arXiv:2610.20810
Syndrome-Resolved Decoding of Qubit Readout Errors
Han Zheng, Vladimir Pešić, Nikolay Gusarov, Andrea Lizzit, Taketo Imaizumi, Chulwoo Ahn, Marco Nicolas Mueller, Raymond Mencia, Edoardo Charbon, Vladimir E. Manucharyan
Abstract
Readout errors in superconducting qubits include incorrect state assignment, bit flips within the computational subspace, and leakage to non-computational states. These errors require different mitigation strategies, yet conventional measurement fidelity and repeatability metrics do not resolve their individual contributions. Previous repeated-readout studies have shown that temporal correlations reveal additional information about readout errors. Here we introduce the Readout Error Decoder (RED), which converts repeated thresholded binary records into a channel-resolved error budget. Controlled error injections validate the channel-specific response of the decoder. We demonstrate RED on both fluxonium and transmon qubits and across distinct leakage-lifetime regimes. RED identifies the error channels limiting readout performance without requiring dedicated discrimination of non-computational states, providing a basis for targeted readout optimization across different qubit modalities.
AI Summary
Conventional readout fidelity and repeatability do not distinguish incorrect assignment, computational-subspace bit flips, and leakage. This paper asks whether repeated thresholded binary measurements, combined with known inter-readout gates, can resolve these contributions without discriminating higher states directly. The Readout Error Decoder (RED) compares observed transitions with those expected under $I$, $X$, or $X$–$\tau$–$X$ operations. Isolated broken correlations estimate bit flips; two consecutive broken correlations estimate assignment errors; longer runs under alternating $X$ gates indicate persistent leakage. Short-lived leakage is inferred from excess isolated breaks under $X$–$\tau$–$X$ relative to an $I$ baseline: decay between the pulses makes leakage observable through a failed first pulse and successful second pulse. Experiments on two fluxonium devices and one transmon test the protocol. Injected bit flips produce near-unit response slopes, while injected leakage gives a slope of 0.887, attributed to incomplete transfer into a short-lived higher state. Duration sweeps show decreasing assignment error and increasing bit-flip error. A fluxonium delay sweep yields a leakage lifetime of $0.44\,\mu\mathrm{s}$. At operating points minimizing decoded total error, totals are 1.79%, 1.88%, and 5.11%, respectively, with different channel compositions. The syndrome mapping is approximate: adjacent errors can mimic other channels, with contamination entering at second order. Simulations show degraded relative accuracy for a weak channel when strongly outweighed by another. Short-lived leakage extraction requires survival until the probing gate and sufficient inter-pulse delay. Measurements use several-microsecond readout without parametric amplification; faster-readout implementations and extensions beyond superconducting qubits remain untested.
Error found: The short-lived leakage estimator measures leakage surviving until the first inter-readout X pulse and returning between the two pulses, without a quantified correction for leakage lost before that probe. Subtracting the I-sequence S1 rate from the X–tau–X rate leaves possible ordinary transition and gate-error contributions, so the difference is not established as a leakage-only rate. Normalizing error counts by successful readouts alone estimates p/(1-p), rather than the per-readout error probability p, even in an independent-error model.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 13 pages, 4 main figures, 4 supplementary figures