We investigate the surface reconstructions of various faces of gold and platinum theoretically with a number of embedded-atom potentials. The potentials examined include conventional models plus a sequence of models where the shape of the embedding function has been varied for electron densities typical of the surface. We find that whereas all models predict virtually equivalent bulk properties, only potentials with a sufficient curvature in the embedding function predict the experimentally observed buckled quasihexagonal reconstruction for the (100) faces. Furthermore, the models that predict this reconstruction also give surface energies in close agreement with experiment, unlike most of the conventional models. Almost all embedded-atom potentials considered predict the observed missing-row structure for the (110) faces, and some the observed missing-row structure for the (311) face of platinum. We take the best combination of Au and Pt potentials and calculate the structure of a Pt(100) substrate in the presence of Au adatoms. In good agreement with experiment, the quasihexagonal reconstruction is lost for coverage greater than 0.4 monolayers.

• Received 17 March 1993

DOI:https://doi.org/10.1103/PhysRevB.48.2611