Electronics · Under the hood · Optional — the deep story · build 2026.09.23-1644
The idea
A switch with no moving parts needs a material that can change its mind about conducting. That material is the semiconductor — and the transistor is a sandwich made of it.
Sort the world's materials by whether they conduct and you get two bins: metals conduct (their outer electrons roam loose), plastics don't (every electron is held tight). Silicon — refined from quartz sand, through heroic purification — sits stubbornly in between. Each silicon atom shares its four outer electrons with four neighbors, locking everything into a tidy crystal where, at room temperature, almost nothing is free to move. Pure silicon is nearly an insulator, and nearly useless.
Its superpower is how easily it can be persuaded.
Swap roughly one silicon atom in a million for a different element — the exact pinch varies wildly by recipe, but always on purpose and with absurd precision — and everything changes. This is called doping:
Neither type is charged
Common trap: N-type silicon isn't negatively charged — each phosphorus atom brought an extra proton along with its extra electron, so the crystal stays neutral. N and P describe which kind of carrier is free to move, not the crystal's overall charge.
Now make a single crystal whose one side is doped P and the other N — one unbroken lattice, not two blocks glued together. At the border inside it:
Right at the border, the roaming electrons fall into the nearby holes, and both vanish as carriers. What's left is a thin strip — the depletion zone — with nothing mobile in it: an insulating moat the crystal built by itself. And the falling-in doesn't spread forever: every intruder atom at the border is now electrically charged (the phosphorus that lost its spare electron sits at +, the boron that filled its seat sits at −), and the pushback from those stranded charges halts the process. That built-in pushback is a real voltage — in silicon, about 0.7 V worth. It's the barrier every crossing has to pay for, and whether current can cross depends entirely on which way you push:
Where the simple model ends
Two honesty notes, because engineers round where physicists wince. The door is not a cliff: forward current grows continuously and steeply — roughly doubling with every small extra push — and "0.7 V" is the engineer's rounding for where it gets big enough to matter; the true number drifts with current and temperature. And "no flow" backwards really means almost none: a tiny leakage always seeps through, and pushed hard enough backwards, any junction breaks down. Every tidy rule in this course is a useful approximation with fine print like this — the fine print is where the physics lives.
Three things you've already met are hiding in that picture:
First, a name. The kit's transistor is a bipolar junction transistor — a BJT — the founding member of the transistor family, and the one this page opens up. (Its field-effect cousins run your phone; they get their moment below.)
NPN means exactly what it spells: a sandwich of N, then P, then N. That makes two junctions back to back — which should mean nothing ever flows from collector to emitter, and with the base untouched, nothing does. The trick that makes it a transistor: the middle P layer is astonishingly thin.
Open the base-emitter door (0.7 V, as always) and the emitter — deliberately doped extra-heavily, so the flood is almost entirely its electrons — pours them into the base. But the crossing isn't a sprint — nothing aims the electrons. They wander (physicists say diffuse: a random shuffle through the crowd), and the base is a hallway so short, with so few holes in it, that about 99 of every 100 stumble out the far side before ever finding a hole — where the collector junction's field sweeps every arrival across. The unlucky 1-in-100 falls into a hole; that, plus a small backwash of holes slipping the opposite way, is the base current.
Read that as currents: collector current ≈ 100 × base current. β is not electronics magic — it's how the sandwich is built: base thickness and doping balance set it, and even temperature and the current of the moment nudge it. That's why β varies so much part to part and reading to reading — it rides on nanometer-scale manufacturing luck. And the multiplication only rules while the collector circuit can supply it: feed the base more than the load allows and the transistor saturates — fully on, a plain closed switch dropping ≈ 0.2 V. (The Transistor's Arithmetic sheet draws that exact boundary; Tolerance & Error explains living with the spread.)
One device, two arrow directions
These pictures follow the electrons, which pour emitter → collector. But "current" on every schematic ever drawn points collector → emitter. Both are right: electrons carry negative charge, so conventional current points opposite to their travel. The convention is a guess Benjamin Franklin made a century before anyone discovered the electron — for wires it came out backwards, and the world kept it anyway. The transistor symbol's emitter arrow points the conventional way.
Flip every layer and you flip every rule: the PNP's flood is made of holes, its currents run the opposite way, and its base current flows out instead of in. Circuits love using NPN and PNP together as mirror-image teammates. This course sticks to NPN — but when you meet a PNP in the wild, you already know its whole story: same sandwich, opposite seasoning.
One honest twist before the scale-up: your phone's billions of transistors are mostly not BJTs. Chips run on the BJT's cousin, the MOSFET — a field-effect transistor that switches with voltage alone: an electric field reaching across a whisker-thin insulator pinches a channel of doped silicon open or shut, with almost no control current at all. Same silicon, same doping, same junctions — a different trick built from the same ingredients, and far thriftier with power, which is why it won the chip. Shrink it to nanometers, pack tens of billions side by side switching billions of times a second, and you have what the news calls "semiconductors." The BJT you built with is the family's founding member — and still everywhere a circuit needs one strong, simple switch.
Where it came from
The first transistor was built in December 1947 at Bell Labs by John Bardeen and Walter Brattain, in a group led by William Shockley — who answered weeks later by inventing the sturdier junction design: the very NPN sandwich on this page. All three shared the 1956 Nobel Prize, and computers stopped needing rooms full of hot glass vacuum tubes. Bardeen later won a second physics Nobel — still the only person ever to do it.