New Collaboration Aims to Accelerate Direct-Write E-Beam Adoption
Some 20 companies throughout the semiconductor supply chain have created the eBeam Initiative to further work in maskless e-beam lithography, particularly techniques that could speed up the process and increase its value to the industry.
Aaron Hand, Executive Editor, Electronic Media -- Semiconductor International, 2/25/2009
In a dyslexic take on Shakespeare’s Hamlet, Aki Fujimura, CEO of D2S Inc. (Yokohama, Japan), posed the question, “To eb or not to eb?” Although perhaps not in quite the same way, many people throughout the semiconductor industry are beginning to ask more questions about whether e-beam lithography may hold some keys to renewing design innovation and to helping small-lot chipmakers survive.
Fujimura was speaking at an event yesterday at SPIE Advanced Lithography in San Jose to announce the launch of the eBeam Initiative, a collaboration of 20 industry players throughout the supply chain that’s aimed at furthering work in direct-write e-beam lithography and an approach known as design for e-beam (DFEB). The launch of the initiative comes at a time when not only is e-beam getting more attention for its suitability for prototyping, derivative designs and high-value/low-volume chip production, but collaboration efforts are becoming increasingly compelling.
Several recent announcements have continued to solidify the success of the IBM Alliance in Fishkill, N.Y.; DRAM manufacturers in Taiwan are teaming up to survive what has become a very challenging market; and the SOI Consortium continues to grow. These are just a few examples of companies — even competitors — pooling their resources to achieve technical goals that they would likely not be able to afford on their own. As the economy and market deteriorate further, this cooperation will become even more essential. As Jan Willis, facilitator for the eBeam Initiative, noted, “Technical innovation is tough enough even in the best of times.”
D2S, which provides software and IP enabling direct-write e-beam lithography, is a lead force behind the eBeam Initiative, along with Fujitsu Microelectronics, which makes use of e-beam in its chip manufacturing operations; Advantest, whose e-beam tool technology comes originally from Fujitsu; and e-Shuttle, a joint venture between Fujitsu and Advantest focused on e-beam prototyping services. Along with these companies, charter members of the new initiative include Alchip Technologies, Altos Design Automation, Cadence Design Systems, CEA-Leti, Dai Nippon Printing, eSilicon Corp., Fastrack Design, Magma Design Automation, Tela Innovations, Toppan Printing, Virage Logic and Vistec Electron Beam Lithography Group.
The long tail
Fujimura explained the need for e-beam’s small-lot capabilities within Chris Anderson’s popular Long Tail theory. Although the term originally referenced the ability of web-based operations like Amazon and iTunes to serve a growing list of niche markets beyond the body of the market that could be reached by brick-and-mortar book and music stores, the concept can also be applied to the semiconductor industry, Fujimura said. Because of increasing concern about mask costs, the semiconductor market is losing its ability to serve the tail, where low volumes make it difficult to recoup the cost of design because of mask costs.
Direct-write e-beam could be useful for startup companies or existing companies with new designs, for example, where “there’s really no reason to pay for a mask,” Fujimura said, until the company knows for sure there’s a market. There’s also an opportunity for derivative designs, which have been limited by mask costs.
But a key roadblock for e-beam lithography in this capacity has been its extremely low throughput. To overcome this adversity, the eBeam Initiative is focused on a technique called Character Projection (CP). While variable shaped beam (VSB) techniques have been able to speed up e-beam lithography over spot-based Gaussian techniques, they are still limited by throughput because of their need to break a pattern into its composite rectangles. In the example below, an E printed with VSB technology would require four shots. With CP, however, the entire E is stenciled onto the aperture plate, enabling just one shot.
|
| Unlike VSB technology, which requires several shots to create a character, Character Projection (CP) can write a cell in a single shot. (Source: Hitachi High-Technologies) |
The amount of cells that could be written with CP could improve the throughput by 3-5×. However, that’s still not enough to make e-beam compelling. Optimizing the design even further through DFEB, write times can be reduced by 10-25×, making e-beam direct write compelling for low-volume applications.
|
| Adding design for e-beam (DFEB) to Character Projection (CP) could reduce e-beam throughput times by 10-25×. (Source: D2S) |
It is this work that the eBeam Initiative is focusing on. Collaboration is already underway, including work among Fujitsu, e-Shuttle and D2S announced last October to prove DFEB design and manufacturing of a 65 nm low-power test chip. Various initiative members have been working to validate maskless manufacturing with successful test wafers for the 45 and 32 nm nodes.
Tuesday afternoon at the Advanced Lithography conference, STMicroelectronics and CEA-Leti presented results from their work with Advantest and D2S. Serdar Manakli reported good results at 65 nm, with CP accuracy comparable to VSB.
“VSB throughput is getting too low because of an increase in design integration,” Manakli said. At 90 nm, it takes ~30 hours to write one wafer, and that’s now reached ~150 hours per wafer at 32 nm. “So we cannot continue like this.” Comparing throughput of the various techniques, with Gaussian at 0.001 wph, VSB is at 0.03 wph. CP is 0.5-1 wph, and the eBeam Initiative aims to get that to 1 wph, which is what the industry has said throughput needs to reach to make e-beam viable. A multi-column cell (MCC) could get that speed to 10 wph, making the proposition more compelling.
The maskless benefit
Although maskless lithography may seem like a threat to maskmakers, in fact the opposite is true, evidenced in part by the involvement of DNP and Toppan in the initiative. Not only does Toppan Photomasks, for example, supply the aperture plates to D2S for use with the CP technique, but direct-write lithography would ideally increase the number of design starts and therefore the need for more masks as products become successful. “In the best case, direct write will increase design activity. And if it increases tape-out activity, I get more business,” said Franklin Kalk, CTO at Toppan Photomasks, in a separate interview. “So in my opinion, direct write is really the way to go.”
Kalk is also a major proponent of the collaborative efforts that are becoming increasingly necessary in the industry, and spoke about what he calls a club approach at a BACUS panel discussion Monday night asking what the reticle industry will look like in five years. “That club — that’s exactly what you have to do these days,” he said Tuesday afternoon. “There has to be a common thread for everyone. And I think that’s a good example — they’re trying to build an ecosystem around direct write.” He applauded the eBeam Initiative, saying that he thought they had a good shot at making direct write successful.
Kalk compared the eBeam Initiative with the efforts from Mapper Lithography, a Dutch company developing multi-column e-beam systems. “If you look at a company like Mapper, their approach is ‘We’re going to build this brute-force, really great direct-write tool.’ And they have some very impressive results. And their strategy appears to be based on getting it into one or two key customers, and then getting traction. And I think that what you’re seeing with D2S is they’re coming on a different axis. Their approach is, ‘We have some technology already, so rather than go off and develop more and more and more technology, let’s build the ecosystem around this technology, and then deploy that way.’ So two completely different axes of approach, but both with the same goal of ‘Let’s get business in this industry for direct write.’” If either technique succeeds, business will likely improve for maskmakers, Kalk added.
Jack Harding, chairman and CEO of eSilicon, is serving as an advisor to the eBeam Initiative. As an ASIC company, which is a key target market for the initiative, eSilicon has no plans to switch to e-beam overnight, he said, but is rather planting a tree for the future. Harding expressed his concern about the design and mask costs, which he called taxes on the supply chain, and whose suppliers make money whether a chip is successful or not. The industry has got to make making chips easier, faster and cheaper or it will not survive, he said. “Anything that eliminates those barriers is in fact goodness.”
Along with Harding, other initiative advisors from the design community include Marty Deneroff, director of engineering at D.E. Shaw Research; Colin Harris, COO of PMC-Sierra; Riko Radojcic, principal engineer and manager at Qualcomm; and Jean-Pierre Geronimi, director of computer-aided design (CAD) at STMicroelectronics.
-
CP is yet another layout limitation, really.
guest - 2/25/2009 9:57:00 PM CST
E-Beam Lithography
06/30/2009E-Beam Lithography
02/09/2010E-Beam Lithography
03/14/2006
