Abstract
The existence of a surface density threshold for star formation is usually explained in terms of the Toomre criterion for gravitational instability in a thin, rotating disk. Here it is shown that it is not gravitational instability that causes the formation of a cold, molecular phase, but vice versa: the transition from the warm ( T ~ 10^4 K) to the cold ( T < 10^3 K) interstellar phase causes the disk to become gravitationally unstable. Rotation does not affect the critical surface density at which the phase transition occurs and cannot stabilize the cold phase in the outer disk.
