MOS structure & operation

ME-1's diode was one junction and could do exactly one thing: conduct one way. The MOSFET — metal-oxide-semiconductor field-effect transistor — is the device that makes everything else possible, because a voltage at one terminal controls the current between the other two. That is amplification and switching in a single sentence, and it is why roughly every transistor manufactured today, by a wide margin, is a MOSFET.

Physically, an n-channel MOSFET starts as a p-type substrate with two heavily doped n-type regions embedded in it: the source and the drain. Between them, on the surface, sits an extremely thin insulating layer of silicon dioxide, and on top of that a conducting gate. That is the whole sandwich, and the name reads it off in order: metal (gate), oxide (insulator), semiconductor (substrate). Note what is missing — the gate touches nothing conductive. It is a capacitor plate, insulated from the silicon beneath it.

With no gate voltage, the source and drain are two n-type islands in a p-type sea: whichever way you apply a drain-source voltage, one of the two junctions is reverse biased, and no current flows. Now raise the gate above the source. The gate-oxide-substrate sandwich is a capacitor, so positive charge on the gate pushes holes away from the surface beneath it and attracts electrons toward it. Below a certain gate voltage this only clears out holes — a depletion layer, still no conduction. But past that voltage, enough electrons accumulate at the surface that the material there stops behaving like p-type at all and INVERTS into a thin n-type channel, bridging source to drain. Current can now flow. The gate voltage at which the channel appears is the threshold voltage $V_{th}$.

The insulated gate has a consequence worth stating on its own: in steady state, the gate draws NO current. None — not a small current, but essentially zero, limited only by leakage through the oxide, measured in picoamperes. A MOSFET is a voltage-controlled device in the purest sense, and this is its great advantage over the BJT of ME-3, which needs a continuous base current to stay on. But the gate IS a capacitor, typically hundreds to thousands of picofarads on a power part, so CHANGING the gate voltage moves charge, and moving charge takes current. Steady state costs nothing; switching costs something — and CA-5's $i = C\,dv/dt$ is the tool that says how much.