WorksPhysics and Astronomy
Cosmic Ray Reacceleration and Gamma-Ray Emission via the Cloud-Wind Interaction in the Galactic Halo
Abstract
The interaction between cosmic rays (CRs) and multiphase gas in the Galactic halo may play an important role in shaping the hard gamma-ray spectrum of the Fermi bubbles. We investigate the interaction between a hot Galactic wind and a cold, dense gas clump with a relative velocity of using two-dimensional hydrodynamic simulations coupled with momentum-dependent CR transport. The system produces a bow shock and wind reacceleration region downstream, where the wind flow is faster than the initial velocity. For a CR pressure comparable to the gas pressure (CR-rich), the wind acceleration and turbulence behind the clump are enhanced. When CR pressure is dynamically weak, strong gas compression leads to significant CR spectral hardening. In contrast, for the CR-rich case, CR pressure gradients weaken the gas compression and suppress further CR reacceleration. We estimate the gamma-ray intensity from CRs, and find that reaccelerated CRs penetrate the dense clump and enhance hadronic gamma-ray emission. The gamma-ray intensity reaches in the 1–10 GeV band, and the gamma-ray spectra become significantly harder, consistent with the Fermi bubbles' emission. These results suggest that an ensemble of dense gas clumps interacting with the Galactic wind may contribute to the gamma-ray emission of the Fermi bubbles.
Introduction
Space telescopes performing direct detections of cosmic rays (CRs) such as AMS-02, CALET, and DAMPE provide precise CR spectra. However, only from those data, we cannot judge whether the CR spectra at the solar neighborhood are representative ones in our galaxy. The complex spectral breaks ([4, 3] (Q. An et al. 2019)) may suggest multiple types of acceleration sources ([31]), or they may indicate complex propagation processes spanning the galactic disk and halo (e.g., [16]). Turbulence in the interstellar medium (ISM), Galactic wind, and fresh gas accretion onto the Galactic halo should affect the CR propagation and reacceleration. Gamma-ray observations provide an important means of probing CR populations beyond the solar neighborhood.
The Fermi bubbles, a pair of giant gamma-ray structures extending approximately 10 kpc above and below the Galactic center ([2] (M. Su et al. 2010; M. Su & D. P. Finkbeiner 2012)), may indicate the dynamical interaction between CRs and the halo gas. The bubbles exhibit a relatively hard gamma-ray spectrum. While many models with explosive activities in the Galactic center, such as AGN jets (e.g., [20]) and starbursts (e.g., [29]), have been proposed, CRs produced by long-term star formation can also be the source of the Fermi bubbles (e.g., [10, 43]).
However, the origin of the hard gamma-ray spectrum of the Fermi bubbles remains an important question for hadronic scenarios. The CR proton spectrum observed in the solar neighborhood ([4, 3]) is softer than that typically inferred from hadronic interpretations of the Fermi bubble emission ([2]). Simple diffusion process of CRs cannot reconcile the observed spectra ([43]). This discrepancy suggests that CR reacceleration within the Galactic halo may play an important role in producing the hard gamma-ray spectrum of the Fermi bubbles. In this context, the multiphase nature of the Galactic halo can provide a natural environment for CR reacceleration. The Galactic halo is expected to contain both inflowing and outflowing gas as a consequence of gas accretion and Galactic winds [39, 45]. Observations of high-velocity clouds provide evidence for gas accretion onto the Galactic disk [53, 32]. Furthermore, cold gas is also observed in the Galactic nuclear wind [11, 33], indicating the coexistence of multiple gas phases in the outflow. Motivated by these findings, several simulations have been performed in the context of gas accretion [e.g. 42] and galactic fountain [e.g. 15].
In the model of J. Shimoda & K. Asano [43], the Fermi bubbles are produced via the interaction between CRs steadily supplied from the Galactic center and gases falling onto the Galactic disk. In this model, the accompanying hot galactic wind, which is heated by CRs, is consistent with the eROSITA bubbles [38], giant structures of diffuse thermal X-ray emission surrounding the Fermi bubbles. Several studies point out that CRs should play an important role in the dynamics of Galactic winds [e.g. 7, 44].
Thus, the interaction between the hot Galactic wind, accreting cold gas, and CRs above the Galactic center is an important research target in the context of both Galaxy evolution and high-energy astrophysics. Recent numerical studies have demonstrated that CRs can substantially modify the multiphase structure of Galactic winds and the evolution of cold gas clouds [6, 51, 55, 41]. Such wind–clump interactions can generate shocks and compressive flows, providing potential sites for CR reacceleration. Previous studies have demonstrated that CR feedback can modify shock structures and drive hydrodynamic instabilities [13, 12, 21]. Recent galaxy-scale simulations have also begun to follow the time evolution of CR energy spectra together with gas dynamics [17, 18]. However, how CR pressure modifies local reacceleration through its dynamical feedback during wind–clump interactions remains poorly understood. To investigate this interplay, it is necessary to follow momentum-dependent CR transport together with gas hydrodynamics.
In this study, we investigate the interaction between a hot Galactic wind and a cold, dense clump using two-dimensional axisymmetric simulations that couple momentum-dependent CR transport with gas hydrodynamics. By comparing models with different CR pressures, we examine how CR pressure affects the shock structure, gas compression, and flow dynamics around the clump. We simultaneously follow the evolution of the CR spatial and energy distributions through spatial diffusion and momentum-space transport associated with gas compression and expansion, including the dynamical feedback of CR pressure on the gas. Using the resulting CR and gas distributions, we calculate the spatial distribution and energy spectra of hadronic gamma-ray emission and investigate how CR feedback affects the emission intensity and morphology. Through these analyses, we explore the role of CR–gas interactions in a multiphase Galactic wind and their possible contribution to the hard gamma-ray spectrum observed in the Fermi bubbles.
This paper is organized as follows. The simulation method to follow the interaction of hot gas and a cold dense cloud with CRs is presented in § 2. The effects of different CR pressures on the gas dynamics, CR spectra, and gamma-ray emission are presented in § 3. § 4 is devoted to discussion, especially for the Fermi bubble. Our conclusions are summarized in § 5.
Simulation Method
In this paper, we consider a dense gas clump accreting within the galactic wind and consistently solve the interaction between the gas clump, wind, and CRs.
Basic equations
We numerically solve the hydrodynamic equations coupled to CR transport in the two-dimensional axisymmetric cylindrical coordinate . Adopting the mass density , pressure , and velocities of the fluid component representing the gas clump and the galactic wind, the conservation equations for gas mass, momentum, and energy are given by
and
respectively. Here, we have neglected wave excitation by CRs, which appears in CR–MHD systems and may heat the background fluid eventually. As can be understood from the energy equation, the CR pressure acts as an external force on the fluid, so that CRs do not directly affect the entropy of the background fluid. The total gas energy density is written as
with an adiabatic index .
The last term in the energy equation represents radiative energy change, which is written with a function
where the first and second terms are radiative cooling and heating terms, respectively. For the cooling function , we adopt the formula proposed in K. Kakiuchi et al. [24]2, where the density and temperature are defined as and . For simplicity, we adopt a constant mean molecular weight of , corresponding to the fully ionized limit. In most parts of our computational domain where the flow is dynamically turbulent, the temperature is well above K, so the difference in the fluid motion resulting from the constant is likely to be small.
For the heating term, we assume photoelectric heating by dust grains and adopt a constant heating rate of erg s, motivated by the enhanced interstellar radiation field in the Galactic center environment. Since we focus on a local region of the Galactic halo, we neglect spatial variations in the heating rate. For simplicity, we also neglect its temperature dependence.
Thermal conduction is included through the conductive energy flux. The standard heat flux is defined as
where is the thermal conductivity. We adopt the thermal conductivities for ionized gas (L. Spitzer [46]) and
for neutral gas (E. N. Parker [37]), in cgs units. To smoothly connect the low- and high-temperature regimes, we use
In a hot and rarefied plasma, the electron mean free path can become comparable to the characteristic temperature-gradient scale, invalidating the classical diffusion approximation. We therefore introduce a saturated heat flux following L. L. Cowie & C. F. McKee [9], , where is a dimensionless coefficient and is the gas sound speed. We adopt , appropriate for a fully ionized gas (J. L. Giuliani [19]; S. Orlando et al. [36]). The conductive flux in equation (4) is then written as
The time evolution of the CR momentum distribution is regulated by the motion of the background fluid. Assuming isotropic momentum distributions, we numerically solve the diffusion–convection equation
where is the spatial diffusion coefficient. For simplicity, we assume here that the diffusion coefficient is spatially uniform, regardless of the properties of the background gas. Motivated by Kolmogorov turbulence, we adopt
with cm s. This coefficient is chosen to be one order of magnitude lower than the canonical Galactic diffusion coefficient to reduce the computational cost. However, considering relatively strong magnetic fields inferred in the Fermi bubble region (G, E. Caretti et al. 2013; M. Ackermann et al. [2]), such a locally suppressed diffusion coefficient is plausible in turbulent regions (A. U. Abysekara et al. 2017).
The CR pressure is calculated as
where
is the particle velocity.
Numerical Procedure
The hydrodynamic equations are solved using the HLL approximate Riemann solver [22]. We adopt the second-order MUSCL scheme [52] with a minmod slope limiter for spatial reconstruction and the second-order Runge–Kutta scheme for time integration. For the MUSCL reconstruction, we choose the logarithms of the gas density and entropy, along with the linear velocity components, as the reconstruction variables. This choice accounts for the adiabatic relation while allowing the reconstruction to handle large density contrasts and maintain positive density and pressure at cell interfaces. The spatial reconstruction is locally reduced to first order in regions with large density contrasts and expanding flows () to maintain numerical stability.
The CR transport equation is solved using an upwind scheme for the advection terms and an explicit scheme for spatial diffusion.
Gamma-Ray Emissivity
Based on the simulation outputs for gas and CRs, we calculate the hadronic gamma-ray emission following the method of K. Nishiwaki et al. [34]. Following S. R. Kelner et al. [27], the gamma-ray emissivity from neutral-pion decay is evaluated as
where is the neutral-pion energy, and
Given , the gamma-ray photons distribute in the energy range of with . The injection rate of neutral pions is calculated from the local gas density and CR proton spectrum as
where is the target gas density, is the CR proton number density per unit momentum, and is the proton velocity. The CR proton energy is given by
For the differential cross section of neutral-pion production , we adopt the formulae of T. Kamae et al. [25, 26].
Simulation Results
Simulation Setup
We consider a local region of the Galactic halo in which a cold gas clump interacts with an upward hot Galactic wind. The computational domain covers pc and pc, with spatial cells and a uniform cell size of 2 pc. The CR distribution is simultaneously evolved over the momentum range GeV/c using 24 momentum bins.
Initially, the cold dense clump with a radius of 50 pc is embedded in a hot rarefied ambient medium. While the gas is divided into two phases, the cold clump and hot wind, the pressure is spatially uniform as eV cm, and heating and cooling are in equilibrium as . To realize the above requirement, we adopt gas number densities and temperatures of the clump and wind as and , respectively. The computation frame is the clump rest frame. The injected wind velocity is . The velocity of the wind flowing from bottom to top corresponds to the relative velocity between the accreting clump and the Galactic wind.
The initial CR energy density is also uniform, and its spectrum is assumed to follow a broken power law in momentum as
The high-momentum index is intended to represent Galactic CRs that have escaped from acceleration sites in the disk and have been softened during propagation into the Galactic halo [47].
We compare two calculations that differ only in the initial CR pressure. In the CR-rich model, the CR pressure is initially equal to the thermal gas pressure,
whereas in the CR-poor model, the initial CR pressure is 100 times smaller,
The latter therefore provides a limiting case in which CR feedback is initially dynamically negligible. At the symmetry axis, , we impose the axisymmetric boundary condition . At the outer radial boundary, pc, we impose an outflow boundary condition. At the upper boundary, pc, we adopt an escape boundary for CRs. At the lower boundary, , the wind is continuously injected with the same gas and CR properties as the initial ambient medium and with an upward velocity km s.
As we have mentioned, the heating rate and diffusion coefficient are uniform and constant with erg s and cm s, respectively.
Gas Dynamics
In this section, we examine how CR pressure affects gas dynamics in the interaction between the Galactic wind and the cold gas clump. Figure 1 and Figure 2 show the evolution of gas dynamics in the CR-poor and CR-rich cases, respectively. In both models, the incoming wind encounters the cold, dense clump and forms a bow shock. As shown by the pressure maps, the curvature of the shock fronts grows with time. In the CR-poor model, the bow shock is characterized by a relatively sharp density jump. In contrast, the CR-rich model exhibits a weaker density jump and the distance from the shock front to the clump head is longer than that in the CR-poor model.

Figure 1. Time evolution of the clump–wind interaction in the CR-poor model. From left to right, the columns show the gas density , vertical velocity , radial velocity , CR pressure , gas pressure , and total pressure . The pressures are shown normalized to their initial values. The simulation is axisymmetric, and each panel is mirrored about the symmetry axis for visualization. Black curves in the density panels show representative streamlines, while the thick white curve indicates the streamline used for the one-dimensional analysis in Figure 3.

Figure 2. Same as Figure 1, but for the CR-rich model. The thick white curve in the density panels indicates the streamline used for the one-dimensional analysis in Figure 4.
The two models also differ in their downstream structures. As shown in the velocity maps ( and ), the gas behind the clump is decelerated, while the outer gas is accelerated faster than its initial velocity. These regions grow more prominently in the CR-rich model, where the downstream flow becomes increasingly disturbed and develops more complex velocity structures. The region of negative velocity () behind the clump significantly grows with time.
The CR pressure maps show horn-like structures behind the clump, where the CR pressure is low. This structure arises because the density of CRs originally present within the clump decreases owing to gas expansion driven by the separation flow from the rear of the clump, generating rarefaction waves. This is the result of the uniform CR distribution in the initial condition. In the CR-rich case, this structure affects the gas motion, resulting in a similar shape of a high gas-pressure region. This spatial anticorrelation between the gas and CR pressures highlights the importance of solving CR transport in simulations rather than imposing a prescribed relation between the two.
In Figure 3 and Figure 4, we also plot the gas and CR properties along the representative streamlines shown in the density maps in Figure 1 and Figure 2. This streamline heads toward the clump from nearly directly in front. After passing through the shock wave, it flows toward the back of the clump, as if bypassing it. However, since this streamline is based on a snapshot, it does not necessarily represent the steady flow in general.

Figure 3. One-dimensional profiles along the representative streamline highlighted in Figure 1 for the CR-poor model at and Myr. The horizontal axes show the distance along the streamline . The left panels show the gas density and the velocity component parallel to the streamline, . The middle panels show the thermal gas pressure and CR pressure , while the right panels show the gas temperature and the net radiative heating rate , where negative values indicate net cooling.

Figure 4. Same as Figure 3, but for the CR-rich model.
In the CR-poor model, a clear shock front is seen, and it has been maintained throughout the simulation. The density further increases behind the shock because the downstream flow is influenced by the clump and radiative cooling. As shown by the radial velocity distributions in Figure 1, the gas passing through the bow shock is strongly deflected radially outward () as it flows around the clump. This two-dimensional flow geometry, radiative cooling (see temperature profile), and the unsteady flow may account for the offset between the velocity minimum and density peak. Behind the density peak position, the gas expands and accelerates. As can be seen from the pressure distribution in panels (b) and (e) of Figure 3, the CR effect is almost negligible in the CR-poor case.
The shocked gas departs from thermal equilibrium and enters a state dominated by radiative cooling. As shown in panels (c) and (f) of Figure 3, the gas evolves almost isothermally. A significant temperature jump at the shock does not appear. The enhanced cooling is accompanied by a decrease in gas temperature and further compression of the post-shock gas.
In the CR-rich model, the gas flow begins to decelerate well upstream of the density peak. As shown in Figure 4, the CR pressure dominates the region behind the shock. The density jump is weak, and the CR pressure distribution gradually evolves into a spatially broader and smoother profile. The CR effect is also evident in the smooth temperature distribution. This behavior is characteristic of a CR-modified shock [13], in which CRs diffusing upstream exert a pressure gradient on the incoming gas and form a precursor ahead of the main compression region. The CR-pressure gradient gradually decelerates the incoming wind before it reaches the main compression front, thereby broadening the velocity transition and weakening the gas compression across the front.
The horn-like structures of the CR pressure behind the clump (Figure 2) induce the dynamical influence on the background gas. The resulting CR pressure gradient acts as an external force and pushes the surrounding gas. This CR effect causes differences in the development of turbulence in the CR-poor and CR-rich cases. As shown in Figure 5, the turbulence behind the clump is enhanced in the CR-rich case compared to the CR-poor case. A swirling flow is formed behind the clump in the CR-rich case.

Figure 5. Baroclinic vorticity source term in the CR-poor (left) and CR-rich (right) models at 1.7 Myr. Black curves show gas streamlines, and yellow arrows indicate the local flow direction.
To clarify the generation mechanism of rotational gas motion, we also show the baroclinic source term in Figure 5. The gas vorticity is defined as . Taking
the curl of the momentum equation gives
which indicates that is the source that drives the gas vorticity. Compared with the CR-poor model, the CR-rich model exhibits a more extended baroclinic structure around the clump and in the downstream region.
Figure 6 shows the time evolution of the volume-integrated gas and CR energies ( and , respectively) within the cylindrical region pc and pc. The gas energy is decomposed into the internal and kinetic components (corresponding to the first and second terms in the right-hand side of equation (5), respectively), , and the total energy is defined as . Note that this is not an energetically closed region, as there are gas inflow and outflow, and radiative cooling and heating.

Figure 6. Time evolution of the volume-integrated gas and CR energies in the CR-poor (left) and CR-rich (right) models, evaluated within the cylindrical region (see text). All the energies are normalized by the initial total gas energy within the same region. The solid curves denote , , and , while the dotted curves denote and . The left and right panels use logarithmic and linear vertical scales, respectively.
A fraction of the gas is accelerated beyond its initial velocity, as shown in Figure 3 and Figure 4. In the CR-rich case, we have seen the enhancement of the turbulence behind the clump and wind acceleration in the outer region. Nevertheless, Figure 6 shows that the gas does work on CRs in the vast majority of the region, causing an increase of in both models. The energy gain associated with gas compression near the shock is expected to dominate over the energy transferred from CRs to the gas.
CR Spectrum
Gas compression can accelerate CRs and modify their energy spectrum. To plot the local CR spectra as a function of the proton kinetic energy, , in Figure 7, we extract the positions along the representative streamlines shown in the left panels. For the CR-poor model, we characterize the sampling position relative to the shock front by

Figure 7. Local CR proton spectra along the representative streamlines in the CR-poor model. The top and bottom rows correspond to and Myr, respectively. The left panels show the gas density distribution, , together with the representative streamline (white line) and the positions at which the CR spectra are plotted (colored filled circles). The right panels show the spectra, , with the same color as the corresponding circle in the left panel.
where is the distance along the streamline and is the shock position. For our parameter sets, pc.
Figure 7 shows that particles begin to be accelerated at the shock front and that the spectra harden downstream. As shown in Figure 3, the gas compression continues downstream, which further accelerates CRs. CR acceleration occurs throughout the broad compression region rather than being concentrated at a localized shock front. The gas at expands and accelerates so that the CR density turns to decline. At 0.56 Myr (upper panel), the CR spectra show irregular behavior. This may be exhibiting transient behavior as the CR distribution approaches a quasi-steady state. The high-energy part ( GeV) of the spectrum at (black), just outside the shock front, is affected by diffusion from the downstream, while the low-energy part remains unchanged from the initial one. At 1.7 Myr (lower panel), though the CR spectra become softer than those at 0.56 Myr, the spectral peak energy is significantly larger than the initial value.
In the CR-rich model, as the shock front is not well-defined, we adopt the same interval between the sampling positions. As shown in Figure 8, The spectral evolution is modest compared with the CR-poor case. This is because the gas compression behind the shock front is weaker than in the CR-poor case.

Figure 8. Same as Figure 7, but for the CR-rich model.
Gamma-Ray Emission
Gamma rays are produced through inelastic collisions between CR protons and ambient gas in the Galactic halo. As shown in Figure 7 and Figure 8, CRs are accelerated by gas compression in the interaction region between the clump and the shock front. The accelerated CRs subsequently diffuse into the dense clump, while rarefaction waves allow CRs initially contained in the clump to escape backward. The penetration of accelerated CRs into the dense clump enhances the hadronic gamma-ray emission.
To obtain the gamma-ray intensity for an observer, we reconstruct the three-dimensional axisymmetric distribution from the two-dimensional simulation and integrate the emissivity along the line of sight for a side-on view. The differential gamma-ray intensity is given by
We calculate the energy-integrated intensity over 1–10 GeV as
We define a fixed projected region of interest (ROI) of 200 200 pc centered on the clump. The same ROI is used at all simulation times and for both CR models.
Figure 9 and Figure 10 show the time evolution of the projected gamma-ray intensity . In both models, the emission is concentrated around the dense clump and becomes brighter with time. The CR-rich model naturally produces a higher overall intensity than in the CR-poor model. However, even with an initial of only 1% relative to the CR-rich model, the CR-poor model yields % of its intensity.

Figure 9. Time evolution of the gamma-ray intensity in the CR-poor model, integrated over 1–10 GeV. The three panels correspond to , , and Myr, from left to right. The image area shown here corresponds to the ROI (see text).

Figure 10. Same as Figure 9, but for the CR-rich model.
In our two models, the gamma-ray intensity reaches around – at the bright regions in the clump, approximately an order of magnitude higher than the observed surface brightness of the Fermi bubbles, – in the same energy band [2]. This suggests that an ensemble of such clumps distributed above the Galactic center account for the observed gamma-ray emission with a projected filling factor of order . Note that the gamma-ray intensity is not saturated yet, especially for the CR-poor case. We can expect brighter intensities from this kind of clumps.
In Figure 11 and Figure 12, we show the volume-averaged CR proton spectra over the three-dimensional region that projects into the fixed ROI, together with the gamma-ray fluxes integrated over the same projected ROI. In the CR-poor model, the average CR spectrum in the ROI becomes progressively harder with time, accompanied by a corresponding hardening. In contrast, the CR-rich model exhibits a modest change in the spectral shape, despite its higher gamma-ray flux.

Figure 11. Evolution of the CR proton and gamma-ray spectra in the CR-poor model. Left: volume-averaged CR spectra within the three-dimensional region corresponding to the projected ROI shown in Figure 9 and Figure 10. Right: gamma-ray spectra integrated over the same projected ROI for an observer at a distance of 8.5 kpc from the clump. The shaded region indicates the 1–3.16 GeV range used to evaluate the photon index in Figure 13.

Figure 12. Same as Figure 11, but for the CR-rich model.
To quantify the spectral hardening, we fit the gamma-ray spectra in the 1–3.16 GeV range with a power law, . Figure 13 shows the time evolution of the photon index . In the CR-poor model, the photon index decreases from initially to at

Figure 13. Evolution of the gamma-ray photon index measured in the 1–3.16 GeV band. The red and blue lines show the CR-poor and CR-rich models, respectively.
the end of the calculation. This hard spectral index is consistent with the observed index in the Fermi bubbles. On the other hand, the CR-rich model shows a smaller change, from to . The difference in the final index is due to the different gas compression degree in the two models.
Discussion
Fermi Bubbles
The Galactic halo is a complex system in which outflowing high-temperature galactic winds, cold accretion flows, and cosmic rays diffusing from the Galactic disk interact. As the two simulations we have demonstrated in this paper use a limited set of parameters, we are still at a preliminary stage in terms of uncovering the gas and CR interactions in the Galactic halo. However, our results indicate that CRs have a significant impact on the galactic wind and accretion flow, and that accreting gas clumps are a promising candidate for the source of the gamma-ray emission from the Fermi bubbles. Especially, hard spectral indices we obtained are encouraging for the Fermi bubbles.
In reality, the heating efficiency and diffusion coefficient depend on the position and environment of the clumps, and their velocities and sizes may be distributed over a wide range. Each clump reaccelerates CRs at a different efficiency. The duration of the clump–wind interaction is much longer than our simulation time. While we have verified gamma-rays from CRs that penetrate the clump directly from the acceleration region, the escaped CRs can also collide with gas in other clumps, emitting gamma-rays. The above facts suggest that the global properties of the emission are yet uncertain, but the gamma-ray emission efficiency can be higher than in our simulations.
If a projected filling factor of clumps is order , the average gamma-ray intensity can be comparable to the Fermi bubbles’ intensity. Recent H I observations have identified hundreds of neutral gas clouds in the Galactic nuclear wind F. J. Lockman et al. [33]. The clump mass in our simulation is , which is much larger than observed clumps (typically , K. A. Noon et al. [35]). To consider typical cases in the Fermi bubble, let us halve the clump radius, leading the mass to . The intensity would be reduced by only half, as the CR density would be almost the same. If the descent velocity of the clump to the Galactic disk is , the mass supply timescale by this clump with a diameter of 50 pc is Myr. Thus, the contribution of such a clump to the gas accretion rate is . To maintain star formation in the Galactic center (), clouds are required to exist in a cylinder with the Fermi bubbles’ radius of 3.5 kpc and a height of 50 pc. In this case, the projected filling factor is .
Although the above estimate is very rough, scattered small spots of gamma-ray emission regions are advantageous for explaining the observed surface brightness distribution. If CRs inside the bubbles emit gamma-rays uniformly, the thicker central region appears brighter, but the actual surface brightness is almost uniform. Considering the distance of 8.5 kpc, a bright region of about 50 pc is roughly the same size as the angular resolution of the Fermi LAT at 3 GeV. The gamma-ray fluxes shown in Figure 11 and Figure 12 are below the detection limit for a point source. Therefore, under the influence of foreground radiation, the gamma-ray spots sparsely distributed throughout the bubbles may not be inconsistent with the analysis that concluded the Fermi bubbles have a nearly uniform surface luminosity.
CR Shear Acceleration and Viscosity
In our simulations, we include only the term of the CR energy change due to compression in the diffusion–convection equation. Velocity shear generated around the clump may also provide an additional mechanism for CR acceleration, particularly at high energies ([14, 40]). As shown in Figure 5, CR pressure modifies the baroclinic generation of vorticity, suggesting that CR feedback can influence the development of complex gas flows. By introducing the contribution of the strain-rate tensor in the equations, we can include shear acceleration and its backreaction on the gas through anisotropic CR pressure. This may therefore modify both CR acceleration and gas evolution ([54]). Although the present results show stronger localized spectral hardening in the CR-poor model, shear acceleration could also enhance the high-energy CR population in the CR-rich model, where compression-driven acceleration is less efficient (Figure 8). On the other hand, the reaction of CRs on a fluid acts as viscosity. The turbulence induced by CRs in the CR-rich case may be softened by the CR viscosity.
Shock Identification
Detecting signatures of the shocks induced by accreting clumps is essential to verify the gas and CR interactions in the Galactic halo. However, in our cases, a prominent temperature jump does not appear at the shock front. The shock identification by X-ray telescopes or line emissions from metals (e.g., D. B. Henley et al. [23]) is difficult. It might actually be more promising to search for evidence of a temperature drop around clumps. While we have neglected electron CRs, they can be accelerated at the bow shocks. However, radio synchrotron emission from electron CRs around clouds is hard to detect (N. Leite et al. [30]). Future simulations including a wider range of cloud velocities, long-term turbulent evolution, and multiwavelength emission may provide a better method to identify the bow shocks.
Summary
We have investigated the interaction between a hot Galactic wind and a cold, dense gas clump using two-dimensional axisymmetric hydrodynamic simulations coupled with CR transport. We have taken into account the CR pressure on the gas, CR acceleration by gas compression, radiative cooling, photoelectric heating, and thermal conduction in our simulations. We have compared two cases: CR-poor () and CR-rich () cases. Our results demonstrate that CR pressure significantly modifies the gas compression and flow structures around the clump. While a sharp isothermal bow shock is formed in the CR-poor case, CR pressure weakens the shock with a smoother and broader pressure profile in the CR-rich case. In both cases, energy was transferred overall from the gas to CRs, where energy transfer via the shock wave is the primary contributor. However, in a part of the downstream, wind gas is accelerated faster than the initial velocity. In the CR-rich model, the wind acceleration and the turbulence behind the cloud are enhanced by the CR effects. These differences in the gas motion reflect on the efficiency of CR reacceleration. In the CR-poor model, strong gas compression behind the bow shock leads to significant CR spectral hardening. In contrast, in the CR-rich model, the weaker compression structure suppresses the efficiency of CR reacceleration.
We also estimate hadronic gamma-ray emission from CRs based on our simulations. CRs accelerated behind the bow shock penetrate into the dense cloud via diffusion. Most of the hadronic gamma-rays are emitted from the cloud. The resulting gamma-ray photon index in the 1–3.16 GeV range decreases from to in the CR-poor model, whereas the CR-rich model shows a more modest hardening to . The gamma-ray intensity around the clump reaches in the 1–10 GeV band, exceeding the characteristic surface brightness of the Fermi bubbles. Even in the CR-poor model, the gamma-ray intensity reaches approximately 10% of that in the CR-rich model, despite an initial CR pressure 100 times smaller.
These results suggest that an ensemble of dense clumps interacting with the Galactic wind may contribute to the gamma-ray emission of the Fermi bubbles. If the projected filling factor of such clumps is of order , their collective emission could reproduce the observed surface brightness. Although the global emission properties depend on the distribution and evolution of clumps, our findings demonstrate the importance of self-consistently treating CR transport and gas dynamics in a multiphase Galactic halo.
Acknowledgments
This work is supported by the joint research program of the Institute for Cosmic Ray Research (ICRR), the University of Tokyo, and KAKENHI grant Nos. 23H04899, 24H00025, and 25K07352 (K.A.), 24K00677, and 25H00394 (J.S.).
References
- [2]Ackermann, M., Albert, A., Atwood, W. B., et al. 2014, ApJ, 793, 64, doi: 10.1088/0004-637X/793/1/64DOI
- [3]Adriani, O., Akaike, Y., Asano, K., et al. 2022, PhRvL, 129, 101102, doi: 10.1103/PhysRevLett.129.101102DOI
- [4]Aguilar, M., Aisa, D., Alpat, B., et al. 2015, PhRvL, 114, 171103, doi: 10.1103/PhysRevLett.114.171103DOI
- [6]Armillotta, L., Ostriker, E. C., Kim, C.-G., & Jiang, Y.-F. 2024, ApJ, 964, 99, doi: 10.3847/1538-4357/ad1e5cDOI
- [7]Breitschwerdt, D., McKenzie, J. F., & Voelk, H. J. 1991, A&A, 245, 79
- [9]Cowie, L. L., & McKee, C. F. 1977, The Astrophysical Journal, 211, 135, doi: 10.1086/154911DOI
- [10]Crocker, R. M., & Aharonian, F. 2011, PhRvL, 106, 101102, doi: 10.1103/PhysRevLett.106.101102DOI
- [11]Di Teodoro, E. M., McClure-Griffiths, N. M., Lockman, F. J., & Armillotta, L. 2020, Nature, 584, 364, doi: 10.1038/s41586-020-2595-zDOI
- [12]Drury, L. O., & Falle, S. A. E. G. 1986, MNRAS, 223, 353, doi: 10.1093/mnras/223.2.353DOI
- [13]Drury, L. O., & Voelk, J. H. 1981, ApJ, 248, 344, doi: 10.1086/159159DOI
- [14]Earl, J. A., Jokipii, J. R., & Morfill, G. 1988, ApJL, 331, L91, doi: 10.1086/185242DOI
- [15]Fraternali, F., Marasco, A., Armillotta, L., & Marinacci, F. 2015, MNRAS, 447, L70, doi: 10.1093/mnrasl/slu182DOI
- [16]Fukumoto, Y., Asano, K., & Shimoda, J. 2026, ApJ, 1003, 114, doi: 10.3847/1538-4357/ae66feDOI
- [17]Girichidis, P., Frommer, C., Pakmor, R., & Springel, V. 2022, MNRAS, 510, 3917, doi: 10.1093/mnras/stab3462DOI
- [18]Girichidis, P., Werhahn, M., Pfrommer, C., Pakmor, R., & Springel, V. 2024, MNRAS, 527, 10897, doi: 10.1093/mnras/stad3628DOI
- [19]Giuliani, J. L. 1984, ApJ, 277, 605, doi: 10.1086/161731
- [20]Guo, F., & Mathews, W. G. 2012, ApJ, 756, 181, doi: 10.1088/0004-637X/756/2/181DOI
- [21]Haggerty, C. C., & Caprioli, D. 2020, ApJ, 905, 1, doi: 10.3847/1538-4357/abbe06DOI
- [22]Harten, A., Lax, P. D., & van Leer, B. 1983, SIAM Review, 25, 35
- [23]Henley, D. B., Shelton, R. L., & Kwak, K. 2014, ApJ, 791, 41, doi: 10.1088/0004-637X/791/1/41DOI
- [24]Kakiuchi, K., Suzuki, T. K., Inutsuka, S.-i., Inoue, T., & Shimoda, J. 2024, The Astrophysical Journal, 966, 230, doi: 10.3847/1538-4357/ad3638DOI
- [25]Kamae, T., Karlsson, N., Mizuno, T., Abe, T., & Koi, T. 2006, The Astrophysical Journal, 647, 692, doi: 10.1086/505189DOI
- [26]Kamae, T., Karlsson, N., Mizuno, T., Abe, T., & Koi, T. 2007, The Astrophysical Journal, 662, 779, doi: 10.1086/513602DOI
- [27]Kelner, S. R., Aharonian, F. A., & Bugayov, V. V. 2006, Physical Review D, 74, 034018, doi: 10.1103/PhysRevD.74.034018DOI
- [28]Koyama, H., & Inutsuka, S.-i. 2002, ApJL, 564, L97, doi: 10.1086/338978DOI
- [29]Lacki, B. C. 2014, MNRAS, 444, L39, doi: 10.1093/mnrasl/slu107DOI
- [30]Leite, N., Reuben, R., Sigl, G., Tytgat, M. H. G., & Vollmann, M. 2016, JCAP, 2016, 021, doi: 10.1088/1475-7516/2016/11/021DOI
- [31]Lhaaso Collaboration, Cao, Z., Aharonian, F., et al. 2025, National Science Review, 12, nwaf496, doi: 10.1093/nsr/nwaf496
- [32]Lockman, F. J., Benjamin, R. A., Heroux, A. J., & Langston, G. I. 2008, ApJL, 679, L21, doi: 10.1086/588838DOI
- [33]Lockman, F. J., Di Teodoro, E. M., Cary, S., & Fox, A. J. 2026, ApJ, 1009, 147, doi: 10.3847/1538-4357/ae99dbDOI
- [34]Nishiwaki, K., Asano, K., & Murase, K. 2021, The Astrophysical Journal, 922, 190, doi: 10.3847/1538-4357/ac1cdbDOI
- [35]Noon, K. A., Krumholz, M. R., McClure-Griffiths, N. M., Di Teodoro, E. M., & Armillotta, L. 2026, MNRAS, 548, stag715, doi: 10.1093/mnras/stag715DOI
- [36]Orlando, S., Peres, G., Reale, F., et al. 2005, A&A, 444, 505, doi: 10.1051/0004-6361:20052896DOI
- [37]Parker, E. N. 1953, ApJ, 117, 431, doi: 10.1086/145707DOI
- [38]Predehl, P., Sunyaev, R. A., Becker, W., et al. 2020, Nature, 588, 227, doi: 10.1038/s41586-020-2979-0DOI
- [39]Putman, M. E., Peek, J. E. G., & Joung, M. R. 2012, ARA&A, 50, 491, doi: 10.1146/annurev-astro-081811-125612DOI
- [40]Rieger, F. M., & Duffy, P. 2019, ApJL, 886, L26, doi: 10.3847/2041-8213/ab563fDOI
- [41]Roy, M., Su, K.-Y., Tonnesen, S., et al. 2026, ApJ, 1004, 147, doi: 10.3847/1538-4357/ae664fDOI
- [42]Schulreich, M. M., Breitschwerdt, D., & Kerp, J. 2026, MNRAS, 551, stag1405, doi: 10.1093/mnras/stag1405DOI
- [43]Shimoda, J., & Asano, K. 2024, ApJ, 973, 78, doi: 10.3847/1538-4357/ad6846DOI
- [44]Shimoda, J., & Inutsuka, S.-i. 2022, ApJ, 926, 8, doi: 10.3847/1538-4357/ac4110DOI
- [45]Shimoda, J., Inutsuka, S.-i., & Nagashima, M. 2024, PASJ, 76, 81, doi: 10.1093/pasj/psad081DOI
- [46]Spitzer, L. 1962, Physics of Fully Ionized Gases, 2nd edn. (New York: Interscience Publishers)
- [47]Strong, A. W., Moskalenko, I. V., & Ptuskin, V. S. 2007, Annual Review of Nuclear and Particle Science, 57, 285, doi: 10.1146/annurev.nucl.57.090506.123011DOI
- [50]Sutherland, R. S., & Dopita, M. A. 1993, ApJS, 88, 253, doi: 10.1006/1/19283
- [51]Thomas, T., Pfrommer, C., & Pakmor, R. 2025, A&A, 698, A104, doi: 10.1051/0004-6361/202450817DOI
- [52]van Leer, B. 1979, Journal of Computational Physics, 32, 101, doi: 10.1016/0021-9991(79)90145-1DOI
- [53]Wakker, B. P., York, D. G., Howk, J. C., et al. 2007, ApJL, 670, L113, doi: 10.1086/524222DOI
- [54]Webb, G. M. 1989, ApJ, 340, 1112, doi: 10.1086/167462DOI
- [55]Weber, M., Thomas, T., Pfrommer, C., & Pakmor, R. 2025, A&A, 698, A125, doi: 10.1051/0004-6361/202553954DOI