What is a transistor, and why does a phone chip hold billions of them?

A transistor is the most basic part of a chip: a tiny switch run by electricity. On means 1, off means 0. One switch can do very little, so it takes huge numbers of them wired together to compute anything, and the smaller the switches get, the more fit in the same space. That's why a phone chip today holds more than ten billion transistors.

What does a transistor look like?

Think of it as a tiny waterway. One end is the entrance, the other end is the exit, and the stretch in between is called the channel. A gate sits over the channel. When the gate opens, electricity flows from the entrance to the exit, and the switch is on. When it closes, nothing gets through, and the switch is off.

Their formal names are the source (the entrance), the drain (the exit) and the gate. All three fit the waterway picture: a source is where water comes from, a drain is where it goes, and a gate, like a floodgate, holds it back. The source and drain are made by adding atoms such as boron or phosphorus to the silicon, most often by ion implantation.

The gate is the key part. It doesn't actually touch the channel below it: a very thin insulating layer, which doesn't conduct electricity, sits in between. Put a voltage on the gate and it pulls a conducting path into the silicon underneath, connecting the source and the drain. Take the voltage away and the path disappears. Silicon can switch between conducting and not conducting like this because it is a semiconductor; for that part, read What is a semiconductor?.

How can a switch compute anything?

A switch is controlled by electricity, and what it passes on is electricity too. So the output of one switch can drive the gate of the next. Wire many switches together according to a design and they can add, compare two numbers or remember one; build those up layer upon layer and you get a chip that takes photos and finds your route. Each switch only ever does one thing, on or off, so the more a chip has to do, the more switches it needs.

Why keep making them smaller?

In the early 1970s, the first chips built to be a computer's "brain" held a few thousand transistors. A phone chip now holds more than ten billion, and shrinking is how it got there. Smaller brings three things:

• More of them: shrink the length and width each to 0.7 times and the area is roughly halved, so the same space holds twice as many switches. On the newest chips, transistors sit side by side just a few tens of nanometers apart, more than a thousand times finer than a hair.
• Less power: every flip means charging the gate with a little electricity and then letting it go, like filling a small cup and emptying it. A smaller transistor is a smaller cup, so each flip takes less.
• More speed: a small cup fills quickly, and the path from source to drain is shorter, so the switch flips faster.

So for decades, engineers have kept finding ways to make it smaller. The "nanometer" numbers in the news say which generation of that effort a chip comes from; see what "2 nm" really means.

What goes wrong when they get too small?

The biggest problem: they don't shut tightly.

The source and drain end up so close that the channel between them is very short, and the gate sits only on top of the channel. It controls the top well, but not the bottom. So even when the gate says "off", a little current still slips from source to drain. That's leakage.

The insulating layer under the gate has the same problem. It got thinner with every generation, and below about two nanometers, the number of electrons going straight through it shoots up, opening another path for leaks. The fix was a new material: a different insulator that can be made thicker while the gate keeps the same control.

One switch leaks so little you could ignore it, but add up more than ten billion of them and your phone uses power while it sits idle, and the chip gets warm. For a long time, each smaller generation also ran at a lower voltage, so chips got faster and more efficient at once. By around the 2000s, leakage had grown so large that the voltage couldn't come down much further, and simply shrinking everything in proportion was no longer enough.

So how do you make it shut tightly?

Imagine squeezing a soft garden hose until no water gets through. Press with one finger from the top and water still leaks underneath. Pinch it from three sides and it's much tighter. Wrap your whole hand around it and it's tightest of all. Transistors changed shape in the same direction:

1. Flat: the channel lies flat on the silicon surface, with the gate only on top. Transistors looked like this for a long time.
2. A thin standing wall: the channel is stood up on its edge as a very thin wall, and the gate grips it from three sides, the top and both sides. Standing it up also saves space, like replacing a one-story house with a taller building. From around the 2010s, the most advanced chips switched to this one after another.
3. Surrounded on all four sides: the channel becomes a few thin sheets stacked on top of each other, and the gate fills in around every sheet: top, bottom and both sides. Only the newest generations, around the ones the news calls 3 nm and 2 nm, have started using it, and different makers switch at different times.

Each change lets the gate hold the channel more firmly and shut it more tightly, so transistors can keep shrinking. And because progress came from changing the shape rather than just scaling everything down, the "nanometer" names drifted further and further from any real size.

In fab24hr

fab24hr's main story, The life of a chip, follows Glowmere, a fictional hearing-aid chip. It runs on a single button battery, so from the R&D line onward, standby leakage is what everyone fears most. In chapter 1 you decide which wafers get transistors made deliberately fast or slow (the corners); the fast ones run quickest and leak the most. In chapter 3 the WAT results come back: current, resistance and leakage measured on the wafer's test keys. In chapter 10 the test engineer decides how high the standby current can be before a chip is stopped. In A day in the fab, as the R&D process engineer for the R&D line's furnace, the morning shift asks how thin to grow the insulating layer under the gate: thinner switches faster, too thin leaks. The companies, chips, people and numbers in the game are fictional; how transistors switch and why they leak come from public, general sources. For other jargon, see Chip fab jargon in plain English.