How a Transistor Switches

Press a key on a keyboard and somewhere inside the machine, a few billion tiny switches flip. None of them move. Nothing slides, clicks, or rotates. The switching is done by transistors, and a transistor is essentially a piece of silicon arranged so that one voltage can decide whether another current flows.

The dominant kind in modern electronics is the MOSFET — the metal-oxide-semiconductor field-effect transistor. It has three main terminals: the source, the drain, and the gate. Current, when it flows, moves between source and drain. The gate decides whether it can. Crucially, the gate does not touch the path the current takes. It sits just above it, separated by a sliver of insulating oxide a few atoms thick. The gate controls the current without ever conducting it.

To see how, picture an n-channel MOSFET sitting in lightly p-doped silicon. P-doped silicon has an excess of holes — places where an electron is missing — and very few free electrons. Between the source and drain regions, which are heavily n-doped and full of free electrons, lies a stretch of this p-type material. Left alone, that stretch is a barrier. Electrons in the source cannot easily cross into the drain because the region in between has almost none of its own to support a current.

Now apply a positive voltage to the gate. The gate, the oxide, and the silicon underneath form something very much like a parallel-plate capacitor. Positive charge accumulates on the gate; an equal negative charge is induced in the silicon directly below. At first, that induced charge just pushes holes away, leaving behind a depletion region. But raise the gate voltage past a particular value — the threshold voltage — and something more dramatic happens: free electrons are pulled out of the source and drain regions and gather in a thin layer right beneath the oxide. That layer is the channel. It is a strip of n-type conductor temporarily conjured into existence inside p-type silicon, and once it forms, source and drain are electrically connected.

Two conditions determine the switch's state. The first is whether the gate-to-source voltage exceeds the threshold; below it, no channel forms and the transistor is off. The second is whether there is a voltage difference between drain and source to push electrons across once the channel exists; with a channel but no drain voltage, the switch is on but no current flows, the way an open valve passes no water unless there is pressure behind it.

This arrangement has consequences that shaped the entire industry. Because the gate is insulated, holding a transistor on costs almost no current — only the tiny leakage through the oxide. Power is consumed mainly when the transistor switches, charging and discharging the gate's small capacitance. That is why CMOS logic, which pairs n-channel and p-channel transistors so that one is always off in any stable state, became the foundation of digital electronics: a circuit at rest draws nearly nothing.

The simplification that a transistor is just a switch is useful but lossy. Real MOSFETs do not snap cleanly between on and off; they pass through a region where current rises exponentially with gate voltage, and another where it saturates. Designers of analog circuits live in those in-between regions. Designers of digital circuits work hard to avoid them, biasing transistors so they spend as little time as possible mid-transition, because a half-on transistor wastes energy and produces heat.

What looks from the outside like a binary device is, underneath, a carefully managed piece of solid-state physics: a capacitor that summons a conductor into being, held in a regime where the summoning is fast and the holding is cheap. The keystroke registers because, for a few nanoseconds, a few atoms of charge on a gate electrode rearranged the electrons in a square of silicon smaller than a virus.