 |
Berkeley breathes
new life into silicon, continued
Back to
Page 1 | Page
2
The problem is that as the gate length is shrunk and the source
and drain are pushed so close together (less than 100 atoms apart),
more electrons can sneak through during the "off" state.
"There's no such thing as zero in engineering, but the goal
is to make the current as minimal as possible" when the transistor
is off, Bokor says. "When chips have hundreds of millions
of transistors, even a tiny amount of leakage per transistor adds
up."
In the FinFET design, a thin vertical silicon "fin"
is built between the source and drain. Then the gate electrode
material is deposited on both sides of the fin resulting in a
double gate structure, one on each side of the channel.
 |
| MicroLab researchers
such as Bokor and King (the lab's director) suit up from head
to toe in "clean suits"(they call them bunny suits)
to prevent even the smallest amount of contamination from
entering the lab or its extremely sensitive equipment. The
lab has more than 300 active users, including researchers
from nine departments on campus and users in private industry.
Peg Skorpinksi photo |
"It's like trying to stop a bleeding vein," Hu says.
"You could press down on the vein, but it would be much more
effective if you could get another finger behind the vein and
pinch it closed."
At IEDM 1999, the group presented a groundbreaking paper demonstrating
that the FinFET design successfully blocked current in the "off
state," even on such a small scale. Last year's encore showed
that the FinFET conducts enough current in the "on"
state to deliver on its promised high performance.
"In industry, they have large fabrication facilities and
lots of engineers to work on optimizing devices," Bokor says.
"It's obviously harder for us. So it took a while, but we
did it."
Hu, Bokor, and King's second transistor design, the Ultra-Thin
Body device, em-ploys a very different engineering innovation
to shorten gate lengths while preventing leakage. In today's transistors,
most of the leakage occurs deep in the body of the transistor
below the gate. The UTB approach is to eliminate that material
except for the top-most portion of the channel that is well-controlled
by the gate.
"To continue the bleeding analogy, the UTB approach is like
closing off a vein by pushing it against a hard surface like a
bone," Hu says. "It's an improvement, but pinching it
like the FinFET does is still the best way."
Once the FinFET and UTB structures were designed, King and Bokor,
along with electrical engineering colleague Vivek Subramanian,
faced another set of challenges inside the clean room where the
transistors are manufactured.
"No one else had really tried to thin down silicon this much
in a controllable manner and make transistors," says King,
the faculty director of Berkeley's Microfabrication Laboratory.
"We had to find ways to even define the features so we could
make these transistors."
For example, in the case of the UTB approach it's desirable to
make the silicon body below the gate as thin as possible to help
control leakage. But this isn't the case on either side of the
gate. There, the material should be as conductive as possible
to prevent a bottleneck in the flow of electricity. Their novel
solution, King explains, was to selectively thicken the areas
around the gate by depositing materials only in those regions.
While King oversees the integration of the entire manufacturing
process, Bokor focuses on one critical element: lithography. Layers
of material in today's chips are patterned using a process where
light shining through a mask (essentially a stencil of a chip's
features) projects the circuit pattern onto a silicon wafer coated
with photoresist, an organic film that hardens when exposed to
light. The shorter the wavelength of light projecting through
the mask, the smaller the features on the chip. In order to make
transistors with features as small as the FinFET and UTB de-vices,
Bokor employed an electron beam, exposing the mask pattern on
the silicon wafer. Fortunately, the best electron beam facility
in the world for this type of work is adjacent to the Berkeley
campus at Lawrence Berkeley National Laboratory (LBNL). Since
1996, Bokor has collaborated with Erik Anderson, director of LBNL's
Center for X-ray Optics, and director of LBNL's "Nanowriter"
facility.
 |
| Probe stations, such
as this one in the Device Characterization Lab, measure devices
as small as a single transistor. Peg Skorpinksi photo |
"This is a laboratory technique though, and not suited for
mass production," Bokor says. "A whole other issue in
this project is how to mass produce transistors on this size-scale."
One part of the answer may lie in Bokor's pioneering research
into extreme ultraviolet lithography (EUV), the industry's leading
candidate for next-generation lithography. It is expected that
EUV will enable the fabrication of chips with 20-nanometer transistors
and smaller, breaking down some of the mass manufacturing barriers
threatening Moore's Law.
"The brick wall is now at 10 nanometers and we're reaching
out to touch it," Bokor says.
With the end of silicon in sight, the Berkeley research into FinFET
and UTB structures is cranking along with an emphasis on honing
the production process for mass manufacture. Indeed, King says,
it's a detail-oriented job.
"On this scale if you have one less atom in a channel, that
can affect the performance," she says. "And we want
to see how these transistor structures perform at the ultimate
size limits."
While computers can simulate the operation of transistors, King
believes that the only way to truly test novel devices is to make
them. Then the simulation models can be updated with real-world
data, and circuits using the FinFET and UTB transistors can be
accurately designed.
Once the advantages are on the table and the manufacturing kinks
ironed out, it's up to the private sector to take the ball and
run with it.
"The brick wall facing today's transistors is still a ways
off," Hu says. "The industry can continue for a while.
But at what point would a company become more competitive to convert
to our new structure? Our challenge is to make that switch more
compelling."
Written by David Pescovitz, a contributing writer to Wired
and creator of Lab
Notes, the College's online research digest. His work has
appeared in Scientific American, New Scientist, the New York Times,
and Salon.
|
