Materials problem snags 157-nm lithography

Materials problem snags 157-nm lithography
SAN FRANCISCO — International Sematech convened an extraordinary meeting here at Semicon West this week to probe ways to fix an unwanted property of calcium fluoride, the material needed to build the lenses of 157-nanometer lithography systems. The jury is out on whether the problem of intrinsic birefringence will delay the arrival of 157-nm tools, which are seen as crucial to deployment of the 70-nm (0.7-micron) semiconductors expected to ramp in 2005.

The problem — in which unpolarized light passes through the calcium fluoride crystalline material at different speeds, creating fuzzy images at the wafer level — was identified in May by John Burnett, a physicist at the National Institute of Standards and Technology (NIST). Burnett measured light passing through a CaFl cube and found that different wavelengths moved through the crystal at speeds that differed by 12 nm per centimeter of calcium fluoride.

The fact that the basic material transmits fuzzy images is not good news for the 157-nm stepper program, but it may not be a showstopper. "We don't see this as a major reason for schedule slippage," said Rich Harbison, the 157-nm lithography program manager at International Sematech (Austin, Texas). "Since May the companies have scrambled to come up with their own strategies, and we still expect pilot systems to appear in 2003."

Nevertheless, lithography vendors this week were talking about 193-nm tweaks as placeholders in case birefringence or other problems delay the 157-nm tools. "We have to keep looking at ways of extending 193-nm [technology]," said Peter Converitito, director of strategic marketing at lithography giant ASML. "The 157-nm [generation] is on the road map, but manufacturers are going to keep finding ways to extend what they have."

"I do see interest in higher numerical apertures for 193-nm optics," concurred ASML product manager Roger Irwin. For example, he said, "You can go to more complex masks or dual exposure methods. We will enable even smaller resolutions." Still, those will be holding techniques. Irwin called the 157-nm migration "a done deal. It's already been factored into road maps."

IC vendors plan to introduce 157-nm scanners to their R&D lines in 2004 and begin early manufacturing the following year, when integrated circuits with 70-nm design rules go into early runs.

The birefringence problem is the latest in a series of challenges in new-generation lithography. Already, shipments of 193-nm lithography systems are running about a year behind schedule, putting the burden on phase-shift masks, which are proving to be highly expensive.

Sematech managers said at least four potential solutions were proposed at this week's workshop, attended by about 50 technical staffers from the major lithography and lens suppliers. Three involve adding other materials to the CaFl; one suggests use of a technique called "clocking."

"Since the NIST report in May, the lithography companies have been working feverishly to come up with solutions to intrinsic birefringence, and the lens designers and tool vendors each have come to different solutions," said Chris Van Pelt, an International Sematech lithography manager.

Answer in barium?

One idea is to create a compound crystal that will mix barium fluoride with calcium fluoride. An unnamed crystal manufacturer expects to produce a cube of such a material by September for evaluation by Burnett at NIST. If the compound results in negligible intrinsic birefringence, that may solve the problem.

Another possibility is to add a barium fluoride element to the lens design. Alternatively, some companies proposed adding magnesium to the mix.

Over the past year, the 157-nm lithography program has been hampered by a severe shortage of the CaFl crystals, which require special crystal-growing furnaces and process controls. Last year, companies in the lithography infrastructure struggled to improve both the quantity and quality of the CaFl production.

"Even if you can get enough of it, you don't know what the quality will be from batch to batch. Is the material repeatable?" asked a manager working the Nikon Precision booth on the Semicon West show floor.

The birefringence problem exists even in good-quality calcium fluoride crystals; hence the intrinsic nature of the challenge. For lenses cut on a 1-1-1 crystal plane, the aberration has three lobes of maximum and minimum transmissivity.

The regular nature of the birefringence may lead to a clocking solution — as used today in 248-nm lenses to adjust for aberrations in the lens elements. Clocking adjusts the lens elements in 60° angles in the horizontal plane. But clocking for intrinsic birefringence in the lens will remove the possibility of using clocking later to adjust for the inevitable imperfections in any lens element.

Tony Yen, a lithography program manager at International Sematech, said the International Semiconductor Technology Roadmap calls for 157-nm tools to be used for the 70-nm node. "All the leading-edge companies know that it will be possible to use 193-nm tools beyond the 100-nm [0.1-micron] node, but just how much below 100 nm we can go is the major question," said Yen, an assignee to the R&D consortium from Taiwan Semiconductor Manufacturing Co. "One belief is that 193-nm scanners with 'hyper-NA' [numerical aperture] lenses will be used for the 70-nm node. But that will require companies to pull a lot of tricks out of the bag at the mask level, and then the question becomes, at what cost?"

Sematech's Harbison said the 157-nm program participants see intrinsic birefringence as just another challenge to surmount. A total 157-nm solution involves other equally testy issues, he said, including finding a usable pellicle that is not destroyed by the 157-nm radiation, developing a resist chemistry and dealing with hydrocarbon contamination of the optical path.

Keeping costs under control is another issue. Ballpark estimates put the expected cost of a 157-nm stepper at $20 million or more, partly because the CaFl material, per kilogram, is priced roughly the same as gold.

Additional reporting by Ian Cameron of Electronics Times, EE Times' sister publication in the United Kingdom.