Zhefu Wang, Rixin Peng, Sherry Xia, Stan Tsai Applied Materials, Inc. Sunnyvale, CA, 94086
INTRODUCTION Tungsten CMP process, extensively used to build via and holes for front end of the line contact in current IC fabrications, holds its own uniqueness in terms of removal mechanism. According to Singh et al., unlike copper CMP which requires longer passivation time and the aid of passivation agents to form the surface layer, tungsten surface can grow a complete thin surface layer in very short time [1]. Lim et al. reported the effects of oxidants on the tungsten CMP process and suggested that fast and effective formation of oxide layer was critical for high removal rate [2].
This work studied the effect of slurry abrasive concentration and process temperature on tungsten CMP using a dual head polishing platform. The removal rate was observed to scale with the cubic root of the abrasive concentration. The temperature controlled experiment showed a linear relation between the tungsten removal rate and the process temperature. Our data suggested that a higher process temperature could lead to a thicker oxide layer and in turn raised the total removal rate.
EXPERIMENTAL DATA The dual head polisher used in this study was the Applied Materials Reflexion GT. Instead of one-head-on-one- platen configuration (single head polishing mode) employed in the current industry, Reflexion GT has two platens, each of which polishes two wafers simultaneously (dual head polishing mode) to enhance the productivity of CMP process. Figure 1 showed the tungsten removal rate for different slurry abrasive concentrations on blanket wafers with all the other process conditions kept the same. An infrared temperature sensor at the top of the polishing head was used to record the process temperature.
In the process temperature study, one single wafer was polished for each experiment split using the single head polishing mode. An in situ heating device based on mechanical friction was used to raise the process temperature on the pad. The process temperature was then tuned independently while the variance of the other process conditions remains the minimal. Figure 2 showed a linear relation between the tungsten removal rate and the process temperature.
ANALYSIS AND DISCUSSION The process temperature study in Figure 2 suggested a linear relationship between the process temperature and the removal rate. Since all the process parameters except temperature were kept approximately unchanged and the process temperature variance in this experiment was assumed to cause no changes in the mechanical facets of the polishing process, the higher temperature tended to lead to stronger surface chemical reaction, then a thicker oxidation layer, and finally a higher removal rate. A simple mathematical derivation of the Arrhenius equation shows that the process temperature difference would be linearly proportional to the final removal rate.
The slurry abrasive concentration test results from Figure 1 was analyzed based on the study of Copper et al. [3]. Figure 3 re-ploted the same data as Figure 1 against the cubic root of the abrasive concentration and showed good agreement with Coppers model. Since the model was purely derived from the relationship between the particle concentration and the collision frequency, the good fitting implied that the removal rate was determined only by mechanical particle collision. Another fact that supported our hypothesis about the oxide layer formation was that the process temperature remained almost constant through all the experimental splits in Figure 1.
In summary, the time constant of the tungsten oxide layer formation tended to be short compared to the particle collision time interval, at least in the process window of our study. However, the effect of process temperature on the tungsten removal might have a different mechanism. Higher temperature led to a thicker oxide layer in the unit time through faster surface reaction rate. Understanding of the effects of these parameters will help tailor process development to meet industry application demands, such as higher throughput, lower cost, and better topography. REFERENCES 1. Singh, Rajiv K. et al., MRS Bulletin, Oct. 2002 2. Lim, Geonja et al., Wear, Feb. 2004, Vol. 257 3. Copper, Kevin et al., Electrochemical and Solid-State Letters, 5 (12), Oct. 2002,
Figure 1: Removal rate and process temperature as function of slurry abrasive concentration.
Figure 2: Tungsten removal rate as a function of process temperature.
Figure 3: Tungsten removal rate as function of (slurry abrasive concentration) 1/3 . Abstract #1971, 220th ECS Meeting, 2011 The Electrochemical Society ecsdl.org/site/terms_use address. Redistribution subject to ECS license or copyright; see 152.135.235.188 Downloaded on 2013-05-21 to IP