Stöber-process silicas (a.k.a. "sol") produced from
TEOS or TMOS have been tested for the final barrier/cap removal CMP step in
DD-Cu flows. Typically more expensive that precipitated silica but usable at
low solids contents, what performance improvements have been seen?
I have heard from several IC development people that eventually we'll need the higher purity of "sol" silica for barrier slurries. Is there evidence of performance improvement with less metallics in the slurry (e.g., colloidal has significantly higher concentrations of metals)?
A couple of years back we worked on characterization of smaller-particle, high-purity colloidal silica CMP slurries designed specifically for ULK dielectric layers. Comparative performance of 2 slurries was presented in terms of wafer polishing rate, NU, and particle defectivity employing different CMP tools, pads, and competitive slurries. These newly designed next generation slurries provided precise and consistent removal rates, minimal defectivity, and maximum planarity across the wafer surface. More details of this study (see attached file)were presented in our paper at CMP-MIC 08.
The selection of the right particle for CMP depends on both cost and performance. Ultra-high-purity silica made from the decomposition of TMOS is expensive, much more so than particles made from the neutralization of sodium silicate coupled with ion-exchange. While there sometimes is a modicum of polishing improvement with the TMOS particles, I believe the key driver for the use of this particle is fear of electromigration from the residual sodium ion. Referring to an earlier panel question about overspecification leading to higher costs, this could be a good example. I am not aware of any published rigorous studies that compare the reliability of chips produced from TMOS vs sodium silicate silicate.
One of the questions I wonder about for M1 and above regards the role of copper ion that gets released during CMP. I thought copper ion was also quite prone to electromigration. So why are we worked up about a little sodium ion when copper is all around? Any experts out there?
What about the surface morphology and size-distribution of sol vs. precipitated silicas? Ignoring subtle compositional variations, it seems that the sol process could be more easily controlled to get more uniform morphology and/or tighter distributions. Of course, any such hypothetical "gains" could just be overspecification...
I totally agree with Cliff. There is simply no solid evidence that particles from decomposition of TMOS result in devices that perform better. Regarding potential for EM, the amount of copper metal (itself an H2O2 decomposition catalyst and a fast diffuser) generated during copper CMP outweighs any dissolved metals that may be present in the POU slurry.
Rigorous ion exchange filtration of sodium silicate derived silica slurries may be a good and inexpensive way to go.
Ed you are right.TMOS derived silica has better PSD and can be made smaller in size with better control. But again, I have not seen much published information that this is indeed the way to go. At UA we did several studies on TMOS derived particles for copper CMP. I will try to dig it out and post it for comments.
There are pathways for synthesizing high purity silica that do not require TEOS or TMOS, but some of the businesses promoting these products have not been commercially successful. I have forwarded this forum to one expert in this area in the hopes that he will join the discussion.
A few years back I worked with a fab on two colloidal slurries for ILD. Both were precipitated, both the same pH, supposedly the same particle sizes, wt % etc. One was lower in metallics & the defects (over many product split lots) was better. Now, I admit that we didn't analyze the "standard", so it could be there was a larger particle size range, but it did lead to the thought that ionic contaminations can lead to higher defects. I can imagine that this might change the interaction w/ the particle and surface, or with particle & pad debris ... we never got to the point of proving anything ... but it was interesting. Anyone have any theories either way?
Regarding Ed's question on controlling morphology of sol particles-- the empirical evidence is - yes you can. Fuso has created a very interesting shaped particle for TMOS hydrolysis that they call cocoon-shaped which has beneficial polishing capability.
On the other hand, Nalco has also created particles with controlled shape, size, and surface using silicate-- for application in polishing, ceramics, paper, and ink-receptive coatings.
I suspect that a deep understanding of hydrolysis, nucleation, growth, and agglomeration of particles from either feedstock will yield dividends in terms of CMP materials of interest.
Silica particles manufactured from organic silicate esters using the Stöber process typically have a much smoother and harder exterior compared to particles made from aqueous sodium silicate using Iler's method. In my experience it is easier (but more expensive) to make particles having a denser morphology and more narrow size distribution using silicate ester raw material.
Having made a number of batches of both particle types, I would mention that there exists a tremendous potential for improvement in aqueous silicate particle manufacturing. It is useful to point out that particles from sodium silicate are typically manufactured in large quantities by industrial factories having a preponderance of industrial customers.
Producing really high quality particles from sodium silicate is not a trivial thing. Producing a narrow particle size distribution in aqueous silicate systems requires longer reaction times. Longer reaction (particle growth) times may also produce a smoother, less brittle particle surface but at a correspondingly higher cost. Improvements in reaction kinetics and transport mechanism will allow for faster batch times with improved surface morphology.
One interesting new development relates to the silica source material. Biogenic silica, extracted from a renewable resource such as rice hulls, will be used to make high purity silica particles in the near future and here's why: Millions of tons of rice are picked every year. The hull accounts for about 20% of that weight, the vast majority of which has become a huge landfill problem among the leading rice producers. Large industrial furnaces designed to burn rice hulls are now being put into use to fuel power generating plants. There remained the problem of the resultant ash disposal until it was found that 90% of the rice hull ash is pure silica. Sodium silicate extracted from rice hull ash is both cheaper and much cleaner compared to traditional sand-based sodium silicate. Biogenic silica can also be converted to silicate esters and silicon ingot with improved economics. Anyone interested in more information regarding the process and current IP situation should contact this author.
Hello Bill, thank you very much for your message on silica particle manufacturing. As always! Very insightful information. As you might have noticed, I did share our CMP-MIC paper with the group on this topic (comment #2).
I also enjoyed reading all responses and and as someone who has spent number of years working in this particular area of various sources of silica abrasive for CMP applications, I can only say that I totally agree with Cliff's suggestion below:
"I suspect that a deep understanding of hydrolysis, nucleation, growth, and agglomeration of particles from either feedstock will yield dividends in terms of CMP materials of interest"
What I would also like to add to this is better understanding of surface properties, surface energetics of various silica particles (TEOS-based, TMOS-based, Na silicate-based, etc.). It would be interesting to add biogenic silica to such a study as well.
From the metal impurity perspective, each end user has certain trace metal impurity requirements for chemicals accepted for various steps of the process and as long as a given slurry meets polishing requirements including trace metal levels it is secondary what type of colloidal silica is used in formulating that slurry.
Many thanks to Bill (#10) for joining the discussion with fascinating info about a new feedstock (and thanks also to Mike for inviting Bill to join in), and to Mansour (#13) and Cliff (#9) who remind us that process-structure-properties (classic Materials Engineering fundamentals) still set the dynamics of our work.
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
