Correction: SPARC galaxy results (UFFT / LSFM / FTF)

November 2026

Earlier papers and Zenodo records of mine said that a log-periodic "λ ≈ e" correction explains galaxy rotation curves without dark matter. They gave figures such as 163 of 165 galaxies (98.8%), 150 of 165 (90.9%), about 80% RMSE improvement, ΔAIC ≈ −238 and λ = 2.70–2.72. I am withdrawing these claims.

Why

  • The original analysis code is lost, so the numbers cannot be checked.
  • The per-galaxy table they came from measured its fit against an undisclosed baseline. For one galaxy, UGC 02953, the table gives ΔAIC −4457, while the paper reports −51.8 for the same galaxy.
  • The headline success counts were never actually calculated from that table.
  • Two new tests, with fixed methods set in advance, find no effect:
    • Locked test (113 galaxies): the ripple amplitude is about 1.8%, and e ranks 32nd of 62 trial values. Re-fitting the paper's own equation, only 19.6% of galaxies meet its success rule, and none comes out near e.
    • Held-out test (130 galaxies): fitting the ripple makes predictions worse (ΔBIC +9.8). All five signature tests fail. Ripples of about 2% or larger are ruled out.

Corrected statement

In SPARC rotation curves there is no evidence for a log-periodic correction with λ ≈ e. Any such effect is below about 2%. My earlier work does not show that dark matter is unnecessary.

I am issuing corrected versions of the affected papers and Zenodo records. The full null-result write-up is available on request.