Ripples in the baryon to dark matter ratio in $\hat{I}›$CDM: implications for galaxy formation

Abstract

We use the FLAMINGO galaxy formation model to quantify the impact of baryon-CDM isocurvature perturbations on galaxy formation in $\hat{I}›$CDM. In linear theory, these perturbations represent local, compensated variations in the ratio between the baryon and CDM densities; they freeze in amplitude at late times, with an RMS amplitude of $1.5\%$ on the Lagrangian scale of a $10^{11}{\mathrm{M}}_{\odot }$ halo ($0.85\mathrm{Mpc}$). Although such perturbations arise naturally within $\hat{I}›$CDM, most cosmological simulations and semi-analytic models to date omit them. These perturbations are strongly anti-correlated with the matter overdensity field such that halos form with baryon fractions below the cosmic mean, with earlier-collapsing halos exhibiting stronger baryonic suppression. To isolate the galaxy response, we analyse three hydrodynamical simulations with identical initial matter overdensity fields that: i) include isocurvature modes, ii) omit them, or iii) invert their amplitude. At $z=8$, isocurvature perturbations reduce the mean baryon fraction and star formation rates of resolved halos by $5\%$ and $12\%$, respectively, relative to the null-isocurvature case. These effects are almost independent of halo mass and diminish steadily with time, reaching $0.1\%$ and $1\%$ by $z=0$. We develop a model based on spherical collapse that accurately reproduces the mean baryon fraction suppression. As high-redshift observations become increasingly routine, incorporating isocurvature perturbations into simulations and semi-analytic models will be important for robust predictions of early galaxy and black hole formation in the JWST era.

12 pages, 3 figures. Monthly Notices of the Royal Astronomical Society, Volume 547, Issue 4, April 2026, stag519, https://doi.org/10.1093/mnras/stag519 . Correction: An error following Eq. (9) stated the baryon fraction as 0.016 rather than 0.16. The correct value, 0.16, was used in Eq. (9), so this was a typographical error only and does not affect the results