Abstract

The method we have been developing for many years to reconstruct the profiles of statistically rough surfaces (based on a microdensitometer analysis of electron micrographs of shadowed surface replicas) makes it possible both to know the profile along many lines and to judge the quality of the reconstruction. This method proved to be particularly appropriate for microroughness studies. It is then possible to compute the autocovariance function and the power spectrum, which quantitatively characterize the surface. An analytical fit of the power spectrum is easy to obtain, especially in the high spatial frequencies range. This fit in turn allows us to determine accurately the Gaussian or exponential shape of the autocovariance function. Such a study was carried out on an extensive set of real microrough surfaces (ranging from rolling to ragged structures and from metallic to dielectric films). Parameters such as noise characteristics and level or sampling distance compared with autocovariance length and surface features’ sizes were carefully investigated. Hence, it is possible to answer the question: Gaussian or exponential autocovariance function? and to explain the discrepancies between results obtained with various methods (replica method, optical methods, profilometry…).

© 1991 Optical Society of America