A Start-Condition Ripple in Exact Mukhanov–Sasaki Running, and a Cheap Fix
Exact Mukhanov–Sasaki (MS) integration is the standard reference for inflationary spectra, but its running α_s can be wrong at the 10^-4–10^-3 level. We found that a five-point second difference of ln P(k) gave α_s=+2.2×10^-4, -3.6×10^-3 and -1.3×10^-4 for steps Δln k=0.1, 0.05 and 0.025 on a smooth Starobinsky-type plateau. Tightening the integrator tolerance from 10^-10 to 10^-13 does not change a 21-point fit at all. Moving the Bunch–Davies start from k/aH=200 to 1000 shifts α_s by 3×10^-5. The scatter is therefore a deterministic ripple in P(k) left by starting the modes at finite k/aH, of relative size ~(aH/k)^2, which narrow finite differences amplify. Starting at k/aH=4000 and fitting nine points over ±0.1 in ln k removes it. α_s is then stable to 7×10^-7 under changes of the fit window and of the start, and a third-order slow-roll code agrees with it to ≤4×10^-6. With the potential compiled, one nine-mode spectrum takes 0.9 s instead of ≈570 s. Every test was pre-registered, and the first, failed attempt is reported. Status. Diagnosis and fix: computed under pre-registered locks (U1, its deviations file, U1b, U4); the first compiled attempt failed its α_s test and is reported. The ripple amplitude (aH/k)² is an order-of-magnitude estimate; the evidence for the mechanism is the response to the start and the absence of a response to the integrator tolerance.
Authors
- Danny Eldridge
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-17
- DOI
- https://doi.org/10.5281/zenodo.24330317
- Citations
- 2
- Primary Topic
- Cosmology and Gravitation Theories
- Type
- preprint