2011 Pais Prize Lecture: Shelter Island Revisited

By Silvan S. Schweber

The Shelter Island Conference on Quantum Mechanics was a landmark. It was the first of three small post-World War II conferences on theoretical physics sponsored by the National Academy of Sciences (NAS). The first took place June 2–4, 1947, at the Ram's Head Inn on Shelter Island at the tip of Long Island. The second, the Pocono Conference, was held from March 30 to April 2, 1948, and the third, Oldstone, occurred April 11–14, 1949.

The initial impetus for the Shelter Island Conference came from Duncan MacInnes, a distinguished physical chemist at the Rockefeller Institute, who had suggested to Frank Jewett, then president of the National Academy of Sciences (NAS), that it sponsor a series of small conferences. The participants (see Fig. 1) were primarily theoretical physicists, most of whom had been leaders at the highly successful wartime laboratories: the MIT Radiation Laboratory, Chicago Metallurgical Laboratory, Johns Hopkins Applied Physics Laboratory, and Los Alamos. Also in attendance were several experimental physicists including Willis Lamb and Isidor Rabi, who reported on experiments on the spectrum of hydrogen that had been recently performed at Columbia University, and Bruno Rossi who reported on the results of researches carried out in Rome on the atmospheric absorption of cosmic rays.

Photograph courtesy of AIP Emilio Segrè Visual Archives, Marshak Collection.

Fig. 1. Assembled attendees of the June 1947 Shelter Island Conference on Quantum Mechanics. Left to right: Isidor Rabi, Linus Pauling, John Van Vleck, Willis Lamb, Gregory Breit, Duncan MacInnes, Karl Darrow, George Uhlenbeck, Julian Schwinger, Edward Teller, Bruno Rossi, Arnold Nordsieck, John von Neumann, John Wheeler, Hans Bethe, Robert Serber, Robert Marshak, Abraham Pais, Robert Oppenheimer, David Bohm, Richard Feynman, Victor Weisskopf, Herman Feshbach. Not included here is Hendrik Kramers.

The significance of the conference was soon as evident to the participants as it is now in retrospect. Six months later, in January of 1948, APS secretary Darrow wrote his fellow organizer Duncan MacInnes a brief postcard stating, "I must quote [you] . . . the words of warm commendation used yesterday by I. I. Rabi about your Shelter Island meeting—he said that it has proved much more important than it seemed even at the time, and would be remembered as the 1911 Solvay Congress is remembered, for having been the starting-point of remarkable new developments." Similarly, Richard Feynman many years later recalled, "There have been many conferences in the world since, but I've never felt any to be as important as this."

The meeting turned out to be the most seminal conference held right after the end of World War II; its impact was indeed comparable to that of the Solvay Congress of 1911. Just as that meeting set the stage for all the subsequent developments in quantum theory, similarly Shelter Island provided the initial stimulus for the post-World War II developments in quantum field theory: effective, relativistically invariant, computational methods; Feynman diagrams; and renormalization theory. The conference also witnessed the elucidation of the structure of the mesonic component of cosmic rays.

The Shelter Island, Pocono and Oldstone conferences were small, closed, and elitist in spirit. In a sense they mark the postponed end of an era, that of the 1930s, and its characteristic style of doing physics: small groups and small budgets. None cost more than $1,500. The ravages of inflation are illustrated by the fact that the average expenditure to accommodate all the 25 attendees for one night at the elegant inns where the conferences took place was$200.

Coming after World War II, these conferences reasserted the values of pure research and helped to purify and revitalize the theoretical physics community. They also made evident the new social reality implied by the newly acquired power of the theoreticians, and helped integrate the most outstanding of the younger theoreticians into the elite: Richard Feynman, Julian Schwinger, Robert Marshak, David Bohm, and Abraham Pais at Shelter Island, plus Freeman Dyson at Oldstone.

The Shelter Island, Pocono, and Oldstone conferences were the precursors of the Rochester Conferences on High Energy Physics. But they differed from these later conferences in important ways. While the Shelter Island, Pocono, and Oldstone conferences reflected the style of an earlier era, the Rochester Conferences were more professional and democratic in outlook and had the imprint of the new era: the large group efforts and the large budgets involved in machine physics. And whereas Shelter Island and Pocono looked upon quantum electrodynamics as a self-contained discipline, the Rochester Conferences saw "particle physics" come into its own, with QED as one of its subfields—albeit one with a privileged, paradigmatic position.

The small, elitist character of the Shelter Island, Pocono and Oldstone conferences embodied the ideal of MacInnes, a NAS member and a past president of the New York Academy of Sciences. For several years before becoming its president in 1944, he had helped arrange a series of small conferences sponsored by that organization, "designed to promote active discussion of different scientific topics." For the conferences to achieve their aim, MacInnes felt it was essential that the topics "be in areas in which actual work was in progress and that participation be restricted to currently active investigators in the designated fields." Although an early conference sponsored by the New York Academy had achieved this objective, he believed that the effectiveness of the later ones had been impaired "by their success in attracting too large a crowd." MacInnes felt very strongly that attendance should be limited. In October of 1944 he declined to run again as New York Academy president when a decision on this matter by its council went against him, and in January 1945 he resigned from the Academy over this issue.

In the fall of 1945, MacInnes sent Jewett a proposal for the NAS to sponsor occasional meetings "of the relatively few men doing active research work in each field." These conferences would address a topic on which active research was occurring; have a small number of papers presented that would be distributed before the conference; have much more time for discussion than for presentation of papers; and be limited to at most 25 or 30 participants. The papers, revised in the light of the discussion that took place, would be published as a NAS monograph.

Jewett liked the idea and felt confident that the NAS would be willing to sponsor the undertaking. He suggested that MacInnes pick out one or two problems that seemed promising and use these as "pilot plants." Given this encouragement, MacInnes, after discussion with colleagues, suggested two conferences, one on "The Nature of Biopotentials" and the other on "The Postulates of Quantum Mechanics." Biopotentials were the focus of the research of his colleague and friend W. J. V. Osterhout and were of considerable importance in his own work. The topic of the second conference was an area that intrigued MacInnes, who at the time was studying wave mechanics.

K. K. Darrow—the urbane, and by then perennial, secretary of the American Physical Society, a theoretical physicist who had been at Bell Telephone Laboratories since 1925 and who was a popularizer of science of some stature—offered MacInnes his help in organizing the conference on quantum mechanics. The two of them held extensive discussions with Leon Brillouin and Wolfgang Pauli. However, MacInnes did not like Pauli's suggestions because "Pauli was planning for too many of the older men, and [MacInnes thought] that the best results would be to get out the coming generation." Thus, at Darrow's suggestion, the topics to be discussed and the responsibility for drawing up a list of participants was given over to John Wheeler, who had also attended the meetings with Pauli. Wheeler was indeed representative of the "younger men."[1] During the war he had distinguished himself at the Chicago Metallurgical Laboratory and borne major responsibility for the design and construction of the Hanford reactors. Wheeler's stature in the theoretical physics community was clearly recognized. In the fall of 1945, he was chosen to present a paper on the "Problems and Prospects in Elementary Particle Research" at the Symposium on Atomic Energy and Its Implications, the high point of the first postwar meeting of the NAS, which it sponsored jointly with the American Philosophical Society. At that symposium Wheeler shared the limelight with Enrico Fermi, J. Robert Oppenheimer, Eugene Wigner, Arthur H. Compton, Harold Urey, and Irving Langmuir. His impressive address gave proof that indeed a new generation of American physicists was taking over the intellectual leadership in the emerging field of "elementary particle" physics.

Under the guidance of Jewett, MacInnes, and Darrow, the Shelter Island Conference would reinforce that message. Darrow in his letter inviting Wheeler to oversee the agenda of the conference made explicit what had been implicit with MacInnes and Jewett: the conference was to be an American one, designed to "bring out" the young American theoretical physicists who had played such a large role in the successful war effort. The three organizers wanted the conference to demonstrate that theoretical physics in the United States had come into its own with the younger men who had been born and trained here. The conference was to prove the strength of American theoretical physics not only in wartime activities but also in "pure" physics. These conferences gave further proof of the remarkable intellectual powers of the leading theoreticians. The final list of participants corroborates this intent.

The conference was eventually set to occur on Monday, Tuesday, and Wednesday, June 2–4, 1947, with Darrow serving as its convener and chairman. Its timing was determined to accommodate the presence of Oppenheimer, who was also to write a 500-word paper outlining subjects for discussion. The other two physicists asked to prepare papers for the conference were Victor Weisskopf and Hendrik Kramers—a long-time friend of Wheeler—who at the time was visiting the Institute for Advanced Study after having chaired the UN scientific and technological committee on nuclear energy and taught at Columbia.

Weisskopf's paper outlined the problems faced in elementary particle physics in broad and general terms, and urged that the conference include a discussion of fundamental experiments to be done with the almost-completed "very powerful accelerators in the energy region of 200–300 MeV." Oppenheimer's outline was more narrowly focused and concentrated on cosmic-ray phenomena.

Kramers, for his part, chose to review the difficulties encountered in QED since its inception in 1927 and to indicate one way out of these problems. He pointed to his own work of 1938–1940, and that of his students, Serpe and Opechowski, which had been carried out in 1940, presenting a theory in which all structure effects had been eliminated and describing "how an electron with experimental mass behaves in its interaction with the electromagnetic field."[2]

By the end of May, the preliminary results of Lamb and Robert Retherford—that contrary to what the Dirac equation said about the energy levels of an electron in a Coulomb field, the 2s1/2 level of hydrogen lies 1,000 megacycles above the 2p1/2 level—had been widely circulated. Schwinger and Weisskopf discussed the theoretical implication of the experiment on their train ride from Boston to New York to attend the conference and agreed that the effect was very likely quantum electrodynamic in origin. Schwinger later recalled their also coming to the conclusion that since the electron self-energy was logarithmically divergent in hole theory, the energy difference between these levels would be finite when calculated with that theory.

In fact, the matter had probably been discussed even earlier over the lunches Schwinger and Weisskopf periodically shared, since this hydrogen level shift was one of Weisskopf's current research interests. In the fall of 1946 he gave the problem of a hole-theoretic computation of the 2s–2p level shift in hydrogen to Bruce French, who had been working on it since then. Weisskopf would very likely have told Schwinger of this research, elicited his reaction, and sought his advice on effective computational approaches.

The conferees gathered in New York on Sunday June 1, 1947, at the AIP headquarters on East 55th Street. From there they were taken "on an old and shaky" bus to Greenport at the western end of Long Island. On the final phase of the trip, they were accompanied by a police motorcycle escort, and their bus didn't stop at any traffic lights. As they passed each county line a new police escort would meet them. In Greenport the conferees were wined and dined as guests of the Chamber of Commerce, paid for by John C. White, its president. In an after-dinner speech, he said that he had been a Marine in the Pacific during the war and that he—and many like him—would not be alive were it not for the atomic bomb.

Darrow chaired the conference, but Oppenheimer was the dominant personality and "in absolute charge." MacInnes recorded in his diary that "it was immediately evident that Oppenheimer was the moving spirit of the affair." Darrow in his diary gave a revealing account:

As the conference went on the ascendency of Oppenheimer became more evident—the analysis (often caustic) of nearly every argument, that magnificent English never marred by hesitation or groping for words (I never heard "catharsis" used in a discourse on [physics], or the clever word "mesoniferous" which is probably O's invention), the dry humor, the perpetually-recurring comment that one idea or another (incl. some of his own) was certainly wrong, and the respect with which he was heard.

Next most impressive was Bethe, who on two or three occasions bore out his reputation for hard & thorough work, as in analyzing data on cosmic rays variously obtained.

On the first day of the conference, Lamb presented the most recent data from his experiment with Retherford. In the following discussion, Oppenheimer pointed out that if one calculated the difference in the 2p and 2s energy levels using hole theory, a finite answer might be obtained, given that the divergences encountered in both were logarithmic (see Fig. 2).

Courtesy of Rose Bethe

Fig. 2. Bethe’s handwritten notes of Oppenheimer’s talk.

Lamb was followed by Rabi, who presented the data that J. E. Nafe, E. B. Nelson and he had obtained on the hyperfine structure of H and D. That afternoon Rossi reported on the experiment that had been carried out in Rome by M. Conversi, E. Pancini and O. Piccioni on the absorption of mesons in the atmosphere. The next morning Kramers presented his version of the Lorentz theory of an extended charge in which structural effects had been encapsulated in the experimental mass of the particle. He concentrated on the classical version of his non-relativistic theory, although Bethe's notes make clear that in the last part of his talk he indicated what quantizing the theory would do (see Fig. 3).

Courtesy of Rose Bethe

Fig. 3. Bethe’s notes of Kramers’ Shelter Island talk.

Kramers' talk was clearly influential. His use of the Hertz potential and his derivation of the potential that the dressed electron with experimental mass experiences became Schwinger's point of departure in his quantum electrodynamic calculation of the Lamb shift and of the magnetic moment of the electron.

In the afternoon of the second day, Weisskopf reviewed the divergence difficulties in the hole-theory calculations of the self-energy of an electron. In a paper that he had written in 1939, he had indicated that in hole theory the divergence would be logarithmic to all orders of perturbation theory. According to Breit's notes of the conference, Weisskopf agreed with Oppenheimer's hunch that a hole-theoretic calculation of the difference in the 2p and 2s energy levels would not diverge. During the ensuing discussion, Weisskopf and Schwinger indicated how a hole-theoretic calculation of the level shift might be attempted and suggested further reasons why a finite result might result from applications of Kramers' ideas.

After the conference was over, Bethe performed his famous non-relativistic calculation on the train ride from New York to Schenectady. The paper in which he proved that the level shift would be accounted for quantum electrodynamically was completed by June 9 and circulated to conference participants.[3] Bethe did not acknowledge Kramers' talk in his paper even though this presentation had been crucial in stressing the importance of expressing observables in terms of the experimental mass of the electron.

This omission probably happened because Bethe thought that he had a much simpler way than Kramers to incorporate this insight. Bethe had noted that the quantum electrodynamically calculated self-energy of a free non-relativistic electron could be ascribed to an electromagnetic mass of the electron and —though divergent—had to be added to the mechanical mass of the electron. The only meaningful statements of the theory involve the sum of the electromagnetic and mechanical masses, which in combination is the experimental mass of a free electron. In contrast to Kramers' approach, Bethe's was a model-independent formulation of mass renormalization that did not assume an extended charge distribution of the electron. And in contrast to Schwinger and Weisskopf's initial insight, that a hole-theoretic calculation of the difference between the energies of two levels would be finite, Bethe's approach allowed computing the energy of each level and gave an unambiguous formulation of mass renormalization in the non-relativistic case. Moreover, he knew what was required to formulate an analogous relativistic prescription. Weisskopf and Schwinger, although emphasizing Kramers' insight, could not do so at Shelter Island.

In his 1983 talk commemorating Shelter Island I, Lamb made the remarks [4]:

Kramers was there that year. When I heard Kramers talk at Shelter Island I it seemed to me that he mainly said that we should be applying the methods used by Lorentz for the classical [electron]. I could not see that he indicated how such a program could be carried out, and hence did not derive great inspiration from his talk…* Unlike Weisskopf, who was working on these problems actively before this wonderful time, Kroll and I were only inspired when we found out how Bethe had subtracted the self energy of a free electron.

In his 1947 paper Bethe did acknowledge the comments by Weisskopf and Schwinger that a hole-theoretic calculation of energy level differences would be finite and in a footnote stated: "It was first suggested by Schwinger and Weisskopf that hole theory must be used to obtain convergence in this problem." Their insight justified the high-energy cut-off he introduced in his calculation.

Dresden, in his 1988 biography of Kramers, suggested that he did not receive adequate credit for his Shelter Island contributions. Bethe's notes indicate that Dresden was correct. To Bethe, the pragmatist for whom numbers were always the criterion of good physics, and who had just been so deeply and successfully involved in the war effort calculating numbers that translated into physical effects and measurable empirical data, the challenge was to get the numbers out and account for the magnitude of the 2s–2p shift in hydrogen and for the new values of the hyperfine splitting in H and D that Nafe, Nelson and Rabi had just measured. Accounting for empirical data would be explaining the data. Kramers' approach was too model-dependent, too theoretical, and too far removed from calculating numbers. For Bethe, the value of a novel idea was gauged by whether it could help you calculate numbers that could be compared with empirical data.

The importance of Bethe's calculation is apparent from Weisskopf's reaction to it. He received Bethe's manuscript on June 11 and after studying it wrote Bethe that he was…

…quite enthusiastic about the result. It is a very nice way to estimate the effect and it is most encouraging that it comes out just right. I am very pleased to see that Schwinger's and my approach seems to be the right one after all. Your way of calculating is just an unrelativistic estimate of our effect, as far as I can see.

I am all the more pleased about the result since I tried myself unsuccessfully to estimate the order of magnitude of our expression. I was unable to do this but I got more and more convinced that the method was sound…I would like to talk it over with you especially the `korrespondenz Deutung' of the effect.[6]

Fig. 4. Excerpt from Bethe, "The Electromagnetic Shift of Energy Levels" [5].

Bethe's calculation was a "crucial calculation"—a notion I owe to my Brandeis colleague and friend Howard Schnitzer. By introducing in a simple manner the concept of mass renormalization and its associated meaning and consequences, Bethe's calculation offered a new perspective on how to address quantum electrodynamical calculations and to obtain numbers that could be compared with experimental data. It was crucial because he could convincingly justify the cut-offs he had to introduce in his non-relativistic calculation and obtain a logarithmic expression that agreed with the observed level shift (see Fig. 4).

Only Bethe could have evaluated the logarithmic contribution as quickly as he did. He had encountered similar logarithmic expressions when calculating quantum mechanically the energy loss of fast charged particles traversing matter in his Habilitationschrift in 1929!

Schwinger at the time made another crucial—and perhaps more important—calculation, of the quantum electrodynamic contribution to the magnetic moment of the Dirac electron. His calculation was crucial because it was the first field-theoretic calculation of quantum electrodynamic radiative corrections. Up to this point, all such calculations had been based on hole theory. Schwinger's calculation asserted that a quantum field-theoretic description of both electrons and radiation is the generative and effective way to describe atomic and subatomic processes.

The Shelter Island Conference marked a major transition in the history of theoretical physics. It placed experiments on the spectrum of hydrogen that indicated deviations from the predictions of the Dirac equation at the center of theoretical interest. By obtaining reliable and precise values for these discrepancies, Lamb and Rabi posed a challenge to the theoretical physics community. Renormalization theory, the principal outgrowth of the discussions at Shelter Island, allowed the difficulties that had plagued quantum electrodynamics to be circumvented. The subsequent computational techniques devised by Schwinger, Feynman and Dyson allowed the calculation of the electromagnetic properties of simple atomic systems to previously unheard-of accuracy. These developments dispelled whatever doubts remained about the adequacy of QED and renewed hope that field-theoretic explanations of the nuclear forces would eventually be found.

Footnotes and References

[1] Wheeler, a brilliant young theoretical physicist, had been educated at Johns Hopkins University. As a National Research Council Fellow, he worked in 1933–34 with Gregory Breit at NYU on problems in quantum electrodynamics and spent the 1934–35 academic year at Niels Bohr's Institute in Copenhagen. He made important contributions to nuclear physics, especially in a paper on the scattering matrix and his researches with Bohr on the fission processes. In 1938, at the age of 26, he joined the Princeton Physics Department as an assistant professor.

[2] After receiving Oppenheimer's, Weisskopf's, and Kramers' papers, Bethe wrote Weisskopf in mid-May:

I read your outline and got the impression that we should be careful not to try and do too much. I should like especially to hear from Kramers in great detail about his new theory. Generally, I think we should try to hear from people who have actually got some results and avoid as much as possible discussions in the vacuum. For this reason I am not much in favor of extending discussions of experiments to be done in future accelerating equipment. If our discussion of concrete theoretical problems leads to desirable experiments, this is very good; but I believe we should not continue the war spirit of planned research too much.

Would it be possible to set one day aside for Kramers and have a really good discussion on that, then perhaps have half a day or one day for meson production starting with Oppy's theory, and then have another day devoted to the Piccioni experiment and its interpretation?

[3] Bethe's accompanying letter to Oppenheimer was brief and to the point:

Enclosed I am sending you a preliminary draft of a paper on the line shift. You see it does work out. Also, the second term already gives a finite result and is not zero as we thought during the conference. In fact, its logarithmic divergence makes the order of magnitude correct. It also seems that Vicki and Schwinger are correct that the hole theory is probably [handwritten insertion] important in order to obtain convergence. Finally, I think it shows that Kramers cannot get the right result by his method.

[4] Shelter Island II: Proceedings of the 1983 Shelter Island Conference on Quantum Field Theory and the Fundamental Problems of Physics. Edited by Roman Jackiw, Nicola Khuri, Steven Weinberg, and Edward Witten (Cambridge, MA: MIT Press, 1985). At *, the editors note, Bethe interjected, "He did inspire me!"

[5] Bethe, H.A., "The Electromagnetic Shift of Energy Levels," Physical Review 72 (1947) 339–41, footnote 6.

I do not quite agree with your treatment of the history of the problem in your note. That the 2S1/2–P1/2 split has something to do with radiation theory and hole theory was proposed by Schwinger and myself for quite some time. We did not do too much about it until shortly before the conference. We then proposed to split an infinite mass term from other terms and get a finite term shift, just as I demonstrated it at the conference. Isn't that exactly what you are doing? Your great and everlasting deed is your bright idea to treat this at first unrelativistically. "Es mochte doch schon sein" if this were indicated in some footnote or otherwise.

Bethe had put such a footnote into his paper and answered Weisskopf a few days after receiving his letter. Weisskopf soon acknowledged that Bethe's "abstract" was "harmloser" (much more harmless) than he initially thought and agreed, "Let's forget about patent claims."

Silvan S. Schweber, QED and the Men Who Made It: Dyson, Feynman, Schwinger, and Tomonaga (Princeton University Press, 1994). Chapter 4 discusses in some detail the Shelter Island, Pocono, and Oldstone conferences.

Freeman J. Dyson, Disturbing the Universe (Harper & Row, 1979)

The articles in this issue represent the views of their authors and are not necessarily those of the Forum or APS.