Zone Refining

Part of Semiconductors

Purifying semiconductor material to electronic grade by passing a narrow molten zone repeatedly along a rod.

Why This Matters

Zone refining, developed by William Pfann at Bell Labs in 1952, is the technique that made the transistor revolution possible. It can purify germanium and silicon to levels of contamination below one part per billion — pure enough that a few added impurity atoms per million silicon atoms (doping) controls conductivity, rather than random contamination. Without this level of purity, transistors cannot be reliably fabricated.

The principle is elegant: most impurities are more soluble in the liquid phase of a semiconductor than in the solid phase. A narrow molten zone sweeping along a rod carries impurities with it. After many passes, impurities concentrate at one end; the other end is very pure. No exotic chemistry required — just controlled heating and cooling.

For a rebuilding civilization, zone refining is a pivotal technology. It does not require complex chemical plants. The apparatus is a tube furnace or RF induction heater with a mechanism to move the heater along a rod. Build this, and you can make semiconductor-grade germanium within weeks of recovering crude metal. The quality of everything downstream — every transistor, every diode, every semiconductor device — depends on how well this step is done.

The Segregation Coefficient

Zone refining works because of segregation: when a liquid freezes, the solid has a different composition than the liquid. The segregation coefficient k is:

k = Cs / CL

where Cs is the impurity concentration in the newly frozen solid, and CL is the impurity concentration in the liquid at the same moment.

k < 1: Impurity prefers to stay in liquid. Most common case for metallic impurities in germanium and silicon. Example: copper in germanium, k ≈ 10^-5 (copper strongly prefers liquid — excellent segregation). Iron in germanium, k ≈ 10^-5 to 10^-4. Aluminum in silicon, k ≈ 0.002.

k > 1: Impurity prefers solid. Less common. Example: boron in silicon, k ≈ 0.8 (boron barely segregates — zone refining works poorly for boron). This is why the Siemens trichlorosilane process is needed to remove boron from silicon.

k = 1: No segregation, zone refining useless.

After N passes of zone refining, the impurity concentration at the beginning of the rod (the first part to solidify each pass) is approximately:

C_final / C_initial ≈ k × e^(-k × L/l × N)

where L is rod length and l is zone length. With k = 0.001, L/l = 10 (rod is 10× the zone length), and N = 10 passes: C_final/C_initial ≈ 10^-6. A million-fold purification.

Apparatus Design

The rod: Cast or pressed crude germanium into a long rod. Typical dimensions: 1-2 cm diameter, 20-50 cm long. Rod must be straight (bow causes uneven zone thickness). Support the rod in a horizontal quartz or alumina cradle — the material must not react with molten germanium.

The heating zone: Options in order of accessibility:

• Resistance heater: a wound coil of nichrome or kanthal wire, 2-3 cm wide, surrounding the rod. Power from variable transformer. Zone width 2-4 cm. Adequate for 20-30°C/cm temperature gradient at the zone edges.
• RF induction heater: a copper coil connected to an RF power source (1-5 MHz, 1-5 kW). Induces eddy currents in the molten zone, heating it from within. Better zone definition than resistance heater; no contact with material. More complex to build.

Zone travel: The heater must move along the rod at 1-5 cm/hour. Slow movement ensures the zone is fully liquid and good segregation occurs. Fast movement leaves impurities behind at their original locations. Use a threaded rod driven by a low-RPM motor (clock motor, gearbox-reduced motor) to move the heater carriage at controlled rate.

Atmosphere control: Germanium oxidizes slowly in air at elevated temperature. Preferably zone refine under nitrogen or argon atmosphere. Simplest approach: run a steady flow of nitrogen over the rod in a partially enclosed tube, with small openings for the heater to project through. Not perfectly sealed, but reduces oxidation significantly.

Number of passes: 10-20 passes of the zone from one end to the other. After all passes, cut off the last 15-20% of the rod (impurity-rich end) and remelt it as feed for the next batch.

Multi-zone operation: Pass multiple heaters simultaneously, each a zone width apart. Three zones in parallel do the work of three passes in one mechanical pass. More complex mechanically but faster.

The Float Zone Method for Silicon

For silicon (melting point 1414°C, too high for quartz crucibles to hold without contaminating the melt), the float zone method avoids container contact:

A polysilicon rod is held vertically, and an RF induction coil heats a short section until it melts. Surface tension holds the liquid zone in place between the solid rod sections above and below. The zone moves along the rod as the coil moves. No crucible — silicon purity is not limited by container contamination.

Float zone silicon is the purest silicon produced and is used for detectors and other high-purity applications. The apparatus requires precise RF power control (the liquid zone must be stable — too much power causes the zone to sag and the rod to separate) and careful mechanical design.

For a rebuilding civilization, float zone silicon is an intermediate-term goal. Start with germanium zone refining (simpler thermal requirements, lower temperature, crucible acceptable), establish the process, then apply the lessons to silicon.

Process Control and Quality Verification

Temperature profile: The zone must be fully liquid at the center and solid at both edges. A temperature difference of 50-100°C between zone center (above melting point) and zone edge is adequate. Too-narrow gradient: fuzzy zone boundary, poor segregation. Too-steep gradient: thermal stress fractures the rod on cooling.

Rod support: Horizontal rods sag as temperature increases. Support the rod every 5-10 cm with alumina or quartz saddles. Saddles must be smooth to allow thermal expansion without cracking the rod.

Contamination from apparatus: Quartz cradles and tubes are SiO2 — they introduce silicon into the melt (small amount) and can absorb impurities from the melt during high-temperature processing. Alumina (Al2O3) is preferred for germanium zone refining; it is more chemically stable at operating temperatures.

Measuring progress: After each set of 5 passes, cut a thin slice from the “good” end of the rod. Measure resistivity with four-point probe. Compare to target. If resistivity is still low (high carrier concentration from impurities), continue zone refining. If target resistivity is reached, stop.

For germanium intended for transistors: target resistivity 5-50 Ω·cm (lightly doped). Very pure, undoped germanium would be ~50 Ω·cm at room temperature. Actual refining goal depends on intended final doping.

Lifetime measurement: After zone refining, measure minority carrier lifetime. Photoconductance decay (illuminate sample briefly with LED flash, measure decay of resistance) gives lifetime. Target: > 100 µs for high-gain transistors. Values below 10 µs indicate metallic contamination not removed by zone refining — trace back to the crude material source or the apparatus materials.

Managing the Impurity-Rich End

The last 15-20% of the rod accumulates impurities. Options:

• Recycle: Remelt and recast into a new rod. Zone refine again. Each cycle further purifies the main batch and concentrates impurities into a shrinking end fraction.
• Use as-is: Heavily doped (from impurities, often n or p type). Test its type and concentration. It may be usable as a doping source or for low-grade applications.
• Discard: If contamination is too severe for recycling to be efficient, set aside as starting material for a later more aggressive purification run.

Keep a mass balance: weigh input crude material, weigh purified rod plus impurity-rich cut. Account for any losses to oxidation or spills. Efficiency losses above 10% indicate process improvement is needed.

Zone refining is patient work — a full run takes days — but the reward is material that no amount of chemical processing can match. A well-run zone refinement program is the foundation of a serious semiconductor capability.