Moore's Law over 20 years

From: hal@finney.org
Date: Tue Apr 25 2000 - 17:42:39 MDT


I saw a pointer on slashdot yesterday to this article at EE Times:
http://www.eet.com/story/OEG20000421S0024. It discusses some of the
issues relating to future progress in miniaturization of electronic
circuitry.

According to the chart, chips are currently being made with feature sizes
as small as .18 microns. This is 180 nanometers for what they call a
"node", of which the "gate" is 140 nm. (Unfortunately I have only a
sketchy understanding of the geometry of silicon devices.)

The interesting thing is that the transistors rely on a thin resisting
layer of silicon dioxide, and that is only about 23 angstroms wide,
or 2.3 nm. Atoms themselves are about 2 angstroms in diameter so these
resistive layers are already only about 10 atoms thick.

They want to begin moving to feature sizes of 70 nm (.07 micron), which
requires SiO2 layers only 15 angstroms thick or less. The article is
about how it is now thought that such thin films will be more stable
than had been feared. Previously, workers were concerned that films
thinner than now in use would be damaged by electron flow and get holes,
making the lifetime of the parts too low to be useful. New work suggests
that this problem is not so severe.

The article doesn't say so, but I think the problem is that this
insulating layer has to withstand the voltage used by the chip. This is
now getting down to just over a volt, but across 15 angstroms that is
about a billion volts per meter. Imagine a million volt difference
across a millimeter, and you can see that sparks will fly.

Once they get to the point where SiO2 will break down, they need to
go to a different insulator: zirconium, hafnium or titanium oxides.
I don't know if the semiconductors themselves would still be made of
silicon or whether this would imply going to a whole new chip substrate.
In any case the industry is reluctant to take that step because of their
30 year track record with SiO2.

The other interesting thing about the article was their long term
projection for shrinking feature size. It shows a decrease from 180 to
25 nm over 18 years, a factor of 7.2. If density is proportional to the
square of the feature size, that would be a density increase of 52, about
6 doublings. 6 doublings in 18 years is one doubling every three years.
That would represent a halving of the traditional rate of increase of
Moore's Law, which has been quoted as a doubling of density every 1.5
years (actually variants using 1 and 2 years also exist).

So it appears that the industry does have a road map for at least
approximately maintaining Moore's Law over the next 20 years, without
going to any exotic technology like nanotech. There are many problems
which must be solved to achieve these gains, but it appears that in
principle it can be done.

Hal



This archive was generated by hypermail 2b29 : Thu Jul 27 2000 - 14:09:49 MDT