The trend towards smaller design geometries for microelectronics devices places unprecedented demands on the measurement of these small structures. Scatterometry is a promising new optical metrology method for nondestructive rapid evaluation of many physical quantities of critical interest in microelectronics fabrication. The analysis of scatter data is carried out with the help ofRigorous Coupled Wave Theory (RCWT) which is used to determine the intensity oflight in each diffracted order based on arigorous application ofMaxwell's equations and relevantboundary conditions. Scatterometer data mustbe relatedback to the diffracting structure, the so-called "inverse-problem", and a novel technique for the solution of this problem is presented.Two sample problems are considered. In the first case we predict the shape ofdeveloped photoresistgratings from diffraction dataobtained from an angle scanning scatterometerin which a detector tracks aparticulardiffraction orderas theangle of incidence is varied. A database oftheoreticalscattered field envelopes is createdby solving forthe diffractedfields using RCWT foreach angle of incidence and each set of parameters in the parameter space. Multivariate statistical analysis of the simulated scatter data is used to build estimators that predict grating parameters from a measured optical scatter pattern. Combining scatterometry, computer simulations ofscauer, and statisticalanalysis to constructestimators avoids theproblems ofproducing a setofcalibration samples and finding an accurate and precise reference method to measure the grating parameters of interest.The second problem we consider is the prediction ofdepths ofcylindrical cells etched into Silicon. A dome scatterometer is used to collect the two-dimensional diffraction signal and in this case the experimental data is used to construct a data base of scatter signals. The etch depths are obtained using cross section SEM techniques. Using some ofthe scatter data to build an estimator and the remaining data for testing we obtain excellent agreement between the predicted and measured etched depths. 0-8194-1242-2/93/$6.00 Downloaded From: http://proceedings.spiedigitallibrary.org/ on 07/03/2016 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspxThegrating parameter estimation problem may also be viewedas apatternrecognition problem, whereby a distinct diffraction signal is associated with a specific grating shape. In order for this technique to work alarge calibration database ofdiffraction signals must be generated. It has been shown by Gottscho et. al. [19] that a neural network may be trained to classify the grating from its SPIE Vol. 1992 Miniature and Micro-Optics and Micromechanics (1993) / 171 Downloaded From: http://proceedings.spiedigitallibrary.org/ on 07/03/2016 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx SPIE Vol. 1992 Miniature and Micro-Optics and Micromechanics (1993)! 177 1 6 11 16 Downloaded From: http://proceedings.spiedigitallibrary.org/ on 07/03/2016 Terms of Use: http://spie...