Cosmic Rays Masquerading as Hot CGM Gas: An Inverse-Compton Origin for Diffuse X-ray Emission in the Circumgalactic Medium

TL;DR

This work tackles the hard-to-interpret soft X-ray halos around MW-like and lower-mass galaxies by proposing that inverse Compton scattering of CMB photons by GeV cosmic-ray electrons escaping the ISM produces the observed halos. The authors develop analytic scalings and detailed emissivity calculations, showing the resulting KeV X-ray spectra and $S_X \propto R^{-1}$ profiles match observations for $M_* \lesssim 2\times10^{11} M_\odot$, with a halo extent set by CR transport and IC losses. They further connect the inferred halo CR energy densities to CR injection from SNe and AGN, and show the results are consistent with MW/M31 gamma-ray data and UV absorption constraints, while radio halos would be faint. The work implies CR pressure is a major component of the MW CGM and demonstrates that X-ray brightness directly traces CR lepton energy density in the CGM, offering a new diagnostic of CR transport in galaxy halos.

Abstract

Observations have argued that Milky Way (MW), Andromeda, and lower-mass galaxies exhibit extended soft X-ray diffuse halos to radii $R\gtrsim100\,$kpc in the circumgalactic medium (CGM). If interpreted as thermal emission, the shallow surface brightness profiles $S_{X}\propto R^{-1}$ are difficult to explain and contradict other observations. We show that such halos instead arise from inverse Compton (IC) scattering of CMB photons with GeV cosmic ray (CR) electrons. GeV electrons have ~Gyr lifetimes and escape the galaxy, forming a shallow extended profile out to $\gtrsim100\,$kpc, where IC off the CMB should produce soft, thermal-like X-ray spectra peaked at ~keV. The observed keV halo luminosities and brightness profiles agree well with those expected for CRs observed in the local interstellar medium (LISM) escaping the galaxy, with energetics consistent with known CRs from SNe and/or AGN, around galaxies with stellar masses $M_{\ast}\lesssim2\times 10^{11}\,M_{\odot}$. At higher masses observed X-ray luminosities are larger than predicted from IC and should be dominated by hot gas. In the MW+M31, the same models of escaping CRs reproduce gamma-ray observations if we assume an LISM-like proton-to-electron ratio and CR-pressure-dominated halo. In all other halos, the radio and $γ$-ray brightness is below detectable limits. If true, the observations provide qualitatively new constraints on CGM and CR physics: X-ray brightness directly traces the CR lepton energy density in the CGM. This agrees with LISM values within 10 kpc, which following the profile expected for escaping CRs in the CGM. The inferred CR pressure is a major part of the MW CGM pressure budget. X-ray surface brightness and luminosity allows one to further determine the CGM diffusivity at radii $\sim10-1000\,$kpc. These also agree with LISM values at small radii but increase in the CGM.

Figures (8)

• Figure 1: Observed-frame X-ray spectra produced by IC from CRs off the CMB (§ \ref{['sec:basic']} & \ref{['sec:detailed']}). Top:$E^{2} dN/dE$ over a very broad energy range. Bottom:$dN/dE$ focused on standard X-ray energies $0.2-8$ keV. We show the spectrum from IC only, from the CMB (other cosmic backgrounds are totally negligible at these energies), at $z=0$, from standard expressions. For the CR lepton spectrum, we compare (1) the observed LISM spectrum, and (2) that spectrum corrected for IC losses with finite travel time (i.e. finite distance from the ISM) parameterized by $\tau_{\rm loss}^{0} \equiv t_{\rm travel}/t_{\rm loss} \sim (R/100\,{\rm kpc})\,(100\,{\rm km\,s^{-1}}/v_{\rm st,\,eff})$ at $1\,$GeV. For reference, we compare APEC thermal emission spectra (free-free+line; metallicities $\sim 0.05\,Z_{\odot}$ matching the X-ray absorption upper limits at MW virial radii; § \ref{['sec:thermal']}). CRs leaking from the ISM produce a thermal-like spectrum with most of the luminosity at $\sim$ keV (Eq. \ref{['eqn:peak']}; from $\sim$ GeV leptons, near the LISM spectral peak, with $\sim$ Gyr lifetimes), until most of the CR energy is lost ($\tau_{\rm loss}^{0} \gg 1$, expected at $R \gtrsim 0.1-1$ Mpc distances from the galaxy; § \ref{['sec:losses']}).
• Figure 2: Observed soft X-ray diffuse CGM emission, compared to expected IC from CRs. Shaded range shows eROSITA stacked mean CGM emission at $0.5-2\,$keV, around Milky Way (MW; $10^{10.5-11}\,M_{\odot}$), Andromeda (M31; $10^{11-11.25}\,M_{\odot}$), and twice M31 (2M31; $10^{11.25-11.5}\,M_{\odot}$) stellar-mass central, isolated galaxies (after subtracting point sources/XRBs/AGN and backgrounds; zhang:2024.hot.cgm.around.lstar.galaxies.xray.surface.brightness.profiles). Lines compare the predicted CR-IC emission, using the simple analytic model in § \ref{['sec:basic']} (with IC losses accounted for per § \ref{['sec:losses']}), for values of the injection rate parameter $\dot{E}_{\rm cr,\,\ell} = \dot{E}_{40}\,10^{40}\,{\rm erg\,s^{-1}}$ and CR streaming speed $v_{\rm st,\,eff}=v_{100}\,100\,{\rm km\,s^{-1}}$ ( labeled), expected for different mass galaxies (matching colors). We also compare the prediction from a fiducial cosmological CR-MHD simulation ( cyan line) of a MW-mass galaxy (which used a simple CR diffusion+streaming model calibrated to LISM CR data) from hopkins:cr.mhd.fire2hopkins:2020.cr.transport.model.fx.galform.
• Figure 3: Leptonic CR energy density around $\sim$ GeV required to explain the observed soft X-ray surface brightness, for each of the three observed stacks of $S_{X}$ in Fig. \ref{['fig:profiles']} ( shaded; deprojecting per § \ref{['sec:redshift']}, but approximately given by Eq. \ref{['eqn:ecr']}). The central (galactic) values are similar to the observed LISM values bisschoff:2019.lism.cr.spectra in the MW-mass systems, and those inferred from multi-wavelength modeling of M31's synchrotron plus $\gamma$-ray emission lacki:2010.fir.radio.conspiracy, with a power-law falloff and eventual cutoff from IC losses at larger radii.
• Figure 4: Total CGM halo soft X-ray luminosity, $L_{\rm X,\,kev}^{\rm tot} = \int S_{X}\,dA$, as a function of central galaxy stellar mass $M_{\ast}$, compared to the expected IC luminosities. Observed values from are shown from stacked ROSAT data (anderson:2015.rosat.xray.halo.stacking.scaling.relations CGM only-emission after subtracting background+galaxy+point-source/XRB contributions), as well as eROSITA, using the same brightness profiles from zhang:2024.hot.cgm.around.lstar.galaxies.xray.surface.brightness.profiles in Fig. \ref{['fig:profiles']} and integrating out to radii where any excess emission is detected (i.e. allowing for an IC cutoff; filled; "detected"), or using the stacking method in zhang:2024.erosita.hot.cgm.around.lstar.galaxies.detected.and.scaling.relations which extrapolates the inferred profiles from smaller radii out to $R_{500}$ in all cases ( open; "$R\rightarrow R_{500}$"). We compare the predicted time-and-population-averaged SNe $\sim$ GeV leptonic CR injection rate and AGN injection rate (assuming empirical SFR-$M_{\ast}$ and $\langle M_{\rm BH} \rangle - M_{\ast}$ relations, and that $\sim 10^{-4}\,\dot{M}_{\rm BH}\,c^{2}$ goes into leptons, for reference; see § \ref{['sec:obs']}) from just the central galaxies. These scalings have factor $\sim 2-3$ systematic uncertainties. IC from standard sources (SNe and/or AGN even with very small efficiencies) can naturally explain the halo luminosities at $M_{\ast} \lesssim (1-2)\times 10^{11}\,M_{\odot}$.
• Figure 5: Implied effective CR diffusivity $\kappa_{\rm eff}$ to explain each observed $S_{X}$ profile ( shaded), given the known IC loss timescale and observed injection rate $\dot{E}_{\rm cr} \sim L_{X,\,{\rm keV}}^{\rm tot}$ (§ \ref{['sec:ltot']} & \ref{['sec:obs']}, Eq. \ref{['eqn:kappa.measured']}). Approaching the ISM, the values in MW/M31 are very similar to those inferred from standard Galactic CR modeling of MW LISM observations from e.g. Voyager, AMS-02, etc. (MW LISM point, median and range from the model sets in delaTorre:2021.dragon2.methods.new.model.comparisonkorsmeier:2022.cr.fitting.update.ams02 using GALPROP and DRAGON2). At larger radii $\kappa_{\rm eff}$ rises, as generally expected in the CGM. We compare independent observational constraints on $\kappa_{\rm eff}$ for $\sim$ GeV protons from a completely independent UV absorption-line method applied to cool neutral HI gas (which should not appear in the X-rays) in butsky:2022.cr.kappa.lower.limits.cgm, colored by stellar mass ( blue matching Z24's MW-mass sample, grey for masses outside this).