Pad Hardness

Kim S-D, Hwang I-S, Choi K-S. Hard-pad-based CMP of premetal dielectric planarization. J Electrochem Soc 2003 150(8) G450-G455. 

Qualitatively, since the contact area increases linearly with applied pressure, the effective pressure is constant for a given pad. Soft, compliant pads have a larger contact area and lower effective pressure whereas hard, stiff pads have a smaller contact area and higher effective pressure. Thus soft pads push abrasive particles against the wafer over a larger area but with less force than hard pads do. 

Since the abrasive-wafer surface contact area increases with the force applied, harder pads will remove more material per abrasive than soft pads, but soft pads—with larger contact areas—can remove more material than hard pads at low pressures. 

The limiting behavior, for large Pv, isR = a. When bE Pv, which is the case for hard pads, Equation 5.3 becomes R = (ajbE) Pv, which identifies the Preston constant as fcpreston = ajbE). Equation 5.3 thus predicts that hard pads do obey the Preston equation, whereas soft pads do not, except at very low pressures. This is illustrated by Fig. 5.14 for tungsten CMP. 

The qualitative explanation of Fig. 5.14 is that as pressure increases, the contact area will be much larger for soft pads than for hard pads, so the removal rate of soft pads will increase faster than the removal rate of hard pads. At low pressures, the mechanical removal step dominates the chemical oxidation step because only a very small fraction of the surface is unoxidized... 

The CMP process is composed of a chemical effect from nanosize ceramic particles and a physical effect from the pressed pad. Pads and slnrries are the consumables of a CMP process. The polishing pads consist of polynrethane. Generally two types of pads (hard and soft types) are simultaneously used in the CMP process. A hard pad gives better local (within die) planarity, bnt a soft pad gives better uniformity of material removal across the entire wafer. A hard pad is mounted onto a softer pad to form a stacked pad. Figure 15.1 shows a stacked CMP pad the hard pad is the Rodel 1C-1000 and the soft pad is the Rodel Snba-I V. ... 

Most species of mirmow have a laterally crxnpressed body, a terminal mouth, and relatively large, shiny scales. Mirmows also have pharyngeal teeth in their throat, which are used to grind their food against a hard pad at the base of the skull. Male mirmows are often smaller than females, and many species develop beauti-ftil colors during the spawnir season. 

Suppose a thin hard pad is on top of a thick soft pad. In this case it is concluded from figure 1 the top pad bends easily. As a consequence the bottom pad will be deformed non uniformly. Typical values for the bottom pad are a Young modulus (E) of 20 MPa and a thickness (1) of 1 mm. The maximal change in thickness for the bottom pad can be deduced from the results for the within die non-uniformity. A typical value is 300 nm. In the perfect pad bending assumption this level difference is entirely translated into thickness variations of the bottom pad (Al). Hooke s law allows to calculate the additional pressure resulting from these deformations, i.e. 

Figure 4 SEM cross-section of cache and DensS structure on a 0.13pm vehicle on a hard pad...

Correlation between the standard deviation of nanotopography profile and fhe film fhickness variation before/after CMP are shown in Figure 2.10 (for soff pad) and Figure 2.11 (for hard pad). It is reasonable that film thickness variations before CMP were independent of nanotopography. However, after CMP, the film thickness variation and nanotopography have positive correlation. Here two facts are pointed out. 

FIGURE 2.11 Correlation between standard deviations of nanotopography and film thickness variation for hard pad test. 

FIGURE 2.12 Power spectral densities for (a) soft pad test and (b) hard pad test. 

The process condition of the POC process is almost identical to the STI CMP process condition. Therefore, for most of the cases, STI CMP and POC are processed on the same CMP equipment. As explained in STI CMP, POC removes most of the oxide at the first oxide CMP step. At the second CMP step, the process removes the remaining oxide and stops on the nitride with minimum nitride loss. Sdica slurry with hard pad is typically used as the first oxide step, but ceria slurry is also used at the first oxide step to minimize any pohshing scratches. The ceria slurry is used as the second step CMP slurry to maximize the selectivity between the oxide and the nitride as well. SEM images of the POC wafer at different stages are shown in Figure 1.26. 

Figure 15.1 plots oxide removal rate against slurry flow rate for a blanket silicon dioxide removal process that uses a high solids content fumed silica slurry and a conventional concentrically grooved hard pad. The pad in the experiment is rinsed... 

---