3.1. Radial Distributions
The differences between HI and H2 (or CO) radial distributions in galaxies is striking (cf figure 2). While all components related to star formation, the blue luminosity from stars, the H (gas ionised by young stars), the radio-continuum (synchrotron related to supernovae), and even the CO distribution, follow an exponential distribution, the HI gas alone is extending much beyond the ``optical'' disk, sometines in average by a factor 2 to 4 (RHI = 2-4 Ropt). The HI gas has very often a small deficiency in the center. Would this mean that the atomic gas is transformed in molecular phase in the denser central parts? This is possible in some galaxies, where the HI and CO distribution appear complementary, but it is not the general case, all possibilities have been observed, including a central gaseous hole, both in CO and HI (like in M31, for example).
Figure 2. Radial distributions of various surface densities in a typical spiral galaxy NGC 6946: H2(CO) and HI column densities, Blue, Radio-continuum and H surface densities (adapted from Tacconi & Young, 1986).
3.2. Large-scale Structure
Within the optical disk, where CO is observed easily, there is a very good large-scale correlation between both gas components (see e.g. Neininger et al. 1998). They appear well mixed, and follow the spiral ams with large contrast. This is also true for the ionised gas (HII regions). Of course, this is only at 100pc-1kpc scale; at very small scale the various components can be anti-correlated, the HI gas being found more at the envelopes of dense molecular clouds, the ionised gas being also anti-correlated with the neutral gas.
3.3. Vertical Structure
In our own Galaxy, and in external galaxies seen edge-on, the galaxy disks appear much narrower in CO emission than in HI. This suggests that the molecular gas is more confined to the plane, and that its vertical dynamical oscillations are of less amplitude than for the atomic gas. The consequence should be a vertical velocity dispersion much lower for the molecular gas, since for a given restoring force due to the stellar disk, the maximum height above the plane is proportional to the z-velocity dispersion. Surprisingly, this is not the case: in face-on galaxies both CO (Combes & Becquaert 1997) and HI (Kamphuis 1992) velocity dispersions are observed of similar values (v ~ 6 km/s), and remarkably constant with radius. This is not a saturation effect of the CO lines, since the 13CO spectra show the same. A possible interpretation is that both gas are well mixed, in fact it is the same dynamical component, which changes phase along its vertical oscillations. It is possible that the H2 gas follows the HI, but the CO is photo-dissociated at high altitudes, or not excited. Or even the H2 could disappear, since the chemistry time-scale (~ 105 yr) is much smaller than the dynamical z-time-scale (~ 108 yr).
3.4. Small scale structure of clouds
The molecular component is also characterized by its remarkable self-similar structure, a hierarchical system of clouds, tightly related to a fractal structure. It can be quantified by power-law relations between cloud size and linewidth, or size and mass (Larson, 1981). These relations are observed, whatever the radial distance to the center, and in the HI component as well. The fact that the same structure is observed outside of the star-forming regions is puzzling: the HI gas outside the optical disk displays a very clumpy structure, implying that it is unstable at all scales (spiral arms, self-similar structure of clumps). The fact that these gravitational instabilities do not trigger star formation must be explained.