If you ask any group of school children what temperature does pure water freeze at and you will normally get the answer 0 degrees celsius, which is the standard answer and it is the temperature below which ice is more stable than water, but that isn't the whole story.
It is quite easy to get liquid water below this temperature by just putting a bottle of very clean water in a freezer, because although a large ice crystal is more stable than liquid water, a very small one isn't so most of them shrink and melt before they get big enough to be stable, but how cold can you get liquid water?
Valeria Molinero and Emily Moore from the University of Utah decided to do away with experiments and try and solve the problem in a computer. This isnít easy, as forming ice intrinsically involves a lot of water molecules, and to get a meaningful result they had to model over 30 thousand water molecules interacting with one another, and they had to model water molecules as a single lump rather than as three atoms..
They found that as the water cooled, more and more of the water formed tetrahederal structures which were somewhere between ice and liquid water, in both structure and density which they call intermediate ice. This can then either convert to ice proper, or into a disordered glasslike structure when it finally froze.
After all this work they finally found the theoretical lowest temperature you can cool water to without freezing is -55 celsius.
This might seem quite academic, but supercooled water is very important in many types of cloud, and has been discovered there at -40 celcius, and understanding how water behaves at these temperatures will help understand clouds and so assist with weather and climate predictions.