In this context it is well-known that in certain situations a surface profile obtained by an optical profiler will differ from the real profile. With respect to practical applications such deviations often occur in the vicinity of steep walls and in cases of high aspect ratio.
In this contribution we compare the transfer characteristics of different 3D optical profiler principles, namely white-light interferometry, focus sensing, and confocal microscopy. Experimental results demonstrate that the transfer characteristics do not only depend on the parameters of the optical measurement system (e. g. wavelength and coherence of light, numerical aperture, evaluated signal feature, polarization) but also on the properties of the measuring object such as step height, aspect ratio, material properties and homogeneity, rounding and steepness of the edges, surface roughness. As a result, typical artefacts such as batwings occur for certain parameter combinations, particularly at certain height-to-wavelength ratio (HWR) values. Understanding of the mechanisms behind these phenomena enables to reduce them by an appropriate parameter adaption. However, it is not only the edge artefacts, but also the position of an edge that may be changed due to the properties of the measuring object.
In order to investigate the relevant effects theoretically, several models are introduced. These are based on either an extension of Richards-Wolf modeling or rigorous coupled wave analysis (RCWA). Although these models explain the experimental effects quite well they suffer from different limitations, so that a quantitative correspondence of theoretical modeling and experimental results is hard to achieve.
Nevertheless, these models are used to study the characteristics of the measured signals occurring at edges of different step height compared to signals occurring at plateaus. Moreover, a special calibration sample with continuous step height variation was developed to reduce the impact of unknown sample properties. We analyzed the signals in both, the spatial and the spatial frequency domain, and found systematic signal changes that will be discussed. As a consequence, these simulations will help to interpret measurement results appropriately and to improve them by proper parameter settings and calibration and finally to increase the edge detection accuracy.