So many options, so little time: Why optics is forever in lithography
05/01/2000
Phillip M. Ware
More than 450 semiconductor professionals paid $150 each to attend the official release of the 1999 International Technology Roadmap for Semiconductors (ITRS) in Tokyo last November. Expectations were high that it would chart the future of lithography more accurately than previous Roadmaps. Unfortunately, however, it actually provides lithography engineers and vendors with a less than precise picture of the future.
Options and omissions
The 1999 ITRS proposes up to five technology options for 100nm lithography. At least one of these must be ready for production starts as early as 2002. In addition, it recommends pursuing seven technology options for 70nm lithography over the next five years. No semiconductor equipment vendor could possibly afford to develop all of these technologies: so many options, so little time.
It is also important to note that the 1999 ITRS for lithography overlooks the full potential of optical extension techniques. In reality, most industry players have conceded that creative approaches to extending KrF, ArF, and F2 lithographies will virtually ensure that "optics is forever." Lithography toolmakers anticipate that high-numerical-aperture (i.e., NA >0.7) KrF scanners with optical extensions will take us to sub-120nm lithography. In 2001, the availability of ArF scanners with similarly high NA will establish capability for mass production of sub-100nm critical resolution features. Toolmakers also anticipate that, as early as 2003, beta-class F2 scanners will be available in limited quantities to support resist and process development efforts toward sub-70nm resolutions.
Extending optical lithography
History has taught lithography tool vendors not to underestimate efforts to extend optical technology:
- Our programs, for example, include the development of high-NA KrF, ArF, F2, and reflective EUV optics, as well as optical extension technologies.
- Mask manufacturers and designers, like Numerical Technologies, for example, have also begun to introduce creative approaches to optical extension using double exposures for the critical layers.
- At the 2000 SPIE Microlithography Symposium, engineers from ST Microelectronics and Canon presented results using Canon's innovative double exposure for advanced lithography (IDEAL) method to generate 130nm-resolution, critical-level test-chip patterns using a 0.63 NA KrF stepper (Fig. 1). The potential here is 70nm resolution using ArF.
The "very serious" challenges of making either F2 or EUV lithography feasible are expected to drive lithographers to push the limits of the 193nm ArF wavelength even more aggressively than they are already pushing 248nm KrF lithography. Among the multiple technical challenges listed in the 1999 ITRS, cost control and return on investment are of particular significance to equipment vendors. The escalating development costs and short technology lifecycle will make it hard for developers of F2 tools and materials to recover their investment before some alternative next-generation-lithography (NGL) approach takes over.
Advocates of optical extension predict that super-high-NA (~0.80) ArF scanners may reach the 70nm node. EUV advocates are targeting 70nm as the EUV insertion point. If this scenario comes to pass, there may be no place for 157nm. It is anyone's guess whether or not F2 will be ready in time to play a significant role between the advent of ArF and EUV.
In reality, at its introduction, 157nm F2 may have to be implemented to manufacture 70nm features (half the F2 laser wavelength). This would be the equivalent of having introduced 356nm i-line lithography for 180nm features, an inconceivable feat only a few years ago.
Exposure tool projection lens NA will jump from the ~0.60-NA range of just two years ago to the 0.73-0.75 NA range by the end of 2000. Speculation is already abuzz about ~0.80-NA projection lens ArF systems appearing to fill the gap if F2 lithography or NGL alternatives are delayed.
What next?
Just how high can projection lens NA go? Lens designers believe 0.80 is the largest conceivable NA for mass production lithography. While the resolving power of a lens grows linearly with increasing NA, the lens diameter grows exponentially. Beyond 0.70 we start to see diminishing returns due to the extreme weight and cost of very small increases in NA (Fig. 2). In short, the materials and manufacturing cost of larger diameters for >0.80 NA lenses are expected to be prohibitively high.
One relic from previous roadmaps is the flawed assumption that chip sizes will keep growing with each generation. The manufacturing requirements for chip sizes larger than the traditional 22mm x 22mm square exposure field have not materialized, yet today we have scanners with 4:1 reduction lenses providing 26mm x 33mm exposure fields. Projections indicate we may never need anything larger than 22mm x 22mm fields. At a 4:1 reduction ratio, critical-level mask costs will spiral upward when dimensions become smaller and more complex as phase-shift and optical proximity correction features are included in the design. If we returned to 5:1 or even considered a 6:1 reduction ratio, this would alleviate many of the mask-manufacturing challenges for sub-120nm features at the wafer level.
The industry has only recently begun to consider seriously whether, or at what point, the reduction ratio should be changed. Industry consensus on the need to change the ratio is unlikely to come soon, if at all. Unfortunately, we are in a Catch-22 situation: If we change the reduction ratio, it will delay the Roadmap; if we don't change the ratio, we may not be able to stay on the Roadmap.
Regardless of the future of the projection lens reduction ratio and the implementation of reticle enhancement techniques, the "optics forever" strategy depends heavily on pushing the leading edge for higher NA and ultra-low aberration lenses for low-k1 lithography. Lens performance is critical for stepper-and-scanner makers. It entails continued major investment in lens metrology, aspheric surface measurement, and manufacturing tools. Consider, for example, that Canon is now investing more than $80 million in a new Optics Technology Research Laboratory in Utsunomiya, Japan, to advance stepper-and-scanner exposure tool optics. A Canon subsidiary, Optron, is developing CaF2 materials and manufacturing capability for ArF and F2 lenses and has already provided samples to the MIT Lincoln Lab program sponsored by International Sematech. All this work in optics is in addition to Canon's investments in the building blocks of electron-beam, x-ray, and extreme ultra-violet (EUV).
The lithography section of the 1999 ITRS contains a few flawed assumptions about the future. But recent efforts to include international manufacturers and equipment vendors as committee members should facilitate more active review of issues among the industry players for the 2000 ITRS. The combined resources and zeal of these vendors should ensure that optics will live forever—or at least until we have all retired.
 |
Phillip M. Ware is director and GM of marketing at Canon USA Inc., Semiconductor Equipment Division, 1130 Oak Ridge Place, Enid, OK 73703-3111; ph 972/409-7845, fax 972/409-7849, www.usa.canon.com/steppers.