Why can sand think?

Here's something I never understood . The chip in your phone is a sliver of silicon. And silicon, joined to oxygen, is silica, the same compound as quartz and sand . Rock, basically. So how does rock end up doing maths? To answer that, we have to start with an older question. Why does anything conduct electricity at all? In 1833, Michael Faraday heated a lump of silver sulfide over a lamp and noticed something odd . As it got hotter, it conducted electricity better, rapidly better, and when he took the heat away, it went back . He wrote that this was in direct contrast to what heat does to metals . It was the first recorded semiconductor effect, and nobody could explain it for about a century . Start simple. An electric current is just electric charge on the move . In a metal like copper, that means electrons that are free to travel . Copper has loads of them, about one for every atom . Connect a battery, and that whole crowd of free electrons drifts along the wire. But why are they free? A copper atom has twenty-nine electrons. Twenty-eight sit in full inner shells, and the last one sits alone on the outside . When copper atoms pack together into a metal, that lone outer electron stops belonging to any one atom. It joins a shared sea of electrons that the whole metal holds in common . Glass and salt are different. There, every electron has a job, either holding one particular bond between two atoms, or held tightly by one particular atom . No spare electrons, so almost no current. To see what's really going on, physicists think about energy. A single atom's electrons can only have certain exact energies, like the rungs of a ladder . Bring two atoms close together, and each rung splits into two, a little apart . Four atoms, four rungs. In a crystal with billions of trillions of atoms, each rung becomes billions of trillions of rungs, packed so tightly they blur into a band . The forbidden ranges are simply the gaps between those bands, where no rungs ended up . Two of those bands decide everything. The lower one, the valence band, holds the electrons that are busy in bonds. The upper one, the conduction band, is where electrons are free to roam . The gap between them isn't a place. In a perfect, pure crystal, it's a range of energies that no electron is allowed to have . Think of each band as rows of seats. To carry a current, an electron has to shuffle into a slightly higher empty seat . In a metal like copper, the top band is only partly full, so there's always a free seat next door . In an insulator, the band is completely full, and the next empty row is a long way up . A semiconductor is full too, but the next row is close enough that heat or light can occasionally lift an electron across . For silicon, that gap is about 1.1 electron volts . For germanium, it's about two thirds of one . So what's an electron volt? It isn't a voltage. It's a tiny amount of energy, the energy one electron picks up when it's pushed through one volt . So silicon's gap of 1.1 electron volts doesn't mean it needs 1.1 volts across it . It means a single electron needs that much energy to jump from a bond to freedom . For scale, a typical kick from heat at room temperature is about 0.026 electron volts . Silicon's gap is around forty times bigger, which is why so few electrons make it across . Visible light carries about 1.7 to 3 electron volts per photon, plenty to free an electron . That's why silicon makes good light sensors. It handles visible light well, and it can pick up near-infrared up to a hard limit around 1,100 nanometres . That changed how I think about the word . A semiconductor isn't a material that conducts a bit. In band terms, it's really an insulator whose gap happens to be narrow . At absolute zero, pure silicon wouldn't conduct at all . What makes it useful is how dramatically that can be changed, by heat, by light and, as we'll see next time, by impurities . Here's what heat does. Every so often, a random thermal kick knocks an electron out of its bond and up into the conduction band, where it's free to move . That electron wanders off. But it leaves behind a gap in the bond, a vacancy called a hole . This happens at any temperature above absolute zero. It's just rare when it's cold . So who fills the hole? Usually not the electron that left . An electron from a neighbouring bond hops sideways into it, which leaves a new hole where that electron used to be . Then another neighbour hops in, and another. Each electron only moves one step, but the hole walks across the crystal, behaving like a positive charge . So one freed electron gives you two carriers, the electron and the hole, and both carry current . Eventually a free electron runs into a hole somewhere and drops back in. That's called recombination, and in silicon the energy it gives up mostly ends up as heat . And that's the heart of Faraday's mystery . In a metal, heat makes the atoms vibrate harder, the free electrons collide more often, and resistance creeps up . In a semiconductor, heat does that too, but it also frees new carriers, and their number grows so fast that it swamps everything else . In pure silicon near room temperature, the number of free carriers roughly doubles for every nine degrees of warming . Hot electrons don't move more easily. There are just far more of them . So what's special about silicon? Each silicon atom has four outer electrons . In a crystal, every atom shares a pair of electrons with each of four neighbours, with the bonds pointing to the corners of a tetrahedron . It's the same arrangement as diamond, because silicon sits right under carbon in the periodic table . With every bond full, a perfect silicon crystal would be an insulator, except that heat keeps breaking a few bonds . How few? At room temperature, pure silicon has about ten billion free electrons in every cubic centimetre . The same cube holds about fifty thousand billion billion atoms . That's roughly one free electron for every five trillion atoms . So pure silicon conducts more than a hundred billion times worse than copper . It's a terrible conductor. So the first job is purity, and that's harder than it sounds . Silicon is the second most common element in the Earth's crust, more than a quarter of it by mass . But in nature it's almost never found on its own . It's almost always locked to oxygen, as silica in quartz and sand, or in the silicate minerals that make up more than nine tenths of the crust . Industry doesn't start with any old sand. It starts with quartz, or quartzite rock . Mixed with carbon and heated to as much as 2,000 degrees in an electric arc furnace, the carbon steals the oxygen, and out comes silicon that's about 98 to 99% pure . For a chip, that's filthy . So it's purified again, chemically: turned into a liquid, distilled over and over, then deposited back onto hot silicon rods. That's the Siemens process . For electronics, counting the atoms that matter electrically, it's ten to eleven nines pure . The strictest grade allows only about three stray phosphorus-type atoms for every hundred billion atoms of silicon . But pure isn't enough. That polysilicon is a jumble of tiny crystals, and early transistor makers found they needed one large single crystal instead . The method was invented in 1916 by the Polish chemist Jan Czochralski . The story goes that it began with an accident: he dipped his pen into molten tin instead of his inkwell, and pulled out a thin metal thread . Melt the purified silicon in a quartz crucible, dip in a small seed crystal, and slowly pull it up while turning it . The silicon freezes onto the seed, copying its arrangement of atoms as it grows . Today's ingots are about 300 millimetres across and take days to grow, before they're sliced into wafers . The first transistors weren't silicon at all . They were germanium, which was easier to purify . But remember that narrow gap . At room temperature, heat frees about two thousand times more carriers in germanium than in silicon . In a transistor, those heat-made carriers show up as leakage, a current that keeps flowing even when the device is meant to be off . And it gets worse as it warms up. A long-standing rule of thumb says germanium's leakage roughly doubles every ten degrees or so . Leakage adds to the current through the transistor, the extra current heats it up, and the heat makes more leakage . Left unchecked, that loop, called thermal runaway, can destroy the transistor . Germanium devices were limited to about seventy degrees, while early silicon ones kept going past a hundred and twenty . The military wanted electronics that kept working in extreme conditions, and germanium couldn't take the heat . Texas Instruments went after that market with silicon . In May 1954, Gordon Teal made the point with some theatre . He played a record through a germanium transistor amplifier, dunked the transistor in hot oil, and the music died. He did the same with silicon, and the music kept playing . Heat was the first reason to switch, but not the last . Heat silicon with oxygen or steam and it grows its own skin of glass, silicon dioxide, a tough insulator, right on its surface . The historian Philip Seidenberg calls that oxide the precipitating cause of the switch to silicon around 1960, especially in computers . Germanium's oxide, by comparison, is flimsy, and it even dissolves slowly in water . Within about a decade of Teal's demonstration, silicon had taken over . So now we have it. Silicon purified to ten or eleven nines, grown into one nearly perfect crystal, and on its own almost useless . In 1931 the physicist Wolfgang Pauli called semiconductors "a filthy mess" . The mess was impurities . And the trick that changed the world was to put them back, deliberately, a tiny and precisely chosen amount, in exactly the right places. That's next time.