Molecular-dynamics simulations were made of self-diffusion on (110) surfaces. The metals were modelled for both (110) and (100) geometries by using semi-empirical potentials which were developed within the framework of the second-moment approximation to the tight-binding model. A large number of high-temperature simulations were performed for the (110)-(1 x 1) surface. The energy barriers to the relevant diffusion processes were calculated by means of quenched molecular dynamics, and were compared with available data. This revealed good agreement. The occurrence of long jumps was investigated in detail, and this showed that the various metals behaved quite differently. That is, long jumps were essentially absent in Au, but frequent in Cu. It was found that correlated jump-exchange processes and double exchanges were common in Cu; even at quite low temperatures. They were never observed in Au, while Ag exhibited an intermediate behavior. These differences were attributed to 2 factors. Firstly, the dissipation of the energy of the adatom to the substrate was larger in Au than in Cu and Ag. More importantly, the potential-energy surface at the saddle point was very narrow in Au but wide in Cu. The results demonstrated the sensitivity of long jumps to the details of the interaction between adatom and substrate, and suggested that the use of one-dimensional diffusion models might also be insufficient for face-centered cubic (110) metal surfaces. Another important difference between the 3 metals concerned correlated cross-channel processes. Again, these were more frequent in Cu than in Ag and were absent in Au. Such correlated processes were likely when there was a significant probability of making long jumps and when the energy barrier to cross-channel diffusion was not large. Moreover, their frequency rose sharply at high temperatures, where the amplitude of the lateral vibrations of the row atoms became large. Another important point was the temperature dependence of the jump rate. A huge number of simulations was performed for Cu at 8 temperatures between 300 and 600K. It was found that it was impossible to obtain a good fit to the data by using a simple Arrhenius law, with a temperature-independent activation energy and pre-factor, for the entire temperature range. However, a fit which was restricted to temperatures of up to 400K was excellent and the estimated activation barrier was very close to the static energy barrier which was deduced from quenching. A fit which was restricted to the highest temperatures gave a lower activation energy and a smaller pre-factor. This reduction was attributed to a finite-barrier effect, and perhaps to the influence of lattice dynamics upon diffusion.

Jumps and Concerted Moves in Cu, Ag and Au(110) Adatom Self-Diffusion. F.Montalenti, R.Ferrando: Physical Review B, 1999, 59[8], 5881-91