Features

Building blocks

There are broadly two approaches to replication: the top-down process of controlled assembly or the bottom-up route in which the devices are simply encouraged to grow themselves. For example, prions, the agents that cause mad-cow disease and its human form Creutzfeldt-Jakob disease, are misshapen proteins that seem to replicate by acting as templates for other proteins to follow. Being able to manufacture matrices of computer logic gates or memory cells like that has its appeal. And its dark side.

Current computing circuits are too intricate, too top-down, for this kind of self-assembling growth to be possible, but the thrust in much of the most forward-looking computing research is towards using larger numbers of simpler, less specialised units. This significantly increases the possibility of finding a design that can assemble spontaneously or by following a template.

The smallest, and probably simplest, functional computer circuits so far were created last year at IBM's Almaden Research Center in San José, California. Some 260,000 times smaller than the circuit elements in a modern silicon chip, they make use of molecular cascades, in which individual molecules of carbon monoxide move in sequence across a copper surface, just like rows of toppling dominoes.

Each domino is a trio of carbon monoxide molecules arranged in a chevron pattern. The two outer ones are fixed so that they slightly squeeze the inner molecule, making it easy for it to pop out if lightly pushed. Lines of chevrons are nested together, so pushing a single molecule at the input initiates a cascade that transmits data through the circuit.

On

ADVERTISEMENT

their own, these cascades act like wiring, but by arranging for cascade streams to interact in precise ways they can be combined to create conventional logic gates with multiple inputs and outputs. The IBM team was able to assemble a working two-input sorter and later a three-input sorter. According to Andreas Heinrich, one of the lead researchers on the project, this was 'the first time all of the components necessary for nanoscale computation have been constructed, connected and then made to compute. It's way smaller than any operating circuits made to date'.

Small they may be, but they're not exactly speedy just yet. Not only does it take hours to position each molecule individually, but - just as with dominoes - the trick only works once, then you have to put them all back one by one. Even so, with the principle now established, attention has turned to finding cascades based on different molecular properties, such as atomic spin, that could be reset and used repeatedly.

Could such an arrangement be made to self-assemble? Probably not on the level of a functional computing circuit, but individual units - or dominoes - of some kind may well be feasible. The problem then is how to put them together.

The top-down approach based on assembly of components applies just as effectively to arranging individual atoms and molecules as to building a car. Even the machinery needn't be all that different. The branch of nanotechnology known as MEMS (Micro-Electro-Mechanical Systems) has been forging ahead to produce very precise and effective gears, levers, actuators and even a working steam engine in which an invisibly small drop of water expands to move a piston. These devices are generally made from polycrystalline silicon, using much the same techniques as are employed for integrated circuits.

Although many examples of MEMS follow the industrial revolution paradigm of rotating and reciprocating machines, an alternative line of development aims to produce tiny multipurpose fingers operated by actuators. A field of these might move objects around just as cilia, tiny hairs, do in many living organisms. Or, arranged more specifically, they might act as little assembler robots for other nanomachines.