What is the best filtration for a slurry?

Hello

that's right, slurry tend to settle and agglomerate with time, depending of the particle type (Ceria, alumina....) shear can also have a major impact on your LPC. Filtration is fundamental to protect wafers, improve defectivity performance and have a stable process.
Filters for slurry are basically a depth graded filter with polypro media, wrapped or pleated....
Those filter are available in different size, cartridge or disposable.
Most important parameters are retention efficiency and pressure drop with DI..... Best way to choose a filter is to have a retention curve, this will provide retention efficiency for different particle size. A filter for CMP is not absolute even if you can have 99.9% retention for a certain particle size.
Some slurry may have chemical component who could affect filter lifetime (surfactants....) this is why the best way to choose the right product is to contact your prefered supplier :) and discuss with him.

Amont parameters to consider, we can list:
Solid Weight Content
Average particle size & particle distribution curve (D50, D99)
Particle type: as their stability and their propension to aggregate and agglomerate vary a lot, depending on material & shape.
Flow rate
Pressure budget (some filter designs request a minimum loop pressure to reach an optimized life time).

As a filter supplier we often also have to take in account the cost. End user have the tendency to put this parameter at the top of the list.

Then the following difficulty occurs: Facilities and POU (Equipment dep’t) are 2 different cost centers in the fab. It is almost impossible to satisfy one without to disappoint the other when we propose a filtration solution.

If you propose poor retention efficiency at the facilities, you will often need a good retention filter at a low retention rate and then you will have very low filter lifetime at POU and very long lifetime at the facilities...

If by chance you manage to persuade the end user to consider filtration as a whole from facilities to POU, then you can work on a proper solution:

The first step would be to obtain a slurry quality at POU equivalent to the slurry quality as out of the slurry manufacturing site.
One parameter that would be also important to know is the final filtration rate at slurry manufacturing site with the retention curve of the final filter in this slurry.

Then it is important to identify the weaknesses of the bulk distribution system:
Sensitive locations that will generate particles or agglomerates like pumps, steep elbows, slurry dilution, etc. and place appropriate filtration at this locations.
At POU, The filter retention curve should be as much as possible similar to the slurry manufacturing site filter.

The second step would be a fine tuning of the filtration set by in line particle monitoring of the slurry and correlate with the defectivity rate on the wafer. This will enable to identify the critical particle size that will generate micro scratches. Then you will have the possibility to fine tune the filtration set.

What is important to retain about filters is:

– The retention curve should be as steep as possible
– Retention curves better describe a slurry filter than a mere Retention Rating information
– Retention curves depend upon the abrasive : Filters behave differently with different slurries

Thanks Arkham for your contribution.
I want to remind to the team that the purpose is also to collect papers on the subject. If some of you have papers, please, put them online ;)
alain

to comment Arkham's answer, the problem is that some filter suppliers comes always with normal or absolute ratings. They take advantage of the ignorance that these retention curves do exist and also that filtration is not something easy to catch.
IC makers should compel retention curves for each kind of filters to all suppliers.

alain

A few words about absolute ratings in slurry filtration: there is no such thing...

Illustration: a given filter (polypropylene depth structure) may remove 99,98% of AC Fine Test Dust, which has been used for 30 years in the biopharm industry to test filters. In this sense, it is "absolute" for Pall (Beta ratio of 5000), very good for Entegris (LRV of 3.7) and for other filter suppliers (99.98% retention).
The same filter, when tested with Cabot iCUE 5306 Copper slurry (Alumina particles) removes 92% of these particles in the same channel.
The same filter, when tested with Cabot SS12 Oxide slurry (Silica particles) removes 85% of these particles in the same channel.
The same filter, when tested with Polystyrene Latex Beads (PSL), which have been used in the semiconductor industry for 15 years to test filters, removes 65% of these particles in the same channel.

Hence the questions:
1°) what rating should the filter manufacturer put on the label?
2°) how relevant will this information be for the semiconductor fab in its actual process?
3°) if we are sold this filter as an "absolute" filter, and use it in Cabot SS12, we won't get the expected results on the tool.

As a consequence, this needs clarification. The fab engineers cannot trust the basic information coming from filter suppliers. They need to discuss the application (what slurry, what problem they are experiencing, setup etc.) with their supplier.

Another problem associated with filter efficiency: results depend on the particle counter. It is rather empirical...

Pierre, is it possible to have graphs on the test you are mentionning? (if yes, send them to [email protected] and I'll try to put them online)

Your remark is very interesting. Last year I did a presentation on the filtration of an silica slurry and as I mentionned, the ratings are only to give us where we would like to go and only tests can say if we have found the good size. But, it is also true, that end users needs to know very well the slurry (its behavior...) they use to make sure that the filter size will fit the expectations.

Another point, I totally agree that we cannot trust basic informations on filters BUT, I do not appreciate suppliers that comes and says that the rating is the good one and tries to convince you that a 0.5µ filter will absolutely stop everything equal or higher than 0.5µ at 99,99% (especially when this filter is a depth filter)...it doesn't make sense (I didn't want to say: it's a lie). To continue: I did tests on these filters and the 100% retention curve is arround 0.8µ and arround 65% at 0.5µ!!!

So, it can be a good opportunity for filter suppliers to start to have informations on tests done on several slurries in the same channel.
Then, an evaluation can be started as an...empirical manner to defined the good size.

Hello

I'd like to clarify the comment of Pierre.
First of all, the example Pierre took, are some tests made by Entegris. What he did not mention, and I think its really important its the right context of these slides.
Those slides have been made by Entegris to provide insight and training to customer and highlight the difference of retention with the same filter upon the particles type. And finally explain the importance to choose a filter with the help of the supplier and not only based on retention since there is no standard on retention rating attribution.

Entegris never characterise filters with ACFTD method which is not accurate for retention but only for loading efficiency. PSL method will provide much more accurate information about retention efficiency however the %retention will be higher in a slurry since the number of particles is higher. PSL method will give an idea of the minimum of retention at a given particle size.

Last point, if you refer to my first post submitted on June 3rd, I clearly stated that absolute filter doesn't exist for CMP as filters are graded depth and my recommendation was to have a discussion with the supplier.

I stay at your disposal if you want further informations.

Gilles

As a filter supplier I felt I should contribute to this discussion. I agree that it is important for both slurry suppliers and end users to discuss the specific application with the filter supplier before determining the proper filter for that application. Much of the marketing literature being provided by the filter suppliers today for CMP applications is taking this into account. Product recommendations are being made based upon the applications (or the abrasive type). Some companies are including retention curves in their literature while others are not focusing on retention (other than to state a micron rating) since it is now known that retention is dependent upon many factors including such things as abrasive type; solids loading; and chemistry. The key point is that filter suppliers do have insight into the performance of their products with various slurries and with this knowledge they can provide their customers with valuable information when it comes to a specific product recommendation for a given slurry application.

Although I cannot speak for all individuals that work for filter suppliers in general I can say that no malice (or deception) was meant by filter suppliers when they make comments or claims regarding their filters. They should know how the company rates their products and should be able to explain it. In my case since I work for Pall I can explain the fact that we have traditionally rated our filters using a standardized test which includes a modified version of AC Fine test dust. A known concentration of this "slurry" was used to challenge the filter and particle counts taken both upstream and downstream. Based upon this test method our filter retains 99.98% of particles at or above the stated micron size of the filter. I realize this has no bearing upon the performance of a given filter in a CMP slurry. This rating method was simply used to compare the performance of Pall Profile II filters as they were developed years ago. However, filter suppliers will have data that shows how their products perform with various slurries. It is from these data (along with field information) the present recommendations for various slurry applications come.

All filter suppliers will agree that retention curves are certainly more meaningful than say a beta 5000 rating when it comes to slurry filtration. However, it is also important to note that retention curves can be arbitrary when it comes to actual performance in a slurry also. Retention curves provide insight into the steepness of the curve but are also dependent upon the test method. What type of abrasive is used? How does the particle size distribution compare with a silica (fumed or colloidal); alumina; or ceria slurry? Was the filter primed in any way to either improve or reduce its retention? For example; if you perform a 200nm latex bead challenge on a submicron depth filter your retention will depend upon the amount of surfactant used in the bead solution. In my case I found that a Pall submicron Profile II filter retained nearly 90% of 200nm latex spheres when I merely challenged the filter with the bead solution. I took the same filter; submerged it in the surfactant for wetting purposes and retested it. The retention of the same 200nm beads was now 25%. Tests can be manipulated to obtain more favorable results.

I believe the key point here is for customers to ask questions of the filter suppliers when it comes to CMP slurry filtration. Please also understand that some people in the sales force may not be totally aware of all these intricate details (although we are trying to train them) so don't assume they are trying to deceive you. Filter suppliers are finally able to develop products specifically for CMP because our customers are finally providing us with the necessary information (related to both particle and chemistry) to be successful. The key to future improvements in the area of defect reduction is more and better collaborations between the filter suppliers and their customers (slurry suppliers, OEM, and end users)

I hope you found my comments relevant

Pat

It can be an opportunity for slurry makers to provide the filtration that can fit expectations, but this must not shut in the end user in specific filters. Slurry makers should come with several kind of filters.
But, IC makers are moving more and more to corporate level on consummables to have better price and this means that several fabs of a same company will have the same filtration for the same slurry.
The danger is that each fab are different and a tubing longer than another one, or a sharp elbow in more on a loop can disturb the slurry and generate particles. What I mean by this is that a filter can have a very good behavior for a slurry in a fab but can be a danger in another fab, just because the slurry loop is ... a little bit different.

So, this posting should be closed next week, and I'm really glad to see two filter suppliers online. Friday, I will try to have others involved and especially IC makers.
I will not do the conclusion today, I think that there is still comments to be brought. BUT, I'm still hungry of what my suppliers can bring me terms of informations and mainly how to bring solutions (tests, qualification, where to locate filters compare to the loop, and so on...)

The discussion is still open, waiting your comments...

Bests

Alain

Lets conclude on this posting: The most two important filter supplier have, by their comments, showed us the difficulties to find a one way rating.
It means that it's up to the end user to makes his choice. If ratings are important to fix up minds on what to do, it's difficult to convince people who are not used to jonggle between Beta ratio, LRV, Efficiency, Retention Curve, and so on when you review a qualification plan.
My example: I have taken three filters from three different suppliers (that I will not name by respect) with the same making (depth filter) and the same rating : 0.5µm. When you look at retention curves provided by the suppliers (with one of them, I had to battle to have it...), these curves started to be quite different at the 95% level to be extremly different at 100%. The results of the defect level were different from one to another as well. The choice was made on the filter that the retention curve matched to the defect level.
My viewpoint is that ratings are directly linked to retention curves.
Bests
Alain