UV, X-rays, and radio waves constrain plausible CGM models of an often overlooked type of halo
It is now widely recognized that future studies of the circumgalactic medium (CGM) will rely heavily on multi-wavelength observations. Last year’s Fundamentals of Gaseous Halos program at the Kavli Institute for Theoretical Physics (KITP) ended on an optimistic note with a week dedicated to the question “What future observations will transform our understanding?’’ That question forced participants to think panchromatically about the CGM predictions that cosmological numerical simulations make.
While cosmological simulations of galaxy formation have well-known shortcomings, including limited resolution, rigid sub-grid models, and omission of physical processes that are difficult to code, they do enable something very helpful: These simulations can be used to generate mock observations of galactic atmospheres at many different wavelengths. By comparing those mock observations with each other and also with real data we can then investigate how diverse and complementary multi-wavelength observations can be combined to more deeply understand the astrophysics of galactic atmospheres.
To illustrate the utility of mock CGM observations, we will focus here on a particular type of galactic halo atmosphere selected for three reasons:
1) These halos can currently be observed via multiple wavelengths.
2) They are widely available in many cosmological numerical simulations with enough resolution elements to provide robust predictions.
3) They are highly sensitive to the astrophysical phenomenon that dramatically disrupts our understanding of the CGM, how galaxies assemble, and even cosmology— superwind feedback.
Therefore, we choose the gaseous atmosphere of a ∼1013M⊙ dark matter halo in the evolved Universe.
Figure 1 shows four simulations from the Cosmology and Astrophysics in MachinE Learning Simulations (CAMELS)  project. One of the main aims of CAMELS is to assess the impact of baryonic physics, such as superwind feedback, on statistical inferences of cosmological parameters. The project attempts to overcome some of the shortcomings of numerical simulations by performing thousands of simulations that explore parameter spaces related to both astrophysical feedback and cosmology. The project also employs multiple codes, to investigate algorithm-dependent differences. However, CAMEL simulations do not resolve the microphysics of gaseous galactic halos.
The set of CAMELS simulations in Figure 1 demonstrates how complementary observations in three different wavebands can be combined to distinguish different feedback scenarios. Panels on the left show the UV absorption signal from neutral hydrogen atoms (H I), which is the best tracer of 104 K gas, as evidenced by the extent of the filamentary and sheet-like structures. The central panels show X-ray emission from hot gas, which probes the hot (T>106 K) CGM filling the interiors of massive galactic halos. Panels on the right show electron dispersion measures accessible via observations of distant Fast Radio Bursts (FRBs), which trace ionized baryons beyond the extent of the X-ray emission. Obvious differences from top to bottom among the rows of Figure 1 reflect the how the strength of superwind feedback affects observable CGM properties. A ∼1013M⊙ halo can be identified as the second most obvious overdensity near the center of each panel.
To understand why differences arise in the observable properties of those simulated halos, some more details about the CAMELS project is helpful. Currently, CAMEL simulations come in two main varieties: the AREPO code  with the IllustrisTNG feedback algorithm , and the GIZMO code  with the SIMBA feedback algorithm . Sets of runs labeled “1P” vary one parameter at a time. Figure 1 shows the fiducial IllustrisTNG run (1P_3_0) in the second row and a variant with less stellar feedback energy per unit star formation (1P_3_n5)1 in the top row. The lower two rows show examples of SIMBA runs with low-speed jet mode AGN feedback (1P_6_n5) and high-speed jet mode AGN feedback (1P_6_5). Moser et al. (2022) generated mock observations of the stacked Sunyaev-Zel’dovich (S-Z) measurements of Amodeo et al. (2021) and Schaan et al. (2021) finding these two parameter changes across the eight feedback parameter sets explored exhibited the most sensitivity.
In the overall CAMELS simulation set, there are nine SIMBA runs with incremental changes in AGN jet-model velocity. Figure 1 shows the run with the slowest jets (1P_6_n5) and the run with the fastest jets (1P_6_5). Similarly, there are nine IllustrisTNG runs with incremental changes in stellar feedback strength. However, we do not show the run with the strongest stellar feedback (1P_3_5), opting instead to show the fiducial run (1P_3_0) and the run with the weakest stellar feedback (1P_3_n5).
In our choice not to show the IllustrisTNG run with the strongest stellar feedback (1P_3_5) lies a cautionary tale: Not all CAMELS 1P runs are equally good at reproducing observed galaxies. While the 1P_3_5 run produces the most baryon-rich CGM with strong HI absorption, bright X-rays, and high dispersion measures that are easily distinguishable from other models, it makes galaxies with too few stars. The strong stellar feedback allows only ~3×1010M⊙ of stars to form in the central galaxies of ~1013M⊙ halos, which typically host Luminous Red Galaxies (LRGs) at least 3 times higher in mass.
Figure 2 therefore compares radial profiles of CGM properties among simulations that produce central galaxies that have about the right stellar mass, ranging from ~7×1010M⊙ to ~1.6×1011M⊙ in ~1013M⊙ halos. After all, the IllustrisTNG and SIMBA creators spent immense efforts to calibrate their fiducial models on the key stellar properties of galaxies, the masses of their central black holes, and the gaseous halo properties of massive groups.
Therefore we ask, do the predicted observational profiles in Figure 2 represent a meaningful range of feasible CGM possibilities given that the stellar masses of the central galaxies are within an acceptable range and their star formation rates are mainly negligible, as expected for LRGs? This is a crucial consideration given that future observational analyses will often rely on mock predictions of state-of-the-art simulations, including potentially CAMELS. As mentioned above, it would be deceptive to present the entire range of CAMELS 1P models as plausible models. A clear answer to this question requires more study, including comparisons to observed black hole masses and even satellite galaxy properties. But if Figure 2 does indeed represent a plausible set of models, then it contains intriguingly discernible observational predictions for the CGM around LRGs that can be related to previous CAMELS results.
For example, Moser et al. (2022) found that increasing the IllustrisTNG stellar feedback input energy raised the density and pressure of the CGM as measured via the kinetic and thermal S-Z effects. Likewise, the signal of every observable CGM feature in Figure 2 increases from run 1P_3_n5 to run 1P_3_0. Villaescusa-Navarro (2021) showed using this suite of models that stronger stellar feedback leads to significantly reduced star formation as more gas is ejected into the CGM. The resulting impact on the power spectrum of the cosmic matter correlation function (which is an essential measure for cosmological parameter estimation) is to increase power on scales out to which feedback can substantially redistribute the baryons.
However, the SIMBA jet speed variations have the opposite effect: Increased jet speed reduces the density and pressure of the CGM . Not only does every observable CGM property decrease significantly in Figure 2, but the electron density maps in Figure 1 take on a smoothed appearance. Indeed, AGN jet-mode feedback in the fiducial SIMBA simulation (run 1P_6_0) blows baryons out to distances exceeding ten times the virial radius of our chosen halos , so ratcheting up jet speeds even more, as in the 1P_6_5 run, will push baryons out even farther.
Intriguingly, changing the AGN speed in the SIMBA simulation set does not affect the central galaxy’s the star formation history very much , likely because black hole feedback operates primarily after most of the galaxy’s stars have formed. While the central galaxies in this set may look similar, the matter power spectrum becomes substantially weaker on all scales, even quite large ones , with faster AGN jets. In other words, CAMELS demonstrates that attempts to derive precise cosmological constraints from the matter power spectrum depends strongly on how supermassive black hole jets couple to their gaseous halos.
The combination of observables in Figure 2 provides a roadmap for a comprehensive accounting of baryons within the next several years. Observations of neutral hydrogen account for the CGM’s neutral gas, but the column densities represent just a trace of the total baryonic mass. Also, H I is the most uncertain of the three measures we show due to simulation numerics and resolution plus other systematics, therefore one should treat these values as relative trends and not final predictions. We are more optimistic about the X-ray predictions in Figure 2 being robust, because they trace a massive baryonic reservoir, but X-ray assessments of baryonic mass rely on accurately measuring the metallicity. Based on initial results by Chadayammuri et al. (2022) in the 140 degree2 eFEDS field of eROSITA, we expect that mission to provide statistics on thousands of group-sized halos, with the brighter scenarios suggesting that X-rays from individual ∼1013M⊙ halos may be detectable. Finally, FRBs in principle provide the cleanest measurements of the baryonic mass distribution, but are statistically the most challenging to analyze, because distant FRBs must be cross-correlated with the foreground halos producing the dispersion-measure signal.
In the near term, observations of the S-Z Effect will likely provide a better alternative to FRB observations, thanks to existing dedicated surveys (e.g. ). Interestingly, Moser et al. (2022) concluded that despite the wide range of feedback strengths in the CAMELS simulations, they cannot reproduce the exceedingly high pressures and high densities observed outside the virial radius of group-sized halos (). Indeed, group-sized (1013−1014M⊙) halos have multiple properties that are not well reproduced by the current generation of flagship cosmological simulations like IllustrisTNG, SIMBA, EAGLE, and ROMULUS . Therefore, developing more observational datasets that focus on ∼1013M⊙ halos, which are comparatively data-starved, will provide new and essential stress tests for large simulation suites such as CAMELS and other simulation projects with similar goals.
Acknowledgements: This research was supported in part by the National Science Foundation under Grant No. NSF PHY-1748958. The authors thank Mark Voit, the Founding Editor for detailed guidance in the production of this Galactic Atmospheres Perspectives article.
Header Background Image: Free electron column density maps of CAMEL simulations with incremental changes in feedback strength. The left six panels show IllustrisTNG 1P_3_n5 though 1P_3_0, and the right six panels show SIMBA 1P_6_n5 through 1P_6_0.