The idea in a minute
Last time, we purified silicon until it was one of the purest things people have ever made . And on its own, it was almost useless: a near-insulator, with a gap of about 1.1 electron volts . The physicist Wolfgang Pauli once called semiconductors a filthy mess, and the mess was impurities . Here's the twist. Put the right impurity back, just one atom in a million, and the silicon conducts millions of times better . How can a handful of wrong atoms do that? Quick recap . Every silicon atom has four outer electrons, and every one of them is busy in a bond with a neighbour . To get free, an electron has to jump a gap of about 1.1 electron volts, and at room temperature only about one in five trillion manages it . So pure silicon has almost no free charge to carry a current . In February 1940, a Bell Labs chemist called Russell Ohl was testing a rod of silicon with a crack across it . When light fell on it, the current changed . Less than two weeks later, he showed his research director, Mervin Kelly . He shone a flashlight on the rod, and it produced about half a volt . Walter Brattain, who would later co-invent the transistor, said the effect was one or two orders of magnitude bigger than anything they'd seen. He thought his leg was being pulled . So what was in the rod? Chemical analysis couldn't tell them, because the impurities were too faint to detect . It turned out that as the molten silicon froze, two different impurities had spread out differently, so one side of the crack ended up with more of one, and the other side with more of the other . One side behaved one way, the other side the opposite way. The chemist Henry Theuerer suspected phosphorus on one side, partly because he'd smelled a trace of phosphine, a phosphorus gas . And adding boron made the other side stronger . Two kinds of impurity, with opposite effects. Let's take them one at a time. Start with phosphorus. A phosphorus atom has five outer electrons, one more than silicon . Slip one into the crystal in place of a silicon atom, and four of its electrons fill the four bonds, just like silicon's do . The fifth has no bond to sit in . It's a spare. And that spare electron is barely held at all. It's attracted to just one extra positive charge, on the phosphorus . But the silicon all around it muffles that pull, the way a crowd muffles a voice . And inside the crystal, the electron behaves as if it were lighter than a free electron, which weakens the grip even more . So instead of hugging its atom, the spare electron's orbit balloons out across a thousand or more silicon atoms . In the band picture from last time, the spare electron sits on a level just below the conduction band, only about 0.045 electron volts down . That's around four per cent of the full gap . Freeing a bonded silicon electron is a big leap. Freeing a phosphorus spare is a small step. So at room temperature, heat sets almost every phosphorus spare loose, while it frees only one silicon electron in trillions . Each freed electron leaves its phosphorus behind as a fixed positive ion, locked in place . Atoms that give away an electron like this are called donors . The silicon now carries current mainly with negative electrons, so it's called n-type . But it isn't charged. Every free electron is balanced by a fixed positive ion, so the whole crystal stays neutral . Now boron. Boron has only three outer electrons, so one of its four bonds is missing an electron . A neighbouring electron needs only a small nudge, again about 0.045 electron volts, to hop into that gap . When it does, the boron becomes a fixed negative ion, and the vacancy it leaves is a hole, the same moving hole from last episode, but made on purpose . Atoms like boron are called acceptors, and silicon doped this way carries current mainly with positive holes. It's p-type . Even the names have two stories . The usual one is p for positive and n for negative, after the charge of the carriers . But one textbook says Ohl named the two sides after the polarity of voltage that made each of them conduct, and the link to boron and phosphorus came later . Either way, the labels stuck. So how much does it take? Astonishingly little .
Read the full transcript
Last time, we purified silicon until it was one of the purest things people have ever made . And on its own, it was almost useless: a near-insulator, with a gap of about 1.1 electron volts . The physicist Wolfgang Pauli once called semiconductors a filthy mess, and the mess was impurities . Here's the twist. Put the right impurity back, just one atom in a million, and the silicon conducts millions of times better . How can a handful of wrong atoms do that? Quick recap . Every silicon atom has four outer electrons, and every one of them is busy in a bond with a neighbour . To get free, an electron has to jump a gap of about 1.1 electron volts, and at room temperature only about one in five trillion manages it . So pure silicon has almost no free charge to carry a current . In February 1940, a Bell Labs chemist called Russell Ohl was testing a rod of silicon with a crack across it . When light fell on it, the current changed . Less than two weeks later, he showed his research director, Mervin Kelly . He shone a flashlight on the rod, and it produced about half a volt . Walter Brattain, who would later co-invent the transistor, said the effect was one or two orders of magnitude bigger than anything they'd seen. He thought his leg was being pulled . So what was in the rod? Chemical analysis couldn't tell them, because the impurities were too faint to detect . It turned out that as the molten silicon froze, two different impurities had spread out differently, so one side of the crack ended up with more of one, and the other side with more of the other . One side behaved one way, the other side the opposite way. The chemist Henry Theuerer suspected phosphorus on one side, partly because he'd smelled a trace of phosphine, a phosphorus gas . And adding boron made the other side stronger . Two kinds of impurity, with opposite effects. Let's take them one at a time. Start with phosphorus. A phosphorus atom has five outer electrons, one more than silicon . Slip one into the crystal in place of a silicon atom, and four of its electrons fill the four bonds, just like silicon's do . The fifth has no bond to sit in . It's a spare. And that spare electron is barely held at all. It's attracted to just one extra positive charge, on the phosphorus . But the silicon all around it muffles that pull, the way a crowd muffles a voice . And inside the crystal, the electron behaves as if it were lighter than a free electron, which weakens the grip even more . So instead of hugging its atom, the spare electron's orbit balloons out across a thousand or more silicon atoms . In the band picture from last time, the spare electron sits on a level just below the conduction band, only about 0.045 electron volts down . That's around four per cent of the full gap . Freeing a bonded silicon electron is a big leap. Freeing a phosphorus spare is a small step. So at room temperature, heat sets almost every phosphorus spare loose, while it frees only one silicon electron in trillions . Each freed electron leaves its phosphorus behind as a fixed positive ion, locked in place . Atoms that give away an electron like this are called donors . The silicon now carries current mainly with negative electrons, so it's called n-type . But it isn't charged. Every free electron is balanced by a fixed positive ion, so the whole crystal stays neutral . Now boron. Boron has only three outer electrons, so one of its four bonds is missing an electron . A neighbouring electron needs only a small nudge, again about 0.045 electron volts, to hop into that gap . When it does, the boron becomes a fixed negative ion, and the vacancy it leaves is a hole, the same moving hole from last episode, but made on purpose . Atoms like boron are called acceptors, and silicon doped this way carries current mainly with positive holes. It's p-type . Even the names have two stories . The usual one is p for positive and n for negative, after the charge of the carriers . But one textbook says Ohl named the two sides after the polarity of voltage that made each of them conduct, and the link to boron and phosphorus came later . Either way, the labels stuck. So how much does it take? Astonishingly little . One phosphorus atom for every billion silicon atoms makes the silicon about three and a half thousand times more conductive . One in a billion is like one second in thirty-two years. One phosphorus atom per million makes it about two and a half million times more conductive . That's one second in eleven and a half days. Even the most heavily doped silicon still conducts hundreds of times worse than copper . Here's why that works so well. At one part per billion, the electrons from phosphorus outnumber silicon's own heat-made electrons about five thousand to one . At one part per million, it's millions to one . And the number of dopant carriers is set by how many dopant atoms you put in, not by the temperature . So an engineer can dial in the conductivity. That's also why the purity from last episode had to come first . You can't add exactly one atom in a billion if unknown impurities are already there at that level . Doping does something sneakier, too . Add lots of free electrons, and they keep falling into the few holes, so the holes almost disappear . In silicon with one phosphorus atom per million, there are about fifty million billion free electrons per cubic centimetre, and only about two thousand holes . Those few are called minority carriers, and the plentiful ones are majority carriers . Keep the minority carriers in mind. Next episode, they turn out to matter a lot . And if you add both kinds? Boron and phosphorus cancel each other out . What counts is the difference between the two, and whichever there's more of decides whether the silicon is n-type or p-type . That turns out to be very useful. Add enough of the opposite kind to one patch of silicon, and you can flip just that patch from n-type to p-type . Doping only works within a temperature window . Too cold, and the dopant atoms start holding on to their spare electrons, so the carriers drop away. For lightly doped silicon, that's below about minus 170 degrees . Too hot, and silicon's own heat-made carriers climb until they swamp the dopants, and the careful design is gone . For lightly doped silicon, that happens at around 280 degrees . For germanium doped the same way, it's only about 120 degrees . That's last episode's germanium problem again, seen from the other side. So how do you actually put the atoms in ? The simplest way is to add the dopant to the molten silicon in the crucible, so the whole crystal grows already doped . In 1950, Bell Labs used this trick to grow junctions in germanium, dropping pellets of impurity into the melt while the crystal was growing . One catch: the dopant prefers to stay in the liquid, so the melt gets richer as the crystal grows, and the doping creeps up along the ingot . To dope just a thin layer near the surface, you need something more precise. From 1952, Calvin Fuller at Bell Labs did it by baking crystals in hot gas carrying the dopant . The time and the temperature controlled how much went in and how deep . Diffusion furnaces use gases such as phosphine and diborane at around a thousand degrees or more, with a layer of oxide masking off the areas that shouldn't be doped . That's the silicon dioxide skin from last episode, earning its keep . Today, the standard method is ion implantation . You turn the dopant into charged ions, accelerate them, and fire them into the wafer . Because each ion carries a charge, you can count very precisely how many went in by measuring the charge the beam delivers . Then the wafer is heated to around a thousand degrees, so the dopant atoms settle into proper places in the crystal . Russell Ohl was already trying this not long after the war: first driving phosphorus into silicon with eight hundred volts, then bombarding it with ions . His equipment was so heavy he worried it would fall through the floor onto a friend below, so he was moved to the basement . So go back to Ohl's cracked rod . It was p-type on one side of the crack and n-type on the other . Where the two met, light made a voltage, and current flowed far more easily one way than the other . What's happening at that boundary? That's the junction, and it's the heart of the diode, and then the transistor. That's next time.