Once upon a time, if you wanted to make a junction transistor, you could start with a small seed crystal of pure germanium. Using Czochralski's 1917 methods, the crystal was gradually rotated and pulled out of a lightly doped N-type material and grown into a small boule. After suitable growth, a small amount of acceptor impurity such as gallium was added to the melt, causing the germanium to form a PN junction. This was all consistent with the theory of William Shockley, published in June 1949.
The making of simple PN junctions was not that new. Then, Morgan Sparks of Bell Labs added a stronger impurity (antimony) to make a second closely spaced junction and keep the crystal growing. On April 12, 1950, the first junction (NPN) transistor was born, and all hell broke loose. Hundreds of inventions were added to make better transistors.
But this grown-junction transistor was hard to make, as the base area was barely 25 µm thick and hard to connect to. Many efforts, both theoretical and practical, went into making practical transistors. Yet for a long time, the performance was still lousy, and the yields were still poor—a ß of 40, f ( α) of 15 MHz, and breakdown of 25 V was considered amazing.
SOME WINTER READING
Bo Lojek, a research scientist at Atmel's Colorado Springs facility,
chronicles the amazing stories of all kinds of brilliant research in
his new book, History of Semiconductor Engineering. He also
documents all of the human foibles that mark the industry's
milestones.
Shockley, the "Fairchild Eight," and Texas Instruments are just some of the major players Bo describes. Jean Hoerni's planar process was a major advance, forcing the junctions to be made under a silicon-dioxide layer and greatly improving reliability under severe temperature conditions.
Dozens of minor companies made little advances. Some of these advances were lost forever, mere speedbumps on the way to progress. Dozens of engineers and scientists contributed ideas of varying degrees of helpfulness or uselessness.
The diffusion of gaseous impurities into a silicon wafer, masked by wax (or later by silicon dioxide), was a major factor in these improvements. Bo even includes the diffusion formulae on how you could make your own transistors, just as the original Fairchild researchers did. Planar process? Do it yourself!
Unfortunately, this book neglects to mention the pair of small, tasteful monuments to the contributions of Bob Widlar and Jean Hoerni at the foot of San Gabriel Court in Sunnyvale, right off Kifer Road, near Maxim's headquarters. I think Bo was too modest to mention that he had instigated the building of these monuments to the pioneers of our industry.
The bold Robert J. Widlar gets his own chapter and a half, showing how the "champion's" approach to pioneering improved (linear) circuits at Fairchild and later at National. His collaboration with Dave Talbert, who optimized the diffusion processes (in conjunction with Widlar's needs), is well documented.
When you're a pioneer, you may have to work extra hard to avoid arrows in your back and prove your ideas will really work. Widlar did that.
The book includes many drawings of classic inventions and photos of great people (and sheep), illustrating the stories. It also describes the business practices and human foibles that led to success (or failure), as well as the full panoply of human brilliance and stupidity.
PICK UP A COPY
Where can you get your own History of
Semiconductor Engineering? Go to www.amazon.com and just
punch in Bo Lojek. There are some good comments and reviews
in there. Or, go to ED
Online 14727.
And in the interests of full disclosure, I helped proofread his book and offered many corrections because I was entranced by his storytelling. Bo has a full 6-ft file on Shockley and many dozens of feet of files on the other players. He distilled the best stories down into this 380-page book.
Comments invited! [email protected] —or:
Mail Stop D2597A, National Semiconductor
P.O. Box 58090, Santa Clara, CA 95052-8090