By Griff Resor, Resor Associates, Solid State Technology Editor Advisory Board Member
Immersion lithography may take a little longer to move into mass production than its champions would like, and the path to the 32nm half-pitch node is also not clear. Still, the pursuit of Moore’s Law continues on many fronts, such as finding ways to “cheat” the “laws” of optical physics, or using organic self-assembly techniques to augment optical lithography. These were the takeaway impressions from the recent IEEE Lithography Workshop, held July 31-August 4 in Prince Edward Island.
The IEEE Lithography Workshop, which meets roughly every 18 months, covers the entire gamut of lithography issues, with all-technical talks presented by the experts in each field. Topics are randomized, with no parallel sessions, a format that fosters interesting discussions and assures a lively discourse, with ample time to meet the experts and discuss questions and ideas. There are no published proceedings, a rule designed to encourage presenters and the audience to be more open with their comments. As such, this article contains generalized observations without identifying any single talk.
Computational lithography and immersion lithography are being relied on for the 45nm half-pitch node. Computational models and hardware appear to be up to the task, while everyone now has roughly the same slide for immersion defects. The mechanics of leading-edge air entrapment and trailing edge fluid escape (and subsequent water stains) appear to be well understood. An ample range of dynamic contact angles exists to make immersion lithography work at production scan speeds. One resist supplier has added trace compounds to a basic 193nm resist to engineer the two contact angles without the need for a separate topcoat material — a simplification the industry has been hoping for.
Worldwide efforts are clearly underway to make substantial progress on defects — results looked very good, and are low enough to put immersion into production. Still, overlay errors remain about 2x larger on wet systems than on dry systems, and it is clear that more time will be needed to sort out the root causes and correct them, possibly delaying insertion of immersion lithography into mass production.
The IEEE Lithography Workshop also provided a good look at what lies beyond the 45nm half-pitch node for immersion lithography. Resist suppliers have run enough experiments to know what basic materials systems look best, and are now working to develop sample resists in their chosen material. It looks like high-index resist (n= 1.8-2.2) can be made, and the basic lens forms developed for NA=1.35 can be used if a high-index material can be found for the last lens element, but so far results are not encouraging. The only solution appears to be a complex garnet crystal. Index, transparency, and birefringence specs look good, but no one has grown an optical grade crystal yet. (It’s reminiscent of the early days of the 157nm quest.) At best, a schedule delay of several years seems likely — bad news for chipmakers that are shrinking critical dimensions on a two-year cycle.
Meanwhile, EUV suppliers are working hard to pick up where optical lithography “ends.” There has been steady progress on machines, mask handling, and defects, so the source looks like the largest challenge. One solution to line-edge roughness is to use significantly more photons (slower resist), and the models look good, but this puts even more demand on the EUV source. The current requirement for 100W of at-wavelength power at the intermediate focus could escalate, to up to 500W. Three source technologies are being worked in parallel by different suppliers to meet this need, but it is too early to predict the outcome.
Double-exposure techniques will probably have to fill the gap. The added cost for double exposure seems to be an emotional issue, but there seems to be few who speak quantitatively to the issue. If six layers have to be double-exposed in a 30-layer process, the cost will rise 20%. On the other hand, super-high NA tool prices may increase at the same rate. Of course, the overlay problems reported at this workshop complicate the double exposure effort, but overlay performance likely will be improved in time.
Block co-polymers were the subject of several presentations, and work in this area appears to be very interesting, particularly as a prospect for magnetic disc media. Most of the speakers are using fine structures made with optical lithography to control the self-assembly of these polymers. Using the right mix of polymers, they can create small vertical cylinders that self-assemble into a hexagon pattern over a distance of 10-15 cylinders. The basic engineering models are being worked out — each cylinder can be 20nm in diameter, organized into 10 tracks guided by features that are 200nm apart.
While the IEEE Lithography Workshop’s main focus remains on chasing Moore’s Law, this year’s program included other applications of lithography, such as disk drives and flat-panel displays. The Workshop provided a multi-discipline set of talks that challenged us to look outside the box at new technologies and new markets. — G.R