Austin, Texas Petersen Advanced Lithography (PAL) here is tackling mask-imaging problems that must be solved if optical lithography is to be extended to the 65-nanometer node and beyond. Founded by lithography veteran John Petersen, PAL has developed a hardware/software product called ProLE for Programmable Lithography Engine that performs electromagnetic-field correction a design-for-manufacturing (DFM) step that involves large amounts of data.
At the 90-nm node and beyond, the three-dimensional features on masks with extensive optical-proximity correction (OPC) and phase-shifting features often create image distortions. Avoiding those problems requires more-detailed modeling as well as data sets that break the normal computational power of engineering workstations.
PAL's small engineering staff develops libraries of optical-proximity-corrected images, created with ProLE, which it sells to customers, much as Artisan Components Inc. sells EDA libraries to its customers. Then, using ProLE, PAL's staff examines an optical-proximity-corrected image and refines the mask to compensate for electromagnetic-field distortions.
Ken Rygler, an independent consultant who earlier helped establish DuPont Photomasks Inc., said PAL is somewhat unusual in the DFM field. Most startups (by one estimate there are 61of them in the DFM sector) were founded by design-side veterans of the EDA industry. Relatively few DFM startups employ lithographers who understand resists, masks, light sources, optics and the interactions among them, Rygler said.
"The venture capitalists who are funding DFM startups understand design and EDA, but for the most part they don't understand lithography," Rygler said. PAL's staff understands how light energy is transmitted through the lens and the mask to the resist, and the changes in those energy patterns as they interact with those materials.
With 193-nm tools used for the critical layers of 90-nm designs, some of those distortions begin to become significant. At 65 nm, companies will need to understand "how energy interacts as it moves through those material layers, or they will be in for a rough ride," Rygler said.
"The problems in the image design space are so immense that we can only handle very small areas at a time," Petersen said. That computational burden required Petersen Advanced Lithography to develop a supercomputer based on blades from ClearCube Technology Inc., an Austin-based company. The current ProLE system is a 64-CPU symmetric-multiprocessing system, with about 60 Gflops of compute power to solve Maxwell's equations for electromagnetic fields. The simulation software breaks up the problem into elements manageable enough for each processor to handle.
"The challenge is that even a fairly small mask area, for an area describing a 300-nm pitch with 80-nm contacts, can require as much as 60 Gbytes of memory. We are in an era where analyzing these problems requires terabytes of information, moving at speeds we have never imagined. With our 64-CPU system we can tune [a mask design] overnight that could take months on a conventional system," Petersen said.
Petersen's goal is to move beyond PAL's contract work, which involves providing OPC primitives and then, with the mask manufacturer, refining completed mask designs. The company is seeking enough venture capital funding to turn ProLE into a more-automated tool, with a better user interface, one that customers can use in-house rather than shipping the entire problem to PAL, Petersen said. "People think we are a CAD company, but that's the least of it. We make masks right," he said.
Petersen worked for several years developing lithography process modules at Texas Instruments Inc., and then developed advanced resists for the Shipley Co. (now a division of Rohm and Haas) for 13 years. He was sent as a Shipley assignee to International Sematech, and later joined the Sematech staff as a lithography senior fellow. He founded Petersen Advanced Lithography (www.ADVlitho.com) in late 1998, completing ProLE by 2002.
Petersen last year hired Rod Dudzinski as CEO to bring PAL into commercial DFM.