Keeping big wafers clean
FOUPs are allowing for a degradation of cleanrooms, because the cleanliness is where it will do the most good — right next to the wafers
by Hank Hogan
The 300-millimeter (300-mm) wafer fab news, when it came, warranted some justifiable pride. In early 1999, Semiconductor300, the Siemens-Motorola joint venture, announced something unique.
“We have the first 300-mm pilot line fully equipped to run all the process steps for the 64-meg DRAM,” says Horia Grecu, deputy general manager for Semiconductor300 (Dresden, Germany).
Nestled inside an existing Class 1 200-mm Siemens wafer facility, the Semiconductor300 pilot line occupies 1,800 square meters, or 20,000 square feet, of cleanroom space. It is not a high volume megafab. Indeed, Grecu says that eventual production will be about 300 wafer starts per week, exclusive of engineering runs. Following the initial working material, qualification of the line is expected to be done not until some time in the third or fourth quarter of 1999. Although originally running a 0.25-micron product, the line will process smaller feature sizes.
But for all the hoopla surrounding the first working silicon, this isn`t a 300-mm facility. It`s 300-mm equipment and process tools chugging along in an existing fab. But such pilot lines are what`s needed to work out the details of contamination control in tomorrow`s large wafer megafabs.
FOUP, there it is
Some of the contamination control challenges that such pilot lines will face concern the front opening unified pod, or FOUP. A remote descendent of a wafer clean box, the FOUP shields wafers from particles and chemical contaminants. The concept has won the endorsement of the manufacturing consortium Sematech/International Sematech (Austin, TX). It is the 300-mm technology of choice for nearly every integrated circuit manufacturer in the world. That`s because the FOUP puts cleanliness where it will do the most good, right next to the wafer. Restricting the wafers to a narrow space has benefits, including the ability to run fine linewidth processes in degraded cleanrooms. That could cut costs, and it could also be a boon to operators inside 300-mm fabs.
“The worker ergonomics become a little better because the worker does not have to be mummified before we send him into the fab,” comments Ashwin Ghatalia, director of productivity and infrastructure for the International Sematech 300-mm program.
In practice, there are some details that have to be worked out. For one thing, a FOUP has about 120 components, including an integrated wafer carrier. This makes the whole assembly mechanically stronger. Consequently, the FOUP can mate to process tools more easily. It also is better suited to automated material handling systems. However, this complex assemblage presents some contamination control challenges.
“You`ve got all kinds of places for either dirt, debris, or moisture, if you go through an aqueous washing process, to be trapped,” observes Carol Davis, automated material handling systems program manager for International Sematech.
While it`s possible to clean a FOUP, a further complication arises when it comes to measuring the results of such a cleaning. One method involves liquid particle counters, which depend upon human operators to swirl a liquid inside a FOUP and then take a measurement. Another uses a vacuum wand to swab particles inside the assembly. Of these two, Kerry Kiser, 300-mm program manager at FOUP manufacturer Fluoroware Inc. (Chaska, MN), says the best may be the liquid version, but even that may not be satisfactory.
“Currently, there is no standard on determining how clean is clean,” says Kiser. “This doesn`t even [address] the problem of how you relate cleanliness to things like yield.”
Furthermore, as Davis of International Sematech notes, using a liquid method to measure the results of a liquid clean probably isn`t the best approach. There`s also the problem that some FOUPs have breathable filters. Water doesn`t stay in such filtered FOUPs, with the result that it ends up on the floor or clothing.
Also unresolved is whether or not FOUPs should be subjected to a nitrogen purge. Filling pods with dry nitrogen would remove the moisture that is often the cause of post-etch corrosion. Potentially, nitrogen-filled FOUPs might lead to a larger process window by preventing interaction with oxygen or moisture. Pressurizing a FOUP with nitrogen could also act as a barrier to contaminant entry.
On the downside, such an approach costs money, both during initial build and on an on-going basis. What`s more, injecting nitrogen could stir up contaminants, thereby defeating one of the purposes of a purge. Finally, there`s no consensus about what the purge setup and specifications should be.
For all of these reasons, the initial wave of 300-mm projects isn`t likely to have nitrogen purges of FOUPs.
What these projects will have are minienvironments. Some of these will be relatively simple units, with fan and filters on top, clear partitions on the side, and holes to vent air to the enclosing cleanroom. Some will be complex, with air recalculation and activated carbon filters to remove airborne chemical contaminants. All will be minienvironments, small enclosures around tools that are surrounded by an outside room.
Therein lies another contamination control challenge, as pointed out by Paul Ballentine, chief technical officer of Jenoptik Infab (Austin, TX). A U. S. subsidiary of fab designer and builder M+W-Zander (Stuttgart, Germany), Jenoptik is a 300-mm automation and process control integrator. According to Ballentine, Jenoptik`s German parent supplied 70 percent of the load ports used in Semiconductor300.
In a bay and chase arrangement, part of the facility is under 100 percent laminar flow. Those are the bays that connect tool load docks. It`s in these areas that wafers travel. The rest of the cleanroom is a gray zone, with perhaps 30 percent laminar flow coverage. That`s the chase, which may even be where return air flows. There`s a pressure differential between the bay and chase, and that`s a problem for a minienvironment that straddles both.
“This just will not work with a minienvironment. The minienvironment should be in uniform pressure,” notes Ballentine.
That`s why he favors the ballroom approach. Given the cost of making a ballroom Class 1 or cleaner, it`s easy to see why the bay and chase approach was used. However, with FOUPs, it may be possible to degrade cleanrooms without degrading the ability to produce the latest generation products. Indeed, research at Sematech indicates this possibility isn`t unfounded. Wafers have been passed back and forth from FOUPs and load ports in Class 10,000 cleanrooms. Even then, the process still meets a spec of no more than 0.09 particle of 0.09 micron or larger added per wafer pass.
If you can`t stand the heat, get out of the cleanroom
No one is saying that advanced semiconductor manufacturing can be done in street clothes. For one thing, in a degraded cleanroom, the only protection against contaminants comes from the FOUP itself and the minienvironments. Especially for new tool sets such as those being used for 300-mm, experience has shown that minienvironments may not be bulletproof during the prototype assembly stage.
“When we do the initial tests, we get failure rates on the minienvironment side up to 90 percent,” reports Dick Dryden, a principal at testing concern Dryden Engineering (Fremont, CA), which was recently acquired by Pentagon Technologies.
Dryden says these failures may include actual particles being deposited on a wafer. But other problems may also be present, such as pressure differentials that allow outside air into the minienvironment. There may be outgassing of materials or eddy currents. All of these are independent of the feature size of the process itself.
Dryden does say that with work these initial failure rates can be reduced to perhaps 30 percent. Eventually tight systems that don`t violate basic contamination control principles can be routinely designed and produced. Achieving this takes time, however. He expects the initial 300-mm efforts to run into some problems because of such violations of basic principles.
Understandably, there`s reluctance from manufacturers to degrade cleanrooms too far. After all, the approach of cleaning up the entire area is what the semiconductor industry is built upon. On the other hand, there can be cost savings in degrading a cleanroom. Because those savings will be wiped out if contamination causes a yield loss, the consensus is that the first 300-mm fabs may be a mixed bag. Some will have degraded cleanrooms; some will not.
But there`s another reason not to degrade the cleanroom too far. Manufacturers don`t want to cook their workers.
“You have to maintain a certain number of air exchanges per hour to keep the temperature under control. You don`t want your operators getting uncomfortable, and, moreover, you can`t tolerate temperature swings in the fab. You need that temperature to be uniform,” notes Ballentine of Jenoptik.
How much to degrade a cleanroom, then, becomes a balancing act. There is no one correct answer, although Ballentine says that M+W-Zander recommends cleanrooms be no worse than Class 300.
Analyzing the enemy
There are, to be sure, some contamination control issues that are process dependent. These have nothing to do with wafer size. Their impact on 300-mm contamination control is primarily because of feature size. For instance, the feature size of a process dictates the minimum problem particle size.
One contamination control issue related to particle size involves static charge. For a 0.18-micron process, which is likely to be the first generation of 300-mm production, the particle size of concern is 0.09 micron. Particles this small respond to charge as much as they do to air currents or gravity. So for 300-mm minienvironments, there may need to be electrostatic charge suppression of the interior and of the wafers being processed.
Such fine linewidths may also be more susceptible to airborne molecular contamination (AMC). Studies by Intel indicate that AMC may impact yield somewhat at the 0.25 micron generation. It`s thought that smaller features will be even more susceptible. The problem, as pointed out by Davis of International Sematech, is that no one knows how much AMC is too much. There is no specification for accumulation rates of various types of chemical contaminants. Such guidelines, if needed, will have to be developed.
One possibility is that for 300-mm production, it won`t just be the number of particles that are important. It may also be necessary to classify the type of contaminant, even those that are molecular in nature, on a routine basis. Sematech has used a sonic acoustical wave (SAW) device to measure the accumulation rate on silicon dioxide layer. Further analysis of the resulting film can determine just what`s being deposited. In another approach, systems have been developed that use lasers to vaporize particles for gas chemical analysis. Kiser of Fluoroware thinks that such detailed data gathering will become the norm as feature sizes continue to shrink.
For 300-mm contamination control, it may be that the semiconductor industry`s business woes have a silver lining. What had been a mad rush toward implementing larger wafer sizes has slowed considerably, largely due to the downturn in semiconductor revenues and cost reductions in the current 200-mm manufacturing. But this delay may actually benefit contamination control.
One time director of the 300-mm program at Semiconductor Equipment and Materials International (SEMI) and now an analyst with consulting concern Cap Gemini America, George Lee sums up the situation: “There is still a lot of that out there that has to be worked on. And this delay in 300-mm is letting people take a much closer and more detailed look at the potential needs downstream.”
 |
 |
 |
Left: FOUPs, such as this one from Fluoroware, shield wafers from particles and chemical contaminants.
Right: New total factory integration solutions for 200 and 300 mm manufacturing. Photo courtesy of Jenoptik Infab.
Time for a turnaround
The February Global 300-mm Report newsletter contains an update on 300-mm fab-build schedules. A look at the data shows that pilot production will continue through 2001, with manufacturers working out kinks in the new technology. It`s only in 2002 and beyond that a number of large 300-mm fabs are predicted to come online.
Given the recent slump in the semiconductor business, it`s understandable that skepticism might greet these predictions. According to a 1998 year-end summary from market researcher Dataquest (San Jose, CA), seven of the eleven largest semiconductor makers saw revenues shrink from the previous year. The drop for each was in double digits, with the worst taking a 26 percent hit. Even industry leader Intel had a rough year, with only a four percent growth in earnings.
There are signs, however, that this may be changing. The latest prediction from Dataquest is for 15 percent growth in the semiconductor industry in 1999. Revenue is projected to reach $154.5 billion. Part of the reason for this growth is some expected advance ordering by end users seeking to avoid any year 2000 related delays. In a sign that Dataquest isn`t the only organization optimistic about future growth, SEMI reports its latest equipment book-to-bill ratio at 1.1. That is, for every $100 of semiconductor equipment orders shipped, there were $110 of new orders received. This indicates that semiconductor manufacturers are planning for expansion later this year.
An increase in semiconductor demand is likely to lead to another type of increase in silicon. As old facilities are upgraded and new ones come online, manufacturers will increasingly look toward moving up in wafer size from 200 to 300 mm. So, if the upward trend in demand is real, there`s a chance the switch to 300-mm may happen sooner rather than later. — HH