Oral History Transcript — Dr. Philip Anderson
This transcript may not be quoted, reproduced or redistributed in whole or in part by any means except with the written permission of the American Institute of Physics.
This transcript is based on a tape-recorded interview deposited at the Center for History of Physics of the American Institute of Physics. The AIP's interviews have generally been transcribed from tape, edited by the interviewer for clarity, and then further edited by the interviewee. If this interview is important to you, you should consult earlier versions of the transcript or listen to the original tape. For many interviews, the AIP retains substantial files with further information about the interviewee and the interview itself. Please contact us for information about accessing these materials.
Please bear in mind that: 1) This material is a
transcript of the spoken word rather than a literary product; 2) An interview
must be read with the awareness that different people's memories about an event
will often differ, and that memories can change with time for many reasons
including subsequent experiences, interactions with others, and one's feelings
about an event. Disclaimer: This transcript was scanned from
a typescript, introducing occasional spelling errors. The original typescript
Access form | Project support | How to cite | Print this page
See the catalog record for this interview and search for other interviews in our collection
Hoddeson: Well, I thought we could discuss what should go into an ideal magnetism chapter on the history of solid state physics. Imagine a chapter of about a hundred pages, starting from an arbitrary point, and ending in the fifties and sixties. In the book, we were supposed to end every chapter at about Ď55, but it seems that all of them in fact moved into the early sixties.
Anderson: Thatís not a good break point. The war would have been a good break point after that. There is no natural, break point after that.
Hoddeson: Why do you feel war is a good break point?
Anderson: Everyone stopped work and started working on other things, more or less, and the other thing is that there was this tremendous technological thing which came out of the war. The whole of spectroscopy essentially started after the war. Nothing before that.
Hoddeson: Right. How would you organize it?
Anderson: I donít know how. Well, I suspect now thereís lots of material. I donít know, I havenít seen Ė
Hoddeson: I havenít it looked at it all either. Thereís material on early French work. Thereís some material on early Japanese work. I donít know how well itís going to be integrated into the chapter. Thereís some material on neutron diffraction. Once again I donít know how well thatís going to be integrated. There is material on van Vleck and Stoner and that ó
Anderson: Have they interviewed Charlie Kittel?
Hoddeson: I donít know.
Anderson: He was an absolutely key figure.
Hoddeson: I know. Kittel is a tough one to interview. He doesnít like to be interviewed on account of his stutter.
Anderson: Oh yes, I suppose thatís true. But everything I say, Charlie could say, because he was really my ó he wasnít supposed to be, but he was really my mentor. At Bell?
Hoddeson: Oh, was he?
Anderson: I learned solid state physics from him. I didnít learn anything from vanVleck because I was doing gas phase spectroscopy. And I learned a lot from Shockley, but then, thereís always the personal problem with Shockley competing with you all the time, trying to compete. So Charlie was really a mentor of the first magnitude. And in the end, well, I didnít even take van Vleckís course. I audited it a little bit, and I got bored. I donít know why it wasnít a good solid state course but it wasnít.
Hoddeson: When did you take it?
Anderson: In Ď47 I guess. I listened to Gorter and Brillouin, who were both guests during that period. Gorter of course is terrific, but just talked about his thing, paramagnetic relaxation. So he was in there, and communicating very much. Bloembergen of course came from Gorter, I suppose, or at least he came from that school. Well, I wonder, what would they be? Domain walls were a big thing. I donít know whether theyíve done anything about domain walls.
Hoddeson: Iím not sure.
Anderson: Itís really solid state physics. And I know what the old-fashioned magneticians were like. They were materials people. You put a little bit of nickel into some iron and... Bozorth was that kind of magnetician. And Charlie Kittel got absolutely furious with them.
Anderson: Because he would use people like Williams and Nesbitt, really good physicists, better physicists than Bozorth, and would use them to fill in spots in his phase diagrams. All these alloy systems. Whereas, Williams was inventing the science and studying domain walls. I donít remember what Nesbitt was doing at the time, but it was equally important. Williams in particular was a much better scientist than Bozorth.
Hoddeson: Was Bell really the major center?
Anderson: The major center for the physics.
Hoddeson: The physics of magnetism.
Anderson: Yes. There was Oxford. There was Bell. After Charlie went there, there was Berkeley. Otherwise I donít think there was much of anything but Bell, except for the coherence spectroscopy bit, the magnetic resonance bit. And even that developed only after Charlie went to Berkeley, and a few other centers sprang up. MIT to some extent in the early days. Of course, the Bitter magnet was developed at MIT, though I donít know that they did much with it in spectroscopy. I donít know does the chapter go into paramagnetic resonance? And nuclear NMR?
Hoddeson: NMR he is supposed to be dealing with, but Iíve heard recently it may be dropped.
Anderson: Thereís so much history of NMR. So many complications, all this business of the relaxation problems, and the different kinds of resonance and so on. Charlie, I guess, when he came from MIT, who did he study with? I donít know. Oh, Breit. Yes. Then he went to MIT as a
post-doc. He really was very much influenced by van Vleck, but I donít think he ever studied with van Vleck. He came to Bell and invented paramagnetic resonance with Yager, paramagnetic resonance. There also was an operation going on on studying domain wall motion under Williams and Galt. That was important. The ferrites were invented of course at Phillips. Phillips was a major center. I donít know whether theyíve gotten Phillips.
Hoddeson: I think theyíve done something with ferrites at Phillips.
Anderson: But then they didnít do much physics of ferrites. They studied the paramagnetics and the structures and so on, but they didnít do much with domain wall motion and other subjects like magnetic resonance. Now, Dillon, Ide on Clogston, Suhl, Walker and so on... Van Vleck was the gray eminence in terms of being the one who understood the spectroscopy and the line breadths and all that stuff. Very important.
Hoddeson: By the way, I read your biographical memoir of van Vleck. It was good.
Anderson: Well, you know, somehow people donít emphasize his role in the coherent spectroscopy business. It was so important. Well, neutrons. Of course, the first demonstration of Shull was very good. The rest of them were OK but Shull was really the person. And when Shull came to Bell Labs, the spectroscopy came along and it was in discussion with him that I realized that my ideas about super-exchange were going to explain what a structure was. That stuff came from Charlie Kittel too. Charlie and Gregory Wannier and I were sitting together, for the first months that I stayed there. We arrived simultaneously. And Charlie came in one day and said, ďWhatís all this about anti Perro magnetism we hear about from Neal." So Gregory started think about using models, which was his thing, and I started thinking about other things, and Charlie I guess got me the old Kramers paper on super-exchange. It was just an observation, it wasnít a paper. And so I did a first paper on super-exchange.
Anderson: Ď48, Ď49. And then Shull came around, and we realized that it must be real, because it explained why the structure was the way it was, and didnít have nearest neighbor interaction or next nearest neighbors interaction. Then I did a couple more simple things in field theory, and I got my first invited paper out of it. Shockley didnít like any of this.
Hoddeson: Why not?
Anderson: Oh, because he had hired me to do ferroelectrics and I was damned well going to do ferroelectrics.
Hoddeson: I see. He hired you?
Anderson: Yes, he hired me. So, we then really began to take mean field theory seriously, and of course, ó letís see, I guess what happened was, I went to Japan and taught a course in magnetism while I was there. And a Little Red Book exists on that course. Kubo, of course, knew much of it, but a lot of the people who later were to be Japanese magnetism people were there, Yoshida, and Moria, Kanamorióa whole gang of them. And then Bell Labs asked me to repeat the course, when they started this magnetism group back at Bell. So that was about your cut-off time. They taught Clogston and Suhl and a number of people, Brubeck? The guys who discovered the double membranes were also in that course.
Hoddeson: Did you spend a semester in Japan?
Anderson: A little over a semester, yes.
Hoddeson: A little over a semester. In Tokyo?
Anderson: Yes, with Kubo. Kubo and I had two things in common. One was we both did antiferromagnetic spin waves. My anti-Perromagnetic spin wave paper was before that. Both of us talked about our early anti-Perromagnetic spin waves at a meeting that was reported in REVIEWS OF MODERN PHYSICS that took place in Washington in 1952. A notorious meeting, because John Slaterís delightful personality was particularly evident at this point. We had a panel discussion in the evening, and just about the time that everyone else got interested, Slater got bored and said, ďI think we have to go to bed,Ē and stamped off the stage.
Hoddeson: Oh no.
Anderson: Just before van Vleck was going to talk about his stuff.
Hoddeson: He never did anything beyond what he did in the thirties, did he?
Anderson: Thatís right. Slater? Oh no, he tried to develop, and maybe even did develop a band theory of anit-Perromagnetism, in the late forties.
He didnít like Mott. He didnít like Mott insulators. He didnít like the Peierlís idea of the strong interaction. And somehow, although he invented the Slater integrals, he wiped the integral FO out of his consciousness, as not being what we now call U, a phase of magnetism physics? It didnít fit band theory, so he scrubbed this out. And therefore he couldnít possibly produce an explanation for manganese oxide being an insulator, unless it was anti-ferromagnetic. And he had this theory that it was anti-ferromagnetic and therefore it was an insulator, and got a band splitting on that. Slater would not have it that there was band splitting for any other reason than
periodicity. And he just, although heíd invented much of many body theory, he just insisted on ignoring everything heíd done before the war. He just changed. He had this dream of everything was going to be computed by the new computers, and everything he did after the war was focussed on that dream. And in the meantime, he came up on something which didnít look like it was computable, and he just tried to wipe it off the well. So van Vleckís theory of magnetism, and the Mott theory ó later on, there was a spectacular incident that happened in Ď67, Ď68, very late. Even later. In the early seventies, there was a Gordon Conference. Slater gave one of the final talks. Jim Phillips very carefully asked him, what did he think of the BCS theory of superconductivity? And Slater said, ďNo computer can reach that level of accuracy at this time. Therefore, we have no idea whether or not the BCS theory really is correct.Ē It was just so perfect, to show what his attitude was. This meeting was the first one where there was any kind of hearing for the van Vleck theory type theory of ferromagnetism that later on became more or less the standard theory. And he just didnít want to hear about it. I met Kubo there, and we also had an interest in line broadening. Kubo was picking up on these techniques which worked like, although we didnít know it, fluctuation dissipation theory. He wrote a paper with [???] that was the next step on ideas on [???] exchange [???] and that went into the [???] theory. Because it was Kubo who discovered the periodic boundary conditions.
Hoddeson: Why do you say Slater invented the many body theory?
Anderson: Well, he wrote the first theory of spin waves, didnít he?
Hoddeson: Well, simultaneously with Bloch.
Anderson: The first theory, microscopic theory, you know, real theory in the sense of trying to deal with a real substance. It was the predecessor of Herring and Kittel. And some of the other stuff he did. He invented the SNGs, which was very clever. All these clever techniques for doing wave functions.
Hoddeson: I guess so, I donít think of him as a many body person.
Anderson: Well, he wasnít really a many body person.
Hoddeson: Any more than I think of Heisenberg in the same way for his work that he did on many electron systems.
Anderson: Another person you may want to have some contact with is Herring. Well, after Charlie went to Berkeley, there was the initiation of cyclotron resonance. In the first place, all of the resonance work that they did, magnetic resonance I guess at first was really Oxford. It was a game for Oxford and Bell Labs. Bleagninís was very important in that early stuff. We were competing. I know one thing there was the G-marker business, in paramagnetic resonance. Charlie invented that just by looking at [???] and Bernstein and discovering that they were organic magnetic compounds. And Alan Holden made it, and our resonance group measured it. And itís ever since been used as G-marker. I donít know how important. They must be important in chemistry nowadays, although they werenít then. They were a sideline in chemistry. We did a lot of paramagnetic resonance. We were very much hampered by our equipment, which was lousy. There was a Bitter magnet, that someone had made the wrong decision about the power supply, and it fluttered to about 10 gauss. So we missed the first measurements of hyperfine structure.
The early days of spin-glass started in Ď44. When Charlie first went to ó well, thereís a whole thing of diluted systems, of Charlie, I think basically Foster. Owen, John Owen from Oxford went to Berkeley, and he did some of these dilution studies on magnetic ions of metals, manganese and copper and so on, and then he went back to Oxford and did dilution studies on manganese and zinc oxide, which was important in measuring exchange integrals. We all were confused. It was a very strange set of data on manganese solutions in copper. I mean, some theory by, could it have been Kondo And there was some theory by Marshall. Marshall was the first one who pointed out the scaling laws. And the Neal group took it up and developed this crazy Neal cluster theory. And all of that happened I think before your cut-off date. None of the theories were anywhere near right. The measurements were more or less right. We picked it out as an interesting system to study.
Hoddeson: When did the field of magnetism clearly, become one of the central fields in solid state physics? For a long time, it was sort of off to the side.
Anderson: I think it was the magnetic resonance techniques.
Hoddeson: The resonance techniques?
Anderson: And then for a long time, it really wasnít a central field in solid state physics, it was a central tool of the resonance theorists. Feher, your chapter doesnít have Feher in it, and it should, because Feher was the inventor par excellence of resonance methods as a tool, double resonance of various kinds and so on. And the other name is Portis, who did important studies, early studies of magnetic resonance. So magnetic resonance became a central subject. But it, you know, it never did, or it did for a while and then the focus, one we had cyclotron resonance and Fermi surface studies and so on, the focus suddenly became on Fermi surfaces, and magnetism went out again.
Hoddeson: Thatís an interesting pattern.
Anderson: Itís a central topic of many body theory, for many years.
Hoddeson: Yes. Right from the beginning.
Anderson: Right from the beginning.
Hoddeson: Right from the late twenties.
Anderson: Right. From the late twenties. And a central topic of critical phenomena, because of the neutron. Walter Marshall of the Harwell group ó once neutron diffraction rates became important, then Walter Marshall is group became very central to many body theory
and solid state physics. Then the Brookhaven group. One subject that should be in there is chromium.
Anderson: Yes. Well, you know all the mistakes that the neutron diffraction made, itís in my paper that feeds this neutron diffraction, because of this big expensive ó and had a lot of personnel in it ó was believed, and again and again they were wrong. Chromium is an example
of that. Itís the first metallic antiferromagnetic. There was a lot of work. In the first place, the neutron diffraction people declared it was not antiferromagnetic, and then they gave it the wrong structure. Then finally, with a great deal of pain, they developed the right structure. Magnetic transition rates were found by acoustic spectroscopy. There was Frank Moran [???], who studied all the transition metal oxides and found all these metal insulator transitions, which are not supposed to be Mott transitions but turned out to be, theyíre not magnetic.
Hoddeson: Chromium was ó anything else to say about it?
Anderson: Chromium ó no, no, just, well, itís an interesting case. Itís the first charged spin density wave. The first measurements were very early. Murray Kline who was at Bell Labs at that time, it must have been again in the late fifties.
Hoddeson: The late fifties?
Anderson: Or even early fifties. George Feher came from the Berkeley group, came to Bell Labs, and I guess he was hired during the period I was in Japan, so he was already doing Endor MNR and lovely magnetic resonance work on the silicons.
Hoddeson: When were you in Japan?
Anderson: Ď53. He would have been doing that by Ď54, Ď55. Thatís when they did the magnetic resonance work on the impurities in silicon, which eventually led to Endor and also led to the Shallow impurity in silicon physics. It really was a part of the whole thing. During the period when semiconductors were the name of the game, magnetic resonance was a major technique that people were doing real physics with, and so magnetism was very central at that time. Then, as I say, as soon as cylcotron resonance came along, and then the Fermi surface studies, it took a back seat, and one, electron physics began to be ó The other thing was the development of optical techniques. Very soon, the semiconductor person was a person with an optical spectrometer, and measured conductivity, and that was about it. I can remember coming back into the field of amorphous semiconductors, being shocked to discover that nobody, nobody ever made a magnetic measurement of an amorphous semiconductor.
Anderson: Because by that time, magnetism had dropped out of the bag of tools. And thereís also Pauling, you know. He always made a big fuss about magnetism as the criterion of the bond type, really thinking, you know, was it magnetic or what was the magnetic configuration of a given atom or a given alloy and that told you how many of the electrons were in bonding.
Hoddeson: Did he make any major contributions?
Anderson: To magnetism? No. Except a wrong one, a very wrong one.
Hoddeson: What was that?
Anderson: The Pauling 1944 theory, he still believes in it, he still writes papers about it ó
Hoddeson: Oh, really?
Anderson: He never believed in the theory of metals, you see. He believes only in resonating bonds. So he has a theory of metals along the lines of resonating bonds, and this gives you a curve for the magnetic moment of the iron group of metals, which is what you would get a lot easier by just thinking about band theory a little bit. And he says this is an absolute proof of his theory of magnetism. And that was 1944. It never had any influence whatever in the field of magnetism as far as I know. It was never accepted by physicists. Herring, Herringís key contribution there was that he put the field back together again.
The van Vleck school, the old fashioned school always thought of magnetism as isolated moments moving around according to the Weiss theory and resonance was a phenomenon of individual magnetic moments. And there was the Stoner theory. The Stoner group believed that magnetism was as described by the Stoner theory, just a uniform change in, the discrepancy, in the band system, spin up and spin down. And Herring and I, to some extent, put that together and said, "Look ó"
Hoddeson: óat what point did this happen?
Anderson: Well, that was beginning to happen in the early fifties. I remember talking with Shull and saying, "Look, you ought to look for spin waves in your metallic magnets" because there will be spin waves. There will be fluctuations. Metallic magnets arenít just Stoner theory. Stoner and Wohlfarth was just absolutely against this, "No, no, no, the metallic magnet is just uniform magnetization, thereís no collective behavior." And Herring and Kittel who wrote the first paper really on elective motion in metals either the first or the second, simultaneously and
independently with Bowman and Klineís; invented the theory of spin waves in metals. From that point on, Herring kept at every meeting he would give a talk, kind of give his cocktail speech, you mix so and so much and so and so much and that will be the theory of magnetism. He had given an example of this, in his work with Kittel. And in his own work in calculating the exchange levels. He was one of the very early structure [???] scientists in many body theory. It has to be Ď55, because it certainly pre-dated the great growth of many body theory in the late fifties. It has to be Ď53 or Ď54. So this was one of the sources of many body theory along with the Landon and Brueckner.
Hoddeson: Kittel and Herring?
Anderson: Kittel and Herring. So thatís important, and Herringís role in that is important. Oh, the ferrites and the garnets, the domain wall motion, we talked a little about that. That really didnít get re-integrated into physics until the wave of statistical mechanics of the seventies. Only later, was it realized that the domain wall was the first or the second of the topologicals, the solitons . But it was the first moment where people really thought about this thing as a quantum mechanical object which could move a dynamical object. Quantum mechanics was the dynamics. Larry Walker spent a lot of time thinking about quantum mechanics. Harry Suhl spent a lot of time thinking about it. Another thing that happened at Bell Labs was the nonlinear magnetic resonance business. Bloembergen and Ryan [???] came out with these strange results on the ferromagnetic resonance of high power levels. And then Harry and I found an explanation ó letís see, the first hints were Walker and Suhl, but they had the wrong mechanism. And then I pointed out that dipole interactions could do it, and then Suhl carried that on, and it really was the first work on using nonlinear, well, the first work on parametric oscillators and using nonlinear effects for amplification. Later of course then the laser people got a lot farther with that, but this pre-dated the laser. In fact, Harry was busy inventing paramagnetic oscillators, and then along came the maser and the laser and they were out of business. But it was very important theoretically. Bubble memories followed from that early Bell Labs work, by Joe Dillan. (pause...)
Hoddeson: Where did garnets come from?
Anderson: Where did garnets come from? I know why I associate them, Moran [???] and Seymour Geller [???] ó Moran [???] wanted to do oxide, I guess he came from the chemistry department and he was interested in varistors, variable resistors, and then he got interested in these, looking for materials, oxide materials which had very high temperature coefficients of resistance, because thatís what a varistor is. And he discovered that in all these vanadium and titanium oxides, there were these gigantic conductivity transitions. So he looked at them and said, "Aha, Mott transitions. Iíll bet these are neutron magnetic." Well, actually, most of them werenít, but there was one in there, and again, one of those situations where the neutron people looked and said, "Nonsense, there are no magnetic moments there," and it turned out that they just didnít have the accuracy they thought they did, and ten or fifteen years later, B2 O3 and several of the others turned out to be magnetic. But they still are the major examples of metal insulator transitions, these D-band oxides which now we know are really important, because other things happened, like high TC superconductivity. So Frank Moran was very early in that. Geller worked on some of those things. Geller was a structural X-ray man, and he and Alton Gilio, I guess, were fooling around with various oxide materials, and they discovered garnets, simultaneously with the French group. And then there was this enormous hassle back and forth about who really had the patents on the garnets and so on. Then Moran and Geller went off to North American, and had a checkered career thereafter, which is irrelevant to the history of magnetism but kind of fun.
Hoddeson: What happened?
Anderson: North American was one of these places that thought it would, you know, back in the days when it was fashionable to do so, thought it would re-invent Bell Laboratories. And they really didnít have the resources. Nobody ever quite realizes how much money Bell Laboratories costs. Itís a very expensive luxury. And the minute something like North American begins to have a little cash flow problem, they close their laboratory and thatís the end of that. So North American was the first of the many, many, many successors who tried to copy Bell Labs and didnít succeed. I guess none of those were in the group that I taught in Ď54. But Joe Dillon was actually was close associated with that and actually still works in bubble memories and that kind of thing. Theory of ferromagnetism that of course was what Hubbard was about. That I guess is past your cut-off.
Hoddeson: Well, Iím not sure. Was that about Ď57 or so?
Anderson: A little later, I would say.
Hoddeson: A little later?
Anderson: Yes, Ď59 was when I did my super-exchange theory. That had to do with, actually, my second super-exchange theory, that had to do with the fact that I was asked to write a review article for the Seitz-Turnbull series. I guess it was Seitz-Turnbull ó Iím not even sure whether Turnbull was in on it at that time. But I kept putting it off and putting it off. I would start to write, and I would realize I didnít really understand super-exchange, so I stopped writing. And the summer after I worked on superconductivity, all of a sudden I began to realize what was going on. And that was Ď58, that summer, the super-exchange theory, and that, as far as I know, is the first physical appearance of the Hubbard Hamiltonian. Hubbard came later. Well, one thing which is ó well, in that same time, 1959, after Iíd done the super-exchange paper, there was one of the perennial reports on the health and welfare of physics. I donít know whether this was the Pake one or maybe an earlier one. This was material science physics. It was one of these National Academy reports. There was a panel on magnetism, and I was on that panel, and among other things, there was a meeting in Oxford, which I used that panel to get to, used the money from that panel to get to via NATs, which we used to take in those days, and Bob Schulman was there too, I think also for the panel. That was where I first saw the Friedel theory of resonant states. Friedel talked about it. I had a little theory of ferromagnetism. And it was at that meeting that I made a bet with Walter Marshall that the hyperfine field in iron would turn out to be negative, and won; which was based on that theory. We all were kind of applying these resonance concepts to, and many body concepts to ferromagnetism. It was kind of the genesis of the Hubbard model approach to ferromagnetism, which then Konamori ó Hubbard himself and so on, lots of people applied and went on with. Hubbard was there. And I think it was, no, it was two years later that I talked to John and he showed me the Hubbard model and his first work with that. Could it have been two years later or could it have been in Ď59? Maybe it was in Ď59 when I visited Harwell. But it was already after my super-exchange paper, which used the same Hamiltonian but for different purposes. So everyone was after some kind of the theory of ferrogmagnetism. Thereís a Zener theory. That was great fun. The famous meeting about magnetism where Zener got up and said, someone asked him, "Have you computed these parameters?" and he said, "I donít need to compute these parameters. Nature is my computer." But it turned out Zenerís theory wasnít the theory of ferrogmagnetism, as far as we know...
Anderson: ...The Oxford meeting was very seminal in this magnetism business. From it came my Anderson model paper.
Hoddeson: When was that meeting?
Anderson: Ď59. September Ď59. Braznos College.
Hoddeson: Who was that, Zener?
Anderson: Well, Zener wasnít.
Hoddeson: He wasnít?
Anderson: Friedel, I think Hubbard, Kurti ran the meeting. Sherman. John Owen must have been, Bleeney must have been, Friedel and Bondell. Blondell and Friedel played a fairly important role in the theory of magnetic impurities. Blondell was always under-estimated. Heís done a lot of important work. [???] was too young. Ziman must have been there. And Roger Elliot, of course, because it was Oxford. I donít remember who else. The other thing we were doing, I was doing and Schulman, was ó well, not then but within the next year or two ó was, we did this work on transferred hyperfine structure, as a way of studying the wave functions of the magnetic ions. I guess John Owen was doing the same thing for the isolated magnetic ions in zinc oxide. John Owen was of course at this meeting. I think he talked about putting one or two manganese ions in zinc oxide and measuring the exchange between them, also measuring the way in which the manganese wave functions spread out into the oxide. The concept of hybridization. Proving van Vleckís ideas about Liegard fields and so on. Another line that came from van Vleck is the Liegard field theory. Bauhausen and Argell were the two names in that, especially in using it in chemistry. Thereís a whole book by Bauhausen. You canít really miss that, and Bauhausen has made a big fuss about the history.
Hoddeson: Why do you say that the meeting was so pivotal?
Anderson: Well, I think it, the communication between Friedel and myself, between the two of us and Hubbard, and itís the first point where I think everyone said to himself, well, if we really can think about the UT , what van Vleck called the UT model right, and think about the band theory right, maybe in the end weíll have not only a theory of ferromagnetism but also it was the first meeting where I think that everybody who was there recognized that there is no old-fashioned theory, that the old-fashioned theories of ferromagnetism didnít work. Stoner and Wohlfarth must have been there, and so these would have been an old guy there too, to argue against. It was that, and then there was this other thing, the idea of transfer of hyperfine structure, and detailed measurement of wave functions of magnetic ions. And that led to Shulman and [???] and so on, and Shulmanís beautiful work on NIF3 and KMNF3, which you know really showed that the magnetic ion and such and such an admixture with the chlorine and that then communicated in so and so much to the next magnetic ion and so on. Then I talked recently to Shulman, and he said that Art Friedman spent the rest of his life trying to destroy that work. A propos of Art Friedmanís role in the high TC superconductivity. Bob said, "Is he just as bad as he ever was?" and I said, "Yes." Oh dear.
Hoddeson: So that pointed the way to ó
Anderson: That pointed the way in a couple of directions. It was the genesis of my thinking about the Anderson model. Well, that plus the results on magnetic impurities in superconductors. The question Berndt Matthias kept asking, which was this question of why is iron magnetic in some cases and ruins the superconductivity. And in other cases it enhances it? Isnít there some mysterious thing, he would say. And I would say, "No, no, thereís no mystery. Itís just that we donít know." And I tried to explain. I tried to show that there could be magnetic and non-magnetic impurities.
Hoddeson: Then what next?
Anderson: Well, thatís really at the end of your time span.
Hoddeson: Maybe we could carry is a little further.
Anderson: Bob and I used to say to each other, "Good enough isnít good enough." And in fact, he kind of kept on with all this stuff and did a lot of very good stuff over the years. Really good phenomenology. D-band oxides in general. Itís all good stuff. And understanding which oxide is doing what, under what circumstances, and really clarified a lot of these metal insulator transitions as being coudent to ionic type transitions, things like that. Most of them really are pretty unglamorous. Theyíre just transitions of some sort, whatís called nowadays spin piles. And good enough; so what other subjects are there? Thereís the oxides. And thereís the general subject of physics. Thereís the Hubbard model, based for that for the Mott transitions, which really didnít come into fruition until much later. And for theory of ferromagnetism in general. And that again. Well, it still hasnít come into fruition. Nobody's really got a good theory of ferromagnetism, but modern work with Hertz and Heine and all kinds of people. And with much better band theory and much better ó Iím still very dissatisfied with the way things are done. For instance, itís gotten all cluttered up with people doingótaking their band theory too literally, in the Slater tradition. But that was what was beginning at that time, with Hubbard. Very shortly there was the Goodswilling [???] method and the Kanamori work and Maria by Ď66 had done ó well, did this work with Alexander on two Anderson models together, and then Maria generalized that, and theyíve done a lot of stuff since then, on weak ferromagnetism, which is reasonably satisfactory theory of some varieties of ferromagnetism.
Hoddeson: I want you to comment a little bit on the Anderson model. How was it received and things of that sort?
Anderson: Well, Berndt of course didnít believe a word of it, but Berndt never believed any theory. It was aimed at two things. One was, general microcosmic theory of magnetism, to use the UT model, and van Vleck model, and get something out of it. And to try to make sense of Friedelís ideas. And another was, Berndt Matthias question of why iron is magnetic some times and not others. The genesis was really more that, than anything else.
Hoddeson: When did you start thinking about it?
Anderson: Well, it was at that meeting, the Brasms meeting. Plus, well, itís Berndtís beginning to do magnetic impurities, which I think was started about that time. Then the idea of the resonant state. They just couldnít see ó also, there were some problems of numbers, really just number problems, and we knew more or less how broad Friedelís resonances were, and we knew more or less how big the exchange integrals were, and Friedel was just using conventional exchange, Slater, F-1s and Slater Fs and Gs. Excluding, like Slater, excluding the Fo excluding the U ó he was thinking in old-fashioned Slater terms, that exchange is exchange between arthugarel orbitals, and doesnít involve any of this big term exchange thatís so important in all these Hubbard model ideas. So I was having trouble believing Friedel, just for numerical reasons. Then I put this theory together, and I discovered two things. One was that you could use U instead of ordinary exchange integrals, which gained you an order of magnitude. And the second was that there was this fortuitous factor, Ti squared, and between the two of them I could easily make irons ferromagnetic, make oscillated ions magnetic [???]. So it was numbers. Orders of magnitude. Friedelís theory in principle was OK if you just took the exchange to be a general term, but he really clearly meant exchange integrals in the sense of ferromagnetic exchange between two orthonal orbitals, and not the antiferromagnetic type U Hubbard model, U exchange. And it was also, it is a development from the super-exchange idea, in that the key thing is the kind of a T squared over U question, in this case T squared over U is in a sense, the depth of the level. All these lines of thought came together, but the main one really, as far as I was concerned, was numbers. I could get the numbers to work, and Friedel as usual was too vague for there to be solid numerical data. Not too vague. It isnít Friedel to be too vague. Too offhand, not formal enough. So, thatís where that came from.
Hoddeson: What happened after that?
Anderson: People mostly believed it. You know, it seemed to be highly successful. Well, there was this one thing that was perhaps lost in the history. Bob Schrieffer is a very modest person. It was Bob Schrieffer who first really firmly decided that the Anderson model alone, and Hartree-Fock theory doesnít work. There were serious problems with quantum fluctuations in the model. And there is a paper, I think, somewhere, by Bob. Also we talked about it a lot, and so, in a sense, he was the first to see that the Anderson model would kondolize the theory would be a kondo type phenomenon. But that was maybe five or six years later. For the time being, everyone accepted it, and kind of carried out the current, and we believed that we understood why some irons were magnetic and some werenít.
Hoddeson: What happened then?
Anderson: Not much.
Hoddeson: Is it still at that point, then?
Anderson: Well, no. I mean, then of course there was all this business of, the Anderson model isnít right, because of when you look for it with various kinds of high energy spectroscopy, in terms so and so in of the D band with such and such energies, it looks as though you havenít got a U that there are various wonders involved there and various other interpretations of the data. And I think it has come back again, and people more or less believe that the width is more or less what I said it was and U is more or less what I said it was.
Anderson: And then also there was all this incredible proliferation of theoretical theory on the fluctuation effects and the kondo effect. The seminal meeting for that was this Brazno thing. Now, what else? There was something else that I was going to go back to. (pause....) What other crucial contacts were there? I was trying to remember the other day, there was some meeting, I think it may have been Slater and his gang ran a very early meeting on what could be done with inelastic scattering of neutrons. It was before Brockhaus, and they had, at Brookhaven, planned to do inelastic scattering. And in the end Brockhaus got in ahead of them. They were just talking about what interesting things could be done, and somehow I remember coming back from that with Shaw on a train, perhaps to the center of town, and we were talking about how, about just metallic ferromagnets, and he said, "Well, of course, they wonít have any of these spin waves that they were talking about." And I said, "Nonsense, of course they will." I donít remember whether that was before or after Herringís work, Herring and Kittell. He had the impression ó he was thinking of metallic ferromagnets and ordinary ferrites and things like that, totally different things. There canít be any fluctuations in the metallic ferromagnets because they obey the Stoner theory, and the Stoner theory says itís an exchange constant, itís everywhere in the metal. And all of a sudden, there is exchange everywhere in the metal, thereís no local dependence. I said, "Thatís not true." Actually, I think already either Bloch or Landau had introduced the concept of the exchange constant, as the twist energy, energy necessary for twisting the magnetization, but nobody had really applied this idea to metallic ferromagnets. It was that idea that Herring then ó I didnít talk about it to Herring. He was interested also in the same problem, and Herring and Kittell then applied in their spin wave theory. Herring, as I said, studied the exchange constant in terms of doing microscopic calculation of it. But everyone seems to think that Ginsberg-Landau, was where the idea of the stiffness, an effective free energy with a stiffness constant in it, came from. It seems to have been around most certainly in Herringís mind, certainly in my mind, because of that conversation with Shaw, quite a number of years before Ginsberg-Landau at least broke on the consciousness of the West. Anyway I was using these same ideas in my theories of ferroelectricity. Effective semi-microscopic interactions, looking at things enlarged but not microscopic scale. It is interesting that we all had that in mind. Itís to be found in one form in Herringís work, in another form in the early theory of domain walls, and in another form in Ginsberg-Landau. And that paper on soft modes in ferroelectrics with some similar ideas in it. I had actually finished it up in the early fifties, but I didnít give it until 1958 in Russia.
Hoddeson: When you said Ginsberg-Landau ó
Anderson: Yes, The Ginsberg-Landau is supposed to be the source of the idea that ó
Hoddeson: Are you talking about superconductivity? Thatís 1950.
Anderson: Thatís Ď51. Yet people hadnít really thought, you know ó nowadays, if you go to a meeting about statistical mechanics, people say, "Ginsberg-Laundau free energy" to you. They donít mean Ginsberg-Landau of superconductivity. What they mean is a free energy thatís got a gradient term in it, an effective free energy thatís got a gradient term in it. Landau clearly understood that when he did that paper. Or Ginsberg. Nobody else in the world did. Certainly in the West nobody was paying any attention to it, in those terms. I say nobody, but, very few people. Landau also had the same idea I think in his work on Bloch walls, the effective stiffness constant tells you how the exchange energy varies with gradients of the magnetization. But maybe the Russians understood it and maybe not. I doubt it. Certainly nobody in the West had this feeling. And these people in magnetism, the Stoners and Wahlfartís and all these people that believed in them, just thought, well, itís either ferromagnetic or it isnít, it doesnít twist. So, that the existence of that kind of free energy was not properly understood. And correspondingly when I first began thinking about ferroelectrics when I went to Bell Labs, Ď49, I was deriving things like that without really realizing that it was new even, or that it wasnít part of physics. And thatís an interesting point, I think. Certainly the experimental physicists who were doing neutrons ó the idea that that was a general way of thinking didnít come till much later. So, Landau-Ginsberg was thought of as a specific theory for superconductivity, not as a general concept which it only became much later. I think because Wilson called it that.
And if it had any origin, any origin to the concept, in peopleís minds, it really was not Landau-Ginsberg, but Landauís earlier work on domain walls. And then the stiffness constant, which came into peopleís minds through Herringís work on the stiffness constant, at least came into my mind, and I think I was probably there a long way before anyone else. Landau-Ginsberg just didnít penetrate to the American consciousness, or Europe, the whole Western consciousness, until much much later.
Hoddeson: So it played quite a peculiar role in the history of superconductivity ó
Hoddeson: It doesnít really; one could have left it out, in a sense.
Anderson: Yes, except ó
Hoddeson: ó and still gotten to, well, they could have gotten to BCS without it.
Hoddeson: I mean, it almost looks, in looking at the history, and of course I wasnít there, so I donít ó
Anderson: It came in the wrong order.
Hoddeson: Yes, in some sense itís a little like a Bohr atom theory, you know. If BCS could be reduced to Ginsberg-Landau, then one believed BCS later.
Anderson: Thatís right.
Hoddeson: But you know, if it hadnít happened, I mean.
Anderson: Thatís right, not because it wasnít important, because it was the right phenomenological theory of superconductivity, just as, I donít know, Bloch and Landau were the right phenomenological theory of ferromagnetism. I remember seeing Conyers write down this free energy and ó well, I donít know who was it that first wrote down. The theory of domain walls.
Hoddeson: Bloch. Bloch wrote that.
Anderson: But Landau had a lot to do with the domain walls. Bloch wrote down a domain wall but he didnít say, thereís an energy that depends on the gradient of N. And then I remember, Conyers Herring saying, "There is an energy that depends on a gradient of N." And that was in Ď53 or Ď54, totally independent of Ginsberg-Landau and not referring back to them. Well, itís a tremendously important thing that got re-invented. Itís like sohtons topological defects. It kept getting re-invented for each individual instance. And the cleanest one was [???] who did one of the early ones, and got the name on it. But I think the earliest by far was the Landau theory of domain walls in ferromagnetism, back in the thirties, which nobody thought of as anything new, because in a way he was just writing elasticity theory for magnetism. Really of course the first one was elasticity. Thatís the first Ginsberg-Landau theory. And Conyers I think just really thought of himself as writing down elasticity theory. But now itís called Ginsberg-Landau, for no particular reason. Well, partly because Ginsberg-Landau were the first people who said, "Letís expand around the critical point." And because they were stuck, they didnít have anywhere else to expand about, not because that was really the right place to sit.
Hoddeson: Didnít Landau do that earlier, in his the theory of phase transitions?
Anderson: No. Did he have gradient terms in his [???]
Hoddeson: I donít really remember it well enough.
Anderson: There was a Tisza paper and there was a Landau paper. It may have had gradient terms in it.
Hoddeson: I seem to remember that Landau had a theory of phase change paper.
Anderson: There must have been gradient terms in it. Because he gets Ornstein-Zernicke out of it.
Hoddeson: One has to look it up.
Anderson: Yes. Yes, I guess thatís true. Itís just strange that itís called Ginsberg-Landau rather than Landau-Ginsberg, because it must have been in Landauís theory of phase transformations and it must have been in all kinds of other things.
Hoddeson: Possibly the Bloch domain wall paper has something like that in it.
Anderson: It may. Certainly Conyers called it the Bloch-Wall thing.
Hoddeson: And heís the one who knows.
Anderson: And heís the one who knows.
Hoddeson: The Bloch-Wall paper was in 1932 or Ď33.
Hoddeson: Really early.
Anderson: Yes. Iím sure there was also a Landau paper on ferromagnetic domain walls very early.
Hoddeson: Iíve never read that one.
Anderson: With someone, one of those guys, Peierls or Ginsberg or Bloch or whoever ó because Landau refers back to that when he does Ginsberg-Landau, because itís again looking at the wall between two different phases. It isnít the same, but it also has the wall concept. So, what are the themes? Thereís resonance ó we talked a little bit about Kubo, and exchange narrowing and all that stuff. The first exchange narrowing paper was Gorter and van Vleck, of course, when Gorter was in Harvard in Ď47 giving that lecture course. Mostly narrowing ó of course Ames was Bloembergern and Purcell and Pound. I think that van Vleck must have played a major role in their getting that straight. Bloch wasnít much interested in that kind of thing, and he did magnetic resonance. He was interested more in the dynamics, not in using it as a tool for studying motions in liquids and solids and things like that. Bloembergen did a lot of that, but not all of it. He certainly introduced the concept of motional narrowing. Then I did my exchange narrowing paper, and then I wrote this general paper in Japan on both exchange and motional narrowing. And as I said, Kubo took over. Kubo-Tomida [???] and then he generalized that into a theory of kind of a fluctuation dissipation theory, and Karen Cower and Welton really called the fluctuation, you know ó theyíre right, they wrote down the fluctuation dissipation integral. On the other hand, people had been using these ideas all along, particularly Kubo and various other people. Thereís all are involved in the attempt to understand the various narrowing phenomena. Oh, what about spin echo? Have you got Hahn in there?
Hoddeson: Who ?
Anderson: Hahn, Erwin Hahn. Did he fit in that time frame? I donít remember when spin echo came along. It must have, because we were already ó I donít remember. No, he has to have been just a little later. He did lots of the passage cases and things like that, in the mid-fifties. He was working closely with George Feher. But we really didnít do spin echo nor were we very interested in it when it did come. George was much more interested in doing field physics with resonance techniques, rather than doing these pretty tricks. Now itís a very useful technique for studying relaxation phenomena; initially it just looked like a pretty trick. A lot of magnetic resonance work history was destroyed by Abregam. Abregan wrote this book about magnetic resonance, and he said he didnít approve of trying to do all this reference writing, so he was going to leave all the references out of his book, and just write a book as though the subject was brand new.
Hoddeson: Oh no. So why do you say it was destroyed?
Anderson: And so everyone thereafter referred to Abregam, and all the references disappeared from the literature.
Hoddeson: We may have to re-discover them.
Anderson: It will have to be rediscovered.
Hoddeson: You didnít comment too much on Kittel.
Anderson: I thought I had.
Hoddeson: A little bit. Maybe thatís enough. I donít know.
Anderson: No, he was, I said, he set up the Berkeley group. He was the leader of what magnetism got started in ó well, not the leader, because in principle, Bozorth was the leader at Bell, but the fact is, Bozorth was interested in completing his book, and Charlie was the person who was defending the magnetism effort, against two things. One of them was the overwhelming effect of the transistor, in kind of taking over everything. And the second was Bozorth, who wanted to finish his book. So Charlie defended Williams, he defended me, he defended the little magnetic resonance group that we had, Galt and Yager. I think he had left already in Ď52, didnít he, or something like that, very early. But by then, there was, it was more or less established, and then they set up this Clogston group. I guess it left, the nucleus of that was Clogston and Suhl and Walker. Heíd been working on traveling wave tubes. I think the company had come to the conclusion that the maser and the laser or something like that was going to supercede traveling wave tubes, so they said to these guys, "Retrain yourself in something," and they decided on magnetism, and they were thereafter the Clogston group. They were always thought of as a group of eager young men traveling in lockstep. Not that Larry Walker ever traveled in lockstep with anyone or Harry Suhl, but the rest of them did, Jack Green, Clogston Dillon and so on.
Hoddeson: But you know, Bozorthís whole approach to physics is so different from ó
Anderson: ó yes, heís a materials man.
Hoddeson: I mean, he took a completely classical point of view.
Anderson: Yes. Yes, he wasnít interested in the quantum mechanics of ferromagnetism. He was interested in putting alloys together.
Hoddeson: How could he have had any control at all?
Anderson: Because he was one of the first two or three real solid state physicists at the laboratories. He had a very strong entrance position, and he knew everybody everywhere and was a friend of van Vleckís. And he was a gentleman and a scholar, and he had a delightful wife. Terrible man himself, in fact. But he was just very influential with the powers that be, and heíd been there longer than anyone, and he knew everyone, and he was a friend of Mervyn Kellyís. He actually took up with Mervyn Kellyís wife after he died. You knew that? But what was so important about Kittel was the Berkeley group ó in the first place, in his influence on Galt and myself and also the Berkeley group. He arrived in Berkeley as the successor to all these people who had left because of the oath. He was the first scab. He wasnít really a scab, because he never signed the oath. They had stopped the oath when he arrived. He inherited all these students that these people had, who were very good. There was Oberhauser and there was Mortel Cohen and there was Elihu Abrahams and two or three others who were very good, Fred Keppler [???]. So he had about six or eight students to work with, and he had Kip, who then had a lot of ó Berkeleyís always had great solid state students, so Kip had in rapid succession Portis and Feher. I donít remember, but Walter Knight. So there was a tremendously powerful group of experimentalists who rapidly developed. And it was just an incredible combination of people doing very exciting things. They did the Oberhauser effect. They did the cyclotron resonance work, in which they beat out Fay Soule who was trying to do it in Bell Labs at the same time. Cyclotron resonance in semiconductors, then they happened to get to Russia very early after the thaw, and came home with the Azbel-Kaner idea. They got that going very rapidly. They also were very good at getting post-docs from England and the Englishman of course wants to go to the sunniest possible place. If heís going to come to the United States, he might as well get some sun. So he had Roger Elliot and he had John Owen, whoís terrific, and he had two or three others. He also got a lot of post-docs through just on general principles, and he had this fund that he got set up so that he could hire people like, first, Dyson, and then I went out there in Ď58, and took a summer professorship that he had at his disposal. Dyson worked for two summers on magnetism problems with Charlie. Did you know that?
Hoddeson: No. Absolutely not.
Anderson: Well, thereís two marvelous papers, one on the magnetic relaxation in metals, moving electrons, a very subtle, very beautiful piece of mathematics. And the second one was on spin waves. Justification of the use of spin waves. And also a justification of the fact that the second term after T to the 3/2 is T to the 4th. Itís excessive. You can get the same results in much simpler terms. But at the time it looked pretty wonderful. But the other isnít excessive. You really do need to do that mathematics, and Dyson did it and it was beautiful.
Hoddeson: I must look at those.
Ď56 and Ď57, then in Ď58, Dyson was busy down at General Dynamics, so Charlie brought me in, and that was the year I did my big paper on gauge invariance. Finished it there, and started the paper on super-exchange. So, Charlie had a real powerful thing going, and they started spin glasses. They did the Oberhauser effect. They did the semiconductor and metal cyclotron resonance work. There was the Knight shift work they did. The Oberhauser effect. Feher came from there, was trained there, but then he did his best stuff at Bell Labs, endor? and all that. Portis wrote the paper that stimulated me to do localization. He made the mistake of assuming that the average breadth is equal to the real breadth, and itís equal to 1 over T1. But Portis did this beautiful study of various passage phenomena. A very under-estimated physicist. He was very good, for a while. And the experimental end of it continued to be good, still continues to be good. But Charlie just couldnít ever get along with another theorist of equal power. He couldnít get along with me. And he had a lot of good people in there temporarily. He had Jim Phillips in there temporarily. He had John Hopfield in there temporarily, in his post-doc assistant professor...