Silicon: The Switchable Material

1The question of the week

Take stock. A bit is matter held in one of two states, with an energy barrier in between so thermal kicks cannot flip it by accident. Notes 03 priced the barrier: roughly forty k_BT buys a bit you can trust for years. So far it has been a curve on a plot — a model. This week it has to become a thing you can manufacture.

And the demands are brutal. You need trillions of barriers, identical, on a chip the size of a fingernail, for roughly nothing each. Each must be adjustable at the factory — you choose how conductive every region is during fabrication. And each must be switchable in service: raised and lowered electrically, billions of times a second, for almost no energy per flip. That last demand rules out both obvious categories of material.

• A metal can never hold the barrier up. Charge flows freely everywhere, so any wall you build leaks away instantly. A copper bit forgets itself before you finish writing it.
• An insulator holds a barrier beautifully and can never take it down. Glass does not have an off switch. A glass bit can never be written at all.

What you need sits between the two — nearly an insulator when left alone, but so close to conducting that a small, cheap nudge tips it over. A material whose conductivity is not a fixed property but a design knob. That is what a semiconductor is, and this week is the story of why silicon sits at that sweet spot and how engineers turned the sweet spot into a valve. No band-structure mathematics required: one good picture carries the whole week, and we build it next.

2The parking garage

Here is the picture. An electron inside a crystal is like a car in a parking garage: it must sit on a floor. No hovering between floors. The floors exist for a genuinely quantum reason — electrons in a crystal, like a guitar string, can only "ring" at certain energies — and that borrowed fact is the only quantum mechanics you will be asked to take on trust. The rest of the week is counting cars and counting spaces.

Now add the second rule, and it is the one that does all the work: a car can only move if there is an empty space to move into. Electric current is cars sliding across a floor. So ask of any material: does its garage have both cars and spaces on the same floor?

Read the garages left to right. In an insulator, the ground floor is packed bumper to bumper and the next floor is so far up that no car will ever get the energy to reach it. A full floor carries no current — not because the cars are stuck, but because there is nowhere for any of them to go. In a conductor, the top occupied floor is only partly full: cars and spaces coexist, so the smallest push makes traffic flow.

Silicon is the interesting case. Its ground floor is completely full — on its own, it is an insulator — but the next floor up is close. Close enough that the thermal jostling from Notes 03, the same k_BT agitation that attacks stored bits, occasionally bumps a car upstairs. The moment that happens, you get two mobile things for the price of one kick:

• The promoted car is upstairs on an empty floor. It can drive anywhere. That is a free electron.
• Downstairs, the car left behind an empty spot in a full floor. Watch the spot: the next car slides in, leaving a new empty spot, and another car slides into that. The spot drifts through the garage, opposite to the shuffling cars, behaving for all the world like a positive charge cruising the ground floor. We call it a hole and count it as a carrier in its own right — tracking one moving spot beats tracking a thousand shuffling cars.

That is the whole vocabulary of the week: a full floor, an empty floor, the gap between them, free electrons upstairs, holes downstairs. (The textbook names, for honesty's sake: valence band, conduction band, and band gap, written E_g. Use whichever you like; the pictures are the same.)

3Pure silicon is boring on purpose

How often does heat actually bump a car upstairs? Notes 03 gave you the tool: the chance of a thermal kick clearing an energy step falls off exponentially once the step passes k_BT. Silicon's gap is E_g = 1.12 eV, and at room temperature k_BT = 25.9 meV — the gap is about 43 thermal kicks tall. Boltzmann's exponential treats that like a 43-story climb: not impossible, but very, very rare.

Here is what "very rare" comes to. A sugar cube of silicon holds about 5 × 10²² atoms; at room temperature only about 10¹⁰ electrons are upstairs at any moment. Divide: about one atom in five trillion has contributed a free electron. If every atom were a person, that is the population of five thousand Earths watching a field with a handful of players on it. Copper fields one free electron per atom. Pure silicon is far closer to glass than to metal — a near-insulator with a faint pulse.

Two things about this near-emptiness matter. First, it is exquisitely temperature-sensitive: the count roughly doubles every 8–10 °C, because you are on the steep part of an exponential. A device relying on pure silicon's own carriers would be a thermometer, not a switch. Second — the point of the section — the boringness is the feature. Pure silicon is a nearly blank page: so few native carriers that a whisper of deliberate contamination will completely overwhelm them. A material teeming with its own carriers would drown out anything you added. Silicon's emptiness is what makes the next section possible.

One caution first: thermal promotion is not the switch we want. "Heat it up to turn it on" fails every demand in section 1 — heat is slow, indiscriminate, and expensive. The small gap matters for a different reason, about to appear.

4Doping: the design knob

Now the trick that turns a curiosity into an industry. Silicon's garage is boring — so edit it. Swap a tiny fraction of the atoms for a periodic-table neighbor chosen for having the wrong number of electrons. This is doping, and it comes in exactly two flavors.

Flavor one: an atom with one electron too many. Phosphorus has five outer electrons where silicon has four. Four settle into the crystal's bonds like everyone else's. The fifth finds the ground floor full, so it has nowhere to sit but upstairs — no thermal kick required; the extra car is on the open floor from the moment the crystal is grown. Each phosphorus atom donates one free electron, so it is a donor, and silicon doped this way — extra negative carriers — is n-type.

Flavor two: an atom with one electron too few. Boron has three outer electrons. It takes a parking spot in the crystal but cannot fill it, leaving a ready-made empty space on the ground floor — a hole, manufactured rather than waited for. Boron accepts a neighboring electron into the spot, so it is an acceptor, and the material — carrying positive movers — is p-type.

Now the numbers, because they are the astonishing part. A typical doping level is one dopant atom per million silicon atoms — if the crystal were the population of Earth, you added a small town. But pure silicon only had one carrier per five trillion atoms, so your small town outnumbers the natives five million to one. A parts-per-million sprinkle raises conductivity by six or seven orders of magnitude — and makes it stable, because the carrier count is now set by chemistry, which does not fluctuate, instead of thermal luck, which does.

Step back and look at what you now own. Conductivity is no longer a property the material has; it is a number you choose, region by region, at the factory, across roughly eight orders of magnitude, by metering a sprinkle. You choose the sign of the carriers too: electrons upstairs or holes downstairs. Demand two from section 1 — adjustable at the factory — is met. Play with both knobs on the bench before reading on.

Two honest trends, no fake precision. Undoped, the population rides an exponential — silicon at the weather's mercy. Doped even lightly, temperature barely matters: chemistry outvotes heat. Push the heat far enough, though, and the thermal crowd overtakes any sprinkle — one honest reason hot chips misbehave.

5The pn junction: a barrier you can manufacture

One more move and we reach the week's destination. So far each doped garage is just a resistor with a chosen value. The magic happens at a border. Make one region of a crystal p-type and its neighbor n-type — same crystal, no physical seam, just where the sprinkle changes — and the line between them is a pn junction: the most consequential border in engineering. You can reason out everything it does from the garage.

Think about the instant the two regions meet. On the n side, spare cars upstairs. On the p side, spare spots downstairs. A car near the border does the obvious thing: it falls into a nearby spot. Car and spot annihilate as a pair. This happens all along the border, leaving a stripe of crystal with no mobile carriers at all: the depletion zone. A perfect insulator, a few hundred nanometers wide, in the middle of a conductive crystal.

Why does the falling stop instead of eating the whole crystal? Look at what the carriers left behind. Every donor whose electron fell across is still bolted into the n side with its charge of +; every acceptor that swallowed one sits on the p side holding a −. A wall of fixed positive charge faces a wall of fixed negative charge across the zone, and together they push back on any further electron trying to cross. Each fall makes the push stronger, so the process throttles itself, settling where push-back balances falling — in silicon, when crossing has become an uphill climb of about 0.7 volts. The junction has built its own hill, out of its own charge, with no battery attached.

Now stop and look at frame (c), because you have seen it before. It is the energy-barrier diagram from Notes 02. A carrier at the bottom, a hill it cannot climb, a state that stays put — except this time the barrier is not an abstraction on a slide. It is a real object, about 0.7 V tall, that a crystal will build for you anywhere you draw a doping border, at a cost of nothing, a trillion times per chip, every copy alike. That answers the question of the week. What remains is the on-off part.

Tilting the hill: the one-way valve

The payoff for all that self-assembly: the hill is adjustable from outside. Connect a battery across the junction and you tilt the balance that set its height.

• Push the p side positive ("forward") and the hill gets lower. Boltzmann from Notes 03 does the rest: the fraction of electrons with enough thermal energy to clear a hill grows exponentially as the hill shrinks. Around 0.6–0.7 V of push the hill is nearly gone and current pours — every extra 18 mV roughly doubles it.
• Pull the p side negative ("reverse") and the hill gets taller, the depletion zone wider. Only a picoamp-scale leak flows, and pulling harder barely changes it — a hill that was already unclimbable cannot get more unclimbable in any way that matters.

A device that conducts one way and blocks the other is a diode — a one-way valve for current, useful on its own (every phone charger rectifies with one). But for this course the point is the sentence hiding inside it: we now have a manufactured barrier whose height is set by a voltage. Tilt it yourself on the bench.

Sweep slowly from −2 V to +0.9 V and notice the asymmetry that makes it a valve: two volts of reverse pull buys picoamps, while the last 0.1 V of forward push multiplies the current fifty-fold. Same Boltzmann factor you modeled in Notes 03 — now with a knob on it.

6From junction to switch — a look ahead

The diode has one shortcoming as a computing element: the control signal and the controlled current share the same two terminals — the same door. Logic needs a valve with a separate handle: a third terminal that raises and lowers the barrier without being part of the path.

The fix is almost cheeky. Park a metal plate above the barrier region, separated from the crystal by an insulating film a few atoms thick. No current can cross the insulator — but the plate's voltage reaches through and pushes on the carriers below, just as the depletion zone's fixed charges do. Charge the plate one way and the hill under it grows: path cut, switch off. Charge it the other way and the hill collapses: carriers flood underneath, switch on. The plate is called the gate, and the whole device is a transistor — a barrier from Notes 02, built by section 5, with an electrical handle on top.

That is deliberately all for this week. But notice that every demand from section 1 now has an answer: barriers by the trillion (draw doping borders), set at the factory (meter the sprinkle), switched in service (put a handle over the hill). Next week we open the device up properly — the gate's grip, the operating regions, and the models you will simulate in Project 2.

7Work before Session 7

Session 7 is Monday, September 28, and it opens the MOSFET — built entirely out of this week's parts. Project 1 is due at the end of that week, so the list below is deliberately half review, half project. You now hold every physical ingredient the project asks about: the barrier, its price, and — as of this week — proof that such barriers actually get manufactured.

• Review these notes. The test: draw Fig 5.1's three frames from memory and narrate them out loud — spare carriers, the pairing-off at the seam, the fixed charges, the hill. If any step feels like recitation instead of a story, reread section 5.
• Spend ten minutes on each bench. On Bench 4.1, find the temperature where "1 per 10 million atoms" doping loses control to heat. On Bench 5.1, find the voltages where the current first reaches one microamp, then one milliamp, and check the spacing against the 18 mV doubling rule.
• Project 1, in progress: by this weekend your barrier model should run end to end — error rate out for barrier height, temperature, and hold time in. Commit as you go.
• Add one short paragraph to your Project 1 report draft: in your own words, how is the junction's built-in hill a physical instance of the abstract barrier your model prices? Name an honest difference or two as well (your model's barrier height was a free parameter; the junction's sits near 0.7 V until a voltage moves it).
• Optional, for the curious: pull up the datasheet of the 1N4148 — the most-manufactured diode in history — and compare its forward voltage and reverse leakage with Bench 5.1.

8Exercises

Not graded, as always. Conceptual on purpose — every one of them can be answered with the garage and nothing else. Bring your answers, or your stuck points, to Session 7.

1. The full-floor paradox. A friend objects: "A full floor has the most electrons — how can it carry the least current?" Answer with the garage. Then: when ground-floor cars shuffle left one space at a time, which way does the empty spot travel, and why is it fair to bill the hole as a positive charge moving on its own?
2. The zone that builds itself. The depletion zone forms by carriers falling into spots — and then the falling stops on its own. What stops it? (Your answer should feature the atoms that are bolted down, not the carriers that move.) Then use the same picture twice more: why does reverse voltage make the zone wider, and forward voltage narrower?
3. Diamond. Diamond has silicon's crystal structure but a gap of 5.5 eV — about 210 thermal kicks at room temperature. (a) Using the Notes 03 exponential in words, why is pure diamond one of the best insulators known? (b) A sensor company wants electronics inside a 300 °C drill head. Using "heat eventually outvotes any sprinkle" from section 4, make the case that doped diamond could stay a switchable material at temperatures where silicon has dissolved into a conductor in all directions.

§Symbols and constants