Constraining cosmic ray transport models using circumgalactic medium properties and observables

TL;DR

This paper investigates how cosmic-ray transport affects the circumgalactic medium (CGM) in Milky Way–mass halos by running FIRE-2 simulations with multiple CR transport prescriptions (Constant Diffusivity, Extrinsic Turbulence, and Self-Confinement) and comparing simulated CGM properties to UV absorption and X-ray observations. Post-processing yields H I and O VI column densities and X-ray luminosities, revealing that CRs shift CGM gas toward cooler phases and modify non-thermal pressure, with the CD variant and its Extrinsic Turbulence relatives generally aligning best with observations. However, substantial model-to-model differences persist, and observational comparisons are sensitive to halo mass, redshift, and resolution, limiting strong constraints on CR transport at present. The study demonstrates that CGM observables, especially extended X-ray emission that includes inverse Compton scattering by CR electrons, offer a promising framework for discriminating CR transport models in future, higher-resolution, and multi-halo analyses.

Abstract

Cosmic rays (CRs) are a pivotal non-thermal component of galaxy formation and evolution. However, the intricacies of CR physics, particularly how they propagate in the circumgalactic medium (CGM), remain largely unconstrained. In this work, we study CGM properties in FIRE-2 (Feedback In Realistic Environments) simulations of the same Milky Way (MW)-mass halo at $z=0$ with different CR transport models that produce similar diffuse $\sim$ GeV $γ$-ray emission, as an attempt to further constrain CR transport models. We study the gas morphology and thermal properties, and generate synthetic observations of rest-frame UV ion absorption columns and X-ray emission. CRs lower galaxy masses and star formation rates (SFRs) while supporting more cool CGM gas, which boosts the HI and OVI column densities in the CGM, bringing simulations more in line with observations, but there can be large differences between CR transport models and resolution levels. X-ray emission within and close to galaxies is consistent with thermal (free-free and metal-line) emission plus X-ray binaries, while more extended ($\sim 100\,$kpc) CGM emission is potentially dominated by inverse Compton (IC) scattering, motivating future work on the spatially resolved X-ray profiles. Although comparisons with observations are sensitive to sample selection and mimicking the details of observations, and our analysis did not result in strong constraints on CR models, the differences between simulations are significant and could be used as a framework for future studies.

Figures (11)

• Figure 1: Maps of projected total hydrogen column density, $N_{\rm H}$, of our m12i simulation runs. The projections were taken over the whole simulation volumes. We include Hydro+, MHD+, and CR-CD runs. All maps are at $z=0$. We show the face-on (larger panels) and edge-on (smaller panels) projections. We compare the normal-res run ($m_{\rm gas \, cell}\approx 56000 M_\odot$, upper half) with the high-res run ($m_{\rm gas \, cell}\approx 7000 M_\odot$, lower half) of each simulation. The white dashed circles represent $0.1 R_{\rm vir}$ as listed in Table \ref{['tab:sims']}. We see that all six runs form a galactic disk at $z=0$ to some extent, but while all three high-res runs show clear disks, only CR runs have the same structure of a symmetric disk at normal resolution. Furthermore, by comparing the upper half with the lower half, we see that the resolution has a much less significant effect on the CR-CD run than the Hydro+ and MHD+ runs. Normal-res Hydro+ and MHD+ runs locked almost all available halo baryons into stars (see Table \ref{['tab:sims']}), resulting in a lack of late-time accretion to form a clear disk.
• Figure 2: Similar to Fig. \ref{['fig:proj_resolution_comparison']}, but for all CR runs. All maps are at $z=0$. The models (marked at the top of each set of projection maps) are listed in Table \ref{['tab:sims']} and briefly described in Section \ref{['sec:2.2']}. All simulations are run in normal-res setup for m12i. Solely by eye, we can tell that the two extrinsic turbulence (ET) runs resemble the CR-CD run of the same halo much more than the self-confinement (SC) runs. This is reflected not only in the gas morphology, but also in the central stellar mass (see Table \ref{['tab:sims']}). SC models are abnormal compared to the others, owing to the late-time SC "blowout/runaway" as described in ponnada2024synchrotron.
• Figure 3: Radial profiles of gas number density $n_{\rm H}$ (first row), volume-weighted gas temperature $T_{\rm gas}$ (second row), and volume-weighted pressures (third row). The pressure (third) row includes both the thermal pressure $P_{\rm therm}$ (solid lines) and the CR pressure $P_{\rm cr}$ (dashed lines, for the CR runs in the right panels). We classify all of our runs into two categories as we did in Table \ref{['tab:sims']}: the left panels are for all non-CR runs and the right panels are for all the single-bin CR runs. We label different runs are labelled with different colours and line widths to guide the eye. All runs have a target halo of m12i. The temperature profiles (middle row) were calculated using volume weighting, the same as was done in hopkins2021effects, to emphasize gas cells with large volume (and corresponding high temperature). The most significant differences between different runs are seen in temperature profiles. This is consistent with what was argued in hopkins2020but and ji2020properties that the CRs change the temperature of the CGM gas and therefore alter its phase structure, which is critical for many of the CGM observables.
• Figure 4: Mass-weighted density-temperature ($n_{\rm H}-T$) histograms of CGM gas (crudely defined as regions of $50 \,{\rm kpc} \leq r \leq 300 \,{\rm kpc}$) for six selected representative runs from our simulations (see the main text for details of selection). There is a broad resemblance among the non-CR runs and the variant diffusivity CR runs (ET and SC), while the CR-CD runs exhibit a unique phase structure.
• Figure 5: Column density profiles of the neutral hydrogen ($\text{H}\scaleto{$I$}{1.2ex}$, the upper two panels) and five-times ionized oxygen ($\text{O}\scaleto{$VI$}{1.2ex}$, the lower two panels) as functions of the impact parameter $b$ from the centre of the galaxy. We normalize $b$ with the virial radii of each run described in Table \ref{['tab:sims']}. Similarly to Fig. \ref{['fig:therm_prof']}, we put all non-CR runs on the left panels and all CR runs on the right panels. Solid lines represent the median values among all three projections (see the text for details), and the shaded regions with the same colours as the lines represent the maximum and minimum ranges. The red stars correspond to the observational data from prochaska2017cos ($\text{H}\scaleto{$I$}{1.2ex}$) and werk2013cos ($\text{O}\scaleto{$VI$}{1.2ex}$). These are all observations from star-forming galaxies. In particular, we take data from Fig. 4 of prochaska2017cos and $\text{O}\scaleto{$VI$}{1.2ex}$ data for star-forming galaxies from werk2013cos. We post-process simulations at $z\sim 0.25$ to be consistent with the observational data we are comparing, though no significant difference is seen in the results of our $z=0$ simulations (not shown in this paper).