Origin of rare correlated error events limiting repetition code performance

Determine the physical origin of the rare correlated error events observed approximately once per hour during high-distance repetition code experiments on the 72‑qubit superconducting transmon processor, which impose a logical error-per-cycle floor near 10^−10. Characterize the mechanism(s) responsible for these spatially localized, anisotropic bursts in detector event rates and their exponential decay behavior to enable effective mitigation strategies.

Background

The paper demonstrates below-threshold operation of surface code memories and probes logical error sensitivity using high-distance repetition codes on a 72-qubit superconducting processor. While prior devices were limited by high-energy impact events creating an error floor around 10^−6, the current device surpasses this, yet reveals a new logical error floor near 10^−10.

In the repetition code experiments up to distance-29, the authors identify rare, spatially localized correlated error bursts that occur roughly once per hour. These events differ from previously observed high-energy impacts: they are confined to neighborhoods of about 30 qubits and decay exponentially with a time constant around 400 µs. The cause of these events is explicitly stated as unknown, and understanding them is highlighted as vital for fault-tolerant quantum computing.