Recent results of 1~O tracer experiments on thermal oxidation of Si in dry 02 suggest two oxidizing species are responsible for the growth of SiO2. We examine here a model that reflects these mechanisms in the growth rate, using a sum of two linear-parabolic growth rates. The model gives very close fits to the oxidation kinetics data for oxides ranging from 30• to 1 ~m thick and for temperatures from 800 ~ to 1200~ at both (100) and (111) Si orientations. Only three parameters are necessary in this model to provide excellent fits to the experimental data. A parallel diffusion of 02 and another undetermined species is consistent with the result.It has long been recognized that the linear-parabolic model adequately describes the Si oxidation kinetics for thick oxides but inadequately describes the oxidation kinetics during dry 02 oxidation for oxides thinner than a few hundred angstroms. Numerous mechanisms have been proposed to explain this rapid initial oxidation period for thin oxides. These models include space-charge enhancement of the diffusing ionic species (1), rapid diffusion of oxidant through microchannels and micropores in the oxide (2-5), varying diffusion coefficients caused by a changing oxide structure with thickness (6), diffusion retardation by a blocking layer at the interface (7, 8), parallel oxidation of two diffusing oxidants (9), oxidation rate enhancement from an altered Si surface layer (10), and oxidatiori rate retardation from a buildup of fixed charge at the interface (11, 12).Many new experiments have also been performed to test these models. One technique that proves extremely valuable uses 180 tracers to investigate oxygen transport through the oxide. Various such studies (13-16) already showed that new oxide grows primarily at the SiO2-Si interface, but also appreciably at the oxide surface. The experiment by Rochet et al. (15) indicated the growth of the surface peak arises from the diffusion of an oxidizing species to the surface of the oxide. This mechanism, combined with Deal-Grove's original mechanism for the interface peak, suggests an oxidation model based on the diffusion of two different oxidizing species. Such a mechanism was originally proposed by Hopper et al (9). The values of the parameters could not be obtained, however, because the data used by Hopper was not available over a sufficient range of thickness. Additional data available from in situ ellipsometry measurements of the oxidation rates allow us to test this model more extensively over a wide range of oxidation conditions. Other parallel oxidation models have also been suggested (3, 5) but are based on the premise that the same species is diffusing via two diffusion paths.In this paper, we first summarize the parallel oxidation model which leads to an analytical relationship for oxide thickness vs. oxidation time, and then report the quality of the fits of this relationship to several sets of oxidation rate data over a wide range of thicknesses and temperatures at both (100) and (111) orientations. The fitted v...

No abstract

In this communication we report on an analysis of recent experimental data (i) which clearly shows that transport properties of thermally grown Si02 change as a function of temperature.Our analysis is consistent with other observations where the refractive index (2,3), density (3), and interfacial stress (4) in Si02 vary with temperature.These changes account for the apparent anomalous effect of temperature on the activation energy for the parabolic rate constant, B, reported by several investigators (5-7). Figure i shows this effect in an Arrhenius plot of these data (5-8). The activation energy varies from 0.9 eV at I150~ to 2.6 eV at 780~ Recent experiments by Irene (i) shed considerable light on the mechanism for this change in activation energy. He grew oxides in dry 02 at 1000~ to l~m thicknesses, then recorded oxidation rates for subsequent oxidations at lower temperatures. The rates measured at lower temperatures should have therefore exhibited the properties of a "I000~ '' oxide, so long as the lower temperature oxidations did not significantly affect their structure.In a reanalysis of Irene's data, using an improved lag time formulation, we indeed observe this effect (9). The oxidation rates were analyzed using the linear parabolic rate law:where L is the thickness, B is the parabolic rate constant B/A is the linear rate constant, and dL/dt is the growth rate. Since dL/dt and L are measured quantities, B can be determined if A is known;since L>>A for these experiments, the error caused by an uncertainty in A is not significant. * Electrochemical Society Student Member ** Electrochemical Society Active MemberTo verify this lwe evaluated the value for B in two limits.First, the B/A values from Ref. (5) provided one limit; second, B/A § provided the other limit for the minimum value for B. We plotted these B values for Irene's 800~900~ and 1000~ oxidation rates in Fig. i along with the previously reported B values (5-8). As mentioned above, previous results showed the familiar change in the apparent activation energy as a function of temperature.However, the parabolic rate constants from Irene's sequential experiments at 800~ and 900~ fall alongra straight line tangent to the curve of B values at 1000~ with a slope of i.I eV! During the second oxidation, the oxide appeared to retain the activation energy and the preexponential factor from the 1000~ oxidation.In the 40 hours Irene oxidized his samples, the 1000~ oxide apparently did not change its form to equilibrate with the lower temperature.It retained a "memory" of the past oxidation process.The direction of the change to lower activation energy for higher temperature oxides is also not surprising since the lower density and the stress relaxation for higher temperature oxides would promote the lower activation energy.There is also evidence that the linear rate constant exhibits a "memory effect" as well.Hamasaki (i0) observed enhanced oxidation rates for thinner oxides in a sequential higher temperature/lower temperature oxidation sequence.In that case,...

The kinetics of thermal nitridation of silicon dioxide in ammonia ambient has been studied. SiO2 films of 100-1000~ thick were thermally nitrided at 950~176 for times from 15s to 2h. Our experimental results based on etch rate and Auger electron spectroscopy measurements clearly indicate the multilayer structure of nitrided-oxide films. Nitrogenrich layers are formed at the surface and interface regions at a very early stage of the nitridation process. After a few minutes, the nitridation reaction mainly goes on in the bulk region, with the surface and interface nitrogen content remaining fairly constant. The Auger depth profiles show that the interface moves away from the nitrogen-rich layer as the nitridation proceeds. In addition, our results indicate the formation of an oxygen-rich layer underneath the nitrogenrich layer whose thickness increases with nitridation time. The formation of this oxide-like layer can be attributed to a slow oxidation of the silicon substrate at the nitroxide/silicon interface by the oxygen which is a by-product of the bulk exchange reaction between NH3 (or nitrogen containing species) and SiO2. The results of this work can be qualitatively used to explain effects such as enhanced boron and phosphorus diffusion and growth of stacking faults in the silicon substrate during nitridation of oxide.

Articles you may be interested inInfluence of process parameters on the timedependent dielectric breakdown of rapid thermally nitrided and reoxidized nitrided thin SiO2