Appl. Phys. A 58, 77-80 (1994) Applied so,,.. PhysicsA "" Surfaces © Springer-Verlag 1994 Characterization of SIPOS Films by Spectroscopic Ellipsometry and Transmission Electron Microscopy G. Kragler 1 H. Bender 1, G. Willeke 1, E. Bucher 1, J. Vanhellemont 2 1 Universit~it of Konstanz, Faculty of Physics, P.O. Box 5560, M 500, D-78434 Konstanz, Germany (Fax: +49-7531/88-3895) 2 IMEC, Kapeldreef 75, B-3001 Leuven, Belgium (Fax: +32-16/281501) Received 6 May 1993/Accepted 14 August 1993 Abstract. The structural and compositional properties of undoped SIPOS thin films have been studied by spectro- scopic ellipsometry and transmission electron microscopy. It is shown that in most cases the former method provides fast and reliable results. The growth rate and crystallinity of SIPOS layers are studied as a function of N20 concentration in the gas phase and annealing temperature. PACS: 07.65, 07.80, 68.55 Semi-Insulating Polycrystalline Silicon (SIPOS) has origi- nally been developed as a passivating layer for high voltage devices [1]. Another application is the incorporation of a SIPOS emitter as a passivating layer in conventional sili- con solar cell structures [2]. The efficiency of these devices depends on the optical and structural properties of SIPOS which in turn depend strongly on the deposition conditions. In the present paper Spectroscopic Ellipsometry (SE) in com- bination with Transmission Electron Microscopy (TEM) has been used to study these layers as a function of deposition and annealing conditions. The present work is an elaboration of preliminary results published elsewhere [3]. Experimental Details The SIPOS film depositions were carried out in a conven- tional LPCVD reactor using Sill 4 and N20 as reactants. A constant silane flow of 25 ml/min and a constant deposition time of 10 min were used in all experiments. By adding a flow of nitrous oxide, the ratio 7 of N20 to Sill 4 molecules in the gas phase was varied between 0.0 and 0.2. In the present work only undoped films were studied. A substrate temperature of 620°C and a pressure of 300 mTorr were maintained during deposition. The SIPOS layers were de- posited on quartz and polished silicon substrates positioned horizontally on a quartz support. In order to obtain the necessary low sheet resistances of SIPOS solar cell emitters, an annealing step between 800°C and 1000° C is required. This was performed in a separate quartz tube under an Ar/O 2 (85%/15%) flow. In these experiments the annealing temperature was reached in approximately 20 min, kept for 15 min and then ramped down to 300 ° C during about 2 h. The spectroscopic ellipsometry (SE) measurements were performed on a SOPRA ES4G instrument in an energy range between 1.4 eV and 4.3 eV. The spectra were evaluated with the Bruggeman Effective Medium Approximation (EMA) [4], using reference files from literature for a-Si [5], c-Si and SiO 2 [6]. The TEM studies were carried out on the Jeol 200CX microscope of the University of Antwerp (RUCA). Experimental Results and Discussion SIPOS can be regarded as a three component mixed phase consisting of crystalline and amorphous silicon as well as voids or SiO 2 [7]. In previous work [3] we have shown that it is not possible in SE to distinguish with spectroscopic ellipsometry between SiO 2 and voids in the considered en- ergy range. The respective fits have thus been carried out by adding voids to the polycrystalline silicon matrix. It was found that the measured ellipsometric spectra often do not show the crystalline peak at 3.4 eV as pronounced as ex- pected from the used EMA mixture of phases. This crys- talline feature, which results from interband transitions at the smallest direct bandgap in crystalline silicon, can be taken as a "fingerprint" for the crystalline phase. Optical direct reflectance measurements [8] have shown that this feature is weakened in polycrystalline Si thin films due to light scattering. However, in our spectroscopic measurements this wavelength dependent scattering effect is considered to be of minor importance since the experiment detects changes in polarisation rather than light intensities. Because of the above mentioned disagreement, the deduced results depend more or less sensitively on the fit criteria used and the choice of the energy range. In other words the optimization of a