What is a diode? Why a PN junction lets current flow only one way

A diode is a part that lets electric current flow mostly in one direction: push from one end and current gets through; push from the other end and almost nothing does. At its heart is a PN junction, the boundary where p-type and n-type semiconductor meet. Turning AC into DC, light-emitting diodes (LEDs), solar cells and the static protection next to a chip's pins all rely on that boundary.

What is a diode, and why "mostly"?

A diode has just two ends: the p-type end is the anode and the n-type end is the cathode. In a circuit diagram it is drawn as a triangle pressed against a short bar, and the tip of the triangle points the way current can flow.

The usual comparison is a check valve in a water pipe: push one way and the valve opens; push the other way and it presses shut. "Mostly" is there for three reasons: pushing the right way has to pass a threshold first; pushing the wrong way still lets a tiny leak through; and pushing the wrong way too hard breaks through.

What is a PN junction, and why is there an empty zone at the boundary?

A quick recap: n-type is silicon with extra electrons that can move; p-type has extra empty spots, called holes, which you can think of as moving positive charge. Neither side is charged as a whole (the full story is in What is doping?). Put the two together, and this happens near the boundary, in order:

1. Spreading toward where there are fewer. The n side has more electrons, so electrons spread into the p side; holes spread the other way. It's like a drop of ink spreading in water from where it's dense to where it's thin; this is called diffusion.
2. Meeting and cancelling. An electron that meets a hole fills the empty spot, and neither can move anymore.
3. Charge left behind that can't move. The dopant atoms are locked into the silicon lattice. On the n side near the boundary, phosphorus atoms have given away their extra electron and are positive; on the p side, boron atoms have taken in one extra electron and are negative. Between the positive and the negative an electric field forms, pushing back any more electrons that try to cross.
4. Settling into balance. The pull of spreading and the push of the field even out. On both sides of the boundary there is now a zone with almost no free electrons or holes, called the depletion region.

Picture the depletion region as a slope at the boundary. Electrons are always moving about at random, and only the few with enough energy climb the slope and reach the other side; the higher the slope, the fewer make it.

What happens when you connect it one way, or the other way?

Connecting a battery across the two ends makes that slope lower or higher.

• Forward bias (p to positive, n to negative): electrons and holes are both pushed toward the boundary, the depletion region narrows, the slope drops, and suddenly far more electrons make it over. A silicon diode starts conducting noticeably at several tenths of a volt (under one volt); past that threshold, a little more voltage brings a lot more current.
• Reverse bias (p to negative, n to positive): electrons and holes are pulled away from the boundary, the depletion region widens, the slope rises, and only a tiny leak gets through.
• Pushed too hard in reverse, it breaks down: the current suddenly jumps, and an ordinary diode can overheat and fail.

What are diodes used for?

Rectification: the first step from AC to DC

The AC from a wall outlet goes back and forth fifty or sixty times a second, but the circuits in a phone or computer want DC, which always flows the same way. Rectification goes in this order:

1. One diode blocks the backward half. Without it, the current goes one way, then the other, and circuits that need DC can't use it. But only half is left, so the current comes in pulses.
2. Four diodes in a loop (a bridge rectifier). When current flows one way it passes through two of them, and when it flows the other way it passes through the other two, so both halves come out in the same direction. Without this, half the time nothing is delivered.
3. A capacitor smooths out the bumps. A capacitor works like a small water tower: it fills when the voltage is high and gives back when it's low. Without it, the voltage rises and falls, and the circuits after it are unsteady.

After that, the voltage still has to be brought down to the few volts a phone needs; see What is a power semiconductor?

Protection: blocking a reversed battery, draining a sudden surge

• Reverse-polarity protection: a diode in series after the power input is reverse biased if the battery goes in backward, so it blocks the current and the parts aren't burned; the cost is that in normal use it takes a few tenths of a volt for itself.
• Static protection next to a chip's pins: one common approach puts two diodes next to a pin, one to the power line and one to the ground line. In normal use the pin's voltage sits between the two, both diodes are reverse biased, and the signal isn't affected; if a voltage higher than the supply or lower than ground suddenly hits the pin, one of them becomes forward biased and drains it away (why chips fear static: What is electrostatic discharge (ESD)?).

Light-emitting diodes (LEDs): light when electrons meet holes

Under forward bias, electrons and holes meet in large numbers near the boundary and cancel, and each pair that cancels gives off a little energy. In silicon that energy mostly becomes heat; in some compound semiconductors it comes out as light. The color depends on the material's bandgap, how high the step is that an electron has to jump: the higher the step, the bluer the light (see The news says "wide-bandgap semiconductors". What are SiC and GaN?).

Photodetectors and solar cells: turning light into electricity

It also works in reverse. Light shining near the boundary knocks a bound electron loose and leaves a hole behind. The electric field already inside the depletion region pushes the electron toward the n side and the hole toward the p side; if they weren't separated, they would soon cancel again. Each side now has a little extra charge, and once a wire connects them, current flows.

A photodetector (also called a photodiode) usually has a reverse voltage on it, and how much current flows tells you how bright the light is; the part that turns light back into electricity in The news says "silicon photonics" does exactly this. A solar cell has no voltage applied and simply turns light into electricity it sends out, so you can think of it as a photodiode with a very large area.

Are there PN junctions inside a transistor too?

Yes, everywhere. In a common kind of transistor, the source and drain are n-type and the big block under them is p-type, so each end forms a PN junction with the silicon below; fabs make these regions mostly by ion implantation. The circuit is wired so these junctions are never forward biased, which is why no current gets through while the gate has no voltage on it. How the gate switches it on: What is a transistor?

How do chips use PN junctions to keep things apart?

A CMOS chip needs two kinds of "floor," p-type and n-type, on the same piece of silicon, called wells (see What is CMOS?). The p-type floor is tied to the lowest voltage (ground) and the n-type well to the highest (the supply), so the PN junction between them is always reverse biased, and its depletion region acts like a wall that keeps the regions apart. Between one transistor and the next, the surface is separated by shallow trenches filled with insulator (see What is shallow trench isolation (STI)?).

In fab24hr

There's no stand-alone diode in fab24hr. The closest is "Junction", a term you can tap in the dialogue: the boundary between the layer of implanted atoms gathered in the silicon and the silicon beneath it. When the implanted atoms and the silicon below are opposite types, that boundary is a PN junction.

Play the R&D line's furnace R&D process engineer in A day in the fab: tap "Clock in", choose "A day in the fab", then on the "Your role today" card tap "R&D line", swipe to the R&D process engineer and tap "Furnace". The afternoon shift is ten days after wafer start. Implanted test wafers have been through three anneal conditions, and Hsiao-Wen measured two things: sheet resistance, to see whether the atoms have started to conduct, and junction depth, to see how far down they moved. The higher the temperature and the longer the time, the better the activation and the farther the atoms move: that's the thermal budget. The game says Glowmere runs at a very low voltage; if the junction is too deep, the short transistors can't turn off tightly, and they leak in standby. You choose between the middle temperature and the high temperature, long time the production fab uses; pick the high one, and later Hsiao-Wen also measures the short-channel test structures: they leak more than expected.

The companies, the chip Glowmere, the people, the tools and the figures in the game are fictional; how PN junctions and diodes work follows public, general sources. For other jargon, see Chip fab jargon in plain English.