Polish Rate Results

One of the key performance metrics for a CMP process is the polish rate, including the uniformity of the polish rate across the wafer and from wafer to wafer. In the previous sections and in Chapter 4, material removal is discussed in terms of fundamental principles. This section discusses the polish rate in terms of process variables and emphasizes what tools the process engineer has to improve polish rate and polish rate control. 

One of Ae biggest concerns with oxide polish is the uniformity of the polish rate across the wafer. The across wafer nonuniformity of the polish rate can be considerable leading to a variation in insulator thickness from die-to-die across the wafer. Consider, when removing 1.5 pm of oxide, that a 3-a nonuniform- 

Polish rate of PSG oxide vs. phosphorus concenuation. (From Ref. (18).) 

In addition to the nonuniformity in the oxide polish rate, non-uniformity in the oxide deposition process leads to variations in the final thickness of the oxide. Deposition nonuniformities are compounded by the fact that thick oxides must be deposited prior to CMP. For example, if the oxide CMP process must remove 5(X) nm in order to planarize the surface, with a final target oxide thickness of 500 nm, a 1 pm thick film must be deposited. A 10% variation in film deposition rates transfers to a final thickness variation of 100 nm or 20% of the final film thickness (assuming no variation in CMP rate across the wafer). Alternatively, if the CMP process must remove 1 pm of oxide in order to planarize, the deposited oxide must be 1.5 pm thick. The same 10% variation in film deposition rates now results in a 150 nm thickness variation or 30% of the final thickness. Thus, the CMP process affects the final oxide thickness uniformity by virtue of the planarization rate as well as polish rate uniformity. This nonuniformity is a second reason that ceria-based slurries discussed in Section 5.1.3 are undesirable. 

---

The different polish characteristics of each of the traces demonstrate the significance of layout pattern dependencies in oxide CMP. For example, the L3 profile traces the large step density transitions with dramatic variations in the resulting oxide thickness. In contrast, the blocks along L4 polish at the same rate, underscoring the fact that layout pitch is not a first-order determinant of the polish rate. This is the case for a very large range of pitch... 

Typically, there are three principal parameters in a CMP recipe, namely, the down force, the platen rotation speed, and the carrier rotation speed. If we put the results of the DOE study from Table I into contour plots, it is easier to see the influences of each factor on the polish nonuniformity. Figures 6-8 show the relationships between recipe parameters and the polish rate and nonuniformity. We see, from Fig. 6, that the polish rate and nonuniformity can be simultaneously improved by using a larger down force. However, if the down force is set too high, scratches on a wafer may result. Decreasing the platen speed and the carrier speed decreases nonuniformity, as we can see from Figs. 6 and 7. However, decrease of the platen speed also decreases the polish rate, as shown in Fig. 6. 

Another issue with the end effector is its down force during conditioning. This down force must be as low as possible, as long as the polish rate remains stable. If the down force is set too high, the resultant high wear rate shortens the pad life. Once the grooves on the pad are worn out, the pad can no longer deliver slurry. 

Experimental evidence strongly suggests that material removal in chemical-mechanical polishing (CMP) processes is a result of one or more chemical steps that alter the wafer surface combined with a mechanical step that removes the altered material. Chemical action by itself also removes material by static etching, but generally at a much lower rate than is observed when mechanical action is also present. Similarly, polishing rates observed when a minimally reactive fluid such as water is used instead of slurry are also low. Both chemical and mechanical processes are therefore involved in material removal at commercially practical rates, and the model we describe reflects this dual nature of the process. 

As shown in Fig. 13.25, the oxide rate is almost independent of slurry D99 under all pressure conditions. The rate for slurry with D99 = 0.14 did show a drop in removal rate. This is the result of a change in D50 for the slurry. As shown in Fig. 13.25, the oxide removal rate shows a linear relationship with D50 under various downpressures. The polish rate slope also increases with increasing... 

Oxidizers For metal CMP, most of the chemical reactions are electrochemical in nature. Oxidizers react with metal surfaces to raise the oxidation state of the metal via a reduction-oxidation reaction, resulting in either dissolution of the metal or the formation of a surface film on the metal. For both tungsten and copper, polish rate has been shown to be proportional to the rate of these reduction-oxidation reactions (see Chapters 6 and 7). 

Recently Hayashi et al. have studied the effect of adding an NHj-salt to an NH4OH or acetic-acid-based silica slurry on the polish rate of SiOz films. Figure 4.53 shows their results. A considerable increase in the SiOj polish rate was observed for all slurries (in the pH range of 6-9) by the addition of small amounts of the ammonium salt. The results were explained on the basis that an addition of salt to the slurry of pH=6-7 led to the reduction of the electrical double layer on the abrasive silica particles. This in turn promoted the agglomeration of the abrasive particles and increased the polish rate. 

---