Collisional processes

Suppose you have a collision between two H$_2$ molecules. If the collision is sufficiently violent, you can imagine that it will break the molecular bond and you have converted H $_2\rightarrow 2$HI. Similarly, if you collide two hydrogen atoms (or a HI with an electron) with enough speed, it may lead to ionisation, HI+e $\rightarrow$HII+2e. Since the energy required to break the molecular bond (11eV) is lower than to ionise hydrogen(13.6eV), the second collision needs to be more energetic than the first. And so you expect that the higher the temperature of the gas - and hence the more energetic the collisions - the more you will shift from molecular, to atomic, to ionised gas.

How about the reverse process? When an HI atom collides with another HI, or with an electron and gets ionised, then the system has converted kinetic energy into ionisation energy. When the ion recombines, HII + e $\rightarrow$ HI, it can be that the photon escapes from the cloud, in which case there has been a net loss of energy for the gas. Since the particles are now moving slower, the temperature of the gas $T\propto \langle v^2\rangle$ is lower, or the gas has cooled down. This process is called radiative cooling and it is very important for star formation.

So we understand now why colder gas will tend to be molecular, whereas hotter gas is more likely to be atomic or even ionised. Now, if the pressure between these phases is nearly the same, then the colder gas (in which H$_2$ is favoured) will need to be denser than the hotter gas (atomic or, at higher temperature, ionised). And so you expect to find dense, cold clouds to be molecular, and hot, rarefied gas to be ionised - which is indeed what we observe, except that there is one more important process: photo-ionisation.