3.1. Transverse Absorption-Line Studies
Viewing the CGM in absorption against a bright background source like a quasar offers three major advantages over other methods: (1) sensitivity to extremely low column density, N ≃ 1012 cm−2, (2) access to a wide range of densities, unlike emission-line measures that scale as density squared, and (3) invariance of detection limits to redshift and the luminosity of the host galaxy. These advantages come at a cost, however: absorption provides only projected, pencil-beam measures of gas surface density, usually limited to one sightline per galaxy by the rarity of background quasars. Within the local Universe (a few Mpc) it is possible to use multiple sightlines (e.g., Lehner, Howk & Wakker, 2015, Bowen et al., 2016), and at higher redshift, multiply-lensed images from background quasars (e.g., Rauch & Haehnelt, 2011, Rubin et al., 2015) to constrain the sizes of absorbers. In general, however, CGM maps made from absorption-line measurements are a statistical sampling of gas aggregated from many galaxies. With massive optical spectroscopic surveys, samples have grown to hundreds or thousands in low ions like Mg ii and Ca ii (e.g., Zhu & Ménard, 2013a). Quasar/galaxy pairings have now been extended out to z ∼ 4 and beyond (Turner et al., 2014, Matejek & Simcoe, 2012).
There are three basic ways of building absorber samples. First, “blind” surveys select background quasars on brightness and/or redshift and so are optimal for samples that are unbiased with respect to foreground structure. Ground-based redshift surveys around previously observed quasar sightlines are now a time-honored method for constructing samples of quasar/galaxy pairs (e.g., Chen et al., 1998, Stocke et al., 2006, Rudie et al., 2012). The second, “targeted”, approach chooses background sources because they probe particular foreground structures, such as L* galaxies (Tumlinson et al., 2013), sub-L* galaxies (Bordoloi et al., 2014b), galaxies with known ISM content (Borthakur et al., 2015), or groups and filaments (Wakker et al., 2015, Tejos et al., 2016), by cross-matching the observable quasar with catalogs of these structures. Finally, maps of absorption in the Milky Way's CGM use essentially any quasar (or UV-bright halo stars), sometimes chosen to pass through known halo gas structures and sometimes not. Though most absorption-line work has been in the UV and optical, Chandra and XMM have been used to search for X-ray gas in individual absorbers, constraining the extent of CGM and IGM hot gas (Nicastro et al., 2005).
| LLS : Lyman Limit Systems, NHI > 1016.2 cm−2, the “dense” CGM |
| DLA : Damped Lyman-α Systems, NHI > 2 × 1020 cm−2, generally ISM |
It is useful to distinguish between H i column density regimes that must be, or can be, treated differently in analysis. Lines up to logN ≃ 15 can usually be analyzed with equivalent widths or Voigt profile fitting. The value logN ≃ 15 is high for the Lyα forest but low for the CGM (there are of course a few exceptions, Tumlinson et al. (2013), Johnson et al. (2014), where H i is not seen at < 100 kpc even to low limits). At logN ≃ 16, saturation becomes a major factor and robust column densities (as opposed to lower limits) must come from profile fitting or from the higher Lyman series lines, if the system is redshifted enough. Systems with logN ≃ 16, are partial or complete LLSs. If the Lyman limit is covered (z > 0.24 for Hubble), the flux decrement at λ = 912 (1 + z) Å allows a precise measurement of logNHI and improved ionization and metallicity diagnostics. Above logNHI ≃ 18 (where NHI is the HI column density in cm−2), the Lyman limit is totally opaque, the highest Lyman series lines are saturated, and genuine column densities must come from fitting the Lyα profile for LLS and DLAs.
Massive spectroscopic surveys have enabled another novel method for examining halo gas. “Stacking” of hundreds or thousands of spectra is a powerful way to extract faint signals from absorption-line datasets. This technique requires catalogs of redshifts, for either foreground galaxies or absorbers, so that the spectra of background objects can be shifted to their rest frames and continuum-normalized and then co-added together. The co-addition beats down statistical noise, enabling measurements of weak absorption at the cost of averaging over individual absorber profiles. When the catalogs of foreground galaxies include properties such as mass, radius, star formation rate, color, environment, or orientation, the stacks can be performed with subsets of the data to examine the variation of mean profiles with these properties (York et al., 2006, Zhu & Ménard, 2013b, Bordoloi et al., 2011). Stacking experiments that correlate the reddening of quasars due to foreground galaxy halos in the SDSS survey have revealed large quantities of dust in the CGM of galaxies (Ménard et al., 2010, Peek, Ménard & Corrales, 2015). Stacking techniques can also exploit more numerous, but fainter, sources; for example, Steidel et al. (2010) characterized the CGM of z ∼ 3 galaxies by stacking the spectra of background galaxies. Stacking can detect weak signals in the mean properties of gas absorbers, but at the cost of averaging out kinematic and ionization structure that may contain significant physical meaning.
“Down-the-barrel” spectroscopy uses a galaxy's own starlight as a background source for detecting absorption. This method has been a fruitful one for studying galactic inflows and outflows from spectroscopy of star-forming galaxies. This method is commonly used in optical and near-UV lines such as Ca ii, Na i, Mg ii, and Fe ii (Martin, 2005, Kornei et al., 2012, Bordoloi et al., 2011, Rubin et al., 2014) to study outflows from galaxies out to z ∼ 1, in UV lines for low-redshift star-forming galaxies (Henry et al., 2015, Heckman et al., 2015), or even to examine accretion (Rubin et al., 2012). Down-the-barrel measurements are critical pieces of the CGM puzzle because they directly trace current outflows at galactocentric radii that are inefficiently covered by background sources (because of the R2 scaling of foreground cross-section). While down-the-barrel spectra are key for tracing the accretion and outflows that dominate CGM kinematics, they have the key limitation that the galactocentric radius of any detected absorption is unconstrained — it could be anywhere along the line of sight — complicating mass and covering fraction estimates inferred from these spectra.
Emission-line observations search for photons emitted directly from CGM gas. As the emission measure scales as n2, and the CGM has nh ∼ 10−2 or less, this is a stiff challenge. The MW halo has been extensively mapped for HVCs and other halo structure using radio emission at 21 cm. This technique has been applied to external galaxies (Putman, Peek & Joung, 2012b) but detections are limited to within ∼ 10−20 kpc of the targeted galaxies. The soft X-ray band is optimal for gas at ≳ 1 million K. The extremely low surface brightness of the gas makes these observations challenging and expensive, but a few individual halos have been detected and their hot gas budgets measured by Chandra and/or ROSAT (e.g., Humphrey et al., 2011, Anderson, Churazov & Bregman, 2016). Stacking of individual galaxies techniques has also yielded mass density profiles for hot gas around nearby galaxies (Anderson, Bregman & Dai 2013). When combined with halo size, density, and metallicity constraints from soft X-ray absorption line techniques, these maps have aided in the assessment of the total mass and baryon fraction of the hot CGM.
Emission line maps are also possible at UV/optical wavelengths, though no less challenging than in the X-ray. Recent reports claim a detection of an extended O vi halo (R ∼ 20 kpc) around a low-redshift starburst galaxy (Hayes et al., 2016). Extended Lyα emission has been seen out to ∼ 100 kpc away from z ∼ 2.5 galaxies and QSOs (Cantalupo et al., 2014, Prescott, Martin & Dey, 2015). In another case, an extended filamentary structure connected to a galactic disk was detected using diffuse emission in the optical (Martin et al., 2015). Emission maps can constrain the density profile, morphology, and physical extent of the gas more directly than aggregated pencil-beam sightlines (Corlies & Schiminovich, 2016). For X-ray emission from fully ionized gas, masses can be inferred more directly, avoiding the uncertain ionization corrections that plague absorption-line measurements (Section 4); indeed, the CGM's more massive cousin, galaxy clusters' intracluster medium, has been studied in detail via X-ray emission for decades (Vikhlinin et al., 2006). On the downside, emission line maps are still challenging technically; the surface brightnesses are extremely small compared to sky and detector backgrounds, and surface brightness dimming has a steep increase with redshift. In a recent study using stacks of fiber spectra from SDSS, Zhang et al. (2016) achieved detections of Hα at 50−100 kpc around low-redshift galaxies, demonstrating that very sensitive limits can be reached on galaxies in the aggregate. These observations remain challenging, but as “taking a picture” of an astrophysical object remains the ideal, efforts to improve instrument technology and enable emission line mapping to reach samples of hundreds of galaxies across cosmic time is an important goal.
Physical models and simulations are essential tools for understanding the CGM. In contrast to observations, they provide for controlled environments where physical properties, histories, and futures of gas are all known and can be manipulated to tease insights out of the otherwise unmanageable complexity of a multiphase gaseous medium. As reviewed by Somerville, Popping & Trager (2015), there are many schemes for simulating the development of the cosmic web and galaxies under the influence of dark matter, gravity, and hydrodynamics. The major methods at present are smoothed particle hydrodynamics (SPH, such as Gadget, Ford et al., 2013, Oppenheimer et al., 2016b, Gasoline, Christensen et al., 2016, Gutcke et al., 2017, and GIZMO Muratov et al., 2016), adaptive mesh refinement (AMR, such as Enzo, Hummels et al., 2013, Corlies & Schiminovich, 2016), and moving mesh (Arepo and the Illustris simulation, Suresh et al., 2015). Large-scale cosmological simulations in Mpc-scale boxes can simulate hundreds of galaxies in their proper ΛCDM context (e.g., Oppenheimer & Davé, 2006, Vogelsberger et al., 2014, Ford et al., 2014). At the opposite end of the scale, very high resolution simulations focused on the interaction between dense clouds and diffuse halos (e.g., Heitsch & Putman, 2009, Armillotta et al., 2016) that can reach scales at ≪ parsec. Spanning these two regimes are the so-called “zoom” simulations, which resolve enough of the large scale structure to accurately trace a single galaxy or a subset of galaxies selected out of larger boxes (Figure 2, Schaye et al., 2015). Even zooms must make assumptions about physics that they do not resolve, using “sub-grid” prescriptions to stand in for such complex phenomena as star formation, metal mixing and transport, supernova and AGN feedback, and others. Sub-grid models are parameterized and tuned to yield specific metrics — like the stellar mass function at z = 0 — and then the properties that emerge — such as SFRs, morphology, quenching, and the CGM — are analyzed and compared to data to constrain the physical prescriptions that went in. We will use simulations from a broad range of techniques and groups to look for insights into how the CGM participates in galaxy evolution, and to help interpret data.