Moore's Prophecy: The plot thickens
02/01/2003
At a recent Semi breakfast, Doug Neugold, president of ATMI, suggested that Gordon Moore didn't really propose a Law in 1965. Instead he observed a trend and predicted that the industry would continue to double integrated circuit densities roughly every 18 months. His forecast, which Neugold suggested would better be called Moore's Prophecy, has proven to be valid for nearly 40 years, and appears likely to continue for at least another decade.
As with anything involving the laws of nature, however, the closer you get to physical limits, the trickier it becomes to get everything working together, particularly when quantum effects such as tunneling begin to intrude. The working groups that put out a new International Technology Roadmap for Semiconductors (ITRS) every two years are finding it harder and harder to keep at bay those red blocks they refer to as the "red brick walls," identifying future problems we don't know how to solve.
Yet solutions are being found. It is truly amazing how rapidly the industry can shift gears, reinvent itself, and zoom steadily ahead at this exponential pace. The innovativeness and creativity of thousands of scientists, researchers and engineers worldwide that allow this "Prophecy" to continue to remain true for decades is truly remarkable. Yet much of their work, done in cluttered labs and with lots of scribbling on blackboards, remains unknown to the general public.
While nature and basic physical laws may offer challenges, these are not the only tough constraints. Economics is critical, too. Today's chipmakers want to push ahead with tools and processes geared predominantly to making CMOS-based logic and memory devices. Requirements must be met with minimal added process steps, and with a reasonable process flow, taking into account the heating effects of annealing, for example. Yet researchers succeed in meeting the challenges time after time.
Shrinking transistors will require such thin insulators under the gates that new materials with higher dielectric constants will be essential to curb leakage currents. Researchers began to scour the periodic table for candidates, sometimes needing to measure fundamental properties that had never been examined before. Zirconium and lanthanide (rare earth) oxides, particularly lanthanum and hafnium, quickly proved promising.
During the same week in early December, important progress was reported at the Materials Research Society's Fall Meeting in Boston and the IEEE's International Electron Devices Conference in San Francisco. Materials researchers in the East were exploring how high-k dielectric materials might be modeled and looked at fundamental properties of ZrO2 and HfO2, while in the West there were a number of reports on processes to make better films from high-k candidates such as these.
Soon, metal gates may be needed to replace polysilicon in order to avoid gate depletion effects, and various metal gate processes were reported at the Electron Devices meeting, including some with tunable work functions. The basic CMOS transistor structure does not scale well into the sub-50nm range, so devices based on dual gates,vertical transistors, and "finFETs" were described. For faster, lower-power circuits, some researchers reported work on fully depleted and partially depletedsilicon-on-insulator (SOI) devices, while others favored the use of strained silicon layers to boostcarrier mobilities.
What do we do, though, when things REALLY get small. With only a few atoms of dopants at the junctions, it may become impossible to get uniform device characteristics all over a wafer. That's when thenanotechnologists will have to take over, working at the molecular level. The idea will shift from trying to play off physical effects against each other, as we do now at the macro level, to letting nature do our work for us. At the molecular level, regular structures often emerge naturally through a process of self-assembly. Carbon nanotubes, based on such principles, are already close to being used as field emitters in commercial field emission displays, and work on the possibilities for using them as interconnects and even transistors was reported at IEDM and the MRS Meeting. We are in just the early phases of finding out how we can work with nature rather than around it.
No one yet knows for sure how far we can follow the path laid out for us by Gordon Moore in 1965. Chances are, we will travel far beyond what seems reasonable today. Hear them scribbling on those blackboards? Oops — today it's more likely to be whiteboards (or computer screens). And tomorrow, with organic LEDs, researchers may just jot down their ideas on the walls.
Robert Haavind
Editor in Chief