Could Lithography Remain Ever Optical?

As incredible as it is to watch the lithography community to continue — year after year, generation after generation — to come up with increasingly inventive ways to postpone the death of optical lithography, it's also a very natural thing to see. This industry is always likely to stick with the good thing it's got going, especially when it continues to make the most productive and economic sense. At the same time, chipmakers have got to be prepared for the future and, as incredible as it may seem, that future might not include optical lithography — at least not at the most critical device layers.

Sematech held its second Litho Forum recently to update the lithography community about the leading technologies' readiness for the 32 nm half-pitch. With technology advocates from major chipmakers leading the way, presentations were made from throughout the lithography supply chain on the technical progress of extreme ultraviolet lithography (EUVL), 193 nm immersion (193i), next-generation immersion (193i+, with high-index fluids, resists and/or lenses), electron projection lithography (EPL), nanoimprint, and maskless lithography. That may seem like an awful lot of competition for precious development dollars, but some options certainly have broader appeal than others.

One front-runner for 32 nm consideration is EUVL, with chip giant Intel as its clear supporter. The key remaining critical issues to get EUV ready for production, as outlined by Janice Golda, Intel's director of lithography capital equipment development, are resist resolution, sensitivity and line edge roughness (LER); collector lifetime; defect-free masks; and source power. Other remaining issues include reticle protection during storage, handling and use; and the quality and lifetime of projection and illuminator optics. The most significant concern is the timing and cost/business case for EUVL development, she said — a concern echoed in Sematech's post-forum survey by other chipmakers and suppliers, who ranked cost of ownership as the most significant hurdle for EUV.

Nonetheless, significant progress has been made for EUVL. Both ASML and Nikon, which are each preparing full-field exposure tools, presented their progress. At the Microlithography conference in February, ASML demonstrated the printing of 40 nm dense lines/spaces, 60 nm contacts, and 35 nm isolated lines from its alpha EUV tool, which builds on its existing Twinscan platform. For the toolmaker's update last month, contacts had improved to 55 nm, and isolated lines to 30 nm.

Nikon is aiming for the delivery of its first EUV tool in the first half of 2007, having previously developed a small-field tool, and having gained experience with vacuum design through its EPL program. One of the advances focused on by Nikon was a reticle pod developed in cooperation with Canon as a way to minimize contamination without a reticle. The pod has been demonstrated to add an average of 0-0.3 particles per cycle, which is a significant improvement to EUVL reticle handling.

Meanwhile, 193 nm immersion lithography with water, having been put through more tests, has been alternately overcoming previous challenges and discovering new difficulties. But it is the clear preference for the 45 nm generation, and has a shot at the 32 nm half-pitch as well. "Until another technology emerges, this is really the technology to beat for that node," said David Medeiros of IBM, immersion's technical champion at the Litho Forum.

Various organizations have been making encouraging progress in next-generation immersion as well, such as high-index fluids from JSR Micro and DuPont, high-index resists, and extensive study of the suitability of various higher-index lens materials. Among the presentations at the Litho Forum, the University of Queensland described a high-index polymer for resists with an index of refraction >1.9, and Schott Lithotech showed progress in improving the transmission of its high-index lens material, LuAG, which seems to be the most promising lens material candidate for the job. Nevertheless, survey respondents did not seem to place a lot of faith in the feasibility of 193i+ for future generations.

What has been getting a surprising amount of interest is double-exposure lithography — a concept that had previously been shot down as causing too big a hit in cost and productivity, but is now suddenly being revered as the next savior of a reasonably familiar optical world. Although there are still concerns about increased mask costs and throughput, and very little has been shown in the way of actual results, lithographers seem considerably more open to the idea.

A great deal of research and progress has been going on with regard to both nanoimprint lithography and maskless lithography as well, but they are not getting the kind of support (i.e., money) that has been poured into EUVL, for example. And neither option garnered much support in the post-forum survey for 2012 or 2015 production. But global support has become a very critical component in a technology's survival. As imprint developers have pointed out on more than one occasion, all they really need is some large chipmaker to bankroll their infrastructure the way Intel has been financing everything to do with EUVL.

As much as the industry continually tries to narrow down the field of choices for next-generation lithography, smart engineers keep coming up with ideas that are just begging to be taken seriously. And the next thing we know, the field has widened again. Certainly, there's a balance there. While there's a concern that the industry just doesn't have the necessary resources to invest in so many different technologies, nobody wants to disregard out of hand a new idea that could prove to be a boon to productivity and cost savings.

At the same time, new technology introductions are always later than first projected, and the prognosis continues to look good for the extendibility of optical lithography — one way or another. Freescale's Will Conley is fond of referring to what he calls Sturtevant's Law: "The end of optical lithography is always six to seven years away."