Oral-History:George E. Smith

About George E. Smith

George E. Smith

Smith received his bachelor’s degree in physics from the University of Pennsylvania and his masters and doctorate in physics from the University of Chicago. He then worked for Bell Laboratories from 1959 to his retirement in 1986. He first worked in the pure research department under Willard S. Boyle, then moved with Boyle into applied research—Smith specializing in device concepts. He did not really get promoted much, due to friction with his superiors, but continued research in various areas under various titles. His research included the charge-coupled device (CCD), the picture phone (integrating photolithography and silicon arrays), the electron beam machine, X-ray lithography, far UV lithography, and simulations of devices via the Cray computer. His career generally moved from pure research to applied research, in parallel with Bell Labs general shift of focus, as the parameters of which research avenues were dead ends and which were technologically feasible became clearer. He was involved with the IEEE from 1965, when he went into the device theory area, in the Device Research Conference, and EdCon. He got Electron Device Letters started, an achievement of which he is particularly proud. Smith shared the 2009 Nobel Prize for physics with Boyle and Charles K. Kao. He and Boyle were honored for their work on the CCD.

About the Interview

GEORGE E. SMITH: An Interview Conducted by David Morton, IEEE History Center, 17 January 2001

Interview # 411 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, 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:

George E. Smith, an oral history conducted in 2001 by David Morton, IEEE History Center, New Brunswick, NJ, USA.

Interview

Interview: George E. Smith

Interviewer David Morton

Date: 17 January 2001

Place: Barnegat, New Jersey

Family and educational background

Morton:

Let’s start with when and where you were born and a little about your early life. I am particularly interested to learn what early experiences led you to your involvement in the field of physics.

Smith:

I was born in White Plains, New York and grew up in about seven different states. I went to nine different grade schools and five different high schools.

Morton:

Why did your parents move so often?

Smith:

My father was an insurance underwriter who didn’t like his profession. He kept trying other things but always came back to insurance underwriting when he went broke trying to do something else. He was in one of his down periods when I graduated from high school and I joined the Navy. I signed up for three years. The Korean War tacked on another year, so I was in the Navy for four years. When I got out I went to college with the help of the GI Bill. I was a little behind time, so I got through my undergraduate work in three years at the University of Pennsylvania.

Morton:

How did you end up at Penn?

Smith:

That was the city my parents lived in at the time and was where I had a free place to stay. However my father was true to form and after one year moved away from there. And I got my own apartment.

Morton:

Did they encourage you to pursue scientific interests, or was that something you got into later?

Smith:

No. I was always very good at math. When I was in the Navy I was stationed at an air base near Miami. In my spare time I applied to the University of Miami to take math courses, just because I liked math. I took the entrance tests for college and scored so high on the math part of it that they skipped my first year of math and started me in the second year. I started school wanting to become a mathematician. I majored in math with a minor in physics, but when it came time to do a senior project I wanted to do applied math. However they fired the professor under whom I had wanted to work.

Morton:

Really?

Smith:

Yes. They thought applied math was no good. They thought that good professors had to go into number theory or something very esoteric. That was what caused me to switch to physics. The esoteric math was too esoteric for me. I wanted to have a full reality with it.

Morton:

Did you pick physics because it was more applied? Did you have a vague or specific goal of what you were doing to do?

Smith:

No. I liked physics and it was very similar to my applied math. I was looking more and more towards the practical. In the beginning I was thinking of theoretical physics, but then I realized that I wanted to get my hands dirty and went into experimental physics.

I applied to two different schools and got appointments at Columbia and the University of Chicago when I got out of college. I was married by that time, and since she didn’t want to go to New York City we went to Chicago. And they had a very good physics department. I have forgotten what they call it when schools offer money to go to their graduate school. I had that.

Morton:

Something like a fellowship?

Smith:

Yes. Something. I have forgotten the names. I also had one from Yale to do mathematics, but had long since decided to get out of math. I got through Chicago in four years, going from a master’s to a Ph.D., which was kind of fast. I also had the shortest dissertation on record in the physics department.

Morton:

How short was it?

Smith:

Three pages.

Morton:

That’s short.

Smith:

Yes. This was because I published it under my own name in the Physical Review. They accepted that rather than the big fat dissertations one usually needed. It was a very short article, but was pretty good.

Morton:

That’s nice.

Bell Labs

Graduate school fellowship

Smith:

In my last year of graduate school I had a fellowship from Bell Labs and decided that was where I wanted to go and was absolutely positive I would get a job offer there. I never even interviewed another company.

Morton:

Was this the kind of fellowship that involved going to Bell Labs and working under someone there?

Smith:

No. It was a grant with no strings attached.

Morton:

That’s nice.

Smith:

I also had the GI Bill and the help of the National Science Foundation. I was very good at taking tests.

Morton:

At which laboratory of Bell Labs did you work?

Smith:

The laboratory in Murray Hill, New Jersey.

Morton:

What year was that?

Smith:

That was in 1959.

Willard S. Boyle's research department; electronic structure of solids

Morton:

What kinds of things did you do when you got there?

Smith:

I was in the pure research area. I was put in the department in which Bill (Willard S. Boyle) had become department head.

Morton:

Is he still around?

Smith:

Yes. He is retired and living in Nova Scotia. He was born in Canada.

I started out doing more or less pure research in the electronic structure of solids. I did a lot of early work in semimetals figuring out things and had quite a few publications. I got my first patent while in the research area. That was with Bill Boyle. We always interacted well with one another and we have several patents. Some of them turned out to be good, but more frequently they bombed out.

Morton:

I guess that was a time when there wasn’t a huge amount of pressure to apply everything to the immediate needs of the telephone business.

Smith:

That is correct. This was in the research area, which was less than 10 percent of the Bell Labs budget at the time. Bill Boyle and I were also avid sailors. We sailed together frequently.

Morton:

When did you start sailing?

Smith:

I started sailing when I got out of graduate school. I drove from Chicago to the east coast. My mother-in-law had a place over in Beach Haven on the Jersey shore here, and we drove straight through, driving overnight. Before going to sleep I went out and bought a boat. That was my first boat. It was what I had promised to do once I graduated.

Device Concepts group; semiconductor device and bipolar transistors research

Smith:

Getting back to Bell Labs, after about four or five years in the research area Bill Boyle got promoted and became director of the fundamental development area. It was applied research fundamental development. Shortly after that he offered me the position of department head there to start up a new group called Device Concepts. Essentially we were supposed to see if we could invent the wave of the future. I also had an offer from the research area to become one of their department heads, but felt I wanted to be more on the practical side of things. That was when I got involved in the semiconductor device area in a serious manner.

Morton:

Do you remember any other projects that were going on at the time?

Smith:

<flashmp3>411_-_smith_-_clip_1.mp3</flashmp3>

In my department we had a fellow, Dawon Kahng, who had a group that was largely doing research on semiconducting ferroelectrics where the ferroelectric material was not an insulator but was a semiconductor as well. All sorts of devices could be imagined that might be made out of that. I think the major material we had was potassium tantalate. I had another group under Art D’Asaro that was looking into semiconductor lasers. In fact, very soon after I got in the area he and another fellow developed the first single-mode striped laser.

Morton:

I am not familiar with that.

Smith:

They made semiconductor lasers by cleaving the ends of the chip to make the mirrors on the end. It was a little square thing, and light would bounce back and forth. It was a very multimode cavity since it was big huge crystal that the light would dance around. Art came up with a way to do a diffused stripe down the center of the strip so that there were only a couple of roads going up and down the stripe. It made a much more efficient laser. In addition, it could be run at room temperature. It didn’t have to be cooled down in order to get CW operation which was a big league problem at the time. Nothing came out of the semiconducting ferroelectric work. There was a rule of thumb that if 90 percent of what you were doing was not a failure then you were not being bold enough. We tried hard to keep up to that number.

Back around 1964 there was still some controversy over whether silicon or germanium would be the best thing for bipolar transistors, which were “the thing.” That was just when silicon was starting to win out over germanium even though germanium had a higher mobility. A good oxide could not be grown on it since germanium had a water soluble oxide. Silicon had this beautiful stuff called silicon dioxide on the surface of the chip. That eventually allowed MOS transistors to be made and put bipolar transistors to rest. A supervisor named Dawon Kahng along with another fellow, John Atalla, were the first to actually make an MIS transistor.

Morton:

Was he part of your department?

Smith:

He did this just before I came over to that area. At first it wasn’t really followed up on. The so-called experts said, “That’s a low power device. It will never fly.” In actuality, it is its low power makes the integrated circuit really work. A microprocessor couldn’t be made with technology that uses too much power.

There is one unsung hero of the MIS transistor named of Joe Ligenza. He was principally a chemist and he was the first to be able to make an oxide on silicon that had very low surface states density. In the MIS transistor, the electrons flow from the source to the drain right at the surface of the transistor. If the charge got caught in a lot of traps it made for a bad device. Joe Ligenza was able to very carefully grow an oxide with no traps. That was what made the MOS transistor possible.

Bill Boyle took over as director of my Laboratory from a fellow named Ian Ross, who had gotten promoted. Ian Ross held a patent on the first MOS transistor structure. I’m trying to think of how many years it was before bipolar research faded into the background. This started in 1964, and it wasn’t long before the change took over, certainly before ‘69. The goal was, and always has been, to make things smaller and smaller. Work on integrated circuits also started at that time. Jack Kilby and Bob Noyce are the recognized inventors of the integrated circuit. Getting things smaller and smaller while packing in more things on a chip was still the major goal when I left in 1986.

Electroluminescence

Morton:

Your bio mentions electroluminescence. What was that about?

Smith:

These are essentially the little diodes that you see lighting up as indicators on electronic equipment. Work on these light-emitting diodes was done in Art D’Asaro’s group as well as the work on lasers. You’ve heard about photodiodes for detecting radiation?

Morton:

Yes.

Smith:

Work on those devices was also done in the group. There was another group that was not in my department that was working on liquid crystal displays, but Bell Labs came to the conclusion that they were never going to go anyplace.

Charge-coupled device (CCD) and the picture phone

Morton:

We can’t win them all. Let’s talk about your work leading up to the charge-coupled device (CCD) in the project that was done in conjunction with the picture phone.

Smith:

The picture phone started around ’68 amidst a lot of reorganization. Gene Gordon had the silicon diode array camera tube group that developed the imaging device used in the picture phone. This was a device which used an array of silicon diodes instead of antimony trisulfide, or something similar, that was used in the Vidicon tube, another imaging device. Gene got promoted from his position as department head in charge of that project to a director and I was transferred to his laboratory and, among other things, given the silicon diode array camera tube project that was making millions of diodes on a chip. That was a considerable feat at the time to do with any finite yield.

Morton:

This was a device that was essentially a vacuum tube making use of silicon, wasn’t it?

Smith:

Yes. In fact I have one. This is a silicon diode array camera tube.

Morton:

It’s smaller than I expected.

Smith:

Did you get a copy of this article I wrote for the Stockholm Imaging 2000 Conference this past summer?

Morton:

I don’t know. I got an older article you wrote on the CCD. I don’t think I have anything recent.

Smith:

I think it would be worthwhile to give you a copy. It’s a very concise history of how the CCD was invented, starting with the MOS transistor.

Morton:

Did you follow the development of the picture phone all the way through the implementation phase?

Smith:

Yes, and I was still involved with it when it died.

Morton:

I understand that the image was in low resolution. Was it also a low frame rate?

Smith:

No. There was the TV standard of thirty or sixty frames per second, depending on how you look at the interlacing.

Morton:

Did it look more or less like some of today’s picture phone type technology where the image is jumpy?

Smith:

That’s a different thing. The Bell Picturephone was as good as regular TV in real time, but the image was smaller because of the bandwidth limitations of sending things over a twisted pair to the house. There was only one-fourth the bandwidth of a normal TV, so one-quarter the size of the screen. Instead of 500 lines there were 250 lines.

Morton:

When I spoke to Gene Gordon he said that he believed that AT&T might have had other ideas about what the picture phone project might ultimately have developed. He thought that maybe part of a hidden reason for developing the picture phone was to install a high bandwidth network that could be used for something other than the picture phone – for instance computer data applications. Did you ever hear that theory at the time?

Smith:

Yes, that was one of the side arguments. The main thing was the picture phone. That was long before the worldwide web, though the Arpanet was up at the time.

Morton:

I guess it was up in some form.

Smith:

Yes. It was very esoteric then. I played around with communicating using UNIX back and forth with someone else who knew how to use UNIX – which were rare creatures at that time.

Here is the best photographic film of one portion of the universe. That’s a CCD picture taken from the Hubble telescope at about a factor of 100 times greater sensitivity than the best photographic film.

Morton:

Why is that?

Smith:

The major thing is the quantum efficiency. You can close to 80 percent quantum efficiency in a CCD. What you want is a photon to come in, and for every photon to make one electron that you can count.

Morton:

I see.

Smith:

With the good CCDs, eight out of ten photons coming in would make an electron that can be read out and counted. With the photographic film a hundred photons need to come in before it breaks one of the bonds in the silver halide so that can be developed.

Morton:

Okay. That makes sense. We are getting ahead of ourselves. Back to the silicon array, you jumped in during the midst of that project.

Smith:

Yes, that’s true.

Morton:

What happened then? What were the problems?

Smith:

The problem was getting it into manufacture with any sort of decent yield. This is because of the couple of million perfect diodes needed. We wanted all of them to work, and that’s a very tough thing to do – especially back in those days before we had the modern fabrication technology that allows one to make huge microprocessor chips.

Morton:

My understanding is that this was something similar to an integrated circuit in that there was a single slice on which were made these millions of interconnected diodes.

Smith:

Yes, but only a single photolithography step was involved. I need a blackboard. If this is the surface of the silicon here, an oxide would be put on top and then photolithography would be done. Without going into how photolithography is done, the photoresist goes on top here. The oxide would be etched through, getting the oxide out of the way. Then a diffusion would be done so that there would be a bunch of diodes on the N type substrate. And that left a bunch of holes in the surface oxide.

Morton:

I see.

Smith:

The backside of the slice would be accessed by an electron beam in order for it to be read out. With the electron beam on the back, these diodes would be reverse biased and light would come in. If light came in here, it would discharge this diode. If there were no light coming in here, this would remain a reverse bias. It would take me a long time to describe this in understandable detail, but then, coming along with the electron beam after that, which of these diodes were discharging (those which received light) and which were not (those receiving no light) could be determined by looking at the substrate current in an external lead.

Morton:

That makes sense.

Smith:

The main thing is that it was just a single photolithography step making a simple diode. Making an integrated circuit takes about ten photolithography sets. They all have to be good in order to get a good chip. Making a million of these things was made possible only because it was such a simple type of thing. Even though it was really simple, just one bad diode caused by imperfections in the silicon would short out the diode, and therefore it couldn’t be used as a detector. Trying to get that into manufacture was a real beast.

Morton:

Would one bad diode make it unusable or would it just give a lower or faulty image?

Smith:

On this 250 by 250 array it would look like a white spot there. We never – ever – made a perfect silicon diode array camera tube. There were always some defects. Criteria were made up as to what was an acceptable number of defects in manufacture. The manufacturing was being done at Redding, the Western Electric plant in Reading. Those poor guys were getting headaches and ulcers working on this, and it was our job to help them out as much as we could.

Morton:

How did that interaction work out in practice? It seems like quite an adjustment for you to start dealing with issues of production rather than the more theoretical physical side of it. What were you guys able to do for them? How effective was that kind of cooperation?

Smith:

One of our major tasks was to determine the nature of these defects. There were a whole slew of them. The number of classifications for the chads on the voting ballots for the Presidential election was nothing in comparison to the number of classifications we had for defects. A little chip in the oxide at the edge of the diode will do it. And of course dirt is always a problem. We tried to determine what the defects were and then tried to interact with the manufacturing people to try to eliminate them.

Morton:

What was the relationship between that and the CCD project later?

Smith:

The CCD did not come about because of the silicon diode array camera tube. There was project on magnetic bubbles that influenced that. There were two groups in the fundamental development area. One group, that was just concerned with semiconductor devices, was in Bill Boyle’s division. He had since been moved upstairs to be executive director. Another group had all other devices. This magnetic bubble thing was the rage as the new memory device. They kept asking for more and more money to build up their effort on that.

Jack Morton was the vice president in charge of all this, and he was going to take money from semiconductor people and give it to the bubble people. No good administrator wants to lose funds. Jack Morton told Boyle, “If you want to keep your money, why don’t you invent something in semiconductors that will do the same thing that the magnetic bubbles do?” Namely, make a shift register. Bill and I really enjoyed batting things around. We had many ideas that never came to fruition.

Metal gate transistor for CCD, development and applications

Smith:

I have one side story I should tell you about.

Morton:

Let’s hear it.

Smith:

<flashmp3>411_-_smith_-_clip_2.mp3</flashmp3>

This is with the metal gate transistor, which is like an MOS transistor except that instead of having an oxide and then a metal plate on top of that there was a metal plate directly on the semiconductor. This allows for a device to be turned on and off very quickly. Bill and I came up with the idea to make a transistor like that. That was when both of us first went to the device area. We thought, “Aha. We can make a transistor this way,” and went to the device experts and told them we wanted them to submit a patent on this. I wish I had saved a copy of my notebook on it. They said, “Oh no, that’s no good. It can’t handle power. What you need is power in a semiconductor device.” People were still thinking of individual circuits or individual transistors being wired together. And power is needed to drive the interconnections. Everybody pooh-poohed our idea, so Bill and I said, “Well, we will have to bow to the experts. We won’t even bother sending it to the patent office.”

Another fellow got the patent on it about a year and half later, and it turned out to be a wonderful device for very high-speed front-end detection devices, largely for radar detectors and radios. There is very little power, but those are very low power signals. We kicked ourselves and also, mentally, the expert who told us it was not a good patent. The fellow who got the patent received a major IEEE award for it. I have forgotten his name. I think he’s out on the west coast.

I went to see Bill one afternoon. My memory is a little hazy, but Gene Gordon was out of town. Bill Boyle was the executive director, Gene Gordon was the director and I was a department head. There was some minor administrative problem so I went to see Bill about it. It was taken care of immediately, and then we talked about other things. I said, “Well, let’s invent something.” Then we sat down for a session I wish I had recorded on tape. We discussed some interesting ideas that day.

We don’t have dipoles like those in the magnetic devices. Magnetic Bubble devices have the magnetic dipoles in their substrate and they use the presence or absence of that dipole to represent a zero or a one in the memory device. It’s like in the magnetic tapes. Things are pointing this way or that way, the domain for the zero or the one. It was a yes or a no on the magnetic domain. We don’t have dipoles and semiconductors, but we do have monopoles, called charge. Why not use charge for our dipole? We have to contain this charge somehow or other. How can we contain the charge? Well, we can use capacitors to contain charge. What kind of capacitors can we have in the semiconductor? Well, the MOS capacitor. We make them all the time.

This is one of those things that once it came about, people would say, “Hey, that’s simple. I could have invented that if I had thought of it.” In any case, everything sort of hinges on the capacitor. We started thinking of making a shift register like ones they had in the magnetic bubble devices. This charge has to be moved from one capacitor to another. One could think of a whole bunch of circuits. Just to transfer the charge from here to over there, a transistor needs to be in between, or something like that. It becomes sort of a kluge. Some sort of regeneration is needed due to charge loss from going from one side to the next. And up and make it a true digital device where there would either be charge in there or not. It must be transferred along and regenerated. And that’s not a very good device. All sorts of circuits can be imagined to do that.

Essentially this was the invention. Let’s not have any circuit in between. Let’s just put these things close together. If charge is stored here, just put a voltage over here that’s bigger than the voltage over there and the charge will fall over. This voltage is then reset and you repeat the process. I have a diagram here. It was a couple of incidents of serendipity, if you will. This could be an analog signal, because there can be anything from zero to the full amount of charge. You’re passing along just that charge. If fifteen electrons are sitting all by themselves over in this potential well, you shift those fifteen electrons. Subject to quantum uncertainties of course. Once all this is done, you have a CCD.

Morton:

Is the significance of that the coupling and simplicity of the way the data is moved through the thing?

Smith:

Yes, that and the fact that the charge is naked at the surface. If we go back to the diode array we have the p-n junction here and reverse bias it. When a beam of light is sent in, a photon will make an electron-hole pair. The electron gets lost in the N type substrate. This hole diffuses as a minority carrier over to the p-type diode and reduces the reverse bias, but there are a whole bunch of holes in here. The individuality of the carrier gets lost. And there are noise problems associated with that. I’d be hard-pressed to try and explain that.

In any case, with that and with another device that Philips had called the bucket brigade, which is very similar to a CCD, there was that same argument that the charge they passed along was small in comparison to the host charge. They would essentially store the charge on the junction and not all by itself at the surface of the semiconductor. Both the scientific community and the patent lawyers agreed that the bucket brigade does not read upon the charged couple device, although they tried hard. Philips tried very hard to make the other argument.

There were three applications that we had in my first notebook entry, certainly using it as a digital memory device like the magnetic bubble and also using it as an imaging device. I had the diode array camera tube group. Both Bill and I were fully aware of its use as an imaging device from the word go. One thing that came later was its use for analog signal processing. That was more sophisticated and it wasn’t Bell Labs that came up with that. It was some fellows at the Naval Research Laboratory I think who came up with several different schemes for using it on signal processing. That looked very promising.

Morton:

I read in one of these articles a suggestion to use it as a display device as well.

Smith:

That was Gene Gordon’s invention. Given the CCD imaging device, if there’s light coming in here but none over here, there is charge here and no charge there. Then, when the optical signal has been integrated for a while all this can be shifted out to get an analog electrical signal that corresponds to the incoming optical signal. However, it can be done the other way around with an electrical signal read in that corresponds to the amount of light one wants to get out. There is a charge here and no charge here. And then instead of having voltages on here to define potential wells holding that charge there, those voltages those are collapsed in order to let the charge go into the substrate. If the electrons on the surface are pushed into the P-type substrate, the electrons and holes will recombine. When they recombine the photon will come out, within a certain probability. There are other recombination mechanisms that can take place. And therefore you have a display device. An electrical signal can be put in that corresponds to the optical signal. That is a clever thing to do but if you stick in the numbers, the resulting light output is a couple of orders of magnitude too small in power. A microscope and image intensefier or something like that would be needed.

By the time Bill and my original paper describing the CCD was submitted, Gene’s patent application, “The Display Device CCD Constructed by the Inverse Process,” got to the patent department and it was realized that we would never find a way to get sufficient power so a patent was never submitted. Everybody and their uncles were starting to make inventions based on the CCDs at the time. Dawon Kahng had several. Bill and me had several more. Once people on the outside heard about it, there was a real flood of CCD patents going to the patent office.

Manufacturing and picture phone market; Bell stops CCD research

Morton:

That was a few years after your camera tube project.

Smith:

Yes.

Morton:

Had the production techniques improved to the point where actual manufacture of this was less problematical?

Smith:

They were cranking them out. It was still a pretty low yield process, but the cost of that camera tube was not a major problem. The major problem was selling the picture phone. AT&T had deep pockets back then, and getting this thing off the ground was more important than the cost of doing it. The repeaters that had to be put in the phone lines were the major cost of the system.

Morton:

Did these go immediately into production or did Western Electric sell them?

Smith:

Western Electric was not allowed by the government to sell them on the open market. The same was true with our semiconductor devices at the time.

Morton:

I didn’t realize that.

Smith:

That’s our government.

Morton:

They could use them internally.

Smith:

Yes. They could use them for the Bell System.

Morton:

Was that immediately used in the Bell System, or was it something that had to wait?

Smith:

It was used in the first introduction of the picture phone.

Morton:

The CCD?

Smith:

The picture phone died before the CCD got to picture phone status. We had made CCDs on a laboratory basis with picture phone quality, but by the time we got to that point, the picture phone was nose-diving as a viable commercial product. We couldn’t sell them on the open market either. Then the memory business died, taking the bubbles and CCDs along with it. Random access memory (RAM) and magnetic disks took over the memory thing. Since we didn’t have an internal market for imaging devices and we couldn’t sell on the open market, the Bell CCD effort collapsed.

Morton:

Where did it go after that?

Smith:

Other companies were going like gangbusters. Gil Amelio, who helped make measurements on the first experimental CCD was hired away by Jim Early of Fairchild Semiconductor. He was a big pusher of CCDs and got Fairchild into the business. TI, RCA and Sony certainly were also. Sony is still a commercial maker of CCDs. They supply not just their own products but also their competitors’. I have a Sony CCD in my Nikon camera, even though Nikon has its own CCD line. The better and cheaper devices are from Sony.

Morton:

I guess your research on that stopped pretty soon after that.

Smith:

Yes.

Department management; Fine Device Laboratory

Morton:

What happened after that? You stayed with Bell Labs until 1986.

Smith:

<flashmp3>411_-_smith_-_clip_3.mp3</flashmp3>

I had many departmental names but never got promoted because I was kind of a maverick and didn’t show proper respect for my superiors at times. The MOS device department – which was the final name – had a laboratory headed by Marty Lepselter. It was the Fine Line Device Laboratory. The goal there was to shrink down the dimensions that could be used in design of devices. The laboratory had essentially three things going, and one was to get fine line lithography. We had invented the electron beam machine, EBIS, and we were also getting into X-ray lithography and far UV lithography to get these small lines. We were working not only on machines to make the lines but also the technology to use those lines to make devices.

When there were huge dimensions we had a two-dimensional models for the devices. With the sources and drains on the MOS device there was a big, long path between for the electrons to go tootling down and then drain. As the devices made became smaller and smaller, the source and drain came so close together it became a three-dimensional problem of getting the charge from here over to there. It was easy to do calculations using the two-dimensional model, but now we needed three-dimensional models from which to do finite difference calculations. Solving a set of differential equations required a bunch of small boxes in the computer. It takes a huge amount of computing space.

We were able to talk Bell Labs into buying a Cray computer. There were other reasons for getting the Cray computer, but probably the biggest one – because it had lots of dollars at the end of the rainbow – was to do simulations of devices. In order to do this experimentally it took about ten or twelve photolithography steps. To do them by trial and error, one of these devices would have to be made and then it would be weeks before an experiment was done from start to end. Then one had to try and figure out what was going on and what was going wrong. It is a very slow, laborious and expensive process. With a good simulation program, the Cray could be fed at night and the result would ready the next morning. It was also very, very cheap in comparison. I was involved in running a department that did these simulations.

I also looked at experimental results. We’d come up with a good design and then of course we’d have to try it out and see if it really worked. We had a little experimental fabrication group, the device group and a calculation group, all of which were against the order of things. We were not supposed to be in competition with another group, e.g., we were not supposed to do simulations because we weren’t the simulation group. We did it anyway, and we did good job.

Shifts in device technologies, 1960s-1980s

Morton:

Did there continue to be fundamentally new kinds of devices coming out of Bell Labs in the later seventies and eighties? Or do you think the labs in general paralleled your own career when their creativity seemed to move from invention to design and manufacturing?

Smith:

That was definitely the case. Back in the early sixties there were projects in all sorts of fields like semiconducting, Feroelectrics, a number of optical devices, semiconductor devices and compound semiconductors.

One group worked on gallium arsenide integrated circuits to get high speed. That’s another story. Gallium arsenide had a very high mobility and high speed. It was very good, but when making very small devices there were very large electric fields. When the source and drain were far apart, you’d introduce carriers here and they’d go zipping along – depending upon what mobility they had – from source to drain. It was in the low field regime of the velocity field curve. Looking at how fast an electron goes, the velocity versus the electric field and the low field region, it’s just a straight line. If you double the field, you double the velocity.

The gallium arsenide had a very steep slope and would get a high velocity from a low field. Just like this. Once you get to higher fields, the electrons start to emit photons and it actually saturates. Looking at silicon, that sort of goes like this. Looking at gallium arsenide, it sort of goes like that. Due to the band structure in gallium arsenide, electrons would transfer from a low effective mass portion of the band structure to the high effective mass portion. To explain that in detail I’d have to start talking about band structure and things like that. Looking at it just experimentally, gallium arsenide at the very high fields that you had when using very small dimensions, would actually go slower than silicon. There’s absolutely no point in fine line gallium arsenide technology. That project folded.

It went from the early sixties, where there were all sorts of possible device technologies that might have some fruition, to the late seventies and early eighties where it became clear that in the integrated circuits, it was the MOS transistor. More specifically, CMOS circuits were successful because of the low power. CMOS took over.

Morton:

Do you think that was because the experimenters and inventors felt they were running out of possibilities to explore or do you think it was structural changes in the industry that put less emphasis on exploring things?

Smith:

Back in the early sixties we explored them because we were ignorant about the possibilities. As we learned more and more and more certain things became clear. Once we learned that this takes place, we knew to forget about gallium arsenide. With the conventional magnetic disks and the tapes, there were those that said that we had the magnetic domains down as small as was possible and that they were never going to get any smaller. Therefore we worked on other technologies such as CCD memories and magnetic bubbles. The magnetic domain people wouldn’t roll over and kept coming up with more clever things whereby they could scrunch domains down. Back in those days the magnetic disk was the big thing to compete against.

The other thing was this fine line business of scrunching things down to the submicron range. That made random access memories so much cheaper that any intermediate memory hierarchy between the magnetic disks and the random access just didn’t have a place. Those two technologies started overlapping and squeezing out what was in between such as the magnetic bubbles and CCD memories.

Speculation on areas for additional research

Morton:

Are there any possibilities that you think have not been explored enough and might come back one day?

Smith:

Well, I have been out of the game for fifteen years.

Morton:

It is interesting to me that magnetic bubbles are so completely different from semiconductor devices – to do a similar task. I wonder if there are technologies that were abandoned because they couldn’t compete with semiconductor devices but which might have had potential given millions of dollars and decades of development. I wonder if there is really a different way to do things might come back to boost things to the next level.

Smith:

If you want to speculate, the only thing that comes to mind are the molecular devices. They haven’t been proven dead yet. There are enough people saying that they will never come to anything to make me believe that maybe there is something there.

IEEE; Electron Device Society and Device Research Conference

Morton:

Your involvement with the IEEE goes back a number of years. When did you first become involved with the IEEE?

Smith:

I became a member as soon as I went into the device area. I believe that was in 1965.

Morton:

Was that an expectation based on your new position or did you get recruited by other members?

Smith:

No, it was because all the good device work was done by IEEE people. I was a member of the American Physical Society, but they were too impractical for me.

Morton:

Was founding a new conference your first involvement with IEEE? That’s a pretty heavy level for beginning involvement. There must have been something before that.

Smith:

No, not really. It sort of went like this: Arthur D’Saro’s group was working on photoluminescence, semiconductor lasers and photodiodes and pointed out that there wasn’t any unified conference going at the time. That conference is still alive by the way. I said, “It’s too bad there isn’t one. Let’s make one.” Art and I were talking one day about some conference in Las Vegas. It was one to which we couldn’t boondoggle a trip. We both wanted to go to Las Vegas. I thought, “Let’s take care of both things. Let’s start a new conference and we’ll have it in Las Vegas.”

Morton:

Was there a big group that first year of the conference?

Smith:

It was over a hundred participants, so it was a good-sized small conference.

Morton:

Was it mostly AT&T people?

Smith:

No, no. They came from all companies. It was a general IEEE-sponsored conference. It was amazingly easy to start a conference at that time.

Morton:

Evidently.

Smith:

I got the rules out and sent for the forms that I had to fill out, filled out the forms and flew to Las Vegas and picked a hotel. Advertised it. It was easy.

Morton:

It must have been a really hot field.

Smith:

There was a vacuum there, too. People had interest in it.

Morton:

Shortly after that you got involved in the Device Research Conference. I’m interested in how people get involved with these things.

Smith:

The main conference for the Electron Device Society is an annual event. This Device Research Conference did not start out even as an IEEE conference, come to think of it.

Morton:

Oh, I see. It wasn’t EDS at the time.

Smith:

The Device Research Conference was the far-out conference for semiconductor as well as for all devices at the time. I started going to that every year since 1965 when I first joined the device area. You get to know people, and also we presented the first paper on CCDs there. It was an invited paper, and I was asked to be on the committee. Then I just worked my way up to the chairman of the thing. First you’re on a committee in a group, then you become the leader of the group, then you become vice chairman, then you become chairman and then you become an old-timer.

A very strange coincidence occurred. It started in 1953 or 1954 at the University of Pennsylvania while I was an undergraduate and had a job in the EE department measuring lifetimes in germanium. The first conference was held at the university museum and the professors for whom I was doing these measurements gave papers at the first conference on the lifetimes of germanium. Not only was I not an author, but since I was really an undergraduate they didn’t even give me acknowledgement in the write-up of the conference. I felt badly about that. That was not an IEEE conference then.

Morton:

Okay.

Smith:

It soon became IEEE-sponsored, but was fiercely independent.

Morton:

Sometimes it is a little hard to determine who has sponsored these conferences since they are independent entities in a way. Some of them are obviously and intrinsically linked to the Societies, but not so with others.

Smith:

I’m almost 100 percent certain that it started out as an ad hoc conference.

Morton:

All right. You were asked to do more committee and conference work for the Electron Device Society and you had a continuing role throughout the seventies and eighties. Was there kind of a club or group of people who stuck with it for many years? Or was this the kind of group where most people did just a couple of years of service and then moved on to other things?

Smith:

I guess you are talking about the AdCom.

Morton:

The AdCom, yes.

Smith:

That was more or less – and probably still is – and old boys network type of thing where you have to know someone to get on the committee. It works well. I cannot remember why I was able to get on.

Morton:

You were on the Foundation for a year in ’84.

Smith:

I started out as a member of the Sarnoff Award and got into the awards business with the IEEE. Once you become chairman – and I had a tendency to work my way up in these things – then that automatically makes you a member of the Foundation. When you become chairman of the Awards Board you automatically become a member of the Foundation.

Morton:

I see.

Electron Device Letters

Smith:

Probably the biggest thing I ever did for the Electron Device Society was to get the Electron Device Letters going. I was very proud of that, and I did that out of my own pocketbook almost literally. We wanted a journal with a fast turnaround. If you sent a letter to the Transactions it would be six months before anything got done. We wanted a two-month turnaround from receipt of a paper to the thing showing up in print. The IEEE publication people said, “It can’t be done. We just can’t do it that fast. You should know better.” My response was, “Well, we’ll show you it can be done.”

I got them to give me a guaranteed publication schedule if I gave them camera-ready proof that they could send to the printer, print right out and mail right out. They could set it up in advance so that if I could guarantee that they would get the copy by a certain date in order to mail it out at a certain date. They were pretty good about that part of it, but I had to get camera-ready copy to them. I think they allowed two weeks from my handing them the camera-ready copy. That left me with six weeks.

First, when a paper is received it has to get reviewed. Usually it is sent to the reviewer with a note from the editor that says, “Please return this in two weeks or as fast as you can.” When I sent this out I said, “You have to return this within 48 hours.” Another thing is that I picked three referees, and if I just got two referees’ responses back that was good enough for me.

I had to learn who were the experts in various fields. That was a real learning experience, and I generated a beautiful list of referees. If I didn’t get it back within 48 hours by fax or something similar to Priority Mail, I gave them a telephone call and read them the riot act.

Morton:

That sounds like a lot of work for you.

Smith:

Yes. Fortunately, most referees with this time scale were pretty good about it. And I just needed a two out of three yield. I would send the paper back to the author if there were comments, and there were usually comments. I sent same abominable kind of note to them. It was something like, “Please return this in 3 days or we will consider the paper withdrawn.” That usually got them working pretty fast.

Morton:

Yes.

Smith:

The IEEE had all the mechanisms for taking a manuscript, setting it up and proofreading it, and that tended to take two or three weeks to do. I got a local outfit that would do it almost overnight for me. It cost money. Bell Labs also provided me a full-time secretary to do this. We actually got two-month turnaround times out of this. This went on for close to six months and then I was wearing down. It was no longer the nice challenging hobby. The IEEE’s publication department finally said, “If you can do it, we can do it.” They took over, and they did almost as well as I did. They just had to be shown that it could be done. After I retired Simon Sze, who was one of my supervisors, took over. When he retired another fellow who is the editor now, John Brews, took over. It stayed in the family of my department.

Morton:

That was a big commitment on your part. I imagine Bell Labs was fairly supportive of this and your other professional volunteer work. Did they provide guidance about that kind of thing?

Smith:

They didn’t provide guidance, but they encouraged it.

Morton:

People have said that AT&T and its various subsidiaries supported this kind of involvement. Was there was a point where they drew the line if people got more and more heavily involved? Or did they consider it such an asset that they never really complained?

Smith:

I never caught any hint but that they considered it a plus. As I said, they provided me with a full-time secretary just to take care of the Electron Device Letter business. And a lot of other people did volunteer work. Bob Lucky spent a good fraction of his time on IEEE things.

Retirement

Morton:

Before we wrap it up, are there any final words you would like to say about your career or the IEEE or any subject that you want to get on the record?

Smith:

I started out being born in White Plains, and I guess the only comment is that when I retired, as I had been planning for several years, I jumped in my boat and started sailing around the world. In fact it was five years before I came back to the U.S. for the first time. I flew back for a week. I’ve been sailing slowly around the world ever since, with occasional visits back here.

Morton:

That’s an amazing hobby. You are lucky to be able to do that. There are worse ways to spend your time.

Smith:

Right. I reached seventy this last year and the bones start getting a little creaky after a while. I’m not sure I look forward to storms at sea anywhere.

Morton:

The cold and wet is a bad contribution. I guess it just means you have to sail to warmer places where you never have bad weather.

Smith:

Or do what I’m doing now, spending winters in a nice warm house and summers in the Med.

Morton:

It could be worse.

The oral history program is made possible by donations. If you find these oral histories valuable, please consider supporting the IEEE History Center’s work by making a donation to IEEE Foundation - History Center Fund. There is a box marked “Designations” with an arrow which allows you to assign your gift to the History Center.