Unless we find a replacement for field effect transistors Moore's law will
eventually come to an end, although IBM's recent technique of using copper
instead of aluminum in a IC probably gave us an additional 5 or 10 years
breathing room. Power dissipation in a FET can be a real problem, as is
trying to interconnect billions of different transistors, even worse, when
the FET channel becomes very short quantum effects start to seriously degrade
performance. We can't change the laws of physics so we need a way to turn
Quantum Mechanics from an enemy to a friend.
Back in 1991 Craig Lent proposed we abandon transistors and capacitors, stop
using voltages to encode information and start using individual electrons in
a Quantum Cellular Automation (QCA). His idea was to have each cell of the
automation be made of 4 quantum dots arranged in a square and have each cell
contain 2 extra electrons. The electrons repel each other so they will always
try to get into quantum dots as far apart as possible, that is the dots at
the diagonal of the square. There are only 2 stable states such a cell can
be in:
_____________ _____________
|X 0| |0 X|
| | | |
| STATE 0 | | STATE 1 |
| | | |
|0___________X| |X___________0|
There is no direct connection between the 4 quantum dots but the 2 extra
electrons can move from dot to dot in a cell by quantum tunneling.
The electrons can't tunnel to other cells in the automation (a small increase
in distance means a huge decrease in the probability of successfully
tunneling) but the electrons in each cell can still feel how the charge is
distributed in a neighboring cell and will always stay as far apart as
possible. For example if you have a long row of such cells and you change the
first one to state 0 then that will change the second cell to state 0 and
that will change the third cell to state 0 and so on all the way down the
line. A signal sent in this way uses no current so almost no heat is
generated.
Because a cellular automation interacts with adjacent cells you could
intertwine several such "wires" to process information. Lent also designed
but did not build a quantum cellular automation adder.
Six years ago not many were very impressed with Lent's idea because they felt
the dots could never be made accurately enough and the slightest imperfection
would let the 2 electrons leak away in about a nanosecond. In the August 15
1997 issue of Science Lent and his associates report they have actually build
one cell of the automation. The cell operated just as Lent said it would,
it can be in only 2 states and can stay in one state for many minutes and the
electrons do not leak out.
The quantum dots in this prototype are huge, 8 microns across, so he had to
keep it at cryogenic temperatures, but Lent says in the article "QCA
architecture is scaleable to molecular levels and performance actually
improves as the size of the device is decreased. [...] a molecular sized QCA
would function at room temperature." Unlike transistors, the smaller these
things get the better they work
In an editorial in the same issue of Science Terry Fountain of University
College London call this new work "A new and promising possibility to shrink
computer architecture [...] by factors as much as 50,000 compared to the
smallest feasible transistor. [...] the remaining hurdles though daunting are
largely technical".
John K Clark johnkc@well.com
-----BEGIN PGP SIGNATURE-----
Version: 2.6.i
iQCzAgUBNDr/1X03wfSpid95AQGf1ATvVOSGbQ5ZNJWA+1TgFYUPM21vCYxYVlt9
xO3Q7Xcc53X+yD4U6u3hv9lL2KIKH7AA7fXPjickpK8dvPM8QzwMBbcU3pclDm/O
86EQwiw+kfm5y/39O2jXkjmcX89Yvake5PHmYx5GXk+VkLGo4M+MEHR4lo42kt2a
NFrlX9sU2EQ7P1MCAxGL7TDx1NRkJKzSY8NDergoOr2BRd+Vjs4=
=fK0S
-----END PGP SIGNATURE-----