MOSFET 1959
The 1959 MOSFET was the first working metal-oxide-semiconductor field-effect transistor, made at Bell Labs by Mohamed Atalla and Dawon Kahng in November 1959, demonstrated at the Solid State Device Research Conference in 1960 and published that year. It controls current through a silicon channel with the electric field from a metal gate that is separated from the silicon by a very thin layer of silicon dioxide. Because the gate is insulated, it draws almost no steady current, and because the structure is flat and can be made by the same oxidation and diffusion steps used for other silicon devices, it can be shrunk and replicated on a vast scale. It is today the most widely manufactured electronic device, the basic element of nearly every microprocessor and memory chip.
The idea of controlling current with a field was old, but nobody had made it work until the oxide problem was solved. The field-effect principle had been patented in the 1920s and 1930s, and when Bell Labs' researchers tried to build such devices in the late 1940s, including William Shockley's group, the effect was cancelled by electrons trapped in states at the semiconductor surface, so the applied field never reached the channel. The turning point was Atalla's work on thermally oxidized silicon, in which a silicon dioxide layer grown in a furnace reduced the surface states dramatically. With a stable oxide on the surface, a gate electrode placed on top could finally modulate the silicon underneath. Atalla and Kahng built the device in late 1959, and the work was presented in June 1960.
The device works by turning the silicon surface under the gate from one type to the other. In a typical n-channel enhancement device, two n-type regions, the source and drain, are formed in p-type silicon, and with no gate voltage no current flows between them because one of the two junctions is always reverse biased. When a positive voltage is applied to the gate, the oxide holds the field, and electrons are drawn to the silicon surface, forming a thin conducting inversion layer, the channel, which connects source to drain. The current is then regulated by the gate voltage, with almost no current in the gate itself. The first devices were slow and had a long channel on the order of tens of micrometers, with an oxide perhaps a hundred nanometers thick, which are large by modern standards.
Commercially, the MOSFET was ignored for several years. The mature bipolar transistor was faster and better understood, Bell Labs management did not rush it into production, and early MOS devices suffered from unstable thresholds caused by mobile ion contamination in the oxide, sodium in particular. Engineers at RCA and Fairchild took up the technology in the early to mid 1960s, and the first commercial MOS integrated circuits appeared in the middle of the decade. The MOS structure won for integrated circuits because it takes little area, needs fewer process steps than bipolar, and consumes little power, which is what matters when thousands of devices share a chip.
CMOS completed the story. In 1963 Frank Wanlass at Fairchild, working with Chih-Tang Sah, described complementary MOS logic, which pairs an n-channel and a p-channel device so that one is always off and the circuit draws appreciable current only when switching. This allowed enormous numbers of gates on a chip without overheating, and CMOS became the dominant logic technology from the 1980s onward. Misconceptions persist: the 1959 device was not the first field-effect transistor concept, since earlier ideas and the junction FET existed, and it was not the beginning of integrated circuits, which had begun in 1958 to 1959 with bipolar devices.
Today the MOSFET, in planar form and in newer shapes such as the FinFET and the gate-all-around transistor, is the unit of computing. Billions of them fit on a single processor, and the total number ever manufactured is estimated to be larger than that of any other device. The principle of gate, oxide and channel is unchanged from 1959.
Read the 1959 MOSFET as the moment the oxide-covered silicon surface became a switch: an insulated gate controlling a channel is a simple idea that took a good oxide to realize, and it remains the structure from which nearly every digital chip is made.