Technology News
01/01/2002
Curing blindness, limitations of CMOS are topics at MNC
The International Microprocesses and Nanotechnology Conference (MNC) is Asia's premiere academic technology symposium, and this year's version did not disappoint, with leading-edge topics such as the limitations of CMOS, implantable electronics for correcting blindness, DNA electronics, and supporting R&D efforts in nanotechnology headlining the sessions and papers.
Power dissipation and the limitations of CMOS dominated the plenary talk given by Tadahiro Kuroda of Keio University. Kuroda pointed out that CMOS replaced bipolar technology 20 years ago because high-performance bipolar chips produced an untenable amount of heat. Today, CMOS is in a similar situation, with some chips dissipating more heat/cm2 than a hot plate does. There is, however, no obvious low-power replacement technology coming on line. Thus, one must adopt CMOS design methodologies to reduce power consumption and try to stretch the lifetime of CMOS for another 6-20 years. Kuroda advocated three power-reducing strategies: lowered voltage, reduced capacitance, and reduced switching activity. Variable threshold voltage CMOS technology (VTCMOS) can reduce power consumption at constant performance using multiple supply voltages and on-chip DC-DC converters. Embedded DRAM reduces capacitance by eliminating parasitics and drivers, and conditional flip-flop logic (where the internal clock is activated only when the function is needed) can reduce switching activity. Taken together, these innovations promise to reduce power consumption by 50-90%, extending the lifetime of CMOS by up to 20 years, according to Kuroda.
Wentai Lai of North Carolina State University surveyed the opportunities and difficulties for developing electronic prostheses to correct blindness and other disabilities. The cochlear implant has been a tremendous success in curing deafness, but vision is more complex and delicate.
He proposed an intraocular retinal prosthesis, with only the retinal transducer inside the eyeball and everything else outside communicating with it via wireless technology. Such a system would minimize power dissipation issues, which would be very significant inside a living eyeball. A low-resolution prototype with a 30x30 array of electrodes was reported to be complete. The "vision" produced might be crude, but useful once the brain learns how to interpret the sensations it receives. Solving bioengineering problems requires a multidisciplinary team, according to Lai, and that carries with it coordination and dominance challenges among doctors, scientists, and engineers.
Carbon nanotubes have inspired nanotechnologists, but in spite of their evident promise, have been little used in silicon microelectronics. Wolfgang Hoenlein of Infineon Technologies AG, Munich, Germany, described their potential for conducting electricity and heat in ICs. Infineon has found a catalytic method of growing aligned nanotubes on a substrate to form interconnects and via conductors.
Only contact resistance matters for metallic nanotubes; electrons move along them without scattering even from surfaces. Thus, bundles of nanotubes can have lower resistance than copper for sub-100nm damascene interconnects if the density is sufficient. Other more visionary opportunities include silicon-free 3-D ICs fabricated using metallic and semiconducting nanotubes.
Beyond nanotubes, nanotechnologists envision entire new "molectronics" technologies based on DNA and other biomimetic systems, fabricated by imprinting and self-assembly and which completely avoid the brick wall obstructing semiconductor progress in 2015. C.R.K. Marrian of the US DARPA outlined the National Nanotechnology Initiative (NNI) and its initial successes, perhaps in hopes of inspiring parallel efforts in Japan and other nations. The NNI has funneled most of its resources to academics like those attending the MNC meeting, but part of the increase has resulted from funds diverted from previously approved programs, according to Marrian.
Tomoji Kawai of Osaka University explained some of the reasons for frustration among those working toward DNA electronics. While DNA is described as a "molecule," its electronic properties depend critically on the sequence of bases along the famous double-helix strands as well as the 3-D configuration of the chain. Changing the helix pitch — which can easily happen when DNA lies on a surface or gets probed by an STM tip — radically alters the electronic behavior. The inconsistent results in the literature partly reflect actual differences among the individual molecules probed.
In general, DNA is a wide-gap semiconductor, according to Kawai, with poly(G)-poly(C) dual chains being p-type and poly(A)-poly(T) being n-type. Everything else is more complicated but perhaps more promising. In particular, 50-unit strands of single-chain DNA can be made to braid together into 2-D networks of 3-terminal nodes, potentially useful as neural-nets.
On a somewhat larger scale, K. Hirao of Kyoto University described how femtosecond pulse lasers can be used for debris-free machining and materials modification. Even stainless steel can be drilled cleanly by breaking the molecular bonds with short laser pulses! A femtosecond beam focused into conventional fused silica can precipitate the high-density amorphous phase, creating an embedded waveguide with complex 3-D geometry.
Other sessions addressed such seemingly esoteric topics as MEMS, microfluidics, photonic devices, nanoimprint lithography, metrology, and nanodevices. Today's academic topic, however, becomes tomorrow's leading-edge technology as soon as the students graduate and form their own start-ups. The vitality of the MNC meeting demonstrated that neither persistent recession nor terrorism can stop innovation, so long as brilliant people are willing to enter our field and pursue new useful knowledge. — M.D.L.
____________________________
Diagonals: A shorter path for interconnect
Increased complexity and functionality in semiconductor devices has placed a greater emphasis on interconnects, where delays have become one of the most significant factors affecting chip performance.
Efforts to minimize interconnect delay have largely concentrated on new process technologies such as dual damascene, and new materials, such as copper and low-k dielectrics, rather than on the physical layout.
Today, sophisticated computing resources and innovative development efforts have given rise to a fundamentally different approach to interconnect design, the X Architecture. Jointly developed by Toshiba Corp., Tokyo, Japan, and Simplex Solutions, Sunnyvale, CA, this architectural approach uses diagonals to shorten interconnects.
Ultimately, this could mean smaller, faster devices with decreased power consumption.
For the past 20 years, chip design has been primarily based on the "Manhattan" layout, an architecture built on right angles. X Architecture uses diagonals: In complex, multiple-metal-layer devices, the primary direction of the interconnect can be rotated on the fourth and fifth metal layers by 45°, enabling a reduction in total wiring on the chip by more than 20%. Based on Simplex's "liquid routing" technology, wiring can be routed in eight directions (four diagonal and four Manhattan) on every layer (see figure).
Therefore, a more direct connection between any two transistors can be established, regardless of their proximity to one another, thus minimizing wiring on the chip. Initial evaluations indicate that this wire-length reduction will enhance chip performance by 10%, decrease power consumption by 20%, and result in 30% more working chips on a wafer, said Jan Willis, VP of business development at Simplex and X Initiative steering group facilitator.
To build an infrastructure, a consortium called the X Initiative was formed, which will continue to focus on removing barriers to adopting this new approach to interconnect design. — R.D.