1st World Congress on Industrial Process Tomography, Buxton, Greater Manchester, April 14-17, 1999. 488 Optical Tomography for Dielectric Profiling in Processing Electronic Materials L. Zeni , R. Bernini, R. Pierri Dipartimento di Ingegneria dell’Informazione, Seconda Università di Napoli Via Roma 29, I-81031 Aversa Italy Phone +39 81 5010222, fax +39 81 5037042 e-mail pierri@unina.it Abstract - Optical methods represent a powerful tool for contactless characterisation of materials in industrial processes. In particular, the microelectronic field great impetus has been given to the on-line measurement of the doping profiles in large scale productions in order to increase the overall equipment effectiveness. In this framework, we propose a new technique based on optical tomography able to reconstruct the doping profiles in semiconductor wafers starting from reflected intensity measurements, taken at infrared wavelengths. Several numerical simulations have shown the effectiveness of the proposed approach. In particular, the reconstruction of typical doping profiles, generated by a process simulator, has been performed with relatively high accuracy. Keywords : Optical tomography, Doping profile, Semiconductor material, Contacless measurements 1. INTRODUCTION In the last years a great effort has been devoted to the development of new materials and process characterisation techniques [1]. In fact, the measurement of material characteristic provides both a check on the fabrication processes [2] and a tool to improve the reliability of process simulators as well. In particular, in the microelectronics (semiconductor material and device) and optoelectronics (optical fiber and planar waveguides), where the material properties are determined by the intentional implant or diffusion of impurities, the exact control and measurements of doping profiles, or refractive index variations, is a very crucial problem [3]. Hence, many different methods have been developed to this purpose. However, the most widely used, mainly in semiconductor industries, require the destruction of the sample under test or the realisation of suitable test structures which can change the doping profile to be measured. On the other hand, completely contactless optical techniques have been recently developed, but they only permit to reconstruct doping profiles representable by well know analytical functions [4-5]. In this work we propose a non-destructive and contactless method for the characterisation of one-dimensional doping profiles in semiconductor wafers. Our approach relies on the optical diffraction tomography, where the complex permittivity of a weakly scattering inhomogeneous object, illuminated by a known wavefield at different wavelengths and/or different directions, is reconstructed starting from the measurements of reflected and/or transmitted field. The proposed technique, starting from infrared spectroscopy data, allows to reconstruct one dimensional profiles in semiconductor samples, and could be used for both ex situ and in situ monitoring of technological process [2]. In our approach, we assume that only the field intensity is available. Using the integral relations of the electromagnetic scattering, under the distorted Born approximation, and taking advantage of the linear relationship, which holds true at infrared wavelengths, between the free carriers concentration and the complex permittivity of the semiconductor material, we relate the field intensity reflected by the sample to the doping profile [6]. The reconstruction problem is formulated as the minimisation of a proper non- linear functional representing the error between the measurements of the reflected intensity, at different frequencies, and the model data. In our approach the unknown carriers concentration profile is not described by a “parametric” expression of a known function [4], but an expansion in a finite series of basis function is used. So no “strong” assumption must be made on the functional form of the doping profile (e.g. exponential function, gaussian function, error function etc ). This particular choice of the data and unknowns allows us to tackle a quadratic inverse problem that has already been faced in the literature addressing and solving the problem