VOLUME 81, NUMBER 13 PHYSICAL REVIEW LETTERS 28 SEPTEMBER 1998 Hexagonally Reconstructed Islands and Anisotropic Diffusion for AuAu(100) Miki Nomura and Xiao-Qian Wang Department of Physics and Center for Theoretical Studies of Physical Systems, Clark Atlanta University, 223 James P. Brawley Drive, Atlanta, Georgia 30314 (Received 27 January 1998) The homoepitaxial island growth on hexagonally reconstructed Au(100) is studied using molecular dynamics based on a well-tested many-atom interatomic potential. Our study reveals that the stable islands of rectangular shape are hexagonally reconstructed in conformity with the patterns of the reconstructed Au(100) surface and suggests the “magic” stable width for the reconstructed islands in agreement with experimental observations. Furthermore, our results on the adatom diffusion indicate that the experimentally observed strong anisotropic effect is attributed to the long-range exchange diffusion. [S0031-9007(98)07233-0] PACS numbers: 68.35.Fx, 68.10.Jy, 68.35.Bs, 68.55. – a Surface adatoms and small cluster diffusion on metal clusters have been the subject of recent experimental and simulational attention due to their important role played in controlling the formation of surface layers, thin film growth, and catalysis. The nucleation and growth of Au islands on the reconstructed Au(100) are both in- trinsically interesting and technologically important. It is expected that for the island growth, surface mor- phology depends strongly on the reconstructed substrate. Recently, Günther and co-workers [1] carried out a scanning-tunneling microscope (STM) study of the nu- cleation and growth of Au islands on the hexagonally reconstructed Au(100) surface. It was shown that the reconstructed substrate yields strong anisotropic effects, specifically the rectangular shape of the island (island shape anisotropy) and strongly anisotropic diffusion. The latter was deduced from the rate equation analysis of the experimental flux and temperature dependence of the island density. Two typical sizes of the islands, namely, (i) islands of widths of 30 Å and lengths ranging from 100 to 150 Å at high deposition flux rate and (ii) islands of widths of less than 80 Å and lengths around 700 Å at low deposition flux rate, were reported. In spite of this intriguing picture, there has remained a paucity of theoretical studies on the island growth and anisotropic diffusions for AuAu(100). A clear under- standing of the origin of anisotropic effects has been hindered by the lack of quantitative theoretical calcula- tions. Recently, Bonig, Liu, and Metiu [2] employed the effective-medium theory to investigate the diffusion of adatoms for a m-atom-wide 1 3 5reconstructed island on top of unreconstructed underneath. Their calculation indicates that kinetics favors m 6, in that the adatom diffusion along the long side of such a reconstructed island is much faster than m 6. However, due to the limita- tion of the employed interatomic potential (not capable of predicting the reconstructed surface), their interpretation of experiments needs to be reexamined with more realistic models. Inspired by the experimental results, we have carried out a large-scale molecular-dynamics (MD) simulation study of the island growth and anisotropic diffusion on AuAu(100). Using a well-tested many-atom potential of gold constructed by Ercolessi, Tosatti, and Parrinello [3], our simulation study reveals that the optimal island is “quantized” in concord to the “magic” size of 7, 13, . . . (6n 1 1, where n is an integer) reconstructed rows. The formation of a hexagonally reconstructed island in confor- mity with the pattern of the reconstructed Au(100) surface is energetically favored. Contrary to the proposed hop- ping mechanism for adatom diffusion [2], our calculation reveals that the long-range exchange effect plays a pre- dominant role in the anisotropic diffusion and the growth of hexagonally reconstructed islands. The semiempirical many-body “glue model” [3] be- longs to the same class as the embedded-atom method [4] and “pair-functional” models [5]. The glue potential for gold has been well tested and is known to provide good results for diverse surface properties. Specifically, the model is capable of explaining the structure and phases of the reconstructed Au(100) surface. It is worthwhile to mention that the simulation study of M 3 5reconstruc- tion configuration (i.e., M 1 1 first-layer atoms on top of M second-layer atoms along [011], in addition to the 6- onto-5 registry along 01 ¯ 1), reveals that the energies are very close for the range 20 , M , 35, with a surface en- ergy difference of 0.1 meVÅ 2 [3,6]. In carrying out our analysis, we have performed in- tensive molecular-dynamics simulations, with in-plane periodic boundary conditions, on 16-layer slab systems consisting of 2 53 10 4 atoms. The experimentally observed 28 3 5structure [7,8] is used for the surface throughout this study. The structure can be characterized by edge dislocations along [011] and 01 ¯ 1, respectively, with the dislocation lines corresponding to the corrugated regions in connection with the excessive stress. It is worth noting that the corrugated lines can be identified as parallel stripes observed in STM experiments. 0031-900798 81(13) 2739(4)$15.00 © 1998 The American Physical Society 2739