I do not know if I should place a comment here, but if nobody else is contributing let me place a quick comment. Back in mid 90-s when both Ara and myself were working for Intel, we at our FAB succeeded to introduce a process which was using only 40 cc/min slurry flow rate. I will never forget when some manufacturing manager looked what was polisher doing he almost stopped the run because he was under impression that the machine was mulfunctioning, because the -pad looked dry and the slurry was barely dripping. Our tests showed that the same polishing parameters could be achieved even with 30-40 cc/min.
The funny point was that with the new process the FAB could stay for more than 1 year with just some small amount of spear slurry we had on our shelves. Cabot could not understand what was going on there.
What we've modified significantly was the wafer carrier, which was also providing slurry spreading action and pressing it into the pad pores. It was also intensively "ironing" the pad, minimizing the impact of the pad debris and excessively large asperities. This process was on the manufacturing floor for a number of years, until the FAB has changed it product family and as per CE-rule had to bring in other fabrication precesses.
I strongly believe that proper wafer carrier modification can bring a lot of surprising effects in this direction.
Thank you for sharing your experience with hardware changes that can significantly reduce slurry use; fascinating that this approach was not replicated.
From first principles, this seems to imply that the whole industry has been operating CMP systems in only one of potentially several other general process spaces.
Does anyone else have experience with such radical reductions in slurry use?
Well, what was developed at Intel, stays at Intel. Also in that conservative organization it is hard to implement something radical. The ATD develops their process approach, starting with the vendor'd recommendations and if this works it is recommended for implementation to save the efforts for more complicated cases.
Believe me, it was a very significant effort for us to break the wall and implement this process. Several cycles of pilot runs to demonstrate that there is no yield impact. The big help came when the FAB hd to boost the output, and there was no space for additional CMP machines. Our process, which was called HEPP-2000, the "2000: was because it was meeting the year 2000 CMP road map targets, and that was in 1994.
Data has been generated showing that managing the slurry residence time (as well as the debris residence time) can have a significant impact on slurry consumption requirements. The Confluense Pad Surface Manager (PSM) conditioning arm retrofit vacuums the pad in-situ to control the slurry and debris residence time. Slurry flow rate reductions of 50% have been documented, with higher or equivalent material removal rates, and that may not be the minimum flow rate that can be achieved while maintaining the appropriate MRR.
In the discussion chain above, we have examples of two extremes -- using slurry for a single platen rotation only, and using slurry repeatedly over an extended period of time. I believe the window for extended use has passed, being subject to extremely tight topology and defect specs on sub-100nm technologies. If specs can only be met by using slurry for just a single pass under the wafer, I expect we will happily embrace it, and worry about breaking that embrace only after the fact.
The data I mentioned showing significant slurry flow rate reductions with equivalent or higher MRRs when using PSM is available on the Confluense website (www.confluense.com) and also on the NCCAVS CMPUG website for their 2009 CMPUG meeting at SCW in 2009. The residence time of slurry and debris can be controlled thereby transitioning the CMP process from a dilution mode to a replenishment mode, resulting in a more stable process and lower defects with a lower CoC.
It may be beneficial to review the article in Future Fab International (Issue 31) about the "Greening of CMP" to get a better understanding of the PSM concept and supporting data. Here is a link to the article. http://www.future-fab.com/documents.asp?d_id=4656&login=tried
I have attached an article on CMP slurry health maintenance considerations. In general, with careful application of optimum slurry health management practices during the slurry shipment, blending and replenishment, handling, metrology, filtration, and dispensing, it should be feasible to reduce overall slurry CoO as well as slurry waste generation. Also, the newer generation slurries, with relatively lower weight percentage of solids and advanced abrasives (also improved additives) are relatively more forgiving to the cumulative number of turnovers slurries may undergo before final consumption, without much change in the slurry chemical & abrasive properties.
To my knowledge not too many reports have been published on slurry recylce/re-use. There have been attempts but not too many success stories. If folks on this distribution are aware of any published work, it would be useful to post them here. Thx
A number of years ago, I published this work (see attached). My conclusions were as follows:
Analytically, compared to fresh slurry, the 5-times-recycled slurry showed: – A small drop in pH (11.00 vs. 10.85) – Negligible drops in viscosity (2.41 vs. 2.39 cP) and S.G. (1.0810 vs. 1.0775). – Relatively large increase in metal content ( 1 vs. 3 ppm) – Large increase in mean aggregate size (i.e. by 11 percent) – 50 – fold increase in LPC (but after filtration, LPC increased by only 3 – fold compared to the fresh & filtered slurry) Slurry filtration did not have any impact on the above metrics, except maybe for meanaggregate size Functionally, compared to fresh slurry, the 5-times-recycled slurry showed a 40 % drop in RR (too large to be due to the drop in pH) – Slurry mixing and chemical spiking must be adopted in future studies Near-perfect correlation was observed between RR and COF:– Repeated slurry recycling & filtration caused COF to drop – Likely due to observed increase in MAS & thus less wafer – abrasive contact area New spectral technique developed for determining the extent of ‘stick-slip’ during CMP: – Based on FFT spectral analysis of raw frictional waveforms in time domain – Subsequent integration of the resulting spectrum in frequency domain – Method showed that slurry recycling causes a reduction in ‘stick-slip’
– Demonstrates that an increase in MAS causes less wafer – abrasive contact area which lowers RR
We're now (January 20th, 2010) past the
official ending (the 18th) of this virtual roundtable discussion, after 4695 views of 129 replies to 18 questions. I'll
edit together Interesting discussions from most of the topics into a
summary document that will be posted to the Planarization Lounge.
We'll leave the topic posting open in case there are additional comments...but they would not be included in the summary.
Print
