The demand for higher speed and lower power consumption ICs has motivated research for higher mobility channel materials (1). Compressively strained SiGe is known to feature higher hole mobility than Si and is largely compatible with the Si CMOS manufacturing platform. Besides, the FDSOI transistor architecture allows power consumption reduction (2), thus making SiGe-On-Insulator (SGOI) channels promising. SGOI layers can be fabricated by the so-called condensation technique, which is based on concurrent Si selective thermal oxidation of SiGe and SiGe composition homogenization by Si and Ge interdiffusion (see Figure 1) (3). Therefore, a sound understanding of kinetics of oxidation and interdiffusion of SiGe is required to develop optimized SGOI structures. SiGe dry oxidation rate has been reported to be higher (4) than or equal (5)(6) to the one of Si. It therefore remains unclear. Few groups investigated Rapid Thermal Oxidation (RTO) (4) while most of them looked at furnace oxidation. Moreover, most studies only considered the initial Ge content to compare oxidation kinetics. Indeed, the Ge concentration below the SiGe-oxide interface is strongly varying with time because of two mechanisms: firstly, the Si-selective oxidation of SiGe tends to pile-up Ge below the oxidizing interface; and secondly, interdiffusion of SiGe favors homogenization of the layer (6). This paper focuses on the oxidation rate of SiGe in regards to the varying Ge concentration in the SiGe layer for various oxidation conditions. Thick (> 20 nm) SiGe layers with either 10% or 30% Ge concentrations were epitaxially grown on bulk Si wafers. Oxidation was performed by RTO in 1 atmosphere of pure O2, with different oxidation temperatures and durations. Oxide thickness and SiGe composition versus depth were measured by Spectroscopic Ellipsometry and X-Ray Reflectivity, and by Secondary Ion Mass Spectroscopy respectively. Figure 2 shows the Ge concentration profile below the oxide. A rapid creation of a pile-up layer is observed at all temperatures. Then, three regimes of evolution of the pile-up layer are distinguished: the Ge concentration at the oxidizing interface (a) increases, (b) is constant, and (c) decreases. The regime of evolution of the pile-up layer is determined by a competition between the oxidation speed and the interdiffusion speed. As schematically illustrated Figure 3, the interdiffusion speed overcomes the oxidation one at higher temperatures because the activation energy of interdiffusion in SiGe ([4,5] eV (7)) is well above the one of SiGe oxidation ([2,2.6] eV (4)). Phrased in a different way, the temperature dependence of interdiffusion is higher than the oxidation one. Figure 4 shows the oxide thickness versus the oxidation duration for dry RTO at 900°C, 1000°C and 1100°C. A higher oxidation rate is observed for SiGe compared to Si for all temperatures. Then, to highlight the effect of the Ge concentration at the oxidizing interface, we fitted the curves to extract oxidation rates for each oxidation conditions. We define the Growth Rate Enhancement (GRE) as the ratio of the oxidation rate of SiGe to the one of Si at a given oxide thickness and for identical oxidation conditions. It is indeed important to highlight that comparing oxidation rates at the same oxide thickness is mandatory: if we want to examine the effect of Ge on the oxidation rate, we have to take into consideration the fact that the oxidation rate also decreases with the oxide thickness. Such an approach allows us to freeze the system and to find more easily correlations. We report on Figure 5 four GRE values extracted from our data versus the Ge concentration at the oxidizing interface. We thus observe that the higher the Ge concentration at the oxidizing interface is, the higher the GRE tends to be. We thus showed that Ge redistribution by interdiffusion during SiGe oxidation can follow different regimes. We also evidenced that for dry RTO conditions, the higher the Ge concentration at the oxidizing interface is, the higher the GRE is. Understanding the interdependence of these two mechanisms is essential to get an accurate picture of the condensation process. Therefore, data for dry RTO will be completed and an in-depth discussion will be led. A similar study will also be conducted for a wet oxidation process called In-Situ Steam Generation (ISSG). (1) Pillarisetty, R., Nature 479.7373 (2011): 324-328. (2) Weber, O., et al., VLSIT IEEE, 2014. (3) Tezuka, T., et al., JJAP 40.4S (2001): 2866. (4) Spadafora, M., et al. MSSP 8.1 (2005): 219-224. (5) LeGoues, et al., APL 54.7 (1989): 644-646. (6) Long, E., et al., PSS (a) 209.10 (2012): 1934-1939. (7) Kube, R., et al., JAP 107.7 (2010): 073520. Figure 1
Read full abstract