Supercooling of liquids

I shall concentrate upon reviewing the important recent change in our appreciation of the facts of supercooling which has been brought about particularly by the work of Turnbull at the General Electric Research Laboratory in Schenectady. I suppose that most of us, talking about supercooling a couple of years ago, would have divided substances into two classes, one with simple crystal structures like gold, and all the other ‘good’ metals on the one hand, and those with complex crystal structures, such as glycerol and the silicates on the other; saying that whereas the latter class can be very much supercooled, and will form glasses, the former class can only be supercooled a very few degrees. Then we would have added that there are some ‘ bad ’ metals, with moderately complex crystal structures, such as antimony or bismuth, which can be supercooled some tens of degrees, forming an intermediate class. I think we would then have added that this is quite comprehensible. In particular, that the X-ray diffraction patterns of the monatomic liquids show us that most of the atoms have the right numbers of nearest neighbours in a first co-ordination shell, all ready in place to start the growth of a crystal; which readily explains why these substances cannot be supercooled very much—a nice simple experimental fact, with a straightforward theoretical interpretation—and both are wrong.