Work

Logical-Error Suppression in the Willow Surface-Code Memory: A Lambda Budget Extrapolated to Algorithmic Error Targets

Year: 2026
Type: article
Abstract: The December 2024 report of the Willow processor presented the first superconducting surface-code memory in which enlarging the code reduced the logical error rate: each step from distance three to five to seven divided the error per cycle by a suppression factor Lambda of 2.14, reaching 1.43e-3 per cycle at distance seven. A single measured suppression factor invites extrapolation, and extrapolation deserves an audit. This paper takes the published Lambda and the published distance-seven rate as fixed inputs and computes what they imply for the distances and qubit counts that algorithmic error targets demand. Holding Lambda constant, a per-cycle logical error rate of 1.00e-9 requires distance 45 and 4049 physical qubits in one rotated patch, while 1.00e-12 requires distance 63 and 7937 qubits. The mean waiting time between logical failures at distance seven is only 699 cycles, which at a microsecond-scale cycle time is well under a millisecond, so the demonstrated patch is a physics milestone rather than a computational memory. The budget also exposes the sensitivity of every projection to Lambda: because distance grows logarithmically in the target but the qubit count grows with the square of distance, a ten per cent erosion of Lambda inflates the largest patches by roughly a quarter. Suppression factors, not raw physical error rates, are therefore the currency in which fault-tolerant overhead should be quoted.
Language: English
Open Access status: green
OpenAlex ID: W7233604746
FWCI:
Cites:
Cited by:
Related to: