We review the classical transition state theory (TST) of desorption and employ it to predict the desorption rate kTST for the Xe/Pt(111) system using a realistic gas-surface interaction potential. The Xe/Pt(111) potential surface is derived from a Xe-Pt pair potential with parameters suitably adjusted to give agreement with experimental data on the Xe/Pt(111) vibrational frequency, well corrugation and well depth. The calculated rates agree well with earlier measurements which span 7 orders of magnitude in rate, corresponding to temperatures in the range 80 < T < 160 K. However, we find that the calculated preexponentials vary by a factor of 5 over this range in T, implying that the actual potential well depth for Xe/Pt(111) is 10 meV greater than the energy obtained directly from the experimental Arrhenius plot slope. The effective preexponential given by the 1/T = 0 intercept of an Arrhenius plot of kTST is found to be 1.6 × 1012 s-1, in excellent agreement with the measured value. We then extend this treatment to calculate desorption rates when surface defects are present. Our pair potential is used to calculate the potential in the vicinity of the close-packed step edge chosen as a model defect. This potential and the measured defect site density are used with TST, generalized to include the effect of defects, to predict desorption rates in the defect-dominated regime. The desorption preexponential factor found in this case is > 103 larger than the value describing the ideal Pt (111) surface, consistent with the striking increase found experimentally, while the well depth obtained at step sites is 355 meV, 100 meV deeper than for Xe on terraces, again in reasonable agreement with experiment. © 1990 American Institute of Physics.