The problem of too many parts
By the mid-1950s, the transistor had replaced the vacuum tube in many designs. Transistors were smaller, cooler and more reliable than tubes, but a complex machine still needed thousands of them, together with resistors, capacitors and the wires between them. Each connection was made by hand, usually by soldering, and each was a possible point of failure. Engineers sometimes called this the “tyranny of numbers”: as circuits got more ambitious, the number of parts and joints made them expensive and unreliable.
Several people independently thought the answer was to make all the parts together, in one piece of material, with the connections built in. A British radar engineer, Geoffrey Dummer, described such an idea publicly in 1952 and later tried to build one, without success. The idea was in the air, but nobody had yet shown how to make it work.
Kilby at Texas Instruments
Jack Kilby joined Texas Instruments in Dallas in the summer of 1958. In the commonly told account, he was a new employee without vacation time, and he used the quiet plant to think about the problem. He reasoned that all the needed parts, including resistors and capacitors, could be made from the same semiconductor material as the transistor.
On September 12, 1958, he demonstrated a working circuit to colleagues. It was built on a thin piece of germanium and included transistor and other components connected by fine wires. It was crude, with the parts connected by wires attached by hand, but it showed that a complete circuit could be formed from a single block of semiconductor. A patent application followed in early 1959.
Noyce at Fairchild
Robert Noyce was a co-founder of Fairchild Semiconductor in California. In 1959, he developed a different approach, building on the planar process worked out at Fairchild by Jean Hoerni. The planar process created transistors on a flat silicon surface and covered them with a protective layer of silicon oxide. Noyce realized that the connections between components could be made by depositing thin metal lines over that layer, which meant that there was no need for separate wires. This made the circuit more suited to mass production.
His notebook entries from early 1959 and a patent application followed. Fairchild made its first working integrated circuits in 1960 and 1961.
Two inventors, one invention
Kilby and Noyce each solved part of the problem. Kilby is credited with showing that a whole circuit could be made in one piece of semiconductor. Noyce is credited with the practical method of connecting the parts, which is the basis of the chips used since. A long patent dispute followed, and in the end the two companies agreed to license each other’s patents. Historians generally treat the two as co-inventors, and this is how the invention is often described.
Kilby received the Nobel Prize in Physics in 2000 for his part in the invention. Noyce had died in 1990, and the prize is not awarded after death, but the Nobel committee’s account and most histories mention both men.
Early customers
The first chips were expensive and had few components, so they did not immediately spread. The early customers were those who valued small size and reliability and could pay for them, notably the U.S. military and the space program. Integrated circuits were used in guided missile systems, and the Apollo guidance computer made use of them in the 1960s. That demand helped companies improve manufacturing and reduce prices.
By the mid-1960s, chips were also used in commercial computers and, soon after, in consumer items such as calculators and hearing aids. The price of a simple circuit fell steeply over the following years, as production volume rose and techniques improved.
Shrinking and Moore’s observation
In 1965, Gordon Moore, a colleague of Noyce who later helped found Intel, noticed that the number of components that could be placed on a chip had been roughly doubling every year and expected the trend to continue. In 1975 he adjusted the estimate to a doubling about every two years. This became known as Moore’s law, and although it is an observation and not a law of nature, it served as a target for the industry for decades.
The consequences were large. Making transistors smaller allowed more of them on a chip, which raised speed and lowered the cost per function. The first microprocessor, Intel’s 4004 of 1971, held about two thousand three hundred transistors, a figure widely quoted. Later chips have billions. This steady miniaturization, discussed further in the story of chip miniaturization, is a main reason why computers moved from rooms to desks to pockets.
What a chip is, briefly
A modern chip is made by building layers of patterns on a thin disc of purified silicon, called a wafer. Light is used to transfer patterns, chemicals etch them, and other steps add materials that change the way the silicon conducts. Many chips are made at once on one wafer and then cut apart. The process is demanding, requiring extremely clean factories, and relatively few companies can make the most advanced chips.
Silicon chips are now in vehicles, appliances, medical equipment, phones and many other things, mostly unnoticed. The idea Kilby and Noyce reached in the late 1950s, to make the whole circuit at once, remains the foundation.