Issue



There really is a litho wall


08/01/2006







Optical lithography has transcended so many limits set by the promoters of next-generation lithographies that much of the industry acts as if it will live forever. Long ago, I first became fascinated with litho when Grant Willson, currently professor of chemical engineering at the U. of Texas, Austin, told our group at IBM that the powers-that-be had decreed that optics would die at one micron, to be replaced by direct-write e-beam and X-ray.

That seemed unlikely. I thought of three optical innovations (high-NA projection, excimer lasers, and phase shift) that seemed likely to take us to 100nm resolution (say 200nm pitch). Then came four more innovations that have taken us below 30nm CD and should allow 90nm pitch, which the International Technology Roadmap for Semiconductors calls “45nm hp.” But that is probably it. To get to the 32nm node and beyond, we are going to have to change some paradigms. The old technologies are about played out.

Resist will likely change first. Today’s materials are inhomogeneous on the nanoscale, even before exposure. When the quantum mechanics of absorption liberates protons at random locations-which then follow random trajectories during PEB, deprotecting some of the resin as they go-the result is an agglomeration of material with different solubilities. Upon development, insoluble nano boulders stick fast in resist walls or wash away, as do random pockets of liquefied ooze. On average, the development rate may be at threshold, but the variation gives rise to linewidth roughness and failed contact-hole development. What is needed is a more uniform material-maybe an organic glass-and a development method that doesn’t depend on when something gets wet, maybe employing super-critical CO2 instead of TMAH. There will be a penalty to pay, though, in photospeed.

Then there is the exposure tool. Water immersion allows numerical apertures as large as ~1.3, and that is enough for 90nm pitch with 193nm illumination. Higher index fluids will allow higher NAs, but to get a factor of 2 of areal density, the NA has to be 1.8 or so. That means a fluid, lens material, and resist with an index of refraction above 1.9. Such materials are possible, but their development isn’t very far along. There actually is a fairly plausible lens material: lutetium aluminum garnet. If we had started working on that when we began the 157nm materials development program, it would be ready now. But we didn’t, and it isn’t. So, as of now, there isn’t any plausible way to get to 64nm pitch with a single optical exposure.

Multiple exposures in a scheme I call “interdigitation” remains a possibility for extending 193nm technology. Interdigitation uses multiple resist images with multiple development to decrease pitch by placing narrow dark lines in the middle of wider bright ones, with a hard mask to remember the locations and transfer the summed pattern into the substrate. This system does convert overlay errors among the multiple images to dimension errors on the layer, but not necessarily to critical dimension errors. More seriously, it reduces throughput and increases mask cost, but only for the finest pitch layers. It may be that the industry can accommodate the increased cost of interdigitation to maintain Moore’s law. However, it does change the CoO paradigm. If layers are going to be synthesized in multiple patterning steps, it might be possible to use some lower-cost process (like imprint or interferometric lithography) for some of the steps.

If 193nm is at the end of its run for critical layers, the next thing may have to be EUV. Whatever its technical and infrastructure deficiencies, EUV remains the consensus option for chips made in large quantities for generations beyond 32nm. The challenge is to build a critical mass for the technology, and if that doesn’t start to happen when 193nm immersion begins to peter out, it won’t happen at all. Much of the industry may choose not to make the transition, finding paradigms other than shrinking pitch to increase function and lower costs. That is the litho wall on the horizon, yet again. Maybe this time it’s for real.

Click here to enlarge image

M. David Levenson is the editor-in-chief of Microlithography World magazine and contributing technical editor for Solid State Technology at PennWell, 201 San Antonio Circle, Suite 205, Mountain View, CA 94040; e-mail [email protected].