Mortal Programming
Is it possible to program a computer without knowing what’s
inside? The answer to that question is essential to James Tour, the
Chao Professor of Chemistry, professor of mechanical engineering
and materials science, and professor of computer science, who has
been working for more than a decade to develop molecular computers.
Molecular microchips are an ambitious departure from the norm in computing.
At present, chipmakers use precision lithography to etch transistors into silicon
wafers. For the past 20 years, the number of transistors crammed into a fingertip-sized
chip has doubled every 18 months, and today’s chips contain more than
10 million transistors apiece. To program a computer, computer scientists rely
on the fact that they know exactly where each transistor is and how much current
it will take to switch the transistor from on to off. All computer logic builds
from this simple on-or-off state, and today’s programming relies on the
notion that programmers are omnipotent: Not only do they know what’s
inside the chip and where, they also have the power to change the state of
any single transistor.
 |
Molecular computing is altogether different. It is based on research by chemists,
like Tour, who have shown that individual molecules can perform the same switching
functions as the transistors on silicon chips. Since the molecules are about
one million times smaller than their solid-state counterparts, the potential
processing power of a molecular computer dwarfs anything available today.
One of the chief benefits of the technology is that molecular chips would self-assemble,
meaning chemists would simply mix up some chemicals and pour them onto a flat
film and the chemical transistors would assemble by themselves. That’s
a much cheaper manufacturing process than etching. A state-of-the-art chip
plant today costs $3 billion and is expected to cost $6 billion by the middle
of this
decade.
But critics of molecular computing also have pointed to self-assembly as the
Achilles’ heel of the technology—if computer scientists don’t
know the details of the molecular microchip, how can they program it? Regardless
of how cheap or powerful the computer is, it’s useless if it can’t
be programmed.
Tour has never listened to those critics. Based on his experiences both as
a teacher and a father, he knows it’s possible to teach pupils without understanding
the precise workings of their minds. Tour believes the key to “mortal programming”—programming
a computer without knowing what is inside or having the ability to change the
state of individual transistors—is to find the right mathematical techniques
to simulate the type of learning people experience every day.
“The first time I brought this up in a speech to the Rice faculty three
years ago, several people told me it could never be done,” Tour says. “At
that point, I knew I needed to talk to students, because students don’t
have any preconceived ideas about what is possible or impossible.”
So Tour invited Summer and Chris Husband, newly married incoming graduate students
in applied mathematics, to join his project. “The joke around the office
is that we didn’t know that we were signing up for the impossible,” Summer
Husband said. “Chris and I just thought it was a really intriguing idea,
and as graduate students, we wanted to work on a project that was challenging
and unique.” Tour’s research team also included computer science
graduate student William L. Van Zandt and undergraduate Lauren S. Wilson and
collaborated with Paul D. Franzon and David P. Nackashi of North Carolina State
University.
To find out whether mortal programming was feasible, the Husbands wrote a series
of computer simulations that mimic what goes on inside a nanocell—a one-micron
square of film containing several hundred gold nanoparticles interlaced with
thousands of molecular switches. Using a variety of mathematical techniques,
the team showed that it was possible to train nanocells to serve as simple adders—a
logic circuit capable of adding two whole numbers that are less than or equal
to three—and as logic gates—the elementary building blocks of digital
circuitry that perform specific tasks. The results of their research appear in
a paper titled “Nanocell Logic Gates for Molecular Computing,” published
recently in the IEEE Transactions on Nanotechnology.
The findings are significant because, in addition to being able to self-assemble,
nanocells are up to 50 times smaller than the solid-state-based logic gates.
Solid-state systems also consume more power and generate more heat than nanocells.
The published study simply sought to prove that it is possible to train nanocells
as logic gates. Now the group is asking the really hard question: Can they actually
conduct mortal programming?
Tour expects to publish that work this year.
—Jade Boyd
|
