Richard,
Something we had talked about last year ...
--Duker
Semiconductor Business News, © 1999, CMP Media Inc.
March 1, 1999
semibiznews.com
The move to 'in-situ' metrology will take longer than expected
But work to develop these new in-line measurement and inspection systems goes ahead rapidly on a broad front
By J. Robert Lineback
It's beginning to look like the aggressive roadmap the industry had established for embedded metrology may be too optimistic. Experts have said the new closed-loop control systems and in situ sensors will be essential for advanced IC processing in the next decade.
At the same time, fab experts are warning that many of the industry's tried-and-true measurement and optical inspection techniques are running out of steam. Now a combination of new chip technologies, shrinking device geometries, new process steps, major changes in materials, and the upcoming 300-mm wafer are conspiring to complicate the job of measuring and inspecting processed wafers.
What was earlier seen as an onslaught of in situ metrology has begun to bog down as equipment suppliers wrestle with both long-standing business issues and the emerging technical problems coming from the many changes in process steps, new materials and shrinking geometries.
"The feeling in the past was that metrology would move more and more towards in situ - from off line [stand-alone systems] to in-line [connected to fab networks] to in situ," points out Allan Diebold, fellow and manager of metrology coordination at Sematech. "But the feeling in the metrology community today is that the evolution will take longer and require a few more tool and process generations than previous roadmaps have indicated."
The 1997 National Technology Roadmap for Semiconductors called for a steady migration by chip makers from "fixed-process recipe, open-loop control to closed-loop control via sensor-driven model-based integrated manufacturing." It is not clear whether the 1999 update of the roadmap, which is now in the works, will adjust this target.
But even if the job of developing next-generation metrology gear is going slower than expected, expects still agree that inspection and measurement capabilities must be more tightly coupled to the process equipment for the advanced IC processing that's coming.
What's needed is getting real-time information about potential defects and precise measurements of critical dimensions (CDs) immediately to wafer processing gear. This data would keep tools and processes from straying out of tightly-defined parameters. Reducing defects would drive up fab yields, while improved control over CDs would improve device performance. All this would make fabs more profitable, of course.
But it's much easier said than done. "We're going through a fairly radical transition," acknowledges Ken Monnig, associate director of Sematech's Interconnect Group in Austin, Tex.
"Up until now," he says, "most of the metrology problems were pretty well defined." But, he adds, "it took us 35 years to figure out what we had to measure when wafers came out of aluminum sputtering machines or oxide deposition tools." Now Monnig says that "we're heading into a period of many changes, and we don't really know what are the important properties to measure on a continuous basis in the fab."
Questions also are being raised about how and where measurement and inspection technologies should be used. In some measurements -- surface-flatness and film-thickness measurements for example -- a consensus is building that more metrology needs to be integrated with process equipment.
Integrated metrology recently gained a foothold in chemical mechanical polishing (CMP), experts say, because of the uncertainty over control and processes inside these tools.
And that's just the beginning, developers insist. "In the next decade, the trend will be towards integrated metrology for process control," maintains Giora Dishon, CEO of Nova Measuring Instruments Ltd. The Israeli company, which sells on-line automatic monitoring control systems for CMP and deposition equipment, now claims hundreds of these systems are working in the field, measuring film thickness immediately after a polishing step or chemical vapor deposition.
But Nova had to overcome two big problems to get there. Besides figuring out how to package thin-film measurement systems so they could be integrated into clusters of tools, the Israeli company also had to persuade process equipment makers to permit customers to add third-party metrology to their tools. "The commercial issues are one of the most difficult aspects," Dishon says. "There are questions about who sells the systems, how the process-tool warrantee and service issues are settled, and how the two systems communicate while maintaining the process flow."
To make it easier to settle such business issues, Nova formed the Integrated Measurement Association, which is attempting to develop standard practices for the marketplace. The company says it also has convinced some CMP tool suppliers to support its "NovaReady" field integration format.
"Process equipment and metrology cultures are completely different," maintains Dishon, which is another reason why his company's pioneering efforts were so difficult. But the Nova CEO, as well as other proponents of integrate metrology, now believe the tide is changing. Wafer fabs, they say, are trying to get a better handle on their new processes and to better utilize factory floor space by packing measurement capability into individual tool or clusters of systems.
Another pioneer of integrated metrology is Nanometrics Inc. The Sunnyvale, Calif., company decided two years ago to develop an integrated metrology strategy for film-thickness measurement after getting negative feedback from customers over early prototypes of its stand-alone, 300-mm platforms.
"We had shipped about 25 of these 300-mm tools to customers, who reacted by saying, 'These [systems] are huge!'" recalls CEO John Heaton. "[Even though] our system was actually one of the smaller 300-mm systems [on the market], we decided to go back and think about it."
Nanometrics decided to develop a film-measurement system for integration directly into CMP tool sets. This not only reduced the floor-space requirements, but also allowed the company to use some of the same subsystems serving the CMP tools. "Metrology has reached the point where the actual [measurement] content of the tool is very small and most of the costs are related to robotics, stages, the frame and so on," Heaton explains. "So we started with the basic technology and worked on fitting it into the process tools."
With that work under its belt, Nanometrics struck an exclusive deal last fall with Applied Materials Inc. to offer its dry-wafer thickness measurement technology in CMP tools. "You will see more and more different types of metrology added-on tools," Heaton predicts. "Film thickness is a natural, but we also know other people are working on integrated defect detection." He says that's because "customers want to stop misprocessing wafers and want 'supervisory metrology' to make sure everything stays within the guidelines of particles and thickness."
Another development now driving integrated metrology is the greatly increased throughput of process tools. That's making it harder -- and more expensive -- to serve a group of equipment in a bay with a single, stand-alone metrology system, according to Heaton. "If metrology is integrated, it can keep up and be optimized for the high-speed process tool," he maintains.
But there also are many tradeoffs in integrating metrology and inspection with process tools, cautions Tom Long, vice president of corporate marketing at KLA-Tencor Corp., the world's largest supplier of measurement and inspection systems. The task of detecting ever smaller, critical dimensions in ICs is growing increasingly difficult, he points out, and that often rules out the option of bundling sensor technologies inside process tools due to greater cost and size. Another problem with integrated or in situ metrology, he says, is that all the measurement systems must be calibrated -- something that doesn't have to be done when a stand-alone measurement system is used in a process bay.
"If a process tool already costs $1 million, you wouldn't want to put a $1 million sensor system in it, for sure," argues the KLA-Tencor vice president. "So one key question is whether you can build tools that have an accuracy down to less than 5 nanometers. Then the question is whether it can be cost-effective inside process equipment, or whether it's better to have some combination of integrated and stand-alone in-line systems for quick sampling of wafers."
"One area where it is not as difficult [to do integrated metrology] is in thin-film thickness," Long says. "Here it's possible to take a quick look to see if the process is in control," he notes. "But if you are dealing with measurements of CDs and overlays, that's more difficult."
While KLA-Tencor says it's still an advocate of stand-alone metrology and inspection tools, it is now exploring the use of its technology in integrated and potentially in situ applications. For example, the San Jose company is now working with National Semiconductor Corp., FSI International Inc., Lam Research Corp., as well as a couple of universities to explore the use of integrated metrology in lithography and etch systems.
The R&D effort is being funded with a three-year $18.6 million grant from the U.S. National Institute of Standards and Technology (NIST). One of the goals of the program is to develop an on-line feed-forward control (FFC) system, which would measure critical dimensions after a photolithography step and then pass to an etcher the information on individual wafers in order to improve etching yields.
"We are going to look at a range of metrology and control configurations,' points out Joe Bendik, a staff researcher at National Semiconductor in Santa Clara, Calif. "We will look at some new and novel techniques, such as scatterometry, as well as old techniques, like ELM [electrical linewidth measurement]." Some of this work also is aimed at finding a replacement or a supplement to today's critical-dimension scanning electron microscopes (CD-SEMs), which are expected to lose their ability to spot defects when the process moves to 0.13- to 0.10-micron feature sizes.
Scatterometry would use lasers or a broadband source to flash light on the surface of a wafer inside a tool. Its charged-couple device (CCD) sensors then would detect critical dimensions for real-time measurement. If ELM is pursued, the technique would be modified to determine CDs at 0.13-micron and below feature sizes -- something that's never been done before, Bendik says. ELM has the capability of doing 3,000 CD measurements in an hour, he says, vs. only a couple hundred using today's CD-SEMs.
Once an integrated metrology system is developed, CD information could be passed forward to the next processing tool, which could adjust its work based on the actual dimensions of processed wafer. This information also could be passed back to the tool that just completed the process step so that it can adjust itself for the next wafer.
"It could be many different kinds of loops. We don't know yet what the best strategy will be," Bendik says. "The beauty of this program is that it pulls together what's been missing, and it gets everyone talking about APC [advanced process control] and standards."
If practical feed-forward control systems can be developed, they could help to address one of the biggest, growing problems in wafer fabs, according to experts.
"It's too expensive to discover problems in wafers at the sort and probe, and it's even too expensive now to discover problems at a post-processing inspection point," points out Mel Effron, vice president of consulting and applications services for HPL Inc., a San Jose-based company that specializes in yield management.
"There really needs to be some kind of technology breakthrough to determine yield impact and fault avoidance during the process," says Effron, a 23-year veteran of IBM Corp. in Burlington, Vt., before joining HPL. "And that all has to be part of the equipment."
|
