The reported redshift evolution of the MOND acceleration scale is degenerate with unmodelled molecular gas

Context. Ciocan et al. [2026a] report the first measurement of a redshift evolution of the characteristic acceleration scale a0of the radial acceleration relation (RAR), from 79 star-forming galaxies at 0.33 < z < 1.44 in the MUSE Hubble Ultra DeepField. They obtain a0|z∼1 = 2.38+0.12−0.10 × 10−10 m s−2 and, for a linear parametrisation, a1 = 1.59+0.11−0.10 × 10−10 m s−2 (95% CI). Aims. Their baryonic acceleration is built from a stellar disk and an atomic gas component only; molecular gas is notmodelled. Ciocan et al. flag the resulting mass systematic but test only redshift-independent offsets. Because the moleculargas fraction rises steeply with redshift, the omitted mass carries a gradient, and it is the gradient that produces a spurious trend. Methods. Using the public Ciocan et al. catalogue and the Tacconi et al. [2018] scaling relation evaluated per galaxy fromthe catalogue redshift, stellar mass and star-formation rate, I compute the missing molecular mass and propagate it throughthe RAR. I then remove the dependence on any particular scaling relation by reducing the correction to a single parameter,the logarithmic slope γ ≡ d log µmol/d log(1 + z), and scan it. Results. The missing mass grows from a median MH2 /M⋆ = 0.51 in the lowest-redshift quartile to 0.97 in the highest.The correction removes 82% of the binned trend in absolute a1 and 77% of the fractional evolution a0(z4)/a0(z1), whichfalls from 1.362 to 1.082; the implied power-law index drops from (1 + z) 0.80 to (1 + z) 0.20. A 4000-draw Monte Carlo over the scaling normalisation, its slope, and the assumed atomic ratio gives a median residual of 21% of the published trend, consistent with zero in 97% of realisations and never above 80%. Only γ < 1.06 would leave half the evolution intact; the adopted relation gives γ = 2.84. The correction also brings the implied baryonic Tully–Fisher zero-point shift from −0.10 to −0.02 dex, in line with the no-evolution result of Jeanneau et al. [2026]. Conclusions. a1 is not a robust physical quantity in this dataset. A high-redshift RAR measurement at this precision is limitedby the baryonic mass budget, not by the kinematics.