SIMULTANEOUS REAL-TIME SPECTROSCOPIC ELLIPSOMETRY AND REFLECTANCE FORMONITORING THIN-FILM PREPARATION

An expansion of the capabilities of high-speed, multichannel spectrosc opic ellipsometry (SE) is described that involves simultaneous measure ment of the reflectance spectrum along with the two spectra in the ell ipsometric angles (psi, DELTA). Previously, a novel rotating-polarizer spectroscopic ellipsometer has been perfected that employs a photodio de array detector for high-speed acquisition of (psi, DELTA) spectra, designed for real-time studies of thin-film growth. For a polarizer an gular rotation frequency of omega0, the (psi, DELTA) values at a given photon energy are deduced from the 2omega0 Fourier components of the detector irradiance, normalized by the dc component. A third parameter , the weighted reflectance R(A), can be obtained from the dc component and from a calibration based on the known optical properties of the s ubstrate measured prior to film growth. With (omega0/2pi)=12.5 Hz, thr ee-parameter data sets, [psi(hnu), DELTA(hnu), R(A)(hnu); 1.5 less-tha n-or-equal-to hnu less-than-or-equal-to 4.5 eV], can be acquired with a time resolution as short as 40 ms. Although R(A) provides complement ary information to (psi, DELTA), it has yet to be exploited in conjunc tion with real-time SE until this study. A resilient analysis approach , based on mathematical inversion and least-squares fitting of the rea l-time three-parameter data sets, is designed to yield the film dielec tric functions and thicknesses independently at each time during the e arly stages of thin-film growth. The three-parameter approach has been applied successfully in studies of amorphous silicon (a-Si:H) thin fi lms prepared by plasma-enhanced chemical vapor deposition, and aluminu m and silver films prepared by physical vapor deposition. For the a-Si :H films, R(A)(hnu) exhibits deviations as large as approximately 3% f rom that predicted in modeling (psi, DELTA) alone. The deviations resu lt from light scattering by plasma particulates, and we show how addit ional information can be extracted from the spectral dependence of the scattering loss.