Three deep-in-the-lab technologies will extend PCs'
relentless power boosts.
Since the invention of the transistor, silicon semiconductors
have been king. But now silicon-based transistors are nearing
the limit of their potential. Excess heat and manufacturing
hurdles are impeding the development of ever-faster and
smaller processors. Advances in materials and chip design to
resist extreme heat and move huge amounts of data, quickly,
will be crucial. Experts are exploring three technologies to
overcome these challenges: spintronics, graphene and
memristors. They are what will someday make
ultra-energy-efficient supercomputers small enough to fit
anywhere—even in the palm of your hand.
Rebuilding RAM
Memristors will store large amounts of data and could make
your computer boot instantly.
Accessing data, whether stored in a spinning hard drive or in
flash-based memory, is a time-suck and a power hog. The
dynamic RAM that rapidly delivers data to the processor is
almost maxed out. “Both technologies for the magnetic hard
disk and D-RAM are within a few generations of hitting brick
walls,” says R. Stanley Williams of HP Labs’s Information and
Quantum Systems Lab. He believes that circuits called
memristors could be the solution. Memristors recently joined
the resistor, capacitor and inductor as the fourth
fundamental circuit element. But unlike the others,
a memristor has the unusual ability to remember the last
resistance it held, even when the power is turned off. When
the current starts up again, the resistance of the circuit
will be the same as it was before, providing instant-on
computers. After the memristor had spent some 30 years as
a theory, Williams and his team designed the first one
earlier this year. Five years from now, he says, the chips
could sit in computers between D-RAM and hard disks to
eliminate the boot-up process. Further down the road,
memristors, which have higher storage densities than the best
flash memory and faster write times than D-RAM, could
supplant both technologies in one fell swoop.
Ditching silicon
Graphene sheets could trump silicon for small, fast devices.
Heat is one major stumbling block to smaller, speedier
processors; as heat increases, it becomes harder for
electrons to move through most materials. But a novel twist
on an age-old material might prove to be the key. Graphite is
made of thin layers of interconnected carbon atoms. Four
years ago, researchers at the University of Manchester in
England tested the electronic properties of a single sheet of
this carbon, known as graphene, sparking interest in what may
be an entirely new branch of semiconductors. This past April,
they built the world’s smallest transistor—one atom thick and
10 atoms wide—with the material. Unlike electrons in silicon,
those in graphene can travel unimpeded for long distances.
This efficiency, which is up to 100 times that of silicon,
allows for ultrafast electronic devices that don’t overheat
(collisions cause heat). In addition to transistors,
researchers hope to develop graphene wires that would
transport electrons from one area of a chip to another much
faster than current materials can.
Making electrons carry information
An electron’s spin could encode more data for heavy-duty,
low-power processing
Most electronic devices read data—the 1s and 0s of binary
code—by measuring the presence or absence of an electrical
charge. In the past decade, however, scientists have
suggested that individual electrons could be turned into
single 1s or 0s. Every electron has a magnetic pole and
corresponding “spin.” But instead of clockwise or
counterclockwise, researchers distinguish electron spins by
their orientation: up or down, 1 or 0.
Until recently, scientists had been able to control and
detect electron spin (which allows the devices to operate at
lower power) in semiconductors such as gallium arsenide,
a material used in laser diodes, but not in silicon—a major
hurdle because the entire infrastructure of computer
manufacturing is still ruled by techniques used with silicon.
Finally, last year Ian Appelbaum, now a physics professor at
the University of Maryland, shot groups of electrons with
aligned spins across 350 microns of silicon and determined
the final angle of their spin. In another development, Igor
Zutic and his colleagues at the State University of New York
at Buffalo have shown that if you align the spins of
electrons in a semiconductor laser, it operates at lower
power. Such lasers could transfer up to 1,000 times as much
data between different parts of a computer as the copper
wires used today.
1. Inject electrons
The spin injector comprises two metals, one of which is
magnetic, separated by an ultra-thin insulator. Apply
a voltage between the two metals, and the electrons tunnel
from the non-magnetic metal to the magnetic one.
2. Filter
The electron continues only if the spin matches the direction
of the magnetic material. If the alignment doesn’t match, the
electron will “scatter.”
3. Travel through the silicon
Electrons travel 350 microns through silicon while
maintaining aligned spins—crucial because electrons need to
travel between intrachip devices.
4. Detect 1s and 0s
Electrons are pushed out through a thin film of magnetic
metal like a ski jumper. If the electron’s spin direction
matches the magnetization of the film, it will land on
a nearby semiconductor, where its charge is detected (say,
a “1”). Mismatched electrons are
shunted elsewhere (a “0”).
Aucun commentaire:
Enregistrer un commentaire