From GHN(Difference between revisions)
|(One intermediate revision by one user not shown)|
|Line 173:||Line 173:|
<p>So it was all there, that’s right. But it took Jean Hoerni at Intel (the planar [[Transistors|transistor]]) and [[Robert Noyce|Bob Noyce]] of Intel and [[Jack Kilby|Jack Kilby]] of TI (the [[Integrated Circuits|integrated circuit]]) to put it all together. </p>
<p>So it was all there, that’s right. But it took Jean Hoerniat Intel (the planar [[Transistors|transistor]]) and [[Robert Noyce|Bob Noyce]] of Intel and [[Jack Kilby|Jack Kilby]] of TI (the [[Integrated Circuits|integrated circuit]]) to put it all together. </p>
|Line 285:||Line 285:|
<p>Thank you very much for the interview. </p>
<p>Thank you very much for the interview. </p>
[[Category:Components, circuits, devices & systems|Tanenbaum]] [[Category:Engineered materials & dielectrics|Tanenbaum]] [[Category:Solid state circuits|Tanenbaum]] [[Category:Transistors|Tanenbaum]] [[Category:Metals|Tanenbaum]] [[Category:Metalloids|Tanenbaum]] [[Category:
[[Category:Components, circuits, devices & systems|Tanenbaum]] [[Category:Engineered materials & dielectrics|Tanenbaum]] [[Category:Solid state circuits|Tanenbaum]] [[Category:Transistors|Tanenbaum]] [[Category:Metals|Tanenbaum]] [[Category:Metalloids|Tanenbaum]] [[Category:|Tanenbaum]] [[Category:Scientific disciplines|Tanenbaum]] [[Category:Chemistry|Tanenbaum]] [[Category:News|Tanenbaum]]
Revision as of 17:07, 19 July 2013
About Morris Tanenbaum
Tanenbaum received his Bachelors in chemistry from John Hopkins (1949) and his PhD in physical chemistry from Princeton. He began work at Bell Laboratories in 1952, in the Chemical Physics Department. He worked on the chemical composition of transistors. Ca. 1954, at William Shockley’s prompting, he began work on silicon crystals as possible transistors. With the invaluable collaboration of the technician Ernie Buehler, he made the world’s first silicon transistor in January 1955. (He notes the competing claims of Gordon Teal and Texas Instruments, but believes, and provides evidence for his belief, that his work predated theirs by several months.) The same year he invented the first diffused base silicon transistor. Then, although Bell Labs had a significant technological lead in silicon transistor technology, they stopped doing proper research in the field—partly because it just wasn’t immediately relevant to AT&T’s business—so silicon transistor technology, including the integrated circuit, was done by Intel and Texas Instruments Instead. Around 1960, Tanenbaum, Buehler, and other scientists such as Gene Kunzler, Berndt Matthias, and Rudy Kompfner, invented the first high field superconducting magnets. Tanenbaum later moved into research management, first at Bell Labs, and then elsewhere within the AT&T umbrella as head of the new Western Elrectric Engineering Research Center. He then moved into mainline telepone company managagement, serving as President of New Jersey Bell in the late 1970s and early 1980s. He returned to AT&T at the break up of the Bell System in 1984, serving as President of AT&T Communicatins. He retired in 1989 as Vice Chairman and Chief Financial Officer of AT&T.
About the Interview
Morris Tanenbaum: An Interview Conducted by Robert Colburn, IEEE History Center, 16 September 1999
Interview #363 for the IEEE History Center The Institute of Electrical and Electronics Engineering, Inc.
This manuscript is being made available for research purposes only. All literary rights in the manuscript, including the right to publish, are reserved to the IEEE History Center. No part of the manuscript may be quoted for publication without the written permission of the Director of IEEE History Center.
Request for permission to quote for publication should be addressed to the IEEE History Center Oral History Program, 39 Union Street, New Brunswick, NJ 08901-8538 USA. It should include identification of the specific passages to be quoted, anticipated use of the passages, and identification of the user.
It is recommended that this oral history be cited as follows:
Morris Tanenbaum, an oral history conducted in 1999 by Robert Colburn, IEEE History Center, New Brunswick, NJ, USA.
Interview: Morris Tanenbaum
Interviewer: Robert Colburn
Date: 16 September 1999
Place: Short Hills, New Jersey
If I could just start with taking you back through some of your early background, education, what got you interested in electrical engineering, and in particular fields of study, that sort of thing.
Education, PhD Thesis
I came to electrical engineering through a very indirect path. I was born in Huntington, West Virginia, and lived there through my high school years. I think it was probably during my junior high and high school years that I became interested primarily in chemistry. I had the usual chemistry set and took the usual paint and varnish off some of my mother’s tables, but decided I really wanted to be a chemist. When it came to college time, I looked around and applied at a number of places. I finally ended up going to Johns Hopkins, which had a long-term reputation in chemistry. And my mother, who wanted me to be a medical doctor, thought probably I might become infected there! But it didn’t work out that way. So, I got my bachelor’s degree in chemistry in 1949.
I started to look around for a job, and that was not a very good time to get jobs. It was before the Korean War and things were still winding down from World War II. One of my chemistry professors, Clark Bricker, suggested I really ought to go to graduate school anyway. He had gotten his Ph.D. at Princeton. He was an assistant professor at Johns Hopkins and was offered a position back at Princeton. He suggested that I apply to Princeton, and said he was sure he could provide some way for me to pay my costs if I were to decide to come to Princeton. So I did that, and I worked in his lab in analytical chemistry as a laboratory assistant during my first year. I also did some research work in analytical chemistry during that period that resulted in my first published paper.
By that time, I was thoroughly certain that physical chemistry was what really attracted me. I choose as my thesis advisor a physical chemist named Walter Kauzman. Kauzman was a student of Henry Eyring, the originator of absolute rate theory, which applied thermodynamics to describe the kinetics of chemical reactions.
Walter was interested in kinetic phenomena even beyond chemical processes. He suggested, as a thesis topic for me, a study of an important phenomenon that was a long standing mystery. It appeared to have very little to do with chemistry at that time. The fact was that, if you grow a very pure, single crystal of a metal, you find that it is mechanically much weaker, much more easily deformed, than the same metal in its impure, polycrystalline form. Walter had a graduate student who had built apparatus and was studying the kinetics of that phenomenon. He was just finishing his work. I took over his apparatus, improved its precision and temperature stability, developed a process for purifying and growing single crystals of zinc and began measuring their strength as a function of temperature and stress at very low levels of deformation.
Zinc has a hexagonal crystal structure and its deformation pattern is relatively simple because there is a preferred crystal plane upon which deformation prefers to occur. I found, as I analyzed my data, that very interestingly, the heat of deformation, i.e. the thermal energy dependence, was very small while the stress dependence was very large. If you applied absolute rate theory, you found that the phenomenon that was controlling the deformation had an extraordinarily high entropy factor which suggested that the deformation process depended on a highly organized phenomenon. That was at the time when people were just starting to think about a special kind of crystal defect called a dislocation. My observations were very consistent with that kind of defect as the principal cause of weakness in metal single crystals. That made a good, publishable Ph.D. thesis.
Then I started looking around for a job. I considered some postdoctoral work but, then in my last year of graduate school, I got involved with a program where graduate students would chaperone groups of undergraduates to visit scientific laboratories around the area. On one of these was Bell Laboratories. Frankly, I had never heard of Bell Laboratories. When I thought about jobs, I thought about Dupont, Exxon, Eastman Kodak, and Monsanto--- places like that, and in fact I had done some interviewing at those places. Well, I was absolutely impressed by Bell Laboratories. I came back and talked to Walter and asked him whether that was a reasonable place to work. He said it was a terrific place to work. So I let it be known that I was interested, and was interviewed. Princeton had a number of graduates who had gone to work at Bell Labs. They offered me a job in 1952. It was just about four years after the invention of the transistor at Bell Labs and solid state science and electronics were just beginning to flower.
When I got to Bell Labs, I ran into a great piece of good fortune. It was suggested to me that I simply look around in the research department, see what people were doing, and if I found an area that I’d like to work in and if they were interested in having me work there, that would be just fine. But if I didn’t and had some ideas of my own, we’d sit down and talk about them. If they seemed to have some reasonable connection to the business of the Bell System, that would be just fine, too. I did a lot of looking around, and was finally attracted to the Chemical Physics Department. At that time, the work on semiconductors had focused on germanium which was the only material from which transistors had been made. One of the research managers, Joe Burton, was interested in looking at other elements and compounds to see if we could make any generalizations about what kinds of compounds would be semiconducting and how the chemical composition would determine their properties such as their energy gap and the mobility of electronic carriers, electron and holes, which were critical properties for transistor action. That intrigued me so I began collecting pure elements wherever I could buy them and building apparatus to zone refine and purify them and to grow single crystals. Then I planned to characterize their electronic properties. Just as I was getting that going, there was a report from Siemens Laboratories in Germany. A scientist there, Welker, had discovered that the compounds between the elements in Group III of the periodic table and Group V formed tetrahedral crystal structures similar to that of germanium which is in Group IV and appeared to have interesting electronic properties. In particular, indium antimonide, InSb, appeared to have a very high electron mobility. Welker’s compounds were not very pure and were polycrystalline. I set about to try to make some of those materials and grow pure single crystals. Very soon, we were able to make large single crystals of InSb and measure its electronic and optical properties using the Hall effect, magneto resistance and infrared transmission. We discovered that the mobility of electrons in InSb was extraordinarily high, much higher than that measured by Welker, but that the energy gap was quite low, too low to be of interest in transistors that would work at room temperature. We also grew pure single crystals of gallium antimonide, GaSb. Its electron mobility was somewhat higher than germanium but its energy gap was not large enough to give it any advantage over germanium. At that time I was joined by another physical chemist who had just gotten his Ph.D. at Berkeley and he started looking at the arsenides. We had a pretty good thing going.
After I had been in Joe Burton’s group maybe eighteen months to two years, I was approached by Bill Shockley, who said (approximately), “Look, we’ve got germanium. We can make good transistors with it. But we’re concerned about the fact that it has a relatively low energy gap.” That would be a problem for many of our telephone applications where our switching offices are not air conditioned (as they were not at that time). There was also a lot of surface sensitivity with germanium, so the transistors had to be encapsulated almost as carefully as you would with a vacuum tube and that was a major disadvantage. There had been some attempts to make transistors using elemental silicon which was known to have a higher energy gap and was naturally covered with a relatively impervious thin layer of silicon dioxide. Those efforts had been unsuccessful primarily because of what appeared to be a very low lifetime of minority electronic carriers. Investigators had been able to make grown-junction transistor structures, and they could observe transistor effects, but the alphas, ( a measure related to the amplification factor) were too low, and no one had ever been able to demonstrate power gain in a silicon transistor-like structure. Shockley asked whether I would be interested in coming over to his group to see if we could make useful transistors out of silicon and determine if they were superior to germanium.
I assume this was based on the expectation that the energy gap would be greater.
Had anybody actually seen evidence for it, or at this point was it purely theoretical?
No, there was pretty clear evidence that the energy gap was larger, both optical evidence as well as measurements of conductivity as a function of temperature. The energy gap was known to be about 1.1 electron volts compared to 0.7 for germanium, so it was perfectly clear that silicon would be a preferable material from that point of view. However, silicon had a high melting point—much higher than germanium. Some single crystals had been grown of silicon, but it was hard to find a crucible material to contain the melt while you pulled a crystal by the Czochralski technique. Molten silicon tended to react with just about any high melting point material we tried. So there were a lot of difficult chemical as well as physical problems to overcome. It was also hard to make electrical contact to silicon. As I mentioned, silicon reacts with atmospheric oxygen and covers itself with a thin film of silicon dioxide which is an excellent insulator, so you had to get through the oxide before you could make electrical contact to the elemental silicon. In addition, the electron and hole mobilities were know to be lower than germanium. That was unattractive, but they were still high enough to be interesting if you could make small enough structures. And, as I mentioned above there was the low carrier lifetime problem which would also require very small, precisely controlled structures. Those were the challenges.
At that time, Dupont had gotten interested in silicon as a potentially commercial electronic material and they were making some fairly pure polycrystalline silicon. I worked with a super technician at Bell Labs, Ernie Buehler, who was a master craftsman in building apparatus and growing semiconductor crystals. He used radio frequency heating so that you could melt very high temperature melting point materials in inert atmospheres without introducing impurities as you might with resistance furnaces. Ernie and I started growing silicon crystals using high purity silicon dioxide (quartz) crucibles to contain the molten silicon. During the growth process, tiny amount of doping elements (e.g. phosphorus, aluminum, etc.) were added to produce n-type and p-type regions to make npn or pnp transistor structures. This was the same technique used to make the first germanium grown junction transistors.
We made a lot of transistor structures, but they had all the same inadequate properties. You just could not get a high enough alpha, a high enough transfer of minority carriers through the base region. This was believed to be due to inadequate lifetime of the minority carriers as they diffused across the base region of the transistor. We made independent lifetime measurements on these structures and found that the lifetimes were very low—less than a microsecond half-life. With the doping technique we were using, Ernie could grow base layers (the center layer in a pnp or npn structure) down to about 0.5 mil (mil = one thousandth of an inch) in width. But it became clear more narrow junctions were needed because of the low minority carrier lifetime.
Ernie next tried a technique known as rate-growing. That depends upon the fact that the rate at which an impurity is incorporated into a growing crystal depends on the rate at which the crystal is growing. In addition the magnitude of that effect varies depending on the impurity. Thus, if you are growing a crystal from a melt containing both a donor (n-type) impurity and an acceptor (p-type) impurity, the resulting crystal can be either n-type or p-type, depending on the rate at which the crystal is grown. Since the growth rate can be changed much more rapidly than you can add different kinds of impurities to a melt, you can make transistor structures with much more narrow base layers.
Using that technique, Ernie produced structures with much more narrow base layers and I tried to make transistors from them. They were npn structures and I ran into a major problem when I tried to make electrical contact to the sub- micron base layers. The best way to make good electric contact to p-type silicon was known to be by alloying an aluminum wire to the p-type silicon. Aluminum and silicon are known to form a lower melting eutectic alloy at about 500 C. I found that when I tried to make contact to the p-type base, that the entire region would become n-type and the p-type base layer would disappear. There was a similar phenomenon in germanium, which was eventually traced to the rapid diffusion of copper, which turns out to be an acceptor impurity. We thought maybe that was problem in silicon but experiments demonstrated that it wasn’t copper. It wasn’t any impurity we could determine, but the phenomenon was there.
In parallel, we had also discovered by heating our formerly npn structures at lower temperatures, the p-type region would appear again so that whatever was causing the effect could be annealed away. Through that process and careful annealing, I was able to make npn transistors with high alpha and amplification. I believe these were the first silicon transistors with power gain that were ever fabricated.
It’s certainly given credit for being in Crystal Fire.
TI gets credit for first silicon transistor
Well no, I think they have it wrong, as a matter of fact. They said the first grown junction silicon transistor was made by Texas Instruments. This is a little aside but I want to give you some material on that.
Let me speak about this a little more from the historical point of view. When we made these first transistors, we thought about patenting the process, but determined for two reasons that it wasn’t worth the effort. First of all, the fact that heating silicon to high temperatures would turn p-type silicon to n-type and that annealing at a lower temperature would reverse the process had been observed earlier by others. In addition, the rate growing process had already been patented. Second, and most important, was that we didn’t like the process. It was hard to reproduce. Growing the submicron base layers was difficult even with rate growing. The p-type conversion process was not understood and the annealing process required depended upon the batch of silicon that you used. So, from a manufacturing point of view, it just didn’t look very attractive.
At the same time, we had been pursing an altogether different process for making transistor structures with much greater promise, diffusion techniques. So we decided to publish the rate growing work and set it aside. It was published both informally and formally. At that time there was an annual “off-the-record” meeting that was held under the umbrella of the IEEE. (It may have been the IRE back then; I’m not sure.)
I think it would have been IRE back then, up to ‘63.
The meeting was called the Device Conference. It was a group where people working in semiconductor devices would get together and talk off-the-record and give papers that were not published, with the understanding (and hope) of keeping the information exchange more open and free. I have tried to find a copy of agendas from those meetings, and as far as I can tell even the agendas were off-the-record. At least, there’s nothing in the records I can find. I’ve tried the IEEE records but no one has been able to find anything. As I recall, these meetings were typically held in the late spring/early summer and I gave a paper there in 1954.
Gordon Teal, who had done some of the early work on germanium at Bell Labs, was at that session. He was still at Bell Labs when I got there, but he left very shortly thereafter and went to Texas Instruments. After my paper, he stood up at the Device Conference and said, “We’ve made silicon transistors at Texas Instruments also.” I invited Gordon to tell us about them, and he said he was not permitted to talk about their work. Our paper on the rate grown junction silicon transistor was submitted for formal publication shortly thereafter and appeared in the Journal of Applied Physics in early 1955. I’ve got a copy of that here someplace.
I continued to wonder just exactly when TI had made their first transistors. To my knowledge, their work was never published. I can give you some correspondence on this matter because what has happened recently has made me just a little angry. I only began to understand that there was a question about who had made the first silicon transistor when I attended the celebration at Bell Labs of the 50th anniversary of the invention of the transistor. Ian Ross give a paper there that was later published which said the first grown junction silicon transistor was made by Texas Instruments. I got hold of Ian and said, “Where did you get that information?” He said, “From TI.” So, it was possible, that TI’s work had actually predated mine but that they just did not document it publicly.
Then in 1997 (1998?) a book entitled Crystal Fire was published and the authors described how Teal demonstrated his first silicon transistor to the management of TI in April of 1954. Since my work had been done in January of 1954, it was clear that our work was at least three months prior to his. Here’s a letter I wrote to Steve Adams, an historian, who was hired by Lucent to do a history of Western Electric. I describe it all there. This is a letter that I wrote to the authors of Crystal Fire. I have never heard anything from them, not even an acknowledgment of my letter. (I did eventually heard from Michael Riordan who has since noted in a number of articles that the first silicon transistor was made at Bell Labs.) Here is a reprint of the publication on the grown junction transistor that I mentioned. In growing the crystals we used, Ernie Buehler would produce many growth transients. The Journal of Applied Physics used a photograph of one of Ernie’s crystal as the Cover of the issue that contained our paper. You can see here, the numerous rate-grown junctions. We would cut slices from such crystals and isolate single npn structures with which to do our experiments.
And are those differing lines reflecting the different rate of pull?
Yes. This is an acid stain that will show a change in the electrical conductivity. I’ve forgotten whether it’s the n-type or the p-type that stains dark, but let’s say it’s the p-type. So this would be n-p, n-p, n-p, n-p, n-p, n-p, etc. If you go into the structure here, particularly if you get to the middle of the crystal where the staining is not as dramatic, these will narrow down and a lot of them will disappear, so you can find structures where you can cut out a single n-p-n junction portion with a very narrow p-type layer.
Colburn:Right. That’s wonderful. I’d never seen anything that really showed it. I’d read it described a lot.
New technique for growing silicon single crystals
That’s pretty much the story of the first silicon transistor. I think TI did make similar transistors but I do not know what growing technique they used since, to my knowledge, they never published anything. I believe they did sell some of their transistors to the military. But all of that is merely prelude to the important part of the story.
As I noted earlier, we were working on a diffusion technique that looked like it would be much more promising if we could make it work because it was a much more controllable process. We were also helped by the fact that just around the same time another metallurgist at Bell Labs, Henry Theurer, developed the floating zone technique for growing silicon single crystals and the silicon crystals that were grown by the floating zone method did not show the p-type to n-type conversion on heating that made our grown junction structures so hard to control. We didn’t know why at that time. As it turned out, within the next year or so, another physicist at Bell Labs, Wolfgang Kaiser, did some simply beautiful work that showed that the defect that was giving us so much trouble was oxygen that was dissolved in the molten silicon from the crucible that was used to contain the melt. By heating and cooling the silicon crystal, you could either make the oxygen dissolve or precipitate. By looking through the crystal with infrared, Kaiser could actually see these precipitates forming. In the floating zone technique there was no crucible, therefore no dissolved oxygen and no conversion effect. That simplified life quite a bit. Unfortunately, the floating zone technique did not lend itself to controllable rate growing since the zone required very constant temperature and growth rates or the zone became physically unstable. In fact, over time the floating zone process was replaced by the Czochralski pulling technique using improved crucible materials and I believe all commercial silicon is now grown by pulling.
Two other chemists in the Chemical Physics Department, Cal Fuller, and Jim Struthers had measured diffusion coefficients of impurities in germanium, and Cal was in the process of measuring diffusion coefficients in silicon. Out of that work came the knowledge that Group III acceptor (p-type) impurities diffused more rapidly than Group V donor (n-type) impurities in silicon. If you simultaneously diffused phosphorous and aluminum at the appropriate concentrations into n-type silicon, you automatically obtained an npn structure near the silicon surface. The process was carried out in a high temperature furnace and impurity additions were made by passing a gas carrying the desired impurities over the surface of the silicon crystal. If you start with n-type silicon and pass a gas containing both n–type and p-type impurities (e.g. antimony and aluminum), the aluminum would diffuse faster than the antimony and would turn a thin layer of the original n-type silicon into p-type. The slower diffusing antimony would reconvert a thinner part of the faster moving aluminum back into n-type and you get the desired npn structure. The resulting layers would be very uniform in thickness and Cal could control their thickness to within a tenth of a mil.
How was that control achieved? Because I want to eventually get into the manufacturability aspects of it.
Well, the diffusion coefficients vary exponentially with temperature. You can control temperatures very accurately. So simply by controlling the temperature, the time of diffusion, the concentration of the impurities in the gas steam and knowing the diffusion coefficients, you have marvelous control over the structure. You don’t have to diffuse both impurities simultaneously. You can diffuse one impurity and then diffuse the second impurity. In diffusing the second impurity, the first one would continue to diffuse and move a bit, but you could compensate for that. So you had a really very highly controllable process for making transistor structures.
The real question was, now that I’ve got this nice structure, how do I make electrical contact to it? Making electrical contact to the bottom is very easy. You have a thick piece of silicon on which to make electrical contact. Making electrical contact to the top was not too difficult, although that layer was quite thin because it was prepared by diffusion. Cal wasn’t able to keep the furnaces and apparatuses that he used absolutely oxygen-free, and so we developed a relatively thick silicon dioxide coating on top of the silicon. But you could remove that with a careful etch without removing all the N-type silicon and evaporate or electroplate a metal layer that made good electrical contact to the top n-type layer. Making good electrical contract to the very thin base layer in the middle of the npn structure was the problem, and we really struggled with that.
I still don’t understand why some of the things we tried did not work. The technique that we tried first was to grind a very shallow angle on the silicon. Just through the geometric effect, that would expose the p-type layer and increase its apparent width. You could then bond a thin aluminum wire to the exposed base layer. We did achieve contact to all three layers that way and got transistor-like effects but could never get a high alpha. I could only conclude that we were damaging the active p-type base area in our angle grinding process. We tried to remove any damage with very careful etching, but somehow I could just never get a good contact and a good transistor.
One evening, my wife was having a bridge game with her friends and I decided to come back to the Labs and stare at the stuff for a while. I thought, “Well, what would happen if I forget about all this fancy angle grinding stuff and just tried to bond an aluminum wire through the top n-type layer and make contact to the p-type layer that way?” This was not a brand new idea. I was sure I could make contact to the P layer. The concern was that the aluminum wire was now also in contact with the heavily doped n-type top layer and I’d have a degenerate junction which would act as a short circuit tying the p-type and n-type layers together electrically. But what the heck, I thought, let’s try it. So I tried it and it worked! Apparently, as the silicon-aluminum eutectic solidified, it grew a p-type silicon layer that isolated it from the top n-type layer and the dreaded short circuit did not form. I don’t think I needed a car to get home that evening. I was flying high.
The next morning, I called a bunch of colleagues in, demonstrated the new transistor structure to them and got them to witness and sign my notebook. That was the birth of the diffused base silicon transistor. They were beautiful transistors with high gain and very high frequency response because of the very thin p-type layers. Of at least equal importance, they were made by a process that should be readily controlled for manufacture.
The highly doped junctions made by the aluminum wire between the base and the emitter was a bit of a problem in that it had a much higher capacity than you’d like. When you got up above 150 megahertz, you’d start to loose current just through that junction capacity. That was sort of intrinsic to that structure. However, nobody had ever seen 150-megahertz silicon transistors with alphas of 0.98 before, and so they looked like wonderful devices that could do a lot.
And what was the date of that?
I think that’s documented on this copy of a page from my notebook. [reading] “Now I’ll try the direct approach.” March 17, 1955. “This looks like the transistor we’ve been waiting for. It should be a cinch to make.”
Well good, that segues very nicely into my questions about manufacturing and manufacturability.
I think it was from this point on, though, that we kind of dropped the ball. We had developed in Bell Labs all the techniques required to invent the planer transistor that greatly improved the silicon transistor structure and all the techniques required to invent the integrated circuit. Carl Frosch at Bell Labs made the discovery that you could use the silicon oxide film (that had always seemed like a nuisance) as a mask to control the diffusion process. When he was trying to make silicon solar cells (he was the co-inventor of the silicon solar cell), he was getting uneven diffusion from the two sides of the slices of silicon that he used. He finally tracked that down to the differing thicknesses of the silicon dioxide film on his samples. Jules Andress at Bell Labs was working on photolithography and demonstrated how it could be used to control surface structures during diffusion and etching processes for silicon. I think they were both well ahead of the game; well ahead of anybody else, and obtained good patents of their work.
So, it was all there.
As an AT&T stockholder, that hurts. That’s painful to think about.
We were a regulated company then, and we wouldn’t have made very much money if we had invented the integrated circuit. We were required to license everybody to any of our patents at low royalty rates. But still, we should have done it.
Do you have a feeling for why it didn’t? Was it just a matter of somebody needing to see how all that fit together, or was it other things that made it not worth doing?
No, it was well worth doing. I did a few experiments with a researcher over in Systems Research who wanted to see if we could make passive networks on silicon by engraving thin diffused layers of n-type silicon on a p-type base into the form of resistors and capacitors. We made some circuits and he measured them, and it looked like it could be made to work. But we didn’t have the photolithography techniques at that time. We made them by melting wax and making little wiggly structures for resistors and etching p-n junction areas to use as capacitors.
Do you remember his name?
It was Bob Wallace. He was a very, very clever man. In the early days of the germanium junction transistor, he built a machine to help automate their manufacture. One of the most tedious processes with those transistors was bonding a thin gold wire to the narrow base layer. His machine started with little bars of germanium that contained the npn structures, made electrical contacts to the two ends, and then used a thin gold wire to step automatically along the bar as an electrical probe until it found the voltage discontinuity at the pn junction. It would then send a high current pulse through the gold wire, heating it until it alloyed with the germanium and made electrical contact with the base layer. His machine was called Mister Meticulous, because it meticulously probed along until it found what it sought. But the demise of the germanium grown junction transistor made it obsolete.
Promotion and a later move to the Metallurgical Department
It was just around the time we invented the silicon diffused base transistor that Bill Shockley left Bell Labs. Morgan Sparks, my supervisor, was promoted to take Shockley’s position and I was promoted to take Morgan Sparks’ position. So I pretty much left the lab bench. But I doubt that I would have invented the integrated circuit, even if I had continued to work in the lab. My background was much more chemical and physical than it was electrical, and I just didn’t think very hard about what the electronic possibilities were.
This, by the way, is one of the very early silicon solar batteries. It was made by Lincoln Derrick who was Carl Frosch’s technical assistant. That’s how the first solar cells looked in the lab.
That is essentially the same thing that’s in my little calculator?
Sure, except what’s in your calculator is an awful lot better. It’s probably, in fact, a thin film of silicon. This one is made of relatively thick slices of a silicon crystal.
I think it probably is, because you can actually almost see through it.
But if I were to put probes on the leads to this early battery which is now over 40 years old and hooked it up to a voltmeter, you’d see a voltage. I think the cells are connected in series, so you’d probably see a pretty decent voltage. I plan to will this back to Lucent or AT&T someday.
I think I also mentioned to you super-conducting magnets. That’s the other artifact I have here. This is one of the very early niobium tin test magnets. This coil would be immersed in liquid helium and then energized. It was made about 1960, according to the label I have on it.
To return to the story, a year or two after I was promoted to Morgan Sparks old job, I was offered the opportunity to come back over into things somewhat more chemical and metallurgical and later became the Associate Director of the Metallurgical Department (it was later called the Materials Science Department). That department was, for the most part, engaged in applied metallurgy. There were a few people in the Department who had started a basic science effort and we brought them together to form a unit for which I was responsible. Among them was Gene Kunzler, a physical chemist from Berkeley. He had worked with Giague there, who was very well known for his studies of the thermodynamics of solids at liquid helium temperatures. Gene was interested in measuring the low temperature electrical properties of commercially important metals, most importantly, copper. There were theories of why copper was such a good conductor, but no one had the data that was required to test those theories. Low temperature measurements were needed of very, very pure copper single crystals where the electron motions were dominated not by thermal effects and not by impurity effects, but just by the perfect copper crystal structure. Gene was successful in getting some very good data that permitted the theorists to proceed.
It was during this period that I had a visit from Rudy Kompfner. Rudy was the inventor the traveling wave tube. Trained as an architect, he became interested in electronics. I believe he did the traveling wave work in England and then joined Bell Labs. Rudy was very much involved in a very new device known as the maser. He and his group were trying to build maser amplifiers to use for the detection and measurement of very low microwave signals. Rudy came over and said something like the following; “You know, we have a problem here, and I wonder if there’s a metallurgical way to solve it. We need pretty high magnetic fields to tune our masers. And we need them over a substantial volume in which to immerse out maser crystals. To generate those fields, we construct coils of the best conductors we have (e.g. copper or silver) but they dissipate a lot of energy in the form of heat. The crystals that we use must be cooled to liquid helium temperature if we want a low noise amplifier and we can not tolerate the heat. So we have a problem.” It just occurred to him that since they had to cool their crystals to liquid helium temperatures anyway, perhaps they could use superconducting coils which would not generate heat.I talked to Gene Kunzler about it. Gene was a good scientist, but he also really liked to make things that might have practical applications. He got very excited about the possibilities. One of the problems was that superconductors don’t like high magnetic fields. If you put a superconductor in an electromagnet and increase the magnetic field strength , the superconducting properties disappear at what is called the critical field. Now, I’m talking about low temperature superconductors. “High temperature” (liquid nitrogen temperature) superconductors have similar problems but in a much more complex way. But we are talking about a time long before the discovery of the newer “high temperature” superconductors.Gene started looking first at superconductors with relatively high critical temperatures (somewhat above liquid helium temperatures), because it had been observed that the higher the critical temperature, the higher the critical field in any given material. Gene started with lead-bismuth alloys, which were known to have a relatively high critical temperature. Being in the metallurgical department, it was easy to get the alloy made, have it drawn into wire and insulated. You could insulate it simply by electroplating it with copper because the copper, albeit one of our best conductors, does not become superconducting at any known temperature so it serves as a perfectly good insulator for materials that do become superconducting. With bismuth-lead alloy coils, he produced magnetic fields of one or two thousand gauss which was a record but not high enough for Rudy’s needs. Rudy needed fields of at least ten thousand gauss.
Berndt Matthias was a physicist in the Physics Department at that time. He had done important work in superconductors and had discovered the materials with the highest known critical temperatures. The record holder was a compound of niobium and tin, Nb3Sn. The problem with that material from the magnet point of view was that it is very brittle. It’s a ceramic-like material. That’s a non-trivial matter when you need to make wire coils.
On the other hand, there were other alloys that had intermediate temperatures that could be drawn into wire. As I recall, they were primarily alloys of titanium and vanadium. Gene made some magnets of those and was able to produce fields in the range of several thousand gauss, new records but still not good enough. The question was where do we go from there. It was sometime around then that Gene and I made a deal. I promised him a bottle of scotch for each two thousand gauss by which he could raise the critical field liquid helium temperature above that of the titanium vanadium alloy magnets.Well, the real breakthrough came about with the help of the same super technician who grew the silicon single crystals I described, Ernie Buehler. Ernie was one of these people who had never gone beyond a high school formal education. But he was very bright, self educated and just extraordinarily imaginative with things mechanical. His proposition was to make a mixture of ductile, pure niobium metal and tin metal powders in the proper ratio, stuff the mixture into a tube of a non-superconducting metal such as copper, silver or stainless steel and draw that composite down into a fine wire. If we pick the outer coating with the proper ductility relative to the niobium and tin powder mixture, we can make a coaxial structure with the niobium tin powder in the center. He would then wind that wire composite into a coil, heat it up to a temperature where the niobium and tin powders would react chemically to form the Nb3Sn compound now in the form of the desired coil and insulated with the non-superconducting metal.
The process worked and I have here one of the first Nb3Sn magnets that was made. Stainless steel was used for this one because of its high melting point The first time Gene measured such a coil, he found that his power supply was too weak to generate a magnetic field high enough to quench the superconductor! He had to build a whole new rig, and coil eventually generated a field of over a 100,000 gauss. Divide that by two and you get a fifty bottles of scotch that I owed him. Even higher field were eventually generated but Gene let me off the hook for only two cases of scotch.
That work was done around 1960. By an interesting coincidence, at the same time these first high field superconducting magnets were invented, Ali Javan at Bell Labs invented the first continuous wave laser, the helium-neon laser. We had two major press releases within the same couple of weeks on two very important but independent inventions, the superconducting magnet and the continuous wave helium-neon laser. Those were exciting times.
Yes, it must have been. Getting back to the diffusion based transistor, what was the process for putting something that into production? Once you demonstrated it, once people got all excited about it, then what happened?
I was in the Research Department then and the Transistor Development Department was right down the hall. That Department was part of the responsibilities of Jack Morton, a vice president of Bell Labs. Jack was a very aggressive leader. He had a group in his laboratories also trying to make silicon devices. Within twenty-four hours of the time I had my first results, Jack was in my lab looking at them. He brought a number of his managers over to look at what we had. They picked it up and started running. There were multitudes of questions that had to be answered. What’s the proper mechanical structure for this new device. How reproducible is the process for making it? Let’s understand the diffusion process and think about how we can mass produce the materials we need and control all the processing required.
Which they pretty much then came up?
An open industry
Oh yes. I’m sure some of the first diffused base silicon transistors that rolled off a production line were made by Western Electric at their Allentown plant. We did not sell transistors commercially back then. We could only make them for our own use. Shortly after the diffused silicon transistor was invented, we held a symposium for all of our patent licensees. We gave papers, showed them around the labs, what we did and how we did it. The technology spread rather rapidly. As I’ve indicated, AT&T was required by a legal consent decree to license any of our patents to anyone at a reasonable royalty. We also operated under the conviction that since we would be a major consumer of the technology, the faster and farther it got developed, the better it would be for the Bell System. As a result of those facts and beliefs and the fact that we were the early leaders in inventing and developing semiconductor technology and broadly discussing and licensing it, I think the whole semiconductor area has been much more open in terms of the dissemination and sharing of information than you’ll find, for example, in the chemical industry where trade secrets are much more important.
Right. So you feel that AT&T being sort of a unique corporate regulated company probably has a lot to do with the whole silicon revolution?
Right, and the speed with which it moved forward at the beginning. When we made the first silicon transistor, we immediately presented the results and published a technical paper. TI was unwilling to even discuss their results. Of course, TI was a pretty small company then, and a proprietary position was very important to them. That’s understandable and I don’t criticize it. But the fact that the Bell System was so open created a culture of openness that characterized the industry for a long time.
So we’re probably much further along than we would be.
There’s no question in my mind that in the first several years that was certainly true. Disclosure was sort of the industry norm. I think people generally felt (and I think they were right) that, yes, it’s nice to have patents, nice to collect royalties, but don’t lean too heavily on them because others are always gaining on you. Better run fast yourselves for that’s the only way to maintain a lead.
That’s always good to hear. It’s become clear to me that Bell Labs is a very special place in the history of technology. A lot of things would never have happened without that particular kind of set up.
I think that’s right. The attitude there was that if you have a phenomenon that is useful to you, you want to understand as much as you can about it. Bell Labs was willing to do the most basic research. In solid state, for example, all kinds of basic work was done there. When you look back at it you will observe that it didn’t all lead to useful devices, but it did lead to a much better understanding of the material and what’s really going on in the solid state. It permitted you to recruit very good research people who would later move on to development, engineering and other forms of leadership. It permitted you to do modeling which helped further develop optimum device structures and prevented you from pursing too many blind alleys. There was an awful lot of basic work on superconductivity before there was any application and that application came in an unanticipated area. That happened because you could look at superconducting phenomenon and say “Wow! There’s got to be something useful there!”
These weren’t necessarily large efforts. The initial work on the silicon transistor involved about five people, two researchers and three excellent technicians. The bulk of the initial work on superconductivity was carried on by two or three people. The super-conducting magnet was developed by two principals plus some support work on making alloys and drawing wire. That had just a very specialized application in our business, so we never built up a significant development effort. But it has been awfully important in other areas. It has probably been most important in medical technology. MRI imaging depends on large superconducting magnets. You could not run those machines with high resolution without super-conducting magnets.
Thank you very much for the interview.