Dallas -- Texas Instruments Inc. has sketched out a lithography roadmap that doesn't count on getting an EUV scanner until the end of this decade, said Shane Palmer, manager of TI's lithography external research.
Palmer said a realistic schedule for EUV is to see beta tools in 2007, and commercial systems in 2009 or 2010. In an interview at the inaugural International EUV symposium here, Palmer said TI "fully supports the EUV effort. We plan to do our development effort at the International Sematech center in Albany, doing the testing of the resists, working on mask issues, contamination control and so on."
However, because Texas Instruments has relatively few chip designs which remain in production for long periods of time, such as Intel's MPUs tend to do. With more "short runners," TI is more sensitive to the issue of EUV mask costs, which he said could approach $100,000 per mask.
A combination of 157-nm scanners from ASML, and projection electron beam lithography (EPL) from Nikon Corp. for contacts, vias, and possibly the first metal layers, is expected to be TI's workhorse combination in the latter half of the decade.
And Palmer said he is concerned that contamination on the mirrors used in EUV systems, and the need to deal with the heat generated by the source laser, could result in high operating costs.
Back of the envelope estimates for a single EUV scanner are in the $50 million to $60 million range.
157-nm scanners on order
TI has ordered three 157-nm scanners from ASML for delivery in about two years for development and early production of the 65-nm node devices, which in TI's case will have a 200-nm feature pitch (pitch is the measurement of the line and space). TI will use ASML 157-nm scanners for its C20 (65-nm node) process at TI's new 300-mm DMOS 6 fab here.
TI has been a vocal proponent of the EPL approach. Palmer worked on the Scalpel EPL system that was stillborn when ASML and Applied Materials decided to disband the eLith joint venture in late 2000. The EPL approach has continued with IBM's Prevail projection e-beam optics being incorporated into a Nikon EPL platform. Nikon is expected to demonstrate stitching of the scanned E-beam patterns, a step required to create a full-field image, late this year. While the throughput of the EPL system is limited, it should be fast enough for prototyping and critical layer production, he said.
Though Palmer said TI has not committed to buying an EPL system, an order is likely if the system is shown to be production-worthy.
"We are very optimistic about EPL, and see it being used for the contacts, vias, and possibly some of the metal (mask layers). It is seen as supportive of the 157-nm systems. The affordability of the EPL masks is one reason: Hoya is making EPL masks in Japan that show little diffusion," he said.
Besides the relatively affordable masks, resists for EPL are "a no brainer," Palmer said, and early estimates on the EPL price tag are coming in at $25 million to $30 million.
Is 193-nm underwater?
"We think 193-nm lithography can be used a lot longer than people say," Palmer said. The 193-nm scanners now coming to the factory floor can be extended first with higher NA lenses, with a numerical aperture of 0.9, though those lenses and heavier and much more expensive. Less certain is the chance that the N.A. can pushed to 1.4 by immersing the lens in a fluid, a technique that has long been used in microscopes, he said. Intel, ASML, and the Massachusetts Institute of Technology all have active immersion research programs ongoing, and International Sematech is planning an immersion lithography workshop for December in Burlingame, Calif.
"Water works as the immersion liquid, and while it is not a done deal, with a 1.4 N.A. lens 193 tools could be extended to the 45-nm node," Palmer said. After that, 157-nm scanners can be extended with phase shift masks, immersion, and reticle enhancement techniques.
Palmer said he also is interested in a novel form of lithography being developed at an Austin-based startup called Molecular Imprints, which is a flavor of what is called Step and Flash Imprint Lithography (SFIL). Former IBMer and current University of Texas at Austin professor Grant Willson is one of the founders of Molecular Imprints. The small company is developing a system which presses a mask on a liquid which is flashed and hardened into a pattern on the wafer.
Where does that leave EUV?
Palmer said "there are fundamental problems with EUV that still need to be solved. The power of the source (of EUV radiation) is not even one-tenth of what is needed for high-throughput EUV scanners. And what if you have to replace the optics every couple of months because of contamination? But I've been impressed by the papers being presented (at the EUV symposium). These are very, very smart engineers, and they are talking about all kinds of clever ways to solve some of these issues."