From Sand to Switch — The Semiconductor

What conducts, what does not, and what sits in between

Some things carry electricity well: copper wire. Some do not: a rubber glove. That is why wires are copper wrapped in rubber.

Silicon sits in between. Neither a good conductor nor a good insulator — a 'semi' conductor. Left alone it looks useless, neither one thing nor the other.

Join two pieces of silicon with different characters and you get a door that lets electricity through one way only. Line up three, give the middle one a small nudge of electricity, and the door opens; cut the nudge and it closes. That is a transistor — the 'switch pressed by electricity' from piece one.

How to make tens of billions — print, do not carve

Here is this piece's surprise. Making switches one at a time and joining them could never reach tens of billions. So a different method is used: printing. Like a stamp, like a photograph.

1. Melt sand and pull out a pure silicon column — Slice it thin and you have wafers. Hundreds of chips are made together on each one.
2. Coat the wafer with a light-sensitive layer — Same idea as photographic film: only where light lands does the layer change.
3. Lay a glass plate carrying the circuit drawing over it and shine light — Shadowed spots stay; lit spots wash away, and impurities are planted there.
4. Repeat dozens of times, layer on layer — One layer of switches, then wiring, then more wiring. Tens of billions of switches appear at once.

So a chip is not sculpted; it is photographed into existence with light. One exposure, hundreds of chips per wafer, tens of billions of switches per chip.

The idea began in 1958: instead of making parts separately and wiring them, make all the parts on a single piece.

Why ultraviolet

The light series said light is a wave and colour is the size of the ripple: red is a long ripple, violet a fine one.

To draw a fine line you need a fine ripple — a thick brush cannot write small letters. Visible light cannot draw a line even as thin as a hair. So ultraviolet is used, and now extreme ultraviolet, finer still.

A little further in

Smaller, smaller — and then a wall

Why keep shrinking? Three things improve: more fit on a wafer, so it is cheaper; electricity travels less far, so it is faster; and it uses less power. That is why piece one called half the history of the computer the history of shrinking the switch.

Now a wall is in sight. Today's lines are a few dozen atoms wide. Any thinner and electricity leaks straight through the walls. You cannot build smaller than an atom.

So instead of shrinking sideways, chips now stack upward — like apartment blocks. The next piece tells that story: why the memory chip became the bottleneck of the whole computer, and how stacking solved it.

The question that remainsTens of billions of switches from sand sit on a fingernail. Should we say we 'made' them — or 'printed' them?