Zone Refining
Zone refining is a purification method in which a narrow molten zone is passed slowly along a solid bar of material, sweeping most impurities toward one end. It was invented by William G. Pfann at Bell Labs, who developed the idea in 1951 and patented it in 1952, and it gave the transistor era its purest germanium and silicon. The method works with no solvents or chemical reactions, only heat and the simple fact that impurities prefer to stay in the liquid when a melt freezes.
The physics rests on the segregation coefficient, the ratio of an impurity's concentration in the freezing solid to its concentration in the liquid it freezes from. For most impurities in germanium and silicon this ratio, called k, is below one: the crystal rejects them, so the solid that freezes behind the moving zone is cleaner than the liquid, and the liquid grows richer in impurity as the zone travels. Iron in silicon has k of order 10^-5 and copper about 10^-4, so they are swept away efficiently; phosphorus is moderate at about 0.35, and boron, with k near 0.8, stays and is hard to remove this way. An impurity with k greater than one, which prefers the solid, is driven toward the starting end instead.
Multiple passes are what turn a modest improvement into extraordinary purity. One pass of the zone helps, but repeating it, and eventually running several molten zones along the bar at once, concentrates impurities at the end, until the cleanest section reaches levels of a few parts per billion or better; for germanium, repeated zone refining reached impurity levels around one part in ten billion. That was dramatically better than chemical purification could achieve and made it possible to detect, and then deliberately control, the dopants that give germanium and silicon their electrical behavior. Pfann also showed the opposite trick, zone leveling, in which a zone is passed once to spread a deliberately added dopant evenly, and he described the whole field in his 1958 book Zone Melting.
The germanium of the 1950s transistors depended on this method, and silicon followed through the float-zone variant. In float-zone refining a polycrystalline silicon rod is held vertically and an induction coil melts a short zone that is held in place by surface tension, so the melt never touches a crucible and cannot pick up contamination from one. Henry Theuerer at Bell Labs developed float-zone silicon in the early 1950s; the technique both purifies the rod and, with a seed attached, can convert it into a single crystal. Because silicon's molten state would otherwise dissolve a container, this containerless melt was a significant practical advance, and the resulting material has the very low oxygen and very high resistivity needed in power devices, radiation detectors and some premium solar cells.
Zone refining is no longer the front-line purifier for the silicon in typical chips, a common point of confusion. The bulk of electronic-grade polysilicon is made chemically, by distilling and decomposing trichlorosilane in the Siemens process, which reaches about nine nines purity or better at industrial scale and tonnage that zone refining cannot match. Zone refining remains where the ultimate purity or special materials are needed: ultra-pure germanium for gamma-ray detectors, high-purity metals, organic compounds, and the float-zone silicon described above. The limitation is also physical, since impurities such as boron that segregate weakly are barely moved by the zone, so chemistry has to handle them upstream.
Read zone refining as purification by freezing: because most impurities favor the liquid, a molten zone driven along a bar acts as a broom that sweeps them to one end, and repeated passes produced the parts-per-billion germanium and silicon that let Bell Labs, and later the whole industry, trust that a transistor's behavior came from its dopants and not from stray atoms.