High-resolution low-energy electron diffraction measurements revealed that the (100) surface underwent weak disordering at about 570K; some 30K below the bulk melting point. This disordering was preceded by an anomalous surface expansion at temperatures below 480K, followed by the sudden generation of a very high density of surface vacancies at 510K. This mechanism contrasted with that operating on the (110) surface, where pre-melting was preceded by a surface roughening transition of Kosterlitz-Thouless type, in which multi-layer steps were involved.

H.N.Yang, K.Fang, G.C.Wang, T.M.Lu: Physical Review B, 1991, 44[3], 1306-10