Oxidation modeling of thin SiO₂ using N₂O for deep submicron CMOS technology

NH₃-Nimded SiO₂ has been studied in the past ten years to improve the elecmcal propemes of silicon dioxide layer. A re-oxidation step is required immediately after nitridation step to minimize the high elecnon trapping efficiency and high electron trap generation rate due to a large amount of hydrogen atoms (H) in the oxide. Though reoxidation has been proved to be effective in reducing hydrogen concentration., the optimization of this process is complicated. Recently there has been uemendous interest in N₂O oxidation using either conventional furnace [l] or rapid thermal processor [2] to form high quality gate dielecmcs. The growth of SiO₂ was reported to be self-limiting. However, we have found the non-saturation growth behavior over a wide range of temperatures and different crystal orientations in N20 oxidation [4]. In this work we present an oxidation model and discuss some simulation resuIts. PI20 oxidation was performed in a conventional resistive-heated furnace in high purity gas (moisture less than 0.5 ppm) (99.9995%) for (100), (1 11) and (1 10) oriented Si substrates. Fig. 1 shows oxide thickness versus oxidation time for (111) Si over the temperature from 900 C to 1100 C. (100) and (110) oriented substrates exhibit similar behavior but with different thicknesses. Several important features can be observed from this figure. First, the growth of SiO₂ was found to be not self-limited. Second, the N₂O oxidation rates are much slower than those in dry 02. Third, the apparent higher growth rate at 1100 C is probably caused by the higher degree of nitrous oxide dissociation at high temperatures. The oxide growth behavior in K₂O can be modeled by using oxidation of silicon in diluted dry oxygen. Empirically it was found that the growth rate in pure N20 is essentially identical to that in 10% dry 02-The experimental data of lo%/& came from Ref. 5. The comparison is shown in Fig. 2. Both NzO and 10% 02 in argon follow the same linear-parabolic model. Based on the experimental data, except for the very short oxidation time, the linear rate constant (B/A) and parabolic rate constant (B) can be determined and are plotted as a function of temerature as shown in Fig. 3 and Fig. 4 respectively. Previous study of N20 oxidation covers only a limited temperature range (885 C-975 C) and time ( up to 100 minutes) and oxide thickness (up to 160 A) [6]. Therefore the growh behavior in that cae was characterized only by the linear rate constant (B/A). Our model can fit a much broader tempemture range and oxidation time since higher temperatures and longer oxidation time were considered. It is interesting to note that the activation energ EA of B/A for (100) Si obtained from this work is 1.005 ev, which is the same as 1.0o~bt ained from Ref. 6. In summary the growth kinerics of N₂O oxidation has been studied and an empirical model has been proposed to cwer both linerar and parabolic regions of oxide growth.