ARCHITECTS II: Impact of subgrid physics on the observable properties of the circumgalactic medium

TL;DR

Architect simulations isolate the impact of three subgrid feedback prescriptions on the CGM of a single galaxy at $z\approx1$, using RAMSES-RT, post-processed ion fractions with KROME, and synthetic quasar sightlines with $10^5$ rays to compare HI, MgII, CIV, and OVI against observations. The four ions trace distinct CGM phases and their spatial distributions vary with the subgrid physics: DC (delayed cooling) yields more metals and hotter gas, boosting CIV and OVI, while ME and MT produce different cold and warm gas structures with weaker metal enrichment. Across ions, column densities broadly bracket observed ranges but HI and MgII remain underpredicted in all models, whereas CIV/OVI are better matched by DC but may overpredict covering fractions; large intrinsic scatter and ionisation-state uncertainties limit agreement. The study demonstrates that covering fractions provide stronger constraints than column densities and highlights the need for additional physics (e.g., AGN, cosmic rays), higher resolution, and non-equilibrium thermochemistry to reconcile CGM observables with simulations.

Abstract

Galaxy evolution is driven by star formation and stellar feedback on scales unresolved by current high-resolution cosmological simulations, requiring robust subgrid models. However, these models remain degenerate, often calibrated primarily to match observed stellar masses. To explore these degeneracies, we conduct three state-of-the-art cosmological zoom-in simulations of the same galaxy, each incorporating different subgrid models: mechanical feedback, a combination of mechanical and thermal feedback, and delayed cooling. We compare their circumgalactic media (CGM) through quasar absorption sightlines of HI, MgII, CIV, and OVI. Our findings demonstrate that despite producing galaxies with the same stellar masses, the models lead to distinct feedback modes and CGM properties. Column densities and covering fractions serve as effective diagnostics of subgrid models, with all four ions providing strong constraints as they trace diverse gas phases, exhibit complementary spatial distributions, and originate from different mechanisms. Although all simulations bracket observed column density distributions, direct comparisons are limited by scarce detections and significant scatter in absorption strengths. Covering fractions of weak absorbers provides the most robust constraints. All models fail to reproduce HI and MgII covering fractions, and delayed cooling overproduces OVI covering fractions, while the other models underproduce them. The simulation including mechanical feedback reproduces the observed CIV covering fractions well, whereas the other models show slight offsets. We argue that this discrepancy is likely driven by unresolved thermal structures for HI and MgII, and insufficient metals for CIV and OVI, arising from missing physics such as AGNs or cosmic rays.

Figures (12)

• Figure 1: Phase diagrams of the gas contained in the CGM with ME. The result consists of 100 snapshots stacked over $1\rm\,Gyr$, from $z=1.3$ to $z=1$. From top to bottom, the phase diagram is weighted by the normalised mass of Hi, Mgii, Civ and Ovi. The solid line contour encompasses 90% of the total gas mass within the CGM in ME. On the left side and at the bottom of each panel, we show the stacked, normalised mass-weighted PDF of temperature and density in the CGM for the three simulations. The vertical blue dashed line in the top panel delimits the densities over which self-shielding is applied on-the-fly with an exponential damping factor. Hydrogen on the left part of this diagram can be photoionised by the UVB.
• Figure 2: Phase diagrams of the gas contained within $R_{200}$ in ME (top), MT (top) and DC (bottom), weighted by the normalised mass of Mgii in the halo. The result consists of 100 snapshots stacked over $1\rm\,Gyr$, from $z=1.3$ to $z=1$. The solid line contour encompasses 90% of the full gas mass within $R_{200}$ in each simulation. On the left side and at the bottom of each panel, we show the stacked, normalised, mass-weighted PDF of temperature and density of the corresponding simulation in the whole halo (solid line) and solely in the CGM (dashed line).
• Figure 3: Maps of the galaxy at $z\approx1.1$ for ME, MT, and DC (from left to right). From top to bottom, we show the column density of Hi, Mgii, Civ and Ovi, computed using KROME. The images are $2.5\,R_{200}$ on a side.
• Figure 4: Cartoon showing how column densities are obtained in our simulations. The central galaxy is shown in blue, with its CGM represented by the diffuse red halo. We produce $10^5$ rays sampling uniformly impact parameters up to $2\,R_{200}$. Their direction is given by ${\textbf{k}_\mathrm{obs}}$, and they intersect the galaxy plane at a point $(x_{\perp i}, y_{\perp i})$, with $i$ the index of the $i$-th ray.
• Figure 5: Column density as a function of impact parameter for Hi (top) and Mgii (bottom) stacked over $1\rm\,Gyr$, from $z=1.3$ to $z=1$. The solid lines correspond to the median column density, and the shaded areas show the 15.9 and 84.1 percentiles. We also show the maximum (resp. mean) column densities measured as dashed (resp. dotted) lines. Different markers show galaxy-selected observational points from different references. Upward-pointing arrows denote lower limits and downward-pointing arrows denote upper limits. The horizontal grey line shows the threshold used for the covering fractions in Fig. \ref{['fig:frac_cold']}, and the vertical black line corresponds to the mean $R_{200}$ over our redshift range, $R_{200\rm,avg}\approx98\rm\,kpc$.