There was no doubt about it, point-contact transistors were fidgety. The transistors being made by Bell just didn't work the same way twice, and on top of that, they were noisy. While one lab at Bell was trying to improve those first type-A transistors, William Shockley was working on a whole different design that would eventually get rid of these problems.
Early in 1948, Shockley conceived of a transistor that looked like a sandwich, with two layers of one type of semiconductor surrounding a second kind. This was a completely different setup which didn't have the shaky wires that made the point-contact transistors so hard to control.
A working sandwich transistor would require that electricity travel straight across a crystal instead of around the surface. But Bardeen's theory about how the point-contact transistor worked said that electricity could only travel around the outside of a semiconductor crystal. In February of 1948, some tentative results in the Shockley lab suggested this might not be true. So the first thing Shockley had to do was determine just what was going on.
Careful experiments led by a physicist in the group, Richard Haynes, helped. Haynes put electrodes on both sides of a thin germanium crystal and took very sensitive measurements of the size and speed of the current. Electricity definitely flowed straight through the crystal. That meant Shockley's vision of a new kind of transistor was theoretically possible.
But Haynes also discovered that the layer in the middle of the sandwich had to be very thin and very pure.
The man who paved the way for growing the best crystals was Gordon Teal. He didn't work in Shockley's group, but he kept tabs on what was going on. He'd even been asked to provide crystals for the Solid State team upon occasion. Teal thought transistors should be built from a single crystal-as opposed to cutting a sliver from a larger ingot of many crystals. The boundaries between all the little crystals caused ruts that scattered the current, and Teal had heard of a way to build a large single crystal which wouldn't have all those crags. The method was to take a tiny seed crystal and dip it into the melted germanium. This was then pulled out ever so slowly, as a crystal formed like an icicle below the seed.
Teal knew how to do it, but no one was interested. A number of institutions at the time, Bell included, had a bad habit of not trusting techniques that hadn't been devised at home. Shockley didn't think these single crystals were necessary at all. Jack Morton, head of the transistor-production group, said Teal should go ahead with the research, but didn't throw much support his way.
Luckily, Teal did continue the research, working with engineer John Little. Three months later, in March of 1949, Shockley had to admit he'd been wrong. Current flowing across Teal's semiconductors could last up to one hundred times longer than it had in the old cut crystals.
Nice crystals are all well and good, but a sandwich transistor needed a sandwich crystal. The outer layers had to be a semiconductor with either too many electrons (known as N-type) or too few (known as P-type), while the inner layer was the opposite. Under Shockley's prodding, Teal and Morgan Sparks began adding impurities to the melt while they pulled the crystal out of the melt. Adding impurities is known as "doping," and it's how one turns a semiconductor into N- or P-type.
As they pulled the seed crystal out of an N-type germanium melt, they quickly added some gallium to turn the melt into P-type. As a layer of P-type formed on the ever-lengthening crystal, they added antimony, which compensated for the gallium and turned the melt back into N-type. Once the process was done, there was a single, thin crystal formed into a perfect sandwich.
By etching away the surface of the outside layers, Sparks and Teal left a tiny bit of P-type crystal protruding. To this they attached a fine electrode-creating a circuit the way Shockley had envisioned. On April 12, 1950, they tested what they had built. Without a doubt, more current came out of the sandwich than went in. It was a working amplifier.
The first junction transistor had been born. But It Wasn't a Very Good One . . . Yet
This transistor could amplify electrical signals, but not particularly complicated ones. If the signal changed rapidly, as a voice coming over a phone line does, the transistor couldn't keep up and would garble the output. The problem lay in the middle of the sandwich: it was too easy for electric current to spread out and become unfocused as it crossed the P-type layer. To solve the problem, the layer had to be even thinner.
In January of 1951, Morgan Sparks figured out a way to accomplish that. By pulling the crystal out more slowly than ever, while constantly stirring the melt, he managed to get the middle layer of the sandwich thinner than a sheet of paper.
This new, improved sandwich did all that the researchers hoped. They still weren't up to the point-contact transistor's ability to handle signals that fluctuated extremely rapidly, but in every other way they were superior. They were much more efficient, used very little power to work, and they were so much quieter that they could handle weaker signals than the type-A transistors ever could.
In July of 1951, Bell held another press conference -- this time announcing the invention of a working and efficient junction transistor.
