In process ellipsometer azimuth angle calibration.

Two methods are described which provide in-process calibration of ellipsometer divided circles. Both methods (a residual method and a nonlinear least-squares estimation procedure) utilize multiple measurements on any arbitrary specimen to compute the azimuth angle corrections for the polarizer, compensator, and analyzer divided circles. The results of an experimental comparison of the two methods show (1) the confidence limits on the calibration constants are greater than is generally assumed, (2) that within experimental uncertainties both methods yield the same set of calibration constants, but one method may be more appropriate than the other for a given set of data, (3) the calibration constants can be a function of the wavelength of the light used for measurements, (4) other factors being equal, the maximum accuracy in a conventional ellipsometer system would be obtained by using a fixed-polarizer nulling scheme with the compensator set after reflection.