Is there any end to smaller, faster,
cheaper microprocessors? Well, yes. But the end may be farther out than
previously thought. Silicon-based transistor technology may not run out of steam
for a dozen years, instead of the previous estimate of fewer than six years,
report researchers at Lucent Technologies' Bell Labs.
Conventional wisdom within the
semiconductor industry has identified a chip's silicon-dioxide insulating layer
as the limiting factor for producing increasingly smaller transistors. The
insulating layer on today's chips is 25 atoms thick, but Bell Labs researchers
recently produced a 5-atom layer, the thinnest ever made. They also showed that
a 4-atom layer is the fundamental physical limit for silicon dioxide–based
insulators. These results suggest that an alternative insulating material must
be found before 2012. If alternative insulating materials aren't found, totally
new technologies will be needed.
Chemical Assembly
One of these new technologies, known as chemically assembled electronic
nanocomputers (CAENs), harnesses chemical reactions that may eventually produce
tiny processors that are billions of times faster than those of today and could
be manufactured at a fraction of the cost. As reported in the journal Science
today, scientists at Hewlett-Packard and the University of California, Los
Angeles' Department of Chemistry and Biochemistry recently made a significant
breakthrough by successfully producing a logic-gate--the building block of a
processor--out of a synthetic molecule called rotaxane. The significance of the
experiment lies in the molecular dimensions of the gate.
By sandwiching a layer of rotaxanes
between metal electrodes and electrically charging the molecule, the researchers
observed the molecule behaving as an on/off switch. They also strung several
switches together to function as AND and OR logic gates, the basic computational
language of computers. According to the researchers, "any chemically
prepared system will be ordered (that is, crystalline) and defective (due to
finite chemical reaction yields), whereas any reliable computational machine
requires perfect complexity." The problem, however, could be solved by
software that during the manufacturing process, wires up only the molecules in
good working order. In the experiment, the rotaxane molecules could be set only
once, making them suitable for read-only memory, but not random access memory,
which requires constant switching. The researchers work, though, isn't finished.
If such tiny, cheap processing power
does come to fruition, the researchers imagine all kinds of uses. For example,
in biomedicine, "they'll be able to snuggle up to a bacterium and determine
if it's tuberculosis, and even what type of TB it is," says Phil Kuekes, a
computer architect at HP who headed up the research along with HP chemist Stan
Williams and a UCLA team led by chemistry professor James Heath.
The chemical assembly of processors is
promising for the microscopic scale it enables, which means that Moore's Law
(predicting that transistor density will double every 18 months) will hold true
well into the future. But the technology is also promising because it's cheap.
The little-known Moore's Second Law states that the cost of the manufacturing
plants used to build chips is increasing faster than the growth in demand for
chips. According to some predictions, by 2010, the price tag of a new plant
producing conventional CMOS (complementary metal-oxide semiconductor) chips
could cost between $30 and $50 billion. Says HP's Williams, "It is very
likely that the economic consequences of Moore's Second Law could be the major
factor that causes his first law to end.
