Silicon-Controlled Rectifier
A silicon-controlled rectifier, or SCR, is a four-layer pnpn semiconductor device with three terminals, anode, cathode and gate, that blocks current in both directions until a brief pulse at the gate switches it on, after which it conducts in the forward direction and stays on by itself. It will not turn off again until the current through it falls below a small holding value. That latching behaviour makes it a rugged, high-power electronic switch, and the SCR is the best-known member of the thyristor family.
The first working version came from Bell Labs research and the first commercial one from General Electric. In 1956 a team at Bell Labs, J. L. Moll, M. Tanenbaum, J. M. Goldey and N. Holonyak, published work on pnpn transistor switches, showing the regenerative switching behaviour of a four-layer structure, following earlier four-layer diode proposals including Shockley's. General Electric developed the structure into a practical power device and introduced the silicon-controlled rectifier commercially in 1957 to 1958, aiming at high-current control that had until then used thyratron tubes and magnetic amplifiers. Nick Holonyak moved to General Electric, where he also invented the first visible-light-emitting diode in 1962, but his role here concerns the four-layer switch.
The latching action is easiest to see by imagining the four layers as two bipolar transistors, a pnp and an npn, sharing their middle regions. The collector of each supplies the base current for the other. A small gate current turns the npn on, its collector current feeds the pnp's base, and the pnp's collector current feeds back into the npn, a positive feedback loop that drives both into saturation once the sum of their current gains reaches one. After that the device holds a low forward voltage of roughly 1 to 2 volts regardless of the gate. Excessive voltage across the device or a very fast rising voltage, expressed as dv/dt, can also trigger it unintentionally, and a too-rapid rise in current, di/dt, can concentrate heat in a small area, so datasheets specify limits for both.
The central practical fact is that the gate can start conduction but cannot stop it. To turn an SCR off, the current has to fall below the holding current, a few milliamperes in small devices, long enough for the junctions to recover. In AC circuits this happens automatically at each zero crossing, which is why SCRs are natural for phase control: a trigger circuit, often a UJT or diac, fires the gate at an adjustable angle in each half-cycle, and the average power delivered to the load varies with the delay. In DC circuits, special commutation circuits using capacitors or additional switches must force the current to zero, which complicated early inverter designs and led to gate-turn-off thyristors and then IGBTs for DC switching.
Applications grew from lamp dimmers to the electricity grid. Light dimmers, speed controls for universal motors in drills and appliances, battery chargers, welding controls, crowbar overvoltage protection and industrial DC motor drives were early uses. Large stacks of series-connected thyristors, each handling kilovolts and thousands of amperes, switch power in high-voltage direct current links, static var compensators and motor soft starters, and large devices are still built for these jobs. A closely related device, the triac, joins two opposite-polarity thyristor structures to conduct in both directions of an AC wave and is more common than a pair of SCRs in household dimmers.
Misconceptions usually centre on the word rectifier and the idea of control. The device does rectify, passing current only from anode to cathode when on, but it is a latching switch, not an amplifier: the gate current needed is a tiny fraction of the load current, yet the output does not follow the gate signal in proportion. Another error is to expect that removing the gate drive turns the load off, and a third is to expect a continuous-DC load to switch off by itself. In practice the SCR remains valued for sheer ruggedness, and its surge-current capability is hard for transistors of equal cost to match.
Read silicon-controlled rectifier as a diode with a trigger: four layers that behave as two cross-coupled transistors, switched on by a brief gate pulse and held on by their own feedback until the current drops away.