Abstract
This study is aimed toward identifying the reasons why large discrepancies exist in the literature relative to the kinetic constants which are supposed to describe the thermal oxidation of single crystal silicon (Si) in pure oxygen (O2). In order to obtain sufficient quality and quantity of silicon dioxide (Si02) film growth thickness-time data, an automated ellipsometer was used to measure the SiO2thickness in situ. The resulting data was fitted to the generally accepted linear-parabolic model by several commonly used methods and the results compared. Careful attention was given to eliminating trace amounts of H20 and Na so that the data are representative of oxidation in pure O2 ; the oxidation data was compiled in the temperature range of 780°-980°C. Calculated standard deviation values were used to evaluate the quality of the fit of the data to the model. From this analysis an initial regime of rapid oxidation was identified which does not conform to linear-parabolic kinetics. This regime extends up to about 350A. A best fit of the data to the model was achieved using data above 350A. By using either data below 350A or only data greater than about 1100A, large curve-fitting errors (~50% in terms of rate constants) were obtained. It was concluded that this source of error in combination with impurity effects, insufficient data, and the specific form of the curve-fitting equation could yield the large reported discrepancies. The activation energy calculated from the linear rate constants of this study (1.5 eV) indicates that 0-0 bond breaking is important for linear kinetics and the activation energy for the parabolic rate constants (2.3 eV) is too large to be correlated with a reported value (1.2 eV) for the diffusion of O2 through Si02. © 1976, The Electrochemical Society, Inc. All rights reserved.