The transistor had solved the problem of vacuum tubes, but it had created another. A complex circuit required thousands of transistors, resistors and capacitors, all to be connected by hand with soldered wires. The bigger the circuit, the higher the cost, the size and the likelihood of a faulty connection. Engineers called this problem the "tyranny of numbers". The solution was to build the whole circuit in a single piece of material.
Jack Kilby and germanium
In the summer of 1958 Jack Kilby, newly hired by Texas Instruments, was left almost alone in the laboratory during the company holidays, to which he was not yet entitled. He had time to think through an idea: if transistors, resistors and capacitors could all be made from the same semiconductor material, why not fabricate them together on a single chip?
On 12 September 1958 Kilby demonstrated an oscillator built on a piece of germanium. The components were formed in the crystal itself, but the connections between them were still very fine gold wires soldered on top of the chip. It was crude, but it worked: it was the first integrated circuit.
Robert Noyce and planar silicon
A few months later, in 1959, Robert Noyce, one of the founders of Fairchild Semiconductor, conceived a different and more practical version. It was based on a technique just developed at Fairchild by Jean Hoerni, the planar process.
What the planar process is
In the planar process the surface of the silicon is covered with a protective layer of oxide. Using photographic techniques, windows are opened in the oxide at precise points, and through them the impurities that form the different regions of the transistors are introduced. The surface stays flat and protected.
Noyce's insight was that, on this flat surface, the connections could be made by depositing a thin layer of metal, also defined with photographic techniques. No more hand-soldered wires: components and connections were all produced together, with the same procedure.
Two inventors
| Kilby (TI) | Noyce (Fairchild) | |
|---|---|---|
| Year | 1958 | 1959 |
| Material | Germanium | Silicon |
| Connections | Soldered gold wires | Metal deposited on the surface |
| Manufacturability | Low | Suited to mass production |
Years of patent disputes followed, eventually resolved with cross-licensing agreements. Today both are recognised as inventors of the integrated circuit. Kilby received the Nobel Prize in Physics in 2000; Noyce had died in 1990, and the Nobel is not awarded posthumously.
Early uses
At first integrated circuits were expensive, and the main customers were military and space programmes, where weight and reliability mattered more than price. The Apollo Guidance Computer, the on-board computer of the lunar missions, was built with integrated circuits. This demand helped production grow and costs fall, paving the way for civilian uses.
Gordon Moore and the 1965 article
In 1965 Gordon Moore, then director of research at Fairchild, was asked by Electronics magazine for an article on the future of electronic components. The piece, published on 19 April 1965, was entitled Cramming more components onto integrated circuits.
What he actually said
Moore observed that the number of components per integrated circuit, taken at the point of minimum cost per component, had doubled roughly every year. He predicted that the trend would continue for at least a decade. In 1975 he revised the estimate, indicating a doubling roughly every two years.
A few points about this "law" are worth clarifying:
- It is not a law of physics. It is an empirical observation about an industrial trend, which the industry itself then used as a planning target.
- It concerns density and cost, not directly processor speed, even though for decades the two went hand in hand.
- It has an economic basis. Moore was talking about the number of components it made sense to integrate, not the technical maximum.
Intel
In 1968 Noyce and Moore left Fairchild and founded Intel, aiming to produce semiconductor memories. A few years later Intel would produce the first commercial microprocessor. From that moment the history of the integrated circuit and that of the personal computer are the same story.
Why it still concerns us
Anyone who works with computer systems lives with the consequences of this trend every day. The cost of memory and computing power has fallen by many orders of magnitude, allowing design choices that would once have been unthinkable, such as keeping everything instead of deleting, or computing on the fly instead of storing. In recent years, however, density growth has become more expensive and slower, and performance gains increasingly come from specialised architectures and parallel processing rather than from miniaturisation alone.
In short
- The integrated circuit solved the "tyranny of numbers" by building components and connections on a single chip.
- Jack Kilby demonstrated the first integrated circuit, in germanium, in September 1958; Robert Noyce conceived the planar silicon version in 1959.
- In 1965 Gordon Moore observed an annual doubling of components per chip, revised in 1975 to about two years.
- Moore's law is an economic and industrial observation, not a law of physics.
