By BBC News Online science editor
Dr David Whitehouse
Scientists at Cambridge University
say they have developed a new
generation of magnetic microchips
that may herald the beginning of
faster, more efficient computers and
electronic devices.
Researchers say that the chip stores
data in the form of tiny magnetic
fields and that versions of the chip
currently being tested are up to
40,000 times more efficient than the
electronic chips in use today.
Dr Russell Cowburn said the chips,
which he and colleague Professor
Mark Welland have developed at the
University's Department of
Engineering, were likely to
revolutionise the manufacture of
microchips.
In a paper published in the journal
Science, Dr Cowburn describes the
advantages the new chip has over
existing technology.
"There are two big differences
between our chips and others. The
first is size: most existing electronic
chips can fit up to 6.6 million
transistors (the basic building blocks
of all microchips) into one square
centimetre.
"But we have demonstrated a device
that can fit 5,500 million transistors
into a similar area. We estimate in
future years this could increase to
250,000 million transistors."
"The other big advantage is energy
consumption. Electronic chips use up
energy during operation, whereas a
magnet does not."
"That means computers developed
using magnetic microchips will need
much less power to work. The days of
carrying around heavy batteries for
laptop computers and mobile phones
are numbered!"
Professor Welland added that it
would be several years before the
new technology could be developed
commercially, but he said a
completely new type of computer
would be made using this method
Scientists at the US Department of
Energy's national security facility
have created a micro-machine etched
from amorphous diamond, the
hardest material in the world after
crystalline diamond.
The methods used were the same as
those used with current silicon chip
manufacturing techniques.
The first diamond micro-machine is a
comb drive whose tiny teeth move
forward and back as an electrical
current reverses constantly between
positive and negative. The teeth are
just two thousandths of a millimetre
apart.
The researchers, at the Sandia
National Laboratories in New Mexico,
say diamond has a number of
valuable properties.
Resistant to wear
Its resistance to wear makes
diamond ideal for micro-machines
that need to function for extended
periods of time.
"One estimate in the literature claims
that diamond should last 10,000
times longer than polysilicon," said
group member Tom Friedmann.
Also, diamond is
less susceptible
to "stiction" than
silicon. Stiction -
a combination of
stickiness and
friction - can
render
micro-machines
useless. This is because silicon is
attracted to water, which acts as a
kind of glue. Diamond does not have
the same problem.
Another advantage is that diamond is
biologically benign. A micro-machine
such as a tiny drug dispensing unit
could be used in the body without
triggering an allergic reaction.
There are two kinds of diamond,
crystalline and amorphous. The
Sandia researchers used amorphous
diamond because crystalline diamond
needs far higher temperatures to
synthesise it, and also its surface
roughness makes it unsuitable for
micro-machines.
Amorphous diamond itself had been
impractical because its tremendous
internal stresses had made it
impossible for the material to stand
alone or to coat thickly any but the
strongest surfaces.
However, a process developed by
Friedmann and Sullivan eliminated
that problem.
Silicon MEMs (MicroElectroMechanical
Systems) are already used in a
variety of applications, ranging from
air bags in cars to optical
micro-mirrors intended for possible
deployment on satellites.
It's hoped that diamond MEMs could
eventually offer a harder wearing and
more flexible alternative, and in the
future could replace silicon
micro-machines completely.
-top-
RNA computer clears 10-bit hurdle
By R. Colin Johnson
EE Times
(02/01/00, 11:05 a.m. EST)
PRINCETON, N.J. ; Princeton University researchers claim to have
reached a new level of complexity in DNA computing. The group has
demonstrated an RNA-based computer capable of solving mathematical
problems that were encoded as 10-bit strings.
Strands of RNA containing 1,024 base pairs were encoded with every
possible solution to a specific chess problem. Ribonuclease digestion
progressively narrowed down the possible solutions until only the 43
correct solutions ; plus one incorrect one ; remained.
"Molecules can store more information than silicon chips, and this was
the largest problem ever solved by a molecular computer ; using
either DNA or RNA. We also learned how far we can push this technology
when we discovered why it made a single error," said professor Laura
Landweber, the leading Princeton researcher on the project.
Landweber's colleagues are professor Richard Lipton and post-doctorate
candidates Dirk Faulhammer and Anthony Cukras.
[INLINE]
In 1994, Leonard Adleman of the University of Southern California-Los
Angeles conceived of harnessing DNA to solve tough computation
problems. He demonstrated his concept on the classic seven-city
traveling-salesman problem using the base-four encoding of DNA. After
synthesizing trillions of DNA strands to stand for every possible
solution, his test tube finally came to the correct conclusion after
about a week.
Tough combinatorial problems like that chosen by Landweber's Princeton
team present significant hurdles to any finite-size computer. The
possible solutions to such problems expand so fast, Adleman reasoned,
that even a few variables will result in a problem so complex that
only approximate solutions can be found.
Toy-sized problem
The Princeton research team's problem was toy-sized, having a mere
1,024 possible solutions, but the group claims the basic nature of
the
single error indicates that it can be scaled up to real-world-size
problems. Apparently, the error came from a rare source that will not
increase geometrically with the higher-dimensional solution spaces
of
real-world problems.
"We just had a bit of bad luck ; or more literally two bits,
since it was two single-point errors in a row that foiled our
algorithm," Landweber said. "Two errors in a row are exceedingly rare
and shouldn't become a problem when we scale up."
The Princeton team substituted RNA for DNA to enable the use of a
universal enzyme that targets any part of a molecule. DNA has only
a
limited set of restriction enzymes, so scientists may not be able to
cut the molecule where they want. The group demonstrated that its
streamlined approach using RNA could inherently scale up to
real-world-size problems by virtue of the universal enzyme.
Using a combination of binary RNA libraries and ribonuclease
digestion, the Princeton team created a destructive algorithm that
acts as a "universal restriction enzyme" ; a kind of molecular
scissors that can selectively cut out any specified strand for
digestion (removal). The thermostable reaction proved impervious to
single-bit errors, ensuring the fidelity of the hybridization.
Since the algorithm (by virtue of its universal restriction enzyme)
operates in parallel on every molecule in the test tube, it can make
trillions of parallel computations theoretically possible. In the
Princeton test, for example, it took only a series of five steps to
target all of the enzymes that slashed away those coded strands that
represented the incorrect solutions.
Knight moves
The particular problem posed to the RNA computer was how many ways
there are to place knights on a chess board so that none can take any
other. A 3-by-3 chess board was chosen so that each of its nine
squares could represent a bit in the RNA strand. A 10-bit library of
RNA strands gave a 1-bit buffer, used to correct 1-bit errors
chemically.
There were a total of 43 correct solutions among 512 total
possibilities; the RNA computer found all 43 correct ones and one
incorrect one.
For the future the Princeton team intends to go after problems that,
while not yet approaching real-world proportions, scale beyond the
toy
category. "The class of general satisfiability, where we got this
chess problem, is naturally extendible. We might go to larger boards
or different variations on the chess problem," said Landweber.
Landweber's department, evolutionary biology, tracks the biological
evolution of species, including that of biological computers. Her
particular take has been to apply evolutionary biological principles
to what she calls molecular evolution.
"I want to trace the evolution of the molecules themselves from their
earliest emergence up to the complex organic molecules of today. In
molecular evolution it was the correct solutions to biological
computations that had a better chance at survival," said Landweber.
Her aim is to identify the component molecules, along with their
mutual reactions and their interrelationships, that have evolved to
create the biological computer called the human brain. Armed with the
correct molecules and their known reactions, previously intractable
problems could turn out to be child's play for trillions of
parallel-processing molecular computers.
CMPnet
All material on this site Copyright © 2000 CMP Media Inc.