5.1 Introduction
As so far, we have showed that homogeneous nanograin-size polycrystalline cobalt disilicide (CoSi2) thin film can be formed by oxide-mediated silicidation method and discuss its formation mechanism. But the thickness of CoSi2 thin film formed by this method still was a limitation.
Kim et al. [1,2] proposed that using Ti-capping can form a thicker epitaxial CoSi2
layer directly without repeating Co deposition and annealing processes. It is because Ti can absorb oxygen in Co layer and weaken the SiOx layer to enhance Co diffusion into Si substrate [3,4]. However, this process required sequential deposition of Co and Ti without vacuum break.
In this chapter, we show that uniform nano-nucleus CoSi2 with nano island size of 4 nm can be achieved by using oxide-mediated silicidation process with Ti-capping in a lower vacuum environment process by controlling annealing temperature and time. The underlined mechanism is also discussed.
5.2 Experimental Details
P-type (8-12 ohm-cm) silicon substrates were chemically cleaned and dipped in a boiling HCl:H2O2:H2O=3:1:1 solution for 3 min to form a SiOx layer (Shiraki Oxide) prior to loading the sample into a DC magnetron sputter. Subsequently, a TiN/Co/Ti/Co multiplayer was deposited using argon (99.995% purity) as the sputtering gas after the base pressure of 3 x 10-6 torr was reached and the target of Co or Ti (99.95% purity) had been pre-sputtered for 10 minutes. The multiplayer deposition was intermittence, where vacuum break was performed for target exchange between layers. The thicknesses of the TiN/Co/Ti/Co/SiOx multiplayer were about 10/4/10/4/2 nm, respectively, from a TEM cross sectional image. Ex-situ annealing was carried out in a vacuum chamber at 10-5 torr. Upon annealing, all layers except the reactive products were stripped off by chemical etching, in order to examine the silicide layer in plan-view. The TiN/Ti, the unreacted Co and the SiOx layers can be stripped off by NH4OH:H2O2: H2O=1:1:4 solution at 50 °C, aluminum etching solution (H3PO4 71 wt%, HNO3 2.5 wt%, CH2COOH 12.5 wt%, others H2O) at 75 °C and HF solution, respectively. The phase and nucleus size of the samples were then examined by transmission electron microscope (TEM). The chemical characteristics of the SiOx layer were analyzed by Fourier transform infrared spectrometer (FTIR).
5.3 Results and Discussion
Fig. 5-1 (a) shows a TEM bright field plan-view image and the corresponding diffraction pattern of the reactive silicide upon annealing at 600 °C for 120 sec with all the other layers removed. The diffraction pattern shows that the silicide is CoSi2, which agrees with that of Tung et al. and proved that CoSi2 can be formed directly bypassing CoSi, Co2Si from lower temperature (500~700 °C ) annealing, because the SiOx acts as a diffusion barrier reducing Co effective concentration at the cobalt disilicide growth interface [5,6]. The image shows that the sample is still in the process of grain growth, and exhibits bimodal size distribution with the peaks centered at the grain size of 3 nm and 16 nm as shown in Fig. 5-1 (b). According to the study of Detavernier [3,4], Ti should diffuse through the grain boundaries of the Co layer to the SiOx layer, and reduce SiOx to CoxTiyOz, which then enhances the Co diffusion to form cobalt disilicide. Fig. 5-2 (a) and 5-2 (b) are TEM bright field cross-sectional images from two different areas of the sample as in Fig. 5-1. Both images reveal that all of the interfaces are rough with a few dark areas at the SiOx/Si interface, corresponding to CoSi2. Since the layer is thin, interface roughness may play a significant role in Ti and Co diffusion, rendering the non-uniform nucleus distribution. The bimodal nucleus size distribution may be decided by the
involvement of Ti at various locations.
This bimodal size distribution can be amplified by doubling annealing time, which is exactly the case as shown in Fig. 5-3 (a) from a TEM bright field plan-view image of the sample upon annealing at 600 °C for 240 sec. Compared to Fig. 5-1 (a), nucleus grow further and bimodal distribution becomes more distinct, which will then lead to a continuous film with non-uniform grain size distribution and is not desired for real application. Thus, we employed a two-step annealing as shown in Fig. 5-3 (b), which is a TEM bright field plan-view image of the sample upon annealing at 460 °C for 240 sec followed by 600 °C for 240 sec. Comparing Fig. 5-3 (a) and 5-3 (b), the most striking feature is the discrepancy for the average nucleus size and the size distribution, which are 12±5.8 nm and 4±0.7 nm corresponding to the one-step and two-step annealing, respectively as shown in Fig. 5-3 (c). Apparently, the two-step annealing produces smaller and more homogeneous nucleus size distribution than the one-step annealing although both experience the same annealing at 600 °C for 240 sec.
Why the nucleus size from the two-step annealing become smaller and uniform although it experiences even additional annealing at 460 °C? According to Fitch et. al.
[7], they showed that the SiOx would become denser toward more stoichiometric SiO2
upon annealing and eventually turn into SiO2 at 900 °C for 30 sec. Therefore, the possible chemical stoichiometry was examined by FTIR in Fig. 5-4, which compares
the result from the as-deposited sample with the sample upon annealing at 460 °C for 240 sec. According to Chao et al. [8], the absorbance frequency of SiOx is in a range between 940 and 1075 cm-1 and that of stoichiometric SiO2 is about 1075 cm-1. While a broad peak representing SiOx for the as-deposited sample in Fig. 5-4, a more definite peak at 1075 cm-1 reveals the evidence of forming SiO2 upon annealing at 460
°C. Therefore, the Ti diffusion rate to the SiOx layer is slowed down due to the densified SiO2 microstructure and the transformation to SiO2. In the one-step higher temperature annealing, SiOx should still be able to transform to SiO2 but the ability of Ti diffusion to SiOx increases more significantly and help stabilize the SiOx layer, which causes the larger nucleus size (see Fig. 5-1 (a)). In the two-step annealing process, the SiO2 formation becomes dominant and the induced slow inter-diffusion may effectively render the nucleus size uniform distribution.
To further justify our supposition, another two-step annealing experiment was performed for 460 °C 120 sec followed by 600 °C 240 sec, where the annealing time at 460 °C was half of that used in Fig. 5-3 (b). Fig. 5-5 (a) shows a TEM bright field plan-view image from this sample. Compared to Fig. 5-3, apparently, the nucleus size magnitude and uniformity degree are just between Fig. 5-3 (a) and Fig. 5-3 (b). This is because the shorter 460 °C annealing time renders looser SiOx network than Fig. 5-3 (b), which is consistent with our previous supposition. The average nucleus size of
7±2.9 nm (see Fig. 5-5 (b)) is still smaller than that from the one-step annealing and the nucleus size distribution is also more homogeneous than one-step annealing.
5.4 Conclusions
Homogeneous nucleus size of 4 nm prior to a CoSi2 thin film can be obtained by oxide mediated silicidation with Ti capping in which cobalt was deposited by DC magnetron sputtering on SiOx/Si with the SiOx as a mediated layer followed by ex-situ two-step annealing (460 °C 240 sec and 600 °C 240 sec). Lower temperature annealing at 460 °C can alter the microstructure of the SiOx layer toward more stoichiometric SiO2, whichthen reduces Co and Ti diffusion rate and thus CoSi2
nucleus size leading to homogeneous nucleus size distribution. The homogeneous nucleus morphology remains unchanged even experiencing subsequent annealing at higher temperatures to increase Co diffusion rate.