 |
Air filtration is a both a cleanroom success story and a work in progress. On the success side, the number of airborne particles has been reduced significantly thanks to the application of 99.9995 percent efficiency filtration technology. Sub-Class 1 cleanrooms are now possible. This has even the most demanding users satisfied.
“The particulate matter is pretty much under control by the time you've got to five nines,” says David Hope, a corporate facilities technology development engineer with Intel Corp. (Santa Clara, CA). Hope is currently working on the company's 300-mm wafer effort in Hillsboro, OR.
But cleanroom air filtration also still has work to do in such areas as reducing outgassing, implementing new materials, and lowering operating costs. At the same time, air filtration vendors have to survive an ongoing wave of consolidation. Since clean air is central to cleanrooms, these are developments that affect microcontamination control in every industry that uses cleanrooms.
Pulled in different directions
According to a study done by the McIlvaine Company (Northbrook, IL), electronics is the largest consumer of high efficiency filters, representing about 30 percent of the world market. Within electronics are semiconductor and disk drive manufacturing, along with other operations. The second largest market classification, at about 17 percent, is what McIlvaine calls bioclean. This includes biotechnology labs and pharmaceutical concerns.
The growth rate for the high efficiency filter market is expected to be slightly higher than the eight percent growth rate of the overall filter market according to McIlvaine.
While both electronics and biotechnology industries use air filtration, their reasons vary. Semiconductor manufacturing demands the highest particulate filtering efficiency, almost exclusively for product yield and reliability. Biotechnology labs and pharmaceutical operations operate in higher class, and therefore more particle-laden, cleanrooms. In bioindustries, however, cleanroom operations are driven both by product concerns and government regulations requiring good manufacturing practices (GMP).
Just as there are differences in why air is filtered, there are differences in the makeup of the air filters themselves.
“What we sell to customer A and to customer B may have the same overall appearance, it may have the same performance characteristics, but some of the construction materials may be different,” says Knox Oakley, vice president for sales at Flanders Filters Inc. (Washington, N.C.).
Such material differences are driven by end user needs. In semiconductor manufacturing, minute amounts of boron and other dopants can kill a device. So, semiconductor filter users require that these be reduced. Filter manufacturers do so by using materials that have lower concentrations of these detrimental elements. Disk drive production, by contrast, is most susceptible to silicone. As a result, filters intended for those cleanrooms have to be silicone free. Because of the nature of their products, for the most part bioindustries don't care about such contamination. They do care about organics, however.
The state of the semiconductor union
Oakley notes that a semiconductor wafer fab may require 6,000 to 8,000 high efficiency filters. A pharmaceutical plant, on the other hand, may need 500 or so. However, Oakley also says there is little replacement business for semiconductor filters. Ironically, this is partly a result of the very cleanliness of semiconductor manufacturing operations: There are few particles generated to clog up the filters. As a result, there are wafer facilities that have had the same final HEPA and ULPA filters in place for 15 or 20 years. The low efficiency filters that handle the in-take air, however, are replaced regularly.
For bioindustries, on the other hand, manufacturing tends to be much less clean in a particle sense. High efficiency filters are replaced, perhaps at five year intervals in these industries.
Historically, semiconductor operations are the heaviest and most demanding users of air filtration. Over the years, chip manufacturing has pressed the state-of-the-art so that cleanrooms have moved steadily down in classification, from Class 100 to Class 10 to Class 1. But this march downward hasn't been the result of any revolutionary breakthroughs.
David Jensen is a member of American Micro Devices Inc. (Sunnyvale, CA) technical staff in Austin, TX, and former defect reduction technology program manager at Sematech (Austin, TX), the semiconductor manufacturing research consortium.
“The physics of filtration technology doesn't change. The ten-fold improvements-from Class 1000 to Class 100 to Class 10 and so on-we've seen every generation or every few years has largely been accomplished by moving more air, more uniformly through greater filter coverage, and by improving the isolation of particle sources-people, wafer handling, process tools, and such-within the cleanroom,” he notes.
That may be about to change. Jensen agrees with Hope of Intel that particles are no longer much of a concern in semiconductor cleanrooms. Incoming air is cleaned through a series of pre-filters before hitting the final filters. Besides that, most of the air circulating in a cleanroom is return air. Inside the cleanroom, tools and people have been reduced significantly as a source of particles.
As a result, the consensus seems to be that airborne particles are not a major concern. This is true even for the most advanced technologies, such as those at 0.18 micron feature size and smaller. However, with the shrinking feature sizes, circuits have become more susceptible to airborne molecular contamination (AMC). Traditional filter technology cannot deal with AMC because the contaminant is molecular in nature, rather than particulate.
“What you can do is minimize generation of the AMC from the filter itself, and that's becoming quite important,” says Hope.
The medium is the message
 |
To paraphrase Marshall McLuhan, for semiconductor filters the medium is at least partly responsible for the mess. The standard in filtration media has long been a glass microfiber. This presents a problem. According to Ming Huang, marketing director for global air filtration business at media maker Lydall Technical Papers (Rochester, NH), the borosilicate microfiber is relatively inert and resistant to many chemicals. However, hydrofluoric acid, used extensively in semiconductor cleanrooms, attacks the filter material. This releases boron, freeing it to wreak havoc in semiconductor devices.
As device dimensions shrink, the level of allowable trace boron also shrinks. The challenge of such airborne molecular boron is made worse by another semiconductor trend: minienvironments. This means any escaped boron is now in a small volume of air, raising the concentration, while at the same time being closer to both acid sources and wafer sinks.
There is, however, another media, expanded polytetraflouroethylene (e-PTFE). Camfil reached an agreement last year with W.L. Gore & Associates in which the two companies signed an exclusive cooperation agreement. Unlike glass microfiber media, e-PTFE is resistant to environmental assaults and is boron free. That's important to semiconductor makers.
However, for all cleanrooms e-PTFE filters also offer lower pressure drops and therefore greater energy efficiencies. The medium also has other advantages.
“The e-PTFE media is very resistant to handling. A paper-based HEPA and ULPA product is somewhat fragile and susceptible to handling, touching, transportation, and freight issues that may create damage or leaks, notes Doug Holzherr, vice president for sales and marketing at Camfil Filtra.
The downside is that e-PTFE is much more expensive than microfiber glass. According to Holzherr, the ratio between the two runs at three or four to one in favor of the microfiber glass.
Traditional filter media manufacturers have responded to the e-PTFE threat by making low-boron products. While reducing the boron makes the media more brittle, the manufacturing problems created by this have been overcome.
Consequently, low-boron filter media have met with market approval.
“We do have low boron products which we have sold against membrane quite successfully,” says R. Vijayakumar, director of marketing for high efficiency and specialty filters at the Hollingsworth & Vose Company (East Walpole, MA).
The i300i 300-mm wafer guidelines allow only e-PTFE and low boron microfiber glass in semiconductor minienvironments. Also, manufacturers are working to lower the cost of synthetic membranes. As a result, synthetic membranes like e-PTFE may become the material of choice. Huang, for instance, reports that Lydall Technical Papers is exploring the possibility of using synthetic media.
However, he also notes that research and development continues into microfibers. For instance, over the last five years, the pressure drop required to push air through microfibers has fallen 14 percent. That's an area that Lydall is continuing to work on.
“Operating cost is a major issue to operating the cleanrooms. From the glass media companies have already pushed to the cutting edge of filtration technology, we can't do more for the filtration we already pushed to the cutting edge of the technology. But we can do a lot on lowering the pressure drop,” says Huang.
Monitoring is next to cleanliness
Somewhat related to the question of media and filter construction is the trend toward lifetime filters. These form the basis for air filtration systems that require little scheduled maintenance or unexpected downtime. This is being driven by a number of factors, chief among them the increasing cost of semiconductor tools.
With the switch to 300-mm wafer sizes, tool cost is expected to increase significantly, doubling or even more. Therefore, the dollar output of goods from the tool also has to go up. Devon Kinkead, president and CEO of Extraction Systems Inc. (Franklin, MA), puts the monetary output of a deep UV scanner at about $150,000 an hour. Extraction makes chemical filters designed to remove amines and other AMCs from semiconductor cleanrooms.
Because of the cost of any downtime, Kinkead says his company is creating chemical filters good for three to five year. That's longer than the depreciated life of the processing equipment.
However, behind these lifetime solutions lies another problem. That's pointed out vividly in the 1998 update to International Sematech's International Technology Roadmap for Semiconductors. Table 58 of the document shows that the time to find the source of a problem must shrink from days today to hours by 2002. Likewise the time to recognize a trend must go from last year's weeks to tomorrow's hours. The rationale behind this time compression is that the value of a 300-mm wafer is quite high. Therefore, allowing very many to be produced with any defects is extremely costly.
For air filtration, especially if filters are never to be changed, this means that detection of a failed filter has to be in near real time. That presents a challenge to both the filter vendors, instrument makers, and their customers.
“If you have an in-situ, real-time monitor that is sensitive enough, you can begin seeing degradation and the trends ahead of time, and you can deal with it. That presupposes you've got a monitoring system, probably a third or better lower than the process threshold,” says Extraction System's Kinkead.
Minienvironments and MegaFabs
A final issue, one that affects all cleanrooms, is the ongoing consolidation in the filtration industry. That can be seen in Camfil's acquisition of Gore's filter business. AAF International, once an independent company, is now part of Malaysia's Hong Leong Group. Even Flanders is consolidating.
This is being driven by the increasing globalization of business. The end users of high efficiency filtration, semiconductor and pharmaceutical companies, are increasingly global. A facility may be designed in the U.S., built in Europe, and duplicated in Asia. In such cases, the company that is the end user of the filters probably doesn't want to buy them from a number of different geographic sources. Instead, what is desired is are global sources. Also, the technical requirements are such that smaller operations have trouble competing today. That will be even more the case in the future.
For end users who are concerned with microcontamination, this brings the benefit of added reliability and increased technical expertise. At the same time, it cuts down the number of available choices.
This convergence may be helped along by the switch to minienvironments. Not only are such enclosures showing up in semiconductors, but something similar is happening in the pharmaceutical industry. According to Hope of Intel, a fab with minienvironments will require only 4,000 or so filters versus the 7,000 needed today. Filter coverage will be only 50 percent outside of the minienvironments, and there will only need to be one filter per minienvironment. That and the trend toward fewer, yet larger facilities in all industries, may mean that the market may converge on only a few suppliers. But for those fighting microcontamination, that may not be a bad thing.
As Holzherr of Camfil says, “The economies of scale and the ability to offer a broader and more diverse range of products is the driving force. You might even use the term total filtration management. The ability to manage the filtration needs of key customers is the important service that needs to be offered both currently and for the immediate future.”
by Hank Hogan