WorksPhysics and Astronomy
Ram Pressure Is Unable to Strip All Phases of the Interstellar Medium out of Spirals Falling into Galaxy Clusters
Abstract
Following the expansion of the Universe, spiral galaxies may fall at large speeds into galaxy clusters. Their interaction with the intracluster medium (ICM) may remove part of their gas and reduce their capacity for making new stars, as envisaged by Gunn & Gott. Ram pressure stripping (RPS) is commonly invoked to explain this gas loss and the wakes or tails left behind in-falling galaxies. Here we revise the impact of RPS on individual dense clouds and particularly focus on turbulent viscous stripping (TVS), driven by Kelvin-Helmholtz instabilities at the cloud-ICM interface, that affects the high-density and low-temperature molecular gas. For an extended gaseous body to decelerate, momentum from its exposed surface must be communicated into the gas by a hydrodynamic disturbance able to propagate through it. Three-dimensional hydrodynamic calculations follow this interaction with dense clouds. At cloud number densities cm, ram pressure acting through the transmitted shock strongly compresses and deforms the cloud and produces only a small deceleration; at cm, the slower transmitted shock produces much weaker deformation. At both cloud densities, TVS continuously removes gas from the cloud surface and forms qualitatively indistinguishable trailing wakes, with comparable mass-loss rates, suggesting that TVS acting on individual dense clouds can generate filamentary substructure within jellyfish-galaxy tails.
Introduction
The interstellar medium (ISM) of spiral galaxies falling into galaxy clusters is expected to experience a strong interaction with the so-called intracluster medium (ICM), the hot and low-density X-ray-emitting gas with temperatures ranging from to K (– keV) and densities in the range to ([3]). This gas is kept in place thanks to the enormous gravitational potential caused by the collection of trapped galaxies and their dark matter. Each of these galaxies has contributed during their fall and also during their collision and interaction with other galaxies to the ICM mass through the loss of part of their interstellar medium. However, the large fraction of baryonic matter implies an important amount of a primordial component resulting from the fact that galaxy clusters formed hierarchically by the merging of smaller systems. Galaxy clusters are thus roughly composed of – dark matter, – stars and galaxies and – hot gas ([3]) with an abundance of heavy elements ranging from to solar, implying an important injection of enriched matter out of galaxies ([27]).
On the other hand, spiral galaxies present a large variety of ISM phases and among these is the so-called galactic corona, the low-density, high-temperature gas that surrounds all of the Milky Way and thus all its stars and all phases of the ISM. The coronal gas presents very different physical properties compared to those found in all other ISM phases. The molecular phase, the phase where the stars are born, presents average density and temperature values, as compiled by K. M. Ferrière [13], of – atoms per and – K. The coronal phase is characterized by a density of – atoms per and a temperature of to several times K. Together with the cold atomic, warm atomic, and warm ionized components summarized by K. M. Ferrière [13], these phases span more than eight orders of magnitude in characteristic gas density. Very different are also their scale-heights, which are of the order of pc for the molecular material as opposed to several kpc in the case of the coronal gas. The volume occupied in the galaxy ranges from less than for
Corresponding author: Sergio Martínez-González
Email: [email protected] the molecular gas to about 30% to 70% for the coronal phase.
In NGC 4402, A. Abramson & J. D. P. Kenney [2] identify dense molecular clouds and elongated structures extending over hundreds of parsecs in regions from which most of the lower-density ISM has already been removed, and interpret the dense clouds as decoupled from that material. W. J. Cramer et al. [10] find molecular structures with distinct spatial and kinematic behavior, including a filament approximately 600 pc long, and propose that diffuse molecular gas can be stripped directly, whereas denser molecular gas initially resists direct stripping and decouples from the lower-density stripped material.
In this communication, the numerical calculations are restricted to the dense-cloud regime. At the opposite end of the density and temperature range, the hot coronal phase spans a much larger, kiloparsec-scale regime and will be considered separately.
The structure of the paper is as follows. Section 2 reviews previous numerical calculations of galaxies interacting with the ICM. Section 3 describes the impact of the ICM on the different ISM phases and formulates the conditions for ram pressure stripping (RPS). Section 4 introduces turbulent viscous stripping (TVS). Section 5 presents the numerical simulations and their results for dense clouds. Section 6 summarizes our conclusions.
Previous numerical calculations
In the 3D numerical calculations made this century to consider full spirals falling into galaxy clusters, the ICM was in most cases represented by a gas of constant density and temperature, unless an orbit was selected for the infalling galaxy. The latter allows the ICM conditions to evolve as the galaxy approaches pericenter. The first calculated 3D spiral structure is due to V. Quilis et al. [24], also developed by E. Roediger & M. Brüggen [25, 26] and S. Tonnesen & G. L. Bryan [28]. They considered a massive spiral galaxy with a gas disk followed using adaptive mesh refinement. Radiative cooling and gas self-gravity were included to allow the formation of a structured multiphase ISM in a static galactic potential.
In these calculations, galaxy disks were left to evolve for a few Gyr until they developed a structured ISM, long before the ICM began to ram against them. As discussed by S. Tonnesen & G. L. Bryan [28], “such disks fragment but do not form spiral arms”, producing lumpy gas distributions. A large fraction of the ISM is molecular hydrogen contained in giant molecular clouds with masses of – and sizes of 50–200 pc. The finest resolution in the mentioned 3D simulations was about 40 pc, insufficient to resolve the internal hydrodynamics of a molecular cloud and the smaller structures associated with star formation (C. Federath & S. S. R. Offner 2026). A 50–200 pc molecular cloud is therefore represented by only about 1–5 cells across its diameter. In three-dimensional shock-cloud calculations, 32–64 cells per cloud radius are required to capture the dominant dynamical processes (J. M. Pittard & E. R. Parkin [23]); earlier two-dimensional studies found convergence only at resolutions of order cells per cloud radius (F. Nakamura et al. [21]). For initially uniform clouds, the mean vorticity converges at resolutions of at least 32 cells per cloud radius (W. E. Banda-Barragán et al. [5]). Insufficient resolution limits the range of instability scales that can develop (J. M. Pittard & E. R. Parkin [23]), and the nonlinear evolution of Kelvin–Helmholtz instability depends sensitively on the numerical treatment of the interface (D. Leocanet et al. 2016). The limited spatial resolution of full-galaxy calculations remains important when interpreting narrow galaxy tails and molecular gas observed at large distances behind galaxies.
The general analytic estimate of J. E. Gunn & J. R. Gott [15] states that stripping is possible if the ram pressure caused by the ICM exceeds locally the gravitational anchoring force exerted by the galaxy. This condition establishes where galactic matter may be removed. For this pressure to decelerate an extended gaseous body, its momentum must be communicated from the exposed surface into the gas by a hydrodynamic disturbance able to propagate through it. The hydrodynamic collision between the ICM and the different ISM phases determines their deceleration and the subsequent structure of the wake.
The limited numerical resolution of full-galaxy calculations has also turned attention toward individual gaseous structures, including dense and low-density clouds. In high-resolution wind cloud calculations, W. E. Banda-Barragán et al. [6] find that the transmitted shock compresses the cloud and Kelvin–Helmholtz instabilities remove gas from its envelope, forming a filamentary tail downstream.
The impact of the ICM on the in-falling galaxies' ISM
As spiral galaxies fall into a galaxy cluster, there are two clearly different cases defined by the accretion process and ultimately by the orbit acquired by the entry bodies: one of them happens when the resultant infall velocity of the galaxy () is supersonic with respect to the intracluster medium (, where is the ICM sound speed). This causes a leading large-scale bow shock in the ICM, just ahead of the galaxy; the resultant high pressure acquired by the shocked ICM drives (if the density and temperature of the ISM allow it; see § 3.1) a second shock into the galaxy’s ISM. The two shock waves aim to thermalize the kinetic energy of the interacting gases. A second possibility, the subsonic case, happens if the temperature of the intracluster medium () is large enough as to inhibit the formation of a leading bow shock in the ICM. In this case, the ICM would ram directly against the ISM of the in-falling galaxy, driving, as in the supersonic case whenever the density and temperature of the ISM allow it, a shock able to decelerate the overtaken matter and transmit the ram pressure (, where is the ICM mass density) exerted by the ICM. Note that as the stars present no cross-section to the ICM ram pressure, they can continue to fall with their full speed; the shocks moving into the in-falling galaxy ISM decelerate the overtaken matter to finally force it to acquire very different velocities (see Table 1) determined only by the square root of the density contrast between the two media.
| ISM Phase | ||||
| 0.609 | Coronal | |||
| 1 | H I–H II | |||
| 1 | 1 | H I–H II | ||
| 1 | 10 | H I–H II | ||
| 2.3 | H | |||
| 2.3 | H |
Table 1. The initial ICM–ISM interaction: ram pressure stripping (RPS).
Note—Here is the adopted mean molecular weight, is the ISM number density, and is the post-shock gas velocity. Velocities are given as fractions of . No shock if K.
Energy-driven ram pressure stripping
Our estimates of RPS assume that the galactic ISM consists of constant-density slabs with a finite thickness representing the various phases of the galactic ISM. Let us first assume an infall velocity of the galaxy and a density of the intracluster medium . For example, (), or , where is the ICM number density. Then the pressures behind the two shocks that result from the collision with a galactic ISM of different densities () are: and . The hydrodynamic solution implies that ought to be equal to and this defines the velocity of the shocks driven into the ISM phases; the velocities of the ram-pressure shocks () are:
This implies, first of all, that upon the interaction with the ICM, the lighter ISM gas in galaxies is the one that would suffer the strongest deceleration as it is overtaken by the strongest ram-pressure shock and thus it is likely to be the first left behind the in-falling galaxy forming part of its wake. On the other hand, denser clouds would be traversed by correspondingly much weaker ram-pressure shocks and thus would remain falling together with their in-falling galaxy for a much longer time.
Note—Here is the shock crossing time through the ISM slab and is the distance traveled by the galaxy during . The adopted slab thicknesses are pc for the coronal and atomic phases and pc for H. values correspond to – yr after exiting G.
The densest molecular gas will hardly be decelerated at all. Thus for a galaxy falling face-on into a galaxy cluster, whether it falls at supersonic or subsonic speeds, in both cases the ICM ram pressure would drive decelerating shocks of different strengths into the encountered ISM gas phases. These compress and collect all matter within the full thickness () presented by the ISM gases into thin layers of shocked gas. The shocked ISM gas acquires a pressure equal or comparable to the ICM ram pressure. Pressure equilibrium does not prevent further momentum transfer from the ICM as long as a relative velocity persists. Subsequent deceleration is neglected in the estimates presented below. The shocked matter is then left behind the galaxy.
Gas compressed and decelerated by RPS then progressively lags behind the stellar disk and may appear as H I and H II, H, or even X-ray emitting gas. This decelerated gas, exposed to strong radiative cooling, would initially continue to fall at the reduced speed acquired after being shocked ().
The shocked gas then separates progressively from the stellar disk, which continues to fall at full speed. Table 1 gives the adopted , , , and for the selected ISM phases. Table 2 gives the slab shock crossing time, , through the adopted for each ISM phase, the distance traveled by the galaxy during that time, and the separation after and yr. Lower densities or a smaller ISM thickness at large galactocentric radii favor earlier removal. An inclined galaxy presents a larger effective path through its gaseous layers and correspondingly longer slab crossing times.
| ISM Phase | ||||
| yr | kpc | |||
| Coronal | 1.26 | 388 pc–38.8 kpc | ||
| H I–H II | 5.13 | 95.8 pc–9.58 kpc | ||
| 1 | H I–H II | 16.2 | 30.3 pc–3.03 kpc | |
| 10 | H I–H II | 51.3 | 9.58 pc–958 pc | |
| H | 43.1 | 2.00 pc–200 pc | ||
| H | 136 | 0.632 pc–63.2 pc |
Table 2. On the separation () between in-falling stellar disks and matter decelerated by RP.
With the assumed infall velocity of the galaxy and density of the intracluster medium here we look for the distance of separation () between the undecelerated galactic stellar disk and parcels of ISM gas, decelerated by ram pressure and left behind the galaxy, at time , assuming a constant post-shock velocity. Table 2 presents values of after and yr while galaxies falling with reach, after – yr, 1.23 kpc and 123 kpc. Continued deceleration by the ICM would increase the estimated separation distances.
Clearly, if one considers even higher-density clouds with densities similar to those of star-forming clouds (say those with an or larger) such clouds would hardly be decelerated at all and thus should continue to fall unimpeded by RPS along with their parent galaxy.
Matter detached from the galaxy continues to fall with a velocity smaller than and consequently encounters the ICM at a progressively smaller relative velocity.
The density contrasts in Table 1 and Table 2 imply differential deceleration among the ISM phases. Lower-density gas is decelerated first and lags behind the stellar component, exposing denser structures that can retain a larger fraction of their initial velocity for a longer time.
Other effects caused by the ICM thermal pressure
If one relaxes the slab geometry assumed earlier for the ISM and instead one allows for the existence of clouds, then depending on the in-fall velocity of galaxies (), sooner or later clouds within the galactic ISM would encounter the ICM. Direct contact with the ICM pressure leads to a large variety of possible evolutions with a direct impact on which of the two stripping processes, either RPS or TVS, may become locally dominant. TVS removes surface gas through shear-driven instabilities at the cloud–ICM interface (see Sect. 4). The defining parameter () results from the ratio of thermal pressures between the ICM and that of the considered cloud (, where and are the cloud number density and temperature). Large values of () would lead to cloud implosion and thus to larger densities and smaller dimensions of the main body of the cloud and thus TVS, acting on the full surface of the cloud and its remnants, may become the dominant stripping mechanism. On the other hand, small values of () inhibit implosions and instead promote the expansion of clouds and thus lead to lower densities and larger dimensions, facts which could favor RPS as the main stripping mechanism. However, given the high temperature often assigned to the ICM, the requirement for a small value of would be to select clouds with large densities. Otherwise, values of smaller than 1 will be hard to get. Large enough cloud densities would indeed cause small values of which should favor RPS unless such density values inhibit the formation and propagation of decelerating shocks into the clouds, as required for RPS. Clouds under such conditions will be stripped exclusively by TVS despite their small values of .
Formally, the joint action of ram pressure from the ICM plus its thermal pressure should speed up the propagation of the shock induced into the leading side of the in-falling cloud, which ensures that it reaches the cloud center first. The velocity of the shock would then be:
where is the ICM thermal pressure.
Momentum-driven turbulent viscous stripping (TVS)
Ram pressure compresses and decelerates the galactic ISM through transmitted shocks. The development of thin, elongated tails also requires surface instabilities at the ISM–ICM interface to continuously remove gas and feed the wake (P. Jáchym et al. [16]). Note that RPS acting particularly on low-density media is likely to create holes and gaps in the gaseous galaxy disks, allowing then for viscous ablation of dense material across a wide range of radii, including regions interior to the outer edge of disks. The stripping process outlined in this section is driven by the transfer of momentum from the hot ICM gas to the surface layers of the interstellar medium, particularly dense clouds. This process is relevant if the interface between the hot ICM gas and the interstellar medium becomes turbulent, as may result from Kelvin–Helmholtz instabilities generated by their rapid relative motion. These instabilities drag surface gas away and can establish a continuous ejection of ISM material into the tail of in-falling galaxies. The ICM exerts dynamic pressure, and parallel shearing stresses drag surface gas away. Boundary-layer mixing promotes Kelvin–Helmholtz and Rayleigh–Taylor instabilities. These instabilities agitate the interface between cold galactic gas and the hot ambient plasma, and the entrained ISM gas can form long, multiphase trailing tails (seen in jellyfish galaxies).
The foundational treatment of turbulent viscous stripping is that of P. E. J. Nulsen [22], who derived an expression for the mass-loss rate () from a spherical cloud of radius exposed to the streaming ICM:
which implies that for a constant density ICM and a constant in-fall velocity of the galaxy, the rate of mass loss is only a function of , regardless of the value of the density presented by the target cloud.
For a fixed cloud cross section, this density independence follows from the incident ICM momentum flux, , which supplies the momentum required to decelerate surface gas at a rate of order . A denser cloud therefore loses the same mass per unit time from a thinner surface layer as long as the cloud remains approximately spherical.
The number of parameters that define the final structure of galaxy wakes includes, first of all, the gas density of the ICM and the in-fall velocity of the galaxy (). Central to this issue are also the various phases of the galactic ISM and in particular, local densities, temperatures and the detailed structure of the volumes occupied by each of them within the target galaxy. Also, if one follows galaxies along their orbit one should care about the large possible variations in the density of the ICM and in . Another important parameter is the angle of inclination selected by the in-falling galaxy, which can drastically change the paths or lengths of the volumes confronted with the ICM in each of the different ISM phases.
Finally, the structure left in the wakes behind infalling galaxies is also subject to evolution similar to the one we witness and have learned to recognize in the ISM of galaxies. For example, after self-shielding from galactic ultraviolet radiation, large amounts of H I gas can be converted into molecular () clouds and vice versa, with molecular clouds being dismantled into coronal or H I gas in the presence of new bursts of star formation.
Numerical simulations
We calculate the interaction of the ICM with dense clouds corresponding to the and density cases considered in Table 1 and Table 2.
Each numerical calculation follows the interaction of the ICM with one galactic gas component at a time. The stellar component and the remaining ISM phases are not included.
The simulations were performed with TONANTZINTLA, a three-dimensional finite-volume hydrodynamic code on a uniform Cartesian grid. Our code1 solves the Euler equations in conservative form for an ideal gas with ratio of specific heats . Hydrodynamic fluxes are calculated with the Harten–Lax–van Leer–Einfeldt approximate Riemann solver (B. Einfeldt [11]). The time step is determined by the Courant–Friedrichs–Lewy condition. At each physical boundary, the hydrodynamic state in the boundary cells is copied from the adjacent interior cells, imposing zero normal gradients. A passive tracer is advected conservatively with the hydrodynamic mass fluxes to identify the gas initially associated with the infalling object. The simulations are adiabatic and neglect magnetic fields; no turbulent velocity field is imposed, and the galactic gravitational potential is not included. Radiative cooling can modify the mixing layer and promote cold-gas condensation ([14, 4]), with gas–grain collisions further increasing the cold-gas mass ([19]). Magnetic fields can modify the turbulent energy cascade ([7]) and suppress Kelvin–Helmholtz instabilities at the cloud–ICM interface through magnetic tension, reducing mixing and delaying cloud disruption ([20, 5]). Transverse magnetic fields can compress the cloud and enhance mass loss, and fields aligned with the flow can increase mixing ([9]). Our setup permits us to isolate the short-timescale cloud–ICM interaction and distinguish compression by transmitted shocks from surface stripping by shear-driven instabilities ([8, 23, 6]).
We adopt a temperature-dependent mean molecular weight. It is for , decreases linearly from 2.33 to 1.27 between 300 and 700 K, remains for 700 9300 K, decreases linearly from 1.27 to 0.609 between 9300 and 9700 K, and is for K.
All simulations use a computational domain comoving with the infalling object at its initial velocity. Velocities quoted in the figures and discussion are given in the ICM rest frame. Here , with along .
As the galaxy interacts with the ICM, the lower-density ISM is decelerated first and left behind, producing holes and gaps that expose denser structures to the ICM. The dense-cloud calculations begin once these dense structures are directly exposed. In the ICM rest frame, each cloud enters the ambient medium with .
We consider CLOUD-100 and CLOUD-1000. CLOUD-100 has an initial radius pc, cloud number density , and temperature K. For CLOUD-1000, pc, , and K. Both clouds encounter an ICM with number density and temperature K. The CLOUD-100 parameters were motivated by the adiabatic spherical-cloud model as384 of J. L. Cooper et al. [8]. To study the cloud–ICM interaction at high spatial resolution within computationally feasible runtimes, we adopted an ICM density higher than that used in the analytical estimates of Table 1 and Table 2. Densities of this order occur in the centers of some cool-core clusters (M. W. Kunz et al. [17]). At fixed cloud properties, this shortens both the characteristic Kelvin–Helmholtz growth time and the transmitted shock crossing time by a factor of 10. The initial cloud temperatures were chosen to maintain thermal pressure equilibrium with the denser ICM, avoiding additional compression and allowing the effects of RPS and TVS to be distinguished. Although these temperatures exceed the 10–20 K typical of molecular clouds, the analytical TVS rate in Equation (5) is independent of cloud density and temperature.
The computational domain extends from to 20 pc along and , and from to 50 pc along , with computational cells. The cell sizes are 0.078 pc along and and 0.137 pc along , so the initial 30 pc cloud diameter is sampled by 384 cells along and and about 219 cells along . A convergence test with cubic cells of 0.117 pc is presented below (Figure 3). The cloud–ICM interfaces are initialized with smooth transitions whose widths are set by the density contrast. For CLOUD-100 this gives five grid cells, corresponding to 0.68 pc along . The model parameters are listed in Table 3.
| Model | ||||
| CLOUD-100 | 15 | |||
| CLOUD-1000 | 15 |
Table 3. Initial parameters of the dense-cloud models. Units: in pc, in , in K, and in . Note—The ICM has and K in both calculations; the clouds enter it with and are initially in thermal pressure equilibrium with it.
Figure 3. Outward vortical cloud-tracer flux through the comoving control surface, compared with the TVS prediction of P. E. J. Nulsen [22] (black dashed). The blue solid, red dashed, orange dash-dotted, and green dotted curves show CLOUD-100, its cubic-domain convergence test, the extended-domain CLOUD-100 calculation, and CLOUD-1000, respectively. The CLOUD-100 curves are shown for Myr.
Figure 1 compares the evolution of the two dense-cloud calculations over the first 1.0 Myr. In each panel, the left half shows CLOUD-100 and the right half shows CLOUD-1000. For CLOUD-100, the density maps show the compression and progressive flattening of the cloud. Velocity shear along the cloud–ICM interface generates Kelvin–Helmholtz instabilities and enhanced vorticity that extends into the wake. For the – slices, the vorticity component normal to the displayed plane is
Figure 1. Evolution of the dense-cloud models CLOUD-100 and CLOUD-1000 at , 80 kyr, 400 kyr, and 1.0 Myr. In each time panel, the left half shows CLOUD-100 and the right half shows CLOUD-1000. The upper row shows and the lower row , where is the vorticity component normal to the displayed – plane. White regions in the lower panels mark vorticity amplitudes below the adopted display threshold. The density maps show compression by the transmitted shock and progressive flattening of CLOUD-100. At 1.0 Myr, only a small fraction of CLOUD-1000 has been processed by the transmitted shock. Velocity shear along the cloud–ICM interfaces generates Kelvin–Helmholtz instabilities and enhanced vorticity that extend into the wakes and accompany the continuous removal of gas from the cloud surfaces.
and the lower panels display . At the vorticity is confined to the cloud–ICM interface. As the cloud falls, gas removed from its surface is decelerated and remains behind, building the wake. A narrow wake is already present at 80 kyr. Kelvin–Helmholtz instabilities along the cloud flanks entrain surface gas and broaden the wake. By 400 kyr enhanced vorticity extends throughout much of the local wake. At 1.0 Myr the CLOUD-100 body is strongly flattened and continues to feed a broad turbulent wake.
Equation (1), evaluated with the initial conditions of CLOUD-100, gives . The corresponding cloud shock crossing time across the 30 pc diameter is . In the simulation, the transmitted shock crosses CLOUD-100 in approximately 920 kyr. At 1.0 Myr the cloud is strongly compressed and flattened, and a dense body remains identifiable. Equation (2) gives a post-shock velocity decrement of approximately , about 2% of the initial 1200 infall velocity. RPS strongly compresses and deforms the cloud and produces only a small deceleration of its dense body.
For CLOUD-1000, Equation (1) gives , and the cloud shock crossing time increases to . The transmitted shock advances much more slowly through the denser cloud. By 1.0 Myr, CLOUD-1000 has evolved for only about , so only a small fraction of the cloud has been processed by the transmitted shock. Kelvin–Helmholtz structures and surface-gas removal are nevertheless already well developed along the cloud flanks and wake. TVS is then essentially the sole mechanism removing gas from the cloud surface.
The leading surface contains a layer of enhanced vorticity associated with the compressed flow. Along the cloud flanks, velocity shear generates Kelvin–Helmholtz instabilities that strip surface gas and feed the wake. This surface removal corresponds to the Turbulent Viscous Stripping mechanism described by P. E. J. Nulsen [22]. At , RPS produces the crushing and deformation of the dense body. The morphology and mass-loss history indicate that the continuous removal of gas from the cloud surface is due to TVS.
The two dense-cloud models have the same , , and , so Equation (5) gives the same for CLOUD-100 and CLOUD-1000. During the common interval, their outward vortical mass fluxes increase toward predicted by P. E. J. Nulsen [22]. The comparison with Equation (5) is most direct while the clouds remain approximately spherical.
A stripped parcel with a velocity deficit develops a spatial lag behind the cloud, as in Equation (3). During the first 1.0 Myr the cloud travels approximately 1.23 kpc at its initial infall velocity. Gas stripped early in the interaction is decelerated and progressively lags behind the cloud. Once this gas crosses the rear boundary of the comoving computational domain, it disappears from the density panels. The 70 pc comoving computational domain therefore contains only the local portion of the wake surrounding the cloud.
We also follow CLOUD-100 in an extended computational domain spanning −30 to 30 pc along and and −20 to 200 pc along , with computational cells and the same initial cloud and ICM conditions. Figure 2 shows the density and temperature from this extended-domain calculation at kyr. The extended domain allows the downstream evolution of the wake to be followed over scales of hundreds of parsecs. Figure 3 shows mass-loss-rate estimates based on the outward flux of cloud material in vortical regions (where local rotation dominates deformation) through a control surface moving with the initial cloud velocity. Initially, the surface extends from to pc along and , and from to pc along , relative to the cloud center. The measured fluxes approach the prediction of Equation (5), , particularly for CLOUD-1000.
Figure 2. Mass density (top) and temperature (bottom) in the extended-domain CLOUD-100 calculation at kyr. The – slices span 220 pc along the wake.
The rate at which cloud material leaves the computational domain follows similar temporal trends and is presented in Appendix § 9. The CLOUD-100 curves in both figures are shown up to the analytical shock-crossing time, Myr.
We also performed a convergence test with a pc computational domain and cells. Its spatial resolution is 0.117 pc in all directions, compared with 0.078 pc along and and 0.137 pc along in CLOUD-100. Their outward vortical fluxes show similar overall trends over the calculated interval and agree to within approximately 1% by the end of the common comparison interval.
CLOUD-100 shows ram pressure and TVS acting concurrently: ram pressure acts through the transmitted shock to compress and decelerate the dense cloud, and TVS continuously removes gas from its surface. The dense body remains identifiable at 1.0 Myr despite the continuous loss of material into the wake. In CLOUD-1000 the transmitted shock is much slower, so ram pressure produces only slow compression, deformation, and deceleration of the cloud, whereas gas is continuously removed from its surface by TVS.
Conclusions
The fall of spiral galaxies plunging at large speeds into a vast sea of ICM in galaxy clusters leads to very complex hydrodynamics promoted by the large number of possibilities which determine the route of every fall and by the intrinsic properties presented by a well-structured multiphase ISM.
The large density range among the ISM phases produces different deceleration timescales during the interaction with the ICM. Lower-density gas can lag behind the stellar component before denser structures are appreciably decelerated.
At , a transmitted shock still propagates through the dense cloud and strongly compresses and deforms its dense body. As the cloud density increases, the transmitted shock becomes progressively slower, with a velocity set by the square root of the density contrast between the ICM and the cloud. At higher cloud density, the transmitted shock crosses the cloud on increasingly long timescales, yet a turbulent wake can develop well before the cloud is fully processed. CLOUD-1000 already exhibits a turbulent wake when only a small fraction of the cloud has been crossed by the transmitted shock. CLOUD-100 and CLOUD-1000 have very different shock-crossing times, yet develop qualitatively indistinguishable trailing wakes. Their outward vortical mass fluxes are of the same order as predicted by P. E. J. Nulsen [22]. In Nulsen’s model, Kelvin–Helmholtz stripping progressively thickens the cloud–ICM interface until the instability is suppressed. The resulting mass-loss rate depends on the ICM density, relative velocity, and cloud cross section, with no dependence on cloud density or temperature. Ram pressure acts through the transmitted shock, compressing the dense body and producing only a small deceleration. In both calculations, Kelvin–Helmholtz surface stripping (TVS) removes cloud material and forms the trailing wakes. At still larger ISM densities, such decelerating shocks can be totally inhibited, leaving TVS as the sole mechanism able to continue removing surface gas. TVS can therefore help sustain gaseous tendrils within jellyfish-galaxy tails, connecting the small-scale filamentary substructure resolved in actively stripped dense ISM {(citep)}(A. Abramson et al. 2016) with molecular filamentary structures observed within extended jellyfish tails {(citep)}(P. Jáchym et al. 2019).
Full galaxies falling into galaxy clusters become exposed to a global interaction with the pervasive ICM. This implies that all physical processes and stripping mechanisms described above begin to operate at the same time on all different phases of the ISM. All decelerated ISM gas that eventually detaches from its parent in-falling galaxy becomes part of its wake. When dealing with galaxy wakes one should keep in mind that one is referring to decelerated ISM. Detached gas, despite moving more slowly than the parent galaxy, inevitably continues to fall and interact with the pervasive ICM, causing further evolution of the detached ISM.
Acknowledgments
The authors thank the anonymous referee for a constructive report that helped improve the clarity and content of the manuscript. Sergio Martínez-González acknowledges support from the Fundación Occident and the Instituto de Astrofísica de Canarias under the Visiting Researcher Programme 2025 to 2028 agreed between both institutions. The authors gratefully acknowledge the computing resources, technical expertise, and support provided by the Laboratorio Nacional de Supercómputo del Sureste de México, a member of SECIHTI’s network of national laboratories.
Appendix
Cloud material leaving the computational domain
Figure 4 shows the rate at which cloud material leaves the comoving computational domain. Its evolution is similar to that of the vortical flux in Figure 3. For the extended-domain CLOUD-100 calculation, the mean rate is between and Myr, lower than for CLOUD-100 over the same interval, consistent with the longer transport time. Increased pressure following shock compression can also drive lateral expansion of CLOUD-100 and contribute to domain losses, especially after the shock fully crosses the cloud.
Figure 4. Rate at which cloud material leaves the comoving computational domain. The blue solid, red dashed, orange dash-dotted, and green dotted curves show CLOUD-100, its cubic-domain convergence test, the extended-domain CLOUD-100 calculation, and CLOUD-1000, respectively. The black dashed line marks the analytical TVS prediction of P. E. J. Nulsen [22] for the initial cloud radius. The CLOUD-100 curves are shown for Myr.
References
- [2]Abramson, A., & Kenney, J. D. P. 2014, AJ, 147, 63, doi: 10.1088/0004-6256/147/3/63DOI
- [3]Bahcall, N. A. 1996, arXiv e-prints, astro, doi: 10.48550/arXiv.astro-ph/9611148DOI
- [4]Banda-Barragán, W. E., Brüggen, M., Heesen, V., et al. 2021, MNRAS, 506, 5658, doi: 10.1093/mnras/stab1884
- [5]Banda-Barragán, W. E., Federrath, C., Crocker, R. M., & Bicknell, G. V. 2018, MNRAS, 473, 3454, doi: 10.1093/mnras/stx2541
- [6]Banda-Barragán, W. E., Parkin, E. R., Federrath, C., Crocker, R. M., & Bicknell, G. V. 2016, MNRAS, 455, 1309, doi: 10.1093/mnras/stv2405
- [7]Beattie, J. R., Federrath, C., Klessen, R. S., Cielo, S., & Bhattacharjee, A. 2025, Nature Astronomy, 9, 1195, doi: 10.1038/s41550-025-02551-5DOI
- [8]Cooper, J. L., Bicknell, G. V., Sutherland, R. S., & Bland-Hawthorn, J. 2009, ApJ, 703, 330, doi: 10.1088/0004-637X/703/1/330DOI
- [9]Cottle, J., Scannapieco, E., Brüggen, M., Banda-Barragán, W., & Federrath, C. 2020, ApJ, 892, 59, doi: 10.3847/1538-4357/ab76d1DOI
- [10]Cramer, W. J., Kenney, J. D. P., Cortes, J. R., et al. 2020, ApJ, 901, 95, doi: 10.3847/1538-4357/abaf54DOI
- [11]Einfeldt, B. 1988, SIAM Journal on Numerical Analysis, 25, 294, doi: 10.1137/0725021DOI
- [13]Ferrière, K. M. 2001, Reviews of Modern Physics, 73, 1031, doi: 10.1103/RevModPhys.73.1031DOI
- [14]Gronke, M., & Oh, S. P. 2018, MNRAS, 480, L111, doi: 10.1093/mnrasl/sly131DOI
- [15]Gunn, J. E., & Gott, III, J. R. 1972, ApJ, 176, 1, doi: 10.1086/151605DOI
- [16]Jáchym, P., Kenney, J. D. P., Sun, M., et al. 2019, ApJ, 883, 145, doi: 10.3847/1538-4357/ab3e6cDOI
- [17]Kunz, M. W., Schekochihin, A. A., Cowley, S. C., Binney, J. J., & Sanders, J. S. 2011, MNRAS, 410, 2446, doi: 10.1111/j.1365-2966.2010.17621.xDOI
- [19]Martínez-González, S. 2025, A&A, 702, L6, doi: 10.1051/0004-6361/202556389
- [20]McCourt, M., O’Leary, R. M., Madigan, A.-M., & Quataert, E. 2015, MNRAS, 449, 2, doi: 10.1093/mnras/stv355DOI
- [21]Nakamura, F., McKee, C. F., Klein, R. I., & Fisher, R. T. 2006, ApJS, 164, 477, doi: 10.1086/501530DOI
- [22]Nulsen, P. E. J. 1982, MNRAS, 198, 1007, doi: 10.1093/mnras/198.4.1007DOI
- [23]Pittard, J. M., & Parkin, E. R. 2016, MNRAS, 457, 4470, doi: 10.1093/mnras/stw025DOI
- [24]Quilis, V., Moore, B., & Bower, R. 2000, Science, 288, 1617, doi: 10.1126/science.288.5471.1617DOI
- [25]Roediger, E., & Brüggen, M. 2006, MNRAS, 369, 567, doi: 10.1111/j.1365-2966.2006.10335.xDOI
- [26]Roediger, E., & Brüggen, M. 2007, MNRAS, 380, 1399, doi: 10.1111/j.1365-2966.2007.12241.xDOI
- [27]Sarazin, C. L. 1988, Cambridge Astrophysics, Vol. 11, X-ray emission from clusters of galaxies (Cambridge: Cambridge University Press)
- [28]Tonnesen, S., & Bryan, G. L. 2009, ApJ, 694, 789, doi: 10.1088/0004-637X/694/2/789DOI



