December 4, 2001Working With Fermi at Chicago and Post-war Los Alamos
Richard L. Garwin
University of Chicago
Enrico Fermi Centennial Celebration
September 29, 2001
Jim Cronin: Now we'll proceed with this program, and the first speaker of this first afternoon session is Dick Garwin from IBM, and the title roughly is "Working With Fermi at Chicago and Post-War Los Alamos."
Dick Garwin: I came to Chicago in the Fall of 1947 with my wife Lois, who is here with me, and after some months of getting accustomed to the courses decided that I needed to do some experimental work, which is my interest and my strength. So I got up my courage and went to Prof. Fermi and volunteered to help in the Lab which was a good thing for me to do. I discovered in his laboratory he had a machine shop--his own lathe, cut-off saw, and the like. He had great respect for the ability of the central shops in Chicago, but felt they were too fastidious, and if you put something there you would get it back ten times as accurate and ten times as long delayed as it should be. So he made a lot of his own equipment and this was just my milieu.
I found in his laboratory also Leona Woods Marshall-- Fermi and Leona had just done some work on electron-neutron interaction; and Jack Steinberger who was working away in the corner there with Geiger tubes and piles of graphite and lead doing something with cosmic rays. Fermi and Marshall were doing an experiment themselves using Geiger tubes on positronium; they had built the tubes themselves incorporating a cotton thread soaked with sodium-22 radioactive material, which would give a positron which would then annihilate and they were studying this. But Martin Deutch at MIT was studying too and he, pretty soon, scooped them because he had access to scintillation counters with end- window phototubes from RCA. Eventually Fermi got such phototubes, and we decided we would put them to work.
The electronics that was being used at that time was rather primitive according to my judgment. It used Rossi coincidence circuits from the 1930s which had microsecond resolving time and Geiger tubes with hundred microsecond dead time. And so it was time to make an investment in building things that could be used more efficiently for the future. So I looked into these matters and made some fast pulsers, so I could test some circuits if I ever made them; and some coincidence circuits which were in the few nanoseconds, a few milli-microseconds, range rather than the microsecond range--using so-called current switch technology and a diode clamp. And those were used for quite awhile here and elsewhere.
I remember how sad it was when Fermi's technician moved on to other things. Here's this great man dependent to some extent upon this young man who was working with him. And yet Fermi realized in order to have a career, the young person had to move on and Fermi had to start again with somebody else and show him how to do the work.
Fermi's propensity for building his own equipment was shown when the cyclotron was put into operation. First of all in the construction of the cyclotron-- it was a synchrocyclotron. So a lot of the problems with high-power radio frequencies come in spades; whenever you build such a thing which has the requirement to provide an electric field over a large area, you put it together and you find that all of the power leaks out someplace else, something burns up. And this is even more of a problem when the frequency is swept from a high frequency down to a lower frequency--as is required by the mass of the proton increasing from 938 some MeV at rest to 450 MeV plus that when it has been accelerated. And the frequency changes in a reciprocal fashion.
So Fermi would meet every morning at 7:30 with the engineers, Leroy Schwartz, and so on; get a report on the problems and prospects of the cyclotron activity; talk to them about the ideas and analyses he had made of the progress of the previous day; and then go about his work and hear from them either later that evening or the next morning.
When the cyclotron started to work, obviously there were beams that had to be brought out and targets that needed to be positioned within the cyclotron. So rather than having a large number of targets mounted on probes penetrating the vacuum chamber that could be plunged into position, Fermi decided that he could have a moving target-- a trolley that would move along the rim of the cyclotron. He took advantage of all of the natural behavior of the cyclotron-- it has a large magnetic field so you don't need for a motor a magnetic field, it's there. It's a stronger field than in any motor. The cyclotron had a ridged pole and so he could build a cart himself that would move along the pole in a circumferential manner. And the cart had a little target that could be flipped into the beam or not. After the trolley was made, he decided to put a thermocouple on the target so that one would know how much power was being dissipated there in nuclear interactions. I remember about 35 watts was pretty good behavior of the cyclotron for awhile. And I helped him a little bit there. Although he could control the cyclotron very well, other less capable people were bothered by the time lag between changing the beam incident on the target and the time the target would take to come up to temperature as measured by the thermocouple-- the thermometer on the target. So I made a little circuit diagram which anticipated what the result would be. You could control the cyclotron and instead of having, I guess, a 90-second time lag there was an irreducible couple of second time lag.
You heard of Fermi's suggestion at lunch that one use an analog computer for these molecular calculations. Well, he was interested in nuclear shell structure in the behavior of nuclear particles in very peculiar nuclear potentials-- there were Feenberg models, there were Jastrow models of the nuclear potential, among others. One day, I came into Fermi's lab in Ryerson Hall, and he had a coil (or maybe it was in this case a bar magnet) suspended as a torsion pendulum. He was going to make an analog computer. The analog of the spatial distribution of one dimensional wave function was going to be the time behavior of this torsion pendulum. And the analog of the difference between the energy and the potential energy would be the current fed to the coil as a restoring force. It's a perfect analogy--for small angles anyhow.
Now the problem would be, of course, if we were going to do this, to record the position of the needle with respect to time, and to give it the current as a function of time. And I suggested that this was probably not the best way to do this; that one could have an analog computer which used just operational amplifiers. And, of course, now I had to build this thing, which I did. And Fermi used it some. It was a whole rack of equipment in those days with operational amplifiers that used plus and minus 600 volts supplies, and had a curve follower so that you could put in the potential. After Fermi's death, I think, Clyde Hutchinson used it but nobody else ever did.
Among my major failures was the failure to respond to one of Fermi's suggestions. By the time, in December 1949, I received my Ph.D. and joined the faculty of the Physics Department, I had my own laboratory in the Institute building. I was busy doing experiments on the betatron; planning experiments for the cyclotron; getting ideas where I could--including scintillation counter ideas from Gaurang Yodh and others. Fermi came in one day and he gave me some suggestions as to a theoretical calculation on nuclear wave functions and energy levels. So I thought about that. Two weeks later he came back and asked what progress there had been and I told him none. So he said he was going to talk to Maria Mayer about it. And I just simply lacked the courage to put down what I was doing and take up a new field, where in fact I would probably not have done nearly so well as Maria.
Now in those days, I don't know how it is now, Chicago paid faculty nine-month salaries. They could either starve or get government contracts for the other three months, but I wasn't about to do either. Fermi suggested that I could be a consultant to Los Alamos. Rumor has it, although I don't recall, that I had made some suggestions to him about nuclear weapons and he said that the place to work on such things was Los Alamos. So in 1950, by then Lois and I and our son Jeffrey, went to Los Alamos for three months. Fermi and his wife Laura were there, and, in fact, I shared an office with Enrico Fermi which was a good experience. People would come in and talk to him. During the war he had gone to Los Alamos, not when the laboratory was formed, because he was busy here planning for the plutonium production reactor at Hanford. He went in the Fall of 1944 and stayed at Los Alamos through 1945. He was not in charge of any development group, although there was a Fermi group, but he was a treasured consultant also known as the Pope. Because whatever anybody wanted to know, and couldn't find another way to answer, they would go to Enrico and ask him and one way or another he would either show them how or, in extremis, provide an answer.
Into our office Fred Reines came in 1950 and suggested, maybe with all of these nuclear explosions going off at the test site we could put a detector underground and detect the neutrinos. Fermi talked to him and pointed out that a nuclear reactor-- one of the modern reactors, anyhow-- burns a couple of kilograms of uranium a day and fission of one kilogram of uranium in a nuclear weapon gives 17 kilotons. So lots more neutrinos are available from a reactor and you can get closer to it, so that Fermi's suggestion to Fred Reines, I think, probably led him to do the more feasible continuous experiment at reactors.
Stan Ulam would come in and Fermi and Ulam would work together on calculations for the burning of a cylinder of deuterium (the classical super)--which had been Fermi's original suggestion in 1941, I believe, to Edward Teller--that had caught fire with Teller but not in reality. It turned out to be very difficult. I won't go into the difficulties. But Fermi would have an accountant's spreadsheet and a Marchant calculator, and a sliderule, and would convert the partial differential equations that were involved to the first order differential equations; fill in the first few rows of the spreadsheet-- time would march down the spreadsheet; and he and Stan would talk about the parameters. And then their computer would come in who was named Miriam Caldwell. She would take this and come back the next morning with the results. They would graph them and give her the next problem.
Part of my own work that summer of 1950 was beginning an experiment to measure the deuterium-deuterium deuterium-tritium cross-sections which hadn't been measured for ten years previously at the University of Texas. And I thought it was a pretty weak read to lean on in deciding to build hydrogen bombs or not, when you didn't know what the cross-sections were. So I devised an experiment and began to build it there. When I had to leave at the end of the summer to come back to my responsibilities at Chicago, Fermi encouraged Jerry Kellogg and the laboratory director, Norris Bradbury, to import Jim Tuck from England-- he had been at Los Alamos during the war-- to continue this experiment, which was then published in 1954.
I worked also on diagnostics of nuclear explosions. The first couple of weeks that I was at Los Alamos in 1950, I spent in the classified report library reading the weekly reports of all of the groups during the war and after.
In 1951 I was back for the summer. I didn't share an office with Fermi. The Physics Division decided I would be better off as a consultant to the Theoretical Division and that's where I was ever after in my summers at Los Alamos. Fermmi was concerned that summer with a large number of things, including Taylor instability. If you have a stable interface like this water, you know, you perturb it and it ripples, and whatnot, and everybody knows that there are waves that run on the surface of the water. Everybody knows also to put a card over a full glass of water and you turn it over (I'm not going to do this for respect for the Ida Noyes Hall and the Max Palevsky Cinema) and the water stays in the glass-- it is stable. But it's meta-stable. If you take the card off, the interface is still supported by the air pressure but the water pretty soon falls out. This is a very important phenomenon in nuclear weaponry and had been plaguing the people at Los Alamos ever since they considered implosion weapons in 1944. Actually they were considered in 1942 but they really had to make them in order to use plutonium in 1944.
So Fermi had schematized the problem on his blackboard. Everybody knows that in the beginning stages of Taylor instability you assume a ripple on the surface, and instead of behaving sinusoidally in time it behaves exponentially in time with the same time behavior except it's imaginary instead of real or vice versa. So there is a time in which the amplitude doubles; the next interval it quadruples; the next interval it gets to be eight times as big. And pretty soon, of course, this cannot go on because the energy in the instability exceeds the energy that was driving it; the velocity exceeds the velocity of light. And so the question is what happens at large amplitudes? So Fermi said, let me make a model; I'll have a broad tongue which moves into the dense material; I'll have a narrow tongue that moves away from it and I'll just solve this numerically. So he did some of that but he wasn't quite satisfied with the solution. One afternoon around 4:50 p.m. John von Neumann came by and saw what Fermi had on the blackboard and asked what he was doing. So Enrico told him and John von Neumann said "That's very interesting." He came back about 15 minutes later and gave him the answer. Fermi leaned against his doorpost and told me, "You know that man makes me feel I know no mathematics at all."
There was another time at Chicago where our colleague, Edward Teller, who was also on the faculty (although you couldn't tell it by the amount of time he spent here) came by and told Fermi of his most recent enthusiasm. And as Teller left, Fermi commented, "That's the one monomaniac I know with more than one mania."
Well, I left the University of Chicago in December 1952 to change my focus from particle physics to condensed matter physics. I didn't like having to tell people six weeks in advance what I wanted to do with the cyclotron or to work with a team of six people. Now, of course, it's 60 weeks in advance and 600 people, so I think I made the right choice considering my personality. But I certainly admire what has been done in particle physics since. And I did work on superconductors, and liquid and solid helium, and I continued to work at Los Alamos.
By 1954 I had worked for a year (sort of half-time) on air defense of the United States and had made contact with people in Washington outside the nuclear weapons community. So hearing of Fermi's illness I returned, I think, in October and saw him in his house. He clearly had an inoperable cancer. And we talked. He regretted not having been more concerned about public policy. I was, of course, terribly saddened that Enrico Fermi was taken from us at the age of 53; I just imagine the joy that he would have experienced had he been around to see the evolution of computers and the development of physics, and the role that he might have played if he had been given another 20 or 30 years. Thank you.
Q&A: What was the computing situation at Los Alamos? During the war people had the card-programmed IBM machines which were card punches. The information was stored on punched cards taken from one computer to another. You could change the program by using either a plugboard or an IBM punched card which would hold the input information in. The speed was about ten calculations per second. Now, of course, the PC on your desktop has a thousand million (a billion) operations per second. Even my few-nanosecond coincidence circuits are totally incapable of keeping up with that. Now we have circuitry ten and a hundred times as fast.
At Los Alamos in 1950, I believe, there was under construction a Maniac-- a copy of the machine at Princeton that von Neumann and Herman Goldstein had developed. It probably had 10,000 operations per second or so. And IBM was making the 701 computer. The Maniac was programmed. In principle it had a magnetic tape for programming.but I believe it never worked. You could either program it with paper tape or you could go along with a clip lead. There are 32 registers. You could put numbers into a register and step to the next register. And we had some hexadecimal computers because this machine worked in binary. So they had some Marchant calculators built that would compute in hex so you could check the numbers that the machine would show.
Fermi put the Maniac to good use both in the phase-shift analysis and as you heard, in the work with John Pasta and Stan Ulam on nonlinear behavior and looking at the ergodic theorem--which is very hard to prove, because the more you looked in detail the less things became ergodic and mixed up. It's like stirring a mixture of dye and glycerine and then if you reverse it it comes back to where it was, diffusion being pretty slow. Fermi certainly have enjoyed having the power of programming language and the liberty to do these things himself without involving computing shops.
**Murmur of a voice from the audience.**
A: Ted says that Fermi came back after one of those summers and gave a series of lectures on how to program a computer. I recall also he made a trip to Los Angeles at the time that Dick Feynman introduced the Feynman diagram, because he thought this was an important tool that he really could not fully understand from afar. I'm sure that he then shared his insights with the rest of us.
**End of RLG talk.**