Zone Refining

Three parts, and the hard one is the temperature gradient rather than the temperature. The charge: use tin, or better a tin-lead or bismuth-tin alloy so there is a measurable impurity to sweep. Pure tin melts at 232 degrees Celsius, which is reachable with an ordinary soldering-iron element. Do not start with silicon. Silicon melts at 1414 degrees, needs an inert atmosphere and a crucible that will not contaminate it, and none of that is bench work — the physics you are about to measure is identical in tin and safe to handle. The boat: a length of steel or ceramic channel section, 300 mm long and 15 mm wide, that the charge sits in as a bar. Steel is fine for tin. Line it with graphite foil if you want a clean release. The bar must be a bar of even cross-section, because the maths you will do assumes a constant area — cast it in the boat itself by melting the charge and letting it freeze slowly end to end. The heater: a narrow band, and narrow is the whole point. A 25 mm wide zone on a 300 mm bar gives you twelve zone lengths of travel, and the purification scales with how many zone lengths you sweep. Wrap a resistance heating element — nichrome ribbon, or a cartridge heater in a machined block — so it heats a band no wider than 25 mm, and fit a heat shield either side. A wide, vague heated region is the commonest failure: it melts a long stretch, the sharp freezing interface disappears, and with it the segregation. Cooling matters as much as heating. The bar must be solid a few millimetres either side of the zone, so run a small fan or a cooled aluminium block along the boat. What you are building is not a heater, it is a steep temperature gradient that happens to peak above the melting point. The drive: the zone must travel slowly and steadily. 1 to 10 millimetres per minute is the working range and slower is better. A 12 V geared motor turning an M8 threaded rod through a nut on the heater carriage gives about 1.25 mm per revolution — at 4 rpm that is 5 mm per minute. Add a PWM controller so you can change the rate and measure its effect. Speed is a real trade, not a detail. Too fast and the impurities cannot diffuse out of the freezing interface quickly enough; they get trapped in the solid, the effective k rises towards 1, and the sweep weakens. Too slow and a pass takes all afternoon.

Cast the bar first, and cast it deliberately. Melt the whole charge in the boat and let it freeze from one end so you begin with a bar of even composition and known cross-section. Cut a 5 mm sample from each end and keep them; they are your before. Now run a pass. Bring the heater to about 40 degrees above the melting point of your alloy and let it establish. You want a molten band you can see, 20 to 25 mm long, with solid metal either side of it. Start the drive at 2 mm per minute and let it travel the full length. A 300 mm bar takes about two and a half hours. Watch the zone rather than the clock. It should be a clearly bounded pool with a visible meniscus, travelling steadily. Two things go wrong and both are visible. If the molten region keeps growing you are putting in more heat than the cooling either side can remove, and you will end up with half a molten bar and no interface. If the zone breaks into two pools you have a cold spot, usually a thin patch in the bar where it has necked. Run at least six passes, all in the same direction. Direction matters: the whole point is to sweep impurities towards one end and keep them there. A pass in the opposite direction brings them back. Expect the bar to change appearance. The head end usually becomes visibly brighter and more uniform. The tail often develops a dull, coarse-grained or discoloured section in the last zone length or so — that is where all the impurity has gone, and in industry that end is simply sawn off and remelted. Now measure. Cut a fresh 5 mm sample from each end. For each of your four samples — head before, tail before, head after, tail after — melt it in a small crucible with a thermocouple in the melt, then let it cool slowly while logging temperature every second. Plot the cooling curve. A pure metal shows a flat arrest at its freezing point: the temperature stops falling while latent heat comes out, then resumes. Impurity does two things to that arrest. It lowers the temperature at which freezing begins, by an amount proportional to the impurity concentration for dilute solutions, and it smears the flat plateau into a slope because the freezing happens over a range instead of at a point. So the comparison you want is simple and it is quantitative. The head sample after refining should show a HIGHER and FLATTER arrest than it did before. The tail should show a lower and more sloped one. The difference between head and tail after six passes is your purification, measured on equipment you built, and it is the same measurement Pfann was making in 1952 before anyone had a mass spectrometer good enough to check him.

WHY DO IMPURITIES PREFER THE LIQUID AT ALL? Because a crystal is a specific arrangement with specific spacings, and a foreign atom of the wrong size or valence costs energy to fit into it. A liquid has no long-range order and accommodates almost anything. The system minimises its free energy by leaving the misfit in the phase that does not care. Impurities that happen to fit the lattice well have a k near 1 — which is precisely the case for boron in silicon, where the boron atom is a reasonable substitute for a silicon atom. WHY IS OXYGEN'S k GREATER THAN 1? It means oxygen prefers the SOLID. Silicon dissolves oxygen readily into interstitial sites, so zone refining concentrates oxygen at the head rather than sweeping it to the tail. This is not a curiosity: it is one reason float-zone silicon, grown without a quartz crucible, has far lower oxygen content than Czochralski silicon, and why float-zone material is chosen for high-voltage power devices where oxygen-related defects matter. WHY NOT JUST MAKE THE ZONE VERY SHORT? Purification scales with the number of zone lengths swept, so a shorter zone on the same bar means more zone lengths and better cleaning. In practice the zone must be long enough to stay stable — surface tension has to hold it in float-zone work, and in a boat it has to survive small variations in bar cross-section. Below a few millimetres it breaks up. WHY DOES THE TAIL GET THROWN AWAY RATHER THAN REPROCESSED? It is reprocessed, but not by zone refining. The tail carries a concentrated mixture of everything the bar contained, and running it back through the same process just moves the problem. It goes back to the chemical purification stage upstream. WHAT ABOUT THE k VALUES CHANGING? They do. The segregation coefficient is a thermodynamic quantity that depends on temperature and, at high concentrations, on the concentration itself. The tabulated values assume dilute solution near the melting point, which is exactly the regime that matters here, but a heavily contaminated feedstock will not follow them. COULD YOU ZONE-REFINE SOMETHING OTHER THAN METALS? Yes, and it is done. Organic chemists zone-refine high-purity organics — anthracene, naphthalene, benzoic acid — for scintillators and reference standards, using exactly the same apparatus at much lower temperatures. If your workshop cannot reach 232 degrees safely, naphthalene at 80 degrees demonstrates the identical physics and the cooling-curve measurement works just as well. WHAT DID THIS ACTUALLY UNLOCK? Pfann's process is the reason the transistor left the laboratory. The 1947 point-contact device worked on germanium that was pure by the standards of the day and wildly variable device to device, because nobody could control what was in the crystal. Zone refining made the material reproducible, and reproducibility is what turns a physics demonstration into an industry. Everything in this batch that follows assumes it.