If you think photomask write times are excessive now, just wait until the 20/22-nm node. At that point, the complexity of reticle enhancement techniques needed to make 193-nm wavelength lithography scanners write 22- or 20-nm features will require something like an 8X increase in write times from the 32-nm node for complex masks, according to Aki Fujimura, CEO of D2S Inc.
"The complexity of the shapes that have to be on the 22- or 20-nanometer logic node are getting to the point where the write times that are going to be required to write these masks are going to hit the 80 hour mark," Fujimura said.
Fujimura's company, D2S, is the managing sponsor of the eBeam Initiative, a multi-company effort pushing a design-to-manufacturing approach known as design for e-beam (DFEB). D2S (San Jose, Calif.) offers a computational design platform that maximizes e-beam technology to reduce mask costs for low- and high-volume applications.
According to Fujimura, DFEB technology can help to solve the "80 hour mask" problem by using overlapping variable shaped beam (VSB) and circular shots to cut the number of ebeam shots needed to image a mask dramatically. At the BACUS photomask symposium last week in Monterey, Calif., members of the initiative presented data that validates DFEB's concepts for overlapping VSB shots and model-based mask design preparation, according to Fujimura.
EUV lithography, once predicted to be in production by the 65-nm node, is now not expected to be in production until at least the 22-nm node, if not beyond. The technology has been repeatedly pushed out due to the lack of power sources, resists, defect-free masks and other technologies.
With EUV delayed, leading-edge chip makers are hoping to extend 193-nm immersion as far as possible through double patterning and other techniques. Meanwhile, several private companies and consortium are pursing various e-beam direct-write technologies for prototyping and low-volume production, but tools appear to be several years away. E-beam direct-write lithography tools will be available no sooner than 2015,according to Kurt Ronse, lithography department director at nanoelectronics research center IMEC.
Fujimura said the eBeam Initiative continues to make progress toward making DFEB technology commercially viable. The group now boasts 34 members with the additions last week of Synopsys Inc., Abeam Technologies Inc., EQUIcon Software GmbH Jena and Tool Corp.
"At 22-nm there is a lot of debate about the methodology, and I think people are coming to the conclusion that at 22 or 20 we will have to use 193-nm immersion because EUV is not ready," Fujimura said.
During a panel discussion at BACUS last week, now available online, Fujimura and others stressed that EUV and direct-write are complimentary, not competing, technologies.
"The e-beam technology will be more targeted at lower volume solutions," Fujimura said. "I'm not sure what we are really competing if you look at any given customer problem.
Franklin Kalk, chief technology officer at Toppan Photomasks Inc., said direct-write is actually complementary to EUV. Because photomasks have gotten so expensive to produce, Kalk and others believe direct-write could make it possible to prototype more designs cost effectively. This, in turn, could lead to an increase in the number of designs that make it into mass production.
"If there is a way to make that less expensive, that's a good thing because it will raise the probability that that design will get into production.," Kalk said.
Fujimura declined to say which of the competing direct-write technologies had the best chance of making an impact. But he said funding would be the key. "I really, truly believe that if the same amount of money that is going into EUV was going into e-beam direct write, it would be working today," Fujimura said.
Kalk said it would be beneficial to the industry to have the number of competing e-beam direct-write technologies diminished so that the most likely winners could attain more funding. "But how do we do that?" he said.
Asked in an earlier interview if DFEB would become a must for chip vendors at the 20/22-nm node, Fujimura said. "It's amazing how people always come up with something to keep up with Moore's Law. You can never count people out in terms of how innovative they are. But right now I don't think there is an alternative."
Mapper has updated its concept...The total number of beams in the final mapper system will be 13000*49 (=0.6m) for 22nmHP. And for 10WPH, there will be 10 tools writing simultaneously. You can do the math now!