IUVSTA 15th International Vacuum Congress (IVC-15), AVS 48th International Symposium (AVS-48), 11th International Conference on Solid Surfaces (ICSS-11)
    Electronics Thursday Sessions
       Session EL-ThA

Paper EL-ThA3
Direct Numerical Inversion of Real-time Ellipsometric Data for Monitoring and Control of Optical Filter Deposition

Thursday, November 1, 2001, 2:40 pm, Room 124

Session: In-Situ Semiconductor Characterization
Presenter: B. Drevillon, Ecole Polytechnique CNRS, France
Authors: D. Kouznetsov, Ecole Polytechnique CNRS, France
A. Hofrichter, Ecole Polytechnique CNRS, France
B. Drevillon, Ecole Polytechnique CNRS, France
Correspondent: Click to Email
In situ ellipsometry is well known to be one of the most sensitive, non-disturbing tools for controlling and monitoring the growth of thin films. However the rapid advances in ellipsometric instrumentation, especially the increasing real time spectroscopic capabilities of state of art ellipsometers necessitate the development of new algorithms for an optimal assessment of the available information. In this work we apply a new direct numerical inversion algorithm for the real-time reconstruction of homogeneous and inhomogeneous refractive index profiles for the monitoring and the control of optical thin film depositions. The algorithm is based on a second order Taylor decomposition of the coefficients of the Abeles matrices of the newly grown layer. The variation of the real-time spectroscopic ellipsometry data are expressed as polynomial functions depending on the dielectric constant and the thickness of the newly grown layer. This allows a direct inversion of the ellipsometric signal and assures the high speed of the algorithm. Typical inversion times are 150 ms for 16 wavelength with a typical precision of 0.02 for the refractive index and less than 2% error in the reconstructed thickness. The algorithm is successfully applied for the real-time material characterization of transparent and weakly absorbant silicon oxynitrides deposited by plasma enhanced chemical vapor deposition. The complete process space can thus be explored in one single run. Combined with traditional ellipsometric control algorithm this method allows to grow multilayer and gradient optical coatings with high accuracy in respect to initial design.