Note: To be published in Physics in Perspective, vol. 12 no 3 (2010) pp. 248-265 The transformation of physics from 1900 to 1945 Yves Gingras Canada Research Chair in History and Sociology of Science Université du Québec à Montréal gingras.yves@uqam.ca Historians of science have analyzed in detail the conceptual transformations that gave physics its modern characters in the first half of the 20th century. We know a lot about the details of the emergence of atomic physics, relativity theory, quantum physics, quantum mechanics, nuclear physics and solid-state physics, to name only the major research domains that emerged between 1900 and 1945. These stories usually focused on the physicists who created new concepts, designed new instruments or imagined new experiments as well as on their universities and research laboratories. What is rarely provided in these narratives, however, is a feeling for the global scene of the physics community of which scientists are a part. How many publishing physicists were there around 1900? How many in 1925 or 1939? Do most physicists published as single authors or in teams and how did they compare with chemists or mathematicians in terms of scientific collaborations? What are the links between these disciplines? These questions about the demography and morphology of scientific disciplines and many others such quantitative questions concerning the collective practice of physics can hardly be answered using the usual tools of the historian of science focusing on particular archives or following the careers of a few scientists. Now, thanks to the major efforts of Thomson Reuters, which produces the now well-known Science Citation Index (SCI), scholars have access to the “Century of Science” database comprising detailed bibliographic information on more than 250 scientific journals covering physics and many other scientific disciplines for the period 1900 to 1945. For each paper we thus have not only the names of all authors, its title and the journal in which it was published, but also all the references it contain to previous research (papers, books and any other kind of cited reference). If one defines the field of physics as comprising those scientists who published at least one paper in the major physics journal of the times, one can then reconstruct the evolution of the physics community over the first half of the 20th century and even compare its development with other disciplines like chemistry and mathematics for example. In this paper, we propose such a global analysis of the trends that can be captured through a bibliometric analysis of hundreds of thousands of scientific papers and the more than one million references they contain. The most important scientific journals in chemistry, mathematics and physics are covered in the database and thus provide for the first time the possibility of making a comparative study of the changing dynamics of science over time (see Table 1 for the list of physics journals). Of course, the data base is not perfect and must be considered as a sample of the total and hardly accessible population of scientists active during the first half of the 20th century. And though American, British and German journals are better represented than French and other European journals for example, these are not completely absent and we can consider the trends observed as giving a good order of magnitude of the numbers of papers published as well as the number of researchers active during the period covered. Taken globally these journals

provide a realistic sample of the spectrum of research going on in countries like England, France, Germany and the United States over the period studied. The growth of scientific production and collaboration ANNALEN DER PHYSIK (Germany)

ANNALES DE CHIMIE ET DE PHYSIQUE (France)

ASTROPHYSICAL JOURNAL (USA)

J. RES. OF THE NATIONAL BUREAU OF STANDARDS USA)

JOURNAL OF THE ACOUSTICAL SOCIETY OF AMERICA (USA)

JOURNAL DE PHYSIQUE ET LE RADIUM (France)

JOURNAL OF APPLIED PHYSICS (USA)

JOURNAL OF CHEMICAL PHYSICS (USA)

JOURNAL OF PHYSICS-USSR

JOURNAL OF THE OPTICAL SOCIETY OF AMERICA (USA)

MEMOIRS OF THE ROYAL METROLOGICAL SOCIETY (UK)

MONTHLY NOTICES OF THE ROYAL ASTRONOMICAL SOCIETY (UK)

PHILOSOPHICAL MAGAZINE (UK)

PHILOSOPHICAL TRANSACTIONS OF THE ROYAL SOCIETY OF LONDON SERIES A (UK)

PHYSICA (Netherlands)

PHYSICAL REVIEW (USA)

PHYSIKALISCHE ZEITSCHRIFT (Germany)

PROCEEDINGS OF THE CAMBRIDGE PHILOSOPHICAL SOCIETY (UK)

PROCEEDINGS OF THE PHYSICAL SOCIETY OF LONDON (UK)

PROCEEDINGS OF THE ROYAL SOCIETY OF LONDON SERIES A (UK)

QUARTERLY JOURNAL OF THE ROYAL METEOROLOGICAL SOCIETY (UK)

REVIEW OF SCIENTIFIC INSTRUMENTS (USA)

REVIEWS OF MODERN PHYSICS (USA) SITZUNGSBERICHTE DER PREUSSICHEN AKADEMIE DER WISSENSCHAFTEN PHYSIKALISCH-MATHEMATISCHE KLASSE (Germany)

ZEITSCHRIFT FUR PHYSIK (Germany) Table 1. Physics Journals covered in the Century of Science database. As Figure 1 shows the number of papers in physics and mathematics did not grow substantially during the first two decades of the 20th century compared with chemistry, which doubled its number of publications. Given that the number of papers per author remains relatively stable over the period in each discipline1, the curve of the annual number of publishing scientists follow closely the same pattern. During the whole period, there is about one publishing mathematician for 5 physicists and 14 chemists. After the First World War, chemistry grows faster than physics as a discipline while the growth of mathematics is very slow during the inter war years. This relative scale clearly suggest that scientific research is not only a pure endeavor to understand nature but is also closely related to the social and economic demands of the times with chemistry the most closely linked to industry through the training of chemists and chemical engineers, followed well behind by physics and mathematics, the latter being useful above all for service

courses to science students but not directly answering to any obvious social needs beyond the training of high schools teachers, university professors, statisticians and actuaries needed to calculate the cost of insurances2.

Figure 1. Annual number of papers published by discipline. An important feature of the evolution of scientific research is the development of scientific collaborations through multi-authored articles. As Figures 2 and 3 show, chemistry is a discipline in which, as early as 1900, 30% of the publications are signed by more than one author, a proportion that is rising continuously over the year to 70% in 1944. By comparison, only 10% of physics papers had more than one author in 1900 but the proportion followed a similar trend as chemistry and rised to about 38% by the end of the 1930s. Not surprisingly, mathematicians predominantly published their papers as single authors, though a small trend toward collaboration is also visible over the period with about 10% of papers signed by two authors at the end of the 1930s. In the case of chemistry, where collaboration is the norm, it is the proportion of papers with 3 authors that is growing during the 1920s while this trend appears in physics only a decade later.

Figure 2. Annual proportion of articles with more than one author (3-year moving average)

Figure 3. Evolution of number of authors per paper in Chemistry and Physics. Relations between countries Keeping in mind the gaps and bias of the database, one can nonetheless get a rough measure of the relative weight of countries like Britain, Germany, USA in the evolution of the physics3. Figure 4 shows the annual number of papers published in physics by these three countries in their main national journals. Generally speaking the trend confirms what historian of science already know: German scientists were dominant in physics in terms of number of papers (and scientists) until the 1930s. We see however that US physical research started to grow vigorously after the end of the First World War and became dominant in the mid-1930s. The effect of the Nazi seizure of power in 1933 did not only affect the migration of many German scientists but also the presence of foreigners in German physics journals. For instance, among the 3,400 physicists who

published in American physics journals in the period 1900-1932, 378 (or 11%) also published at least once in German physics journals. For the period 1933-1944 that proportion falls to 91 physicists, less than 2% of the 5,600 publishing physicists for this period. The mid-thirties clearly signal the decline of German physics and the emerging centrality of American physics. The links of the latter with British physics are predictably stronger as 13% of the physicists present in American journals also publish in British journals in the period 1900-1932, a proportion that falls by half in the following period, maybe a sign of the greater internal strength of the American physics community. This seems corroborated by the fact that in the period 1933-1944 the proportion of physicists present in British journals who also publish in American journals is up at 13% from 11% in the first period. As for the links of British physics with Germany, they were weaker than those of American with only 7% of physicists in British journals publishing also in Germany before 1933. The proportion is the same in the other direction, from German physics journals to British physics journals. In general terms, German physics was the less dependent on foreign journals.

Figure 4. Annual number of papers in physics in Germany, UK and US. Though the absolute level of production of papers is of course related to the size of the country4, the distribution of citations to authors from different countries better reflect the relative strength of countries in the discipline as perceived from the citing countries. Figure 5 shows the distribution of the citations by country of the cited journal for physics excluding journal self-citations. For example, a citation to Annalen der Physik is counted as a German citation and one to Physical Review as an American one, and so on for each cited journal5. Though, as we have seen, a small proportion of papers from foreign authors were published in national journals, the vast majority of the papers, more than 90% in fact, come from nationals during the period covered.6 We see that American physics journals raise their proportion of citation over the period. Globally, the German scientific journals remain the most cited during the period despite a

continuous relative decline after World War One. Unsurprisingly, French journals are less cited in physics than American or British journals and their proportion falls after the war.

Figure 5. Country of origin of cited papers in physics (journal self-citations excluded). Of course, as shown in Fig 6, scientists tend to cite their own national journals before those of other countries. Nonetheless, starting in the mid 1920s, we see that German physics journals give more attention to papers published in American Journals, while the proportion of the British ones remain constant, a sure sign of the rise of American physics7. We observe the same general trends in the other disciplines.

Figure 6. Citations to different countries by German, US and UK national journals (self-citations excluded). Relations between physics journals Even though the database covers only the major physics journals, the papers they contain do cite thousands of different local journals. Figure 7 shows the network of relations between the major co-cited journals over the period 1900-1924. The pattern for the following period 1925-1944 (not shown) is similar except that Physical review now occupies the central position.

Figure 7: Network of relations between cocited journals in physics over the period 1900-1924. As could be expected, there are strong relations between all the most important physics journals of the time and we also observe that astrophysics and astronomy form a sub-group of three closely connected journals with the Astrophysical journal making the link with other astronomical publications and general physics journals. Table 2 shows the ten most central8 journals for the two periods and we see that Physical review moves from 7th position to first position, again a clear indication of the rapid growth of American physics after the 1920s. The major German physics journals, Annalen der Physik keeps a central position over the entire period but Physikalische Zeitschrift is displaced by Zeitschrift für Physik, created in 1920, which occupies the second position during the period 1925-1944. It also outranked the Verhandlungen der Physikalischen Gesellschaft of Berlin which was in the 6th position during the first period. The British journals, Philosophical Magazine and the Proceedings of the Royal Society also remain central over the entire period. Among the French journals the Comptes rendus of the Paris Academy of Sciences moves down a bit but stays among the top ten most central journals cited by physicists though the Annales de chimie et de physique whose publication stops in 1913 (while the Journal de physique et le radium starts in 1920) goes down to a much lower position (19th). Also, the rise of chemical physics, a specialty that developed during the 1910s, is clear through the emergence of a German journal in that field followed by the Journal of the American

Chemical Society. Interestingly, the next ten in the list comprise specialty journals in astrophysics, chemical physics (with the Journal of Chemical Physics being in 19th position) and optics.

Table 2 : Ten most central journals cited by physics journals for two time periods. Relations between disciplines By identifying the discipline of cited journals, we can measure the relations between disciplines. Here, the country bias of the database does not really affects the results since the choice of the cited journal is made by the scientists and the thousands of papers provide a strong sample of all physics papers. Also, eliminating the self-citations of a journal to itself, we should get a good idea of the exchanges between disciplines. An analysis of chemistry, mathematics and physics, shows that each discipline is pretty closed on itself and cite essentially disciplinary journals with no clear trend in time toward opening up to non disciplinary journals9. In mathematics for example only 6% of the citations go to physics journals while it is even much less in physics where mathematics journals are rarely cited (less than 1%). The relations between chemistry and physics are of course stronger because of the existence of the fields of chemical physics and physical chemistry, which act as boundary connections (Figure 8). As could be expected journals of biology and medicine are only important for chemistry (with 9% of the citations) and marginal in physics and mathematics as are engineering journals, which are cited only in physics (3%).

Figure 8. Distributions of citations in physics journals by discipline of the cited journals If we look at authors publishing in journals from different disciplines, the results are similar over the period 1900-1944 as we find few multidisciplinary authors except, again, in chemistry. Among a population of about 19,300 physicists and 4,500 mathematicians who published at least

1900-1924 1925-1944

ANNALEN DER PHYSIK PHYS REV

PHYS Z ZEITSCHRIFT FUR PHYSIK

PHILOSOPHICAL MAGAZINE ANNALEN DER PHYSIK

SITZUNGSBERICHTE DER KONIGLICH PREUSSISCHEN AKADEMIE DER WISSENSCHAFTEN PROCEEDINGS OF THE ROYAL SOCIETY OF LONDON

COMPTES RENDUS DE L ACADEMIE DES SCIENCES PHILOSOPHICAL MAGAZINE

VERH PHYSL GES PHYS Z

PHYS REV NATURE

PROCEEDINGS OF THE ROYAL SOCIETY OF LONDON COMPTES RENDUS DE L ACADEMIE DES SCIENCES

ANNALES DE CHIMIE ET DE PHYSIQUE ZEITSCHRIFT FUR PHYSIKALISCHE CHEMIE

PHILOSOPHICAL TRANSACTIONS OF THE ROYAL SOCIETY OF LONDON JOURNAL OF THE AMERICAN CHEMICAL SOCIETY

one paper over the entire period, only 760 of them publish in journals of both disciplines. Unsurprisingly that number is lower for chemists as we find only 500 among 62,300 who publish in mathematics in addition to chemistry journals. Again, the connections between chemistry and physics journals are much stronger, in large part because of physical chemistry and chemical physics, as nearly 4,800 scientists are found publishing in both types of journals. In short: multidisciplinary research is a very marginal practice during the first half of the 20th century and disciplines are essentially autonomous sub-systems of science. Changing generations and conceptual networks Bibliometric data can also provide insights into the changing centrality of research topics in physics. Analyzing the references contained in the papers, we can construct co-citation matrices of authors who are often cited together in the papers. Then, using techniques of network analysis, as we did above for journals, we can map the resulting conceptual networks10. Figure 9-12 show such maps for the central periods of physics between 1900 and 1944: quantum physics (1900-1904), relativity (1905-1911), quantum mechanics (1925-1930), nuclear physics and solid state physics (1937-1944) respectively. Looking at them makes pretty clear who were the central scientists of the times: J.J. Thomson and his work on the electron and propagation of electricity through gases and the physicists working on the black body problem all linked to Max Planck. For the period 1905-1911, we clearly see a reorganization of the relations between the main actors with Einstein becoming linked to H.A. Lorenz, and Minkowski, for their work on relativity theory, while Rutherford is central in the field of radioactivity and closely connected to the Curies. The centrality of an author may be due to his published papers as well as to having written classic textbooks such as Paul Drude’s Lehrbuch der Optik or Sommerfeld’s Atombau und Spektrallinien both translated in English and who long remained classic references. When we move to the mid-twenties, we clearly see a major shift in the most central actors with quantum mechanics and spectroscopy having replaced the study of relativity, electron theory and black body theory as the central topics of the field. The successive maps also make visible a densification of physics as well as its subdivision into more specialties. As we have shown elsewhere, applying techniques of community detection to the whole network recover specialties in the discipline. For the period 1905-1911, for example, we recover the sub set of physicists whose work touched upon electron theory and relativity as well as other sub groups corresponding to different specialties of physics11.

Table 3: Changes in the ten most central physicists in networks of cocitations during the period 1900-1944.

1900-1904 1905-1911 1912-1918 1919-1924 1925-1930 1931-1936 1937-1944

THOMSON, JJ THOMSON, JJ STARK, J SOMMERFELD, A HEISENBERG, W COMPTON, AH BETHE, HA

DRUDE, P DRUDE, P THOMSON, JJ BOHR, N SOMMERFELD, A DEBYE, P BHABHA, HJ

WIEN, W STARK, J DEBYE, P DEBYE, P HUND, F SLATER, JC BREIT, G

STARK, J LENARD, P EINSTEIN, A BORN, M BORN, M HEISENBERG, W BLACKETT, PMS

LENARD, P PLANCK, M PLANCK, M FRANCK, J MULLIKEN, RS DIRAC, PAM MOTT, NF

PLANCK, M WARBURG, E NERNST, W EINSTEIN, A PAULI, W MULLIKEN, RS HEISENBERG, W

LUMMER, O ABRAHAM, M WIEN, W THOMSON, JJ BIRGE, RT WALLER, I FERMI, E

WIEDEMANN, E LORENTZ, HA LENARD, P STARK, J SCHRODINGER, E HARTREE, DR ROSSI, B

WARBURG, E VOIGT, W SOMMERFELD, A LENARD, P COMPTON, AH FERMI, E WIGNER, E

PASCHEN, F EINSTEIN, A VOIGT, W PLANCK, M DEBYE, P JAUNCEY, GEM JOHNSON, TH

These changes in the centrality of authors are best summarized in Table 3, which shows the ten most central physicists for each five-year period. The rapid rise of Einstein as a central figure in physics, for example, is directly visible in the evolution of his degree of centrality: while he was of course absent in the period 1900-1904, he jumped already to the 10th position in the 1905-1911 period and raised to the 4th in the 1912-1918 period and slightly down (to the 6th) in the next period (1919-1924). The decline in centrality of Lorentz in the period 1912-1918 suggests that theory of the electron had become obsolete with the rise of Einstein theory of relativity. Einstein continued to be a central actor through his development of general relativity as well as his contributions to quantum theory. Like many of his generation, he rapidly declined in the period 1925-1930 and disappeared from the network of the most highly co-cited authors. Just looking at the names in fact suggests the changes in central topics as mentioned above. While the two decades preceding the First World War were dominated by problems related to electron and atomic physics, the mid-1920s was clearly focused on quantum mechanics followed by nuclear physics during the mid-1930s (with names like Hans Bethe, Homi Bhabha and Gregory Breit). Finally, an interesting feature of this global analysis of the scientific fields is that generational changes are also made visible in these maps: no major physicist kept his central position for more than 15 or 20 years. As vacant places are occupied by younger researchers, new research topics emerge and physics as a discipline is thus continually transformed.

Figure 9. Co-citation network of physicists, 1900-1904 (More than 8 co-citations) Line’s thickness is proportional to number of links.

Figure 10. Co-citation network of physicists, 1905-1911 (More than 8 co-citations).

Figure 11. Co-citation network of physicists, 1925-1930 (More than 15 co-citations)

Figure 12. Co-citation network of physicists, 1937-1944 (More than 22 co-citations) 1 The average number of papers per author over the period 1900-1945 is about 1.7 for Chemistry, 1.3 for mathematics and 1.4 for physics. These figures are similar to the value of about 1.6 for physicists around 1900 obtained by P. Forman, J. L. Heilbron and S. Weart “Physics circa 1900”, Historical studies in the physical sciences, vol. 5, 119 (1975). 2 On the growth of mathematics see K. H. Parshall and A. C. Rice (Eds), Mathematics Unbound : The Evolution of an International Mathematical Research Community, 1800-1945, History of Mathematics, vol. 23, American Mathematical Society, (2002). 3 It should be kept in mind that France is clearly under represented in the database with only two disciplinary journals and the Comptes rendus of the Paris Academy of sciences that cover all disciplines. Physics and chemistry (combined) are represented by the Annales de chimie et de physique (1900-1913) followed by the Journal de physique et le radium (1920-1944). To compensate this under estimation, we attribute the total number of papers in these journals to physics. As for the Comptes rendus papers, we have analyzed a sample of about 7000 papers over the period 1900-1944 and found that a third of them pertain to physics. In their study of the state of physics around 1900, Forman, Heilbron and Weart, using four different samples from British and German index of scientific papers, found that Germany then accounts for about 29% of the published papers, UK for 21%, France for 18% and USA for 12%. If we attribute 1/3 of the

papers in the Comptes rendus of the Paris Academy to physics and add the results to those of the Annales de chimie et physique, we get comparable data within their suggested error of 20% (Forman et al, op. cit., note 1, p. 10). 4 Forman, Helibron and Weart find that around 1900 the difference in physics productivity of the major producing country is essentially due to their differences in per capita national incomes (op. cit., note 1, p. 10) 5 Though there are also books and other kind of documents cited, we have retained only the journals and the % are calculated on the total of the sample of the cited journals for which we have attributed a country of origin. We also excluded the self-citations of journals to themselves and kept only inter-journal citations. 6 Following the pioneering work of Paul Forman, John Heilbron and Spencer Weart, we attribute to a given country all the papers appearing in their national journals. As noted by these authors, this procedure introduces “no appreciable error”; see “Physics circa 1900”, (op. cit. note 1, p 115). 7 On the rise of American physics see D. Kevles, The Physicists. The History of a Scientific Community in Modern America, Alfred A. Knopf, New York (1978); S. Weart, “The Physics Business in America, 1919-1940 : A Statistical Reconnaissance”, in N. Reingold (Ed), The Sciences in the American Contetxt: New Perspectives, Smithsonian Institution Press, Washington D.C. , pp. 295-358 (1979). 8 In networks analysis, the centrality of a node is calculated as the sum of all the links with which it is connected. The most central nodes are the ones with the largest number of relations with other nodes and the least central ones are at the periphery of the network. On centrality, see See Freeman, L. C. “Centrality in Social Networks. Conceptual Clarification”. Social Networks, 1, 215-239 (1978/1979). 9 The class of “general” journals include non disciplinary ones like Nature, Science and Proceedings of Academies that cover many scientific disciplines. 10 Gingras, Yves, “Mapping the Changing Centrality of Physicists (1900-1944)”, in Proceedings of the 11th Conference of the International Society for Scientometrics and Informetrics (ISSI), Madrid, Spain, , pp.314-320 (2007). 11 Wallace, Matthew L., Yves Gingras, Russell Duhon, “A new approach for detecting scientific specialties from raw cocitation networks”, JASIST, 60, no 2, 240-246 (2009).