Field-Effect Transistor
A field-effect transistor, or FET, is a transistor in which a voltage on an insulated or junction-isolated gate electrode creates an electric field that controls the current flowing through a channel between two other terminals, the source and the drain. Because the control is by field, the gate draws almost no steady current. This single device type, in its metal-oxide-semiconductor form, is the building block of almost every modern chip.
The terminals and the operating logic follow one simple chain. Carriers enter the channel at the source and leave at the drain; the gate sits over the channel, separated by a thin insulator in the MOSFET. In an n-channel MOSFET on p-type silicon, a gate voltage above a threshold attracts electrons to the surface and forms a thin conducting layer, an inversion channel, so current flows between source and drain. Below the threshold the channel is absent and the device is largely off. Used as a switch, it gives the zero and one of logic; used within its active range, the drain current follows the gate voltage and the device amplifies. A p-channel device works with holes and opposite polarities.
Two families developed, with separate histories. Shockley published the theory of a unipolar field-effect transistor in 1952, and the first working junction FET, the JFET, was reported in 1953 by George Dacey and Ian Ross at Bell Labs. In a JFET, the gate is a reverse-biased pn junction whose depletion region pinches the channel; it keeps the field away from the troublesome surface. The MOSFET came from Mohamed Atalla and Dawon Kahng, who built the first working device at Bell Labs in 1959 and presented it in 1960, relying on Atalla's earlier work on thermally grown silicon dioxide, which made the oxide-silicon interface clean enough for the gate field to reach the channel. That made real the field-effect concept that decades of attempts, from Lilienfeld's patents on, had failed to realise.
The FET dominates because its properties suit large-scale integration. The control is by voltage with near-zero gate current, so a gate can drive many others with little static power. The device is self-isolating and simple in layout, needing no large base region. Pairing an n-channel and a p-channel MOSFET in complementary form, CMOS, which Frank Wanlass and Chih-Tang Sah described at Fairchild in 1963, yields logic that draws appreciable current mainly while switching. Scaling the gate length down improved speed, density and energy per switch for decades in line with Moore's observations; today a processor or AI accelerator holds billions to tens of billions of such devices, and the memory cells of DRAM and flash are themselves FET-based.
Short channels forced geometry changes, and the newest generations are variations on the same theme. As gate lengths fell, the gate lost control of a channel the drain could influence, so current leaked even when the device was off. To restore control, the FinFET, developed in research from the 1990s and used in volume production from around 2011, wraps the gate over a raised fin of silicon on three sides. The gate-all-around or nanosheet transistor, entering production in the 2020s, surrounds stacked channels completely. A related misconception is that the FET is a different species from the bipolar transistor in its reliance on field: the bipolar device also depends on internal fields, but its control is by injected current, while a FET controls by gate voltage.
A further pitfall is believing the gate draws no current at all. The steady current is negligible, but switching a gate means charging and discharging its capacitance every cycle, which is the main source of dynamic power in chips, and very thin oxides allow tunnelling leakage, a problem that drove the adoption of high-permittivity gate insulators.
Read the field-effect transistor as a voltage-controlled valve for charge: a gate field across a thin insulator opens or pinches a channel, the principle that waited from 1925 until 1960 and now runs nearly every chip.