Documenting the Coming Singularity

Sunday, April 01, 2007

OK Nanotubes, Get in Line!

Have you ever heard of Moore's Law? Gordon E. Moore, a co-founder of Intel, made the observation in 1965 that the number of transistors on an integrated circuit for minimum component cost doubles every 24 months. According to Wikipedia: "However, it is also common to cite Moore's Law to refer to the rapidly continuing advance in computing power per unit cost, because increase in transistor count is also a rough measure of computer processing power."

It is because of Moore's Law that we've been getting faster and more powerful computers, at the same price, at the rate we have enjoyed thus far. But people have been predicting, correctly, that Moore's Law will soon reach the limits of a set of more powerful laws: the laws of physics. Eventually, trying to pack more transistors onto a silicon chip will result in an untenable leakage of current. Intel speculates that that limit will be reached in 2015.

Don't worry though, because the history of the accelerating rate of technological progress shows that whenever a paradigm approaches its ultimate limitations, pressure increases (meaning more resources are invested) for the developments of new paradigms that will continue the exponentially growing pace of progress. And this is exactly what is happening with Moore's Law.

Several new paradigms are currently (no pun intended) in development. They are: 1. Optical Processing, which uses photons rather than electrons to transmit information; 2. Quantum Computing, which makes further use of quantum effects to radically increase processing speeds; 3-D Layouts, which uses the third dimension (building chips in the upward direction, rather than merely length-and width-wise; and 4. Carbon Nanotube Transistors.

Up until now, a major roadblock in the development of nanotubes for use as transistors has been getting them aligned. According to this article at MIT's Technology Review:

Until now, making transistors with multiple carbon nanotubes meant depositing electrodes on mesh-like layers of unaligned carbon nanotubes, Rogers says. But since the randomly arranged carbon nanotubes cross one another, at each crossing, flowing charges face a resistance, which reduces the device current. The perfectly aligned array solves this problem because there are "absolutely no tube-tube overlap junctions," Rogers says.

The research team makes the arrays by patterning thin strips of an iron catalyst on quartz crystals and then growing nanometer-wide carbon nanotubes along those strips using conventional carbon vapor deposition. The quartz crystal aligns the nanotubes. Then the researchers can make transistors by depositing source, drain, and gate electrodes using conventional photolithography.

Researchers have not been able to grow well-aligned nanotube arrays until now, according to Robert Hauge, a chemistry professor who studies carbon nanotubes at Rice University. Indeed, "alignment is no longer a showstopper," says Ali Javey, an assistant professor of electrical engineering and computer sciences at the University of California, Berkeley.

Making a well-ordered array in which parallel nanotubes are connected between the source and drain electrodes is a big achievement, says Richard Martel, a chemistry professor at the University of Montreal. The new work allows a true comparison between nanotube transistors and silicon transistors because an array of nanotubes gives a planar structure similar to silicon devices, he says. "They did exactly what needed to be done, and it's a significant step."

So, Moore's Law will continue to apply as these new paradigms come online. It's a beautiful thing. What do you think?

3 comments :

Spaceman Spiff said...

On quantum computing:

I went to a talk on quantum computing once and the impression I got was that it wouldn't really increase processing speed. I think that quantum computing would merely reduce the time required for a very specific kind of problem from increasing exponentially to increasing linearly with the number of pieces of data.

Because of superposition, it is believed that quantum states could be used to more quickly solve "travelling salesmen" type problems, which conventionally require the computer to try all possible solutions individually with no clues as to which solutions might be best and all possible solutions take roughly the same amount of time to calculate.

Included in this set of problems are finding the next prime number, solving a minesweeper puzzle, password cracking, etc.

Barry Mahfood said...

Spaceman,

Quantum computing would be used for the types of problems you mentioned, since it can try all possible solutions in an instant and test for the correct one, and as part of a system working in conjuction with conventional computing.

CinnamongirlFla said...

I don't know what I think cuz I don't get what you are talking about.