By M. David Levenson, Contributing editor
The laws of physics and the challenges of economics are the same around the world, but different countries seem to pursue different technology strategies in the worldwide semiconductor marketplace. Japan's once dominant IC industry now must compete with lower cost producers in the DRAM market and with entrenched monopolies in the microprocessor market. Where can a creative country find unoccupied market niches?
Sophisticated consumer electronics seems a good bet to some technology leaders in Japan. The problem is that the ASIC and system-on-a-chip circuits required will be made in such limited quantities that the lithography strategies developed to support MPUs and DRAMs may prove uneconomical. That is one reason that Advantest, Hitachi, and Japanese R&D consortia are developing cell-projection e-beam lithography, whereas US manufacturers are emphasizing EUV and SCALPEL, both of which require potentially expensive masks.
The Association of Super-Advanced Electronics Technologies (ASET) has completed its program on direct-write e-beam lithography and transferred sponsorship to Selete. Electron-beam work is now concentrated on mask writing in support of other pattern replication technologies. Even though ASET is meant to expire as an organization in 2001, it continues to undertake long-term projects.
For example, ASET is actively pursuing EUV under the direction of Shinji Okazaki — formerly “Mr. Lithography” at Hitachi. The promise of EUV is high-throughput, which may not be possible with any charged particle lithography, according to Okazaki. The ASET EUV program is emphasizing resists and photomasks, while relying on Tinsley Corp. to fabricate the optics and outside universities and institutes for the EUV source. An experimental three-mirror projection system is being built at Himeji Institute of Technology with an electron undulator as the radiation source.
According to Okazaki, 13nm EUV has an advantage over 157nm VUV exposure in that the known resists are substantially more transparent at the shorter wavelength. “Etch resistant high-carbon-content resins — even Novolac — can be used as single layer EUV resists with reasonable thicknesses,” states Okazaki. “That is not true of 157nm, and perhaps never will be.”
There is currently a race to insert either EUV or VUV as the first post-193nm lithography, probably in 2005. The EUV R&D must be accelerated by two years to head off the 157nm program championed by Masaru Sasago of Matsushita Electronics. “While EUV costs more than VUV to develop, it will last two or three generations while VUV will support only one generation at best. Thus the cost per generation is less,” reports Okazaki.
The yield of the best exposure tool is zero without adequate reticles, which has been a concern for manufacturers dependent on the US photomask industry. The situation is somewhat different in Japan. Dai Nippon Printing, the premier mask making company, means to do whatever is necessary to support industry progress, according to Naoya Hayashi, general manager of research at the Semiconductor Components Laboratory. DNP, like the other Japanese mask makers, avoided the most destructive consequences of the price war that raged in the United States and preserved independent R&D capability. Today a leading edge (10-hour write) binary mask would sell for $20,000 in Japan, but a low-end reticle would still be above $3,000.
The DNP Kamifukuoka site has 36 pattern generators (with six more in Kyoto), with tools from the recent acquisition of the Hitachi mask house soon moving in. Having so much production at a single site facilitates discipline and consistency. Inside the clean rooms, workers wear specialized “intermediate clothing” under their bunny suits and work as if they were in a wafer fab.
Beyond conventional binary masks, DNP's roadmap includes embedded-attenuated PSM technology for 248 and 193nm DUV with CrFx material, perhaps overlaying partially transparent chrome. Tri-tone masks, with a patternable opaque layer above the semi-transparent phase-shifting layer and high-transmission CrFx for specific applications are also on the agenda. Like other mask makers, DNP is endeavoring to develop an economical and consistent process for alternating-aperture PSMs, which they see as being implemented after other options have been exhausted. A recent symposium on alternating PSM design tools in Tokyo attracted 80 participants. For next generation lithography, DNP is emphasizing stencil masks with 20µm substrate thickness for cell-projection e-beam over thin membrane technologies for SCALPEL.
In spite of the economic downturn and the uncertainty about future technology directions, Japan's research and vendor organizations have maintained a high level of quality and responsiveness. They are well positioned to capitalize on any technological or market opportunity and have learned not to blindly follow consensus roadmaps. In particular, the SIA-led International Technology Roadmap for Semiconductors is perceived in Japan as “Supporting Intel's Architecture.” The lithographers in Japan are preparing to go their own way, whichever direction it may go.
Levenson is also the editor of SST's sister publication, Microlithography World.