On the Use of Pad Mechanical Properties as a Proxy for Conditioning Response

A proliferation of pad materials in recent years, coupled with the increasing complexity of process integration and limited resources available for process development, has highlighted the need for a more in depth understanding of conditioning interactions and better predictive tools for process and consumable design. We have previously noted a range of conditioning response as illustrated in Figure 1. As the figure illustrates, on some materials the surface height distribution imparted by the pad conditioner is shifted to the left relative to the exponential signature of the pad native porosity which is used as a reference. Denoted regime 1, these surfaces are characterized by the prevalence of “eyelid” features which obscure the line-of-sight path to the bottoms of the pores. In regime 2, the surface height distribution is more balanced, with a smooth transition between the exponential porosity signature and the conditioning signature. In regime 3, there is an abrupt disconnect between the exponential porosity signature and the conditioner signature. Such surfaces are characterized by deep gouges. We are investigating pad material tensile properties as a proxy to help understand and predict these conditioning responses. Figure 2 illustrates the range of typical tensile behavior for polymers. While Figure 2 invokes temperature as the variable that drives these behaviors, note that strain rate can also drive tensile behavior across this range with high temperature being analogous to low strain rate and vice versa. Figure 3 and Table 1 summarize some results from our testing of IC1000, VP5000, VP6000, IK4250H and IK4350L. We have determined that grooved pad materials at their nominal dimensions are adequate for testing according to ASTM D638: Standard Test Method for Tensile Properties of Plastics. Typical delivered pad thickness is within the specifications for sample thickness, and as long as the grooves are oriented to be substantially parallel with the strain axis (pulling direction) reproducible results can be obtained. A range of properties spanning at least about 2x is observed in these pads. Figure 4 gives one example of how this information can be directly applied. The figure illustrates a manifestation of the threshold behavior that is typically observed when conditioning with a pad conditioner with uniform tip height, where all the cutting points are controlled within a height range of about 10 µm. As opposed to a conventional conditioning platform, where there is a much wider range of tip heights, the uniform height-controlled conditioner exhibits a threshold behavior where no measurable cutting occurs below a limiting condition down-force (cdf). Figure 4 illustrates the disconnect between a high down-force, accelerated cutting test, using 14 lbs cdf and designed to return the “CR” value for a conditioner, and typical process cdf levels, in the range of 5 lbs cdf. Note that CR values are generally about an order of magnitude higher than comparable polishing tool level pad wear rates. Whereas conventional platforms yield a linear response between CR and tool level pad wear rates, the uniform height design exhibits an offset, such that at low CR levels, low pad wear rates are observed. The position of this threshold is a function of pad type. These data suggest that this threshold can be correlated with the mechanical strength of the pad material, as determined by the tensile test results. We have also explored the effect of strain rate on tensile response. Strain rates in a typical CMP process may be as high as 10 5 s -1 , and as shown in Figure 2, strain rate can have a significant effect on material behavior. Figure 5 illustrates typical results, illustrating the extremes of response for this group of pads, and Figure 6 shows the normalized response of offset yield and tensile strength as a function of strain rate. Of these materials, only IK4350L exhibits a divergent response, and as shown in Figure 7, IK4350L typically exhibits the most abrupt transition between the exponential porosity signature and the conditioner cutting signature, indicating a tendency towards the gross over-conditioning behavior characteristic of regime 3. Over the range of strain rates explored here, all of these materials can be characterized as “rubbery”. This rubbery behavior can be understood in the context of reported property measurements of CMP pads, which indicate distinct glass transitions related to the soft and hard phases of the polyurethane from which the pad is made. This analysis suggests that pad behavior is dominated by the “rubbery plateau” of the soft segment of the polymer, with the glass transition of the hard segment occurring above typical polishing temperatures. Figure 1