In conventional CMP, asperities on the pad surface contact the wafer, and with the assistance of polishing with slurry containing chemistry and abrasive particles, remove material from the wafer surface. Because polyurethane (PU) pad asperities are irreversibly deformed due to wafer contact and are also abraded by slurry particles, the pad surface must be continuously renewed with a diamond disc conditioner in order to sustain a stable process. Diamonds cut the pad surface at a slow rate, eliminating old asperities and creating new ones. Conditioning also gradually thins the pad, which eventually forces its replacement. As such, conventional CMP processes have several weaknesses including: (a) large amount of waste due to frequent replacement of pads and diamond disc conditioners, and (b) poorly controlled pad asperities and highly variable contact area distributions which cause wafer to wafer variations in material removal rate and increase in the number of wafer-level defects.
Researchers at Araca, in collaboration with the University of Arizona and MIT, have developed a novel solution that eliminates both the pad and the conditioner. The term “pad-in-a-bottle”(PIB), coined by our CTO Dr. Len Borucki, and taken after “genie-in-a-bottle” replaces the pad asperities by high quality micron-size PU beads (Figure 1) that are comparable in size to conventional pad asperities. The beads are suspended in the polishing slurry (and hence bottled up) using a surfactant and come into contact with the wafer surface by a means described below to promote polishing in much the same way as conventional asperities.
Since we can select the size and the concentration of the beads, we are able to achieve better control of the height, curvature, and area density of the “summits” that come in contact with the wafer. This substantially reduces the process variability associated with conventional asperity contact. Use of beads requires a second surface, or a counter-face, for polishing to occur. Active beads (those involved in polishing) are compressed between the wafer and the counter-face and slide or roll along the wafer surface. The counter-face material is made of polycarbonate with a mildly roughened surface for providing some “teeth” for temporarily holding the beads in place to allow momentum transfer to the wafer. Figure 2 below describes our novel method.
From an environmental perspective, pad debris in the liquid waste stream that has a distribution of sizes ranging from submicron to tens of microns in the conventional CMP processes is traded for PU micro-beads in the waste stream that have a much tighter size distribution. The latter will therefore be more easily filtered or made to float, via anti-wetting agents, thus allowing easy recovery of the beads. In addition, it is possible to hold the mass fraction of the beads to a value no greater than the mass fraction of pad debris currently in the waste slurry. Therefore, the rate of waste production in the modified process will not increase over the current process. Controlling the size and usage rate, and filtering or removing the waste, will help ensure that the beads do not enter downstream water supplies.
Since the modified process also does not require periodical disposal of two-thirds of the original pad (including the sub-pad and adhesive layers), we will produce a net reduction of the PU usage by 3X. Micro-beads are also cheaper and much easier to produce and transport than PU polishing pads, so there are additional economic and environmental benefits to their use.
It must be noted that pads, which are 78 or so cm in diameter, are strongly adhered to the platen as a result of wafer pressures exerted on the pad during the 2,000 or more polishes. As such their removal from the polishing platen can cause back injuries if not done properly. Furthermore, as 450 mm wafer CMP requires pads that are 115 cm in diameter, replacing the pad entirely with PU beads removes the current and future ergonomic concerns entirely.
Lastly, with conventional pads, both the asperities as well as the land and valley areas act as absorption sites for unwanted and environmentally harmful polishing by-products such as copper, gold, ruthenium and in the future, arsenic (for compound semiconductor CMP). These elements remain within the polymer matrix after the pad has been decommissioned and discarded in landfills. On the other hand, the beads, having a much lower number of absorption sites for the metallic by-products, will likely force most of the by-products to remain in solution and be subsequently treated by electro-coagulation and other chelating, sedimentation and biological means.
Similarly, the modified process will replace a technologically complex diamond conditioner that uses artificial diamonds and is made using high EHS footprint processes that include electroplating, sintering, brazing or chemical vapor deposition, with an environmentally friendly soft brush to clean the counter-face plate.
At the upcoming International Conference on Planarization/CMP Technology 2015 in September (Phoenix, AZ), Prof. Duane Boning of MIT will present (for the first time ever in a public forum) our latest results re: PIB technology used to polish blanket 200-mm oxide wafers with PU beads with median diameters of 15 and 35 micron that were dispersed in a CMC SS25® slurry with a surfactant known as Silsurf®. In these studies, the PU bead content was kept constant and a polycarbonate counter-face with concentric grooves was used. Oxide removal rates of up to 150 nm/min were observed. Previous blanket wafer PIB modeling had indicated that beads must be stacked (rather than monolayer dispersed) in order to achieve this usable removal rate.
Patterned SKW3-2 STI test wafers were then polished using PIB and a new model was fit to measured dishing and erosion data. Based on the extracted model, full chip dishing and erosion simulation studies were performed to understand factors that were dominant in achieving dishing and erosion at nitride clear time, compared to conventional CMP. The stiff counter-face was found to be most influential in achieving improved within-chip uniformity as compared to conventional CMP.
Currently we are looking for an IC maker and/or a slurry supplier to enter into a JDP with us to further improve, commercialize and proliferate our PIB technology for insulation and metal CMP applications.