About thirty five well-polished professionals (…pun intended…) attended the May 16th meeting of the NCCAVS CMP Users Group at SEMI headquarters in San Jose. The agenda comprised several topics with the overall theme of CMP consumables and their management. Presentations will be posted in the coming weeks at CMP User Group Proceedings.
The meeting kicked off with Mike Fury (yes, he’s related to me) of Vantage Technology speaking on "The Sampling Statistics of Low Slurry Particle Counts." Slurry contains millions of particles/ml below 0.5µm by design, with a rapid decrease in particle counts as the particle size increases above that. In a well-managed slurry, low particle counts are expected above 1µm. Line control using periodic slurry sample sizes of 1ml down to 0.25ml provides a good statistical representation of small particles which are plentiful, but the sampling error for large particles quickly increases over 100% for average particle counts lower than 10 particles/ml. Periodic sampling and dilution measurements do not provide a good statistical representation of large particle counts (LPC). Continuous sampling improves this representation considerably. More importantly, the availability of continuous LPC data allows the implementation of six-sigma line control practices to identify, track down and eliminate the root causes of the LPC drifts or spikes to ultimately reduce scratch defect yield losses.
Paul Bernatis, DuPont EKC Technology, presented "Tackling Critical Requirements for Advanced Post-CMP Cleans" and described the design of EKC PCMP 5600, a 4th generation PCMP cleaning solution product. The challenge for formulators continues to be a vanishingly small process window in which the chemistry can perform all of the cleaning functions required without contributing to corrosion or delamination. Impact on TDDB was studied closely in several copper / low-κ dielectric systems. Electrostatic attraction of residues to the wafer surface is another issue that needs to be moderated by the formulation. Electrochemical studies showed the successful avoidance of corrosion conditions in experimental Cu on Co barrier systems. Mitigation of Cu dendrite growth after processing was based on prior DuPont experience in the prevention of Ag migration.
John Mendonca, Asahi Glass Corp. spoke on "Front-end, back-end and substrate slurries: Key process data", providing an overview of the ceria and silica slurries and contrasting their key properties. Ceria is finding applications beyond STI, including ILD, polySi, poly stop and backside TSV. Emerging applications include sapphire and SiC polishing. Ceria abrasives are migrating from a D50 of 0.18µm down to 0.14µm. The poly stop slurry has a TEOS/poly selectivity >500:1. A silica slurry for sapphire has been used successfully with continuous recycling and reuse.
Floyd McClung, founder of CV Nanotechnology (CV being Castro Valley, CA and not capacitance-voltage) talked about his current incarnation, "Advances in optical polishing abrasives and slurries." Semiconductor CMP and silicon wafer polishing, after all, had their roots in optical polishing. Slurry recycling is common, slurry filtration is not, and chemistries are much simpler and less aggressive than chip CMP. Ceria is the most commonly used abrasive for its high removal rate on the silicon oxides that are most commonly polished. The most widely used PCMP solution is Dawn, the same stuff as under your kitchen sink. Some of the slurries used are quite large, with D50 ~0.8µm and a tail beyond 5µm. Applying the process advances in CMP to the optical polishing industry is likely to prove beneficial to them. To hear Floyd describe it, retiring to optical polishing after a career in CMP is a bit like kicking your feet up on the porch with a cool adult beverage in your hands.
The organizers may not have realized that the speaker lineup for this meeting had a heavy bias toward EKC genealogy, as Floyd and I also have roots there, as did a few attendees. An alumni reunion was hastily organized and enjoyed by all.