• The Listener
  • North & South
  • Noted
  • RNZ

Size matters

Smart circuits and why your trainers are cooler than ever.

There are some questions in computing that never go away, and one of them is the future of Moore's Law.

The "law" - actually, it's really an observation that pretty much holds good - was proposed by Gordon Moore, a co-founder of Intel, in a 1965 issue of Electronics magazine. It decrees that the number of transistors (and hence processing power, and the cost of performance) on microprocessors will double every 12 to 18 months.

As I noted, the observation has since held good, or better. But processor technology, and microelectronics in general, seems to hit a roadblock every now and then, prompting speculation that the law has reached the end of its useful life. Then somebody unveils a breakthrough and it's all on again.

Roadblocks usually involve process: currently there are limits to how many million features can be mapped onto a wafer of silicon. The lithographic tools that do that job are big, expensive and inevitably limited in what they can do. Never mind actually building more complex chips, it's painfully difficult and expensive just to design them.

So what say you sidestepped the roadblock - the lithographic design process - and found a way for circuits to design themselves? That's effectively what IBM announced recently, in the form of a research paper showing how specially developed polymer molecules could be induced to assemble themselves into tiny, precise and predictable patterns.

IBM isn't the only organisation to be working with so-called nanocrystals to build new kinds of chips - several other companies have been able to deposit the tiny structures along lines drawn with conventional lithography - but its twist is that the hexagonal patterns formed by the polymers aren't themselves circuits that have to be connected to larger wires, but molecule-sized stencils through which light can be shone to create circuitry on silicon.

This is quite close to the way chips are made now: by shining a pattern on a piece of silicon, then using a chemical process to carve out the spaces in between - the raised parts left become the "wires" that carry electrical current across the chip. The advantage in the method described by IBM is that the hard part - the design stencil - is created through a natural process in the polymer. This works quite well with existing manufacturing processes, and is likely to be relatively cheap and practical.

IBM's proof of concept wasn't a processor, but a flash memory chip. Such chips store data in tiny cells that are either on or off - or, if you prefer, represent a one or a zero. IBM's demonstration chip holds cells 20 nanometres in diameter and 40 nanometres apart (a nanometre is one billionth of a metre). That's one hundredth the size of the cells that currently hold a bit in a memory chip.

Result: vastly increased storage on smaller chips. Alternatively, you could get around current problems with reliability (the electrical charges that signify the data can leak out of chips when the power in a device is off) by writing 100 instances of every piece of data.

These amazing new polymers may already be part of your life - on your feet. IBM is working with diblock copolymers that consist of two types of polymers bonded together which would normally repel each other. The counterbalancing forces allow IBM's researchers to control their exact position. Diblock copolymers are currently used to glue the soles on trainers.

Pilot production with this new process isn't likely to start for three to five years. Assuming that the first real products to emerge are flash memory chips - with more complex microprocessors to follow - we could expect radical improvements in the kinds of devices that use such chips: handheld computers, mobile phones and digital cameras.

I already have an Apple iPod based around a tiny conventional hard drive that can store more data than my big old computer. It's about the size of a pack of cards, only thinner, and it holds nearly 10 days worth of music. It's reasonable to assume that in 10 years' time, I'll be looking at something that will be more the size of my wristwatch.

The thing is, circuit sizes aren't going to stop at the molecular level. Research already in train is seeking to develop transistors at sub-atomic sizes. At the same time as we acquire this vastly greater capacity to store and process information, we will also become more connected than ever, bringing into play another principle: Metcalfe's Law, which holds that the value of a network increases by the square of the number of its nodes. (So the Internet was exponentially more significant when it reached, say, 10 million users than it had been with just one million.)

Find it hard to grasp? Don't fret. Just enjoy the idea that your new trainers are even cooler than you thought.