Editor's Page

More questions than answers

You can talk thermal, etch, planarization, or all the other fab processes until you're red in the face, but without lithography there would be no chips. The semiconductor industry's future depends on the continuing development of the photo cluster as a viable, reliable manufacturing process, both with the existing optically based tools and the proposed "next-generation lithography," or NGL, technologies. The recent SPIE Microlithography symposium and accompanying exhibition, while offering a wide-angle snapshot of the litho community's state of affairs, also raised many troubling questions without offering any definitive answers.

NGL—which a few folks call SFL, for science fiction lithography—garnered a healthy amount of attention at the event. Aside from the obvious question of which NGL or NGLs will win out, the main quandary was, what will be the insertion point for the leading NGL contenders, extreme ultraviolet lithography (EUV) and the Scalpel electron beam technology? Plenary speaker William Arnold and others put that point in the 2007—2011 time frame, with pilot-line work at the 70-nm node and production at 50 nm. But several severe technological challenges remain. How do you make a scanner into an oxygen-free, particle-free vacuum tool, as the EUV alternative requires? Can the new photo tools ever ramp to the same throughput levels as today's steppers? There are different core competencies at work, different ways of doing business from what the industry is accustomed to.

On a parallel track, another often-asked question centered on the lifespan of optical technologies. Who would have thought a few years ago that the 157-nm node would be reached with optical technology, ahead of the timelines set in the industry roadmaps? But how long will 157 last, and for how many nodes beyond that can optical be pushed out? In what he called the "optics forever" scenario, Arnold sees optical extended all the way to 70-nm for some layers, intersecting with nonoptical at that point. Soon there will be fabs supporting 248-, 193-, and 157-nm lithography, with one of the NGL options in development. How will supplier companies deal with the strain of servicing these kinds of multiple-technology toolsets? Will defectivity, already a major headache in DUV lithography operations, threaten reasonable attempts at profitable yields?

The mask side of the challenge may push all the other concerns off the table. If you don't have the reticles, it doesn't matter what kinds of whiz-bang scanners you develop. For example, how do you make defect-free reflection masks at a reasonable cost? As Sam Harrell, the other plenary speaker, noted, "There's no way out of the mask squeeze by going to NGL." So if, as Harrell said, reticle costs will drive the economics for all the new technologies, who's going to pay for it? The prevailing opinion holds that since the suppliers can't handle it all, IC makers must take on some of the development risks. This has already happened with the recently formed alliances between chipmakers and the tool and materials companies, but will the partners bail when the boom times come again, harking back to the more selfish, control-freak tendencies of the not-so-distant past?

Harrell's notion of the "subwavelength gap," where feature size falls below the lithography wavelength, shows how the economic and technological demands mount and how efforts to achieve high yields, low defectivity, and moderate cost of ownership become excruciatingly difficult. He sees the cost of reticles moving from a consumables model to one more akin to capital tools, comparable to equipment costs in the smokestack industries. The critical dimension controls must tighten, leading to more stringent CD specifications, which push yields down, pump costs up, and increase the need for and complexity of inspection equipment.

Given the enormous pressures, both technological and economic, to deal successfully with the difficult challenges of turn-of-the-century, or TOTC, lithography, the answer to the question, "Who's going to pay for it?" seems to be, "The entire industry."

Tom Cheyney
Editor
tom.cheyney@cancom.com