Annu. Rev. Astron. Astrophys. 2012. 50: 491-529 Copyright © 2012 by Annual Reviews. All rights reserved

4. ORIGINS

Spiral galaxy halo gas is likely to originate from multiple sources, including from the intergalactic medium, satellite galaxies, and the disk of the galaxy itself. This section discusses these origins in the context of the observational results presented in Section 2 and Section 3. Viable origin scenarios should be consistent with the following key observational features.

• Large scale flows of multiphase gas are prevalent in galaxy halos. This is especially evident from the spatial and kinematic link between the cold, warm, and warm-hot gas in the Milky Way halo (see Figures 1 and 2).

• The densest halo gas is closest to the disk in all phases. Exceptions (and the largest cold halo features) are largely linked to satellite accretion (see Figures 3 and 6).
• Gas at the disk-halo interface shows a clear temperature gradient (cold to hot), and kinematic gradient (decreasing rotation) with increasing z-height (see Figure 5). This gas is also more enriched than gas further out in the halo and more extensive in galaxies with higher SFRs.

• Diffuse gas extends to radii > 100 kpc in spiral galaxy halos. This is evident for the Milky Way from the properties of clouds associated with the Magellanic System and stripped dwarf galaxies, and for other galaxies from absorption line results that show large covering fractions.

• For the Milky Way, the mass in warm and cold halo gas is < 10⁹ M, and, though largely model-dependent, the hot gas mass is < 10¹⁰ M. These measurements are consistent with the HI and x-ray observations for other low redshift spiral galaxies of similar mass.

• The halo gas appears to be bound to the dark matter halos of the spiral galaxies. This gas is also found to be enriched with metals at some level (i.e., non-primordial).

Throughout this section we refer to and make connections to a cosmological grid simulation of a Milky Way-mass galaxy at z = 0 (referred to as MW simulation; Joung et al. 2012a, Fernández et al. 2012). We use this simulation not as the final word on the state of the Milky Way halo, but rather as a guide to our understanding of the important physical processes at play. The simulation, performed with an adaptive mesh refinement (AMR) code, Enzo (Bryan & Norman 2000, O'Shea et al. 2004), has sufficiently high spatial (136-272 pc comoving or better at all times) and mass ( ~ 10⁵ M) resolutions to allow us to study and track the spatial and kinematic distribution of the multiphase gas in the halo in detail. The simulation includes metallicity-dependent cooling, metagalactic UV background, shielding of UV radiation by neutral hydrogen, photoelectric heating for the diffuse gas, star formation and stellar feedback (but not AGN feedback). Figure 7 displays a snapshot of the MW simulation in HI at z = 0.3 and zooms in on halo gas origins relevant to each of the below sections. Figure 8 shows a cartoon based on the simulation that is split to show the main accretion and feedback mechanisms distinctly. It also depicts the density-temperature diagram of simulation cells at z = 0.

Figure 7. A cosmological AMR simulation of a Milky Way galaxy at z = 0.3 (Joung et al. 2012a). We chose t = 10.3 Gyr as a relatively low redshift time slice, as it clearly shows examples of three different origins of neutral gas in the halo as indicated by the numbered boxes in the top panel. Box 1 represents the interaction between outflowing gas and accreting halo gas, box 2 shows a filamentary cool flow feeding the disk, and box 3 represents the accretion of three satellites. The cut-outs show the total gas density (N), temperature (T), stellar density (*) and metallicity (Z). The features become less distinct in the total gas density plots due to the large amount of multiphase gas in the halo.

Figure 8. Top: A diagram based on the MW simulation at z = 0 to a radius of 100 kpc where the top hemisphere shows the outflow due to SN-driven winds (pink) and the bottom shows the inflow of cooling (green to blue) and heating (green to red) filamentary flows and a satellite (black). Bottom: The density-temperature diagram for cells in the MW simulation at z = 0 with color coding corresponding to the top panel and main features noted. The directions of increasing pressure and entropy are shown with the arrows in the upper right. The cells were randomly selected from the simulation with the probability of selection proportional to gas density.

4.1. Accretion from the IGM

Under CDM, the large majority of the baryons enter the virial radius of a galactic halo in gaseous form along cosmic filaments. This is in contrast to the classical picture that posited that the incoming gas would spherically collapse and be shock-heated close to the virial temperature of pressure-supported gas within a galaxy's halo (White & Rees 1978, White & Frenk 1991). Analytic arguments (Birnboim & Dekel 2003, Dekel & Birnboim 2006, Binney 1977) and SPH simulations have shown that some of the gas will never be shock-heated to the virial temperature of the halo and be accreted via a "cold mode" (Keres et al. 2005, Brooks et al. 2009). This mode is found to dominate in galaxies at high redshift (z 2), as well as in present day low-mass galaxies (M_halo 10¹² M). We note that this mode is referred to as cold, but in reality the gas is at temperatures 10^5.5 K, so much of it is not cold under our observational definition in Section 2.

The distinction between hot and cold mode accretion is based on the maximum temperature that a given gas particle attains along its trajectory as it enters the virial radius. At z ~ 0, for galaxies with masses similar to the Milky Way, a large fraction of the accreted gas is hot mode and does not connect directly to the disk. The hot mode gas is heated either by one strong shock close to the virial radius or via a series of weak shocks as the kinetic energy of the outer portions of the filaments is converted to thermal energy due to compression from the existing halo gas (Keres et al. 2005, Joung et al. 2012a). In the MW simulation, ~ 70% of the mass inflow at the virial radius is consistent with coming in along filaments. The cold mode of the accretion at z ~ 0 is found in the core regions of the compressed filaments at temperatures 10^5.5 K. This gas stays cold because it is shielded from interacting with the existing halo medium by the hot mode gas that surrounds it in the filamentary flow (Joung et al. 2012a). The accretion of the hot and cold mode gas along filamentary flows is depicted in the top panel of Figure 8 and box 2 of Figure 7.

The properties of the filamentary flows can be examined further in the MW simulation. In the density-temperature diagram of Figure 8, the filamentary flows generally have intermediate specific entropy between that of the disk gas and the outflows. The flows come in at temperatures of ~ 10^4.5-5.5 K and the outer portions are heated toward ~ 10⁶ K and the inner portions cool toward ~ 10⁴ K. The gas at < 10^5.5 K has a relatively high neutral fraction (generally f_neut > 10^-4), and extends from radii of approximately 25 - 200 kpc. Some of the cold cores of the filaments have HI densities comparable to those of gas stripped from simulated satellites, but they generally have lower metallicities (Z / Z 0.2; see Figure 7). Details on the HI properties of the halo gas in the MW simulation are outlined in Fernández et al. (2012).

In terms of observations, the filamentary flows could be represented by the diffuse warm and warm-hot gas detected with absorption lines near the velocity of the galaxy (Section 3.2); the majority of these filaments are not detectable in HI with our current sensitivities until close to the disk (see below). The covering fraction of the observed warm halo gas approaches 100%, and this is consistent with the covering fraction in the MW simulation when low enough column density gas is selected (< 5 × 10¹³ cm^-2; Fernández et al. 2012). The gas that is heated to temperatures > 10^5.5 K largely becomes part of the general hot halo medium, and may be related to the fact that we see evidence for a diffuse hot halo medium out to large radii (however see also Section 4.2). The MW simulation finds 2 × 10¹⁰ M in hot gas within the virial radius and this is consistent with the simulation of Sommer-Larsen (2006). This is greater than the amount found by observations thus far.

For the gas in the filamentary flows to feed the star-forming disk, the gas needs to cool further; the gas potentially forms clouds within the halo (Maller & Bullock 2004, Connors et al. 2006a, Kaufmann et al. 2006, Kaufmann et al. 2009, Lin & Murray 2000). The cold mode gas contains density enhancements that can cool as long as the cooling time is shorter than the disruption time (Joung et al. 2012b, Binney et al. 2009). Linear fluctuations were not able to cool in these detailed studies and cooling without significant density enhancements is also not seen in the MW simulation (Figures 7 & 8). The gas that does cool may not survive the trip to the disk without disrupting further (see Section 5). It remains somewhat unclear how halo gas accreted in hot mode can feed the disk. Possibly closer to the disk, where it is denser, the hot gas may mix with feedback material and become part of the disk-halo interface (Section 6.2). In any case, in the MW simulation the gas is only observable as HI clouds within ~ 50 kpc, and this is consistent with observations and previous simulations (Sommer-Larsen 2006, Kaufmann et al. 2006, Keres & Hernquist 2009). The mass in observed warm and cold clouds is 10^8-9 M in both the observations and simulations when the same type of gas is selected (Peek et al. 2008, Fernández et al. 2012).

4.2. Feedback

After baryons collect at the centers of galaxies and form stars, the stars and central black holes act back on their environments and can affect the gaseous galactic halos. These "feedback" processes include radiation from young stars that heat and ionize the halo gas (Bland-Hawthorn & Maloney 1999, Bland-Hawthorn & Maloney 2002, Bland-Hawthorn & Putman 2001), mechanical energy from supernova explosions (e.g., Mac Low & Klessen 2004, Veilleux et al. 2005) and AGNs (Antonuccio-Delogu & Silk 2010, Di Matteo et al. 2005, Di Matteo et al. 2008), as well as momentum-driven winds due to radiation pressure from central sources on dust grains (Murray et al. 2005) and a galactic fountain as enriched hot gas rises up from the disk (e.g., Shapiro & Field 1976, Joung & Mac Low 2006). We will discuss each of these mechanisms in turn. Feedback is required in cosmological simulations to reproduce the fundamentals of observed galaxies; however, despite its importance, it is described by phenomenological prescriptions partially due to our lack of understanding of the dominant process(es) (Oppenheimer & Davé 2006, Springel & Hernquist 2003, Stinson et al. 2006, Dalla Vecchia & Schaye 2008). Halo gas is in many ways a more direct consequence of disk feedback than the stellar observables of galaxies and is therefore an important tool for understanding these processes.

Radiation escaping from a galaxy's disk ionizes and heats halo gas. H observations of cold Galactic halo clouds have been used to estimate that 6% of the Milky Way's ionizing photons escape normal to the disk (f_esc = 1-2% when averaged over the solid angle; Bland-Hawthorn & Maloney 1999 and see Section 2.2 and Section 2.1.1). For other spiral galaxies, there are mainly upper limits (e.g., Tumlinson et al. 1999, Deharveng et al. 2001), and the measurements are consistent with escape fractions of less than a few percent as long as starbursts are excluded. As expected, if f_esc is estimated with a galaxy's diffuse ionized gas it is highly dependent on the clumpiness of the galaxy's ISM (Clarke & Oey 2002, Zurita et al. 2002, Fernandez & Shull 2011). Radiation escaping from galaxy disks is only starting to be included in cosmological galaxy formation simulations. Constraints on the escape fraction impact both halo gas and the ionization of the IGM as a whole, in particular at higher redshift (e.g., Ciardi et al. 2003, Gnedin 2000).

Mechanical feedback mechanisms, such as SN-driven or past AGN-driven outflows, are the most relevant in the context of the origin of a spiral galaxy's halo gas. For the Milky Way, a diffuse hot halo medium is evident out to the distance of the Magellanic System ( ~ 50-100 kpc), and this indicates either relatively strong Galactic winds and/or gas left from hot mode accretion is present. The detected x-ray and gamma-ray 'bubbles' are also consistent with the Milky Way having feedback from a central engine in the past 10-15 Myr (Su et al. 2010). For other galaxies, outflows are consistent with the enriched warm-hot gas at distances on the order of the galaxy's virial radius (Section 3.2). Finally, the baryon fraction within the virial radius of Milky Way-like spirals is substantially smaller than the cosmic mean ( ~ 0.4 for the Milky Way), and this is attributed to powerful outflow mechanisms (Bregman 2007, McGaugh et al. 2010), as is the ability to reproduce the basic properties of a galaxy in cosmological simulations (Benson et al. 2003, McCarthy et al. 2010, McCarthy et al. 2011).

Some of the material from large scale galactic winds is likely to escape the galaxy's halo (potentially explaining the lack of baryons), and a fraction is also thought to feed the hot halo. In Figure 8 we show a cartoon based on the MW simulation that depicts outflow due to SN-driven winds. This wind extends from velocities of 200 km s^-1 to over 1000 km s^-1. The fastest gas escapes, while the lower velocity gas feeds the hot halo high entropy, enriched material. Observationally, the lower velocity material may be some of the low density absorption line systems in the Milky Way and other galaxy halos (Section 2.2 & Section 3.2); in particular, the gas at high positive velocity in the general direction of the Galactic Poles shown in Figure 2. The SN-wind gas at higher velocities is too hot and low density to be observable. Figure 8 also shows that the MW simulation confirms previous simulation expectations that supernova feedback can be thought of as injecting high entropy gas into the system. In particular, once launched, the specific entropy of this enriched gas remains unchanged for a large range of the galactocentric radius, suggesting adiabatic expansion.

Momentum driven feedback is also commonly discussed as a method of launching material from the disk. This mechanism is likely to be relevant only in galaxies with powerful starbursts or AGNs (Oppenheimer & Davé 2006, Murray et al. 2007) that may have operated in the more violent, formative years of present-day spiral galaxies like the Milky Way (z ~ 2-3; e.g., Kim et al. 2011). In addition, this mechanism may have difficulty launching material as, even if one assumes the maximal efficiency of converting mass to energy and that all the available momentum is used to drive outward motion of the surrounding gas, the outflow velocity would be too low to eject gas for reasonable values of the mass loading factor. Momentum driven winds have also been invoked as a method to launch substantial amounts of dust into galaxy halos (Aguirre et al. 2001); however, energy-driven mechanisms may be equally effective through the coupling of dust and gas. Dust observations thus far could be consistent with either mechanism.

A galactic fountain is distinct from the powerful galactic winds discussed above in that it is a disk-wide phenomenon in which gas is driven upward by the cumulative effect of supernova explosions and adiabatic expansion, and subsequently falls back down due to radiative cooling (Shapiro & Field 1976, Bregman 1980). The galactic fountain phenomena is most likely to affect the detailed structure and dynamics of gas at the disk-halo interface. Kiloparsec-scale ISM simulations in vertically elongated boxes (de Avillez 2000, de Avillez & Breitschwerdt 2005, Joung & Mac Low 2006, Booth & Theuns 200, Hill et al. 2012) report ejecta cooling and condensing out at a few kpc heights having morphology and kinematics similar to those of observed IVCs (Section 2.5). The near-solar metallicities and dustiness of many IVCs are also consistent with a disk origin. The temperature gradient of the disk-halo interface gas is expected with hot gas from the SN rising to higher z-heights, and the kinematic lag suggests a combination of this type of feedback and fueling (see Section 6.2). The large population of small disk-halo clouds may also be linked to star formation processes in the disk (Ford et al. 2010). The cartoon shown in Figure 5 summarizes some of the observational findings for gas at the disk-halo interface and has many similarities to the fountain simulations.

Feedback material mixes and interacts with the existing halo gas. In box 1 of Figure 7, we show an example in the MW simulation of neutral halo clouds that have originated from the interaction of a supernova-driven wind and incoming cooler flows. The gas has condensed at distances of 50-100 kpc from the disk and has metallicities around 0.3 Z, which is greater than many of the other simulated HI halo features, and lower than that of the wind itself. The HI clouds fragment further and impact the disk ~ 300 Myr later than the timestep shown. The higher metallicities of some regions in HVC Complex C may be representative of this mixing process closer to the disk. HVC interaction with feedback material may also be evident in the H brightness of some complexes and the mixture of photo- and collisional ionization required to model the line ratios measured for halo gas.

4.3. Satellite Accretion

Satellites lose their gas as they move through a spiral galaxy's halo (e.g., Putman et al. 2003b, Grcevich & Putman 2009). The gas is primarily ram pressure stripped from the dwarf in models, although tidal forces also play a role (Mayer et al. 2007, Mayer et al. 2006). CDM simulations find 40% of L_* galaxies host a 0.1 L_* satellite within its virial radius and this is consistent with results from SDSS (Tollerud et al. 2011, Liu et al. 2011). This percentage drops significantly at smaller radii (e.g., 12% within 75 kpc) and also when only blue and star forming satellites like the LMC are considered (8.2% within 100 kpc; Tollerud et al. 2011, James & Ivory 2011). We know from the Local Group that the vast majority of the satellites have < 0.1 L_*, however the amount of gas they bring is substantially smaller ( ~ 10^5-8 M each). For instance in the Local Group, the Magellanic System contains ~ 30-50% of the HI mass of the Milky Way, and M33 will eventually provide~ 25% of M31's HI mass (Putman et al. 2003b, Putman et al. 2009c). The lower mass satellites within the virial radius of the Milky Way will only have provided ~ 1%, or if we include the large number of yet unobserved satellites predicted by CDM, ~ 10% (Grcevich & Putman 2009).

The cumulative effect of the accretion of satellites will leave behind a large amount of gas that will eventually feed the disk. Since most galaxies are likely to be stripped at radii > 20 kpc (Grcevich & Putman 2009, Mayer et al. 2006, Nichols & Bland-Hawthorn 2011), much of the cold gas will be integrated into the diffuse galactic halo before eventually recooling at the disk (see Section 5). This is consistent with multiphase flows being found throughout galaxy halos and helps to sync the angular momentum distribution of the satellite gas with the spiral's halo gas. This integration of satellite material with the halo medium can be seen in box 3 of Figure 7 which depicts the accretion of a triple system of dwarf galaxies. The stripped gas shows a temperature gradient as it integrates into the hot halo material, and the column density and metallicity of the gas also gradual decreases (see also Figure 8). The cold and warm satellite gas has on average higher radial velocities than other features in the halo. This material becomes untraceable several hundred Myrs after being stripped from the satellite core, but may leave behind overdensities that help seed the cool filaments closer to the disk (see Section 5 and Section 6.2).

The system shown in Figure 7 has many similarities to what is observed for the Magellanic System of the Milky Way, which has given us a detailed look at the relationship between gaseous satellites and a spiral galaxy's halo gas (see Figures 1 - 3 and 6). The Magellanic System shows a column density gradient along its length and multiphase gas surrounding it, particularly at the tail where it is being integrated into the hot halo medium. This system can also account for the highest negative and positive velocity halo gas in the Milky Way halo. The interaction of the satellite gas with an existing halo medium, and the high covering fraction of the warm and warm-hot gas does make it relatively clear that mechanisms besides satellite accretion are also feeding the halo. In addition, for the Milky Way, most of the HVCs (besides those associated with the Magellanic System) do not show any clear link to satellites or stellar streams. There is a large body of literature describing the Magellanic System and it is thought to be the result of a combination of a tidal interaction between the two satellites and their subsequent interaction with the Milky Way (Connors et al. 2006b, Mastropietro et al. 2005, Besla et al. 2010, Gardiner 1999, Diaz & Bekki 2012, Nidever et al. 2008).

In general if the HI gas observed in galaxy halos is directly from a satellite, it is likely to have been relatively recently stripped from the satellite. This is consistent with the link between large HI features in spiral galaxy halos and a relatively recent interaction event (see Section 3.1). Satellite HI features are also likely to have higher metallicities than gas that originates from the IGM (e.g., Fernández et al. 2012), although given the likely mixing of halo gas components, deep maps of the stellar components of galaxy halos may remain the best method to identify a satellite origin for the gas (e.g., Martínez-Delgado et al. 2010). The Milky Way has a large number of HVCs at z-heights 10 kpc that do not seem to be related to stellar halo features. These clouds may be a combination of satellite gas and halo gas of other origins that is recooling close to the disk.