Integrated circuits consist of millions of transistors "printed" onto a silicon wafer. The transistors exist in either of two states. They are either fully switched on or fully switched off. When switched on they represent the binary digit (or bit) one, and when switched off they represent the binary digit zero. The processing power of the chip depends on the speed at which the transistors can flip between the two states, and also the total number of transistors printed on it.
In 1965, Gordon Moore, research director of Fairchild Semiconductor, wrote an article for Electronics magazine attempting to predict the future of the microchip industry. At the time, the chip with the highest density of components in the world was a Fairchild chip with 64 transistors on board.
Moore guessed that the number of transistors per chip would double every year. He later admitted he was only trying to be upbeat about the chip industry having a big future. But it was astonishing how accurate his forecast turned out to be, and it became known as Moore's Law.
Three years later, Moore and some colleagues from Fairchild broke away and formed Intel. From that point onwards, Intel, IBM and Motorola led the way in doubling chip density every year for the next 20 years. By 1971, there were 2,300 transistors per chip. By 1990, there were 1,200,000. Industry terms such as VLSI and ULSI - very large and ultra large-scale integration - quickly became redundant. Intel's Pentium IV released to market in November 2000 has 42 million transistors.
Moore's law is not a law of nature of course, more a rule of thumb for hardware market analysts. In the last few years, the law has been revised with Moore's blessing to an 18-month doubling period. This is a sign that the increase in component density is slowing down somewhat, and although the past is littered with discredited seers predicting the demise of Moore's Law, the laws of physics will eventually have their way.
The latest chips have pitches as small as 180 nanometres. A nanometre is one millionth of a millimetre. Using Moore's law, we can predict that the pitches need to be of the order of 100 nanometres by 2005. Shrinking the transistors doesn't reduce the amount of charge they must hold, so you must increase the concentration of the dopant to achieve the same effect. Above a certain concentration, the dopants start to clump, creating regions that are electrically inactive.
Second, the gates that control the flow of electrons in chips are now so small they are susceptible to quantum mechanical tunnelling. Basically this means electrons can burrow through closed gates, and a further reduction in gate size would increase the tunnelling effect such as to render the component useless.
Researchers are looking at alternatives to silicon and gallium arsenide. One possibility is using particle spin. Some subatomic particles have a spin property that is either up or down. This means that the binary one and zero can be assigned to the two states. Such bits have now been named qubits. But there are many problems. First, there is the quantum mechanical problem of superposition. According to this theory, certain quantum mechanical effects can exist in two states at the same time. This was first put forward by Schrodinger in his famous cat analogy, whereby if governed by quantum mechanical principles, a cat could be both dead and alive at the same time. Some scientists claim this phenomenon can be exploited to produce a more powerful and more diminutive processing power. IBM recently announced it had achieved a simple prototype, but only a few molecules were involved.
Another line of research that is exploiting superposition involves a super-conducting loop that is simultaneously passing current in both directions. Firstly a super-conducting loop is set up, and then a microwave type of laser called a maser is used to induce the superposition state.
One thing is certain, though: by the time 2005 comes, and if by that time physical laws prevent further miniaturisation of existing wafer technology, there will be no working model from the current quantum mechanical lines of research capable of stepping into the breach. But it is much more likely a solution will emerge by exploiting the current technology in a different way.
Fintan Gibney is an IT consultant with SmartForce. He can be contacted at gibney@ireland.com
So what's what?
Pitch: the distance between dots on screen, printed material, etc. Used with relation to monitors, printers and processors
Dopant/to dope: add elements into a semi-conductor material during the manufacturing process to increase its conductivity.
Gallium arsenide: semi-conductor material used as an alternative to silicon