The contamination control industry believes it's inching closer to the fabled “blue-jean fab.” But will it ever fully realize the dream?
by Neil Savage
A production engineer for a global semiconductor manufacturer is out for his Sunday morning run when his pager goes off, informing him that there's a minor glitch at his company's 300-mm wafer fab. Because he's just down the street, he jogs over to the fab, walks into the production area, wipes his sweaty hands on his T-shirt, pushes a few buttons on a processing tool, and leaves to finish his run. Wafer production continues unimpeded.
 |
Obviously, this scenario is a fantasy.
No one in his right mind would go anywhere near a wafer processing tool without first changing into the industry-standard bunny suit to keep from contaminating the clean environment.
And of course, there is only one 300 mm line in production at this time, the Infineon Technologies (Munich, Germany) plant in Dresden, Germany. However pilot lines are being run by Samsung Electronics (Kiheung, Korea) and by Motorola (Schaumburg, IL). Texas Instruments (Dallas), UMC (Hsinchu, Taiwan) and TSMC (Hsinchu, Taiwan) expect to have full-volume 300 mm fabs running before the end of this year.
But is this scenario—the so-called “blue-jean fab” in which workers wearing their street clothes can interact with the wafers and the tools—a realistic possibility?
It's certainly getting closer to reality as the move to 300 mm wafers pushes the industry toward both greater cleanliness requirements and greater automation.
“It's still a far way from that blue-jeans vision,” says Monty Stranski, industrial engineering manager at IDC (Portland, OR), a design and engineering firm that builds fabs for semiconductor manufacturers.
But as manufacturers move to 300 mm production, they're also shrinking the geometries of chips, which, in turn, has intensified the desire for the ultimate hands-off production scenario.
“As the line widths get down around 0.18, 0.15 [µm], the rules change a bit,” says Michael Brain, vice president of systems and software solutions at Asyst Technologies (Fremont, CA). As line widths shrink, particle contamination becomes a greater concern. Add to that the fact that such designs use copper interconnects, and copper contamination becomes a worry. The wafers therefore need a cleanroom environment of ISO Class 3 (Fed-Std-209E Class 1) or better.
“The industry is not stopping at 0.15,” Brain says. “They're headed to 0.10 and below.”
Tool-to-tool wafer handling
Maintaining every inch of a fab at ISO Class 3 (Class 1) or cleaner levels is logistically daunting. “You can't control a 100,000-square-foot space with people in it,” says Ed Guttes, technology manager at IDC. “Even if you could, the cost could prove prohibitive.”
The clean environment is shrinking, getting down to the level of the wafers and the tools, and human control is moving outside of it. The 300 mm fabs will use an automated material handling system (AMHS) that isolates the wafers in front opening unified pods (FOUPs). The FOUPs move the wafers from tool to tool, delivering them into clean environments within each tool. The FOUPs themselves can be moved around the factory floor with little regard to cleanliness, transported by either automated systems or humans pushing carts. The tools, meanwhile, will become more modular.
“The goal is really to provide direct tool-to-tool wafer handling,” says Jeff Nestel-Patt, director of corporate communications at PRI Automation (Billerica, MA). As that happens, the clean area shrinks to surround the wafers, reducing cleanliness control in the rest of the fab. “Overall, the manufacturers are trying to minimize the amount of cleanroom floor space, because it's very expensive to maintain,” Nestel-Patt says. “The less footprint you have to create like that in your factory, the less costly it is to build and maintain.”
The FOUPs make factory design a little easier, as manufacturers no longer have to worry about overhead monorail transport systems shedding particles onto the wafers. But the need for routine maintenance and manual intervention in the tools presents a problem. “How can you have a stable environment if somebody's ripping the guts out of the machine every six minutes?” Brain asks.
What needs to happen is that the mean time between failure or assist has to increase dramatically, Brain notes, and that is already happening. But the automation also needs to be improved to reduce the need for intervention, with more intelligence in the diagnostics and tracking. “The automated equipment has to be able to fix its own problems to the extent it can, and call for help when it can't,” he says.
In fact, says Stransky, one of the challenges will be improving the level of control and integration of all the various pieces of equipment in a factory. A fab can contain 400 to 500 pieces of equipment in 30 or 40 classes, such as photolithography machines, wet etch equipment and dry etch devices. What he sees happening is a move away from discrete manufacturing, in which every machine does just one job, to a more integrated style, in which one machine performs several steps to complete a process, such as creating an integrated circuit layer. Such integration, he says, is probably five to 10 years in the future.
Automating widespread automation
The information technology necessary to control widespread automation is becoming more sophisticated as the challenges grow. PRI, for instance, is developing scheduling software to get the materials to the right tools at the right time. “A lot of scheduling is still done on spreadsheets, but that won't work in a highly automated 300 mm factory,” says Nestel-Patt. “A wafer will travel 8 to 10 miles through a factory and it will take 30 to 40 days to complete its manufacturing cycle.”
“Software for advanced planning and scheduling is critical if automated fabs are to run smoothly and remove the contaminating presence of humans from the process,” says Ed Czupryna, general manager of Ortems Direct (Warrenville, IL), which develops such software. By managing the complicated changes in what needs to happen in a fab, and by getting feedback from the tools and altering activities based on that information, software is moving the supervision of the fab outside the clean area to managers' desk tops.
 This rendering shows several bays inside a fab with automation equipment moving wafers. Courtesy of PRI Automation. |
Automation has to improve not only for reasons of cleanliness, Nestel-Patt says. Fabs have to increase their throughput to stay economically viable. Eventually they'll reach a point of diminishing returns from geometry changes and will have to increase their yield by making scheduling and manufacturing more efficient. Automation also reduces the need for production line workers and gives fabs the ability to run around the clock. Semiconductor manufacturing will have to become more like the automotive and consumer electronics industries, where automation is more mature.
“If those industries were operating at the efficiency of fabs today, they wouldn't be profitable,” Nestel-Patt says.
As to whether a blue jeans fab is realistic, Czupryna of Ortems says, “It's probably a goal that the companies are striving for, but I don't think we'll ever reach it.”
Creature comforts
Stransky says the FOUPs will be used in ISO Class 4 (Class 10) cleanrooms at first, eventually moving to ISO Class 5 (Class 100) or ISO Class 6 (Class 1000) cleanrooms with workers wearing smocks instead of the full bunny suits. Already some Japanese manufacturers are moving 8-inch wafers around ISO Class 6 (Class 1000) cleanrooms in pods, and, as microenvironments become the norm in new 300 mm fabs, the same thing is likely to happen there.
“There's a lot of value to making the environment more friendly to humans,” Brain says. For one thing, if the manufacturing area is more hospitable, that will tend to get the engineers out of their cubicles and conference rooms and into the fab, where they can get a feel for how well things are working.
“Right now, engineers are reluctant to go into the fab,” Brain says. “When they do, they tend to be distracted by the requirements of the cleanroom—the noise of air circulating through suits, the masks over their faces, the gloves on their hands, the strange things the ionizers do to their hair. It prevents them from thinking and observing well. It's not a comfortable human environment,” he says.
On the other hand, Brain adds that there is a psychological benefit to making people jump through cleanliness hoops. “By telling people they have to put on bunny suits, you're telling them, 'Watch out for particles. We don't want dirt on the wafers,'” Brain says.
Although human bodies are the single biggest source of contamination in fabs, you can't just take them out of the equation.
 The inside of a fab at UMC (Hsinchu, Taiwan) showing Mattson Technology tools. (Courtesy of UMC and Mattson Technology) |
“A manufacturing operation is a system, and people are a major component of that system,” Brain says. The human-centered engineering and design process has to work with the automated, ultraclean manufacturing process. “Even though you want to isolate those two environments, you also want to connect them,” he says.
And certain precautions are almost inevitable. “Personally, I don't like gloves, but gloves are probably going to be the most difficult thing to get rid of,” says Brain. “Touching anything with bare hands deposits skin oils, which attract dirt, and it's hard to see how to get around that problem.”
Even as the automation gets better and the wafers spend more of their lives in isolation, there is always going to be some interaction between workers and the clean environment when the tools undergo preventive maintenance, says Ben Ighani, industrial engineering manager at IDC.
“Is it going to be 100 percent blue jeans?
I think what we're trying to say is it's going to get very close, but it's not going to get all the way there,” Ighani adds.
Neil Savage is a freelance writer in Lowell, MA.