CAMELS Environments: The Impact of Local Neighbours on Galaxy Evolution across the SIMBA, IllustrisTNG, ASTRID, and Swift-EAGLE Simulations

TL;DR

CAMELS Environments investigates how local environment shapes galaxy and halo properties across four galaxy formation models (SIMBA, IllustrisTNG, ASTRID, Swift-EAGLE). By leveraging the CAMELS CV sets and environment metrics $ abla$10 and $D_{1,1}$, the study shows that satellites are consistently quenched in overdense regions while centrals exhibit mass- and redshift-dependent responses, with halo baryon content and CGM fractions following similarly nuanced environmental trends. The work reveals strong model dependence: the same environmental metric can imply opposite effects on $f_{ m B}$ and $f_{ m CGM}$ for different subgrid prescriptions, and redshift evolution can flip trends observed at $z=0$. These results highlight the need to interpret environmental influences within the context of the underlying feedback physics and motivate future CAMELS runs with larger volumes to capture richer environments and tighten observational comparisons.

Abstract

Internal feedback from massive stars and active galactic nuclei (AGN) play a key role in galaxy evolution, but external environmental effects can also strongly influence galaxies. We investigate the impact of environment on galaxy evolution, and its dependence on baryonic physics implementation, using Cosmology and Astrophysics with MachinE Learning Simulations (CAMELS) spanning a wide range of stellar and AGN feedback implementations in the SIMBA, IllustrisTNG, ASTRID, and Swift-EAGLE galaxy formation models. We show that satellite galaxies are significantly affected by the environment in all simulation models, with their gas fraction and star formation rate (SFR) suppressed in overdense regions compared to similar mass satellites in underdense environments at $z=0$. Central galaxies are less sensitive to environment but tend to show lower gas fraction and SFR in overdense regions at low stellar mass, transitioning to higher gas fraction and SFR for massive galaxies in higher-density environments. Halo baryon fraction ($f_{\rm B}$) and circumgalactic medium mass fraction ($f_{\rm CGM}$) at $z=0$ show clear environmental effects. In SIMBA, low-mass haloes in overdense regions have systematically lower $f_{\rm B}$ and $f_{\rm CGM}$ at fixed halo mass, while Swift-EAGLE haloes in overdense regions have systematically higher $f_{\rm B}$ and $f_{\rm CGM}$ across the full halo mass range, and IllustrisTNG and ASTRID show opposite trends at the low and high mass ends. Environmental effects can flip at higher redshift, with SFR and $f_{\rm B}$ increasing with local density in low-mass haloes before quenching at an increasing overdensity threshold. Our results demonstrate that the impact of environment on galaxy evolution depends significantly on galaxy formation model, and higher-density environments can either suppress or enhance star formation depending on galaxy mass and cosmic epoch.

Figures (13)

• Figure 1: Local galaxy environmental density at $z=0$ as a function of halo mass for the $\delta_{10}$ (left) and $D_{1,1}$ (right) environment definitions (see Section 2.2 for details) in the SIMBA CV simulations. The black points represent the median overdensity in each mass bin, and the blue region represents the 25$^{\rm th}$ to 75$^{\rm th}$ percentile range quantifying cosmic variance across the CV set. There is a clear positive correlation between halo mass and $\delta_{10}$, with the most massive haloes usually forming in the densest regions. The reduced y-axis range in the right panel indicates that the definition of $D_{1,1}$ is far less sensitive to $M_{\rm halo}$ than $\delta_{10}$.
• Figure 2: Distibution of galaxies for the CV_0 simulation from the SIMBA (top left), IllustrisTNG (top right), ASTRID (bottom left), and Swift-EAGLE (bottom right) simulation suites in CAMELS. The background gray scale represents the projected dark-matter mass distribution in each $(25\,h^{-1}{\rm Mpc})^3$ volume. Central galaxies are represented as circles with size equal to the virial radius of the corresponding host dark matter halo while satellite galaxies are represented by the plus signs, with their colors indicating each galaxy specific SFR. The most massive galaxies tend to be quenched and reside in the densest regions while the smaller and more actively star forming galaxies are more abundant in lower density regions. Satellite galaxies of massive haloes often display low sSFR. Despite these general trends, there are significant differences between models. SIMBA and IllustrisTNG similarly quench the most massive galaxies but SIMBA predicts a higher abundance of low-mass, star-forming galaxies. ASTRID and Swift-EAGLE are less efficient at suppressing star formation in massive galaxies while more efficiently quenching low-mass galaxies in under-dense environments.
• Figure 3: Visual illustration of central galaxy overdensities (top row) and $D_{1,1}$ (bottom row) for three representative simulations from the SIMBA suite (CV_0, CV_1, and CV_2 from left to right). The background gray scale represents the projected dark matter distribution and the color of each halo indicates the $\delta_{10}$ or $D_{1,1}$ environment value of its central galaxy. The most massive haloes reside in the most overdense regions and have higher $\delta_{10}$ values/lower $D_{1,1}$ values.
• Figure 4: Baryon fraction, $f_{\rm B}$, as a function of halo mass at $z=0$ in SIMBA, IllustrisTNG, ASTRID, and Swift-EAGLE (from left to right) for all simulations in each CAMELS CV set. We show the median $f_{\rm B}$ in each halo mass bin (black), the median $f_{\rm B}$ of underdense haloes with $\delta_{10}$ below the $25^{\rm th}$ percentile in each bin (blue), and the median $f_{\rm B}$ of overdense haloes with $\delta_{10}$ above the $75^{\rm th}$ percentile in each bin (red). The blue and red shaded regions indicate the $25^{\rm th}$ to $75^{\rm th}$ percentile ranges of underdense and overdense haloes, respectively. Error bars depict 90% confidence intervals of bootstrap medians and the bottom panels show Mann-Whitney-U test p-values between underdense and overdense haloes as a function of $M_{\rm halo}$. In SIMBA, low-mass haloes ($M_{\rm halo} \lesssim 10^{12} {\rm M}_{\rm \odot}$) have higher $f_{\rm B}$ in lower density environments but this trend reverses for intermediate mass haloes ($10^{12} {\rm M}_{\rm \odot} \lesssim M_{\rm halo} \lesssim 10^{13} {\rm M}_{\rm \odot}$) and disappears for higher mass haloes. In IllustrisTNG and ASTRID, low-mass haloes are roughly independent of environment, showing instead a weak trend for higher $f_{\rm B}$ in overdense regions for massive haloes. In contrast, Swift-EAGLE haloes in overdense regions have systematically higher $f_{\rm B}$ across the full halo mass range with the trend weakening for haloes with $10^{13} {\rm M}_{\rm \odot} \lesssim M_{\rm halo} \lesssim 10^{13.5} {\rm M}_{\rm \odot}$.
• Figure 5: Circumgalactic medium fraction, $f_{\rm CGM}$, as a function of halo mass at $z=0$ in SIMBA, IllustrisTNG, ASTRID, and Swift-EAGLE (from left to right) for all simulations in each CAMELS CV set. Black, blue, and red lines show median trends for all, underdense, and overdense haloes respectively, with 90% bootstrap error bars and Mann-Whitney-U p-values shown as in Figure \ref{['fig:f_b']}. In SIMBA, haloes with $M_{\rm halo} < 10^{11.5}\,{\rm M}_\odot$ have higher $f_{\rm CGM}$ in underdense environments compared to overdense environments. This trend reverses in the $10^{12}\,{\rm M}_\odot$--$10^{13}\,{\rm M}_\odot$ halo mass regime, and there is no trend with environment for higher masses. In IllustrisTNG and ASTRID, haloes with $M_{\rm halo} > 10^{12.5}\,{\rm M}_\odot$ have higher $f_{\rm CGM}$ in high density environments but there is only a weak or absent environmental impact at lower $M_{\rm halo}$. In Swift-EAGLE, haloes in overdense regions have systematically higher $f_{\rm CGM}$ compared to lower density regions until they reach a halo mass of $10^{13}\,{\rm M}_\odot$ with the trend reappearing at $M_{\rm halo} \sim 10^{14}\,{\rm M}_\odot$.