An upper limit on narrowband periodic modulation of the decoherence rate in superconducting qubits: a line search in Google's surface-code memory data

Within the Stochastic Rupture (SR) framework, the only channel through which an external influence could act on quantum measurement without violating the elimination theorem for the Born rule is the rate of decoherence Γ, never the individual outcomes. This makes a concrete, falsifiable prediction available to test on existing hardware: a narrow spectral line in the time series of detection-event rates of a quantum error-correction (QEC) memory experiment. We perform such a search on the publicly released data of Google’s 105-qubit Willow surface-code memory experiments (Zenodo 13273331), analysing the r = 250 slice: 28 experiments (18 × d=3, 8 × d=5, 2 × d=7), 5 × 104 shots each, 1.4 × 106 shots in total, covering 22–451 kHz. The result is null. No frequency exceeds the Bonferronicorrected threshold zcrit = 3.76 in any per-distance or global stack, and no frequency shows z > 2 in more than five of the 28 experiments. The pipeline is not insensitive: it recovers three expected structures in the same data — a broadband low-frequency red-noise ramp (z falling monotonically from 27.7 to 6.0 over five contiguous bins), a coloured background ∼1.7× above shot noise from the code’s round-to-round detector correlations, and long memory across shots (HDFA = 0.632 ± 0.097, 0.757 ± 0.024, 0.771 ± 0.026 for d = 3, 5, 7). Calibrating by injection into the real data, where the response is quadratic (∆z = c a2) with c rising steeply with code distance, we set what is to our knowledge the first upper limit of this kind: any narrowband periodic modulation of Γ in the band 22–451 kHz has fractional amplitude a < 0.99% at 99% confidence. The negative result and its limit are reported as the primary deliverable; the search is intended to be repeated on the remaining 392 experiments and replicated on an independent dataset.