All,
Semiconductor Business News, CMP Media Inc.
July 1997
Fab strategies: Lithography decision time
The options must soon be narrowed for 0.15-micron technologies
By Jack Robertson
The countdown begins. While the Semiconductor Technology Roadmap
doesn't call for wafers with feature sizes of 0.15 micron or less to be introduced until 2003, it's mega-decision time now in the semiconductor industry.
What the industry has to do quickly is decide on the one or two lithography system candidates that could turn out this generation of wafers. The answer is needed now because lithography tools must be ready several years before commercial products are introduced, according to Sematech officials.
That means the ongoing debate between optical, X-ray, electron beam, ion beam, and extreme ultraviolet lithography has to be pretty much wrapped up this year. Device makers and their suppliers must reach a consensus by the end of the year on winnowing down the list to one or two potential 0.15-micron tools.
"We're running out of time," declared John Canning, Sema-tech director of lithography. "By the end of this year we've got to settle on one or two promising approaches and move full speed ahead with development."
No clear leader
There still is no clear leader among the half-dozen contending systems -- each has major technical hurdles to overcome. So Sematech is running a series of workshops across the country to analyze the pros and cons of each of the half-dozen contending lithography systems.
Sematech's goal is to resolve as many "open questions" as possible on all the various lithography candidates by the Nov. 4 summit it is sponsoring at Colorado Springs, Colo. Attending the global meeting will be all major device makers, lithography tool and infrastructure suppliers. Once they choose one or two likely next-generation technologies, a global working group will begin laying out the technology path in 1998 to develop prototypes that will be ready to go soon after the turn of the century, Canning said.
Sematech's Canning listed the following as contenders, as well as noting each technology's pros and cons:
Optical systems can extend deep-UV wavelengths into the
sub-0.15-micron frontier, supporters claim. Argon flouride 193-nm excimer laser systems can process chips with 0.15-micron feature sizes. Optical advocates claim that with new phase-shift masks, the same excimer laser should be able to handle 0.13- and 0.11-micron geometries. An F2 flouride 157-nm laser now in development might be able to extend optical down to 0.10-micron feature sizes.
PRO: Chip makers potentially could build on optical lithography processes they already know, possibly using some of the same equipment. Optical systems for making the critical wafer layers would integrate well with existing optical tools processing noncritical layers.
CON: Phase-shift masks for 0.15-micron and below feature sizes have yet to be demonstrated. Lenses at these minute geometries are a potential show-stopper and the 157-nm wavelength excimer laser is unproved. Sematech already is writing it off as a viable candidate, even though the Massachusetts Institute of Technology and SVG Lithography are pushing the F2 flouride laser hard.
X-ray lithography, which has been around for more than a decade, has
burst on the scene as a serious candidate with the promise of new
photomasks and lower-cost, next-generation synchrotron X-ray sources.
PRO: X-ray has been in development far longer than rival systems, which have jumped into contention only in the past several years. Using a new tantalum mask, IBM Microelectronics has fabricated a laboratory 1-gigabit DRAM chip on its X-ray synchrotron line.
CON: The tantalum mask has only been in use since April and needs more
lifetime testing data on reliability and stability. A commercial supplier of the new mask blanks and mask writer is required to give chip makers greater confidence of supply. Up to now, IBM has been the principal supplier of X-ray masks, using its own e-beam mask writer. X-ray's 1:1 mask exposure can't be done with the equipment being used for the 4:1 ratio now in optical lithography and requires entirely new hardware.
Scalpel direct-write e-beam now in development at Lucent Technologies,
Murray Hill, N.J., has passed a "proof-of-concept" demonstration. Lucent's Bell Laboratory division predicts that a preproduction Scalpel tool could be on the market by 1999-2000.
PRO: Scalpel can utilize the resists now used on wafers in deep-UV optical systems, as well as commercially available masks from DuPont Photomasks or Photronics. The 4:1 exposure reduction is similar to optical lithography. Lucent officials claim Scalpel works well with step-and-scan lithography, and they expect it to be the predominant critical-layer process for feature sizes of 0.15 micron and less.
CON: Scalpel moves the wafer under a stationary e-beam source instead of the conventional mask staging across the wafer. This even Lucent conceded could be a throughput problem. This also means that each exposure made on the wafer as it passes under the source must be stitched precisely to the previously exposed pattern -- a process that also has to be more fully demonstrated. Also needing a demo is the new kind of membrane mask.
Microcolumn e-beam systems, championed by Etec Systems, use
IBM-developed technology. Working on this system is Stanford University's Solid State Electronics Laboratory. The tool would integrate hundreds of miniature e-beam writers in a phased array to reach potentially 0.1-micron wafer geometries.
PRO: Advocates claim this concept offers the extremely fine-line patterning of e-beam combined with the potential of far greater throughput. Operating at 1 kilovolt, it is far less than Scalpel's 100-kV requirement.
CON: The concept hasn't been demonstrated. The multibeam array may
require more complex logic than can be quickly developed to control
hundreds of miniature e-beam writers.
Extreme Ultraviolet (EUV) uses a 13-nm wavelength pulsed xenon gas
laser and has the capability of operating below 0.1-micron linewidths. It is being pushed by Intel Corp., which still hopes to form a consortium of device makers to develop the lithography tool. Ultratech Stepper, San Jose, is not part of Intel's consortium, but is actively researching UV technology on its own.
PRO: Intel claims chip makers need only make the capital investment in EUV systems one time, which would cover all succeeding wafer generations from 0.15-micron to sub-0.1-micron capability. A new pulsed supersonic jet xenon gas laser under development at Sandia National Laboratory is a high-efficiency source that doesn't produce the particle scattering that occurs in solid- state lasers at this wavelength.
CON: The reflective optics need a new multilayer mask with as many as 50 to 60 thin-film layers, a manufacturing process that hasn't yet been demonstrated. The Sandia laser needs a higher power driver, up to 1 kilovolt. High-resolution resists with thicknesses of 0.5 micron also need to be developed.
Ion beam projection: Germany's Siemens AG is executive manager of a
European-sponsored ion-beam projection system that's aimed at a tool able to pattern feature sizes below 0.1 micron.
PRO: Siemens officials said that a large 60-mm X 60-mm field size is
possible that has superior depth of focus and can run at a throughput of 30 wafers per hour. A prototype system could be ready for 8-in. wafers by the end of 1999 and for 12-in. wafers by 2001.
CON: Ion-beam projection requires a new type of stencil mask that has yet to be demonstrated as a commercial product. Sematech's Canning said getting rid of distortion in wafer patterns, beam blur, and stitching has yet to be demonstrated.
~~~~end~~~~ |
