Technology Research News
The ultimate in miniaturization is the
atom -- there are 10 million billion of them in a single grain of salt.
The scientist Richard Feynman suggested several decades ago that it would
be possible to use single atoms to store bits of data. Researchers from
the University of Wisconsin at Madison have taken a large step toward
making the idea a reality with a prototype that uses single silicon atoms
to represent the 1s and 0s of computing.
Practical atomic-scale memory would increase the amount of information
that could be stored per square inch of recording material by several
The researchers realized they had hit upon a mechanism for atomic memory
when they discovered that scattering gold atoms on a silicon wafer caused
the silicon atoms to assemble into tracks exactly five atoms wide. The
pattern resembled the microstructure of a CD.
Making the tracks turned out to be relatively easy. "We can actually get
atoms to assemble themselves... precisely, without any type of lithography,"
said Franz Himpsel, a professor of physics at the University of Wisconsin.
"It is actually quite simple, and my graduate students make the surfaces
routinely now," he said.
The breakthrough that made the prototype possible was working out a practical
way to write data into the memory, Himpsel said. "In general, it is difficult
to work with an individual atom in a controlled way, without affecting
neighboring atoms," he said.
The researchers initially tried to write information to the memory by
moving atoms along the tracks. "That works well at very low temperatures
with loosely-bound atoms, but not at room temperature where we wanted
the memory to operate," said Himpsel. Eventually, "instead of moving them,
we [picked] up the atoms," using a scanning tunneling microscope, he said.
The researchers found that the optimum spacing for each bit is a four-atom
section of track. This makes the bit spacing 1.5 nanometers along the
tracks and 1.7 nanometers between tracks, which amounts to a data storage
density of 250 trillion bits per square inch. This is equivalent to storing
the contents of 6,650 DVDs in one square inch of material. A nanometer
is one millionth of a millimeter.
The researchers formatted the memory by depositing extra silicon atoms
onto the tracks to make every bit a 1, then wrote data by removing atoms
to represent 0s.
To read the data, they scanned along the tracks looking for the presence
or absence of the extra atoms.
The approach bridges two broad camps of work on data storage, said Himpsel.
"One group of researchers manipulates atoms with great precision at very
low temperature," he said. The other group improves existing memory devices,
which use thousands of atoms per bit. "We attempt to bridge the gap between
the two camps by sticking to the atomic density limit, but... obtaining
realistic numbers for the ultimate performance limits of a memory, such
as speed, error rate, and stability" at room temperature, he said.
The researcher's prototype resembles the way nature stores data in DNA,
said Himpsel. The memory structure self-assembles into the tracks. In
addition, "the density and readout speed of DNA [is] quite similar to
our silicon memory," he said. While DNA uses 32 atoms to store one bit
using one of four base molecules, the researcher's silicon memory uses
20 atoms including the atoms between the individual atoms that store the
bits, said Himpsel.
The prototype is clever, said Phillip First, an associate professor of
physics at the Georgia Institute of technology.
Writing atomic bits is impractically slow at present, but the work is
a realistic analysis of bit stability, which is good, recording density,
which is high, and readout speed, said First. It is "a very impressive
demonstration of the practical limits of two-dimensional data storage
using single-atom bits," he said.
Although the prototype is very slow at reading data compared with readout
speeds of magnetic disks, the researcher's work shows that this can be
improved, said First. The fundamental limit of readout speeds has to do
with signal-to-noise ratio, and the researchers showed that this limit
is substantially higher than the speed they were able to achieve with
the current prototype. If the researcher's storage media were combined
with microelectromechanical systems that allowed for parallel data readout
from the memory, "the ultimate data rate could be comparable to magnetic
disks," according to First.
The researchers intend to boost the readout speed next, said Himpsel.
"An immediate next step is the use of fast scanning electronics to speed
up the readout to its theoretical limit, which is 100,000 times faster
than our simple electronics allowed," he said.
Because the readout speed decreases as the bit size gets smaller, when
practical atomic storage devices are eventually made, scientists are likely
to strike a balance between storage density and performance that falls
short of the 20-atoms per bit density the researchers have proven is possible,
said Himpsel. "Somewhere on the way to the atomic limit is an optimum
combination of density and speed."
The researchers are also working on finding the optimum coding and signal
filtering methods for the memory in order to cut down on the error rate,
Himpsel said. Using more than a single atom per bit could also reduce
And the researchers eventually need to tackle the as yet unsolved problem
of making the memory work outside of a vacuum, said Himpsel.
Although the research shows that it is possible to store bits of data
in single atoms, there is a lot of work to be done before the technology
could be made practical, said Himpsel. It will take "decades," he said.
"I will be retired by the time we have atomic-scale memory in use."
This type of memory may eventually become useful for storing vast amounts
of data, but because the stability of each bit of information depends
on one or a few atoms, it likely to be used for applications where a small
number of errors can be tolerated. "I would not want to trust my bank
account to a memory where a single atom could wipe out my savings," said
Himpsel. "However, for pattern recognition -- face or handwriting recognition
-- it is not critical that all the pixels are stored perfectly," he said.
Himpsel's research colleagues were Roland Bennewitz of the University
of Wisconsin at Madison and the University of Basel in Switzerland, and
Jason Crain, Armen Kirakosian, Jia-Ling Lin, Jessica McChesney and Dmitri
Petrovykh of the University of Wisconsin at Madison. They published the
research in the July 4, 2002 issue of the journal Nanotechnology. The
research was funded by the National Science Foundation.
Timeline: > 20 years
TRN Categories: Data Storage Technology; Nanotechnology;
Story Type: News
Related Elements: Technical paper, "Atomic Scale Memory
at a Silicon Surface," Nanotechnology, July 4, 2002.
Ultimate memory demoed
makes bugs sing
Nanotubes grown in place
Quantum secrets ride
Chip keeps atoms in line
Research News Roundup
Research Watch blog
View from the High Ground Q&A
How It Works
News | Blog
Buy an ad link