3.1. Introduction: atomic and molecular interstellar gas
Studying interstellar gas at galactic scales is essential to understand the structure and dynamics of galaxies, and the way they form stars. Except for the impeding factor of interstellar dust extinction, the bulk of the insterstellar medium is not easily accessible to optical astronomy because of its lack of emission at low temperature, the lack of adequate optical absorption bands and the blurring of absorption lines by large dust extinction. Its exploration has thus relied on longer wavelengths, essentially radio astronomy. First, since its discovery (Ewen & Purcell 1951, Muller & Oort 1951, Pawseyl 1951), the 21-cm line of atomic hydrogen, HI, has shown the perfect tool for tracing the atomic gas, even in the mainly molecular medium. The ubiquity of HI which is generally the most massive component of the ISM in galaxies, the sensitivity of the 21-cm line, its lack of interstellar and atmospheric important absorption, together with relatively easy technical developments, have made HI 21-cm studies one of the main drivers of the spectacular development of radio astronomy. Its multiple achievements are invaluable for tracing the interstellar gas and its various properties: amount and distribution in the Milky Way and various types of galaxies; kinematics tracing rotation and global masses including dark matter; details of structures and gravitational potentials including bars, spiral structure, warps; major gas complexes driving massive star formation; perturbed dynamics in galaxy interactions; exchanges with the intergalactic medium in inflows and outflows; magnetic field; etc (see e.g. Hartmann & Burton 1994 and Kalberla et al. 2005 and references therein for reviews and data about HI in the Milky Way, and Rogstad et al. 1973, Scodeggio & Gavazzi 1993 and Hoffman et al. 1996 in galaxies).
Molecular gas is the next important component of the interstellar medium. As it traces the densest parts of the gas, its importance increases on average with significant concentrations of gas, especially in the disks of spiral galaxies such as the Milky Way. Immediately after the discovery of interstellar CO (Wilson et al. 1970), it became clear that molecular gas is ubiquitous in the Galactic disk, and it was rapidly identified in external galaxies (Rickard et al. 1975). This has completely renewed our views about star formation in galaxies. The amount, distribution and properties of molecular gas are thus essential for understanding the present star formation and evolution of galaxies. The distribution of the molecular gas in the Milky Way, mostly within giant molecular clouds, has been established by CO surveys of the Galactic disk. It may serve as a benchmark to compare with molecular gas in other spiral galaxies which display a quite significant range of molecular material. Star formation is the key issue related to the molecular medium. Its global rate eventually relies much on feeding by extra-galactic gas, and it may be greatly enhanced by galaxy interaction and merging, leading to violent starbursts (Section 5). The amount of molecular gas is generally smaller in galaxies other than spirals. It is also more or less correlated to the total amount of gas and star formation activity. The molecular gas has special properties in central regions of galaxies, where it may display high concentrations and strong starbursts, and may have a close relationship with bars and eventually the activity of the nucleus.
3.2. Molecular gas in the Milky Way at galactic scale
While H2 constitutes at least 99% of the molecular gas, its lack of permanent dipole moment and the cold temperature, well below the excited energy levels, make it almost impossible to directly observe in most of the molecular medium (see Section 10). Therefore, the observation of the molecular gas essentially relies on other molecules and especially on CO (indeed most often the main isotope variety 12C16O) which proves to be readily observable even in quite tenuous molecular gas. CO lines, and in particular the J = 1-0 transition at 115 GHz, are thus somewhat equivalent to the 21-cm line for the study of the molecular interstellar gas. CO is practically always the easiest molecule to detect in all molecular clouds of the Milky Way, as well as in any galaxy at any redshift, and it is systematically used to estimate masses in molecular gas.
We have a good view of the overall amount of CO, the remarkable statistical properties of molecular clouds and their distribution in the Milky Way, from the various Galactic 12CO and 13CO surveys (see e.g. Dame et al. 1987, 2001, Sanders, Solomon & Scoville 1984, Clemens et al. 1988, and other early references in Combes 1991, and papers in Clemens et al. 2004 for recent surveys including BU-FCRAO and NANTEN [e.g. Tachihara et al. 2002]).
Most of the molecular mass ( 90%) appears to be in massive structures, distributed in clumps, the giant molecular clouds (GMC) with diameter ~ 50 pc, masses ~ 105-6 M and densities ~ 102 cm-3 (see e.g. Scoville & Solomon 1974, Williams et al. 2000, Evans 1999 and references therein). A striking property is the relations between the internal linewidth of the clouds, their size R and their mass M (Larson 1981), namely Rs and M Rm, with s close to 0.5 and m close to two. They imply that the GMCs appear to have a constant average surface density and are in some way close to virial equilibrium. However, Larson's relations are indeed valid over six orders of magnitude, covering a much broader range than GMCs, and they are reminiscent of the Kolmogorov scaling law and suggestive of a fractal medium. Similar properties are revealed in the power spectrum of widespread HI emission, and at many, maybe most, scales, cloud boundaries are likely to be dynamic and transient. It is today well agreed that the ISM and its structure are dominated at all scales by a complex variety of turbulent processes combining self-gravity, stellar pressures and magnetic fields. Understanding the turbulent ISM is constantly progressing from observations and simulations (see e.g. references in Williams et al. 2000, Falgarone et al. 2005a, and Elmegreen & Scalo 2004 whose conclusions give a good idea of the great difficulties which have to be overcome to properly understand interstellar turbulence).
Because of the large isotope abundance ratio 12CO / 13CO 60-70, but only an average factor ~ 5 for the line ratio, 12CO and 13CO surveys are complementary to map the Galactic molecular gas. As 12CO lines are completely saturated in the densest parts of molecular clouds, observations of 13CO are better able to trace such regions and to provide detailed cloud structures in surveys such as BU-FCRAO (Jackson et al. 2006) and AST/RO (Martin et al. 2004), although in many cases 13CO proves to be still optically thick and C18O should be better. However, the higher optical thickness of the 12CO(1-0) line and the low gas density needed to excite this line are better matched to trace the large-scale envelopes containing most of the mass of the clouds. Therefore, the most comprehensive molecular survey of the Milky Way, practically complete for Galactic CO in GMCs, was carried out in this line during more than two decades by the two 1.2 m telescopes of the CFA group (Dame et al. 2001). This survey provides information on individual molecular clouds in most regions and displays the main structural features of the molecular Galaxy. As known since the early surveys (e.g. Burton & Gordon 1978 and Sanders et al. 1984 and references therein), most of the molecular mass is concentrated in the region of Galactic radius between 3 and 7 kpc, known as the `5 kpc Molecular Ring' (Scoville & Solomon 1975). The origin of such a distribution, which is indeed exceptional among other galaxies (Section 4.1), might be related to the Galactic bar (Combes 1991) and its relation to the Galactic inner bulge (Blitz & Spergel 1991). Except for a few GMCs at high Galactic latitude, most of the H2 mass is concentrated within a couple of degrees from the Galactic plane. The average H2 column density decreases by a factor ~ 30 from |b| = 5° to 30° (Dame et al. 2001).
The 12CO lines are generally optically thick, and so care must be placed in interpreting the quantities derived. However, it has been empirically proven that the velocity-integrated CO(1-0) intensity, ICO = T*bdv, is not only the most sensitive qualitative tracer of the molecular gas, but also a good quantitative tracer, providing the column density NH2 of H2 to a factor of a few. The conversion `X-factor'
may be determined for CO emission in the solar neighbourhood by different ways: correlation with diffuse gamma-ray emission, with far-infrared plus 21-cm surveys, optical or X-ray extinction (Combes 1991, Dame et al. 2001, Lequeux 2005). Although X varies by a factor of a few, in particular with the galactic latitude b, it is remarkable that this variation is not larger. Indeed, it shows little systematic variation from the value of
for large clouds out of the Galactic plane (|b| > 5°) (Dame et al. 2001). See also the very good concordance with other derivations of X from rays for local clouds 1.74 ± 0.03 (Grenier et al. 2005), or Galactic averages 1.9 ± 0.2 (Strong et al. 2004) and
(Hunter et al. 1997) which is probably the best average Milky Way estimate from ray emission.
It is difficult to analyse in detail the origin of such a tight correlation between the intensity of 12CO emission and the mass of molecular gas since the CO line intensity results from a complex combination of the CO abundance, its excitation, and radiative transfer with random cloud size and velocity distribution (see e.g. Scoville & Solomon 1974). However, a qualitative insight may be gained by stressing that (see e.g. Solomon et al. 1987, Maloney & Black 1988): 1) for 12CO Galactic and extragalactic surveys, telescope beams encompass an ensemble of quasi-virialised molecular clumps obeying the general scaling relations; 2) most of their mass is located out of their cores in regions optically thin in the CO(1-0) line, and there is at most one clump on a line of sight emitting at given radial velocity; such a low filling factor avoids saturation effects when one increases the clump density; 3) for abundances close to solar, the photodissociation boundary of CO is close to that of H2, and a nearly constant fraction of the cooling emission occurs in the CO(1-0) line, so that for each clump the CO emission is proportional to the H2 mass if the heating by cosmic rays remains standard.
The total derived mass of H2 has the same uncertainty as the X-factor. For the Milky Way, it could be ~ 1.0 × 109 M, excluding He, (Blitz 1996), which is significantly smaller than the HI mass, ~ 5 × 109 M (Wouterloot et al. 1990).
While most of the molecular mass in the Milky Way and other galaxies is accounted for by molecular clouds and especially GMCs, it is important to remind that H2 and other molecules are also found in a number of other Galactic environments in amounts much smaller, but quite significant for the variety and the importance of the media implied, and their wide spread nature at Galactic scales. They include atmospheres of cold stars, giants and dwarfs, and brown dwarfs; AGB envelopes and planetary nebulae; planetary atmospheres; supernova remnants; various stellar outflows, extragalactic outflows; etc. The most relevant case for a general description of the interstellar medium is that of the molecules of the diffuse `atomic' gas, i.e. of all the HI regions. Despite the fact that molecules are rapidly photodissociated by the ambient UV radiation, reformation of H2 on grains maintains a certain abundance of H2, which initiates the formation of small amounts of other simple molecules through cosmic ray reactions and photochemistry. As the diffuse gas is not UV optically thick, absorption spectroscopy of strong UV lines is the main tool for studying first H2 and a few other molecules such as CO and OH, provided there are lines of sight with strong background UV sources such as hot stars or quasars. Let us also recall the importance of molecules of photodissociation regions at the boundary of the molecular and atomic media, and of the very large aromatic molecules (PAHs) which pervade the photodissociation regions and the whole diffuse medium (Section 8).
It has been proposed that in addition to the large mass of H2 in regular molecular clouds traced by CO millimetre emission, an even larger mass of H2 might be hidden in the shape of extremely cold H2 in the outer Galactic disk (Pfenniger, Combes & Martinet 1994, Pfenniger & Combes 1994, Wardle & Walker 1999, Pfenniger 2004, Combes 2006, Bell et al. 2006). It has even been considered that such cold H2 could explain part or the totality of the Galactic dark matter implied by the Galactic rotation curve, especially in the outer disk. There are interesting arguments in favour of such an assumption. They include: an apparent excess of very cold dust with respect of the amount of HI traced by the 21 cm line and H2 traced by CO emission (e.g. Cambresy et al. 2001); a similar excess in ray emission with respect to HI + H2 (Dixon et al. 1998, Grenier et al. 2005) [but the lacking gas needed to match cold dust or rays could as well be in the shape of additional cold HI gas difficult to measure with precision]; the interest of a large reservoir of gas to explain the evolution of spiral galaxies; etc. However, direct evidence of such very cold H2 is still lacking. Its wide distribution is ruled out by the lack of significant detection by absorption of H2 UV lines, despite the number of lines of sight studied, in particular by FUSE. The only possibility would be that H2 is condensed in very small globules of mass and core radius maybe as small as the Earth (Pfenniger 2004), or more diffuse gaseous H2 with ~ 10-3 M and ~ 10 AU (Rafikov & Draine 2001). Such globules should be very difficult to detect because of their very small filling factor (~ 1%) and their opacity to UV radiation. The lack of absorption evidence in the search for microlensing events in the direction of the Large Magellanic Cloud constrains the properties of the proposed globules (Alcock et al. 2000, Lasserre et al. 2000) constrains the properties of the proposed globules, but does not completely rule them out (Rafikov & Draine 2001). Clumps of ionized and HI interstellar gas have been observed in radioastronomy in `Extreme Scattering Events' in front of radio quasars, and in VLBI observations of the 21 cm absorption line. They have the right size and could be ionized and atomic envelopes of such molecular globules immersed in the interstellar UV radiation; but the implied individual atomic or ionized gas masses have no common measure with the much higher masses invoked in self-gravitating H2 globules.
To summarise, apart from elusive cold H2 globules, we have a very good view of the properties and distribution of molecular gas in the Milky Way from CO surveys. The bulk of this gas is distributed in GMCs along the thin Galactic disk, where most of young stars form. Despite the complex structure of GMCs, the integrated intensity of the CO(1-0) line provides the most widely used quantitative estimate of the amount of H2 through the `H2 / CO conversion factor'. However, the physical origin of this tight empirical relation turns out to be complex. Besides, complementary 13CO surveys are needed to trace the densest parts of molecular clouds.