On this page
A transistor only works if the semiconductor inside it is extremely pure, with impurities then added back on purpose in tiny, controlled amounts. In the early 1950s that first step, getting germanium or silicon clean enough, was a real obstacle. William G. Pfann, a metallurgist at Bell Telephone Laboratories, solved it with a simple idea: melt only a thin slice of the bar and move that slice along. This post explains how zone refining works, runs the numbers from Pfann's patent, and draws a lesson from his approach.
Who was William Pfann?
Pfann lived from 1917 to 1982, as the title of his MRS Bulletin obituary records. He worked at Bell Telephone Laboratories, the assignee of his zone-melting patent, US 2,739,088. The patent was filed on 16 November 1951 and granted on 20 March 1956. It refers back to earlier applications of his, filed in 1948 and 1950, on fusing metal electrodes into germanium, so he was already working on germanium devices.
He set out the theory in a paper called Principles of Zone-Melting, published in the Journal of Metals in July 1952.
Why do semiconductors need to be so pure?
NCERT's Class 12 Physics chapter on semiconductor electronics explains that adding a few parts per million (ppm) of a suitable impurity to a pure semiconductor changes its conductivity a great deal. That deliberate addition is doping. Now think about the reverse. If the starting crystal already holds unknown impurities at a similar level, you cannot control what the doping does.
Pfann's patent states the aim directly: the process can be adapted to refining semiconductor materials such as germanium and silicon "so as to produce a degree of purity heretofore unobtainable".
How does a moving molten zone remove impurities?
The key quantity is what the patent calls the distribution coefficient, defined as the ratio of the solute concentration in the freezing solid to the concentration in the liquid. Today it is usually written k = Cs/Cl. The patent uses the Greek letter gamma.
When k is less than 1, the solid that freezes out is purer than the liquid it freezes from, so the impurity stays behind in the melt. If you freeze a whole molten bar slowly from one end, the impurity piles up in the last part to freeze. The patent calls this normal freezing and cites earlier metallurgy work by Hayes and Chipman for its equation.
Pfann's step was to melt only a short zone. In his apparatus a graphite ring heater, heated by induction, made molten zones "as short as one-half inch" in germanium, inside a quartz tube with a hydrogen atmosphere. As the bar moves through the heater, the zone melts fresh solid at its front edge and freezes purer solid at its back edge, so impurity is carried along to the far end. That end is cut off and discarded. The patent also describes multiple-pass refining, with several heaters or repeated passes, because one pass is not enough.
Worked example: how much purer is the first slice?
The patent lists measured values in germanium: about 0.12 for gallium, about 0.07 for aluminium and about 0.07 for arsenic. Take gallium, k = 0.12. The bar below uses sample numbers for illustration: a uniform starting concentration C0 = 1.0 ppm of gallium.
The first solid to freeze has Cs = k × C0 = 0.12 × 1.0 = 0.12 ppm. That slice is about 8 times purer than the starting bar.
The impurity that did not freeze into the solid stays in the zone, so the liquid gets richer as the zone moves.
The patent explains that the liquid keeps enriching until it reaches C0/k = 1.0/0.12 ≈ 8.3 ppm. At that point the solid freezing out has concentration k × 8.3 ≈ 1.0 ppm, the same as the solid being melted, and nothing changes any more.
The last zone-length of the bar freezes by normal freezing and ends up with most of the gallium.
So a single pass cleans only the starting end. The middle of the bar stays at about C0. Repeated passes extend the clean region, which is why Pfann built multipass equipment.
Impurities with k greater than 1 move the other way. The patent notes that boron in germanium has a value of about 5, so boron collects near the starting end instead.
What mistakes do students make with this idea?
Assuming one pass purifies the whole bar. Step 3 shows the middle reaches a steady state at the original concentration.
Writing k upside down. It is solid over liquid. With k = 0.12 the solid is purer; if you invert it you will predict the opposite direction.
Forgetting impurities with k above 1. They are pushed toward the start, not the end.
If ratios like Cs/Cl or ppm conversions feel shaky, the Fermi Concept Library covers every concept in physics, chemistry, biology, mathematics and computing, so you can revise the one idea that is blocking you.
Why did zone melting matter beyond germanium?
The patent points out another advantage: zone melting can be used "in cases where no suitable mold or crucible exists for the material being purified", because the solid parts of the bar hold up the molten part. In his 1962 review in Science, Zone Melting, Pfann wrote that there was "no basic reason why tons rather than pounds cannot be treated", and that continuous refiners had been designed and operated.
To see where pure, doped crystals end up, read our posts on semiconductors and logic gates and on John Bardeen and the transistor.
What can a student learn from how Pfann worked?
Segregation of impurities during freezing was already described in metallurgy papers, such as the 1939 Hayes and Chipman work his patent cites. Pfann turned that effect into a purification tool. He started from an existing equation, defined one clear quantity (the distribution coefficient), measured it for the impurities that mattered, and only then designed simple apparatus around it. When something in an experiment keeps "going wrong" in the same direction, it is worth asking whether that pattern can be put to work.
Redo the worked example for aluminium with k = 0.07 and find the first-slice concentration and the steady-state zone concentration.
