The thin oxide film present on most metals and semiconductors is difficult to detect and, if uncorrected for, can cause serious errors in the ellipsometric determination of optical constants. The reflectance at normal incidence calculated from these incorrect optical constants will, in general, be higher than that measured experimentally for a film-covered sample. A comparison of the approximate, linear, thin-film relations of Drude and Archer with the exact equations shows that the Drude relations are more inaccurate than are usually reported; in some cases serious errors result for films as thin as 10 Å. Archer’s relations are more accurate than Drude’s in most cases. However, if either of these sets of relations is used to correct measured ellipsometric values to the no-film condition, considerable errors in determining the optical constants may occur even if the film thickness is known. The variation in the ellipsometric parameters ψ and Δ with angle of incidence cannot be used to determine the presence of a thin surface film, since calculations show that the variation of these parameters with angle of incidence for a film-covered surface is practically identical with that for a fictitious film-free surface with appropriate optical constants. Also, the requirement that Rs2= Rp at 45° angle of incidence is only marginally useful in detecting a surface film.
D. K. BURGE and H. E. BENNETT, "Effect of a Thin Surface Film on the Ellipsometric Determination of Optical Constants," J. Opt. Soc. Am. 54, 1428-1433 (1964)