Semiconductors and Electronics

In a solid, electrons can only have energies in certain ranges called bands. The valence band holds electrons tied up in bonds. The conduction band above it holds electrons that are free to move and carry current. The energy between them is the band gap.

In a metal, the bands overlap, so plenty of electrons are always free: metals conduct well. In an insulator such as diamond, the gap is big, about 5.5 eV, and almost no electrons can jump it. A semiconductor such as silicon has a small gap, about 1.1 eV. A few electrons get across thanks to heat, and more do when it's warmer, so a semiconductor conducts better as it heats up, the opposite of a metal.

Silicon has four outer electrons, and each atom shares them with four neighbors in a crystal. Germanium, with a gap of about 0.67 eV, behaves the same way and was used in the earliest transistors.

Pure silicon conducts poorly, so chipmakers dope it, replacing a tiny fraction of its atoms, often around one in a million or fewer, with another element.

n-type: phosphorus has five outer electrons, one more than silicon. Four go into bonds and the fifth is easily freed, so n-type silicon has extra free electrons, which are negative charge carriers.

p-type: boron has three outer electrons, one fewer than silicon, so one bond is missing an electron. That gap is a hole. A neighboring electron can hop into it, which moves the hole the other way, so holes behave like positive charges moving through the crystal.

Both kinds of doped silicon are still electrically neutral overall. Doping just adds charges that are free to move.

Put p-type and n-type silicon side by side and some free electrons near the boundary drift across and fill holes. That leaves a thin depletion zone with no free charges, and a built-in electric field across it.

Connect a battery with + on the p side (forward bias) and, once the voltage is above about 0.6–0.7 V for silicon, the zone shrinks and current flows easily. Connect it the other way (reverse bias) and the zone widens, so almost no current flows. That one-way device is a diode.

Diodes are everywhere: phone chargers use them to turn AC from the wall into DC (see Unit 18), and they protect circuits from batteries put in backwards.

In a light-emitting diode, electrons crossing the junction drop from the conduction band into holes in the valence band. Each drop releases a photon whose energy is about the band gap. Using E = hf and c = fλ,

λ = hc ÷ E ≈ 1240 eV·nm ÷ E

So a bigger gap gives higher-energy, shorter-wavelength light. Red LEDs use materials with gaps near 1.9 eV; blue LEDs need about 2.7 eV, which took decades to achieve. White LEDs are usually blue LEDs with a yellow phosphor coating.

Silicon's gap is too small, and its electrons rarely give off light anyway, so LEDs use compounds such as gallium arsenide phosphide or indium gallium nitride. An LED also needs a series resistor to limit the current: R = (V_supply − V_LED) ÷ I.

A transistor is a sandwich of n-type and p-type regions with three connections. In the kind used in almost every chip, a MOSFET, a voltage on the gate pulls charges into a thin channel under it. With no gate voltage the channel is closed and no current flows; with the gate on, current flows. A small signal controls a bigger current.

Used as a switch, the two states stand for the binary digits 0 and 1. Used as an amplifier, a small changing signal on the gate makes a larger copy of it in the output current, which is how a microphone's tiny signal gets loud enough to drive a speaker.

Wire a few transistor switches together and you get logic gates. An AND gate outputs 1 only if both inputs are 1; an OR gate outputs 1 if either is; a NOT gate flips its input. Gates combine into circuits that add numbers, store bits and run programs.

An integrated circuit builds transistors and wiring together on one slice of silicon by printing patterns with light, layer after layer. In 1965 Gordon Moore noticed that the number of components on a chip was doubling every year; in 1975 he revised it to about every two years. That trend became known as Moore's law.

It held up remarkably well. Intel's 4004 of 1971 had 2,250 transistors. Apple's M3 Ultra, released in 2025, has 184 billion, on two joined dies. That's about 26 doublings in 54 years, close to one every two years.

Moore's law is an observation about an industry, not a law of physics, and it's slowing. Chip designers report progress has run slightly below the old pace since around 2010, as transistors approach the size of a few dozen atoms. Experts disagree about when it will end.

A solar cell is a large p-n junction, usually silicon, with a thin n-type layer on top. When a photon with more energy than the band gap is absorbed near the junction, it frees an electron and leaves a hole. The junction's built-in field pushes the electron one way and the hole the other, so current flows through an outside circuit. It's an LED running in reverse.

A silicon cell gives about 0.5–0.6 V, so panels connect dozens of cells in series. Photons with less energy than the gap, wavelengths longer than about 1 100 nm, pass through unused, and any energy above the gap is lost as heat. Those losses limit a single-junction cell to about 33% efficiency (the Shockley–Queisser limit). The best commercial panels now reach about 24%. Panels make DC; an inverter turns it into AC for the grid (see Unit 18).

Before transistors, electronics ran on glass vacuum tubes that were big, hot and burned out. Semiconductors changed that in a few decades.