Precision surface metrology in recent semiconductor industries is grouped under two major fields of study. These are optical properties and thickness measurements and surface profile measurements. In the former, Ellipsometry is the typical method and in the latter, various measurement methods, for example phase-shifting interferometry, white-light scanning interferometry, confocal microscopy, etc have been developed. But these methods cannot measure thickness distribution or surface profile of thin-film samples whose top layers are transparent below a few micron thickness and have patterns on the surface in micro orders.
This dissertation shows a new measuring method that can measure thickness distribution and surface profile of these samples simultaneously. So it can reconstruct the 3-D thickness profile same as real structure of transparent thin-film layer
In the previous research concerning the white-light scanning interferometry, only the phase changes due to the optical differences between the reference plane and measuring points are used for profile measurement. But there are not only phase changes due to the opticaI pass differences but also those due to the reflection on a surface in interferograms and the latter includes informations about the structures of measuring points. So, it is possible to extract these informations, for example heights and thickness, from analyzing phases acquired by Fourier transformation of the interferograms.
To get these informations, new algorithm is developed as follows; First step, practical phase data are calculated from interferograms acquired using white-light scanning interferometry. Second step, theoretical phase model in consideration of multiple reflection phenomena in transparent thin-film is defined. The last step, heights and thickness values are searched by comparing the phase model and practical phase data. This phase model mentioned above comprises linear and nonlinear terms; the former represents phase change due to optical pass differences, the latter due to surface structure, such as the thickness, of measuring points. So this model has two variables, thickness and height.
Finally, these values at each measuring points are calculated by fitting the phase model to corresponding phase data using optimization technique, especially nonlinear least squares method.
This dissertation shows the practical measuring algorithm and analyzes the error sources in the algorithm through various simulations. It also shows the $SiO_2$ thickness profile of various samples having different structures.