300mm contamination control data points to concerns in the tools, the infrastructure and the silicon itself
by Hank Hogan
The world's first high-volume production 300mm wafer fab has already been built on paper. It's a computer model of a factory churning out 20,000 300mm wafers a month, using a generic 180 nanometer linewidth process flow and standardized tool performance. Constructed by International Sematech (Austin, TX), the model incorporates an automated material handling system (AMHS), complete with front opening unified pods (FOUPs), for wafer transport between and within process bays.
Initially an AMHS research tool, the model doesn't include such things as the impact of contamination control. For instance, the material movement assumes no segregation zones, such as might separate doped from undoped areas. Nor does it answer such manufacturing questions as ” How frequently do you swap out your FOUPs? How frequently do they have to be cleaned? Those things are not currently included in this model,” notes Jim Ammenheuser, a program manager at International Sematech.
Accounting for contamination control requires not only a change to the model but also data. Fortunately, 300mm contamination control data is starting to emerge, pointing out areas of concern in the tools, the infrastructure and the silicon itself.
Pilot line lessons
Several full-blown 300mm production lines have recently been announced, but today there are only two pilot lines running to supply the missing contamination control data. The lesser known of the two is operated by Samsung Electronics in Kiheung, Korea. That anonymity is by design.
“Information regarding our 300mm business is quite restricted. We are operating a pilot line here in Kiheung, and this year we will be going into phase two of pilot operation. And our mass production line is mapped out for 2001 or 2002,” says Sunghae Park, a spokeswoman for the semiconductor business sector of Samsung Electronics.
The other pilot line is considerably better known. Dubbed Semiconductor300, or SC300, it's a joint project between Motorola Inc. (Schaumburg, IL) and Siemens spin-off Infineon Technologies AG (Munich, Germany). Located in Dresden, Germany, the 300mm pilot line uses FOUPs and other components of the automated material handling system. The line also uses 300mm tools, but it does run inside a 200mm factory and an ISO Class 3 (Class 1) cleanroom.
With about 1,800 square meters (20,000 square feet) of cleanroom floor space, the 300mm line manufactures 64-megabit dynamic RAMs on about 300 wafer starts a week. The product was qualified in September 1999 and is being sold. All SC300 will say publicly is that the yield of the 64-megabit memory chips is above 60 percent. Plans call for the production of 256-megabit devices, as well as scaling down the process geometry to less than 0.18 micron.
There's good news and bad news in what little SC300 will say about contamination control. The tools contribute most of the contamination, while the FOUPs and other infrastructure add little to the load. That outcome isn't entirely unexpected.
Clint Haris, an automation engineer at SC300 notes that operating a FOUP is like opening and closing a door. The key to minimizing FOUP-generated particles is to ensure a proper and abrasion-free fit of the “door” into its frame. A process tool, on the other hand, either deposits or removes a film. Therefore, a FOUP is inherently cleaner and generates fewer particles than the process too1.
Contradictory tool results
For instance, Haris works with tool manufacturers on minienvironments, which use filters and air flow to suppress contaminants. Haris reports good progress in a crucial area: the delivery of a minienvironment that meets manufacturers' needs for environmental and contamination control.
“The tools have evolved to the state where today, usually, purchased tools have minienvironments that meet our needs,” says Haris.
That doesn't exactly agree with re-sults from Pentagon Technologies (Fremont, CA). According to Dick Dryden, who founded the Dryden Engineering component of Pentagon Technologies, some 30 to 80 percent of all minienvironments fail some specification during initial testing. Dryden says these failures may include particles being deposited on a wafer. But other problems include pressure differentials that allow outside air into the minienvironment. There may also be outgassing of materials or eddy currents.
However, these failures may not result in contamination. Nor do the failures mean that the minienvironments don't meet a semiconductor manufacturer's needs. For that reason, it may well be that there is no discrepancy between what's seen at SC300 and elsewhere.
Other results are harder to reconcile. For instance, Semiconductor Leading Edge Technologies Inc. (Selete) of Yokohama, Japan, is engaged in 300mm tool characterization. Some of this involves contamination control testing. In January, Selete reported problems with process equipment. One area cited particle and metallic contamination in chemical vapor deposition (CVD) systems and wet bench particle control.
But not everyone agrees that these tool contamination problems exist. “We haven't seen any of that,” notes Doug Keenan, 300mm manager at Motorola's advanced products research and development laboratory (APRDL) in Austin, TX. Keenan and his group are heavily involved in the SC300 pilot line.
“I was surprised to see that, although on the other hand, we're involved with Selete and we've had very good results there,” echoes Dan Syverson, product marketing manager for the surface conditioning division at FSI International Inc. (Chaska, MN). Syverson notes that Selete cited six or seven primary obstacles in 300mm tools, not all of them contamination related, in its report. When contacted, Selete declined to comment.
Trapped in a FOUP
Another area where contamination control data is emerging is in the infrastructure, such as the AMHS and the FOUPs. SC300 started with 13-wafer FOUPs, but moved in September 1999 to the 25-wafer variety. This was done largely to improve manufacturing efficiency.
For their part, FOUPs shield wafers from contamination. Anecdotal reports indicate that when properly sealed, FOUPs and related enclosures can protect wafers for weeks at a time from ordinary, unfiltered, particle-laden air. Stories from 200mm technology tell of engineers leaving wafers sitting on their desks for extended periods with no ill effects. However, there is some concern about the FOUPs themselves in the area of airborne molecular contamination (AMC).
“The FOUP is usually made of polycarbonate (PC), and it produces AMC outgassing. Also, since the FOUP is sealed, it can trap the AMC or other contaminants on its walls from the processed wafers or process equipment, if a tool's minienvironment is not working,” says Randy Goodall, associate director of productivity and infrastructure at International Sematech.
Goodall notes that tests for AMC generated by the FOUPs themselves is one of the things his group plans in its upcoming carrier project test phase. Others see problems with AMC due to the use of FOUPs.
“They're going to be more receptive, have more exposure, to the airborne molecular contamination at that level because of the smallness of the volume and the choice of materials,” predicts Dryden of Pentagon Technologies.
However, there are no reports from any of the 300mm pilot lines of AMC problems. Why this is so is unclear. Dryden thinks that the problem will only crop up in large-volume manufacturing, something the pilot lines cannot do. A typical solution for AMCs is the use of chemical filters. Putting these inside the FOUP itself would be costly. For that reason chip manufacturers are likely to search for materials that don't generate molecular contaminants, if they do indeed turn out to be a problem.
A somewhat related concern is how to tell if a FOUP is clean. Today's accepted method, according to International Sematech's Goodall, is to use a wet test. Liquid is poured into a FOUP, the container is given a few shakes, and then a liquid particle counter is used to count what's in the liquid. Aside from being a somewhat messy affair, the method suffers from a lack of repeatability. Since the liquid washes out a majority of the particles, there's no way to redo the test.
Keenan of Motorola notes there's another solution to the problem.
“If wafers stay clean in the FOUP, then the FOUP is clean enough,” he says.
Counting the number of particles a FOUP puts down during repeated open and close cycles on a set of ultraclean test wafers establishes a baseline. After that, the same test can be repeated as needed. A rise in the count indicates a problem with the FOUP, which then could be cleaned. The whole process could then start over.
Through the barrier
The other area of concern in the infrastructure is cross-contamination. In theory, a FOUP could dock with a processing tool, pick up a contaminant either from the tool or wafers, and then transfer it to the next tool or the next wafer batch. The classic example of this cross-contamination is in doped and undoped process areas. It's long been known that doped material, typically at the back end of the line, has to be kept separate from the undoped areas at the beginning of the line.
For tomorrow's technologies, an added requirement may be to keep copper metallization out of the non-copper areas that are found earlier in the manufacturing flow. Currently, the SC300 pilot runs an aluminum metallization process. So there's no need to worry about copper contamination.
The accepted approach is to segregate areas and put material pass-through barriers in place. The wafers can enter and exit the area, with positive air pressure, physical barriers, and particle and chemical filters to ensure that nothing else does. The 300mm facilities will differ from earlier wafer sizes only by the presence of FOUPs and a great deal of automation. In practice, neither wafers nor FOUPs will be allowed to travel freely from a dirty zone to a clean one.
Unlike earlier technologies, however, 300mm will not be able to swap carriers because the FOUP itself is the carrier. That's a problem that PRI Automation Inc. (Billerica, MA) is working on. PRI supplied the automation system to SC300.
“If it's 300mm, you can do a batch wafer transfer,” says Joe Reiss, director of product marketing for the factory systems division at PRI. “The only thing that moves are the wafers between one zone and the other. The FOUPs remain within their area.”
That could have an impact on cycle time through the factory, as well as driving the need for more robots, stockers, and wafer staging areas.
On the flip side
There are some contamination issues that relate specifically to the 300mm wafers themselves. One arises because 300mm wafers have a polished backside. This by itself doesn't add contamination, but it makes it possible to measure backside particles. That in turn may make it possible to control those particles and eliminate contamination problems. Like other contamination concerns, backside particles may not impact pilot lines. Instead the problem may appear only in full production factories.
“It's a loading effect. If you've got backside particles they could get transferred throughout the line, and they can rain down on the front surface of the wafer if they get built up or loaded up in a particular process tool,” comments Gary Rosen, implant development manager with the implant systems division at Eaton Corp. (Cleveland, OH).
Rosen says that in the past identifying such contaminants required indirect means. One approach was to map out wafer thickness and flatness, inferring backside particles by frontside dips and bumps. A polished backside, on the other hand, allows a direct particle count. That may well result in a cleaner frontside.
A final contamination problem may be the wafers themselves acting as particle magnets when charged. Ion Systems Inc. (Berkeley, CA) manufactures equipment for static charge control. Because of its greater size, a 300mm wafer will create a larger electric field for the same charge than a 200mm wafer would. Ion Systems' theory is that the larger electric field above a larger wafer will actually attract more particles. To test this idea, the company has conducted experiments with wafers sitting inside minienvironments for long periods of time.
“We've seen that the larger wafer will collect somewhere around four to five times more particles than the smaller wafer not just an increased total number of particles but an increased number of particles per area,” says John O'Reilly, semiconductor sector manager for Ion Systems.
Ion's answer to this problem is, as might be expected, static charge control. This control would also help other electrostatic discharge (ESD) problems, such as reticle damage. However, Keenan of Motorola notes that some aspects of 300mm manufacturing may help alleviate this concern. In particular he notes that minienvironments make it possible to cut back on room air velocity because of the ability to run at a higher class (less clean) cleanroom.
“We've seen a high incidence of ESD where you have high velocity air flow. But in these rooms, because you're in a minienvironment, there is going to be a tendency to throttle back the cleanroom class,” Keenan says.
Process improvements expected from the integration of minienvironments
by Bruce MacGibbon, Ph.D.
In the past (200mm era), only process equipment operating at atmospheric pressure observed process improvements as a result of implementing a minienvironment system. But with the onset of 300mm, all wafer handling must be automated. Therefore, all 300mm process equipment, even vacuum tools, will have an interface to the tool that includes a FOUP (front opening unified pod) dock and a robot to transfer the wafer from the FOUP to the process portion of the tool. This transfer area, often called a factory interface or front end, will be integrated with a minienvironment.
Since all 300mm equipment will have some form of minienvironment, whether simply a factory interface on a vacuum tool or a fully integrated minienvironment on an atmospheric tool, the entire manufacturing process will be less dependent on the cleanroom environment. Therefore process results should be more repeatable from site to site (i.e. independent of cleanroom). Minienvironments remove the variability caused by the difference in fab conditions around the world and help ensure consistent process results.
Minienvironments integrated with 300mm equipment will benefit from the knowledge gained on 200mm minienvironment systems. All of the industry standards and guidelines for 300mm minienvironment design have been developed from studies and experiences with 200mm systems over the last decade.
The following example illustrates how minienvironments have improved processes. It is assumed that the minienvironments are located in a typical cleanroom. The typical cleanroom is defined as having a cleanliness level of ISO Class 3 to 7 (Class 1 to 10,000), temperature control of ±1 degree C, and humidity control of ±5 percent RH.
Process equipment that have general requirements (contamination, airflow management, and static control) have seen process improvements in the form of improved particles per wafer pass (PWP) when minienvironments were integrated. Examples of minienvironments of this nature would be for metrology equipment and factory interfaces. The minienvironment is designed to reduce particle deposition by 1) using ULPA filters, 2) designing the airflow so no particle generators are upstream of the wafer, and 3) pressurizing the minienvironment so no particles enter through gaps or openings. Studies1, 2 have indicated that minienvironments must be pressurized to at least +0.003 inches w.c. relative to the ambient environment. If they are not pressurized to this level, contaminants can infiltrate the minienvironment through gaps or openings and PWP will increase. The minienvironment also contains static dissipative materials and possibly air ionizers to reduce static buildup and prevent static discharge.
Lithography equipment has specific requirements that are designed in to the minienvironment system to provide process improvements. In addition to the general requirements for contamination control, airflow in the lithography equipment must be controlled to ±0.1 degree C and ±0.5 percent RH. The minienvironment is designed with a local temperature and humidity controller. Recall it is much easier to tightly control air “quality” for a small volume of air. For DUV lithography, amine levels must be controlled to 1 to 10 ppb (typical fab levels are 30 ppb) to prevent T-topping3, 4. The die yields are significantly affected if these requirements are not achieved. For DUV systems, special chemical filters are used to remove alkaline molecules from the air.
Wet benches with integrated minienvironments provide process improvements and safety features5. A wet bench typically consists of multiple tanks that contain acids, bases, and solvents. The integrated minienvironment improves the process control by tightly controlling the airflow direction and velocity, and thereby not allowing vapors to migrate between acidic and basic tanks. The migration of vapors to other process modules allows the generation of salts that can deposit on wafers. Evaporation rates at the tanks and wafers are held constant by controlled airflow velocities in the minienvironment. The minienvironment also improves safety by providing a physical barrier between the operator and hazardous chemicals. Safety is also promoted by maintaining constant airflow characteristics (no turbulence) in the environment.
Chemical mechanical polishers with integrated post process cleaners (CMP dry-in/dry-out) have been incorporated in a single minienvironment with separate zones of cleanliness 6. CMP is an inherently dirty process; however the input/output area of the tool must be maintained at Class ISO 2. The minienvironment does not allow contaminants in the polisher to escape to the cleanroom, while at the same time not allowing contaminants from the cleanroom to enter the input/output area. The minienvironment improves the process by lowering PWP and also makes the CMP-cleaning process more cost effective by clustering multiple processes.
Integrating minienvironments with 300mm equipment will improve processes by 1) providing better control of environmental parameters, and 2) providing repeatable process results, independent of cleanroom environment.
Bruce MacGibbon, Ph.D., is director of research and development at Huntair Inc. Huntair is a Portland, OR-based company that specializes in integrated minienvironments, environmental control systems and airflow management for semiconductor equipment manufacturers. He received his doctorate from Clarkson University in 1995.
- Liu, B.Y.H. and Yoo, S.H. “Isolation Ratio and Particle Performance Measurement of a SMIF System”, J. of IEST, 1997.
- Pariseau, D. “The Effects of Differential Pressure on Particulate Levels in Minienvironments”, Proc. 41st ATM, IEST, 1995.
- Hinsberg, W.D., MacDonald, S.A., Clecak, N.J., and Synder, C.D. “Quantitation of Airborne Chemical Contamination of Chemically Amplified Resists using Radiochemcial Analysis”. SPIE Vol. 1672, 1992.
- Sematech Technology Tranfer #95052812A-TR, “Forecast of Airborne Molecular Contamination Limits for the 0.25 Micron High Performance Logic Process.”
- MacGibbon, B., Marvell, G., Benson, D., Busnaina, A.A. “Airflow and Chemical Migration in a Wetstation Minienvironment”. Proc. 41st ATM, IEST, 1995.
- MacGibbon, B. and Cleary, T. “Cleanliness Performance in a Dry-in/Dry-out CMP Tool”. Proc. Symposium on Contamination Free Manufacturing for Semiconductor Processing, SEMI, 1998.
Is the 300mm rush on?
The two 300mm pilot lines could soon have plenty of company. Taiwan-based TSMC, for instance, broke ground in mid-December on its Fab 12. The plan for the company's first 300mm factory is to have the superstructure done by the end of 2000, with equipment move-in slated for 2001. Estimated initial capacity will be 25,000 300mm wafers per month. Pilot production will be on a 0.15 or 0.13 micron process, with eventual evolution to a 0.1 micron process.
According to company spokesman Chuck Byers, the factory will produce logic, memory, and chipset products. The emphasis will be on logic. Byers notes that details of the fab layout are deliberately vague at this point, but that production will begin in early 2002.
When those first wafers roll off the assembly line, they'll be racing Intel. The Santa Clara, CA-based chip giant revived its 300mm program in mid-1999. Intel also plans to begin production in 2002, using a 0.13 micron copper metallization process. This will be done at the company's D1C development fab in Hillsboro, OR, at an ISO Class 3 (Class 1) facility with 120,000 square feet of cleanroom.
What's more, those aren't the only 300mm fabs on the horizon. Besides the obvious possibilities of both Motorola and Infineon, there are also reports that Japanese chip maker NEC plans to build a 300mm facility in California. The surge in activity hasn't escaped the notice of those in the industry.
“The 300mm industry slacked off over the last two or three years. It's my opinion that it's picking up and it will pick up quite significantly this year,” says Dan Syverson of FSI.
Others agree and point to history as a guide.
“We're reentering that period that is the equivalent of maybe 1990, 1991 when six-inch wafers were dominant but eight inch was coming on. Now eight inch is dominant, and 300mm will be coming on,” predicts Goodall of International Sematech.
If history does repeat itself, the transition to 300mm will take years. However, all of the cutting-edge technology for both processing and contamination control will be at the larger wafer size. HH