As highlighted in an earlier blog by Dr. Len Borucki, our research team has investigated several key questions resulting from CMP wafer size scale-up by simulating hypothetical 300 and 450 mm processes on rotary polishers. Here, I’d like to describe how, in collaboration with Intel, we’ve managed to mechanically modify an existing 300 mm polisher to polish both 300 and 450 mm wafers, and demonstrate experimentally whether any differences exist in the tribological and thermal characteristics of the two processes. Even though we did not perform any removal rate tests, inferences were made regarding removal rate differences between the two systems by coupling their tribological and thermal fingerprints with highly successful kinetics models for copper removal.
In order to accommodate both wafer sizes, an interface module (480 mm in diameter) made of anodized aluminum is manufactured and attached to the carrier. The backside of the module is fitted with an o-ring which ensures a secure fit onto the carrier head using the existing vacuum conditions in the head. In addition, 300 or 450 mm compatible backing films and retaining ring assemblies are attached onto the front side of the interface module. The wetted backing film is used to securely hold the proper size wafer via capillary forces. Since our polishing pad has a diameter of 762 mm, and since no adjustments could be made to the distance between the center of the platen and the center of the carrier head, when processing 450 mm wafers, at any given time, approximately 13.7 percent of the wafer hangs off the edge of the pad and hence does not contact the pad during polishing. We did not think this was a major issue given the overall objective of our work.
All wafers are polished for 1 minute on a concentrically grooved pad at 3 pressures and 3 sliding velocities. Since the distance between the center of the platen and the center of the carrier head is the same regardless of the size of wafers being polished, the rotational velocity of the head and the platen need not be adjusted for a given linear sliding velocity. This is important to note since in a real 450 mm polisher, the distance between the center of the platen and the center of the carrier head would be larger and hence a lower rotational velocity would be required to match the linear velocity to that of a 300 mm process. For instance, a real 450 mm polisher would have a platen that is 1,094 mm in diameter. Therefore, to achieve a sliding velocity of 1 m/s, the larger platen would require an angular velocity of only 33 RPM compared to 42 RPM for the 300 mm system.
Experimental results, when adjusted for the lower platen speed needed in an actual 450 mm polisher, suggest that within the ranges of parameter investigated, the two systems behave similarly in terms of their coefficients of friction and lubrication regimes. Additionally, it is shown that the 450 mm process runs only slightly warmer (by about 1 to 2 C) than its 300 mm counterpart. A result that is not consistent with simulations reported earlier (the prediction in Len Borucki’s blog was for the two systems to have similar temperatures) possibly due to differences in the way the wafers were polished and/or the assumptions made in the numerical analysis.
Experimental data, coupled with copper removal rate simulations show that the wafer surface reaction temperature of the 450 mm process is slightly higher (by 1 to 2 C) than the 300 mm process. Consequently, simulated copper removal rates for the 450 mm process are slightly also higher (2 to 13 percent) than those of the 300 mm process at most polishing conditions. Our results indicate that when the current 300 mm CMP process is scaled up to 450 mm, the tribological, thermal, and kinetic attributes of the process remain similar and do not undergo significant changes.
Please contact us if you are interested in processing 450 mm wafers in our facility.
For a copy of our published article on this subject, please click here.