High-resolution scanning tunnelling microscopic data on reconstructed (111) surfaces were presented. These gave a comprehensive picture of the atomic structure, the long-range ordering, and the interaction between reconstruction and surface defects on the reconstructed surface. On the basis of the atomically resolved structure, a stacking-fault domain model which involved periodic transitions from a face-centered cubic to an hexagonal close-packed stacking of the top-layer atoms was confirmed. An essentially uniform contraction in the surface layer along [1¯10] indicated that previously proposed soliton theories were not appropriate for the face-centered cubic to hexagonal close-packed stacking transition. The lateral displacement of about 0.09nm in the (22-102) unit cell, along [11¯2], was in good agreement with the transition between face-centered cubic and hexagonal close-packed stackings. The vertical displacement (0.02nm) in the transition regions was largely independent of the tunnelling parameters, while the atomic corrugation (average: 0.02nm, maximum: 0.1nm) depended markedly upon the tunnelling parameters and tip conditions. The 2 different stacking regions within the unit cell were deduced directly from the domain pattern at step edges. Thus, face-centered cubic stacking was deduced for wider areas and were energetically more favorable. A new long-range superstructure was reported. It was created by a correlated periodic bending, of the parallel corrugation lines, by ±120º every 25nm. That is, rotational domains were arranged in a zig-zag pattern. Interactions on this scale suggested the existence of a long-range elastic lattice strain. This structure reflected an overall tendency to isotropic contraction, and combined a locally favorable uniaxial contraction with an effective isotropic contraction at a larger scale. The boundaries of rotational domains could also form by termination of the reconstruction lines. Individual corrugation lines, separating differing stacking regions, could not disappear. The termination occurred in well-ordered U-shaped connections of neighboring lines, or via a complicated pattern of entangled corrugation lines. Steps and bulk defects did not inhibit reconstruction, but could affect the local reconstruction pattern. In most cases, steps were crossed by the reconstruction lines. A strict correlation of the reconstruction pattern on the terraces, both in phase and orientation, reflected an interaction over the step edge. The reconstruction pattern at the steps sometimes resembled those which were to be found at rotational domain boundaries.

J.V.Barth, H.Brune, G.Ertl, R.J.Behm: Physical Review B, 1990, 42[15], 9307-18