Right at the freezing point, both the liquid and the solid are stable.
They're in equilibrium with each other, meaning that the free-energy
per molecule is the same. At any other temperature, one or the other
has lower free energy, and is more stable.
In addition to the issue of true long-term stability, there's the
question of stability for practical purposes. Pure water cooled a
little below the freezing point won't actually freeze any time soon.
The reason is that a nucleus of ice has to form before the freezing
really gets going. Just below the freezing point, those ice nuclei
don't become more stable than the liquid until they have a lot of
molecules in them. Structures like that happen only very rarely by
chance, so it takes a long time for the freezing to occur. A liquid
that is still liquid even though its temperature is below its freezing
point is called "supercooled".
Even if there is a little crystal of ice present to get things
started, and even if the temperature is a tiny bit below the freezing
point, water won't 'immediately turn to ice'. The reason is that as it
turns to ice, heat is released. That keeps the nearby liquid from
freezing until enough time has passed for the heat to diffuse away and
the temperature to drop down to a low enough value. In practice, of
course, this is not really a start-and-stop process but rather a slow
steady process limited by the rate a which heat diffuses away.
Going the other way, ice melts without a nucleation delay when
heated above the freezing point. The reason is said to be a layer of
liquid-like molecules on the surface, which serve as a nucleus for the
liquid phase. However, melting soaks up heat, so the process still
proceeds at a rate limited by how fast heat flows in.
(republished on 07/25/06)