PASADENA, Calif. — The future of computing may be in its past.
The
silicon transistor, the tiny switch that is the building block of
modern microelectronics, replaced the vacuum tube in many consumer
products in the 1970s. Now as shrinking transistors to even more
Lilliputian dimensions is becoming vastly more challenging, the vacuum
tube may be on the verge of a comeback.
In
a darkened laboratory here, two stories beneath the California
Institute of Technology campus, two students stare through the walls of a
thick plastic vacuum chamber at what they hope will be the next small
thing — a computer chip made from circuits like vacuum tubes whose
dimensions are each roughly one-thousandth the size of a red blood cell.
At
stake is the future of what electronic engineers call scaling, the
ability to continue to shrink the size of electronic circuits, which is
becoming harder to do as they become as small as viruses.
It
has been more than half a century since the physicist Richard Feynman
predicted the rise of microelectronics, noting “there’s plenty of room
at the bottom.” He used the phrase in 1959 when he speculated about
engineering with individual atoms. Several years later, Gordon Moore,
co-founder of Intel, wrote that the number of transistors that could be
etched into silicon wafers would double at regular intervals for the
foreseeable future.
Now, however, there is growing evidence that space, if still available, is increasingly at a premium.
Progress is slowing down. The time between each new chip generation is
stretching out, and the cost of individual transistors, although
infinitesimal, is no longer falling. The tiny transistors also bedevil
chip designers because as they get smaller, they generate unwanted heat.
For Axel Scherer, who heads the Nanofabrication Group
at Caltech, that means going back to the future. With his students Max
Jones and Daniil Lukin, he is pursuing what is in effect an ultrasmall
vacuum tube as a candidate to replace the transistor. In their
laboratory here, they have fabricated circuits that function like vacuum
tubes but are a millionth the size of that 100-year-old technology.
“Computer technologies seem to work in cycles,” said Alan Huang, a former electrical engineer for Bell Laboratories. “Some of the same algorithms that were developed for the last generation can sometimes be used for the next generation.”
Continue reading the main story
The
last time researchers explored vacuum tubes was in the 1990s, when they
were a promising option for building flat-panel displays. The
technology failed to take off, however, because of cheaper and more
efficient liquid crystal displays.
“The vacuum tube comes back about every decade,” Dr. Scherer said with a laugh.
And
for decades, that has been the story of vacuum tubes: There has always
been a better option. Transistors replaced vacuum tubes because they
were more compact, did not generate skin-burning heat and did not need a
vacuum — the absence of atmosphere made it possible for electrons to
jump between positively and negatively charged elements.
The
vacuum tubes the Caltech researchers are looking at are nothing like
the bulky objects that hummed in the old family radio and even early
computers. Both transistors and vacuum tubes — the British called the
devices valves — control the flow of electricity, but they do so
differently.
The
researchers have created a tiny tube formed from metal and capable of
turning on and off the flow of electrons between four even smaller
probes, which under an electron microscope appear like the tips of four
ballpoint pens almost touching one another.
The
Achilles’ heel of today’s transistors is the smaller they get, the more
they leak electrons. In modern computer chips, as much as half of the
power consumed is lost to electrons leaking from transistors that are
only dozens of atoms wide. Those electrons waste energy and generate
heat.
In
contrast, Dr. Scherer’s miniature vacuum tube switches perform a
jujitsu move by using the same mechanism that causes leakage in
transistors — known by physicists as quantum tunneling — to switch on
and off the flow of electrons without leakage. As a result, he believes
that modern vacuum tube circuits have the potential to use less power
and work faster than today’s transistor-based chips.
“Effects
that are currently problems in scaling are precisely those that we
would like to use for switching in these next-generation devices,” Dr.
Scherer said, noting that while there are efforts to redesign
semiconductor-based transistors around the tunneling effect, his
approach is significantly simpler.
Vacuum
tubes are one of a range of ideas that engineers are looking at as they
work to create chips that can do more while using less power. Other
promising approaches include exotic materials such as carbon nanotubes
and even microscopic mechanical switches that can be opened and closed
just like an electronic gate.
The
Caltech researchers returned to the idea of vacuum tubes several years
ago after they had begun experimenting with the idea of making
ultrasmall incandescent light bulbs no larger than a modern transistor
that would be bright enough to be seen by the naked eye from across a
room.
The
group previously worked in research areas like quantum dots, nanoscale
structures now used in television displays to produce precise colors,
and optoelectronics, a field that explores the use of lasers in
electronic circuits. But they decided to look for new research areas
that were less crowded with competitors.
Today,
semiconductor companies like Intel are making silicon chips with
minimum dimensions between 10 and 20 nanometers. (A strand of DNA is
roughly 2.5 nanometers in diameter.) Once the industry shrinks below 10
nanometers, Dr. Scherer expects that researchers will be surprised by
the behavior of silicon at such atomic dimensions.
For
one thing, silicon emits light below 10 nanometers, he said. More
significantly, it also becomes remarkably elastic as it becomes that
small.
“It’s
a different material, and it gives you this different behavior,” he
said. He sees the future in other materials and in old ideas that would
be made new again.
In
contrast to silicon, a semiconductor, which can either conduct or
insulate, depending on how it is chemically modified, Dr. Scherer’s
tubes can be made from a range of conducting metals, such as tungsten,
molybdenum, gold and platinum. This will be an advantage because it will
significantly simplify the tiny switches at the atomic scale.
Dr.
Scherer does not think the tiny tube will immediately replace the
transistor, but the possibility of applications in space and aviation
has caught the attention of Boeing, which is financing the research. Such specialty chips might be ready commercially before the end of the decade.
“Ten years ago, silicon transistors could meet all of our demands,” he said. “In the next decade, that will no longer be true.”
No comments:
Post a Comment