Abstract

Spectroscopic ellipsometry is a well established technique for the optical characterization of bulk materials, deposited thin films and surface layers [1]. It is based on the measurement of the change of the polarization state of light after reflection on a sample. This change is directly related to the complex ratio ρ defined by : ρ = r^p / r^s = tanψe^jΔ , where r^p and r^s are respectively the reflection coefficients of the waves parallel and perpendicular to the plane of incidence. A. Rosëler has been the first to combine Fourier transform spectrometer (FTS) to a classical rotating polariser or analyser ellipsometer [2]. The optical set-up of such an apparatus consists of a light source, an Michelson interferometer, a polariser, a sample, an analyser and a detector set. The accuracy of this ellipsometer is affected by polarization defects of the different components, particulary those of the wire grid used as polariser and analyser. Many procedures have been proposed recently to take into account imperfect components [2,3,4] such as interferometric residual polarization, polariser and analyser non nul attenuation coefficients and detector set dichroism. However, those procedures have some limitations and do not take in consideration the detector set non linearity. Our new ellipsometer is based on SOPRA’s compagny rotating analyser ellipsometer VASE-FTIR model 3 [5]. The use of special polariser and analyser allowed us to get an accuracy better than 1% in straight line without sample on the sample holder. We have observed however a slight shift on the cosΔ spectrum. This behaviour is induced by the detector set non linearity. Assuming a quadratic non linearity we have been able to correct the ellipsometric spectrum in real time. The accuracy in straight line reaches a few 1/1000 over the whole spectral range, which is very good in the IR.

© 1997 Optical Society of America