Tracer and chemical surface diffusion coefficients for oxygen chemisor
bed on W(110) have been investigated by means of Monte Carlo modeling,
taking full advantage of the numerical power of a supermassive parall
el supercomputer. The simulations were performed for a lattice gas wit
h up to fourth nearest neighbor interactions. These were chosen to rep
roduce the experimentally determined O/W(110) phase diagram for theta
less than or equal to 0.5 which shows a dominating p(2 x 1) phase arou
nd half coverage. It was found that the p(2 x 1) ordering strongly inf
luences surface diffusion at low temperatures. The tracer diffusion co
efficient shows a strong minimum slightly below half coverage and a ve
ry small maximum above half coverage. This behavior is attributed to t
he change of the dominating lattice defects of the p(2 x 1) phase from
vacancies in filled rows below theta = 0.5 to interstitials in empty
rows above theta = 0.5. For the p(2 x 1) ordered lattice gas phase the
chemical diffusion coefficient exhibits a strong maximum which become
s more pronounced as the temperature is lowered. This is attributed to
the behavior of mean square fluctuations <(delta N)(2) >/< N >, i.e.
the inverse of the thermodynamic factor, which dominates the chemical
diffusion coefficient under these circumstances. The p(2 x 1) ordering
causes the surface diffusion to become highly anisotropic. Attempts t
o simulate the increase in activation energy with coverage found exper
imentally showed that this was not possible without invoking adsorbate
induced changes in the saddle point energies for diffusion. (C) 1997
Elsevier Science B.V.