This article first appeared in Personal Computer World magazine, September 1999.
TIME IS RUNNING OUT for silicon chips as we know them. Although rapid advances in manufacturing technology are bringing us smaller, faster, cheaper and more densely-packed chips, there's a fundamental physical limit looming. And when we hit it, reckoned to be around 2020 or so, we'll need to turn to a quite different technology if computers are to continue to grow in speed and shrink in size. Recent research results are pointing to a novel computer technology that needs almost no power, operates at speeds we can only dream about today, and is vanishingly small.
The new approach is dauntingly named "quantum cellular automata", QCA for short, and it's implemented on a semiconductor chip using gallium arsenide technology. QCA's basic component is a "cell", which contains four "quantum dots", arranged in a square pattern. A quantum dot is a kind of artificial atom, which provides a comfortable spot for a passing electron to rest in. During manufacture, two free electrons are inserted into the cell. The electrons can move around within the cell, but can't escape from it. Because they repel each other, the electrons try to get as far apart as possible, and end up sitting on diagonally-opposed quantum dots. This leads to two possible configurations for the cell - whcih can represent a "0" or a "1". Effectively, each cell is a memory element holding one bit. Incredibly, you can actually see an image of electrons in a quantum dot.
Something interesting happens when you place these cells very close together, in a line from left to right. Suppose we give the first cell on the left a tiny electrical nudge to set it to encode a "1", with its electrons sitting on the top-right and bottom-left dots. Although the electrons can't leave the first cell, their charges are felt by the electrons in the next cell along, which get repelled to its top-right and bottom-left dots. So, the second cell is now also a binary 1. As more cells are lined up, the same thing happens, like a collapsing line of dominos. What you effectively get is a "wire", where the state of the last cell in the wire is a copy of the state of the first cell. In this case a "1" has been transmitted down the wire, but in contrast to a conventional wire no current has actually flowed. You can experiment with a nice java animation of this effect on Craig Lent's QCA pages .
Taking the principle further, groups of cells in arrays and other configurations can be used to perform the basic logic functions of AND, OR and NOT. And crucially, logic elements can be interconnected using the current-less QCA wires. Recent work at Indiana's Notre Dame University has demonstrated that this really does work.
What makes QCA so exciting is that it operates on an incredibly tiny scale, allowing vast numbers of logic elements to be crammed into chips. And because it requires only miniscule power to prod a cell into an initial configuration, after which the domino effect performs the computation with almost no power dissipation, QCA chips will run cool.
For now, QCA exists only in research laboratories. There are many unsolved practical problems, not least of which is the slightly annoying constraint that current systems only work at temperatures just above absolute zero. A QCA-based notebook PC might run forever on a single tiny battery, although lugging around enough liquid helium to keep it cool enough might be awkward. But researchers are optimistic that room-temperature QCA will eventually be possible, and then we'll have some seriously cool hardware. And by 2020, you never know, we might even have some seriously cool software. But I kinda doubt it.
Toby Howard teaches at the University of Manchester.