamonds.
