Oral-History:James Early

About James Early

Early worked at Bell’s Murray Hill lab 1951-64, the Allentown Lab (final-development oriented) till 1969, than at Fairchild Cameron. He discovered the Early Effect (figuring out that junction transistors are node circuits, not loop circuits) in 1952. He personally designed the transistors for the Vanguard satellite, but generally did more theoretical work. At Allentown he set the goal of packaging silicon devices properly. At Fairchild Cameron he set up the program to work on the buried channel charge coupled device (CCD). His involvement with the IEEE and Electron Devices Society included work on the IEEE Standards Committee, doing transistor definitions, and papers at the Electron Device Conference and International Solid State Circuits Conference. He values the organizational framework provided by the IEEE. He speculates in general terms about the future of the field. Early died in 2004.

About the Interview

JAMES EARLY: An Interview Conducted by David Hochfelder, IEEE History Center, 22 December 1999

Interview #384 for the IEEE History Center, The Institute of Electrical and Electronics Engineers, Inc.

Copyright Statement

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, IEEE History Center, 445 Hoes Lane, Piscataway, NJ 08854 USA or ieee-history@ieee.org. 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:

Dr. James Early, an oral history conducted in 1999 by David Hochfelder, IEEE History Center, Piscataway, NJ, USA.

Interview

Interview: Dr. James (Jim) Early

Interviewer: David Hochfelder

Date: 22 December 1999

Place: Palo Alto, California

Summary of advances in electron devices; the Early Effect

Hochfelder:

Good afternoon. I’d like to start by asking you to give a summary of the major advances in the field of electron devices in the past fifty years.

Early:

Audio File

MP3 Audio
(384 - early - clip 1.mp3)

The field of electron devices is where I’ve spent most of my professional life. There has been a recurrent pattern in which there were paradigms. There were idealized models that were accepted by the majority of the leading workers in the field until there was evidence, an invention or an experiment that broke the paradigm and led to a new resolution. Take the case of semiconductors, which are of particular interest to me. Before the transistor was invented it was believed by executives at Bell Laboratories, or at least by Dr. Kelly, that buried somewhere in the possibilities of semiconductor rectifiers – including germanium, silicon and galena – was a semiconductor amplifier. Bardeen and Brattain were working under that paradigm when they discovered the transistor effect in December of 1947.

At that time Bardeen interpreted the transistor effect in terms of a surface effect model. It is believed that that surface effect model governed his work for a short period of time, although he didn’t say. Certainly this was the accepted view of most of Bardeen’s colleagues at Bell – with the solitary exception of Shockley. After being informed of and seeing the results, Shockley secreted himself off for a period of several weeks and came up with the theory of bulk transport of injected minority carriers and its resulting bipolar transistor action. Shockley postulated and invented the bipolar transistor and the related PN junction diode in a classic patent known at Bell Laboratories as Shockley Case No. 12. I believe it was published as a paper in June of 1948 as the "Theory of the Junction Transistor." That theory was revealed to Bardeen, Brattain and other colleagues in February of 1948 less than two months after the original invention of the point contact. This became the new paradigm, and very large R&D forces were thrown at the problem.

Interestingly, there was a subordinate paradigm. Along with the point contact transistor there was an equivalent circuit, T, which was commonly accepted. It was a loop circuit. As far as I know that combined paradigm rode along until January of 1952. At that time I had the good fortune to differentiate a simple equation. It was the equation that said collector current = emitter current Η transport factor ∀ + a leakage current. I was bored and had been assigned to work on junction transistors. I thought about the problem as I measured and looked at the data.

Finally on one January day I realized that I should not be looking at it as a loop circuit but as a node circuit, because the leakage that was occurring was in parallel with whatever else was happening at the collector. It was obvious there was something else happening. I differentiated the equation, and immediately a term stuck out at me. This is differentiated with a miter current constant. A miter current Η the parcel derivative of ∀ with respect to collector voltage, and bingo. Well of course I knew that. The base layer of the transistor was not fixed in width. As the collector voltage rose, the collector depletion layer got wider and infringed on the base thus narrowing it and allowing more emitter current to pass. Therefore the collector current or fixed emitter current would rise as the base narrowed. This of course decreased the amount of recombination in the associated base.

That was the discovery that later came to be called the Early Effect, although I called it the Effects of Space Charge Layer Widening in Junction Transistors. That was a breach of the Shockley paradigm, because Shockley’s equations were so beautiful and the transistors basically behaved as he said they would. This was almost incredible. Then there were devices that worked across many orders of magnitude in the response of current to voltage. It was almost incredible, and just beautiful. And here is deviation.

At about the same time, other deviations were found. Webster at RCA found that high miter currents produced conductivity modulation in the base with the effect that diffusion of the minority carriers doubled and transit time was cut in half. Because the majority carrier was unable to diffuse, it piled up and built up a field that accelerated the minority carriers. There were other effects, but I won’t go into all of those here.

Then about a year later McKye and MacAffee were looking at the breakdowns in transistors and diodes. They postulated what should have been obvious to us all but was not. That is that what we were seeing was not thermal effects but electric field effects. As we went to higher electric fields, we started getting impact ionization of carriers. Carried far enough, that could result in impact ionization by both kinds of carriers so that a hole created by an electron impact could go back and create an electron. Thus there would be a cumulative discharge called a runaway that was the source of avalanche breakdown. Those were major paradigms.

Then a new paradigm came along when, under instructions from Shockley, Lee, Tanenbaum and their colleagues at Bell Laboratories created diffused base transistors. The idea was that a very thin base could be built in consequence of discovering this space charge layer widening. That thin base was of the sort that I had proposed. I didn’t mention it, but I had at the same time proposed that.

The fact that you could build these very thin bases gave a new and practical way of building junction transistors. They were very high speed and had problems, but the basic diffusion technology was adopted widely by the industry. Fascinatingly, this technology was adapted by each company in accordance with its background. Texas Instruments immediately went to its germanium grown junction transistor lines and introduced the idea of putting the base material in with the emitter in the growing of the junction. In letting the base diffuse a little further than the emitter material, a diffused base resulted that was a grown junction diffused base transistor. It was beautiful, and they got to market with it in a hurry. Then they turned around and did the same thing in silicon. They were the first to market a silicon transistor of any kind. I think RCA did some experiments somewhat in parallel, making alloy transistors with a diffused base. They put them on the one-one-one plane, which was required for the alloy transistor.

Hochfelder:

Would this be the crystal orientation?

Early:

Yes. When you alloy into the one-one-one plane, the alloying end is of one-one-one planes. This was very important in alloy transistors, because you would get the emitter and the collector to come in parallel with one another.

Hochfelder:

What do you mean by an alloy transistor?

Early:

This is a germanium transistor made by putting a small globule of indium, or indium plus other materials, on one side and a small globule on the other side while holding them together in a fixture. This is done with many transistors at a time. Then this is put through a furnace at which temperature allows the indium – or indium and other materials – to dissolve the germanium. On cooling the germanium doped with indium is recrystallized. This is where indium, and maybe other materials, are the dominant impurities – particularly P type. This made first the GE and later the RCA transistor the transistor of choice for economy. This was the PNP germanium alloy transistor. RCA improved the GE transistor by making the collector larger than the emitter so that most of the emitted carriers were transmitted through the base and collected by the collector. They then tried to build diffused base transistors with an alloy miter/alloy collector.

Philco had been working on a separate line of germanium transistors in which it at first tried to simply plate electrodes. Perhaps it had started out doing the plating with the intention of alloying the plated material – indium never plated — into the semiconductor. Fortunately or unfortunately, they tried biasing structures before alloying, and found that the transistors actually worked without the indium being alloyed. Apparently, it worked simply by injecting from the metallic indium through the N type germanium to the somewhat larger plated indium collector. I’m not sure that anybody ever really fully understood how those devices worked, but from a practical standpoint they were junction transistors. However, Philco found that they got better results by alloying those deposited emitters and collectors and went to that.

Later on when Bell Labs invented the diffused transistor, they proceeded to build an alloy diffused transistor called the micro-alloy diffused transistor. They had called their previous transistor the micro-alloy transistor. The Philco transistors were made with one transistor per little piece of germanium by a jet action spray of electrolyte. They were made with a stream of electrolyte from both sides of the little piece of germanium. They were observing and measuring the thickness of the resulting germanium by some external means without touching. I think they used infrared. After getting it to the desired thickness, after the action, they would plate. They did that of course for the micro-alloying and the micro-alloy diffused transistor. All of those side issues – the micro-alloy diffused transistor, Texas Instrument’s grown junction transistor, and the RCA version – fell before Bell’s diffused base transistor, because the diffused base transistor could be made an entire wafer at a time in that form.

Hochfelder:

Was the diffused base transistor Shockley’s idea?

Early:

Shockley never claimed credit for it, but he specifically instructed Lee and Tannenbaum as to what they were to do. He gave them their assignments. Based on the timing of his instructions to them, my suspicion – with no validation – is that he had been deeply influenced by my proposals for thin base. However, they made no reference to it when they got their results, so apparently he made no reference to my work or proposals.

Principles of transistor operation

Hochfelder:

For the benefit of those who are not familiar with transistor technology and electron devices, would you please briefly explain the basic principles of transistor operation and its basic parts?

Early:

Audio File

MP3 Audio
(384 - early - clip 2.mp3)

Speaking of the original junction transistor Shockley proposed, here we have an electron device. A semiconductor material of the modern sort is a bit like a gas in which we can do something very strange. There are two kinds of global charge carriers in it. There are mobile electrons and mobile holes or electron vacancies. The electrons in Shockley’s analogy are riding around on the upper level of a parking garage. Down where the holes are located there is a level parked solidly with cars. All that can be done is to move the vacancies from one spot to another. Imagine the cars can move in both axes in the manner of some modern Japanese cars that move sideways as well as back and forward. That’s a pure semiconductor with an equal number of those mobile charges. By displacing some of the silicon or germanium that are four-valent atoms with five-valent atoms, for example phosphorus, four of those valences can tie into the silicons. The fifth thing is just sitting there, and the associated electron goes wandering off. We have a nucleus that has five charges instead of only four. Effectively, that phosphorus atom is a +1 charge. The mobile charge of the electron up above is balancing it. At the same time, at normal temperatures and without external disturbance, the concentration of holes lessens substantially because there are so many electrons around. Wherever a hole pops up, they try to fill it. That’s a simple way of looking at it.

What is a transistor? In Shockley’s view we start with a first region that is very heavily anti-pour [correct word?] electron doped, for instance phosphorus, with this five-valent material. Then we have a thin region which perhaps in Shockley’s time was originally a thousandth of an inch or two-thousandths of an inch thick. This is enormous by today’s standards. In that region, deliberately, is a positive three-valent atom that tries to make bonds in four directions. In this process it acquires locally another electron which sits there and doesn’t do anything. It is stealing an electron from somewhere, so there is a hole. What should be right at that boron atom is wandering all over in this lower level that is parked solid. And there are a lot of those. This material is P type, and it doesn’t have very many electrons. This P type region doesn’t have nearly as many holes per unit volume as the emitter had. The first region had electrons. Then a next region has still fewer electrons, but it’s N type again. These five-valent atoms are much fewer than in the base, which of course is still lower than the emitter.

When we apply a voltage between the emitter and base regions it will make it negative on the base and positive on the emitter. It will try to send electrons into the base hole and positive charges into the emitter. There are not many holes in there because not very many holes will go into the emitter. Normally there are a great many more electrons in the base than there are holes in the emitter and there will be a significant flow of electrons into the base. Those electrons go through, wander about and will get over to the collector region. If a voltage is put on that which is plus on the N type material and negative on the base, those electrons that get over and near the collector go “wham” to the positive electrode and create what is known as transistor action.

Controlling that little voltage between the emitter and base – and it can’t get very big, it is a fraction of a volt –if the current going into the base changes very rapidly with respect to the voltage a current will go through that is almost independent of the collector voltage. If that voltage is reversed, is biased far enough and doesn’t have a big resistance in series with it, it will take it as long as more current is supplied..

Hochfelder:

What is an equivalent circuit and what is the difference between a loop circuit and a node circuit? I assume that what you mean by the loop circuit and node circuit is a voltage equivalent versus a current equivalent.

Early:

A loop circuit is one in which the assumed independent variable for the analysis is a current. Let me carry it further. A physical circuit, and its model in terms of the branches – at first in terms of the nodes – looks the same whether it is going to be regarded as a loop circuit or a node circuit. The model looks the same to the eye and you use the same diagram. However, first you have these nodes which are places where the voltage can be defined. It can be said that a voltage is at ground, up at node 1, up at node 2. Those are with respect to ground and so forth. Between any node and any other node – and this includes nodes to ground – there can be branches. There can be more than one branch and they can be in parallel. By “branch” I mean a resistance, capacitance, inductance or some complex impedance. There could also be generators of current or voltage that are dependent or independent.

A loop circuit is any group of connected nodes where it can be drawn from ground to node 1 to node 2 and back to ground and it could be said that there is a loop that can be drawn there. I am assuming here that there is a branch between node 2 and ground. Where there is a node 3 there could be another loop from ground to node 2 over to node 3 and back to ground. Those are loops. On the other hand, when defined in terms of nodes what is looked at is a voltage from ground to node 1, the voltage from ground to node 2, and so forth. Simplifications are made. If looking at a loop circuit, one could automatically put the elements between two nodes into series with each other. For example there may be a series resistance and a voltage generator. They may be that way, and in fact that is the way the transistor was modeled. It had a very important modeling effect.

In the node circuit, when you treated as nodes, that same effect would be treated almost certainly as a conductance – a very high resistance from one node to the other but directly connecting them. In parallel with that would be a current generator. This is a very different concept. For those who are familiar with electrical parlance, the two are equivalent circuits – series resistance and voltage generator in series, and the parallel conductance and the current generator in parallel.

Hochfelder:

In electron devices theory a typical method is to develop a model of a transistor that treats it as a very complex circuit composed of resistors, capacitors, voltage generators or current generators?

Early:

That’s right.

Hochfelder:

Your crucial insight that improved upon Shockley’s work was that you looked at the transistor in terms of the node circuit.

Early:

That’s right.

Hochfelder:

And the virtue of that is that current becomes the independent variable?

Early:

Yes.

Hochfelder:

Shockley’s analysis used a loop circuit to model the transistor and in that sense the voltage was the independent variable.

Early:

Yes. Peculiarly, the result I got was implicit in what Shockley had done. However, he and everybody else were prisoners of the T equivalent circuit, the loop equivalent circuit. It was that part of the total paradigm that obscured vision of the truth. As soon as the problem was looked at the other way and the most primitive analysis was done, something fell out in a few seconds. I was sitting in one of those student chairs with an arm and was at a meeting and bored.

Hochfelder:

You differentiated the equation for collector current?

Early:

Yes – and bingo. I was completely distracted. I don’t have the vaguest idea of what that meeting was about. I couldn’t quite believe what had happened. There were a lot of smart people working this problem. I wondered how I could be seeing something nobody else had seen. Yet as the days went on, I came to believe that was the case. I got lucky.

Bell Labs working environment, 1940s-1950s

Hochfelder:

In the late forties and early fifties there were a lot of talented people at Bell Labs. Would you talk about some of those engineers and physicists with whom you had the privilege to work? What were your general impressions of that period and of the work environment there?

Early:

One example is what a son of my first supervisor at Bell said. Some years afterward he commented about going to the laboratory to pick up his father. He said one of his most powerful recollections was “watching the men come out, their faces filled with thought.” My dad came down to visit one time, and I took him to lunch in the dining room at the Murray Hill Laboratory. My father was then in his sixties. He’d been a newspaperman since age 20, served six months during the war on the line in France, and had done many other things. My father told me that he had never in his professional experience seen the look of thought and intellect he saw in that dining room at Murray Hill. He was shocked.

On perhaps three occasions during the early fifties while I was at Murray Hill I had lunch with C. V. L. Hartley – of the Hartley Oscillator. That may give you an idea what it was like at Bell Laboratories.

Hochfelder:

Was Hartley employed by Bell Labs at that time?

Early:

No, he was retired. He was living alone in a hotel in Summit, New Jersey. He had never married. He was a very tall, straight man and looked a bit like a British colonel. But there he was. And I met others. I met Frank Gray, who was still working when I joined. In 1931 he had invented the basic scheme of putting the color signals at the high end of the television transmission band. There was no use for this until after the patent expired in 1948. Talk about being ahead of your time. It was obvious that the recruiting strategy at Bell was to get the best people that could be found.

I remember talking to Sam Weiss. He later went off to become vice president and then president of Autodyne or one of the other aerospace houses in the LA area. He and I were talking at a wine and cheese party in the early fifties. We agreed that Bell’s strategy was the high quality of its researchers, and it was an obvious success. There was a department of theoretical physics that probably had as its primary objective the creation of a nucleus that would draw others of comparable ability – or perhaps lesser ability but the desire to rub elbows. It was absolutely marvelous. My guess is that Stanford Benet; [IQ equivalency scores] in excess of 165 was commonplace. I really mean that. That’s a fairly extraordinary level.

Conyers Harring was a research physicist, a theoretician. I remember a meeting up in a little room on the sixth floor. It was an attic room with a sloping ceiling and things were crowded. A visiting professor was explaining something like Van Allen waves – which I didn’t know then and still don’t know. In that meeting the professor got his explanation wrong and I saw Conyers Harring challenge his conclusion very mildly and discreetly. He was clever in his questioning and led this man to reverse the result without embarrassing him. I was impressed by the quality of his mind and heart. That was extraordinarily beautiful. It was brilliant on the one hand and modest and self-effacing on the other.

Hochfelder:

A rare combination.

Early:

I’m not sure it’s that rare. When I talked to Conyers about it years later he denied any recollection of it. It may have been his natural mode that he wouldn’t even think about it; that it was automatic not to injure others, particularly in their self-esteem.

And there were the quarrels. We know that there was a major disagreement between Shockley and Bardeen. I don’t know what happened. I plan to write to Fred Sykes to find out if he knows anything. Sytes claims to have known both of them very well. Nick Holonyak at Illinois feels certain that Bardeen was on the right track and Shockley deliberately steered him off. I say this based on some cryptic remarks from Bardeen to Holonyak in his later years. However, Bardeen was a very kind man and when I questioned him on the subject he wouldn’t talk. When I asked Shockley about the rift, he professed not to be fully aware there was a rift. My suspicion is that Shockley gave Bardeen and Bratton sailing directions and didn’t say a word to them about the direction in which he was thinking. Then when he discovered the truth he failed to impart it to them before the public meeting inside Bell Laboratories that was held three or four weeks after he discovered it. Shockley made a splash at that meeting. If Shockley had given it careful and kind forethought, he would have given an earlier and more private disclosure to Bardeen and Brattain. However I don’t really know.

Allentown laboratory, Bell Labs, 1964-1969; etching silicon nitride

Hochfelder:

You worked at Bell Labs from 1951 to 1964 and then moved over to the Allentown laboratory in 1964.

Early:

That’s right.

Hochfelder:

How did your work at Allentown differ from that at Murray Hill?

Early:

Audio File

MP3 Audio
(384 - early - clip 3.mp3)

The responsibility at Allentown was for final development. Our final output was specifications, procedures and methods of fabrication in a specification detail supplied to Western Electric. We were supplied process specifications to which Western Electric was expected to adhere. It would be used as the Bible if things didn’t go right. Commonplace in such work was that a really smart and experienced Bell Labs director would, when confronted with a problem and a quarrel about what it meant, ask to see the first issue of the drawings. This is because Western Electric would commonly change the drawings, and sometimes in a way that made no sense. The real differences could be found by referring to the first issue.

Of course the number of variables interacting in the fabrication of semiconductors is so large that it is very easy to make a move from a previously found optimum point to a nearby point from fragmentary evidence. That may look all right but soon prove not to be quite right. Then another set of data can lead to still another point, and one can end up wandering all over the place. The further one strays, the worse it gets. It’s not nice. There is no substitute for experimentation. That was basically the job, and in that I had an extraordinary opportunity to play a part in the development of transistors and integrated circuits. My responsibilities involved everything from old vacuum and gas tubes to transistors and bipolar integrated circuits. I had some MLS work too.

The most useful thing I did in that time period, when I had been there less than a year, was to make a proposal to my department heads. I proposed that we assume we would solve the problem of sealing silicon devices against moisture and against ion attack and arrange it so we could encapsulate them in plastic rather than vacuum tight containers. That went along through the summer of 1965 and early autumn with indifferent results. There were various modifications of silicon dioxide, none of which were fully acceptable. Phosphorating the glass seemed to help by providing some sort of ion barrier, but it had a little problem in that it tended to make the hydroscopic. Phosphoric acid is not something one wants to have around metals.

We were working away at that when Jim Godfrey, who I think was the department head of the chemical area, came back from a meeting of the Electrochemical Society. At that meeting IBM had revealed the story on a new material, silicon nitride, which was absolutely beautiful. It was totally impervious to ions even in rather thin layers of a few hundred angstroms. It did have a problem, however, in that it was unetchable. Jim laughed and said, “The unetchable material is like the immovable object and the irresistible force.” I asked him, “In what relation, Jim?” He said, “It doesn’t exist.”

Hochfelder:

It’s like the massless spring in the [inaudible word] surface.

Early:

Yes, exactly. What are you going to about it? He put his people to work, and before the end of the year they had three different methods of etching silicon nitride. I don’t know how many of those methods are in use today. I think silicon nitride is still in use to protect integrated circuits. It’s a very simple scheme. The joker is that the silicon nitride was IBM’s discovery, but they didn’t know what to do with it. They let an imaginary obstacle block them because they did not have the vision of what was needed. However that stimulus of knowing what was needed was already moving our people. They had butted their heads against what wouldn’t work. The idea that here was something that would work and all they had to do was etch it, the attitude was, “Use it.”

Theory vs. practical applications; the Vanguard transistor

Hochfelder:

In your career at Bell Labs you went from working on theoretical aspects modeling transistor behavior and structure to practical applications including how to produce and package transistors and semiconductors. Is that correct?

Early:

I don’t really think of it as a career, just a job. It’s complicated, because I was always interested in the theory. I was always looking for the generalization. This is unconscious. It’s just the way my mind works. However, yes, I worked on the experimental side almost from the beginning. I wasn’t terribly good at that silicon nitride etching development. All I really did on that was say, “Guys, here’s the goal.” Godfrey came back with the results and Godfrey’s people learned how to etch it. I did nothing except to say at the start that we needed this desperately. I had been involved in making alloy transistors similar to the RCA structure and design of the jigging to make them. I actually designed the little phosphorus springs and the rest of it to make the diffused base transistor that flew in the first U.S. satellite – the little basketball that went up and came down. That transistor was made at Murray Hill.

Hochfelder:

Is that Vanguard?

Early:

That sounds right. I personally designed the Vanguard transistor. At Murray Hill I had responsibility in the design and fabrication of transistors and solar cells. I was the department head during the time of Telstar and I had most of the transistors and solar cells for Vanguard. The solar cells were made at Allentown, but one of my groups was responsible under Friedhoff Schmidt for the detail, design and so forth. I was mixed up in these things a long way, but I was always fascinated by the theoretical and worked on theory off and on.

One side issue was that we once took data on 3,000 of these solar cells regarding the short circuit current. We plotted it on log of current versus the probability coordinate, the bell-shaped curve. This is the percent of those having shown across. It’s the percent of those having a current less than that shown by the vertical on the large scale. What resulted was a marvelous straight line. It was a log normal distribution, and it was a little tail at each end. We went further into this, and discovered that the little tail at the low end was cells showing excessive damage – obvious gross damage. On the high end the cells showed an unusual staining. It was an oxidation to a different color increasing the light penetration and absorption. Aside from that, it was a normal distribution. These tails were only maybe 1 percent, very small, but large enough for the data to be real.

Employment at Fairchild Cameron, 1969; charge-coupled devices

Hochfelder:

After you had been at Allentown for five years went to work for Fairchild in 1969. That was Fairchild Cameron, not Fairchild Semiconductor.

Early:

That’s right. Fairchild Semiconductor Corporation – a wholly owned subsidiary – had been absorbed into Fairchild Cameron with substantial payouts to the founders of Fairchild Semiconductor. Fairchild Semiconductor existed as an unused corporate shell on the shelf inside Fairchild Cameron. I came out as a vice president in research and development. I was used to this kind of world yet ill fitted for it and failed at it. I did not disburse the 600 people to the operating divisions and did not cut the laboratory back to a more reasonable size considering market conditions and so on. Consequently I was demoted to director of research and development. That was a bit hard on the ego but cost me nothing in salary. However my salary did not rise rapidly subsequent to that.

Nevertheless I clung to my new position and tried to use the opportunities it offered. I discovered an entirely different and very wonderful mode of operation. My individual subordinates had their own responsibilities for the most part and relatively little coordination between them was required. Therefore I could discuss their individual goals with them and look over their shoulders. As a result I was able to make specific personal contributions to several projects. The evidence that the contributions were personal is that quite a few of them resulted in patents, and most of those patents are in my name alone. I’m suspicious when I see a patent with a lot of names on it and one of them is the boss, because that’s very easy to do. I never did that and I’m glad I didn’t. I was asked and sometimes put my name on patents after discussion with the patent attorneys. This is because the patent law is strict: only the inventors and all of the inventors’ names must be on the patent. A violation in either direction invalidates the patent, so one must be careful.

Hochfelder:

Did you work on charge-coupled devices at Fairchild?

Early:

Yes. That is a fascinating story. I had left Bell Labs at about the time the charge-coupled device was discovered. I suppose there was a report of it at Device Research in the summer of 1970, though I don’t remember that. What I remember vividly was a phone conversation with E. I. (Gene) Gordon at Bell Laboratories in March of ‘71. He said that the charge-coupled integer is the most sensitive photosensor known to man. Wow. That was a powerful statement.

Hochfelder:

The avalanched photodiode?

Early:

No. The joker is that you can get the same result. A charge-coupled device can have a lower capacitance [inaudible word] area, and charge for an interval can be accumulated and then passed along. Almost immediately I talked to a fellow at the Space and Defense Division of Fairchild on Long Island. I can’t recall his name now. We agreed that we would get a charge coupled program going, and we did. I should also mention that just a few weeks after that conversation with Gene Gordon, Rudy Dyke and I traveled east to Bell Laboratories and got insight into their CCD work, which was beautiful. Rudy Dyke subsequently ran the program.

Buried channel charge-coupled devices

Early:

I think it was on the plane on the way there that I conceived –conceived, not invented – of the buried channel charge-coupled device. I didn’t know it had already been invented by one of the people at Bell Laboratories a few months earlier. However having conceived of it, I believed in it. The objective of our program was to build buried channel devices right from the start. As soon as Dyke and I returned from Bell Labs we set up the program.

Toward the end of that summer someone called and told me that Gil Emelio of [inaudible word] was looking around. I gave Emelio a call and we arranged to meet at the TWA Building at Kennedy. We had a two-hour or three-hour dinner and agreed that he would come out and join us. He came out as an engineer, not as a supervisor. That was not because he was an unknown quantity. I was sure he could do a supervisor’s job. It was because it is psychologically better not to bring in a lower level boss from Bell. It was much better to have him brought in as an engineer and let him cut it along with the others. Gil came in September of 1971. Within two months CCD workers were lining up outside his door to talk to him. At that point I double promoted Emelio and moved him right up past Rudy Dyke. That didn’t bother Dyke a bit, because it was obvious that Gil was the “Joe Who Knows” and the man who should be in charge. He had the vision, the knowledge, the experience and everything it took to do it. That team, led by Emelio, produced the first working buried channel devices, which also had unheard of low noise performance. Are you familiar with the idea of the buried channel device?

Hochfelder:

No.

Early:

It’s a scheme whereby the semiconductor is doped properly and there is a suitable bias between electrodes and overlying electrodes as in MLS devices. Charge can be generated at one point and moved a centimeter or more as a packet of charge without being mixed with other charges. Without losing track of it and taking quite a long time, a charge can be moved a centimeter or more through a solid. It’s the unbelievable engineer’s mobile charge in a ball and moving inside a solid. The reason it is able to move as a ball is because it is sitting in a region that has a higher fixed charge density of an opposite kind to the mobile charge density. Of course at the fringes where one is out of mobile charge the other charge takes over. The thing works like a bloody charm, and is the basis of the endless camcorders.

There were several patents I did in there, and one of them was is dispute with the courts. The firm Loral now owns the patent and there are two Japanese manufacturers on it. The case was won by a jury trial in Brooklyn, and the judge reversed the decision. The Appeals Court upheld his reversal. There is old boy network involved. The latest is that the attorneys for Loral have appealed it and asked for a certiorari from the Supreme Court for a hearing at an appellate court level or whatever. The jury, purely and simply and without any holdups, found for the plaintiffs. Judges are not supposed to reverse jury verdicts on civil matters. There are two patents involved and one of them was mine. It rose out of my background in bipolar devices.

IEEE, Electron Devices Society

Hochfelder:

Would you talk about your involvement with the IEEE and the Electron Devices Society?

Early:

Sure. The first thing I’ll say is that I was involved with matters now considered IEEE before there was an IEEE. I was a member of both IRE and IEEE from when I was a graduate student onward until the two organizations amalgamated around 1960 or thereabouts. First of all my primary involvement was with the IEEE Standards Committee. I was a member of the committee on transistor definitions. I don’t remember whether I was involved methods or tests. Someone would have to go back and look at records. I was never one to keep a CV in the usual academic sense. I gave papers at the Electron Device Conference that pre-existed IEEE and at the International Solid State Circuits Conference. And at one time and another I was on papers committees for both. I think what is now the International Electron Device meeting that was originally held only at Washington probably started as a local initiative by local IRE chapters in the Washington area. At one time and another I was session chairman and different things. I still have a minor IEEE capacity in which I serve as an IEEE representative on an Engineer of the Year type of award. It’s an award named after a steelmaker and my role involves very little responsibility.

Hochfelder:

Would you talk about the role of the Electron Devices Society in helping to advance the state of the art in your field? What has been the value of the IEEE and Electron Devices Society in career?

Early:

The IEEE and the Electron Devices Society in particular have provided the organizational framework and sponsorship for very important technical meetings. In the years I was at Bell – and the same thing applies today to a lesser degree – there was not any question that the most important meeting from a research standpoint was the Electron Device Conference. It was traditionally held at a university in June. Many very important advances have been revealed at those meetings. An example is the classic meeting at Pittsburgh in 1960. The final version of the junction transistor, the epitaxial junction transistor where the collector was an epitaxial substrate, was revealed. This was at the same session that the MLS transistor was revealed. That was a double whammy from Bell.

I remember very well that Philco had just sold its micro-alloy diffusion technology to National Transistor or General Transistor in Long Island. It was probably General Transistor. One of the fellows from General Transistor was sitting in the audience. I came down the aisle from the front and asked him, “How are things going?” He said, “Oh, we’ve been stung.” They had bought a dead horse. That’s an old Syracuse story where somebody raffled off a dead horse and refunded his payment to the man who won. He can’t complain. He got his money back.

Predictions for the future of electron devices

Hochfelder:

What do you foresee as the sort of electron devices that will be available in the next ten or twenty years? What sort of technical challenges will engineers face?

Early:

The obvious is the exploitation is what is already known to be possible. At the same time, the bringing to life of anything that appears to be really significantly better. When I say what’s known to be possible, I am referring to the known physical limits on junction transistors, my old love, and the MLS device – particularly synapse as we know it now. Also, what could be done if fabrication was not a limit? There are physical limitations.

Hochfelder:

For instance as if minimum feature size were not a problem?

Early:

Audio File

MP3 Audio
(384 - early - clip 4.mp3)

Yes. Suppose minimum feature size is not a problem from a fabrication standpoint. It still can be a problem from an operating standpoint. It can get you into anomalies where for example supposing that it is not a problem, structures still have to be built which work at high speeds. The conductors have to bring the current to them and take it away. This means that conductors have to be built that are many times higher than they are wide or thick and many times closer than they are high. Now you have a conductor-to-conductor capacitance that is going to be a dominant feature. There are nightmares involved in this.

Hochfelder:

It will slow down the times.

Early:

Oh yes. Or, and you say, “Then let’s not make them as high.” That’s fine, except then not as much current can be put in because it can’t be gotten over there and back and so forth. There are real limitations. When it comes to something that does the same job that the electron devices do in silicon and related materials, there is a prodigious problem. The problem is, “What is the energy form, how can it be gotten to the spot where it interacts, and how can it be gotten away?” I hear words about the quantum devices and the 1-electron transistor, but I don’t hear the story of how these are going to work at a system level. By a system level I mean a chip level, a chip system level at high speeds. If a completely different fabrication technology is proposed there will be a real issue of how to get that fabrication technology up and running and debugged. Where is the niche application that will allow one to get started in that and drive it sufficiently far through? It’s a very nasty problem. I have a saying that pioneers solve the easy problems. An example of this is that Columbus discovered new continents. Shakespeare wrote fantastic plays but the English language was brand new, and Goethe’s plays were written when German was coming together as a language. Similarly, those of us who contributed to transistors in the early days solved easy problems.

However with every solution comes a space of extinction of alternate solutions. Each of the great symphonies occupies a space in the complex multidimensional space of music that excludes something that’s closely similar. In fact it has to be quite radically and unmistakably different. Where is this alternative going to get started? How will the incredibly fine technology be proven from a manufacturing standpoint? What functions will it perform that are so important that someone will be willing to pay the enormous bills that will be required? In short, the problem for tomorrow’s engineer is, “Your fathers did that. What are you going to do for an encore?” That is the nasty one. I’m very conscious of the S curve, and this is what we are talking about.

Hochfelder:

Are there any final thoughts you would like to share?

Early:

It has been great fun. There are a lot of things I would have done differently if I knew then what I know now, but I can’t complain. The biggest lesson I have learned in all the years is the importance of love. The truth is that we human beings are meant to love one another. Sometimes, sadly, it involves injuring others – but mostly it does not.

Hochfelder:

That’s a good concluding thought. Thank you very much.