Abstract
We consider the nucleation of a crystal phase, on a crystalline surface of a different substance. Sixty years ago, Turnbull and Vonnegut predicted that a crystalline surface is best at inducing nucleation of another crystal when there is a perfect epitaxial match between the two bulk lattices. We use computer simulation to show that this is not quite right. In fact, the crystal lattice of a finite nucleus is strained from that in the bulk, and nucleation is fastest when the surface matches this strained lattice. We also find that the approach of Hillier and Ward predicts when nucleation is epitaxial. © Copyright EPLA, 2014.